1204 files changed, 153042 insertions, 61356 deletions

diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c
index 1750e976b8..aea4894448 100644
--- a/src/egl/drivers/demo/demo.c
+++ b/src/egl/drivers/demo/demo.c
@@ -49,9 +49,8 @@ typedef struct demo_context
 
 
 static EGLBoolean
-demoInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
+demoInitialize(_EGLDriver *drv, _EGLDisplay *disp, EGLint *major, EGLint *minor)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    _EGLScreen *scrn;
    EGLint i;
 
@@ -85,7 +84,9 @@ demoInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
       _eglAddConfig(disp, config);
    }
 
-   drv->Initialized = EGL_TRUE;
+   /* enable supported extensions */
+   disp->Extensions.MESA_screen_surface = EGL_TRUE;
+   disp->Extensions.MESA_copy_context = EGL_TRUE;
 
    *major = 1;
    *minor = 0;
@@ -95,71 +96,56 @@ demoInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 
 
 static EGLBoolean
-demoTerminate(_EGLDriver *drv, EGLDisplay dpy)
+demoTerminate(_EGLDriver *drv, _EGLDisplay *dpy)
 {
    /*DemoDriver *demo = DEMO_DRIVER(dpy);*/
-   free(drv);
    return EGL_TRUE;
 }
 
 
 static DemoContext *
-LookupDemoContext(EGLContext ctx)
+LookupDemoContext(_EGLContext *c)
 {
-   _EGLContext *c = _eglLookupContext(ctx);
    return (DemoContext *) c;
 }
 
 
 static DemoSurface *
-LookupDemoSurface(EGLSurface surf)
+LookupDemoSurface(_EGLSurface *s)
 {
-   _EGLSurface *s = _eglLookupSurface(surf);
    return (DemoSurface *) s;
 }
 
 
-
-static EGLContext
-demoCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list)
+static _EGLContext *
+demoCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list)
 {
-   _EGLConfig *conf;
    DemoContext *c;
    int i;
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   if (!conf) {
-      _eglError(EGL_BAD_CONFIG, "eglCreateContext");
-      return EGL_NO_CONTEXT;
-   }
-
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
       switch (attrib_list[i]) {
          /* no attribs defined for now */
       default:
          _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext");
-         return EGL_NO_CONTEXT;
+         return NULL;
       }
    }
 
    c = (DemoContext *) calloc(1, sizeof(DemoContext));
    if (!c)
-      return EGL_NO_CONTEXT;
+      return NULL;
 
-   _eglInitContext(drv, dpy, &c->Base, config, attrib_list);
+   _eglInitContext(drv, &c->Base, conf, attrib_list);
    c->DemoStuff = 1;
    printf("demoCreateContext\n");
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
-   assert(_eglGetContextHandle(&c->Base));
-
-   return _eglGetContextHandle(&c->Base);
+   return &c->Base;
 }
 
 
-static EGLSurface
-demoCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list)
+static _EGLSurface *
+demoCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list)
 {
    int i;
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
@@ -167,107 +153,88 @@ demoCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, Nativ
          /* no attribs at this time */
       default:
          _eglError(EGL_BAD_ATTRIBUTE, "eglCreateWindowSurface");
-         return EGL_NO_SURFACE;
+         return NULL;
       }
    }
    printf("eglCreateWindowSurface()\n");
    /* XXX unfinished */
 
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
-static EGLSurface
-demoCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list)
+static _EGLSurface *
+demoCreatePixmapSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list)
 {
-   _EGLConfig *conf;
    EGLint i;
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   if (!conf) {
-      _eglError(EGL_BAD_CONFIG, "eglCreatePixmapSurface");
-      return EGL_NO_SURFACE;
-   }
-
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
       switch (attrib_list[i]) {
          /* no attribs at this time */
       default:
          _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePixmapSurface");
-         return EGL_NO_SURFACE;
+         return NULL;
       }
    }
 
    if (conf->Attrib[EGL_SURFACE_TYPE - FIRST_ATTRIB] == 0) {
       _eglError(EGL_BAD_MATCH, "eglCreatePixmapSurface");
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
    printf("eglCreatePixmapSurface()\n");
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
-static EGLSurface
-demoCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+static _EGLSurface *
+demoCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                          const EGLint *attrib_list)
 {
    DemoSurface *surf = (DemoSurface *) calloc(1, sizeof(DemoSurface));
+
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
    /* a real driver would allocate the pbuffer memory here */
 
-   return surf->Base.Handle;
+   return &surf->Base;
 }
 
 
 static EGLBoolean
-demoDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+demoDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface)
 {
    DemoSurface *fs = LookupDemoSurface(surface);
-   _eglRemoveSurface(&fs->Base);
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!_eglIsSurfaceBound(&fs->Base))
       free(fs);
-   }
    return EGL_TRUE;
 }
 
 
 static EGLBoolean
-demoDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
+demoDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *context)
 {
    DemoContext *fc = LookupDemoContext(context);
-   _eglRemoveContext(&fc->Base);
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!_eglIsContextBound(&fc->Base))
       free(fc);
-   }
    return EGL_TRUE;
 }
 
 
 static EGLBoolean
-demoMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context)
+demoMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *drawSurf, _EGLSurface *readSurf, _EGLContext *ctx)
 {
    /*DemoDriver *demo = DEMO_DRIVER(dpy);*/
-   DemoSurface *readSurf = LookupDemoSurface(read);
-   DemoSurface *drawSurf = LookupDemoSurface(draw);
-   DemoContext *ctx = LookupDemoContext(context);
    EGLBoolean b;
 
-   b = _eglMakeCurrent(drv, dpy, draw, read, context);
+   b = _eglMakeCurrent(drv, dpy, drawSurf, readSurf, ctx);
    if (!b)
       return EGL_FALSE;
 
@@ -281,12 +248,19 @@ demoMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface rea
 }
 
 
+static void
+demoUnload(_EGLDriver *drv)
+{
+   free(drv);
+}
+
+
 /**
  * The bootstrap function.  Return a new DemoDriver object and
  * plug in API functions.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *dpy, const char *args)
+_eglMain(const char *args)
 {
    DemoDriver *demo;
 
@@ -308,9 +282,8 @@ _eglMain(_EGLDisplay *dpy, const char *args)
    demo->Base.API.DestroySurface = demoDestroySurface;
    demo->Base.API.DestroyContext = demoDestroyContext;
 
-   /* enable supported extensions */
-   demo->Base.Extensions.MESA_screen_surface = EGL_TRUE;
-   demo->Base.Extensions.MESA_copy_context = EGL_TRUE;
+   demo->Base.Name = "egl/demo";
+   demo->Base.Unload = demoUnload;
 
    return &demo->Base;
 }
diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile
index 4041d5c906..7339c97c77 100644
--- a/src/egl/drivers/dri/Makefile
+++ b/src/egl/drivers/dri/Makefile
@@ -50,11 +50,12 @@ $(TOP)/$(LIB_DIR)/libEGLdri.so: $(OBJECTS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGLdri.so $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libEGLdri.so $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o
 	-rm -f *.so
+	-rm -f depend depend.bak
 
 depend: $(SOURCES) $(HEADERS)
 	@ echo "running $(MKDEP)"
diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c
index 57661cc3ab..9e400be624 100644
--- a/src/egl/drivers/dri/egldri.c
+++ b/src/egl/drivers/dri/egldri.c
@@ -171,7 +171,10 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!c)
       return EGL_NO_CONTEXT;
 
-   if (!_eglInitContext(drv, dpy, &c->Base, config, attrib_list)) {
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
+   if (!_eglInitContext(drv, &c->Base, conf, attrib_list)) {
       free(c);
       return EGL_NO_CONTEXT;
    }
@@ -189,8 +192,6 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    else
       sharePriv = NULL;
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   assert(conf);
    _eglConfigToContextModesRec(conf, &visMode);
 
    c->driContext.private = disp->driScreen.createNewContext(disp, &visMode,
@@ -200,8 +201,8 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_FALSE;
    }
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
+   /* link to display */
+   _eglLinkContext(&c->Base, &disp->Base);
 
    return _eglGetContextHandle(&c->Base);
 }
@@ -237,14 +238,18 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
                             const EGLint *attrib_list)
 {
    driSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
 
    surf = (driSurface *) calloc(1, sizeof(*surf));
    if (!surf) {
       return EGL_NO_SURFACE;
    }
 
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
@@ -275,7 +280,7 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 #endif
    }
 
-   _eglSaveSurface(&surf->Base);
+   _eglLinkSurface(&surf->Base, _eglLookupDisplay(dpy));
 
    return surf->Base.Handle;
 }
@@ -287,16 +292,12 @@ _eglDRIDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
    driDisplay *disp = Lookup_driDisplay(dpy);
    driSurface *fs = Lookup_driSurface(surface);
 
-   _eglRemoveSurface(&fs->Base);
+   _eglUnlinkSurface(&fs->Base);
 
    fs->drawable.destroyDrawable(disp, fs->drawable.private);
 
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!_eglIsSurfaceBound(&fs->Base))
       free(fs);
-   }
    return EGL_TRUE;
 }
 
@@ -307,16 +308,12 @@ _eglDRIDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
    driDisplay *disp = Lookup_driDisplay(dpy);
    driContext *fc = Lookup_driContext(context);
 
-   _eglRemoveContext(&fc->Base);
+   _eglUnlinkContext(&fc->Base);
 
    fc->driContext.destroyContext(disp, 0, fc->driContext.private);
 
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!_eglIsContextBound(&fc->Base))
       free(fc);
-   }
    return EGL_TRUE;
 }
 
@@ -340,13 +337,13 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    }
 
    /* init base class, do error checking, etc. */
-   if (!_eglInitSurface(drv, dpy, &surface->Base, EGL_SCREEN_BIT_MESA,
-                        cfg, attrib_list)) {
+   if (!_eglInitSurface(drv, &surface->Base, EGL_SCREEN_BIT_MESA,
+                        config, attrib_list)) {
       free(surface);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(&surface->Base);
+   _eglLinkSurface(&surface->Base &disp->Base);
 
 
    /*
@@ -363,7 +360,7 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    if (!disp->driScreen.createNewDrawable(disp, &visMode, drawBuf,
                                           &surface->drawable, GLX_WINDOW_BIT,
                                           empty_attribute_list)) {
-      _eglRemoveSurface(&surface->Base);
+      _eglUnlinkSurface(&surface->Base);
       free(surface);
       return EGL_NO_SURFACE;
    }
diff --git a/src/egl/drivers/glx/Makefile b/src/egl/drivers/glx/Makefile
index 5f041a268f..20ef0352ad 100644
--- a/src/egl/drivers/glx/Makefile
+++ b/src/egl/drivers/glx/Makefile
@@ -58,7 +58,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	rm -f *.o
diff --git a/src/egl/drivers/glx/egl_glx.c b/src/egl/drivers/glx/egl_glx.c
index 155caa413c..4685f600e2 100644
--- a/src/egl/drivers/glx/egl_glx.c
+++ b/src/egl/drivers/glx/egl_glx.c
@@ -50,14 +50,11 @@
 #include <GL/gl.h>
 #include "glxclient.h"
 
-#define _EGL_PLATFORM_X
-
 #include "eglconfig.h"
 #include "eglcontext.h"
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
-#include "eglhash.h"
 #include "egllog.h"
 #include "eglsurface.h"
 
@@ -106,7 +103,13 @@ struct visual_attribs
 struct GLX_egl_driver
 {
    _EGLDriver Base;   /**< base class */
+};
 
+
+/** driver data of _EGLDisplay */
+struct GLX_egl_display
+{
+   Display *dpy;
    XVisualInfo *visuals;
    GLXFBConfig *fbconfigs;
 
@@ -136,6 +139,7 @@ struct GLX_egl_surface
 struct GLX_egl_config
 {
    _EGLConfig Base;   /**< base class */
+   int index;
 };
 
 /** cast wrapper */
@@ -145,6 +149,12 @@ GLX_egl_driver(_EGLDriver *drv)
    return (struct GLX_egl_driver *) drv;
 }
 
+static struct GLX_egl_display *
+GLX_egl_display(_EGLDisplay *dpy)
+{
+   return (struct GLX_egl_display *) dpy->DriverData;
+}
+
 static struct GLX_egl_context *
 GLX_egl_context(_EGLContext *ctx)
 {
@@ -157,6 +167,12 @@ GLX_egl_surface(_EGLSurface *surf)
    return (struct GLX_egl_surface *) surf;
 }
 
+static int
+GLX_egl_config_index(_EGLConfig *conf)
+{
+   return ((struct GLX_egl_config *) conf)->index;
+}
+
 static GLboolean
 get_visual_attribs(Display *dpy, XVisualInfo *vInfo,
                    struct visual_attribs *attribs)
@@ -267,6 +283,7 @@ glx_token_to_visual_class(int visual_type)
       return None;
    }
 }
+
 static int
 get_fbconfig_attribs(Display *dpy, GLXFBConfig fbconfig,
 		     struct visual_attribs *attribs)
@@ -345,23 +362,22 @@ get_fbconfig_attribs(Display *dpy, GLXFBConfig fbconfig,
 #endif
 
 static EGLBoolean
-create_configs(_EGLDisplay *disp, struct GLX_egl_driver *GLX_drv)
+create_configs(_EGLDisplay *disp, struct GLX_egl_display *GLX_dpy)
 {
    XVisualInfo theTemplate;
    int numVisuals;
    long mask;
    int i;
-   int egl_configs = 1;
    struct visual_attribs attribs;
 
-   GLX_drv->fbconfigs = NULL;
+   GLX_dpy->fbconfigs = NULL;
 
 #ifdef GLX_VERSION_1_3
    /* get list of all fbconfigs on this screen */
-   GLX_drv->fbconfigs = glXGetFBConfigs(disp->Xdpy, DefaultScreen(disp->Xdpy), &numVisuals);
+   GLX_dpy->fbconfigs = glXGetFBConfigs(GLX_dpy->dpy, DefaultScreen(GLX_dpy->dpy), &numVisuals);
 
    if (numVisuals == 0) {
-      GLX_drv->fbconfigs = NULL;
+      GLX_dpy->fbconfigs = NULL;
       goto xvisual;
    }
 
@@ -369,12 +385,13 @@ create_configs(_EGLDisplay *disp, struct GLX_egl_driver *GLX_drv)
       struct GLX_egl_config *config;
       int bit;
 
-      bit = get_fbconfig_attribs(disp->Xdpy, GLX_drv->fbconfigs[i], &attribs);
+      bit = get_fbconfig_attribs(GLX_dpy->dpy, GLX_dpy->fbconfigs[i], &attribs);
       if (!bit)
          continue;
 
       config = CALLOC_STRUCT(GLX_egl_config);
 
+      config->index = i;
       _eglInitConfig(&config->Base, (i+1));
       SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, attribs.id);
       SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, attribs.bufferSize);
@@ -400,18 +417,19 @@ create_configs(_EGLDisplay *disp, struct GLX_egl_driver *GLX_drv)
 
 xvisual:
    /* get list of all visuals on this screen */
-   theTemplate.screen = DefaultScreen(disp->Xdpy);
+   theTemplate.screen = DefaultScreen(GLX_dpy->dpy);
    mask = VisualScreenMask;
-   GLX_drv->visuals = XGetVisualInfo(disp->Xdpy, mask, &theTemplate, &numVisuals);
+   GLX_dpy->visuals = XGetVisualInfo(GLX_dpy->dpy, mask, &theTemplate, &numVisuals);
 
    for (i = 0; i < numVisuals; i++) {
       struct GLX_egl_config *config;
 
-      if (!get_visual_attribs(disp->Xdpy, &GLX_drv->visuals[i], &attribs))
+      if (!get_visual_attribs(GLX_dpy->dpy, &GLX_dpy->visuals[i], &attribs))
 	 continue;
 
       config = CALLOC_STRUCT(GLX_egl_config);
 
+      config->index = i;
       _eglInitConfig(&config->Base, (i+1));
       SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, attribs.id);
       SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, attribs.bufferSize);
@@ -441,76 +459,60 @@ end:
  * Called via eglInitialize(), GLX_drv->API.Initialize().
  */
 static EGLBoolean
-GLX_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
-                   EGLint *minor, EGLint *major)
+GLX_eglInitialize(_EGLDriver *drv, _EGLDisplay *disp,
+                   EGLint *major, EGLint *minor)
 {
-   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_display *GLX_dpy;
 
-   _eglLog(_EGL_DEBUG, "GLX: eglInitialize");
+   GLX_dpy = CALLOC_STRUCT(GLX_egl_display);
+   if (!GLX_dpy)
+      return _eglError(EGL_BAD_ALLOC, "eglInitialize");
 
-   if (!disp->Xdpy) {
-      disp->Xdpy = XOpenDisplay(NULL);
-      if (!disp->Xdpy) {
+   GLX_dpy->dpy = (Display *) disp->NativeDisplay;
+   if (!GLX_dpy->dpy) {
+      GLX_dpy->dpy = XOpenDisplay(NULL);
+      if (!GLX_dpy->dpy) {
          _eglLog(_EGL_WARNING, "GLX: XOpenDisplay failed");
+         free(GLX_dpy);
          return EGL_FALSE;
       }
-   } 
+   }
 
-   glXQueryVersion(disp->Xdpy, &GLX_drv->glx_maj, &GLX_drv->glx_min);
-   
-   GLX_drv->Base.Initialized = EGL_TRUE;
+   disp->DriverData = (void *) GLX_dpy;
+   disp->ClientAPIsMask = all_apis;
 
-   GLX_drv->Base.Name = "GLX";
+   glXQueryVersion(GLX_dpy->dpy, &GLX_dpy->glx_maj, &GLX_dpy->glx_min);
 
    /* we're supporting EGL 1.4 */
-   *minor = 1;
-   *major = 4;
+   *major = 1;
+   *minor = 4;
 
-   create_configs(disp, GLX_drv);
+   create_configs(disp, GLX_dpy);
 
    return EGL_TRUE;
 }
 
-/*
- * Do some clean-up that normally occurs in XCloseDisplay().
- * We do this here because we're about to unload a dynamic library
- * that has added some per-display extension data and callbacks.
- * If we don't do this here we'll crash in XCloseDisplay() because it'll
- * try to call functions that went away when the driver library was unloaded.
- */
-static void
-FreeDisplayExt(Display *dpy)
-{
-   _XExtension *ext, *next;
-
-   for (ext = dpy->ext_procs; ext; ext = next) {
-      next = ext->next;
-      if (ext->close_display) {
-         ext->close_display(dpy, &ext->codes);
-         ext->close_display = NULL;
-      }
-      if (ext->name)
-         Xfree(ext->name);
-      Xfree(ext);
-   }
-   dpy->ext_procs = NULL;
-
-   _XFreeExtData (dpy->ext_data);
-   dpy->ext_data = NULL;
-}
-
 /**
  * Called via eglTerminate(), drv->API.Terminate().
  */
 static EGLBoolean
-GLX_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+GLX_eglTerminate(_EGLDriver *drv, _EGLDisplay *disp)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
 
-   _eglLog(_EGL_DEBUG, "GLX: eglTerminate");
+   _eglReleaseDisplayResources(drv, disp);
+   _eglCleanupDisplay(disp);
 
-   FreeDisplayExt(disp->Xdpy);
+   if (GLX_dpy->visuals)
+      XFree(GLX_dpy->visuals);
+   if (GLX_dpy->fbconfigs)
+      XFree(GLX_dpy->fbconfigs);
+
+   if (!disp->NativeDisplay)
+      XCloseDisplay(GLX_dpy->dpy);
+   free(GLX_dpy);
+
+   disp->DriverData = NULL;
 
    return EGL_TRUE;
 }
@@ -519,52 +521,54 @@ GLX_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
 /**
  * Called via eglCreateContext(), drv->API.CreateContext().
  */
-static EGLContext
-GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
-                      EGLContext share_list, const EGLint *attrib_list)
+static _EGLContext *
+GLX_eglCreateContext(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
+                      _EGLContext *share_list, const EGLint *attrib_list)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    struct GLX_egl_context *GLX_ctx = CALLOC_STRUCT(GLX_egl_context);
-   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
-   struct GLX_egl_context *GLX_ctx_shared = NULL;
-   _EGLConfig *conf;
-
-   if (!GLX_ctx)
-      return EGL_NO_CONTEXT;
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
+   struct GLX_egl_context *GLX_ctx_shared = GLX_egl_context(share_list);
 
-   if (!_eglInitContext(drv, dpy, &GLX_ctx->Base, config, attrib_list)) {
-      free(GLX_ctx);
-      return EGL_NO_CONTEXT;
+   if (!GLX_ctx) {
+      _eglError(EGL_BAD_ALLOC, "eglCreateContext");
+      return NULL;
    }
 
-   if (share_list != EGL_NO_CONTEXT) {
-      _EGLContext *shareCtx = _eglLookupContext(share_list);
-      if (!shareCtx) {
-         _eglError(EGL_BAD_CONTEXT, "eglCreateContext(share_list)");
-         return EGL_FALSE;
-      }
-      GLX_ctx_shared = GLX_egl_context(shareCtx);
+   if (!_eglInitContext(drv, &GLX_ctx->Base, conf, attrib_list)) {
+      free(GLX_ctx);
+      return NULL;
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   assert(conf);
-
 #ifdef GLX_VERSION_1_3
-   if (GLX_drv->fbconfigs)
-      GLX_ctx->context = glXCreateNewContext(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], GLX_RGBA_TYPE, GLX_ctx_shared ? GLX_ctx_shared->context : NULL, GL_TRUE);
+   if (GLX_dpy->fbconfigs)
+      GLX_ctx->context =
+         glXCreateNewContext(GLX_dpy->dpy,
+                             GLX_dpy->fbconfigs[GLX_egl_config_index(conf)],
+                             GLX_RGBA_TYPE,
+                             GLX_ctx_shared ? GLX_ctx_shared->context : NULL,
+                             GL_TRUE);
    else
 #endif
-      GLX_ctx->context = glXCreateContext(disp->Xdpy, &GLX_drv->visuals[(int)config-1], GLX_ctx_shared ? GLX_ctx_shared->context : NULL, GL_TRUE);
-   if (!GLX_ctx->context)
-      return EGL_FALSE;
+      GLX_ctx->context =
+         glXCreateContext(GLX_dpy->dpy,
+                          &GLX_dpy->visuals[GLX_egl_config_index(conf)],
+                          GLX_ctx_shared ? GLX_ctx_shared->context : NULL,
+                          GL_TRUE);
+   if (!GLX_ctx->context) {
+      free(GLX_ctx);
+      return NULL;
+   }
 
 #if 1
    /* (maybe?) need to have a direct rendering context */
-   if (!glXIsDirect(disp->Xdpy, GLX_ctx->context))
-      return EGL_FALSE;
+   if (!glXIsDirect(GLX_dpy->dpy, GLX_ctx->context)) {
+      glXDestroyContext(GLX_dpy->dpy, GLX_ctx->context);
+      free(GLX_ctx);
+      return NULL;
+   }
 #endif
 
-   return _eglGetContextHandle(&GLX_ctx->Base);
+   return &GLX_ctx->Base;
 }
 
 
@@ -572,27 +576,32 @@ GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * Called via eglMakeCurrent(), drv->API.MakeCurrent().
  */
 static EGLBoolean
-GLX_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
-                    EGLSurface r, EGLContext context)
+GLX_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *dsurf,
+                   _EGLSurface *rsurf, _EGLContext *ctx)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLContext *ctx = _eglLookupContext(context);
-   _EGLSurface *dsurf = _eglLookupSurface(d);
-   _EGLSurface *rsurf = _eglLookupSurface(r);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_dsurf = GLX_egl_surface(dsurf);
    struct GLX_egl_surface *GLX_rsurf = GLX_egl_surface(rsurf);
    struct GLX_egl_context *GLX_ctx = GLX_egl_context(ctx);
+   GLXDrawable ddraw, rdraw;
+   GLXContext cctx;
 
-   if (!_eglMakeCurrent(drv, dpy, d, r, context))
+   if (!_eglMakeCurrent(drv, disp, dsurf, rsurf, ctx))
       return EGL_FALSE;
 
+   ddraw = (GLX_dsurf) ? GLX_dsurf->drawable : None;
+   rdraw = (GLX_rsurf) ? GLX_rsurf->drawable : None;
+   cctx = (GLX_ctx) ? GLX_ctx->context : NULL;
+
 #ifdef GLX_VERSION_1_3
-   if (!glXMakeContextCurrent(disp->Xdpy, GLX_dsurf ? GLX_dsurf->drawable : 0, GLX_rsurf ? GLX_rsurf->drawable : 0, GLX_ctx ? GLX_ctx->context : NULL))
+   if (glXMakeContextCurrent(GLX_dpy->dpy, ddraw, rdraw, cctx))
+      return EGL_TRUE;
 #endif
-      if (!glXMakeCurrent(disp->Xdpy, GLX_dsurf ? GLX_dsurf->drawable : 0, GLX_ctx ? GLX_ctx->context : NULL))
-         return EGL_FALSE;
 
-   return EGL_TRUE;
+   if (ddraw == rdraw && glXMakeCurrent(GLX_dpy->dpy, ddraw, cctx))
+      return EGL_TRUE;
+
+   return EGL_FALSE;
 }
 
 /** Get size of given window */
@@ -612,92 +621,106 @@ get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
 /**
  * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
  */
-static EGLSurface
-GLX_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+static _EGLSurface *
+GLX_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
                             NativeWindowType window, const EGLint *attrib_list)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_surf;
    uint width, height;
 
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
-   if (!GLX_surf)
-      return EGL_NO_SURFACE;
+   if (!GLX_surf) {
+      _eglError(EGL_BAD_ALLOC, "eglCreateWindowSurface");
+      return NULL;
+   }
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_WINDOW_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
-      return EGL_FALSE;
+      return NULL;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
-
    GLX_surf->drawable = window;
-   get_drawable_size(disp->Xdpy, window, &width, &height);
+   get_drawable_size(GLX_dpy->dpy, window, &width, &height);
    GLX_surf->Base.Width = width;
    GLX_surf->Base.Height = height;
 
-   return _eglGetSurfaceHandle(&GLX_surf->Base);
+   return &GLX_surf->Base;
 }
 
 #ifdef GLX_VERSION_1_3
-static EGLSurface
-GLX_eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, 
+static _EGLSurface *
+GLX_eglCreatePixmapSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
 			   NativePixmapType pixmap, const EGLint *attrib_list)
 {
-   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_surf;
    int i;
 
+   /* GLX must >= 1.3 */
+   if (!(GLX_dpy->glx_maj == 1 && GLX_dpy->glx_min >= 3))
+      return NULL;
+
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
-   if (!GLX_surf)
-      return EGL_NO_SURFACE;
+   if (!GLX_surf) {
+      _eglError(EGL_BAD_ALLOC, "eglCreatePixmapSurface");
+      return NULL;
+   }
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PIXMAP_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_PIXMAP_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
-      return EGL_FALSE;
+      return NULL;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
-
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
       switch (attrib_list[i]) {
          /* no attribs at this time */
       default:
          _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePixmapSurface");
-         return EGL_NO_SURFACE;
+         free(GLX_surf);
+         return NULL;
       }
    }
 
-   GLX_surf->drawable = glXCreatePixmap(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], pixmap, NULL);
+   GLX_surf->drawable =
+      glXCreatePixmap(GLX_dpy->dpy,
+                      GLX_dpy->fbconfigs[GLX_egl_config_index(conf)],
+                      pixmap, NULL);
+   if (!GLX_surf->drawable) {
+      free(GLX_surf);
+      return NULL;
+   }
 
-   return _eglGetSurfaceHandle(&GLX_surf->Base);
+   return &GLX_surf->Base;
 }
 
-static EGLSurface
-GLX_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
-                            const EGLint *attrib_list)
+static _EGLSurface *
+GLX_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *disp,
+                            _EGLConfig *conf, const EGLint *attrib_list)
 {
-   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_surf;
    int attribs[5];
    int i = 0, j = 0;
 
+   /* GLX must >= 1.3 */
+   if (!(GLX_dpy->glx_maj == 1 && GLX_dpy->glx_min >= 3))
+      return NULL;
+
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
-   if (!GLX_surf)
-      return EGL_NO_SURFACE;
+   if (!GLX_surf) {
+      _eglError(EGL_BAD_ALLOC, "eglCreatePbufferSurface");
+      return NULL;
+   }
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
-
    while(attrib_list[i] != EGL_NONE) {
       switch (attrib_list[i]) {
          case EGL_WIDTH:
@@ -713,80 +736,81 @@ GLX_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    }
    attribs[j++] = 0;
 
-   GLX_surf->drawable = glXCreatePbuffer(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], attribs);
+   GLX_surf->drawable =
+      glXCreatePbuffer(GLX_dpy->dpy,
+                       GLX_dpy->fbconfigs[GLX_egl_config_index(conf)],
+                       attribs);
+   if (!GLX_surf->drawable) {
+      free(GLX_surf);
+      return NULL;
+   }
 
-   return _eglGetSurfaceHandle(&GLX_surf->Base);
+   return &GLX_surf->Base;
 }
 #endif
 
 static EGLBoolean
-GLX_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+GLX_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *surf)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLSurface *surf = _eglLookupSurface(surface);
-   return EGL_TRUE;
-   if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->IsBound) {
-         surf->DeletePending = EGL_TRUE;
-      }
-      else {
-         free(surf);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
+   if (!_eglIsSurfaceBound(surf)) {
+      struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
+      switch (surf->Type) {
+      case EGL_PBUFFER_BIT:
+         glXDestroyPbuffer(GLX_dpy->dpy, GLX_surf->drawable);
+         break;
+      case EGL_PIXMAP_BIT:
+         glXDestroyPixmap(GLX_dpy->dpy, GLX_surf->drawable);
+         break;
+      default:
+         break;
       }
-
-      return EGL_TRUE;
-   }
-   else {
-      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
-      return EGL_FALSE;
+      free(surf);
    }
+
+   return EGL_TRUE;
 }
 
 
 static EGLBoolean
-GLX_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+GLX_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *surf,
                      EGLint buffer)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLSurface *surf = _eglLookupSurface(surface);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
 
    /* buffer ?? */
-   glXBindTexImageEXT(disp->Xdpy, GLX_surf->drawable, GLX_FRONT_LEFT_EXT, NULL);
+   glXBindTexImageEXT(GLX_dpy->dpy, GLX_surf->drawable,
+                      GLX_FRONT_LEFT_EXT, NULL);
 
    return EGL_TRUE;
 }
 
 
 static EGLBoolean
-GLX_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+GLX_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *surf,
                         EGLint buffer)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLSurface *surf = _eglLookupSurface(surface);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
    struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
 
    /* buffer ?? */
-   glXReleaseTexImageEXT(disp->Xdpy, GLX_surf->drawable, GLX_FRONT_LEFT_EXT);
+   glXReleaseTexImageEXT(GLX_dpy->dpy, GLX_surf->drawable,
+                         GLX_FRONT_LEFT_EXT);
 
    return EGL_TRUE;
 }
 
 
 static EGLBoolean
-GLX_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+GLX_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *draw)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLSurface *surf = _eglLookupSurface(draw);
-   struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
+   struct GLX_egl_display *GLX_dpy = GLX_egl_display(disp);
+   struct GLX_egl_surface *GLX_surf = GLX_egl_surface(draw);
 
    _eglLog(_EGL_DEBUG, "GLX: EGL SwapBuffers 0x%x",draw);
 
-   /* error checking step: */
-   if (!_eglSwapBuffers(drv, dpy, draw))
-      return EGL_FALSE;
-
-   glXSwapBuffers(disp->Xdpy, GLX_surf->drawable);
+   glXSwapBuffers(GLX_dpy->dpy, GLX_surf->drawable);
 
    return EGL_TRUE;
 }
@@ -802,28 +826,35 @@ GLX_eglGetProcAddress(const char *procname)
     * some point.
     */
    _EGLProc (*get_proc_addr)(const char *procname);
+   _EGLProc proc_addr;
    get_proc_addr = dlsym(NULL, "st_get_proc_address");
    if (get_proc_addr)
       return get_proc_addr(procname);
 
-   get_proc_addr = glXGetProcAddress((const GLubyte *)procname);
-   if (get_proc_addr)
-      return get_proc_addr(procname);
+   proc_addr = glXGetProcAddress((const GLubyte *)procname);
+   if (proc_addr)
+      return proc_addr;
 
    return (_EGLProc)dlsym(NULL, procname);
 }
 
 
+static void
+GLX_Unload(_EGLDriver *drv)
+{
+   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
+   free(GLX_drv);
+}
+
+
 /**
  * This is the main entrypoint into the driver, called by libEGL.
  * Create a new _EGLDriver object and init its dispatch table.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *disp, const char *args)
+_eglMain(const char *args)
 {
    struct GLX_egl_driver *GLX_drv = CALLOC_STRUCT(GLX_egl_driver);
-   char *env;
-   int maj = 0, min = 0;
 
    if (!GLX_drv)
       return NULL;
@@ -835,13 +866,8 @@ _eglMain(_EGLDisplay *disp, const char *args)
    GLX_drv->Base.API.MakeCurrent = GLX_eglMakeCurrent;
    GLX_drv->Base.API.CreateWindowSurface = GLX_eglCreateWindowSurface;
 #ifdef GLX_VERSION_1_3
-   if (GLX_drv->glx_maj == 1 && GLX_drv->glx_min >= 3) {
-      GLX_drv->Base.API.CreatePixmapSurface = GLX_eglCreatePixmapSurface;
-      GLX_drv->Base.API.CreatePbufferSurface = GLX_eglCreatePbufferSurface;
-      printf("GLX: Pbuffer and Pixmap support\n");
-   } else {
-      printf("GLX: No pbuffer or pixmap support\n");
-   }
+   GLX_drv->Base.API.CreatePixmapSurface = GLX_eglCreatePixmapSurface;
+   GLX_drv->Base.API.CreatePbufferSurface = GLX_eglCreatePbufferSurface;
 #endif
    GLX_drv->Base.API.DestroySurface = GLX_eglDestroySurface;
    GLX_drv->Base.API.BindTexImage = GLX_eglBindTexImage;
@@ -849,22 +875,8 @@ _eglMain(_EGLDisplay *disp, const char *args)
    GLX_drv->Base.API.SwapBuffers = GLX_eglSwapBuffers;
    GLX_drv->Base.API.GetProcAddress = GLX_eglGetProcAddress;
 
-   GLX_drv->Base.ClientAPIsMask = all_apis;
    GLX_drv->Base.Name = "GLX";
-
-   _eglLog(_EGL_DEBUG, "GLX: main(%s)", args);
-
-   /* set new DRI path to pick up EGL version (which doesn't contain any mesa 
-    * code), but don't override if one is already set.
-    */
-   env = getenv("LIBGL_DRIVERS_PATH");
-   if (env) {
-      if (!strstr(env, "egl")) {
-         sprintf(env, "%s/egl", env);
-         setenv("LIBGL_DRIVERS_PATH", env, 1);
-      }
-   } else
-      setenv("LIBGL_DRIVERS_PATH", DEFAULT_DRIVER_DIR"/egl", 0);
+   GLX_drv->Base.Unload = GLX_Unload;
 
    return &GLX_drv->Base;
 }
diff --git a/src/egl/drivers/xdri/Makefile b/src/egl/drivers/xdri/Makefile
index eb83867b71..4c1fc9071c 100644
--- a/src/egl/drivers/xdri/Makefile
+++ b/src/egl/drivers/xdri/Makefile
@@ -16,19 +16,24 @@ INCLUDE_DIRS = \
 	$(shell pkg-config --cflags-only-I libdrm) \
 	-I$(TOP)/include \
 	-I$(TOP)/include/GL/internal \
+	-I$(TOP)/src/mesa \
 	-I$(TOP)/src/mesa/glapi \
-	-I$(TOP)/src/mesa/drivers/dri/common \
 	-I$(TOP)/src/egl/main \
 	-I$(TOP)/src/glx/x11
 
-SOURCES = egl_xdri.c
+HEADERS = glxinit.h driinit.h
+SOURCES = egl_xdri.c glxinit.c driinit.c
+
+DRI_SOURCES = dri_common.c XF86dri.c dri2.c dri2_glx.c dri_glx.c
+DRI_SOURCES := $(addprefix ../../../glx/x11/,$(DRI_SOURCES))
+
+SOURCES += $(DRI_SOURCES)
 
 OBJECTS = $(SOURCES:.c=.o)
 
 DRM_LIB = `pkg-config --libs libdrm`
 
-MISC_LIBS = -ldl -lXext -lGL
-
+CFLAGS += -DGLX_DIRECT_RENDERING
 
 .c.o:
 	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
@@ -50,11 +55,11 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
 		-major 1 -minor 0 \
 		-L$(TOP)/$(LIB_DIR) \
 		-install $(TOP)/$(LIB_DIR) \
-		$(OBJECTS) $(DRM_LIB) $(MISC_LIBS)
+		$(OBJECTS) $(DRM_LIB) $(GL_LIB_DEPS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	rm -f *.o
diff --git a/src/egl/drivers/xdri/driinit.c b/src/egl/drivers/xdri/driinit.c
new file mode 100644
index 0000000000..12da1bcd24
--- /dev/null
+++ b/src/egl/drivers/xdri/driinit.c
@@ -0,0 +1,67 @@
+/**
+ * DRI initialization.  The DRI loaders are defined in src/glx/x11/.
+ */
+
+#include <sys/time.h>
+
+#include "glxclient.h"
+#include "driinit.h"
+
+/* for __DRI_SYSTEM_TIME extension */
+_X_HIDDEN int
+__glXGetUST(int64_t * ust)
+{
+   struct timeval tv;
+
+   if (ust == NULL) {
+      return -EFAULT;
+   }
+
+   if (gettimeofday(&tv, NULL) == 0) {
+      ust[0] = (tv.tv_sec * 1000000) + tv.tv_usec;
+      return 0;
+   }
+   else {
+      return -errno;
+   }
+}
+
+_X_HIDDEN GLboolean
+__driGetMscRateOML(__DRIdrawable * draw,
+                   int32_t * numerator, int32_t * denominator, void *private)
+{
+   return GL_FALSE;
+}
+
+/* ignore glx extensions */
+_X_HIDDEN void
+__glXEnableDirectExtension(__GLXscreenConfigs * psc, const char *name)
+{
+}
+
+_X_HIDDEN __GLXDRIdisplay *
+__driCreateDisplay(__GLXdisplayPrivate *dpyPriv, int *version)
+{
+   __GLXDRIdisplay *driDisplay;
+   int ver = 0;
+
+   /* try DRI2 first */
+   driDisplay = dri2CreateDisplay(dpyPriv->dpy);
+   if (driDisplay) {
+      /* fill in the required field */
+      dpyPriv->dri2Display = driDisplay;
+      ver = 2;
+   }
+   else {
+      /* try DRI */
+      driDisplay = driCreateDisplay(dpyPriv->dpy);
+      if (driDisplay) {
+         dpyPriv->driDisplay = driDisplay;
+         ver = 1;
+      }
+   }
+
+   if (version)
+      *version = ver;
+   return driDisplay;
+}
diff --git a/src/egl/drivers/xdri/driinit.h b/src/egl/drivers/xdri/driinit.h
new file mode 100644
index 0000000000..6ea05cebef
--- /dev/null
+++ b/src/egl/drivers/xdri/driinit.h
@@ -0,0 +1,9 @@
+#ifndef DRIINIT_INCLUDED
+#define DRIINIT_INCLUDED
+
+#include "glxclient.h"
+
+extern __GLXDRIdisplay *
+__driCreateDisplay(__GLXdisplayPrivate *dpyPriv, int *version);
+
+#endif /* DRIINIT_INCLUDED */
diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c
index 3b3e312746..518091a2d1 100644
--- a/src/egl/drivers/xdri/egl_xdri.c
+++ b/src/egl/drivers/xdri/egl_xdri.c
@@ -39,63 +39,40 @@
  * Authors: Brian Paul
  */
 
-
 #include <assert.h>
-#include <stdio.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include "dlfcn.h"
+#include <stdlib.h>
 #include <X11/Xlib.h>
-#include <GL/gl.h>
-#include "xf86dri.h"
-#include "glxclient.h"
-#include "dri_util.h"
-#include "drm_sarea.h"
 
-#define _EGL_PLATFORM_X
+#include "glxinit.h"
+#include "driinit.h"
+#include "glapi/glapi.h" /* for glapi functions */
 
 #include "eglconfig.h"
 #include "eglcontext.h"
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
-#include "eglhash.h"
 #include "egllog.h"
 #include "eglsurface.h"
 
-#include <GL/gl.h>
-
-typedef void (*glGetIntegerv_t)(GLenum, GLint *);
-typedef void (*glBindTexture_t)(GLenum, GLuint);
-typedef void (*glCopyTexImage2D_t)(GLenum, GLint, GLenum, GLint, GLint,
-                                   GLint, GLint, GLint);
-
-
 #define CALLOC_STRUCT(T)   (struct T *) calloc(1, sizeof(struct T))
 
-
 /** subclass of _EGLDriver */
 struct xdri_egl_driver
 {
    _EGLDriver Base;   /**< base class */
-
-   const char *dri_driver_name;  /**< name of DRI driver to load */
-   void *dri_driver_handle;      /**< returned by dlopen(dri_driver_name) */
-
-   __GLXdisplayPrivate *glx_priv;
+};
 
 
-   /* XXX we're not actually using these at this time: */
-   int chipset;
-   int minor;
-   int drmFD;
+/** driver data of _EGLDisplay */
+struct xdri_egl_display
+{
+   Display *dpy;
+   __GLXdisplayPrivate *dpyPriv;
+   __GLXDRIdisplay *driDisplay;
 
-   __DRIframebuffer framebuffer;
-   drm_handle_t hSAREA;
-   drmAddress pSAREA;
-   char *busID;
-   drm_magic_t magic;
+   __GLXscreenConfigs *psc;
+   EGLint scr;
 };
 
 
@@ -104,9 +81,10 @@ struct xdri_egl_context
 {
    _EGLContext Base;   /**< base class */
 
-   __DRIcontext driContext;
+   /* just enough info to create dri contexts */
+   GLXContext dummy_gc;
 
-   GLint bound_tex_object;
+   __GLXDRIcontext *driContext;
 };
 
 
@@ -115,8 +93,8 @@ struct xdri_egl_surface
 {
    _EGLSurface Base;   /**< base class */
 
-   __DRIid driDrawable;  /**< DRI surface */
-   drm_drawable_t hDrawable;
+   Drawable drawable;
+   __GLXDRIdrawable *driDrawable;
 };
 
 
@@ -131,46 +109,44 @@ struct xdri_egl_config
 
 
 /** cast wrapper */
-static struct xdri_egl_driver *
+static INLINE struct xdri_egl_driver *
 xdri_egl_driver(_EGLDriver *drv)
 {
    return (struct xdri_egl_driver *) drv;
 }
 
 
+static INLINE struct xdri_egl_display *
+lookup_display(_EGLDisplay *dpy)
+{
+   return (struct xdri_egl_display *) dpy->DriverData;
+}
+
+
 /** Map EGLSurface handle to xdri_egl_surface object */
-static struct xdri_egl_surface *
-lookup_surface(EGLSurface surf)
+static INLINE struct xdri_egl_surface *
+lookup_surface(_EGLSurface *surface)
 {
-   _EGLSurface *surface = _eglLookupSurface(surf);
    return (struct xdri_egl_surface *) surface;
 }
 
 
 /** Map EGLContext handle to xdri_egl_context object */
-static struct xdri_egl_context *
-lookup_context(EGLContext c)
+static INLINE struct xdri_egl_context *
+lookup_context(_EGLContext *context)
 {
-   _EGLContext *context = _eglLookupContext(c);
    return (struct xdri_egl_context *) context;
 }
 
-static struct xdri_egl_context *
-current_context(void)
-{
-   return (struct xdri_egl_context *) _eglGetCurrentContext();
-}
 
 /** Map EGLConfig handle to xdri_egl_config object */
-static struct xdri_egl_config *
-lookup_config(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
+static INLINE struct xdri_egl_config *
+lookup_config(_EGLConfig *conf)
 {
-   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
    return (struct xdri_egl_config *) conf;
 }
 
 
-
 /** Get size of given window */
 static Status
 get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
@@ -188,22 +164,18 @@ get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
 
 /**
  * Produce a set of EGL configs.
- * Note that we get the list of GLcontextModes from the GLX library.
- * This dependency on GLX lib will be removed someday.
  */
-static void
-create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv)
+static EGLint
+create_configs(_EGLDisplay *disp, const __GLcontextModes *m, EGLint first_id)
 {
    static const EGLint all_apis = (EGL_OPENGL_ES_BIT |
                                    EGL_OPENGL_ES2_BIT |
                                    EGL_OPENVG_BIT |
                                    EGL_OPENGL_BIT);
-   __GLXscreenConfigs *scrn = glx_priv->screenConfigs;
-   const __GLcontextModes *m;
-   int id = 1;
+   int id = first_id;
 
-   for (m = scrn->configs; m; m = m->next) {
-      /* EGL requires double-buffered configs */
+   for (; m; m = m->next) {
+      /* add double buffered visual */
       if (m->doubleBufferMode) {
          struct xdri_egl_config *config = CALLOC_STRUCT(xdri_egl_config);
 
@@ -222,9 +194,7 @@ create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv)
          SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType);
          SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis);
          SET_CONFIG_ATTRIB(&config->Base, EGL_RENDERABLE_TYPE, all_apis);
-         /* XXX only window rendering allowed ATM */
-         SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE,
-                           (EGL_WINDOW_BIT | EGL_PBUFFER_BIT));
+         SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE, EGL_WINDOW_BIT);
 
          /* XXX possibly other things to init... */
 
@@ -234,431 +204,76 @@ create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv)
          _eglAddConfig(disp, &config->Base);
       }
    }
-}
-
-
-/**
- * Called via __DRIinterfaceMethods object
- */
-static __DRIfuncPtr
-dri_get_proc_address(const char * proc_name)
-{
-   return NULL;
-}
-
-
-static void
-dri_context_modes_destroy(__GLcontextModes *modes)
-{
-   _eglLog(_EGL_DEBUG, "%s", __FUNCTION__);
-
-   while (modes) {
-      __GLcontextModes * const next = modes->next;
-      free(modes);
-      modes = next;
-   }
-}
-
-
-/**
- * Create a linked list of 'count' GLcontextModes.
- * These are used during the client/server visual negotiation phase,
- * then discarded.
- */
-static __GLcontextModes *
-dri_context_modes_create(unsigned count, size_t minimum_size)
-{
-   /* This code copied from libGLX, and modified */
-   const size_t size = (minimum_size > sizeof(__GLcontextModes))
-       ? minimum_size : sizeof(__GLcontextModes);
-   __GLcontextModes * head = NULL;
-   __GLcontextModes ** next;
-   unsigned   i;
-
-   next = & head;
-   for (i = 0 ; i < count ; i++) {
-      *next = (__GLcontextModes *) calloc(1, size);
-      if (*next == NULL) {
-	 dri_context_modes_destroy(head);
-	 head = NULL;
-	 break;
-      }
-      
-      (*next)->doubleBufferMode = 1;
-      (*next)->visualID = GLX_DONT_CARE;
-      (*next)->visualType = GLX_DONT_CARE;
-      (*next)->visualRating = GLX_NONE;
-      (*next)->transparentPixel = GLX_NONE;
-      (*next)->transparentRed = GLX_DONT_CARE;
-      (*next)->transparentGreen = GLX_DONT_CARE;
-      (*next)->transparentBlue = GLX_DONT_CARE;
-      (*next)->transparentAlpha = GLX_DONT_CARE;
-      (*next)->transparentIndex = GLX_DONT_CARE;
-      (*next)->xRenderable = GLX_DONT_CARE;
-      (*next)->fbconfigID = GLX_DONT_CARE;
-      (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
-      (*next)->bindToTextureRgb = GLX_DONT_CARE;
-      (*next)->bindToTextureRgba = GLX_DONT_CARE;
-      (*next)->bindToMipmapTexture = GLX_DONT_CARE;
-      (*next)->bindToTextureTargets = 0;
-      (*next)->yInverted = GLX_DONT_CARE;
-
-      next = & ((*next)->next);
-   }
-
-   return head;
-}
-
-
-static __DRIscreen *
-dri_find_dri_screen(__DRInativeDisplay *ndpy, int scrn)
-{
-   __GLXdisplayPrivate *priv = __glXInitialize(ndpy);
-   __GLXscreenConfigs *scrnConf = priv->screenConfigs;
-   return &scrnConf->driScreen;
-}
-
-
-static GLboolean
-dri_window_exists(__DRInativeDisplay *ndpy, __DRIid draw)
-{
-   return EGL_TRUE;
-}
-
-
-static GLboolean
-dri_create_context(__DRInativeDisplay *ndpy, int screenNum, int configID,
-                   void * contextID, drm_context_t * hw_context)
-{
-   assert(configID >= 0);
-   return XF86DRICreateContextWithConfig(ndpy, screenNum,
-                                         configID, contextID, hw_context);
-}
-
-
-static GLboolean
-dri_destroy_context(__DRInativeDisplay * ndpy, int screen, __DRIid context)
-{
-   return XF86DRIDestroyContext(ndpy, screen, context);
-}
-
-
-static GLboolean
-dri_create_drawable(__DRInativeDisplay * ndpy, int screen,
-                    __DRIid drawable, drm_drawable_t * hHWDrawable)
-{
-   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
-
-   /* Create DRI drawable for given window ID (drawable) */
-   if (!XF86DRICreateDrawable(ndpy, screen, drawable, hHWDrawable))
-      return EGL_FALSE;
-
-   return EGL_TRUE;
-}
-
-
-static GLboolean
-dri_destroy_drawable(__DRInativeDisplay * ndpy, int screen, __DRIid drawable)
-{
-   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
-   return XF86DRIDestroyDrawable(ndpy, screen, drawable);
-}
-
-
-static GLboolean
-dri_get_drawable_info(__DRInativeDisplay *ndpy, int scrn,
-                      __DRIid draw, unsigned int * index, unsigned int * stamp,
-                      int * x, int * y, int * width, int * height,
-                      int * numClipRects, drm_clip_rect_t ** pClipRects,
-                      int * backX, int * backY,
-                      int * numBackClipRects,
-                      drm_clip_rect_t ** pBackClipRects)
-{
-   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
-
-   if (!XF86DRIGetDrawableInfo(ndpy, scrn, draw, index, stamp,
-                               x, y, width, height,
-                               numClipRects, pClipRects,
-                               backX, backY,
-                               numBackClipRects, pBackClipRects)) {
-      return EGL_FALSE;
-   }
 
-   return EGL_TRUE;
+   return id;
 }
 
 
 /**
- * Table of functions exported by the loader to the driver.
+ * Called via eglInitialize(), xdri_dpy->API.Initialize().
  */
-static const __DRIinterfaceMethods interface_methods = {
-   dri_get_proc_address,
-
-   dri_context_modes_create,
-   dri_context_modes_destroy,
-
-   dri_find_dri_screen,
-   dri_window_exists,
-
-   dri_create_context,
-   dri_destroy_context,
-
-   dri_create_drawable,
-   dri_destroy_drawable,
-   dri_get_drawable_info,
-
-   NULL,/*__eglGetUST,*/
-   NULL,/*__eglGetMSCRate,*/
-};
-
-
-
 static EGLBoolean
-init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp)
+xdri_eglInitialize(_EGLDriver *drv, _EGLDisplay *dpy,
+                   EGLint *minor, EGLint *major)
 {
-   __DRIversion ddx_version;
-   __DRIversion dri_version;
-   __DRIversion drm_version;
-   drmVersionPtr version;
-   drm_handle_t  hFB;
-   int newlyopened;
-   int status;
-   int scrn = DefaultScreen(disp->Xdpy);
-
-#if 0
-   createNewScreen = (PFNCREATENEWSCREENFUNC)
-      dlsym(xdri_drv->dri_driver_handle, createNewScreenName);
-   if (!createNewScreen) {
-      _eglLog(_EGL_WARNING, "XDRI: Couldn't find %s function in the driver.",
-              createNewScreenName);
-      return EGL_FALSE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: Found %s", createNewScreenName);
-   }
-#endif
-
-   /*
-    * Get the DRI X extension version.
-    */
-   dri_version.major = 4;
-   dri_version.minor = 0;
-   dri_version.patch = 0;
-
-   if (!XF86DRIOpenConnection(disp->Xdpy, scrn,
-                              &xdri_drv->hSAREA, &xdri_drv->busID)) {
-      _eglLog(_EGL_WARNING, "XF86DRIOpenConnection failed");
-   }
-
-   xdri_drv->drmFD = drmOpenOnce(NULL, xdri_drv->busID, &newlyopened);
-   if (xdri_drv->drmFD < 0) {
-      perror("drmOpenOnce failed: ");
-      return EGL_FALSE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: drmOpenOnce returned %d", xdri_drv->drmFD);
-   }
-
-
-   if (drmGetMagic(xdri_drv->drmFD, &xdri_drv->magic)) {
-      perror("drmGetMagic failed: ");
-      return EGL_FALSE;
-   }
-
-   version = drmGetVersion(xdri_drv->drmFD);
-   if (version) {
-      drm_version.major = version->version_major;
-      drm_version.minor = version->version_minor;
-      drm_version.patch = version->version_patchlevel;
-      drmFreeVersion(version);
-      _eglLog(_EGL_DEBUG, "XDRI: Got DRM version %d.%d.%d",
-              drm_version.major,
-              drm_version.minor,
-              drm_version.patch);
-   }
-   else {
-      drm_version.major = -1;
-      drm_version.minor = -1;
-      drm_version.patch = -1;
-      _eglLog(_EGL_WARNING, "XDRI: drmGetVersion() failed");
-      return EGL_FALSE;
-   }
-
-   /* Authenticate w/ server.
-    */
-   if (!XF86DRIAuthConnection(disp->Xdpy, scrn, xdri_drv->magic)) {
-      _eglLog(_EGL_WARNING, "XDRI: XF86DRIAuthConnection() failed");
-      return EGL_FALSE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: XF86DRIAuthConnection() success");
-   }
-
-   /* Get ddx version.
-    */
-   {
-      char *driverName;
-
-      /*
-       * Get device name (like "tdfx") and the ddx version
-       * numbers.  We'll check the version in each DRI driver's
-       * "createNewScreen" function.
-       */
-      if (!XF86DRIGetClientDriverName(disp->Xdpy, scrn,
-                                     &ddx_version.major,
-                                     &ddx_version.minor,
-                                     &ddx_version.patch,
-                                     &driverName)) {
-         _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetClientDriverName failed");
-         return EGL_FALSE;
-      }
-      else {
-         _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetClientDriverName returned %s", driverName);
+   struct xdri_egl_display *xdri_dpy;
+   __GLXdisplayPrivate *dpyPriv;
+   __GLXDRIdisplay *driDisplay;
+   __GLXscreenConfigs *psc;
+   EGLint first_id = 1;
+   int scr;
+
+   xdri_dpy = CALLOC_STRUCT(xdri_egl_display);
+   if (!xdri_dpy)
+      return _eglError(EGL_BAD_ALLOC, "eglInitialize");
+
+   xdri_dpy->dpy = (Display *) dpy->NativeDisplay;
+   if (!xdri_dpy->dpy) {
+      xdri_dpy->dpy = XOpenDisplay(NULL);
+      if (!xdri_dpy->dpy) {
+         free(xdri_dpy);
+         return _eglError(EGL_NOT_INITIALIZED, "eglInitialize");
       }
    }
 
-   /* Get framebuffer info.
-    */
-   {
-      int junk;
-      if (!XF86DRIGetDeviceInfo(disp->Xdpy, scrn,
-                                &hFB,
-                                &junk,
-                                &xdri_drv->framebuffer.size,
-                                &xdri_drv->framebuffer.stride,
-                                &xdri_drv->framebuffer.dev_priv_size,
-                                &xdri_drv->framebuffer.dev_priv)) {
-         _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetDeviceInfo() failed");
-         return EGL_FALSE;
-      }
-      else {
-         _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetDeviceInfo() success");
-      }
-      xdri_drv->framebuffer.width = DisplayWidth(disp->Xdpy, scrn);
-      xdri_drv->framebuffer.height = DisplayHeight(disp->Xdpy, scrn);
-   }
-
-   /* Map the framebuffer region. (this may not be needed)
-    */
-   status = drmMap(xdri_drv->drmFD, hFB, xdri_drv->framebuffer.size, 
-                   (drmAddressPtr) &xdri_drv->framebuffer.base);
-   if (status != 0) {
-      _eglLog(_EGL_WARNING, "XDRI: drmMap(framebuffer) failed");
-      return EGL_FALSE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: drmMap(framebuffer) success");
-   }
-
-   /* Map the SAREA region.
-    */
-   status = drmMap(xdri_drv->drmFD, xdri_drv->hSAREA, SAREA_MAX, &xdri_drv->pSAREA);
-   if (status != 0) {
-      _eglLog(_EGL_WARNING, "XDRI: drmMap(sarea) failed");
-      return EGL_FALSE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: drmMap(sarea) success");
-   }
-
-   return EGL_TRUE;
-}
-
-
-/**
- * Load the DRI driver named by "xdri_drv->dri_driver_name".
- * Basically, dlopen() the library to set "xdri_drv->dri_driver_handle".
- *
- * Later, we'll call dlsym(createNewScreenName) to get a pointer to
- * the driver's createNewScreen() function which is the bootstrap function.
- *
- * \return EGL_TRUE for success, EGL_FALSE for failure
- */
-static EGLBoolean
-load_dri_driver(struct xdri_egl_driver *xdri_drv)
-{
-   char filename[100];
-   int flags = RTLD_NOW;
-
-   /* try "egl_xxx_dri.so" first */
-   snprintf(filename, sizeof(filename), "egl_%s.so", xdri_drv->dri_driver_name);
-   _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename);
-   xdri_drv->dri_driver_handle = dlopen(filename, flags);
-   if (xdri_drv->dri_driver_handle) {
-      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename);
-      return EGL_TRUE;
-   }
-   else {
-      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) fail (%s)", filename, dlerror());
+   dpyPriv = __glXInitialize(xdri_dpy->dpy);
+   if (!dpyPriv) {
+      _eglLog(_EGL_WARNING, "failed to create GLX display");
+      free(xdri_dpy);
+      return _eglError(EGL_NOT_INITIALIZED, "eglInitialize");
    }
 
-   /* try regular "xxx_dri.so" next */
-   snprintf(filename, sizeof(filename), "%s.so", xdri_drv->dri_driver_name);
-   _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename);
-   xdri_drv->dri_driver_handle = dlopen(filename, flags);
-   if (xdri_drv->dri_driver_handle) {
-      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename);
-      return EGL_TRUE;
+   driDisplay = __driCreateDisplay(dpyPriv, NULL);
+   if (!driDisplay) {
+      _eglLog(_EGL_WARNING, "failed to create DRI display");
+      free(xdri_dpy);
+      return _eglError(EGL_NOT_INITIALIZED, "eglInitialize");
    }
 
-   _eglLog(_EGL_WARNING, "XDRI Could not open %s (%s)", filename, dlerror());
-   return EGL_FALSE;
-}
-
-
-/**
- * Called via eglInitialize(), xdri_drv->API.Initialize().
- */
-static EGLBoolean
-xdri_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
-                   EGLint *minor, EGLint *major)
-{
-   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   static char name[100];
+   scr = DefaultScreen(xdri_dpy->dpy);
+   psc = &dpyPriv->screenConfigs[scr];
 
-   _eglLog(_EGL_DEBUG, "XDRI: eglInitialize");
+   xdri_dpy->dpyPriv = dpyPriv;
+   xdri_dpy->driDisplay = driDisplay;
+   xdri_dpy->psc = psc;
+   xdri_dpy->scr = scr;
 
-   if (!disp->Xdpy) {
-      disp->Xdpy = XOpenDisplay(NULL);
-      if (!disp->Xdpy) {
-         _eglLog(_EGL_WARNING, "XDRI: XOpenDisplay failed");
-         return EGL_FALSE;
-      }
+   psc->driScreen = driDisplay->createScreen(psc, scr, dpyPriv);
+   if (!psc->driScreen) {
+      _eglLog(_EGL_WARNING, "failed to create DRI screen #%d", scr);
+      free(xdri_dpy);
+      return _eglError(EGL_NOT_INITIALIZED, "eglInitialize");
    }
 
-#if 0
-   /* choose the DRI driver to load */
-   xdri_drv->dri_driver_name = _eglChooseDRMDriver(0);
-   if (!load_dri_driver(xdri_drv))
-      return EGL_FALSE;
-#else
-   (void) load_dri_driver;
-#endif
-
-#if 0
-   if (!init_drm(xdri_drv, disp))
-      return EGL_FALSE;
-#else
-   (void) init_drm;
-#endif
-
-   /*
-    * NOTE: this call to __glXInitialize() bootstraps the whole GLX/DRI
-    * interface, loads the DRI driver, etc.
-    * This replaces the load_dri_driver()  and init_drm() code above.
-    */
-   xdri_drv->glx_priv = __glXInitialize(disp->Xdpy);
-
-   create_configs(disp, xdri_drv->glx_priv);
+   /* add visuals and fbconfigs */
+   first_id = create_configs(dpy, psc->visuals, first_id);
+   create_configs(dpy, psc->configs, first_id);
 
-   xdri_drv->Base.Initialized = EGL_TRUE;
-
-   if (xdri_drv->dri_driver_name)
-      snprintf(name, sizeof(name), "X/DRI:%s", xdri_drv->dri_driver_name);
-   else
-      snprintf(name, sizeof(name), "X/DRI");
-   xdri_drv->Base.Name = name;
+   dpy->DriverData = xdri_dpy;
+   dpy->ClientAPIsMask = (EGL_OPENGL_BIT |
+                          EGL_OPENGL_ES_BIT |
+                          EGL_OPENGL_ES2_BIT |
+                          EGL_OPENVG_BIT);
 
    /* we're supporting EGL 1.4 */
    *minor = 1;
@@ -668,57 +283,37 @@ xdri_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
 }
 
 
-/*
- * Do some clean-up that normally occurs in XCloseDisplay().
- * We do this here because we're about to unload a dynamic library
- * that has added some per-display extension data and callbacks.
- * If we don't do this here we'll crash in XCloseDisplay() because it'll
- * try to call functions that went away when the driver library was unloaded.
- */
-static void
-FreeDisplayExt(Display *dpy)
-{
-   _XExtension *ext, *next;
-
-   for (ext = dpy->ext_procs; ext; ext = next) {
-      next = ext->next;
-      if (ext->close_display) {
-         ext->close_display(dpy, &ext->codes);
-         ext->close_display = NULL;
-      }
-      if (ext->name)
-         Xfree(ext->name);
-      Xfree(ext);
-   }
-   dpy->ext_procs = NULL;
-
-   _XFreeExtData (dpy->ext_data);
-   dpy->ext_data = NULL;
-}
-
-
 /**
  * Called via eglTerminate(), drv->API.Terminate().
  */
 static EGLBoolean
-xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+xdri_eglTerminate(_EGLDriver *drv, _EGLDisplay *dpy)
 {
-   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_display *xdri_dpy = lookup_display(dpy);
+   __GLXscreenConfigs *psc;
 
-   _eglLog(_EGL_DEBUG, "XDRI: eglTerminate");
+   _eglReleaseDisplayResources(drv, dpy);
+   _eglCleanupDisplay(dpy);
 
-   _eglLog(_EGL_DEBUG, "XDRI: Closing %s", xdri_drv->dri_driver_name);
-
-   FreeDisplayExt(disp->Xdpy);
+   psc = xdri_dpy->psc;
+   if (psc->driver_configs) {
+      unsigned int i;
+      for (i = 0; psc->driver_configs[i]; i++)
+         free((__DRIconfig *) psc->driver_configs[i]);
+      free(psc->driver_configs);
+      psc->driver_configs = NULL;
+   }
+   if (psc->driScreen) {
+      psc->driScreen->destroyScreen(psc);
+      free(psc->driScreen);
+      psc->driScreen = NULL;
+   }
 
-#if 0
-   /* this causes a segfault for some reason */
-   dlclose(xdri_drv->dri_driver_handle);
-#endif
-   xdri_drv->dri_driver_handle = NULL;
+   xdri_dpy->driDisplay->destroyDisplay(xdri_dpy->driDisplay);
+   __glXRelease(xdri_dpy->dpyPriv);
 
-   free((void*) xdri_drv->dri_driver_name);
+   free(xdri_dpy);
+   dpy->DriverData = NULL;
 
    return EGL_TRUE;
 }
@@ -730,71 +325,77 @@ xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
 static _EGLProc
 xdri_eglGetProcAddress(const char *procname)
 {
-#if 0
-   _EGLDriver *drv = NULL;
-
-   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-   /*_EGLDisplay *disp = _eglLookupDisplay(dpy);*/
-   _EGLProc *proc = xdri_drv->driScreen.getProcAddress(procname);
-   return proc;
-#elif 1
-   /* This is a bit of a hack to get at the gallium/Mesa state tracker
-    * function st_get_proc_address().  This will probably change at
-    * some point.
-    */
-   _EGLProc (*st_get_proc_addr)(const char *procname);
-   st_get_proc_addr = dlsym(NULL, "st_get_proc_address");
-   if (st_get_proc_addr) {
-      return st_get_proc_addr(procname);
-   }
-   return NULL;   
-#else
-   return NULL;
-#endif
+   /* the symbol is defined in libGL.so */
+   return (_EGLProc) _glapi_get_proc_address(procname);
 }
 
 
 /**
  * Called via eglCreateContext(), drv->API.CreateContext().
  */
-static EGLContext
-xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
-                      EGLContext share_list, const EGLint *attrib_list)
-{
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config);
-   void *shared = NULL;
+static _EGLContext *
+xdri_eglCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
+                      _EGLContext *share_list, const EGLint *attrib_list)
+{
+   struct xdri_egl_display *xdri_dpy = lookup_display(dpy);
+   struct xdri_egl_config *xdri_config = lookup_config(conf);
+   struct xdri_egl_context *shared = lookup_context(share_list);
+   __GLXscreenConfigs *psc = xdri_dpy->psc;
    int renderType = GLX_RGBA_BIT;
+   struct xdri_egl_context *xdri_ctx;
 
-   struct xdri_egl_context *xdri_ctx = CALLOC_STRUCT(xdri_egl_context);
-   if (!xdri_ctx)
-      return EGL_NO_CONTEXT;
+   xdri_ctx = CALLOC_STRUCT(xdri_egl_context);
+   if (!xdri_ctx) {
+      _eglError(EGL_BAD_ALLOC, "eglCreateContext");
+      return NULL;
+   }
 
-   if (!_eglInitContext(drv, dpy, &xdri_ctx->Base, config, attrib_list)) {
+   xdri_ctx->dummy_gc = CALLOC_STRUCT(__GLXcontextRec);
+   if (!xdri_ctx->dummy_gc) {
+      _eglError(EGL_BAD_ALLOC, "eglCreateContext");
       free(xdri_ctx);
-      return EGL_NO_CONTEXT;
+      return NULL;
    }
 
-   assert(xdri_config);
-
-   {
-      struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-      __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
-      xdri_ctx->driContext.private = 
-         scrnConf->driScreen.createNewContext(disp->Xdpy,
-                                              xdri_config->mode, renderType,
-                                              shared, &xdri_ctx->driContext);
+   if (!_eglInitContext(drv, &xdri_ctx->Base, &xdri_config->Base, attrib_list)) {
+      free(xdri_ctx->dummy_gc);
+      free(xdri_ctx);
+      return NULL;
    }
 
-   if (!xdri_ctx->driContext.private) {
-      _eglLog(_EGL_DEBUG, "driScreen.createNewContext failed");
+   xdri_ctx->driContext =
+      psc->driScreen->createContext(psc,
+                                    xdri_config->mode,
+                                    xdri_ctx->dummy_gc,
+                                    (shared) ? shared->dummy_gc : NULL,
+                                    renderType);
+   if (!xdri_ctx->driContext) {
+      free(xdri_ctx->dummy_gc);
       free(xdri_ctx);
-      return EGL_NO_CONTEXT;
+      return NULL;
    }
 
-   xdri_ctx->driContext.mode = xdri_config->mode;
+   /* fill in the required field */
+   xdri_ctx->dummy_gc->driContext = xdri_ctx->driContext;
 
-   return _eglGetContextHandle(&xdri_ctx->Base);
+   return &xdri_ctx->Base;
+}
+
+
+static EGLBoolean
+xdri_eglDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx)
+{
+   struct xdri_egl_display *xdri_dpy = lookup_display(dpy);
+   struct xdri_egl_context *xdri_ctx = lookup_context(ctx);
+
+   if (!_eglIsContextBound(ctx)) {
+      xdri_ctx->driContext->destroyContext(xdri_ctx->driContext,
+                                           xdri_dpy->psc, xdri_dpy->dpy);
+      free(xdri_ctx->dummy_gc);
+      free(xdri_ctx);
+   }
+
+   return EGL_TRUE;
 }
 
 
@@ -802,25 +403,32 @@ xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * Called via eglMakeCurrent(), drv->API.MakeCurrent().
  */
 static EGLBoolean
-xdri_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
-                    EGLSurface r, EGLContext context)
+xdri_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *d,
+                    _EGLSurface *r, _EGLContext *context)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    struct xdri_egl_context *xdri_ctx = lookup_context(context);
-   struct xdri_egl_surface *xdri_draw = lookup_surface(d);
-   struct xdri_egl_surface *xdri_read = lookup_surface(r);
-   __DRIid draw = xdri_draw ? xdri_draw->driDrawable : 0;
-   __DRIid read = xdri_read ? xdri_read->driDrawable : 0;
-   int scrn = DefaultScreen(disp->Xdpy);
+   struct xdri_egl_surface *draw = lookup_surface(d);
+   struct xdri_egl_surface *read = lookup_surface(r);
 
    if (!_eglMakeCurrent(drv, dpy, d, r, context))
       return EGL_FALSE;
 
+   /* the symbol is defined in libGL.so */
+   _glapi_check_multithread();
 
-   if (xdri_ctx &&
-       !xdri_ctx->driContext.bindContext(disp->Xdpy, scrn, draw, read,
-                                         &xdri_ctx->driContext)) {
-      return EGL_FALSE;
+   if (xdri_ctx) {
+      if (!xdri_ctx->driContext->bindContext(xdri_ctx->driContext,
+                                             draw->driDrawable,
+                                             read->driDrawable)) {
+         return EGL_FALSE;
+      }
+   }
+   else {
+      _EGLContext *old = _eglGetCurrentContext();
+      if (old) {
+         xdri_ctx = lookup_context(old);
+         xdri_ctx->driContext->unbindContext(xdri_ctx->driContext);
+      }
    }
 
    return EGL_TRUE;
@@ -830,309 +438,106 @@ xdri_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
 /**
  * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
  */
-static EGLSurface
-xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+static _EGLSurface *
+xdri_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                             NativeWindowType window, const EGLint *attrib_list)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_display *xdri_dpy = lookup_display(dpy);
+   struct xdri_egl_config *xdri_config = lookup_config(conf);
    struct xdri_egl_surface *xdri_surf;
-   int scrn = DefaultScreen(disp->Xdpy);
    uint width, height;
 
    xdri_surf = CALLOC_STRUCT(xdri_egl_surface);
-   if (!xdri_surf)
-      return EGL_NO_SURFACE;
+   if (!xdri_surf) {
+      _eglError(EGL_BAD_ALLOC, "eglCreateWindowSurface");
+      return NULL;
+   }
 
-   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &xdri_surf->Base, EGL_WINDOW_BIT,
+                        &xdri_config->Base, attrib_list)) {
       free(xdri_surf);
-      return EGL_FALSE;
+      return NULL;
    }
 
-   if (!XF86DRICreateDrawable(disp->Xdpy, scrn, window, &xdri_surf->hDrawable)) {
+   xdri_surf->driDrawable =
+      xdri_dpy->psc->driScreen->createDrawable(xdri_dpy->psc,
+                                               (XID) window,
+                                               (GLXDrawable) window,
+                                               xdri_config->mode);
+   if (!xdri_surf->driDrawable) {
       free(xdri_surf);
-      return EGL_FALSE;
+      return NULL;
    }
 
-   xdri_surf->driDrawable = window;
+   xdri_surf->drawable = (Drawable) window;
 
-   _eglSaveSurface(&xdri_surf->Base);
-
-   get_drawable_size(disp->Xdpy, window, &width, &height);
+   get_drawable_size(xdri_dpy->dpy, window, &width, &height);
    xdri_surf->Base.Width = width;
    xdri_surf->Base.Height = height;
 
-   _eglLog(_EGL_DEBUG,
-           "XDRI: CreateWindowSurface win 0x%x  handle %d  hDrawable %d",
-           (int) window, _eglGetSurfaceHandle(&xdri_surf->Base),
-           (int) xdri_surf->hDrawable);
-
-   return _eglGetSurfaceHandle(&xdri_surf->Base);
+   return &xdri_surf->Base;
 }
 
 
 /**
  * Called via eglCreatePbufferSurface(), drv->API.CreatePbufferSurface().
  */
-static EGLSurface
-xdri_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+static _EGLSurface *
+xdri_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                              const EGLint *attrib_list)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   struct xdri_egl_surface *xdri_surf;
-   struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config);
-   int scrn = DefaultScreen(disp->Xdpy);
-   Window window;
-
-   xdri_surf = CALLOC_STRUCT(xdri_egl_surface);
-   if (!xdri_surf)
-      return EGL_NO_SURFACE;
-
-   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
-      free(xdri_surf);
-      return EGL_FALSE;
-   }
-
-   /* Create a dummy X window */
-   {
-      Window root = RootWindow(disp->Xdpy, scrn);
-      XSetWindowAttributes attr;
-      XVisualInfo *visInfo, visTemplate;
-      unsigned mask;
-      int nvis;
-
-      visTemplate.visualid = xdri_config->mode->visualID;
-      visInfo = XGetVisualInfo(disp->Xdpy, VisualIDMask, &visTemplate, &nvis);
-      if (!visInfo) {
-         return EGL_NO_SURFACE;
-      }
-
-      attr.background_pixel = 0;
-      attr.border_pixel = 0;
-      attr.colormap = XCreateColormap(disp->Xdpy, root,
-                                      visInfo->visual, AllocNone);
-      attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
-      mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
-      
-      window = XCreateWindow(disp->Xdpy, root, 0, 0,
-                             xdri_surf->Base.Width, xdri_surf->Base.Height,
-                             0, visInfo->depth, InputOutput,
-                             visInfo->visual, mask, &attr);
-
-      /*XMapWindow(disp->Xdpy, window);*/
-      XFree(visInfo);
-
-      /* set hints and properties */
-      /*
-      sizehints.width = xdri_surf->Base.Width;
-      sizehints.height = xdri_surf->Base.Height;
-      sizehints.flags = USPosition;
-      XSetNormalHints(disp->Xdpy, window, &sizehints);
-      */
-   }
-
-   if (!XF86DRICreateDrawable(disp->Xdpy, scrn, window, &xdri_surf->hDrawable)) {
-      free(xdri_surf);
-      return EGL_FALSE;
-   }
-
-   xdri_surf->driDrawable = window;
-
-   _eglSaveSurface(&xdri_surf->Base);
-
-   _eglLog(_EGL_DEBUG,
-           "XDRI: CreatePbufferSurface handle %d  hDrawable %d",
-           _eglGetSurfaceHandle(&xdri_surf->Base),
-           (int) xdri_surf->hDrawable);
-
-   return _eglGetSurfaceHandle(&xdri_surf->Base);
+   return NULL;
 }
 
 
 
 static EGLBoolean
-xdri_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+xdri_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface)
 {
    struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
-   if (xdri_surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (xdri_surf->Base.IsBound) {
-         xdri_surf->Base.DeletePending = EGL_TRUE;
-      }
-      else {
-         /*
-         st_unreference_framebuffer(surf->Framebuffer);
-         */
-         free(xdri_surf);
-      }
-      return EGL_TRUE;
-   }
-   else {
-      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
-      return EGL_FALSE;
+
+   if (!_eglIsSurfaceBound(&xdri_surf->Base)) {
+      xdri_surf->driDrawable->destroyDrawable(xdri_surf->driDrawable);
+      free(xdri_surf);
    }
+
+   return EGL_TRUE;
 }
 
 
 static EGLBoolean
-xdri_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+xdri_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf,
                      EGLint buffer)
 {
-   typedef int (*bind_teximage)(__DRInativeDisplay *dpy,
-                                __DRIid surface, __DRIscreen *psc,
-                                int buffer, int target, int format,
-                                int level, int mipmap);
-
-   bind_teximage egl_dri_bind_teximage;
-
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-
-   struct xdri_egl_context *xdri_ctx = current_context();
-   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-   struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
-
-   __DRIid dri_surf = xdri_surf ? xdri_surf->driDrawable : 0;
-
-   __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
-   __DRIscreen *psc = &scrnConf->driScreen;
-
-   /* this call just does error checking */
-   if (!_eglBindTexImage(drv, dpy, surface, buffer)) {
-      return EGL_FALSE;
-   }
-
-   egl_dri_bind_teximage =
-      (bind_teximage) dlsym(NULL, "egl_dri_bind_teximage");
-   if (egl_dri_bind_teximage) {
-      return egl_dri_bind_teximage(disp->Xdpy, dri_surf, psc,
-                                   buffer,
-                                   xdri_surf->Base.TextureTarget,
-                                   xdri_surf->Base.TextureFormat,
-                                   xdri_surf->Base.MipmapLevel,
-                                   xdri_surf->Base.MipmapTexture);
-   }
-   else {
-      /* fallback path based on glCopyTexImage() */
-      /* Get/save currently bound 2D texobj name */
-      glGetIntegerv_t glGetIntegerv_func = 
-         (glGetIntegerv_t) dlsym(NULL, "glGetIntegerv");
-      GLint curTexObj = 0;
-      if (glGetIntegerv_func) {
-         (*glGetIntegerv_func)(GL_TEXTURE_BINDING_2D, &curTexObj);
-      }
-      xdri_ctx->bound_tex_object = curTexObj;
-   }
-
    return EGL_FALSE;
 }
 
 
 static EGLBoolean
-xdri_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+xdri_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf,
                         EGLint buffer)
 {
-   typedef int (*release_teximage)(__DRInativeDisplay *dpy,
-                                   __DRIid surface, __DRIscreen *psc,
-                                   int buffer, int target, int format,
-                                   int level, int mipmap);
-   release_teximage egl_dri_release_teximage;
-
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-
-   struct xdri_egl_context *xdri_ctx = current_context();
-   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-   struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
-
-   __DRIid dri_surf = xdri_surf ? xdri_surf->driDrawable : 0;
-
-   __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
-   __DRIscreen *psc = &scrnConf->driScreen;
-
-   /* this call just does error checking */
-   if (!_eglReleaseTexImage(drv, dpy, surface, buffer)) {
-      return EGL_FALSE;
-   }
-
-   egl_dri_release_teximage =
-      (release_teximage) dlsym(NULL, "egl_dri_release_teximage");
-   if (egl_dri_release_teximage) {
-      return egl_dri_release_teximage(disp->Xdpy, dri_surf, psc,
-                                      buffer,
-                                      xdri_surf->Base.TextureTarget,
-                                      xdri_surf->Base.TextureFormat,
-                                      xdri_surf->Base.MipmapLevel,
-                                      xdri_surf->Base.MipmapTexture);
-   }
-   else {
-      /* fallback path based on glCopyTexImage() */
-      glGetIntegerv_t glGetIntegerv_func = 
-         (glGetIntegerv_t) dlsym(NULL, "glGetIntegerv");
-      glBindTexture_t glBindTexture_func = 
-         (glBindTexture_t) dlsym(NULL, "glBindTexture");
-      glCopyTexImage2D_t glCopyTexImage2D_func = 
-         (glCopyTexImage2D_t) dlsym(NULL, "glCopyTexImage2D");
-      GLint curTexObj;
-      GLenum intFormat;
-      GLint level, width, height;
-
-      if (xdri_surf->Base.TextureFormat == EGL_TEXTURE_RGBA)
-         intFormat = GL_RGBA;
-      else
-         intFormat = GL_RGB;
-      level = xdri_surf->Base.MipmapLevel;
-      width = xdri_surf->Base.Width >> level;
-      height = xdri_surf->Base.Height >> level;
-
-      if (width > 0 && height > 0 &&
-          glGetIntegerv_func && glBindTexture_func && glCopyTexImage2D_func) {
-         glGetIntegerv_func(GL_TEXTURE_BINDING_2D, &curTexObj);
-         /* restore texobj from time of eglBindTexImage() call */
-         if (curTexObj != xdri_ctx->bound_tex_object)
-            glBindTexture_func(GL_TEXTURE_2D, xdri_ctx->bound_tex_object);
-         /* copy pbuffer image to texture */
-         glCopyTexImage2D_func(GL_TEXTURE_2D,
-                               level,
-                               intFormat,
-                               0, 0, width, height, 0);
-         /* restore current texture */
-         if (curTexObj != xdri_ctx->bound_tex_object)
-            glBindTexture_func(GL_TEXTURE_2D, curTexObj);
-      }
-      xdri_ctx->bound_tex_object = -1;
-   }
-
    return EGL_FALSE;
 }
 
 
 static EGLBoolean
-xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+xdri_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_display *xdri_dpy = lookup_display(dpy);
+   struct xdri_egl_surface *xdri_surf = lookup_surface(draw);
 
-   _eglLog(_EGL_DEBUG, "XDRI: EGL SwapBuffers");
+   xdri_dpy->psc->driScreen->swapBuffers(xdri_surf->driDrawable);
 
-   /* error checking step: */
-   if (!_eglSwapBuffers(drv, dpy, draw))
-      return EGL_FALSE;
+   return EGL_TRUE;
+}
 
-   {
-      struct xdri_egl_surface *xdri_surf = lookup_surface(draw);
-      struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
-      __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
-      __DRIscreen *psc = &scrnConf->driScreen;
-      __DRIdrawable * const pdraw = psc->getDrawable(disp->Xdpy,
-                                                     xdri_surf->driDrawable,
-                                                     psc->private);
-
-      if (pdraw)
-         pdraw->swapBuffers(disp->Xdpy, pdraw->private);
-      else
-         _eglLog(_EGL_WARNING, "pdraw is null in SwapBuffers");
-   }
 
-   return EGL_TRUE;
+static void
+xdri_Unload(_EGLDriver *drv)
+{
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   free(xdri_drv);
 }
 
 
@@ -1141,15 +546,12 @@ xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
  * Create a new _EGLDriver object and init its dispatch table.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *disp, const char *args)
+_eglMain(const char *args)
 {
    struct xdri_egl_driver *xdri_drv = CALLOC_STRUCT(xdri_egl_driver);
    if (!xdri_drv)
       return NULL;
 
-   /* Tell libGL to prefer the EGL drivers over regular DRI drivers */
-   __glXPreferEGL(1);
-
    _eglInitDriverFallbacks(&xdri_drv->Base);
    xdri_drv->Base.API.Initialize = xdri_eglInitialize;
    xdri_drv->Base.API.Terminate = xdri_eglTerminate;
@@ -1157,6 +559,7 @@ _eglMain(_EGLDisplay *disp, const char *args)
    xdri_drv->Base.API.GetProcAddress = xdri_eglGetProcAddress;
 
    xdri_drv->Base.API.CreateContext = xdri_eglCreateContext;
+   xdri_drv->Base.API.DestroyContext = xdri_eglDestroyContext;
    xdri_drv->Base.API.MakeCurrent = xdri_eglMakeCurrent;
    xdri_drv->Base.API.CreateWindowSurface = xdri_eglCreateWindowSurface;
    xdri_drv->Base.API.CreatePbufferSurface = xdri_eglCreatePbufferSurface;
@@ -1165,13 +568,8 @@ _eglMain(_EGLDisplay *disp, const char *args)
    xdri_drv->Base.API.ReleaseTexImage = xdri_eglReleaseTexImage;
    xdri_drv->Base.API.SwapBuffers = xdri_eglSwapBuffers;
 
-   xdri_drv->Base.ClientAPIsMask = (EGL_OPENGL_BIT |
-                                    EGL_OPENGL_ES_BIT |
-                                    EGL_OPENGL_ES2_BIT |
-                                    EGL_OPENVG_BIT);
    xdri_drv->Base.Name = "X/DRI";
-
-   _eglLog(_EGL_DEBUG, "XDRI: main(%s)", args);
+   xdri_drv->Base.Unload = xdri_Unload;
 
    return &xdri_drv->Base;
 }
diff --git a/src/egl/drivers/xdri/glxinit.c b/src/egl/drivers/xdri/glxinit.c
new file mode 100644
index 0000000000..7775009394
--- /dev/null
+++ b/src/egl/drivers/xdri/glxinit.c
@@ -0,0 +1,626 @@
+/**
+ * GLX initialization.  Code based on glxext.c, glx_query.c, and
+ * glcontextmodes.c under src/glx/x11/.  The major difference is that no DRI
+ * related code here.
+ *
+ */
+
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/Xproto.h>
+#include <X11/extensions/Xext.h>
+#include <X11/extensions/extutil.h>
+#include <sys/time.h>
+
+#include "glxinit.h"
+
+typedef struct GLXGenericGetString
+{
+   CARD8 reqType;
+   CARD8 glxCode;
+   CARD16 length B16;
+   CARD32 for_whom B32;
+   CARD32 name B32;
+} xGLXGenericGetStringReq;
+
+#define sz_xGLXGenericGetStringReq 12
+#define X_GLXGenericGetString 0
+
+/* Extension required boiler plate */
+
+static char *__glXExtensionName = GLX_EXTENSION_NAME;
+static XExtensionInfo *__glXExtensionInfo = NULL;
+
+static /* const */ XExtensionHooks __glXExtensionHooks = { NULL };
+static
+XEXT_GENERATE_FIND_DISPLAY(__glXFindDisplay, __glXExtensionInfo,
+                           __glXExtensionName, &__glXExtensionHooks,
+                           __GLX_NUMBER_EVENTS, NULL)
+
+static GLint
+_gl_convert_from_x_visual_type(int visualType)
+{
+#define NUM_VISUAL_TYPES   6
+   static const int glx_visual_types[NUM_VISUAL_TYPES] = {
+      GLX_STATIC_GRAY, GLX_GRAY_SCALE,
+      GLX_STATIC_COLOR, GLX_PSEUDO_COLOR,
+      GLX_TRUE_COLOR, GLX_DIRECT_COLOR
+   };
+
+   return ((unsigned) visualType < NUM_VISUAL_TYPES)
+      ? glx_visual_types[visualType] : GLX_NONE;
+}
+
+static __GLcontextModes *
+_gl_context_modes_create(unsigned count, size_t minimum_size)
+{
+   const size_t size = (minimum_size > sizeof(__GLcontextModes))
+      ? minimum_size : sizeof(__GLcontextModes);
+   __GLcontextModes *base = NULL;
+   __GLcontextModes **next;
+   unsigned i;
+
+   next = &base;
+   for (i = 0; i < count; i++) {
+      *next = (__GLcontextModes *) Xmalloc(size);
+      if (*next == NULL) {
+         _gl_context_modes_destroy(base);
+         base = NULL;
+         break;
+      }
+
+      memset(*next, 0, size);
+      (*next)->visualID = GLX_DONT_CARE;
+      (*next)->visualType = GLX_DONT_CARE;
+      (*next)->visualRating = GLX_NONE;
+      (*next)->transparentPixel = GLX_NONE;
+      (*next)->transparentRed = GLX_DONT_CARE;
+      (*next)->transparentGreen = GLX_DONT_CARE;
+      (*next)->transparentBlue = GLX_DONT_CARE;
+      (*next)->transparentAlpha = GLX_DONT_CARE;
+      (*next)->transparentIndex = GLX_DONT_CARE;
+      (*next)->xRenderable = GLX_DONT_CARE;
+      (*next)->fbconfigID = GLX_DONT_CARE;
+      (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
+      (*next)->bindToTextureRgb = GLX_DONT_CARE;
+      (*next)->bindToTextureRgba = GLX_DONT_CARE;
+      (*next)->bindToMipmapTexture = GLX_DONT_CARE;
+      (*next)->bindToTextureTargets = GLX_DONT_CARE;
+      (*next)->yInverted = GLX_DONT_CARE;
+
+      next = &((*next)->next);
+   }
+
+   return base;
+}
+
+_X_HIDDEN void
+_gl_context_modes_destroy(__GLcontextModes * modes)
+{
+   while (modes != NULL) {
+      __GLcontextModes *const next = modes->next;
+
+      Xfree(modes);
+      modes = next;
+   }
+}
+
+_X_HIDDEN char *
+__glXQueryServerString(Display * dpy, int opcode, CARD32 screen, CARD32 name)
+{
+   xGLXGenericGetStringReq *req;
+   xGLXSingleReply reply;
+   int length;
+   int numbytes;
+   char *buf;
+   CARD32 for_whom = screen;
+   CARD32 glxCode = X_GLXQueryServerString;
+
+
+   LockDisplay(dpy);
+
+
+   /* All of the GLX protocol requests for getting a string from the server
+    * look the same.  The exact meaning of the for_whom field is usually
+    * either the screen number (for glXQueryServerString) or the context tag
+    * (for GLXSingle).
+    */
+
+   GetReq(GLXGenericGetString, req);
+   req->reqType = opcode;
+   req->glxCode = glxCode;
+   req->for_whom = for_whom;
+   req->name = name;
+
+   _XReply(dpy, (xReply *) & reply, 0, False);
+
+   length = reply.length * 4;
+   numbytes = reply.size;
+
+   buf = (char *) Xmalloc(numbytes);
+   if (buf != NULL) {
+      _XRead(dpy, buf, numbytes);
+      length -= numbytes;
+   }
+
+   _XEatData(dpy, length);
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return buf;
+}
+
+/************************************************************************/
+/*
+** Free the per screen configs data as well as the array of
+** __glXScreenConfigs.
+*/
+static void
+FreeScreenConfigs(__GLXdisplayPrivate * priv)
+{
+   __GLXscreenConfigs *psc;
+   GLint i, screens;
+
+   /* Free screen configuration information */
+   psc = priv->screenConfigs;
+   screens = ScreenCount(priv->dpy);
+   for (i = 0; i < screens; i++, psc++) {
+      if (psc->configs) {
+         _gl_context_modes_destroy(psc->configs);
+         psc->configs = NULL;   /* NOTE: just for paranoia */
+      }
+      if (psc->visuals) {
+         _gl_context_modes_destroy(psc->visuals);
+         psc->visuals = NULL;   /* NOTE: just for paranoia */
+      }
+      Xfree((char *) psc->serverGLXexts);
+   }
+   XFree((char *) priv->screenConfigs);
+   priv->screenConfigs = NULL;
+}
+
+/************************************************************************/
+
+/*
+** Query the version of the GLX extension.  This procedure works even if
+** the client extension is not completely set up.
+*/
+static Bool
+QueryVersion(Display * dpy, int opcode, int *major, int *minor)
+{
+   xGLXQueryVersionReq *req;
+   xGLXQueryVersionReply reply;
+
+   /* Send the glXQueryVersion request */
+   LockDisplay(dpy);
+   GetReq(GLXQueryVersion, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXQueryVersion;
+   req->majorVersion = GLX_MAJOR_VERSION;
+   req->minorVersion = GLX_MINOR_VERSION;
+   _XReply(dpy, (xReply *) & reply, 0, False);
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   if (reply.majorVersion != GLX_MAJOR_VERSION) {
+      /*
+       ** The server does not support the same major release as this
+       ** client.
+       */
+      return GL_FALSE;
+   }
+   *major = reply.majorVersion;
+   *minor = min(reply.minorVersion, GLX_MINOR_VERSION);
+   return GL_TRUE;
+}
+
+_X_HIDDEN void
+__glXInitializeVisualConfigFromTags(__GLcontextModes * config, int count,
+                                    const INT32 * bp, Bool tagged_only,
+                                    Bool fbconfig_style_tags)
+{
+   int i;
+
+   if (!tagged_only) {
+      /* Copy in the first set of properties */
+      config->visualID = *bp++;
+
+      config->visualType = _gl_convert_from_x_visual_type(*bp++);
+
+      config->rgbMode = *bp++;
+
+      config->redBits = *bp++;
+      config->greenBits = *bp++;
+      config->blueBits = *bp++;
+      config->alphaBits = *bp++;
+      config->accumRedBits = *bp++;
+      config->accumGreenBits = *bp++;
+      config->accumBlueBits = *bp++;
+      config->accumAlphaBits = *bp++;
+
+      config->doubleBufferMode = *bp++;
+      config->stereoMode = *bp++;
+
+      config->rgbBits = *bp++;
+      config->depthBits = *bp++;
+      config->stencilBits = *bp++;
+      config->numAuxBuffers = *bp++;
+      config->level = *bp++;
+
+      count -= __GLX_MIN_CONFIG_PROPS;
+   }
+
+   /*
+    ** Additional properties may be in a list at the end
+    ** of the reply.  They are in pairs of property type
+    ** and property value.
+    */
+
+#define FETCH_OR_SET(tag) \
+    config-> tag = ( fbconfig_style_tags ) ? *bp++ : 1
+
+   for (i = 0; i < count; i += 2) {
+      switch (*bp++) {
+      case GLX_RGBA:
+         FETCH_OR_SET(rgbMode);
+         break;
+      case GLX_BUFFER_SIZE:
+         config->rgbBits = *bp++;
+         break;
+      case GLX_LEVEL:
+         config->level = *bp++;
+         break;
+      case GLX_DOUBLEBUFFER:
+         FETCH_OR_SET(doubleBufferMode);
+         break;
+      case GLX_STEREO:
+         FETCH_OR_SET(stereoMode);
+         break;
+      case GLX_AUX_BUFFERS:
+         config->numAuxBuffers = *bp++;
+         break;
+      case GLX_RED_SIZE:
+         config->redBits = *bp++;
+         break;
+      case GLX_GREEN_SIZE:
+         config->greenBits = *bp++;
+         break;
+      case GLX_BLUE_SIZE:
+         config->blueBits = *bp++;
+         break;
+      case GLX_ALPHA_SIZE:
+         config->alphaBits = *bp++;
+         break;
+      case GLX_DEPTH_SIZE:
+         config->depthBits = *bp++;
+         break;
+      case GLX_STENCIL_SIZE:
+         config->stencilBits = *bp++;
+         break;
+      case GLX_ACCUM_RED_SIZE:
+         config->accumRedBits = *bp++;
+         break;
+      case GLX_ACCUM_GREEN_SIZE:
+         config->accumGreenBits = *bp++;
+         break;
+      case GLX_ACCUM_BLUE_SIZE:
+         config->accumBlueBits = *bp++;
+         break;
+      case GLX_ACCUM_ALPHA_SIZE:
+         config->accumAlphaBits = *bp++;
+         break;
+      case GLX_VISUAL_CAVEAT_EXT:
+         config->visualRating = *bp++;
+         break;
+      case GLX_X_VISUAL_TYPE:
+         config->visualType = *bp++;
+         break;
+      case GLX_TRANSPARENT_TYPE:
+         config->transparentPixel = *bp++;
+         break;
+      case GLX_TRANSPARENT_INDEX_VALUE:
+         config->transparentIndex = *bp++;
+         break;
+      case GLX_TRANSPARENT_RED_VALUE:
+         config->transparentRed = *bp++;
+         break;
+      case GLX_TRANSPARENT_GREEN_VALUE:
+         config->transparentGreen = *bp++;
+         break;
+      case GLX_TRANSPARENT_BLUE_VALUE:
+         config->transparentBlue = *bp++;
+         break;
+      case GLX_TRANSPARENT_ALPHA_VALUE:
+         config->transparentAlpha = *bp++;
+         break;
+      case GLX_VISUAL_ID:
+         config->visualID = *bp++;
+         break;
+      case GLX_DRAWABLE_TYPE:
+         config->drawableType = *bp++;
+         break;
+      case GLX_RENDER_TYPE:
+         config->renderType = *bp++;
+         break;
+      case GLX_X_RENDERABLE:
+         config->xRenderable = *bp++;
+         break;
+      case GLX_FBCONFIG_ID:
+         config->fbconfigID = *bp++;
+         break;
+      case GLX_MAX_PBUFFER_WIDTH:
+         config->maxPbufferWidth = *bp++;
+         break;
+      case GLX_MAX_PBUFFER_HEIGHT:
+         config->maxPbufferHeight = *bp++;
+         break;
+      case GLX_MAX_PBUFFER_PIXELS:
+         config->maxPbufferPixels = *bp++;
+         break;
+      case GLX_OPTIMAL_PBUFFER_WIDTH_SGIX:
+         config->optimalPbufferWidth = *bp++;
+         break;
+      case GLX_OPTIMAL_PBUFFER_HEIGHT_SGIX:
+         config->optimalPbufferHeight = *bp++;
+         break;
+      case GLX_VISUAL_SELECT_GROUP_SGIX:
+         config->visualSelectGroup = *bp++;
+         break;
+      case GLX_SWAP_METHOD_OML:
+         config->swapMethod = *bp++;
+         break;
+      case GLX_SAMPLE_BUFFERS_SGIS:
+         config->sampleBuffers = *bp++;
+         break;
+      case GLX_SAMPLES_SGIS:
+         config->samples = *bp++;
+         break;
+      case GLX_BIND_TO_TEXTURE_RGB_EXT:
+         config->bindToTextureRgb = *bp++;
+         break;
+      case GLX_BIND_TO_TEXTURE_RGBA_EXT:
+         config->bindToTextureRgba = *bp++;
+         break;
+      case GLX_BIND_TO_MIPMAP_TEXTURE_EXT:
+         config->bindToMipmapTexture = *bp++;
+         break;
+      case GLX_BIND_TO_TEXTURE_TARGETS_EXT:
+         config->bindToTextureTargets = *bp++;
+         break;
+      case GLX_Y_INVERTED_EXT:
+         config->yInverted = *bp++;
+         break;
+      case None:
+         i = count;
+         break;
+      default:
+         break;
+      }
+   }
+
+   config->renderType =
+      (config->rgbMode) ? GLX_RGBA_BIT : GLX_COLOR_INDEX_BIT;
+
+   config->haveAccumBuffer = ((config->accumRedBits +
+                               config->accumGreenBits +
+                               config->accumBlueBits +
+                               config->accumAlphaBits) > 0);
+   config->haveDepthBuffer = (config->depthBits > 0);
+   config->haveStencilBuffer = (config->stencilBits > 0);
+}
+
+static __GLcontextModes *
+createConfigsFromProperties(Display * dpy, int nvisuals, int nprops,
+                            int screen, GLboolean tagged_only)
+{
+   INT32 buf[__GLX_TOTAL_CONFIG], *props;
+   unsigned prop_size;
+   __GLcontextModes *modes, *m;
+   int i;
+
+   if (nprops == 0)
+      return NULL;
+
+   /* FIXME: Is the __GLX_MIN_CONFIG_PROPS test correct for FBconfigs? */
+
+   /* Check number of properties */
+   if (nprops < __GLX_MIN_CONFIG_PROPS || nprops > __GLX_MAX_CONFIG_PROPS)
+      return NULL;
+
+   /* Allocate memory for our config structure */
+   modes = _gl_context_modes_create(nvisuals, sizeof(__GLcontextModes));
+   if (!modes)
+      return NULL;
+
+   prop_size = nprops * __GLX_SIZE_INT32;
+   if (prop_size <= sizeof(buf))
+      props = buf;
+   else
+      props = Xmalloc(prop_size);
+
+   /* Read each config structure and convert it into our format */
+   m = modes;
+   for (i = 0; i < nvisuals; i++) {
+      _XRead(dpy, (char *) props, prop_size);
+      /* Older X servers don't send this so we default it here. */
+      m->drawableType = GLX_WINDOW_BIT;
+      __glXInitializeVisualConfigFromTags(m, nprops, props,
+                                     tagged_only, GL_TRUE);
+      m->screen = screen;
+      m = m->next;
+   }
+
+   if (props != buf)
+      Xfree(props);
+
+   return modes;
+}
+
+static GLboolean
+getVisualConfigs(Display * dpy, __GLXdisplayPrivate * priv, int screen)
+{
+   xGLXGetVisualConfigsReq *req;
+   __GLXscreenConfigs *psc;
+   xGLXGetVisualConfigsReply reply;
+
+   LockDisplay(dpy);
+
+   psc = priv->screenConfigs + screen;
+   psc->visuals = NULL;
+   GetReq(GLXGetVisualConfigs, req);
+   req->reqType = priv->majorOpcode;
+   req->glxCode = X_GLXGetVisualConfigs;
+   req->screen = screen;
+
+   if (!_XReply(dpy, (xReply *) & reply, 0, False))
+      goto out;
+
+   psc->visuals = createConfigsFromProperties(dpy,
+                                              reply.numVisuals,
+                                              reply.numProps,
+                                              screen, GL_FALSE);
+
+ out:
+   UnlockDisplay(dpy);
+   return psc->visuals != NULL;
+}
+
+static GLboolean
+getFBConfigs(Display * dpy, __GLXdisplayPrivate * priv, int screen)
+{
+   xGLXGetFBConfigsReq *fb_req;
+   xGLXGetFBConfigsSGIXReq *sgi_req;
+   xGLXVendorPrivateWithReplyReq *vpreq;
+   xGLXGetFBConfigsReply reply;
+   __GLXscreenConfigs *psc;
+
+   psc = priv->screenConfigs + screen;
+   psc->serverGLXexts =
+      __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
+
+   LockDisplay(dpy);
+
+   psc->configs = NULL;
+   if (atof(priv->serverGLXversion) >= 1.3) {
+      GetReq(GLXGetFBConfigs, fb_req);
+      fb_req->reqType = priv->majorOpcode;
+      fb_req->glxCode = X_GLXGetFBConfigs;
+      fb_req->screen = screen;
+   }
+   else if (strstr(psc->serverGLXexts, "GLX_SGIX_fbconfig") != NULL) {
+      GetReqExtra(GLXVendorPrivateWithReply,
+                  sz_xGLXGetFBConfigsSGIXReq +
+                  sz_xGLXVendorPrivateWithReplyReq, vpreq);
+      sgi_req = (xGLXGetFBConfigsSGIXReq *) vpreq;
+      sgi_req->reqType = priv->majorOpcode;
+      sgi_req->glxCode = X_GLXVendorPrivateWithReply;
+      sgi_req->vendorCode = X_GLXvop_GetFBConfigsSGIX;
+      sgi_req->screen = screen;
+   }
+   else
+      goto out;
+
+   if (!_XReply(dpy, (xReply *) & reply, 0, False))
+      goto out;
+
+   psc->configs = createConfigsFromProperties(dpy,
+                                              reply.numFBConfigs,
+                                              reply.numAttribs * 2,
+                                              screen, GL_TRUE);
+
+ out:
+   UnlockDisplay(dpy);
+   return psc->configs != NULL;
+}
+
+static GLboolean
+AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
+{
+   __GLXscreenConfigs *psc;
+   GLint i, screens;
+
+   /*
+    ** First allocate memory for the array of per screen configs.
+    */
+   screens = ScreenCount(dpy);
+   psc = (__GLXscreenConfigs *) Xmalloc(screens * sizeof(__GLXscreenConfigs));
+   if (!psc) {
+      return GL_FALSE;
+   }
+   memset(psc, 0, screens * sizeof(__GLXscreenConfigs));
+   priv->screenConfigs = psc;
+
+   priv->serverGLXversion =
+      __glXQueryServerString(dpy, priv->majorOpcode, 0, GLX_VERSION);
+   if (priv->serverGLXversion == NULL) {
+      FreeScreenConfigs(priv);
+      return GL_FALSE;
+   }
+
+   for (i = 0; i < screens; i++, psc++) {
+      getFBConfigs(dpy, priv, i);
+      getVisualConfigs(dpy, priv, i);
+      psc->scr = i;
+      psc->dpy = dpy;
+   }
+
+   SyncHandle();
+
+   return GL_TRUE;
+}
+
+_X_HIDDEN void
+__glXRelease(__GLXdisplayPrivate *dpyPriv)
+{
+   FreeScreenConfigs(dpyPriv);
+
+   if (dpyPriv->serverGLXvendor) {
+      Xfree((char *) dpyPriv->serverGLXvendor);
+      dpyPriv->serverGLXvendor = NULL;
+   }
+   if (dpyPriv->serverGLXversion) {
+      Xfree((char *) dpyPriv->serverGLXversion);
+      dpyPriv->serverGLXversion = NULL;
+   }
+
+   Xfree(dpyPriv);
+}
+
+_X_HIDDEN __GLXdisplayPrivate *
+__glXInitialize(Display * dpy)
+{
+   XExtDisplayInfo *info = __glXFindDisplay(dpy);
+   __GLXdisplayPrivate *dpyPriv;
+   int major, minor;
+
+   if (!XextHasExtension(info))
+      return NULL;
+
+   /* See if the versions are compatible */
+   if (!QueryVersion(dpy, info->codes->major_opcode, &major, &minor))
+      return NULL;
+
+   dpyPriv = (__GLXdisplayPrivate *) Xcalloc(1, sizeof(__GLXdisplayPrivate));
+   if (!dpyPriv)
+      return NULL;
+
+   /*
+    ** Init the display private and then read in the screen config
+    ** structures from the server.
+    */
+   dpyPriv->majorOpcode = info->codes->major_opcode;
+   dpyPriv->majorVersion = major;
+   dpyPriv->minorVersion = minor;
+   dpyPriv->dpy = dpy;
+
+   dpyPriv->serverGLXvendor = NULL;
+   dpyPriv->serverGLXversion = NULL;
+
+   if (!AllocAndFetchScreenConfigs(dpy, dpyPriv)) {
+      Xfree(dpyPriv);
+      return NULL;
+   }
+
+   return dpyPriv;
+}
diff --git a/src/egl/drivers/xdri/glxinit.h b/src/egl/drivers/xdri/glxinit.h
new file mode 100644
index 0000000000..57206e627b
--- /dev/null
+++ b/src/egl/drivers/xdri/glxinit.h
@@ -0,0 +1,14 @@
+#ifndef GLXINIT_INCLUDED
+#define GLXINIT_INCLUDED
+
+#include <X11/Xlib.h>
+#include "glxclient.h"
+
+/* this is used by DRI loaders */
+extern void
+_gl_context_modes_destroy(__GLcontextModes * modes);
+
+extern void
+__glXRelease(__GLXdisplayPrivate *dpyPriv);
+
+#endif /* GLXINIT_INCLUDED */
diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile
index fab8f16089..c951b070f1 100644
--- a/src/egl/main/Makefile
+++ b/src/egl/main/Makefile
@@ -7,38 +7,38 @@ include $(TOP)/configs/current
 INCLUDE_DIRS = -I$(TOP)/include -I$(TOP)/src/mesa/glapi $(X11_INCLUDES)
 
 HEADERS = \
+	eglcompiler.h \
 	eglconfig.h \
 	eglconfigutil.h \
 	eglcontext.h \
+	eglcurrent.h \
 	egldefines.h \
 	egldisplay.h \
 	egldriver.h \
 	eglglobals.h \
 	egllog.h \
-	eglhash.h \
 	eglmisc.h \
 	eglmode.h \
+	eglmutex.h \
 	eglscreen.h \
 	eglstring.h \
-	eglsurface.h \
-	eglx.h
+	eglsurface.h
 
 SOURCES = \
 	eglapi.c \
 	eglconfig.c \
 	eglconfigutil.c \
 	eglcontext.c \
+	eglcurrent.c \
 	egldisplay.c \
 	egldriver.c \
 	eglglobals.c \
 	egllog.c \
-	eglhash.c \
 	eglmisc.c \
 	eglmode.c \
 	eglscreen.c \
 	eglstring.c \
-	eglsurface.c \
-	eglx.c
+	eglsurface.c
 
 OBJECTS = $(SOURCES:.c=.o)
 
@@ -66,11 +66,12 @@ $(TOP)/$(LIB_DIR)/libEGL.so: $(OBJECTS)
 
 install: default
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGL.so* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libEGL.so* $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o *.so*
 	-rm -f core.*
+	-rm -f depend depend.bak
 
 
 depend: $(SOURCES) $(HEADERS)
diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c
index 9df938e188..29617b7aff 100644
--- a/src/egl/main/eglapi.c
+++ b/src/egl/main/eglapi.c
@@ -29,7 +29,6 @@
  */
 
 
-
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -39,7 +38,9 @@
 #include "eglglobals.h"
 #include "egldriver.h"
 #include "eglsurface.h"
-
+#include "eglconfig.h"
+#include "eglscreen.h"
+#include "eglmode.h"
 
 
 /**
@@ -50,8 +51,12 @@ EGLDisplay EGLAPIENTRY
 eglGetDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy;
-   _eglInitGlobals();
-   dpy = _eglNewDisplay(nativeDisplay);
+   dpy = _eglFindDisplay(nativeDisplay);
+   if (!dpy) {
+      dpy = _eglNewDisplay(nativeDisplay);
+      if (dpy)
+         _eglLinkDisplay(dpy);
+   }
    return _eglGetDisplayHandle(dpy);
 }
 
@@ -63,232 +68,484 @@ eglGetDisplay(NativeDisplayType nativeDisplay)
 EGLBoolean EGLAPIENTRY
 eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor)
 {
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLDriver *drv;
    EGLint major_int, minor_int;
 
-   if (dpy) {
-      EGLBoolean retVal;
-      _EGLDisplay *dpyPriv = _eglLookupDisplay(dpy);
-      if (!dpyPriv) {
-         return EGL_FALSE;
-      }
-      dpyPriv->Driver = _eglOpenDriver(dpyPriv,
-                                       dpyPriv->DriverName,
-                                       dpyPriv->DriverArgs);
-      if (!dpyPriv->Driver) {
-         return EGL_FALSE;
-      }
-      /* Initialize the particular driver now */
-      retVal = dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy,
-                                               &major_int, &minor_int);
-
-      dpyPriv->Driver->APImajor = major_int;
-      dpyPriv->Driver->APIminor = minor_int;
-      snprintf(dpyPriv->Driver->Version, sizeof(dpyPriv->Driver->Version),
-               "%d.%d (%s)", major_int, minor_int, dpyPriv->Driver->Name);
-
-      /* Update applications version of major and minor if not NULL */
-      if((major != NULL) && (minor != NULL))
-      {
-         *major = major_int;
-         *minor = minor_int;
+   if (!disp)
+      return _eglError(EGL_BAD_DISPLAY, __FUNCTION__);
+
+   drv = disp->Driver;
+   if (!drv) {
+      drv = _eglOpenDriver(disp);
+      if (!drv)
+         return _eglError(EGL_NOT_INITIALIZED, __FUNCTION__);
+
+      /* Initialize the particular display now */
+      if (!drv->API.Initialize(drv, disp, &major_int, &minor_int)) {
+         _eglCloseDriver(drv, disp);
+         return _eglError(EGL_NOT_INITIALIZED, __FUNCTION__);
       }
 
-      return retVal;
+      disp->APImajor = major_int;
+      disp->APIminor = minor_int;
+      snprintf(disp->Version, sizeof(disp->Version),
+               "%d.%d (%s)", major_int, minor_int, drv->Name);
+
+      /* update the global notion of supported APIs */
+      _eglGlobal.ClientAPIsMask |= disp->ClientAPIsMask;
+
+      disp->Driver = drv;
+   } else {
+      major_int = disp->APImajor;
+      minor_int = disp->APIminor;
+   }
+
+   /* Update applications version of major and minor if not NULL */
+   if ((major != NULL) && (minor != NULL)) {
+      *major = major_int;
+      *minor = minor_int;
    }
-   return EGL_FALSE;
+
+   return EGL_TRUE;
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglTerminate(EGLDisplay dpy)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return _eglCloseDriver(drv, dpy);
-   else
-      return EGL_FALSE;
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLDriver *drv;
+
+   if (!disp)
+      return _eglError(EGL_BAD_DISPLAY, __FUNCTION__);
+
+   drv = disp->Driver;
+   if (drv) {
+      drv->API.Terminate(drv, disp);
+      _eglCloseDriver(drv, disp);
+      disp->Driver = NULL;
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * A bunch of check functions and declare macros to simply error checking.
+ */
+static INLINE _EGLDriver *
+_eglCheckDisplay(_EGLDisplay *disp, const char *msg)
+{
+   if (!disp) {
+      _eglError(EGL_BAD_DISPLAY, msg);
+      return NULL;
+   }
+   if (!disp->Driver) {
+      _eglError(EGL_NOT_INITIALIZED, msg);
+      return NULL;
+   }
+   return disp->Driver;
+}
+
+
+static INLINE _EGLDriver *
+_eglCheckSurface(_EGLDisplay *disp, _EGLSurface *surf, const char *msg)
+{
+   _EGLDriver *drv = _eglCheckDisplay(disp, msg);
+   if (!drv)
+      return NULL;
+   if (!surf) {
+      _eglError(EGL_BAD_SURFACE, msg);
+      return NULL;
+   }
+   return drv;
+}
+
+
+static INLINE _EGLDriver *
+_eglCheckContext(_EGLDisplay *disp, _EGLContext *context, const char *msg)
+{
+   _EGLDriver *drv = _eglCheckDisplay(disp, msg);
+   if (!drv)
+      return NULL;
+   if (!context) {
+      _eglError(EGL_BAD_CONTEXT, msg);
+      return NULL;
+   }
+   return drv;
+}
+
+
+static INLINE _EGLDriver *
+_eglCheckConfig(_EGLDisplay *disp, _EGLConfig *conf, const char *msg)
+{
+   _EGLDriver *drv = _eglCheckDisplay(disp, msg);
+   if (!drv)
+      return NULL;
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, msg);
+      return NULL;
+   }
+   return drv;
 }
 
 
+#define _EGL_DECLARE_DD(dpy)                                   \
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);                 \
+   _EGLDriver *drv;                                            \
+   do {                                                        \
+      drv = _eglCheckDisplay(disp, __FUNCTION__);              \
+      if (!drv)                                                \
+         return EGL_FALSE;                                     \
+   } while (0)
+
+
+#define _EGL_DECLARE_DD_AND_SURFACE(dpy, surface)              \
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);                 \
+   _EGLSurface *surf = _eglLookupSurface((surface), disp);     \
+   _EGLDriver *drv;                                            \
+   do {                                                        \
+      drv = _eglCheckSurface(disp, surf, __FUNCTION__);        \
+      if (!drv)                                                \
+         return EGL_FALSE;                                     \
+   } while (0)
+
+
+#define _EGL_DECLARE_DD_AND_CONTEXT(dpy, ctx)                  \
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);                 \
+   _EGLContext *context = _eglLookupContext((ctx), disp);      \
+   _EGLDriver *drv;                                            \
+   do {                                                        \
+      drv = _eglCheckContext(disp, context, __FUNCTION__);     \
+      if (!drv)                                                \
+         return EGL_FALSE;                                     \
+   } while (0)
+
+
+#ifdef EGL_MESA_screen_surface
+
+
+static INLINE _EGLDriver *
+_eglCheckScreen(_EGLDisplay *disp, _EGLScreen *scrn, const char *msg)
+{
+   _EGLDriver *drv = _eglCheckDisplay(disp, msg);
+   if (!drv)
+      return NULL;
+   if (!scrn) {
+      _eglError(EGL_BAD_SCREEN_MESA, msg);
+      return NULL;
+   }
+   return drv;
+}
+
+
+static INLINE _EGLDriver *
+_eglCheckMode(_EGLDisplay *disp, _EGLMode *m, const char *msg)
+{
+   _EGLDriver *drv = _eglCheckDisplay(disp, msg);
+   if (!drv)
+      return NULL;
+   if (!m) {
+      _eglError(EGL_BAD_MODE_MESA, msg);
+      return NULL;
+   }
+   return drv;
+}
+
+
+#define _EGL_DECLARE_DD_AND_SCREEN(dpy, screen)                \
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);                 \
+   _EGLScreen *scrn = _eglLookupScreen((screen), disp);        \
+   _EGLDriver *drv;                                            \
+   do {                                                        \
+      drv = _eglCheckScreen(disp, scrn, __FUNCTION__);         \
+      if (!drv)                                                \
+         return EGL_FALSE;                                     \
+   } while (0)
+
+
+#define _EGL_DECLARE_DD_AND_MODE(dpy, mode)                    \
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);                 \
+   _EGLMode *m = _eglLookupMode((mode), disp);                 \
+   _EGLDriver *drv;                                            \
+   do {                                                        \
+      drv = _eglCheckMode(disp, m, __FUNCTION__);              \
+      if (!drv)                                                \
+         return EGL_FALSE;                                     \
+   } while (0)
+
+
+#endif /* EGL_MESA_screen_surface */
+
+
 const char * EGLAPIENTRY
 eglQueryString(EGLDisplay dpy, EGLint name)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.QueryString(drv, dpy, name);
-   else
-      return NULL;
+   _EGL_DECLARE_DD(dpy);
+   return drv->API.QueryString(drv, disp, name);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglGetConfigs(EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config)
+eglGetConfigs(EGLDisplay dpy, EGLConfig *configs,
+              EGLint config_size, EGLint *num_config)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   /* XXX check drv for null in remaining functions */
-   return drv->API.GetConfigs(drv, dpy, configs, config_size, num_config);
+   _EGL_DECLARE_DD(dpy);
+   return drv->API.GetConfigs(drv, disp, configs, config_size, num_config);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config)
+eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs,
+                EGLint config_size, EGLint *num_config)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.ChooseConfig(drv, dpy, attrib_list, configs, config_size, num_config);
+   _EGL_DECLARE_DD(dpy);
+   return drv->API.ChooseConfig(drv, disp, attrib_list, configs,
+                                config_size, num_config);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *value)
+eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config,
+                   EGLint attribute, EGLint *value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.GetConfigAttrib(drv, dpy, config, attribute, value);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_FALSE;
+
+   return drv->API.GetConfigAttrib(drv, disp, conf, attribute, value);
 }
 
 
 EGLContext EGLAPIENTRY
-eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list)
-{
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreateContext(drv, dpy, config, share_list, attrib_list);
+eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list,
+                 const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLContext *share = _eglLookupContext(share_list, disp);
+   _EGLDriver *drv;
+   _EGLContext *context;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_CONTEXT;
+   if (!share && share_list != EGL_NO_CONTEXT) {
+      _eglError(EGL_BAD_CONTEXT, __FUNCTION__);
+      return EGL_NO_CONTEXT;
+   }
+
+   context = drv->API.CreateContext(drv, disp, conf, share, attrib_list);
+   if (context)
+      return _eglLinkContext(context, disp);
+   else
+      return EGL_NO_CONTEXT;
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglDestroyContext(EGLDisplay dpy, EGLContext ctx)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.DestroyContext(drv, dpy, ctx);
+   _EGL_DECLARE_DD_AND_CONTEXT(dpy, ctx);
+   _eglUnlinkContext(context);
+   return drv->API.DestroyContext(drv, disp, context);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx)
+eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read,
+               EGLContext ctx)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.MakeCurrent(drv, dpy, draw, read, ctx);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLContext *context = _eglLookupContext(ctx, disp);
+   _EGLSurface *draw_surf = _eglLookupSurface(draw, disp);
+   _EGLSurface *read_surf = _eglLookupSurface(read, disp);
+   _EGLDriver *drv;
+
+   drv = _eglCheckDisplay(disp, __FUNCTION__);
+   if (!drv)
+      return EGL_FALSE;
+   if (!context && ctx != EGL_NO_CONTEXT)
+      return _eglError(EGL_BAD_CONTEXT, __FUNCTION__);
+   if ((!draw_surf && draw != EGL_NO_SURFACE) ||
+       (!read_surf && read != EGL_NO_SURFACE))
+      return _eglError(EGL_BAD_SURFACE, __FUNCTION__);
+
+   return drv->API.MakeCurrent(drv, disp, draw_surf, read_surf, context);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglQueryContext(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value)
+eglQueryContext(EGLDisplay dpy, EGLContext ctx,
+                EGLint attribute, EGLint *value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QueryContext(drv, dpy, ctx, attribute, value);
+   _EGL_DECLARE_DD_AND_CONTEXT(dpy, ctx);
+   return drv->API.QueryContext(drv, disp, context, attribute, value);
 }
 
 
 EGLSurface EGLAPIENTRY
-eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list)
+eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config,
+                       NativeWindowType window, const EGLint *attrib_list)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreateWindowSurface(drv, dpy, config, window, attrib_list);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_SURFACE;
+
+   surf = drv->API.CreateWindowSurface(drv, disp, conf, window, attrib_list);
+   if (surf)
+      return _eglLinkSurface(surf, disp);
+   else
+      return EGL_NO_SURFACE;
 }
 
 
 EGLSurface EGLAPIENTRY
-eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list)
+eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config,
+                       NativePixmapType pixmap, const EGLint *attrib_list)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreatePixmapSurface(drv, dpy, config, pixmap, attrib_list);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_SURFACE;
+
+   surf = drv->API.CreatePixmapSurface(drv, disp, conf, pixmap, attrib_list);
+   if (surf)
+      return _eglLinkSurface(surf, disp);
+   else
+      return EGL_NO_SURFACE;
 }
 
 
 EGLSurface EGLAPIENTRY
-eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list)
+eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config,
+                        const EGLint *attrib_list)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreatePbufferSurface(drv, dpy, config, attrib_list);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_SURFACE;
+
+   surf = drv->API.CreatePbufferSurface(drv, disp, conf, attrib_list);
+   if (surf)
+      return _eglLinkSurface(surf, disp);
+   else
+      return EGL_NO_SURFACE;
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglDestroySurface(EGLDisplay dpy, EGLSurface surface)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.DestroySurface(drv, dpy, surface);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   _eglUnlinkSurface(surf);
+   return drv->API.DestroySurface(drv, disp, surf);
 }
 
-
 EGLBoolean EGLAPIENTRY
-eglQuerySurface(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *value)
+eglQuerySurface(EGLDisplay dpy, EGLSurface surface,
+                EGLint attribute, EGLint *value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QuerySurface(drv, dpy, surface, attribute, value);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.QuerySurface(drv, disp, surf, attribute, value);
 }
 
-
 EGLBoolean EGLAPIENTRY
-eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value)
+eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface,
+                 EGLint attribute, EGLint value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.SurfaceAttrib(drv, dpy, surface, attribute, value);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.SurfaceAttrib(drv, disp, surf, attribute, value);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglBindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.BindTexImage(drv, dpy, surface, buffer);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.BindTexImage(drv, disp, surf, buffer);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglReleaseTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.ReleaseTexImage(drv, dpy, surface, buffer);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.ReleaseTexImage(drv, disp, surf, buffer);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglSwapInterval(EGLDisplay dpy, EGLint interval)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.SwapInterval(drv, dpy, interval);
+   _EGL_DECLARE_DD(dpy);
+   return drv->API.SwapInterval(drv, disp, interval);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglSwapBuffers(EGLDisplay dpy, EGLSurface draw)
+eglSwapBuffers(EGLDisplay dpy, EGLSurface surface)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.SwapBuffers(drv, dpy, draw);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.SwapBuffers(drv, disp, surf);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglCopyBuffers(EGLDisplay dpy, EGLSurface surface, NativePixmapType target)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CopyBuffers(drv, dpy, surface, target);
+   _EGL_DECLARE_DD_AND_SURFACE(dpy, surface);
+   return drv->API.CopyBuffers(drv, disp, surf, target);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglWaitGL(void)
 {
-   EGLDisplay dpy = eglGetCurrentDisplay();
-   if (dpy != EGL_NO_DISPLAY) {
-      _EGLDriver *drv = _eglLookupDriver(dpy);
-      return drv->API.WaitGL(drv, dpy);
-   }
-   else
-      return EGL_FALSE;
+   _EGLDisplay *disp = _eglGetCurrentDisplay();
+   _EGLDriver *drv;
+
+   if (!disp)
+      return EGL_TRUE;
+
+   /* a current display is always initialized */
+   drv = disp->Driver;
+
+   return drv->API.WaitGL(drv, disp);
 }
 
 
 EGLBoolean EGLAPIENTRY
 eglWaitNative(EGLint engine)
 {
-   EGLDisplay dpy = eglGetCurrentDisplay();
-   if (dpy != EGL_NO_DISPLAY) {
-      _EGLDriver *drv = _eglLookupDriver(dpy);
-      return drv->API.WaitNative(drv, dpy, engine);
-   }
-   else
-      return EGL_FALSE;
+   _EGLDisplay *disp = _eglGetCurrentDisplay();
+   _EGLDriver *drv;
+
+   if (!disp)
+      return EGL_TRUE;
+
+   /* a current display is always initialized */
+   drv = disp->Driver;
+
+   return drv->API.WaitNative(drv, disp, engine);
 }
 
 
@@ -321,7 +578,8 @@ eglGetError(void)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
    EGLint e = t->LastError;
-   t->LastError = EGL_SUCCESS;
+   if (!_eglIsCurrentThreadDummy())
+      t->LastError = EGL_SUCCESS;
    return e;
 }
 
@@ -415,111 +673,168 @@ eglChooseModeMESA(EGLDisplay dpy, EGLScreenMESA screen,
                   const EGLint *attrib_list, EGLModeMESA *modes,
                   EGLint modes_size, EGLint *num_modes)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.ChooseModeMESA(drv, dpy, screen, attrib_list, modes, modes_size, num_modes);
-   else
-      return EGL_FALSE;
+   _EGL_DECLARE_DD_AND_SCREEN(dpy, screen);
+   return drv->API.ChooseModeMESA(drv, disp, scrn, attrib_list,
+                                  modes, modes_size, num_modes);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint mode_size, EGLint *num_mode)
+eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes,
+                EGLint mode_size, EGLint *num_mode)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.GetModesMESA(drv, dpy, screen, modes, mode_size, num_mode);
-   else
-      return EGL_FALSE;
+   _EGL_DECLARE_DD_AND_SCREEN(dpy, screen);
+   return drv->API.GetModesMESA(drv, disp, scrn, modes, mode_size, num_mode);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint *value)
+eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode,
+                     EGLint attribute, EGLint *value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.GetModeAttribMESA(drv, dpy, mode, attribute, value);
-   else
-      return EGL_FALSE;
+   _EGL_DECLARE_DD_AND_MODE(dpy, mode);
+   return drv->API.GetModeAttribMESA(drv, disp, m, attribute, value);
 }
 
 
 EGLBoolean EGLAPIENTRY
-eglCopyContextMESA(EGLDisplay dpy, EGLContext source, EGLContext dest, EGLint mask)
-{
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.CopyContextMESA(drv, dpy, source, dest, mask);
-   else
+eglCopyContextMESA(EGLDisplay dpy, EGLContext source, EGLContext dest,
+                   EGLint mask)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLContext *source_context = _eglLookupContext(source, disp);
+   _EGLContext *dest_context = _eglLookupContext(dest, disp);
+   _EGLDriver *drv;
+
+   drv = _eglCheckContext(disp, source_context, __FUNCTION__);
+   if (!drv || !dest_context) {
+      if (drv)
+         _eglError(EGL_BAD_CONTEXT, __FUNCTION__);
       return EGL_FALSE;
+   }
+
+   return drv->API.CopyContextMESA(drv, disp, source_context, dest_context,
+                                   mask);
 }
 
 
 EGLBoolean
-eglGetScreensMESA(EGLDisplay dpy, EGLScreenMESA *screens, EGLint max_screens, EGLint *num_screens)
+eglGetScreensMESA(EGLDisplay dpy, EGLScreenMESA *screens,
+                  EGLint max_screens, EGLint *num_screens)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   if (drv)
-      return drv->API.GetScreensMESA(drv, dpy, screens, max_screens, num_screens);
-   else
-      return EGL_FALSE;
+   _EGL_DECLARE_DD(dpy);
+   return drv->API.GetScreensMESA(drv, disp, screens,
+                                  max_screens, num_screens);
 }
 
 
 EGLSurface
-eglCreateScreenSurfaceMESA(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list)
+eglCreateScreenSurfaceMESA(EGLDisplay dpy, EGLConfig config,
+                           const EGLint *attrib_list)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreateScreenSurfaceMESA(drv, dpy, config, attrib_list);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_SURFACE;
+
+   surf = drv->API.CreateScreenSurfaceMESA(drv, disp, conf, attrib_list);
+   if (surf)
+      return _eglLinkSurface(surf, disp);
+   else
+      return EGL_NO_SURFACE;
 }
 
 
 EGLBoolean
-eglShowScreenSurfaceMESA(EGLDisplay dpy, EGLint screen, EGLSurface surface, EGLModeMESA mode)
+eglShowScreenSurfaceMESA(EGLDisplay dpy, EGLint screen,
+                         EGLSurface surface, EGLModeMESA mode)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.ShowScreenSurfaceMESA(drv, dpy, screen, surface, mode);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLScreen *scrn = _eglLookupScreen((EGLScreenMESA) screen, disp);
+   _EGLSurface *surf = _eglLookupSurface(surface, disp);
+   _EGLMode *m = _eglLookupMode(mode, disp);
+   _EGLDriver *drv;
+
+   drv = _eglCheckScreen(disp, scrn, __FUNCTION__);
+   if (!drv)
+      return EGL_FALSE;
+   if (!surf && surface != EGL_NO_SURFACE)
+      return _eglError(EGL_BAD_SURFACE, __FUNCTION__);
+   if (!m && mode != EGL_NO_MODE_MESA)
+      return _eglError(EGL_BAD_MODE_MESA, __FUNCTION__);
+
+   return drv->API.ShowScreenSurfaceMESA(drv, disp, scrn, surf, m);
 }
 
 
 EGLBoolean
 eglScreenPositionMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLint x, EGLint y)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.ScreenPositionMESA(drv, dpy, screen, x, y);
+   _EGL_DECLARE_DD_AND_SCREEN(dpy, screen);
+   return drv->API.ScreenPositionMESA(drv, disp, scrn, x, y);
 }
 
 
 EGLBoolean
-eglQueryScreenMESA( EGLDisplay dpy, EGLScreenMESA screen, EGLint attribute, EGLint *value)
+eglQueryScreenMESA(EGLDisplay dpy, EGLScreenMESA screen,
+                   EGLint attribute, EGLint *value)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QueryScreenMESA(drv, dpy, screen, attribute, value);
+   _EGL_DECLARE_DD_AND_SCREEN(dpy, screen);
+   return drv->API.QueryScreenMESA(drv, disp, scrn, attribute, value);
 }
 
 
 EGLBoolean
-eglQueryScreenSurfaceMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLSurface *surface)
+eglQueryScreenSurfaceMESA(EGLDisplay dpy, EGLScreenMESA screen,
+                          EGLSurface *surface)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QueryScreenSurfaceMESA(drv, dpy, screen, surface);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLScreen *scrn = _eglLookupScreen((EGLScreenMESA) screen, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckScreen(disp, scrn, __FUNCTION__);
+   if (!drv)
+      return EGL_FALSE;
+
+   if (drv->API.QueryScreenSurfaceMESA(drv, disp, scrn, &surf) != EGL_TRUE)
+      surf = NULL;
+   if (surface)
+      *surface = _eglGetSurfaceHandle(surf);
+   return (surf != NULL);
 }
 
 
 EGLBoolean
 eglQueryScreenModeMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *mode)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QueryScreenModeMESA(drv, dpy, screen, mode);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLScreen *scrn = _eglLookupScreen((EGLScreenMESA) screen, disp);
+   _EGLDriver *drv;
+   _EGLMode *m;
+
+   drv = _eglCheckScreen(disp, scrn, __FUNCTION__);
+   if (!drv)
+      return EGL_FALSE;
+
+   if (drv->API.QueryScreenModeMESA(drv, disp, scrn, &m) != EGL_TRUE)
+      m = NULL;
+   if (mode)
+      *mode = m->Handle;
+
+   return (m != NULL);
 }
 
 
 const char *
 eglQueryModeStringMESA(EGLDisplay dpy, EGLModeMESA mode)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.QueryModeStringMESA(drv, dpy, mode);
+   _EGL_DECLARE_DD_AND_MODE(dpy, mode);
+   return drv->API.QueryModeStringMESA(drv, disp, m);
 }
 
 
@@ -546,11 +861,17 @@ eglBindAPI(EGLenum api)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
 
+   if (_eglIsCurrentThreadDummy())
+      return _eglError(EGL_BAD_ALLOC, "eglBindAPI");
+
+   if (!_eglIsApiValid(api))
+      return _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
+
    switch (api) {
 #ifdef EGL_VERSION_1_4
    case EGL_OPENGL_API:
       if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
@@ -558,14 +879,14 @@ eglBindAPI(EGLenum api)
 #endif
    case EGL_OPENGL_ES_API:
       if (_eglGlobal.ClientAPIsMask & (EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT)) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
       return EGL_FALSE;
    case EGL_OPENVG_API:
       if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
@@ -585,7 +906,7 @@ eglQueryAPI(void)
 {
    /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */
    _EGLThreadInfo *t = _eglGetCurrentThread();
-   return t->CurrentAPI;
+   return _eglConvertApiFromIndex(t->CurrentAPIIndex);
 }
 
 
@@ -594,24 +915,38 @@ eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype,
                                  EGLClientBuffer buffer, EGLConfig config,
                                  const EGLint *attrib_list)
 {
-   _EGLDriver *drv = _eglLookupDriver(dpy);
-   return drv->API.CreatePbufferFromClientBuffer(drv, dpy, buftype, buffer,
-                                                 config, attrib_list);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(config, disp);
+   _EGLDriver *drv;
+   _EGLSurface *surf;
+
+   drv = _eglCheckConfig(disp, conf, __FUNCTION__);
+   if (!drv)
+      return EGL_NO_SURFACE;
+
+   surf = drv->API.CreatePbufferFromClientBuffer(drv, disp, buftype, buffer,
+                                                 conf, attrib_list);
+   if (surf)
+      return _eglLinkSurface(surf, disp);
+   else
+      return EGL_NO_SURFACE;
 }
 
 
 EGLBoolean
 eglReleaseThread(void)
 {
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   EGLDisplay dpy = eglGetCurrentDisplay();
-   if (dpy) {
-      _EGLDriver *drv = _eglLookupDriver(dpy);
-      /* unbind context */
-      (void) drv->API.MakeCurrent(drv, dpy, EGL_NO_SURFACE,
-                                  EGL_NO_SURFACE, EGL_NO_CONTEXT);
+   /* unbind current context */
+   if (!_eglIsCurrentThreadDummy()) {
+      _EGLDisplay *disp = _eglGetCurrentDisplay();
+      _EGLDriver *drv;
+      if (disp) {
+         drv = disp->Driver;
+         (void) drv->API.MakeCurrent(drv, disp, NULL, NULL, NULL);
+      }
    }
-   _eglDeleteThreadData(t);
+
+   _eglDestroyCurrentThread();
    return EGL_TRUE;
 }
 
@@ -619,13 +954,17 @@ eglReleaseThread(void)
 EGLBoolean
 eglWaitClient(void)
 {
-   EGLDisplay dpy = eglGetCurrentDisplay();
-   if (dpy != EGL_NO_DISPLAY) {
-      _EGLDriver *drv = _eglLookupDriver(dpy);
-      return drv->API.WaitClient(drv, dpy);
-   }
-   else
-      return EGL_FALSE;
+   _EGLDisplay *disp = _eglGetCurrentDisplay();
+   _EGLDriver *drv;
+
+   if (!disp)
+      return EGL_TRUE;
+
+   /* a current display is always initialized */
+   drv = disp->Driver;
+
+   return drv->API.WaitClient(drv, disp);
 }
 
+
 #endif /* EGL_VERSION_1_2 */
diff --git a/src/egl/main/eglapi.h b/src/egl/main/eglapi.h
index f6163a0c7a..6081e58892 100644
--- a/src/egl/main/eglapi.h
+++ b/src/egl/main/eglapi.h
@@ -12,61 +12,61 @@ typedef void (*_EGLProc)();
  */
 
 /* driver funcs */
-typedef EGLBoolean (*Initialize_t)(_EGLDriver *, EGLDisplay dpy, EGLint *major, EGLint *minor);
-typedef EGLBoolean (*Terminate_t)(_EGLDriver *, EGLDisplay dpy);
+typedef EGLBoolean (*Initialize_t)(_EGLDriver *, _EGLDisplay *dpy, EGLint *major, EGLint *minor);
+typedef EGLBoolean (*Terminate_t)(_EGLDriver *, _EGLDisplay *dpy);
 
 /* config funcs */
-typedef EGLBoolean (*GetConfigs_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config);
-typedef EGLBoolean (*ChooseConfig_t)(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config);
-typedef EGLBoolean (*GetConfigAttrib_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *value);
+typedef EGLBoolean (*GetConfigs_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config);
+typedef EGLBoolean (*ChooseConfig_t)(_EGLDriver *drv, _EGLDisplay *dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config);
+typedef EGLBoolean (*GetConfigAttrib_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, EGLint attribute, EGLint *value);
 
 /* context funcs */
-typedef EGLContext (*CreateContext_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list);
-typedef EGLBoolean (*DestroyContext_t)(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx);
-typedef EGLBoolean (*MakeCurrent_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx);
-typedef EGLBoolean (*QueryContext_t)(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value);
+typedef _EGLContext *(*CreateContext_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, _EGLContext *share_list, const EGLint *attrib_list);
+typedef EGLBoolean (*DestroyContext_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx);
+typedef EGLBoolean (*MakeCurrent_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *ctx);
+typedef EGLBoolean (*QueryContext_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx, EGLint attribute, EGLint *value);
 
 /* surface funcs */
-typedef EGLSurface (*CreateWindowSurface_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list);
-typedef EGLSurface (*CreatePixmapSurface_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list);
-typedef EGLSurface (*CreatePbufferSurface_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
-typedef EGLBoolean (*DestroySurface_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface);
-typedef EGLBoolean (*QuerySurface_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *value);
-typedef EGLBoolean (*SurfaceAttrib_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value);
-typedef EGLBoolean (*BindTexImage_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer);
-typedef EGLBoolean (*ReleaseTexImage_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer);
-typedef EGLBoolean (*SwapInterval_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint interval);
-typedef EGLBoolean (*SwapBuffers_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw);
-typedef EGLBoolean (*CopyBuffers_t)(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, NativePixmapType target);
+typedef _EGLSurface *(*CreateWindowSurface_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, NativeWindowType window, const EGLint *attrib_list);
+typedef _EGLSurface *(*CreatePixmapSurface_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, NativePixmapType pixmap, const EGLint *attrib_list);
+typedef _EGLSurface *(*CreatePbufferSurface_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, const EGLint *attrib_list);
+typedef EGLBoolean (*DestroySurface_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface);
+typedef EGLBoolean (*QuerySurface_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, EGLint attribute, EGLint *value);
+typedef EGLBoolean (*SurfaceAttrib_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, EGLint attribute, EGLint value);
+typedef EGLBoolean (*BindTexImage_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, EGLint buffer);
+typedef EGLBoolean (*ReleaseTexImage_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, EGLint buffer);
+typedef EGLBoolean (*SwapInterval_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLint interval);
+typedef EGLBoolean (*SwapBuffers_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw);
+typedef EGLBoolean (*CopyBuffers_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, NativePixmapType target);
 
 /* misc funcs */
-typedef const char *(*QueryString_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint name);
-typedef EGLBoolean (*WaitGL_t)(_EGLDriver *drv, EGLDisplay dpy);
-typedef EGLBoolean (*WaitNative_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint engine);
+typedef const char *(*QueryString_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLint name);
+typedef EGLBoolean (*WaitGL_t)(_EGLDriver *drv, _EGLDisplay *dpy);
+typedef EGLBoolean (*WaitNative_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLint engine);
 
 typedef _EGLProc (*GetProcAddress_t)(const char *procname);
 
 
 
 #ifdef EGL_MESA_screen_surface
-typedef EGLBoolean (*ChooseModeMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, const EGLint *attrib_list, EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes);
-typedef EGLBoolean (*GetModesMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint mode_size, EGLint *num_mode);
-typedef EGLBoolean (*GetModeAttribMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint *value);
-typedef EGLBoolean (*CopyContextMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLContext source, EGLContext dest, EGLint mask);
-typedef EGLBoolean (*GetScreensMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA *screens, EGLint max_screens, EGLint *num_screens);
-typedef EGLSurface (*CreateScreenSurfaceMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
-typedef EGLBoolean (*ShowScreenSurfaceMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface surface, EGLModeMESA mode);
-typedef EGLBoolean (*ScreenPositionMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLint x, EGLint y);
-typedef EGLBoolean (*QueryScreenMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLint attribute, EGLint *value);
-typedef EGLBoolean (*QueryScreenSurfaceMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface *surface);
-typedef EGLBoolean (*QueryScreenModeMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *mode);
-typedef const char * (*QueryModeStringMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLModeMESA mode);
+typedef EGLBoolean (*ChooseModeMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, const EGLint *attrib_list, EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes);
+typedef EGLBoolean (*GetModesMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, EGLModeMESA *modes, EGLint mode_size, EGLint *num_mode);
+typedef EGLBoolean (*GetModeAttribMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *mode, EGLint attribute, EGLint *value);
+typedef EGLBoolean (*CopyContextMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *source, _EGLContext *dest, EGLint mask);
+typedef EGLBoolean (*GetScreensMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLScreenMESA *screens, EGLint max_screens, EGLint *num_screens);
+typedef _EGLSurface *(*CreateScreenSurfaceMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *config, const EGLint *attrib_list);
+typedef EGLBoolean (*ShowScreenSurfaceMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, _EGLSurface *surface, _EGLMode *mode);
+typedef EGLBoolean (*ScreenPositionMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, EGLint x, EGLint y);
+typedef EGLBoolean (*QueryScreenMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, EGLint attribute, EGLint *value);
+typedef EGLBoolean (*QueryScreenSurfaceMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, _EGLSurface **surface);
+typedef EGLBoolean (*QueryScreenModeMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, _EGLMode **mode);
+typedef const char * (*QueryModeStringMESA_t)(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *mode);
 #endif /* EGL_MESA_screen_surface */
 
 
 #ifdef EGL_VERSION_1_2
-typedef EGLBoolean (*WaitClient_t)(_EGLDriver *drv, EGLDisplay dpy);
-typedef EGLSurface (*CreatePbufferFromClientBuffer_t)(_EGLDriver *drv, EGLDisplay dpy, EGLenum buftype, EGLClientBuffer buffer, EGLConfig config, const EGLint *attrib_list);
+typedef EGLBoolean (*WaitClient_t)(_EGLDriver *drv, _EGLDisplay *dpy);
+typedef _EGLSurface *(*CreatePbufferFromClientBuffer_t)(_EGLDriver *drv, _EGLDisplay *dpy, EGLenum buftype, EGLClientBuffer buffer, _EGLConfig *config, const EGLint *attrib_list);
 #endif /* EGL_VERSION_1_2 */
 
 
diff --git a/src/egl/main/eglcompiler.h b/src/egl/main/eglcompiler.h
new file mode 100644
index 0000000000..6b639b75c6
--- /dev/null
+++ b/src/egl/main/eglcompiler.h
@@ -0,0 +1,64 @@
+#ifndef EGLCOMPILER_INCLUDED
+#define EGLCOMPILER_INCLUDED
+
+
+/**
+ * Get standard integer types
+ */
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L)
+#  include <stdint.h>
+#elif defined(_MSC_VER)
+   typedef __int8             int8_t;
+   typedef unsigned __int8    uint8_t;
+   typedef __int16            int16_t;
+   typedef unsigned __int16   uint16_t;
+#  ifndef __eglplatform_h_
+     typedef __int32            int32_t;
+#  endif
+   typedef unsigned __int32   uint32_t;
+   typedef __int64            int64_t;
+   typedef unsigned __int64   uint64_t;
+
+#  if defined(_WIN64)
+     typedef __int64            intptr_t;
+     typedef unsigned __int64   uintptr_t;
+#  else
+     typedef __int32            intptr_t;
+     typedef unsigned __int32   uintptr_t;
+#  endif
+
+#  define INT64_C(__val) __val##i64
+#  define UINT64_C(__val) __val##ui64
+#else
+/* hope the best instead of adding a bunch of ifdef's */
+#  include <stdint.h>
+#endif
+
+
+/**
+ * Function inlining
+ */
+#if defined(__GNUC__)
+#  define INLINE __inline__
+#elif defined(__MSC__)
+#  define INLINE __inline
+#elif defined(_MSC_VER)
+#  define INLINE __inline
+#elif defined(__ICL)
+#  define INLINE __inline
+#elif defined(__INTEL_COMPILER)
+#  define INLINE inline
+#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100)
+#  define INLINE __inline
+#elif defined(__SUNPRO_C) && defined(__C99FEATURES__)
+#  define INLINE inline
+#  define __inline inline
+#  define __inline__ inline
+#elif (__STDC_VERSION__ >= 199901L) /* C99 */
+#  define INLINE inline
+#else
+#  define INLINE
+#endif
+
+
+#endif /* EGLCOMPILER_INCLUDED */
diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c
index f2f32585c7..d47b99eed4 100644
--- a/src/egl/main/eglconfig.c
+++ b/src/egl/main/eglconfig.c
@@ -34,7 +34,7 @@ void
 _eglInitConfig(_EGLConfig *config, EGLint id)
 {
    memset(config, 0, sizeof(*config));
-   config->Handle = (EGLConfig) id;
+   config->Handle = (EGLConfig) _eglUIntToPointer((unsigned int) id);
    _eglSetConfigAttrib(config, EGL_CONFIG_ID,               id);
    _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGB,     EGL_DONT_CARE);
    _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGBA,    EGL_DONT_CARE);
@@ -71,10 +71,9 @@ _eglGetConfigHandle(_EGLConfig *config)
  * This is the inverse of _eglGetConfigHandle().
  */
 _EGLConfig *
-_eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
+_eglLookupConfig(EGLConfig config, _EGLDisplay *disp)
 {
    EGLint i;
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    for (i = 0; i < disp->NumConfigs; i++) {
       if (disp->Configs[i]->Handle == config) {
           return disp->Configs[i];
@@ -319,10 +318,9 @@ _eglCompareConfigs(const void *a, const void *b)
  * Typical fallback routine for eglChooseConfig
  */
 EGLBoolean
-_eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list,
+_eglChooseConfig(_EGLDriver *drv, _EGLDisplay *disp, const EGLint *attrib_list,
                  EGLConfig *configs, EGLint config_size, EGLint *num_configs)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    _EGLConfig **configList, criteria;
    EGLint i, count;
 
@@ -367,10 +365,9 @@ _eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list,
  * Fallback for eglGetConfigAttrib.
  */
 EGLBoolean
-_eglGetConfigAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_eglGetConfigAttrib(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                     EGLint attribute, EGLint *value)
 {
-   const _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
    const EGLint k = attribute - FIRST_ATTRIB;
    if (k >= 0 && k < MAX_ATTRIBS) {
       *value = conf->Attrib[k];
@@ -387,16 +384,9 @@ _eglGetConfigAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * Fallback for eglGetConfigs.
  */
 EGLBoolean
-_eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs,
+_eglGetConfigs(_EGLDriver *drv, _EGLDisplay *disp, EGLConfig *configs,
                EGLint config_size, EGLint *num_config)
 {
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-
-   if (!drv->Initialized) {
-      _eglError(EGL_NOT_INITIALIZED, "eglGetConfigs");
-      return EGL_FALSE;
-   }
-
    if (configs) {
       EGLint i;
       *num_config = MIN2(disp->NumConfigs, config_size);
diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h
index db1c4c10e0..36ed96ae95 100644
--- a/src/egl/main/eglconfig.h
+++ b/src/egl/main/eglconfig.h
@@ -34,7 +34,7 @@ _eglGetConfigHandle(_EGLConfig *config);
 
 
 extern _EGLConfig *
-_eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config);
+_eglLookupConfig(EGLConfig config, _EGLDisplay *dpy);
 
 
 extern _EGLConfig *
@@ -46,15 +46,15 @@ _eglParseConfigAttribs(_EGLConfig *config, const EGLint *attrib_list);
 
 
 extern EGLBoolean
-_eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config);
+_eglChooseConfig(_EGLDriver *drv, _EGLDisplay *dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config);
 
 
 extern EGLBoolean
-_eglGetConfigAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *value);
+_eglGetConfigAttrib(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, EGLint attribute, EGLint *value);
 
 
 extern EGLBoolean
-_eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config);
+_eglGetConfigs(_EGLDriver *drv, _EGLDisplay *dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config);
 
 
 extern void
diff --git a/src/egl/main/eglconfigutil.c b/src/egl/main/eglconfigutil.c
index 138dc729e7..c9d00e7982 100644
--- a/src/egl/main/eglconfigutil.c
+++ b/src/egl/main/eglconfigutil.c
@@ -156,6 +156,7 @@ _eglFillInConfigs(_EGLConfig * configs,
             {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000},  /* 8_8_8_8_REV */
          };
 
+#if 0
    static const uint32_t masks_table_bgr[8][4] = {
             {0x00000000, 0x00000000, 0x00000000, 0x00000000},
             {0x00000000, 0x00000000, 0x00000000, 0x00000000},
@@ -177,6 +178,7 @@ _eglFillInConfigs(_EGLConfig * configs,
             {0x00000000, 0x00000000, 0x00000000, 0x00000000},
             {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000},  /* 8_8_8_8_REV */
          };
+#endif
 
    static const uint8_t bytes_per_pixel[8] = {
             0, 0, 0, 2, 2, 4, 0, 4
diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c
index 461679db09..b094f49bfc 100644
--- a/src/egl/main/eglcontext.c
+++ b/src/egl/main/eglcontext.c
@@ -14,11 +14,9 @@
  * in the attrib_list.
  */
 EGLBoolean
-_eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
-                EGLConfig config, const EGLint *attrib_list)
+_eglInitContext(_EGLDriver *drv, _EGLContext *ctx,
+                _EGLConfig *conf, const EGLint *attrib_list)
 {
-   _EGLConfig *conf;
-   _EGLDisplay *display = _eglLookupDisplay(dpy);
    EGLint i;
    const EGLenum api = eglQueryAPI();
 
@@ -27,12 +25,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
       return EGL_FALSE;
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   if (!conf) {
-      _eglError(EGL_BAD_CONFIG, "_eglInitContext");
-      return EGL_FALSE;
-   }
-
    memset(ctx, 0, sizeof(_EGLContext));
 
    ctx->ClientVersion = 1; /* the default, per EGL spec */
@@ -49,7 +41,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
       }
    }
 
-   ctx->Display = display;
    ctx->Config = conf;
    ctx->DrawSurface = EGL_NO_SURFACE;
    ctx->ReadSurface = EGL_NO_SURFACE;
@@ -60,88 +51,27 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
 
 
 /**
- * Save a new _EGLContext into the hash table.
- */
-void
-_eglSaveContext(_EGLContext *ctx)
-{
-   /* no-op.
-    * Public EGLContext handle and private _EGLContext are the same.
-    */
-}
-
-
-/**
- * Remove the given _EGLContext object from the hash table.
- */
-void
-_eglRemoveContext(_EGLContext *ctx)
-{
-   /* no-op.
-    * Public EGLContext handle and private _EGLContext are the same.
-    */
-}
-
-
-/**
- * Return the public handle for the given private context ptr.
- * This is the inverse of _eglLookupContext().
- */
-EGLContext
-_eglGetContextHandle(_EGLContext *ctx)
-{
-   /* just a cast! */
-   return (EGLContext) ctx;
-}
-
-
-/**
- * Return the _EGLContext object that corresponds to the given
- * EGLContext handle.
- * This is the inverse of _eglGetContextHandle().
- */
-_EGLContext *
-_eglLookupContext(EGLContext ctx)
-{
-   /* just a cast since EGLContext is just a void ptr */
-   return (_EGLContext *) ctx;
-}
-
-
-/**
- * Return the currently bound _EGLContext object, or NULL.
- */
-_EGLContext *
-_eglGetCurrentContext(void)
-{
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   return t->CurrentContext;
-}
-
-
-/**
  * Just a placeholder/demo function.  Real driver will never use this!
  */
-EGLContext
-_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
-                  EGLContext share_list, const EGLint *attrib_list)
+_EGLContext *
+_eglCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
+                  _EGLContext *share_list, const EGLint *attrib_list)
 {
 #if 0 /* example code */
    _EGLContext *context;
 
    context = (_EGLContext *) calloc(1, sizeof(_EGLContext));
    if (!context)
-      return EGL_NO_CONTEXT;
+      return NULL;
 
-   if (!_eglInitContext(drv, dpy, context, config, attrib_list)) {
+   if (!_eglInitContext(drv, context, conf, attrib_list)) {
       free(context);
-      return EGL_NO_CONTEXT;
+      return NULL;
    }
 
-   _eglSaveContext(context);
-   return (EGLContext) context;
+   return context;
 #endif
-   return EGL_NO_CONTEXT;
+   return NULL;
 }
 
 
@@ -149,39 +79,21 @@ _eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * Default fallback routine - drivers should usually override this.
  */
 EGLBoolean
-_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
+_eglDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx)
 {
-   _EGLContext *context = _eglLookupContext(ctx);
-   if (context) {
-      if (context->IsBound) {
-         context->DeletePending = EGL_TRUE;
-      }
-      else {
-         free(context);
-      }
-      return EGL_TRUE;
-   }
-   else {
-      _eglError(EGL_BAD_CONTEXT, "eglDestroyContext");
-      return EGL_TRUE;
-   }
+   if (!_eglIsContextBound(ctx))
+      free(ctx);
+   return EGL_TRUE;
 }
 
 
 EGLBoolean
-_eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx,
+_eglQueryContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *c,
                  EGLint attribute, EGLint *value)
 {
-   _EGLContext *c = _eglLookupContext(ctx);
-
    (void) drv;
    (void) dpy;
 
-   if (!c) {
-      _eglError(EGL_BAD_CONTEXT, "eglQueryContext");
-      return EGL_FALSE;
-   }
-
    switch (attribute) {
    case EGL_CONFIG_ID:
       *value = GET_CONFIG_ATTRIB(c->Config, EGL_CONFIG_ID);
@@ -203,90 +115,100 @@ _eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx,
 
 /**
  * Drivers will typically call this to do the error checking and
- * update the various IsBound and DeletePending flags.
+ * update the various flags.
  * Then, the driver will do its device-dependent Make-Current stuff.
  */
 EGLBoolean
-_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
-                EGLSurface r, EGLContext context)
+_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw,
+                _EGLSurface *read, _EGLContext *ctx)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
-   _EGLContext *ctx = _eglLookupContext(context);
-   _EGLSurface *draw = _eglLookupSurface(d);
-   _EGLSurface *read = _eglLookupSurface(r);
+   _EGLContext *oldContext = NULL;
+   _EGLSurface *oldDrawSurface = NULL;
+   _EGLSurface *oldReadSurface = NULL;
+   EGLint apiIndex;
 
-   _EGLContext *oldContext = _eglGetCurrentContext();
-   _EGLSurface *oldDrawSurface = _eglGetCurrentSurface(EGL_DRAW);
-   _EGLSurface *oldReadSurface = _eglGetCurrentSurface(EGL_READ);
+   if (_eglIsCurrentThreadDummy())
+      return _eglError(EGL_BAD_ALLOC, "eglMakeCurrent");
 
-   /* error checking */
    if (ctx) {
-      if (draw == NULL || read == NULL) {
-         _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
-         return EGL_FALSE;
-      }
-      if (draw->Config != ctx->Config) {
-         _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
-         return EGL_FALSE;
-      }
-      if (read->Config != ctx->Config) {
-         _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
-         return EGL_FALSE;
+      /* error checking */
+      if (ctx->Binding && ctx->Binding != t)
+         return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+      if (draw == NULL || read == NULL)
+         return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
+      if (draw->Config != ctx->Config || read->Config != ctx->Config)
+         return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
+      if ((draw->Binding && draw->Binding->Binding != t) ||
+          (read->Binding && read->Binding->Binding != t))
+         return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+
+#ifdef EGL_VERSION_1_4
+      /* OpenGL and OpenGL ES are conflicting */
+      switch (ctx->ClientAPI) {
+      case EGL_OPENGL_ES_API:
+         if (t->CurrentContexts[_eglConvertApiToIndex(EGL_OPENGL_API)])
+            return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+         break;
+      case EGL_OPENGL_API:
+         if (t->CurrentContexts[_eglConvertApiToIndex(EGL_OPENGL_ES_API)])
+            return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+         break;
+      default:
+         break;
       }
+#endif
+      apiIndex = _eglConvertApiToIndex(ctx->ClientAPI);
+   }
+   else {
+      if (draw != NULL || read != NULL)
+         return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
+      apiIndex = t->CurrentAPIIndex;
    }
 
-   /*
-    * check if the old context or surfaces need to be deleted
-    */
-   if (oldDrawSurface != NULL) {
-      oldDrawSurface->IsBound = EGL_FALSE;
-      if (oldDrawSurface->DeletePending) {
-         /* make sure we don't try to rebind a deleted surface */
-         if (draw == oldDrawSurface || draw == oldReadSurface) {
-            draw = NULL;
-         }
-         /* really delete surface now */
-         drv->API.DestroySurface(drv, dpy, oldDrawSurface->Handle);
+   oldContext = t->CurrentContexts[apiIndex];
+   if (oldContext) {
+      oldDrawSurface = oldContext->DrawSurface;
+      oldReadSurface = oldContext->ReadSurface;
+      assert(oldDrawSurface);
+      assert(oldReadSurface);
+
+      /* break old bindings */
+      t->CurrentContexts[apiIndex] = NULL;
+      oldContext->Binding = NULL;
+      oldContext->DrawSurface = NULL;
+      oldContext->ReadSurface = NULL;
+      oldDrawSurface->Binding = NULL;
+      oldReadSurface->Binding = NULL;
+
+      /*
+       * check if the old context or surfaces need to be deleted
+       */
+      if (!_eglIsSurfaceLinked(oldDrawSurface)) {
+         assert(draw != oldDrawSurface && read != oldDrawSurface);
+         drv->API.DestroySurface(drv, dpy, oldDrawSurface);
       }
-   }
-   if (oldReadSurface != NULL && oldReadSurface != oldDrawSurface) {
-      oldReadSurface->IsBound = EGL_FALSE;
-      if (oldReadSurface->DeletePending) {
-         /* make sure we don't try to rebind a deleted surface */
-         if (read == oldDrawSurface || read == oldReadSurface) {
-            read = NULL;
-         }
-         /* really delete surface now */
-         drv->API.DestroySurface(drv, dpy, oldReadSurface->Handle);
+      if (oldReadSurface != oldDrawSurface &&
+          !_eglIsSurfaceLinked(oldReadSurface)) {
+         assert(draw != oldReadSurface && read != oldReadSurface);
+         drv->API.DestroySurface(drv, dpy, oldReadSurface);
       }
-   }
-   if (oldContext != NULL) {
-      oldContext->IsBound = EGL_FALSE;
-      if (oldContext->DeletePending) {
-         /* make sure we don't try to rebind a deleted context */
-         if (ctx == oldContext) {
-            ctx = NULL;
-         }
-         /* really delete context now */
-         drv->API.DestroyContext(drv, dpy, _eglGetContextHandle(oldContext));
+      if (!_eglIsContextLinked(oldContext)) {
+         assert(ctx != oldContext);
+         drv->API.DestroyContext(drv, dpy, oldContext);
       }
    }
 
+   /* build new bindings */
    if (ctx) {
-      /* check read/draw again, in case we deleted them above */
-      if (draw == NULL || read == NULL) {
-         _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
-         return EGL_FALSE;
-      }
+      t->CurrentContexts[apiIndex] = ctx;
+      ctx->Binding = t;
       ctx->DrawSurface = draw;
       ctx->ReadSurface = read;
-      ctx->IsBound = EGL_TRUE;
-      draw->IsBound = EGL_TRUE;
-      read->IsBound = EGL_TRUE;
+      draw->Binding = ctx;
+      read->Binding = ctx;
    }
 
-   t->CurrentContext = ctx;
-
    return EGL_TRUE;
 }
 
diff --git a/src/egl/main/eglcontext.h b/src/egl/main/eglcontext.h
index 34fee9c637..647f24488f 100644
--- a/src/egl/main/eglcontext.h
+++ b/src/egl/main/eglcontext.h
@@ -11,15 +11,16 @@
  */
 struct _egl_context
 {
-   _EGLDisplay *Display; /* who do I belong to? */
-
-   _EGLConfig *Config;
+   /* Managed by EGLDisplay for linking */
+   _EGLDisplay *Display;
+   _EGLContext *Next;
 
+   /* The bound status of the context */
+   _EGLThreadInfo *Binding;
    _EGLSurface *DrawSurface;
    _EGLSurface *ReadSurface;
 
-   EGLBoolean IsBound;
-   EGLBoolean DeletePending;
+   _EGLConfig *Config;
 
    EGLint ClientAPI; /**< EGL_OPENGL_ES_API, EGL_OPENGL_API, EGL_OPENVG_API */
    EGLint ClientVersion; /**< 1 = OpenGLES 1.x, 2 = OpenGLES 2.x */
@@ -27,47 +28,38 @@ struct _egl_context
 
 
 extern EGLBoolean
-_eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
-                EGLConfig config, const EGLint *attrib_list);
-
-
-extern void
-_eglSaveContext(_EGLContext *ctx);
-
-
-extern void
-_eglRemoveContext(_EGLContext *ctx);
-
-
-extern EGLContext
-_eglGetContextHandle(_EGLContext *ctx);
+_eglInitContext(_EGLDriver *drv, _EGLContext *ctx,
+                _EGLConfig *config, const EGLint *attrib_list);
 
 
 extern _EGLContext *
-_eglLookupContext(EGLContext ctx);
- 
-
-extern _EGLContext *
-_eglGetCurrentContext(void);
-
-
-extern EGLContext
-_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list);
+_eglCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list);
 
 
 extern EGLBoolean
-_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx);
+_eglDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx);
 
 
 extern EGLBoolean
-_eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value);
+_eglQueryContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx, EGLint attribute, EGLint *value);
 
 
 extern EGLBoolean
-_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx);
+_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *ctx);
 
 
 extern EGLBoolean
 _eglCopyContextMESA(_EGLDriver *drv, EGLDisplay dpy, EGLContext source, EGLContext dest, EGLint mask);
 
+
+/**
+ * Return true if the context is bound to a thread.
+ */
+static INLINE EGLBoolean
+_eglIsContextBound(_EGLContext *ctx)
+{
+   return (ctx->Binding != NULL);
+}
+
+
 #endif /* EGLCONTEXT_INCLUDED */
diff --git a/src/egl/main/eglcurrent.c b/src/egl/main/eglcurrent.c
new file mode 100644
index 0000000000..4431f964f6
--- /dev/null
+++ b/src/egl/main/eglcurrent.c
@@ -0,0 +1,342 @@
+#include <stdlib.h>
+#include <string.h>
+#include "eglcurrent.h"
+#include "eglcontext.h"
+#include "egllog.h"
+#include "eglmutex.h"
+#include "eglglobals.h"
+
+
+/* This should be kept in sync with _eglInitThreadInfo() */
+#define _EGL_THREAD_INFO_INITIALIZER \
+   { EGL_SUCCESS, { NULL }, 1 }
+
+/* a fallback thread info to guarantee that every thread always has one */
+static _EGLThreadInfo dummy_thread = _EGL_THREAD_INFO_INITIALIZER;
+
+
+#ifdef GLX_USE_TLS
+static __thread const _EGLThreadInfo *_egl_TSD;
+   __attribute__ ((tls_model("initial-exec")));
+
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
+{
+   _egl_TSD = t;
+}
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) _egl_TSD;
+}
+
+static INLINE void _eglFiniTSD(void)
+{
+}
+
+static INLINE EGLBoolean _eglInitTSD(void (*dtor)(_EGLThreadInfo *))
+{
+   /* TODO destroy TSD */
+   (void) dtor;
+   (void) _eglFiniTSD;
+   return EGL_TRUE;
+}
+
+#elif PTHREADS
+#include <pthread.h>
+
+static _EGL_DECLARE_MUTEX(_egl_TSDMutex);
+static EGLBoolean _egl_TSDInitialized;
+static pthread_key_t _egl_TSD;
+static void (*_egl_FreeTSD)(_EGLThreadInfo *);
+
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
+{
+   pthread_setspecific(_egl_TSD, (const void *) t);
+}
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) pthread_getspecific(_egl_TSD);
+}
+
+static INLINE void _eglFiniTSD(void)
+{
+   _eglLockMutex(&_egl_TSDMutex);
+   if (_egl_TSDInitialized) {
+      _EGLThreadInfo *t = _eglGetTSD();
+
+      _egl_TSDInitialized = EGL_FALSE;
+      if (t && _egl_FreeTSD)
+         _egl_FreeTSD((void *) t);
+      pthread_key_delete(_egl_TSD);
+   }
+   _eglUnlockMutex(&_egl_TSDMutex);
+}
+
+static INLINE EGLBoolean _eglInitTSD(void (*dtor)(_EGLThreadInfo *))
+{
+   if (!_egl_TSDInitialized) {
+      _eglLockMutex(&_egl_TSDMutex);
+
+      /* check again after acquiring lock */
+      if (!_egl_TSDInitialized) {
+         if (pthread_key_create(&_egl_TSD, (void (*)(void *)) dtor) != 0) {
+            _eglUnlockMutex(&_egl_TSDMutex);
+            return EGL_FALSE;
+         }
+         _egl_FreeTSD = dtor;
+         _eglAddAtExitCall(_eglFiniTSD);
+         _egl_TSDInitialized = EGL_TRUE;
+      }
+
+      _eglUnlockMutex(&_egl_TSDMutex);
+   }
+
+   return EGL_TRUE;
+}
+
+#else /* PTHREADS */
+static const _EGLThreadInfo *_egl_TSD;
+static void (*_egl_FreeTSD)(_EGLThreadInfo *);
+
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
+{
+   _egl_TSD = t;
+}
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) _egl_TSD;
+}
+
+static INLINE void _eglFiniTSD(void)
+{
+   if (_egl_FreeTSD && _egl_TSD)
+      _egl_FreeTSD((_EGLThreadInfo *) _egl_TSD);
+}
+
+static INLINE EGLBoolean _eglInitTSD(void (*dtor)(_EGLThreadInfo *))
+{
+   if (!_egl_FreeTSD && dtor) {
+      _egl_FreeTSD = dtor;
+      _eglAddAtExitCall(_eglFiniTSD);
+   }
+   return EGL_TRUE;
+}
+
+#endif /* !PTHREADS */
+
+
+static void
+_eglInitThreadInfo(_EGLThreadInfo *t)
+{
+   memset(t, 0, sizeof(*t));
+   t->LastError = EGL_SUCCESS;
+   /* default, per EGL spec */
+   t->CurrentAPIIndex = _eglConvertApiToIndex(EGL_OPENGL_ES_API);
+}
+
+
+/**
+ * Allocate and init a new _EGLThreadInfo object.
+ */
+static _EGLThreadInfo *
+_eglCreateThreadInfo(void)
+{
+   _EGLThreadInfo *t = (_EGLThreadInfo *) calloc(1, sizeof(_EGLThreadInfo));
+   if (t)
+      _eglInitThreadInfo(t);
+   else
+      t = &dummy_thread;
+   return t;
+}
+
+
+/**
+ * Delete/free a _EGLThreadInfo object.
+ */
+static void
+_eglDestroyThreadInfo(_EGLThreadInfo *t)
+{
+   if (t != &dummy_thread)
+      free(t);
+}
+
+
+/**
+ * Make sure TSD is initialized and return current value.
+ */
+static INLINE _EGLThreadInfo *
+_eglCheckedGetTSD(void)
+{
+   if (_eglInitTSD(&_eglDestroyThreadInfo) != EGL_TRUE) {
+      _eglLog(_EGL_FATAL, "failed to initialize \"current\" system");
+      return NULL;
+   }
+
+   return _eglGetTSD();
+}
+
+
+/**
+ * Return the calling thread's thread info.
+ * If the calling thread nevers calls this function before, or if its thread
+ * info was destroyed, a new one is created.  This function never returns NULL.
+ * In the case allocation fails, a dummy one is returned.  See also
+ * _eglIsCurrentThreadDummy.
+ */
+_EGLThreadInfo *
+_eglGetCurrentThread(void)
+{
+   _EGLThreadInfo *t = _eglCheckedGetTSD();
+   if (!t) {
+      t = _eglCreateThreadInfo();
+      _eglSetTSD(t);
+   }
+
+   return t;
+}
+
+
+/**
+ * Destroy the calling thread's thread info.
+ */
+void
+_eglDestroyCurrentThread(void)
+{
+   _EGLThreadInfo *t = _eglCheckedGetTSD();
+   if (t) {
+      _eglDestroyThreadInfo(t);
+      _eglSetTSD(NULL);
+   }
+}
+
+
+/**
+ * Return true if the calling thread's thread info is dummy.
+ * A dummy thread info is shared by all threads and should not be modified.
+ * Functions like eglBindAPI or eglMakeCurrent should check for dummy-ness
+ * before updating the thread info.
+ */
+EGLBoolean
+_eglIsCurrentThreadDummy(void)
+{
+   _EGLThreadInfo *t = _eglCheckedGetTSD();
+   return (!t || t == &dummy_thread);
+}
+
+
+/**
+ * Return the currently bound context, or NULL.
+ */
+_EGLContext *
+_eglGetCurrentContext(void)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   return t->CurrentContexts[t->CurrentAPIIndex];
+}
+
+
+/**
+ * Return the display of the currently bound context, or NULL.
+ */
+_EGLDisplay *
+_eglGetCurrentDisplay(void)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLContext *ctx = t->CurrentContexts[t->CurrentAPIIndex];
+   if (ctx)
+      return ctx->Display;
+   else
+      return NULL;
+}
+
+
+/**
+ * Return the read or write surface of the currently bound context, or NULL.
+ */
+_EGLSurface *
+_eglGetCurrentSurface(EGLint readdraw)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLContext *ctx = t->CurrentContexts[t->CurrentAPIIndex];
+   if (ctx) {
+      switch (readdraw) {
+      case EGL_DRAW:
+         return ctx->DrawSurface;
+      case EGL_READ:
+         return ctx->ReadSurface;
+      default:
+         return NULL;
+      }
+   }
+   return NULL;
+}
+
+
+/**
+ * Record EGL error code.
+ */
+EGLBoolean
+_eglError(EGLint errCode, const char *msg)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   const char *s;
+
+   if (t == &dummy_thread)
+      return EGL_FALSE;
+
+   if (t->LastError == EGL_SUCCESS) {
+      t->LastError = errCode;
+
+      switch (errCode) {
+      case EGL_BAD_ACCESS:
+         s = "EGL_BAD_ACCESS";
+         break;
+      case EGL_BAD_ALLOC:
+         s = "EGL_BAD_ALLOC";
+         break;
+      case EGL_BAD_ATTRIBUTE:
+         s = "EGL_BAD_ATTRIBUTE";
+         break;
+      case EGL_BAD_CONFIG:
+         s = "EGL_BAD_CONFIG";
+         break;
+      case EGL_BAD_CONTEXT:
+         s = "EGL_BAD_CONTEXT";
+         break;
+      case EGL_BAD_CURRENT_SURFACE:
+         s = "EGL_BAD_CURRENT_SURFACE";
+         break;
+      case EGL_BAD_DISPLAY:
+         s = "EGL_BAD_DISPLAY";
+         break;
+      case EGL_BAD_MATCH:
+         s = "EGL_BAD_MATCH";
+         break;
+      case EGL_BAD_NATIVE_PIXMAP:
+         s = "EGL_BAD_NATIVE_PIXMAP";
+         break;
+      case EGL_BAD_NATIVE_WINDOW:
+         s = "EGL_BAD_NATIVE_WINDOW";
+         break;
+      case EGL_BAD_PARAMETER:
+         s = "EGL_BAD_PARAMETER";
+         break;
+      case EGL_BAD_SURFACE:
+         s = "EGL_BAD_SURFACE";
+         break;
+      case EGL_BAD_SCREEN_MESA:
+         s = "EGL_BAD_SCREEN_MESA";
+         break;
+      case EGL_BAD_MODE_MESA:
+         s = "EGL_BAD_MODE_MESA";
+         break;
+      default:
+         s = "other";
+      }
+      _eglLog(_EGL_DEBUG, "EGL user error 0x%x (%s) in %s\n", errCode, s, msg);
+   }
+
+   return EGL_FALSE;
+}
diff --git a/src/egl/main/eglcurrent.h b/src/egl/main/eglcurrent.h
new file mode 100644
index 0000000000..8eb241029e
--- /dev/null
+++ b/src/egl/main/eglcurrent.h
@@ -0,0 +1,84 @@
+#ifndef EGLCURRENT_INCLUDED
+#define EGLCURRENT_INCLUDED
+
+#include "egltypedefs.h"
+
+
+#define _EGL_API_NUM_INDICES \
+   (EGL_OPENGL_API - EGL_OPENGL_ES_API + 2) /* idx 0 is for EGL_NONE */
+
+
+/**
+ * Per-thread info
+ */
+struct _egl_thread_info
+{
+   EGLint LastError;
+   _EGLContext *CurrentContexts[_EGL_API_NUM_INDICES];
+   /* use index for fast access to current context */
+   EGLint CurrentAPIIndex;
+};
+
+
+/**
+ * Return true if a client API enum can be converted to an index.
+ */
+static INLINE EGLBoolean
+_eglIsApiValid(EGLenum api)
+{
+   return ((api >= EGL_OPENGL_ES_API && api <= EGL_OPENGL_API) ||
+           api == EGL_NONE);
+}
+
+
+/**
+ * Convert a client API enum to an index, for use by thread info.
+ * The client API enum is assumed to be valid.
+ */
+static INLINE EGLint
+_eglConvertApiToIndex(EGLenum api)
+{
+   return (api != EGL_NONE) ? api - EGL_OPENGL_ES_API + 1 : 0;
+}
+
+
+/**
+ * Convert an index, used by thread info, to a client API enum.
+ * The index is assumed to be valid.
+ */
+static INLINE EGLenum
+_eglConvertApiFromIndex(EGLint idx)
+{
+   return (idx) ? EGL_OPENGL_ES_API + idx - 1 : EGL_NONE;
+}
+
+
+extern _EGLThreadInfo *
+_eglGetCurrentThread(void);
+
+
+extern void
+_eglDestroyCurrentThread(void);
+
+
+extern EGLBoolean
+_eglIsCurrentThreadDummy(void);
+
+
+extern _EGLContext *
+_eglGetCurrentContext(void);
+
+
+extern _EGLDisplay *
+_eglGetCurrentDisplay(void);
+
+
+extern _EGLSurface *
+_eglGetCurrentSurface(EGLint readdraw);
+
+
+extern EGLBoolean
+_eglError(EGLint errCode, const char *msg);
+
+
+#endif /* EGLCURRENT_INCLUDED */
diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c
index 47a2323eaf..896d60dbe1 100644
--- a/src/egl/main/egldisplay.c
+++ b/src/egl/main/egldisplay.c
@@ -1,4 +1,3 @@
-
 /**
  * Functions related to EGLDisplay.
  */
@@ -7,11 +6,67 @@
 #include <stdlib.h>
 #include <string.h>
 #include "eglcontext.h"
+#include "eglsurface.h"
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
-#include "eglhash.h"
 #include "eglstring.h"
+#include "eglmutex.h"
+#include "egllog.h"
+
+
+/**
+ * Finish display management.
+ */
+void
+_eglFiniDisplay(void)
+{
+   _EGLDisplay *dpyList, *dpy;
+
+   /* atexit function is called with global mutex locked */
+   dpyList = _eglGlobal.DisplayList;
+   while (dpyList) {
+      /* pop list head */
+      dpy = dpyList;
+      dpyList = dpyList->Next;
+
+      if (dpy->ContextList || dpy->SurfaceList)
+         _eglLog(_EGL_DEBUG, "Display %p is destroyed with resources", dpy);
+
+      free(dpy);
+   }
+   _eglGlobal.DisplayList = NULL;
+}
+
+
+/**
+ * If the first character is '!' we interpret it as specific driver name
+ * (i.e. "!r200" or "!i830").  Whatever follows ':' is interpreted as
+ * arguments.
+ *
+ * The caller may free() the returned driver name.
+ */
+char *
+_eglSplitDisplayString(const char *dpyString, const char **args)
+{
+   char *drv, *p;
+
+   if (!dpyString || dpyString[0] != '!')
+      return NULL;
+   drv = _eglstrdup(dpyString + 1);
+   if (!drv)
+      return NULL;
+
+   p = strchr(dpyString, ':');
+   if (p) {
+      drv[p - dpyString] = '\0';
+      p++;
+   }
+   if (args)
+      *args = p;
+
+   return drv;
+}
 
 
 /**
@@ -25,17 +80,9 @@ _eglNewDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy = (_EGLDisplay *) calloc(1, sizeof(_EGLDisplay));
    if (dpy) {
-      EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
-
-      dpy->Handle = (EGLDisplay) key;
-      _eglHashInsert(_eglGlobal.Displays, key, dpy);
-
       dpy->NativeDisplay = nativeDisplay;
-#if defined(_EGL_PLATFORM_X)
-      dpy->Xdpy = (Display *) nativeDisplay;
-#endif
 
-      dpy->DriverName = _eglChooseDriver(dpy);
+      dpy->DriverName = _eglPreloadDriver(dpy);
       if (!dpy->DriverName) {
          free(dpy);
          return NULL;
@@ -46,54 +93,107 @@ _eglNewDisplay(NativeDisplayType nativeDisplay)
 
 
 /**
- * Return the public handle for an internal _EGLDisplay.
- * This is the inverse of _eglLookupDisplay().
+ * Link a display to itself and return the handle of the link.
+ * The handle can be passed to client directly.
  */
 EGLDisplay
-_eglGetDisplayHandle(_EGLDisplay *display)
+_eglLinkDisplay(_EGLDisplay *dpy)
 {
-   if (display)
-      return display->Handle;
-   else
-      return EGL_NO_DISPLAY;
+   _eglLockMutex(_eglGlobal.Mutex);
+
+   dpy->Next = _eglGlobal.DisplayList;
+   _eglGlobal.DisplayList = dpy;
+
+   _eglUnlockMutex(_eglGlobal.Mutex);
+
+   return (EGLDisplay) dpy;
 }
 
- 
+
 /**
- * Return the _EGLDisplay object that corresponds to the given public/
- * opaque display handle.
- * This is the inverse of _eglGetDisplayHandle().
+ * Unlink a linked display from itself.
+ * Accessing an unlinked display should generate EGL_BAD_DISPLAY error.
  */
-_EGLDisplay *
-_eglLookupDisplay(EGLDisplay dpy)
+void
+_eglUnlinkDisplay(_EGLDisplay *dpy)
 {
-   EGLuint key = (EGLuint) dpy;
-   if (!_eglGlobal.Displays)
-      return NULL;
-   return (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key);
+   _EGLDisplay *prev;
+
+   _eglLockMutex(_eglGlobal.Mutex);
+
+   prev = _eglGlobal.DisplayList;
+   if (prev != dpy) {
+      while (prev) {
+         if (prev->Next == dpy)
+            break;
+         prev = prev->Next;
+      }
+      assert(prev);
+      prev->Next = dpy->Next;
+   }
+   else {
+      _eglGlobal.DisplayList = dpy->Next;
+   }
+
+   _eglUnlockMutex(_eglGlobal.Mutex);
 }
 
 
-void
-_eglSaveDisplay(_EGLDisplay *dpy)
+/**
+ * Find the display corresponding to the specified native display id in all
+ * linked displays.
+ */
+_EGLDisplay *
+_eglFindDisplay(NativeDisplayType nativeDisplay)
 {
-   EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
-   assert(dpy);
-   assert(!dpy->Handle);
-   dpy->Handle = (EGLDisplay) key;
-   assert(dpy->Handle);
-   _eglHashInsert(_eglGlobal.Displays, key, dpy);
+   _EGLDisplay *dpy;
+
+   _eglLockMutex(_eglGlobal.Mutex);
+
+   dpy = _eglGlobal.DisplayList;
+   while (dpy) {
+      if (dpy->NativeDisplay == nativeDisplay) {
+         _eglUnlockMutex(_eglGlobal.Mutex);
+         return dpy;
+      }
+      dpy = dpy->Next;
+   }
+
+   _eglUnlockMutex(_eglGlobal.Mutex);
+
+   return NULL;
 }
 
 
-_EGLDisplay *
-_eglGetCurrentDisplay(void)
+/**
+ * Destroy the contexts and surfaces that are linked to the display.
+ */
+void
+_eglReleaseDisplayResources(_EGLDriver *drv, _EGLDisplay *display)
 {
-   _EGLContext *ctx = _eglGetCurrentContext();
-   if (ctx)
-      return ctx->Display;
-   else
-      return NULL;
+   _EGLContext *contexts;
+   _EGLSurface *surfaces;
+
+   contexts = display->ContextList;
+   surfaces = display->SurfaceList;
+
+   while (contexts) {
+      _EGLContext *ctx = contexts;
+      contexts = contexts->Next;
+
+      _eglUnlinkContext(ctx);
+      drv->API.DestroyContext(drv, display, ctx);
+   }
+   assert(!display->ContextList);
+
+   while (surfaces) {
+      _EGLSurface *surf = surfaces;
+      surfaces = surfaces->Next;
+
+      _eglUnlinkSurface(surf);
+      drv->API.DestroySurface(drv, display, surf);
+   }
+   assert(!display->SurfaceList);
 }
 
 
@@ -106,16 +206,166 @@ _eglCleanupDisplay(_EGLDisplay *disp)
 {
    EGLint i;
 
-   for (i = 0; i < disp->NumConfigs; i++) {
-      free(disp->Configs[i]);
+   if (disp->Configs) {
+      for (i = 0; i < disp->NumConfigs; i++)
+         free(disp->Configs[i]);
+      free(disp->Configs);
+      disp->Configs = NULL;
+      disp->NumConfigs = 0;
    }
-   free(disp->Configs);
-   disp->Configs = NULL;
 
    /* XXX incomplete */
+}
 
-   free((void *) disp->DriverName);
-   disp->DriverName = NULL;
 
-   /* driver deletes the _EGLDisplay object */
+/**
+ * Link a context to a display and return the handle of the link.
+ * The handle can be passed to client directly.
+ */
+EGLContext
+_eglLinkContext(_EGLContext *ctx, _EGLDisplay *dpy)
+{
+   ctx->Display = dpy;
+   ctx->Next = dpy->ContextList;
+   dpy->ContextList = ctx;
+   return (EGLContext) ctx;
+}
+
+
+/**
+ * Unlink a linked context from its display.
+ * Accessing an unlinked context should generate EGL_BAD_CONTEXT error.
+ */
+void
+_eglUnlinkContext(_EGLContext *ctx)
+{
+   _EGLContext *prev;
+
+   prev = ctx->Display->ContextList;
+   if (prev != ctx) {
+      while (prev) {
+         if (prev->Next == ctx)
+            break;
+         prev = prev->Next;
+      }
+      assert(prev);
+      prev->Next = ctx->Next;
+   }
+   else {
+      ctx->Display->ContextList = ctx->Next;
+   }
+
+   ctx->Next = NULL;
+   ctx->Display = NULL;
+}
+
+
+/**
+ * Link a surface to a display and return the handle of the link.
+ * The handle can be passed to client directly.
+ */
+EGLSurface
+_eglLinkSurface(_EGLSurface *surf, _EGLDisplay *dpy)
+{
+   surf->Display = dpy;
+   surf->Next = dpy->SurfaceList;
+   dpy->SurfaceList = surf;
+   return (EGLSurface) surf;
 }
+
+
+/**
+ * Unlink a linked surface from its display.
+ * Accessing an unlinked surface should generate EGL_BAD_SURFACE error.
+ */
+void
+_eglUnlinkSurface(_EGLSurface *surf)
+{
+   _EGLSurface *prev;
+
+   prev = surf->Display->SurfaceList;
+   if (prev != surf) {
+      while (prev) {
+         if (prev->Next == surf)
+            break;
+         prev = prev->Next;
+      }
+      assert(prev);
+      prev->Next = surf->Next;
+   }
+   else {
+      prev = NULL;
+      surf->Display->SurfaceList = surf->Next;
+   }
+
+   surf->Next = NULL;
+   surf->Display = NULL;
+}
+
+
+#ifndef _EGL_SKIP_HANDLE_CHECK
+
+
+/**
+ * Return EGL_TRUE if the given handle is a valid handle to a display.
+ */
+EGLBoolean
+_eglCheckDisplayHandle(EGLDisplay dpy)
+{
+   _EGLDisplay *cur;
+
+   _eglLockMutex(_eglGlobal.Mutex);
+   cur = _eglGlobal.DisplayList;
+   while (cur) {
+      if (cur == (_EGLDisplay *) dpy)
+         break;
+      cur = cur->Next;
+   }
+   _eglUnlockMutex(_eglGlobal.Mutex);
+   return (cur != NULL);
+}
+
+
+/**
+ * Return EGL_TRUE if the given handle is a valid handle to a context.
+ */
+EGLBoolean
+_eglCheckContextHandle(EGLContext ctx, _EGLDisplay *dpy)
+{
+   _EGLContext *cur = NULL;
+
+   if (dpy)
+      cur = dpy->ContextList;
+   while (cur) {
+      if (cur == (_EGLContext *) ctx) {
+         assert(cur->Display == dpy);
+         break;
+      }
+      cur = cur->Next;
+   }
+   return (cur != NULL);
+}
+
+
+/**
+ * Return EGL_TRUE if the given handle is a valid handle to a surface.
+ */
+EGLBoolean
+_eglCheckSurfaceHandle(EGLSurface surf, _EGLDisplay *dpy)
+{
+   _EGLSurface *cur = NULL;
+
+   if (dpy)
+      cur = dpy->SurfaceList;
+   while (cur) {
+      if (cur == (_EGLSurface *) surf) {
+         assert(cur->Display == dpy);
+         break;
+      }
+      cur = cur->Next;
+   }
+   return (cur != NULL);
+}
+
+
+#endif /* !_EGL_SKIP_HANDLE_CHECK */
diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h
index ff623ee1c6..6575fdf198 100644
--- a/src/egl/main/egldisplay.h
+++ b/src/egl/main/egldisplay.h
@@ -1,21 +1,45 @@
 #ifndef EGLDISPLAY_INCLUDED
 #define EGLDISPLAY_INCLUDED
 
-#ifdef _EGL_PLATFORM_X
-#include <X11/Xlib.h>
-#endif
-
 #include "egltypedefs.h"
+#include "egldefines.h"
+#include "eglcontext.h"
+#include "eglsurface.h"
+
+
+/**
+ * Optional EGL extensions info.
+ */
+struct _egl_extensions
+{
+   EGLBoolean MESA_screen_surface;
+   EGLBoolean MESA_copy_context;
+
+   char String[_EGL_MAX_EXTENSIONS_LEN];
+};
 
 
 struct _egl_display 
 {
+   /* used to link displays */
+   _EGLDisplay *Next;
+
    EGLNativeDisplayType NativeDisplay;
-   EGLDisplay Handle;
 
    const char *DriverName;
-   const char *DriverArgs;
    _EGLDriver *Driver;
+   void *DriverData; /* private to driver */
+
+   int APImajor, APIminor; /**< as returned by eglInitialize() */
+   char Version[1000];     /**< initialized from APImajor/minor, DriverName */
+
+   /** Bitmask of supported APIs (EGL_xx_BIT) set by the driver during init */
+   EGLint ClientAPIsMask;
+   char ClientAPIs[1000];   /**< updated by eglQueryString */
+
+   _EGLExtensions Extensions;
+
+   int LargestPbuffer;
 
    EGLint NumScreens;
    _EGLScreen **Screens;  /* array [NumScreens] */
@@ -23,39 +47,226 @@ struct _egl_display
    EGLint NumConfigs;
    _EGLConfig **Configs;  /* array [NumConfigs] of ptr to _EGLConfig */
 
-#ifdef _EGL_PLATFORM_X
-   Display *Xdpy;
-#endif
+   /* lists of linked contexts and surface */
+   _EGLContext *ContextList;
+   _EGLSurface *SurfaceList;
 };
 
 
-extern _EGLDisplay *
-_eglNewDisplay(NativeDisplayType displayName);
+extern void
+_eglFiniDisplay(void);
 
 
-EGLDisplay
-_eglGetDisplayHandle(_EGLDisplay *display);
+extern char *
+_eglSplitDisplayString(const char *dpyString, const char **args);
 
 
 extern _EGLDisplay *
-_eglLookupDisplay(EGLDisplay dpy);
+_eglNewDisplay(NativeDisplayType displayName);
+
+
+extern EGLDisplay
+_eglLinkDisplay(_EGLDisplay *dpy);
 
 
 extern void
-_eglSaveDisplay(_EGLDisplay *dpy);
+_eglUnlinkDisplay(_EGLDisplay *dpy);
 
 
 extern _EGLDisplay *
-_eglGetCurrentDisplay(void);
+_eglFindDisplay(NativeDisplayType nativeDisplay);
+
+
+extern void
+_eglReleaseDisplayResources(_EGLDriver *drv, _EGLDisplay *dpy);
 
 
 extern void
 _eglCleanupDisplay(_EGLDisplay *disp);
 
 
-extern EGLBoolean 
-_eglQueryDisplayMESA(_EGLDriver *drv, EGLDisplay dpy, EGLint attrib, EGLint *value);
+extern EGLContext
+_eglLinkContext(_EGLContext *ctx, _EGLDisplay *dpy);
+
+
+extern void
+_eglUnlinkContext(_EGLContext *ctx);
+
+
+extern EGLSurface
+_eglLinkSurface(_EGLSurface *surf, _EGLDisplay *dpy);
+
+
+extern void
+_eglUnlinkSurface(_EGLSurface *surf);
+
+
+#ifndef _EGL_SKIP_HANDLE_CHECK
+
+
+extern EGLBoolean
+_eglCheckDisplayHandle(EGLDisplay dpy);
+
+
+extern EGLBoolean
+_eglCheckContextHandle(EGLContext ctx, _EGLDisplay *dpy);
+
+
+extern EGLBoolean
+_eglCheckSurfaceHandle(EGLSurface surf, _EGLDisplay *dpy);
+
+
+#else /* !_EGL_SKIP_HANDLE_CHECK */
+
+/* Only do a quick check.  This is NOT standard compliant. */
+
+static INLINE EGLBoolean
+_eglCheckDisplayHandle(EGLDisplay dpy)
+{
+   return ((_EGLDisplay *) dpy != NULL);
+}
 
 
+static INLINE EGLBoolean
+_eglCheckContextHandle(EGLContext ctx, _EGLDisplay *dpy)
+{
+   _EGLContext *c = (_EGLContext *) ctx;
+   return (dpy && c && c->Display == dpy);
+}
+
+
+static INLINE EGLBoolean
+_eglCheckSurfaceHandle(EGLSurface surf, _EGLDisplay *dpy)
+{
+   _EGLSurface *s = (_EGLSurface *) surf;
+   return (dpy && s && s->Display == dpy);
+}
+
+
+#endif /* _EGL_SKIP_HANDLE_CHECK */
+
+
+/**
+ * Lookup a handle to find the linked display.
+ * Return NULL if the handle has no corresponding linked display.
+ */
+static INLINE _EGLDisplay *
+_eglLookupDisplay(EGLDisplay display)
+{
+   _EGLDisplay *dpy = (_EGLDisplay *) display;
+   if (!_eglCheckDisplayHandle(display))
+      dpy = NULL;
+   return dpy;
+}
+
+
+/**
+ * Return the handle of a linked display, or EGL_NO_DISPLAY.
+ */
+static INLINE EGLDisplay
+_eglGetDisplayHandle(_EGLDisplay *dpy)
+{
+   return (EGLDisplay) ((dpy) ? dpy : EGL_NO_DISPLAY);
+}
+
+
+/**
+ * Return true if the display is linked.
+ */
+static INLINE EGLBoolean
+_eglIsDisplayLinked(_EGLDisplay *dpy)
+{
+   return (EGLBoolean) (_eglGetDisplayHandle(dpy) != EGL_NO_DISPLAY);
+}
+
+
+/**
+ * Lookup a handle to find the linked context.
+ * Return NULL if the handle has no corresponding linked context.
+ */
+static INLINE _EGLContext *
+_eglLookupContext(EGLContext context, _EGLDisplay *dpy)
+{
+   _EGLContext *ctx = (_EGLContext *) context;
+   if (!_eglCheckContextHandle(context, dpy))
+      ctx = NULL;
+   return ctx;
+}
+
+
+/**
+ * Return the handle of a linked context, or EGL_NO_CONTEXT.
+ */
+static INLINE EGLContext
+_eglGetContextHandle(_EGLContext *ctx)
+{
+   return (EGLContext) ((ctx && ctx->Display) ? ctx : EGL_NO_CONTEXT);
+}
+
+
+/**
+ * Return true if the context is linked to a display.
+ */
+static INLINE EGLBoolean
+_eglIsContextLinked(_EGLContext *ctx)
+{
+   return (EGLBoolean) (_eglGetContextHandle(ctx) != EGL_NO_CONTEXT);
+}
+
+
+/**
+ * Lookup a handle to find the linked surface.
+ * Return NULL if the handle has no corresponding linked surface.
+ */
+static INLINE _EGLSurface *
+_eglLookupSurface(EGLSurface surface, _EGLDisplay *dpy)
+{
+   _EGLSurface *surf = (_EGLSurface *) surface;
+   if (!_eglCheckSurfaceHandle(surf, dpy))
+      surf = NULL;
+   return surf;
+}
+
+
+/**
+ * Return the handle of a linked surface, or EGL_NO_SURFACE.
+ */
+static INLINE EGLSurface
+_eglGetSurfaceHandle(_EGLSurface *surf)
+{
+   return (EGLSurface) ((surf && surf->Display) ? surf : EGL_NO_SURFACE);
+}
+
+
+/**
+ * Return true if the surface is linked to a display.
+ */
+static INLINE EGLBoolean
+_eglIsSurfaceLinked(_EGLSurface *surf)
+{
+   return (EGLBoolean) (_eglGetSurfaceHandle(surf) != EGL_NO_SURFACE);
+}
+
+
+/**
+ * Cast an unsigned int to a pointer.
+ */
+static INLINE void *
+_eglUIntToPointer(unsigned int v)
+{
+   return (void *) ((uintptr_t) v);
+}
+
+
+/**
+ * Cast a pointer to an unsigned int.  The pointer must be one that is
+ * returned by _eglUIntToPointer.
+ */
+static INLINE unsigned int
+_eglPointerToUInt(const void *p)
+{
+   return (unsigned int) ((uintptr_t) p);
+}
+
 
 #endif /* EGLDISPLAY_INCLUDED */
diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c
index 43b1f51903..87786e36bb 100644
--- a/src/egl/main/egldriver.c
+++ b/src/egl/main/egldriver.c
@@ -22,7 +22,6 @@
 
 #if defined(_EGL_PLATFORM_X)
 #include <dlfcn.h>
-#include "eglx.h"
 #elif defined(_EGL_PLATFORM_WINDOWS)
 /* Use static linking on Windows for now */
 #define WINDOWS_STATIC_LINK
@@ -38,7 +37,6 @@
    /* XXX Need to decide how to do dynamic name lookup on Windows */
    static const char *DefaultDriverName = "TBD";
 #endif
-   static const char *SysFS = NULL;
    typedef HMODULE lib_handle;
 
    static HMODULE
@@ -61,8 +59,7 @@
    }
 
 #elif defined(_EGL_PLATFORM_X)
-   static const char *DefaultDriverName = ":0";
-   static const char *SysFS = "/sys/class";
+   static const char *DefaultDriverName = "egl_softpipe";
 
    typedef void * lib_handle;
 
@@ -80,142 +77,69 @@
    
 #endif
 
-/**
- * Given a card number, use sysfs to determine the DRI driver name.
- */
-const char *
-_eglChooseDRMDriver(int card)
-{
-#if 0
-   return _eglstrdup("libEGLdri");
-#else
-   char path[2000], driverName[2000];
-   FILE *f;
-   int length;
-
-   snprintf(path, sizeof(path), "%s/drm/card%d/dri_library_name", SysFS, card);
-
-   f = fopen(path, "r");
-   if (!f)
-      return NULL;
-
-   fgets(driverName, sizeof(driverName), f);
-   fclose(f);
-
-   if ((length = strlen(driverName)) > 1) {
-      /* remove the trailing newline from sysfs */
-      driverName[length - 1] = '\0';
-      strncat(driverName, "_dri", sizeof(driverName));
-      return _eglstrdup(driverName);
-   }
-   else {
-      return NULL;
-   }   
-#endif
-}
 
 /**
- * XXX this function is totally subject change!!!
- *
- *
- * Determine/return the name of the driver to use for the given _EGLDisplay.
- *
- * Try to be clever and determine if nativeDisplay is an Xlib Display
- * ptr or a string (naming a driver or screen number, etc).
- *
- * If the first character is ':' we interpret it as a screen or card index
- * number (i.e. ":0" or ":1", etc)
- * Else if the first character is '!' we interpret it as specific driver name
- * (i.e. "!r200" or "!i830".
- *
- * Whatever follows ':' is copied and put into dpy->DriverArgs.
- *
- * The caller may free() the returned string.
+ * Choose a driver for a given display.
+ * The caller may free() the returned strings.
  */
-const char *
-_eglChooseDriver(_EGLDisplay *dpy)
+static char *
+_eglChooseDriver(_EGLDisplay *dpy, char **argsRet)
 {
-   /* Under Windows, the NativeDisplay is an HDC handle, therefore */
-   /* it can't be interpreted as a string or a pointer. */
-#if defined(_EGL_PLATFORM_WINDOWS)
-   const char *displayString = NULL;
-#else
-   const char *displayString = (const char *) dpy->NativeDisplay;
-#endif
-   const char *driverName = NULL;
+   char *path = NULL;
+   const char *args = NULL;
+   const char *suffix = NULL;
+   const char *p;
 
-   (void) DefaultDriverName;
+   path = getenv("EGL_DRIVER");
+   if (path)
+      path = _eglstrdup(path);
 
 #if defined(_EGL_PLATFORM_X)
-   /* First, if the EGL_DRIVER env var is set, use that */
-   driverName = getenv("EGL_DRIVER");
-   if (driverName)
-      return _eglstrdup(driverName);
-#endif
-
-#if 0
-   if (!displayString) {
-      /* choose a default */
-      displayString = DefaultDriverName;
+   if (!path && dpy->NativeDisplay) {
+      /* assume (wrongly!) that the native display is a display string */
+      path = _eglSplitDisplayString((const char *) dpy->NativeDisplay, &args);
    }
-#endif
-   /* extract default DriverArgs = whatever follows ':' */
-   if (displayString &&
-       (displayString[0] == '!' ||
-        displayString[0] == ':')) {
-      const char *args = strchr(displayString, ':');
-      if (args)
-         dpy->DriverArgs = _eglstrdup(args + 1);
-   }
-
-   /* determine driver name now */
-   if (displayString && displayString[0] == ':' &&
-       (displayString[1] >= '0' && displayString[1] <= '9') &&
-       !displayString[2]) {
-      int card = atoi(displayString + 1);
-      driverName = _eglChooseDRMDriver(card);
-   }
-   else if (displayString && displayString[0] == '!') {
-      /* use user-specified driver name */
-      driverName = _eglstrdup(displayString + 1);
-      /* truncate driverName at ':' if present */
-      {
-         char *args = strchr(driverName, ':');
-         if (args) {
-            *args = 0;
-         }
+   suffix = "so";
+#elif defined(_EGL_PLATFORM_WINDOWS)
+   suffix = "dll";
+#endif /* _EGL_PLATFORM_X */
+
+   if (!path)
+      path = _eglstrdup(DefaultDriverName);
+
+   /* append suffix if there isn't */
+   p = strrchr(path, '.');
+   if (!p && suffix) {
+      size_t len = strlen(path);
+      char *tmp = malloc(len + strlen(suffix) + 2);
+      if (tmp) {
+         memcpy(tmp, path, len);
+         tmp[len++] = '.';
+         tmp[len] = '\0';
+         strcat(tmp + len, suffix);
+
+         free(path);
+         path = tmp;
       }
    }
-   else 
-   {
-      /* NativeDisplay is not a string! */
-#if defined(_EGL_PLATFORM_X)
-      driverName = _xeglChooseDriver(dpy);
-#else
-      driverName = DefaultDriverName;
-#endif
-   }
 
-   return driverName;
+   if (argsRet)
+      *argsRet = (args) ? _eglstrdup(args) : NULL;
+
+   return path;
 }
 
 
 /**
- * Open/load the named driver and call its bootstrap function: _eglMain().
- * By the time this function is called, the dpy->DriverName should have
- * been determined.
- *
- * \return  new _EGLDriver object.
+ * Open the named driver and find its bootstrap function: _eglMain().
  */
-_EGLDriver *
-_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args)
+static _EGLMain_t
+_eglOpenLibrary(const char *driverPath, lib_handle *handle)
 {
-   _EGLDriver *drv;
    _EGLMain_t mainFunc;
    lib_handle lib;
-   char driverFilename[1000];
 
-   assert(driverName);
+   assert(driverPath);
 
 #if defined(_EGL_PLATFORM_WINDOWS)
 /* Use static linking on Windows for now */
@@ -224,85 +148,190 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args)
    mainFunc = (_EGLMain_t)_eglMain;
 #else
    /* XXX untested */
-   sprintf(driverFilename, "%s.dll", driverName);
-   _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename);
-   lib = open_library(driverFilename);
+   _eglLog(_EGL_DEBUG, "dlopen(%s)", driverPath);
+   lib = open_library(driverPath);
    if (!lib) {
       _eglLog(_EGL_WARNING, "Could not open %s",
-              driverFilename);
+              driverPath);
       return NULL;
    }
    mainFunc = (_EGLMain_t) GetProcAddress(lib, "_eglMain");
 #endif
 #elif defined(_EGL_PLATFORM_X)
-   /* XXX also prepend a directory path??? */
-   sprintf(driverFilename, "%s.so", driverName);
-   _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename);
-   lib = open_library(driverFilename);
+   _eglLog(_EGL_DEBUG, "dlopen(%s)", driverPath);
+   lib = open_library(driverPath);
    if (!lib) {
       _eglLog(_EGL_WARNING, "Could not open %s (%s)",
-              driverFilename, dlerror());
+              driverPath, dlerror());
+      if (!getenv("EGL_DRIVER"))
+         _eglLog(_EGL_WARNING,
+                 "The driver can be overridden by setting EGL_DRIVER");
       return NULL;
    }
    mainFunc = (_EGLMain_t) dlsym(lib, "_eglMain");
 #endif
 
    if (!mainFunc) {
-      _eglLog(_EGL_WARNING, "_eglMain not found in %s", driverFilename);
-      close_library(lib);
+      _eglLog(_EGL_WARNING, "_eglMain not found in %s", driverPath);
+      if (lib)
+         close_library(lib);
       return NULL;
    }
 
-   drv = mainFunc(dpy, args);
+   *handle = lib;
+   return mainFunc;
+}
+
+
+/**
+ * Load the named driver.  The path and args passed will be
+ * owned by the driver and freed.
+ */
+static _EGLDriver *
+_eglLoadDriver(char *path, char *args)
+{
+   _EGLMain_t mainFunc;
+   lib_handle lib;
+   _EGLDriver *drv = NULL;
+
+   mainFunc = _eglOpenLibrary(path, &lib);
+   if (!mainFunc)
+      return NULL;
+
+   drv = mainFunc(args);
    if (!drv) {
-      close_library(lib);
+      if (lib)
+         close_library(lib);
       return NULL;
    }
 
-   /* with a recurvise open you want the inner most handle */
-   if (!drv->LibHandle) {
-      drv->LibHandle = lib;
-   }
-   else {
-      close_library(lib);
+   if (!drv->Name) {
+      _eglLog(_EGL_WARNING, "Driver loaded from %s has no name", path);
+      drv->Name = "UNNAMED";
    }
 
-   /* update the global notion of supported APIs */
-   _eglGlobal.ClientAPIsMask |= drv->ClientAPIsMask;
-
-   _eglSaveDriver(drv);
+   drv->Path = path;
+   drv->Args = args;
+   drv->LibHandle = lib;
 
    return drv;
 }
 
 
-EGLBoolean
-_eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy)
+/**
+ * Match a display to a preloaded driver.
+ */
+static _EGLDriver *
+_eglMatchDriver(_EGLDisplay *dpy)
 {
-   void *handle = drv->LibHandle;
-   EGLBoolean b;
+   _EGLDriver *defaultDriver = NULL;
+   EGLint i;
 
-   _eglLog(_EGL_DEBUG, "Closing %s", drv->Name);
+   for (i = 0; i < _eglGlobal.NumDrivers; i++) {
+      _EGLDriver *drv = _eglGlobal.Drivers[i];
 
-   /*
-    * XXX check for currently bound context/surfaces and delete them?
-    */
+      /* display specifies a driver */
+      if (dpy->DriverName) {
+         if (strcmp(dpy->DriverName, drv->Name) == 0)
+            return drv;
+      }
+      else if (drv->Probe) {
+         if (drv->Probe(drv, dpy))
+            return drv;
+      }
+      else {
+         if (!defaultDriver)
+            defaultDriver = drv;
+      }
+   }
+
+   return defaultDriver;
+}
+
+
+/**
+ * Load a driver and save it.
+ */
+const char *
+_eglPreloadDriver(_EGLDisplay *dpy)
+{
+   char *path, *args;
+   _EGLDriver *drv;
+   EGLint i;
+
+   path = _eglChooseDriver(dpy, &args);
+   if (!path)
+      return NULL;
+
+   for (i = 0; i < _eglGlobal.NumDrivers; i++) {
+      drv = _eglGlobal.Drivers[i];
+      if (strcmp(drv->Path, path) == 0) {
+         _eglLog(_EGL_DEBUG, "Driver %s is already preloaded",
+                 drv->Name);
+         free(path);
+         if (args)
+            free(args);
+         return drv->Name;
+      }
+   }
+
+   drv = _eglLoadDriver(path, args);
+   if (!drv)
+      return NULL;
 
-   b = drv->API.Terminate(drv, dpy);
+   _eglGlobal.Drivers[_eglGlobal.NumDrivers++] = drv;
+
+   return drv->Name;
+}
+
+
+/**
+ * Open a preloaded driver.
+ */
+_EGLDriver *
+_eglOpenDriver(_EGLDisplay *dpy)
+{
+   _EGLDriver *drv = _eglMatchDriver(dpy);
+   return drv;
+}
 
-   close_library(handle);
 
-   return b;
+/**
+ * Close a preloaded driver.
+ */
+EGLBoolean
+_eglCloseDriver(_EGLDriver *drv, _EGLDisplay *dpy)
+{
+   return EGL_TRUE;
 }
 
 
 /**
- * Save the given driver pointer in the list of all known drivers.
+ * Unload preloaded drivers.
  */
 void
-_eglSaveDriver(_EGLDriver *drv)
+_eglUnloadDrivers(void)
 {
-   _eglGlobal.Drivers[ _eglGlobal.NumDrivers++ ] = drv;
+   EGLint i;
+   for (i = 0; i < _eglGlobal.NumDrivers; i++) {
+      _EGLDriver *drv = _eglGlobal.Drivers[i];
+      lib_handle handle = drv->LibHandle;
+
+      if (drv->Path)
+         free((char *) drv->Path);
+      if (drv->Args)
+         free((char *) drv->Args);
+
+      /* destroy driver */
+      if (drv->Unload)
+         drv->Unload(drv);
+
+      if (handle)
+         close_library(handle);
+      _eglGlobal.Drivers[i] = NULL;
+   }
+
+   _eglGlobal.NumDrivers = 0;
 }
 
 
diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h
index 4066c6ec1d..6c848eb35e 100644
--- a/src/egl/main/egldriver.h
+++ b/src/egl/main/egldriver.h
@@ -4,19 +4,6 @@
 
 #include "egltypedefs.h"
 #include "eglapi.h"
-#include "egldefines.h"
-
-
-/**
- * Optional EGL extensions info.
- */
-struct _egl_extensions
-{
-   EGLBoolean MESA_screen_surface;
-   EGLBoolean MESA_copy_context;
-
-   char String[_EGL_MAX_EXTENSIONS_LEN];
-};
 
 
 /**
@@ -24,46 +11,37 @@ struct _egl_extensions
  */
 struct _egl_driver
 {
-   EGLBoolean Initialized; /**< set by driver after initialized */
-
    void *LibHandle; /**< dlopen handle */
+   const char *Path;  /**< path to this driver */
+   const char *Args;  /**< args to load this driver */
 
    const char *Name;  /**< name of this driver */
-
-   int APImajor, APIminor; /**< as returned by eglInitialize() */
-   char Version[1000];       /**< initialized from APImajor/minor, Name */
-
-   /** Bitmask of supported APIs (EGL_xx_BIT) set by the driver during init */
-   EGLint ClientAPIsMask;
+   /**< probe a display to see if it is supported */
+   EGLBoolean (*Probe)(_EGLDriver *drv, _EGLDisplay *dpy);
+   /**< called before dlclose to release this driver */
+   void (*Unload)(_EGLDriver *drv);
 
    _EGLAPI API;  /**< EGL API dispatch table */
-
-   _EGLExtensions Extensions;
-
-   int LargestPbuffer;
 };
 
 
-extern _EGLDriver *_eglMain(_EGLDisplay *dpy, const char *args);
+extern _EGLDriver *_eglMain(const char *args);
 
 
 extern const char *
-_eglChooseDRMDriver(int card);
-
-extern const char *
-_eglChooseDriver(_EGLDisplay *dpy);
+_eglPreloadDriver(_EGLDisplay *dpy);
 
 
 extern _EGLDriver *
-_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args);
+_eglOpenDriver(_EGLDisplay *dpy);
 
 
 extern EGLBoolean
-_eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy);
+_eglCloseDriver(_EGLDriver *drv, _EGLDisplay *dpy);
 
 
-extern void
-_eglSaveDriver(_EGLDriver *drv);
+void
+_eglUnloadDrivers(void);
 
 
 extern _EGLDriver *
diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c
index b770e55dbd..3ae4c1ad3a 100644
--- a/src/egl/main/eglglobals.c
+++ b/src/egl/main/eglglobals.c
@@ -1,146 +1,56 @@
-#include <stdio.h>
 #include <stdlib.h>
+#include <assert.h>
 #include "eglglobals.h"
+#include "egldriver.h"
+#include "egllog.h"
+#include "eglmutex.h"
 
-struct _egl_global _eglGlobal = 
-{
-   EGL_FALSE
-};
-
-/**
- * Init the fields in the _eglGlobal struct
- * May be safely called more than once.
- */
-void
-_eglInitGlobals(void)
-{
-   if (!_eglGlobal.Initialized) {
-      _eglGlobal.Displays = _eglNewHashTable();
-      _eglGlobal.Surfaces = _eglNewHashTable();
-      _eglGlobal.FreeScreenHandle = 1;
-      _eglGlobal.Initialized = EGL_TRUE;
-
-      _eglGlobal.ClientAPIsMask = 0x0;
 
-      /* XXX temporary */
-      _eglGlobal.ThreadInfo = _eglNewThreadInfo();
-   }
-}
+#define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
 
 
-/**
- * Should call this via an atexit handler.
- */
-void
-_eglDestroyGlobals(void)
+static _EGL_DECLARE_MUTEX(_eglGlobalMutex);
+struct _egl_global _eglGlobal =
 {
-   /* XXX TODO walk over table entries, deleting each */
-   _eglDeleteHashTable(_eglGlobal.Displays);
-   _eglDeleteHashTable(_eglGlobal.Surfaces);
-}
+   &_eglGlobalMutex,       /* Mutex */
+   NULL,                   /* DisplayList */
+   1,                      /* FreeScreenHandle */
+   0x0,                    /* ClientAPIsMask */
+   0,                      /* NumDrivers */
+   { NULL },               /* Drivers */
+   2,                      /* NumAtExitCalls */
+   {                       /* AtExitCalls */
+      _eglFiniDisplay,
+      _eglUnloadDrivers
+   },
+};
 
 
-/**
- * Allocate and init a new _EGLThreadInfo object.
- */
-_EGLThreadInfo *
-_eglNewThreadInfo(void)
+static void
+_eglAtExit(void)
 {
-   _EGLThreadInfo *t = (_EGLThreadInfo *) calloc(1, sizeof(_EGLThreadInfo));
-   if (t) {
-      t->CurrentContext = EGL_NO_CONTEXT;
-      t->LastError = EGL_SUCCESS;
-      t->CurrentAPI = EGL_OPENGL_ES_API;  /* default, per EGL spec */
-   }
-   return t;
+   EGLint i;
+   for (i = _eglGlobal.NumAtExitCalls - 1; i >= 0; i--)
+      _eglGlobal.AtExitCalls[i]();
 }
 
 
-/**
- * Delete/free a _EGLThreadInfo object.
- */
 void
-_eglDeleteThreadData(_EGLThreadInfo *t)
+_eglAddAtExitCall(void (*func)(void))
 {
-   free(t);
-}
-
-
+   if (func) {
+      static EGLBoolean registered = EGL_FALSE;
 
-/**
- * Return pointer to calling thread's _EGLThreadInfo object.
- * Create a new one if needed.
- * Should never return NULL.
- */
-_EGLThreadInfo *
-_eglGetCurrentThread(void)
-{
-   _eglInitGlobals();
-
-   /* XXX temporary */
-   return _eglGlobal.ThreadInfo;
-}
+      _eglLockMutex(_eglGlobal.Mutex);
 
+      if (!registered) {
+         atexit(_eglAtExit);
+         registered = EGL_TRUE;
+      }
 
-/**
- * Record EGL error code.
- */
-void
-_eglError(EGLint errCode, const char *msg)
-{
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   const char *s;
-
-   if (t->LastError == EGL_SUCCESS) {
-      t->LastError = errCode;
+      assert(_eglGlobal.NumAtExitCalls < ARRAY_SIZE(_eglGlobal.AtExitCalls));
+      _eglGlobal.AtExitCalls[_eglGlobal.NumAtExitCalls++] = func;
 
-      switch (errCode) {
-      case EGL_BAD_ACCESS:
-         s = "EGL_BAD_ACCESS";
-         break;
-      case EGL_BAD_ALLOC:
-         s = "EGL_BAD_ALLOC";
-         break;
-      case EGL_BAD_ATTRIBUTE:
-         s = "EGL_BAD_ATTRIBUTE";
-         break;
-      case EGL_BAD_CONFIG:
-         s = "EGL_BAD_CONFIG";
-         break;
-      case EGL_BAD_CONTEXT:
-         s = "EGL_BAD_CONTEXT";
-         break;
-      case EGL_BAD_CURRENT_SURFACE:
-         s = "EGL_BAD_CURRENT_SURFACE";
-         break;
-      case EGL_BAD_DISPLAY:
-         s = "EGL_BAD_DISPLAY";
-         break;
-      case EGL_BAD_MATCH:
-         s = "EGL_BAD_MATCH";
-         break;
-      case EGL_BAD_NATIVE_PIXMAP:
-         s = "EGL_BAD_NATIVE_PIXMAP";
-         break;
-      case EGL_BAD_NATIVE_WINDOW:
-         s = "EGL_BAD_NATIVE_WINDOW";
-         break;
-      case EGL_BAD_PARAMETER:
-         s = "EGL_BAD_PARAMETER";
-         break;
-      case EGL_BAD_SURFACE:
-         s = "EGL_BAD_SURFACE";
-         break;
-      case EGL_BAD_SCREEN_MESA:
-         s = "EGL_BAD_SCREEN_MESA";
-         break;
-      case EGL_BAD_MODE_MESA:
-         s = "EGL_BAD_MODE_MESA";
-         break;
-      default:
-         s = "other";
-      }
-      /* XXX temporary */
-      fprintf(stderr, "EGL user error 0x%x (%s) in %s\n", errCode, s, msg);
+      _eglUnlockMutex(_eglGlobal.Mutex);
    }
 }
diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h
index 14d8ea487a..58511076d4 100644
--- a/src/egl/main/eglglobals.h
+++ b/src/egl/main/eglglobals.h
@@ -2,18 +2,9 @@
 #define EGLGLOBALS_INCLUDED
 
 #include "egltypedefs.h"
-#include "eglhash.h"
-
-
-/**
- * Per-thread info
- */
-struct _egl_thread_info
-{
-   EGLint LastError;
-   _EGLContext *CurrentContext;
-   EGLenum CurrentAPI;
-};
+#include "egldisplay.h"
+#include "eglcurrent.h"
+#include "eglmutex.h"
 
 
 /**
@@ -21,23 +12,21 @@ struct _egl_thread_info
  */
 struct _egl_global
 {
-   EGLBoolean Initialized;
+   _EGLMutex *Mutex;
 
-   _EGLHashtable *Displays;
-   _EGLHashtable *Surfaces;
+   /* the list of all displays */
+   _EGLDisplay *DisplayList;
 
    EGLScreenMESA FreeScreenHandle;
 
    /* bitmaks of supported APIs (supported by _some_ driver) */
    EGLint ClientAPIsMask;
 
-   char ClientAPIs[1000];   /**< updated by eglQueryString */
-
-   /* XXX temporary - should be thread-specific data (TSD) */
-   _EGLThreadInfo *ThreadInfo;
-
    EGLint NumDrivers;
    _EGLDriver *Drivers[10];
+
+   EGLint NumAtExitCalls;
+   void (*AtExitCalls[10])(void);
 };
 
 
@@ -45,27 +34,7 @@ extern struct _egl_global _eglGlobal;
 
 
 extern void
-_eglInitGlobals(void);
-
-
-extern void
-_eglDestroyGlobals(void);
-
-
-extern _EGLThreadInfo *
-_eglNewThreadInfo(void);
-
-
-extern void
-_eglDeleteThreadData(_EGLThreadInfo *t);
-
-
-extern _EGLThreadInfo *
-_eglGetCurrentThread(void);
-
-
-extern void
-_eglError(EGLint errCode, const char *msg);
+_eglAddAtExitCall(void (*func)(void));
 
 
 #endif /* EGLGLOBALS_INCLUDED */
diff --git a/src/egl/main/eglhash.c b/src/egl/main/eglhash.c
deleted file mode 100644
index 8e3da2e906..0000000000
--- a/src/egl/main/eglhash.c
+++ /dev/null
@@ -1,347 +0,0 @@
-/**
- * \file hash.c
- * Generic hash table. 
- *
- * This code taken from Mesa and adapted.
- */
-
-#include <assert.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include "eglhash.h"
-
-
-#define TABLE_SIZE 1023  /**< Size of lookup table/array */
-
-#define HASH_FUNC(K)  ((K) % TABLE_SIZE)
-
-
-/*
- * Unfinished mutex stuff
- */
-
-typedef int _EGLMutex;
-
-static void
-_eglInitMutex(_EGLMutex m)
-{
-}
-
-static void
-_eglDestroyMutex(_EGLMutex m)
-{
-}
-
-static void
-_eglLockMutex(_EGLMutex m)
-{
-}
-
-static void
-_eglUnlockMutex(_EGLMutex m)
-{
-}
-
-
-
-typedef struct _egl_hashentry _EGLHashentry;
-
-struct _egl_hashentry
-{
-   EGLuint Key;            /**< the entry's key */
-   void *Data;             /**< the entry's data */
-   _EGLHashentry *Next;    /**< pointer to next entry */
-};
-
-
-struct _egl_hashtable
-{
-   _EGLHashentry *Table[TABLE_SIZE];  /**< the lookup table */
-   EGLuint MaxKey;                    /**< highest key inserted so far */
-   _EGLMutex Mutex;                   /**< mutual exclusion lock */
-};
-
-
-/**
- * Create a new hash table.
- * 
- * \return pointer to a new, empty hash table.
- */
-_EGLHashtable *
-_eglNewHashTable(void)
-{
-   _EGLHashtable *table = (_EGLHashtable *) calloc(1, sizeof(_EGLHashtable));
-   if (table) {
-      _eglInitMutex(table->Mutex);
-      table->MaxKey = 1;
-   }
-   return table;
-}
-
-
-
-/**
- * Delete a hash table.
- * Frees each entry on the hash table and then the hash table structure itself.
- * Note that the caller should have already traversed the table and deleted
- * the objects in the table (i.e. We don't free the entries' data pointer).
- *
- * \param table the hash table to delete.
- */
-void
-_eglDeleteHashTable(_EGLHashtable *table)
-{
-   EGLuint i;
-   assert(table);
-   for (i = 0; i < TABLE_SIZE; i++) {
-      _EGLHashentry *entry = table->Table[i];
-      while (entry) {
-	 _EGLHashentry *next = entry->Next;
-	 free(entry);
-	 entry = next;
-      }
-   }
-   _eglDestroyMutex(table->Mutex);
-   free(table);
-}
-
-
-
-/**
- * Lookup an entry in the hash table.
- * 
- * \param table the hash table.
- * \param key the key.
- * 
- * \return pointer to user's data or NULL if key not in table
- */
-void *
-_eglHashLookup(const _EGLHashtable *table, EGLuint key)
-{
-   EGLuint pos;
-   const _EGLHashentry *entry;
-
-   assert(table);
-
-   if (!key)
-      return NULL;
-
-   pos = HASH_FUNC(key);
-   entry = table->Table[pos];
-   while (entry) {
-      if (entry->Key == key) {
-	 return entry->Data;
-      }
-      entry = entry->Next;
-   }
-   return NULL;
-}
-
-
-
-/**
- * Insert a key/pointer pair into the hash table.  
- * If an entry with this key already exists we'll replace the existing entry.
- * 
- * \param table the hash table.
- * \param key the key (not zero).
- * \param data pointer to user data.
- */
-void
-_eglHashInsert(_EGLHashtable *table, EGLuint key, void *data)
-{
-   /* search for existing entry with this key */
-   EGLuint pos;
-   _EGLHashentry *entry;
-
-   assert(table);
-   assert(key);
-
-   _eglLockMutex(table->Mutex);
-
-   if (key > table->MaxKey)
-      table->MaxKey = key;
-
-   pos = HASH_FUNC(key);
-   entry = table->Table[pos];
-   while (entry) {
-      if (entry->Key == key) {
-         /* replace entry's data */
-	 entry->Data = data;
-         _eglUnlockMutex(table->Mutex);
-	 return;
-      }
-      entry = entry->Next;
-   }
-
-   /* alloc and insert new table entry */
-   entry = (_EGLHashentry *) malloc(sizeof(_EGLHashentry));
-   entry->Key = key;
-   entry->Data = data;
-   entry->Next = table->Table[pos];
-   table->Table[pos] = entry;
-
-   _eglUnlockMutex(table->Mutex);
-}
-
-
-
-/**
- * Remove an entry from the hash table.
- * 
- * \param table the hash table.
- * \param key key of entry to remove.
- *
- * While holding the hash table's lock, searches the entry with the matching
- * key and unlinks it.
- */
-void
-_eglHashRemove(_EGLHashtable *table, EGLuint key)
-{
-   EGLuint pos;
-   _EGLHashentry *entry, *prev;
-
-   assert(table);
-   assert(key);
-
-   _eglLockMutex(table->Mutex);
-
-   pos = HASH_FUNC(key);
-   prev = NULL;
-   entry = table->Table[pos];
-   while (entry) {
-      if (entry->Key == key) {
-         /* found it! */
-         if (prev) {
-            prev->Next = entry->Next;
-         }
-         else {
-            table->Table[pos] = entry->Next;
-         }
-         free(entry);
-         _eglUnlockMutex(table->Mutex);
-	 return;
-      }
-      prev = entry;
-      entry = entry->Next;
-   }
-
-   _eglUnlockMutex(table->Mutex);
-}
-
-
-
-/**
- * Get the key of the "first" entry in the hash table.
- * 
- * This is used in the course of deleting all display lists when
- * a context is destroyed.
- * 
- * \param table the hash table
- * 
- * \return key for the "first" entry in the hash table.
- *
- * While holding the lock, walks through all table positions until finding
- * the first entry of the first non-empty one.
- */
-EGLuint
-_eglHashFirstEntry(_EGLHashtable *table)
-{
-   EGLuint pos;
-   assert(table);
-   _eglLockMutex(table->Mutex);
-   for (pos = 0; pos < TABLE_SIZE; pos++) {
-      if (table->Table[pos]) {
-         _eglUnlockMutex(table->Mutex);
-         return table->Table[pos]->Key;
-      }
-   }
-   _eglUnlockMutex(table->Mutex);
-   return 0;
-}
-
-
-/**
- * Given a hash table key, return the next key.  This is used to walk
- * over all entries in the table.  Note that the keys returned during
- * walking won't be in any particular order.
- * \return next hash key or 0 if end of table.
- */
-EGLuint
-_eglHashNextEntry(const _EGLHashtable *table, EGLuint key)
-{
-   const _EGLHashentry *entry;
-   EGLuint pos;
-
-   assert(table);
-   assert(key);
-
-   /* Find the entry with given key */
-   pos = HASH_FUNC(key);
-   entry = table->Table[pos];
-   while (entry) {
-      if (entry->Key == key) {
-         break;
-      }
-      entry = entry->Next;
-   }
-
-   if (!entry) {
-      /* the key was not found, we can't find next entry */
-      return 0;
-   }
-
-   if (entry->Next) {
-      /* return next in linked list */
-      return entry->Next->Key;
-   }
-   else {
-      /* look for next non-empty table slot */
-      pos++;
-      while (pos < TABLE_SIZE) {
-         if (table->Table[pos]) {
-            return table->Table[pos]->Key;
-         }
-         pos++;
-      }
-      return 0;
-   }
-}
-
-
-/**
- * Dump contents of hash table for debugging.
- * 
- * \param table the hash table.
- */
-void
-_eglHashPrint(const _EGLHashtable *table)
-{
-   EGLuint i;
-   assert(table);
-   for (i = 0; i < TABLE_SIZE; i++) {
-      const _EGLHashentry *entry = table->Table[i];
-      while (entry) {
-	 printf("%u %p\n", entry->Key, entry->Data);
-	 entry = entry->Next;
-      }
-   }
-}
-
-
-
-/**
- * Return a new, unused hash key.
- */
-EGLuint
-_eglHashGenKey(_EGLHashtable *table)
-{
-   EGLuint k;
-
-   _eglLockMutex(table->Mutex);
-   k = table->MaxKey;
-   table->MaxKey++;
-   _eglUnlockMutex(table->Mutex);
-   return k;
-}
-
diff --git a/src/egl/main/eglhash.h b/src/egl/main/eglhash.h
deleted file mode 100644
index 1d6db9598c..0000000000
--- a/src/egl/main/eglhash.h
+++ /dev/null
@@ -1,39 +0,0 @@
-/**
- * \file eglhash.h
- * Generic hash table. 
- */
-
-
-#ifndef EGLHASH_INCLUDED
-#define EGLHASH_INCLUDED
-
-
-/* XXX move this? */
-typedef unsigned int EGLuint;
-
-
-typedef struct _egl_hashtable _EGLHashtable;
-
-
-extern _EGLHashtable *_eglNewHashTable(void);
-
-extern void _eglDeleteHashTable(_EGLHashtable *table);
-
-extern void *_eglHashLookup(const _EGLHashtable *table, EGLuint key);
-
-extern void _eglHashInsert(_EGLHashtable *table, EGLuint key, void *data);
-
-extern void _eglHashRemove(_EGLHashtable *table, EGLuint key);
-
-extern EGLuint _eglHashFirstEntry(_EGLHashtable *table);
-
-extern EGLuint _eglHashNextEntry(const _EGLHashtable *table, EGLuint key);
-
-extern void _eglHashPrint(const _EGLHashtable *table);
-
-extern EGLuint _eglHashGenKey(_EGLHashtable *table);
-
-extern void _egltest_hash_functions(void);
-
-
-#endif /* EGLHASH_INCLUDED */
diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c
index b5bdc3ea4b..b37213faf1 100644
--- a/src/egl/main/eglmisc.c
+++ b/src/egl/main/eglmisc.c
@@ -35,6 +35,7 @@
 #include <string.h>
 #include "eglglobals.h"
 #include "eglmisc.h"
+#include "egldisplay.h"
 
 
 /**
@@ -42,42 +43,47 @@
  * the driver's Extensions string.
  */
 static void
-_eglUpdateExtensionsString(_EGLDriver *drv)
+_eglUpdateExtensionsString(_EGLDisplay *dpy)
 {
-   drv->Extensions.String[0] = 0;
+   char *exts = dpy->Extensions.String;
 
-   if (drv->Extensions.MESA_screen_surface)
-      strcat(drv->Extensions.String, "EGL_MESA_screen_surface ");
-   if (drv->Extensions.MESA_copy_context)
-      strcat(drv->Extensions.String, "EGL_MESA_copy_context ");
-   assert(strlen(drv->Extensions.String) < _EGL_MAX_EXTENSIONS_LEN);
+   if (exts[0])
+      return;
+
+   if (dpy->Extensions.MESA_screen_surface)
+      strcat(exts, "EGL_MESA_screen_surface ");
+   if (dpy->Extensions.MESA_copy_context)
+      strcat(exts, "EGL_MESA_copy_context ");
+   assert(strlen(exts) < _EGL_MAX_EXTENSIONS_LEN);
 }
 
 
 static void
-_eglUpdateAPIsString(_EGLDriver *drv)
+_eglUpdateAPIsString(_EGLDisplay *dpy)
 {
-   _eglGlobal.ClientAPIs[0] = 0;
+   char *apis = dpy->ClientAPIs;
 
-   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT)
-      strcat(_eglGlobal.ClientAPIs, "OpenGL ");
+   if (apis[0] || !dpy->ClientAPIsMask)
+      return;
 
-   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES_BIT)
-      strcat(_eglGlobal.ClientAPIs, "OpenGL_ES ");
+   if (dpy->ClientAPIsMask & EGL_OPENGL_BIT)
+      strcat(apis, "OpenGL ");
 
-   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES2_BIT)
-      strcat(_eglGlobal.ClientAPIs, "OpenGL_ES2 ");
+   if (dpy->ClientAPIsMask & EGL_OPENGL_ES_BIT)
+      strcat(apis, "OpenGL_ES ");
 
-   if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT)
-      strcat(_eglGlobal.ClientAPIs, "OpenVG ");
+   if (dpy->ClientAPIsMask & EGL_OPENGL_ES2_BIT)
+      strcat(apis, "OpenGL_ES2 ");
 
-   assert(strlen(_eglGlobal.ClientAPIs) < sizeof(_eglGlobal.ClientAPIs));
-}
+   if (dpy->ClientAPIsMask & EGL_OPENVG_BIT)
+      strcat(apis, "OpenVG ");
 
+   assert(strlen(apis) < sizeof(dpy->ClientAPIs));
+}
 
 
 const char *
-_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name)
+_eglQueryString(_EGLDriver *drv, _EGLDisplay *dpy, EGLint name)
 {
    (void) drv;
    (void) dpy;
@@ -85,14 +91,14 @@ _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name)
    case EGL_VENDOR:
       return _EGL_VENDOR_STRING;
    case EGL_VERSION:
-      return drv->Version;
+      return dpy->Version;
    case EGL_EXTENSIONS:
-      _eglUpdateExtensionsString(drv);
-      return drv->Extensions.String;
+      _eglUpdateExtensionsString(dpy);
+      return dpy->Extensions.String;
 #ifdef EGL_VERSION_1_2
    case EGL_CLIENT_APIS:
-      _eglUpdateAPIsString(drv);
-      return _eglGlobal.ClientAPIs;
+      _eglUpdateAPIsString(dpy);
+      return dpy->ClientAPIs;
 #endif
    default:
       _eglError(EGL_BAD_PARAMETER, "eglQueryString");
@@ -102,7 +108,7 @@ _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name)
 
 
 EGLBoolean
-_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy)
+_eglWaitGL(_EGLDriver *drv, _EGLDisplay *dpy)
 {
    /* just a placeholder */
    (void) drv;
@@ -112,7 +118,7 @@ _eglWaitGL(_EGLDriver *drv, EGLDisplay dpy)
 
 
 EGLBoolean
-_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine)
+_eglWaitNative(_EGLDriver *drv, _EGLDisplay *dpy, EGLint engine)
 {
    /* just a placeholder */
    (void) drv;
diff --git a/src/egl/main/eglmisc.h b/src/egl/main/eglmisc.h
index 4e2a40ea99..a15c839be2 100644
--- a/src/egl/main/eglmisc.h
+++ b/src/egl/main/eglmisc.h
@@ -33,15 +33,15 @@
 
 
 extern const char *
-_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name);
+_eglQueryString(_EGLDriver *drv, _EGLDisplay *dpy, EGLint name);
 
 
 extern EGLBoolean
-_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy);
+_eglWaitGL(_EGLDriver *drv, _EGLDisplay *dpy);
 
 
 extern EGLBoolean
-_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine);
+_eglWaitNative(_EGLDriver *drv, _EGLDisplay *dpy, EGLint engine);
 
 
 #endif /* EGLMISC_INCLUDED */
diff --git a/src/egl/main/eglmode.c b/src/egl/main/eglmode.c
index 786432234b..0f3ba6e5c0 100644
--- a/src/egl/main/eglmode.c
+++ b/src/egl/main/eglmode.c
@@ -34,9 +34,8 @@ my_strdup(const char *s)
  * or null if non-existant.
  */
 _EGLMode *
-_eglLookupMode(EGLDisplay dpy, EGLModeMESA mode)
+_eglLookupMode(EGLModeMESA mode, _EGLDisplay *disp)
 {
-   const _EGLDisplay *disp = _eglLookupDisplay(dpy);
    EGLint scrnum;
 
    /* loop over all screens on the display */
@@ -272,19 +271,13 @@ _eglCompareModes(const void *a, const void *b)
  * Called via eglChooseModeMESA API function.
  */
 EGLBoolean
-_eglChooseModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
+_eglChooseModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
                    const EGLint *attrib_list, EGLModeMESA *modes,
                    EGLint modes_size, EGLint *num_modes)
 {
-   const _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
    _EGLMode **modeList, min;
    EGLint i, count;
 
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglChooseModeMESA");
-      return EGL_FALSE;
-   }
-
    if (!_eglParseModeAttribs(&min, attrib_list)) {
       /* error code will have been recorded */
       return EGL_FALSE;
@@ -326,16 +319,9 @@ _eglChooseModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
  * Called via eglGetModesMESA() API function.
  */
 EGLBoolean
-_eglGetModesMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
+_eglGetModesMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
                  EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes)
 {
-   _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglGetModesMESA");
-      return EGL_FALSE;
-   }
-
    if (modes) {
       EGLint i;
       *num_modes = MIN2(scrn->NumModes, modes_size);
@@ -356,17 +342,11 @@ _eglGetModesMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
  * Query an attribute of a mode.
  */
 EGLBoolean
-_eglGetModeAttribMESA(_EGLDriver *drv, EGLDisplay dpy,
-                      EGLModeMESA mode, EGLint attribute, EGLint *value)
+_eglGetModeAttribMESA(_EGLDriver *drv, _EGLDisplay *dpy,
+                      _EGLMode *m, EGLint attribute, EGLint *value)
 {
-   _EGLMode *m = _eglLookupMode(dpy, mode);
    EGLint v;
 
-   if (!m) {
-      _eglError(EGL_BAD_MODE_MESA, "eglGetModeAttribMESA");
-      return EGL_FALSE;
-   }
-
    v = getModeAttrib(m, attribute);
    if (v < 0) {
       _eglError(EGL_BAD_ATTRIBUTE, "eglGetModeAttribMESA");
@@ -382,13 +362,8 @@ _eglGetModeAttribMESA(_EGLDriver *drv, EGLDisplay dpy,
  * This is the default function called by eglQueryModeStringMESA().
  */
 const char *
-_eglQueryModeStringMESA(_EGLDriver *drv, EGLDisplay dpy, EGLModeMESA mode)
+_eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m)
 {
-   _EGLMode *m = _eglLookupMode(dpy, mode);
-   if (!m) {
-      _eglError(EGL_BAD_MODE_MESA, "eglQueryModeStringMESA");
-      return NULL;
-   }
    return m->Name;
 }
 
diff --git a/src/egl/main/eglmode.h b/src/egl/main/eglmode.h
index 52d4875676..af7c2c56d3 100644
--- a/src/egl/main/eglmode.h
+++ b/src/egl/main/eglmode.h
@@ -26,7 +26,7 @@ struct _egl_mode
 
 
 extern _EGLMode *
-_eglLookupMode(EGLDisplay dpy, EGLModeMESA mode);
+_eglLookupMode(EGLModeMESA mode, _EGLDisplay *dpy);
 
 
 extern _EGLMode *
@@ -35,23 +35,23 @@ _eglAddNewMode(_EGLScreen *screen, EGLint width, EGLint height,
 
 
 extern EGLBoolean
-_eglChooseModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
+_eglChooseModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
                    const EGLint *attrib_list, EGLModeMESA *modes,
                    EGLint modes_size, EGLint *num_modes);
 
 
 extern EGLBoolean
-_eglGetModesMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
+_eglGetModesMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
                  EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes);
 
 
 extern EGLBoolean
-_eglGetModeAttribMESA(_EGLDriver *drv, EGLDisplay dpy, EGLModeMESA mode,
+_eglGetModeAttribMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m,
                       EGLint attribute, EGLint *value);
 
 
 extern const char *
-_eglQueryModeStringMESA(_EGLDriver *drv, EGLDisplay dpy, EGLModeMESA mode);
+_eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m);
 
 
 #endif /* EGLMODE_INCLUDED */
diff --git a/src/egl/main/eglmutex.h b/src/egl/main/eglmutex.h
new file mode 100644
index 0000000000..29faba0f24
--- /dev/null
+++ b/src/egl/main/eglmutex.h
@@ -0,0 +1,52 @@
+#ifndef EGLMUTEX_INCLUDED
+#define EGLMUTEX_INCLUDED
+
+#include "eglcompiler.h"
+
+#ifdef PTHREADS
+#include <pthread.h>
+
+typedef pthread_mutex_t _EGLMutex;
+
+static INLINE void _eglInitMutex(_EGLMutex *m)
+{
+   pthread_mutex_init(m, NULL);
+}
+
+static INLINE void
+_eglDestroyMutex(_EGLMutex *m)
+{
+   pthread_mutex_destroy(m);
+}
+
+static INLINE void
+_eglLockMutex(_EGLMutex *m)
+{
+   pthread_mutex_lock(m);
+}
+
+static INLINE void
+_eglUnlockMutex(_EGLMutex *m)
+{
+   pthread_mutex_unlock(m);
+}
+
+#define _EGL_MUTEX_INITIALIZER PTHREAD_MUTEX_INITIALIZER
+#define _EGL_DECLARE_MUTEX(m) \
+   _EGLMutex m = _EGL_MUTEX_INITIALIZER
+
+#else
+
+typedef int _EGLMutex;
+static INLINE void _eglInitMutex(_EGLMutex *m) { (void) m; }
+static INLINE void _eglDestroyMutex(_EGLMutex *m) { (void) m; }
+static INLINE void _eglLockMutex(_EGLMutex *m) { (void) m; }
+static INLINE void _eglUnlockMutex(_EGLMutex *m) { (void) m; }
+
+#define _EGL_MUTEX_INITIALIZER 0
+#define _EGL_DECLARE_MUTEX(m) \
+   _EGLMutex m = _EGL_MUTEX_INITIALIZER
+
+#endif
+
+#endif /* EGLMUTEX_INCLUDED */
diff --git a/src/egl/main/eglscreen.c b/src/egl/main/eglscreen.c
index 9c9a8377bf..14a1e9f8fe 100644
--- a/src/egl/main/eglscreen.c
+++ b/src/egl/main/eglscreen.c
@@ -52,13 +52,9 @@ _eglInitScreen(_EGLScreen *screen)
  * Given a public screen handle, return the internal _EGLScreen object.
  */
 _EGLScreen *
-_eglLookupScreen(EGLDisplay dpy, EGLScreenMESA screen)
+_eglLookupScreen(EGLScreenMESA screen, _EGLDisplay *display)
 {
    EGLint i;
-   _EGLDisplay *display = _eglLookupDisplay(dpy);
-
-   if (!display)
-      return NULL;
 
    for (i = 0; i < display->NumScreens; i++) {
       if (display->Screens[i]->Handle == screen)
@@ -89,17 +85,11 @@ _eglAddScreen(_EGLDisplay *display, _EGLScreen *screen)
 
 
 EGLBoolean
-_eglGetScreensMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA *screens,
+_eglGetScreensMESA(_EGLDriver *drv, _EGLDisplay *display, EGLScreenMESA *screens,
                    EGLint max_screens, EGLint *num_screens)
 {
-   _EGLDisplay *display = _eglLookupDisplay(dpy);
    EGLint n;
 
-   if (!display) {
-      _eglError(EGL_BAD_DISPLAY, "eglGetScreensMESA");
-      return EGL_FALSE;
-   }
-
    if (display->NumScreens > max_screens) {
       n = max_screens;
    }
@@ -122,8 +112,8 @@ _eglGetScreensMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA *screens,
 /**
  * Example function - drivers should do a proper implementation.
  */
-EGLSurface
-_eglCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_EGLSurface *
+_eglCreateScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                             const EGLint *attrib_list)
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
@@ -131,19 +121,17 @@ _eglCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_SCREEN_BIT_MESA,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_SCREEN_BIT_MESA,
+                        conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return surf;
 #endif
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
@@ -155,28 +143,15 @@ _eglCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * this with code that _really_ shows the surface.
  */
 EGLBoolean
-_eglShowScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy,
-                          EGLScreenMESA screen, EGLSurface surface,
-                          EGLModeMESA m)
+_eglShowScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy,
+                          _EGLScreen *scrn, _EGLSurface *surf,
+                          _EGLMode *mode)
 {
-   _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-   _EGLMode *mode = _eglLookupMode(dpy, m);
-
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglShowSurfaceMESA");
-      return EGL_FALSE;
-   }
-   if (!mode && (m != EGL_NO_MODE_MESA )) {
-      _eglError(EGL_BAD_MODE_MESA, "eglShowSurfaceMESA");
-      return EGL_FALSE;
-   }
-
-   if (surface == EGL_NO_SURFACE) {
+   if (!surf) {
       scrn->CurrentSurface = NULL;
    }
    else {
-      _EGLSurface *surf = _eglLookupSurface(surface);
-      if (!surf || surf->Type != EGL_SCREEN_BIT_MESA) {
+      if (surf->Type != EGL_SCREEN_BIT_MESA) {
          _eglError(EGL_BAD_SURFACE, "eglShowSurfaceMESA");
          return EGL_FALSE;
       }
@@ -201,18 +176,10 @@ _eglShowScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy,
  * this with code that _really_ sets the mode.
  */
 EGLBoolean
-_eglScreenModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
-                   EGLModeMESA mode)
+_eglScreenModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
+                   _EGLMode *m)
 {
-   _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglScreenModeMESA");
-      return EGL_FALSE;
-   }
-
-   scrn->CurrentMode = _eglLookupMode(dpy, mode);
-
+   scrn->CurrentMode = m;
    return EGL_TRUE;
 }
 
@@ -221,15 +188,9 @@ _eglScreenModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
  * Set a screen's surface origin.
  */
 EGLBoolean
-_eglScreenPositionMESA(_EGLDriver *drv, EGLDisplay dpy,
-                       EGLScreenMESA screen, EGLint x, EGLint y)
+_eglScreenPositionMESA(_EGLDriver *drv, _EGLDisplay *dpy,
+                       _EGLScreen *scrn, EGLint x, EGLint y)
 {
-   _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglScreenPositionMESA");
-      return EGL_FALSE;
-   }
-
    scrn->OriginX = x;
    scrn->OriginY = y;
 
@@ -241,14 +202,10 @@ _eglScreenPositionMESA(_EGLDriver *drv, EGLDisplay dpy,
  * Query a screen's current surface.
  */
 EGLBoolean
-_eglQueryScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy,
-                           EGLScreenMESA screen, EGLSurface *surface)
+_eglQueryScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy,
+                           _EGLScreen *scrn, _EGLSurface **surf)
 {
-   const _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-   if (scrn->CurrentSurface)
-      *surface = scrn->CurrentSurface->Handle;
-   else
-      *surface = EGL_NO_SURFACE;
+   *surf = scrn->CurrentSurface;
    return EGL_TRUE;
 }
 
@@ -257,29 +214,18 @@ _eglQueryScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy,
  * Query a screen's current mode.
  */
 EGLBoolean
-_eglQueryScreenModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
-                        EGLModeMESA *mode)
+_eglQueryScreenModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
+                        _EGLMode **m)
 {
-   const _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-   if (scrn->CurrentMode)
-      *mode = scrn->CurrentMode->Handle;
-   else
-      *mode = EGL_NO_MODE_MESA;
+   *m = scrn->CurrentMode;
    return EGL_TRUE;
 }
 
 
 EGLBoolean
-_eglQueryScreenMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen,
+_eglQueryScreenMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn,
                     EGLint attribute, EGLint *value)
 {
-   const _EGLScreen *scrn = _eglLookupScreen(dpy, screen);
-
-   if (!scrn) {
-      _eglError(EGL_BAD_SCREEN_MESA, "eglQueryScreenMESA");
-      return EGL_FALSE;
-   }
-
    switch (attribute) {
    case EGL_SCREEN_POSITION_MESA:
       value[0] = scrn->OriginX;
diff --git a/src/egl/main/eglscreen.h b/src/egl/main/eglscreen.h
index 833439b410..8860a2aa7f 100644
--- a/src/egl/main/eglscreen.h
+++ b/src/egl/main/eglscreen.h
@@ -35,7 +35,7 @@ _eglInitScreen(_EGLScreen *screen);
 
 
 extern _EGLScreen *
-_eglLookupScreen(EGLDisplay dpy, EGLScreenMESA screen);
+_eglLookupScreen(EGLScreenMESA screen, _EGLDisplay *dpy);
 
 
 extern void
@@ -43,40 +43,40 @@ _eglAddScreen(_EGLDisplay *display, _EGLScreen *screen);
 
 
 extern EGLBoolean
-_eglGetScreensMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA *screens, EGLint max_screens, EGLint *num_screens);
+_eglGetScreensMESA(_EGLDriver *drv, _EGLDisplay *dpy, EGLScreenMESA *screens, EGLint max_screens, EGLint *num_screens);
 
 
-extern EGLSurface
-_eglCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
+extern _EGLSurface *
+_eglCreateScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list);
 
 
 extern EGLBoolean
-_eglShowScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface surface, EGLModeMESA mode);
+_eglShowScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn, _EGLSurface *surf, _EGLMode *m);
 
 
 extern EGLBoolean
-_eglScreenModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA mode);
+_eglScreenModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn, _EGLMode *m);
 
 
 extern EGLBoolean
-_eglScreenPositionMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLint x, EGLint y);
+_eglScreenPositionMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn, EGLint x, EGLint y);
 
 
 extern EGLBoolean
-_eglQueryDisplayMESA(_EGLDriver *drv, EGLDisplay dpy, EGLint attribute, EGLint *value);
+_eglQueryDisplayMESA(_EGLDriver *drv, _EGLDisplay *dpy, EGLint attribute, EGLint *value);
 
 
 extern EGLBoolean
-_eglQueryScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy,
-                           EGLScreenMESA screen, EGLSurface *surface);
+_eglQueryScreenSurfaceMESA(_EGLDriver *drv, _EGLDisplay *dpy,
+                           _EGLScreen *scrn, _EGLSurface **surface);
 
 
 extern EGLBoolean
-_eglQueryScreenModeMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *mode);
+_eglQueryScreenModeMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn, _EGLMode **m);
 
 
 extern EGLBoolean
-_eglQueryScreenMESA(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLint attribute, EGLint *value);
+_eglQueryScreenMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *scrn, EGLint attribute, EGLint *value);
 
 
 extern void
diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c
index 6905acac50..e7a1a8329e 100644
--- a/src/egl/main/eglsurface.c
+++ b/src/egl/main/eglsurface.c
@@ -6,11 +6,11 @@
 #include <assert.h>
 #include <stdlib.h>
 #include <string.h>
+#include "egldisplay.h"
 #include "eglcontext.h"
 #include "eglconfig.h"
 #include "egldriver.h"
 #include "eglglobals.h"
-#include "eglhash.h"
 #include "egllog.h"
 #include "eglsurface.h"
 
@@ -20,14 +20,13 @@
  * \return EGL_TRUE if no errors, EGL_FALSE otherwise.
  */
 EGLBoolean
-_eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
-                _EGLSurface *surf, EGLint type, EGLConfig config,
-                const EGLint *attrib_list)
+_eglInitSurface(_EGLDriver *drv, _EGLSurface *surf, EGLint type,
+                _EGLConfig *conf, const EGLint *attrib_list)
 {
    const char *func;
-   _EGLConfig *conf;
    EGLint width = 0, height = 0, largest = 0;
-   EGLint texFormat = 0, texTarget = 0, mipmapTex = 0;
+   EGLint texFormat = EGL_NO_TEXTURE, texTarget = EGL_NO_TEXTURE;
+   EGLint mipmapTex = EGL_FALSE;
    EGLint renderBuffer = EGL_BACK_BUFFER;
 #ifdef EGL_VERSION_1_2
    EGLint colorspace = EGL_COLORSPACE_sRGB;
@@ -55,7 +54,6 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
       return EGL_FALSE;
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
    if (!conf) {
       _eglError(EGL_BAD_CONFIG, func);
       return EGL_FALSE;
@@ -211,89 +209,16 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
 }
 
 
-void
-_eglSaveSurface(_EGLSurface *surf)
-{
-   EGLuint key = _eglHashGenKey(_eglGlobal.Surfaces);
-   assert(surf);
-   assert(!surf->Handle);
-   surf->Handle = (EGLSurface) key;
-   assert(surf->Handle);
-   _eglHashInsert(_eglGlobal.Surfaces, key, surf);
-}
-
-
-void
-_eglRemoveSurface(_EGLSurface *surf)
-{
-   _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surf->Handle);
-}
-
-
-
-/**
- * Return the public handle for an internal _EGLSurface.
- * This is the inverse of _eglLookupSurface().
- */
-EGLSurface
-_eglGetSurfaceHandle(_EGLSurface *surface)
-{
-   if (surface)
-      return surface->Handle;
-   else
-      return EGL_NO_SURFACE;
-}
-
-
-/**
- * Return the private _EGLSurface which corresponds to a public EGLSurface
- * handle.
- * This is the inverse of _eglGetSurfaceHandle().
- */
-_EGLSurface *
-_eglLookupSurface(EGLSurface surf)
-{
-   _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces,
-                                                   (EGLuint) surf);
-   return c;
-}
-
-
-_EGLSurface *
-_eglGetCurrentSurface(EGLint readdraw)
-{
-   _EGLContext *ctx = _eglGetCurrentContext();
-   if (ctx) {
-      switch (readdraw) {
-      case EGL_DRAW:
-         return ctx->DrawSurface;
-      case EGL_READ:
-         return ctx->ReadSurface;
-      default:
-         return NULL;
-      }
-   }
-   return NULL;
-}
-
-
 EGLBoolean
-_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf)
 {
-   /* Basically just do error checking here.  Drivers have to do the
-    * actual buffer swap.
-    */
-   _EGLSurface *surface = _eglLookupSurface(draw);
-   if (surface == NULL) {
-      _eglError(EGL_BAD_SURFACE, "eglSwapBuffers");
-      return EGL_FALSE;
-   }
+   /* Drivers have to do the actual buffer swap.  */
    return EGL_TRUE;
 }
 
 
 EGLBoolean
-_eglCopyBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+_eglCopyBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf,
                 NativePixmapType target)
 {
    /* copy surface to native pixmap */
@@ -303,14 +228,9 @@ _eglCopyBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
 
 
 EGLBoolean
-_eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+_eglQuerySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface,
                  EGLint attribute, EGLint *value)
 {
-   _EGLSurface *surface = _eglLookupSurface(surf);
-   if (surface == NULL) {
-      _eglError(EGL_BAD_SURFACE, "eglQuerySurface");
-      return EGL_FALSE;
-   }
    switch (attribute) {
    case EGL_WIDTH:
       *value = surface->Width;
@@ -322,7 +242,7 @@ _eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
       *value = GET_CONFIG_ATTRIB(surface->Config, EGL_CONFIG_ID);
       return EGL_TRUE;
    case EGL_LARGEST_PBUFFER:
-      *value = drv->LargestPbuffer;
+      *value = dpy->LargestPbuffer;
       return EGL_TRUE;
    case EGL_SURFACE_TYPE:
       *value = surface->Type;
@@ -379,8 +299,8 @@ _eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
 /**
  * Example function - drivers should do a proper implementation.
  */
-EGLSurface
-_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_EGLSurface *
+_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                         NativeWindowType window, const EGLint *attrib_list)
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
@@ -388,26 +308,24 @@ _eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_WINDOW_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_WINDOW_BIT, conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return surf;
 #endif
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
 /**
  * Example function - drivers should do a proper implementation.
  */
-EGLSurface
-_eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_EGLSurface *
+_eglCreatePixmapSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                         NativePixmapType pixmap, const EGLint *attrib_list)
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
@@ -415,26 +333,24 @@ _eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_PIXMAP_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_PIXMAP_BIT, conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return surf;
 #endif
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
 /**
  * Example function - drivers should do a proper implementation.
  */
-EGLSurface
-_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_EGLSurface *
+_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                          const EGLint *attrib_list)
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
@@ -442,18 +358,16 @@ _eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_PBUFFER_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_PBUFFER_BIT, conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return NULL;
 #endif
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 
@@ -461,23 +375,11 @@ _eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
  * Default fallback routine - drivers should usually override this.
  */
 EGLBoolean
-_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf)
 {
-   _EGLSurface *surf = _eglLookupSurface(surface);
-   if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->IsBound) {
-         surf->DeletePending = EGL_TRUE;
-      }
-      else {
-         free(surf);
-      }
-      return EGL_TRUE;
-   }
-   else {
-      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
-      return EGL_FALSE;
-   }
+   if (!_eglIsSurfaceBound(surf))
+      free(surf);
+   return EGL_TRUE;
 }
 
 
@@ -485,16 +387,9 @@ _eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
  * Default fallback routine - drivers might override this.
  */
 EGLBoolean
-_eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+_eglSurfaceAttrib(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface,
                   EGLint attribute, EGLint value)
 {
-   _EGLSurface *surface = _eglLookupSurface(surf);
-
-   if (surface == NULL) {
-      _eglError(EGL_BAD_SURFACE, "eglSurfaceAttrib");
-      return EGL_FALSE;
-   }
-
    switch (attribute) {
    case EGL_MIPMAP_LEVEL:
       surface->MipmapLevel = value;
@@ -508,15 +403,14 @@ _eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
 
 
 EGLBoolean
-_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface,
                  EGLint buffer)
 {
    /* Just do basic error checking and return success/fail.
     * Drivers must implement the real stuff.
     */
-   _EGLSurface *surface = _eglLookupSurface(surf);
 
-   if (!surface || surface->Type != EGL_PBUFFER_BIT) {
+   if (surface->Type != EGL_PBUFFER_BIT) {
       _eglError(EGL_BAD_SURFACE, "eglBindTexImage");
       return EGL_FALSE;
    }
@@ -538,15 +432,14 @@ _eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
 
 
 EGLBoolean
-_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface,
                     EGLint buffer)
 {
    /* Just do basic error checking and return success/fail.
     * Drivers must implement the real stuff.
     */
-   _EGLSurface *surface = _eglLookupSurface(surf);
 
-   if (!surface || surface->Type != EGL_PBUFFER_BIT) {
+   if (surface->Type != EGL_PBUFFER_BIT) {
       _eglError(EGL_BAD_SURFACE, "eglBindTexImage");
       return EGL_FALSE;
    }
@@ -573,14 +466,11 @@ _eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
 
 
 EGLBoolean
-_eglSwapInterval(_EGLDriver *drv, EGLDisplay dpy, EGLint interval)
+_eglSwapInterval(_EGLDriver *drv, _EGLDisplay *dpy, EGLint interval)
 {
    _EGLSurface *surf = _eglGetCurrentSurface(EGL_DRAW);
-   if (surf == NULL) {
-      _eglError(EGL_BAD_SURFACE, "eglSwapInterval");
-      return EGL_FALSE;
-   }
-   surf->SwapInterval = interval;
+   if (surf)
+      surf->SwapInterval = interval;
    return EGL_TRUE;
 }
 
@@ -590,17 +480,17 @@ _eglSwapInterval(_EGLDriver *drv, EGLDisplay dpy, EGLint interval)
 /**
  * Example function - drivers should do a proper implementation.
  */
-EGLSurface
-_eglCreatePbufferFromClientBuffer(_EGLDriver *drv, EGLDisplay dpy,
+_EGLSurface *
+_eglCreatePbufferFromClientBuffer(_EGLDriver *drv, _EGLDisplay *dpy,
                                   EGLenum buftype, EGLClientBuffer buffer,
-                                  EGLConfig config, const EGLint *attrib_list)
+                                  _EGLConfig *conf, const EGLint *attrib_list)
 {
    if (buftype != EGL_OPENVG_IMAGE) {
       _eglError(EGL_BAD_PARAMETER, "eglCreatePbufferFromClientBuffer");
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   return EGL_NO_SURFACE;
+   return NULL;
 }
 
 #endif /* EGL_VERSION_1_2 */
diff --git a/src/egl/main/eglsurface.h b/src/egl/main/eglsurface.h
index 50f965b5cb..f6d44b5922 100644
--- a/src/egl/main/eglsurface.h
+++ b/src/egl/main/eglsurface.h
@@ -10,14 +10,16 @@
  */
 struct _egl_surface
 {
-   EGLSurface Handle;  /* The public/opaque handle which names this object */
-   _EGLConfig *Config;
+   /* Managed by EGLDisplay for linking */
+   _EGLDisplay *Display;
+   _EGLSurface *Next;
 
-   /* May need reference counting here */
-   EGLBoolean IsBound;
-   EGLBoolean DeletePending;
+   /* The bound status of the surface */
+   _EGLContext *Binding;
    EGLBoolean BoundToTexture;
 
+   _EGLConfig *Config;
+
    EGLint Type; /* one of EGL_WINDOW_BIT, EGL_PIXMAP_BIT or EGL_PBUFFER_BIT */
    EGLint Width, Height;
    EGLint TextureFormat, TextureTarget;
@@ -39,84 +41,74 @@ struct _egl_surface
 
 
 extern EGLBoolean
-_eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
-                _EGLSurface *surf, EGLint type, EGLConfig config,
-                const EGLint *attrib_list);
-
-
-extern void
-_eglSaveSurface(_EGLSurface *surf);
-
-
-extern void
-_eglRemoveSurface(_EGLSurface *surf);
-
-
-extern EGLSurface
-_eglGetSurfaceHandle(_EGLSurface *surface);
-
-
-extern _EGLSurface *
-_eglLookupSurface(EGLSurface surf);
- 
-
-extern _EGLSurface *
-_eglGetCurrentSurface(EGLint readdraw);
+_eglInitSurface(_EGLDriver *drv, _EGLSurface *surf, EGLint type,
+                _EGLConfig *config, const EGLint *attrib_list);
 
 
 extern EGLBoolean
-_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw);
+_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf);
 
 
 extern EGLBoolean
-_eglCopyBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, NativePixmapType target);
+_eglCopyBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf, NativePixmapType target);
 
 
 extern EGLBoolean
-_eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *value);
+_eglQuerySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf, EGLint attribute, EGLint *value);
 
 
-extern EGLSurface
-_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list);
+extern _EGLSurface *
+_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list);
 
 
-extern EGLSurface
-_eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list);
+extern _EGLSurface *
+_eglCreatePixmapSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list);
 
 
-extern EGLSurface
-_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
+extern _EGLSurface *
+_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list);
 
 
 extern EGLBoolean
-_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface);
+_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf);
 
 
 extern EGLBoolean
-_eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value);
+_eglSurfaceAttrib(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf, EGLint attribute, EGLint value);
 
 
 extern EGLBoolean
-_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf, EGLint buffer);
 
 
 extern EGLBoolean
-_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf, EGLint buffer);
 
 
 extern EGLBoolean
-_eglSwapInterval(_EGLDriver *drv, EGLDisplay dpy, EGLint interval);
+_eglSwapInterval(_EGLDriver *drv, _EGLDisplay *dpy, EGLint interval);
 
 
 #ifdef EGL_VERSION_1_2
 
-extern EGLSurface
-_eglCreatePbufferFromClientBuffer(_EGLDriver *drv, EGLDisplay dpy,
+extern _EGLSurface *
+_eglCreatePbufferFromClientBuffer(_EGLDriver *drv, _EGLDisplay *dpy,
                                   EGLenum buftype, EGLClientBuffer buffer,
-                                  EGLConfig config, const EGLint *attrib_list);
+                                  _EGLConfig *conf, const EGLint *attrib_list);
 
 #endif /* EGL_VERSION_1_2 */
 
 
+/**
+ * Return true if the surface is bound to a thread.
+ * A surface bound to a texutre is not considered bound by
+ * this function.
+ */
+static INLINE EGLBoolean
+_eglIsSurfaceBound(_EGLSurface *surf)
+{
+   return (surf->Binding != NULL);
+}
+
 
 #endif /* EGLSURFACE_INCLUDED */
diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h
index 9fbc55352c..4461440b9b 100644
--- a/src/egl/main/egltypedefs.h
+++ b/src/egl/main/egltypedefs.h
@@ -6,6 +6,7 @@
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
 
+#include "eglcompiler.h"
 
 typedef struct _egl_api _EGLAPI;
 
@@ -28,7 +29,7 @@ typedef struct _egl_surface _EGLSurface;
 typedef struct _egl_thread_info _EGLThreadInfo;
 
 
-typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy, const char *args);
+typedef _EGLDriver *(*_EGLMain_t)(const char *args);
 
 
 #endif /* EGLTYPEDEFS_INCLUDED */
diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c
deleted file mode 100644
index 50acc3a24f..0000000000
--- a/src/egl/main/eglx.c
+++ /dev/null
@@ -1,100 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-
-/**
- * X-specific EGL code.
- *
- * Any glue code needed to make EGL work with X is placed in this file.
- */
-
-
-#include <assert.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <X11/Xlib.h>
-
-#include "egldriver.h"
-#include "egllog.h"
-#include "eglstring.h"
-#include "eglx.h"
-
-
-static const char *DefaultGLXDriver = "egl_glx";
-static const char *DefaultSoftDriver = "egl_softpipe";
-
-
-/**
- * Given an X Display ptr (at dpy->Xdpy) try to determine the appropriate
- * device driver.  Return its name.
- *
- * This boils down to whether to use the egl_glx.so driver which will
- * load a DRI driver or the egl_softpipe.so driver that'll do software
- * rendering on Xlib.
- */
-const char *
-_xeglChooseDriver(_EGLDisplay *dpy)
-{
-#ifdef _EGL_PLATFORM_X
-   _XPrivDisplay xdpy;
-   int screen;
-   const char *driverName;
-
-   assert(dpy);
-
-   if (!dpy->Xdpy) {
-      dpy->Xdpy = XOpenDisplay(NULL);
-      if (!dpy->Xdpy) {
-         /* can't open X display -> can't use X-based driver */
-         return NULL;
-      }
-   }
-   xdpy = (_XPrivDisplay) dpy->Xdpy;
-
-   assert(dpy->Xdpy);
-
-   screen = DefaultScreen(dpy->Xdpy);
-
-   /* See if we can choose a DRI/DRM driver */
-   driverName = _eglChooseDRMDriver(screen);
-   if (driverName) {
-      free((void *) driverName);
-      driverName = _eglstrdup(DefaultGLXDriver);
-   }
-   else {
-      driverName = _eglstrdup(DefaultSoftDriver);
-   }
-
-   _eglLog(_EGL_DEBUG, "_xeglChooseDriver: %s", driverName);
-
-   return driverName;
-#else
-   return NULL;
-#endif
-}
-
-
diff --git a/src/egl/main/eglx.h b/src/egl/main/eglx.h
deleted file mode 100644
index 4323d55838..0000000000
--- a/src/egl/main/eglx.h
+++ /dev/null
@@ -1,12 +0,0 @@
-#ifndef EGLX_INCLUDED
-#define EGLX_INCLUDED
-
-
-#include "egldisplay.h"
-
-
-extern const char *
-_xeglChooseDriver(_EGLDisplay *dpy);
-
-
-#endif /* EGLX_INCLUDED */
diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template
index 98487d43bd..63983c5220 100644
--- a/src/gallium/Makefile.template
+++ b/src/gallium/Makefile.template
@@ -31,8 +31,8 @@ INCLUDES = \
 
 default: depend lib$(LIBNAME).a
 
-lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template
-	$(MKLIB) -o $(LIBNAME) -static $(OBJECTS)
+lib$(LIBNAME).a: $(OBJECTS) $(EXTRA_OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template
+	$(MKLIB) -o $(LIBNAME) -static $(OBJECTS) $(EXTRA_OBJECTS)
 
 depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
 	rm -f depend
@@ -41,7 +41,7 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
 
 # Emacs tags
 tags:
-	etags `find . -name \*.[ch]` `find ../include`
+	etags `find . -name \*.[ch]` `find $(TOP)/src/gallium/include -name \*.h`
 
 # Remove .o and backup files
 clean:
diff --git a/src/gallium/SConscript b/src/gallium/SConscript
index b6ceaf3edf..89c69d7205 100644
--- a/src/gallium/SConscript
+++ b/src/gallium/SConscript
@@ -22,6 +22,7 @@ SConscript([
 	'auxiliary/draw/SConscript',
 	'auxiliary/pipebuffer/SConscript',
 	'auxiliary/indices/SConscript',
+	'auxiliary/rbug/SConscript',
 ])
 
 for driver in env['drivers']:
@@ -30,6 +31,7 @@ for driver in env['drivers']:
 SConscript('state_trackers/python/SConscript')
 SConscript('state_trackers/glx/xlib/SConscript')
 SConscript('state_trackers/dri/SConscript')
+SConscript('state_trackers/xorg/SConscript')
 
 if platform == 'windows':
 	SConscript('state_trackers/wgl/SConscript')
diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c
index 0bc77a5728..e6dce3f0e5 100644
--- a/src/gallium/auxiliary/cso_cache/cso_cache.c
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.c
@@ -361,6 +361,10 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type,
 void cso_cache_delete(struct cso_cache *sc)
 {
    assert(sc);
+
+   if (!sc)
+      return;
+
    /* delete driver data */
    cso_for_each_state(sc, CSO_BLEND, delete_blend_state, 0);
    cso_for_each_state(sc, CSO_DEPTH_STENCIL_ALPHA, delete_depth_stencil_state, 0);
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index f388bf5d95..36c882acb7 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -55,14 +55,14 @@ struct cso_context {
    void *samplers[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers;
 
-   void *samplers_saved[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers_saved;
+   void *samplers_saved[PIPE_MAX_SAMPLERS];
 
    struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
    uint nr_textures;
 
-   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
    uint nr_textures_saved;
+   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
 
    /** Current and saved state.
     * The saved state is used as a 1-deep stack.
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
index 3cfae2de83..1c6d657297 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.c
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -248,22 +248,22 @@ void draw_pipeline_run( struct draw_context *draw,
                 ( DRAW_PIPE_RESET_STIPPLE |                      \
                   DRAW_PIPE_EDGE_FLAG_0 |                        \
                   DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i1),                          \
+                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i1),                           \
                 verts + stride * (i3));                          \
-      do_triangle( draw,                                         \
-                   ( DRAW_PIPE_EDGE_FLAG_0 |                     \
-                     DRAW_PIPE_EDGE_FLAG_1 ),                    \
-                 verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i2),                          \
-                 verts + stride * (i3))
+   do_triangle( draw,                                            \
+                ( DRAW_PIPE_EDGE_FLAG_0 |                        \
+                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
+                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i2),                           \
+                verts + stride * (i3))
 
 #define TRIANGLE(flags,i0,i1,i2)                                 \
    do_triangle( draw,                                            \
                 flags,  /* flags */                              \
-                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i1),                          \
-                 verts + stride * (i2))
+                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i1),                           \
+                verts + stride * (i2))
 
 #define LINE(flags,i0,i1)                                   \
    do_line( draw,                                           \
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index 30a6d2919d..283502cdf3 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -256,7 +256,10 @@ pstip_transform_inst(struct tgsi_transform_context *ctx,
          uint size = 4;
          immed = tgsi_default_full_immediate();
          immed.Immediate.NrTokens = 1 + size; /* one for the token itself */
-         immed.u.Pointer = (void *) value;
+         immed.u[0].Float = value[0];
+         immed.u[1].Float = value[1];
+         immed.u[2].Float = value[2];
+         immed.u[3].Float = value[3];
          ctx->emit_immediate(ctx, &immed);
       }
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index a5d840b96e..1a5269c0de 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -159,8 +159,19 @@ vbuf_tri( struct draw_stage *stage,
 
    check_space( vbuf, 3 );
 
-   for (i = 0; i < 3; i++) {
-      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+   if (vbuf->stage.draw->rasterizer->flatshade_first) {
+      /* Put provoking vertex in position expected by the driver.
+       * Emit last provoking vertex in first pos.
+       * Swap verts 0 & 1 to preserve polygon winding.
+       */
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[2] );
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] );
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[1] );
+   }
+   else {
+      for (i = 0; i < 3; i++) {
+         vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+      }
    }
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index d84bab9eaa..7d76a7dbf3 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -28,6 +28,30 @@
 /* Authors:  Keith Whitwell <keith@tungstengraphics.com>
  */
 
+/**
+ * Notes on wide points and sprite mode:
+ *
+ * In wide point/sprite mode we effectively need to convert each incoming
+ * vertex into four outgoing vertices specifying the corners of a quad.
+ * Since we don't (yet) have geometry shaders, we have to handle this here
+ * in the draw module.
+ *
+ * For sprites, it also means that this is where we have to handle texcoords
+ * for the vertices of the quad.  OpenGL's GL_COORD_REPLACE state specifies
+ * if/how enabled texcoords are automatically generated for sprites.  We pass
+ * that info through gallium in the pipe_rasterizer_state::sprite_coord_mode
+ * array.
+ *
+ * Additionally, GLSL's gl_PointCoord fragment attribute has to be handled
+ * here as well.  This is basically an additional texture/generic attribute
+ * that varies .x from 0 to 1 horizontally across the point and varies .y
+ * vertically from 0 to 1 down the sprite.
+ *
+ * With geometry shaders, the state tracker could create a GS to do
+ * most/all of this.
+ */
+
+
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "pipe/p_defines.h"
@@ -52,7 +76,7 @@ struct widepoint_stage {
 
    int psize_slot;
 
-   int point_coord_fs_input;  /**< input for pointcoord (and fog) */
+   int point_coord_fs_input;  /**< input for pointcoord */
 };
 
 
@@ -64,8 +88,6 @@ widepoint_stage( struct draw_stage *stage )
 }
 
 
-
-
 /**
  * Set the vertex texcoords for sprite mode.
  * Coords may be left untouched or set to a right-side-up or upside-down
@@ -89,10 +111,12 @@ static void set_texcoords(const struct widepoint_stage *wide,
    }
 
    if (wide->point_coord_fs_input >= 0) {
-      /* put gl_PointCoord into extra vertex output's zw components */
-      uint k = wide->stage.draw->extra_vp_outputs.slot;
-      v->data[k][2] = tc[0];
-      v->data[k][3] = tc[1];
+      /* put gl_PointCoord into the extra vertex slot */
+      uint slot = wide->stage.draw->extra_vp_outputs.slot;
+      v->data[slot][0] = tc[0];
+      v->data[slot][1] = tc[1];
+      v->data[slot][2] = 0.0F;
+      v->data[slot][3] = 1.0F;
    }
 }
 
@@ -181,6 +205,16 @@ static void widepoint_point( struct draw_stage *stage,
 }
 
 
+static int
+find_pntc_input_attrib(struct draw_context *draw)
+{
+   /* Scan the fragment program's input decls to find the pointcoord
+    * attribute.  The xy components will store the point coord.
+    */
+   return 0; /* XXX fix this */
+}
+
+
 static void widepoint_first_point( struct draw_stage *stage, 
 			      struct prim_header *header )
 {
@@ -219,8 +253,8 @@ static void widepoint_first_point( struct draw_stage *stage,
       }
       wide->num_texcoords = j;
 
-      /* find fragment shader PointCoord/Fog input */
-      wide->point_coord_fs_input = 0; /* XXX fix this! */
+      /* find fragment shader PointCoord input */
+      wide->point_coord_fs_input = find_pntc_input_attrib(draw);
 
       /* setup extra vp output (point coord implemented as a texcoord) */
       draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 81e4eae401..41fcb16a0a 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -44,7 +44,6 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 
-#include "tgsi/tgsi_exec.h"
 #include "tgsi/tgsi_scan.h"
 
 
@@ -55,6 +54,8 @@ struct draw_vertex_shader;
 struct draw_context;
 struct draw_stage;
 struct vbuf_render;
+struct tgsi_exec_machine;
+struct tgsi_sampler;
 
 
 /**
@@ -185,7 +186,7 @@ struct draw_context
       uint position_output;
 
       /** TGSI program interpreter runtime state */
-      struct tgsi_exec_machine machine;
+      struct tgsi_exec_machine *machine;
 
       uint num_samplers;
       struct tgsi_sampler **samplers;
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 9ea0cbe599..dbb5ac7182 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -36,6 +36,7 @@
 #include "draw/draw_vs.h"
 #include "tgsi/tgsi_dump.h"
 #include "util/u_math.h"
+#include "util/u_prim.h"
 
 static unsigned trim( unsigned count, unsigned first, unsigned incr )
 {
@@ -278,7 +279,7 @@ void
 draw_arrays(struct draw_context *draw, unsigned prim,
             unsigned start, unsigned count)
 {
-   unsigned reduced_prim = draw_pt_reduced_prim(prim);
+   unsigned reduced_prim = u_reduced_prim(prim);
    if (reduced_prim != draw->reduced_prim) {
       draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
       draw->reduced_prim = reduced_prim;
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 8ef0ea8011..7a17a9fb6b 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -228,7 +228,6 @@ void draw_pt_post_vs_destroy( struct pt_post_vs *pvs );
  * Utils: 
  */
 void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr);
-unsigned draw_pt_reduced_prim(unsigned prim);
 
 
 #endif
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
index 3fb0695687..4ca5b52020 100644
--- a/src/gallium/auxiliary/draw/draw_pt_decompose.h
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -47,10 +47,19 @@ static void FUNC( ARGS,
 
    case PIPE_PRIM_TRIANGLES:
       for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                   (i + 0),
-                   (i + 1),
-                   (i + 2 ));
+         if (flatfirst) {
+            /* put provoking vertex in last pos for clipper */
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 1),
+                      (i + 2),
+                      (i + 0 ));
+         }
+         else {
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 0),
+                      (i + 1),
+                      (i + 2 ));
+         }
       }
       break;
 
@@ -58,9 +67,9 @@ static void FUNC( ARGS,
       if (flatfirst) {
          for (i = 0; i+2 < count; i++) {
             TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0),
                       (i + 1 + (i&1)),
-                      (i + 2 - (i&1)));
+                      (i + 2 - (i&1)),
+                      (i + 0) );
          }
       }
       else {
@@ -78,9 +87,9 @@ static void FUNC( ARGS,
          if (flatfirst) {
             for (i = 0; i+2 < count; i++) {
                TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (i + 1),
                          (i + 2),
-                         (0 ));
+                         0,
+                         (i + 1) );
             }
          }
          else {
@@ -96,20 +105,40 @@ static void FUNC( ARGS,
 
 
    case PIPE_PRIM_QUADS:
-      for (i = 0; i+3 < count; i += 4) {
-         QUAD( (i + 0),
-               (i + 1),
-               (i + 2),
-               (i + 3));
+      if (flatfirst) {
+         for (i = 0; i+3 < count; i += 4) {
+            QUAD( (i + 1),
+                  (i + 2),
+                  (i + 3),
+                  (i + 0) );
+         }
+      }
+      else {
+         for (i = 0; i+3 < count; i += 4) {
+            QUAD( (i + 0),
+                  (i + 1),
+                  (i + 2),
+                  (i + 3));
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      for (i = 0; i+3 < count; i += 2) {
-         QUAD( (i + 2),
-               (i + 0),
-               (i + 1),
-               (i + 3));
+      if (flatfirst) {
+         for (i = 0; i+3 < count; i += 2) {
+            QUAD( (i + 1),
+                  (i + 3),
+                  (i + 2),
+                  (i + 0) );
+         }
+      }
+      else {
+         for (i = 0; i+3 < count; i += 2) {
+            QUAD( (i + 2),
+                  (i + 0),
+                  (i + 1),
+                  (i + 3));
+         }
       }
       break;
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
index 6b7d02a19b..e7fe6b3b76 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -130,6 +130,10 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
       unsigned output_format;
 
       switch (vinfo->attrib[i].emit) {
+      case EMIT_4UB:
+	 output_format = PIPE_FORMAT_R8G8B8A8_UNORM;
+	 emit_sz = 4 * sizeof(unsigned char);
+         break;
       case EMIT_4F:
 	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 	 emit_sz = 4 * sizeof(float);
@@ -153,6 +157,8 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
 	 output_format = PIPE_FORMAT_R32_FLOAT;
 	 emit_sz = 1 * sizeof(float);
          break;
+      case EMIT_OMIT:
+         continue;
       default:
          assert(0);
 	 output_format = PIPE_FORMAT_NONE;
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
index 3bc7939c55..b61fa29143 100644
--- a/src/gallium/auxiliary/draw/draw_pt_util.c
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -75,28 +75,3 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       break;
    }
 }
-
-
-unsigned draw_pt_reduced_prim(unsigned prim)
-{
-   switch (prim) {
-   case PIPE_PRIM_POINTS:
-      return PIPE_PRIM_POINTS;
-   case PIPE_PRIM_LINES:
-   case PIPE_PRIM_LINE_STRIP:
-   case PIPE_PRIM_LINE_LOOP:
-      return PIPE_PRIM_LINES;
-   case PIPE_PRIM_TRIANGLES:
-   case PIPE_PRIM_TRIANGLE_STRIP:
-   case PIPE_PRIM_TRIANGLE_FAN:
-   case PIPE_PRIM_POLYGON:
-   case PIPE_PRIM_QUADS:
-   case PIPE_PRIM_QUAD_STRIP:
-      return PIPE_PRIM_TRIANGLES;
-   default:
-      assert(0);
-      return PIPE_PRIM_POINTS;
-   }
-}
-
-
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 5d268a2226..1a0527be63 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -31,6 +31,7 @@
   */
 
 #include "util/u_memory.h"
+#include "util/u_prim.h"
 #include "draw/draw_context.h"
 #include "draw/draw_private.h"
 #include "draw/draw_pt.h"
@@ -193,16 +194,30 @@ vcache_ef_quad( struct vcache_frontend *vcache,
                 unsigned i2,
                 unsigned i3 )
 {
-   vcache_triangle_flags( vcache,
-                          ( DRAW_PIPE_RESET_STIPPLE |
-                            DRAW_PIPE_EDGE_FLAG_0 |
-                            DRAW_PIPE_EDGE_FLAG_2 ),
-                          i0, i1, i3 );
-
-   vcache_triangle_flags( vcache,
-                          ( DRAW_PIPE_EDGE_FLAG_0 |
-                            DRAW_PIPE_EDGE_FLAG_1 ),
-                          i1, i2, i3 );
+   if (vcache->draw->rasterizer->flatshade_first) {
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_RESET_STIPPLE |
+                               DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i0, i1, i2 );
+
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_EDGE_FLAG_2 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i0, i2, i3 );
+   }
+   else {
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_RESET_STIPPLE |
+                               DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_2 ),
+                             i0, i1, i3 );
+
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i1, i2, i3 );
+   }
 }
 
 /* At least for now, we're back to using a template include file for
@@ -453,7 +468,7 @@ vcache_prepare( struct draw_pt_front_end *frontend,
    }
 
    vcache->input_prim = prim;
-   vcache->output_prim = draw_pt_reduced_prim(prim);
+   vcache->output_prim = u_reduced_prim(prim);
 
    vcache->middle = middle;
    vcache->opt = opt;
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
index ec05bbeab4..62822a3d56 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -118,21 +118,39 @@ static void FUNC( struct draw_pt_front_end *frontend,
 
    case PIPE_PRIM_QUADS:
       for (i = 0; i+3 < count; i += 4) {
-         QUAD( vcache,
-               get_elt(elts, i + 0),
-               get_elt(elts, i + 1),
-               get_elt(elts, i + 2),
-               get_elt(elts, i + 3));
+         if (flatfirst) {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 3) );
+         }
+         else {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 3) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 0; i+3 < count; i += 2) {
-         QUAD( vcache,
-               get_elt(elts, i + 2),
-               get_elt(elts, i + 0),
-               get_elt(elts, i + 1),
-               get_elt(elts, i + 3));
+         if (flatfirst) {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 3),
+                  get_elt(elts, i + 2) );
+         }
+         else {
+            QUAD( vcache,
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 3) );
+         }
       }
       break;
 
@@ -144,19 +162,38 @@ static void FUNC( struct draw_pt_front_end *frontend,
          const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
          const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
          const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+         ushort edge_next, edge_finish;
 
-         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+         if (flatfirst) {
+            flags = DRAW_PIPE_RESET_STIPPLE | edge_middle | edge_last;
+            edge_next = edge_last;
+            edge_finish = edge_first;
+         }
+         else {
+            flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+            edge_next = edge_middle;
+            edge_finish = edge_last;
+         }
 
-	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
+	 for (i = 0; i+2 < count; i++, flags = edge_next) {
 
             if (i + 3 == count)
-               flags |= edge_last;
+               flags |= edge_finish;
 
-	    TRIANGLE( vcache,
-                      flags,
-                      get_elt(elts, i + 1),
-                      get_elt(elts, i + 2),
-                      get_elt(elts, 0));
+            if (flatfirst) {
+               TRIANGLE( vcache,
+                         flags,
+                         get_elt(elts, 0),
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2) );
+            }
+            else {
+               TRIANGLE( vcache,
+                         flags,
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2),
+                         get_elt(elts, 0));
+            }
 	 }
       }
       break;
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index c057cd67fd..790e89ed82 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -43,6 +43,8 @@
 #include "translate/translate.h"
 #include "translate/translate_cache.h"
 
+#include "tgsi/tgsi_exec.h"
+
 
 
 
@@ -146,16 +148,8 @@ draw_delete_vertex_shader(struct draw_context *draw,
 boolean 
 draw_vs_init( struct draw_context *draw )
 {
-   tgsi_exec_machine_init(&draw->vs.machine);
-
-   /* FIXME: give this machine thing a proper constructor:
-    */
-   draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
-   if (!draw->vs.machine.Inputs)
-      return FALSE;
-
-   draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
-   if (!draw->vs.machine.Outputs)
+   draw->vs.machine = tgsi_exec_machine_create();
+   if (!draw->vs.machine)
       return FALSE;
 
    draw->vs.emit_cache = translate_cache_create();
@@ -178,12 +172,6 @@ draw_vs_init( struct draw_context *draw )
 void
 draw_vs_destroy( struct draw_context *draw )
 {
-   if (draw->vs.machine.Inputs)
-      align_free(draw->vs.machine.Inputs);
-
-   if (draw->vs.machine.Outputs)
-      align_free(draw->vs.machine.Outputs);
-
    if (draw->vs.fetch_cache)
       translate_cache_destroy(draw->vs.fetch_cache);
 
@@ -196,8 +184,7 @@ draw_vs_destroy( struct draw_context *draw )
    if (draw->vs.aligned_constant_storage)
       align_free((void*)draw->vs.aligned_constant_storage);
 
-   tgsi_exec_machine_free_data(&draw->vs.machine);
-
+   tgsi_exec_machine_destroy(draw->vs.machine);
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c
index 9e37a26c1e..62e04a65f3 100644
--- a/src/gallium/auxiliary/draw/draw_vs_aos.c
+++ b/src/gallium/auxiliary/draw/draw_vs_aos.c
@@ -1758,24 +1758,24 @@ emit_instruction( struct aos_compilation *cp,
    case TGSI_OPCODE_SUB:
       return emit_SUB(cp, inst);
  
-   case TGSI_OPCODE_LERP:
+   case TGSI_OPCODE_LRP:
 //      return emit_LERP(cp, inst);
       return FALSE;
 
-   case TGSI_OPCODE_FRAC:
+   case TGSI_OPCODE_FRC:
       return emit_FRC(cp, inst);
 
    case TGSI_OPCODE_CLAMP:
 //      return emit_CLAMP(cp, inst);
       return FALSE;
 
-   case TGSI_OPCODE_FLOOR:
+   case TGSI_OPCODE_FLR:
       return emit_FLR(cp, inst);
 
    case TGSI_OPCODE_ROUND:
       return emit_RND(cp, inst);
 
-   case TGSI_OPCODE_EXPBASE2:
+   case TGSI_OPCODE_EX2:
 #if FAST_MATH
       return emit_EXPBASE2(cp, inst);
 #elif 0
@@ -1787,13 +1787,13 @@ emit_instruction( struct aos_compilation *cp,
       return FALSE;
 #endif
 
-   case TGSI_OPCODE_LOGBASE2:
+   case TGSI_OPCODE_LG2:
       return emit_LG2(cp, inst);
 
-   case TGSI_OPCODE_POWER:
+   case TGSI_OPCODE_POW:
       return emit_POW(cp, inst);
 
-   case TGSI_OPCODE_CROSSPRODUCT:
+   case TGSI_OPCODE_XPD:
       return emit_XPD(cp, inst);
 
    case TGSI_OPCODE_ABS:
@@ -1891,8 +1891,9 @@ static boolean note_immediate( struct aos_compilation *cp,
    unsigned pos = cp->num_immediates++;
    unsigned j;
 
+   assert( imm->Immediate.NrTokens <= 4 + 1 );
    for (j = 0; j < imm->Immediate.NrTokens - 1; j++) {
-      cp->vaos->machine->immediate[pos][j] = imm->u.ImmediateFloat32[j].Float;
+      cp->vaos->machine->immediate[pos][j] = imm->u[j].Float;
    }
 
    return TRUE;
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index dbbc33fffa..41cc802613 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -41,6 +41,7 @@
 
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_scan.h"
+#include "tgsi/tgsi_exec.h"
 
 
 struct exec_vertex_shader {
@@ -114,6 +115,12 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
 #endif
 
          for (slot = 0; slot < shader->info.num_inputs; slot++) {
+#if 0
+            assert(!util_is_inf_or_nan(input[slot][0]));
+            assert(!util_is_inf_or_nan(input[slot][1]));
+            assert(!util_is_inf_or_nan(input[slot][2]));
+            assert(!util_is_inf_or_nan(input[slot][3]));
+#endif
             machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
             machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
             machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
@@ -195,7 +202,7 @@ draw_create_vs_exec(struct draw_context *draw,
    vs->base.run_linear = vs_exec_run_linear;
    vs->base.delete = vs_exec_delete;
    vs->base.create_varient = draw_vs_varient_generic;
-   vs->machine = &draw->vs.machine;
+   vs->machine = draw->vs.machine;
 
    return &vs->base;
 }
diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
index 727977bc3a..b3535c0e48 100644
--- a/src/gallium/auxiliary/draw/draw_vs_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c
@@ -119,7 +119,7 @@ draw_create_vs_llvm(struct draw_context *draw,
    vs->base.create_varient = draw_vs_varient_generic;
    vs->base.run_linear = vs_llvm_run_linear;
    vs->base.delete = vs_llvm_delete;
-   vs->machine = &draw->vs.machine;
+   vs->machine = draw->vs.machine;
 
    {
       struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS);
diff --git a/src/gallium/auxiliary/draw/draw_vs_ppc.c b/src/gallium/auxiliary/draw/draw_vs_ppc.c
index d35db57d57..ad184bd696 100644
--- a/src/gallium/auxiliary/draw/draw_vs_ppc.c
+++ b/src/gallium/auxiliary/draw/draw_vs_ppc.c
@@ -48,6 +48,7 @@
 #include "rtasm/rtasm_ppc.h"
 #include "tgsi/tgsi_ppc.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
 
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 77ba5152f9..702051387a 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -48,27 +48,16 @@
 #include "rtasm/rtasm_x86sse.h"
 #include "tgsi/tgsi_sse2.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
 
 #define SSE_MAX_VERTICES 4
 
-typedef void (PIPE_CDECL *codegen_function) (
-   const struct tgsi_exec_vector *input, /* 1 */
-   struct tgsi_exec_vector *output, /* 2 */
-   float (*constant)[4],        /* 3 */
-   struct tgsi_exec_vector *temporary, /* 4 */
-   float (*immediates)[4],      /* 5 */
-   const float (*aos_input)[4], /* 6 */
-   uint num_inputs,             /* 7 */
-   uint input_stride,           /* 8 */
-   float (*aos_output)[4],      /* 9 */
-   uint num_outputs,            /* 10 */
-   uint output_stride );        /* 11 */
 
 struct draw_sse_vertex_shader {
    struct draw_vertex_shader base;
    struct x86_function sse2_program;
 
-   codegen_function func;
+   tgsi_sse2_vs_func func;
    
    struct tgsi_exec_machine *machine;
 };
@@ -78,6 +67,10 @@ static void
 vs_sse_prepare( struct draw_vertex_shader *base,
 		struct draw_context *draw )
 {
+   struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base;
+   struct tgsi_exec_machine *machine = shader->machine;
+
+   machine->Samplers = draw->vs.samplers;
 }
 
 
@@ -118,11 +111,9 @@ vs_sse_run_linear( struct draw_vertex_shader *base,
 
       /* run compiled shader
        */
-      shader->func(machine->Inputs,
-		   machine->Outputs,
-		   (float (*)[4])constants,
-		   machine->Temps,
-		   (float (*)[4])shader->base.immediates,
+      shader->func(machine,
+		   constants,
+		   shader->base.immediates,
                    input,
                    base->info.num_inputs,
                    input_stride,
@@ -184,7 +175,7 @@ draw_create_vs_sse(struct draw_context *draw,
    vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 *
                                       sizeof(float), 16);
 
-   vs->machine = &draw->vs.machine;
+   vs->machine = draw->vs.machine;
    
    x86_init_func( &vs->sse2_program );
 
@@ -194,7 +185,7 @@ draw_create_vs_sse(struct draw_context *draw,
                         TRUE )) 
       goto fail;
       
-   vs->func = (codegen_function) x86_get_func( &vs->sse2_program );
+   vs->func = (tgsi_sse2_vs_func) x86_get_func( &vs->sse2_program );
    if (!vs->func) {
       goto fail;
    }
diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
index 925e948763..721b7d2d83 100644
--- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp
+++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
@@ -24,6 +24,8 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  **************************************************************************/
+
+#include <cstdio>
 #include "instructionssoa.h"
 
 #include "storagesoa.h"
@@ -126,7 +128,7 @@ void InstructionsSoa::createFunctionMap()
    m_functionsMap[TGSI_OPCODE_DP4]   = "dp4";
    m_functionsMap[TGSI_OPCODE_MIN]   = "min";
    m_functionsMap[TGSI_OPCODE_MAX]   = "max";
-   m_functionsMap[TGSI_OPCODE_POWER] = "pow";
+   m_functionsMap[TGSI_OPCODE_POW]   = "pow";
    m_functionsMap[TGSI_OPCODE_LIT]   = "lit";
    m_functionsMap[TGSI_OPCODE_RSQ]   = "rsq";
    m_functionsMap[TGSI_OPCODE_SLT]   = "slt";
@@ -309,7 +311,7 @@ std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> i
 std::vector<llvm::Value*> InstructionsSoa::pow(const std::vector<llvm::Value*> in1,
                                                const std::vector<llvm::Value*> in2)
 {
-   llvm::Function *func = function(TGSI_OPCODE_POWER);
+   llvm::Function *func = function(TGSI_OPCODE_POW);
    return callBuiltin(func, in1, in2);
 }
 
diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
index 5b08200d14..bf84401e11 100644
--- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
+++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
@@ -160,10 +160,11 @@ translate_immediate(Storage *storage,
 {
    float vec[4];
    int i;
+   assert( imm->Immediate.NrTokens <= 4 + 1 );
    for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) {
       switch (imm->Immediate.DataType) {
       case TGSI_IMM_FLOAT32:
-         vec[i] = imm->u.ImmediateFloat32[i].Float;
+         vec[i] = imm->u[i].Float;
          break;
       default:
          assert(0);
@@ -179,10 +180,11 @@ translate_immediateir(StorageSoa *storage,
 {
    float vec[4];
    int i;
+   assert( imm->Immediate.NrTokens <= 4 + 1 );
    for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) {
       switch (imm->Immediate.DataType) {
       case TGSI_IMM_FLOAT32:
-         vec[i] = imm->u.ImmediateFloat32[i].Float;
+         vec[i] = imm->u[i].Float;
          break;
       default:
          assert(0);
@@ -336,7 +338,7 @@ translate_instruction(llvm::Module *module,
       out = instr->sub(inputs[0], inputs[1]);
    }
       break;
-   case TGSI_OPCODE_LERP: {
+   case TGSI_OPCODE_LRP: {
       out = instr->lerp(inputs[0], inputs[1], inputs[2]);
    }
       break;
@@ -348,17 +350,11 @@ translate_instruction(llvm::Module *module,
       out = instr->cnd0(inputs[0], inputs[1], inputs[2]);
    }
       break;
-   case TGSI_OPCODE_DOT2ADD: {
+   case TGSI_OPCODE_DP2A: {
       out = instr->dot2add(inputs[0], inputs[1], inputs[2]);
    }
       break;
-   case TGSI_OPCODE_INDEX:
-      break;
-   case TGSI_OPCODE_NEGATE: {
-      out = instr->neg(inputs[0]);
-   }
-      break;
-   case TGSI_OPCODE_FRAC: {
+   case TGSI_OPCODE_FRC: {
       out = instr->frc(inputs[0]);
    }
       break;
@@ -366,30 +362,28 @@ translate_instruction(llvm::Module *module,
       out = instr->clamp(inputs[0]);
    }
       break;
-   case TGSI_OPCODE_FLOOR: {
+   case TGSI_OPCODE_FLR: {
       out = instr->floor(inputs[0]);
    }
       break;
    case TGSI_OPCODE_ROUND:
       break;
-   case TGSI_OPCODE_EXPBASE2: {
+   case TGSI_OPCODE_EX2: {
       out = instr->ex2(inputs[0]);
    }
       break;
-   case TGSI_OPCODE_LOGBASE2: {
+   case TGSI_OPCODE_LG2: {
       out = instr->lg2(inputs[0]);
    }
       break;
-   case TGSI_OPCODE_POWER: {
+   case TGSI_OPCODE_POW: {
       out = instr->pow(inputs[0], inputs[1]);
    }
       break;
-   case TGSI_OPCODE_CROSSPRODUCT: {
+   case TGSI_OPCODE_XPD: {
       out = instr->cross(inputs[0], inputs[1]);
    }
       break;
-   case TGSI_OPCODE_MULTIPLYMATRIX:
-      break;
    case TGSI_OPCODE_ABS: {
       out = instr->abs(inputs[0]);
    }
@@ -522,7 +516,7 @@ translate_instruction(llvm::Module *module,
       return;  //just update the state
    }
       break;
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
       break;
    case TGSI_OPCODE_REP:
       break;
@@ -538,7 +532,7 @@ translate_instruction(llvm::Module *module,
       return; //just update the state
    }
       break;
-   case TGSI_OPCODE_ENDLOOP:
+   case TGSI_OPCODE_ENDFOR:
       break;
    case TGSI_OPCODE_ENDREP:
       break;
@@ -580,7 +574,7 @@ translate_instruction(llvm::Module *module,
       break;
    case TGSI_OPCODE_ENDPRIM:
       break;
-   case TGSI_OPCODE_BGNLOOP2: {
+   case TGSI_OPCODE_BGNLOOP: {
       instr->beginLoop();
       storage->setCurrentBlock(instr->currentBlock());
       return;
@@ -593,7 +587,7 @@ translate_instruction(llvm::Module *module,
       return;
    }
       break;
-   case TGSI_OPCODE_ENDLOOP2: {
+   case TGSI_OPCODE_ENDLOOP: {
       instr->endLoop();
       storage->setCurrentBlock(instr->currentBlock());
       return;
@@ -617,14 +611,6 @@ translate_instruction(llvm::Module *module,
       break;
    case TGSI_OPCODE_NOP:
       break;
-   case TGSI_OPCODE_M4X3:
-      break;
-   case TGSI_OPCODE_M3X4:
-      break;
-   case TGSI_OPCODE_M3X3:
-      break;
-   case TGSI_OPCODE_M3X2:
-      break;
    case TGSI_OPCODE_CALLNZ:
       break;
    case TGSI_OPCODE_IFC:
@@ -778,44 +764,38 @@ translate_instructionir(llvm::Module *module,
       out = instr->sub(inputs[0], inputs[1]);
    }
       break;
-   case TGSI_OPCODE_LERP: {
+   case TGSI_OPCODE_LRP: {
    }
       break;
    case TGSI_OPCODE_CND:
       break;
    case TGSI_OPCODE_CND0:
       break;
-   case TGSI_OPCODE_DOT2ADD:
+   case TGSI_OPCODE_DP2A:
       break;
-   case TGSI_OPCODE_INDEX:
-      break;
-   case TGSI_OPCODE_NEGATE:
-      break;
-   case TGSI_OPCODE_FRAC: {
+   case TGSI_OPCODE_FRC: {
    }
       break;
    case TGSI_OPCODE_CLAMP:
       break;
-   case TGSI_OPCODE_FLOOR: {
+   case TGSI_OPCODE_FLR: {
    }
       break;
    case TGSI_OPCODE_ROUND:
       break;
-   case TGSI_OPCODE_EXPBASE2: {
+   case TGSI_OPCODE_EX2: {
    }
       break;
-   case TGSI_OPCODE_LOGBASE2: {
+   case TGSI_OPCODE_LG2: {
    }
       break;
-   case TGSI_OPCODE_POWER: {
+   case TGSI_OPCODE_POW: {
       out = instr->pow(inputs[0], inputs[1]);
    }
       break;
-   case TGSI_OPCODE_CROSSPRODUCT: {
+   case TGSI_OPCODE_XPD: {
    }
       break;
-   case TGSI_OPCODE_MULTIPLYMATRIX:
-      break;
    case TGSI_OPCODE_ABS: {
       out = instr->abs(inputs[0]);
    }
@@ -910,7 +890,7 @@ translate_instructionir(llvm::Module *module,
    case TGSI_OPCODE_IF: {
    }
       break;
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
       break;
    case TGSI_OPCODE_REP:
       break;
@@ -920,7 +900,7 @@ translate_instructionir(llvm::Module *module,
    case TGSI_OPCODE_ENDIF: {
    }
       break;
-   case TGSI_OPCODE_ENDLOOP:
+   case TGSI_OPCODE_ENDFOR:
       break;
    case TGSI_OPCODE_ENDREP:
       break;
@@ -961,13 +941,13 @@ translate_instructionir(llvm::Module *module,
       break;
    case TGSI_OPCODE_ENDPRIM:
       break;
-   case TGSI_OPCODE_BGNLOOP2: {
+   case TGSI_OPCODE_BGNLOOP: {
    }
       break;
    case TGSI_OPCODE_BGNSUB: {
    }
       break;
-   case TGSI_OPCODE_ENDLOOP2: {
+   case TGSI_OPCODE_ENDLOOP: {
    }
       break;
    case TGSI_OPCODE_ENDSUB: {
@@ -983,14 +963,6 @@ translate_instructionir(llvm::Module *module,
       break;
    case TGSI_OPCODE_NOP:
       break;
-   case TGSI_OPCODE_M4X3:
-      break;
-   case TGSI_OPCODE_M3X4:
-      break;
-   case TGSI_OPCODE_M3X3:
-      break;
-   case TGSI_OPCODE_M3X2:
-      break;
    case TGSI_OPCODE_NRM4:
       break;
    case TGSI_OPCODE_CALLNZ:
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
index 0760d60716..6bdce5fcb0 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
@@ -53,6 +53,8 @@ static INLINE struct malloc_buffer *
 malloc_buffer(struct pb_buffer *buf)
 {
    assert(buf);
+   if (!buf)
+      return NULL;
    assert(buf->vtbl == &malloc_buffer_vtbl);
    return (struct malloc_buffer *)buf;
 }
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
index bc3093f620..cb32d25136 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
@@ -74,6 +74,8 @@ static INLINE struct pb_ondemand_buffer *
 pb_ondemand_buffer(struct pb_buffer *buf)
 {
    assert(buf);
+   if (!buf)
+      return NULL;
    assert(buf->vtbl == &pb_ondemand_buffer_vtbl);
    return (struct pb_ondemand_buffer *)buf;
 }
diff --git a/src/gallium/auxiliary/rbug/Makefile b/src/gallium/auxiliary/rbug/Makefile
new file mode 100644
index 0000000000..cd12e8468f
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/Makefile
@@ -0,0 +1,14 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = rbug
+
+C_SOURCES = \
+	rbug_connection.c \
+	rbug_core.c \
+	rbug_texture.c \
+	rbug_context.c \
+	rbug_shader.c \
+	rbug_demarshal.c
+
+include ../../Makefile.template
diff --git a/src/gallium/auxiliary/rbug/README b/src/gallium/auxiliary/rbug/README
new file mode 100644
index 0000000000..33d76371de
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/README
@@ -0,0 +1,21 @@
+                 GALLIUM REMOTE DEBUGGING COMMON CODE
+
+= About =
+
+This directory contains the common code for the Gallium 3D remote debugging
+driver and clients. The code is two parts the connection managment code and
+the (de)marsheller.
+
+The code currently uses tcp and ip4v for connections.
+
+Information about driver integration can be found in:
+
+src/gallium/drivers/trace/README
+
+for information about applications look in:
+
+progs/rbug/README
+
+
+--
+Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/auxiliary/rbug/SConscript b/src/gallium/auxiliary/rbug/SConscript
new file mode 100644
index 0000000000..4a9afb45d3
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/SConscript
@@ -0,0 +1,14 @@
+Import('*')
+
+rbug = env.ConvenienceLibrary(
+	target = 'rbug',
+	source = [
+		'rbug_core.c',
+		'rbug_shader.c',
+		'rbug_context.c',
+		'rbug_texture.c',
+		'rbug_demarshal.c',
+		'rbug_connection.c',
+	])
+
+auxiliaries.insert(0, rbug)
diff --git a/src/gallium/auxiliary/rbug/rbug.h b/src/gallium/auxiliary/rbug/rbug.h
new file mode 100644
index 0000000000..259bfc6c79
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Include all for users the remote debugger protocol code.
+ */
+
+#include "rbug/rbug_core.h"
+#include "rbug/rbug_shader.h"
+#include "rbug/rbug_context.h"
+#include "rbug/rbug_texture.h"
+#include "rbug/rbug_connection.h"
diff --git a/src/gallium/auxiliary/rbug/rbug_connection.c b/src/gallium/auxiliary/rbug/rbug_connection.c
new file mode 100644
index 0000000000..52acb700af
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_connection.c
@@ -0,0 +1,167 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "rbug/rbug.h"
+#include "rbug/rbug_internal.h"
+
+#include "util/u_network.h"
+
+struct rbug_connection
+{
+   int socket;
+   uint32_t send_serial;
+   uint32_t recv_serial;
+   enum rbug_opcode opcode;
+};
+
+/**
+ * Create a rbug connection from a socket created with u_socket.
+ *
+ * Result:
+ *    A new allocated connection using socket as communication path
+ */
+struct rbug_connection *
+rbug_from_socket(int socket)
+{
+   struct rbug_connection *c = CALLOC_STRUCT(rbug_connection);
+   c->socket = socket;
+   return c;
+}
+
+/**
+ * Free a connection, also closes socket.
+ */
+void
+rbug_disconnect(struct rbug_connection *c)
+{
+   u_socket_close(c->socket);
+   FREE(c);
+}
+
+/**
+ * Waits for a message to be fully received.
+ * Also returns the serial for the message, serial is not touched for replys.
+ *
+ * Result:
+ *    demarshaled message on success, NULL on connection error
+ */
+struct rbug_header *
+rbug_get_message(struct rbug_connection *c, uint32_t *serial)
+{
+   struct rbug_proto_header header;
+   struct rbug_header *out;
+   struct rbug_proto_header *data;
+   size_t length = 0;
+   size_t read = 0;
+   int ret;
+
+
+   ret = u_socket_peek(c->socket, &header, sizeof(header));
+   if (ret <= 0) {
+      return NULL;
+   }
+
+   length = (size_t)header.length * 4;
+   data = MALLOC(length);
+   if (!data) {
+      return NULL;
+   }
+
+   do {
+      uint8_t *ptr = ((uint8_t*)data) + read;
+      ret = u_socket_recv(c->socket, ptr, length - read);
+
+      if (ret <= 0) {
+         FREE(data);
+         return NULL;
+      }
+
+      read += ret;
+   } while(read < length);
+
+   out = rbug_demarshal(data);
+   if (!out)
+      FREE(data);
+   else if (serial)
+      *serial = c->recv_serial++;
+   else
+      c->recv_serial++;
+
+   return out;
+}
+
+/**
+ * Frees a message and associated data.
+ */
+void
+rbug_free_header(struct rbug_header *header)
+{
+   if (!header)
+      return;
+
+   FREE(header->__message);
+   FREE(header);
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Start sending a message.
+ */
+int
+rbug_connection_send_start(struct rbug_connection *c, enum rbug_opcode opcode, uint32_t length)
+{
+   c->opcode = opcode;
+   return 0;
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Write data to the socket.
+ */
+int
+rbug_connection_write(struct rbug_connection *c, void *to, uint32_t size)
+{
+   int ret = u_socket_send(c->socket, to, size);
+   return ret;
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Finish writeing data to the socket.
+ * Ups the send_serial and sets the serial argument if supplied.
+ */
+int rbug_connection_send_finish(struct rbug_connection *c, uint32_t *serial)
+{
+   if (c->opcode < 0)
+      return 0;
+   else if (serial)
+      *serial = c->send_serial++;
+   else
+      c->send_serial++;
+
+   return 0;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_connection.h b/src/gallium/auxiliary/rbug/rbug_connection.h
new file mode 100644
index 0000000000..1f2c9ff347
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_connection.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file contains the function defentions for connection see c file for
+ * more comments covering function use.
+ */
+
+#ifndef _RBUG_CONNECTION_H_
+#define _RBUG_CONNECTION_H_
+
+#include "rbug/rbug_proto.h"
+
+struct rbug_connection * rbug_from_socket(int socket);
+
+void rbug_disconnect(struct rbug_connection *c);
+
+struct rbug_header * rbug_get_message(struct rbug_connection *c, uint32_t *serial);
+
+void rbug_free_header(struct rbug_header *header);
+
+struct rbug_header * rbug_demarshal(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_context.c b/src/gallium/auxiliary/rbug/rbug_context.c
new file mode 100644
index 0000000000..1832425658
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_context.c
@@ -0,0 +1,759 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_context.h"
+
+int rbug_send_context_list(struct rbug_connection *__con,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_info(struct rbug_connection *__con,
+                           rbug_context_t context,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_block(struct rbug_connection *__con,
+                                 rbug_context_t context,
+                                 rbug_block_t block,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCK));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCK, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_step(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_block_t step,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* step */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_STEP));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, step); /* step */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_STEP, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_unblock(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t unblock,
+                                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* unblock */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, unblock); /* unblock */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_UNBLOCK, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_rule(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_shader_t vertex,
+                                rbug_shader_t fragment,
+                                rbug_texture_t texture,
+                                rbug_texture_t surface,
+                                rbug_block_t block,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* vertex */
+	LEN(8); /* fragment */
+	LEN(8); /* texture */
+	LEN(8); /* surface */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_RULE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, vertex); /* vertex */
+	WRITE(8, rbug_shader_t, fragment); /* fragment */
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(8, rbug_texture_t, surface); /* surface */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_RULE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_flush(struct rbug_connection *__con,
+                            rbug_context_t context,
+                            int32_t flags,
+                            uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* flags */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_FLUSH));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, int32_t, flags); /* flags */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_FLUSH, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_context_t *contexts,
+                                 uint32_t contexts_len,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, contexts); /* contexts */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_context_t, contexts); /* contexts */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_shader_t vertex,
+                                 rbug_shader_t fragment,
+                                 rbug_texture_t *texs,
+                                 uint32_t texs_len,
+                                 rbug_texture_t *cbufs,
+                                 uint32_t cbufs_len,
+                                 rbug_texture_t zsbuf,
+                                 rbug_block_t blocker,
+                                 rbug_block_t blocked,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(8); /* vertex */
+	LEN(8); /* fragment */
+	LEN_ARRAY(8, texs); /* texs */
+	LEN_ARRAY(8, cbufs); /* cbufs */
+	LEN(8); /* zsbuf */
+	LEN(4); /* blocker */
+	LEN(4); /* blocked */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(8, rbug_shader_t, vertex); /* vertex */
+	WRITE(8, rbug_shader_t, fragment); /* fragment */
+	WRITE_ARRAY(8, rbug_texture_t, texs); /* texs */
+	WRITE_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */
+	WRITE(8, rbug_texture_t, zsbuf); /* zsbuf */
+	WRITE(4, rbug_block_t, blocker); /* blocker */
+	WRITE(4, rbug_block_t, blocked); /* blocked */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_blocked(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t block,
+                                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCKED));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCKED, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_block *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCK)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_step *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_STEP)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, step); /* step */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_unblock *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, unblock); /* unblock */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_rule *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_RULE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, vertex); /* vertex */
+	READ(8, rbug_shader_t, fragment); /* fragment */
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(8, rbug_texture_t, surface); /* surface */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
+
+struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_flush *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_FLUSH)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, int32_t, flags); /* flags */
+
+	return ret;
+}
+
+struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_context_t, contexts); /* contexts */
+
+	return ret;
+}
+
+struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(8, rbug_shader_t, vertex); /* vertex */
+	READ(8, rbug_shader_t, fragment); /* fragment */
+	READ_ARRAY(8, rbug_texture_t, texs); /* texs */
+	READ_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */
+	READ(8, rbug_texture_t, zsbuf); /* zsbuf */
+	READ(4, rbug_block_t, blocker); /* blocker */
+	READ(4, rbug_block_t, blocked); /* blocked */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_blocked *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCKED)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_context.h b/src/gallium/auxiliary/rbug/rbug_context.h
new file mode 100644
index 0000000000..da61c2365b
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_context.h
@@ -0,0 +1,212 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_CONTEXT_H_
+#define _RBUG_PROTO_CONTEXT_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+typedef enum
+{
+	RBUG_BLOCK_BEFORE = 1,
+	RBUG_BLOCK_AFTER = 2,
+	RBUG_BLOCK_RULE = 4,
+	RBUG_BLOCK_MASK = 7,
+} rbug_block_t;
+
+struct rbug_proto_context_list
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_context_info
+{
+	struct rbug_header header;
+	rbug_context_t context;
+};
+
+struct rbug_proto_context_draw_block
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t block;
+};
+
+struct rbug_proto_context_draw_step
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t step;
+};
+
+struct rbug_proto_context_draw_unblock
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t unblock;
+};
+
+struct rbug_proto_context_draw_rule
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t vertex;
+	rbug_shader_t fragment;
+	rbug_texture_t texture;
+	rbug_texture_t surface;
+	rbug_block_t block;
+};
+
+struct rbug_proto_context_flush
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	int32_t flags;
+};
+
+struct rbug_proto_context_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_context_t *contexts;
+	uint32_t contexts_len;
+};
+
+struct rbug_proto_context_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_shader_t vertex;
+	rbug_shader_t fragment;
+	rbug_texture_t *texs;
+	uint32_t texs_len;
+	rbug_texture_t *cbufs;
+	uint32_t cbufs_len;
+	rbug_texture_t zsbuf;
+	rbug_block_t blocker;
+	rbug_block_t blocked;
+};
+
+struct rbug_proto_context_draw_blocked
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t block;
+};
+
+int rbug_send_context_list(struct rbug_connection *__con,
+                           uint32_t *__serial);
+
+int rbug_send_context_info(struct rbug_connection *__con,
+                           rbug_context_t context,
+                           uint32_t *__serial);
+
+int rbug_send_context_draw_block(struct rbug_connection *__con,
+                                 rbug_context_t context,
+                                 rbug_block_t block,
+                                 uint32_t *__serial);
+
+int rbug_send_context_draw_step(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_block_t step,
+                                uint32_t *__serial);
+
+int rbug_send_context_draw_unblock(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t unblock,
+                                   uint32_t *__serial);
+
+int rbug_send_context_draw_rule(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_shader_t vertex,
+                                rbug_shader_t fragment,
+                                rbug_texture_t texture,
+                                rbug_texture_t surface,
+                                rbug_block_t block,
+                                uint32_t *__serial);
+
+int rbug_send_context_flush(struct rbug_connection *__con,
+                            rbug_context_t context,
+                            int32_t flags,
+                            uint32_t *__serial);
+
+int rbug_send_context_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_context_t *contexts,
+                                 uint32_t contexts_len,
+                                 uint32_t *__serial);
+
+int rbug_send_context_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_shader_t vertex,
+                                 rbug_shader_t fragment,
+                                 rbug_texture_t *texs,
+                                 uint32_t texs_len,
+                                 rbug_texture_t *cbufs,
+                                 uint32_t cbufs_len,
+                                 rbug_texture_t zsbuf,
+                                 rbug_block_t blocker,
+                                 rbug_block_t blocked,
+                                 uint32_t *__serial);
+
+int rbug_send_context_draw_blocked(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t block,
+                                   uint32_t *__serial);
+
+struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header);
+
+struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header);
+
+struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header);
+
+struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_core.c b/src/gallium/auxiliary/rbug/rbug_core.c
new file mode 100644
index 0000000000..876ae5a0ce
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_core.c
@@ -0,0 +1,359 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_core.h"
+
+int rbug_send_noop(struct rbug_connection *__con,
+                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_NOOP));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_NOOP, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_ping(struct rbug_connection *__con,
+                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_PING));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_PING, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_error(struct rbug_connection *__con,
+                    uint32_t error,
+                    uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* error */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, error); /* error */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_ERROR, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_ping_reply(struct rbug_connection *__con,
+                         uint32_t serial,
+                         uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_PING_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_PING_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_error_reply(struct rbug_connection *__con,
+                          uint32_t serial,
+                          uint32_t error,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* error */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, error); /* error */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_ERROR_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_noop *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_NOOP)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_ping *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_PING)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_error *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_ERROR)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, error); /* error */
+
+	return ret;
+}
+
+struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_ping_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_PING_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+
+	return ret;
+}
+
+struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_error_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_ERROR_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, error); /* error */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_core.h b/src/gallium/auxiliary/rbug/rbug_core.h
new file mode 100644
index 0000000000..99a36a0163
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_core.h
@@ -0,0 +1,105 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_CORE_H_
+#define _RBUG_PROTO_CORE_H_
+
+#include "rbug/rbug_proto.h"
+
+typedef uint64_t rbug_shader_t;
+typedef uint64_t rbug_context_t;
+typedef uint64_t rbug_texture_t;
+
+struct rbug_proto_noop
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_ping
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_error
+{
+	struct rbug_header header;
+	uint32_t error;
+};
+
+struct rbug_proto_ping_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+};
+
+struct rbug_proto_error_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t error;
+};
+
+int rbug_send_noop(struct rbug_connection *__con,
+                   uint32_t *__serial);
+
+int rbug_send_ping(struct rbug_connection *__con,
+                   uint32_t *__serial);
+
+int rbug_send_error(struct rbug_connection *__con,
+                    uint32_t error,
+                    uint32_t *__serial);
+
+int rbug_send_ping_reply(struct rbug_connection *__con,
+                         uint32_t serial,
+                         uint32_t *__serial);
+
+int rbug_send_error_reply(struct rbug_connection *__con,
+                          uint32_t serial,
+                          uint32_t error,
+                          uint32_t *__serial);
+
+struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header);
+
+struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header);
+
+struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header);
+
+struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_demarshal.c b/src/gallium/auxiliary/rbug/rbug_demarshal.c
new file mode 100644
index 0000000000..47390fbcee
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_demarshal.c
@@ -0,0 +1,93 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "rbug.h"
+
+/**
+ * Small function that looks at the proto_header and selects the correct
+ * demarshal functions and return the result.
+ */
+struct rbug_header * rbug_demarshal(struct rbug_proto_header *header)
+{
+	switch(header->opcode) {
+	case RBUG_OP_NOOP:
+		return (struct rbug_header *)rbug_demarshal_noop(header);
+	case RBUG_OP_PING:
+		return (struct rbug_header *)rbug_demarshal_ping(header);
+	case RBUG_OP_ERROR:
+		return (struct rbug_header *)rbug_demarshal_error(header);
+	case RBUG_OP_PING_REPLY:
+		return (struct rbug_header *)rbug_demarshal_ping_reply(header);
+	case RBUG_OP_ERROR_REPLY:
+		return (struct rbug_header *)rbug_demarshal_error_reply(header);
+	case RBUG_OP_TEXTURE_LIST:
+		return (struct rbug_header *)rbug_demarshal_texture_list(header);
+	case RBUG_OP_TEXTURE_INFO:
+		return (struct rbug_header *)rbug_demarshal_texture_info(header);
+	case RBUG_OP_TEXTURE_WRITE:
+		return (struct rbug_header *)rbug_demarshal_texture_write(header);
+	case RBUG_OP_TEXTURE_READ:
+		return (struct rbug_header *)rbug_demarshal_texture_read(header);
+	case RBUG_OP_TEXTURE_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_list_reply(header);
+	case RBUG_OP_TEXTURE_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_info_reply(header);
+	case RBUG_OP_TEXTURE_READ_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_read_reply(header);
+	case RBUG_OP_CONTEXT_LIST:
+		return (struct rbug_header *)rbug_demarshal_context_list(header);
+	case RBUG_OP_CONTEXT_INFO:
+		return (struct rbug_header *)rbug_demarshal_context_info(header);
+	case RBUG_OP_CONTEXT_DRAW_BLOCK:
+		return (struct rbug_header *)rbug_demarshal_context_draw_block(header);
+	case RBUG_OP_CONTEXT_DRAW_STEP:
+		return (struct rbug_header *)rbug_demarshal_context_draw_step(header);
+	case RBUG_OP_CONTEXT_DRAW_UNBLOCK:
+		return (struct rbug_header *)rbug_demarshal_context_draw_unblock(header);
+	case RBUG_OP_CONTEXT_DRAW_RULE:
+		return (struct rbug_header *)rbug_demarshal_context_draw_rule(header);
+	case RBUG_OP_CONTEXT_FLUSH:
+		return (struct rbug_header *)rbug_demarshal_context_flush(header);
+	case RBUG_OP_CONTEXT_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_context_list_reply(header);
+	case RBUG_OP_CONTEXT_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_context_info_reply(header);
+	case RBUG_OP_CONTEXT_DRAW_BLOCKED:
+		return (struct rbug_header *)rbug_demarshal_context_draw_blocked(header);
+	case RBUG_OP_SHADER_LIST:
+		return (struct rbug_header *)rbug_demarshal_shader_list(header);
+	case RBUG_OP_SHADER_INFO:
+		return (struct rbug_header *)rbug_demarshal_shader_info(header);
+	case RBUG_OP_SHADER_DISABLE:
+		return (struct rbug_header *)rbug_demarshal_shader_disable(header);
+	case RBUG_OP_SHADER_REPLACE:
+		return (struct rbug_header *)rbug_demarshal_shader_replace(header);
+	case RBUG_OP_SHADER_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_shader_list_reply(header);
+	case RBUG_OP_SHADER_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_shader_info_reply(header);
+	default:
+		return NULL;
+	}
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_internal.h b/src/gallium/auxiliary/rbug/rbug_internal.h
new file mode 100644
index 0000000000..4aba1a810f
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_internal.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file is internal to the rbug protocol code, and contains asorted
+ * features needed by the code.
+ */
+
+#ifndef _RBUG_INTERNAL_H_
+#define _RBUG_INTERNAL_H_
+
+#include "rbug/rbug_proto.h"
+
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include <errno.h>
+
+int rbug_connection_send_start(struct rbug_connection *con, enum rbug_opcode opcode, uint32_t length);
+int rbug_connection_write(struct rbug_connection *con, void *data, uint32_t size);
+int rbug_connection_send_finish(struct rbug_connection *con, uint32_t *c);
+
+/**
+ * Only works with multiples of 2
+ */
+#define PAD(from, to)                       \
+do {                                        \
+	from = (from + to - 1) & ~(to - 1); \
+} while(0)
+
+#define LEN(size)         \
+do {                      \
+	PAD(__len, size); \
+	__len += size;    \
+} while(0)
+
+#define LEN_ARRAY(size, name)       \
+do {                                \
+	LEN(4);                     \
+	PAD(__len, size);           \
+	__len += size * name##_len; \
+} while(0)
+
+#define WRITE(size, type, name)             \
+do {                                        \
+	PAD(__pos, size);                   \
+	*((type *)(&__data[__pos])) = name; \
+	__pos += size;                      \
+} while(0)
+
+#define WRITE_ARRAY(size, type, name)                    \
+do {                                                     \
+	WRITE(4, uint32_t, name##_len);                  \
+	PAD(__pos, size);                                \
+	memcpy(&__data[__pos], name, size * name##_len); \
+	__pos += size * name##_len;                      \
+} while(0)
+
+#define READ(size, type, name)                      \
+do {                                                \
+	PAD(pos, size);                             \
+	pos += size;                                \
+	if (pos > len)                              \
+		break;                              \
+	ret->name = *((type *)(&data[pos - size])); \
+} while(0)
+
+#define READ_ARRAY(size, type, name)                             \
+do {                                                             \
+	READ(4, uint32_t, name##_len);                           \
+	if (pos > len)                                           \
+		break;                                           \
+	PAD(pos, size);                                          \
+	pos += size * ret->name##_len;                           \
+	if (pos > len)                                           \
+		break;                                           \
+	ret->name = (type *)&data[pos - size * ret->name##_len]; \
+} while(0)
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_proto.h b/src/gallium/auxiliary/rbug/rbug_proto.h
new file mode 100644
index 0000000000..d273be0166
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_proto.h
@@ -0,0 +1,94 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds common definitions of the gallium remote debugging protocol.
+ */
+
+#ifndef _RBUG_PROTO_H_
+#define _RBUG_PROTO_H_
+
+#include "pipe/p_compiler.h"
+
+/**
+ * Uniqe indentifier for each command.
+ *
+ * Replys are designated by negative.
+ */
+enum rbug_opcode
+{
+	RBUG_OP_NOOP = 0,
+	RBUG_OP_PING = 1,
+	RBUG_OP_ERROR = 2,
+	RBUG_OP_PING_REPLY = -1,
+	RBUG_OP_ERROR_REPLY = -2,
+	RBUG_OP_TEXTURE_LIST = 256,
+	RBUG_OP_TEXTURE_INFO = 257,
+	RBUG_OP_TEXTURE_WRITE = 258,
+	RBUG_OP_TEXTURE_READ = 259,
+	RBUG_OP_TEXTURE_LIST_REPLY = -256,
+	RBUG_OP_TEXTURE_INFO_REPLY = -257,
+	RBUG_OP_TEXTURE_READ_REPLY = -259,
+	RBUG_OP_CONTEXT_LIST = 512,
+	RBUG_OP_CONTEXT_INFO = 513,
+	RBUG_OP_CONTEXT_DRAW_BLOCK = 514,
+	RBUG_OP_CONTEXT_DRAW_STEP = 515,
+	RBUG_OP_CONTEXT_DRAW_UNBLOCK = 516,
+	RBUG_OP_CONTEXT_DRAW_RULE = 518,
+	RBUG_OP_CONTEXT_FLUSH = 519,
+	RBUG_OP_CONTEXT_LIST_REPLY = -512,
+	RBUG_OP_CONTEXT_INFO_REPLY = -513,
+	RBUG_OP_CONTEXT_DRAW_BLOCKED = 517,
+	RBUG_OP_SHADER_LIST = 768,
+	RBUG_OP_SHADER_INFO = 769,
+	RBUG_OP_SHADER_DISABLE = 770,
+	RBUG_OP_SHADER_REPLACE = 771,
+	RBUG_OP_SHADER_LIST_REPLY = -768,
+	RBUG_OP_SHADER_INFO_REPLY = -769,
+};
+
+/**
+ * Header for demarshaled message.
+ */
+struct rbug_header
+{
+	enum rbug_opcode opcode;
+	void *__message;
+};
+
+/**
+ * Header for a message in wire format.
+ */
+struct rbug_proto_header
+{
+	int32_t opcode;
+	uint32_t length;
+};
+
+/**
+ * Forward declare connection here, as this file is included by all users.
+ */
+struct rbug_connection;
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_shader.c b/src/gallium/auxiliary/rbug/rbug_shader.c
new file mode 100644
index 0000000000..fccd2f55ef
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_shader.c
@@ -0,0 +1,468 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_shader.h"
+
+int rbug_send_shader_list(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_info(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          rbug_shader_t shader,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_disable(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint8_t disable,
+                             uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+	LEN(1); /* disable */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_DISABLE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+	WRITE(1, uint8_t, disable); /* disable */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_DISABLE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_replace(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint32_t *tokens,
+                             uint32_t tokens_len,
+                             uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+	LEN_ARRAY(4, tokens); /* tokens */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_REPLACE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+	WRITE_ARRAY(4, uint32_t, tokens); /* tokens */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_REPLACE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_list_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                rbug_shader_t *shaders,
+                                uint32_t shaders_len,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, shaders); /* shaders */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_shader_t, shaders); /* shaders */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_info_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                uint32_t *original,
+                                uint32_t original_len,
+                                uint32_t *replaced,
+                                uint32_t replaced_len,
+                                uint8_t disabled,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(4, original); /* original */
+	LEN_ARRAY(4, replaced); /* replaced */
+	LEN(1); /* disabled */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(4, uint32_t, original); /* original */
+	WRITE_ARRAY(4, uint32_t, replaced); /* replaced */
+	WRITE(1, uint8_t, disabled); /* disabled */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+
+	return ret;
+}
+
+struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+
+	return ret;
+}
+
+struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_disable *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_DISABLE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+	READ(1, uint8_t, disable); /* disable */
+
+	return ret;
+}
+
+struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_replace *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_REPLACE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+	READ_ARRAY(4, uint32_t, tokens); /* tokens */
+
+	return ret;
+}
+
+struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_shader_t, shaders); /* shaders */
+
+	return ret;
+}
+
+struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(4, uint32_t, original); /* original */
+	READ_ARRAY(4, uint32_t, replaced); /* replaced */
+	READ(1, uint8_t, disabled); /* disabled */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_shader.h b/src/gallium/auxiliary/rbug/rbug_shader.h
new file mode 100644
index 0000000000..b5d886781d
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_shader.h
@@ -0,0 +1,142 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_SHADER_H_
+#define _RBUG_PROTO_SHADER_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+struct rbug_proto_shader_list
+{
+	struct rbug_header header;
+	rbug_context_t context;
+};
+
+struct rbug_proto_shader_info
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+};
+
+struct rbug_proto_shader_disable
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+	uint8_t disable;
+};
+
+struct rbug_proto_shader_replace
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+	uint32_t *tokens;
+	uint32_t tokens_len;
+};
+
+struct rbug_proto_shader_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_shader_t *shaders;
+	uint32_t shaders_len;
+};
+
+struct rbug_proto_shader_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t *original;
+	uint32_t original_len;
+	uint32_t *replaced;
+	uint32_t replaced_len;
+	uint8_t disabled;
+};
+
+int rbug_send_shader_list(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          uint32_t *__serial);
+
+int rbug_send_shader_info(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          rbug_shader_t shader,
+                          uint32_t *__serial);
+
+int rbug_send_shader_disable(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint8_t disable,
+                             uint32_t *__serial);
+
+int rbug_send_shader_replace(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint32_t *tokens,
+                             uint32_t tokens_len,
+                             uint32_t *__serial);
+
+int rbug_send_shader_list_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                rbug_shader_t *shaders,
+                                uint32_t shaders_len,
+                                uint32_t *__serial);
+
+int rbug_send_shader_info_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                uint32_t *original,
+                                uint32_t original_len,
+                                uint32_t *replaced,
+                                uint32_t replaced_len,
+                                uint8_t disabled,
+                                uint32_t *__serial);
+
+struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_texture.c b/src/gallium/auxiliary/rbug/rbug_texture.c
new file mode 100644
index 0000000000..5a918fe6bc
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_texture.c
@@ -0,0 +1,631 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_texture.h"
+
+int rbug_send_texture_list(struct rbug_connection *__con,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_info(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_write(struct rbug_connection *__con,
+                            rbug_texture_t texture,
+                            uint32_t face,
+                            uint32_t level,
+                            uint32_t zslice,
+                            uint32_t x,
+                            uint32_t y,
+                            uint32_t w,
+                            uint32_t h,
+                            uint8_t *data,
+                            uint32_t data_len,
+                            uint32_t stride,
+                            uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+	LEN(4); /* face */
+	LEN(4); /* level */
+	LEN(4); /* zslice */
+	LEN(4); /* x */
+	LEN(4); /* y */
+	LEN(4); /* w */
+	LEN(4); /* h */
+	LEN_ARRAY(1, data); /* data */
+	LEN(4); /* stride */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_WRITE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(4, uint32_t, face); /* face */
+	WRITE(4, uint32_t, level); /* level */
+	WRITE(4, uint32_t, zslice); /* zslice */
+	WRITE(4, uint32_t, x); /* x */
+	WRITE(4, uint32_t, y); /* y */
+	WRITE(4, uint32_t, w); /* w */
+	WRITE(4, uint32_t, h); /* h */
+	WRITE_ARRAY(1, uint8_t, data); /* data */
+	WRITE(4, uint32_t, stride); /* stride */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_WRITE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_read(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t face,
+                           uint32_t level,
+                           uint32_t zslice,
+                           uint32_t x,
+                           uint32_t y,
+                           uint32_t w,
+                           uint32_t h,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+	LEN(4); /* face */
+	LEN(4); /* level */
+	LEN(4); /* zslice */
+	LEN(4); /* x */
+	LEN(4); /* y */
+	LEN(4); /* w */
+	LEN(4); /* h */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(4, uint32_t, face); /* face */
+	WRITE(4, uint32_t, level); /* level */
+	WRITE(4, uint32_t, zslice); /* zslice */
+	WRITE(4, uint32_t, x); /* x */
+	WRITE(4, uint32_t, y); /* y */
+	WRITE(4, uint32_t, w); /* w */
+	WRITE(4, uint32_t, h); /* h */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_texture_t *textures,
+                                 uint32_t textures_len,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, textures); /* textures */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_texture_t, textures); /* textures */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t target,
+                                 uint32_t format,
+                                 uint32_t *width,
+                                 uint32_t width_len,
+                                 uint32_t *height,
+                                 uint32_t height_len,
+                                 uint32_t *depth,
+                                 uint32_t depth_len,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint32_t last_level,
+                                 uint32_t nr_samples,
+                                 uint32_t tex_usage,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* target */
+	LEN(4); /* format */
+	LEN_ARRAY(4, width); /* width */
+	LEN_ARRAY(4, height); /* height */
+	LEN_ARRAY(4, depth); /* depth */
+	LEN(4); /* blockw */
+	LEN(4); /* blockh */
+	LEN(4); /* blocksize */
+	LEN(4); /* last_level */
+	LEN(4); /* nr_samples */
+	LEN(4); /* tex_usage */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, target); /* target */
+	WRITE(4, uint32_t, format); /* format */
+	WRITE_ARRAY(4, uint32_t, width); /* width */
+	WRITE_ARRAY(4, uint32_t, height); /* height */
+	WRITE_ARRAY(4, uint32_t, depth); /* depth */
+	WRITE(4, uint32_t, blockw); /* blockw */
+	WRITE(4, uint32_t, blockh); /* blockh */
+	WRITE(4, uint32_t, blocksize); /* blocksize */
+	WRITE(4, uint32_t, last_level); /* last_level */
+	WRITE(4, uint32_t, nr_samples); /* nr_samples */
+	WRITE(4, uint32_t, tex_usage); /* tex_usage */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_read_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t format,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint8_t *data,
+                                 uint32_t data_len,
+                                 uint32_t stride,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* format */
+	LEN(4); /* blockw */
+	LEN(4); /* blockh */
+	LEN(4); /* blocksize */
+	LEN_ARRAY(1, data); /* data */
+	LEN(4); /* stride */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, format); /* format */
+	WRITE(4, uint32_t, blockw); /* blockw */
+	WRITE(4, uint32_t, blockh); /* blockh */
+	WRITE(4, uint32_t, blocksize); /* blocksize */
+	WRITE_ARRAY(1, uint8_t, data); /* data */
+	WRITE(4, uint32_t, stride); /* stride */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+
+	return ret;
+}
+
+struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_write *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_WRITE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(4, uint32_t, face); /* face */
+	READ(4, uint32_t, level); /* level */
+	READ(4, uint32_t, zslice); /* zslice */
+	READ(4, uint32_t, x); /* x */
+	READ(4, uint32_t, y); /* y */
+	READ(4, uint32_t, w); /* w */
+	READ(4, uint32_t, h); /* h */
+	READ_ARRAY(1, uint8_t, data); /* data */
+	READ(4, uint32_t, stride); /* stride */
+
+	return ret;
+}
+
+struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_read *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(4, uint32_t, face); /* face */
+	READ(4, uint32_t, level); /* level */
+	READ(4, uint32_t, zslice); /* zslice */
+	READ(4, uint32_t, x); /* x */
+	READ(4, uint32_t, y); /* y */
+	READ(4, uint32_t, w); /* w */
+	READ(4, uint32_t, h); /* h */
+
+	return ret;
+}
+
+struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_texture_t, textures); /* textures */
+
+	return ret;
+}
+
+struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, target); /* target */
+	READ(4, uint32_t, format); /* format */
+	READ_ARRAY(4, uint32_t, width); /* width */
+	READ_ARRAY(4, uint32_t, height); /* height */
+	READ_ARRAY(4, uint32_t, depth); /* depth */
+	READ(4, uint32_t, blockw); /* blockw */
+	READ(4, uint32_t, blockh); /* blockh */
+	READ(4, uint32_t, blocksize); /* blocksize */
+	READ(4, uint32_t, last_level); /* last_level */
+	READ(4, uint32_t, nr_samples); /* nr_samples */
+	READ(4, uint32_t, tex_usage); /* tex_usage */
+
+	return ret;
+}
+
+struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_read_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, format); /* format */
+	READ(4, uint32_t, blockw); /* blockw */
+	READ(4, uint32_t, blockh); /* blockh */
+	READ(4, uint32_t, blocksize); /* blocksize */
+	READ_ARRAY(1, uint8_t, data); /* data */
+	READ(4, uint32_t, stride); /* stride */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_texture.h b/src/gallium/auxiliary/rbug/rbug_texture.h
new file mode 100644
index 0000000000..fbb247e1d4
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_texture.h
@@ -0,0 +1,206 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_TEXTURE_H_
+#define _RBUG_PROTO_TEXTURE_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+struct rbug_proto_texture_list
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_texture_info
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+};
+
+struct rbug_proto_texture_write
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+	uint32_t face;
+	uint32_t level;
+	uint32_t zslice;
+	uint32_t x;
+	uint32_t y;
+	uint32_t w;
+	uint32_t h;
+	uint8_t *data;
+	uint32_t data_len;
+	uint32_t stride;
+};
+
+struct rbug_proto_texture_read
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+	uint32_t face;
+	uint32_t level;
+	uint32_t zslice;
+	uint32_t x;
+	uint32_t y;
+	uint32_t w;
+	uint32_t h;
+};
+
+struct rbug_proto_texture_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_texture_t *textures;
+	uint32_t textures_len;
+};
+
+struct rbug_proto_texture_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t target;
+	uint32_t format;
+	uint32_t *width;
+	uint32_t width_len;
+	uint32_t *height;
+	uint32_t height_len;
+	uint32_t *depth;
+	uint32_t depth_len;
+	uint32_t blockw;
+	uint32_t blockh;
+	uint32_t blocksize;
+	uint32_t last_level;
+	uint32_t nr_samples;
+	uint32_t tex_usage;
+};
+
+struct rbug_proto_texture_read_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t format;
+	uint32_t blockw;
+	uint32_t blockh;
+	uint32_t blocksize;
+	uint8_t *data;
+	uint32_t data_len;
+	uint32_t stride;
+};
+
+int rbug_send_texture_list(struct rbug_connection *__con,
+                           uint32_t *__serial);
+
+int rbug_send_texture_info(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t *__serial);
+
+int rbug_send_texture_write(struct rbug_connection *__con,
+                            rbug_texture_t texture,
+                            uint32_t face,
+                            uint32_t level,
+                            uint32_t zslice,
+                            uint32_t x,
+                            uint32_t y,
+                            uint32_t w,
+                            uint32_t h,
+                            uint8_t *data,
+                            uint32_t data_len,
+                            uint32_t stride,
+                            uint32_t *__serial);
+
+int rbug_send_texture_read(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t face,
+                           uint32_t level,
+                           uint32_t zslice,
+                           uint32_t x,
+                           uint32_t y,
+                           uint32_t w,
+                           uint32_t h,
+                           uint32_t *__serial);
+
+int rbug_send_texture_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_texture_t *textures,
+                                 uint32_t textures_len,
+                                 uint32_t *__serial);
+
+int rbug_send_texture_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t target,
+                                 uint32_t format,
+                                 uint32_t *width,
+                                 uint32_t width_len,
+                                 uint32_t *height,
+                                 uint32_t height_len,
+                                 uint32_t *depth,
+                                 uint32_t depth_len,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint32_t last_level,
+                                 uint32_t nr_samples,
+                                 uint32_t tex_usage,
+                                 uint32_t *__serial);
+
+int rbug_send_texture_read_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t format,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint8_t *data,
+                                 uint32_t data_len,
+                                 uint32_t stride,
+                                 uint32_t *__serial);
+
+struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
index 57fcf6de2a..1acf3c373e 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
@@ -993,6 +993,15 @@ void sse_pmovmskb( struct x86_function *p,
    emit_modrm(p, dst, src);
 }
 
+void sse_movmskps( struct x86_function *p,
+                   struct x86_reg dst,
+                   struct x86_reg src)
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x50);
+   emit_modrm(p, dst, src);
+}
+
 /***********************************************************************
  * SSE2 instructions
  */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
index 1b5eaaca85..731a651796 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
@@ -223,6 +223,7 @@ void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg sr
 void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src );
 void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movmskps( struct x86_function *p, struct x86_reg dst, struct x86_reg src);
 
 void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile
index b4900e8dba..5f0a580b09 100644
--- a/src/gallium/auxiliary/tgsi/Makefile
+++ b/src/gallium/auxiliary/tgsi/Makefile
@@ -16,6 +16,7 @@ C_SOURCES = \
 	tgsi_sse2.c \
 	tgsi_text.c \
 	tgsi_transform.c \
+	tgsi_ureg.c \
 	tgsi_util.c
 
 include ../../Makefile.template
diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript
index 8200cce42f..b6bc2924f0 100644
--- a/src/gallium/auxiliary/tgsi/SConscript
+++ b/src/gallium/auxiliary/tgsi/SConscript
@@ -16,6 +16,7 @@ tgsi = env.ConvenienceLibrary(
 		'tgsi_sse2.c',
 		'tgsi_text.c',
 		'tgsi_transform.c',
+		'tgsi_ureg.c',
 		'tgsi_util.c',
 	])
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
index a3f4947c73..eb492076b7 100644
--- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
+++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
@@ -187,11 +187,7 @@ TGSI Instruction Specification
 
 1.2.6  CND0 - Condition Zero
 
-  dst.x = (src2.x >= 0.0) ? src0.x : src1.x
-  dst.y = (src2.y >= 0.0) ? src0.y : src1.y
-  dst.z = (src2.z >= 0.0) ? src0.z : src1.z
-  dst.w = (src2.w >= 0.0) ? src0.w : src1.w
-
+       Removed.  Use (CMP src2, src1, src0) instead.
 
 1.2.7  DOT2ADD - 2-component Dot Product And Add
 
@@ -665,9 +661,18 @@ TGSI Instruction Specification
   TBD
 
 
-1.9.8  LOOP - Loop
+1.9.8  BGNFOR - Begin a For-Loop
 
-  TBD
+  dst.x = floor(src.x)
+  dst.y = floor(src.y)
+  dst.z = floor(src.z)
+
+  if (dst.y <= 0)
+    pc = [matching ENDFOR] + 1
+  endif
+
+  Note: The destination must be a loop register.
+        The source must be a constant register.
 
 
 1.9.9  REP - Repeat
@@ -685,9 +690,16 @@ TGSI Instruction Specification
   TBD
 
 
-1.9.12  ENDLOOP - End Loop
+1.9.12  ENDFOR - End a For-Loop
 
-  TBD
+  dst.x = dst.x + dst.z
+  dst.y = dst.y - 1.0
+
+  if (dst.y > 0)
+    pc = [matching BGNFOR instruction] + 1
+  endif
+
+  Note: The destination must be a loop register.
 
 
 1.9.13  ENDREP - End Repeat
@@ -840,7 +852,7 @@ TGSI Instruction Specification
 ----------
 
 
-1.13.1  BGNLOOP2 - Begin Loop
+1.13.1  BGNLOOP - Begin a Loop
 
   TBD
 
@@ -850,7 +862,7 @@ TGSI Instruction Specification
   TBD
 
 
-1.13.3  ENDLOOP2 - End Loop
+1.13.3  ENDLOOP - End a Loop
 
   TBD
 
@@ -1015,12 +1027,12 @@ TGSI Instruction Specification
 
 1.18.1  EXPP - Approximate Exponential Base 2
 
-  Alias for EXP.
+  Use EXP. See also 1.19.3.
 
 
 1.18.2  LOGP - Logarithm Base 2
 
-  Alias for LG2.
+  Use LOG. See also 1.19.4.
 
 
 1.19  vs_2_0
@@ -1037,6 +1049,16 @@ TGSI Instruction Specification
   Alias for ARR.
 
 
+1.19.3  EXPP - Approximate Exponential Base 2
+
+  Use EX2.
+
+
+1.19.4  LOGP - Logarithm Base 2
+
+  Use LG2.
+
+
 2  Explanation of symbols used
 ==============================
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
index 94d9c638e1..e0cfc54420 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -335,7 +335,10 @@ tgsi_default_full_immediate( void )
    struct tgsi_full_immediate fullimm;
 
    fullimm.Immediate = tgsi_default_immediate();
-   fullimm.u.Pointer = (void *) 0;
+   fullimm.u[0].Float = 0.0f;
+   fullimm.u[1].Float = 0.0f;
+   fullimm.u[2].Float = 0.0f;
+   fullimm.u[3].Float = 0.0f;
 
    return fullimm;
 }
@@ -352,19 +355,19 @@ immediate_grow(
    header_bodysize_grow( header );
 }
 
-struct tgsi_immediate_float32
+union tgsi_immediate_data
 tgsi_build_immediate_float32(
    float value,
    struct tgsi_immediate *immediate,
    struct tgsi_header *header )
 {
-   struct tgsi_immediate_float32 immediate_float32;
+   union tgsi_immediate_data immediate_data;
 
-   immediate_float32.Float = value;
+   immediate_data.Float = value;
 
    immediate_grow( immediate, header );
 
-   return immediate_float32;
+   return immediate_data;
 }
 
 unsigned
@@ -384,16 +387,18 @@ tgsi_build_full_immediate(
 
    *immediate = tgsi_build_immediate( header );
 
+   assert( full_imm->Immediate.NrTokens <= 4 + 1 );
+
    for( i = 0; i < full_imm->Immediate.NrTokens - 1; i++ ) {
-      struct tgsi_immediate_float32 *if32;
+      union tgsi_immediate_data *data;
 
       if( maxsize <= size )
          return  0;
-      if32 = (struct tgsi_immediate_float32 *) &tokens[size];
+      data = (union tgsi_immediate_data *) &tokens[size];
       size++;
 
-      *if32 = tgsi_build_immediate_float32(
-         full_imm->u.ImmediateFloat32[i].Float,
+      *data = tgsi_build_immediate_float32(
+         full_imm->u[i].Float,
          immediate,
          header );
    }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h
index 9a3a077cf2..17d977b059 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.h
@@ -119,7 +119,7 @@ tgsi_build_immediate(
 struct tgsi_full_immediate
 tgsi_default_full_immediate( void );
 
-struct tgsi_immediate_float32
+union tgsi_immediate_data
 tgsi_build_immediate_float32(
    float value,
    struct tgsi_immediate *immediate,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index a6994ecd48..111d95b666 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -33,12 +33,20 @@
 #include "tgsi_info.h"
 #include "tgsi_iterate.h"
 
+
+/** Number of spaces to indent for IF/LOOP/etc */
+static const int indent_spaces = 3;
+
+
 struct dump_ctx
 {
    struct tgsi_iterate_context iter;
 
    uint instno;
+   int indent;
    
+   uint indentation;
+
    void (*printf)(struct dump_ctx *ctx, const char *format, ...);
 };
 
@@ -82,7 +90,7 @@ static const char *processor_type_names[] =
    "GEOM"
 };
 
-static const char *file_names[] =
+static const char *file_names[TGSI_FILE_COUNT] =
 {
    "NULL",
    "CONST",
@@ -91,7 +99,8 @@ static const char *file_names[] =
    "TEMP",
    "SAMP",
    "ADDR",
-   "IMM"
+   "IMM",
+   "LOOP"
 };
 
 static const char *interpolate_names[] =
@@ -295,10 +304,12 @@ iter_immediate(
    ENM( imm->Immediate.DataType, immediate_type_names );
 
    TXT( " { " );
+
+   assert( imm->Immediate.NrTokens <= 4 + 1 );
    for (i = 0; i < imm->Immediate.NrTokens - 1; i++) {
       switch (imm->Immediate.DataType) {
       case TGSI_IMM_FLOAT32:
-         FLT( imm->u.ImmediateFloat32[i].Float );
+         FLT( imm->u[i].Float );
          break;
       default:
          assert( 0 );
@@ -332,13 +343,19 @@ iter_instruction(
 {
    struct dump_ctx *ctx = (struct dump_ctx *) iter;
    uint instno = ctx->instno++;
-   
+   const struct tgsi_opcode_info *info = tgsi_get_opcode_info( inst->Instruction.Opcode );
    uint i;
    boolean first_reg = TRUE;
 
    INSTID( instno );
    TXT( ": " );
-   TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic );
+   
+   ctx->indent -= info->pre_dedent;
+   for(i = 0; (int)i < ctx->indent; ++i)
+      TXT( "  " );
+   ctx->indent += info->post_indent;
+   
+   TXT( info->mnemonic );
 
    switch (inst->Instruction.Saturate) {
    case TGSI_SAT_NONE:
@@ -470,14 +487,22 @@ iter_instruction(
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_IF:
    case TGSI_OPCODE_ELSE:
-   case TGSI_OPCODE_BGNLOOP2:
-   case TGSI_OPCODE_ENDLOOP2:
+   case TGSI_OPCODE_BGNLOOP:
+   case TGSI_OPCODE_ENDLOOP:
    case TGSI_OPCODE_CAL:
       TXT( " :" );
       UID( inst->InstructionExtLabel.Label );
       break;
    }
 
+   /* update indentation */
+   if (inst->Instruction.Opcode == TGSI_OPCODE_IF ||
+       inst->Instruction.Opcode == TGSI_OPCODE_ELSE ||
+       inst->Instruction.Opcode == TGSI_OPCODE_BGNFOR ||
+       inst->Instruction.Opcode == TGSI_OPCODE_BGNLOOP) {
+      ctx->indentation += indent_spaces;
+   }
+
    EOL();
 
    return TRUE;
@@ -491,7 +516,9 @@ tgsi_dump_instruction(
    struct dump_ctx ctx;
 
    ctx.instno = instno;
+   ctx.indent = 0;
    ctx.printf = dump_ctx_printf;
+   ctx.indentation = 0;
 
    iter_instruction( &ctx.iter, (struct tgsi_full_instruction *)inst );
 }
@@ -523,7 +550,9 @@ tgsi_dump(
    ctx.iter.epilog = NULL;
 
    ctx.instno = 0;
+   ctx.indent = 0;
    ctx.printf = dump_ctx_printf;
+   ctx.indentation = 0;
 
    tgsi_iterate_shader( tokens, &ctx.iter );
 }
@@ -575,7 +604,9 @@ tgsi_dump_str(
    ctx.base.iter.epilog = NULL;
 
    ctx.base.instno = 0;
+   ctx.base.indent = 0;
    ctx.base.printf = &str_dump_ctx_printf;
+   ctx.base.indentation = 0;
 
    ctx.str = str;
    ctx.str[0] = 0;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
index 3dc61c48ca..4a9c02b141 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
@@ -69,7 +69,7 @@ static const char *TGSI_TOKEN_TYPES[] =
    "TOKEN_TYPE_INSTRUCTION"
 };
 
-static const char *TGSI_FILES[] =
+static const char *TGSI_FILES[TGSI_FILE_COUNT] =
 {
    "FILE_NULL",
    "FILE_CONSTANT",
@@ -78,7 +78,8 @@ static const char *TGSI_FILES[] =
    "FILE_TEMPORARY",
    "FILE_SAMPLER",
    "FILE_ADDRESS",
-   "FILE_IMMEDIATE"
+   "FILE_IMMEDIATE",
+   "FILE_LOOP"
 };
 
 static const char *TGSI_INTERPOLATES[] =
@@ -283,12 +284,13 @@ dump_immediate_verbose(
       UIX( imm->Immediate.Padding );
    }
 
+   assert( imm->Immediate.NrTokens <= 4 + 1 );
    for( i = 0; i < imm->Immediate.NrTokens - 1; i++ ) {
       EOL();
       switch( imm->Immediate.DataType ) {
       case TGSI_IMM_FLOAT32:
          TXT( "\nFloat: " );
-         FLT( imm->u.ImmediateFloat32[i].Float );
+         FLT( imm->u[i].Float );
          break;
 
       default:
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 259894e72e..c79c56debd 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -304,14 +304,14 @@ tgsi_exec_machine_bind_shader(
       case TGSI_TOKEN_TYPE_IMMEDIATE:
          {
             uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
-            assert( size % 4 == 0 );
-            assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES );
+            assert( size <= 4 );
+            assert( mach->ImmLimit + 1 <= TGSI_EXEC_NUM_IMMEDIATES );
 
             for( i = 0; i < size; i++ ) {
-               mach->Imms[mach->ImmLimit + i / 4][i % 4] = 
-		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+               mach->Imms[mach->ImmLimit][i] = 
+		  parse.FullToken.FullImmediate.u[i].Float;
             }
-            mach->ImmLimit += size / 4;
+            mach->ImmLimit += 1;
          }
          break;
 
@@ -373,13 +373,18 @@ tgsi_exec_machine_bind_shader(
 }
 
 
-void
-tgsi_exec_machine_init(
-   struct tgsi_exec_machine *mach )
+struct tgsi_exec_machine *
+tgsi_exec_machine_create( void )
 {
+   struct tgsi_exec_machine *mach;
    uint i;
 
-   mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps);
+   mach = align_malloc( sizeof *mach, 16 );
+   if (!mach)
+      goto fail;
+
+   memset(mach, 0, sizeof(*mach));
+
    mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR];
 
    /* Setup constants. */
@@ -401,22 +406,24 @@ tgsi_exec_machine_init(
    (void) print_chan;
    (void) print_temp;
 #endif
+
+   return mach;
+
+fail:
+   align_free(mach);
+   return NULL;
 }
 
 
 void
-tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach)
+tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach)
 {
-   if (mach->Instructions) {
+   if (mach) {
       FREE(mach->Instructions);
-      mach->Instructions = NULL;
-      mach->NumInstructions = 0;
-   }
-   if (mach->Declarations) {
       FREE(mach->Declarations);
-      mach->Declarations = NULL;
-      mach->NumDeclarations = 0;
    }
+
+   align_free(mach);
 }
 
 
@@ -2015,8 +2022,7 @@ exec_instruction(
 
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_ARL:
-   case TGSI_OPCODE_FLOOR:
-   /* TGSI_OPCODE_FLR */
+   case TGSI_OPCODE_FLR:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          micro_flr( &r[0], &r[0] );
@@ -2299,8 +2305,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LERP:
-   /* TGSI_OPCODE_LRP */
+   case TGSI_OPCODE_LRP:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
@@ -2324,18 +2329,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_CND0:
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         FETCH(&r[0], 0, chan_index);
-         FETCH(&r[1], 1, chan_index);
-         FETCH(&r[2], 2, chan_index);
-         micro_le(&r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[2], &r[0], &r[1]);
-         STORE(&r[0], 0, chan_index);
-      }
-      break;
-
-   case TGSI_OPCODE_DOT2ADD:
-   /* TGSI_OPCODE_DP2A */
+   case TGSI_OPCODE_DP2A:
       FETCH( &r[0], 0, CHAN_X );
       FETCH( &r[1], 1, CHAN_X );
       micro_mul( &r[0], &r[0], &r[1] );
@@ -2353,18 +2347,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_INDEX:
-      /* XXX: considered for removal */
-      assert (0);
-      break;
-
-   case TGSI_OPCODE_NEGATE:
-      /* XXX: considered for removal */
-      assert (0);
-      break;
-
-   case TGSI_OPCODE_FRAC:
-   /* TGSI_OPCODE_FRC */
+   case TGSI_OPCODE_FRC:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          micro_frc( &r[0], &r[0] );
@@ -2392,8 +2375,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_EXPBASE2:
-   /* TGSI_OPCODE_EX2 */
+   case TGSI_OPCODE_EX2:
       FETCH(&r[0], 0, CHAN_X);
 
 #if FAST_MATH
@@ -2407,8 +2389,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LOGBASE2:
-   /* TGSI_OPCODE_LG2 */
+   case TGSI_OPCODE_LG2:
       FETCH( &r[0], 0, CHAN_X );
       micro_lg2( &r[0], &r[0] );
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
@@ -2416,8 +2397,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_POWER:
-   /* TGSI_OPCODE_POW */
+   case TGSI_OPCODE_POW:
       FETCH(&r[0], 0, CHAN_X);
       FETCH(&r[1], 1, CHAN_X);
 
@@ -2428,8 +2408,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_CROSSPRODUCT:
-   /* TGSI_OPCODE_XPD */
+   case TGSI_OPCODE_XPD:
       FETCH(&r[0], 0, CHAN_Y);
       FETCH(&r[1], 1, CHAN_Z);
 
@@ -2471,11 +2450,6 @@ exec_instruction(
       }
       break;
 
-    case TGSI_OPCODE_MULTIPLYMATRIX:
-       /* XXX: considered for removal */
-       assert (0);
-       break;
-
     case TGSI_OPCODE_ABS:
        FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
           FETCH(&r[0], 0, chan_index);
@@ -2782,19 +2756,32 @@ exec_instruction(
       if (mach->ExecMask) {
          /* do the call */
 
-         /* push the Cond, Loop, Cont stacks */
+         /* First, record the depths of the execution stacks.
+          * This is important for deeply nested/looped return statements.
+          * We have to unwind the stacks by the correct amount.  For a
+          * real code generator, we could determine the number of entries
+          * to pop off each stack with simple static analysis and avoid
+          * implementing this data structure at run time.
+          */
+         mach->CallStack[mach->CallStackTop].CondStackTop = mach->CondStackTop;
+         mach->CallStack[mach->CallStackTop].LoopStackTop = mach->LoopStackTop;
+         mach->CallStack[mach->CallStackTop].ContStackTop = mach->ContStackTop;
+         /* note that PC was already incremented above */
+         mach->CallStack[mach->CallStackTop].ReturnAddr = *pc;
+
+         mach->CallStackTop++;
+
+         /* Second, push the Cond, Loop, Cont, Func stacks */
          assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
          mach->CondStack[mach->CondStackTop++] = mach->CondMask;
          assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
          mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
          assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
          mach->ContStack[mach->ContStackTop++] = mach->ContMask;
-
          assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING);
          mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask;
 
-         /* note that PC was already incremented above */
-         mach->CallStack[mach->CallStackTop++] = *pc;
+         /* Finally, jump to the subroutine */
          *pc = inst->InstructionExtLabel.Label;
       }
       break;
@@ -2811,18 +2798,24 @@ exec_instruction(
             *pc = -1;
             return;
          }
-         *pc = mach->CallStack[--mach->CallStackTop];
 
-         /* pop the Cond, Loop, Cont stacks */
-         assert(mach->CondStackTop > 0);
-         mach->CondMask = mach->CondStack[--mach->CondStackTop];
-         assert(mach->LoopStackTop > 0);
-         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
-         assert(mach->ContStackTop > 0);
-         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+         assert(mach->CallStackTop > 0);
+         mach->CallStackTop--;
+
+         mach->CondStackTop = mach->CallStack[mach->CallStackTop].CondStackTop;
+         mach->CondMask = mach->CondStack[mach->CondStackTop];
+
+         mach->LoopStackTop = mach->CallStack[mach->CallStackTop].LoopStackTop;
+         mach->LoopMask = mach->LoopStack[mach->LoopStackTop];
+
+         mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop;
+         mach->ContMask = mach->ContStack[mach->ContStackTop];
+
          assert(mach->FuncStackTop > 0);
          mach->FuncMask = mach->FuncStack[--mach->FuncStackTop];
 
+         *pc = mach->CallStack[mach->CallStackTop].ReturnAddr;
+
          UPDATE_EXEC_MASK(mach);
       }
       break;
@@ -3119,26 +3112,81 @@ exec_instruction(
       mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
       break;
 
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
+      assert(mach->LoopCounterStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      for (chan_index = 0; chan_index < 3; chan_index++) {
+         FETCH( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[chan_index], 0, chan_index );
+      }
+      STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X );
+      ++mach->LoopCounterStackTop;
       /* fall-through (for now) */
-   case TGSI_OPCODE_BGNLOOP2:
+   case TGSI_OPCODE_BGNLOOP:
       /* push LoopMask and ContMasks */
       assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
       mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
       assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
       mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+      assert(mach->LoopLabelStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      mach->LoopLabelStack[mach->LoopLabelStackTop++] = *pc - 1;
+      break;
+
+   case TGSI_OPCODE_ENDFOR:
+      assert(mach->LoopCounterStackTop > 0);
+      micro_sub( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 
+                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X],
+                 &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
+      /* update LoopMask */
+      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[0] <= 0) {
+         mach->LoopMask &= ~0x1;
+      }
+      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[1] <= 0 ) {
+         mach->LoopMask &= ~0x2;
+      }
+      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[2] <= 0 ) {
+         mach->LoopMask &= ~0x4;
+      }
+      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[3] <= 0 ) {
+         mach->LoopMask &= ~0x8;
+      }
+      micro_add( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], 
+                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], 
+                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]);
+      assert(mach->LoopLabelStackTop > 0);
+      inst = mach->Instructions + mach->LoopLabelStack[mach->LoopLabelStackTop - 1];
+      STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X );
+      /* Restore ContMask, but don't pop */
+      assert(mach->ContStackTop > 0);
+      mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
+      UPDATE_EXEC_MASK(mach);
+      if (mach->ExecMask) {
+         /* repeat loop: jump to instruction just past BGNLOOP */
+         assert(mach->LoopLabelStackTop > 0);
+         *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1;
+      }
+      else {
+         /* exit loop: pop LoopMask */
+         assert(mach->LoopStackTop > 0);
+         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
+         /* pop ContMask */
+         assert(mach->ContStackTop > 0);
+         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+         assert(mach->LoopLabelStackTop > 0);
+         --mach->LoopLabelStackTop;
+         assert(mach->LoopCounterStackTop > 0);
+         --mach->LoopCounterStackTop;
+      }
+      UPDATE_EXEC_MASK(mach);
       break;
-
+      
    case TGSI_OPCODE_ENDLOOP:
-      /* fall-through (for now at least) */
-   case TGSI_OPCODE_ENDLOOP2:
       /* Restore ContMask, but don't pop */
       assert(mach->ContStackTop > 0);
       mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
       UPDATE_EXEC_MASK(mach);
       if (mach->ExecMask) {
          /* repeat loop: jump to instruction just past BGNLOOP */
-         *pc = inst->InstructionExtLabel.Label + 1;
+         assert(mach->LoopLabelStackTop > 0);
+         *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1;
       }
       else {
          /* exit loop: pop LoopMask */
@@ -3147,6 +3195,8 @@ exec_instruction(
          /* pop ContMask */
          assert(mach->ContStackTop > 0);
          mach->ContMask = mach->ContStack[--mach->ContStackTop];
+         assert(mach->LoopLabelStackTop > 0);
+         --mach->LoopLabelStackTop;
       }
       UPDATE_EXEC_MASK(mach);
       break;
@@ -3214,7 +3264,6 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
    mach->FuncMask = 0xf;
    mach->ExecMask = 0xf;
 
-   mach->CondStackTop = 0; /* temporarily subvert this assertion */
    assert(mach->CondStackTop == 0);
    assert(mach->LoopStackTop == 0);
    assert(mach->ContStackTop == 0);
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index 182e5e228b..c72f76809d 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -29,6 +29,7 @@
 #define TGSI_EXEC_H
 
 #include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
 
 #if defined __cplusplus
 extern "C" {
@@ -94,7 +95,6 @@ struct tgsi_exec_labels
 
 
 #define TGSI_EXEC_NUM_TEMPS       128
-#define TGSI_EXEC_NUM_TEMP_EXTRAS   6
 #define TGSI_EXEC_NUM_IMMEDIATES  256
 
 /*
@@ -162,9 +162,14 @@ struct tgsi_exec_labels
 #define TGSI_EXEC_MASK_I            (TGSI_EXEC_NUM_TEMPS + 3)
 #define TGSI_EXEC_MASK_C            2
 
+/* 4 register buffer for various purposes */
 #define TGSI_EXEC_TEMP_R0           (TGSI_EXEC_NUM_TEMPS + 4)
+#define TGSI_EXEC_NUM_TEMP_R        4
+
+#define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 8)
+#define TGSI_EXEC_NUM_ADDRS         1
+#define TGSI_EXEC_NUM_TEMP_EXTRAS   9
 
-#define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 5)
 
 
 #define TGSI_EXEC_MAX_COND_NESTING  20
@@ -181,30 +186,38 @@ struct tgsi_exec_labels
  */
 #define TGSI_EXEC_MAX_CONST_BUFFER  4096
 
+
+/** function call/activation record */
+struct tgsi_call_record
+{
+   uint CondStackTop;
+   uint LoopStackTop;
+   uint ContStackTop;
+   uint ReturnAddr;
+};
+
+
 /**
  * Run-time virtual machine state for executing TGSI shader.
  */
 struct tgsi_exec_machine
 {
    /* Total = program temporaries + internal temporaries
-    *         + 1 padding to align to 16 bytes
     */
-   struct tgsi_exec_vector       _Temps[TGSI_EXEC_NUM_TEMPS +
-                                        TGSI_EXEC_NUM_TEMP_EXTRAS + 1];
+   struct tgsi_exec_vector       Temps[TGSI_EXEC_NUM_TEMPS +
+                                       TGSI_EXEC_NUM_TEMP_EXTRAS];
+
+   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
+
+   struct tgsi_exec_vector       Inputs[PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Outputs[PIPE_MAX_ATTRIBS];
 
-   /*
-    * This will point to _Temps after aligning to 16B boundary.
-    */
-   struct tgsi_exec_vector       *Temps;
    struct tgsi_exec_vector       *Addrs;
 
    struct tgsi_sampler           **Samplers;
 
-   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
    unsigned                      ImmLimit;
    const float                   (*Consts)[4];
-   struct tgsi_exec_vector       *Inputs;
-   struct tgsi_exec_vector       *Outputs;
    const struct tgsi_token       *Tokens;   /**< Declarations, instructions */
    unsigned                      Processor; /**< TGSI_PROCESSOR_x */
 
@@ -230,6 +243,14 @@ struct tgsi_exec_machine
    uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int LoopStackTop;
 
+   /** Loop label stack */
+   uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int LoopLabelStackTop;
+
+   /** Loop counter stack (x = count, y = current, z = step) */
+   struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int LoopCounterStackTop;
+   
    /** Loop continue mask stack (see comments in tgsi_exec.c) */
    uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int ContStackTop;
@@ -239,7 +260,7 @@ struct tgsi_exec_machine
    int FuncStackTop;
 
    /** Function call stack for saving/restoring the program counter */
-   uint CallStack[TGSI_EXEC_MAX_CALL_NESTING];
+   struct tgsi_call_record CallStack[TGSI_EXEC_MAX_CALL_NESTING];
    int CallStackTop;
 
    struct tgsi_full_instruction *Instructions;
@@ -251,9 +272,11 @@ struct tgsi_exec_machine
    struct tgsi_exec_labels Labels;
 };
 
+struct tgsi_exec_machine *
+tgsi_exec_machine_create( void );
+
 void
-tgsi_exec_machine_init(
-   struct tgsi_exec_machine *mach );
+tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach);
 
 
 void 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c
index 37f2b66d1f..17af4cb7ad 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_info.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.c
@@ -26,136 +26,156 @@
  **************************************************************************/
 
 #include "util/u_debug.h"
+#include "util/u_memory.h"
 #include "tgsi_info.h"
 
 static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
 {
-   { 1, 1, 0, 0, "ARL", NULL, NULL },
-   { 1, 1, 0, 0, "MOV", NULL, NULL },
-   { 1, 1, 0, 0, "LIT", NULL, NULL },
-   { 1, 1, 0, 0, "RCP", "RECIP", NULL },
-   { 1, 1, 0, 0, "RSQ", "RECIPSQRT", NULL },
-   { 1, 1, 0, 0, "EXP", "EXPP", NULL },
-   { 1, 1, 0, 0, "LOG", NULL, NULL },
-   { 1, 2, 0, 0, "MUL", NULL, NULL },
-   { 1, 2, 0, 0, "ADD", NULL, NULL },
-   { 1, 2, 0, 0, "DP3", "DOT3", NULL },
-   { 1, 2, 0, 0, "DP4", "DOT4", NULL },
-   { 1, 2, 0, 0, "DST", NULL, NULL },
-   { 1, 2, 0, 0, "MIN", NULL, NULL },
-   { 1, 2, 0, 0, "MAX", NULL, NULL },
-   { 1, 2, 0, 0, "SLT", "SETLT", NULL },
-   { 1, 2, 0, 0, "SGE", "SETGE", NULL },
-   { 1, 3, 0, 0, "MAD", "MADD", NULL },
-   { 1, 2, 0, 0, "SUB", NULL, NULL },
-   { 1, 3, 0, 0, "LRP", "LERP", NULL },
-   { 1, 3, 0, 0, "CND", NULL, NULL },
-   { 1, 3, 0, 0, "CND0", NULL, NULL },
-   { 1, 3, 0, 0, "DP2A", "DP2ADD", "DOT2ADD" },
-   { 1, 2, 0, 0, "INDEX", NULL, NULL },
-   { 1, 1, 0, 0, "NEGATE", NULL, NULL },
-   { 1, 1, 0, 0, "FRC", "FRAC", NULL },
-   { 1, 3, 0, 0, "CLAMP", NULL, NULL },
-   { 1, 1, 0, 0, "FLR", "FLOOR", NULL },
-   { 1, 1, 0, 0, "ROUND", NULL, NULL },
-   { 1, 1, 0, 0, "EX2", "EXPBASE2", NULL },
-   { 1, 1, 0, 0, "LG2", "LOGBASE2", "LOGP" },
-   { 1, 2, 0, 0, "POW", "POWER", NULL },
-   { 1, 2, 0, 0, "XPD", "CRS", "CROSSPRODUCT" },
-   { 1, 2, 0, 0, "M4X4", "MULTIPLYMATRIX", NULL },
-   { 1, 1, 0, 0, "ABS", NULL, NULL },
-   { 1, 1, 0, 0, "RCC", NULL, NULL },
-   { 1, 2, 0, 0, "DPH", NULL, NULL },
-   { 1, 1, 0, 0, "COS", NULL, NULL },
-   { 1, 1, 0, 0, "DDX", "DSX", NULL },
-   { 1, 1, 0, 0, "DDY", "DSY", NULL },
-   { 0, 0, 0, 0, "KILP", NULL, NULL },
-   { 1, 1, 0, 0, "PK2H", NULL, NULL },
-   { 1, 1, 0, 0, "PK2US", NULL, NULL },
-   { 1, 1, 0, 0, "PK4B", NULL, NULL },
-   { 1, 1, 0, 0, "PK4UB", NULL, NULL },
-   { 1, 2, 0, 0, "RFL", NULL, NULL },
-   { 1, 2, 0, 0, "SEQ", NULL, NULL },
-   { 1, 2, 0, 0, "SFL", NULL, NULL },
-   { 1, 2, 0, 0, "SGT", NULL, NULL },
-   { 1, 1, 0, 0, "SIN", NULL, NULL },
-   { 1, 2, 0, 0, "SLE", NULL, NULL },
-   { 1, 2, 0, 0, "SNE", NULL, NULL },
-   { 1, 2, 0, 0, "STR", NULL, NULL },
-   { 1, 2, 1, 0, "TEX", "TEXLD", NULL },
-   { 1, 4, 1, 0, "TXD", "TEXLDD", NULL },
-   { 1, 2, 1, 0, "TXP", NULL, NULL },
-   { 1, 1, 0, 0, "UP2H", NULL, NULL },
-   { 1, 1, 0, 0, "UP2US", NULL, NULL },
-   { 1, 1, 0, 0, "UP4B", NULL, NULL },
-   { 1, 1, 0, 0, "UP4UB", NULL, NULL },
-   { 1, 3, 0, 0, "X2D", NULL, NULL },
-   { 1, 1, 0, 0, "ARA", NULL, NULL },
-   { 1, 1, 0, 0, "ARR", "MOVA", NULL },
-   { 0, 1, 0, 0, "BRA", NULL, NULL },
-   { 0, 0, 0, 1, "CAL", "CALL", NULL },
-   { 0, 0, 0, 0, "RET", NULL, NULL },
-   { 1, 1, 0, 0, "SGN", "SSG", NULL },
-   { 1, 3, 0, 0, "CMP", NULL, NULL },
-   { 1, 1, 0, 0, "SCS", "SINCOS", NULL },
-   { 1, 2, 1, 0, "TXB", "TEXLDB", NULL },
-   { 1, 1, 0, 0, "NRM", NULL, NULL },
-   { 1, 2, 0, 0, "DIV", NULL, NULL },
-   { 1, 2, 0, 0, "DP2", NULL, NULL },
-   { 1, 2, 1, 0, "TXL", NULL, NULL },
-   { 0, 0, 0, 0, "BRK", "BREAK", NULL },
-   { 0, 1, 0, 1, "IF", NULL, NULL },
-   { 0, 0, 0, 0, "LOOP", NULL, NULL },
-   { 0, 1, 0, 0, "REP", NULL, NULL },
-   { 0, 0, 0, 1, "ELSE", NULL, NULL },
-   { 0, 0, 0, 0, "ENDIF", NULL, NULL },
-   { 0, 0, 0, 0, "ENDLOOP", NULL, NULL },
-   { 0, 0, 0, 0, "ENDREP", NULL, NULL },
-   { 0, 1, 0, 0, "PUSHA", NULL, NULL },
-   { 1, 0, 0, 0, "POPA", NULL, NULL },
-   { 1, 1, 0, 0, "CEIL", NULL, NULL },
-   { 1, 1, 0, 0, "I2F", NULL, NULL },
-   { 1, 1, 0, 0, "NOT", NULL, NULL },
-   { 1, 1, 0, 0, "INT", "TRUNC", NULL },
-   { 1, 2, 0, 0, "SHL", NULL, NULL },
-   { 1, 2, 0, 0, "SHR", NULL, NULL },
-   { 1, 2, 0, 0, "AND", NULL, NULL },
-   { 1, 2, 0, 0, "OR", NULL, NULL },
-   { 1, 2, 0, 0, "MOD", NULL, NULL },
-   { 1, 2, 0, 0, "XOR", NULL, NULL },
-   { 1, 3, 0, 0, "SAD", NULL, NULL },
-   { 1, 2, 1, 0, "TXF", NULL, NULL },
-   { 1, 2, 1, 0, "TXQ", NULL, NULL },
-   { 0, 0, 0, 0, "CONT", NULL, NULL },
-   { 0, 0, 0, 0, "EMIT", NULL, NULL },
-   { 0, 0, 0, 0, "ENDPRIM", NULL, NULL },
-   { 0, 0, 0, 1, "BGNLOOP2", NULL, NULL },
-   { 0, 0, 0, 0, "BGNSUB", NULL, NULL },
-   { 0, 0, 0, 1, "ENDLOOP2", NULL, NULL },
-   { 0, 0, 0, 0, "ENDSUB", NULL, NULL },
-   { 1, 1, 0, 0, "NOISE1", NULL, NULL },
-   { 1, 1, 0, 0, "NOISE2", NULL, NULL },
-   { 1, 1, 0, 0, "NOISE3", NULL, NULL },
-   { 1, 1, 0, 0, "NOISE4", NULL, NULL },
-   { 0, 0, 0, 0, "NOP", NULL, NULL },
-   { 1, 2, 0, 0, "M4X3", NULL, NULL },
-   { 1, 2, 0, 0, "M3X4", NULL, NULL },
-   { 1, 2, 0, 0, "M3X3", NULL, NULL },
-   { 1, 2, 0, 0, "M3X2", NULL, NULL },
-   { 1, 1, 0, 0, "NRM4", NULL, NULL },
-   { 0, 1, 0, 0, "CALLNZ", NULL, NULL },
-   { 0, 1, 0, 0, "IFC", NULL, NULL },
-   { 0, 1, 0, 0, "BREAKC", NULL, NULL },
-   { 0, 1, 0, 0, "KIL", "TEXKILL", NULL },
-   { 0, 0, 0, 0, "END", NULL, NULL },
-   { 1, 1, 0, 0, "SWZ", NULL, NULL }
+   { 1, 1, 0, 0, 0, 0, "ARL", TGSI_OPCODE_ARL },
+   { 1, 1, 0, 0, 0, 0, "MOV", TGSI_OPCODE_MOV },
+   { 1, 1, 0, 0, 0, 0, "LIT", TGSI_OPCODE_LIT },
+   { 1, 1, 0, 0, 0, 0, "RCP", TGSI_OPCODE_RCP },
+   { 1, 1, 0, 0, 0, 0, "RSQ", TGSI_OPCODE_RSQ },
+   { 1, 1, 0, 0, 0, 0, "EXP", TGSI_OPCODE_EXP },
+   { 1, 1, 0, 0, 0, 0, "LOG", TGSI_OPCODE_LOG },
+   { 1, 2, 0, 0, 0, 0, "MUL", TGSI_OPCODE_MUL },
+   { 1, 2, 0, 0, 0, 0, "ADD", TGSI_OPCODE_ADD },
+   { 1, 2, 0, 0, 0, 0, "DP3", TGSI_OPCODE_DP3 },
+   { 1, 2, 0, 0, 0, 0, "DP4", TGSI_OPCODE_DP4 },
+   { 1, 2, 0, 0, 0, 0, "DST", TGSI_OPCODE_DST },
+   { 1, 2, 0, 0, 0, 0, "MIN", TGSI_OPCODE_MIN },
+   { 1, 2, 0, 0, 0, 0, "MAX", TGSI_OPCODE_MAX },
+   { 1, 2, 0, 0, 0, 0, "SLT", TGSI_OPCODE_SLT },
+   { 1, 2, 0, 0, 0, 0, "SGE", TGSI_OPCODE_SGE },
+   { 1, 3, 0, 0, 0, 0, "MAD", TGSI_OPCODE_MAD },
+   { 1, 2, 0, 0, 0, 0, "SUB", TGSI_OPCODE_SUB },
+   { 1, 3, 0, 0, 0, 0, "LRP", TGSI_OPCODE_LRP },
+   { 1, 3, 0, 0, 0, 0, "CND", TGSI_OPCODE_CND },
+   { 0, 0, 0, 0, 0, 0, "", 20 },      /* removed */
+   { 1, 3, 0, 0, 0, 0, "DP2A", TGSI_OPCODE_DP2A },
+   { 0, 0, 0, 0, 0, 0, "", 22 },      /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 23 },      /* removed */
+   { 1, 1, 0, 0, 0, 0, "FRC", TGSI_OPCODE_FRC },
+   { 1, 3, 0, 0, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP },
+   { 1, 1, 0, 0, 0, 0, "FLR", TGSI_OPCODE_FLR },
+   { 1, 1, 0, 0, 0, 0, "ROUND", TGSI_OPCODE_ROUND },
+   { 1, 1, 0, 0, 0, 0, "EX2", TGSI_OPCODE_EX2 },
+   { 1, 1, 0, 0, 0, 0, "LG2", TGSI_OPCODE_LG2 },
+   { 1, 2, 0, 0, 0, 0, "POW", TGSI_OPCODE_POW },
+   { 1, 2, 0, 0, 0, 0, "XPD", TGSI_OPCODE_XPD },
+   { 0, 0, 0, 0, 0, 0, "", 32 },      /* removed */
+   { 1, 1, 0, 0, 0, 0, "ABS", TGSI_OPCODE_ABS },
+   { 1, 1, 0, 0, 0, 0, "RCC", TGSI_OPCODE_RCC },
+   { 1, 2, 0, 0, 0, 0, "DPH", TGSI_OPCODE_DPH },
+   { 1, 1, 0, 0, 0, 0, "COS", TGSI_OPCODE_COS },
+   { 1, 1, 0, 0, 0, 0, "DDX", TGSI_OPCODE_DDX },
+   { 1, 1, 0, 0, 0, 0, "DDY", TGSI_OPCODE_DDY },
+   { 0, 0, 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP },
+   { 1, 1, 0, 0, 0, 0, "PK2H", TGSI_OPCODE_PK2H },
+   { 1, 1, 0, 0, 0, 0, "PK2US", TGSI_OPCODE_PK2US },
+   { 1, 1, 0, 0, 0, 0, "PK4B", TGSI_OPCODE_PK4B },
+   { 1, 1, 0, 0, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB },
+   { 1, 2, 0, 0, 0, 0, "RFL", TGSI_OPCODE_RFL },
+   { 1, 2, 0, 0, 0, 0, "SEQ", TGSI_OPCODE_SEQ },
+   { 1, 2, 0, 0, 0, 0, "SFL", TGSI_OPCODE_SFL },
+   { 1, 2, 0, 0, 0, 0, "SGT", TGSI_OPCODE_SGT },
+   { 1, 1, 0, 0, 0, 0, "SIN", TGSI_OPCODE_SIN },
+   { 1, 2, 0, 0, 0, 0, "SLE", TGSI_OPCODE_SLE },
+   { 1, 2, 0, 0, 0, 0, "SNE", TGSI_OPCODE_SNE },
+   { 1, 2, 0, 0, 0, 0, "STR", TGSI_OPCODE_STR },
+   { 1, 2, 1, 0, 0, 0, "TEX", TGSI_OPCODE_TEX },
+   { 1, 4, 1, 0, 0, 0, "TXD", TGSI_OPCODE_TXD },
+   { 1, 2, 1, 0, 0, 0, "TXP", TGSI_OPCODE_TXP },
+   { 1, 1, 0, 0, 0, 0, "UP2H", TGSI_OPCODE_UP2H },
+   { 1, 1, 0, 0, 0, 0, "UP2US", TGSI_OPCODE_UP2US },
+   { 1, 1, 0, 0, 0, 0, "UP4B", TGSI_OPCODE_UP4B },
+   { 1, 1, 0, 0, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB },
+   { 1, 3, 0, 0, 0, 0, "X2D", TGSI_OPCODE_X2D },
+   { 1, 1, 0, 0, 0, 0, "ARA", TGSI_OPCODE_ARA },
+   { 1, 1, 0, 0, 0, 0, "ARR", TGSI_OPCODE_ARR },
+   { 0, 1, 0, 0, 0, 0, "BRA", TGSI_OPCODE_BRA },
+   { 0, 0, 0, 1, 0, 0, "CAL", TGSI_OPCODE_CAL },
+   { 0, 0, 0, 0, 0, 0, "RET", TGSI_OPCODE_RET },
+   { 1, 1, 0, 0, 0, 0, "SSG", TGSI_OPCODE_SSG },
+   { 1, 3, 0, 0, 0, 0, "CMP", TGSI_OPCODE_CMP },
+   { 1, 1, 0, 0, 0, 0, "SCS", TGSI_OPCODE_SCS },
+   { 1, 2, 1, 0, 0, 0, "TXB", TGSI_OPCODE_TXB },
+   { 1, 1, 0, 0, 0, 0, "NRM", TGSI_OPCODE_NRM },
+   { 1, 2, 0, 0, 0, 0, "DIV", TGSI_OPCODE_DIV },
+   { 1, 2, 0, 0, 0, 0, "DP2", TGSI_OPCODE_DP2 },
+   { 1, 2, 1, 0, 0, 0, "TXL", TGSI_OPCODE_TXL },
+   { 0, 0, 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK },
+   { 0, 1, 0, 1, 0, 1, "IF", TGSI_OPCODE_IF },
+   { 1, 1, 0, 0, 0, 1, "BGNFOR", TGSI_OPCODE_BGNFOR },
+   { 0, 1, 0, 0, 0, 1, "REP", TGSI_OPCODE_REP },
+   { 0, 0, 0, 1, 1, 1, "ELSE", TGSI_OPCODE_ELSE },
+   { 0, 0, 0, 0, 1, 0, "ENDIF", TGSI_OPCODE_ENDIF },
+   { 1, 0, 0, 0, 1, 0, "ENDFOR", TGSI_OPCODE_ENDFOR },
+   { 0, 0, 0, 0, 1, 0, "ENDREP", TGSI_OPCODE_ENDREP },
+   { 0, 1, 0, 0, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA },
+   { 1, 0, 0, 0, 0, 0, "POPA", TGSI_OPCODE_POPA },
+   { 1, 1, 0, 0, 0, 0, "CEIL", TGSI_OPCODE_CEIL },
+   { 1, 1, 0, 0, 0, 0, "I2F", TGSI_OPCODE_I2F },
+   { 1, 1, 0, 0, 0, 0, "NOT", TGSI_OPCODE_NOT },
+   { 1, 1, 0, 0, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC },
+   { 1, 2, 0, 0, 0, 0, "SHL", TGSI_OPCODE_SHL },
+   { 1, 2, 0, 0, 0, 0, "SHR", TGSI_OPCODE_SHR },
+   { 1, 2, 0, 0, 0, 0, "AND", TGSI_OPCODE_AND },
+   { 1, 2, 0, 0, 0, 0, "OR", TGSI_OPCODE_OR },
+   { 1, 2, 0, 0, 0, 0, "MOD", TGSI_OPCODE_MOD },
+   { 1, 2, 0, 0, 0, 0, "XOR", TGSI_OPCODE_XOR },
+   { 1, 3, 0, 0, 0, 0, "SAD", TGSI_OPCODE_SAD },
+   { 1, 2, 1, 0, 0, 0, "TXF", TGSI_OPCODE_TXF },
+   { 1, 2, 1, 0, 0, 0, "TXQ", TGSI_OPCODE_TXQ },
+   { 0, 0, 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT },
+   { 0, 0, 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT },
+   { 0, 0, 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM },
+   { 0, 0, 0, 1, 0, 1, "BGNLOOP", TGSI_OPCODE_BGNLOOP },
+   { 0, 0, 0, 0, 0, 1, "BGNSUB", TGSI_OPCODE_BGNSUB },
+   { 0, 0, 0, 1, 1, 0, "ENDLOOP", TGSI_OPCODE_ENDLOOP },
+   { 0, 0, 0, 0, 1, 0, "ENDSUB", TGSI_OPCODE_ENDSUB },
+   { 1, 1, 0, 0, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 },
+   { 1, 1, 0, 0, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 },
+   { 1, 1, 0, 0, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 },
+   { 1, 1, 0, 0, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 },
+   { 0, 0, 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP },
+   { 0, 0, 0, 0, 0, 0, "", 108 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 109 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 110 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 111 },     /* removed */
+   { 1, 1, 0, 0, 0, 0, "NRM4", TGSI_OPCODE_NRM4 },
+   { 0, 1, 0, 0, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ },
+   { 0, 1, 0, 0, 0, 0, "IFC", TGSI_OPCODE_IFC },
+   { 0, 1, 0, 0, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC },
+   { 0, 1, 0, 0, 0, 0, "KIL", TGSI_OPCODE_KIL },
+   { 0, 0, 0, 0, 0, 0, "END", TGSI_OPCODE_END },
+   { 1, 1, 0, 0, 0, 0, "SWZ", TGSI_OPCODE_SWZ }
 };
 
 const struct tgsi_opcode_info *
 tgsi_get_opcode_info( uint opcode )
 {
+   static boolean firsttime = 1;
+
+   if (firsttime) {
+      unsigned i;
+      firsttime = 0;
+      for (i = 0; i < Elements(opcode_info); i++)
+         assert(opcode_info[i].opcode == i);
+   }
+   
    if (opcode < TGSI_OPCODE_LAST)
       return &opcode_info[opcode];
+
    assert( 0 );
    return NULL;
 }
+
+
+const char *
+tgsi_get_opcode_name( uint opcode )
+{
+   const struct tgsi_opcode_info *info = tgsi_get_opcode_info(opcode);
+   return info->mnemonic;
+}
+
diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h
index 077e25acd7..74713c3b98 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_info.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.h
@@ -36,18 +36,23 @@ extern "C" {
 
 struct tgsi_opcode_info
 {
-   uint num_dst;
-   uint num_src;
-   boolean is_tex;
-   boolean is_branch;
+   unsigned num_dst:3;
+   unsigned num_src:3;
+   unsigned is_tex:1;
+   unsigned is_branch:1;
+   int pre_dedent:2;
+   int post_indent:2;
    const char *mnemonic;
-   const char *alt_mnemonic1;
-   const char *alt_mnemonic2;
+   uint opcode;
 };
 
 const struct tgsi_opcode_info *
 tgsi_get_opcode_info( uint opcode );
 
+const char *
+tgsi_get_opcode_name( uint opcode );
+
+
 #if defined __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
new file mode 100644
index 0000000000..e7bcf4bf75
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
@@ -0,0 +1,172 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+#ifndef OP12_TEX
+#define OP12_TEX(a) OP12(a)
+#endif
+
+#ifndef OP14_TEX
+#define OP14_TEX(a) OP14(a)
+#endif
+
+#ifndef OP00_LBL
+#define OP00_LBL(a) OP00(a)
+#endif
+
+#ifndef OP01_LBL
+#define OP01_LBL(a) OP01(a)
+#endif
+
+OP11(ARL)
+OP11(MOV)
+OP11(LIT)
+OP11(RCP)
+OP11(RSQ)
+OP11(EXP)
+OP11(LOG)
+OP12(MUL)
+OP12(ADD)
+OP12(DP3)
+OP12(DP4)
+OP12(DST)
+OP12(MIN)
+OP12(MAX)
+OP12(SLT)
+OP12(SGE)
+OP13(MAD)
+OP12(SUB)
+OP13(LRP)
+OP13(CND)
+OP13(DP2A)
+OP11(FRC)
+OP13(CLAMP)
+OP11(FLR)
+OP11(ROUND)
+OP11(EX2)
+OP11(LG2)
+OP12(POW)
+OP12(XPD)
+OP11(ABS)
+OP11(RCC)
+OP12(DPH)
+OP11(COS)
+OP11(DDX)
+OP11(DDY)
+OP00(KILP)
+OP11(PK2H)
+OP11(PK2US)
+OP11(PK4B)
+OP11(PK4UB)
+OP12(RFL)
+OP12(SEQ)
+OP12(SFL)
+OP12(SGT)
+OP11(SIN)
+OP12(SLE)
+OP12(SNE)
+OP12(STR)
+OP12_TEX(TEX)
+OP14_TEX(TXD)
+OP12_TEX(TXP)
+OP11(UP2H)
+OP11(UP2US)
+OP11(UP4B)
+OP11(UP4UB)
+OP13(X2D)
+OP11(ARA)
+OP11(ARR)
+OP01(BRA)
+OP00_LBL(CAL)
+OP00(RET)
+OP11(SSG)
+OP13(CMP)
+OP11(SCS)
+OP12_TEX(TXB)
+OP11(NRM)
+OP12(DIV)
+OP12(DP2)
+OP12_TEX(TXL)
+OP00(BRK)
+OP01_LBL(IF)
+OP11(BGNFOR)
+OP01(REP)
+OP00_LBL(ELSE)
+OP00(ENDIF)
+OP10(ENDFOR)
+OP00(ENDREP)
+OP01(PUSHA)
+OP10(POPA)
+OP11(CEIL)
+OP11(I2F)
+OP11(NOT)
+OP11(TRUNC)
+OP12(SHL)
+OP12(SHR)
+OP12(AND)
+OP12(OR)
+OP12(MOD)
+OP12(XOR)
+OP13(SAD)
+OP12_TEX(TXF)
+OP12_TEX(TXQ)
+OP00(CONT)
+OP00(EMIT)
+OP00(ENDPRIM)
+OP00_LBL(BGNLOOP)
+OP00(BGNSUB)
+OP00_LBL(ENDLOOP)
+OP00(ENDSUB)
+OP11(NOISE1)
+OP11(NOISE2)
+OP11(NOISE3)
+OP11(NOISE4)
+OP00(NOP)
+OP11(NRM4)
+OP01(CALLNZ)
+OP01(IFC)
+OP01(BREAKC)
+OP01(KIL)
+OP00(END)
+OP11(SWZ)
+
+
+#undef OP00
+#undef OP01
+#undef OP10
+#undef OP11
+#undef OP12
+#undef OP13
+
+#ifdef OP14
+#undef OP14
+#endif
+
+#undef OP00_LBL
+#undef OP01_LBL
+
+#undef OP12_TEX
+#undef OP14_TEX
+
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c
index 7f2cfb7988..4870f82b6b 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_parse.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c
@@ -42,9 +42,6 @@ void
 tgsi_full_token_free(
    union tgsi_full_token *full_token )
 {
-   if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) {
-      FREE( (void *) full_token->FullImmediate.u.Pointer );
-   }
 }
 
 unsigned
@@ -156,14 +153,8 @@ tgsi_parse_token(
       case TGSI_IMM_FLOAT32:
          {
             uint imm_count = imm->Immediate.NrTokens - 1;
-            struct tgsi_immediate_float32 *data;
-
-            data = (struct tgsi_immediate_float32 *) MALLOC(sizeof(struct tgsi_immediate_float32) * imm_count);
-            if (data) {
-               for (i = 0; i < imm_count; i++) {
-                  next_token(ctx, &data[i]);
-               }
-               imm->u.ImmediateFloat32 = data;
+            for (i = 0; i < imm_count; i++) {
+               next_token(ctx, &imm->u[i]);
             }
          }
          break;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h
index 5268c02900..a26ee5ba86 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_parse.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h
@@ -73,11 +73,7 @@ struct tgsi_full_declaration
 struct tgsi_full_immediate
 {
    struct tgsi_immediate   Immediate;
-   union
-   {
-      const void                          *Pointer;
-      const struct tgsi_immediate_float32 *ImmediateFloat32;
-   } u;
+   union tgsi_immediate_data u[4];
 };
 
 #define TGSI_FULL_MAX_DST_REGISTERS 2
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
index 8466d9bc22..4b1c7d4e01 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
@@ -38,6 +38,7 @@
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "util/u_sse.h"
+#include "tgsi/tgsi_info.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 #include "tgsi_dump.h"
@@ -619,17 +620,17 @@ emit_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst)
             ppc_vandc(gen->f, v1, v0, bit31_vec); /* v1 = v0 & ~bit31 */
          }
          break;
-      case TGSI_OPCODE_FLOOR:
+      case TGSI_OPCODE_FLR:
          ppc_vrfim(gen->f, v1, v0);         /* v1 = floor(v0) */
          break;
-      case TGSI_OPCODE_FRAC:
+      case TGSI_OPCODE_FRC:
          ppc_vrfim(gen->f, v1, v0);      /* tmp = floor(v0) */
          ppc_vsubfp(gen->f, v1, v0, v1); /* v1 = v0 - v1 */
          break;
-      case TGSI_OPCODE_EXPBASE2:
+      case TGSI_OPCODE_EX2:
          ppc_vexptefp(gen->f, v1, v0);     /* v1 = 2^v0 */
          break;
-      case TGSI_OPCODE_LOGBASE2:
+      case TGSI_OPCODE_LG2:
          /* XXX this may be broken! */
          ppc_vlogefp(gen->f, v1, v0);      /* v1 = log2(v0) */
          break;
@@ -1120,10 +1121,10 @@ emit_instruction(struct gen_context *gen,
    case TGSI_OPCODE_MOV:
    case TGSI_OPCODE_SWZ:
    case TGSI_OPCODE_ABS:
-   case TGSI_OPCODE_FLOOR:
-   case TGSI_OPCODE_FRAC:
-   case TGSI_OPCODE_EXPBASE2:
-   case TGSI_OPCODE_LOGBASE2:
+   case TGSI_OPCODE_FLR:
+   case TGSI_OPCODE_FRC:
+   case TGSI_OPCODE_EX2:
+   case TGSI_OPCODE_LG2:
       emit_unaryop(gen, inst);
       break;
    case TGSI_OPCODE_RSQ:
@@ -1326,8 +1327,10 @@ tgsi_emit_ppc(const struct tgsi_token *tokens,
          ok = emit_instruction(&gen, &parse.FullToken.FullInstruction);
 
 	 if (!ok) {
-	    debug_printf("failed to translate tgsi opcode %d to PPC (%s)\n", 
-			 parse.FullToken.FullInstruction.Instruction.Opcode,
+            uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode;
+	    debug_printf("failed to translate tgsi opcode %d (%s) to PPC (%s)\n", 
+			 opcode,
+                         tgsi_get_opcode_name(opcode),
                          parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ?
                          "vertex shader" : "fragment shader");
 	 }
@@ -1342,7 +1345,7 @@ tgsi_emit_ppc(const struct tgsi_token *tokens,
             assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES);
             for (i = 0; i < size; i++) {
                immediates[num_immediates][i] =
-		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+		  parse.FullToken.FullImmediate.u[i].Float;
             }
             num_immediates++;
          }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index 6f1f5c2b4b..8a13885da9 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -131,7 +131,7 @@ is_register_used(
    return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
 }
 
-static const char *file_names[] =
+static const char *file_names[TGSI_FILE_COUNT] =
 {
    "NULL",
    "CONST",
@@ -140,7 +140,8 @@ static const char *file_names[] =
    "TEMP",
    "SAMP",
    "ADDR",
-   "IMM"
+   "IMM",
+   "LOOP"
 };
 
 static boolean
@@ -198,10 +199,10 @@ iter_instruction(
    }
 
    if (info->num_dst != inst->Instruction.NumDstRegs) {
-      report_error( ctx, "Invalid number of destination operands, should be %u", info->num_dst );
+      report_error( ctx, "%s: Invalid number of destination operands, should be %u", info->mnemonic, info->num_dst );
    }
    if (info->num_src != inst->Instruction.NumSrcRegs) {
-      report_error( ctx, "Invalid number of source operands, should be %u", info->num_src );
+      report_error( ctx, "%s: Invalid number of source operands, should be %u", info->mnemonic, info->num_src );
    }
 
    /* Check destination and source registers' validity.
@@ -234,9 +235,29 @@ iter_instruction(
             index,
             "indirect",
             FALSE );
-         if (file != TGSI_FILE_ADDRESS || index != 0)
-            report_warning( ctx, "Indirect register not ADDR[0]" );
+         if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) {
+            report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]");
+         }
+      }
+   }
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_BGNFOR:
+   case TGSI_OPCODE_ENDFOR:
+      if (inst->FullDstRegisters[0].DstRegister.File != TGSI_FILE_LOOP ||
+          inst->FullDstRegisters[0].DstRegister.Index != 0) {
+         report_error(ctx, "Destination register must be LOOP[0]");
+      }
+      break;
+   }
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_BGNFOR:
+      if (inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_CONSTANT &&
+          inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_IMMEDIATE) {
+         report_error(ctx, "Source register file must be either CONST or IMM");
       }
+      break;
    }
 
    ctx->num_instructions++;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index a13368c80c..3cdf8b9f35 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -32,9 +32,11 @@
 #include "util/u_debug.h"
 #include "pipe/p_shader_tokens.h"
 #include "util/u_math.h"
+#include "util/u_memory.h"
 #if defined(PIPE_ARCH_SSE)
 #include "util/u_sse.h"
 #endif
+#include "tgsi/tgsi_info.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 #include "tgsi_exec.h"
@@ -100,37 +102,55 @@ get_const_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_CX );
+      reg_AX );
 }
 
 static struct x86_reg
-get_input_base( void )
+get_machine_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_AX );
+      reg_CX );
+}
+
+static struct x86_reg
+get_input_base( void )
+{
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Inputs) );
 }
 
 static struct x86_reg
 get_output_base( void )
 {
-   return x86_make_reg(
-      file_REG32,
-      reg_DX );
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Outputs) );
 }
 
 static struct x86_reg
 get_temp_base( void )
 {
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Temps) );
+}
+
+static struct x86_reg
+get_coef_base( void )
+{
    return x86_make_reg(
       file_REG32,
       reg_BX );
 }
 
 static struct x86_reg
-get_coef_base( void )
+get_sampler_base( void )
 {
-   return get_output_base();
+   return x86_make_reg(
+      file_REG32,
+      reg_DI );
 }
 
 static struct x86_reg
@@ -138,7 +158,7 @@ get_immediate_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_DI );
+      reg_DX );
 }
 
 
@@ -168,6 +188,15 @@ get_const(
 }
 
 static struct x86_reg
+get_sampler_ptr(
+   unsigned unit )
+{
+   return x86_make_disp(
+      get_sampler_base(),
+      unit * sizeof( struct tgsi_sampler * ) );
+}
+
+static struct x86_reg
 get_input(
    unsigned vec,
    unsigned chan )
@@ -241,12 +270,14 @@ emit_const(
       /* 'vec' is the offset from the address register's value.
        * We're loading CONST[ADDR+vec] into an xmm register.
        */
-      struct x86_reg r0 = get_input_base();
-      struct x86_reg r1 = get_output_base();
+      struct x86_reg r0 = get_immediate_base();
+      struct x86_reg r1 = get_coef_base();
       uint i;
 
       assert( indirectFile == TGSI_FILE_ADDRESS );
       assert( indirectIndex == 0 );
+      assert( r0.mod == mod_REG );
+      assert( r1.mod == mod_REG );
 
       x86_push( func, r0 );
       x86_push( func, r1 );
@@ -520,24 +551,15 @@ emit_coef_dady(
  * that the stack pointer is 16 byte aligned, as expected.
  */
 static void
-emit_func_call_dst(
+emit_func_call(
    struct x86_function *func,
-   unsigned xmm_save,
-   unsigned xmm_dst,
+   unsigned xmm_save_mask,
+   const struct x86_reg *arg,
+   unsigned nr_args,
    void (PIPE_CDECL *code)() )
 {
    struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
    unsigned i, n;
-   unsigned xmm_mask;
-   
-   /* Bitmask of the xmm registers to save */
-   xmm_mask = (1 << xmm_save) - 1;
-   xmm_mask &= ~(1 << xmm_dst);
-
-   sse_movaps(
-      func,
-      get_temp( TEMP_R0, 0 ),
-      make_xmm( xmm_dst ) );
 
    x86_push(
       func,
@@ -549,8 +571,10 @@ emit_func_call_dst(
       func,
       x86_make_reg( file_REG32, reg_DX) );
    
+   /* Store XMM regs to the stack
+    */
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i))
+      if(xmm_save_mask & (1 << i))
          ++n;
    
    x86_sub_imm(
@@ -559,26 +583,42 @@ emit_func_call_dst(
       n*16);
 
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i)) {
+      if(xmm_save_mask & (1 << i)) {
          sse_movups(
             func,
             x86_make_disp( x86_make_reg( file_REG32, reg_SP ), n*16 ),
             make_xmm( i ) );
          ++n;
       }
+
+   for (i = 0; i < nr_args; i++) {
+      /* Load the address of the buffer we use for passing arguments and
+       * receiving results:
+       */
+      x86_lea(
+	 func,
+	 ecx,
+	 arg[i] );
    
-   x86_lea(
-      func,
-      ecx,
-      get_temp( TEMP_R0, 0 ) );
-   
-   x86_push( func, ecx );
+      /* Push actual function arguments (currently just the pointer to
+       * the buffer above), and call the function:
+       */
+      x86_push( func, ecx );
+   }
+
    x86_mov_reg_imm( func, ecx, (unsigned long) code );
    x86_call( func, ecx );
-   x86_pop(func, ecx );
-   
+
+   /* Pop the arguments (or just add an immediate to esp)
+    */
+   for (i = 0; i < nr_args; i++) {
+      x86_pop(func, ecx );
+   }
+
+   /* Pop the saved XMM regs:
+    */
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i)) {
+      if(xmm_save_mask & (1 << i)) {
          sse_movups(
             func,
             make_xmm( i ),
@@ -602,34 +642,86 @@ emit_func_call_dst(
    x86_pop(
       func,
       x86_make_reg( file_REG32, reg_AX) );
+}
+
+static void
+emit_func_call_dst_src1(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst,
+   unsigned xmm_src0,
+   void (PIPE_CDECL *code)() )
+{
+   struct x86_reg store = get_temp( TEMP_R0, 0 );
+   unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst);
+   
+   /* Store our input parameters (in xmm regs) to the buffer we use
+    * for passing arguments.  We will pass a pointer to this buffer as
+    * the actual function argument.
+    */
+   sse_movaps(
+      func,
+      store,
+      make_xmm( xmm_src0 ) );
+
+   emit_func_call( func,
+                   xmm_mask,
+                   &store,
+                   1,
+                   code );
 
    sse_movaps(
       func,
       make_xmm( xmm_dst ),
-      get_temp( TEMP_R0, 0 ) );
+      store );
 }
 
+
 static void
-emit_func_call_dst_src(
+emit_func_call_dst_src2(
    struct x86_function *func,
    unsigned xmm_save, 
    unsigned xmm_dst,
-   unsigned xmm_src,
+   unsigned xmm_src0,
+   unsigned xmm_src1,
    void (PIPE_CDECL *code)() )
 {
+   struct x86_reg store = get_temp( TEMP_R0, 0 );
+   unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst);
+
+   /* Store two inputs to parameter buffer.
+    */
    sse_movaps(
       func,
-      get_temp( TEMP_R0, 1 ),
-      make_xmm( xmm_src ) );
+      store,
+      make_xmm( xmm_src0 ) );
 
-   emit_func_call_dst(
+   sse_movaps(
       func,
-      xmm_save,
-      xmm_dst,
-      code );
+      x86_make_disp( store, 4 * sizeof(float) ),
+      make_xmm( xmm_src1 ) );
+
+
+   /* Emit the call
+    */
+   emit_func_call( func,
+                   xmm_mask,
+                   &store,
+                   1,
+                   code );
+
+   /* Retrieve the results:
+    */
+   sse_movaps(
+      func,
+      make_xmm( xmm_dst ),
+      store );
 }
 
 
+
+
+
 #if defined(PIPE_ARCH_SSE)
 
 /*
@@ -782,10 +874,11 @@ emit_cos(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save, 
       xmm_dst,
+      xmm_dst,
       cos4f );
 }
 
@@ -812,10 +905,11 @@ emit_ex2(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       ex24f );
 }
 
@@ -857,10 +951,11 @@ emit_flr(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       flr4f );
 }
 
@@ -880,10 +975,11 @@ emit_frc(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       frc4f );
 }
 
@@ -910,10 +1006,11 @@ emit_lg2(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       lg24f );
 }
 
@@ -975,13 +1072,15 @@ emit_pow(
    struct x86_function *func,
    unsigned xmm_save, 
    unsigned xmm_dst,
-   unsigned xmm_src )
+   unsigned xmm_src0,
+   unsigned xmm_src1 )
 {
-   emit_func_call_dst_src(
+   emit_func_call_dst_src2(
       func,
       xmm_save,
       xmm_dst,
-      xmm_src,
+      xmm_src0,
+      xmm_src1,
       pow4f );
 }
 
@@ -1017,10 +1116,11 @@ emit_rnd(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       rnd4f );
 }
 
@@ -1099,10 +1199,11 @@ emit_sgn(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       sgn4f );
 }
 
@@ -1121,10 +1222,11 @@ emit_sin (struct x86_function *func,
           unsigned xmm_save, 
           unsigned xmm_dst)
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       sin4f );
 }
 
@@ -1140,6 +1242,12 @@ emit_sub(
       make_xmm( xmm_src ) );
 }
 
+
+
+
+
+
+
 /**
  * Register fetch.
  */
@@ -1298,20 +1406,164 @@ emit_store(
 #define STORE( FUNC, INST, XMM, INDEX, CHAN )\
    emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN )
 
+
+static void PIPE_CDECL
+fetch_texel( struct tgsi_sampler **sampler,
+             float *store )
+{
+#if 0
+   uint j;
+
+   debug_printf("%s sampler: %p (%p) store: %p\n", 
+                __FUNCTION__,
+                sampler, *sampler,
+                store );
+
+   debug_printf("lodbias %f\n", store[12]);
+
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d texcoord %f %f\n", 
+                   j, 
+                   store[0+j],
+                   store[4+j]);
+#endif
+
+   {
+      float rgba[NUM_CHANNELS][QUAD_SIZE];
+      (*sampler)->get_samples(*sampler, 
+                              &store[0], 
+                              &store[4], 
+                              &store[8], 
+                              0.0f, /*store[12],  lodbias */
+                              rgba);
+
+      memcpy( store, rgba, 16 * sizeof(float));
+   }
+
+#if 0
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d result %f %f %f %f\n", 
+                   j, 
+                   store[0+j],
+                   store[4+j],
+                   store[8+j],
+                   store[12+j]);
+#endif
+}
+
 /**
  * High-level instruction translators.
  */
 
 static void
+emit_tex( struct x86_function *func,
+          const struct tgsi_full_instruction *inst,
+          boolean lodbias,
+          boolean projected)
+{
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   struct x86_reg args[2];
+   unsigned count;
+   unsigned i;
+
+   switch (inst->InstructionExtTexture.Texture) {
+   case TGSI_TEXTURE_1D:
+      count = 1;
+      break;
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+      count = 2;
+      break;
+   case TGSI_TEXTURE_SHADOW1D:
+   case TGSI_TEXTURE_SHADOW2D:
+   case TGSI_TEXTURE_SHADOWRECT:
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+      count = 3;
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+   if (lodbias) {
+      FETCH( func, *inst, 3, 0, 3 );
+   }
+   else {
+      emit_tempf(
+         func,
+         3,
+         TGSI_EXEC_TEMP_00000000_I,
+         TGSI_EXEC_TEMP_00000000_C );
+
+   }
+
+   /* store lodbias whether enabled or not -- fetch_texel currently
+    * respects it always.
+    */
+   sse_movaps( func,
+               get_temp( TEMP_R0, 3 ),
+               make_xmm( 3 ) );
+
+   
+   if (projected) {
+      FETCH( func, *inst, 3, 0, 3 );
+
+      emit_rcp( func, 3, 3 );
+   }
+
+   for (i = 0; i < count; i++) {
+      FETCH( func, *inst, i, 0, i );
+
+      if (projected) {
+         sse_mulps(
+            func,
+            make_xmm( i ),
+            make_xmm( 3 ) );
+      }
+      
+      /* Store in the argument buffer:
+       */
+      sse_movaps(
+         func,
+         get_temp( TEMP_R0, i ),
+         make_xmm( i ) );
+   }
+
+   args[0] = get_temp( TEMP_R0, 0 );
+   args[1] = get_sampler_ptr( unit );
+
+
+   emit_func_call( func,
+                   0,
+                   args,
+                   Elements(args),
+                   fetch_texel );
+
+   /* If all four channels are enabled, could use a pointer to
+    * dst[0].x instead of TEMP_R0 for store?
+    */
+   FOR_EACH_DST0_ENABLED_CHANNEL( *inst, i ) {
+
+      sse_movaps(
+         func,
+         make_xmm( 0 ),
+         get_temp( TEMP_R0, i ) );
+
+      STORE( func, *inst, 0, 0, i );
+   }
+}
+
+
+static void
 emit_kil(
    struct x86_function *func,
    const struct tgsi_full_src_register *reg )
 {
    unsigned uniquemask;
-   unsigned registers[4];
-   unsigned nextregister = 0;
-   unsigned firstchan = ~0;
+   unsigned unique_count = 0;
    unsigned chan_index;
+   unsigned i;
 
    /* This mask stores component bits that were already tested. Note that
     * we test if the value is less than zero, so 1.0 and 0.0 need not to be
@@ -1331,18 +1583,11 @@ emit_kil(
          uniquemask |= 1 << swizzle;
 
          /* allocate register */
-         registers[chan_index] = nextregister;
          emit_fetch(
             func,
-            nextregister,
+            unique_count++,
             reg,
             chan_index );
-         nextregister++;
-
-         /* mark the first channel used */
-         if( firstchan == ~0 ) {
-            firstchan = chan_index;
-         }
       }
    }
 
@@ -1353,32 +1598,32 @@ emit_kil(
       func,
       x86_make_reg( file_REG32, reg_DX ) );
 
-   FOR_EACH_CHANNEL( chan_index ) {
-      if( uniquemask & (1 << chan_index) ) {
-         sse_cmpps(
+   for (i = 0 ; i < unique_count; i++ ) {
+      struct x86_reg dataXMM = make_xmm(i);
+
+      sse_cmpps(
+         func,
+         dataXMM,
+         get_temp(
+            TGSI_EXEC_TEMP_00000000_I,
+            TGSI_EXEC_TEMP_00000000_C ),
+         cc_LessThan );
+      
+      if( i == 0 ) {
+         sse_movmskps(
             func,
-            make_xmm( registers[chan_index] ),
-            get_temp(
-               TGSI_EXEC_TEMP_00000000_I,
-               TGSI_EXEC_TEMP_00000000_C ),
-            cc_LessThan );
-
-         if( chan_index == firstchan ) {
-            sse_pmovmskb(
-               func,
-               x86_make_reg( file_REG32, reg_AX ),
-               make_xmm( registers[chan_index] ) );
-         }
-         else {
-            sse_pmovmskb(
-               func,
-               x86_make_reg( file_REG32, reg_DX ),
-               make_xmm( registers[chan_index] ) );
-            x86_or(
-               func,
-               x86_make_reg( file_REG32, reg_AX ),
-               x86_make_reg( file_REG32, reg_DX ) );
-         }
+            x86_make_reg( file_REG32, reg_AX ),
+            dataXMM );
+      }
+      else {
+         sse_movmskps(
+            func,
+            x86_make_reg( file_REG32, reg_DX ),
+            dataXMM );
+         x86_or(
+            func,
+            x86_make_reg( file_REG32, reg_AX ),
+            x86_make_reg( file_REG32, reg_DX ) );
       }
    }
 
@@ -1581,7 +1826,7 @@ emit_instruction(
                get_temp(
                   TGSI_EXEC_TEMP_MINUS_128_I,
                   TGSI_EXEC_TEMP_MINUS_128_C ) );
-            emit_pow( func, 3, 1, 2 );
+            emit_pow( func, 3, 1, 1, 2 );
             FETCH( func, *inst, 0, 0, CHAN_X );
             sse_xorps(
                func,
@@ -1828,8 +2073,7 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LERP:
-   /* TGSI_OPCODE_LRP */
+   case TGSI_OPCODE_LRP:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 0, 0, chan_index );
          FETCH( func, *inst, 1, 1, chan_index );
@@ -1845,12 +2089,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_CND0:
-      return 0;
-      break;
-
-   case TGSI_OPCODE_DOT2ADD:
-   /* TGSI_OPCODE_DP2A */
+   case TGSI_OPCODE_DP2A:
       FETCH( func, *inst, 0, 0, CHAN_X );  /* xmm0 = src[0].x */
       FETCH( func, *inst, 1, 1, CHAN_X );  /* xmm1 = src[1].x */
       emit_mul( func, 0, 1 );              /* xmm0 = xmm0 * xmm1 */
@@ -1865,16 +2104,7 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_INDEX:
-      return 0;
-      break;
-
-   case TGSI_OPCODE_NEGATE:
-      return 0;
-      break;
-
-   case TGSI_OPCODE_FRAC:
-   /* TGSI_OPCODE_FRC */
+   case TGSI_OPCODE_FRC:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 0, 0, chan_index );
          emit_frc( func, 0, 0 );
@@ -1886,8 +2116,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_FLOOR:
-   /* TGSI_OPCODE_FLR */
+   case TGSI_OPCODE_FLR:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 0, 0, chan_index );
          emit_flr( func, 0, 0 );
@@ -1903,8 +2132,7 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_EXPBASE2:
-   /* TGSI_OPCODE_EX2 */
+   case TGSI_OPCODE_EX2:
       FETCH( func, *inst, 0, 0, CHAN_X );
       emit_ex2( func, 0, 0 );
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
@@ -1912,8 +2140,7 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LOGBASE2:
-   /* TGSI_OPCODE_LG2 */
+   case TGSI_OPCODE_LG2:
       FETCH( func, *inst, 0, 0, CHAN_X );
       emit_lg2( func, 0, 0 );
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
@@ -1921,18 +2148,16 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_POWER:
-   /* TGSI_OPCODE_POW */
+   case TGSI_OPCODE_POW:
       FETCH( func, *inst, 0, 0, CHAN_X );
       FETCH( func, *inst, 1, 1, CHAN_X );
-      emit_pow( func, 0, 0, 1 );
+      emit_pow( func, 0, 0, 0, 1 );
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          STORE( func, *inst, 0, 0, chan_index );
       }
       break;
 
-   case TGSI_OPCODE_CROSSPRODUCT:
-   /* TGSI_OPCODE_XPD */
+   case TGSI_OPCODE_XPD:
       if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
           IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
          FETCH( func, *inst, 1, 1, CHAN_Z );
@@ -1978,10 +2203,6 @@ emit_instruction(
       }
       break;
 
-   case TGSI_OPCODE_MULTIPLYMATRIX:
-      return 0;
-      break;
-
    case TGSI_OPCODE_ABS:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 0, 0, chan_index );
@@ -2094,21 +2315,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TEX:
-      if (0) {
-	 /* Disable dummy texture code: 
-	  */
-	 emit_tempf(
-	    func,
-	    0,
-	    TEMP_ONE_I,
-	    TEMP_ONE_C );
-	 FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
-	    STORE( func, *inst, 0, 0, chan_index );
-	 }
-      }
-      else {
-	 return 0;
-      }
+      emit_tex( func, inst, FALSE, FALSE );
       break;
 
    case TGSI_OPCODE_TXD:
@@ -2206,7 +2413,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXB:
-      return 0;
+      emit_tex( func, inst, TRUE, FALSE );
       break;
 
    case TGSI_OPCODE_NRM:
@@ -2314,9 +2521,13 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXL:
-      return 0;
+      emit_tex( func, inst, TRUE, FALSE );
       break;
 
+   case TGSI_OPCODE_TXP:
+      emit_tex( func, inst, FALSE, TRUE );
+      break;
+      
    case TGSI_OPCODE_BRK:
       return 0;
       break;
@@ -2325,7 +2536,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
       return 0;
       break;
 
@@ -2341,7 +2552,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_ENDLOOP:
+   case TGSI_OPCODE_ENDFOR:
       return 0;
       break;
 
@@ -2496,7 +2707,7 @@ emit_declaration(
 
 static void aos_to_soa( struct x86_function *func, 
                         uint arg_aos,
-                        uint arg_soa, 
+                        uint arg_machine, 
                         uint arg_num, 
                         uint arg_stride )
 {
@@ -2511,7 +2722,10 @@ static void aos_to_soa( struct x86_function *func,
    x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
 
    x86_mov( func, aos_input,  x86_fn_arg( func, arg_aos ) );
-   x86_mov( func, soa_input,  x86_fn_arg( func, arg_soa ) );
+   x86_mov( func, soa_input,  x86_fn_arg( func, arg_machine ) );
+   x86_lea( func, soa_input,  
+	    x86_make_disp( soa_input, 
+			   Offset(struct tgsi_exec_machine, Inputs) ) );
    x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) );
    x86_mov( func, stride,     x86_fn_arg( func, arg_stride ) );
 
@@ -2553,28 +2767,30 @@ static void aos_to_soa( struct x86_function *func,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, aos_input );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 }
 
-static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride )
+static void soa_to_aos( struct x86_function *func, 
+			uint arg_aos, 
+			uint arg_machine, 
+			uint arg_num, 
+			uint arg_stride )
 {
-   struct x86_reg soa_output;
-   struct x86_reg aos_output;
-   struct x86_reg num_outputs;
-   struct x86_reg temp;
+   struct x86_reg soa_output = x86_make_reg( file_REG32, reg_AX );
+   struct x86_reg aos_output = x86_make_reg( file_REG32, reg_BX );
+   struct x86_reg num_outputs = x86_make_reg( file_REG32, reg_CX );
+   struct x86_reg temp = x86_make_reg( file_REG32, reg_DX );
    int inner_loop;
 
-   soa_output = x86_make_reg( file_REG32, reg_AX );
-   aos_output = x86_make_reg( file_REG32, reg_BX );
-   num_outputs = x86_make_reg( file_REG32, reg_CX );
-   temp = x86_make_reg( file_REG32, reg_DX );
-
    /* Save EBX */
-   x86_push( func, aos_output );
+   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
 
-   x86_mov( func, soa_output, x86_fn_arg( func, soa ) );
-   x86_mov( func, aos_output, x86_fn_arg( func, aos ) );
-   x86_mov( func, num_outputs, x86_fn_arg( func, num ) );
+   x86_mov( func, aos_output, x86_fn_arg( func, arg_aos ) );
+   x86_mov( func, soa_output, x86_fn_arg( func, arg_machine ) );
+   x86_lea( func, soa_output, 
+	    x86_make_disp( soa_output, 
+			   Offset(struct tgsi_exec_machine, Outputs) ) );
+   x86_mov( func, num_outputs, x86_fn_arg( func, arg_num ) );
 
    /* do */
    inner_loop = x86_get_label( func );
@@ -2591,7 +2807,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
       sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) );
       sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) );
 
-      x86_mov( func, temp, x86_fn_arg( func, stride ) );
+      x86_mov( func, temp, x86_fn_arg( func, arg_stride ) );
       x86_push( func, aos_output );
       sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) );
       sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) );
@@ -2615,20 +2831,13 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, aos_output );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 }
 
 /**
  * Translate a TGSI vertex/fragment shader to SSE2 code.
  * Slightly different things are done for vertex vs. fragment shaders.
  *
- * Note that fragment shaders are responsible for interpolating shader
- * inputs. Because on x86 we have only 4 GP registers, and here we
- * have 5 shader arguments (input, output, const, temp and coef), the
- * code is split into two phases -- DECLARATION and INSTRUCTION phase.
- * GP register holding the output argument is aliased with the coeff
- * argument, as outputs are not needed in the DECLARATION phase.
- *
  * \param tokens  the TGSI input shader
  * \param func  the output SSE code/function
  * \param immediates  buffer to place immediates, later passed to SSE func
@@ -2642,7 +2851,6 @@ tgsi_emit_sse2(
    boolean do_swizzles )
 {
    struct tgsi_parse_context parse;
-   boolean instruction_phase = FALSE;
    unsigned ok = 1;
    uint num_immediates = 0;
 
@@ -2654,74 +2862,48 @@ tgsi_emit_sse2(
 
    /* Can't just use EDI, EBX without save/restoring them:
     */
-   x86_push(
-      func,
-      get_immediate_base() );
-
-   x86_push(
-      func,
-      get_temp_base() );
-
+   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
+   x86_push( func, x86_make_reg( file_REG32, reg_DI ) );
 
    /*
     * Different function args for vertex/fragment shaders:
     */
-   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
-      /* DECLARATION phase, do not load output argument. */
-      x86_mov(
-         func,
-         get_input_base(),
-         x86_fn_arg( func, 1 ) );
-      /* skipping outputs argument here */
-      x86_mov(
-         func,
-         get_const_base(),
-         x86_fn_arg( func, 3 ) );
-      x86_mov(
-         func,
-         get_temp_base(),
-         x86_fn_arg( func, 4 ) );
-      x86_mov(
-         func,
-         get_coef_base(),
-         x86_fn_arg( func, 5 ) );
-      x86_mov(
-         func,
-         get_immediate_base(),
-         x86_fn_arg( func, 6 ) );
-   }
-   else {
-      assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX);
-
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
       if (do_swizzles)
          aos_to_soa( func, 
-                     6,         /* aos_input */
-                     1,         /* machine->input */
-                     7,         /* num_inputs */
-                     8 );       /* input_stride */
+                     4,         /* aos_input */
+                     1,         /* machine */
+                     5,         /* num_inputs */
+                     6 );       /* input_stride */
+   }
 
+   x86_mov(
+      func,
+      get_machine_base(),
+      x86_fn_arg( func, 1 ) );
+   x86_mov(
+      func,
+      get_const_base(),
+      x86_fn_arg( func, 2 ) );
+   x86_mov(
+      func,
+      get_immediate_base(),
+      x86_fn_arg( func, 3 ) );
+
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
       x86_mov(
-         func,
-         get_input_base(),
-         x86_fn_arg( func, 1 ) );
-      x86_mov(
-         func,
-         get_output_base(),
-         x86_fn_arg( func, 2 ) );
-      x86_mov(
-         func,
-         get_const_base(),
-         x86_fn_arg( func, 3 ) );
-      x86_mov(
-         func,
-         get_temp_base(),
-         x86_fn_arg( func, 4 ) );
-      x86_mov(
-         func,
-         get_immediate_base(),
-         x86_fn_arg( func, 5 ) );
+	 func,
+	 get_coef_base(),
+	 x86_fn_arg( func, 4 ) );
    }
 
+   x86_mov(
+      func,
+      get_sampler_base(),
+      x86_make_disp( get_machine_base(),
+                     Offset( struct tgsi_exec_machine, Samplers ) ) );
+
+
    while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
       tgsi_parse_token( &parse );
 
@@ -2735,24 +2917,15 @@ tgsi_emit_sse2(
          break;
 
       case TGSI_TOKEN_TYPE_INSTRUCTION:
-         if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
-            if( !instruction_phase ) {
-               /* INSTRUCTION phase, overwrite coeff with output. */
-               instruction_phase = TRUE;
-               x86_mov(
-                  func,
-                  get_output_base(),
-                  x86_fn_arg( func, 2 ) );
-            }
-         }
-
          ok = emit_instruction(
             func,
             &parse.FullToken.FullInstruction );
 
 	 if (!ok) {
-	    debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", 
-			 parse.FullToken.FullInstruction.Instruction.Opcode,
+            uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode;
+	    debug_printf("failed to translate tgsi opcode %d (%s) to SSE (%s)\n", 
+			 opcode,
+                         tgsi_get_opcode_name(opcode),
                          parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ?
                          "vertex shader" : "fragment shader");
 	 }
@@ -2767,7 +2940,7 @@ tgsi_emit_sse2(
             assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES);
             for( i = 0; i < size; i++ ) {
                immediates[num_immediates][i] =
-		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+		  parse.FullToken.FullImmediate.u[i].Float;
             }
 #if 0
             debug_printf("SSE FS immediate[%d] = %f %f %f %f\n",
@@ -2789,18 +2962,17 @@ tgsi_emit_sse2(
 
    if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
       if (do_swizzles)
-         soa_to_aos( func, 9, 2, 10, 11 );
+         soa_to_aos( func, 
+		     7, 	/* aos_output */
+		     1, 	/* machine */
+		     8, 	/* num_outputs */
+		     9 );	/* output_stride */
    }
 
    /* Can't just use EBX, EDI without save/restoring them:
     */
-   x86_pop(
-      func,
-      get_temp_base() );
-
-   x86_pop(
-      func,
-      get_immediate_base() );
+   x86_pop( func, x86_make_reg( file_REG32, reg_DI ) );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 
    emit_ret( func );
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
index af838b2a25..d81ee3d00e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
@@ -34,6 +34,7 @@ extern "C" {
 
 struct tgsi_token;
 struct x86_function;
+struct tgsi_interp_coef;
 
 unsigned
 tgsi_emit_sse2(
@@ -42,6 +43,33 @@ tgsi_emit_sse2(
    float (*immediates)[4],
    boolean do_swizzles );
 
+
+/* This is the function prototype generated when do_swizzles is false
+ * -- effectively for fragment shaders.
+ */
+typedef void (PIPE_CDECL *tgsi_sse2_fs_function) (
+   struct tgsi_exec_machine *machine, /* 1 */
+   const float (*constant)[4],		    /* 2 */
+   const float (*immediate)[4],		    /* 3 */
+   const struct tgsi_interp_coef *coef	    /* 4 */
+   );
+
+
+/* This is the function prototype generated when do_swizzles is true
+ * -- effectively for vertex shaders.
+ */
+typedef void (PIPE_CDECL *tgsi_sse2_vs_func) (
+   struct tgsi_exec_machine *machine, /* 1 */
+   const float (*constant)[4],        /* 2 */
+   const float (*immediate)[4],       /* 3 */
+   const float (*aos_input)[4], /* 4 */
+   uint num_inputs,             /* 5 */
+   uint input_stride,           /* 6 */
+   float (*aos_output)[4],      /* 7 */
+   uint num_outputs,            /* 8 */
+   uint output_stride );        /* 9 */
+
+
 #if defined __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index a76bbc9140..d438450b1e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -231,7 +231,8 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "TEMP",
    "SAMP",
    "ADDR",
-   "IMM"
+   "IMM",
+   "LOOP"
 };
 
 static boolean
@@ -789,16 +790,6 @@ match_inst_mnemonic(const char **pcur,
    if (str_match_no_case(pcur, info->mnemonic)) {
       return TRUE;
    }
-   if (info->alt_mnemonic1) {
-      if (str_match_no_case(pcur, info->alt_mnemonic1)) {
-         return TRUE;
-      }
-      if (info->alt_mnemonic2) {
-         if (str_match_no_case(pcur, info->alt_mnemonic2)) {
-            return TRUE;
-         }
-      }
-   }
    return FALSE;
 }
 
@@ -1091,7 +1082,10 @@ static boolean parse_immediate( struct translate_ctx *ctx )
    imm = tgsi_default_full_immediate();
    imm.Immediate.NrTokens += 4;
    imm.Immediate.DataType = TGSI_IMM_FLOAT32;
-   imm.u.Pointer = values;
+   imm.u[0].Float = values[0];
+   imm.u[1].Float = values[1];
+   imm.u[2].Float = values[2];
+   imm.u[3].Float = values[3];
 
    advance = tgsi_build_full_immediate(
       &imm,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
new file mode 100644
index 0000000000..f7096bd8e2
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
@@ -0,0 +1,938 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_ureg.h"
+#include "tgsi/tgsi_info.h"
+#include "tgsi/tgsi_dump.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+union tgsi_any_token {
+   struct tgsi_version version;
+   struct tgsi_header header;
+   struct tgsi_processor processor;
+   struct tgsi_token token;
+   struct tgsi_declaration decl;
+   struct tgsi_declaration_range decl_range;
+   struct tgsi_declaration_semantic decl_semantic;
+   struct tgsi_immediate imm;
+   union  tgsi_immediate_data imm_data;
+   struct tgsi_instruction insn;
+   struct tgsi_instruction_ext_nv insn_ext_nv;
+   struct tgsi_instruction_ext_label insn_ext_label;
+   struct tgsi_instruction_ext_texture insn_ext_texture;
+   struct tgsi_instruction_ext_predicate insn_ext_predicate;
+   struct tgsi_src_register src;
+   struct tgsi_src_register_ext_swz src_ext_swz;
+   struct tgsi_src_register_ext_mod src_ext_mod;
+   struct tgsi_dimension dim;
+   struct tgsi_dst_register dst;
+   struct tgsi_dst_register_ext_concode dst_ext_code;
+   struct tgsi_dst_register_ext_modulate dst_ext_mod;
+   struct tgsi_dst_register_ext_predicate dst_ext_pred;
+   unsigned value;
+};
+
+
+struct ureg_tokens {
+   union tgsi_any_token *tokens;
+   unsigned size;
+   unsigned order;
+   unsigned count;
+};
+
+#define UREG_MAX_INPUT PIPE_MAX_ATTRIBS
+#define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS
+#define UREG_MAX_IMMEDIATE 32
+#define UREG_MAX_TEMP 256
+#define UREG_MAX_ADDR 2
+
+#define DOMAIN_DECL 0
+#define DOMAIN_INSN 1
+
+struct ureg_program
+{
+   unsigned processor;
+   struct pipe_context *pipe;
+
+   struct {
+      unsigned semantic_name;
+      unsigned semantic_index;
+      unsigned interp;
+   } input[UREG_MAX_INPUT];
+   unsigned nr_inputs;
+
+   struct {
+      unsigned semantic_name;
+      unsigned semantic_index;
+   } output[UREG_MAX_OUTPUT];
+   unsigned nr_outputs;
+
+   struct {
+      float v[4];
+      unsigned nr;
+   } immediate[UREG_MAX_IMMEDIATE];
+   unsigned nr_immediates;
+
+   struct ureg_src sampler[PIPE_MAX_SAMPLERS];
+   unsigned nr_samplers;
+
+   unsigned temps_active[UREG_MAX_TEMP / 32];
+   unsigned nr_temps;
+
+   unsigned nr_addrs;
+
+   unsigned nr_constants;
+   unsigned nr_instructions;
+
+   struct ureg_tokens domain[2];
+};
+
+static union tgsi_any_token error_tokens[32];
+
+static void tokens_error( struct ureg_tokens *tokens )
+{
+   tokens->tokens = error_tokens;
+   tokens->size = Elements(error_tokens);
+   tokens->count = 0;
+}
+
+
+static void tokens_expand( struct ureg_tokens *tokens,
+                           unsigned count )
+{
+   unsigned old_size = tokens->size * sizeof(unsigned);
+
+   if (tokens->tokens == error_tokens)
+      goto fail;
+
+   while (tokens->count + count > tokens->size) {
+      tokens->size = (1 << ++tokens->order);
+   }
+
+   tokens->tokens = REALLOC(tokens->tokens, 
+                            old_size,
+                            tokens->size * sizeof(unsigned));
+   if (tokens->tokens == NULL) 
+      goto fail;
+
+   return;
+          
+fail:
+   tokens_error(tokens);
+}
+
+static void set_bad( struct ureg_program *ureg )
+{
+   tokens_error(&ureg->domain[0]);
+}
+
+
+
+static union tgsi_any_token *get_tokens( struct ureg_program *ureg,
+                                         unsigned domain,
+                                         unsigned count )
+{
+   struct ureg_tokens *tokens = &ureg->domain[domain];
+   union tgsi_any_token *result;
+
+   if (tokens->count + count > tokens->size) 
+      tokens_expand(tokens, count);
+
+   result = &tokens->tokens[tokens->count];
+   tokens->count += count;
+   return result;
+}
+
+
+static union tgsi_any_token *retrieve_token( struct ureg_program *ureg,
+                                            unsigned domain,
+                                            unsigned nr )
+{
+   if (ureg->domain[domain].tokens == error_tokens)
+      return &error_tokens[0];
+
+   return &ureg->domain[domain].tokens[nr];
+}
+
+
+
+static INLINE struct ureg_dst
+ureg_dst_register( unsigned file,
+                   unsigned index )
+{
+   struct ureg_dst dst;
+
+   dst.File      = file;
+   dst.WriteMask = TGSI_WRITEMASK_XYZW;
+   dst.Indirect  = 0;
+   dst.IndirectIndex = 0;
+   dst.IndirectSwizzle = 0;
+   dst.Saturate  = 0;
+   dst.Index     = index;
+   dst.Pad1      = 0;
+   dst.Pad2      = 0;
+
+   return dst;
+}
+
+static INLINE struct ureg_src 
+ureg_src_register( unsigned file,
+                   unsigned index )
+{
+   struct ureg_src src;
+
+   src.File     = file;
+   src.SwizzleX = TGSI_SWIZZLE_X;
+   src.SwizzleY = TGSI_SWIZZLE_Y;
+   src.SwizzleZ = TGSI_SWIZZLE_Z;
+   src.SwizzleW = TGSI_SWIZZLE_W;
+   src.Pad      = 0;
+   src.Indirect = 0;
+   src.IndirectIndex = 0;
+   src.IndirectSwizzle = 0;
+   src.Absolute = 0;
+   src.Index    = index;
+   src.Negate   = 0;
+
+   return src;
+}
+
+
+
+
+static struct ureg_src 
+ureg_DECL_input( struct ureg_program *ureg,
+                 unsigned name,
+                 unsigned index,
+                 unsigned interp_mode )
+{
+   unsigned i;
+
+   for (i = 0; i < ureg->nr_inputs; i++) {
+      if (ureg->input[i].semantic_name == name &&
+          ureg->input[i].semantic_index == index) 
+         goto out;
+   }
+
+   if (ureg->nr_inputs < UREG_MAX_INPUT) {
+      ureg->input[i].semantic_name = name;
+      ureg->input[i].semantic_index = index;
+      ureg->input[i].interp = interp_mode;
+      ureg->nr_inputs++;
+   }
+   else {
+      set_bad( ureg );
+   }
+
+out:
+   return ureg_src_register( TGSI_FILE_INPUT, i );
+}
+
+
+
+struct ureg_src 
+ureg_DECL_fs_input( struct ureg_program *ureg,
+                    unsigned name,
+                    unsigned index,
+                    unsigned interp )
+{
+   assert(ureg->processor == TGSI_PROCESSOR_FRAGMENT);
+   return ureg_DECL_input( ureg, name, index, interp );
+}
+
+
+struct ureg_src 
+ureg_DECL_vs_input( struct ureg_program *ureg,
+                    unsigned name,
+                    unsigned index )
+{
+   assert(ureg->processor == TGSI_PROCESSOR_VERTEX);
+   return ureg_DECL_input( ureg, name, index, TGSI_INTERPOLATE_CONSTANT );
+}
+
+
+struct ureg_dst 
+ureg_DECL_output( struct ureg_program *ureg,
+                  unsigned name,
+                  unsigned index )
+{
+   unsigned i;
+
+   for (i = 0; i < ureg->nr_outputs; i++) {
+      if (ureg->output[i].semantic_name == name &&
+          ureg->output[i].semantic_index == index) 
+         goto out;
+   }
+
+   if (ureg->nr_outputs < UREG_MAX_OUTPUT) {
+      ureg->output[i].semantic_name = name;
+      ureg->output[i].semantic_index = index;
+      ureg->nr_outputs++;
+   }
+   else {
+      set_bad( ureg );
+   }
+
+out:
+   return ureg_dst_register( TGSI_FILE_OUTPUT, i );
+}
+
+
+/* Returns a new constant register.  Keep track of which have been
+ * referred to so that we can emit decls later.
+ *
+ * There is nothing in this code to bind this constant to any tracked
+ * value or manage any constant_buffer contents -- that's the
+ * resposibility of the calling code.
+ */
+struct ureg_src ureg_DECL_constant(struct ureg_program *ureg )
+{
+   return ureg_src_register( TGSI_FILE_CONSTANT, ureg->nr_constants++ );
+}
+
+
+/* Allocate a new temporary.  Temporaries greater than UREG_MAX_TEMP
+ * are legal, but will not be released.
+ */
+struct ureg_dst ureg_DECL_temporary( struct ureg_program *ureg )
+{
+   unsigned i;
+
+   for (i = 0; i < UREG_MAX_TEMP; i += 32) {
+      int bit = ffs(~ureg->temps_active[i/32]);
+      if (bit != 0) {
+         i += bit - 1;
+         goto out;
+      }
+   }
+
+   /* No reusable temps, so allocate a new one:
+    */
+   i = ureg->nr_temps++;
+
+out:
+   if (i < UREG_MAX_TEMP)
+      ureg->temps_active[i/32] |= 1 << (i % 32);
+
+   if (i >= ureg->nr_temps)
+      ureg->nr_temps = i + 1;
+
+   return ureg_dst_register( TGSI_FILE_TEMPORARY, i );
+}
+
+
+void ureg_release_temporary( struct ureg_program *ureg,
+                             struct ureg_dst tmp )
+{
+   if(tmp.File == TGSI_FILE_TEMPORARY)
+      if (tmp.Index < UREG_MAX_TEMP)
+         ureg->temps_active[tmp.Index/32] &= ~(1 << (tmp.Index % 32));
+}
+
+
+/* Allocate a new address register.
+ */
+struct ureg_dst ureg_DECL_address( struct ureg_program *ureg )
+{
+   if (ureg->nr_addrs < UREG_MAX_ADDR)
+      return ureg_dst_register( TGSI_FILE_ADDRESS, ureg->nr_addrs++ );
+
+   assert( 0 );
+   return ureg_dst_register( TGSI_FILE_ADDRESS, 0 );
+}
+
+/* Allocate a new sampler.
+ */
+struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg,
+                                   unsigned nr )
+{
+   unsigned i;
+
+   for (i = 0; i < ureg->nr_samplers; i++)
+      if (ureg->sampler[i].Index == nr)
+         return ureg->sampler[i];
+   
+   if (i < PIPE_MAX_SAMPLERS) {
+      ureg->sampler[i] = ureg_src_register( TGSI_FILE_SAMPLER, nr );
+      ureg->nr_samplers++;
+      return ureg->sampler[i];
+   }
+
+   assert( 0 );
+   return ureg->sampler[0];
+}
+
+
+
+
+static int match_or_expand_immediate( const float *v,
+                                      unsigned nr,
+                                      float *v2,
+                                      unsigned *nr2,
+                                      unsigned *swizzle )
+{
+   unsigned i, j;
+   
+   *swizzle = 0;
+
+   for (i = 0; i < nr; i++) {
+      boolean found = FALSE;
+
+      for (j = 0; j < *nr2 && !found; j++) {
+         if (v[i] == v2[j]) {
+            *swizzle |= j << (i * 2);
+            found = TRUE;
+         }
+      }
+
+      if (!found) {
+         if (*nr2 >= 4) 
+            return FALSE;
+
+         v2[*nr2] = v[i];
+         *swizzle |= *nr2 << (i * 2);
+         (*nr2)++;
+      }
+   }
+
+   return TRUE;
+}
+
+
+
+
+struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, 
+                                     const float *v,
+                                     unsigned nr )
+{
+   unsigned i, j;
+   unsigned swizzle;
+
+   /* Could do a first pass where we examine all existing immediates
+    * without expanding.
+    */
+
+   for (i = 0; i < ureg->nr_immediates; i++) {
+      if (match_or_expand_immediate( v, 
+                                     nr,
+                                     ureg->immediate[i].v,
+                                     &ureg->immediate[i].nr, 
+                                     &swizzle ))
+         goto out;
+   }
+
+   if (ureg->nr_immediates < UREG_MAX_IMMEDIATE) {
+      i = ureg->nr_immediates++;
+      if (match_or_expand_immediate( v,
+                                     nr,
+                                     ureg->immediate[i].v,
+                                     &ureg->immediate[i].nr, 
+                                     &swizzle ))
+         goto out;
+   }
+
+   set_bad( ureg );
+
+out:
+   /* Make sure that all referenced elements are from this immediate.
+    * Has the effect of making size-one immediates into scalars.
+    */
+   for (j = nr; j < 4; j++)
+      swizzle |= (swizzle & 0x3) << (j * 2);
+
+   return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ),
+                        (swizzle >> 0) & 0x3,
+                        (swizzle >> 2) & 0x3,
+                        (swizzle >> 4) & 0x3,
+                        (swizzle >> 6) & 0x3);
+}
+
+
+void 
+ureg_emit_src( struct ureg_program *ureg,
+               struct ureg_src src )
+{
+   unsigned size = (1 + 
+                    (src.Absolute ? 1 : 0) +
+                    (src.Indirect ? 1 : 0));
+
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size );
+   unsigned n = 0;
+
+   assert(src.File != TGSI_FILE_NULL);
+   assert(src.File != TGSI_FILE_OUTPUT);
+   assert(src.File < TGSI_FILE_COUNT);
+   
+   out[n].value = 0;
+   out[n].src.File = src.File;
+   out[n].src.SwizzleX = src.SwizzleX;
+   out[n].src.SwizzleY = src.SwizzleY;
+   out[n].src.SwizzleZ = src.SwizzleZ;
+   out[n].src.SwizzleW = src.SwizzleW;
+   out[n].src.Index = src.Index;
+   out[n].src.Negate = src.Negate;
+   n++;
+   
+   if (src.Absolute) {
+      out[0].src.Extended = 1;
+      out[0].src.Negate = 0;
+      out[n].value = 0;
+      out[n].src_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD;
+      out[n].src_ext_mod.Absolute = 1;
+      out[n].src_ext_mod.Negate = src.Negate;
+      n++;
+   }
+
+   if (src.Indirect) {
+      out[0].src.Indirect = 1;
+      out[n].value = 0;
+      out[n].src.File = TGSI_FILE_ADDRESS;
+      out[n].src.SwizzleX = src.IndirectSwizzle;
+      out[n].src.SwizzleY = src.IndirectSwizzle;
+      out[n].src.SwizzleZ = src.IndirectSwizzle;
+      out[n].src.SwizzleW = src.IndirectSwizzle;
+      out[n].src.Index = src.IndirectIndex;
+      n++;
+   }
+
+   assert(n == size);
+}
+
+
+void 
+ureg_emit_dst( struct ureg_program *ureg,
+               struct ureg_dst dst )
+{
+   unsigned size = (1 + 
+                    (dst.Indirect ? 1 : 0));
+
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size );
+   unsigned n = 0;
+
+   assert(dst.File != TGSI_FILE_NULL);
+   assert(dst.File != TGSI_FILE_CONSTANT);
+   assert(dst.File != TGSI_FILE_INPUT);
+   assert(dst.File != TGSI_FILE_SAMPLER);
+   assert(dst.File != TGSI_FILE_IMMEDIATE);
+   assert(dst.File < TGSI_FILE_COUNT);
+   
+   out[n].value = 0;
+   out[n].dst.File = dst.File;
+   out[n].dst.WriteMask = dst.WriteMask;
+   out[n].dst.Indirect = dst.Indirect;
+   out[n].dst.Index = dst.Index;
+   n++;
+   
+   if (dst.Indirect) {
+      out[n].value = 0;
+      out[n].src.File = TGSI_FILE_ADDRESS;
+      out[n].src.SwizzleX = dst.IndirectSwizzle;
+      out[n].src.SwizzleY = dst.IndirectSwizzle;
+      out[n].src.SwizzleZ = dst.IndirectSwizzle;
+      out[n].src.SwizzleW = dst.IndirectSwizzle;
+      out[n].src.Index = dst.IndirectIndex;
+      n++;
+   }
+
+   assert(n == size);
+}
+
+
+
+unsigned
+ureg_emit_insn(struct ureg_program *ureg,
+               unsigned opcode,
+               boolean saturate,
+               unsigned num_dst,
+               unsigned num_src )
+{
+   union tgsi_any_token *out;
+
+   out = get_tokens( ureg, DOMAIN_INSN, 1 );
+   out[0].value = 0;
+   out[0].insn.Type = TGSI_TOKEN_TYPE_INSTRUCTION;
+   out[0].insn.NrTokens = 0;
+   out[0].insn.Opcode = opcode;
+   out[0].insn.Saturate = saturate;
+   out[0].insn.NumDstRegs = num_dst;
+   out[0].insn.NumSrcRegs = num_src;
+   out[0].insn.Padding = 0;
+   out[0].insn.Extended = 0;
+   
+   ureg->nr_instructions++;
+   
+   return ureg->domain[DOMAIN_INSN].count - 1;
+}
+
+
+void
+ureg_emit_label(struct ureg_program *ureg,
+                unsigned insn_token,
+                unsigned *label_token )
+{
+   union tgsi_any_token *out, *insn;
+
+   if(!label_token)
+      return;
+
+   out = get_tokens( ureg, DOMAIN_INSN, 1 );
+   insn = retrieve_token( ureg, DOMAIN_INSN, insn_token );
+
+   insn->insn.Extended = 1;
+
+   out[0].value = 0;
+   out[0].insn_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL;
+   
+   *label_token = ureg->domain[DOMAIN_INSN].count - 1;
+}
+
+/* Will return a number which can be used in a label to point to the
+ * next instruction to be emitted.
+ */
+unsigned
+ureg_get_instruction_number( struct ureg_program *ureg )
+{
+   return ureg->nr_instructions;
+}
+
+/* Patch a given label (expressed as a token number) to point to a
+ * given instruction (expressed as an instruction number).
+ */
+void
+ureg_fixup_label(struct ureg_program *ureg,
+                 unsigned label_token,
+                 unsigned instruction_number )
+{
+   union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, label_token );
+
+   assert(out->insn_ext_label.Type == TGSI_INSTRUCTION_EXT_TYPE_LABEL);
+   out->insn_ext_label.Label = instruction_number;
+}
+
+
+void
+ureg_emit_texture(struct ureg_program *ureg,
+                  unsigned insn_token,
+                  unsigned target )
+{
+   union tgsi_any_token *out, *insn;
+
+   out = get_tokens( ureg, DOMAIN_INSN, 1 );
+   insn = retrieve_token( ureg, DOMAIN_INSN, insn_token );
+
+   insn->insn.Extended = 1;
+
+   out[0].value = 0;
+   out[0].insn_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE;
+   out[0].insn_ext_texture.Texture = target;
+}
+
+
+void
+ureg_fixup_insn_size(struct ureg_program *ureg,
+                     unsigned insn )
+{
+   union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, insn );
+
+   assert(out->insn.Type == TGSI_TOKEN_TYPE_INSTRUCTION);
+   out->insn.NrTokens = ureg->domain[DOMAIN_INSN].count - insn - 1;
+}
+
+
+void
+ureg_insn(struct ureg_program *ureg,
+          unsigned opcode,
+          const struct ureg_dst *dst,
+          unsigned nr_dst,
+          const struct ureg_src *src,
+          unsigned nr_src )
+{
+   unsigned insn, i;
+   boolean saturate;
+
+#ifdef DEBUG
+   {
+      const struct tgsi_opcode_info *info = tgsi_get_opcode_info( opcode );
+      assert(info);
+      if(info) {
+         assert(nr_dst == info->num_dst);
+         assert(nr_src == info->num_src);
+      }
+   }
+#endif
+   
+   saturate = nr_dst ? dst[0].Saturate : FALSE;
+
+   insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src );
+
+   for (i = 0; i < nr_dst; i++)
+      ureg_emit_dst( ureg, dst[i] );
+
+   for (i = 0; i < nr_src; i++)
+      ureg_emit_src( ureg, src[i] );
+
+   ureg_fixup_insn_size( ureg, insn );
+}
+
+
+
+static void emit_decl( struct ureg_program *ureg,
+                       unsigned file,
+                       unsigned index,
+                       unsigned semantic_name,
+                       unsigned semantic_index,
+                       unsigned interp )
+{
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 );
+
+   out[0].value = 0;
+   out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+   out[0].decl.NrTokens = 3;
+   out[0].decl.File = file;
+   out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW; /* FIXME! */
+   out[0].decl.Interpolate = interp;
+   out[0].decl.Semantic = 1;
+
+   out[1].value = 0;
+   out[1].decl_range.First = 
+      out[1].decl_range.Last = index;
+
+   out[2].value = 0;
+   out[2].decl_semantic.SemanticName = semantic_name;
+   out[2].decl_semantic.SemanticIndex = semantic_index;
+
+}
+
+
+static void emit_decl_range( struct ureg_program *ureg,
+                             unsigned file,
+                             unsigned first,
+                             unsigned count )
+{
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 2 );
+
+   out[0].value = 0;
+   out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+   out[0].decl.NrTokens = 2;
+   out[0].decl.File = file;
+   out[0].decl.UsageMask = 0xf;
+   out[0].decl.Interpolate = TGSI_INTERPOLATE_CONSTANT;
+   out[0].decl.Semantic = 0;
+
+   out[1].value = 0;
+   out[1].decl_range.First = first;
+   out[1].decl_range.Last = first + count - 1;
+}
+
+static void emit_immediate( struct ureg_program *ureg,
+                            const float *v )
+{
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 5 );
+
+   out[0].value = 0;
+   out[0].imm.Type = TGSI_TOKEN_TYPE_IMMEDIATE;
+   out[0].imm.NrTokens = 5;
+   out[0].imm.DataType = TGSI_IMM_FLOAT32;
+   out[0].imm.Padding = 0;
+   out[0].imm.Extended = 0;
+
+   out[1].imm_data.Float = v[0];
+   out[2].imm_data.Float = v[1];
+   out[3].imm_data.Float = v[2];
+   out[4].imm_data.Float = v[3];
+}
+
+
+
+
+static void emit_decls( struct ureg_program *ureg )
+{
+   unsigned i;
+
+   for (i = 0; i < ureg->nr_inputs; i++) {
+      emit_decl( ureg, 
+                 TGSI_FILE_INPUT, 
+                 i,
+                 ureg->input[i].semantic_name,
+                 ureg->input[i].semantic_index,
+                 ureg->input[i].interp );
+   }
+
+   for (i = 0; i < ureg->nr_outputs; i++) {
+      emit_decl( ureg, 
+                 TGSI_FILE_OUTPUT, 
+                 i,
+                 ureg->output[i].semantic_name,
+                 ureg->output[i].semantic_index,
+                 TGSI_INTERPOLATE_CONSTANT );
+   }
+
+   for (i = 0; i < ureg->nr_samplers; i++) {
+      emit_decl_range( ureg, 
+                       TGSI_FILE_SAMPLER,
+                       ureg->sampler[i].Index, 1 );
+   }
+
+   if (ureg->nr_constants) {
+      emit_decl_range( ureg,
+                       TGSI_FILE_CONSTANT,
+                       0, ureg->nr_constants );
+   }
+
+   if (ureg->nr_temps) {
+      emit_decl_range( ureg,
+                       TGSI_FILE_TEMPORARY,
+                       0, ureg->nr_temps );
+   }
+
+   if (ureg->nr_addrs) {
+      emit_decl_range( ureg,
+                       TGSI_FILE_ADDRESS,
+                       0, ureg->nr_addrs );
+   }
+
+   for (i = 0; i < ureg->nr_immediates; i++) {
+      emit_immediate( ureg,
+                      ureg->immediate[i].v );
+   }
+}
+
+/* Append the instruction tokens onto the declarations to build a
+ * contiguous stream suitable to send to the driver.
+ */
+static void copy_instructions( struct ureg_program *ureg )
+{
+   unsigned nr_tokens = ureg->domain[DOMAIN_INSN].count;
+   union tgsi_any_token *out = get_tokens( ureg, 
+                                           DOMAIN_DECL, 
+                                           nr_tokens );
+
+   memcpy(out, 
+          ureg->domain[DOMAIN_INSN].tokens, 
+          nr_tokens * sizeof out[0] );
+}
+
+
+static void
+fixup_header_size(struct ureg_program *ureg)
+{
+   union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 1 );
+
+   out->header.BodySize = ureg->domain[DOMAIN_DECL].count - 3;
+}
+
+
+static void
+emit_header( struct ureg_program *ureg )
+{
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 );
+
+   out[0].version.MajorVersion = 1;
+   out[0].version.MinorVersion = 1;
+   out[0].version.Padding = 0;
+
+   out[1].header.HeaderSize = 2;
+   out[1].header.BodySize = 0;
+
+   out[2].processor.Processor = ureg->processor;
+   out[2].processor.Padding = 0;
+}
+
+
+const struct tgsi_token *ureg_finalize( struct ureg_program *ureg )
+{
+   const struct tgsi_token *tokens;
+
+   emit_header( ureg );
+   emit_decls( ureg );
+   copy_instructions( ureg );
+   fixup_header_size( ureg );
+   
+   if (ureg->domain[0].tokens == error_tokens ||
+       ureg->domain[1].tokens == error_tokens) {
+      debug_printf("%s: error in generated shader\n", __FUNCTION__);
+      assert(0);
+      return NULL;
+   }
+
+   tokens = &ureg->domain[DOMAIN_DECL].tokens[0].token;
+
+   if (0) {
+      debug_printf("%s: emitted shader %d tokens:\n", __FUNCTION__, 
+                   ureg->domain[DOMAIN_DECL].count);
+      tgsi_dump( tokens, 0 );
+   }
+   
+   return tokens;
+}
+
+
+void *ureg_create_shader( struct ureg_program *ureg,
+                          struct pipe_context *pipe )
+{
+   struct pipe_shader_state state;
+
+   state.tokens = ureg_finalize(ureg);
+   if(!state.tokens)
+      return NULL;
+
+   if (ureg->processor == TGSI_PROCESSOR_VERTEX)
+      return pipe->create_vs_state( pipe, &state );
+   else
+      return pipe->create_fs_state( pipe, &state );
+}
+
+
+
+
+struct ureg_program *ureg_create( unsigned processor )
+{
+   struct ureg_program *ureg = CALLOC_STRUCT( ureg_program );
+   if (ureg == NULL)
+      return NULL;
+
+   ureg->processor = processor;
+   return ureg;
+}
+
+
+void ureg_destroy( struct ureg_program *ureg )
+{
+   unsigned i;
+
+   for (i = 0; i < Elements(ureg->domain); i++) {
+      if (ureg->domain[i].tokens && 
+          ureg->domain[i].tokens != error_tokens)
+         FREE(ureg->domain[i].tokens);
+   }
+   
+   FREE(ureg);
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
new file mode 100644
index 0000000000..acbca59040
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
@@ -0,0 +1,590 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_UREG_H
+#define TGSI_UREG_H
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_shader_tokens.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+   
+struct ureg_program;
+
+/* Almost a tgsi_src_register, but we need to pull in the Absolute
+ * flag from the _ext token.  Indirect flag always implies ADDR[0].
+ */
+struct ureg_src
+{
+   unsigned File        : 4;  /* TGSI_FILE_ */
+   unsigned SwizzleX    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleY    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleZ    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleW    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned Pad         : 1;  /* BOOL */
+   unsigned Indirect    : 1;  /* BOOL */
+   unsigned Absolute    : 1;  /* BOOL */
+   int      Index       : 16; /* SINT */
+   unsigned Negate      : 1;  /* BOOL */
+   int      IndirectIndex   : 16; /* SINT */
+   int      IndirectSwizzle : 2;  /* TGSI_SWIZZLE_ */
+};
+
+/* Very similar to a tgsi_dst_register, removing unsupported fields
+ * and adding a Saturate flag.  It's easier to push saturate into the
+ * destination register than to try and create a _SAT varient of each
+ * instruction function.
+ */
+struct ureg_dst
+{
+   unsigned File        : 4;  /* TGSI_FILE_ */
+   unsigned WriteMask   : 4;  /* TGSI_WRITEMASK_ */
+   unsigned Indirect    : 1;  /* BOOL */
+   unsigned Saturate    : 1;  /* BOOL */
+   int      Index       : 16; /* SINT */
+   unsigned Pad1        : 5;
+   unsigned Pad2        : 1;  /* BOOL */
+   int      IndirectIndex   : 16; /* SINT */
+   int      IndirectSwizzle : 2;  /* TGSI_SWIZZLE_ */
+};
+
+struct pipe_context;
+
+struct ureg_program *
+ureg_create( unsigned processor );
+
+const struct tgsi_token *
+ureg_finalize( struct ureg_program * );
+
+void *
+ureg_create_shader( struct ureg_program *,
+                    struct pipe_context *pipe );
+
+void 
+ureg_destroy( struct ureg_program * );
+
+
+/***********************************************************************
+ * Convenience routine:
+ */
+static INLINE void *
+ureg_create_shader_and_destroy( struct ureg_program *p,
+                                struct pipe_context *pipe )
+{
+   void *result = ureg_create_shader( p, pipe );
+   ureg_destroy( p );
+   return result;
+}
+
+
+
+/***********************************************************************
+ * Build shader declarations:
+ */
+
+struct ureg_src
+ureg_DECL_fs_input( struct ureg_program *,
+                    unsigned semantic_name,
+                    unsigned semantic_index,
+                    unsigned interp_mode );
+
+struct ureg_src
+ureg_DECL_vs_input( struct ureg_program *,
+                    unsigned semantic_name,
+                    unsigned semantic_index );
+
+struct ureg_dst
+ureg_DECL_output( struct ureg_program *,
+                  unsigned semantic_name,
+                  unsigned semantic_index );
+
+struct ureg_src
+ureg_DECL_immediate( struct ureg_program *,
+                     const float *v,
+                     unsigned nr );
+
+struct ureg_src
+ureg_DECL_constant( struct ureg_program * );
+
+struct ureg_dst
+ureg_DECL_temporary( struct ureg_program * );
+
+void 
+ureg_release_temporary( struct ureg_program *ureg,
+                        struct ureg_dst tmp );
+
+struct ureg_dst
+ureg_DECL_address( struct ureg_program * );
+
+/* Supply an index to the sampler declaration as this is the hook to
+ * the external pipe_sampler state.  Users of this function probably
+ * don't want just any sampler, but a specific one which they've set
+ * up state for in the context.
+ */
+struct ureg_src
+ureg_DECL_sampler( struct ureg_program *,
+                   unsigned index );
+
+
+static INLINE struct ureg_src
+ureg_imm4f( struct ureg_program *ureg,
+                       float a, float b,
+                       float c, float d)
+{
+   float v[4];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   v[3] = d;
+   return ureg_DECL_immediate( ureg, v, 4 );
+}
+
+static INLINE struct ureg_src
+ureg_imm3f( struct ureg_program *ureg,
+                       float a, float b,
+                       float c)
+{
+   float v[3];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   return ureg_DECL_immediate( ureg, v, 3 );
+}
+
+static INLINE struct ureg_src
+ureg_imm2f( struct ureg_program *ureg,
+                       float a, float b)
+{
+   float v[2];
+   v[0] = a;
+   v[1] = b;
+   return ureg_DECL_immediate( ureg, v, 2 );
+}
+
+static INLINE struct ureg_src
+ureg_imm1f( struct ureg_program *ureg,
+                       float a)
+{
+   float v[1];
+   v[0] = a;
+   return ureg_DECL_immediate( ureg, v, 1 );
+}
+
+/***********************************************************************
+ * Functions for patching up labels
+ */
+
+
+/* Will return a number which can be used in a label to point to the
+ * next instruction to be emitted.
+ */
+unsigned
+ureg_get_instruction_number( struct ureg_program *ureg );
+
+
+/* Patch a given label (expressed as a token number) to point to a
+ * given instruction (expressed as an instruction number).
+ *
+ * Labels are obtained from instruction emitters, eg ureg_CAL().
+ * Instruction numbers are obtained from ureg_get_instruction_number(),
+ * above.
+ */
+void
+ureg_fixup_label(struct ureg_program *ureg,
+                 unsigned label_token,
+                 unsigned instruction_number );
+
+
+/* Generic instruction emitter.  Use if you need to pass the opcode as
+ * a parameter, rather than using the emit_OP() varients below.
+ */
+void
+ureg_insn(struct ureg_program *ureg,
+          unsigned opcode,
+          const struct ureg_dst *dst,
+          unsigned nr_dst,
+          const struct ureg_src *src,
+          unsigned nr_src );
+
+
+/***********************************************************************
+ * Internal instruction helpers, don't call these directly:
+ */
+
+unsigned
+ureg_emit_insn(struct ureg_program *ureg,
+               unsigned opcode,
+               boolean saturate,
+               unsigned num_dst,
+               unsigned num_src );
+
+void
+ureg_emit_label(struct ureg_program *ureg,
+                unsigned insn_token,
+                unsigned *label_token );
+
+void
+ureg_emit_texture(struct ureg_program *ureg,
+                  unsigned insn_token,
+                  unsigned target );
+
+void 
+ureg_emit_dst( struct ureg_program *ureg,
+               struct ureg_dst dst );
+
+void 
+ureg_emit_src( struct ureg_program *ureg,
+               struct ureg_src src );
+
+void
+ureg_fixup_insn_size(struct ureg_program *ureg,
+                     unsigned insn );
+
+
+#define OP00( op )                                              \
+static INLINE void ureg_##op( struct ureg_program *ureg )       \
+{                                                               \
+   unsigned opcode = TGSI_OPCODE_##op;                          \
+   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \
+   ureg_fixup_insn_size( ureg, insn );                          \
+}
+
+#define OP01( op )                                              \
+static INLINE void ureg_##op( struct ureg_program *ureg,        \
+                              struct ureg_src src )             \
+{                                                               \
+   unsigned opcode = TGSI_OPCODE_##op;                          \
+   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \
+   ureg_emit_src( ureg, src );                                  \
+   ureg_fixup_insn_size( ureg, insn );                          \
+}
+
+#define OP00_LBL( op )                                          \
+static INLINE void ureg_##op( struct ureg_program *ureg,        \
+                              unsigned *label_token )           \
+{                                                               \
+   unsigned opcode = TGSI_OPCODE_##op;                          \
+   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \
+   ureg_emit_label( ureg, insn, label_token );                  \
+   ureg_fixup_insn_size( ureg, insn );                          \
+}
+
+#define OP01_LBL( op )                                          \
+static INLINE void ureg_##op( struct ureg_program *ureg,        \
+                              struct ureg_src src,              \
+                              unsigned *label_token )          \
+{                                                               \
+   unsigned opcode = TGSI_OPCODE_##op;                          \
+   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \
+   ureg_emit_label( ureg, insn, label_token );                  \
+   ureg_emit_src( ureg, src );                                  \
+   ureg_fixup_insn_size( ureg, insn );                          \
+}
+
+#define OP10( op )                                                      \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst )                     \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 0 );  \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+
+#define OP11( op )                                                      \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst,                      \
+                              struct ureg_src src )                     \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 1 );  \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_emit_src( ureg, src );                                          \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+#define OP12( op )                                                      \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst,                      \
+                              struct ureg_src src0,                     \
+                              struct ureg_src src1 )                    \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 );  \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_emit_src( ureg, src0 );                                         \
+   ureg_emit_src( ureg, src1 );                                         \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+#define OP12_TEX( op )                                                  \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst,                      \
+                              unsigned target,                          \
+                              struct ureg_src src0,                     \
+                              struct ureg_src src1 )                    \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 );  \
+   ureg_emit_texture( ureg, insn, target );                             \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_emit_src( ureg, src0 );                                         \
+   ureg_emit_src( ureg, src1 );                                         \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+#define OP13( op )                                                      \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst,                      \
+                              struct ureg_src src0,                     \
+                              struct ureg_src src1,                     \
+                              struct ureg_src src2 )                    \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 3 );  \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_emit_src( ureg, src0 );                                         \
+   ureg_emit_src( ureg, src1 );                                         \
+   ureg_emit_src( ureg, src2 );                                         \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+#define OP14_TEX( op )                                                  \
+static INLINE void ureg_##op( struct ureg_program *ureg,                \
+                              struct ureg_dst dst,                      \
+                              unsigned target,                          \
+                              struct ureg_src src0,                     \
+                              struct ureg_src src1,                     \
+                              struct ureg_src src2,                     \
+                              struct ureg_src src3 )                    \
+{                                                                       \
+   unsigned opcode = TGSI_OPCODE_##op;                                  \
+   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 4 );  \
+   ureg_emit_texture( ureg, insn, target );                             \
+   ureg_emit_dst( ureg, dst );                                          \
+   ureg_emit_src( ureg, src0 );                                         \
+   ureg_emit_src( ureg, src1 );                                         \
+   ureg_emit_src( ureg, src2 );                                         \
+   ureg_emit_src( ureg, src3 );                                         \
+   ureg_fixup_insn_size( ureg, insn );                                  \
+}
+
+
+/* Use a template include to generate a correctly-typed ureg_OP()
+ * function for each TGSI opcode:
+ */
+#include "tgsi_opcode_tmp.h"
+
+
+/***********************************************************************
+ * Inline helpers for manipulating register structs:
+ */
+static INLINE struct ureg_src 
+ureg_negate( struct ureg_src reg )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   reg.Negate ^= 1;
+   return reg;
+}
+
+static INLINE struct ureg_src
+ureg_abs( struct ureg_src reg )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   reg.Absolute = 1;
+   reg.Negate = 0;
+   return reg;
+}
+
+static INLINE struct ureg_src 
+ureg_swizzle( struct ureg_src reg, 
+              int x, int y, int z, int w )
+{
+   unsigned swz = ( (reg.SwizzleX << 0) |
+                    (reg.SwizzleY << 2) |
+                    (reg.SwizzleZ << 4) |
+                    (reg.SwizzleW << 6));
+
+   assert(reg.File != TGSI_FILE_NULL);
+   assert(x < 4);
+   assert(y < 4);
+   assert(z < 4);
+   assert(w < 4);
+
+   reg.SwizzleX = (swz >> (x*2)) & 0x3;
+   reg.SwizzleY = (swz >> (y*2)) & 0x3;
+   reg.SwizzleZ = (swz >> (z*2)) & 0x3;
+   reg.SwizzleW = (swz >> (w*2)) & 0x3;
+   return reg;
+}
+
+static INLINE struct ureg_src
+ureg_scalar( struct ureg_src reg, int x )
+{
+   return ureg_swizzle(reg, x, x, x, x);
+}
+
+static INLINE struct ureg_dst 
+ureg_writemask( struct ureg_dst reg,
+                unsigned writemask )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   reg.WriteMask &= writemask;
+   return reg;
+}
+
+static INLINE struct ureg_dst 
+ureg_saturate( struct ureg_dst reg )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   reg.Saturate = 1;
+   return reg;
+}
+
+static INLINE struct ureg_dst 
+ureg_dst_indirect( struct ureg_dst reg, struct ureg_src addr )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   assert(addr.File == TGSI_FILE_ADDRESS);
+   reg.Indirect = 1;
+   reg.IndirectIndex = addr.Index;
+   reg.IndirectSwizzle = addr.SwizzleX;
+   return reg;
+}
+
+static INLINE struct ureg_src 
+ureg_src_indirect( struct ureg_src reg, struct ureg_src addr )
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   assert(addr.File == TGSI_FILE_ADDRESS);
+   reg.Indirect = 1;
+   reg.IndirectIndex = addr.Index;
+   reg.IndirectSwizzle = addr.SwizzleX;
+   return reg;
+}
+
+static INLINE struct ureg_dst
+ureg_dst( struct ureg_src src )
+{
+   struct ureg_dst dst;
+
+   dst.File      = src.File;
+   dst.WriteMask = TGSI_WRITEMASK_XYZW;
+   dst.Indirect  = src.Indirect;
+   dst.IndirectIndex = src.IndirectIndex;
+   dst.IndirectSwizzle = src.IndirectSwizzle;
+   dst.Saturate  = 0;
+   dst.Index     = src.Index;
+   dst.Pad1      = 0;
+   dst.Pad2      = 0;
+
+   return dst;
+}
+
+static INLINE struct ureg_src
+ureg_src( struct ureg_dst dst )
+{
+   struct ureg_src src;
+
+   src.File      = dst.File;
+   src.SwizzleX  = TGSI_SWIZZLE_X;
+   src.SwizzleY  = TGSI_SWIZZLE_Y;
+   src.SwizzleZ  = TGSI_SWIZZLE_Z;
+   src.SwizzleW  = TGSI_SWIZZLE_W;
+   src.Pad       = 0;
+   src.Indirect  = dst.Indirect;
+   src.IndirectIndex = dst.IndirectIndex;
+   src.IndirectSwizzle = dst.IndirectSwizzle;
+   src.Absolute  = 0;
+   src.Index     = dst.Index;
+   src.Negate    = 0;
+
+   return src;
+}
+
+
+
+static INLINE struct ureg_dst
+ureg_dst_undef( void )
+{
+   struct ureg_dst dst;
+
+   dst.File      = TGSI_FILE_NULL;
+   dst.WriteMask = 0;
+   dst.Indirect  = 0;
+   dst.IndirectIndex = 0;
+   dst.IndirectSwizzle = 0;
+   dst.Saturate  = 0;
+   dst.Index     = 0;
+   dst.Pad1      = 0;
+   dst.Pad2      = 0;
+
+   return dst;
+}
+
+static INLINE struct ureg_src
+ureg_src_undef( void )
+{
+   struct ureg_src src;
+
+   src.File      = TGSI_FILE_NULL;
+   src.SwizzleX  = 0;
+   src.SwizzleY  = 0;
+   src.SwizzleZ  = 0;
+   src.SwizzleW  = 0;
+   src.Pad       = 0;
+   src.Indirect  = 0;
+   src.IndirectIndex = 0;
+   src.IndirectSwizzle = 0;
+   src.Absolute  = 0;
+   src.Index     = 0;
+   src.Negate    = 0;
+   
+   return src;
+}
+
+static INLINE boolean
+ureg_src_is_undef( struct ureg_src src )
+{
+   return src.File == TGSI_FILE_NULL;
+}
+
+static INLINE boolean
+ureg_dst_is_undef( struct ureg_dst dst )
+{
+   return dst.File == TGSI_FILE_NULL;
+}
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h
index 7877f34558..21eb656327 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_util.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.h
@@ -32,6 +32,10 @@
 extern "C" {
 #endif
 
+struct tgsi_src_register;
+struct tgsi_src_register_ext_swz;
+struct tgsi_full_src_register;
+
 void *
 tgsi_align_128bit(
    void *unaligned );
diff --git a/src/gallium/auxiliary/util/.gitignore b/src/gallium/auxiliary/util/.gitignore
new file mode 100644
index 0000000000..29c586c9b5
--- /dev/null
+++ b/src/gallium/auxiliary/util/.gitignore
@@ -0,0 +1,2 @@
+u_format_access.c
+u_format_table.c
diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile
index 2995aba1b9..ae8d330a78 100644
--- a/src/gallium/auxiliary/util/Makefile
+++ b/src/gallium/auxiliary/util/Makefile
@@ -5,17 +5,22 @@ LIBNAME = util
 
 C_SOURCES = \
 	u_debug.c \
+	u_debug_dump.c \
 	u_debug_symbol.c \
 	u_debug_stack.c \
 	u_blit.c \
 	u_cache.c \
 	u_draw_quad.c \
+	u_format.c \
+	u_format_access.c \
+	u_format_table.c \
 	u_gen_mipmap.c \
 	u_handle_table.c \
 	u_hash_table.c \
 	u_hash.c \
 	u_keymap.c \
 	u_linear.c \
+	u_network.c \
 	u_math.c \
 	u_mm.c \
 	u_rect.c \
@@ -31,3 +36,9 @@ C_SOURCES = \
 	u_simple_screen.c
 
 include ../../Makefile.template
+
+u_format_table.c: u_format_table.py u_format_parse.py u_format.csv
+	python u_format_table.py u_format.csv > $@
+
+u_format_access.c: u_format_access.py u_format_parse.py u_format.csv
+	python u_format_access.py u_format.csv > $@
diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript
index d3ac7f747f..28a5ab4256 100644
--- a/src/gallium/auxiliary/util/SConscript
+++ b/src/gallium/auxiliary/util/SConscript
@@ -1,5 +1,23 @@
 Import('*')
 
+env.Clone()
+
+env.Append(CPPPATH = ['.'])
+
+env.CodeGenerate(
+	target = 'u_format_table.c',
+	script = 'u_format_table.py',
+	source = ['u_format.csv'],
+	command = 'python $SCRIPT $SOURCE > $TARGET'
+)
+
+env.CodeGenerate(
+	target = 'u_format_access.c',
+	script = 'u_format_access.py',
+	source = ['u_format.csv'],
+	command = 'python $SCRIPT $SOURCE > $TARGET'
+)
+
 util = env.ConvenienceLibrary(
 	target = 'util',
 	source = [
@@ -7,16 +25,21 @@ util = env.ConvenienceLibrary(
 		'u_blit.c',
 		'u_cache.c',
 		'u_debug.c',
+		'u_debug_dump.c',
 		'u_debug_memory.c',
 		'u_debug_profile.c',
 		'u_debug_stack.c',
 		'u_debug_symbol.c',
 		'u_draw_quad.c',
+		'u_format.c',
+		'u_format_access.c',
+		'u_format_table.c',
 		'u_gen_mipmap.c',
 		'u_handle_table.c',
 		'u_hash.c',
 		'u_hash_table.c',
 		'u_keymap.c',
+		'u_network.c',
 		'u_math.c',
 		'u_mm.c',
 		'u_rect.c',
diff --git a/src/gallium/auxiliary/util/u_cache.c b/src/gallium/auxiliary/util/u_cache.c
index 41cd38171f..47c16b1c92 100644
--- a/src/gallium/auxiliary/util/u_cache.c
+++ b/src/gallium/auxiliary/util/u_cache.c
@@ -137,6 +137,8 @@ util_cache_set(struct util_cache *cache,
    struct util_cache_entry *entry;
 
    assert(cache);
+   if (!cache)
+      return;
 
    entry = util_cache_entry_get(cache, key);
    util_cache_entry_destroy(cache, entry);
@@ -158,6 +160,8 @@ util_cache_get(struct util_cache *cache,
    struct util_cache_entry *entry;
 
    assert(cache);
+   if (!cache)
+      return NULL;
 
    entry = util_cache_entry_get(cache, key);
    if(!entry->key && !entry->value)
@@ -176,7 +180,9 @@ util_cache_clear(struct util_cache *cache)
    uint32_t i;
 
    assert(cache);
-   
+   if (!cache)
+      return;
+
    for(i = 0; i < cache->size; ++i)
       util_cache_entry_destroy(cache, &cache->entries[i]);
 }
@@ -186,6 +192,8 @@ void
 util_cache_destroy(struct util_cache *cache)
 {
    assert(cache);
+   if (!cache)
+      return;
 
 #ifdef DEBUG
    if(cache->count >= 20*cache->size) {
diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h
index d42b65ce28..1380d98d7e 100644
--- a/src/gallium/auxiliary/util/u_debug.h
+++ b/src/gallium/auxiliary/util/u_debug.h
@@ -88,6 +88,7 @@ _debug_printf(const char *format, ...)
  * - avoid outputing large strings (512 bytes is the current maximum length 
  * that is guaranteed to be printed in all platforms)
  */
+#if !defined(PIPE_OS_HAIKU)
 static INLINE void
 debug_printf(const char *format, ...)
 {
@@ -101,6 +102,7 @@ debug_printf(const char *format, ...)
 #endif
 }
 
+#endif /* !PIPE_OS_HAIKU */
 
 /*
  * ... isn't portable so we need to pass arguments in parentheses.
diff --git a/src/gallium/auxiliary/util/u_debug_dump.c b/src/gallium/auxiliary/util/u_debug_dump.c
new file mode 100644
index 0000000000..6bdecde048
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_debug_dump.c
@@ -0,0 +1,189 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_debug.h" 
+#include "util/u_debug_dump.h" 
+
+
+#define DEBUG_DUMP_INVALID_NAME "<invalid>"
+
+
+#if 0
+static const char *
+debug_dump_strip_prefix(const char *name,
+                        const char *prefix) 
+{
+   const char *stripped;
+   assert(name);
+   assert(prefix);
+   stripped = name;
+   while(*prefix) {
+      if(*stripped != *prefix)
+	 return name;
+
+      ++stripped;
+      ++prefix;
+   }
+   return stripped;
+}
+#endif
+
+static const char *
+debug_dump_enum_continuous(unsigned value, 
+                           unsigned num_names,
+                           const char **names)
+{
+   if (value >= num_names)
+      return DEBUG_DUMP_INVALID_NAME;
+   return names[value];
+}
+
+
+#define DEFINE_DEBUG_DUMP_CONTINUOUS(_name) \
+   const char * \
+   debug_dump_##_name(unsigned value, boolean shortened) \
+   { \
+      if(shortened) \
+         return debug_dump_enum_continuous(value, Elements(debug_dump_##_name##_short_names), debug_dump_##_name##_short_names); \
+      else \
+         return debug_dump_enum_continuous(value, Elements(debug_dump_##_name##_names), debug_dump_##_name##_names); \
+   }
+
+
+static const char *
+debug_dump_blend_factor_names[] = {
+   DEBUG_DUMP_INVALID_NAME, /* 0x0 */
+   "PIPE_BLENDFACTOR_ONE",
+   "PIPE_BLENDFACTOR_SRC_COLOR",
+   "PIPE_BLENDFACTOR_SRC_ALPHA",
+   "PIPE_BLENDFACTOR_DST_ALPHA",
+   "PIPE_BLENDFACTOR_DST_COLOR",
+   "PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE",
+   "PIPE_BLENDFACTOR_CONST_COLOR",
+   "PIPE_BLENDFACTOR_CONST_ALPHA",
+   "PIPE_BLENDFACTOR_SRC1_COLOR",
+   "PIPE_BLENDFACTOR_SRC1_ALPHA",
+   DEBUG_DUMP_INVALID_NAME, /* 0x0b */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0c */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0d */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0e */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0f */
+   DEBUG_DUMP_INVALID_NAME, /* 0x10 */
+   "PIPE_BLENDFACTOR_ZERO",
+   "PIPE_BLENDFACTOR_INV_SRC_COLOR",
+   "PIPE_BLENDFACTOR_INV_SRC_ALPHA",
+   "PIPE_BLENDFACTOR_INV_DST_ALPHA",
+   "PIPE_BLENDFACTOR_INV_DST_COLOR",
+   DEBUG_DUMP_INVALID_NAME, /* 0x16 */
+   "PIPE_BLENDFACTOR_INV_CONST_COLOR",
+   "PIPE_BLENDFACTOR_INV_CONST_ALPHA",
+   "PIPE_BLENDFACTOR_INV_SRC1_COLOR",
+   "PIPE_BLENDFACTOR_INV_SRC1_ALPHA"
+};
+
+static const char *
+debug_dump_blend_factor_short_names[] = {
+   DEBUG_DUMP_INVALID_NAME, /* 0x0 */
+   "one",
+   "src_color",
+   "src_alpha",
+   "dst_alpha",
+   "dst_color",
+   "src_alpha_saturate",
+   "const_color",
+   "const_alpha",
+   "src1_color",
+   "src1_alpha",
+   DEBUG_DUMP_INVALID_NAME, /* 0x0b */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0c */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0d */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0e */
+   DEBUG_DUMP_INVALID_NAME, /* 0x0f */
+   DEBUG_DUMP_INVALID_NAME, /* 0x10 */
+   "zero",
+   "inv_src_color",
+   "inv_src_alpha",
+   "inv_dst_alpha",
+   "inv_dst_color",
+   DEBUG_DUMP_INVALID_NAME, /* 0x16 */
+   "inv_const_color",
+   "inv_const_alpha",
+   "inv_src1_color",
+   "inv_src1_alpha"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(blend_factor)
+
+
+static const char *
+debug_dump_blend_func_names[] = {
+   "PIPE_BLEND_ADD",
+   "PIPE_BLEND_SUBTRACT",
+   "PIPE_BLEND_REVERSE_SUBTRACT",
+   "PIPE_BLEND_MIN",
+   "PIPE_BLEND_MAX"
+};
+
+static const char *
+debug_dump_blend_func_short_names[] = {
+   "add",
+   "sub",
+   "rev_sub",
+   "min",
+   "max"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(blend_func)
+
+
+static const char *
+debug_dump_func_names[] = {
+   "PIPE_FUNC_NEVER",
+   "PIPE_FUNC_LESS",
+   "PIPE_FUNC_EQUAL",
+   "PIPE_FUNC_LEQUAL",
+   "PIPE_FUNC_GREATER",
+   "PIPE_FUNC_NOTEQUAL",
+   "PIPE_FUNC_GEQUAL",
+   "PIPE_FUNC_ALWAYS"
+};
+
+static const char *
+debug_dump_func_short_names[] = {
+   "never",
+   "less",
+   "equal",
+   "less_equal",
+   "greater",
+   "not_equal",
+   "greater_equal",
+   "always"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(func)
diff --git a/src/gallium/auxiliary/util/u_debug_dump.h b/src/gallium/auxiliary/util/u_debug_dump.h
new file mode 100644
index 0000000000..102935559c
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_debug_dump.h
@@ -0,0 +1,65 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Dump data in human/machine readable format.
+ * 
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef U_DEBUG_DUMP_H_
+#define U_DEBUG_DUMP_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+
+const char *
+debug_dump_blend_factor(unsigned value, boolean shortened);
+
+const char *
+debug_dump_blend_func(unsigned value, boolean shortened);
+
+const char *
+debug_dump_func(unsigned value, boolean shortened);
+
+
+/* FIXME: Move the other debug_dump_xxx functions out of u_debug.h into here. */
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* U_DEBUG_H_ */
diff --git a/src/gallium/auxiliary/util/u_format.c b/src/gallium/auxiliary/util/u_format.c
new file mode 100644
index 0000000000..98ea13b60b
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format.c
@@ -0,0 +1,46 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Vmware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "u_format.h"
+
+
+const struct util_format_description *
+util_format_description(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description_table;
+
+   while(TRUE) {
+      if(desc->format == format)
+         return desc;
+
+      if(desc->format == PIPE_FORMAT_NONE)
+         return NULL;
+
+      ++desc;
+   };
+}
diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv
new file mode 100644
index 0000000000..00a46d0cc4
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format.csv
@@ -0,0 +1,99 @@
+PIPE_FORMAT_A8R8G8B8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , zyxw, rgb
+PIPE_FORMAT_X8R8G8B8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , zyx1, rgb
+PIPE_FORMAT_B8G8R8A8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , yzwx, rgb
+PIPE_FORMAT_B8G8R8X8_UNORM        , arith , 1, 1, un8 , un8 , un8 , un8 , yzw1, rgb
+PIPE_FORMAT_A1R5G5B5_UNORM        , arith , 1, 1, un1 , un5 , un5 , un5 , zyxw, rgb
+PIPE_FORMAT_A4R4G4B4_UNORM        , arith , 1, 1, un4 , un4 , un4 , un4 , zyxw, rgb
+PIPE_FORMAT_R5G6B5_UNORM          , arith , 1, 1, un5 , un6 , un5 ,     , zyx1, rgb
+PIPE_FORMAT_A2B10G10R10_UNORM     , arith , 1, 1, un10, un10, un10, un2 , xyzw, rgb
+PIPE_FORMAT_L8_UNORM              , arith , 1, 1, un8 ,     ,     ,     , xxx1, rgb
+PIPE_FORMAT_A8_UNORM              , arith , 1, 1, un8 ,     ,     ,     , 000x, rgb
+PIPE_FORMAT_I8_UNORM              , arith , 1, 1, un8 ,     ,     ,     , xxxx, rgb
+PIPE_FORMAT_A8L8_UNORM            , arith , 1, 1, un8 , un8 ,     ,     , xxxy, rgb
+PIPE_FORMAT_L16_UNORM             , arith , 1, 1, un16,     ,     ,     , xxx1, rgb
+PIPE_FORMAT_Z16_UNORM             , array , 1, 1, un16,     ,     ,     , x___, zs 
+PIPE_FORMAT_Z32_UNORM             , array , 1, 1, un32,     ,     ,     , x___, zs 
+PIPE_FORMAT_Z32_FLOAT             , array , 1, 1, f32 ,     ,     ,     , x___, zs 
+PIPE_FORMAT_S8Z24_UNORM           , arith , 1, 1, un8 , un24,     ,     , yx__, zs 
+PIPE_FORMAT_Z24S8_UNORM           , arith , 1, 1, un24, un8 ,     ,     , xy__, zs 
+PIPE_FORMAT_X8Z24_UNORM           , arith , 1, 1, un8 , un24,     ,     , y___, zs 
+PIPE_FORMAT_Z24X8_UNORM           , arith , 1, 1, un24, un8 ,     ,     , x___, zs 
+PIPE_FORMAT_S8_UNORM              , array , 1, 1, un8 ,     ,     ,     , _x__, zs 
+PIPE_FORMAT_R64_FLOAT             , array , 1, 1, f64 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R64G64_FLOAT          , array , 1, 1, f64 , f64 ,     ,     , xy01, rgb
+PIPE_FORMAT_R64G64B64_FLOAT       , array , 1, 1, f64 , f64 , f64 ,     , xyz1, rgb
+PIPE_FORMAT_R64G64B64A64_FLOAT    , array , 1, 1, f64 , f64 , f64 , f64 , xyzw, rgb
+PIPE_FORMAT_R32_FLOAT             , array , 1, 1, f32 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_FLOAT          , array , 1, 1, f32 , f32 ,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_FLOAT       , array , 1, 1, f32 , f32 , f32 ,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_FLOAT    , array , 1, 1, f32 , f32 , f32 , f32 , xyzw, rgb
+PIPE_FORMAT_R32_UNORM             , array , 1, 1, un32,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_UNORM          , array , 1, 1, un32, un32,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_UNORM       , array , 1, 1, un32, un32, un32,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_UNORM    , array , 1, 1, un32, un32, un32, un32, xyzw, rgb
+PIPE_FORMAT_R32_USCALED           , array , 1, 1, u32 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_USCALED        , array , 1, 1, u32 , u32 ,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_USCALED     , array , 1, 1, u32 , u32 , u32 ,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_USCALED  , array , 1, 1, u32 , u32 , u32 , u32 , xyzw, rgb
+PIPE_FORMAT_R32_SNORM             , array , 1, 1, sn32,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_SNORM          , array , 1, 1, sn32, sn32,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_SNORM       , array , 1, 1, sn32, sn32, sn32,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_SNORM    , array , 1, 1, sn32, sn32, sn32, sn32, xyzw, rgb
+PIPE_FORMAT_R32_SSCALED           , array , 1, 1, s32 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_SSCALED        , array , 1, 1, s32 , s32 ,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_SSCALED     , array , 1, 1, s32 , s32 , s32 ,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_SSCALED  , array , 1, 1, s32 , s32 , s32 , s32 , xyzw, rgb
+PIPE_FORMAT_R16_UNORM             , array , 1, 1, un16,     ,     ,     , x001, rgb
+PIPE_FORMAT_R16G16_UNORM          , array , 1, 1, un16, un16,     ,     , xy01, rgb
+PIPE_FORMAT_R16G16B16_UNORM       , array , 1, 1, un16, un16, un16,     , xyz1, rgb
+PIPE_FORMAT_R16G16B16A16_UNORM    , array , 1, 1, un16, un16, un16, un16, xyzw, rgb
+PIPE_FORMAT_R16_USCALED           , array , 1, 1, u16 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R16G16_USCALED        , array , 1, 1, u16 , u16 ,     ,     , xy01, rgb
+PIPE_FORMAT_R16G16B16_USCALED     , array , 1, 1, u16 , u16 , u16 ,     , xyz1, rgb
+PIPE_FORMAT_R16G16B16A16_USCALED  , array , 1, 1, u16 , u16 , u16 , u16 , xyzw, rgb
+PIPE_FORMAT_R16_SNORM             , array , 1, 1, sn16,     ,     ,     , x001, rgb
+PIPE_FORMAT_R16G16_SNORM          , array , 1, 1, sn16, sn16,     ,     , xy01, rgb
+PIPE_FORMAT_R16G16B16_SNORM       , array , 1, 1, sn16, sn16, sn16,     , xyz1, rgb
+PIPE_FORMAT_R16G16B16A16_SNORM    , array , 1, 1, sn16, sn16, sn16, sn16, xyzw, rgb
+PIPE_FORMAT_R16_SSCALED           , array , 1, 1, s16 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R16G16_SSCALED        , array , 1, 1, s16 , s16 ,     ,     , xy01, rgb
+PIPE_FORMAT_R16G16B16_SSCALED     , array , 1, 1, s16 , s16 , s16 ,     , xyz1, rgb
+PIPE_FORMAT_R16G16B16A16_SSCALED  , array , 1, 1, s16 , s16 , s16 , s16 , xyzw, rgb
+PIPE_FORMAT_R8_UNORM              , array , 1, 1, un8 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R8G8_UNORM            , array , 1, 1, un8 , un8 ,     ,     , xy01, rgb
+PIPE_FORMAT_R8G8B8_UNORM          , array , 1, 1, un8 , un8 , un8 ,     , xyz1, rgb
+PIPE_FORMAT_R8G8B8A8_UNORM        , array , 1, 1, un8 , un8 , un8 , un8 , xyzw, rgb
+PIPE_FORMAT_R8G8B8X8_UNORM        , array , 1, 1, un8 , un8 , un8 , un8 , xyz1, rgb
+PIPE_FORMAT_R8_USCALED            , array , 1, 1, u8  ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R8G8_USCALED          , array , 1, 1, u8  , u8  ,     ,     , xy01, rgb
+PIPE_FORMAT_R8G8B8_USCALED        , array , 1, 1, u8  , u8  , u8  ,     , xyz1, rgb
+PIPE_FORMAT_R8G8B8A8_USCALED      , array , 1, 1, u8  , u8  , u8  , u8  , xyzw, rgb
+PIPE_FORMAT_R8G8B8X8_USCALED      , array , 1, 1, u8  , u8  , u8  , u8  , xyz1, rgb
+PIPE_FORMAT_R8_SNORM              , array , 1, 1, sn8 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R8G8_SNORM            , array , 1, 1, sn8 , sn8 ,     ,     , xy01, rgb
+PIPE_FORMAT_R8G8B8_SNORM          , array , 1, 1, sn8 , sn8 , sn8 ,     , xyz1, rgb
+PIPE_FORMAT_R8G8B8A8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyzw, rgb
+PIPE_FORMAT_R8G8B8X8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyz1, rgb
+PIPE_FORMAT_B6G5R5_SNORM          , arith , 1, 1, sn5 , sn5 , sn6 ,     , zyx1, rgb
+PIPE_FORMAT_A8B8G8R8_SNORM        , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyxw, rgb
+PIPE_FORMAT_X8B8G8R8_SNORM        , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyx1, rgb
+PIPE_FORMAT_R8_SSCALED            , array , 1, 1, s8  ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R8G8_SSCALED          , array , 1, 1, s8  , s8  ,     ,     , xy01, rgb
+PIPE_FORMAT_R8G8B8_SSCALED        , array , 1, 1, s8  , s8  , s8  ,     , xyz1, rgb
+PIPE_FORMAT_R8G8B8A8_SSCALED      , array , 1, 1, s8  , s8  , s8  , s8  , xyzw, rgb
+PIPE_FORMAT_R8G8B8X8_SSCALED      , array , 1, 1, s8  , s8  , s8  , s8  , xyz1, rgb
+PIPE_FORMAT_R32_FIXED             , array , 1, 1, h32 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R32G32_FIXED          , array , 1, 1, h32 , h32 ,     ,     , xy01, rgb
+PIPE_FORMAT_R32G32B32_FIXED       , array , 1, 1, h32 , h32 , h32 ,     , xyz1, rgb
+PIPE_FORMAT_R32G32B32A32_FIXED    , array , 1, 1, h32 , h32 , h32 , h32 , xyzw, rgb
+PIPE_FORMAT_L8_SRGB               , arith , 1, 1, u8  ,     ,     ,     , xxx1, srgb 
+PIPE_FORMAT_A8L8_SRGB             , arith , 1, 1, u8  , u8  ,     ,     , xxxy, srgb 
+PIPE_FORMAT_R8G8B8_SRGB           , arith , 1, 1, u8  , u8  , u8  ,     , xyz1, srgb 
+PIPE_FORMAT_R8G8B8A8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , xyzw, srgb 
+PIPE_FORMAT_R8G8B8X8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , xyz1, srgb 
+PIPE_FORMAT_A8R8G8B8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , wxyz, srgb 
+PIPE_FORMAT_X8R8G8B8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , 1xyz, srgb 
+PIPE_FORMAT_B8G8R8A8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , zyxw, srgb 
+PIPE_FORMAT_B8G8R8X8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , zyx1, srgb 
+PIPE_FORMAT_X8UB8UG8SR8S_NORM     , arith , 1, 1, sn8 , sn8 , un8 , x8  , 1zyx, rgb
+PIPE_FORMAT_B6UG5SR5S_NORM        , arith , 1, 1, sn5 , sn5 , un6 ,     , xyz1, rgb
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
new file mode 100644
index 0000000000..7b5b7fcda5
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Vmware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef U_FORMAT_H
+#define U_FORMAT_H
+
+
+#include "pipe/p_format.h"
+
+
+enum util_format_layout {
+   UTIL_FORMAT_LAYOUT_SCALAR = 0,
+   UTIL_FORMAT_LAYOUT_ARITH = 1,
+   UTIL_FORMAT_LAYOUT_ARRAY = 2,
+   UTIL_FORMAT_LAYOUT_YUV = 3,
+   UTIL_FORMAT_LAYOUT_DXT = 4
+};
+
+
+struct util_format_block
+{
+   /** Block width in pixels */
+   unsigned width;
+   
+   /** Block height in pixels */
+   unsigned height;
+
+   /** Block size in bytes */
+   unsigned bits;
+};
+
+
+enum util_format_type {
+   UTIL_FORMAT_TYPE_VOID = 0,
+   UTIL_FORMAT_TYPE_UNSIGNED = 1,
+   UTIL_FORMAT_TYPE_SIGNED = 2,
+   UTIL_FORMAT_TYPE_FIXED = 3,
+   UTIL_FORMAT_TYPE_FLOAT = 4
+};
+
+
+enum util_format_swizzle {
+   UTIL_FORMAT_SWIZZLE_X = 0,
+   UTIL_FORMAT_SWIZZLE_Y = 1,
+   UTIL_FORMAT_SWIZZLE_Z = 2,
+   UTIL_FORMAT_SWIZZLE_W = 3,
+   UTIL_FORMAT_SWIZZLE_0 = 4,
+   UTIL_FORMAT_SWIZZLE_1 = 5,
+   UTIL_FORMAT_SWIZZLE_NONE = 6
+};
+
+
+enum util_format_colorspace {
+   UTIL_FORMAT_COLORSPACE_RGB = 0,
+   UTIL_FORMAT_COLORSPACE_SRGB = 1,
+   UTIL_FORMAT_COLORSPACE_YUV = 2,
+   UTIL_FORMAT_COLORSPACE_ZS = 3,
+};
+
+
+struct util_format_channel_description
+{
+   unsigned type:6;
+   unsigned normalized:1;
+   unsigned size:9;
+};
+
+
+struct util_format_description
+{
+   enum pipe_format format;
+   const char *name;
+   struct util_format_block block;
+   enum util_format_layout layout;
+   struct util_format_channel_description channel[4];
+   unsigned char swizzle[4];
+   enum util_format_colorspace colorspace;
+};
+
+
+extern const struct util_format_description 
+util_format_description_table[];
+
+
+const struct util_format_description *
+util_format_description(enum pipe_format format);
+
+
+void
+util_format_read_4f(enum pipe_format format,
+                    float *dst, unsigned dst_stride, 
+                    const void *src, unsigned src_stride, 
+                    unsigned x, unsigned y, unsigned w, unsigned h);
+
+void
+util_format_write_4f(enum pipe_format format,
+                     const float *src, unsigned src_stride, 
+                     void *dst, unsigned dst_stride, 
+                     unsigned x, unsigned y, unsigned w, unsigned h);
+
+void
+util_format_read_4ub(enum pipe_format format,
+                     uint8_t *dst, unsigned dst_stride, 
+                     const void *src, unsigned src_stride, 
+                     unsigned x, unsigned y, unsigned w, unsigned h);
+
+void
+util_format_write_4ub(enum pipe_format format,
+                      const uint8_t *src, unsigned src_stride, 
+                      void *dst, unsigned dst_stride, 
+                      unsigned x, unsigned y, unsigned w, unsigned h);
+
+#endif /* ! U_FORMAT_H */
diff --git a/src/gallium/auxiliary/util/u_format_access.py b/src/gallium/auxiliary/util/u_format_access.py
new file mode 100644
index 0000000000..eeb1a9657f
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_access.py
@@ -0,0 +1,498 @@
+#!/usr/bin/env python
+
+'''
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Pixel format accessor functions.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+'''
+
+
+import sys
+
+from u_format_parse import *
+
+
+def short_name(format):
+    '''Make up a short norm for a format, suitable to be used as suffix in
+    function names.'''
+
+    name = format.name
+    if name.startswith('PIPE_FORMAT_'):
+        name = name[len('PIPE_FORMAT_'):]
+    name = name.lower()
+    return name
+
+
+def is_format_supported(format):
+    '''Determines whether we actually have the plumbing necessary to generate the 
+    to read/write to/from this format.'''
+
+    # FIXME: Ideally we would support any format combination here.
+
+    # XXX: It should be straightforward to support srgb
+    if format.colorspace not in ('rgb', 'zs'):
+        return False
+
+    if format.layout not in (ARITH, ARRAY):
+        return False
+
+    for i in range(4):
+        type = format.in_types[i]
+        if type.kind not in (VOID, UNSIGNED, FLOAT):
+            return False
+
+    # We can only read a color from a depth/stencil format if the depth channel is present
+    if format.colorspace == 'zs' and format.out_swizzle[0] == SWIZZLE_NONE:
+        return False
+
+    return True
+
+
+def native_type(format):
+    '''Get the native appropriate for a format.'''
+
+    if format.layout == ARITH:
+        # For arithmetic pixel formats return the integer type that matches the whole pixel
+        return 'uint%u_t' % format.block_size()
+    elif format.layout == ARRAY:
+        # For array pixel formats return the integer type that matches the color channel
+        type = format.in_types[0]
+        if type.kind == UNSIGNED:
+            return 'uint%u_t' % type.size
+        elif type.kind == SIGNED:
+            return 'int%u_t' % type.size
+        elif type.kind == FLOAT:
+            if type.size == 32:
+                return 'float'
+            elif type.size == 64:
+                return 'double'
+            else:
+                assert False
+        else:
+            assert False
+    else:
+        assert False
+
+
+def intermediate_native_type(bits, sign):
+    '''Find a native type adequate to hold intermediate results of the request bit size.'''
+
+    bytes = 4 # don't use anything smaller than 32bits
+    while bytes * 8 < bits:
+        bytes *= 2
+    bits = bytes*8
+
+    if sign:
+        return 'int%u_t' % bits
+    else:
+        return 'uint%u_t' % bits
+
+
+def get_one_shift(type):
+    '''Get the number of the bit that matches unity for this type.'''
+    if type.kind == 'FLOAT':
+        assert False
+    if not type.norm:
+        return 0
+    if type.kind == UNSIGNED:
+        return type.size
+    if type.kind == SIGNED:
+        return type.size - 1
+    if type.kind == FIXED:
+        return type.size / 2
+    assert False
+
+
+def get_one(type):
+    '''Get the value of unity for this type.'''
+    if type.kind == 'FLOAT' or not type.norm:
+        return 1
+    else:
+        return (1 << get_one_shift(type)) - 1
+
+
+def generate_clamp():
+    '''Code generate the clamping functions for each type.
+
+    We don't use a macro so that arguments with side effects, 
+    like *src_pixel++ are correctly handled.
+    '''
+
+    for suffix, native_type in [
+        ('', 'double'),
+        ('f', 'float'),
+        ('ui', 'unsigned int'),
+        ('si', 'int'),
+    ]:
+        print 'static INLINE %s' % native_type
+        print 'clamp%s(%s value, %s lbound, %s ubound)' % (suffix, native_type, native_type, native_type)
+        print '{'
+        print '   if(value < lbound)'
+        print '      return lbound;'
+        print '   if(value > ubound)'
+        print '      return ubound;'
+        print '   return value;'
+        print '}'
+        print
+
+
+def clamp_expr(src_type, dst_type, dst_native_type, value):
+    '''Generate the expression to clamp the value in the source type to the
+    destination type range.'''
+
+    if src_type == dst_type:
+        return value
+
+    # Pick the approriate clamp function
+    if src_type.kind == FLOAT:
+        if src_type.size == 32:
+            func = 'clampf'
+        elif src_type.size == 64:
+            func = 'clamp'
+        else:
+            assert False
+    elif src_type.kind == UNSIGNED:
+        func = 'clampui'
+    elif src_type.kind == SIGNED:
+        func = 'clampsi'
+    else:
+        assert False
+
+    # Clamp floats to [-1, 1] or [0, 1] range
+    if src_type.kind == FLOAT and dst_type.norm:
+        max = 1
+        if src_type.sign and dst_type.sign:
+            min = -1
+        else:
+            min = 0
+        return '%s(%s, %s, %s)' % (func, value, min, max)
+                
+    # FIXME: Also clamp scaled values
+
+    return value
+
+
+def conversion_expr(src_type, dst_type, dst_native_type, value):
+    '''Generate the expression to convert a value between two types.'''
+
+    if src_type == dst_type:
+        return value
+
+    if src_type.kind == FLOAT and dst_type.kind == FLOAT:
+        return '(%s)%s' % (dst_native_type, value)
+    
+    if not src_type.norm and not dst_type.norm:
+        return '(%s)%s' % (dst_native_type, value)
+
+    value = clamp_expr(src_type, dst_type, dst_native_type, value)
+
+    if dst_type.kind == FLOAT:
+        if src_type.norm:
+            one = get_one(src_type)
+            if src_type.size <= 23:
+                scale = '(1.0f/0x%x)' % one
+            else:
+                # bigger than single precision mantissa, use double
+                scale = '(1.0/0x%x)' % one
+            value = '(%s * %s)' % (value, scale)
+        return '(%s)%s' % (dst_native_type, value)
+
+    if src_type.kind == FLOAT:
+        if dst_type.norm:
+            dst_one = get_one(dst_type)
+            if dst_type.size <= 23:
+                scale = '0x%x' % dst_one
+            else:
+                # bigger than single precision mantissa, use double
+                scale = '(double)0x%x' % dst_one
+            value = '(%s * %s)' % (value, scale)
+        return '(%s)%s' % (dst_native_type, value)
+
+    if src_type.kind == dst_type.kind:
+        src_one = get_one(src_type)
+        dst_one = get_one(dst_type)
+
+        if src_one > dst_one and src_type.norm and dst_type.norm:
+            # We can just bitshift
+            src_shift = get_one_shift(src_type)
+            dst_shift = get_one_shift(dst_type)
+            value = '(%s >> %s)' % (value, src_shift - dst_shift)
+        else:
+            # We need to rescale using an intermediate type big enough to hold the multiplication of both
+            tmp_native_type = intermediate_native_type(src_type.size + dst_type.size, src_type.sign and dst_type.sign)
+            value = '(%s)%s' % (tmp_native_type, value)
+            value = '%s * 0x%x / 0x%x' % (value, dst_one, src_one)
+        value = '(%s)%s' % (dst_native_type, value)
+        return value
+
+    assert False
+
+
+def generate_format_read(format, dst_type, dst_native_type, dst_suffix):
+    '''Generate the function to read pixels from a particular format'''
+
+    name = short_name(format)
+
+    src_native_type = native_type(format)
+
+    print 'static void'
+    print 'util_format_%s_read_%s(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, dst_suffix, dst_native_type)
+    print '{'
+    print '   unsigned x, y;'
+    print '   const uint8_t *src_row = src + y0*src_stride;'
+    print '   %s *dst_row = dst;' % dst_native_type
+    print '   for (y = 0; y < h; ++y) {'
+    print '      const %s *src_pixel = (const %s *)(src_row + x0*%u);' % (src_native_type, src_native_type, format.stride())
+    print '      %s *dst_pixel = dst_row;' %dst_native_type
+    print '      for (x = 0; x < w; ++x) {'
+
+    names = ['']*4
+    if format.colorspace == 'rgb':
+        for i in range(4):
+            swizzle = format.out_swizzle[i]
+            if swizzle < 4:
+                names[swizzle] += 'rgba'[i]
+    elif format.colorspace == 'zs':
+        swizzle = format.out_swizzle[0]
+        if swizzle < 4:
+            names[swizzle] = 'z'
+        else:
+            assert False
+    else:
+        assert False
+
+    if format.layout == ARITH:
+        print '         %s pixel = *src_pixel++;' % src_native_type
+        shift = 0;
+        for i in range(4):
+            src_type = format.in_types[i]
+            width = src_type.size
+            if names[i]:
+                value = 'pixel'
+                mask = (1 << width) - 1
+                if shift:
+                    value = '(%s >> %u)' % (value, shift)
+                if shift + width < format.block_size():
+                    value = '(%s & 0x%x)' % (value, mask)
+                value = conversion_expr(src_type, dst_type, dst_native_type, value)
+                print '         %s %s = %s;' % (dst_native_type, names[i], value)
+            shift += width
+    elif format.layout == ARRAY:
+        for i in range(4):
+            src_type = format.in_types[i]
+            if names[i]:
+                value = '(*src_pixel++)'
+                value = conversion_expr(src_type, dst_type, dst_native_type, value)
+                print '         %s %s = %s;' % (dst_native_type, names[i], value)
+    else:
+        assert False
+
+    for i in range(4):
+        if format.colorspace == 'rgb':
+            swizzle = format.out_swizzle[i]
+            if swizzle < 4:
+                value = names[swizzle]
+            elif swizzle == SWIZZLE_0:
+                value = '0'
+            elif swizzle == SWIZZLE_1:
+                value = '1'
+            else:
+                assert False
+        elif format.colorspace == 'zs':
+            if i < 3:
+                value = 'z'
+            else:
+                value = '1'
+        else:
+            assert False
+        print '         *dst_pixel++ = %s; /* %s */' % (value, 'rgba'[i])
+
+    print '      }'
+    print '      src_row += src_stride;'
+    print '      dst_row += dst_stride/sizeof(%s);' % dst_native_type
+    print '   }'
+    print '}'
+    print
+    
+
+def generate_format_write(format, src_type, src_native_type, src_suffix):
+    '''Generate the function to write pixels to a particular format'''
+
+    name = short_name(format)
+
+    dst_native_type = native_type(format)
+
+    print 'static void'
+    print 'util_format_%s_write_%s(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, src_suffix, src_native_type)
+    print '{'
+    print '   unsigned x, y;'
+    print '   uint8_t *dst_row = dst + y0*dst_stride;'
+    print '   const %s *src_row = src;' % src_native_type
+    print '   for (y = 0; y < h; ++y) {'
+    print '      %s *dst_pixel = (%s *)(dst_row + x0*%u);' % (dst_native_type, dst_native_type, format.stride())
+    print '      const %s *src_pixel = src_row;' %src_native_type
+    print '      for (x = 0; x < w; ++x) {'
+
+    inv_swizzle = [None]*4
+    if format.colorspace == 'rgb':
+        for i in range(4):
+            swizzle = format.out_swizzle[i]
+            if swizzle < 4:
+                inv_swizzle[swizzle] = i
+    elif format.colorspace == 'zs':
+        swizzle = format.out_swizzle[0]
+        if swizzle < 4:
+            inv_swizzle[swizzle] = 0
+    else:
+        assert False
+
+    if format.layout == ARITH:
+        print '         %s pixel = 0;' % dst_native_type
+        shift = 0;
+        for i in range(4):
+            dst_type = format.in_types[i]
+            width = dst_type.size
+            if inv_swizzle[i] is not None:
+                value = 'src_pixel[%u]' % inv_swizzle[i]
+                value = conversion_expr(src_type, dst_type, dst_native_type, value)
+                if shift:
+                    value = '(%s << %u)' % (value, shift)
+                print '         pixel |= %s;' % value
+            shift += width
+        print '         *dst_pixel++ = pixel;'
+    elif format.layout == ARRAY:
+        for i in range(4):
+            dst_type = format.in_types[i]
+            if inv_swizzle[i] is not None:
+                value = 'src_pixel[%u]' % inv_swizzle[i]
+                value = conversion_expr(src_type, dst_type, dst_native_type, value)
+                print '         *dst_pixel++ = %s;' % value
+    else:
+        assert False
+    print '         src_pixel += 4;'
+
+    print '      }'
+    print '      dst_row += dst_stride;'
+    print '      src_row += src_stride/sizeof(%s);' % src_native_type
+    print '   }'
+    print '}'
+    print
+    
+
+def generate_read(formats, dst_type, dst_native_type, dst_suffix):
+    '''Generate the dispatch function to read pixels from any format'''
+
+    for format in formats:
+        if is_format_supported(format):
+            generate_format_read(format, dst_type, dst_native_type, dst_suffix)
+
+    print 'void'
+    print 'util_format_read_%s(enum pipe_format format, %s *dst, unsigned dst_stride, const void *src, unsigned src_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (dst_suffix, dst_native_type)
+    print '{'
+    print '   void (*func)(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % dst_native_type
+    print '   switch(format) {'
+    for format in formats:
+        if is_format_supported(format):
+            print '   case %s:' % format.name
+            print '      func = &util_format_%s_read_%s;' % (short_name(format), dst_suffix)
+            print '      break;'
+    print '   default:'
+    print '      debug_printf("unsupported format\\n");'
+    print '      return;'
+    print '   }'
+    print '   func(dst, dst_stride, (const uint8_t *)src, src_stride, x, y, w, h);'
+    print '}'
+    print
+
+
+def generate_write(formats, src_type, src_native_type, src_suffix):
+    '''Generate the dispatch function to write pixels to any format'''
+
+    for format in formats:
+        if is_format_supported(format):
+            generate_format_write(format, src_type, src_native_type, src_suffix)
+
+    print 'void'
+    print 'util_format_write_%s(enum pipe_format format, const %s *src, unsigned src_stride, void *dst, unsigned dst_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (src_suffix, src_native_type)
+    
+    print '{'
+    print '   void (*func)(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % src_native_type
+    print '   switch(format) {'
+    for format in formats:
+        if is_format_supported(format):
+            print '   case %s:' % format.name
+            print '      func = &util_format_%s_write_%s;' % (short_name(format), src_suffix)
+            print '      break;'
+    print '   default:'
+    print '      debug_printf("unsupported format\\n");'
+    print '      return;'
+    print '   }'
+    print '   func(src, src_stride, (uint8_t *)dst, dst_stride, x, y, w, h);'
+    print '}'
+    print
+
+
+def main():
+    formats = []
+    for arg in sys.argv[1:]:
+        formats.extend(parse(arg))
+
+    print '/* This file is autogenerated by u_format_access.py from u_format.csv. Do not edit directly. */'
+    print
+    # This will print the copyright message on the top of this file
+    print __doc__.strip()
+    print
+    print '#include "pipe/p_compiler.h"'
+    print '#include "u_format.h"'
+    print '#include "u_math.h"'
+    print
+
+    generate_clamp()
+
+    type = Type(FLOAT, False, 32)
+    native_type = 'float'
+    suffix = '4f'
+
+    generate_read(formats, type, native_type, suffix)
+    generate_write(formats, type, native_type, suffix)
+
+    type = Type(UNSIGNED, True, 8)
+    native_type = 'uint8_t'
+    suffix = '4ub'
+
+    generate_read(formats, type, native_type, suffix)
+    generate_write(formats, type, native_type, suffix)
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/auxiliary/util/u_format_parse.py b/src/gallium/auxiliary/util/u_format_parse.py
new file mode 100755
index 0000000000..493aff7112
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_parse.py
@@ -0,0 +1,141 @@
+#!/usr/bin/env python
+
+'''
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+'''
+
+
+import sys
+
+
+VOID, UNSIGNED, SIGNED, FIXED, FLOAT = range(5)
+
+SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_0, SWIZZLE_1, SWIZZLE_NONE, = range(7)
+
+ARITH = 'arith'
+ARRAY = 'array'
+
+
+class Type:
+    '''Describe the type of a color channel.'''
+    
+    def __init__(self, kind, norm, size):
+        self.kind = kind
+        self.norm = norm
+        self.size = size
+        self.sign = kind in (SIGNED, FIXED, FLOAT)
+
+    def __str__(self):
+        s = str(self.kind)
+        if self.norm:
+            s += 'n'
+        s += str(self.size)
+        return s
+
+    def __eq__(self, other):
+        return self.kind == other.kind and self.norm == other.norm and self.size == other.size
+
+
+class Format:
+    '''Describe a pixel format.'''
+
+    def __init__(self, name, layout, block_width, block_height, in_types, out_swizzle, colorspace):
+        self.name = name
+        self.layout = layout
+        self.block_width = block_width
+        self.block_height = block_height
+        self.in_types = in_types
+        self.out_swizzle = out_swizzle
+        self.name = name
+        self.colorspace = colorspace
+
+    def __str__(self):
+        return self.name
+
+    def block_size(self):
+        size = 0
+        for type in self.in_types:
+            size += type.size
+        return size
+
+    def stride(self):
+        return self.block_size()/8
+
+
+_kind_parse_map = {
+    '':  VOID,
+    'x': VOID,
+    'u': UNSIGNED,
+    's': SIGNED,
+    'h': FIXED,
+    'f': FLOAT,
+}
+
+_swizzle_parse_map = {
+    'x': SWIZZLE_X,
+    'y': SWIZZLE_Y,
+    'z': SWIZZLE_Z,
+    'w': SWIZZLE_W,
+    '0': SWIZZLE_0,
+    '1': SWIZZLE_1,
+    '_': SWIZZLE_NONE,
+}
+
+def parse(filename):
+    '''Parse the format descrition in CSV format in terms of the 
+    Type and Format classes above.'''
+
+    stream = open(filename)
+    formats = []
+    for line in stream:
+        line = line.rstrip()
+        fields = [field.strip() for field in line.split(',')]
+        name = fields[0]
+        layout = fields[1]
+        block_width, block_height = map(int, fields[2:4])
+        in_types = []
+        for field in fields[4:8]:
+            if field:
+                kind = _kind_parse_map[field[0]]
+                if field[1] == 'n':
+                    norm = True
+                    size = int(field[2:])
+                else:
+                    norm = False
+                    size = int(field[1:])
+            else:
+                kind = VOID
+                norm = False
+                size = 0
+            in_type = Type(kind, norm, size)
+            in_types.append(in_type)
+        out_swizzle = [_swizzle_parse_map[swizzle] for swizzle in fields[8]]
+        colorspace = fields[9]
+        formats.append(Format(name, layout, block_width, block_height, in_types, out_swizzle, colorspace))
+    return formats
+
diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py
new file mode 100755
index 0000000000..8834568e8e
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_table.py
@@ -0,0 +1,144 @@
+#!/usr/bin/env python
+
+'''
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+'''
+
+
+import sys
+
+from u_format_parse import *
+
+
+def layout_map(layout):
+    return 'UTIL_FORMAT_LAYOUT_' + str(layout).upper()
+
+
+def colorspace_map(colorspace):
+    return 'UTIL_FORMAT_COLORSPACE_' + str(colorspace).upper()
+
+
+colorspace_channels_map = {
+    'rgb': 'rgba',
+    'rgba': 'rgba',
+    'zs': 'zs',
+    'yuv': ['y1', 'y2', 'u', 'v'],
+    'dxt': []
+}
+
+
+kind_map = {
+    VOID:     "UTIL_FORMAT_TYPE_VOID",
+    UNSIGNED: "UTIL_FORMAT_TYPE_UNSIGNED",
+    SIGNED:   "UTIL_FORMAT_TYPE_SIGNED",
+    FIXED:    "UTIL_FORMAT_TYPE_FIXED",
+    FLOAT:    "UTIL_FORMAT_TYPE_FLOAT",
+}
+
+
+def bool_map(value):
+    if value:
+        return "TRUE"
+    else:
+        return "FALSE"
+
+
+swizzle_map = {
+    SWIZZLE_X:    "UTIL_FORMAT_SWIZZLE_X",
+    SWIZZLE_Y:    "UTIL_FORMAT_SWIZZLE_Y",
+    SWIZZLE_Z:    "UTIL_FORMAT_SWIZZLE_Z",
+    SWIZZLE_W:    "UTIL_FORMAT_SWIZZLE_W",
+    SWIZZLE_0:    "UTIL_FORMAT_SWIZZLE_0",
+    SWIZZLE_1:    "UTIL_FORMAT_SWIZZLE_1",
+    SWIZZLE_NONE: "UTIL_FORMAT_SWIZZLE_NONE",
+}
+
+
+def write_format_table(formats):
+    print '/* This file is autogenerated by u_format_table.py from u_format.csv. Do not edit directly. */'
+    print
+    # This will print the copyright message on the top of this file
+    print __doc__.strip()
+    print
+    print '#include "u_format.h"'
+    print
+    print 'const struct util_format_description'
+    print 'util_format_description_table[] = '
+    print "{"
+    for format in formats:
+        print "   {"
+        print "      %s," % (format.name,)
+        print "      \"%s\"," % (format.name,)
+        print "      {%u, %u, %u}, /* block */" % (format.block_width, format.block_height, format.block_size())
+        print "      %s," % (layout_map(format.layout),)
+        print "      {"
+        for i in range(4):
+            type = format.in_types[i]
+            if i < 3:
+                sep = ","
+            else:
+                sep = ""
+            print "         {%s, %s, %u}%s /* %s */" % (kind_map[type.kind], bool_map(type.norm), type.size, sep, "xyzw"[i])
+        print "      },"
+        print "      {"
+        for i in range(4):
+            swizzle = format.out_swizzle[i]
+            if i < 3:
+                sep = ","
+            else:
+                sep = ""
+            try:
+                comment = layout_channels_map[format.layout][i]
+            except:
+                comment = 'ignored'
+            print "         %s%s /* %s */" % (swizzle_map[swizzle], sep, comment)
+        print "      },"
+        print "      %s," % (colorspace_map(format.colorspace),)
+        print "   },"
+    print "   {"
+    print "      PIPE_FORMAT_NONE,"
+    print "      \"PIPE_FORMAT_NONE\","
+    print "      {0, 0, 0},"
+    print "      0,"
+    print "      {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}},"
+    print "      {0, 0, 0, 0},"
+    print "      0"
+    print "   },"
+    print "};"
+
+
+def main():
+
+    formats = []
+    for arg in sys.argv[1:]:
+        formats.extend(parse(arg))
+    write_format_table(formats)
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 833c0b8338..f06c0e463d 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -46,10 +46,6 @@
 #include "util/u_gen_mipmap.h"
 #include "util/u_simple_shaders.h"
 
-#include "tgsi/tgsi_build.h"
-#include "tgsi/tgsi_dump.h"
-#include "tgsi/tgsi_parse.h"
-
 #include "cso_cache/cso_context.h"
 
 
diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c
index 6da7353e25..3703718a62 100644
--- a/src/gallium/auxiliary/util/u_handle_table.c
+++ b/src/gallium/auxiliary/util/u_handle_table.c
@@ -87,6 +87,8 @@ handle_table_set_destroy(struct handle_table *ht,
                          void (*destroy)(void *object))
 {
    assert(ht);
+   if (!ht)
+      return;
    ht->destroy = destroy;
 }
 
@@ -155,7 +157,7 @@ handle_table_add(struct handle_table *ht,
    
    assert(ht);
    assert(object);
-   if(!object)
+   if(!object || !ht)
       return 0;
 
    /* linear search for an empty handle */
@@ -193,7 +195,7 @@ handle_table_set(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle)
+   if(!handle || !ht)
       return 0;
 
    assert(object);
@@ -222,7 +224,7 @@ handle_table_get(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle || handle > ht->size)
+   if(!handle || !ht || handle > ht->size)
       return NULL;
 
    object = ht->objects[handle - 1];
@@ -240,7 +242,7 @@ handle_table_remove(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle || handle > ht->size)
+   if(!handle || !ht || handle > ht->size)
       return;
 
    index = handle - 1;
@@ -283,6 +285,9 @@ handle_table_destroy(struct handle_table *ht)
    unsigned index;
    assert(ht);
 
+   if (!ht)
+      return;
+
    if(ht->destroy)
       for(index = 0; index < ht->size; ++index)
          handle_table_clear(ht, index);
diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c
index 2f83e318e4..8c2a8f454c 100644
--- a/src/gallium/auxiliary/util/u_hash_table.c
+++ b/src/gallium/auxiliary/util/u_hash_table.c
@@ -148,6 +148,8 @@ hash_table_set(struct hash_table *ht,
    struct cso_hash_iter iter;
 
    assert(ht);
+   if (!ht)
+      return PIPE_ERROR_BAD_INPUT;
 
    key_hash = ht->hash(key);
 
@@ -183,6 +185,8 @@ hash_table_get(struct hash_table *ht,
    struct hash_table_item *item;
 
    assert(ht);
+   if (!ht)
+      return NULL;
 
    key_hash = ht->hash(key);
 
@@ -203,6 +207,8 @@ hash_table_remove(struct hash_table *ht,
    struct hash_table_item *item;
 
    assert(ht);
+   if (!ht)
+      return;
 
    key_hash = ht->hash(key);
 
@@ -225,7 +231,9 @@ hash_table_clear(struct hash_table *ht)
    struct hash_table_item *item;
 
    assert(ht);
-   
+   if (!ht)
+      return;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
@@ -243,9 +251,11 @@ hash_table_foreach(struct hash_table *ht,
    struct cso_hash_iter iter;
    struct hash_table_item *item;
    enum pipe_error result;
-   
+
    assert(ht);
-   
+   if (!ht)
+      return PIPE_ERROR_BAD_INPUT;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_iter_data(iter);
@@ -264,9 +274,11 @@ hash_table_destroy(struct hash_table *ht)
 {
    struct cso_hash_iter iter;
    struct hash_table_item *item;
-   
+
    assert(ht);
-   
+   if (!ht)
+      return;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_iter_data(iter);
diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c
index 3f70809efd..508a2ee063 100644
--- a/src/gallium/auxiliary/util/u_keymap.c
+++ b/src/gallium/auxiliary/util/u_keymap.c
@@ -194,6 +194,8 @@ util_keymap_insert(struct keymap *map, const void *key,
    struct cso_hash_iter iter;
 
    assert(map);
+   if (!map)
+      return FALSE;
 
    key_hash = hash(key, map->key_size);
 
@@ -234,6 +236,8 @@ util_keymap_lookup(const struct keymap *map, const void *key)
    struct keymap_item *item;
 
    assert(map);
+   if (!map)
+      return NULL;
 
    key_hash = hash(key, map->key_size);
 
@@ -258,6 +262,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user)
    struct keymap_item *item;
 
    assert(map);
+   if (!map)
+      return;
 
    key_hash = hash(key, map->key_size);
 
@@ -267,6 +273,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user)
    
    item = hash_table_item(iter);
    assert(item);
+   if (!item)
+      return;
    map->delete_func(map, item->key, item->value, user);
    FREE(item->key);
    FREE(item);
@@ -288,7 +296,9 @@ util_keymap_remove_all(struct keymap *map, void *user)
    struct keymap_item *item;
 
    assert(map);
-   
+   if (!map)
+      return;
+
    iter = cso_hash_first_node(map->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct keymap_item *)
diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h
index e5003af01d..4c6c2bc00e 100644
--- a/src/gallium/auxiliary/util/u_math.h
+++ b/src/gallium/auxiliary/util/u_math.h
@@ -53,11 +53,11 @@ __inline double ceil(double val)
 {
    double ceil_val;
 
-   if((val - (long) val) == 0) {
+   if ((val - (long) val) == 0) {
       ceil_val = val;
    }
    else {
-      if(val > 0) {
+      if (val > 0) {
          ceil_val = (long) val + 1;
       }
       else {
@@ -73,11 +73,11 @@ __inline double floor(double val)
 {
    double floor_val;
 
-   if((val - (long) val) == 0) {
+   if ((val - (long) val) == 0) {
       floor_val = val;
    }
    else {
-      if(val > 0) {
+      if (val > 0) {
          floor_val = (long) val;
       }
       else {
@@ -189,7 +189,10 @@ static INLINE double log2( double x )
 extern float pow2_table[POW2_TABLE_SIZE];
 
 
-
+/**
+ * Initialize math module.  This should be called before using any
+ * other functions in this module.
+ */
 extern void
 util_init_math(void);
 
@@ -216,23 +219,24 @@ util_fast_exp2(float x)
    int32_t ipart;
    float fpart, mpart;
    union fi epart;
-   
+
    if(x > 129.00000f)
       return 3.402823466e+38f;
-   
-   if(x < -126.99999f)
+
+   if (x < -126.99999f)
       return 0.0f;
 
    ipart = (int32_t) x;
    fpart = x - (float) ipart;
-   
+
    /* same as
     *   epart.f = (float) (1 << ipart)
-    * but faster and without integer overflow for ipart > 31 */
+    * but faster and without integer overflow for ipart > 31
+    */
    epart.i = (ipart + 127 ) << 23;
-   
+
    mpart = pow2_table[POW2_TABLE_OFFSET + (int)(fpart * POW2_TABLE_SCALE)];
-   
+
    return epart.f * mpart;
 }
 
@@ -254,6 +258,9 @@ util_fast_exp(float x)
 extern float log2_table[LOG2_TABLE_SIZE];
 
 
+/**
+ * Fast approximation to log2(x).
+ */
 static INLINE float
 util_fast_log2(float x)
 {
@@ -267,6 +274,9 @@ util_fast_log2(float x)
 }
 
 
+/**
+ * Fast approximation to x^y.
+ */
 static INLINE float
 util_fast_pow(float x, float y)
 {
@@ -274,7 +284,6 @@ util_fast_pow(float x, float y)
 }
 
 
-
 /**
  * Floor(x), returned as int.
  */
@@ -284,8 +293,8 @@ util_ifloor(float f)
    int ai, bi;
    double af, bf;
    union fi u;
-   af = (3 << 22) + 0.5 + (double)f;
-   bf = (3 << 22) + 0.5 - (double)f;
+   af = (3 << 22) + 0.5 + (double) f;
+   bf = (3 << 22) + 0.5 - (double) f;
    u.f = (float) af;  ai = u.i;
    u.f = (float) bf;  bi = u.i;
    return (ai - bi) >> 1;
@@ -305,9 +314,9 @@ util_iround(float f)
 #elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)
    int r;
    _asm {
-	 fld f
-	 fistp r
-	}
+      fld f
+      fistp r
+   }
    return r;
 #else
    if (f >= 0.0f)
@@ -340,7 +349,7 @@ static INLINE
 unsigned long ffs( unsigned long u )
 {
    unsigned long i;
-   if(_BitScanForward(&i, u))
+   if (_BitScanForward(&i, u))
       return i + 1;
    else
       return 0;
@@ -351,7 +360,7 @@ unsigned ffs( unsigned u )
 {
    unsigned i;
 
-   if( u == 0 ) {
+   if (u == 0) {
       return 0;
    }
 
@@ -365,6 +374,22 @@ unsigned ffs( unsigned u )
 #define ffs __builtin_ffs
 #endif
 
+#ifdef __MINGW32__
+#define ffs __builtin_ffs
+#endif
+
+
+/* Could also binary search for the highest bit.
+ */
+static INLINE unsigned
+util_unsigned_logbase2(unsigned n)
+{
+   unsigned log2 = 0;
+   while (n >>= 1)
+      ++log2;
+   return log2;
+}
+
 
 /**
  * Return float bits.
@@ -378,7 +403,10 @@ fui( float f )
 }
 
 
-
+/**
+ * Convert ubyte to float in [0, 1].
+ * XXX a 256-entry lookup table would be slightly faster.
+ */
 static INLINE float
 ubyte_to_float(ubyte ub)
 {
@@ -409,7 +437,23 @@ float_to_ubyte(float f)
 }
 
 
+/**
+ * Calc log base 2
+ */
+static INLINE unsigned
+util_logbase2(unsigned n)
+{
+   unsigned log2 = 0;
+   while (n >>= 1)
+      ++log2;
+   return log2;
+}
+
 
+/**
+ * Clamp X to [MIN, MAX].
+ * This is a macro to allow float, int, uint, etc. types.
+ */
 #define CLAMP( X, MIN, MAX )  ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) )
 
 #define MIN2( A, B )   ( (A)<(B) ? (A) : (B) )
@@ -422,6 +466,11 @@ align(int value, int alignment)
    return (value + alignment - 1) & ~(alignment - 1);
 }
 
+static INLINE unsigned
+minify(unsigned value)
+{
+    return MAX2(1, value >> 1);
+}
 
 #ifndef COPY_4V
 #define COPY_4V( DST, SRC )         \
diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h
index 0b18d043ad..c3f8c91833 100644
--- a/src/gallium/auxiliary/util/u_memory.h
+++ b/src/gallium/auxiliary/util/u_memory.h
@@ -100,8 +100,14 @@ ExFreePool(void *P);
 #define MALLOC( SIZE )  malloc( SIZE )
 #define CALLOC( COUNT, SIZE )   calloc( COUNT, SIZE )
 #define FREE( PTR )  free( PTR )
-#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE )  realloc( OLDPTR, NEWSIZE )
 
+static INLINE void *
+_REALLOC( void *old_ptr, unsigned old_size, unsigned new_size )
+{
+   (void) old_size;
+   return realloc(old_ptr, new_size);
+}
+#define REALLOC( a, b, c ) _REALLOC( a, b, c )
 #endif
 
 
diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c
index 151a480d34..4b75d4ba1d 100644
--- a/src/gallium/auxiliary/util/u_mm.c
+++ b/src/gallium/auxiliary/util/u_mm.c
@@ -33,30 +33,32 @@
 void
 u_mmDumpMemInfo(const struct mem_block *heap)
 {
-   debug_printf("Memory heap %p:\n", (void *)heap);
+   debug_printf("Memory heap %p:\n", (void *) heap);
    if (heap == 0) {
       debug_printf("  heap == 0\n");
-   } else {
+   }
+   else {
       const struct mem_block *p;
 
-      for(p = heap->next; p != heap; p = p->next) {
-	 debug_printf("  Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size,
-		 p->free ? 'F':'.',
-		 p->reserved ? 'R':'.');
+      for (p = heap->next; p != heap; p = p->next) {
+	 debug_printf("  Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size,
+                      p->free ? 'F':'.',
+                      p->reserved ? 'R':'.');
       }
 
       debug_printf("\nFree list:\n");
 
-      for(p = heap->next_free; p != heap; p = p->next_free) {
-	 debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size,
-		 p->free ? 'F':'.',
-		 p->reserved ? 'R':'.');
+      for (p = heap->next_free; p != heap; p = p->next_free) {
+	 debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size,
+                      p->free ? 'F':'.',
+                      p->reserved ? 'R':'.');
       }
 
    }
    debug_printf("End of memory blocks\n");
 }
 
+
 struct mem_block *
 u_mmInit(int ofs, int size)
 {
diff --git a/src/gallium/auxiliary/util/u_mm.h b/src/gallium/auxiliary/util/u_mm.h
index ce20e48763..6b158aae6e 100644
--- a/src/gallium/auxiliary/util/u_mm.h
+++ b/src/gallium/auxiliary/util/u_mm.h
@@ -84,7 +84,7 @@ extern struct mem_block *u_mmFindBlock(struct mem_block *heap, int start);
 extern void u_mmDestroy(struct mem_block *mmInit);
 
 /**
- * For debuging purpose.
+ * For debugging purposes.
  */
 extern void u_mmDumpMemInfo(const struct mem_block *mmInit);
 
diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c
new file mode 100644
index 0000000000..bc4b758406
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_network.c
@@ -0,0 +1,188 @@
+
+#include "pipe/p_compiler.h"
+#include "util/u_network.h"
+#include "util/u_debug.h"
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  include <winsock2.h>
+#  include <windows.h>
+#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU)
+#  include <sys/socket.h>
+#  include <netinet/in.h>
+#  include <unistd.h>
+#  include <fcntl.h>
+#  include <netdb.h>
+#else
+#  warning "No socket implementation"
+#endif
+
+boolean
+u_socket_init()
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   WORD wVersionRequested;
+   WSADATA wsaData;
+   int err;
+
+   /* Use the MAKEWORD(lowbyte, highbyte) macro declared in Windef.h */
+   wVersionRequested = MAKEWORD(1, 1);
+
+   err = WSAStartup(wVersionRequested, &wsaData);
+   if (err != 0) {
+      debug_printf("WSAStartup failed with error: %d\n", err);
+      return FALSE;
+   }
+   return TRUE;
+#elif defined(PIPE_HAVE_SOCKETS)
+   return TRUE;
+#else
+   return FALSE;
+#endif
+}
+
+void
+u_socket_stop()
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   WSACleanup();
+#endif
+}
+
+void
+u_socket_close(int s)
+{
+   if (s < 0)
+      return;
+
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU)
+   shutdown(s, SHUT_RDWR);
+   close(s);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   shutdown(s, SD_BOTH);
+   closesocket(s);
+#else
+   assert(0);
+#endif
+}
+
+int u_socket_accept(int s)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return accept(s, NULL, NULL);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_send(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return send(s, data, size, 0);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_peek(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return recv(s, data, size, MSG_PEEK);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_recv(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return recv(s, data, size, 0);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_connect(const char *hostname, uint16_t port)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   int s;
+   struct sockaddr_in sa;
+   struct hostent *host = NULL;
+
+   memset(&sa, 0, sizeof(struct sockaddr_in));
+   host = gethostbyname(hostname);
+   if (!host)
+      return -1;
+
+   memcpy((char *)&sa.sin_addr,host->h_addr,host->h_length);
+   sa.sin_family= host->h_addrtype;
+   sa.sin_port = htons(port);
+
+   s = socket(host->h_addrtype, SOCK_STREAM, IPPROTO_TCP);
+   if (s < 0)
+      return -1;
+
+   if (connect(s, (struct sockaddr *)&sa, sizeof(sa))) {
+      u_socket_close(s);
+      return -1;
+   }
+
+   return s;
+#else
+   assert(0);
+   return -1;
+#endif
+}
+
+int
+u_socket_listen_on_port(uint16_t portnum)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   int s;
+   struct sockaddr_in sa;
+   memset(&sa, 0, sizeof(struct sockaddr_in));
+
+   sa.sin_family = AF_INET;
+   sa.sin_port = htons(portnum);
+
+   s = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+   if (s < 0)
+      return -1;
+
+   if (bind(s, (struct sockaddr *)&sa, sizeof(struct sockaddr_in)) == -1) {
+      u_socket_close(s);
+      return -1;
+   }
+
+   listen(s, 0);
+
+   return s;
+#else
+   assert(0);
+   return -1;
+#endif
+}
+
+void
+u_socket_block(int s, boolean block)
+{
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU)
+   int old = fcntl(s, F_GETFL, 0);
+   if (old == -1)
+      return;
+
+   /* TODO obey block */
+   if (block)
+      fcntl(s, F_SETFL, old & ~O_NONBLOCK);
+   else
+      fcntl(s, F_SETFL, old | O_NONBLOCK);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   u_long iMode = block ? 0 : 1;
+   ioctlsocket(s, FIONBIO, &iMode);
+#else
+   assert(0);
+#endif
+}
diff --git a/src/gallium/auxiliary/util/u_network.h b/src/gallium/auxiliary/util/u_network.h
new file mode 100644
index 0000000000..8c778f492c
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_network.h
@@ -0,0 +1,24 @@
+
+#ifndef _U_NETWORK_H_
+#define _U_NETWORK_H_
+
+#include "pipe/p_compiler.h"
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  define PIPE_HAVE_SOCKETS
+#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU)
+#  define PIPE_HAVE_SOCKETS
+#endif
+
+boolean u_socket_init(void);
+void u_socket_stop(void);
+void u_socket_close(int s);
+int u_socket_listen_on_port(uint16_t portnum);
+int u_socket_accept(int s);
+int u_socket_connect(const char *host, uint16_t port);
+int u_socket_send(int s, void *data, size_t size);
+int u_socket_peek(int s, void *data, size_t size);
+int u_socket_recv(int s, void *data, size_t size);
+void u_socket_block(int s, boolean block);
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
index d7c3995dbf..a9b533eea7 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -119,7 +119,7 @@ static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr )
 }
 
 
-static INLINE boolean u_reduced_prim( unsigned pipe_prim )
+static INLINE unsigned u_reduced_prim( unsigned pipe_prim )
 {
    switch (pipe_prim) {
    case PIPE_PRIM_POINTS:
diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c
index 74259d453b..9866b6fc8a 100644
--- a/src/gallium/auxiliary/util/u_rect.c
+++ b/src/gallium/auxiliary/util/u_rect.c
@@ -43,7 +43,7 @@
  * src_pitch may be negative to do vertical flip of pixels from source.
  */
 void
-pipe_copy_rect(ubyte * dst,
+util_copy_rect(ubyte * dst,
                const struct pipe_format_block *block,
                unsigned dst_stride,
                unsigned dst_x,
@@ -91,7 +91,7 @@ pipe_copy_rect(ubyte * dst,
 }
 
 void
-pipe_fill_rect(ubyte * dst,
+util_fill_rect(ubyte * dst,
                const struct pipe_format_block *block,
                unsigned dst_stride,
                unsigned dst_x,
@@ -204,7 +204,7 @@ util_surface_copy(struct pipe_context *pipe,
 
    if (src_map && dst_map) {
       /* If do_flip, invert src_y position and pass negative src stride */
-      pipe_copy_rect(dst_map,
+      util_copy_rect(dst_map,
                      &dst_trans->block,
                      dst_trans->stride,
                      0, 0,
@@ -263,7 +263,7 @@ util_surface_fill(struct pipe_context *pipe,
       case 1:
       case 2:
       case 4:
-         pipe_fill_rect(dst_map, &dst_trans->block, dst_trans->stride,
+         util_fill_rect(dst_map, &dst_trans->block, dst_trans->stride,
                         0, 0, width, height, value);
          break;
       case 8:
diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h
index 59e842e16d..daa50834d3 100644
--- a/src/gallium/auxiliary/util/u_rect.h
+++ b/src/gallium/auxiliary/util/u_rect.h
@@ -42,13 +42,13 @@ struct pipe_surface;
 
 
 extern void
-pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block,
+util_copy_rect(ubyte * dst, const struct pipe_format_block *block,
                unsigned dst_stride, unsigned dst_x, unsigned dst_y,
                unsigned width, unsigned height, const ubyte * src,
                int src_stride, unsigned src_x, int src_y);
 
 extern void
-pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block,
+util_fill_rect(ubyte * dst, const struct pipe_format_block *block,
                unsigned dst_stride, unsigned dst_x, unsigned dst_y,
                unsigned width, unsigned height, uint32_t value);
 
diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c
index 8114b53cd0..f01296b40f 100644
--- a/src/gallium/auxiliary/util/u_simple_screen.c
+++ b/src/gallium/auxiliary/util/u_simple_screen.c
@@ -65,12 +65,13 @@ pass_surface_buffer_create(struct pipe_screen *screen,
                            unsigned width, unsigned height,
                            enum pipe_format format,
                            unsigned usage,
+                           unsigned tex_usage,
                            unsigned *stride)
 {
    struct pipe_buffer *buffer =
       screen->winsys->surface_buffer_create(screen->winsys,
                                                 width, height,
-                                                format, usage, stride);
+                                                format, usage, tex_usage, stride);
 
    buffer->screen = screen;
 
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c
index acc5b83c62..ab754296fa 100644
--- a/src/gallium/auxiliary/util/u_simple_shaders.c
+++ b/src/gallium/auxiliary/util/u_simple_shaders.c
@@ -42,9 +42,7 @@
 #include "util/u_memory.h"
 #include "util/u_simple_shaders.h"
 
-#include "tgsi/tgsi_build.h"
-#include "tgsi/tgsi_dump.h"
-#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_ureg.h"
 
 
 
@@ -58,93 +56,31 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
                                     const uint *semantic_indexes)
                                     
 {
-   struct pipe_shader_state shader;
-   struct tgsi_token tokens[100];
-   struct tgsi_header *header;
-   struct tgsi_processor *processor;
-   struct tgsi_full_declaration decl;
-   struct tgsi_full_instruction inst;
-   const uint procType = TGSI_PROCESSOR_VERTEX;
-   uint ti, i;
+   struct ureg_program *ureg;
+   uint i;
 
-   /* shader header
-    */
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+   ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
+   if (ureg == NULL)
+      return NULL;
 
-   header = (struct tgsi_header *) &tokens[1];
-   *header = tgsi_build_header();
-
-   processor = (struct tgsi_processor *) &tokens[2];
-   *processor = tgsi_build_processor( procType, header );
-
-   ti = 3;
-
-   /* declare inputs */
-   for (i = 0; i < num_attribs; i++) {
-      decl = tgsi_default_full_declaration();
-      decl.Declaration.File = TGSI_FILE_INPUT;
-
-      decl.Declaration.Semantic = 1;
-      decl.Semantic.SemanticName = semantic_names[i];
-      decl.Semantic.SemanticIndex = semantic_indexes[i];
-
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = i;
-      ti += tgsi_build_full_declaration(&decl,
-                                        &tokens[ti],
-                                        header,
-                                        Elements(tokens) - ti);
-   }
-
-   /* declare outputs */
    for (i = 0; i < num_attribs; i++) {
-      decl = tgsi_default_full_declaration();
-      decl.Declaration.File = TGSI_FILE_OUTPUT;
-      decl.Declaration.Semantic = 1;
-      decl.Semantic.SemanticName = semantic_names[i];
-      decl.Semantic.SemanticIndex = semantic_indexes[i];
-      decl.DeclarationRange.First = 
-         decl.DeclarationRange.Last = i;
-      ti += tgsi_build_full_declaration(&decl,
-                                        &tokens[ti],
-                                        header,
-                                        Elements(tokens) - ti);
+      struct ureg_src src;
+      struct ureg_dst dst;
+
+      src = ureg_DECL_vs_input( ureg,
+                                semantic_names[i],
+                                semantic_indexes[i]);
+      
+      dst = ureg_DECL_output( ureg,
+                              semantic_names[i],
+                              semantic_indexes[i]);
+      
+      ureg_MOV( ureg, dst, src );
    }
 
-   /* emit MOV instructions */
-   for (i = 0; i < num_attribs; i++) {
-      /* MOVE out[i], in[i]; */
-      inst = tgsi_default_full_instruction();
-      inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-      inst.Instruction.NumDstRegs = 1;
-      inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      inst.FullDstRegisters[0].DstRegister.Index = i;
-      inst.Instruction.NumSrcRegs = 1;
-      inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      inst.FullSrcRegisters[0].SrcRegister.Index = i;
-      ti += tgsi_build_full_instruction(&inst,
-                                        &tokens[ti],
-                                        header,
-                                        Elements(tokens) - ti );
-   }
-
-   /* END instruction */
-   inst = tgsi_default_full_instruction();
-   inst.Instruction.Opcode = TGSI_OPCODE_END;
-   inst.Instruction.NumDstRegs = 0;
-   inst.Instruction.NumSrcRegs = 0;
-   ti += tgsi_build_full_instruction(&inst,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti );
+   ureg_END( ureg );
 
-#if 0 /*debug*/
-   tgsi_dump(tokens, 0);
-#endif
-
-   shader.tokens = tokens;
-
-   return pipe->create_vs_state(pipe, &shader);
+   return ureg_create_shader_and_destroy( ureg, pipe );
 }
 
 
@@ -161,130 +97,29 @@ void *
 util_make_fragment_tex_shader_writemask(struct pipe_context *pipe,
                                         unsigned writemask )
 {
-   struct pipe_shader_state shader;
-   struct tgsi_token tokens[100];
-   struct tgsi_header *header;
-   struct tgsi_processor *processor;
-   struct tgsi_full_declaration decl;
-   struct tgsi_full_instruction inst;
-   const uint procType = TGSI_PROCESSOR_FRAGMENT;
-   uint ti;
-
-   /* shader header
-    */
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-
-   header = (struct tgsi_header *) &tokens[1];
-   *header = tgsi_build_header();
-
-   processor = (struct tgsi_processor *) &tokens[2];
-   *processor = tgsi_build_processor( procType, header );
-
-   ti = 3;
-
-   /* declare TEX[0] input */
-   decl = tgsi_default_full_declaration();
-   decl.Declaration.File = TGSI_FILE_INPUT;
-   /* XXX this could be linear... */
-   decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
-   decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-   decl.Semantic.SemanticIndex = 0;
-   decl.DeclarationRange.First = 
-   decl.DeclarationRange.Last = 0;
-   ti += tgsi_build_full_declaration(&decl,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti);
-
-   /* declare color[0] output */
-   decl = tgsi_default_full_declaration();
-   decl.Declaration.File = TGSI_FILE_OUTPUT;
-   decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-   decl.Semantic.SemanticIndex = 0;
-   decl.DeclarationRange.First = 
-   decl.DeclarationRange.Last = 0;
-   ti += tgsi_build_full_declaration(&decl,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti);
-
-   /* declare sampler */
-   decl = tgsi_default_full_declaration();
-   decl.Declaration.File = TGSI_FILE_SAMPLER;
-   decl.DeclarationRange.First = 
-   decl.DeclarationRange.Last = 0;
-   ti += tgsi_build_full_declaration(&decl,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti);
-
-
-   if (writemask != TGSI_WRITEMASK_XYZW) {
-      struct tgsi_full_immediate imm;
-      static const float value[4] = { 0, 0, 0, 1 };
-
-      imm = tgsi_default_full_immediate();
-      imm.Immediate.NrTokens += 4;
-      imm.Immediate.DataType = TGSI_IMM_FLOAT32;
-      imm.u.Pointer = value;
-
-      ti += tgsi_build_full_immediate(&imm,
-                                      &tokens[ti],
-                                      header,
-                                      Elements(tokens) - ti );
-
-      /* MOV instruction */
-      inst = tgsi_default_full_instruction();
-      inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-      inst.Instruction.NumDstRegs = 1;
-      inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      inst.FullDstRegisters[0].DstRegister.Index = 0;
-      inst.Instruction.NumSrcRegs = 1;
-      inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_IMMEDIATE;
-      inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-      ti += tgsi_build_full_instruction(&inst,
-                                        &tokens[ti],
-                                        header,
-                                        Elements(tokens) - ti );
-   }
+   struct ureg_program *ureg;
+   struct ureg_src sampler;
+   struct ureg_src tex;
+   struct ureg_dst out;
 
-   /* TEX instruction */
-   inst = tgsi_default_full_instruction();
-   inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-   inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-   inst.FullDstRegisters[0].DstRegister.Index = 0;
-   inst.FullDstRegisters[0].DstRegister.WriteMask = writemask;
-   inst.Instruction.NumSrcRegs = 2;
-   inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-   inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-   inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-   inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-   inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-   ti += tgsi_build_full_instruction(&inst,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti );
-
-   /* END instruction */
-   inst = tgsi_default_full_instruction();
-   inst.Instruction.Opcode = TGSI_OPCODE_END;
-   inst.Instruction.NumDstRegs = 0;
-   inst.Instruction.NumSrcRegs = 0;
-   ti += tgsi_build_full_instruction(&inst,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti );
-
-#if 0 /*debug*/
-   tgsi_dump(tokens, 0);
-#endif
-
-   shader.tokens = tokens;
-
-   return pipe->create_fs_state(pipe, &shader);
+   ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
+   if (ureg == NULL)
+      return NULL;
+   
+   sampler = ureg_DECL_sampler( ureg, 0 );
+
+   tex = ureg_DECL_fs_input( ureg, 
+                             TGSI_SEMANTIC_GENERIC, 0, 
+                             TGSI_INTERPOLATE_PERSPECTIVE );
+
+   out = ureg_DECL_output( ureg, 
+                           TGSI_SEMANTIC_COLOR,
+                           0 );
+
+   ureg_TEX( ureg, out, TGSI_TEXTURE_2D, tex, sampler );
+   ureg_END( ureg );
+
+   return ureg_create_shader_and_destroy( ureg, pipe );
 }
 
 void *
@@ -296,95 +131,29 @@ util_make_fragment_tex_shader(struct pipe_context *pipe )
 
 
 
-
-
 /**
  * Make simple fragment color pass-through shader.
  */
 void *
 util_make_fragment_passthrough_shader(struct pipe_context *pipe)
 {
-   struct pipe_shader_state shader;
-   struct tgsi_token tokens[40];
-   struct tgsi_header *header;
-   struct tgsi_processor *processor;
-   struct tgsi_full_declaration decl;
-   struct tgsi_full_instruction inst;
-   const uint procType = TGSI_PROCESSOR_FRAGMENT;
-   uint ti;
-
-   /* shader header
-    */
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-
-   header = (struct tgsi_header *) &tokens[1];
-   *header = tgsi_build_header();
-
-   processor = (struct tgsi_processor *) &tokens[2];
-   *processor = tgsi_build_processor( procType, header );
-
-   ti = 3;
-
-   /* declare input */
-   decl = tgsi_default_full_declaration();
-   decl.Declaration.File = TGSI_FILE_INPUT;
-   decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-   decl.Semantic.SemanticIndex = 0;
-   decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = 0;
-   ti += tgsi_build_full_declaration(&decl,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti);
-
-   /* declare output */
-   decl = tgsi_default_full_declaration();
-   decl.Declaration.File = TGSI_FILE_OUTPUT;
-   decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-   decl.Semantic.SemanticIndex = 0;
-   decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = 0;
-   ti += tgsi_build_full_declaration(&decl,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti);
-
-
-   /* MOVE out[0], in[0]; */
-   inst = tgsi_default_full_instruction();
-   inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-   inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-   inst.FullDstRegisters[0].DstRegister.Index = 0;
-   inst.Instruction.NumSrcRegs = 1;
-   inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-   inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-   ti += tgsi_build_full_instruction(&inst,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti );
-
-   /* END instruction */
-   inst = tgsi_default_full_instruction();
-   inst.Instruction.Opcode = TGSI_OPCODE_END;
-   inst.Instruction.NumDstRegs = 0;
-   inst.Instruction.NumSrcRegs = 0;
-   ti += tgsi_build_full_instruction(&inst,
-                                     &tokens[ti],
-                                     header,
-                                     Elements(tokens) - ti );
-
-   assert(ti < Elements(tokens));
-
-#if 0 /*debug*/
-   tgsi_dump(tokens, 0);
-#endif
-
-   shader.tokens = tokens;
-
-   return pipe->create_fs_state(pipe, &shader);
+   struct ureg_program *ureg;
+   struct ureg_src src;
+   struct ureg_dst dst;
+
+   ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
+   if (ureg == NULL)
+      return NULL;
+
+   src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0, 
+                             TGSI_INTERPOLATE_PERSPECTIVE );
+
+   dst = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 );
+
+   ureg_MOV( ureg, dst, src );
+   ureg_END( ureg );
+
+   return ureg_create_shader_and_destroy( ureg, pipe );
 }
 
 
diff --git a/src/gallium/auxiliary/util/u_stream_stdc.c b/src/gallium/auxiliary/util/u_stream_stdc.c
index d8f648e5dd..5cd05b2904 100644
--- a/src/gallium/auxiliary/util/u_stream_stdc.c
+++ b/src/gallium/auxiliary/util/u_stream_stdc.c
@@ -32,7 +32,7 @@
 
 #include "pipe/p_config.h"
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
 
 #include <stdio.h>
 
diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
index aa16fe771f..1235a67d26 100644
--- a/src/gallium/auxiliary/util/u_tile.c
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -62,7 +62,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt,
    if(!src)
       return;
 
-   pipe_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y);
+   util_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y);
 
    screen->transfer_unmap(screen, pt);
 }
@@ -90,7 +90,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt,
    if(!dst)
       return;
 
-   pipe_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0);
+   util_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0);
 
    screen->transfer_unmap(screen, pt);
 }
diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c
index 5268cbf79c..b958a98635 100644
--- a/src/gallium/auxiliary/util/u_time.c
+++ b/src/gallium/auxiliary/util/u_time.c
@@ -35,7 +35,7 @@
 
 #include "pipe/p_config.h"
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
 #include <sys/time.h>
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
 #include <windows.h>
@@ -77,7 +77,7 @@ util_time_get_frequency(void)
 void 
 util_time_get(struct util_time *t)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    gettimeofday(&t->tv, NULL);
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
    LONGLONG temp;
@@ -102,7 +102,7 @@ util_time_add(const struct util_time *t1,
               int64_t usecs,
               struct util_time *t2)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000;
    t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
@@ -124,7 +124,7 @@ int64_t
 util_time_diff(const struct util_time *t1, 
                const struct util_time *t2)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    return (t2->tv.tv_usec - t1->tv.tv_usec) + 
           (t2->tv.tv_sec - t1->tv.tv_sec)*1000000;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
@@ -144,7 +144,7 @@ util_time_micros( void )
    
    util_time_get(&t1);
    
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    return t1.tv.tv_usec + t1.tv.tv_sec*1000000LL;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
    util_time_get_frequency();
@@ -166,7 +166,7 @@ static INLINE int
 util_time_compare(const struct util_time *t1, 
                   const struct util_time *t2)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    if (t1->tv.tv_sec < t2->tv.tv_sec)
       return -1;
    else if(t1->tv.tv_sec > t2->tv.tv_sec)
diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h
index 6bca6077a2..a6189a247b 100644
--- a/src/gallium/auxiliary/util/u_time.h
+++ b/src/gallium/auxiliary/util/u_time.h
@@ -38,11 +38,16 @@
 
 #include "pipe/p_config.h"
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE)
 #include <time.h> /* timeval */
 #include <unistd.h> /* usleep */
 #endif
 
+#if defined(PIPE_OS_HAIKU)
+#include <sys/time.h> /* timeval */
+#include <unistd.h>
+#endif
+
 #include "pipe/p_compiler.h"
 
 
@@ -58,7 +63,7 @@ extern "C" {
  */
 struct util_time 
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    struct timeval tv;
 #else
    int64_t counter;
@@ -89,7 +94,7 @@ util_time_timeout(const struct util_time *start,
                   const struct util_time *end,
                   const struct util_time *curr);
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
 #define util_time_sleep usleep
 #else
 void
diff --git a/src/gallium/auxiliary/util/u_timed_winsys.c b/src/gallium/auxiliary/util/u_timed_winsys.c
index 77b2a3a1c8..178acdca4d 100644
--- a/src/gallium/auxiliary/util/u_timed_winsys.c
+++ b/src/gallium/auxiliary/util/u_timed_winsys.c
@@ -212,13 +212,14 @@ timed_surface_buffer_create(struct pipe_winsys *winsys,
                               unsigned width, unsigned height,
                               enum pipe_format format, 
                               unsigned usage,
+                              unsigned tex_usage,
                               unsigned *stride)
 {
    struct pipe_winsys *backend = timed_winsys(winsys)->backend;
    uint64_t start = time_start();
 
    struct pipe_buffer *ret = backend->surface_buffer_create( backend, width, height, 
-                                                             format, usage, stride );
+                                                             format, usage, tex_usage, stride );
 
    time_finish(winsys, start, 7, __FUNCTION__);
    
diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h
index 1f6860da11..d5f5c7bbba 100644
--- a/src/gallium/drivers/cell/common.h
+++ b/src/gallium/drivers/cell/common.h
@@ -327,7 +327,7 @@ struct cell_command_sampler
    opcode_t opcode;         /**< CELL_CMD_STATE_SAMPLER */
    uint unit;
    struct pipe_sampler_state state;
-   uint32_t pad_[1];
+   uint32_t pad_[2];
 };
 
 
diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fp.c b/src/gallium/drivers/cell/ppu/cell_gen_fp.c
index 5a889a6119..312621fd53 100644
--- a/src/gallium/drivers/cell/ppu/cell_gen_fp.c
+++ b/src/gallium/drivers/cell/ppu/cell_gen_fp.c
@@ -1767,7 +1767,7 @@ emit_instruction(struct codegen *gen,
    case TGSI_OPCODE_MAD:
       return emit_MAD(gen, inst);
    case TGSI_OPCODE_LERP:
-      return emit_LERP(gen, inst);
+      return emit_LRP(gen, inst);
    case TGSI_OPCODE_DP3:
       return emit_DP3(gen, inst);
    case TGSI_OPCODE_DP4:
@@ -1810,9 +1810,9 @@ emit_instruction(struct codegen *gen,
       return emit_function_call(gen, inst, "spu_sin", 1, TRUE);
    case TGSI_OPCODE_POW:
       return emit_function_call(gen, inst, "spu_pow", 2, TRUE);
-   case TGSI_OPCODE_EXPBASE2:
+   case TGSI_OPCODE_EX2:
       return emit_function_call(gen, inst, "spu_exp2", 1, TRUE);
-   case TGSI_OPCODE_LOGBASE2:
+   case TGSI_OPCODE_LG2:
       return emit_function_call(gen, inst, "spu_log2", 1, TRUE);
    case TGSI_OPCODE_TEX:
       /* fall-through for now */
@@ -1834,9 +1834,9 @@ emit_instruction(struct codegen *gen,
    case TGSI_OPCODE_ENDIF:
       return emit_ENDIF(gen, inst);
 
-   case TGSI_OPCODE_BGNLOOP2:
+   case TGSI_OPCODE_BGNLOOP:
       return emit_BGNLOOP(gen, inst);
-   case TGSI_OPCODE_ENDLOOP2:
+   case TGSI_OPCODE_ENDLOOP:
       return emit_ENDLOOP(gen, inst);
    case TGSI_OPCODE_BRK:
       return emit_BRK(gen, inst);
@@ -1875,9 +1875,9 @@ emit_immediate(struct codegen *gen, const struct tgsi_full_immediate *immed)
    assert(gen->num_imm < MAX_TEMPS);
 
    for (ch = 0; ch < 4; ch++) {
-      float val = immed->u.ImmediateFloat32[ch].Float;
+      float val = immed->u[ch].Float;
 
-      if (ch > 0 && val == immed->u.ImmediateFloat32[ch - 1].Float) {
+      if (ch > 0 && val == immed->u[ch - 1].Float) {
          /* re-use previous register */
          gen->imm_regs[gen->num_imm][ch] = gen->imm_regs[gen->num_imm][ch - 1];
       }
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index e26594448f..6a63a0e6ce 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -44,13 +44,6 @@
 
 
 
-static unsigned
-minify(unsigned d)
-{
-   return MAX2(1, d>>1);
-}
-
-
 static void
 cell_texture_layout(struct cell_texture *ct)
 {
@@ -424,7 +417,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer)
    if (!ctrans->map)
       return NULL; /* out of memory */
 
-   if (transfer->usage & PIPE_TRANSFER_READ) {
+   if (transfer->usage == PIPE_TRANSFER_READ ||
+       transfer->usage == PIPE_TRANSFER_READ_WRITE) {
       /* need to untwiddle the texture to make a linear version */
       const uint bpp = pf_get_size(ct->base.format);
       if (bpp == 4) {
@@ -465,7 +459,8 @@ cell_transfer_unmap(struct pipe_screen *screen,
                                    PIPE_BUFFER_USAGE_CPU_READ);
    }
 
-   if (transfer->usage & PIPE_TRANSFER_WRITE) {
+   if (transfer->usage == PIPE_TRANSFER_WRITE ||
+       transfer->usage == PIPE_TRANSFER_READ_WRITE) {
       /* The user wrote new texture data into the mapped buffer.
        * We need to convert the new linear data into the twiddled/tiled format.
        */
diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c
index e27df2dfb3..0eaae2e451 100644
--- a/src/gallium/drivers/cell/spu/spu_exec.c
+++ b/src/gallium/drivers/cell/spu/spu_exec.c
@@ -952,7 +952,6 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_RCP:
-   /* TGSI_OPCODE_RECIP */
       FETCH( &r[0], 0, CHAN_X );
       r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q);
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
@@ -961,7 +960,6 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_RSQ:
-   /* TGSI_OPCODE_RECIPSQRT */
       FETCH( &r[0], 0, CHAN_X );
       r[0].q = micro_sqrt(r[0].q);
       r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q);
@@ -1115,7 +1113,6 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_MAD:
-   /* TGSI_OPCODE_MADD */
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          FETCH( &r[1], 1, chan_index );
@@ -1136,8 +1133,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LERP:
-   /* TGSI_OPCODE_LRP */
+   case TGSI_OPCODE_LRP:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
@@ -1154,25 +1150,11 @@ exec_instruction(
       ASSERT (0);
       break;
 
-   case TGSI_OPCODE_CND0:
+   case TGSI_OPCODE_DP2A:
       ASSERT (0);
       break;
 
-   case TGSI_OPCODE_DOT2ADD:
-      /* TGSI_OPCODE_DP2A */
-      ASSERT (0);
-      break;
-
-   case TGSI_OPCODE_INDEX:
-      ASSERT (0);
-      break;
-
-   case TGSI_OPCODE_NEGATE:
-      ASSERT (0);
-      break;
-
-   case TGSI_OPCODE_FRAC:
-   /* TGSI_OPCODE_FRC */
+   case TGSI_OPCODE_FRC:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          r[0].q = micro_frc(r[0].q);
@@ -1184,8 +1166,7 @@ exec_instruction(
       ASSERT (0);
       break;
 
-   case TGSI_OPCODE_FLOOR:
-   /* TGSI_OPCODE_FLR */
+   case TGSI_OPCODE_FLR:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          r[0].q = micro_flr(r[0].q);
@@ -1201,8 +1182,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_EXPBASE2:
-    /* TGSI_OPCODE_EX2 */
+   case TGSI_OPCODE_EX2:
       FETCH(&r[0], 0, CHAN_X);
 
       r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q);
@@ -1212,8 +1192,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_LOGBASE2:
-   /* TGSI_OPCODE_LG2 */
+   case TGSI_OPCODE_LG2:
       FETCH( &r[0], 0, CHAN_X );
       r[0].q = micro_lg2(r[0].q);
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
@@ -1221,8 +1200,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_POWER:
-      /* TGSI_OPCODE_POW */
+   case TGSI_OPCODE_POW:
       FETCH(&r[0], 0, CHAN_X);
       FETCH(&r[1], 1, CHAN_X);
 
@@ -1233,7 +1211,7 @@ exec_instruction(
       }
       break;
 
-   case TGSI_OPCODE_CROSSPRODUCT:
+   case TGSI_OPCODE_XPD:
       /* TGSI_OPCODE_XPD */
       FETCH(&r[0], 0, CHAN_Y);
       FETCH(&r[1], 1, CHAN_Z);
@@ -1275,10 +1253,6 @@ exec_instruction(
       }
       break;
 
-    case TGSI_OPCODE_MULTIPLYMATRIX:
-       ASSERT (0);
-       break;
-
     case TGSI_OPCODE_ABS:
        FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
           FETCH(&r[0], 0, chan_index);
@@ -1780,9 +1754,9 @@ exec_instruction(
       mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
       break;
 
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
       /* fall-through (for now) */
-   case TGSI_OPCODE_BGNLOOP2:
+   case TGSI_OPCODE_BGNLOOP:
       /* push LoopMask and ContMasks */
       ASSERT(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
       mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
@@ -1790,9 +1764,9 @@ exec_instruction(
       mach->ContStack[mach->ContStackTop++] = mach->ContMask;
       break;
 
-   case TGSI_OPCODE_ENDLOOP:
+   case TGSI_OPCODE_ENDFOR:
       /* fall-through (for now at least) */
-   case TGSI_OPCODE_ENDLOOP2:
+   case TGSI_OPCODE_ENDLOOP:
       /* Restore ContMask, but don't pop */
       ASSERT(mach->ContStackTop > 0);
       mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c
index 7c225e2f27..5ffb7073ab 100644
--- a/src/gallium/drivers/cell/spu/spu_render.c
+++ b/src/gallium/drivers/cell/spu/spu_render.c
@@ -32,6 +32,7 @@
 
 #include "spu_main.h"
 #include "spu_render.h"
+#include "spu_shuffle.h"
 #include "spu_tri.h"
 #include "spu_tile.h"
 #include "cell/common.h"
@@ -267,15 +268,75 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr)
 
       uint drawn = 0;
 
-      /* loop over tris */
-      for (j = 0; j < render->num_indexes; j += 3) {
-         const float *v0, *v1, *v2;
-
-         v0 = (const float *) (vertices + indexes[j+0] * vertex_size);
-         v1 = (const float *) (vertices + indexes[j+1] * vertex_size);
-         v2 = (const float *) (vertices + indexes[j+2] * vertex_size);
-
-         drawn += tri_draw(v0, v1, v2, tx, ty);
+      const qword vertex_sizes = (qword)spu_splats(vertex_size);
+      const qword verticess = (qword)spu_splats((uint)vertices);
+
+      ASSERT_ALIGN16(&indexes[0]);
+
+      const uint num_indexes = render->num_indexes;
+
+      /* loop over tris
+	   * &indexes[0] will be 16 byte aligned.  This loop is heavily unrolled
+	   * avoiding variable rotates when extracting vertex indices.
+	   */
+      for (j = 0; j < num_indexes; j += 24) {
+         /* Load three vectors, containing 24 ushort indices */
+         const qword* lower_qword = (qword*)&indexes[j];
+         const qword indices0 = lower_qword[0];
+         const qword indices1 = lower_qword[1];
+         const qword indices2 = lower_qword[2];
+
+         /* stores three indices for each tri n in slots 0, 1 and 2 of vsn */
+		 /* Straightforward rotates for these */
+         qword vs0 = indices0;
+         qword vs1 = si_shlqbyi(indices0, 6);
+         qword vs3 = si_shlqbyi(indices1, 2);
+         qword vs4 = si_shlqbyi(indices1, 8);
+         qword vs6 = si_shlqbyi(indices2, 4);
+         qword vs7 = si_shlqbyi(indices2, 10);
+
+         /* For tri 2 and 5, the three indices are split across two machine
+		  * words - rotate and combine */
+         const qword tmp2a = si_shlqbyi(indices0, 12);
+         const qword tmp2b = si_rotqmbyi(indices1, 12|16);
+         qword vs2 = si_selb(tmp2a, tmp2b, si_fsmh(si_from_uint(0x20)));
+
+         const qword tmp5a = si_shlqbyi(indices1, 14);
+         const qword tmp5b = si_rotqmbyi(indices2, 14|16);
+         qword vs5 = si_selb(tmp5a, tmp5b, si_fsmh(si_from_uint(0x60)));
+
+         /* unpack indices from halfword slots to word slots */
+         vs0 = si_shufb(vs0, vs0, SHUFB8(0,A,0,B,0,C,0,0));
+         vs1 = si_shufb(vs1, vs1, SHUFB8(0,A,0,B,0,C,0,0));
+         vs2 = si_shufb(vs2, vs2, SHUFB8(0,A,0,B,0,C,0,0));
+         vs3 = si_shufb(vs3, vs3, SHUFB8(0,A,0,B,0,C,0,0));
+         vs4 = si_shufb(vs4, vs4, SHUFB8(0,A,0,B,0,C,0,0));
+         vs5 = si_shufb(vs5, vs5, SHUFB8(0,A,0,B,0,C,0,0));
+         vs6 = si_shufb(vs6, vs6, SHUFB8(0,A,0,B,0,C,0,0));
+         vs7 = si_shufb(vs7, vs7, SHUFB8(0,A,0,B,0,C,0,0));
+
+         /* Calculate address of vertex in vertices[] */
+         vs0 = si_mpya(vs0, vertex_sizes, verticess);
+         vs1 = si_mpya(vs1, vertex_sizes, verticess);
+         vs2 = si_mpya(vs2, vertex_sizes, verticess);
+         vs3 = si_mpya(vs3, vertex_sizes, verticess);
+         vs4 = si_mpya(vs4, vertex_sizes, verticess);
+         vs5 = si_mpya(vs5, vertex_sizes, verticess);
+         vs6 = si_mpya(vs6, vertex_sizes, verticess);
+         vs7 = si_mpya(vs7, vertex_sizes, verticess);
+
+         /* Select the appropriate call based on the number of vertices 
+		  * remaining */
+         switch(num_indexes - j) {
+            default: drawn += tri_draw(vs7, tx, ty);
+            case 21: drawn += tri_draw(vs6, tx, ty);
+            case 18: drawn += tri_draw(vs5, tx, ty);
+            case 15: drawn += tri_draw(vs4, tx, ty);
+            case 12: drawn += tri_draw(vs3, tx, ty);
+            case 9:  drawn += tri_draw(vs2, tx, ty);
+            case 6:  drawn += tri_draw(vs1, tx, ty);
+            case 3:  drawn += tri_draw(vs0, tx, ty);
+         }
       }
 
       //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3);
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index d727268475..58be001be4 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -133,7 +133,15 @@ struct setup_stage {
 
    uint tx, ty;  /**< position of current tile (x, y) */
 
-   int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy;
+   union {
+      struct {
+         int cliprect_minx;
+         int cliprect_miny;
+         int cliprect_maxx;
+         int cliprect_maxy;
+      };
+      qword cliprect;
+   };
 
    struct interp_coef coef[PIPE_MAX_SHADER_INPUTS];
 
@@ -432,6 +440,41 @@ print_vertex(const struct vertex_header *v)
 }
 #endif
 
+/* Returns the minimum of each slot of two vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n]);
+ */
+static qword
+minfq(qword q0, qword q1)
+{
+   const qword q0q1m = si_fcgt(q0, q1);
+   return si_selb(q0, q1, q0q1m);
+}
+
+/* Returns the minimum of each slot of three vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+min3fq(qword q0, qword q1, qword q2)
+{
+   return minfq(minfq(q0, q1), q2);
+}
+
+/* Returns the maximum of each slot of two vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+maxfq(qword q0, qword q1) {
+   const qword q0q1m = si_fcgt(q0, q1);
+   return si_selb(q1, q0, q0q1m);
+}
+
+/* Returns the maximum of each slot of three vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+max3fq(qword q0, qword q1, qword q2) {
+   return maxfq(maxfq(q0, q1), q2);
+}
 
 /**
  * Sort vertices from top to bottom.
@@ -440,9 +483,7 @@ print_vertex(const struct vertex_header *v)
  * \return  FALSE if tri is totally outside tile, TRUE otherwise
  */
 static boolean
-setup_sort_vertices(const struct vertex_header *v0,
-                    const struct vertex_header *v1,
-                    const struct vertex_header *v2)
+setup_sort_vertices(const qword vs)
 {
    float area, sign;
 
@@ -455,57 +496,57 @@ setup_sort_vertices(const struct vertex_header *v0,
    }
 #endif
 
-   /* determine bottom to top order of vertices */
    {
+      /* Load the float values for various processing... */
+      const qword f0 = (qword)(((const struct vertex_header*)si_to_ptr(vs))->data[0]);
+      const qword f1 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 4)))->data[0]);
+      const qword f2 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 8)))->data[0]);
+
+      /* Check if triangle is completely outside the tile bounds
+       * Find the min and max x and y positions of the three poits */
+      const qword minf = min3fq(f0, f1, f2);
+      const qword maxf = max3fq(f0, f1, f2);
+
+      /* Compare min and max against cliprect vals */
+      const qword maxsmins = si_shufb(maxf, minf, SHUFB4(A,B,a,b));
+      const qword outside = si_fcgt(maxsmins, si_csflt(setup.cliprect, 0));
+
+      /* Use a little magic to work out of the tri is visible or not */
+      if(si_to_uint(si_xori(si_gb(outside), 0xc))) return FALSE;
+
+      /* determine bottom to top order of vertices */
       /* A table of shuffle patterns for putting vertex_header pointers into
          correct order.  Quite magical. */
-      const vec_uchar16 sort_order_patterns[] = {
-         SHUFFLE4(A,B,C,C),
-         SHUFFLE4(C,A,B,C),
-         SHUFFLE4(A,C,B,C),
-         SHUFFLE4(B,C,A,C),
-         SHUFFLE4(B,A,C,C),
-         SHUFFLE4(C,B,A,C) };
-
-      /* The vertex_header pointers, packed for easy shuffling later */
-      const vec_uint4 vs = {(unsigned)v0, (unsigned)v1, (unsigned)v2};
+      const qword sort_order_patterns[] = {
+         SHUFB4(A,B,C,C),
+         SHUFB4(C,A,B,C),
+         SHUFB4(A,C,B,C),
+         SHUFB4(B,C,A,C),
+         SHUFB4(B,A,C,C),
+         SHUFB4(C,B,A,C) };
 
       /* Collate y values into two vectors for comparison.
          Using only one shuffle constant! ;) */
-      const vec_float4 y_02_ = spu_shuffle(v0->data[0], v2->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_10_ = spu_shuffle(v1->data[0], v0->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_012 = spu_shuffle(y_02_, v1->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_120 = spu_shuffle(y_10_, v2->data[0], SHUFFLE4(0,B,b,C));
+      const qword y_02_ = si_shufb(f0, f2, SHUFB4(0,B,b,C));
+      const qword y_10_ = si_shufb(f1, f0, SHUFB4(0,B,b,C));
+      const qword y_012 = si_shufb(y_02_, f1, SHUFB4(0,B,b,C));
+      const qword y_120 = si_shufb(y_10_, f2, SHUFB4(0,B,b,C));
 
       /* Perform comparison: {y0,y1,y2} > {y1,y2,y0} */
-      const vec_uint4 compare = spu_cmpgt(y_012, y_120);
+      const qword compare = si_fcgt(y_012, y_120);
       /* Compress the result of the comparison into 4 bits */
-      const vec_uint4 gather = spu_gather(compare);
+      const qword gather = si_gb(compare);
       /* Subtract one to attain the index into the LUT.  Magical. */
-      const unsigned int index = spu_extract(gather, 0) - 1;
+      const unsigned int index = si_to_uint(gather) - 1;
 
       /* Load the appropriate pattern and construct the desired vector. */
-      setup.vertex_headers = (qword)spu_shuffle(vs, vs, sort_order_patterns[index]);
+      setup.vertex_headers = si_shufb(vs, vs, sort_order_patterns[index]);
 
       /* Using the result of the comparison, set sign.
          Very magical. */
-      sign = ((si_to_uint(si_cntb((qword)gather)) == 2) ? 1.0f : -1.0f);
+      sign = ((si_to_uint(si_cntb(gather)) == 2) ? 1.0f : -1.0f);
    }
 
-   /* Check if triangle is completely outside the tile bounds */
-   if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy)
-      return FALSE;
-   if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny)
-      return FALSE;
-   if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx &&
-       spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx &&
-       spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx)
-      return FALSE;
-   if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx &&
-       spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx &&
-       spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx)
-      return FALSE;
-
    setup.ebot.ds = spu_sub(setup.vmid->data[0], setup.vmin->data[0]);
    setup.emaj.ds = spu_sub(setup.vmax->data[0], setup.vmin->data[0]);
    setup.etop.ds = spu_sub(setup.vmax->data[0], setup.vmid->data[0]);
@@ -761,21 +802,19 @@ subtriangle(struct edge *eleft, struct edge *eright, unsigned lines)
  * The tile data should have already been fetched.
  */
 boolean
-tri_draw(const float *v0, const float *v1, const float *v2,
+tri_draw(const qword vs,
          uint tx, uint ty)
 {
    setup.tx = tx;
    setup.ty = ty;
 
    /* set clipping bounds to tile bounds */
-   setup.cliprect_minx = tx * TILE_SIZE;
-   setup.cliprect_miny = ty * TILE_SIZE;
-   setup.cliprect_maxx = (tx + 1) * TILE_SIZE;
-   setup.cliprect_maxy = (ty + 1) * TILE_SIZE;
-
-   if (!setup_sort_vertices((struct vertex_header *) v0,
-                            (struct vertex_header *) v1,
-                            (struct vertex_header *) v2)) {
+   const qword clipbase = (qword)((vec_uint4){tx, ty});
+   const qword clipmin = si_mpyui(clipbase, TILE_SIZE);
+   const qword clipmax = si_ai(clipmin, TILE_SIZE);
+   setup.cliprect = si_shufb(clipmin, clipmax, SHUFB4(A,B,a,b));
+
+   if(!setup_sort_vertices(vs)) {
       return FALSE; /* totally clipped */
    }
 
diff --git a/src/gallium/drivers/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h
index aa694dd7c9..82e3b19ad7 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.h
+++ b/src/gallium/drivers/cell/spu/spu_tri.h
@@ -31,7 +31,7 @@
 
 
 extern boolean
-tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty);
+tri_draw(const qword vs, uint tx, uint ty);
 
 
 #endif /* SPU_TRI_H */
diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile
index 8870b39866..fb533c1796 100644
--- a/src/gallium/drivers/i915simple/Makefile
+++ b/src/gallium/drivers/i915simple/Makefile
@@ -5,6 +5,7 @@ LIBNAME = i915simple
 
 C_SOURCES = \
 	i915_blit.c \
+	i915_buffer.c \
 	i915_clear.c \
 	i915_flush.c \
 	i915_context.c \
diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript
index 2366e1247f..778c4ed0fd 100644
--- a/src/gallium/drivers/i915simple/SConscript
+++ b/src/gallium/drivers/i915simple/SConscript
@@ -6,6 +6,7 @@ i915simple = env.ConvenienceLibrary(
 	target = 'i915simple',
 	source = [
 		'i915_blit.c',
+		'i915_buffer.c',
 		'i915_clear.c',
 		'i915_context.c',
 		'i915_debug.c',
diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h
index a433cf054d..b813784723 100644
--- a/src/gallium/drivers/i915simple/i915_batch.h
+++ b/src/gallium/drivers/i915simple/i915_batch.h
@@ -28,89 +28,20 @@
 #ifndef I915_BATCH_H
 #define I915_BATCH_H
 
-#include "i915_winsys.h"
+#include "intel_batchbuffer.h"
 
-struct i915_batchbuffer
-{
-   struct pipe_buffer *buffer;
-   struct i915_winsys *winsys;
+#define BEGIN_BATCH(dwords, relocs) \
+   (intel_batchbuffer_check(i915->batch, dwords, relocs))
 
-   unsigned char *map;
-   unsigned char *ptr;
+#define OUT_BATCH(dword) \
+   intel_batchbuffer_dword(i915->batch, dword)
 
-   size_t size;
-   size_t actual_size;
+#define OUT_RELOC(buf, usage, offset) \
+   intel_batchbuffer_reloc(i915->batch, buf, usage, offset)
 
-   size_t relocs;
-   size_t max_relocs;
-};
-
-static INLINE boolean
-i915_batchbuffer_check( struct i915_batchbuffer *batch,
-			size_t dwords,
-			size_t relocs )
-{
-   /** TODO JB: Check relocs */
-   return dwords * 4 <= batch->size - (batch->ptr - batch->map);
-}
-
-static INLINE size_t
-i915_batchbuffer_space( struct i915_batchbuffer *batch )
-{
-   return batch->size - (batch->ptr - batch->map);
-}
-
-static INLINE void
-i915_batchbuffer_dword( struct i915_batchbuffer *batch,
-			unsigned dword )
-{
-   if (i915_batchbuffer_space(batch) < 4)
-      return;
-
-   *(unsigned *)batch->ptr = dword;
-   batch->ptr += 4;
-}
-
-static INLINE void
-i915_batchbuffer_write( struct i915_batchbuffer *batch,
-			void *data,
-			size_t size )
-{
-   if (i915_batchbuffer_space(batch) < size)
-      return;
-
-   memcpy(data, batch->ptr, size);
-   batch->ptr += size;
-}
-
-static INLINE void
-i915_batchbuffer_reloc( struct i915_batchbuffer *batch,
-			struct pipe_buffer *buffer,
-			size_t flags,
-			size_t offset )
-{
-   batch->winsys->batch_reloc( batch->winsys, buffer, flags, offset );
-}
-
-static INLINE void
-i915_batchbuffer_flush( struct i915_batchbuffer *batch,
-			struct pipe_fence_handle **fence )
-{
-   batch->winsys->batch_flush( batch->winsys, fence );
-}
-
-#define BEGIN_BATCH( dwords, relocs ) \
-   (i915_batchbuffer_check( i915->batch, dwords, relocs ))
-
-#define OUT_BATCH( dword ) \
-   i915_batchbuffer_dword( i915->batch, dword )
-
-#define OUT_RELOC( buf, flags, delta ) \
-   i915_batchbuffer_reloc( i915->batch, buf, flags, delta )
-
-#define FLUSH_BATCH(fence) do {				\
-   i915->winsys->batch_flush( i915->winsys, fence );	\
-   i915->hardware_dirty = ~0;				\
+#define FLUSH_BATCH(fence) do {                 \
+   intel_batchbuffer_flush(i915->batch, fence); \
+   i915->hardware_dirty = ~0;                   \
 } while (0)
 
 #endif
diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c
index 448a4708ce..83dfc33528 100644
--- a/src/gallium/drivers/i915simple/i915_blit.c
+++ b/src/gallium/drivers/i915simple/i915_blit.c
@@ -26,8 +26,6 @@
  **************************************************************************/
 
 
-#include "i915_context.h"
-#include "i915_winsys.h"
 #include "i915_blit.h"
 #include "i915_reg.h"
 #include "i915_batch.h"
@@ -37,33 +35,33 @@
 
 void
 i915_fill_blit(struct i915_context *i915,
-	       unsigned cpp,
-	       unsigned short dst_pitch,
-	       struct pipe_buffer *dst_buffer,
-	       unsigned dst_offset,
-	       short x, short y, 
-	       short w, short h, 
-	       unsigned color)
+               unsigned cpp,
+               unsigned short dst_pitch,
+               struct intel_buffer *dst_buffer,
+               unsigned dst_offset,
+               short x, short y, 
+               short w, short h, 
+               unsigned color)
 {
    unsigned BR13, CMD;
 
 
    I915_DBG(i915,
-       "%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
-       __FUNCTION__,
-       dst_buffer, dst_pitch, dst_offset, x, y, w, h);
+      "%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
+      __FUNCTION__,
+      dst_buffer, dst_pitch, dst_offset, x, y, w, h);
 
    switch (cpp) {
    case 1:
    case 2:
    case 3:
       BR13 = (((int) dst_pitch) & 0xffff) |
-	 (0xF0 << 16) | (1 << 24);
+         (0xF0 << 16) | (1 << 24);
       CMD = XY_COLOR_BLT_CMD;
       break;
    case 4:
       BR13 = (((int) dst_pitch) & 0xffff) |
-	 (0xF0 << 16) | (1 << 24) | (1 << 25);
+         (0xF0 << 16) | (1 << 24) | (1 << 25);
       CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA |
              XY_COLOR_BLT_WRITE_RGB);
       break;
@@ -79,25 +77,24 @@ i915_fill_blit(struct i915_context *i915,
    OUT_BATCH(BR13);
    OUT_BATCH((y << 16) | x);
    OUT_BATCH(((y + h) << 16) | (x + w));
-   OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset);
+   OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset);
    OUT_BATCH(color);
    FLUSH_BATCH(NULL);
 }
 
-
 void
-i915_copy_blit( struct i915_context *i915,
-                  unsigned do_flip,
-                  unsigned cpp,
-                  unsigned short src_pitch,
-                  struct pipe_buffer *src_buffer,
-                  unsigned src_offset,
-                  unsigned short dst_pitch,
-                  struct pipe_buffer *dst_buffer,
-                  unsigned dst_offset,
-                  short src_x, short src_y,
-                  short dst_x, short dst_y, 
-		  short w, short h )
+i915_copy_blit(struct i915_context *i915,
+               unsigned do_flip,
+               unsigned cpp,
+               unsigned short src_pitch,
+               struct intel_buffer *src_buffer,
+               unsigned src_offset,
+               unsigned short dst_pitch,
+               struct intel_buffer *dst_buffer,
+               unsigned dst_offset,
+               short src_x, short src_y,
+               short dst_x, short dst_y, 
+               short w, short h)
 {
    unsigned CMD, BR13;
    int dst_y2 = dst_y + h;
@@ -105,32 +102,30 @@ i915_copy_blit( struct i915_context *i915,
 
 
    I915_DBG(i915,
-       "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
-       __FUNCTION__,
-       src_buffer, src_pitch, src_offset, src_x, src_y,
-       dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
+      "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
+      __FUNCTION__,
+      src_buffer, src_pitch, src_offset, src_x, src_y,
+      dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
 
    switch (cpp) {
    case 1:
    case 2:
    case 3:
       BR13 = (((int) dst_pitch) & 0xffff) |
-	 (0xCC << 16) | (1 << 24);
+         (0xCC << 16) | (1 << 24);
       CMD = XY_SRC_COPY_BLT_CMD;
       break;
    case 4:
       BR13 = (((int) dst_pitch) & 0xffff) |
-	 (0xCC << 16) | (1 << 24) | (1 << 25);
-      CMD =
-         (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA |
-          XY_SRC_COPY_BLT_WRITE_RGB);
+             (0xCC << 16) | (1 << 24) | (1 << 25);
+      CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA |
+            XY_SRC_COPY_BLT_WRITE_RGB);
       break;
    default:
       return;
    }
 
-   if (dst_y2 < dst_y || 
-       dst_x2 < dst_x) {
+   if (dst_y2 < dst_y || dst_x2 < dst_x) {
       return;
    }
 
@@ -140,7 +135,6 @@ i915_copy_blit( struct i915_context *i915,
     */
    assert (dst_pitch > 0 && src_pitch > 0);
 
-
    if (!BEGIN_BATCH(8, 2)) {
       FLUSH_BATCH(NULL);
       assert(BEGIN_BATCH(8, 2));
@@ -149,11 +143,9 @@ i915_copy_blit( struct i915_context *i915,
    OUT_BATCH(BR13);
    OUT_BATCH((dst_y << 16) | dst_x);
    OUT_BATCH((dst_y2 << 16) | dst_x2);
-   OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset);
+   OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset);
    OUT_BATCH((src_y << 16) | src_x);
    OUT_BATCH(((int) src_pitch & 0xffff));
-   OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset);
+   OUT_RELOC(src_buffer, INTEL_USAGE_2D_SOURCE, src_offset);
    FLUSH_BATCH(NULL);
 }
-
-
diff --git a/src/gallium/drivers/i915simple/i915_blit.h b/src/gallium/drivers/i915simple/i915_blit.h
index 0bb3453861..8ce3220cfd 100644
--- a/src/gallium/drivers/i915simple/i915_blit.h
+++ b/src/gallium/drivers/i915simple/i915_blit.h
@@ -32,24 +32,24 @@
 
 extern void i915_copy_blit(struct i915_context *i915,
                            unsigned do_flip,
-			   unsigned cpp,
-			   unsigned short src_pitch,
-			   struct pipe_buffer *src_buffer,
-			   unsigned src_offset,
-			   unsigned short dst_pitch,
-			   struct pipe_buffer *dst_buffer,
-			   unsigned dst_offset,
-			   short srcx, short srcy,
-			   short dstx, short dsty,
-			   short w, short h );
+                           unsigned cpp,
+                           unsigned short src_pitch,
+                           struct intel_buffer *src_buffer,
+                           unsigned src_offset,
+                           unsigned short dst_pitch,
+                           struct intel_buffer *dst_buffer,
+                           unsigned dst_offset,
+                           short srcx, short srcy,
+                           short dstx, short dsty,
+                           short w, short h);
 
 extern void i915_fill_blit(struct i915_context *i915,
-			   unsigned cpp,
-			   unsigned short dst_pitch,
-			   struct pipe_buffer *dst_buffer,
-			   unsigned dst_offset,
-			   short x, short y,
-			   short w, short h, unsigned color);
+                           unsigned cpp,
+                           unsigned short dst_pitch,
+                           struct intel_buffer *dst_buffer,
+                           unsigned dst_offset,
+                           short x, short y,
+                           short w, short h, unsigned color);
 
 
 #endif
diff --git a/src/gallium/drivers/i915simple/i915_buffer.c b/src/gallium/drivers/i915simple/i915_buffer.c
new file mode 100644
index 0000000000..effeba1297
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_buffer.c
@@ -0,0 +1,136 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 Jakob Bornecrantz
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "i915_screen.h"
+#include "i915_buffer.h"
+
+struct intel_buffer;
+
+struct i915_buffer
+{
+   struct pipe_buffer base;
+
+   struct intel_buffer *ibuf; /** hw buffer */
+
+   void *data; /**< user and malloc data */
+   boolean own; /**< we own the data incase of malloc */
+};
+
+static INLINE struct i915_buffer *
+i915_buffer(struct pipe_buffer *buffer)
+{
+   return (struct i915_buffer *)buffer;
+}
+
+static struct pipe_buffer *
+i915_buffer_create(struct pipe_screen *screen,
+                   unsigned alignment,
+                   unsigned usage,
+                   unsigned size)
+{
+   struct i915_buffer *buf = CALLOC_STRUCT(i915_buffer);
+
+   if (!buf)
+      return NULL;
+
+   pipe_reference_init(&buf->base.reference, 1);
+   buf->base.alignment = alignment;
+   buf->base.screen = screen;
+   buf->base.usage = usage;
+   buf->base.size = size;
+   buf->data = MALLOC(size);
+   buf->own = TRUE;
+
+   if (!buf->data)
+      goto err;
+
+   return &buf->base;
+
+err:
+   FREE(buf);
+   return NULL;
+}
+
+static struct pipe_buffer *
+i915_user_buffer_create(struct pipe_screen *screen,
+                        void *ptr,
+                        unsigned bytes)
+{
+   struct i915_buffer *buf = CALLOC_STRUCT(i915_buffer);
+
+   if (!buf)
+      return NULL;
+
+   pipe_reference_init(&buf->base.reference, 1);
+   buf->base.alignment = 0;
+   buf->base.screen = screen;
+   buf->base.usage = 0;
+   buf->base.size = bytes;
+   buf->data = ptr;
+   buf->own = FALSE;
+
+   return &buf->base;
+}
+
+static void *
+i915_buffer_map(struct pipe_screen *screen,
+                struct pipe_buffer *buffer,
+                unsigned usage)
+{
+   struct i915_buffer *buf = i915_buffer(buffer);
+   assert(!buf->ibuf);
+   return buf->data;
+}
+
+static void
+i915_buffer_unmap(struct pipe_screen *screen,
+                  struct pipe_buffer *buffer)
+{
+   struct i915_buffer *buf = i915_buffer(buffer);
+   assert(!buf->ibuf);
+}
+
+static void
+i915_buffer_destroy(struct pipe_buffer *buffer)
+{
+   struct i915_buffer *buf = i915_buffer(buffer);
+   assert(!buf->ibuf);
+
+   if (buf->own)
+      FREE(buf->data);
+   FREE(buf);
+}
+
+void i915_init_screen_buffer_functions(struct i915_screen *screen)
+{
+   screen->base.buffer_create = i915_buffer_create;
+   screen->base.user_buffer_create = i915_user_buffer_create;
+   screen->base.buffer_map = i915_buffer_map;
+   screen->base.buffer_map_range = NULL;
+   screen->base.buffer_flush_mapped_range = NULL;
+   screen->base.buffer_unmap = i915_buffer_unmap;
+   screen->base.buffer_destroy = i915_buffer_destroy;
+}
diff --git a/src/gallium/drivers/i915simple/i915_buffer.h b/src/gallium/drivers/i915simple/i915_buffer.h
new file mode 100644
index 0000000000..80fda7c62f
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_buffer.h
@@ -0,0 +1,31 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 Jakob Bornecrantz
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef I915_BUFFER_H
+#define I915_BUFFER_H
+
+void i915_init_screen_buffer_functions(struct i915_screen *screen);
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c
index ccf9bb31fb..b43f735245 100644
--- a/src/gallium/drivers/i915simple/i915_context.c
+++ b/src/gallium/drivers/i915simple/i915_context.c
@@ -26,8 +26,8 @@
  **************************************************************************/
 
 #include "i915_context.h"
-#include "i915_winsys.h"
 #include "i915_state.h"
+#include "i915_screen.h"
 #include "i915_batch.h"
 #include "i915_texture.h"
 #include "i915_reg.h"
@@ -40,66 +40,58 @@
 #include "pipe/p_screen.h"
 
 
-static void i915_destroy( struct pipe_context *pipe )
-{
-   struct i915_context *i915 = i915_context( pipe );
-
-   draw_destroy( i915->draw );
-   
-   if(i915->winsys->destroy)
-      i915->winsys->destroy(i915->winsys);
-
-   FREE( i915 );
-}
+/*
+ * Draw functions
+ */
 
 
 static boolean
 i915_draw_range_elements(struct pipe_context *pipe,
-			     struct pipe_buffer *indexBuffer,
-			     unsigned indexSize,
-			     unsigned min_index,
-			     unsigned max_index,
-			     unsigned prim, unsigned start, unsigned count)
+                         struct pipe_buffer *indexBuffer,
+                         unsigned indexSize,
+                         unsigned min_index,
+                         unsigned max_index,
+                         unsigned prim, unsigned start, unsigned count)
 {
-   struct i915_context *i915 = i915_context( pipe );
+   struct i915_context *i915 = i915_context(pipe);
    struct draw_context *draw = i915->draw;
    unsigned i;
 
    if (i915->dirty)
-      i915_update_derived( i915 );
+      i915_update_derived(i915);
 
    /*
     * Map vertex buffers
     */
    for (i = 0; i < i915->num_vertex_buffers; i++) {
-      void *buf
-         = pipe_buffer_map(pipe->screen,
-                                    i915->vertex_buffer[i].buffer,
-                                    PIPE_BUFFER_USAGE_CPU_READ);
+      void *buf = pipe_buffer_map(pipe->screen, i915->vertex_buffer[i].buffer,
+                                  PIPE_BUFFER_USAGE_CPU_READ);
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
-   /* Map index buffer, if present */
+
+   /*
+    * Map index buffer, if present
+    */
    if (indexBuffer) {
-      void *mapped_indexes
-         = pipe_buffer_map(pipe->screen, indexBuffer,
-                                    PIPE_BUFFER_USAGE_CPU_READ);
+      void *mapped_indexes = pipe_buffer_map(pipe->screen, indexBuffer,
+                                             PIPE_BUFFER_USAGE_CPU_READ);
       draw_set_mapped_element_buffer_range(draw, indexSize,
-					   min_index,
-					   max_index,
-					   mapped_indexes);
-   }
-   else {
-      /* no index/element buffer */
+                                           min_index,
+                                           max_index,
+                                           mapped_indexes);
+   } else {
       draw_set_mapped_element_buffer(draw, 0, NULL);
    }
 
 
    draw_set_mapped_constant_buffer(draw,
                                    i915->current.constants[PIPE_SHADER_VERTEX],
-                                   ( i915->current.num_user_constants[PIPE_SHADER_VERTEX] * 
-                                     4 * sizeof(float) ));
+                                   (i915->current.num_user_constants[PIPE_SHADER_VERTEX] * 
+                                      4 * sizeof(float)));
 
-   /* draw! */
+   /*
+    * Do the drawing
+    */
    draw_arrays(i915->draw, prim, start, count);
 
    /*
@@ -109,6 +101,7 @@ i915_draw_range_elements(struct pipe_context *pipe,
       pipe_buffer_unmap(pipe->screen, i915->vertex_buffer[i].buffer);
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
+
    if (indexBuffer) {
       pipe_buffer_unmap(pipe->screen, indexBuffer);
       draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
@@ -118,51 +111,81 @@ i915_draw_range_elements(struct pipe_context *pipe,
 }
 
 static boolean
-i915_draw_elements( struct pipe_context *pipe,
-                    struct pipe_buffer *indexBuffer,
-                    unsigned indexSize,
-                    unsigned prim, unsigned start, unsigned count)
+i915_draw_elements(struct pipe_context *pipe,
+                   struct pipe_buffer *indexBuffer,
+                   unsigned indexSize,
+                   unsigned prim, unsigned start, unsigned count)
 {
-   return i915_draw_range_elements( pipe, indexBuffer,
-					indexSize,
-					0, 0xffffffff,
-					prim, start, count );
+   return i915_draw_range_elements(pipe, indexBuffer,
+                                   indexSize,
+                                   0, 0xffffffff,
+                                   prim, start, count);
 }
 
-static boolean i915_draw_arrays( struct pipe_context *pipe,
-				 unsigned prim, unsigned start, unsigned count)
+static boolean
+i915_draw_arrays(struct pipe_context *pipe,
+                 unsigned prim, unsigned start, unsigned count)
 {
    return i915_draw_elements(pipe, NULL, 0, prim, start, count);
 }
 
 
+/*
+ * Is referenced functions
+ */
+
+
 static unsigned int
-i915_is_texture_referenced( struct pipe_context *pipe,
-			    struct pipe_texture *texture,
-			    unsigned face, unsigned level)
+i915_is_texture_referenced(struct pipe_context *pipe,
+                           struct pipe_texture *texture,
+                           unsigned face, unsigned level)
 {
    /**
-    * FIXME: Optimize.
+    * FIXME: Return the corrent result. We can't alays return referenced
+    *        since it causes a double flush within the vbo module.
     */
-
+#if 0
    return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
+#else
+   return 0;
+#endif
 }
 
 static unsigned int
-i915_is_buffer_referenced( struct pipe_context *pipe,
-			   struct pipe_buffer *buf)
+i915_is_buffer_referenced(struct pipe_context *pipe,
+                          struct pipe_buffer *buf)
 {
    /**
-    * FIXME: Optimize.
+    * FIXME: Return the corrent result. We can't alays return referenced
+    *        since it causes a double flush within the vbo module.
     */
-
+#if 0
    return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
+#else
+   return 0;
+#endif
 }
 
 
-struct pipe_context *i915_create_context( struct pipe_screen *screen,
-                                          struct pipe_winsys *pipe_winsys,
-                                          struct i915_winsys *i915_winsys )
+/*
+ * Generic context functions
+ */
+
+
+static void i915_destroy(struct pipe_context *pipe)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   draw_destroy(i915->draw);
+   
+   if(i915->batch)
+      i915->iws->batchbuffer_destroy(i915->batch);
+
+   FREE(i915);
+}
+
+struct pipe_context *
+i915_create_context(struct pipe_screen *screen)
 {
    struct i915_context *i915;
 
@@ -170,21 +193,20 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen,
    if (i915 == NULL)
       return NULL;
 
-   i915->winsys = i915_winsys;
-   i915->pipe.winsys = pipe_winsys;
-   i915->pipe.screen = screen;
+   i915->iws = i915_screen(screen)->iws;
+   i915->base.winsys = NULL;
+   i915->base.screen = screen;
 
-   i915->pipe.destroy = i915_destroy;
+   i915->base.destroy = i915_destroy;
 
-   i915->pipe.clear = i915_clear;
+   i915->base.clear = i915_clear;
 
+   i915->base.draw_arrays = i915_draw_arrays;
+   i915->base.draw_elements = i915_draw_elements;
+   i915->base.draw_range_elements = i915_draw_range_elements;
 
-   i915->pipe.draw_arrays = i915_draw_arrays;
-   i915->pipe.draw_elements = i915_draw_elements;
-   i915->pipe.draw_range_elements = i915_draw_range_elements;
-
-   i915->pipe.is_texture_referenced = i915_is_texture_referenced;
-   i915->pipe.is_buffer_referenced = i915_is_buffer_referenced;
+   i915->base.is_texture_referenced = i915_is_texture_referenced;
+   i915->base.is_buffer_referenced = i915_is_buffer_referenced;
 
    /*
     * Create drawing context and plug our rendering stage into it.
@@ -193,27 +215,23 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen,
    assert(i915->draw);
    if (!debug_get_bool_option("I915_NO_VBUF", FALSE)) {
       draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915));
-   }
-   else {
+   } else {
       draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915));
    }
 
    i915_init_surface_functions(i915);
    i915_init_state_functions(i915);
    i915_init_flush_functions(i915);
-   i915_init_texture_functions(i915);
 
-   draw_install_aaline_stage(i915->draw, &i915->pipe);
-   draw_install_aapoint_stage(i915->draw, &i915->pipe);
+   draw_install_aaline_stage(i915->draw, &i915->base);
+   draw_install_aapoint_stage(i915->draw, &i915->base);
 
    i915->dirty = ~0;
    i915->hardware_dirty = ~0;
 
    /* Batch stream debugging is a bit hacked up at the moment:
     */
-   i915->batch = i915_winsys->batch_get(i915_winsys);
-   i915->batch->winsys = i915_winsys;
+   i915->batch = i915->iws->batchbuffer_create(i915->iws);
 
-   return &i915->pipe;
+   return &i915->base;
 }
-
diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h
index b6983ba86e..234b441ce6 100644
--- a/src/gallium/drivers/i915simple/i915_context.h
+++ b/src/gallium/drivers/i915simple/i915_context.h
@@ -38,6 +38,11 @@
 #include "tgsi/tgsi_scan.h"
 
 
+struct intel_winsys;
+struct intel_buffer;
+struct intel_batchbuffer;
+
+
 #define I915_TEX_UNITS 8
 
 #define I915_DYNAMIC_MODES4       0
@@ -182,7 +187,6 @@ struct i915_sampler_state {
    unsigned maxlod;
 };
 
-
 struct i915_texture {
    struct pipe_texture base;
 
@@ -192,7 +196,8 @@ struct i915_texture {
    unsigned depth_stride;          /* per-image on i945? */
    unsigned total_nblocksy;
 
-   unsigned tiled;
+   unsigned sw_tiled; /**< tiled with software flags */
+   unsigned hw_tiled; /**< tiled with hardware fences */
 
    unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
 
@@ -206,15 +211,15 @@ struct i915_texture {
 
    /* The data is held here:
     */
-   struct pipe_buffer *buffer;
+   struct intel_buffer *buffer;
 };
 
-struct i915_batchbuffer;
-
 struct i915_context
 {
-   struct pipe_context pipe;
-   struct i915_winsys *winsys;
+   struct pipe_context base;
+
+   struct intel_winsys *iws;
+
    struct draw_context *draw;
 
    /* The most recent drawing state as set by the driver:
@@ -243,10 +248,10 @@ struct i915_context
    unsigned num_vertex_elements;
    unsigned num_vertex_buffers;
 
-   struct i915_batchbuffer *batch;
+   struct intel_batchbuffer *batch;
 
    /** Vertex buffer */
-   struct pipe_buffer *vbo;
+   struct intel_buffer *vbo;
    size_t vbo_offset;
    unsigned vbo_flushed;
 
diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c
index e08582efab..ce92d1af9a 100644
--- a/src/gallium/drivers/i915simple/i915_debug.c
+++ b/src/gallium/drivers/i915simple/i915_debug.c
@@ -27,7 +27,6 @@
 
 #include "i915_reg.h"
 #include "i915_context.h"
-#include "i915_winsys.h"
 #include "i915_debug.h"
 #include "i915_batch.h"
 #include "pipe/internal/p_winsys_screen.h"
@@ -864,7 +863,7 @@ static boolean i915_debug_packet( struct debug_stream *stream )
 
 
 void
-i915_dump_batchbuffer( struct i915_batchbuffer *batch )
+i915_dump_batchbuffer( struct intel_batchbuffer *batch )
 {
    struct debug_stream stream;
    unsigned *start = (unsigned*)batch->map;
diff --git a/src/gallium/drivers/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h
index 16ca7277c7..dd9b86e17b 100644
--- a/src/gallium/drivers/i915simple/i915_debug.h
+++ b/src/gallium/drivers/i915simple/i915_debug.h
@@ -104,9 +104,9 @@ I915_DBG(
 #endif
 
 
-struct i915_batchbuffer;
+struct intel_batchbuffer;
 
-void i915_dump_batchbuffer( struct i915_batchbuffer *i915 );
+void i915_dump_batchbuffer( struct intel_batchbuffer *i915 );
 
 void i915_debug_init( struct i915_context *i915 );
 
diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c
index 472e0ab774..1582168eba 100644
--- a/src/gallium/drivers/i915simple/i915_flush.c
+++ b/src/gallium/drivers/i915simple/i915_flush.c
@@ -45,6 +45,7 @@ static void i915_flush( struct pipe_context *pipe,
 
    draw_flush(i915->draw);
 
+#if 0
    /* Do we need to emit an MI_FLUSH command to flush the hardware
     * caches?
     */
@@ -63,6 +64,13 @@ static void i915_flush( struct pipe_context *pipe,
       }
       OUT_BATCH( flush );
    }
+#endif
+
+#if 0
+   if (i915->batch->map == i915->batch->ptr) {
+      return;
+   }
+#endif
 
    /* If there are no flags, just flush pending commands to hardware:
     */
@@ -74,5 +82,5 @@ static void i915_flush( struct pipe_context *pipe,
 
 void i915_init_flush_functions( struct i915_context *i915 )
 {
-   i915->pipe.flush = i915_flush;
+   i915->base.flush = i915_flush;
 }
diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c
index 961c1bf213..89504ced27 100644
--- a/src/gallium/drivers/i915simple/i915_fpc_translate.c
+++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c
@@ -975,8 +975,9 @@ i915_translate_instructions(struct i915_fp_compile *p,
                = &parse.FullToken.FullImmediate;
             const uint pos = p->num_immediates++;
             uint j;
+            assert( imm->Immediate.NrTokens <= 4 + 1 );
             for (j = 0; j < imm->Immediate.NrTokens - 1; j++) {
-               p->immediates[pos][j] = imm->u.ImmediateFloat32[j].Float;
+               p->immediates[pos][j] = imm->u[j].Float;
             }
          }
          break;
diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c
index 8f1f58b2dd..d9a5c40ab9 100644
--- a/src/gallium/drivers/i915simple/i915_prim_emit.c
+++ b/src/gallium/drivers/i915simple/i915_prim_emit.c
@@ -32,7 +32,6 @@
 #include "util/u_pack_color.h"
 
 #include "i915_context.h"
-#include "i915_winsys.h"
 #include "i915_reg.h"
 #include "i915_state.h"
 #include "i915_batch.h"
diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
index 9bdd91f288..508f4560e4 100644
--- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c
+++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
@@ -42,13 +42,11 @@
 #include "draw/draw_vbuf.h"
 #include "util/u_debug.h"
 #include "pipe/p_inlines.h"
-#include "pipe/internal/p_winsys_screen.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
 #include "i915_context.h"
 #include "i915_reg.h"
-#include "i915_winsys.h"
 #include "i915_batch.h"
 #include "i915_state.h"
 
@@ -59,7 +57,7 @@
 struct i915_vbuf_render {
    struct vbuf_render base;
 
-   struct i915_context *i915;   
+   struct i915_context *i915;
 
    /** Vertex size in bytes */
    size_t vertex_size;
@@ -74,7 +72,7 @@ struct i915_vbuf_render {
    unsigned fallback;
 
    /* Stuff for the vbo */
-   struct pipe_buffer *vbo;
+   struct intel_buffer *vbo;
    size_t vbo_size;
    size_t vbo_offset;
    void *vbo_ptr;
@@ -87,36 +85,34 @@ struct i915_vbuf_render {
  * Basically a cast wrapper.
  */
 static INLINE struct i915_vbuf_render *
-i915_vbuf_render( struct vbuf_render *render )
+i915_vbuf_render(struct vbuf_render *render)
 {
    assert(render);
    return (struct i915_vbuf_render *)render;
 }
 
-
 static const struct vertex_info *
-i915_vbuf_render_get_vertex_info( struct vbuf_render *render )
+i915_vbuf_render_get_vertex_info(struct vbuf_render *render)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
 
    if (i915->dirty) {
       /* make sure we have up to date vertex layout */
-      i915_update_derived( i915 );
+      i915_update_derived(i915);
    }
 
    return &i915->current.vertex_info;
 }
 
-
 static boolean
-i915_vbuf_render_allocate_vertices( struct vbuf_render *render,
-                                    ushort vertex_size,
-                                    ushort nr_vertices )
+i915_vbuf_render_allocate_vertices(struct vbuf_render *render,
+                                   ushort vertex_size,
+                                   ushort nr_vertices)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
-   struct pipe_screen *screen = i915->pipe.screen;
+   struct intel_winsys *iws = i915->iws;
    size_t size = (size_t)vertex_size * (size_t)nr_vertices;
 
    /* FIXME: handle failure */
@@ -125,17 +121,17 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render,
    if (i915_render->vbo_size > size + i915_render->vbo_offset && !i915->vbo_flushed) {
    } else {
       i915->vbo_flushed = 0;
-      if (i915_render->vbo)
-         pipe_buffer_reference(&i915_render->vbo, NULL);
+      if (i915_render->vbo) {
+         iws->buffer_destroy(iws, i915_render->vbo);
+         i915_render->vbo = NULL;
+      }
    }
 
    if (!i915_render->vbo) {
       i915_render->vbo_size = MAX2(size, i915_render->vbo_alloc_size);
       i915_render->vbo_offset = 0;
-      i915_render->vbo = pipe_buffer_create(screen,
-                                            64,
-                                            I915_BUFFER_USAGE_LIT_VERTEX,
-                                            i915_render->vbo_size);
+      i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size, 64,
+                                            INTEL_NEW_VERTEX);
 
    }
 
@@ -149,40 +145,37 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render,
    return TRUE;
 }
 
-
 static void *
-i915_vbuf_render_map_vertices( struct vbuf_render *render )
+i915_vbuf_render_map_vertices(struct vbuf_render *render)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
-   struct pipe_screen *screen = i915->pipe.screen;
+   struct intel_winsys *iws = i915->iws;
 
    if (i915->vbo_flushed)
       debug_printf("%s bad vbo flush occured stalling on hw\n");
 
-   i915_render->vbo_ptr = pipe_buffer_map(screen,
-                                          i915_render->vbo,
-                                          PIPE_BUFFER_USAGE_CPU_WRITE);
+   i915_render->vbo_ptr = iws->buffer_map(iws, i915_render->vbo, TRUE);
 
    return (unsigned char *)i915_render->vbo_ptr + i915->vbo_offset;
 }
 
 static void
-i915_vbuf_render_unmap_vertices( struct vbuf_render *render,
-                                 ushort min_index,
-                                 ushort max_index )
+i915_vbuf_render_unmap_vertices(struct vbuf_render *render,
+                                ushort min_index,
+                                ushort max_index)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
-   struct pipe_screen *screen = i915->pipe.screen;
+   struct intel_winsys *iws = i915->iws;
 
    i915_render->vbo_max_used = MAX2(i915_render->vbo_max_used, i915_render->vertex_size * (max_index + 1));
-   pipe_buffer_unmap(screen, i915_render->vbo);
+   iws->buffer_unmap(iws, i915_render->vbo);
 }
 
 static boolean
-i915_vbuf_render_set_primitive( struct vbuf_render *render, 
-                                unsigned prim )
+i915_vbuf_render_set_primitive(struct vbuf_render *render, 
+                               unsigned prim)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    i915_render->prim = prim;
@@ -234,15 +227,13 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render,
    }
 }
 
-
-
 /**
  * Used for fallbacks in draw_arrays
  */
 static void
-draw_arrays_generate_indices( struct vbuf_render *render,
-                              unsigned start, uint nr,
-                              unsigned type )
+draw_arrays_generate_indices(struct vbuf_render *render,
+                             unsigned start, uint nr,
+                             unsigned type)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
@@ -251,29 +242,29 @@ draw_arrays_generate_indices( struct vbuf_render *render,
    switch(type) {
    case 0:
       for (i = start; i+1 < end; i += 2)
-	 OUT_BATCH( (i+0) | (i+1) << 16 );
+         OUT_BATCH((i+0) | (i+1) << 16);
       if (i < end)
-	 OUT_BATCH( i );
+         OUT_BATCH(i);
       break;
    case PIPE_PRIM_LINE_LOOP:
       if (nr >= 2) {
-	 for (i = start + 1; i < end; i++)
-	    OUT_BATCH( (i-0) | (i+0) << 16 );
-	 OUT_BATCH( (i-0) | (  start) << 16 );
+         for (i = start + 1; i < end; i++)
+            OUT_BATCH((i-0) | (i+0) << 16);
+         OUT_BATCH((i-0) | ( start) << 16);
       }
       break;
    case PIPE_PRIM_QUADS:
       for (i = start; i + 3 < end; i += 4) {
-	 OUT_BATCH( (i+0) | (i+1) << 16 );
-	 OUT_BATCH( (i+3) | (i+1) << 16 );
-	 OUT_BATCH( (i+2) | (i+3) << 16 );
+         OUT_BATCH((i+0) | (i+1) << 16);
+         OUT_BATCH((i+3) | (i+1) << 16);
+         OUT_BATCH((i+2) | (i+3) << 16);
       }
       break;
    case PIPE_PRIM_QUAD_STRIP:
       for (i = start; i + 3 < end; i += 2) {
-	 OUT_BATCH( (i+0) | (i+1) << 16 );
-	 OUT_BATCH( (i+3) | (i+2) << 16 );
-	 OUT_BATCH( (i+0) | (i+3) << 16 );
+         OUT_BATCH((i+0) | (i+1) << 16);
+         OUT_BATCH((i+3) | (i+2) << 16);
+         OUT_BATCH((i+0) | (i+3) << 16);
       }
       break;
    default:
@@ -282,16 +273,16 @@ draw_arrays_generate_indices( struct vbuf_render *render,
 }
 
 static unsigned
-draw_arrays_calc_nr_indices( uint nr, unsigned type )
+draw_arrays_calc_nr_indices(uint nr, unsigned type)
 {
    switch (type) {
    case 0:
       return nr;
    case PIPE_PRIM_LINE_LOOP:
       if (nr >= 2)
-	 return nr * 2;
+         return nr * 2;
       else
-	 return 0;
+         return 0;
    case PIPE_PRIM_QUADS:
       return (nr / 4) * 6;
    case PIPE_PRIM_QUAD_STRIP:
@@ -303,64 +294,64 @@ draw_arrays_calc_nr_indices( uint nr, unsigned type )
 }
 
 static void
-draw_arrays_fallback( struct vbuf_render *render,
-                      unsigned start,
-                      uint nr )
+draw_arrays_fallback(struct vbuf_render *render,
+                     unsigned start,
+                     uint nr)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
    unsigned nr_indices;
 
    if (i915->dirty)
-      i915_update_derived( i915 );
+      i915_update_derived(i915);
 
    if (i915->hardware_dirty)
-      i915_emit_hardware_state( i915 );
+      i915_emit_hardware_state(i915);
 
-   nr_indices = draw_arrays_calc_nr_indices( nr, i915_render->fallback );
+   nr_indices = draw_arrays_calc_nr_indices(nr, i915_render->fallback);
    if (!nr_indices)
       return;
 
-   if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+   if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) {
       FLUSH_BATCH(NULL);
 
       /* Make sure state is re-emitted after a flush:
        */
-      i915_update_derived( i915 );
-      i915_emit_hardware_state( i915 );
+      i915_update_derived(i915);
+      i915_emit_hardware_state(i915);
       i915->vbo_flushed = 1;
 
-      if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
-	 assert(0);
-	 goto out;
+      if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) {
+         assert(0);
+         goto out;
       }
    }
-   OUT_BATCH( _3DPRIMITIVE |
-	      PRIM_INDIRECT |
-	      i915_render->hwprim |
-	      PRIM_INDIRECT_ELTS |
-	      nr_indices );
+   OUT_BATCH(_3DPRIMITIVE |
+             PRIM_INDIRECT |
+             i915_render->hwprim |
+             PRIM_INDIRECT_ELTS |
+             nr_indices);
 
-   draw_arrays_generate_indices( render, start, nr, i915_render->fallback );
+   draw_arrays_generate_indices(render, start, nr, i915_render->fallback);
 
 out:
    return;
 }
 
 static void
-i915_vbuf_render_draw_arrays( struct vbuf_render *render,
-                              unsigned start,
-                              uint nr )
+i915_vbuf_render_draw_arrays(struct vbuf_render *render,
+                             unsigned start,
+                             uint nr)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
 
    if (i915_render->fallback) {
-      draw_arrays_fallback( render, start, nr );
+      draw_arrays_fallback(render, start, nr);
       return;
    }
 
    /* JB: TODO submit direct cmds */
-   draw_arrays_fallback( render, start, nr );
+   draw_arrays_fallback(render, start, nr);
 }
 
 /**
@@ -368,10 +359,10 @@ i915_vbuf_render_draw_arrays( struct vbuf_render *render,
  * If type is zero normal operation assumed.
  */
 static void
-draw_generate_indices( struct vbuf_render *render,
-                       const ushort *indices,
-                       uint nr_indices,
-                       unsigned type )
+draw_generate_indices(struct vbuf_render *render,
+                      const ushort *indices,
+                      uint nr_indices,
+                      unsigned type)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
@@ -380,31 +371,31 @@ draw_generate_indices( struct vbuf_render *render,
    switch(type) {
    case 0:
       for (i = 0; i + 1 < nr_indices; i += 2) {
-	 OUT_BATCH( indices[i] | indices[i+1] << 16 );
+         OUT_BATCH(indices[i] | indices[i+1] << 16);
       }
       if (i < nr_indices) {
-	 OUT_BATCH( indices[i] );
+         OUT_BATCH(indices[i]);
       }
       break;
    case PIPE_PRIM_LINE_LOOP:
       if (nr_indices >= 2) {
-	 for (i = 1; i < nr_indices; i++)
-	    OUT_BATCH( indices[i-1] | indices[i] << 16 );
-	 OUT_BATCH( indices[i-1] | indices[0] << 16 );
+         for (i = 1; i < nr_indices; i++)
+            OUT_BATCH(indices[i-1] | indices[i] << 16);
+         OUT_BATCH(indices[i-1] | indices[0] << 16);
       }
       break;
    case PIPE_PRIM_QUADS:
       for (i = 0; i + 3 < nr_indices; i += 4) {
-	 OUT_BATCH( indices[i+0] | indices[i+1] << 16 );
-	 OUT_BATCH( indices[i+3] | indices[i+1] << 16 );
-	 OUT_BATCH( indices[i+2] | indices[i+3] << 16 );
+         OUT_BATCH(indices[i+0] | indices[i+1] << 16);
+         OUT_BATCH(indices[i+3] | indices[i+1] << 16);
+         OUT_BATCH(indices[i+2] | indices[i+3] << 16);
       }
       break;
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 0; i + 3 < nr_indices; i += 2) {
-	 OUT_BATCH( indices[i+0] | indices[i+1] << 16 );
-	 OUT_BATCH( indices[i+3] | indices[i+2] << 16 );
-	 OUT_BATCH( indices[i+0] | indices[i+3] << 16 );
+         OUT_BATCH(indices[i+0] | indices[i+1] << 16);
+         OUT_BATCH(indices[i+3] | indices[i+2] << 16);
+         OUT_BATCH(indices[i+0] | indices[i+3] << 16);
       }
       break;
    default:
@@ -414,16 +405,16 @@ draw_generate_indices( struct vbuf_render *render,
 }
 
 static unsigned
-draw_calc_nr_indices( uint nr_indices, unsigned type )
+draw_calc_nr_indices(uint nr_indices, unsigned type)
 {
    switch (type) {
    case 0:
       return nr_indices;
    case PIPE_PRIM_LINE_LOOP:
       if (nr_indices >= 2)
-	 return nr_indices * 2;
+         return nr_indices * 2;
       else
-	 return 0;
+         return 0;
    case PIPE_PRIM_QUADS:
       return (nr_indices / 4) * 6;
    case PIPE_PRIM_QUAD_STRIP:
@@ -435,9 +426,9 @@ draw_calc_nr_indices( uint nr_indices, unsigned type )
 }
 
 static void 
-i915_vbuf_render_draw( struct vbuf_render *render,
-                       const ushort *indices,
-                       uint nr_indices)
+i915_vbuf_render_draw(struct vbuf_render *render,
+                      const ushort *indices,
+                      uint nr_indices)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
@@ -445,48 +436,47 @@ i915_vbuf_render_draw( struct vbuf_render *render,
 
    save_nr_indices = nr_indices;
 
-   nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback );
+   nr_indices = draw_calc_nr_indices(nr_indices, i915_render->fallback);
    if (!nr_indices)
       return;
 
    if (i915->dirty)
-      i915_update_derived( i915 );
+      i915_update_derived(i915);
 
    if (i915->hardware_dirty)
-      i915_emit_hardware_state( i915 );
+      i915_emit_hardware_state(i915);
 
-   if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+   if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) {
       FLUSH_BATCH(NULL);
 
       /* Make sure state is re-emitted after a flush: 
        */
-      i915_update_derived( i915 );
-      i915_emit_hardware_state( i915 );
+      i915_update_derived(i915);
+      i915_emit_hardware_state(i915);
       i915->vbo_flushed = 1;
 
-      if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
-	 assert(0);
-     goto out;
+      if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) {
+         assert(0);
+         goto out;
       }
    }
 
-   OUT_BATCH( _3DPRIMITIVE |
-	      PRIM_INDIRECT |
-	      i915_render->hwprim |
-	      PRIM_INDIRECT_ELTS |
-	      nr_indices );
-   draw_generate_indices( render,
-			  indices,
-			  save_nr_indices,
-			  i915_render->fallback );
+   OUT_BATCH(_3DPRIMITIVE |
+             PRIM_INDIRECT |
+             i915_render->hwprim |
+             PRIM_INDIRECT_ELTS |
+             nr_indices);
+   draw_generate_indices(render,
+                         indices,
+                         save_nr_indices,
+                         i915_render->fallback);
 
 out:
    return;
 }
 
-
 static void
-i915_vbuf_render_release_vertices( struct vbuf_render *render )
+i915_vbuf_render_release_vertices(struct vbuf_render *render)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    struct i915_context *i915 = i915_render->i915;
@@ -499,23 +489,21 @@ i915_vbuf_render_release_vertices( struct vbuf_render *render )
    i915->dirty |= I915_NEW_VBO;
 }
 
-
 static void
-i915_vbuf_render_destroy( struct vbuf_render *render )
+i915_vbuf_render_destroy(struct vbuf_render *render)
 {
    struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
    FREE(i915_render);
 }
 
-
 /**
  * Create a new primitive render.
  */
 static struct vbuf_render *
-i915_vbuf_render_create( struct i915_context *i915 )
+i915_vbuf_render_create(struct i915_context *i915)
 {
    struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render);
-   struct pipe_screen *screen = i915->pipe.screen;
+   struct intel_winsys *iws = i915->iws;
 
    i915_render->i915 = i915;
    
@@ -539,23 +527,19 @@ i915_vbuf_render_create( struct i915_context *i915 )
    i915_render->vbo_alloc_size = 128 * 4096;
    i915_render->vbo_size = i915_render->vbo_alloc_size;
    i915_render->vbo_offset = 0;
-   i915_render->vbo = pipe_buffer_create(screen,
-                                         64,
-                                         I915_BUFFER_USAGE_LIT_VERTEX,
-                                         i915_render->vbo_size);
-   i915_render->vbo_ptr = pipe_buffer_map(screen,
-                                          i915_render->vbo,
-                                          PIPE_BUFFER_USAGE_CPU_WRITE);
-   pipe_buffer_unmap(screen, i915_render->vbo);
+   i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size, 64,
+                                         INTEL_NEW_VERTEX);
+   /* TODO JB: is this realy needed? */
+   i915_render->vbo_ptr = iws->buffer_map(iws, i915_render->vbo, TRUE);
+   iws->buffer_unmap(iws, i915_render->vbo);
 
    return &i915_render->base;
 }
 
-
 /**
  * Create a new primitive vbuf/render stage.
  */
-struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 )
+struct draw_stage *i915_draw_vbuf_stage(struct i915_context *i915)
 {
    struct vbuf_render *render;
    struct draw_stage *stage;
@@ -564,7 +548,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 )
    if(!render)
       return NULL;
    
-   stage = draw_vbuf_stage( i915->draw, render );
+   stage = draw_vbuf_stage(i915->draw, render);
    if(!stage) {
       render->destroy(render);
       return NULL;
diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c
index f4aa8e60d8..9f017a14cc 100644
--- a/src/gallium/drivers/i915simple/i915_screen.c
+++ b/src/gallium/drivers/i915simple/i915_screen.c
@@ -26,33 +26,36 @@
  **************************************************************************/
 
 
-#include "util/u_memory.h"
-#include "util/u_simple_screen.h"
-#include "pipe/internal/p_winsys_screen.h"
 #include "pipe/p_inlines.h"
+#include "util/u_memory.h"
 #include "util/u_string.h"
 
 #include "i915_reg.h"
 #include "i915_context.h"
 #include "i915_screen.h"
+#include "i915_buffer.h"
 #include "i915_texture.h"
-#include "i915_winsys.h"
+#include "intel_winsys.h"
+
+
+/*
+ * Probe functions
+ */
 
 
 static const char *
-i915_get_vendor( struct pipe_screen *pscreen )
+i915_get_vendor(struct pipe_screen *screen)
 {
    return "Tungsten Graphics, Inc.";
 }
 
-
 static const char *
-i915_get_name( struct pipe_screen *pscreen )
+i915_get_name(struct pipe_screen *screen)
 {
    static char buffer[128];
    const char *chipset;
 
-   switch (i915_screen(pscreen)->pci_id) {
+   switch (i915_screen(screen)->pci_id) {
    case PCI_CHIP_I915_G:
       chipset = "915G";
       break;
@@ -86,7 +89,6 @@ i915_get_name( struct pipe_screen *pscreen )
    return buffer;
 }
 
-
 static int
 i915_get_param(struct pipe_screen *screen, int param)
 {
@@ -122,7 +124,6 @@ i915_get_param(struct pipe_screen *screen, int param)
    }
 }
 
-
 static float
 i915_get_paramf(struct pipe_screen *screen, int param)
 {
@@ -148,13 +149,12 @@ i915_get_paramf(struct pipe_screen *screen, int param)
    }
 }
 
-
 static boolean
-i915_is_format_supported( struct pipe_screen *screen,
-                          enum pipe_format format, 
-                          enum pipe_texture_target target,
-                          unsigned tex_usage, 
-                          unsigned geom_flags )
+i915_is_format_supported(struct pipe_screen *screen,
+                         enum pipe_format format, 
+                         enum pipe_texture_target target,
+                         unsigned tex_usage, 
+                         unsigned geom_flags)
 {
    static const enum pipe_format tex_supported[] = {
       PIPE_FORMAT_R8G8B8A8_UNORM,
@@ -173,7 +173,6 @@ i915_is_format_supported( struct pipe_screen *screen,
       PIPE_FORMAT_A8R8G8B8_UNORM,
       PIPE_FORMAT_R5G6B5_UNORM,
       PIPE_FORMAT_S8Z24_UNORM,
-      /*PIPE_FORMAT_R16G16B16A16_SNORM,*/
       PIPE_FORMAT_NONE  /* list terminator */
    };
    const enum pipe_format *list;
@@ -193,120 +192,73 @@ i915_is_format_supported( struct pipe_screen *screen,
 }
 
 
-static void
-i915_destroy_screen( struct pipe_screen *screen )
-{
-   struct pipe_winsys *winsys = screen->winsys;
-
-   if(winsys->destroy)
-      winsys->destroy(winsys);
-
-   FREE(screen);
-}
+/*
+ * Fence functions
+ */
 
 
-static struct pipe_transfer*
-i915_get_tex_transfer(struct pipe_screen *screen,
-                      struct pipe_texture *texture,
-                      unsigned face, unsigned level, unsigned zslice,
-                      enum pipe_transfer_usage usage, unsigned x, unsigned y,
-                      unsigned w, unsigned h)
+static void
+i915_fence_reference(struct pipe_screen *screen,
+                     struct pipe_fence_handle **ptr,
+                     struct pipe_fence_handle *fence)
 {
-   struct i915_texture *tex = (struct i915_texture *)texture;
-   struct i915_transfer *trans;
-   unsigned offset;  /* in bytes */
+   struct i915_screen *is = i915_screen(screen);
 
-   if (texture->target == PIPE_TEXTURE_CUBE) {
-      offset = tex->image_offset[level][face];
-   }
-   else if (texture->target == PIPE_TEXTURE_3D) {
-      offset = tex->image_offset[level][zslice];
-   }
-   else {
-      offset = tex->image_offset[level][0];
-      assert(face == 0);
-      assert(zslice == 0);
-   }
-
-   trans = CALLOC_STRUCT(i915_transfer);
-   if (trans) {
-      pipe_texture_reference(&trans->base.texture, texture);
-      trans->base.format = trans->base.format;
-      trans->base.width = w;
-      trans->base.height = h;
-      trans->base.block = texture->block;
-      trans->base.nblocksx = texture->nblocksx[level];
-      trans->base.nblocksy = texture->nblocksy[level];
-      trans->base.stride = tex->stride;
-      trans->offset = offset;
-      trans->base.usage = usage;
-   }
-   return &trans->base;
+   is->iws->fence_reference(is->iws, ptr, fence);
 }
 
-static void
-i915_tex_transfer_destroy(struct pipe_transfer *trans)
+static int
+i915_fence_signalled(struct pipe_screen *screen,
+                     struct pipe_fence_handle *fence,
+                     unsigned flags)
 {
-   pipe_texture_reference(&trans->texture, NULL);
-   FREE(trans);
+   struct i915_screen *is = i915_screen(screen);
+
+   return is->iws->fence_signalled(is->iws, fence);
 }
 
-static void *
-i915_transfer_map( struct pipe_screen *screen,
-                   struct pipe_transfer *transfer )
+static int
+i915_fence_finish(struct pipe_screen *screen,
+                  struct pipe_fence_handle *fence,
+                  unsigned flags)
 {
-   struct i915_texture *tex = (struct i915_texture *)transfer->texture;
-   char *map;
-   unsigned flags = 0;
+   struct i915_screen *is = i915_screen(screen);
 
-   if (transfer->usage != PIPE_TRANSFER_WRITE)
-      flags |= PIPE_BUFFER_USAGE_CPU_READ;
+   return is->iws->fence_finish(is->iws, fence);
+}
 
-   if (transfer->usage != PIPE_TRANSFER_READ)
-      flags |= PIPE_BUFFER_USAGE_CPU_WRITE;
 
-   map = pipe_buffer_map( screen, tex->buffer, flags );
-   if (map == NULL)
-      return NULL;
+/*
+ * Generic functions
+ */
 
-   if (transfer->texture &&
-       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
-   {
-      /* Do something to notify contexts of a texture change.  
-       */
-      /* i915_screen(screen)->timestamp++; */
-   }
-   
-   return map + i915_transfer(transfer)->offset +
-      transfer->y / transfer->block.height * transfer->stride +
-      transfer->x / transfer->block.width * transfer->block.size;
-}
 
 static void
-i915_transfer_unmap(struct pipe_screen *screen,
-                    struct pipe_transfer *transfer)
+i915_destroy_screen(struct pipe_screen *screen)
 {
-   struct i915_texture *tex = (struct i915_texture *)transfer->texture;
-   pipe_buffer_unmap( screen, tex->buffer );
-}
+   struct i915_screen *is = i915_screen(screen);
 
+   if (is->iws)
+      is->iws->destroy(is->iws);
 
+   FREE(is);
+}
 
 /**
  * Create a new i915_screen object
  */
 struct pipe_screen *
-i915_create_screen(struct pipe_winsys *winsys, uint pci_id)
+i915_create_screen(struct intel_winsys *iws, uint pci_id)
 {
-   struct i915_screen *i915screen = CALLOC_STRUCT(i915_screen);
+   struct i915_screen *is = CALLOC_STRUCT(i915_screen);
 
-   if (!i915screen)
+   if (!is)
       return NULL;
 
    switch (pci_id) {
    case PCI_CHIP_I915_G:
    case PCI_CHIP_I915_GM:
-      i915screen->is_i945 = FALSE;
+      is->is_i945 = FALSE;
       break;
 
    case PCI_CHIP_I945_G:
@@ -315,7 +267,7 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id)
    case PCI_CHIP_G33_G:
    case PCI_CHIP_Q33_G:
    case PCI_CHIP_Q35_G:
-      i915screen->is_i945 = TRUE;
+      is->is_i945 = TRUE;
       break;
 
    default:
@@ -324,24 +276,25 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id)
       return NULL;
    }
 
-   i915screen->pci_id = pci_id;
+   is->pci_id = pci_id;
+   is->iws = iws;
+
+   is->base.winsys = NULL;
 
-   i915screen->screen.winsys = winsys;
+   is->base.destroy = i915_destroy_screen;
 
-   i915screen->screen.destroy = i915_destroy_screen;
+   is->base.get_name = i915_get_name;
+   is->base.get_vendor = i915_get_vendor;
+   is->base.get_param = i915_get_param;
+   is->base.get_paramf = i915_get_paramf;
+   is->base.is_format_supported = i915_is_format_supported;
 
-   i915screen->screen.get_name = i915_get_name;
-   i915screen->screen.get_vendor = i915_get_vendor;
-   i915screen->screen.get_param = i915_get_param;
-   i915screen->screen.get_paramf = i915_get_paramf;
-   i915screen->screen.is_format_supported = i915_is_format_supported;
-   i915screen->screen.get_tex_transfer = i915_get_tex_transfer;
-   i915screen->screen.tex_transfer_destroy = i915_tex_transfer_destroy;
-   i915screen->screen.transfer_map = i915_transfer_map;
-   i915screen->screen.transfer_unmap = i915_transfer_unmap;
+   is->base.fence_reference = i915_fence_reference;
+   is->base.fence_signalled = i915_fence_signalled;
+   is->base.fence_finish = i915_fence_finish;
 
-   i915_init_screen_texture_functions(&i915screen->screen);
-   u_simple_screen_init(&i915screen->screen);
+   i915_init_screen_texture_functions(is);
+   i915_init_screen_buffer_functions(is);
 
-   return &i915screen->screen;
+   return &is->base;
 }
diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h
index 5284c32595..5126485caa 100644
--- a/src/gallium/drivers/i915simple/i915_screen.h
+++ b/src/gallium/drivers/i915simple/i915_screen.h
@@ -25,17 +25,14 @@
  * 
  **************************************************************************/
 
-
 #ifndef I915_SCREEN_H
 #define I915_SCREEN_H
 
-
+#include "pipe/p_state.h"
 #include "pipe/p_screen.h"
 
 
-#ifdef __cplusplus
-extern "C" {
-#endif
+struct intel_winsys;
 
 
 /**
@@ -43,13 +40,14 @@ extern "C" {
  */
 struct i915_screen
 {
-   struct pipe_screen screen;
+   struct pipe_screen base;
+
+   struct intel_winsys *iws;
 
    boolean is_i945;
    uint pci_id;
 };
 
-
 /**
  * Subclass of pipe_transfer
  */
@@ -61,7 +59,11 @@ struct i915_transfer
 };
 
 
-/** cast wrappers */
+/*
+ * Cast wrappers
+ */
+
+
 static INLINE struct i915_screen *
 i915_screen(struct pipe_screen *pscreen)
 {
@@ -69,14 +71,10 @@ i915_screen(struct pipe_screen *pscreen)
 }
 
 static INLINE struct i915_transfer *
-i915_transfer( struct pipe_transfer *transfer )
+i915_transfer(struct pipe_transfer *transfer)
 {
    return (struct i915_transfer *)transfer;
 }
 
 
-#ifdef __cplusplus
-}
-#endif
-
 #endif /* I915_SCREEN_H */
diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c
index 273e74002a..0087dfa410 100644
--- a/src/gallium/drivers/i915simple/i915_state.c
+++ b/src/gallium/drivers/i915simple/i915_state.c
@@ -518,7 +518,7 @@ static void i915_set_constant_buffer(struct pipe_context *pipe,
                                      const struct pipe_constant_buffer *buf)
 {
    struct i915_context *i915 = i915_context(pipe);
-   struct pipe_winsys *ws = pipe->winsys;
+   struct pipe_screen *screen = pipe->screen;
    draw_flush(i915->draw);
 
    assert(shader < PIPE_SHADER_TYPES);
@@ -536,10 +536,10 @@ static void i915_set_constant_buffer(struct pipe_context *pipe,
    if (buf) {
       void *mapped;
       if (buf->buffer && buf->buffer->size &&
-          (mapped = ws->buffer_map(ws, buf->buffer,
+          (mapped = pipe_buffer_map(screen, buf->buffer,
                                     PIPE_BUFFER_USAGE_CPU_READ))) {
          memcpy(i915->current.constants[shader], mapped, buf->buffer->size);
-         ws->buffer_unmap(ws, buf->buffer);
+         pipe_buffer_unmap(screen, buf->buffer);
          i915->current.num_user_constants[shader]
             = buf->buffer->size / (4 * sizeof(float));
       }
@@ -751,38 +751,38 @@ static void i915_set_edgeflags(struct pipe_context *pipe,
 void
 i915_init_state_functions( struct i915_context *i915 )
 {
-   i915->pipe.set_edgeflags = i915_set_edgeflags;
-   i915->pipe.create_blend_state = i915_create_blend_state;
-   i915->pipe.bind_blend_state = i915_bind_blend_state;
-   i915->pipe.delete_blend_state = i915_delete_blend_state;
-
-   i915->pipe.create_sampler_state = i915_create_sampler_state;
-   i915->pipe.bind_sampler_states = i915_bind_sampler_states;
-   i915->pipe.delete_sampler_state = i915_delete_sampler_state;
-
-   i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state;
-   i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state;
-   i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state;
-
-   i915->pipe.create_rasterizer_state = i915_create_rasterizer_state;
-   i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state;
-   i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state;
-   i915->pipe.create_fs_state = i915_create_fs_state;
-   i915->pipe.bind_fs_state = i915_bind_fs_state;
-   i915->pipe.delete_fs_state = i915_delete_fs_state;
-   i915->pipe.create_vs_state = i915_create_vs_state;
-   i915->pipe.bind_vs_state = i915_bind_vs_state;
-   i915->pipe.delete_vs_state = i915_delete_vs_state;
-
-   i915->pipe.set_blend_color = i915_set_blend_color;
-   i915->pipe.set_clip_state = i915_set_clip_state;
-   i915->pipe.set_constant_buffer = i915_set_constant_buffer;
-   i915->pipe.set_framebuffer_state = i915_set_framebuffer_state;
-
-   i915->pipe.set_polygon_stipple = i915_set_polygon_stipple;
-   i915->pipe.set_scissor_state = i915_set_scissor_state;
-   i915->pipe.set_sampler_textures = i915_set_sampler_textures;
-   i915->pipe.set_viewport_state = i915_set_viewport_state;
-   i915->pipe.set_vertex_buffers = i915_set_vertex_buffers;
-   i915->pipe.set_vertex_elements = i915_set_vertex_elements;
+   i915->base.set_edgeflags = i915_set_edgeflags;
+   i915->base.create_blend_state = i915_create_blend_state;
+   i915->base.bind_blend_state = i915_bind_blend_state;
+   i915->base.delete_blend_state = i915_delete_blend_state;
+
+   i915->base.create_sampler_state = i915_create_sampler_state;
+   i915->base.bind_sampler_states = i915_bind_sampler_states;
+   i915->base.delete_sampler_state = i915_delete_sampler_state;
+
+   i915->base.create_depth_stencil_alpha_state = i915_create_depth_stencil_state;
+   i915->base.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state;
+   i915->base.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state;
+
+   i915->base.create_rasterizer_state = i915_create_rasterizer_state;
+   i915->base.bind_rasterizer_state = i915_bind_rasterizer_state;
+   i915->base.delete_rasterizer_state = i915_delete_rasterizer_state;
+   i915->base.create_fs_state = i915_create_fs_state;
+   i915->base.bind_fs_state = i915_bind_fs_state;
+   i915->base.delete_fs_state = i915_delete_fs_state;
+   i915->base.create_vs_state = i915_create_vs_state;
+   i915->base.bind_vs_state = i915_bind_vs_state;
+   i915->base.delete_vs_state = i915_delete_vs_state;
+
+   i915->base.set_blend_color = i915_set_blend_color;
+   i915->base.set_clip_state = i915_set_clip_state;
+   i915->base.set_constant_buffer = i915_set_constant_buffer;
+   i915->base.set_framebuffer_state = i915_set_framebuffer_state;
+
+   i915->base.set_polygon_stipple = i915_set_polygon_stipple;
+   i915->base.set_scissor_state = i915_set_scissor_state;
+   i915->base.set_sampler_textures = i915_set_sampler_textures;
+   i915->base.set_viewport_state = i915_set_viewport_state;
+   i915->base.set_vertex_buffers = i915_set_vertex_buffers;
+   i915->base.set_vertex_elements = i915_set_vertex_elements;
 }
diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c
index 1e1fb968b4..a3d4e3b04e 100644
--- a/src/gallium/drivers/i915simple/i915_state_emit.c
+++ b/src/gallium/drivers/i915simple/i915_state_emit.c
@@ -28,7 +28,6 @@
 
 #include "i915_reg.h"
 #include "i915_context.h"
-#include "i915_winsys.h"
 #include "i915_batch.h"
 #include "i915_reg.h"
 
@@ -107,7 +106,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
                              6 
                            ) * 3/2; /* plus 50% margin */
    const unsigned relocs = ( I915_TEX_UNITS +
-	                     3
+                             3
                            ) * 3/2; /* plus 50% margin */
 
 #if 0
@@ -123,9 +122,9 @@ i915_emit_hardware_state(struct i915_context *i915 )
    if (i915->hardware_dirty & I915_HW_INVARIENT)
    {
       OUT_BATCH(_3DSTATE_AA_CMD |
-		AA_LINE_ECAAR_WIDTH_ENABLE |
-		AA_LINE_ECAAR_WIDTH_1_0 |
-		AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0);
+                AA_LINE_ECAAR_WIDTH_ENABLE |
+                AA_LINE_ECAAR_WIDTH_1_0 |
+                AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0);
 
       OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD);
       OUT_BATCH(0);
@@ -137,24 +136,24 @@ i915_emit_hardware_state(struct i915_context *i915 )
       OUT_BATCH(0);
 
       OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS |
-		CSB_TCB(0, 0) |
-		CSB_TCB(1, 1) |
-		CSB_TCB(2, 2) |
-		CSB_TCB(3, 3) |
-		CSB_TCB(4, 4) | 
-		CSB_TCB(5, 5) | 
-		CSB_TCB(6, 6) | 
-		CSB_TCB(7, 7));
+                CSB_TCB(0, 0) |
+                CSB_TCB(1, 1) |
+                CSB_TCB(2, 2) |
+                CSB_TCB(3, 3) |
+                CSB_TCB(4, 4) | 
+                CSB_TCB(5, 5) | 
+                CSB_TCB(6, 6) | 
+                CSB_TCB(7, 7));
 
       OUT_BATCH(_3DSTATE_RASTER_RULES_CMD |
-		ENABLE_POINT_RASTER_RULE |
-		OGL_POINT_RASTER_RULE |
-		ENABLE_LINE_STRIP_PROVOKE_VRTX |
-		ENABLE_TRI_FAN_PROVOKE_VRTX |
-		LINE_STRIP_PROVOKE_VRTX(1) |
-		TRI_FAN_PROVOKE_VRTX(2) | 
-		ENABLE_TEXKILL_3D_4D | 
-		TEXKILL_4D);
+                ENABLE_POINT_RASTER_RULE |
+                OGL_POINT_RASTER_RULE |
+                ENABLE_LINE_STRIP_PROVOKE_VRTX |
+                ENABLE_TRI_FAN_PROVOKE_VRTX |
+                LINE_STRIP_PROVOKE_VRTX(1) |
+                TRI_FAN_PROVOKE_VRTX(2) | 
+                ENABLE_TEXKILL_3D_4D | 
+                TEXKILL_4D);
 
       /* Need to initialize this to zero.
        */
@@ -173,21 +172,21 @@ i915_emit_hardware_state(struct i915_context *i915 )
    if (i915->hardware_dirty & I915_HW_IMMEDIATE)
    {
       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | 
-		I1_LOAD_S(0) |
-		I1_LOAD_S(1) |
-		I1_LOAD_S(2) |
-		I1_LOAD_S(4) |
-		I1_LOAD_S(5) |
-		I1_LOAD_S(6) | 
-		(5));
+                I1_LOAD_S(0) |
+                I1_LOAD_S(1) |
+                I1_LOAD_S(2) |
+                I1_LOAD_S(4) |
+                I1_LOAD_S(5) |
+                I1_LOAD_S(6) | 
+                (5));
       
       if(i915->vbo)
          OUT_RELOC(i915->vbo,
-                   I915_BUFFER_ACCESS_READ,
+                   INTEL_USAGE_VERTEX,
                    i915->current.immediate[I915_IMMEDIATE_S0]);
       else
-	 /* FIXME: we should not do this */
-	 OUT_BATCH(0);
+         /* FIXME: we should not do this */
+         OUT_BATCH(0);
       OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]);
       OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]);
       OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]);
@@ -200,7 +199,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
    {
       int i;
       for (i = 0; i < I915_MAX_DYNAMIC; i++) {
-	 OUT_BATCH(i915->current.dynamic[i]);
+         OUT_BATCH(i915->current.dynamic[i]);
       }
    }
    
@@ -211,68 +210,68 @@ i915_emit_hardware_state(struct i915_context *i915 )
       struct pipe_surface *depth_surface = i915->framebuffer.zsbuf;
 
       if (cbuf_surface) {
-	 unsigned ctile = BUF_3D_USE_FENCE;
+         unsigned ctile = BUF_3D_USE_FENCE;
          struct i915_texture *tex = (struct i915_texture *)
                                     cbuf_surface->texture;
          assert(tex);
 
-	 if (tex && tex->tiled) {
-	    ctile = BUF_3D_TILED_SURFACE;
-	 }
+         if (tex && tex->sw_tiled) {
+            ctile = BUF_3D_TILED_SURFACE;
+         }
 
-	 OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
+         OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
 
-	 OUT_BATCH(BUF_3D_ID_COLOR_BACK |
-		   BUF_3D_PITCH(tex->stride) |  /* pitch in bytes */
-		   ctile);
+         OUT_BATCH(BUF_3D_ID_COLOR_BACK |
+                   BUF_3D_PITCH(tex->stride) |  /* pitch in bytes */
+                   ctile);
 
-	 OUT_RELOC(tex->buffer,
-		   I915_BUFFER_ACCESS_WRITE,
-		   cbuf_surface->offset);
+         OUT_RELOC(tex->buffer,
+                   INTEL_USAGE_RENDER,
+                   cbuf_surface->offset);
       }
 
       /* What happens if no zbuf??
        */
       if (depth_surface) {
-	 unsigned ztile = BUF_3D_USE_FENCE;
+         unsigned ztile = BUF_3D_USE_FENCE;
          struct i915_texture *tex = (struct i915_texture *)
                                     depth_surface->texture;
          assert(tex);
 
-	 if (tex && tex->tiled) {
-	    ztile = BUF_3D_TILED_SURFACE;
-	 }
+         if (tex && tex->sw_tiled) {
+            ztile = BUF_3D_TILED_SURFACE;
+         }
 
-	 OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
+         OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
 
-	 OUT_BATCH(BUF_3D_ID_DEPTH |
-		   BUF_3D_PITCH(tex->stride) |  /* pitch in bytes */
-		   ztile);
+         OUT_BATCH(BUF_3D_ID_DEPTH |
+                   BUF_3D_PITCH(tex->stride) |  /* pitch in bytes */
+                   ztile);
 
-	 OUT_RELOC(tex->buffer,
-		   I915_BUFFER_ACCESS_WRITE,
-		   depth_surface->offset);
+         OUT_RELOC(tex->buffer,
+                   INTEL_USAGE_RENDER,
+                   depth_surface->offset);
       }
    
       {
-	 unsigned cformat, zformat = 0;
+         unsigned cformat, zformat = 0;
       
-	 if (cbuf_surface)
+         if (cbuf_surface)
             cformat = cbuf_surface->format;
          else
             cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */
          cformat = translate_format(cformat);
 
-	 if (depth_surface) 
-	    zformat = translate_depth_format( i915->framebuffer.zsbuf->format );
+         if (depth_surface) 
+            zformat = translate_depth_format( i915->framebuffer.zsbuf->format );
 
-	 OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD);
-	 OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */
-		   DSTORG_VERT_BIAS(0x8) | /* .5 */
-		   LOD_PRECLAMP_OGL |
-		   TEX_DEFAULT_COLOR_OGL |
-		   cformat |
-		   zformat );
+         OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD);
+         OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */
+                   DSTORG_VERT_BIAS(0x8) | /* .5 */
+                   LOD_PRECLAMP_OGL |
+                   TEX_DEFAULT_COLOR_OGL |
+                   cformat |
+                   zformat );
       }
    }
 
@@ -290,16 +289,13 @@ i915_emit_hardware_state(struct i915_context *i915 )
             OUT_BATCH(enabled);
             for (unit = 0; unit < I915_TEX_UNITS; unit++) {
                if (enabled & (1 << unit)) {
-                  struct pipe_buffer *buf =
-                     i915->texture[unit]->buffer;
+                  struct intel_buffer *buf = i915->texture[unit]->buffer;
                   uint offset = 0;
                   assert(buf);
 
                   count++;
 
-                  OUT_RELOC(buf,
-                            I915_BUFFER_ACCESS_READ,
-                            offset);
+                  OUT_RELOC(buf, INTEL_USAGE_SAMPLER, offset);
                   OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */
                   OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */
                }
@@ -315,20 +311,20 @@ i915_emit_hardware_state(struct i915_context *i915 )
    if (i915->hardware_dirty & I915_HW_SAMPLER) 
    {
       if (i915->current.sampler_enable_nr) {
-	 int i;
-	 
-	 OUT_BATCH( _3DSTATE_SAMPLER_STATE | 
-		    (3 * i915->current.sampler_enable_nr) );
-
-	 OUT_BATCH( i915->current.sampler_enable_flags );
-
-	 for (i = 0; i < I915_TEX_UNITS; i++) {
-	    if (i915->current.sampler_enable_flags & (1<<i)) {
-	       OUT_BATCH( i915->current.sampler[i][0] );
-	       OUT_BATCH( i915->current.sampler[i][1] );
-	       OUT_BATCH( i915->current.sampler[i][2] );
-	    }
-	 }
+         int i;
+         
+         OUT_BATCH( _3DSTATE_SAMPLER_STATE | 
+                    (3 * i915->current.sampler_enable_nr) );
+
+         OUT_BATCH( i915->current.sampler_enable_flags );
+
+         for (i = 0; i < I915_TEX_UNITS; i++) {
+            if (i915->current.sampler_enable_flags & (1<<i)) {
+               OUT_BATCH( i915->current.sampler[i][0] );
+               OUT_BATCH( i915->current.sampler[i][1] );
+               OUT_BATCH( i915->current.sampler[i][2] );
+            }
+         }
       }
    }
 #endif
diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c
index 3667ed1afa..c5e9084d12 100644
--- a/src/gallium/drivers/i915simple/i915_state_sampler.c
+++ b/src/gallium/drivers/i915simple/i915_state_sampler.c
@@ -247,7 +247,7 @@ i915_update_texture(struct i915_context *i915,
    assert(format);
    assert(pitch);
 
-   if (tex->tiled) {
+   if (tex->sw_tiled) {
       assert(!((pitch - 1) & pitch));
       tiled = MS3_TILED_SURFACE;
    }
diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c
index 09b2c499b8..ab8331f3e6 100644
--- a/src/gallium/drivers/i915simple/i915_surface.c
+++ b/src/gallium/drivers/i915simple/i915_surface.c
@@ -89,6 +89,6 @@ i915_surface_fill(struct pipe_context *pipe,
 void
 i915_init_surface_functions(struct i915_context *i915)
 {
-   i915->pipe.surface_copy = i915_surface_copy;
-   i915->pipe.surface_fill = i915_surface_fill;
+   i915->base.surface_copy = i915_surface_copy;
+   i915->base.surface_fill = i915_surface_fill;
 }
diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c
index ca8e87af8d..6a6c654271 100644
--- a/src/gallium/drivers/i915simple/i915_texture.c
+++ b/src/gallium/drivers/i915simple/i915_texture.c
@@ -42,12 +42,14 @@
 #include "i915_texture.h"
 #include "i915_debug.h"
 #include "i915_screen.h"
-#include "i915_winsys.h"
+#include "intel_winsys.h"
+
 
 /*
  * Helper function and arrays
  */
 
+
 /**
  * Initial offset for Cube map.
  */
@@ -72,11 +74,6 @@ static const int step_offsets[6][2] = {
    {-1, 1}
 };
 
-static unsigned minify( unsigned d )
-{
-   return MAX2(1, d>>1);
-}
-
 static unsigned
 power_of_two(unsigned x)
 {
@@ -138,7 +135,7 @@ i915_miptree_set_level_info(struct i915_texture *tex,
 
 static void
 i915_miptree_set_image_offset(struct i915_texture *tex,
-			      unsigned level, unsigned img, unsigned x, unsigned y)
+                              unsigned level, unsigned img, unsigned x, unsigned y)
 {
    if (img == 0 && level == 0)
       assert(x == 0 && y == 0);
@@ -155,49 +152,189 @@ i915_miptree_set_image_offset(struct i915_texture *tex,
 
 
 /*
- * Layout functions
+ * i915 layout functions, some used by i945
  */
 
 
 /**
- * Special case to deal with display targets.
+ * Special case to deal with scanout textures.
  */
 static boolean
-i915_displaytarget_layout(struct i915_texture *tex)
+i915_scanout_layout(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
 
    if (pt->last_level > 0 || pt->block.size != 4)
       return 0;
 
-   i915_miptree_set_level_info( tex, 0, 1,
-                                tex->base.width[0],
-                                tex->base.height[0],
-                                1 );
-   i915_miptree_set_image_offset( tex, 0, 0, 0, 0 );
+   i915_miptree_set_level_info(tex, 0, 1,
+                               tex->base.width[0],
+                               tex->base.height[0],
+                               1);
+   i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   if (tex->base.width[0] >= 128) {
+   if (tex->base.width[0] >= 240) {
+      tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
+      tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
+      tex->hw_tiled = INTEL_TILE_X;
+   } else if (tex->base.width[0] == 64 && tex->base.height[0] == 64) {
       tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
       tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
-#if 0 /* used for tiled display targets */
-      tex->tiled = 1;
-#endif
    } else {
-      tex->stride = round_up(tex->base.nblocksx[0] * pt->block.size, 64);
-      tex->total_nblocksy = tex->base.nblocksy[0];
+      return FALSE;
    }
 
-   /*
-   printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
+   debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
       tex->base.width[0], tex->base.height[0], pt->block.size,
       tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
-   */
 
-   return 1;
+   return TRUE;
+}
+
+static void
+i915_miptree_layout_2d(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned level;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned nblocksx = pt->nblocksx[0];
+   unsigned nblocksy = pt->nblocksy[0];
+
+   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->total_nblocksy = 0;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      i915_miptree_set_level_info(tex, level, 1, width, height, 1);
+      i915_miptree_set_image_offset(tex, level, 0, 0, tex->total_nblocksy);
+
+      nblocksy = round_up(MAX2(2, nblocksy), 2);
+
+      tex->total_nblocksy += nblocksy;
+
+      width = minify(width);
+      height = minify(height);
+      nblocksx = pf_get_nblocksx(&pt->block, width);
+      nblocksy = pf_get_nblocksy(&pt->block, height);
+   }
+}
+
+static void
+i915_miptree_layout_3d(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned level;
+
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned depth = pt->depth[0];
+   unsigned nblocksx = pt->nblocksx[0];
+   unsigned nblocksy = pt->nblocksy[0];
+   unsigned stack_nblocksy = 0;
+
+   /* Calculate the size of a single slice. 
+    */
+   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+
+   /* XXX: hardware expects/requires 9 levels at minimum.
+    */
+   for (level = 0; level <= MAX2(8, pt->last_level); level++) {
+      i915_miptree_set_level_info(tex, level, depth, width, height, depth);
+
+      stack_nblocksy += MAX2(2, nblocksy);
+
+      width = minify(width);
+      height = minify(height);
+      depth = minify(depth);
+      nblocksx = pf_get_nblocksx(&pt->block, width);
+      nblocksy = pf_get_nblocksy(&pt->block, height);
+   }
+
+   /* Fixup depth image_offsets: 
+    */
+   depth = pt->depth[0];
+   for (level = 0; level <= pt->last_level; level++) {
+      unsigned i;
+      for (i = 0; i < depth; i++) 
+         i915_miptree_set_image_offset(tex, level, i, 0, i * stack_nblocksy);
+
+      depth = minify(depth);
+   }
+
+   /* Multiply slice size by texture depth for total size.  It's
+    * remarkable how wasteful of memory the i915 texture layouts
+    * are.  They are largely fixed in the i945.
+    */
+   tex->total_nblocksy = stack_nblocksy * pt->depth[0];
+}
+
+static void
+i915_miptree_layout_cube(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned width = pt->width[0], height = pt->height[0];
+   const unsigned nblocks = pt->nblocksx[0];
+   unsigned level;
+   unsigned face;
+
+   assert(width == height); /* cubemap images are square */
+
+   /* double pitch for cube layouts */
+   tex->stride = round_up(nblocks * pt->block.size * 2, 4);
+   tex->total_nblocksy = nblocks * 4;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      i915_miptree_set_level_info(tex, level, 6, width, height, 1);
+      width /= 2;
+      height /= 2;
+   }
+
+   for (face = 0; face < 6; face++) {
+      unsigned x = initial_offsets[face][0] * nblocks;
+      unsigned y = initial_offsets[face][1] * nblocks;
+      unsigned d = nblocks;
+
+      for (level = 0; level <= pt->last_level; level++) {
+         i915_miptree_set_image_offset(tex, level, face, x, y);
+         d >>= 1;
+         x += step_offsets[face][0] * d;
+         y += step_offsets[face][1] * d;
+      }
+   }
+}
+
+static boolean
+i915_miptree_layout(struct i915_texture * tex)
+{
+   struct pipe_texture *pt = &tex->base;
+
+   switch (pt->target) {
+   case PIPE_TEXTURE_1D:
+   case PIPE_TEXTURE_2D:
+      i915_miptree_layout_2d(tex);
+      break;
+   case PIPE_TEXTURE_3D:
+      i915_miptree_layout_3d(tex);
+      break;
+   case PIPE_TEXTURE_CUBE:
+      i915_miptree_layout_cube(tex);
+      break;
+   default:
+      assert(0);
+      return FALSE;
+   }
+
+   return TRUE;
 }
 
+
+/*
+ * i945 layout functions
+ */
+
+
 static void
-i945_miptree_layout_2d( struct i915_texture *tex )
+i945_miptree_layout_2d(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
    const int align_x = 2, align_y = 4;
@@ -209,10 +346,10 @@ i945_miptree_layout_2d( struct i915_texture *tex )
    unsigned nblocksx = pt->nblocksx[0];
    unsigned nblocksy = pt->nblocksy[0];
 
-   /* used for tiled display targets */
-   if (0)
-      if (i915_displaytarget_layout(tex))
-	 return;
+   /* used for scanouts that need special layouts */
+   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
+      if (i915_scanout_layout(tex))
+         return;
 
    tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
 
@@ -223,11 +360,11 @@ i945_miptree_layout_2d( struct i915_texture *tex )
     */
    if (pt->last_level > 0) {
       unsigned mip1_nblocksx 
-	 = align(pf_get_nblocksx(&pt->block, minify(width)), align_x)
+         = align(pf_get_nblocksx(&pt->block, minify(width)), align_x)
          + pf_get_nblocksx(&pt->block, minify(minify(width)));
 
       if (mip1_nblocksx > nblocksx)
-	 tex->stride = mip1_nblocksx * pt->block.size;
+         tex->stride = mip1_nblocksx * pt->block.size;
    }
 
    /* Pitch must be a whole number of dwords
@@ -249,10 +386,10 @@ i945_miptree_layout_2d( struct i915_texture *tex )
       /* Layout_below: step right after second mipmap level.
        */
       if (level == 1) {
-	 x += align(nblocksx, align_x);
+         x += align(nblocksx, align_x);
       }
       else {
-	 y += nblocksy;
+         y += nblocksy;
       }
 
       width  = minify(width);
@@ -263,6 +400,63 @@ i945_miptree_layout_2d( struct i915_texture *tex )
 }
 
 static void
+i945_miptree_layout_3d(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned depth = pt->depth[0];
+   unsigned nblocksx = pt->nblocksx[0];
+   unsigned nblocksy = pt->nblocksy[0];
+   unsigned pack_x_pitch, pack_x_nr;
+   unsigned pack_y_pitch;
+   unsigned level;
+
+   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+   tex->total_nblocksy = 0;
+
+   pack_y_pitch = MAX2(pt->nblocksy[0], 2);
+   pack_x_pitch = tex->stride / pt->block.size;
+   pack_x_nr = 1;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      int x = 0;
+      int y = 0;
+      unsigned q, j;
+
+      i915_miptree_set_level_info(tex, level, depth, width, height, depth);
+
+      for (q = 0; q < depth;) {
+         for (j = 0; j < pack_x_nr && q < depth; j++, q++) {
+            i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy);
+            x += pack_x_pitch;
+         }
+
+         x = 0;
+         y += pack_y_pitch;
+      }
+
+      tex->total_nblocksy += y;
+
+      if (pack_x_pitch > 4) {
+         pack_x_pitch >>= 1;
+         pack_x_nr <<= 1;
+         assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride);
+      }
+
+      if (pack_y_pitch > 2) {
+         pack_y_pitch >>= 1;
+      }
+
+      width = minify(width);
+      height = minify(height);
+      depth = minify(depth);
+      nblocksx = pf_get_nblocksx(&pt->block, width);
+      nblocksy = pf_get_nblocksy(&pt->block, height);
+   }
+}
+
+static void
 i945_miptree_layout_cube(struct i915_texture *tex)
 {
    struct pipe_texture *pt = &tex->base;
@@ -364,226 +558,44 @@ i945_miptree_layout_cube(struct i915_texture *tex)
 }
 
 static boolean
-i915_miptree_layout(struct i915_texture * tex)
-{
-   struct pipe_texture *pt = &tex->base;
-   unsigned level;
-
-   switch (pt->target) {
-   case PIPE_TEXTURE_CUBE: {
-         const unsigned nblocks = pt->nblocksx[0];
-         unsigned face;
-         unsigned width = pt->width[0], height = pt->height[0];
-
-         assert(width == height); /* cubemap images are square */
-
-         /* double pitch for cube layouts */
-         tex->stride = round_up(nblocks * pt->block.size * 2, 4);
-         tex->total_nblocksy = nblocks * 4;
-
-         for (level = 0; level <= pt->last_level; level++) {
-            i915_miptree_set_level_info(tex, level, 6,
-                                         width, height,
-                                         1);
-            width /= 2;
-            height /= 2;
-         }
-
-         for (face = 0; face < 6; face++) {
-            unsigned x = initial_offsets[face][0] * nblocks;
-            unsigned y = initial_offsets[face][1] * nblocks;
-            unsigned d = nblocks;
-
-            for (level = 0; level <= pt->last_level; level++) {
-               i915_miptree_set_image_offset(tex, level, face, x, y);
-               d >>= 1;
-               x += step_offsets[face][0] * d;
-               y += step_offsets[face][1] * d;
-            }
-         }
-         break;
-      }
-   case PIPE_TEXTURE_3D:{
-         unsigned width = pt->width[0];
-         unsigned height = pt->height[0];
-         unsigned depth = pt->depth[0];
-         unsigned nblocksx = pt->nblocksx[0];
-         unsigned nblocksy = pt->nblocksy[0];
-         unsigned stack_nblocksy = 0;
-
-         /* Calculate the size of a single slice. 
-          */
-         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
-
-         /* XXX: hardware expects/requires 9 levels at minimum.
-          */
-         for (level = 0; level <= MAX2(8, pt->last_level);
-              level++) {
-            i915_miptree_set_level_info(tex, level, depth,
-                                        width, height, depth);
-
-
-            stack_nblocksy += MAX2(2, nblocksy);
-
-            width = minify(width);
-            height = minify(height);
-            depth = minify(depth);
-            nblocksx = pf_get_nblocksx(&pt->block, width);
-            nblocksy = pf_get_nblocksy(&pt->block, height);
-         }
-
-         /* Fixup depth image_offsets: 
-          */
-         depth = pt->depth[0];
-         for (level = 0; level <= pt->last_level; level++) {
-            unsigned i;
-            for (i = 0; i < depth; i++) 
-               i915_miptree_set_image_offset(tex, level, i,
-                                             0, i * stack_nblocksy);
-
-            depth = minify(depth);
-         }
-
-
-         /* Multiply slice size by texture depth for total size.  It's
-          * remarkable how wasteful of memory the i915 texture layouts
-          * are.  They are largely fixed in the i945.
-          */
-         tex->total_nblocksy = stack_nblocksy * pt->depth[0];
-         break;
-      }
-
-   default:{
-         unsigned width = pt->width[0];
-         unsigned height = pt->height[0];
-         unsigned nblocksx = pt->nblocksx[0];
-         unsigned nblocksy = pt->nblocksy[0];
-
-         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
-         tex->total_nblocksy = 0;
-
-         for (level = 0; level <= pt->last_level; level++) {
-            i915_miptree_set_level_info(tex, level, 1,
-                                        width, height, 1);
-            i915_miptree_set_image_offset(tex, level, 0,
-                                          0, tex->total_nblocksy);
-
-            nblocksy = round_up(MAX2(2, nblocksy), 2);
-
-	    tex->total_nblocksy += nblocksy;
-
-            width = minify(width);
-            height = minify(height);
-            nblocksx = pf_get_nblocksx(&pt->block, width);
-            nblocksy = pf_get_nblocksy(&pt->block, height);
-         }
-         break;
-      }
-   }
-   /*
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
-       tex->pitch,
-       tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy);
-   */
-
-   return TRUE;
-}
-
-
-static boolean
 i945_miptree_layout(struct i915_texture * tex)
 {
    struct pipe_texture *pt = &tex->base;
-   unsigned level;
 
    switch (pt->target) {
+   case PIPE_TEXTURE_1D:
+   case PIPE_TEXTURE_2D:
+      i945_miptree_layout_2d(tex);
+      break;
+   case PIPE_TEXTURE_3D:
+      i945_miptree_layout_3d(tex);
+      break;
    case PIPE_TEXTURE_CUBE:
       i945_miptree_layout_cube(tex);
       break;
-   case PIPE_TEXTURE_3D:{
-         unsigned width = pt->width[0];
-         unsigned height = pt->height[0];
-         unsigned depth = pt->depth[0];
-         unsigned nblocksx = pt->nblocksx[0];
-         unsigned nblocksy = pt->nblocksy[0];
-         unsigned pack_x_pitch, pack_x_nr;
-         unsigned pack_y_pitch;
-
-         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
-         tex->total_nblocksy = 0;
-
-         pack_y_pitch = MAX2(pt->nblocksy[0], 2);
-         pack_x_pitch = tex->stride / pt->block.size;
-         pack_x_nr = 1;
-
-         for (level = 0; level <= pt->last_level; level++) {
-            unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6;
-            int x = 0;
-            int y = 0;
-            unsigned q, j;
-
-            i915_miptree_set_level_info(tex, level, nr_images,
-                                        width, height, depth);
-
-            for (q = 0; q < nr_images;) {
-               for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) {
-                  i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy);
-                  x += pack_x_pitch;
-               }
-
-               x = 0;
-               y += pack_y_pitch;
-            }
-
-
-            tex->total_nblocksy += y;
-
-            if (pack_x_pitch > 4) {
-               pack_x_pitch >>= 1;
-               pack_x_nr <<= 1;
-               assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride);
-            }
-
-            if (pack_y_pitch > 2) {
-               pack_y_pitch >>= 1;
-            }
-
-            width = minify(width);
-            height = minify(height);
-            depth = minify(depth);
-            nblocksx = pf_get_nblocksx(&pt->block, width);
-            nblocksy = pf_get_nblocksy(&pt->block, height);
-         }
-         break;
-      }
-
-   case PIPE_TEXTURE_1D:
-   case PIPE_TEXTURE_2D:
-//   case PIPE_TEXTURE_RECTANGLE:
-         i945_miptree_layout_2d(tex);
-         break;
    default:
       assert(0);
       return FALSE;
    }
 
-   /*
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
-       tex->pitch,
-       tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy);
-   */
-
    return TRUE;
 }
 
 
+/*
+ * Screen texture functions
+ */
+
+
 static struct pipe_texture *
 i915_texture_create(struct pipe_screen *screen,
                     const struct pipe_texture *templat)
 {
-   struct i915_screen *i915screen = i915_screen(screen);
+   struct i915_screen *is = i915_screen(screen);
+   struct intel_winsys *iws = is->iws;
    struct i915_texture *tex = CALLOC_STRUCT(i915_texture);
    size_t tex_size;
+   unsigned buf_usage = 0;
 
    if (!tex)
       return NULL;
@@ -595,23 +607,35 @@ i915_texture_create(struct pipe_screen *screen,
    tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]);
    tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]);
    
-   if (i915screen->is_i945) {
+   if (is->is_i945) {
       if (!i945_miptree_layout(tex))
-	 goto fail;
+         goto fail;
    } else {
       if (!i915_miptree_layout(tex))
-	 goto fail;
+         goto fail;
    }
 
    tex_size = tex->stride * tex->total_nblocksy;
 
-   tex->buffer = screen->buffer_create(screen, 64,
-                                    PIPE_BUFFER_USAGE_PIXEL,
-                                    tex_size);
 
+
+   /* for scanouts and cursors, cursors arn't scanouts */
+   if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width[0] != 64)
+      buf_usage = INTEL_NEW_SCANOUT;
+   else
+      buf_usage = INTEL_NEW_TEXTURE;
+
+   tex->buffer = iws->buffer_create(iws, tex_size, 64, buf_usage);
    if (!tex->buffer)
       goto fail;
 
+   /* setup any hw fences */
+   if (tex->hw_tiled) {
+      assert(tex->sw_tiled == INTEL_TILE_NONE);
+      iws->buffer_set_fence_reg(iws, tex->buffer, tex->stride, tex->hw_tiled);
+   }
+
+   
 #if 0
    void *ptr = ws->buffer_map(ws, tex->buffer,
       PIPE_BUFFER_USAGE_CPU_WRITE);
@@ -626,18 +650,56 @@ fail:
    return NULL;
 }
 
+static struct pipe_texture *
+i915_texture_blanket(struct pipe_screen * screen,
+                     const struct pipe_texture *base,
+                     const unsigned *stride,
+                     struct pipe_buffer *buffer)
+{
+#if 0
+   struct i915_texture *tex;
+   assert(screen);
+
+   /* Only supports one type */
+   if (base->target != PIPE_TEXTURE_2D ||
+       base->last_level != 0 ||
+       base->depth[0] != 1) {
+      return NULL;
+   }
+
+   tex = CALLOC_STRUCT(i915_texture);
+   if (!tex)
+      return NULL;
+
+   tex->base = *base;
+   pipe_reference_init(&tex->base.reference, 1);
+   tex->base.screen = screen;
+
+   tex->stride = stride[0];
+
+   i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1);
+   i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
+
+   pipe_buffer_reference(&tex->buffer, buffer);
+
+   return &tex->base;
+#else
+   return NULL;
+#endif
+}
 
 static void
 i915_texture_destroy(struct pipe_texture *pt)
 {
    struct i915_texture *tex = (struct i915_texture *)pt;
+   struct intel_winsys *iws = i915_screen(pt->screen)->iws;
    uint i;
 
    /*
      DBG("%s deleting %p\n", __FUNCTION__, (void *) tex);
    */
 
-   pipe_buffer_reference(&tex->buffer, NULL);
+   iws->buffer_destroy(iws, tex->buffer);
 
    for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++)
       if (tex->image_offset[i])
@@ -646,6 +708,12 @@ i915_texture_destroy(struct pipe_texture *pt)
    FREE(tex);
 }
 
+
+/*
+ * Screen surface functions
+ */
+
+
 static struct pipe_surface *
 i915_get_tex_surface(struct pipe_screen *screen,
                      struct pipe_texture *pt,
@@ -681,11 +749,122 @@ i915_get_tex_surface(struct pipe_screen *screen,
    return ps;
 }
 
-static struct pipe_texture *
-i915_texture_blanket(struct pipe_screen * screen,
-                     const struct pipe_texture *base,
-                     const unsigned *stride,
-                     struct pipe_buffer *buffer)
+static void
+i915_tex_surface_destroy(struct pipe_surface *surf)
+{
+   pipe_texture_reference(&surf->texture, NULL);
+   FREE(surf);
+}
+
+
+/*
+ * Screen transfer functions
+ */
+
+
+static struct pipe_transfer*
+i915_get_tex_transfer(struct pipe_screen *screen,
+                      struct pipe_texture *texture,
+                      unsigned face, unsigned level, unsigned zslice,
+                      enum pipe_transfer_usage usage, unsigned x, unsigned y,
+                      unsigned w, unsigned h)
+{
+   struct i915_texture *tex = (struct i915_texture *)texture;
+   struct i915_transfer *trans;
+   unsigned offset;  /* in bytes */
+
+   if (texture->target == PIPE_TEXTURE_CUBE) {
+      offset = tex->image_offset[level][face];
+   }
+   else if (texture->target == PIPE_TEXTURE_3D) {
+      offset = tex->image_offset[level][zslice];
+   }
+   else {
+      offset = tex->image_offset[level][0];
+      assert(face == 0);
+      assert(zslice == 0);
+   }
+
+   trans = CALLOC_STRUCT(i915_transfer);
+   if (trans) {
+      pipe_texture_reference(&trans->base.texture, texture);
+      trans->base.format = trans->base.format;
+      trans->base.x = x;
+      trans->base.y = y;
+      trans->base.width = w;
+      trans->base.height = h;
+      trans->base.block = texture->block;
+      trans->base.nblocksx = texture->nblocksx[level];
+      trans->base.nblocksy = texture->nblocksy[level];
+      trans->base.stride = tex->stride;
+      trans->offset = offset;
+      trans->base.usage = usage;
+   }
+   return &trans->base;
+}
+
+static void *
+i915_transfer_map(struct pipe_screen *screen,
+                  struct pipe_transfer *transfer)
+{
+   struct i915_texture *tex = (struct i915_texture *)transfer->texture;
+   struct intel_winsys *iws = i915_screen(tex->base.screen)->iws;
+   char *map;
+   boolean write = FALSE;
+
+   if (transfer->usage != PIPE_TRANSFER_READ)
+      write = TRUE;
+
+   map = iws->buffer_map(iws, tex->buffer, write);
+   if (map == NULL)
+      return NULL;
+
+   return map + i915_transfer(transfer)->offset +
+      transfer->y / transfer->block.height * transfer->stride +
+      transfer->x / transfer->block.width * transfer->block.size;
+}
+
+static void
+i915_transfer_unmap(struct pipe_screen *screen,
+                    struct pipe_transfer *transfer)
+{
+   struct i915_texture *tex = (struct i915_texture *)transfer->texture;
+   struct intel_winsys *iws = i915_screen(tex->base.screen)->iws;
+   iws->buffer_unmap(iws, tex->buffer);
+}
+
+static void
+i915_tex_transfer_destroy(struct pipe_transfer *trans)
+{
+   pipe_texture_reference(&trans->texture, NULL);
+   FREE(trans);
+}
+
+
+/*
+ * Other texture functions
+ */
+
+
+void
+i915_init_screen_texture_functions(struct i915_screen *is)
+{
+   is->base.texture_create = i915_texture_create;
+   is->base.texture_blanket = i915_texture_blanket;
+   is->base.texture_destroy = i915_texture_destroy;
+   is->base.get_tex_surface = i915_get_tex_surface;
+   is->base.tex_surface_destroy = i915_tex_surface_destroy;
+   is->base.get_tex_transfer = i915_get_tex_transfer;
+   is->base.transfer_map = i915_transfer_map;
+   is->base.transfer_unmap = i915_transfer_unmap;
+   is->base.tex_transfer_destroy = i915_tex_transfer_destroy;
+}
+
+struct pipe_texture *
+i915_texture_blanket_intel(struct pipe_screen *screen,
+                           struct pipe_texture *base,
+                           unsigned stride,
+                           struct intel_buffer *buffer)
 {
    struct i915_texture *tex;
    assert(screen);
@@ -705,52 +884,28 @@ i915_texture_blanket(struct pipe_screen * screen,
    pipe_reference_init(&tex->base.reference, 1);
    tex->base.screen = screen;
 
-   tex->stride = stride[0];
+   tex->stride = stride;
 
    i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1);
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   pipe_buffer_reference(&tex->buffer, buffer);
+   tex->buffer = buffer;
 
    return &tex->base;
 }
 
-void
-i915_init_texture_functions(struct i915_context *i915)
-{
-//   i915->pipe.texture_update = i915_texture_update;
-}
-
-static void
-i915_tex_surface_destroy(struct pipe_surface *surf)
-{
-   pipe_texture_reference(&surf->texture, NULL);
-   FREE(surf);
-}
-
-void
-i915_init_screen_texture_functions(struct pipe_screen *screen)
-{
-   screen->texture_create = i915_texture_create;
-   screen->texture_destroy = i915_texture_destroy;
-   screen->get_tex_surface = i915_get_tex_surface;
-   screen->texture_blanket = i915_texture_blanket;
-   screen->tex_surface_destroy = i915_tex_surface_destroy;
-}
-
-boolean i915_get_texture_buffer( struct pipe_texture *texture,
-                                 struct pipe_buffer **buf,
-                                 unsigned *stride )
+boolean
+i915_get_texture_buffer_intel(struct pipe_texture *texture,
+                              struct intel_buffer **buffer,
+                              unsigned *stride)
 {
    struct i915_texture *tex = (struct i915_texture *)texture;
 
-   if (!tex)
+   if (!texture)
       return FALSE;
 
-   pipe_buffer_reference(buf, tex->buffer);
-
-   if (stride)
-      *stride = tex->stride;
+   *stride = tex->stride;
+   *buffer = tex->buffer;
 
    return TRUE;
 }
diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h
index 7225016a9f..51a1dd984c 100644
--- a/src/gallium/drivers/i915simple/i915_texture.h
+++ b/src/gallium/drivers/i915simple/i915_texture.h
@@ -28,16 +28,9 @@
 #ifndef I915_TEXTURE_H
 #define I915_TEXTURE_H
 
-struct i915_context;
-struct pipe_screen;
-
+struct i915_screen;
 
 extern void
-i915_init_texture_functions(struct i915_context *i915);
-
-
-extern void
-i915_init_screen_texture_functions(struct pipe_screen *screen);
-
+i915_init_screen_texture_functions(struct i915_screen *is);
 
 #endif /* I915_TEXTURE_H */
diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h
deleted file mode 100644
index ff5b34f193..0000000000
--- a/src/gallium/drivers/i915simple/i915_winsys.h
+++ /dev/null
@@ -1,143 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-/**
- * \file
- * This is the interface that i915simple requires any window system
- * hosting it to implement.  This is the only include file in i915simple
- * which is public.
- *
- * This isn't currently true as the winsys needs i915_batchbuffer.h
- */
-
-#ifndef I915_WINSYS_H
-#define I915_WINSYS_H
-
-
-#include "pipe/p_defines.h"
-
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-   
-/* Pipe drivers are independent of both GL and the window system.
- * The window system provides a buffer manager and a set of additional
- * hooks for things like command buffer submission, etc.
- *
- * There clearly has to be some agreement between the window system
- * driver and the hardware driver about the format of command buffers,
- * etc.
- */
-
-struct i915_batchbuffer;
-struct pipe_texture;
-struct pipe_buffer;
-struct pipe_fence_handle;
-struct pipe_winsys;
-struct pipe_screen;
-
-
-/**
- * Additional winsys interface for i915simple.
- *
- * It is an over-simple batchbuffer mechanism.  Will want to improve the
- * performance of this, perhaps based on the cmdstream stuff.  It
- * would be pretty impossible to implement swz on top of this
- * interface.
- *
- * Will also need additions/changes to implement static/dynamic
- * indirect state.
- */
-struct i915_winsys {
-
-   void (*destroy)( struct i915_winsys *sws );
-   
-   /**
-    * Get the current batch buffer from the winsys.
-    */
-   struct i915_batchbuffer *(*batch_get)( struct i915_winsys *sws );
-
-   /**
-    * Emit a relocation to a buffer.
-    * 
-    * Used not only when the buffer addresses are not pinned, but also to 
-    * ensure refered buffers will not be destroyed until the current batch 
-    * buffer execution is finished.
-    *
-    * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and 
-    * I915_BUFFER_ACCESS_READ macros.
-    */
-   void (*batch_reloc)( struct i915_winsys *sws,
-			struct pipe_buffer *buf,
-			unsigned access_flags,
-			unsigned delta );
-
-   /**
-    * Flush the batch.
-    */
-   void (*batch_flush)( struct i915_winsys *sws,
-                        struct pipe_fence_handle **fence );
-};
-
-#define I915_BUFFER_ACCESS_WRITE   0x1 
-#define I915_BUFFER_ACCESS_READ    0x2
-
-#define I915_BUFFER_USAGE_LIT_VERTEX  (PIPE_BUFFER_USAGE_CUSTOM << 0)
-
-
-/**
- * Create i915 pipe_screen.
- */
-struct pipe_screen *i915_create_screen( struct pipe_winsys *winsys,
-                                        uint pci_id );
-
-/**
- * Create a i915 pipe_context.
- */
-struct pipe_context *i915_create_context( struct pipe_screen *screen,
-                                          struct pipe_winsys *winsys,
-                                          struct i915_winsys *i915 );
-
-/**
- * Used for the winsys to get the buffer used for a texture
- * and also the stride used for the texture.
- *
- * Buffer is referenced for you so you need to unref after use.
- *
- * This is needed for example kms.
- */
-boolean i915_get_texture_buffer( struct pipe_texture *texture,
-                                 struct pipe_buffer **buf,
-                                 unsigned *stride );
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif
diff --git a/src/gallium/drivers/i915simple/intel_batchbuffer.h b/src/gallium/drivers/i915simple/intel_batchbuffer.h
new file mode 100644
index 0000000000..db12dfd2ac
--- /dev/null
+++ b/src/gallium/drivers/i915simple/intel_batchbuffer.h
@@ -0,0 +1,87 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_BATCH_H
+#define INTEL_BATCH_H
+
+#include "intel_winsys.h"
+
+static INLINE boolean
+intel_batchbuffer_check(struct intel_batchbuffer *batch,
+                        size_t dwords,
+                        size_t relocs)
+{
+   return dwords * 4 <= batch->size - (batch->ptr - batch->map) &&
+          relocs <= (batch->max_relocs - batch->relocs);
+}
+
+static INLINE size_t
+intel_batchbuffer_space(struct intel_batchbuffer *batch)
+{
+   return batch->size - (batch->ptr - batch->map);
+}
+
+static INLINE void
+intel_batchbuffer_dword(struct intel_batchbuffer *batch,
+                        unsigned dword)
+{
+   if (intel_batchbuffer_space(batch) < 4)
+      return;
+
+   *(unsigned *)batch->ptr = dword;
+   batch->ptr += 4;
+}
+
+static INLINE void
+intel_batchbuffer_write(struct intel_batchbuffer *batch,
+                        void *data,
+                        size_t size)
+{
+   if (intel_batchbuffer_space(batch) < size)
+      return;
+
+   memcpy(data, batch->ptr, size);
+   batch->ptr += size;
+}
+
+static INLINE int
+intel_batchbuffer_reloc(struct intel_batchbuffer *batch,
+                        struct intel_buffer *buffer,
+                        enum intel_buffer_usage usage,
+                        size_t offset)
+{
+   return batch->iws->batchbuffer_reloc(batch, buffer, usage, offset);
+}
+
+static INLINE void
+intel_batchbuffer_flush(struct intel_batchbuffer *batch,
+                        struct pipe_fence_handle **fence)
+{
+   batch->iws->batchbuffer_flush(batch, fence);
+}
+
+#endif
diff --git a/src/gallium/drivers/i915simple/intel_winsys.h b/src/gallium/drivers/i915simple/intel_winsys.h
new file mode 100644
index 0000000000..f949f52a9c
--- /dev/null
+++ b/src/gallium/drivers/i915simple/intel_winsys.h
@@ -0,0 +1,219 @@
+/**************************************************************************
+ *
+ * Copyright © 2009 Jakob Bornecrantz
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef INTEL_WINSYS_H
+#define INTEL_WINSYS_H
+
+#include "pipe/p_compiler.h"
+
+struct intel_winsys;
+struct intel_buffer;
+struct intel_batchbuffer;
+struct pipe_texture;
+struct pipe_fence_handle;
+
+enum intel_buffer_usage
+{
+   /* use on textures */
+   INTEL_USAGE_RENDER    = 0x01,
+   INTEL_USAGE_SAMPLER   = 0x02,
+   INTEL_USAGE_2D_TARGET = 0x04,
+   INTEL_USAGE_2D_SOURCE = 0x08,
+   /* use on vertex */
+   INTEL_USAGE_VERTEX    = 0x10,
+};
+
+enum intel_buffer_type
+{
+   INTEL_NEW_TEXTURE,
+   INTEL_NEW_SCANOUT, /**< a texture used for scanning out from */
+   INTEL_NEW_VERTEX,
+};
+
+enum intel_buffer_tile
+{
+   INTEL_TILE_NONE,
+   INTEL_TILE_X,
+   INTEL_TILE_Y,
+};
+
+struct intel_batchbuffer {
+
+   struct intel_winsys *iws;
+
+   /**
+    * Values exported to speed up the writing the batchbuffer,
+    * instead of having to go trough a accesor function for
+    * each dword written.
+    */
+   /*{@*/
+   uint8_t *map;
+   uint8_t *ptr;
+   size_t size;
+
+   size_t relocs;
+   size_t max_relocs;
+   /*@}*/
+};
+
+struct intel_winsys {
+
+   /**
+    * Batchbuffer functions.
+    */
+   /*@{*/
+   /**
+    * Create a new batchbuffer.
+    */
+   struct intel_batchbuffer *(*batchbuffer_create)(struct intel_winsys *iws);
+
+   /**
+    * Emit a relocation to a buffer.
+    * Target position in batchbuffer is the same as ptr.
+    *
+    * @batch
+    * @reloc buffer address to be inserted into target.
+    * @usage how is the hardware going to use the buffer.
+    * @offset add this to the reloc buffers address
+    * @target buffer where to write the address, null for batchbuffer.
+    */
+   int (*batchbuffer_reloc)(struct intel_batchbuffer *batch,
+                            struct intel_buffer *reloc,
+                            enum intel_buffer_usage usage,
+                            unsigned offset);
+
+   /**
+    * Flush a bufferbatch.
+    */
+   void (*batchbuffer_flush)(struct intel_batchbuffer *batch,
+                             struct pipe_fence_handle **fence);
+
+   /**
+    * Destroy a batchbuffer.
+    */
+   void (*batchbuffer_destroy)(struct intel_batchbuffer *batch);
+   /*@}*/
+
+
+   /**
+    * Buffer functions.
+    */
+   /*@{*/
+   /**
+    * Create a buffer.
+    */
+   struct intel_buffer *(*buffer_create)(struct intel_winsys *iws,
+                                         unsigned size, unsigned alignment,
+                                         enum intel_buffer_type type);
+
+   /**
+    * Fence a buffer with a fence reg.
+    * Not to be confused with pipe_fence_handle.
+    */
+   int (*buffer_set_fence_reg)(struct intel_winsys *iws,
+                               struct intel_buffer *buffer,
+                               unsigned stride,
+                               enum intel_buffer_tile tile);
+
+   /**
+    * Map a buffer.
+    */
+   void *(*buffer_map)(struct intel_winsys *iws,
+                       struct intel_buffer *buffer,
+                       boolean write);
+
+   /**
+    * Unmap a buffer.
+    */
+   void (*buffer_unmap)(struct intel_winsys *iws,
+                        struct intel_buffer *buffer);
+
+   void (*buffer_destroy)(struct intel_winsys *iws,
+                          struct intel_buffer *buffer);
+   /*@}*/
+
+
+   /**
+    * Fence functions.
+    */
+   /*@{*/
+   /**
+    * Reference fence and set ptr to fence.
+    */
+   void (*fence_reference)(struct intel_winsys *iws,
+                           struct pipe_fence_handle **ptr,
+                           struct pipe_fence_handle *fence);
+
+   /**
+    * Check if a fence has finished.
+    */
+   int (*fence_signalled)(struct intel_winsys *iws,
+                          struct pipe_fence_handle *fence);
+
+   /**
+    * Wait on a fence to finish.
+    */
+   int (*fence_finish)(struct intel_winsys *iws,
+                       struct pipe_fence_handle *fence);
+   /*@}*/
+
+
+   /**
+    * Destroy the winsys.
+    */
+   void (*destroy)(struct intel_winsys *iws);
+};
+
+
+/**
+ * Create i915 pipe_screen.
+ */
+struct pipe_screen *i915_create_screen(struct intel_winsys *iws, unsigned pci_id);
+
+/**
+ * Create a i915 pipe_context.
+ */
+struct pipe_context *i915_create_context(struct pipe_screen *screen);
+
+/**
+ * Get the intel_winsys buffer backing the texture.
+ *
+ * TODO UGLY
+ */
+boolean i915_get_texture_buffer_intel(struct pipe_texture *texture,
+                                      struct intel_buffer **buffer,
+                                      unsigned *stride);
+
+/**
+ * Wrap a intel_winsys buffer with a texture blanket.
+ *
+ * TODO UGLY
+ */
+struct pipe_texture * i915_texture_blanket_intel(struct pipe_screen *screen,
+                                                 struct pipe_texture *tmplt,
+                                                 unsigned pitch,
+                                                 struct intel_buffer *buffer);
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c
index 648aaa0da5..49d80cb41c 100644
--- a/src/gallium/drivers/i965simple/brw_draw.c
+++ b/src/gallium/drivers/i965simple/brw_draw.c
@@ -35,6 +35,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/internal/p_winsys_screen.h"
+#include "util/u_prim.h"
 
 static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = {
    _3DPRIM_POINTLIST,
@@ -50,20 +51,6 @@ static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = {
 };
 
 
-static const int reduced_prim[PIPE_PRIM_POLYGON+1] = {
-   PIPE_PRIM_POINTS,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES
-};
-
-
 /* When the primitive changes, set a state bit and re-validate.  Not
  * the nicest and would rather deal with this by having all the
  * programs be immune to the active primitive (ie. cope with all
@@ -85,8 +72,8 @@ static void brw_set_prim(struct brw_context *brw, int prim)
       brw->primitive = prim;
       brw->state.dirty.brw |= BRW_NEW_PRIMITIVE;
 
-      if (reduced_prim[prim] != brw->reduced_primitive) {
-	 brw->reduced_primitive = reduced_prim[prim];
+      if (u_reduced_prim(prim) != brw->reduced_primitive) {
+	 brw->reduced_primitive = u_reduced_prim(prim);
 	 brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE;
       }
 
diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c
index 511779dbfa..724a69b2ee 100644
--- a/src/gallium/drivers/i965simple/brw_surface.c
+++ b/src/gallium/drivers/i965simple/brw_surface.c
@@ -60,7 +60,7 @@ brw_surface_copy(struct pipe_context *pipe,
                                                        src,
                                                        PIPE_BUFFER_USAGE_CPU_READ );
       
-      pipe_copy_rect(dst_map,
+      util_copy_rect(dst_map,
                      &dst->block,
                      dst->stride,
                      dstx, dsty,
@@ -99,7 +99,7 @@ brw_surface_fill(struct pipe_context *pipe,
                                                  dst,
                                                  PIPE_BUFFER_USAGE_CPU_WRITE );
 
-      pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value);
+      util_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value);
 
       pipe->screen->surface_unmap(pipe->screen, dst);
    }
diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c
index 8aea8c0558..998ffaeac4 100644
--- a/src/gallium/drivers/i965simple/brw_tex_layout.c
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.c
@@ -65,11 +65,6 @@ unsigned intel_compressed_alignment(unsigned internalFormat)
 }
 #endif
 
-static unsigned minify( unsigned d )
-{
-   return MAX2(1, d>>1);
-}
-
 
 static void intel_miptree_set_image_offset(struct brw_texture *tex,
                                            unsigned level,
diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c
index e03d653482..3ee82d95b3 100644
--- a/src/gallium/drivers/i965simple/brw_vs_emit.c
+++ b/src/gallium/drivers/i965simple/brw_vs_emit.c
@@ -1294,10 +1294,10 @@ void brw_vs_emit(struct brw_vs_compile *c)
       case TGSI_TOKEN_TYPE_IMMEDIATE: {
          struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate;
          assert(imm->Immediate.NrTokens == 4 + 1);
-         c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float;
-         c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float;
-         c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float;
-         c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u[0].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u[1].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u[2].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u[3].Float;
          c->prog_data.num_imm++;
       }
          break;
diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c
index ab6410aa60..db75963932 100644
--- a/src/gallium/drivers/i965simple/brw_wm_glsl.c
+++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c
@@ -947,7 +947,7 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c,
 #endif
 
       break;
-   case TGSI_OPCODE_LOOP:
+   case TGSI_OPCODE_BGNFOR:
       c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8);
       break;
    case TGSI_OPCODE_BRK:
@@ -958,11 +958,11 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c,
       brw_CONT(p);
       brw_set_predicate_control(p, BRW_PREDICATE_NONE);
       break;
-   case TGSI_OPCODE_ENDLOOP:
+   case TGSI_OPCODE_ENDFOR:
       c->loop_insn--;
       c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]);
       /* patch all the BREAK instructions from
-	 last BEGINLOOP */
+         last BGNFOR */
       while (c->inst0 > c->loop_inst[c->loop_insn]) {
 	 c->inst0--;
 	 if (c->inst0->header.opcode == BRW_OPCODE_BREAK) {
diff --git a/src/gallium/drivers/identity/Makefile b/src/gallium/drivers/identity/Makefile
new file mode 100644
index 0000000000..e32b9102e5
--- /dev/null
+++ b/src/gallium/drivers/identity/Makefile
@@ -0,0 +1,12 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = identity
+
+C_SOURCES = \
+	id_objects.c \
+	id_context.c \
+	id_screen.c \
+	id_drm.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/identity/SConscript b/src/gallium/drivers/identity/SConscript
new file mode 100644
index 0000000000..7f079dd0a8
--- /dev/null
+++ b/src/gallium/drivers/identity/SConscript
@@ -0,0 +1,13 @@
+Import('*')
+
+env = env.Clone()
+
+identity = env.ConvenienceLibrary(
+	target = 'identity',
+	source = [
+		'id_screen.c',
+		'id_context.c',
+		'id_objects.c',
+	])
+
+Export('identity')
diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c
new file mode 100644
index 0000000000..4e700089e3
--- /dev/null
+++ b/src/gallium/drivers/identity/id_context.c
@@ -0,0 +1,719 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_context.h"
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_context.h"
+#include "id_objects.h"
+
+
+static void
+identity_destroy(struct pipe_context *_pipe)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->destroy(pipe);
+
+   free(id_pipe);
+}
+
+static void
+identity_set_edgeflags(struct pipe_context *_pipe,
+                       const unsigned *bitfield)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_edgeflags(pipe,
+                       bitfield);
+}
+
+static boolean
+identity_draw_arrays(struct pipe_context *_pipe,
+                     unsigned prim,
+                     unsigned start,
+                     unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->draw_arrays(pipe,
+                            prim,
+                            start,
+                            count);
+}
+
+static boolean
+identity_draw_elements(struct pipe_context *_pipe,
+                       struct pipe_buffer *_indexBuffer,
+                       unsigned indexSize,
+                       unsigned prim,
+                       unsigned start,
+                       unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_indexBuffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *indexBuffer = id_buffer->buffer;
+
+   return pipe->draw_elements(pipe,
+                              indexBuffer,
+                              indexSize,
+                              prim,
+                              start,
+                              count);
+}
+
+static boolean
+identity_draw_range_elements(struct pipe_context *_pipe,
+                             struct pipe_buffer *_indexBuffer,
+                             unsigned indexSize,
+                             unsigned minIndex,
+                             unsigned maxIndex,
+                             unsigned mode,
+                             unsigned start,
+                             unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_indexBuffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *indexBuffer = id_buffer->buffer;
+
+   return pipe->draw_range_elements(pipe,
+                                    indexBuffer,
+                                    indexSize,
+                                    minIndex,
+                                    maxIndex,
+                                    mode,
+                                    start,
+                                    count);
+}
+
+static struct pipe_query *
+identity_create_query(struct pipe_context *_pipe,
+                      unsigned query_type)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_query(pipe,
+                             query_type);
+}
+
+static void
+identity_destroy_query(struct pipe_context *_pipe,
+                       struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->destroy_query(pipe,
+                       query);
+}
+
+static void
+identity_begin_query(struct pipe_context *_pipe,
+                     struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->begin_query(pipe,
+                     query);
+}
+
+static void
+identity_end_query(struct pipe_context *_pipe,
+                   struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->end_query(pipe,
+                   query);
+}
+
+static boolean
+identity_get_query_result(struct pipe_context *_pipe,
+                          struct pipe_query *query,
+                          boolean wait,
+                          uint64_t *result)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->get_query_result(pipe,
+                                 query,
+                                 wait,
+                                 result);
+}
+
+static void *
+identity_create_blend_state(struct pipe_context *_pipe,
+                            const struct pipe_blend_state *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_blend_state(pipe,
+                                   blend);
+}
+
+static void
+identity_bind_blend_state(struct pipe_context *_pipe,
+                          void *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_blend_state(pipe,
+                              blend);
+}
+
+static void
+identity_delete_blend_state(struct pipe_context *_pipe,
+                            void *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_blend_state(pipe,
+                            blend);
+}
+
+static void *
+identity_create_sampler_state(struct pipe_context *_pipe,
+                              const struct pipe_sampler_state *sampler)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_sampler_state(pipe,
+                                     sampler);
+}
+
+static void
+identity_bind_sampler_states(struct pipe_context *_pipe,
+                             unsigned num,
+                             void **samplers)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_sampler_states(pipe,
+                             num,
+                             samplers);
+}
+
+static void
+identity_delete_sampler_state(struct pipe_context *_pipe,
+                              void *sampler)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_sampler_state(pipe,
+                              sampler);
+}
+
+static void *
+identity_create_rasterizer_state(struct pipe_context *_pipe,
+                                 const struct pipe_rasterizer_state *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_rasterizer_state(pipe,
+                                        rasterizer);
+}
+
+static void
+identity_bind_rasterizer_state(struct pipe_context *_pipe,
+                               void *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_rasterizer_state(pipe,
+                               rasterizer);
+}
+
+static void
+identity_delete_rasterizer_state(struct pipe_context *_pipe,
+                                 void *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_rasterizer_state(pipe,
+                                 rasterizer);
+}
+
+static void *
+identity_create_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                          const struct pipe_depth_stencil_alpha_state *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_depth_stencil_alpha_state(pipe,
+                                                 depth_stencil_alpha);
+}
+
+static void
+identity_bind_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                        void *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_depth_stencil_alpha_state(pipe,
+                                        depth_stencil_alpha);
+}
+
+static void
+identity_delete_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                          void *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_depth_stencil_alpha_state(pipe,
+                                          depth_stencil_alpha);
+}
+
+static void *
+identity_create_fs_state(struct pipe_context *_pipe,
+                         const struct pipe_shader_state *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_fs_state(pipe,
+                                fs);
+}
+
+static void
+identity_bind_fs_state(struct pipe_context *_pipe,
+                       void *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_fs_state(pipe,
+                       fs);
+}
+
+static void
+identity_delete_fs_state(struct pipe_context *_pipe,
+                         void *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_fs_state(pipe,
+                         fs);
+}
+
+static void *
+identity_create_vs_state(struct pipe_context *_pipe,
+                         const struct pipe_shader_state *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_vs_state(pipe,
+                                vs);
+}
+
+static void
+identity_bind_vs_state(struct pipe_context *_pipe,
+                       void *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_vs_state(pipe,
+                       vs);
+}
+
+static void
+identity_delete_vs_state(struct pipe_context *_pipe,
+                         void *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_vs_state(pipe,
+                         vs);
+}
+
+static void
+identity_set_blend_color(struct pipe_context *_pipe,
+                         const struct pipe_blend_color *blend_color)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_blend_color(pipe,
+                         blend_color);
+}
+
+static void
+identity_set_clip_state(struct pipe_context *_pipe,
+                        const struct pipe_clip_state *clip)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_clip_state(pipe,
+                        clip);
+}
+
+static void
+identity_set_constant_buffer(struct pipe_context *_pipe,
+                             uint shader,
+                             uint index,
+                             const struct pipe_constant_buffer *_buffer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_constant_buffer unwrapped_buffer;
+   struct pipe_constant_buffer *buffer = NULL;
+
+   /* unwrap the input state */
+   if (_buffer) {
+      unwrapped_buffer.buffer = identity_buffer_unwrap(_buffer->buffer);
+      buffer = &unwrapped_buffer;
+   }
+
+   pipe->set_constant_buffer(pipe,
+                             shader,
+                             index,
+                             buffer);
+}
+
+static void
+identity_set_framebuffer_state(struct pipe_context *_pipe,
+                               const struct pipe_framebuffer_state *_state)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_framebuffer_state unwrapped_state;
+   struct pipe_framebuffer_state *state = NULL;
+   unsigned i;
+
+   /* unwrap the input state */
+   if (_state) {
+      memcpy(&unwrapped_state, _state, sizeof(unwrapped_state));
+      for(i = 0; i < _state->nr_cbufs; i++)
+         unwrapped_state.cbufs[i] = identity_surface_unwrap(_state->cbufs[i]);
+      for (; i < PIPE_MAX_COLOR_BUFS; i++)
+         unwrapped_state.cbufs[i] = NULL;
+      unwrapped_state.zsbuf = identity_surface_unwrap(_state->zsbuf);
+      state = &unwrapped_state;
+   }
+
+   pipe->set_framebuffer_state(pipe,
+                               state);
+}
+
+static void
+identity_set_polygon_stipple(struct pipe_context *_pipe,
+                             const struct pipe_poly_stipple *poly_stipple)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_polygon_stipple(pipe,
+                             poly_stipple);
+}
+
+static void
+identity_set_scissor_state(struct pipe_context *_pipe,
+                           const struct pipe_scissor_state *scissor)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_scissor_state(pipe,
+                           scissor);
+}
+
+static void
+identity_set_viewport_state(struct pipe_context *_pipe,
+                            const struct pipe_viewport_state *viewport)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_viewport_state(pipe,
+                            viewport);
+}
+
+static void
+identity_set_sampler_textures(struct pipe_context *_pipe,
+                              unsigned num_textures,
+                              struct pipe_texture **_textures)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
+   struct pipe_texture **textures = NULL;
+   unsigned i;
+
+   if (_textures) {
+      for (i = 0; i < num_textures; i++)
+         unwrapped_textures[i] = identity_texture_unwrap(_textures[i]);
+      for (; i < PIPE_MAX_SAMPLERS; i++)
+         unwrapped_textures[i] = NULL;
+
+      textures = unwrapped_textures;
+   }
+
+   pipe->set_sampler_textures(pipe,
+                              num_textures,
+                              textures);
+}
+
+static void
+identity_set_vertex_buffers(struct pipe_context *_pipe,
+                            unsigned num_buffers,
+                            const struct pipe_vertex_buffer *_buffers)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_vertex_buffer unwrapped_buffers[PIPE_MAX_SHADER_INPUTS];
+   struct pipe_vertex_buffer *buffers = NULL;
+   unsigned i;
+
+   if (num_buffers) {
+      memcpy(unwrapped_buffers, _buffers, num_buffers * sizeof(*_buffers));
+      for (i = 0; i < num_buffers; i++)
+         unwrapped_buffers[i].buffer = identity_buffer_unwrap(_buffers[i].buffer);
+      buffers = unwrapped_buffers;
+   }
+
+   pipe->set_vertex_buffers(pipe,
+                            num_buffers,
+                            buffers);
+}
+
+static void
+identity_set_vertex_elements(struct pipe_context *_pipe,
+                             unsigned num_elements,
+                             const struct pipe_vertex_element *vertex_elements)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_vertex_elements(pipe,
+                             num_elements,
+                             vertex_elements);
+}
+
+static void
+identity_surface_copy(struct pipe_context *_pipe,
+                      struct pipe_surface *_dst,
+                      unsigned dstx,
+                      unsigned dsty,
+                      struct pipe_surface *_src,
+                      unsigned srcx,
+                      unsigned srcy,
+                      unsigned width,
+                      unsigned height)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_surface *id_surface_dst = identity_surface(_dst);
+   struct identity_surface *id_surface_src = identity_surface(_src);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_surface *dst = id_surface_dst->surface;
+   struct pipe_surface *src = id_surface_src->surface;
+
+   pipe->surface_copy(pipe,
+                      dst,
+                      dstx,
+                      dsty,
+                      src,
+                      srcx,
+                      srcy,
+                      width,
+                      height);
+}
+
+static void
+identity_surface_fill(struct pipe_context *_pipe,
+                      struct pipe_surface *_dst,
+                      unsigned dstx,
+                      unsigned dsty,
+                      unsigned width,
+                      unsigned height,
+                      unsigned value)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_surface *id_surface_dst = identity_surface(_dst);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_surface *dst = id_surface_dst->surface;
+
+   pipe->surface_fill(pipe,
+                      dst,
+                      dstx,
+                      dsty,
+                      width,
+                      height,
+                      value);
+}
+
+static void
+identity_clear(struct pipe_context *_pipe,
+               unsigned buffers,
+               const float *rgba,
+               double depth,
+               unsigned stencil)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->clear(pipe,
+               buffers,
+               rgba,
+               depth,
+               stencil);
+}
+
+static void
+identity_flush(struct pipe_context *_pipe,
+               unsigned flags,
+               struct pipe_fence_handle **fence)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->flush(pipe,
+               flags,
+               fence);
+}
+
+static unsigned int
+identity_is_texture_referenced(struct pipe_context *_pipe,
+                               struct pipe_texture *_texture,
+                               unsigned face,
+                               unsigned level)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_texture *texture = id_texture->texture;
+
+   return pipe->is_texture_referenced(pipe,
+                                      texture,
+                                      face,
+                                      level);
+}
+
+static unsigned int
+identity_is_buffer_referenced(struct pipe_context *_pipe,
+                              struct pipe_buffer *_buffer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return pipe->is_buffer_referenced(pipe,
+                                     buffer);
+}
+
+struct pipe_context *
+identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
+{
+   struct identity_context *id_pipe;
+   (void)identity_screen(_screen);
+
+   id_pipe = CALLOC_STRUCT(identity_context);
+   if (!id_pipe) {
+      return NULL;
+   }
+
+   id_pipe->base.winsys = NULL;
+   id_pipe->base.screen = _screen;
+   id_pipe->base.priv = pipe->priv;
+   id_pipe->base.draw = NULL;
+
+   id_pipe->base.destroy = identity_destroy;
+   id_pipe->base.set_edgeflags = identity_set_edgeflags;
+   id_pipe->base.draw_arrays = identity_draw_arrays;
+   id_pipe->base.draw_elements = identity_draw_elements;
+   id_pipe->base.draw_range_elements = identity_draw_range_elements;
+   id_pipe->base.create_query = identity_create_query;
+   id_pipe->base.destroy_query = identity_destroy_query;
+   id_pipe->base.begin_query = identity_begin_query;
+   id_pipe->base.end_query = identity_end_query;
+   id_pipe->base.get_query_result = identity_get_query_result;
+   id_pipe->base.create_blend_state = identity_create_blend_state;
+   id_pipe->base.bind_blend_state = identity_bind_blend_state;
+   id_pipe->base.delete_blend_state = identity_delete_blend_state;
+   id_pipe->base.create_sampler_state = identity_create_sampler_state;
+   id_pipe->base.bind_sampler_states = identity_bind_sampler_states;
+   id_pipe->base.delete_sampler_state = identity_delete_sampler_state;
+   id_pipe->base.create_rasterizer_state = identity_create_rasterizer_state;
+   id_pipe->base.bind_rasterizer_state = identity_bind_rasterizer_state;
+   id_pipe->base.delete_rasterizer_state = identity_delete_rasterizer_state;
+   id_pipe->base.create_depth_stencil_alpha_state = identity_create_depth_stencil_alpha_state;
+   id_pipe->base.bind_depth_stencil_alpha_state = identity_bind_depth_stencil_alpha_state;
+   id_pipe->base.delete_depth_stencil_alpha_state = identity_delete_depth_stencil_alpha_state;
+   id_pipe->base.create_fs_state = identity_create_fs_state;
+   id_pipe->base.bind_fs_state = identity_bind_fs_state;
+   id_pipe->base.delete_fs_state = identity_delete_fs_state;
+   id_pipe->base.create_vs_state = identity_create_vs_state;
+   id_pipe->base.bind_vs_state = identity_bind_vs_state;
+   id_pipe->base.delete_vs_state = identity_delete_vs_state;
+   id_pipe->base.set_blend_color = identity_set_blend_color;
+   id_pipe->base.set_clip_state = identity_set_clip_state;
+   id_pipe->base.set_constant_buffer = identity_set_constant_buffer;
+   id_pipe->base.set_framebuffer_state = identity_set_framebuffer_state;
+   id_pipe->base.set_polygon_stipple = identity_set_polygon_stipple;
+   id_pipe->base.set_scissor_state = identity_set_scissor_state;
+   id_pipe->base.set_viewport_state = identity_set_viewport_state;
+   id_pipe->base.set_sampler_textures = identity_set_sampler_textures;
+   id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers;
+   id_pipe->base.set_vertex_elements = identity_set_vertex_elements;
+   id_pipe->base.surface_copy = identity_surface_copy;
+   id_pipe->base.surface_fill = identity_surface_fill;
+   id_pipe->base.clear = identity_clear;
+   id_pipe->base.flush = identity_flush;
+   id_pipe->base.is_texture_referenced = identity_is_texture_referenced;
+   id_pipe->base.is_buffer_referenced = identity_is_buffer_referenced;
+
+   id_pipe->pipe = pipe;
+
+   return &id_pipe->base;
+}
diff --git a/src/gallium/drivers/identity/id_context.h b/src/gallium/drivers/identity/id_context.h
new file mode 100644
index 0000000000..75b73fc7df
--- /dev/null
+++ b/src/gallium/drivers/identity/id_context.h
@@ -0,0 +1,48 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_CONTEXT_H
+#define ID_CONTEXT_H
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+
+
+struct identity_context {
+   struct pipe_context base;  /**< base class */
+
+   struct pipe_context *pipe;
+};
+
+
+static INLINE struct identity_context *
+identity_context(struct pipe_context *pipe)
+{
+   return (struct identity_context *)pipe;
+}
+
+#endif /* ID_CONTEXT_H */
diff --git a/src/gallium/drivers/identity/id_drm.c b/src/gallium/drivers/identity/id_drm.c
new file mode 100644
index 0000000000..14f68ac0d0
--- /dev/null
+++ b/src/gallium/drivers/identity/id_drm.c
@@ -0,0 +1,173 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "state_tracker/drm_api.h"
+
+#include "util/u_memory.h"
+#include "identity/id_drm.h"
+#include "identity/id_screen.h"
+#include "identity/id_public.h"
+#include "identity/id_screen.h"
+#include "identity/id_objects.h"
+
+struct identity_drm_api
+{
+   struct drm_api base;
+
+   struct drm_api *api;
+};
+
+static INLINE struct identity_drm_api *
+identity_drm_api(struct drm_api *_api)
+{
+   return (struct identity_drm_api *)_api;
+}
+
+static struct pipe_screen *
+identity_drm_create_screen(struct drm_api *_api, int fd,
+                           struct drm_create_screen_arg *arg)
+{
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct drm_api *api = id_api->api;
+   struct pipe_screen *screen;
+
+   if (arg && arg->mode != DRM_CREATE_NORMAL)
+      return NULL;
+
+   screen = api->create_screen(api, fd, arg);
+
+   return identity_screen_create(screen);
+}
+
+static struct pipe_context *
+identity_drm_create_context(struct drm_api *_api,
+                            struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct drm_api *api = id_api->api;
+   struct pipe_context *pipe;
+
+   pipe = api->create_context(api, screen);
+
+   pipe = identity_context_create(_screen, pipe);
+
+   return pipe;
+}
+
+static struct pipe_texture *
+identity_drm_texture_from_shared_handle(struct drm_api *_api,
+                                        struct pipe_screen *_screen,
+                                        struct pipe_texture *templ,
+                                        const char *name,
+                                        unsigned stride,
+                                        unsigned handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct drm_api *api = id_api->api;
+   struct pipe_texture *result;
+
+   result = api->texture_from_shared_handle(api, screen, templ, name, stride, handle);
+
+   result = identity_texture_create(identity_screen(_screen), result);
+
+   return result;
+}
+
+static boolean
+identity_drm_shared_handle_from_texture(struct drm_api *_api,
+                                        struct pipe_screen *_screen,
+                                        struct pipe_texture *_texture,
+                                        unsigned *stride,
+                                        unsigned *handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct drm_api *api = id_api->api;
+
+   return api->shared_handle_from_texture(api, screen, texture, stride, handle);
+}
+
+static boolean
+identity_drm_local_handle_from_texture(struct drm_api *_api,
+                                       struct pipe_screen *_screen,
+                                       struct pipe_texture *_texture,
+                                       unsigned *stride,
+                                       unsigned *handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct drm_api *api = id_api->api;
+
+   return api->local_handle_from_texture(api, screen, texture, stride, handle);
+}
+
+static void
+identity_drm_destroy(struct drm_api *_api)
+{
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct drm_api *api = id_api->api;
+   api->destroy(api);
+
+   free(id_api);
+}
+
+struct drm_api *
+identity_drm_create(struct drm_api *api)
+{
+   struct identity_drm_api *id_api;
+
+   if (!api)
+      goto error;
+
+   id_api = CALLOC_STRUCT(identity_drm_api);
+
+   if (!id_api)
+      goto error;
+
+   id_api->base.create_screen = identity_drm_create_screen;
+   id_api->base.create_context = identity_drm_create_context;
+   id_api->base.texture_from_shared_handle = identity_drm_texture_from_shared_handle;
+   id_api->base.shared_handle_from_texture = identity_drm_shared_handle_from_texture;
+   id_api->base.local_handle_from_texture = identity_drm_local_handle_from_texture;
+   id_api->base.destroy = identity_drm_destroy;
+   id_api->api = api;
+
+   return &id_api->base;
+
+error:
+   return api;
+}
diff --git a/src/gallium/drivers/identity/id_drm.h b/src/gallium/drivers/identity/id_drm.h
new file mode 100644
index 0000000000..cf2ad2ce07
--- /dev/null
+++ b/src/gallium/drivers/identity/id_drm.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_DRM_H
+#define ID_DRM_H
+
+struct drm_api;
+
+struct drm_api* identity_drm_create(struct drm_api *api);
+
+#endif /* ID_DRM_H */
diff --git a/src/gallium/drivers/identity/id_objects.c b/src/gallium/drivers/identity/id_objects.c
new file mode 100644
index 0000000000..e893e59940
--- /dev/null
+++ b/src/gallium/drivers/identity/id_objects.c
@@ -0,0 +1,182 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_screen.h"
+#include "id_objects.h"
+
+struct pipe_buffer *
+identity_buffer_create(struct identity_screen *id_screen,
+                       struct pipe_buffer *buffer)
+{
+   struct identity_buffer *id_buffer;
+
+   if(!buffer)
+      goto error;
+
+   assert(buffer->screen == id_screen->screen);
+
+   id_buffer = CALLOC_STRUCT(identity_buffer);
+   if(!id_buffer)
+      goto error;
+
+   memcpy(&id_buffer->base, buffer, sizeof(struct pipe_buffer));
+
+   pipe_reference_init(&id_buffer->base.reference, 1);
+   id_buffer->base.screen = &id_screen->base;
+   id_buffer->buffer = buffer;
+
+   return &id_buffer->base;
+
+error:
+   pipe_buffer_reference(&buffer, NULL);
+   return NULL;
+}
+
+void
+identity_buffer_destroy(struct identity_buffer *id_buffer)
+{
+   pipe_buffer_reference(&id_buffer->buffer, NULL);
+   FREE(id_buffer);
+}
+
+
+struct pipe_texture *
+identity_texture_create(struct identity_screen *id_screen,
+                        struct pipe_texture *texture)
+{
+   struct identity_texture *id_texture;
+
+   if(!texture)
+      goto error;
+
+   assert(texture->screen == id_screen->screen);
+
+   id_texture = CALLOC_STRUCT(identity_texture);
+   if(!id_texture)
+      goto error;
+
+   memcpy(&id_texture->base, texture, sizeof(struct pipe_texture));
+
+   pipe_reference_init(&id_texture->base.reference, 1);
+   id_texture->base.screen = &id_screen->base;
+   id_texture->texture = texture;
+
+   return &id_texture->base;
+
+error:
+   pipe_texture_reference(&texture, NULL);
+   return NULL;
+}
+
+void
+identity_texture_destroy(struct identity_texture *id_texture)
+{
+   pipe_texture_reference(&id_texture->texture, NULL);
+   FREE(id_texture);
+}
+
+
+struct pipe_surface *
+identity_surface_create(struct identity_texture *id_texture,
+                        struct pipe_surface *surface)
+{
+   struct identity_surface *id_surface;
+
+   if(!surface)
+      goto error;
+
+   assert(surface->texture == id_texture->texture);
+
+   id_surface = CALLOC_STRUCT(identity_surface);
+   if(!id_surface)
+      goto error;
+
+   memcpy(&id_surface->base, surface, sizeof(struct pipe_surface));
+
+   pipe_reference_init(&id_surface->base.reference, 1);
+   id_surface->base.texture = NULL;
+   pipe_texture_reference(&id_surface->base.texture, &id_texture->base);
+   id_surface->surface = surface;
+
+   return &id_surface->base;
+
+error:
+   pipe_surface_reference(&surface, NULL);
+   return NULL;
+}
+
+void
+identity_surface_destroy(struct identity_surface *id_surface)
+{
+   pipe_texture_reference(&id_surface->base.texture, NULL);
+   pipe_surface_reference(&id_surface->surface, NULL);
+   FREE(id_surface);
+}
+
+
+struct pipe_transfer *
+identity_transfer_create(struct identity_texture *id_texture,
+                         struct pipe_transfer *transfer)
+{
+   struct identity_transfer *id_transfer;
+
+   if(!transfer)
+      goto error;
+
+   assert(transfer->texture == id_texture->texture);
+
+   id_transfer = CALLOC_STRUCT(identity_transfer);
+   if(!id_transfer)
+      goto error;
+
+   memcpy(&id_transfer->base, transfer, sizeof(struct pipe_transfer));
+
+   id_transfer->base.texture = NULL;
+   pipe_texture_reference(&id_transfer->base.texture, &id_texture->base);
+   id_transfer->transfer = transfer;
+   assert(id_transfer->base.texture == &id_texture->base);
+
+   return &id_transfer->base;
+
+error:
+   transfer->texture->screen->tex_transfer_destroy(transfer);
+   return NULL;
+}
+
+void
+identity_transfer_destroy(struct identity_transfer *id_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(id_transfer->base.texture->screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   pipe_texture_reference(&id_transfer->base.texture, NULL);
+   screen->tex_transfer_destroy(id_transfer->transfer);
+   FREE(id_transfer);
+}
diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h
new file mode 100644
index 0000000000..ce58faa3c7
--- /dev/null
+++ b/src/gallium/drivers/identity/id_objects.h
@@ -0,0 +1,169 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_OBJECTS_H
+#define ID_OBJECTS_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+#include "id_screen.h"
+
+
+struct identity_buffer
+{
+   struct pipe_buffer base;
+
+   struct pipe_buffer *buffer;
+};
+
+
+struct identity_texture
+{
+   struct pipe_texture base;
+
+   struct pipe_texture *texture;
+};
+
+
+struct identity_surface
+{
+   struct pipe_surface base;
+
+   struct pipe_surface *surface;
+};
+
+
+struct identity_transfer
+{
+   struct pipe_transfer base;
+
+   struct pipe_transfer *transfer;
+};
+
+
+static INLINE struct identity_buffer *
+identity_buffer(struct pipe_buffer *_buffer)
+{
+   if(!_buffer)
+      return NULL;
+   (void)identity_screen(_buffer->screen);
+   return (struct identity_buffer *)_buffer;
+}
+
+static INLINE struct identity_texture *
+identity_texture(struct pipe_texture *_texture)
+{
+   if(!_texture)
+      return NULL;
+   (void)identity_screen(_texture->screen);
+   return (struct identity_texture *)_texture;
+}
+
+static INLINE struct identity_surface *
+identity_surface(struct pipe_surface *_surface)
+{
+   if(!_surface)
+      return NULL;
+   (void)identity_texture(_surface->texture);
+   return (struct identity_surface *)_surface;
+}
+
+static INLINE struct identity_transfer *
+identity_transfer(struct pipe_transfer *_transfer)
+{
+   if(!_transfer)
+      return NULL;
+   (void)identity_texture(_transfer->texture);
+   return (struct identity_transfer *)_transfer;
+}
+
+
+static INLINE struct pipe_buffer *
+identity_buffer_unwrap(struct pipe_buffer *_buffer)
+{
+   if(!_buffer)
+      return NULL;
+   return identity_buffer(_buffer)->buffer;
+}
+
+static INLINE struct pipe_texture *
+identity_texture_unwrap(struct pipe_texture *_texture)
+{
+   if(!_texture)
+      return NULL;
+   return identity_texture(_texture)->texture;
+}
+
+static INLINE struct pipe_surface *
+identity_surface_unwrap(struct pipe_surface *_surface)
+{
+   if(!_surface)
+      return NULL;
+   return identity_surface(_surface)->surface;
+}
+
+static INLINE struct pipe_transfer *
+identity_transfer_unwrap(struct pipe_transfer *_transfer)
+{
+   if(!_transfer)
+      return NULL;
+   return identity_transfer(_transfer)->transfer;
+}
+
+
+struct pipe_buffer *
+identity_buffer_create(struct identity_screen *id_screen,
+                       struct pipe_buffer *buffer);
+
+void
+identity_buffer_destroy(struct identity_buffer *id_buffer);
+
+struct pipe_texture *
+identity_texture_create(struct identity_screen *id_screen,
+                        struct pipe_texture *texture);
+
+void
+identity_texture_destroy(struct identity_texture *id_texture);
+
+struct pipe_surface *
+identity_surface_create(struct identity_texture *id_texture,
+                        struct pipe_surface *surface);
+
+void
+identity_surface_destroy(struct identity_surface *id_surface);
+
+struct pipe_transfer *
+identity_transfer_create(struct identity_texture *id_texture,
+                         struct pipe_transfer *transfer);
+
+void
+identity_transfer_destroy(struct identity_transfer *id_transfer);
+
+
+#endif /* ID_OBJECTS_H */
diff --git a/src/gallium/drivers/identity/id_public.h b/src/gallium/drivers/identity/id_public.h
new file mode 100644
index 0000000000..cac14cfd60
--- /dev/null
+++ b/src/gallium/drivers/identity/id_public.h
@@ -0,0 +1,40 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_PUBLIC_H
+#define ID_PUBLIC_H
+
+struct pipe_screen;
+struct pipe_context;
+
+struct pipe_screen *
+identity_screen_create(struct pipe_screen *screen);
+
+struct pipe_context *
+identity_context_create(struct pipe_screen *screen, struct pipe_context *pipe);
+
+#endif /* PT_PUBLIC_H */
diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c
new file mode 100644
index 0000000000..26439637d0
--- /dev/null
+++ b/src/gallium/drivers/identity/id_screen.c
@@ -0,0 +1,483 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_screen.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_screen.h"
+#include "id_objects.h"
+
+
+static void
+identity_screen_destroy(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   screen->destroy(screen);
+
+   FREE(id_screen);
+}
+
+static const char *
+identity_screen_get_name(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_name(screen);
+}
+
+static const char *
+identity_screen_get_vendor(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_vendor(screen);
+}
+
+static int
+identity_screen_get_param(struct pipe_screen *_screen,
+                          int param)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_param(screen,
+                            param);
+}
+
+static float
+identity_screen_get_paramf(struct pipe_screen *_screen,
+                           int param)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_paramf(screen,
+                             param);
+}
+
+static boolean
+identity_screen_is_format_supported(struct pipe_screen *_screen,
+                                    enum pipe_format format,
+                                    enum pipe_texture_target target,
+                                    unsigned tex_usage,
+                                    unsigned geom_flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->is_format_supported(screen,
+                                      format,
+                                      target,
+                                      tex_usage,
+                                      geom_flags);
+}
+
+static struct pipe_texture *
+identity_screen_texture_create(struct pipe_screen *_screen,
+                               const struct pipe_texture *templat)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *result;
+
+   result = screen->texture_create(screen,
+                                   templat);
+
+   if (result)
+      return identity_texture_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_texture *
+identity_screen_texture_blanket(struct pipe_screen *_screen,
+                                const struct pipe_texture *templat,
+                                const unsigned *stride,
+                                struct pipe_buffer *_buffer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+   struct pipe_texture *result;
+
+   result = screen->texture_blanket(screen,
+                                    templat,
+                                    stride,
+                                    buffer);
+
+   if (result)
+      return identity_texture_create(id_screen, result);
+   return NULL;
+}
+
+static void
+identity_screen_texture_destroy(struct pipe_texture *_texture)
+{
+   identity_texture_destroy(identity_texture(_texture));
+}
+
+static struct pipe_surface *
+identity_screen_get_tex_surface(struct pipe_screen *_screen,
+                                struct pipe_texture *_texture,
+                                unsigned face,
+                                unsigned level,
+                                unsigned zslice,
+                                unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct pipe_surface *result;
+
+   result = screen->get_tex_surface(screen,
+                                    texture,
+                                    face,
+                                    level,
+                                    zslice,
+                                    usage);
+
+   if (result)
+      return identity_surface_create(id_texture, result);
+   return NULL;
+}
+
+static void
+identity_screen_tex_surface_destroy(struct pipe_surface *_surface)
+{
+   identity_surface_destroy(identity_surface(_surface));
+}
+
+static struct pipe_transfer *
+identity_screen_get_tex_transfer(struct pipe_screen *_screen,
+                                 struct pipe_texture *_texture,
+                                 unsigned face,
+                                 unsigned level,
+                                 unsigned zslice,
+                                 enum pipe_transfer_usage usage,
+                                 unsigned x,
+                                 unsigned y,
+                                 unsigned w,
+                                 unsigned h)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct pipe_transfer *result;
+
+   result = screen->get_tex_transfer(screen,
+                                     texture,
+                                     face,
+                                     level,
+                                     zslice,
+                                     usage,
+                                     x,
+                                     y,
+                                     w,
+                                     h);
+
+   if (result)
+      return identity_transfer_create(id_texture, result);
+   return NULL;
+}
+
+static void
+identity_screen_tex_transfer_destroy(struct pipe_transfer *_transfer)
+{
+   identity_transfer_destroy(identity_transfer(_transfer));
+}
+
+static void *
+identity_screen_transfer_map(struct pipe_screen *_screen,
+                             struct pipe_transfer *_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_transfer *id_transfer = identity_transfer(_transfer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_transfer *transfer = id_transfer->transfer;
+
+   return screen->transfer_map(screen,
+                               transfer);
+}
+
+static void
+identity_screen_transfer_unmap(struct pipe_screen *_screen,
+                               struct pipe_transfer *_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_transfer *id_transfer = identity_transfer(_transfer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_transfer *transfer = id_transfer->transfer;
+
+   screen->transfer_unmap(screen,
+                          transfer);
+}
+
+static struct pipe_buffer *
+identity_screen_buffer_create(struct pipe_screen *_screen,
+                              unsigned alignment,
+                              unsigned usage,
+                              unsigned size)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->buffer_create(screen,
+                                  alignment,
+                                  usage,
+                                  size);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_buffer *
+identity_screen_user_buffer_create(struct pipe_screen *_screen,
+                                   void *ptr,
+                                   unsigned bytes)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->user_buffer_create(screen,
+                                       ptr,
+                                       bytes);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_buffer *
+identity_screen_surface_buffer_create(struct pipe_screen *_screen,
+                                      unsigned width,
+                                      unsigned height,
+                                      enum pipe_format format,
+                                      unsigned usage,
+                                      unsigned tex_usage,
+                                      unsigned *stride)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->surface_buffer_create(screen,
+                                          width,
+                                          height,
+                                          format,
+                                          usage,
+                                          tex_usage,
+                                          stride);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static void *
+identity_screen_buffer_map(struct pipe_screen *_screen,
+                           struct pipe_buffer *_buffer,
+                           unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return screen->buffer_map(screen,
+                             buffer,
+                             usage);
+}
+
+static void *
+identity_screen_buffer_map_range(struct pipe_screen *_screen,
+                                 struct pipe_buffer *_buffer,
+                                 unsigned offset,
+                                 unsigned length,
+                                 unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return screen->buffer_map_range(screen,
+                                   buffer,
+                                   offset,
+                                   length,
+                                   usage);
+}
+
+static void
+identity_screen_buffer_flush_mapped_range(struct pipe_screen *_screen,
+                                          struct pipe_buffer *_buffer,
+                                          unsigned offset,
+                                          unsigned length)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   screen->buffer_flush_mapped_range(screen,
+                                     buffer,
+                                     offset,
+                                     length);
+}
+
+static void
+identity_screen_buffer_unmap(struct pipe_screen *_screen,
+                             struct pipe_buffer *_buffer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   screen->buffer_unmap(screen,
+                        buffer);
+}
+
+static void
+identity_screen_buffer_destroy(struct pipe_buffer *_buffer)
+{
+   identity_buffer_destroy(identity_buffer(_buffer));
+}
+
+static void
+identity_screen_flush_frontbuffer(struct pipe_screen *_screen,
+                                  struct pipe_surface *_surface,
+                                  void *context_private)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_surface *id_surface = identity_surface(_surface);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_surface *surface = id_surface->surface;
+
+   screen->flush_frontbuffer(screen,
+                             surface,
+                             context_private);
+}
+
+static void
+identity_screen_fence_reference(struct pipe_screen *_screen,
+                                struct pipe_fence_handle **ptr,
+                                struct pipe_fence_handle *fence)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   screen->fence_reference(screen,
+                           ptr,
+                           fence);
+}
+
+static int
+identity_screen_fence_signalled(struct pipe_screen *_screen,
+                                struct pipe_fence_handle *fence,
+                                unsigned flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->fence_signalled(screen,
+                                  fence,
+                                  flags);
+}
+
+static int
+identity_screen_fence_finish(struct pipe_screen *_screen,
+                             struct pipe_fence_handle *fence,
+                             unsigned flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->fence_finish(screen,
+                               fence,
+                               flags);
+}
+
+struct pipe_screen *
+identity_screen_create(struct pipe_screen *screen)
+{
+   struct identity_screen *id_screen;
+
+   id_screen = CALLOC_STRUCT(identity_screen);
+   if (!id_screen) {
+      return NULL;
+   }
+
+   id_screen->base.winsys = NULL;
+
+   id_screen->base.destroy = identity_screen_destroy;
+   id_screen->base.get_name = identity_screen_get_name;
+   id_screen->base.get_vendor = identity_screen_get_vendor;
+   id_screen->base.get_param = identity_screen_get_param;
+   id_screen->base.get_paramf = identity_screen_get_paramf;
+   id_screen->base.is_format_supported = identity_screen_is_format_supported;
+   id_screen->base.texture_create = identity_screen_texture_create;
+   id_screen->base.texture_blanket = identity_screen_texture_blanket;
+   id_screen->base.texture_destroy = identity_screen_texture_destroy;
+   id_screen->base.get_tex_surface = identity_screen_get_tex_surface;
+   id_screen->base.tex_surface_destroy = identity_screen_tex_surface_destroy;
+   id_screen->base.get_tex_transfer = identity_screen_get_tex_transfer;
+   id_screen->base.tex_transfer_destroy = identity_screen_tex_transfer_destroy;
+   id_screen->base.transfer_map = identity_screen_transfer_map;
+   id_screen->base.transfer_unmap = identity_screen_transfer_unmap;
+   id_screen->base.buffer_create = identity_screen_buffer_create;
+   id_screen->base.user_buffer_create = identity_screen_user_buffer_create;
+   id_screen->base.surface_buffer_create = identity_screen_surface_buffer_create;
+   if (screen->buffer_map)
+      id_screen->base.buffer_map = identity_screen_buffer_map;
+   if (screen->buffer_map_range)
+      id_screen->base.buffer_map_range = identity_screen_buffer_map_range;
+   if (screen->buffer_flush_mapped_range)
+      id_screen->base.buffer_flush_mapped_range = identity_screen_buffer_flush_mapped_range;
+   if (screen->buffer_unmap)
+      id_screen->base.buffer_unmap = identity_screen_buffer_unmap;
+   id_screen->base.buffer_destroy = identity_screen_buffer_destroy;
+   id_screen->base.flush_frontbuffer = identity_screen_flush_frontbuffer;
+   id_screen->base.fence_reference = identity_screen_fence_reference;
+   id_screen->base.fence_signalled = identity_screen_fence_signalled;
+   id_screen->base.fence_finish = identity_screen_fence_finish;
+
+   id_screen->screen = screen;
+
+   return &id_screen->base;
+}
diff --git a/src/gallium/drivers/identity/id_screen.h b/src/gallium/drivers/identity/id_screen.h
new file mode 100644
index 0000000000..2c4f129089
--- /dev/null
+++ b/src/gallium/drivers/identity/id_screen.h
@@ -0,0 +1,48 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_SCREEN_H
+#define ID_SCREEN_H
+
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+
+
+struct identity_screen {
+   struct pipe_screen base;
+
+   struct pipe_screen *screen;
+};
+
+
+static INLINE struct identity_screen *
+identity_screen(struct pipe_screen *screen)
+{
+   return (struct identity_screen *)screen;
+}
+
+#endif /* ID_SCREEN_H */
diff --git a/src/gallium/drivers/llvmpipe/Makefile b/src/gallium/drivers/llvmpipe/Makefile
new file mode 100644
index 0000000000..5ac09de79e
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/Makefile
@@ -0,0 +1,54 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = llvmpipe
+
+CFLAGS += -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS
+
+C_SOURCES = \
+	lp_bld_alpha.c \
+	lp_bld_arit.c \
+	lp_bld_blend_aos.c \
+	lp_bld_blend_logicop.c \
+	lp_bld_blend_soa.c \
+	lp_bld_const.c \
+	lp_bld_conv.c \
+	lp_bld_debug.c \
+	lp_bld_depth.c \
+	lp_bld_flow.c \
+	lp_bld_format_aos.c \
+	lp_bld_interp.c \
+	lp_bld_intr.c \
+	lp_bld_logic.c \
+	lp_bld_swizzle.c \
+	lp_bld_struct.c \
+	lp_bld_tgsi_soa.c \
+	lp_bld_type.c \
+	lp_buffer.c \
+	lp_clear.c \
+	lp_context.c \
+	lp_draw_arrays.c \
+	lp_flush.c \
+	lp_jit.c \
+	lp_prim_setup.c \
+	lp_prim_vbuf.c \
+	lp_setup.c \
+	lp_query.c \
+	lp_screen.c \
+	lp_state_blend.c \
+	lp_state_clip.c \
+	lp_state_derived.c \
+	lp_state_fs.c \
+	lp_state_rasterizer.c \
+	lp_state_sampler.c \
+	lp_state_surface.c \
+	lp_state_vertex.c \
+	lp_state_vs.c \
+	lp_surface.c \
+	lp_tex_cache.c \
+	lp_tex_sample.c \
+	lp_texture.c \
+	lp_tile_cache.c \
+	lp_tile_soa.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/llvmpipe/README b/src/gallium/drivers/llvmpipe/README
new file mode 100644
index 0000000000..498d21dea6
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/README
@@ -0,0 +1,142 @@
+LLVMPIPE -- a fork of softpipe that employs LLVM for code generation.
+
+
+Status
+======
+
+Done so far is:
+
+ - the whole fragment pipeline is code generated in a single function
+ 
+   - depth testing
+ 
+   - fragment shader TGSI translation
+     - same level of support as the TGSI SSE2 exec machine, with the exception
+       we don't fallback to TGSI interpretation when an unsupported opcode is
+       found, but just ignore it
+     - texture sampling via an intrinsic call
+     - done in SoA layout
+     - input interpolation also code generated
+ 
+   - alpha testing
+ 
+   - blend (including logic ops)
+     - both in SoA and AoS layouts, but only the former used for now
+ 
+ - code is generic
+   - intermediates can be vectors of floats, ubytes, fixed point, etc, and of
+     any width and length
+   - not all operations are implemented for these types yet though
+
+Most mesa/progs/demos/* work. Speed is on par with Keith's softpipe-opt branch,
+which includes hand written fast implementations for common cases.
+
+To do (probably by this order):
+
+ - code generate stipple and stencil testing
+
+ - code generate texture sampling
+
+ - translate TGSI control flow instructions, and all other remaining opcodes
+
+ - code generate the triangle setup and rasterization
+
+
+Requirements
+============
+
+ - Linux
+ 
+ - udis86, http://udis86.sourceforge.net/ . Use my repository, which decodes
+   opcodes not yet supported by upstream.
+ 
+     git clone git://people.freedesktop.org/~jrfonseca/udis86
+     cd udis86
+     ./configure --with-pic
+     make
+     sudo make install
+ 
+ - LLVM 2.5. On Debian based distributions do:
+ 
+     aptitude install llvm-dev
+
+   There is a typo in one of the llvm-dev 2.5 headers, that causes compilation
+   errors in the debug build:
+
+     --- /usr/include/llvm-c/Core.h.orig	2009-08-10 15:38:54.000000000 +0100
+     +++ /usr/include/llvm-c/Core.h	2009-08-10 15:38:25.000000000 +0100
+     @@ -831,7 +831,7 @@
+        template<typename T>
+        inline T **unwrap(LLVMValueRef *Vals, unsigned Length) {
+          #if DEBUG
+     -    for (LLVMValueRef *I = Vals, E = Vals + Length; I != E; ++I)
+     +    for (LLVMValueRef *I = Vals, *E = Vals + Length; I != E; ++I)
+            cast<T>(*I);
+          #endif
+          return reinterpret_cast<T**>(Vals);
+ 
+ - A x86 or amd64 processor with support for sse2, sse3, and sse4.1 SIMD
+   instructions. This is necessary because we emit several SSE intrinsics for
+   convenience. See /proc/cpuinfo to know what your CPU supports.
+ 
+ - scons
+
+
+Building
+========
+
+To build everything invoke scons as:
+
+  scons debug=yes statetrackers=mesa drivers=llvmpipe winsys=xlib dri=false -k
+
+Alternatively, you can build it with GNU make, if you prefer, by invoking it as
+
+  make linux-llvm
+
+but the rest of these instructions assume scons is used.
+
+
+Using
+=====
+
+Building will create a drop-in alternative for libGL.so. To use it set the
+environment variables:
+
+  export LD_LIBRARY_PATH=$PWD/build/linux-x86_64-debug/lib:$LD_LIBRARY_PATH
+
+or
+
+  export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/lib:$LD_LIBRARY_PATH
+
+
+Unit testing
+============
+
+Building will also create several unit tests in
+build/linux-???-debug/gallium/drivers/llvmpipe:
+
+ - lp_test_blend: blending
+ - lp_test_conv: SIMD vector conversion
+ - lp_test_format: pixel unpacking/packing
+
+Some of this tests can output results and benchmarks to a tab-seperated-file
+for posterior analysis, e.g.:
+
+  build/linux-x86_64-debug/gallium/drivers/llvmpipe/lp_test_blend -o blend.tsv
+
+
+Development Notes
+=================
+
+- When looking to this code by the first time start in lp_state_fs.c, and 
+  then skim through the lp_bld_* functions called in there, and the comments
+  at the top of the lp_bld_*.c functions.  
+
+- All lp_bld_*.[ch] are isolated from the rest of the driver, and could/may be 
+  put in a standalone Gallium state -> LLVM IR translation module.
+
+- We use LLVM-C bindings for now. They are not documented, but follow the C++
+  interfaces very closely, and appear to be complete enough for code
+  generation. See 
+  http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html
+  for a standalone example.
diff --git a/src/gallium/drivers/llvmpipe/SConscript b/src/gallium/drivers/llvmpipe/SConscript
new file mode 100644
index 0000000000..5c29bdac56
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/SConscript
@@ -0,0 +1,81 @@
+Import('*')
+
+env = env.Clone()
+
+env.Tool('llvm')
+if 'LLVM_VERSION' not in env:
+    print 'warning: LLVM not found: not building llvmpipe'
+    Return()
+
+env.Tool('udis86')
+
+llvmpipe = env.ConvenienceLibrary(
+	target = 'llvmpipe',
+	source = [
+		'lp_bld_alpha.c',
+		'lp_bld_arit.c',
+		'lp_bld_blend_aos.c',
+		'lp_bld_blend_logicop.c',
+		'lp_bld_blend_soa.c',
+		'lp_bld_const.c',
+		'lp_bld_conv.c',
+		'lp_bld_debug.c',
+		'lp_bld_depth.c',
+		'lp_bld_flow.c',
+		'lp_bld_format_aos.c',
+		'lp_bld_interp.c',
+		'lp_bld_intr.c',
+		'lp_bld_struct.c',
+		'lp_bld_logic.c',
+		'lp_bld_swizzle.c',
+		'lp_bld_tgsi_soa.c',		
+		'lp_bld_type.c',
+		'lp_buffer.c',
+		'lp_clear.c',
+		'lp_context.c',
+		'lp_draw_arrays.c',
+		'lp_flush.c',
+		'lp_jit.c',
+		'lp_prim_setup.c',
+		'lp_prim_vbuf.c',
+		'lp_setup.c',
+		'lp_query.c',
+		'lp_screen.c',
+		'lp_state_blend.c',
+		'lp_state_clip.c',
+		'lp_state_derived.c',
+		'lp_state_fs.c',
+		'lp_state_rasterizer.c',
+		'lp_state_sampler.c',
+		'lp_state_surface.c',
+		'lp_state_vertex.c',
+		'lp_state_vs.c',
+		'lp_surface.c',
+		'lp_tex_cache.c',
+		'lp_tex_sample.c',
+		'lp_texture.c',
+		'lp_tile_cache.c',
+		'lp_tile_soa.c',
+	])
+
+
+env = env.Clone()
+
+env.Prepend(LIBS = [llvmpipe] + auxiliaries)
+
+env.Program(
+    target = 'lp_test_format',
+    source = ['lp_test_format.c'],
+)
+
+env.Program(
+    target = 'lp_test_blend',
+    source = ['lp_test_blend.c', 'lp_test_main.c'],
+)
+
+env.Program(
+    target = 'lp_test_conv',
+    source = ['lp_test_conv.c', 'lp_test_main.c'],
+)
+
+Export('llvmpipe')
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_alpha.c b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c
new file mode 100644
index 0000000000..49c2f911af
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c
@@ -0,0 +1,64 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Alpha testing to LLVM IR translation.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "pipe/p_state.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_flow.h"
+#include "lp_bld_debug.h"
+#include "lp_bld_alpha.h"
+
+
+void
+lp_build_alpha_test(LLVMBuilderRef builder,
+                    const struct pipe_alpha_state *state,
+                    union lp_type type,
+                    struct lp_build_mask_context *mask,
+                    LLVMValueRef alpha,
+                    LLVMValueRef ref)
+{
+   struct lp_build_context bld;
+
+   lp_build_context_init(&bld, builder, type);
+
+   if(state->enabled) {
+      LLVMValueRef test = lp_build_cmp(&bld, state->func, alpha, ref);
+
+      lp_build_name(test, "alpha_mask");
+
+      lp_build_mask_update(mask, test);
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_alpha.h b/src/gallium/drivers/llvmpipe/lp_bld_alpha.h
new file mode 100644
index 0000000000..9dbcdb4daa
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_alpha.h
@@ -0,0 +1,54 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Alpha testing to LLVM IR translation.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_BLD_ALPHA_H
+#define LP_BLD_ALPHA_H
+
+
+#include <llvm-c/Core.h>  
+
+struct pipe_alpha_state;
+union lp_type;
+struct lp_build_mask_context;
+
+
+void
+lp_build_alpha_test(LLVMBuilderRef builder,
+                    const struct pipe_alpha_state *state,
+                    union lp_type type,
+                    struct lp_build_mask_context *mask,
+                    LLVMValueRef alpha,
+                    LLVMValueRef ref);
+
+
+#endif /* !LP_BLD_ALPHA_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
new file mode 100644
index 0000000000..09a57ff33d
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
@@ -0,0 +1,963 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Helper
+ *
+ * LLVM IR doesn't support all basic arithmetic operations we care about (most
+ * notably min/max and saturated operations), and it is often necessary to
+ * resort machine-specific intrinsics directly. The functions here hide all
+ * these implementation details from the other modules.
+ *
+ * We also do simple expressions simplification here. Reasons are:
+ * - it is very easy given we have all necessary information readily available
+ * - LLVM optimization passes fail to simplify several vector expressions
+ * - We often know value constraints which the optimization passes have no way
+ *   of knowing, such as when source arguments are known to be in [0, 1] range.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_string.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_arit.h"
+
+
+/**
+ * Generate min(a, b)
+ * No checks for special case values of a or b = 1 or 0 are done.
+ */
+static LLVMValueRef
+lp_build_min_simple(struct lp_build_context *bld,
+                    LLVMValueRef a,
+                    LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+   const char *intrinsic = NULL;
+   LLVMValueRef cond;
+
+   /* TODO: optimize the constant case */
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width * type.length == 128) {
+      if(type.floating) {
+         if(type.width == 32)
+            intrinsic = "llvm.x86.sse.min.ps";
+         if(type.width == 64)
+            intrinsic = "llvm.x86.sse2.min.pd";
+      }
+      else {
+         if(type.width == 8 && !type.sign)
+            intrinsic = "llvm.x86.sse2.pminu.b";
+         if(type.width == 8 && type.sign)
+            intrinsic = "llvm.x86.sse41.pminsb";
+         if(type.width == 16 && !type.sign)
+            intrinsic = "llvm.x86.sse41.pminuw";
+         if(type.width == 16 && type.sign)
+            intrinsic = "llvm.x86.sse2.pmins.w";
+         if(type.width == 32 && !type.sign)
+            intrinsic = "llvm.x86.sse41.pminud";
+         if(type.width == 32 && type.sign)
+            intrinsic = "llvm.x86.sse41.pminsd";
+      }
+   }
+#endif
+
+   if(intrinsic)
+      return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b);
+
+   cond = lp_build_cmp(bld, PIPE_FUNC_LESS, a, b);
+   return lp_build_select(bld, cond, a, b);
+}
+
+
+/**
+ * Generate max(a, b)
+ * No checks for special case values of a or b = 1 or 0 are done.
+ */
+static LLVMValueRef
+lp_build_max_simple(struct lp_build_context *bld,
+                    LLVMValueRef a,
+                    LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+   const char *intrinsic = NULL;
+   LLVMValueRef cond;
+
+   /* TODO: optimize the constant case */
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width * type.length == 128) {
+      if(type.floating) {
+         if(type.width == 32)
+            intrinsic = "llvm.x86.sse.max.ps";
+         if(type.width == 64)
+            intrinsic = "llvm.x86.sse2.max.pd";
+      }
+      else {
+         if(type.width == 8 && !type.sign)
+            intrinsic = "llvm.x86.sse2.pmaxu.b";
+         if(type.width == 8 && type.sign)
+            intrinsic = "llvm.x86.sse41.pmaxsb";
+         if(type.width == 16 && !type.sign)
+            intrinsic = "llvm.x86.sse41.pmaxuw";
+         if(type.width == 16 && type.sign)
+            intrinsic = "llvm.x86.sse2.pmaxs.w";
+         if(type.width == 32 && !type.sign)
+            intrinsic = "llvm.x86.sse41.pmaxud";
+         if(type.width == 32 && type.sign)
+            intrinsic = "llvm.x86.sse41.pmaxsd";
+      }
+   }
+#endif
+
+   if(intrinsic)
+      return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b);
+
+   cond = lp_build_cmp(bld, PIPE_FUNC_GREATER, a, b);
+   return lp_build_select(bld, cond, a, b);
+}
+
+
+/**
+ * Generate 1 - a, or ~a depending on bld->type.
+ */
+LLVMValueRef
+lp_build_comp(struct lp_build_context *bld,
+              LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+
+   if(a == bld->one)
+      return bld->zero;
+   if(a == bld->zero)
+      return bld->one;
+
+   if(type.norm && !type.floating && !type.fixed && !type.sign) {
+      if(LLVMIsConstant(a))
+         return LLVMConstNot(a);
+      else
+         return LLVMBuildNot(bld->builder, a, "");
+   }
+
+   if(LLVMIsConstant(a))
+      return LLVMConstSub(bld->one, a);
+   else
+      return LLVMBuildSub(bld->builder, bld->one, a, "");
+}
+
+
+/**
+ * Generate a + b
+ */
+LLVMValueRef
+lp_build_add(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+   LLVMValueRef res;
+
+   if(a == bld->zero)
+      return b;
+   if(b == bld->zero)
+      return a;
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   if(bld->type.norm) {
+      const char *intrinsic = NULL;
+
+      if(a == bld->one || b == bld->one)
+        return bld->one;
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+      if(type.width * type.length == 128 &&
+         !type.floating && !type.fixed) {
+         if(type.width == 8)
+            intrinsic = type.sign ? "llvm.x86.sse2.padds.b" : "llvm.x86.sse2.paddus.b";
+         if(type.width == 16)
+            intrinsic = type.sign ? "llvm.x86.sse2.padds.w" : "llvm.x86.sse2.paddus.w";
+      }
+#endif
+   
+      if(intrinsic)
+         return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b);
+   }
+
+   if(LLVMIsConstant(a) && LLVMIsConstant(b))
+      res = LLVMConstAdd(a, b);
+   else
+      res = LLVMBuildAdd(bld->builder, a, b, "");
+
+   /* clamp to ceiling of 1.0 */
+   if(bld->type.norm && (bld->type.floating || bld->type.fixed))
+      res = lp_build_min_simple(bld, res, bld->one);
+
+   /* XXX clamp to floor of -1 or 0??? */
+
+   return res;
+}
+
+
+/**
+ * Generate a - b
+ */
+LLVMValueRef
+lp_build_sub(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+   LLVMValueRef res;
+
+   if(b == bld->zero)
+      return a;
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+   if(a == b)
+      return bld->zero;
+
+   if(bld->type.norm) {
+      const char *intrinsic = NULL;
+
+      if(b == bld->one)
+        return bld->zero;
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+      if(type.width * type.length == 128 &&
+         !type.floating && !type.fixed) {
+         if(type.width == 8)
+            intrinsic = type.sign ? "llvm.x86.sse2.psubs.b" : "llvm.x86.sse2.psubus.b";
+         if(type.width == 16)
+            intrinsic = type.sign ? "llvm.x86.sse2.psubs.w" : "llvm.x86.sse2.psubus.w";
+      }
+#endif
+   
+      if(intrinsic)
+         return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b);
+   }
+
+   if(LLVMIsConstant(a) && LLVMIsConstant(b))
+      res = LLVMConstSub(a, b);
+   else
+      res = LLVMBuildSub(bld->builder, a, b, "");
+
+   if(bld->type.norm && (bld->type.floating || bld->type.fixed))
+      res = lp_build_max_simple(bld, res, bld->zero);
+
+   return res;
+}
+
+
+/**
+ * Build shuffle vectors that match PUNPCKLxx and PUNPCKHxx instructions.
+ */
+static LLVMValueRef 
+lp_build_unpack_shuffle(unsigned n, unsigned lo_hi)
+{
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i, j;
+
+   assert(n <= LP_MAX_VECTOR_LENGTH);
+   assert(lo_hi < 2);
+
+   for(i = 0, j = lo_hi*n/2; i < n; i += 2, ++j) {
+      elems[i + 0] = LLVMConstInt(LLVMInt32Type(), 0 + j, 0);
+      elems[i + 1] = LLVMConstInt(LLVMInt32Type(), n + j, 0);
+   }
+
+   return LLVMConstVector(elems, n);
+}
+
+
+/**
+ * Build constant int vector of width 'n' and value 'c'.
+ */
+static LLVMValueRef 
+lp_build_const_vec(LLVMTypeRef type, unsigned n, long long c)
+{
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(n <= LP_MAX_VECTOR_LENGTH);
+
+   for(i = 0; i < n; ++i)
+      elems[i] = LLVMConstInt(type, c, 0);
+
+   return LLVMConstVector(elems, n);
+}
+
+
+/**
+ * Normalized 8bit multiplication.
+ *
+ * - alpha plus one
+ *
+ *     makes the following approximation to the division (Sree)
+ *    
+ *       a*b/255 ~= (a*(b + 1)) >> 256
+ *    
+ *     which is the fastest method that satisfies the following OpenGL criteria
+ *    
+ *       0*0 = 0 and 255*255 = 255
+ *
+ * - geometric series
+ *
+ *     takes the geometric series approximation to the division
+ *
+ *       t/255 = (t >> 8) + (t >> 16) + (t >> 24) ..
+ *
+ *     in this case just the first two terms to fit in 16bit arithmetic
+ *
+ *       t/255 ~= (t + (t >> 8)) >> 8
+ *
+ *     note that just by itself it doesn't satisfies the OpenGL criteria, as
+ *     255*255 = 254, so the special case b = 255 must be accounted or roundoff
+ *     must be used
+ *
+ * - geometric series plus rounding
+ *
+ *     when using a geometric series division instead of truncating the result
+ *     use roundoff in the approximation (Jim Blinn)
+ *
+ *       t/255 ~= (t + (t >> 8) + 0x80) >> 8
+ *
+ *     achieving the exact results
+ *
+ * @sa Alvy Ray Smith, Image Compositing Fundamentals, Tech Memo 4, Aug 15, 1995, 
+ *     ftp://ftp.alvyray.com/Acrobat/4_Comp.pdf
+ * @sa Michael Herf, The "double blend trick", May 2000, 
+ *     http://www.stereopsis.com/doubleblend.html
+ */
+static LLVMValueRef
+lp_build_mul_u8n(LLVMBuilderRef builder,
+                 LLVMValueRef a, LLVMValueRef b)
+{
+   static LLVMValueRef c01 = NULL;
+   static LLVMValueRef c08 = NULL;
+   static LLVMValueRef c80 = NULL;
+   LLVMValueRef ab;
+
+   if(!c01) c01 = lp_build_const_vec(LLVMInt16Type(), 8, 0x01);
+   if(!c08) c08 = lp_build_const_vec(LLVMInt16Type(), 8, 0x08);
+   if(!c80) c80 = lp_build_const_vec(LLVMInt16Type(), 8, 0x80);
+   
+#if 0
+   
+   /* a*b/255 ~= (a*(b + 1)) >> 256 */
+   b = LLVMBuildAdd(builder, b, c01, "");
+   ab = LLVMBuildMul(builder, a, b, "");
+
+#else
+   
+   /* t/255 ~= (t + (t >> 8) + 0x80) >> 8 */
+   ab = LLVMBuildMul(builder, a, b, "");
+   ab = LLVMBuildAdd(builder, ab, LLVMBuildLShr(builder, ab, c08, ""), "");
+   ab = LLVMBuildAdd(builder, ab, c80, "");
+
+#endif
+   
+   ab = LLVMBuildLShr(builder, ab, c08, "");
+
+   return ab;
+}
+
+
+/**
+ * Generate a * b
+ */
+LLVMValueRef
+lp_build_mul(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+
+   if(a == bld->zero)
+      return bld->zero;
+   if(a == bld->one)
+      return b;
+   if(b == bld->zero)
+      return bld->zero;
+   if(b == bld->one)
+      return a;
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   if(!type.floating && !type.fixed && type.norm) {
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+      if(type.width == 8 && type.length == 16) {
+         LLVMTypeRef i16x8 = LLVMVectorType(LLVMInt16Type(), 8);
+         LLVMTypeRef i8x16 = LLVMVectorType(LLVMInt8Type(), 16);
+         static LLVMValueRef ml = NULL;
+         static LLVMValueRef mh = NULL;
+         LLVMValueRef al, ah, bl, bh;
+         LLVMValueRef abl, abh;
+         LLVMValueRef ab;
+         
+         if(!ml) ml = lp_build_unpack_shuffle(16, 0);
+         if(!mh) mh = lp_build_unpack_shuffle(16, 1);
+
+         /*  PUNPCKLBW, PUNPCKHBW */
+         al = LLVMBuildShuffleVector(bld->builder, a, bld->zero, ml, "");
+         bl = LLVMBuildShuffleVector(bld->builder, b, bld->zero, ml, "");
+         ah = LLVMBuildShuffleVector(bld->builder, a, bld->zero, mh, "");
+         bh = LLVMBuildShuffleVector(bld->builder, b, bld->zero, mh, "");
+
+         /* NOP */
+         al = LLVMBuildBitCast(bld->builder, al, i16x8, "");
+         bl = LLVMBuildBitCast(bld->builder, bl, i16x8, "");
+         ah = LLVMBuildBitCast(bld->builder, ah, i16x8, "");
+         bh = LLVMBuildBitCast(bld->builder, bh, i16x8, "");
+
+         /* PMULLW, PSRLW, PADDW */
+         abl = lp_build_mul_u8n(bld->builder, al, bl);
+         abh = lp_build_mul_u8n(bld->builder, ah, bh);
+
+         /* PACKUSWB */
+         ab = lp_build_intrinsic_binary(bld->builder, "llvm.x86.sse2.packuswb.128" , i16x8, abl, abh);
+
+         /* NOP */
+         ab = LLVMBuildBitCast(bld->builder, ab, i8x16, "");
+         
+         return ab;
+      }
+#endif
+
+      /* FIXME */
+      assert(0);
+   }
+
+   if(LLVMIsConstant(a) && LLVMIsConstant(b))
+      return LLVMConstMul(a, b);
+
+   return LLVMBuildMul(bld->builder, a, b, "");
+}
+
+
+/**
+ * Generate a / b
+ */
+LLVMValueRef
+lp_build_div(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+
+   if(a == bld->zero)
+      return bld->zero;
+   if(a == bld->one)
+      return lp_build_rcp(bld, b);
+   if(b == bld->zero)
+      return bld->undef;
+   if(b == bld->one)
+      return a;
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   if(LLVMIsConstant(a) && LLVMIsConstant(b))
+      return LLVMConstFDiv(a, b);
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width == 32 && type.length == 4)
+      return lp_build_mul(bld, a, lp_build_rcp(bld, b));
+#endif
+
+   return LLVMBuildFDiv(bld->builder, a, b, "");
+}
+
+
+/**
+ * Generate min(a, b)
+ * Do checks for special cases.
+ */
+LLVMValueRef
+lp_build_min(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   if(a == b)
+      return a;
+
+   if(bld->type.norm) {
+      if(a == bld->zero || b == bld->zero)
+         return bld->zero;
+      if(a == bld->one)
+         return b;
+      if(b == bld->one)
+         return a;
+   }
+
+   return lp_build_min_simple(bld, a, b);
+}
+
+
+/**
+ * Generate max(a, b)
+ * Do checks for special cases.
+ */
+LLVMValueRef
+lp_build_max(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   if(a == b)
+      return a;
+
+   if(bld->type.norm) {
+      if(a == bld->one || b == bld->one)
+         return bld->one;
+      if(a == bld->zero)
+         return b;
+      if(b == bld->zero)
+         return a;
+   }
+
+   return lp_build_max_simple(bld, a, b);
+}
+
+
+/**
+ * Generate abs(a)
+ */
+LLVMValueRef
+lp_build_abs(struct lp_build_context *bld,
+             LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+
+   if(!type.sign)
+      return a;
+
+   /* XXX: is this really necessary? */
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(!type.floating && type.width*type.length == 128) {
+      LLVMTypeRef vec_type = lp_build_vec_type(type);
+      if(type.width == 8)
+         return lp_build_intrinsic_unary(bld->builder, "llvm.x86.ssse3.pabs.b.128", vec_type, a);
+      if(type.width == 16)
+         return lp_build_intrinsic_unary(bld->builder, "llvm.x86.ssse3.pabs.w.128", vec_type, a);
+      if(type.width == 32)
+         return lp_build_intrinsic_unary(bld->builder, "llvm.x86.ssse3.pabs.d.128", vec_type, a);
+   }
+#endif
+
+   return lp_build_max(bld, a, LLVMBuildNeg(bld->builder, a, ""));
+}
+
+
+LLVMValueRef
+lp_build_sqrt(struct lp_build_context *bld,
+              LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   char intrinsic[32];
+
+   /* TODO: optimize the constant case */
+   /* TODO: optimize the constant case */
+
+   assert(type.floating);
+   util_snprintf(intrinsic, sizeof intrinsic, "llvm.sqrt.v%uf%u", type.length, type.width);
+
+   return lp_build_intrinsic_unary(bld->builder, intrinsic, vec_type, a);
+}
+
+
+LLVMValueRef
+lp_build_rcp(struct lp_build_context *bld,
+             LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+
+   if(a == bld->zero)
+      return bld->undef;
+   if(a == bld->one)
+      return bld->one;
+   if(a == bld->undef)
+      return bld->undef;
+
+   assert(type.floating);
+
+   if(LLVMIsConstant(a))
+      return LLVMConstFDiv(bld->one, a);
+
+   /* XXX: is this really necessary? */
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width == 32 && type.length == 4)
+      return lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", lp_build_vec_type(type), a);
+#endif
+
+   return LLVMBuildFDiv(bld->builder, bld->one, a, "");
+}
+
+
+/**
+ * Generate 1/sqrt(a)
+ */
+LLVMValueRef
+lp_build_rsqrt(struct lp_build_context *bld,
+               LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+
+   assert(type.floating);
+
+   /* XXX: is this really necessary? */
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width == 32 && type.length == 4)
+      return lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rsqrt.ps", lp_build_vec_type(type), a);
+#endif
+
+   return lp_build_rcp(bld, lp_build_sqrt(bld, a));
+}
+
+
+/**
+ * Generate cos(a)
+ */
+LLVMValueRef
+lp_build_cos(struct lp_build_context *bld,
+              LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   char intrinsic[32];
+
+   /* TODO: optimize the constant case */
+
+   assert(type.floating);
+   util_snprintf(intrinsic, sizeof intrinsic, "llvm.cos.v%uf%u", type.length, type.width);
+
+   return lp_build_intrinsic_unary(bld->builder, intrinsic, vec_type, a);
+}
+
+
+/**
+ * Generate sin(a)
+ */
+LLVMValueRef
+lp_build_sin(struct lp_build_context *bld,
+              LLVMValueRef a)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   char intrinsic[32];
+
+   /* TODO: optimize the constant case */
+
+   assert(type.floating);
+   util_snprintf(intrinsic, sizeof intrinsic, "llvm.sin.v%uf%u", type.length, type.width);
+
+   return lp_build_intrinsic_unary(bld->builder, intrinsic, vec_type, a);
+}
+
+
+/**
+ * Generate pow(x, y)
+ */
+LLVMValueRef
+lp_build_pow(struct lp_build_context *bld,
+             LLVMValueRef x,
+             LLVMValueRef y)
+{
+   /* TODO: optimize the constant case */
+   if(LLVMIsConstant(x) && LLVMIsConstant(y))
+      debug_printf("%s: inefficient/imprecise constant arithmetic\n");
+
+   return lp_build_exp2(bld, lp_build_mul(bld, lp_build_log2(bld, x), y));
+}
+
+
+/**
+ * Generate exp(x)
+ */
+LLVMValueRef
+lp_build_exp(struct lp_build_context *bld,
+             LLVMValueRef x)
+{
+   /* log2(e) = 1/log(2) */
+   LLVMValueRef log2e = lp_build_const_scalar(bld->type, 1.4426950408889634);
+
+   return lp_build_mul(bld, log2e, lp_build_exp2(bld, x));
+}
+
+
+/**
+ * Generate log(x)
+ */
+LLVMValueRef
+lp_build_log(struct lp_build_context *bld,
+             LLVMValueRef x)
+{
+   /* log(2) */
+   LLVMValueRef log2 = lp_build_const_scalar(bld->type, 1.4426950408889634);
+
+   return lp_build_mul(bld, log2, lp_build_exp2(bld, x));
+}
+
+
+#define EXP_POLY_DEGREE 3
+#define LOG_POLY_DEGREE 5
+
+
+/**
+ * Generate polynomial.
+ * Ex:  x^2 * coeffs[0] + x * coeffs[1] + coeffs[2].
+ */
+static LLVMValueRef
+lp_build_polynomial(struct lp_build_context *bld,
+                    LLVMValueRef x,
+                    const double *coeffs,
+                    unsigned num_coeffs)
+{
+   const union lp_type type = bld->type;
+   LLVMValueRef res = NULL;
+   unsigned i;
+
+   /* TODO: optimize the constant case */
+   if(LLVMIsConstant(x))
+      debug_printf("%s: inefficient/imprecise constant arithmetic\n");
+
+   for (i = num_coeffs; i--; ) {
+      LLVMValueRef coeff = lp_build_const_scalar(type, coeffs[i]);
+      if(res)
+         res = lp_build_add(bld, coeff, lp_build_mul(bld, x, res));
+      else
+         res = coeff;
+   }
+
+   if(res)
+      return res;
+   else
+      return bld->undef;
+}
+
+
+/**
+ * Minimax polynomial fit of 2**x, in range [-0.5, 0.5[
+ */
+const double lp_build_exp2_polynomial[] = {
+#if EXP_POLY_DEGREE == 5
+   9.9999994e-1, 6.9315308e-1, 2.4015361e-1, 5.5826318e-2, 8.9893397e-3, 1.8775767e-3
+#elif EXP_POLY_DEGREE == 4
+   1.0000026, 6.9300383e-1, 2.4144275e-1, 5.2011464e-2, 1.3534167e-2
+#elif EXP_POLY_DEGREE == 3
+   9.9992520e-1, 6.9583356e-1, 2.2606716e-1, 7.8024521e-2
+#elif EXP_POLY_DEGREE == 2
+   1.0017247, 6.5763628e-1, 3.3718944e-1
+#else
+#error
+#endif
+};
+
+
+void
+lp_build_exp2_approx(struct lp_build_context *bld,
+                     LLVMValueRef x,
+                     LLVMValueRef *p_exp2_int_part,
+                     LLVMValueRef *p_frac_part,
+                     LLVMValueRef *p_exp2)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+   LLVMValueRef ipart = NULL;
+   LLVMValueRef fpart = NULL;
+   LLVMValueRef expipart = NULL;
+   LLVMValueRef expfpart = NULL;
+   LLVMValueRef res = NULL;
+
+   if(p_exp2_int_part || p_frac_part || p_exp2) {
+      /* TODO: optimize the constant case */
+      if(LLVMIsConstant(x))
+         debug_printf("%s: inefficient/imprecise constant arithmetic\n");
+
+      assert(type.floating && type.width == 32);
+
+      x = lp_build_min(bld, x, lp_build_const_scalar(type,  129.0));
+      x = lp_build_max(bld, x, lp_build_const_scalar(type, -126.99999));
+
+      /* ipart = int(x - 0.5) */
+      ipart = LLVMBuildSub(bld->builder, x, lp_build_const_scalar(type, 0.5f), "");
+      ipart = LLVMBuildFPToSI(bld->builder, ipart, int_vec_type, "");
+
+      /* fpart = x - ipart */
+      fpart = LLVMBuildSIToFP(bld->builder, ipart, vec_type, "");
+      fpart = LLVMBuildSub(bld->builder, x, fpart, "");
+   }
+
+   if(p_exp2_int_part || p_exp2) {
+      /* expipart = (float) (1 << ipart) */
+      expipart = LLVMBuildAdd(bld->builder, ipart, lp_build_int_const_scalar(type, 127), "");
+      expipart = LLVMBuildShl(bld->builder, expipart, lp_build_int_const_scalar(type, 23), "");
+      expipart = LLVMBuildBitCast(bld->builder, expipart, vec_type, "");
+   }
+
+   if(p_exp2) {
+      expfpart = lp_build_polynomial(bld, fpart, lp_build_exp2_polynomial,
+                                     Elements(lp_build_exp2_polynomial));
+
+      res = LLVMBuildMul(bld->builder, expipart, expfpart, "");
+   }
+
+   if(p_exp2_int_part)
+      *p_exp2_int_part = expipart;
+
+   if(p_frac_part)
+      *p_frac_part = fpart;
+
+   if(p_exp2)
+      *p_exp2 = res;
+}
+
+
+LLVMValueRef
+lp_build_exp2(struct lp_build_context *bld,
+              LLVMValueRef x)
+{
+   LLVMValueRef res;
+   lp_build_exp2_approx(bld, x, NULL, NULL, &res);
+   return res;
+}
+
+
+/**
+ * Minimax polynomial fit of log2(x)/(x - 1), for x in range [1, 2[
+ * These coefficients can be generate with
+ * http://www.boost.org/doc/libs/1_36_0/libs/math/doc/sf_and_dist/html/math_toolkit/toolkit/internals2/minimax.html
+ */
+const double lp_build_log2_polynomial[] = {
+#if LOG_POLY_DEGREE == 6
+   3.11578814719469302614, -3.32419399085241980044, 2.59883907202499966007, -1.23152682416275988241, 0.318212422185251071475, -0.0344359067839062357313
+#elif LOG_POLY_DEGREE == 5
+   2.8882704548164776201, -2.52074962577807006663, 1.48116647521213171641, -0.465725644288844778798, 0.0596515482674574969533
+#elif LOG_POLY_DEGREE == 4
+   2.61761038894603480148, -1.75647175389045657003, 0.688243882994381274313, -0.107254423828329604454
+#elif LOG_POLY_DEGREE == 3
+   2.28330284476918490682, -1.04913055217340124191, 0.204446009836232697516
+#else
+#error
+#endif
+};
+
+
+/**
+ * See http://www.devmaster.net/forums/showthread.php?p=43580
+ */
+void
+lp_build_log2_approx(struct lp_build_context *bld,
+                     LLVMValueRef x,
+                     LLVMValueRef *p_exp,
+                     LLVMValueRef *p_floor_log2,
+                     LLVMValueRef *p_log2)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+
+   LLVMValueRef expmask = lp_build_int_const_scalar(type, 0x7f800000);
+   LLVMValueRef mantmask = lp_build_int_const_scalar(type, 0x007fffff);
+   LLVMValueRef one = LLVMConstBitCast(bld->one, int_vec_type);
+
+   LLVMValueRef i = NULL;
+   LLVMValueRef exp = NULL;
+   LLVMValueRef mant = NULL;
+   LLVMValueRef logexp = NULL;
+   LLVMValueRef logmant = NULL;
+   LLVMValueRef res = NULL;
+
+   if(p_exp || p_floor_log2 || p_log2) {
+      /* TODO: optimize the constant case */
+      if(LLVMIsConstant(x))
+         debug_printf("%s: inefficient/imprecise constant arithmetic\n");
+
+      assert(type.floating && type.width == 32);
+
+      i = LLVMBuildBitCast(bld->builder, x, int_vec_type, "");
+
+      /* exp = (float) exponent(x) */
+      exp = LLVMBuildAnd(bld->builder, i, expmask, "");
+   }
+
+   if(p_floor_log2 || p_log2) {
+      logexp = LLVMBuildLShr(bld->builder, exp, lp_build_int_const_scalar(type, 23), "");
+      logexp = LLVMBuildSub(bld->builder, logexp, lp_build_int_const_scalar(type, 127), "");
+      logexp = LLVMBuildSIToFP(bld->builder, logexp, vec_type, "");
+   }
+
+   if(p_log2) {
+      /* mant = (float) mantissa(x) */
+      mant = LLVMBuildAnd(bld->builder, i, mantmask, "");
+      mant = LLVMBuildOr(bld->builder, mant, one, "");
+      mant = LLVMBuildSIToFP(bld->builder, mant, vec_type, "");
+
+      logmant = lp_build_polynomial(bld, mant, lp_build_log2_polynomial,
+                                    Elements(lp_build_log2_polynomial));
+
+      /* This effectively increases the polynomial degree by one, but ensures that log2(1) == 0*/
+      logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildMul(bld->builder, mant, bld->one, ""), "");
+
+      res = LLVMBuildAdd(bld->builder, logmant, logexp, "");
+   }
+
+   if(p_exp)
+      *p_exp = exp;
+
+   if(p_floor_log2)
+      *p_floor_log2 = logexp;
+
+   if(p_log2)
+      *p_log2 = res;
+}
+
+
+LLVMValueRef
+lp_build_log2(struct lp_build_context *bld,
+              LLVMValueRef x)
+{
+   LLVMValueRef res;
+   lp_build_log2_approx(bld, x, NULL, NULL, &res);
+   return res;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.h b/src/gallium/drivers/llvmpipe/lp_bld_arit.h
new file mode 100644
index 0000000000..fc8cb25966
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.h
@@ -0,0 +1,143 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper arithmetic functions.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_ARIT_H
+#define LP_BLD_ARIT_H
+
+
+#include <llvm-c/Core.h>  
+
+
+union lp_type type;
+struct lp_build_context;
+
+
+/**
+ * Complement, i.e., 1 - a.
+ */
+LLVMValueRef
+lp_build_comp(struct lp_build_context *bld,
+              LLVMValueRef a);
+
+LLVMValueRef
+lp_build_add(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_sub(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_mul(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_div(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_min(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_max(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_abs(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_sqrt(struct lp_build_context *bld,
+              LLVMValueRef a);
+
+LLVMValueRef
+lp_build_rcp(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_rsqrt(struct lp_build_context *bld,
+               LLVMValueRef a);
+
+LLVMValueRef
+lp_build_cos(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_sin(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_pow(struct lp_build_context *bld,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+LLVMValueRef
+lp_build_exp(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_log(struct lp_build_context *bld,
+             LLVMValueRef a);
+
+LLVMValueRef
+lp_build_exp2(struct lp_build_context *bld,
+              LLVMValueRef a);
+
+LLVMValueRef
+lp_build_log2(struct lp_build_context *bld,
+              LLVMValueRef a);
+
+void
+lp_build_exp2_approx(struct lp_build_context *bld,
+                     LLVMValueRef x,
+                     LLVMValueRef *p_exp2_int_part,
+                     LLVMValueRef *p_frac_part,
+                     LLVMValueRef *p_exp2);
+
+void
+lp_build_log2_approx(struct lp_build_context *bld,
+                     LLVMValueRef x,
+                     LLVMValueRef *p_exp,
+                     LLVMValueRef *p_floor_log2,
+                     LLVMValueRef *p_log2);
+
+#endif /* !LP_BLD_ARIT_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend.h b/src/gallium/drivers/llvmpipe/lp_bld_blend.h
new file mode 100644
index 0000000000..d19e18846c
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_blend.h
@@ -0,0 +1,107 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef LP_BLD_BLEND_H
+#define LP_BLD_BLEND_H
+
+
+/**
+ * @file
+ * LLVM IR building helpers interfaces.
+ *
+ * We use LLVM-C bindings for now. They are not documented, but follow the C++
+ * interfaces very closely, and appear to be complete enough for code
+ * genration. See
+ * http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html
+ * for a standalone example.
+ */
+
+#include <llvm-c/Core.h>  
+ 
+#include "pipe/p_format.h"
+
+
+struct pipe_blend_state;
+union lp_type;
+struct lp_build_context;
+
+
+/**
+ * Whether the blending function is commutative or not.
+ */
+boolean
+lp_build_blend_func_commutative(unsigned func);
+
+
+/**
+ * Whether the blending functions are the reverse of each other.
+ */
+boolean
+lp_build_blend_func_reverse(unsigned rgb_func, unsigned alpha_func);
+
+
+LLVMValueRef
+lp_build_blend_func(struct lp_build_context *bld,
+                    unsigned func,
+                    LLVMValueRef term1,
+                    LLVMValueRef term2);
+
+
+LLVMValueRef
+lp_build_blend_aos(LLVMBuilderRef builder,
+                   const struct pipe_blend_state *blend,
+                   union lp_type type,
+                   LLVMValueRef src,
+                   LLVMValueRef dst,
+                   LLVMValueRef const_,
+                   unsigned alpha_swizzle);
+
+
+void
+lp_build_blend_soa(LLVMBuilderRef builder,
+                   const struct pipe_blend_state *blend,
+                   union lp_type type,
+                   LLVMValueRef src[4],
+                   LLVMValueRef dst[4],
+                   LLVMValueRef const_[4],
+                   LLVMValueRef res[4]);
+
+
+/**
+ * Apply a logic op.
+ *
+ * src/dst parameters are packed values. It should work regardless the inputs
+ * are scalars, or a vector.
+ */
+LLVMValueRef
+lp_build_logicop(LLVMBuilderRef builder,
+                 unsigned logicop_func,
+                 LLVMValueRef src,
+                 LLVMValueRef dst);
+
+
+#endif /* !LP_BLD_BLEND_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend_aos.c b/src/gallium/drivers/llvmpipe/lp_bld_blend_aos.c
new file mode 100644
index 0000000000..c11a9398f8
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_blend_aos.c
@@ -0,0 +1,356 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Blend LLVM IR generation -- AoS layout.
+ *
+ * AoS blending is in general much slower than SoA, but there are some cases
+ * where it might be faster. In particular, if a pixel is rendered only once
+ * then the overhead of tiling and untiling will dominate over the speedup that
+ * SoA gives. So we might want to detect such cases and fallback to AoS in the
+ * future, but for now this function is here for historical/benchmarking
+ * purposes.
+ *
+ * Run lp_blend_test after any change to this file.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "pipe/p_state.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_swizzle.h"
+#include "lp_bld_blend.h"
+#include "lp_bld_debug.h"
+
+
+/**
+ * We may the same values several times, so we keep them here to avoid
+ * recomputing them. Also reusing the values allows us to do simplifications
+ * that LLVM optimization passes wouldn't normally be able to do.
+ */
+struct lp_build_blend_aos_context
+{
+   struct lp_build_context base;
+   
+   LLVMValueRef src;
+   LLVMValueRef dst;
+   LLVMValueRef const_;
+
+   LLVMValueRef inv_src;
+   LLVMValueRef inv_dst;
+   LLVMValueRef inv_const;
+   LLVMValueRef saturate;
+
+   LLVMValueRef rgb_src_factor;
+   LLVMValueRef alpha_src_factor;
+   LLVMValueRef rgb_dst_factor;
+   LLVMValueRef alpha_dst_factor;
+};
+
+
+static LLVMValueRef
+lp_build_blend_factor_unswizzled(struct lp_build_blend_aos_context *bld,
+                                 unsigned factor,
+                                 boolean alpha)
+{
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ZERO:
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_ONE:
+      return bld->base.one;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      return bld->src;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      return bld->dst;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      if(alpha)
+         return bld->base.one;
+      else {
+         if(!bld->inv_dst)
+            bld->inv_dst = lp_build_comp(&bld->base, bld->dst);
+         if(!bld->saturate)
+            bld->saturate = lp_build_min(&bld->base, bld->src, bld->inv_dst);
+         return bld->saturate;
+      }
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      return bld->const_;
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      if(!bld->inv_src)
+         bld->inv_src = lp_build_comp(&bld->base, bld->src);
+      return bld->inv_src;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      if(!bld->inv_dst)
+         bld->inv_dst = lp_build_comp(&bld->base, bld->dst);
+      return bld->inv_dst;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      if(!bld->inv_const)
+         bld->inv_const = lp_build_comp(&bld->base, bld->const_);
+      return bld->inv_const;
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   default:
+      assert(0);
+      return bld->base.zero;
+   }
+}
+
+
+enum lp_build_blend_swizzle {
+   LP_BUILD_BLEND_SWIZZLE_RGBA = 0,
+   LP_BUILD_BLEND_SWIZZLE_AAAA = 1,
+};
+
+
+/**
+ * How should we shuffle the base factor.
+ */
+static enum lp_build_blend_swizzle
+lp_build_blend_factor_swizzle(unsigned factor)
+{
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ONE:
+   case PIPE_BLENDFACTOR_ZERO:
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+   case PIPE_BLENDFACTOR_DST_COLOR:
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+      return LP_BUILD_BLEND_SWIZZLE_RGBA;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+      return LP_BUILD_BLEND_SWIZZLE_AAAA;
+   default:
+      assert(0);
+      return LP_BUILD_BLEND_SWIZZLE_RGBA;
+   }
+}
+
+
+static LLVMValueRef
+lp_build_blend_swizzle(struct lp_build_blend_aos_context *bld,
+                       LLVMValueRef rgb, 
+                       LLVMValueRef alpha, 
+                       enum lp_build_blend_swizzle rgb_swizzle,
+                       unsigned alpha_swizzle)
+{
+   if(rgb == alpha) {
+      if(rgb_swizzle == LP_BUILD_BLEND_SWIZZLE_RGBA)
+         return rgb;
+      if(rgb_swizzle == LP_BUILD_BLEND_SWIZZLE_AAAA)
+         return lp_build_broadcast_aos(&bld->base, rgb, alpha_swizzle);
+   }
+   else {
+      if(rgb_swizzle == LP_BUILD_BLEND_SWIZZLE_RGBA) {
+         boolean cond[4] = {0, 0, 0, 0};
+         cond[alpha_swizzle] = 1;
+         return lp_build_select_aos(&bld->base, alpha, rgb, cond);
+      }
+      if(rgb_swizzle == LP_BUILD_BLEND_SWIZZLE_AAAA) {
+         unsigned char swizzle[4];
+         swizzle[0] = alpha_swizzle;
+         swizzle[1] = alpha_swizzle;
+         swizzle[2] = alpha_swizzle;
+         swizzle[3] = alpha_swizzle;
+         swizzle[alpha_swizzle] += 4;
+         return lp_build_swizzle2_aos(&bld->base, rgb, alpha, swizzle);
+      }
+   }
+   assert(0);
+   return bld->base.undef;
+}
+
+
+/**
+ * @sa http://www.opengl.org/sdk/docs/man/xhtml/glBlendFuncSeparate.xml
+ */
+static LLVMValueRef
+lp_build_blend_factor(struct lp_build_blend_aos_context *bld,
+                      LLVMValueRef factor1,
+                      unsigned rgb_factor,
+                      unsigned alpha_factor,
+                      unsigned alpha_swizzle)
+{
+   LLVMValueRef rgb_factor_;
+   LLVMValueRef alpha_factor_;
+   LLVMValueRef factor2;
+   enum lp_build_blend_swizzle rgb_swizzle;
+
+   rgb_factor_   = lp_build_blend_factor_unswizzled(bld, rgb_factor,   FALSE);
+   alpha_factor_ = lp_build_blend_factor_unswizzled(bld, alpha_factor, TRUE);
+
+   rgb_swizzle = lp_build_blend_factor_swizzle(rgb_factor);
+
+   factor2 = lp_build_blend_swizzle(bld, rgb_factor_, alpha_factor_, rgb_swizzle, alpha_swizzle);
+
+   return lp_build_mul(&bld->base, factor1, factor2);
+}
+
+
+boolean
+lp_build_blend_func_commutative(unsigned func)
+{
+   switch (func) {
+   case PIPE_BLEND_ADD:
+   case PIPE_BLEND_MIN:
+   case PIPE_BLEND_MAX:
+      return TRUE;
+   case PIPE_BLEND_SUBTRACT:
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      return FALSE;
+   default:
+      assert(0);
+      return TRUE;
+   }
+}
+
+
+boolean
+lp_build_blend_func_reverse(unsigned rgb_func, unsigned alpha_func)
+{
+   if(rgb_func == alpha_func)
+      return FALSE;
+   if(rgb_func == PIPE_BLEND_SUBTRACT && alpha_func == PIPE_BLEND_REVERSE_SUBTRACT)
+      return TRUE;
+   if(rgb_func == PIPE_BLEND_REVERSE_SUBTRACT && alpha_func == PIPE_BLEND_SUBTRACT)
+      return TRUE;
+   return FALSE;
+}
+
+
+/**
+ * @sa http://www.opengl.org/sdk/docs/man/xhtml/glBlendEquationSeparate.xml
+ */
+LLVMValueRef
+lp_build_blend_func(struct lp_build_context *bld,
+                    unsigned func,
+                    LLVMValueRef term1, 
+                    LLVMValueRef term2)
+{
+   switch (func) {
+   case PIPE_BLEND_ADD:
+      return lp_build_add(bld, term1, term2);
+      break;
+   case PIPE_BLEND_SUBTRACT:
+      return lp_build_sub(bld, term1, term2);
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      return lp_build_sub(bld, term2, term1);
+   case PIPE_BLEND_MIN:
+      return lp_build_min(bld, term1, term2);
+   case PIPE_BLEND_MAX:
+      return lp_build_max(bld, term1, term2);
+   default:
+      assert(0);
+      return bld->zero;
+   }
+}
+
+
+LLVMValueRef
+lp_build_blend_aos(LLVMBuilderRef builder,
+                   const struct pipe_blend_state *blend,
+                   union lp_type type,
+                   LLVMValueRef src,
+                   LLVMValueRef dst,
+                   LLVMValueRef const_,
+                   unsigned alpha_swizzle)
+{
+   struct lp_build_blend_aos_context bld;
+   LLVMValueRef src_term;
+   LLVMValueRef dst_term;
+
+   /* FIXME */
+   assert(blend->colormask == 0xf);
+
+   if(!blend->blend_enable)
+      return src;
+
+   /* It makes no sense to blend unless values are normalized */
+   assert(type.norm);
+
+   /* Setup build context */
+   memset(&bld, 0, sizeof bld);
+   lp_build_context_init(&bld.base, builder, type);
+   bld.src = src;
+   bld.dst = dst;
+   bld.const_ = const_;
+
+   /* TODO: There are still a few optimization opportunities here. For certain
+    * combinations it is possible to reorder the operations and therefore saving
+    * some instructions. */
+
+   src_term = lp_build_blend_factor(&bld, src, blend->rgb_src_factor, blend->alpha_src_factor, alpha_swizzle);
+   dst_term = lp_build_blend_factor(&bld, dst, blend->rgb_dst_factor, blend->alpha_dst_factor, alpha_swizzle);
+
+   lp_build_name(src_term, "src_term");
+   lp_build_name(dst_term, "dst_term");
+
+   if(blend->rgb_func == blend->alpha_func) {
+      return lp_build_blend_func(&bld.base, blend->rgb_func, src_term, dst_term);
+   }
+   else {
+      /* Seperate RGB / A functions */
+
+      LLVMValueRef rgb;
+      LLVMValueRef alpha;
+
+      rgb   = lp_build_blend_func(&bld.base, blend->rgb_func,   src_term, dst_term);
+      alpha = lp_build_blend_func(&bld.base, blend->alpha_func, src_term, dst_term);
+
+      return lp_build_blend_swizzle(&bld, rgb, alpha, LP_BUILD_BLEND_SWIZZLE_RGBA, alpha_swizzle);
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend_logicop.c b/src/gallium/drivers/llvmpipe/lp_bld_blend_logicop.c
new file mode 100644
index 0000000000..88321f62a2
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_blend_logicop.c
@@ -0,0 +1,108 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Blend LLVM IR generation -- logic ops.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "pipe/p_state.h"
+
+#include "lp_bld_blend.h"
+
+
+LLVMValueRef
+lp_build_logicop(LLVMBuilderRef builder,
+                 unsigned logicop_func,
+                 LLVMValueRef src,
+                 LLVMValueRef dst)
+{
+   LLVMTypeRef type;
+   LLVMValueRef res;
+
+   type = LLVMTypeOf(src);
+
+   switch (logicop_func) {
+   case PIPE_LOGICOP_CLEAR:
+      res = LLVMConstNull(type);
+      break;
+   case PIPE_LOGICOP_NOR:
+      res = LLVMBuildNot(builder, LLVMBuildOr(builder, src, dst, ""), "");
+      break;
+   case PIPE_LOGICOP_AND_INVERTED:
+      res = LLVMBuildAnd(builder, LLVMBuildNot(builder, src, ""), dst, "");
+      break;
+   case PIPE_LOGICOP_COPY_INVERTED:
+      res = LLVMBuildNot(builder, src, "");
+      break;
+   case PIPE_LOGICOP_AND_REVERSE:
+      res = LLVMBuildAnd(builder, src, LLVMBuildNot(builder, dst, ""), "");
+      break;
+   case PIPE_LOGICOP_INVERT:
+      res = LLVMBuildNot(builder, dst, "");
+      break;
+   case PIPE_LOGICOP_XOR:
+      res = LLVMBuildXor(builder, src, dst, "");
+      break;
+   case PIPE_LOGICOP_NAND:
+      res = LLVMBuildNot(builder, LLVMBuildAnd(builder, src, dst, ""), "");
+      break;
+   case PIPE_LOGICOP_AND:
+      res = LLVMBuildAnd(builder, src, dst, "");
+      break;
+   case PIPE_LOGICOP_EQUIV:
+      res = LLVMBuildNot(builder, LLVMBuildXor(builder, src, dst, ""), "");
+      break;
+   case PIPE_LOGICOP_NOOP:
+      res = dst;
+      break;
+   case PIPE_LOGICOP_OR_INVERTED:
+      res = LLVMBuildOr(builder, LLVMBuildNot(builder, src, ""), dst, "");
+      break;
+   case PIPE_LOGICOP_COPY:
+      res = src;
+      break;
+   case PIPE_LOGICOP_OR_REVERSE:
+      res = LLVMBuildOr(builder, src, LLVMBuildNot(builder, dst, ""), "");
+      break;
+   case PIPE_LOGICOP_OR:
+      res = LLVMBuildOr(builder, src, dst, "");
+      break;
+   case PIPE_LOGICOP_SET:
+      res = LLVMConstAllOnes(type);
+      break;
+   default:
+      assert(0);
+      res = src;
+   }
+
+   return res;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c
new file mode 100644
index 0000000000..b92254a7d6
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c
@@ -0,0 +1,298 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Blend LLVM IR generation -- SoA layout.
+ *
+ * Blending in SoA is much faster than AoS, especially when separate rgb/alpha
+ * factors/functions are used, since no channel masking/shuffling is necessary
+ * and we can achieve the full throughput of the SIMD operations. Furthermore
+ * the fragment shader output is also in SoA, so it fits nicely with the rest of
+ * the fragment pipeline.
+ *
+ * The drawback is that to be displayed the color buffer needs to be in AoS
+ * layout, so we need to tile/untile the color buffer before/after rendering.
+ * A color buffer like
+ *
+ *  R11 G11 B11 A11 R12 G12 B12 A12  R13 G13 B13 A13 R14 G14 B14 A14  ...
+ *  R21 G21 B21 A21 R22 G22 B22 A22  R23 G23 B23 A23 R24 G24 B24 A24  ...
+ *
+ *  R31 G31 B31 A31 R32 G32 B32 A32  R33 G33 B33 A33 R34 G34 B34 A34  ...
+ *  R41 G41 B41 A41 R42 G42 B42 A42  R43 G43 B43 A43 R44 G44 B44 A44  ...
+ *
+ *  ... ... ... ... ... ... ... ...  ... ... ... ... ... ... ... ...  ...
+ *
+ * will actually be stored in memory as
+ *
+ *  R11 R12 R21 R22 R13 R14 R23 R24 ... G11 G12 G21 G22 G13 G14 G23 G24 ... B11 B12 B21 B22 B13 B14 B23 B24 ... A11 A12 A21 A22 A13 A14 A23 A24 ...
+ *  R31 R32 R41 R42 R33 R34 R43 R44 ... G31 G32 G41 G42 G33 G34 G43 G44 ... B31 B32 B41 B42 B33 B34 B43 B44 ... A31 A32 A41 A42 A33 A34 A43 A44 ...
+ *  ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
+ *
+ * NOTE: Run lp_blend_test after any change to this file.
+ *
+ * You can also run lp_blend_test to obtain AoS vs SoA benchmarks. Invoking it
+ * as:
+ *
+ *  lp_blend_test -o blend.tsv
+ *
+ * will generate a tab-seperated-file with the test results and performance
+ * measurements.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "pipe/p_state.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_blend.h"
+
+
+/**
+ * We may the same values several times, so we keep them here to avoid
+ * recomputing them. Also reusing the values allows us to do simplifications
+ * that LLVM optimization passes wouldn't normally be able to do.
+ */
+struct lp_build_blend_soa_context
+{
+   struct lp_build_context base;
+
+   LLVMValueRef src[4];
+   LLVMValueRef dst[4];
+   LLVMValueRef con[4];
+
+   LLVMValueRef inv_src[4];
+   LLVMValueRef inv_dst[4];
+   LLVMValueRef inv_con[4];
+
+   LLVMValueRef src_alpha_saturate;
+
+   /**
+    * We store all factors in a table in order to eliminate redundant
+    * multiplications later.
+    */
+   LLVMValueRef factor[2][2][4];
+
+   /**
+    * Table with all terms.
+    */
+   LLVMValueRef term[2][4];
+};
+
+
+static LLVMValueRef
+lp_build_blend_soa_factor(struct lp_build_blend_soa_context *bld,
+                          unsigned factor, unsigned i)
+{
+   /*
+    * Compute src/first term RGB
+    */
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      return bld->base.one;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      return bld->src[i];
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      return bld->src[3];
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      return bld->dst[i];
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      return bld->dst[3];
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      if(i == 3)
+         return bld->base.one;
+      else {
+         if(!bld->inv_dst[3])
+            bld->inv_dst[3] = lp_build_comp(&bld->base, bld->dst[3]);
+         if(!bld->src_alpha_saturate)
+            bld->src_alpha_saturate = lp_build_min(&bld->base, bld->src[3], bld->inv_dst[3]);
+         return bld->src_alpha_saturate;
+      }
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      return bld->con[i];
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      return bld->con[3];
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_ZERO:
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      if(!bld->inv_src[i])
+         bld->inv_src[i] = lp_build_comp(&bld->base, bld->src[i]);
+      return bld->inv_src[i];
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      if(!bld->inv_src[3])
+         bld->inv_src[3] = lp_build_comp(&bld->base, bld->src[3]);
+      return bld->inv_src[3];
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      if(!bld->inv_dst[i])
+         bld->inv_dst[i] = lp_build_comp(&bld->base, bld->dst[i]);
+      return bld->inv_dst[i];
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      if(!bld->inv_dst[3])
+         bld->inv_dst[3] = lp_build_comp(&bld->base, bld->dst[3]);
+      return bld->inv_dst[3];
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      if(!bld->inv_con[i])
+         bld->inv_con[i] = lp_build_comp(&bld->base, bld->con[i]);
+      return bld->inv_con[i];
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      if(!bld->inv_con[3])
+         bld->inv_con[3] = lp_build_comp(&bld->base, bld->con[3]);
+      return bld->inv_con[3];
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+      /* TODO */
+      assert(0);
+      return bld->base.zero;
+   default:
+      assert(0);
+      return bld->base.zero;
+   }
+}
+
+
+/**
+ * Generate blend code in SOA mode.
+ * \param src  src/fragment color
+ * \param dst  dst/framebuffer color
+ * \param con  constant blend color
+ * \param res  the result/output
+ */
+void
+lp_build_blend_soa(LLVMBuilderRef builder,
+                   const struct pipe_blend_state *blend,
+                   union lp_type type,
+                   LLVMValueRef src[4],
+                   LLVMValueRef dst[4],
+                   LLVMValueRef con[4],
+                   LLVMValueRef res[4])
+{
+   struct lp_build_blend_soa_context bld;
+   unsigned i, j, k;
+
+   /* Setup build context */
+   memset(&bld, 0, sizeof bld);
+   lp_build_context_init(&bld.base, builder, type);
+   for (i = 0; i < 4; ++i) {
+      bld.src[i] = src[i];
+      bld.dst[i] = dst[i];
+      bld.con[i] = con[i];
+   }
+
+   for (i = 0; i < 4; ++i) {
+      if (blend->colormask & (1 << i)) {
+         if (blend->logicop_enable) {
+            if(!type.floating) {
+               res[i] = lp_build_logicop(builder, blend->logicop_func, src[i], dst[i]);
+            }
+            else
+               res[i] = dst[i];
+         }
+         else if (blend->blend_enable) {
+            unsigned src_factor = i < 3 ? blend->rgb_src_factor : blend->alpha_src_factor;
+            unsigned dst_factor = i < 3 ? blend->rgb_dst_factor : blend->alpha_dst_factor;
+            unsigned func = i < 3 ? blend->rgb_func : blend->alpha_func;
+            boolean func_commutative = lp_build_blend_func_commutative(func);
+
+            /* It makes no sense to blend unless values are normalized */
+            assert(type.norm);
+
+            /*
+             * Compute src/dst factors.
+             */
+
+            bld.factor[0][0][i] = src[i];
+            bld.factor[0][1][i] = lp_build_blend_soa_factor(&bld, src_factor, i);
+            bld.factor[1][0][i] = dst[i];
+            bld.factor[1][1][i] = lp_build_blend_soa_factor(&bld, dst_factor, i);
+
+            /*
+             * Compute src/dst terms
+             */
+
+            for(k = 0; k < 2; ++k) {
+               /* See if this multiplication has been previously computed */
+               for(j = 0; j < i; ++j) {
+                  if((bld.factor[k][0][j] == bld.factor[k][0][i] &&
+                      bld.factor[k][1][j] == bld.factor[k][1][i]) ||
+                     (bld.factor[k][0][j] == bld.factor[k][1][i] &&
+                      bld.factor[k][1][j] == bld.factor[k][0][i]))
+                     break;
+               }
+
+               if(j < i)
+                  bld.term[k][i] = bld.term[k][j];
+               else
+                  bld.term[k][i] = lp_build_mul(&bld.base, bld.factor[k][0][i], bld.factor[k][1][i]);
+            }
+
+            /*
+             * Combine terms
+             */
+
+            /* See if this function has been previously applied */
+            for(j = 0; j < i; ++j) {
+               unsigned prev_func = j < 3 ? blend->rgb_func : blend->alpha_func;
+               unsigned func_reverse = lp_build_blend_func_reverse(func, prev_func);
+
+               if((!func_reverse &&
+                   bld.term[0][j] == bld.term[0][i] &&
+                   bld.term[1][j] == bld.term[1][i]) ||
+                  ((func_commutative || func_reverse) &&
+                   bld.term[0][j] == bld.term[1][i] &&
+                   bld.term[1][j] == bld.term[0][i]))
+                  break;
+            }
+
+            if(j < i)
+               res[i] = res[j];
+            else
+               res[i] = lp_build_blend_func(&bld.base, func, bld.term[0][i], bld.term[1][i]);
+         }
+         else {
+            res[i] = src[i];
+         }
+      }
+      else {
+         res[i] = dst[i];
+      }
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_const.c b/src/gallium/drivers/llvmpipe/lp_bld_const.c
new file mode 100644
index 0000000000..21487365ea
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_const.c
@@ -0,0 +1,369 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Helper functions for constant building.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include <float.h>
+
+#include "util/u_debug.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+
+
+unsigned
+lp_mantissa(union lp_type type)
+{
+   assert(type.floating);
+
+   if(type.floating) {
+      switch(type.width) {
+      case 32:
+         return 23;
+      case 64:
+         return 53;
+      default:
+         assert(0);
+         return 0;
+      }
+   }
+   else {
+      if(type.sign)
+         return type.width - 1;
+      else
+         return type.width;
+   }
+}
+
+
+/**
+ * Shift of the unity.
+ *
+ * Same as lp_const_scale(), but in terms of shifts.
+ */
+unsigned
+lp_const_shift(union lp_type type)
+{
+   if(type.floating)
+      return 0;
+   else if(type.fixed)
+      return type.width/2;
+   else if(type.norm)
+      return type.sign ? type.width - 1 : type.width;
+   else
+      return 0;
+}
+
+
+unsigned
+lp_const_offset(union lp_type type)
+{
+   if(type.floating || type.fixed)
+      return 0;
+   else if(type.norm)
+      return 1;
+   else
+      return 0;
+}
+
+
+/**
+ * Scaling factor between the LLVM native value and its interpretation.
+ *
+ * This is 1.0 for all floating types and unnormalized integers, and something
+ * else for the fixed points types and normalized integers.
+ */
+double
+lp_const_scale(union lp_type type)
+{
+   unsigned long long llscale;
+   double dscale;
+
+   llscale = (unsigned long long)1 << lp_const_shift(type);
+   llscale -= lp_const_offset(type);
+   dscale = (double)llscale;
+   assert((unsigned long long)dscale == llscale);
+
+   return dscale;
+}
+
+
+/**
+ * Minimum value representable by the type.
+ */
+double
+lp_const_min(union lp_type type)
+{
+   unsigned bits;
+
+   if(!type.sign)
+      return 0.0;
+
+   if(type.norm)
+      return -1.0;
+
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         return -FLT_MAX;
+      case 64:
+         return -DBL_MAX;
+      default:
+         assert(0);
+         return 0.0;
+      }
+   }
+
+   if(type.fixed)
+      /* FIXME: consider the fractional bits? */
+      bits = type.width / 2 - 1;
+   else
+      bits = type.width - 1;
+
+   return (double)-((long long)1 << bits);
+}
+
+
+/**
+ * Maximum value representable by the type.
+ */
+double
+lp_const_max(union lp_type type)
+{
+   unsigned bits;
+
+   if(type.norm)
+      return 1.0;
+
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         return FLT_MAX;
+      case 64:
+         return DBL_MAX;
+      default:
+         assert(0);
+         return 0.0;
+      }
+   }
+
+   if(type.fixed)
+      bits = type.width / 2;
+   else
+      bits = type.width;
+
+   if(type.sign)
+      bits -= 1;
+
+   return (double)(((unsigned long long)1 << bits) - 1);
+}
+
+
+double
+lp_const_eps(union lp_type type)
+{
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         return FLT_EPSILON;
+      case 64:
+         return DBL_EPSILON;
+      default:
+         assert(0);
+         return 0.0;
+      }
+   }
+   else {
+      double scale = lp_const_scale(type);
+      return 1.0/scale;
+   }
+}
+
+
+LLVMValueRef
+lp_build_undef(union lp_type type)
+{
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   return LLVMGetUndef(vec_type);
+}
+               
+
+LLVMValueRef
+lp_build_zero(union lp_type type)
+{
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   return LLVMConstNull(vec_type);
+}
+               
+
+LLVMValueRef
+lp_build_one(union lp_type type)
+{
+   LLVMTypeRef elem_type;
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(type.length <= LP_MAX_VECTOR_LENGTH);
+
+   elem_type = lp_build_elem_type(type);
+
+   if(type.floating)
+      elems[0] = LLVMConstReal(elem_type, 1.0);
+   else if(type.fixed)
+      elems[0] = LLVMConstInt(elem_type, 1LL << (type.width/2), 0);
+   else if(!type.norm)
+      elems[0] = LLVMConstInt(elem_type, 1, 0);
+   else if(type.sign)
+      elems[0] = LLVMConstInt(elem_type, (1LL << (type.width - 1)) - 1, 0);
+   else {
+      /* special case' -- 1.0 for normalized types is more easily attained if
+       * we start with a vector consisting of all bits set */
+      LLVMTypeRef vec_type = LLVMVectorType(elem_type, type.length);
+      LLVMValueRef vec = LLVMConstAllOnes(vec_type);
+
+#if 0
+      if(type.sign)
+         /* TODO: Unfortunately this caused "Tried to create a shift operation
+          * on a non-integer type!" */
+         vec = LLVMConstLShr(vec, lp_build_int_const_scalar(type, 1));
+#endif
+
+      return vec;
+   }
+
+   for(i = 1; i < type.length; ++i)
+      elems[i] = elems[0];
+
+   return LLVMConstVector(elems, type.length);
+}
+               
+
+LLVMValueRef
+lp_build_const_scalar(union lp_type type,
+                      double val)
+{
+   LLVMTypeRef elem_type = lp_build_elem_type(type);
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(type.length <= LP_MAX_VECTOR_LENGTH);
+
+   if(type.floating) {
+      elems[0] = LLVMConstReal(elem_type, val);
+   }
+   else {
+      double dscale = lp_const_scale(type);
+
+      elems[0] = LLVMConstInt(elem_type, val*dscale + 0.5, 0);
+   }
+
+   for(i = 1; i < type.length; ++i)
+      elems[i] = elems[0];
+
+   return LLVMConstVector(elems, type.length);
+}
+
+
+LLVMValueRef
+lp_build_int_const_scalar(union lp_type type,
+                          long long val)
+{
+   LLVMTypeRef elem_type = lp_build_int_elem_type(type);
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(type.length <= LP_MAX_VECTOR_LENGTH);
+
+   for(i = 0; i < type.length; ++i)
+      elems[i] = LLVMConstInt(elem_type, val, type.sign ? 1 : 0);
+
+   return LLVMConstVector(elems, type.length);
+}
+
+
+LLVMValueRef
+lp_build_const_aos(union lp_type type, 
+                   double r, double g, double b, double a, 
+                   const unsigned char *swizzle)
+{
+   const unsigned char default_swizzle[4] = {0, 1, 2, 3};
+   LLVMTypeRef elem_type;
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(type.length % 4 == 0);
+   assert(type.length <= LP_MAX_VECTOR_LENGTH);
+
+   elem_type = lp_build_elem_type(type);
+
+   if(swizzle == NULL)
+      swizzle = default_swizzle;
+
+   if(type.floating) {
+      elems[swizzle[0]] = LLVMConstReal(elem_type, r);
+      elems[swizzle[1]] = LLVMConstReal(elem_type, g);
+      elems[swizzle[2]] = LLVMConstReal(elem_type, b);
+      elems[swizzle[3]] = LLVMConstReal(elem_type, a);
+   }
+   else {
+      double dscale = lp_const_scale(type);
+
+      elems[swizzle[0]] = LLVMConstInt(elem_type, r*dscale + 0.5, 0);
+      elems[swizzle[1]] = LLVMConstInt(elem_type, g*dscale + 0.5, 0);
+      elems[swizzle[2]] = LLVMConstInt(elem_type, b*dscale + 0.5, 0);
+      elems[swizzle[3]] = LLVMConstInt(elem_type, a*dscale + 0.5, 0);
+   }
+
+   for(i = 4; i < type.length; ++i)
+      elems[i] = elems[i % 4];
+
+   return LLVMConstVector(elems, type.length);
+}
+
+
+LLVMValueRef
+lp_build_const_mask_aos(union lp_type type,
+                        boolean cond[4])
+{
+   LLVMTypeRef elem_type = LLVMIntType(type.width);
+   LLVMValueRef masks[LP_MAX_VECTOR_LENGTH];
+   unsigned i, j;
+
+   assert(type.length <= LP_MAX_VECTOR_LENGTH);
+
+   for(j = 0; j < type.length; j += 4)
+      for(i = 0; i < 4; ++i)
+         masks[j + i] = LLVMConstInt(elem_type, cond[i] ? ~0 : 0, 0);
+
+   return LLVMConstVector(masks, type.length);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_const.h b/src/gallium/drivers/llvmpipe/lp_bld_const.h
new file mode 100644
index 0000000000..1934530ea3
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_const.h
@@ -0,0 +1,108 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for constant building.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_CONST_H
+#define LP_BLD_CONST_H
+
+
+#include <llvm-c/Core.h>  
+
+#include <pipe/p_compiler.h>
+
+
+union lp_type type;
+
+
+unsigned
+lp_mantissa(union lp_type type);
+
+
+unsigned
+lp_const_shift(union lp_type type);
+
+
+unsigned
+lp_const_offset(union lp_type type);
+
+
+double
+lp_const_scale(union lp_type type);
+
+double
+lp_const_min(union lp_type type);
+
+
+double
+lp_const_max(union lp_type type);
+
+
+double
+lp_const_eps(union lp_type type);
+
+
+LLVMValueRef
+lp_build_undef(union lp_type type);
+
+
+LLVMValueRef
+lp_build_zero(union lp_type type);
+
+
+LLVMValueRef
+lp_build_one(union lp_type type);
+
+
+LLVMValueRef
+lp_build_const_scalar(union lp_type type,
+                      double val);
+
+
+LLVMValueRef
+lp_build_int_const_scalar(union lp_type type,
+                          long long val);
+
+
+LLVMValueRef
+lp_build_const_aos(union lp_type type, 
+                   double r, double g, double b, double a, 
+                   const unsigned char *swizzle);
+
+
+LLVMValueRef
+lp_build_const_mask_aos(union lp_type type,
+                        boolean cond[4]);
+
+
+#endif /* !LP_BLD_CONST_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_conv.c b/src/gallium/drivers/llvmpipe/lp_bld_conv.c
new file mode 100644
index 0000000000..c8954c8a34
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_conv.c
@@ -0,0 +1,702 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Helper functions for type conversions.
+ *
+ * We want to use the fastest type for a given computation whenever feasible.
+ * The other side of this is that we need to be able convert between several
+ * types accurately and efficiently.
+ *
+ * Conversion between types of different bit width is quite complex since a 
+ *
+ * To remember there are a few invariants in type conversions:
+ *
+ * - register width must remain constant:
+ *
+ *     src_type.width * src_type.length == dst_type.width * dst_type.length
+ *
+ * - total number of elements must remain constant:
+ *
+ *     src_type.length * num_srcs == dst_type.length * num_dsts
+ *
+ * It is not always possible to do the conversion both accurately and
+ * efficiently, usually due to lack of adequate machine instructions. In these
+ * cases it is important not to cut shortcuts here and sacrifice accuracy, as
+ * there this functions can be used anywhere. In the future we might have a
+ * precision parameter which can gauge the accuracy vs efficiency compromise,
+ * but for now if the data conversion between two stages happens to be the
+ * bottleneck, then most likely should just avoid converting at all and run
+ * both stages with the same type.
+ *
+ * Make sure to run lp_test_conv unit test after any change to this file.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "util/u_debug.h"
+#include "util/u_math.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_conv.h"
+
+
+/**
+ * Special case for converting clamped IEEE-754 floats to unsigned norms.
+ *
+ * The mathematical voodoo below may seem excessive but it is actually
+ * paramount we do it this way for several reasons. First, there is no single
+ * precision FP to unsigned integer conversion Intel SSE instruction. Second,
+ * secondly, even if there was, since the FP's mantissa takes only a fraction
+ * of register bits the typically scale and cast approach would require double
+ * precision for accurate results, and therefore half the throughput
+ *
+ * Although the result values can be scaled to an arbitrary bit width specified
+ * by dst_width, the actual result type will have the same width.
+ */
+LLVMValueRef
+lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
+                                        union lp_type src_type,
+                                        unsigned dst_width,
+                                        LLVMValueRef src)
+{
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(src_type);
+   LLVMValueRef res;
+   unsigned mantissa;
+   unsigned n;
+   unsigned long long ubound;
+   unsigned long long mask;
+   double scale;
+   double bias;
+
+   assert(src_type.floating);
+
+   mantissa = lp_mantissa(src_type);
+
+   /* We cannot carry more bits than the mantissa */
+   n = MIN2(mantissa, dst_width);
+
+   /* This magic coefficients will make the desired result to appear in the
+    * lowest significant bits of the mantissa.
+    */
+   ubound = ((unsigned long long)1 << n);
+   mask = ubound - 1;
+   scale = (double)mask/ubound;
+   bias = (double)((unsigned long long)1 << (mantissa - n));
+
+   res = LLVMBuildMul(builder, src, lp_build_const_scalar(src_type, scale), "");
+   res = LLVMBuildAdd(builder, res, lp_build_const_scalar(src_type, bias), "");
+   res = LLVMBuildBitCast(builder, res, int_vec_type, "");
+
+   if(dst_width > n) {
+      int shift = dst_width - n;
+      res = LLVMBuildShl(builder, res, lp_build_int_const_scalar(src_type, shift), "");
+
+      /* Fill in the empty lower bits for added precision? */
+#if 0
+      {
+         LLVMValueRef msb;
+         msb = LLVMBuildLShr(builder, res, lp_build_int_const_scalar(src_type, dst_width - 1), "");
+         msb = LLVMBuildShl(builder, msb, lp_build_int_const_scalar(src_type, shift), "");
+         msb = LLVMBuildSub(builder, msb, lp_build_int_const_scalar(src_type, 1), "");
+         res = LLVMBuildOr(builder, res, msb, "");
+      }
+#elif 0
+      while(shift > 0) {
+         res = LLVMBuildOr(builder, res, LLVMBuildLShr(builder, res, lp_build_int_const_scalar(src_type, n), ""), "");
+         shift -= n;
+         n *= 2;
+      }
+#endif
+   }
+   else
+      res = LLVMBuildAnd(builder, res, lp_build_int_const_scalar(src_type, mask), "");
+
+   return res;
+}
+
+
+/**
+ * Inverse of lp_build_clamped_float_to_unsigned_norm above.
+ */
+LLVMValueRef
+lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
+                                unsigned src_width,
+                                union lp_type dst_type,
+                                LLVMValueRef src)
+{
+   LLVMTypeRef vec_type = lp_build_vec_type(dst_type);
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(dst_type);
+   LLVMValueRef bias_;
+   LLVMValueRef res;
+   unsigned mantissa;
+   unsigned n;
+   unsigned long long ubound;
+   unsigned long long mask;
+   double scale;
+   double bias;
+
+   mantissa = lp_mantissa(dst_type);
+
+   n = MIN2(mantissa, src_width);
+
+   ubound = ((unsigned long long)1 << n);
+   mask = ubound - 1;
+   scale = (double)ubound/mask;
+   bias = (double)((unsigned long long)1 << (mantissa - n));
+
+   res = src;
+
+   if(src_width > mantissa) {
+      int shift = src_width - mantissa;
+      res = LLVMBuildLShr(builder, res, lp_build_int_const_scalar(dst_type, shift), "");
+   }
+
+   bias_ = lp_build_const_scalar(dst_type, bias);
+
+   res = LLVMBuildOr(builder,
+                     res,
+                     LLVMBuildBitCast(builder, bias_, int_vec_type, ""), "");
+
+   res = LLVMBuildBitCast(builder, res, vec_type, "");
+
+   res = LLVMBuildSub(builder, res, bias_, "");
+   res = LLVMBuildMul(builder, res, lp_build_const_scalar(dst_type, scale), "");
+
+   return res;
+}
+
+
+/**
+ * Build shuffle vectors that match PUNPCKLxx and PUNPCKHxx instructions.
+ */
+static LLVMValueRef
+lp_build_const_unpack_shuffle(unsigned n, unsigned lo_hi)
+{
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i, j;
+
+   assert(n <= LP_MAX_VECTOR_LENGTH);
+   assert(lo_hi < 2);
+
+   /* TODO: cache results in a static table */
+
+   for(i = 0, j = lo_hi*n/2; i < n; i += 2, ++j) {
+      elems[i + 0] = LLVMConstInt(LLVMInt32Type(), 0 + j, 0);
+      elems[i + 1] = LLVMConstInt(LLVMInt32Type(), n + j, 0);
+   }
+
+   return LLVMConstVector(elems, n);
+}
+
+
+/**
+ * Build shuffle vectors that match PACKxx instructions.
+ */
+static LLVMValueRef
+lp_build_const_pack_shuffle(unsigned n)
+{
+   LLVMValueRef elems[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   assert(n <= LP_MAX_VECTOR_LENGTH);
+
+   /* TODO: cache results in a static table */
+
+   for(i = 0; i < n; ++i)
+      elems[i] = LLVMConstInt(LLVMInt32Type(), 2*i, 0);
+
+   return LLVMConstVector(elems, n);
+}
+
+
+/**
+ * Expand the bit width.
+ *
+ * This will only change the number of bits the values are represented, not the
+ * values themselved.
+ */
+static void
+lp_build_expand(LLVMBuilderRef builder,
+               union lp_type src_type,
+               union lp_type dst_type,
+               LLVMValueRef src,
+               LLVMValueRef *dst, unsigned num_dsts)
+{
+   unsigned num_tmps;
+   unsigned i;
+
+   /* Register width must remain constant */
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length == dst_type.length * num_dsts);
+
+   num_tmps = 1;
+   dst[0] = src;
+
+   while(src_type.width < dst_type.width) {
+      union lp_type new_type = src_type;
+      LLVMTypeRef new_vec_type;
+
+      new_type.width *= 2;
+      new_type.length /= 2;
+      new_vec_type = lp_build_vec_type(new_type);
+
+      for(i = num_tmps; i--; ) {
+         LLVMValueRef zero;
+         LLVMValueRef shuffle_lo;
+         LLVMValueRef shuffle_hi;
+         LLVMValueRef lo;
+         LLVMValueRef hi;
+
+         zero = lp_build_zero(src_type);
+         shuffle_lo = lp_build_const_unpack_shuffle(src_type.length, 0);
+         shuffle_hi = lp_build_const_unpack_shuffle(src_type.length, 1);
+
+         /*  PUNPCKLBW, PUNPCKHBW */
+         lo = LLVMBuildShuffleVector(builder, dst[i], zero, shuffle_lo, "");
+         hi = LLVMBuildShuffleVector(builder, dst[i], zero, shuffle_hi, "");
+
+         dst[2*i + 0] = LLVMBuildBitCast(builder, lo, new_vec_type, "");
+         dst[2*i + 1] = LLVMBuildBitCast(builder, hi, new_vec_type, "");
+      }
+
+      src_type = new_type;
+
+      num_tmps *= 2;
+   }
+
+   assert(num_tmps == num_dsts);
+}
+
+
+/**
+ * Non-interleaved pack.
+ *
+ * This will move values as
+ *
+ *   lo =   __ l0 __ l1 __ l2 __..  __ ln
+ *   hi =   __ h0 __ h1 __ h2 __..  __ hn
+ *   res =  l0 l1 l2 .. ln h0 h1 h2 .. hn
+ *
+ * TODO: handle saturation consistently.
+ */
+static LLVMValueRef
+lp_build_pack2(LLVMBuilderRef builder,
+               union lp_type src_type,
+               union lp_type dst_type,
+               boolean clamped,
+               LLVMValueRef lo,
+               LLVMValueRef hi)
+{
+   LLVMTypeRef src_vec_type = lp_build_vec_type(src_type);
+   LLVMTypeRef dst_vec_type = lp_build_vec_type(dst_type);
+   LLVMValueRef shuffle;
+   LLVMValueRef res;
+
+   /* Register width must remain constant */
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length * 2 == dst_type.length);
+
+   assert(!src_type.floating);
+   assert(!dst_type.floating);
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(src_type.width * src_type.length == 128) {
+      /* All X86 non-interleaved pack instructions all take signed inputs and
+       * saturate them, so saturate beforehand. */
+      if(!src_type.sign && !clamped) {
+         struct lp_build_context bld;
+         unsigned dst_bits = dst_type.sign ? dst_type.width - 1 : dst_type.width;
+         LLVMValueRef dst_max = lp_build_int_const_scalar(src_type, ((unsigned long long)1 << dst_bits) - 1);
+         lp_build_context_init(&bld, builder, src_type);
+         lo = lp_build_min(&bld, lo, dst_max);
+         hi = lp_build_min(&bld, hi, dst_max);
+      }
+
+      switch(src_type.width) {
+      case 32:
+         if(dst_type.sign)
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packssdw.128", src_vec_type, lo, hi);
+         else
+            /* PACKUSDW is the only instrinsic with a consistent signature */
+            return lp_build_intrinsic_binary(builder, "llvm.x86.sse41.packusdw", dst_vec_type, lo, hi);
+         break;
+
+      case 16:
+         if(dst_type.sign)
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packsswb.128", src_vec_type, lo, hi);
+         else
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packuswb.128", src_vec_type, lo, hi);
+         break;
+
+      default:
+         assert(0);
+         return LLVMGetUndef(dst_vec_type);
+         break;
+      }
+
+      res = LLVMBuildBitCast(builder, res, dst_vec_type, "");
+      return res;
+   }
+#endif
+
+   lo = LLVMBuildBitCast(builder, lo, dst_vec_type, "");
+   hi = LLVMBuildBitCast(builder, hi, dst_vec_type, "");
+
+   shuffle = lp_build_const_pack_shuffle(dst_type.length);
+
+   res = LLVMBuildShuffleVector(builder, lo, hi, shuffle, "");
+
+   return res;
+}
+
+
+/**
+ * Truncate the bit width.
+ *
+ * TODO: Handle saturation consistently.
+ */
+static LLVMValueRef
+lp_build_trunc(LLVMBuilderRef builder,
+               union lp_type src_type,
+               union lp_type dst_type,
+               boolean clamped,
+               const LLVMValueRef *src, unsigned num_srcs)
+{
+   LLVMValueRef tmp[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   /* Register width must remain constant */
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length * num_srcs == dst_type.length);
+
+   for(i = 0; i < num_srcs; ++i)
+      tmp[i] = src[i];
+
+   while(src_type.width > dst_type.width) {
+      union lp_type new_type = src_type;
+
+      new_type.width /= 2;
+      new_type.length *= 2;
+
+      /* Take in consideration the sign changes only in the last step */
+      if(new_type.width == dst_type.width)
+         new_type.sign = dst_type.sign;
+
+      num_srcs /= 2;
+
+      for(i = 0; i < num_srcs; ++i)
+         tmp[i] = lp_build_pack2(builder, src_type, new_type, clamped,
+                                 tmp[2*i + 0], tmp[2*i + 1]);
+
+      src_type = new_type;
+   }
+
+   assert(num_srcs == 1);
+
+   return tmp[0];
+}
+
+
+/**
+ * Generic type conversion.
+ *
+ * TODO: Take a precision argument, or even better, add a new precision member
+ * to the lp_type union.
+ */
+void
+lp_build_conv(LLVMBuilderRef builder,
+              union lp_type src_type,
+              union lp_type dst_type,
+              const LLVMValueRef *src, unsigned num_srcs,
+              LLVMValueRef *dst, unsigned num_dsts)
+{
+   union lp_type tmp_type;
+   LLVMValueRef tmp[LP_MAX_VECTOR_LENGTH];
+   unsigned num_tmps;
+   unsigned i;
+
+   /* Register width must remain constant */
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length * num_srcs == dst_type.length * num_dsts);
+
+   assert(src_type.length <= LP_MAX_VECTOR_LENGTH);
+   assert(dst_type.length <= LP_MAX_VECTOR_LENGTH);
+
+   tmp_type = src_type;
+   for(i = 0; i < num_srcs; ++i)
+      tmp[i] = src[i];
+   num_tmps = num_srcs;
+
+   /*
+    * Clamp if necessary
+    */
+
+   if(src_type.value != dst_type.value) {
+      struct lp_build_context bld;
+      double src_min = lp_const_min(src_type);
+      double dst_min = lp_const_min(dst_type);
+      double src_max = lp_const_max(src_type);
+      double dst_max = lp_const_max(dst_type);
+      LLVMValueRef thres;
+
+      lp_build_context_init(&bld, builder, tmp_type);
+
+      if(src_min < dst_min) {
+         if(dst_min == 0.0)
+            thres = bld.zero;
+         else
+            thres = lp_build_const_scalar(src_type, dst_min);
+         for(i = 0; i < num_tmps; ++i)
+            tmp[i] = lp_build_max(&bld, tmp[i], thres);
+      }
+
+      if(src_max > dst_max) {
+         if(dst_max == 1.0)
+            thres = bld.one;
+         else
+            thres = lp_build_const_scalar(src_type, dst_max);
+         for(i = 0; i < num_tmps; ++i)
+            tmp[i] = lp_build_min(&bld, tmp[i], thres);
+      }
+   }
+
+   /*
+    * Scale to the narrowest range
+    */
+
+   if(dst_type.floating) {
+      /* Nothing to do */
+   }
+   else if(tmp_type.floating) {
+      if(!dst_type.fixed && !dst_type.sign && dst_type.norm) {
+         for(i = 0; i < num_tmps; ++i) {
+            tmp[i] = lp_build_clamped_float_to_unsigned_norm(builder,
+                                                             tmp_type,
+                                                             dst_type.width,
+                                                             tmp[i]);
+         }
+         tmp_type.floating = FALSE;
+      }
+      else {
+         double dst_scale = lp_const_scale(dst_type);
+         LLVMTypeRef tmp_vec_type;
+
+         if (dst_scale != 1.0) {
+            LLVMValueRef scale = lp_build_const_scalar(tmp_type, dst_scale);
+            for(i = 0; i < num_tmps; ++i)
+               tmp[i] = LLVMBuildMul(builder, tmp[i], scale, "");
+         }
+
+         /* Use an equally sized integer for intermediate computations */
+         tmp_type.floating = FALSE;
+         tmp_vec_type = lp_build_vec_type(tmp_type);
+         for(i = 0; i < num_tmps; ++i) {
+#if 0
+            if(dst_type.sign)
+               tmp[i] = LLVMBuildFPToSI(builder, tmp[i], tmp_vec_type, "");
+            else
+               tmp[i] = LLVMBuildFPToUI(builder, tmp[i], tmp_vec_type, "");
+#else
+           /* FIXME: there is no SSE counterpart for LLVMBuildFPToUI */
+            tmp[i] = LLVMBuildFPToSI(builder, tmp[i], tmp_vec_type, "");
+#endif
+         }
+      }
+   }
+   else {
+      unsigned src_shift = lp_const_shift(src_type);
+      unsigned dst_shift = lp_const_shift(dst_type);
+
+      /* FIXME: compensate different offsets too */
+      if(src_shift > dst_shift) {
+         LLVMValueRef shift = lp_build_int_const_scalar(tmp_type, src_shift - dst_shift);
+         for(i = 0; i < num_tmps; ++i)
+            if(src_type.sign)
+               tmp[i] = LLVMBuildAShr(builder, tmp[i], shift, "");
+            else
+               tmp[i] = LLVMBuildLShr(builder, tmp[i], shift, "");
+      }
+   }
+
+   /*
+    * Truncate or expand bit width
+    */
+
+   assert(!tmp_type.floating || tmp_type.width == dst_type.width);
+
+   if(tmp_type.width > dst_type.width) {
+      assert(num_dsts == 1);
+      tmp[0] = lp_build_trunc(builder, tmp_type, dst_type, TRUE, tmp, num_tmps);
+      tmp_type.width = dst_type.width;
+      tmp_type.length = dst_type.length;
+      num_tmps = 1;
+   }
+
+   if(tmp_type.width < dst_type.width) {
+      assert(num_tmps == 1);
+      lp_build_expand(builder, tmp_type, dst_type, tmp[0], tmp, num_dsts);
+      tmp_type.width = dst_type.width;
+      tmp_type.length = dst_type.length;
+      num_tmps = num_dsts;
+   }
+
+   assert(tmp_type.width == dst_type.width);
+   assert(tmp_type.length == dst_type.length);
+   assert(num_tmps == num_dsts);
+
+   /*
+    * Scale to the widest range
+    */
+
+   if(src_type.floating) {
+      /* Nothing to do */
+   }
+   else if(!src_type.floating && dst_type.floating) {
+      if(!src_type.fixed && !src_type.sign && src_type.norm) {
+         for(i = 0; i < num_tmps; ++i) {
+            tmp[i] = lp_build_unsigned_norm_to_float(builder,
+                                                     src_type.width,
+                                                     dst_type,
+                                                     tmp[i]);
+         }
+         tmp_type.floating = TRUE;
+      }
+      else {
+         double src_scale = lp_const_scale(src_type);
+         LLVMTypeRef tmp_vec_type;
+
+         /* Use an equally sized integer for intermediate computations */
+         tmp_type.floating = TRUE;
+         tmp_type.sign = TRUE;
+         tmp_vec_type = lp_build_vec_type(tmp_type);
+         for(i = 0; i < num_tmps; ++i) {
+#if 0
+            if(dst_type.sign)
+               tmp[i] = LLVMBuildSIToFP(builder, tmp[i], tmp_vec_type, "");
+            else
+               tmp[i] = LLVMBuildUIToFP(builder, tmp[i], tmp_vec_type, "");
+#else
+            /* FIXME: there is no SSE counterpart for LLVMBuildUIToFP */
+            tmp[i] = LLVMBuildSIToFP(builder, tmp[i], tmp_vec_type, "");
+#endif
+          }
+
+          if (src_scale != 1.0) {
+             LLVMValueRef scale = lp_build_const_scalar(tmp_type, 1.0/src_scale);
+             for(i = 0; i < num_tmps; ++i)
+                tmp[i] = LLVMBuildMul(builder, tmp[i], scale, "");
+          }
+      }
+    }
+    else {
+       unsigned src_shift = lp_const_shift(src_type);
+       unsigned dst_shift = lp_const_shift(dst_type);
+
+       /* FIXME: compensate different offsets too */
+       if(src_shift < dst_shift) {
+          LLVMValueRef shift = lp_build_int_const_scalar(tmp_type, dst_shift - src_shift);
+          for(i = 0; i < num_tmps; ++i)
+             tmp[i] = LLVMBuildShl(builder, tmp[i], shift, "");
+       }
+    }
+
+   for(i = 0; i < num_dsts; ++i)
+      dst[i] = tmp[i];
+}
+
+
+/**
+ * Bit mask conversion.
+ *
+ * This will convert the integer masks that match the given types.
+ *
+ * The mask values should 0 or -1, i.e., all bits either set to zero or one.
+ * Any other value will likely cause in unpredictable results.
+ *
+ * This is basically a very trimmed down version of lp_build_conv.
+ */
+void
+lp_build_conv_mask(LLVMBuilderRef builder,
+                   union lp_type src_type,
+                   union lp_type dst_type,
+                   const LLVMValueRef *src, unsigned num_srcs,
+                   LLVMValueRef *dst, unsigned num_dsts)
+{
+   /* Register width must remain constant */
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length * num_srcs == dst_type.length * num_dsts);
+
+   /*
+    * Drop
+    *
+    * We assume all values are 0 or -1
+    */
+
+   src_type.floating = FALSE;
+   src_type.fixed = FALSE;
+   src_type.sign = TRUE;
+   src_type.norm = FALSE;
+
+   dst_type.floating = FALSE;
+   dst_type.fixed = FALSE;
+   dst_type.sign = TRUE;
+   dst_type.norm = FALSE;
+
+   /*
+    * Truncate or expand bit width
+    */
+
+   if(src_type.width > dst_type.width) {
+      assert(num_dsts == 1);
+      dst[0] = lp_build_trunc(builder, src_type, dst_type, TRUE, src, num_srcs);
+   }
+   else if(src_type.width < dst_type.width) {
+      assert(num_srcs == 1);
+      lp_build_expand(builder, src_type, dst_type, src[0], dst, num_dsts);
+   }
+   else {
+      assert(num_srcs == num_dsts);
+      memcpy(dst, src, num_dsts * sizeof *dst);
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_conv.h b/src/gallium/drivers/llvmpipe/lp_bld_conv.h
new file mode 100644
index 0000000000..05c1ef2a10
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_conv.h
@@ -0,0 +1,73 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for type conversions.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_CONV_H
+#define LP_BLD_CONV_H
+
+
+#include <llvm-c/Core.h>  
+
+
+union lp_type type;
+
+
+LLVMValueRef
+lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
+                                        union lp_type src_type,
+                                        unsigned dst_width,
+                                        LLVMValueRef src);
+
+LLVMValueRef
+lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
+                                unsigned src_width,
+                                union lp_type dst_type,
+                                LLVMValueRef src);
+
+
+void
+lp_build_conv(LLVMBuilderRef builder,
+              union lp_type src_type,
+              union lp_type dst_type,
+              const LLVMValueRef *srcs, unsigned num_srcs,
+              LLVMValueRef *dsts, unsigned num_dsts);
+
+void
+lp_build_conv_mask(LLVMBuilderRef builder,
+                   union lp_type src_type,
+                   union lp_type dst_type,
+                   const LLVMValueRef *src, unsigned num_srcs,
+                   LLVMValueRef *dst, unsigned num_dsts);
+
+#endif /* !LP_BLD_CONV_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_debug.c b/src/gallium/drivers/llvmpipe/lp_bld_debug.c
new file mode 100644
index 0000000000..30925b5f41
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_debug.c
@@ -0,0 +1,108 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifdef HAVE_UDIS86
+#include <udis86.h>
+#endif
+
+#include "util/u_debug.h"
+#include "lp_bld_debug.h"
+
+
+void
+lp_disassemble(const void* func)
+{
+#ifdef HAVE_UDIS86
+   ud_t ud_obj;
+   uint64_t max_jmp_pc;
+
+   ud_init(&ud_obj);
+
+   ud_set_input_buffer(&ud_obj, (void*)func, 0xffff);
+
+   max_jmp_pc = (uint64_t) (uintptr_t) func;
+   ud_set_pc(&ud_obj, max_jmp_pc);
+
+#ifdef PIPE_ARCH_X86
+   ud_set_mode(&ud_obj, 32);
+#endif
+#ifdef PIPE_ARCH_X86_64
+   ud_set_mode(&ud_obj, 64);
+#endif
+
+   ud_set_syntax(&ud_obj, UD_SYN_ATT);
+
+   while (ud_disassemble(&ud_obj)) {
+
+#ifdef PIPE_ARCH_X86
+      debug_printf("%08lx: ", (unsigned long)ud_insn_off(&ud_obj));
+#endif
+#ifdef PIPE_ARCH_X86_64
+      debug_printf("%016llx: ", (unsigned long long)ud_insn_off(&ud_obj));
+#endif
+
+#if 0
+      debug_printf("%-16s ", ud_insn_hex(&ud_obj));
+#endif
+
+      debug_printf("%s\n", ud_insn_asm(&ud_obj));
+
+      if(ud_obj.mnemonic != UD_Icall) {
+         unsigned i;
+         for(i = 0; i < 3; ++i) {
+            const struct ud_operand *op = &ud_obj.operand[i];
+            if (op->type == UD_OP_JIMM){
+               uint64_t pc = ud_obj.pc;
+
+               switch (op->size) {
+               case 8:
+                  pc += op->lval.sbyte;
+                  break;
+               case 16:
+                  pc += op->lval.sword;
+                  break;
+               case 32:
+                  pc += op->lval.sdword;
+                  break;
+               default:
+                  break;
+               }
+               if(pc > max_jmp_pc)
+                  max_jmp_pc = pc;
+            }
+         }
+      }
+
+      if (ud_insn_off(&ud_obj) >= max_jmp_pc && ud_obj.mnemonic == UD_Iret)
+         break;
+   }
+   debug_printf("\n");
+#else
+   (void)func;
+#endif
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_debug.h b/src/gallium/drivers/llvmpipe/lp_bld_debug.h
new file mode 100644
index 0000000000..ecdafef76d
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_debug.h
@@ -0,0 +1,60 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef LP_BLD_DEBUG_H
+#define LP_BLD_DEBUG_H
+
+
+#include <llvm-c/Core.h>
+
+#include "pipe/p_compiler.h"
+#include "util/u_string.h"
+
+
+static INLINE void
+lp_build_name(LLVMValueRef val, const char *format, ...)
+{
+#ifdef DEBUG
+   char name[32];
+   va_list ap;
+   va_start(ap, format);
+   util_vsnprintf(name, sizeof name, format, ap);
+   va_end(ap);
+   LLVMSetValueName(val, name);
+#else
+   (void)val;
+   (void)format;
+#endif
+}
+
+
+void
+lp_disassemble(const void* func);
+
+
+#endif /* !LP_BLD_DEBUG_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_depth.c b/src/gallium/drivers/llvmpipe/lp_bld_depth.c
new file mode 100644
index 0000000000..2cd6e6b921
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_depth.c
@@ -0,0 +1,214 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Depth/stencil testing to LLVM IR translation.
+ *
+ * To be done accurately/efficiently the depth/stencil test must be done with
+ * the same type/format of the depth/stencil buffer, which implies massaging
+ * the incoming depths to fit into place. Using a more straightforward
+ * type/format for depth/stencil values internally and only convert when
+ * flushing would avoid this, but it would most likely result in depth fighting
+ * artifacts.
+ *
+ * We are free to use a different pixel layout though. Since our basic
+ * processing unit is a quad (2x2 pixel block) we store the depth/stencil
+ * values tiled, a quad at time. That is, a depth buffer containing 
+ *
+ *  Z11 Z12 Z13 Z14 ...
+ *  Z21 Z22 Z23 Z24 ...
+ *  Z31 Z32 Z33 Z34 ...
+ *  Z41 Z42 Z43 Z44 ...
+ *  ... ... ... ... ...
+ *
+ * will actually be stored in memory as
+ *
+ *  Z11 Z12 Z21 Z22 Z13 Z14 Z23 Z24 ...
+ *  Z31 Z32 Z41 Z42 Z33 Z34 Z43 Z44 ...
+ *  ... ... ... ... ... ... ... ... ...
+ *
+ * FIXME: Code generate stencil test
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "pipe/p_state.h"
+#include "util/u_format.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_flow.h"
+#include "lp_bld_debug.h"
+#include "lp_bld_depth.h"
+
+
+/**
+ * Return a type appropriate for depth/stencil testing.
+ */
+union lp_type
+lp_depth_type(const struct util_format_description *format_desc,
+              unsigned length)
+{
+   union lp_type type;
+   unsigned swizzle;
+
+   assert(format_desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS);
+   assert(format_desc->block.width == 1);
+   assert(format_desc->block.height == 1);
+
+   swizzle = format_desc->swizzle[0];
+   assert(swizzle < 4);
+
+   type.value = 0;
+   type.width = format_desc->block.bits;
+
+   if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_FLOAT) {
+      type.floating = TRUE;
+      assert(swizzle = 0);
+      assert(format_desc->channel[swizzle].size == format_desc->block.bits);
+   }
+   else if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_UNSIGNED) {
+      assert(format_desc->block.bits <= 32);
+      if(format_desc->channel[swizzle].normalized)
+         type.norm = TRUE;
+   }
+   else
+      assert(0);
+
+   assert(type.width <= length);
+   type.length = length / type.width;
+
+   return type;
+}
+
+
+/**
+ * Depth test.
+ */
+void
+lp_build_depth_test(LLVMBuilderRef builder,
+                    const struct pipe_depth_state *state,
+                    union lp_type type,
+                    const struct util_format_description *format_desc,
+                    struct lp_build_mask_context *mask,
+                    LLVMValueRef src,
+                    LLVMValueRef dst_ptr)
+{
+   struct lp_build_context bld;
+   unsigned z_swizzle;
+   LLVMValueRef dst;
+   LLVMValueRef z_bitmask = NULL;
+   LLVMValueRef test;
+
+   if(!state->enabled)
+      return;
+
+   assert(format_desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS);
+   assert(format_desc->block.width == 1);
+   assert(format_desc->block.height == 1);
+
+   z_swizzle = format_desc->swizzle[0];
+   if(z_swizzle == UTIL_FORMAT_SWIZZLE_NONE)
+      return;
+
+   /* Sanity checking */
+   assert(z_swizzle < 4);
+   assert(format_desc->block.bits == type.width);
+   if(type.floating) {
+      assert(z_swizzle == 0);
+      assert(format_desc->channel[z_swizzle].type == UTIL_FORMAT_TYPE_FLOAT);
+      assert(format_desc->channel[z_swizzle].size == format_desc->block.bits);
+   }
+   else {
+      assert(format_desc->channel[z_swizzle].type == UTIL_FORMAT_TYPE_UNSIGNED);
+      assert(format_desc->channel[z_swizzle].normalized);
+      assert(!type.fixed);
+      assert(!type.sign);
+      assert(type.norm);
+   }
+
+   /* Setup build context */
+   lp_build_context_init(&bld, builder, type);
+
+   dst = LLVMBuildLoad(builder, dst_ptr, "");
+
+   lp_build_name(dst, "zsbuf");
+
+   /* Align the source depth bits with the destination's, and mask out any
+    * stencil or padding bits from both */
+   if(format_desc->channel[z_swizzle].size == format_desc->block.bits) {
+      assert(z_swizzle == 0);
+      /* nothing to do */
+   }
+   else {
+      unsigned padding_left;
+      unsigned padding_right;
+      unsigned chan;
+
+      assert(format_desc->layout == UTIL_FORMAT_LAYOUT_ARITH);
+      assert(format_desc->channel[z_swizzle].type == UTIL_FORMAT_TYPE_UNSIGNED);
+      assert(format_desc->channel[z_swizzle].size <= format_desc->block.bits);
+      assert(format_desc->channel[z_swizzle].normalized);
+
+      padding_right = 0;
+      for(chan = 0; chan < z_swizzle; ++chan)
+         padding_right += format_desc->channel[chan].size;
+      padding_left = format_desc->block.bits - format_desc->channel[z_swizzle].size;
+
+      if(padding_left || padding_right) {
+         const long long mask_left = ((long long)1 << (format_desc->block.bits - padding_left)) - 1;
+         const long long mask_right = ((long long)1 << (padding_right)) - 1;
+         z_bitmask = lp_build_int_const_scalar(type, mask_left & mask_right);
+      }
+
+      if(padding_left)
+         src = LLVMBuildLShr(builder, src, lp_build_int_const_scalar(type, padding_left), "");
+      if(padding_right)
+         src = LLVMBuildAnd(builder, src, z_bitmask, "");
+      if(padding_left || padding_right)
+         dst = LLVMBuildAnd(builder, dst, z_bitmask, "");
+   }
+
+   lp_build_name(dst, "zsbuf.z");
+
+   test = lp_build_cmp(&bld, state->func, src, dst);
+   lp_build_mask_update(mask, test);
+
+   if(state->writemask) {
+      if(z_bitmask)
+         z_bitmask = LLVMBuildAnd(builder, mask->value, z_bitmask, "");
+      else
+         z_bitmask = mask->value;
+
+      dst = lp_build_select(&bld, z_bitmask, src, dst);
+      LLVMBuildStore(builder, dst, dst_ptr);
+   }
+
+   assert(!state->occlusion_count);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_depth.h b/src/gallium/drivers/llvmpipe/lp_bld_depth.h
new file mode 100644
index 0000000000..5d2e042fcc
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_depth.h
@@ -0,0 +1,63 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * Depth/stencil testing to LLVM IR translation.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_BLD_DEPTH_H
+#define LP_BLD_DEPTH_H
+
+
+#include <llvm-c/Core.h>  
+
+ 
+struct pipe_depth_state;
+struct util_format_description;
+union lp_type;
+struct lp_build_mask_context;
+
+
+union lp_type
+lp_depth_type(const struct util_format_description *format_desc,
+              unsigned length);
+
+
+void
+lp_build_depth_test(LLVMBuilderRef builder,
+                    const struct pipe_depth_state *state,
+                    union lp_type type,
+                    const struct util_format_description *format_desc,
+                    struct lp_build_mask_context *mask,
+                    LLVMValueRef src,
+                    LLVMValueRef dst_ptr);
+
+
+#endif /* !LP_BLD_DEPTH_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_flow.c b/src/gallium/drivers/llvmpipe/lp_bld_flow.c
new file mode 100644
index 0000000000..9d99e1a9d9
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_flow.c
@@ -0,0 +1,173 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * LLVM control flow build helpers.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "util/u_debug.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_flow.h"
+
+
+void
+lp_build_mask_begin(struct lp_build_mask_context *mask,
+                    LLVMBuilderRef builder,
+                    union lp_type type,
+                    LLVMValueRef value)
+{
+   memset(mask, 0, sizeof *mask);
+
+   mask->builder = builder;
+   mask->reg_type = LLVMIntType(type.width * type.length);
+   mask->value = value;
+}
+
+
+void
+lp_build_mask_update(struct lp_build_mask_context *mask,
+                     LLVMValueRef value)
+{
+
+   LLVMValueRef cond;
+   LLVMBasicBlockRef current_block;
+   LLVMBasicBlockRef next_block;
+   LLVMBasicBlockRef new_block;
+
+   if(mask->value)
+      mask->value = LLVMBuildAnd(mask->builder, mask->value, value, "");
+   else
+      mask->value = value;
+
+   /* FIXME: disabled until we have proper control flow helpers */
+#if 0
+   cond = LLVMBuildICmp(mask->builder,
+                        LLVMIntEQ,
+                        LLVMBuildBitCast(mask->builder, mask->value, mask->reg_type, ""),
+                        LLVMConstNull(mask->reg_type),
+                        "");
+
+   current_block = LLVMGetInsertBlock(mask->builder);
+
+   if(!mask->skip_block) {
+      LLVMValueRef function = LLVMGetBasicBlockParent(current_block);
+      mask->skip_block = LLVMAppendBasicBlock(function, "skip");
+
+      mask->phi = LLVMBuildPhi(mask->builder, LLVMTypeOf(mask->value), "");
+   }
+
+   next_block = LLVMGetNextBasicBlock(current_block);
+   assert(next_block);
+   if(next_block) {
+      new_block = LLVMInsertBasicBlock(next_block, "");
+   }
+   else {
+      LLVMValueRef function = LLVMGetBasicBlockParent(current_block);
+      new_block = LLVMAppendBasicBlock(function, "");
+   }
+
+   LLVMAddIncoming(mask->phi, &mask->value, &current_block, 1);
+   LLVMBuildCondBr(mask->builder, cond, mask->skip_block, new_block);
+
+   LLVMPositionBuilderAtEnd(mask->builder, new_block);
+#endif
+}
+
+
+LLVMValueRef
+lp_build_mask_end(struct lp_build_mask_context *mask)
+{
+   if(mask->skip_block) {
+      LLVMBasicBlockRef current_block = LLVMGetInsertBlock(mask->builder);
+
+      LLVMAddIncoming(mask->phi, &mask->value, &current_block, 1);
+      LLVMBuildBr(mask->builder, mask->skip_block);
+
+      LLVMPositionBuilderAtEnd(mask->builder, mask->skip_block);
+
+      mask->value = mask->phi;
+      mask->phi = NULL;
+      mask->skip_block = NULL;
+   }
+
+   return mask->value;
+}
+
+
+
+void
+lp_build_loop_begin(LLVMBuilderRef builder,
+                    LLVMValueRef start,
+                    struct lp_build_loop_state *state)
+{
+   LLVMBasicBlockRef block = LLVMGetInsertBlock(builder);
+   LLVMValueRef function = LLVMGetBasicBlockParent(block);
+
+   state->block = LLVMAppendBasicBlock(function, "loop");
+
+   LLVMBuildBr(builder, state->block);
+
+   LLVMPositionBuilderAtEnd(builder, state->block);
+
+   state->counter = LLVMBuildPhi(builder, LLVMTypeOf(start), "");
+
+   LLVMAddIncoming(state->counter, &start, &block, 1);
+
+}
+
+
+void
+lp_build_loop_end(LLVMBuilderRef builder,
+                  LLVMValueRef end,
+                  LLVMValueRef step,
+                  struct lp_build_loop_state *state)
+{
+   LLVMBasicBlockRef block = LLVMGetInsertBlock(builder);
+   LLVMValueRef function = LLVMGetBasicBlockParent(block);
+   LLVMValueRef next;
+   LLVMValueRef cond;
+   LLVMBasicBlockRef after_block;
+
+   if (!step)
+      step = LLVMConstInt(LLVMTypeOf(end), 1, 0);
+
+   next = LLVMBuildAdd(builder, state->counter, step, "");
+
+   cond = LLVMBuildICmp(builder, LLVMIntNE, next, end, "");
+
+   after_block = LLVMAppendBasicBlock(function, "");
+
+   LLVMBuildCondBr(builder, cond, after_block, state->block);
+
+   LLVMAddIncoming(state->counter, &next, &block, 1);
+
+   LLVMPositionBuilderAtEnd(builder, after_block);
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_flow.h b/src/gallium/drivers/llvmpipe/lp_bld_flow.h
new file mode 100644
index 0000000000..1b634ff038
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_flow.h
@@ -0,0 +1,103 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * LLVM control flow build helpers.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_BLD_FLOW_H
+#define LP_BLD_FLOW_H
+
+
+#include <llvm-c/Core.h>  
+
+
+union lp_type;
+
+
+struct lp_build_mask_context
+{
+   LLVMBuilderRef builder;
+
+   LLVMTypeRef reg_type;
+
+   LLVMValueRef value;
+
+   LLVMValueRef phi;
+
+   LLVMBasicBlockRef skip_block;
+};
+
+
+void
+lp_build_mask_begin(struct lp_build_mask_context *mask,
+                    LLVMBuilderRef builder,
+                    union lp_type type,
+                    LLVMValueRef value);
+
+/**
+ * Bitwise AND the mask with the given value, if a previous mask was set.
+ */
+void
+lp_build_mask_update(struct lp_build_mask_context *mask,
+                     LLVMValueRef value);
+
+LLVMValueRef
+lp_build_mask_end(struct lp_build_mask_context *mask);
+
+
+/**
+ * LLVM's IR doesn't represent for-loops directly. Furthermore it
+ * it requires creating code blocks, branches, phi variables, so it
+ * requires a fair amount of code.
+ *
+ * @sa http://www.llvm.org/docs/tutorial/LangImpl5.html#for
+ */
+struct lp_build_loop_state
+{
+  LLVMBasicBlockRef block;
+  LLVMValueRef counter;
+};
+
+
+void
+lp_build_loop_begin(LLVMBuilderRef builder,
+                    LLVMValueRef start,
+                    struct lp_build_loop_state *state);
+
+
+void
+lp_build_loop_end(LLVMBuilderRef builder,
+                  LLVMValueRef end,
+                  LLVMValueRef step,
+                  struct lp_build_loop_state *state);
+
+
+
+#endif /* !LP_BLD_FLOW_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_format.h b/src/gallium/drivers/llvmpipe/lp_bld_format.h
new file mode 100644
index 0000000000..01c8a752d1
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_format.h
@@ -0,0 +1,101 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef LP_BLD_H
+#define LP_BLD_H
+
+
+/**
+ * @file
+ * LLVM IR building helpers interfaces.
+ *
+ * We use LLVM-C bindings for now. They are not documented, but follow the C++
+ * interfaces very closely, and appear to be complete enough for code
+ * genration. See
+ * http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html
+ * for a standalone example.
+ */
+
+#include <llvm-c/Core.h>  
+ 
+#include "pipe/p_format.h"
+
+
+union lp_type;
+
+
+/**
+ * Unpack a pixel into its RGBA components.
+ *
+ * @param packed integer.
+ *
+ * @return RGBA in a 4 floats vector.
+ */
+LLVMValueRef
+lp_build_unpack_rgba(LLVMBuilderRef builder,
+                     enum pipe_format format, 
+                     LLVMValueRef packed);
+
+
+/**
+ * Pack a pixel.
+ *
+ * @param rgba 4 float vector with the unpacked components.
+ */
+LLVMValueRef
+lp_build_pack_rgba(LLVMBuilderRef builder,
+                   enum pipe_format format,
+                   LLVMValueRef rgba);
+
+
+/**
+ * Load a pixel into its RGBA components.
+ *
+ * @param ptr value with the pointer to the packed pixel. Pointer type is
+ * irrelevant.
+ *
+ * @return RGBA in a 4 floats vector.
+ */
+LLVMValueRef
+lp_build_load_rgba(LLVMBuilderRef builder,
+                   enum pipe_format format, 
+                   LLVMValueRef ptr);
+
+
+/**
+ * Store a pixel.
+ *
+ * @param rgba 4 float vector with the unpacked components.
+ */
+void 
+lp_build_store_rgba(LLVMBuilderRef builder,
+                    enum pipe_format format,
+                    LLVMValueRef ptr,
+                    LLVMValueRef rgba);
+
+
+#endif /* !LP_BLD_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c
new file mode 100644
index 0000000000..dcbc0076c7
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c
@@ -0,0 +1,303 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_format.h"
+
+#include "lp_bld_format.h"
+
+
+LLVMValueRef
+lp_build_unpack_rgba(LLVMBuilderRef builder,
+                     enum pipe_format format,
+                     LLVMValueRef packed)
+{
+   const struct util_format_description *desc;
+   LLVMTypeRef type;
+   LLVMValueRef shifted, casted, scaled, masked;
+   LLVMValueRef shifts[4];
+   LLVMValueRef masks[4];
+   LLVMValueRef scales[4];
+   LLVMValueRef swizzles[4];
+   LLVMValueRef aux[4];
+   bool normalized;
+   int empty_channel;
+   unsigned shift;
+   unsigned i;
+
+   desc = util_format_description(format);
+
+   /* FIXME: Support more formats */
+   assert(desc->layout == UTIL_FORMAT_LAYOUT_ARITH);
+   assert(desc->block.width == 1);
+   assert(desc->block.height == 1);
+   assert(desc->block.bits <= 32);
+
+   type = LLVMIntType(desc->block.bits);
+
+   /* Do the intermediate integer computations with 32bit integers since it
+    * matches floating point size */
+   if (desc->block.bits < 32)
+      packed = LLVMBuildZExt(builder, packed, LLVMInt32Type(), "");
+
+   /* Broadcast the packed value to all four channels */
+   packed = LLVMBuildInsertElement(builder,
+                                   LLVMGetUndef(LLVMVectorType(LLVMInt32Type(), 4)),
+                                   packed,
+                                   LLVMConstNull(LLVMInt32Type()),
+                                   "");
+   packed = LLVMBuildShuffleVector(builder,
+                                   packed,
+                                   LLVMGetUndef(LLVMVectorType(LLVMInt32Type(), 4)),
+                                   LLVMConstNull(LLVMVectorType(LLVMInt32Type(), 4)),
+                                   "");
+
+   /* Initialize vector constants */
+   normalized = FALSE;
+   empty_channel = -1;
+   shift = 0;
+   for (i = 0; i < 4; ++i) {
+      unsigned bits = desc->channel[i].size;
+
+      if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
+         shifts[i] = LLVMGetUndef(LLVMInt32Type());
+         masks[i] = LLVMConstNull(LLVMInt32Type());
+         scales[i] =  LLVMConstNull(LLVMFloatType());
+         empty_channel = i;
+      }
+      else {
+         unsigned mask = (1 << bits) - 1;
+
+         assert(desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED);
+         assert(bits < 32);
+
+         shifts[i] = LLVMConstInt(LLVMInt32Type(), shift, 0);
+         masks[i] = LLVMConstInt(LLVMInt32Type(), mask, 0);
+
+         if (desc->channel[i].normalized) {
+            scales[i] = LLVMConstReal(LLVMFloatType(), 1.0/mask);
+            normalized = TRUE;
+         }
+         else
+            scales[i] =  LLVMConstReal(LLVMFloatType(), 1.0);
+      }
+
+      shift += bits;
+   }
+
+   shifted = LLVMBuildLShr(builder, packed, LLVMConstVector(shifts, 4), "");
+   masked = LLVMBuildAnd(builder, shifted, LLVMConstVector(masks, 4), "");
+   // UIToFP can't be expressed in SSE2
+   casted = LLVMBuildSIToFP(builder, masked, LLVMVectorType(LLVMFloatType(), 4), "");
+
+   if (normalized)
+      scaled = LLVMBuildMul(builder, casted, LLVMConstVector(scales, 4), "");
+   else
+      scaled = casted;
+
+   for (i = 0; i < 4; ++i)
+      aux[i] = LLVMGetUndef(LLVMFloatType());
+
+   for (i = 0; i < 4; ++i) {
+      enum util_format_swizzle swizzle = desc->swizzle[i];
+
+      switch (swizzle) {
+      case UTIL_FORMAT_SWIZZLE_X:
+      case UTIL_FORMAT_SWIZZLE_Y:
+      case UTIL_FORMAT_SWIZZLE_Z:
+      case UTIL_FORMAT_SWIZZLE_W:
+         swizzles[i] = LLVMConstInt(LLVMInt32Type(), swizzle, 0);
+         break;
+      case UTIL_FORMAT_SWIZZLE_0:
+         assert(empty_channel >= 0);
+         swizzles[i] = LLVMConstInt(LLVMInt32Type(), empty_channel, 0);
+         break;
+      case UTIL_FORMAT_SWIZZLE_1:
+         swizzles[i] = LLVMConstInt(LLVMInt32Type(), 4, 0);
+         aux[0] = LLVMConstReal(LLVMFloatType(), 1.0);
+         break;
+      case UTIL_FORMAT_SWIZZLE_NONE:
+         swizzles[i] = LLVMGetUndef(LLVMFloatType());
+         assert(0);
+         break;
+      }
+   }
+
+   return LLVMBuildShuffleVector(builder, scaled, LLVMConstVector(aux, 4), LLVMConstVector(swizzles, 4), "");
+}
+
+
+LLVMValueRef
+lp_build_pack_rgba(LLVMBuilderRef builder,
+                   enum pipe_format format,
+                   LLVMValueRef rgba)
+{
+   const struct util_format_description *desc;
+   LLVMTypeRef type;
+   LLVMValueRef packed = NULL;
+   LLVMValueRef swizzles[4];
+   LLVMValueRef shifted, casted, scaled, unswizzled;
+   LLVMValueRef shifts[4];
+   LLVMValueRef scales[4];
+   bool normalized;
+   unsigned shift;
+   unsigned i, j;
+
+   desc = util_format_description(format);
+
+   assert(desc->layout == UTIL_FORMAT_LAYOUT_ARITH);
+   assert(desc->block.width == 1);
+   assert(desc->block.height == 1);
+
+   type = LLVMIntType(desc->block.bits);
+
+   /* Unswizzle the color components into the source vector. */
+   for (i = 0; i < 4; ++i) {
+      for (j = 0; j < 4; ++j) {
+         if (desc->swizzle[j] == i)
+            break;
+      }
+      if (j < 4)
+         swizzles[i] = LLVMConstInt(LLVMInt32Type(), j, 0);
+      else
+         swizzles[i] = LLVMGetUndef(LLVMInt32Type());
+   }
+
+   unswizzled = LLVMBuildShuffleVector(builder, rgba,
+                                       LLVMGetUndef(LLVMVectorType(LLVMFloatType(), 4)),
+                                       LLVMConstVector(swizzles, 4), "");
+
+   normalized = FALSE;
+   shift = 0;
+   for (i = 0; i < 4; ++i) {
+      unsigned bits = desc->channel[i].size;
+
+      if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
+         shifts[i] = LLVMGetUndef(LLVMInt32Type());
+         scales[i] =  LLVMGetUndef(LLVMFloatType());
+      }
+      else {
+         unsigned mask = (1 << bits) - 1;
+
+         assert(desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED);
+         assert(bits < 32);
+
+         shifts[i] = LLVMConstInt(LLVMInt32Type(), shift, 0);
+
+         if (desc->channel[i].normalized) {
+            scales[i] = LLVMConstReal(LLVMFloatType(), mask);
+            normalized = TRUE;
+         }
+         else
+            scales[i] =  LLVMConstReal(LLVMFloatType(), 1.0);
+      }
+
+      shift += bits;
+   }
+
+   if (normalized)
+      scaled = LLVMBuildMul(builder, unswizzled, LLVMConstVector(scales, 4), "");
+   else
+      scaled = unswizzled;
+
+   casted = LLVMBuildFPToSI(builder, scaled, LLVMVectorType(LLVMInt32Type(), 4), "");
+
+   shifted = LLVMBuildShl(builder, casted, LLVMConstVector(shifts, 4), "");
+   
+   /* Bitwise or all components */
+   for (i = 0; i < 4; ++i) {
+      if (desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED) {
+         LLVMValueRef component = LLVMBuildExtractElement(builder, shifted, LLVMConstInt(LLVMInt32Type(), i, 0), "");
+         if (packed)
+            packed = LLVMBuildOr(builder, packed, component, "");
+         else
+            packed = component;
+      }
+   }
+
+   if (!packed)
+      packed = LLVMGetUndef(LLVMInt32Type());
+
+   if (desc->block.bits < 32)
+      packed = LLVMBuildTrunc(builder, packed, type, "");
+
+   return packed;
+}
+
+
+LLVMValueRef
+lp_build_load_rgba(LLVMBuilderRef builder,
+                   enum pipe_format format,
+                   LLVMValueRef ptr)
+{
+   const struct util_format_description *desc;
+   LLVMTypeRef type;
+   LLVMValueRef packed;
+
+   desc = util_format_description(format);
+
+   /* FIXME: Support more formats */
+   assert(desc->layout == UTIL_FORMAT_LAYOUT_ARITH);
+   assert(desc->block.width == 1);
+   assert(desc->block.height == 1);
+   assert(desc->block.bits <= 32);
+
+   type = LLVMIntType(desc->block.bits);
+
+   ptr = LLVMBuildBitCast(builder, ptr, LLVMPointerType(type, 0), "");
+
+   packed = LLVMBuildLoad(builder, ptr, "");
+
+   return lp_build_unpack_rgba(builder, format, packed);
+}
+
+
+void
+lp_build_store_rgba(LLVMBuilderRef builder,
+                    enum pipe_format format,
+                    LLVMValueRef ptr,
+                    LLVMValueRef rgba)
+{
+   const struct util_format_description *desc;
+   LLVMTypeRef type;
+   LLVMValueRef packed;
+
+   desc = util_format_description(format);
+
+   assert(desc->layout == UTIL_FORMAT_LAYOUT_ARITH);
+   assert(desc->block.width == 1);
+   assert(desc->block.height == 1);
+
+   type = LLVMIntType(desc->block.bits);
+
+   packed = lp_build_pack_rgba(builder, format, rgba);
+
+   ptr = LLVMBuildBitCast(builder, ptr, LLVMPointerType(type, 0), "");
+
+   LLVMBuildStore(builder, packed, ptr);
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_interp.c b/src/gallium/drivers/llvmpipe/lp_bld_interp.c
new file mode 100644
index 0000000000..cfe20a0d75
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_interp.c
@@ -0,0 +1,377 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Position and shader input interpolation.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "pipe/p_shader_tokens.h"
+#include "util/u_debug.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "tgsi/tgsi_parse.h"
+#include "lp_bld_debug.h"
+#include "lp_bld_const.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_swizzle.h"
+#include "lp_bld_interp.h"
+
+
+static void
+attrib_name(LLVMValueRef val, unsigned attrib, unsigned chan, const char *suffix)
+{
+   if(attrib == 0)
+      lp_build_name(val, "pos.%c%s", "xyzw"[chan], suffix);
+   else
+      lp_build_name(val, "input%u.%c%s", attrib - 1, "xyzw"[chan], suffix);
+}
+
+
+static void
+coeffs_init(struct lp_build_interp_soa_context *bld,
+            LLVMValueRef a0_ptr,
+            LLVMValueRef dadx_ptr,
+            LLVMValueRef dady_ptr)
+{
+   LLVMBuilderRef builder = bld->base.builder;
+   unsigned attrib;
+   unsigned chan;
+
+   for(attrib = 0; attrib < bld->num_attribs; ++attrib) {
+      unsigned mask = bld->mask[attrib];
+      unsigned mode = bld->mode[attrib];
+      for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+         if(mask & (1 << chan)) {
+            LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), attrib*NUM_CHANNELS + chan, 0);
+            LLVMValueRef a0 = NULL;
+            LLVMValueRef dadx = NULL;
+            LLVMValueRef dady = NULL;
+
+            switch( mode ) {
+            case TGSI_INTERPOLATE_PERSPECTIVE:
+               /* fall-through */
+
+            case TGSI_INTERPOLATE_LINEAR:
+               dadx = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dadx_ptr, &index, 1, ""), "");
+               dady = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dady_ptr, &index, 1, ""), "");
+               dadx = lp_build_broadcast_scalar(&bld->base, dadx);
+               dady = lp_build_broadcast_scalar(&bld->base, dady);
+               attrib_name(dadx, attrib, chan, ".dadx");
+               attrib_name(dady, attrib, chan, ".dady");
+               /* fall-through */
+
+            case TGSI_INTERPOLATE_CONSTANT:
+               a0 = LLVMBuildLoad(builder, LLVMBuildGEP(builder, a0_ptr, &index, 1, ""), "");
+               a0 = lp_build_broadcast_scalar(&bld->base, a0);
+               attrib_name(a0, attrib, chan, ".dady");
+               break;
+
+            default:
+               assert(0);
+               break;
+            }
+
+            bld->a0  [attrib][chan] = a0;
+            bld->dadx[attrib][chan] = dadx;
+            bld->dady[attrib][chan] = dady;
+         }
+      }
+   }
+}
+
+
+/**
+ * Small vector x scale multiplication optimization.
+ *
+ * TODO: Should be elsewhere.
+ */
+static LLVMValueRef
+coeff_multiply(struct lp_build_interp_soa_context *bld,
+               LLVMValueRef coeff,
+               int step)
+{
+   LLVMValueRef factor;
+
+   switch(step) {
+   case 0:
+      return bld->base.zero;
+   case 1:
+      return coeff;
+   case 2:
+      return lp_build_add(&bld->base, coeff, coeff);
+   default:
+      factor = lp_build_const_scalar(bld->base.type, (double)step);
+      return lp_build_mul(&bld->base, coeff, factor);
+   }
+}
+
+
+/**
+ * Multiply the dadx and dady with the xstep and ystep respectively.
+ */
+static void
+coeffs_update(struct lp_build_interp_soa_context *bld)
+{
+   unsigned attrib;
+   unsigned chan;
+
+   for(attrib = 0; attrib < bld->num_attribs; ++attrib) {
+      unsigned mask = bld->mask[attrib];
+      unsigned mode = bld->mode[attrib];
+      if (mode != TGSI_INTERPOLATE_CONSTANT) {
+         for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+            if(mask & (1 << chan)) {
+               bld->dadx[attrib][chan] = coeff_multiply(bld, bld->dadx[attrib][chan], bld->xstep);
+               bld->dady[attrib][chan] = coeff_multiply(bld, bld->dady[attrib][chan], bld->ystep);
+            }
+         }
+      }
+   }
+}
+
+
+static void
+attribs_init(struct lp_build_interp_soa_context *bld)
+{
+   LLVMValueRef x = bld->pos[0];
+   LLVMValueRef y = bld->pos[1];
+   LLVMValueRef oow = NULL;
+   unsigned attrib;
+   unsigned chan;
+
+   for(attrib = 0; attrib < bld->num_attribs; ++attrib) {
+      unsigned mask = bld->mask[attrib];
+      unsigned mode = bld->mode[attrib];
+      for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+         if(mask & (1 << chan)) {
+            LLVMValueRef a0   = bld->a0  [attrib][chan];
+            LLVMValueRef dadx = bld->dadx[attrib][chan];
+            LLVMValueRef dady = bld->dady[attrib][chan];
+            LLVMValueRef res;
+
+            res = a0;
+
+            if (mode != TGSI_INTERPOLATE_CONSTANT) {
+               res = lp_build_add(&bld->base, res, lp_build_mul(&bld->base, x, dadx));
+               res = lp_build_add(&bld->base, res, lp_build_mul(&bld->base, y, dady));
+            }
+
+            /* Keep the value of the attribue before perspective divide for faster updates */
+            bld->attribs_pre[attrib][chan] = res;
+
+            if (mode == TGSI_INTERPOLATE_PERSPECTIVE) {
+               LLVMValueRef w = bld->pos[3];
+               assert(attrib != 0);
+               if(!oow)
+                  oow = lp_build_rcp(&bld->base, w);
+               res = lp_build_mul(&bld->base, res, oow);
+            }
+
+            attrib_name(res, attrib, chan, "");
+
+            bld->attribs[attrib][chan] = res;
+         }
+      }
+   }
+}
+
+
+static void
+attribs_update(struct lp_build_interp_soa_context *bld)
+{
+   LLVMValueRef oow = NULL;
+   unsigned attrib;
+   unsigned chan;
+
+   for(attrib = 0; attrib < bld->num_attribs; ++attrib) {
+      unsigned mask = bld->mask[attrib];
+      unsigned mode = bld->mode[attrib];
+
+      if (mode != TGSI_INTERPOLATE_CONSTANT) {
+         for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+            if(mask & (1 << chan)) {
+               LLVMValueRef dadx = bld->dadx[attrib][chan];
+               LLVMValueRef dady = bld->dady[attrib][chan];
+               LLVMValueRef res;
+
+               res = bld->attribs_pre[attrib][chan];
+
+               if(bld->xstep)
+                  res = lp_build_add(&bld->base, res, dadx);
+
+               if(bld->ystep)
+                  res = lp_build_add(&bld->base, res, dady);
+
+               bld->attribs_pre[attrib][chan] = res;
+
+               if (mode == TGSI_INTERPOLATE_PERSPECTIVE) {
+                  LLVMValueRef w = bld->pos[3];
+                  assert(attrib != 0);
+                  if(!oow)
+                     oow = lp_build_rcp(&bld->base, w);
+                  res = lp_build_mul(&bld->base, res, oow);
+               }
+
+               attrib_name(res, attrib, chan, "");
+
+               bld->attribs[attrib][chan] = res;
+            }
+         }
+      }
+   }
+}
+
+
+/**
+ * Generate the position vectors.
+ *
+ * Parameter x0, y0 are the integer values with the quad upper left coordinates.
+ */
+static void
+pos_init(struct lp_build_interp_soa_context *bld,
+         LLVMValueRef x0,
+         LLVMValueRef y0)
+{
+   lp_build_name(x0, "pos.x");
+   lp_build_name(y0, "pos.y");
+
+   bld->attribs[0][0] = x0;
+   bld->attribs[0][1] = y0;
+}
+
+
+static void
+pos_update(struct lp_build_interp_soa_context *bld)
+{
+   LLVMValueRef x = bld->attribs[0][0];
+   LLVMValueRef y = bld->attribs[0][1];
+
+   if(bld->xstep)
+      x = lp_build_add(&bld->base, x, lp_build_const_scalar(bld->base.type, bld->xstep));
+
+   if(bld->ystep)
+      y = lp_build_add(&bld->base, y, lp_build_const_scalar(bld->base.type, bld->ystep));
+
+   lp_build_name(x, "pos.x");
+   lp_build_name(y, "pos.y");
+
+   bld->attribs[0][0] = x;
+   bld->attribs[0][1] = y;
+}
+
+
+void
+lp_build_interp_soa_init(struct lp_build_interp_soa_context *bld,
+                         const struct tgsi_token *tokens,
+                         LLVMBuilderRef builder,
+                         union lp_type type,
+                         LLVMValueRef a0_ptr,
+                         LLVMValueRef dadx_ptr,
+                         LLVMValueRef dady_ptr,
+                         LLVMValueRef x0,
+                         LLVMValueRef y0,
+                         int xstep,
+                         int ystep)
+{
+   struct tgsi_parse_context parse;
+   struct tgsi_full_declaration *decl;
+
+   memset(bld, 0, sizeof *bld);
+
+   lp_build_context_init(&bld->base, builder, type);
+
+   /* For convenience */
+   bld->pos = bld->attribs[0];
+   bld->inputs = (const LLVMValueRef (*)[NUM_CHANNELS]) bld->attribs[1];
+
+   /* Position */
+   bld->num_attribs = 1;
+   bld->mask[0] = TGSI_WRITEMASK_ZW;
+   bld->mode[0] = TGSI_INTERPOLATE_LINEAR;
+
+   /* Inputs */
+   tgsi_parse_init( &parse, tokens );
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         decl = &parse.FullToken.FullDeclaration;
+         if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+            unsigned first, last, mask;
+            unsigned attrib;
+
+            first = decl->DeclarationRange.First;
+            last = decl->DeclarationRange.Last;
+            mask = decl->Declaration.UsageMask;
+
+            for( attrib = first; attrib <= last; ++attrib ) {
+               bld->mask[1 + attrib] = mask;
+               bld->mode[1 + attrib] = decl->Declaration.Interpolate;
+            }
+
+            bld->num_attribs = MAX2(bld->num_attribs, 1 + last + 1);
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+   tgsi_parse_free( &parse );
+
+   coeffs_init(bld, a0_ptr, dadx_ptr, dady_ptr);
+
+   pos_init(bld, x0, y0);
+
+   attribs_init(bld);
+
+   bld->xstep = xstep;
+   bld->ystep = ystep;
+
+   coeffs_update(bld);
+}
+
+
+/**
+ * Advance the position and inputs with the xstep and ystep.
+ */
+void
+lp_build_interp_soa_update(struct lp_build_interp_soa_context *bld)
+{
+   pos_update(bld);
+
+   attribs_update(bld);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_interp.h b/src/gallium/drivers/llvmpipe/lp_bld_interp.h
new file mode 100644
index 0000000000..9194f6233a
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_interp.h
@@ -0,0 +1,99 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Position and shader input interpolation.
+ *
+ * Special attention is given to the interpolation of side by side quads.
+ * Multiplications are made only for the first quad. Interpolation of
+ * inputs for posterior quads are done exclusively with additions, and
+ * perspective divide if necessary.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_BLD_INTERP_H
+#define LP_BLD_INTERP_H
+
+
+#include <llvm-c/Core.h>
+
+#include "tgsi/tgsi_exec.h"
+
+#include "lp_bld_type.h"
+
+
+struct tgsi_token;
+
+
+struct lp_build_interp_soa_context
+{
+   struct lp_build_context base;
+
+   unsigned num_attribs;
+   unsigned mask[1 + PIPE_MAX_SHADER_INPUTS];
+   unsigned mode[1 + PIPE_MAX_SHADER_INPUTS];
+
+   LLVMValueRef a0  [1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+   LLVMValueRef dadx[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+   LLVMValueRef dady[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+
+   int xstep;
+   int ystep;
+
+   /* Attribute values before perspective divide */
+   LLVMValueRef attribs_pre[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+
+   LLVMValueRef attribs[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+
+   /*
+    * Convenience pointers. Callers may access this one.
+    */
+   const LLVMValueRef *pos;
+   const LLVMValueRef (*inputs)[NUM_CHANNELS];
+};
+
+
+void
+lp_build_interp_soa_init(struct lp_build_interp_soa_context *bld,
+                         const struct tgsi_token *tokens,
+                         LLVMBuilderRef builder,
+                         union lp_type type,
+                         LLVMValueRef a0_ptr,
+                         LLVMValueRef dadx_ptr,
+                         LLVMValueRef dady_ptr,
+                         LLVMValueRef x0,
+                         LLVMValueRef y0,
+                         int xstep,
+                         int ystep);
+
+void
+lp_build_interp_soa_update(struct lp_build_interp_soa_context *bld);
+
+
+#endif /* LP_BLD_INTERP_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_intr.c b/src/gallium/drivers/llvmpipe/lp_bld_intr.c
new file mode 100644
index 0000000000..9895749d56
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_intr.c
@@ -0,0 +1,192 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Helpers for emiting intrinsic calls.
+ *
+ * LLVM vanilla IR doesn't represent all basic arithmetic operations we care
+ * about, and it is often necessary to resort target-specific intrinsics for
+ * performance, convenience.
+ *
+ * Ideally we would like to stay away from target specific intrinsics and
+ * move all the instruction selection logic into upstream LLVM where it belongs.
+ *
+ * These functions are also used for calling C functions provided by us from
+ * generated LLVM code.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "util/u_debug.h"
+
+#include "lp_bld_intr.h"
+
+
+LLVMValueRef
+lp_declare_intrinsic(LLVMModuleRef module,
+                     const char *name,
+                     LLVMTypeRef ret_type,
+                     LLVMTypeRef *arg_types,
+                     unsigned num_args)
+{
+   LLVMTypeRef function_type;
+   LLVMValueRef function;
+
+   assert(!LLVMGetNamedFunction(module, name));
+
+   function_type = LLVMFunctionType(ret_type, arg_types, num_args, 0);
+   function = LLVMAddFunction(module, name, function_type);
+
+   LLVMSetFunctionCallConv(function, LLVMCCallConv);
+   LLVMSetLinkage(function, LLVMExternalLinkage);
+
+   assert(LLVMIsDeclaration(function));
+
+   if(name[0] == 'l' &&
+      name[1] == 'l' &&
+      name[2] == 'v' &&
+      name[3] == 'm' &&
+      name[4] == '.')
+      assert(LLVMGetIntrinsicID(function));
+
+   return function;
+}
+
+
+LLVMValueRef
+lp_build_intrinsic(LLVMBuilderRef builder,
+                   const char *name,
+                   LLVMTypeRef ret_type,
+                   LLVMValueRef *args,
+                   unsigned num_args)
+{
+   LLVMModuleRef module = LLVMGetGlobalParent(LLVMGetBasicBlockParent(LLVMGetInsertBlock(builder)));
+   LLVMValueRef function;
+
+   function = LLVMGetNamedFunction(module, name);
+   if(!function) {
+      LLVMTypeRef arg_types[LP_MAX_FUNC_ARGS];
+      unsigned i;
+
+      assert(num_args <= LP_MAX_FUNC_ARGS);
+
+      for(i = 0; i < num_args; ++i) {
+         assert(args[i]);
+         arg_types[i] = LLVMTypeOf(args[i]);
+      }
+
+      function = lp_declare_intrinsic(module, name, ret_type, arg_types, num_args);
+   }
+
+   return LLVMBuildCall(builder, function, args, num_args, "");
+}
+
+
+LLVMValueRef
+lp_build_intrinsic_unary(LLVMBuilderRef builder,
+                         const char *name,
+                         LLVMTypeRef ret_type,
+                         LLVMValueRef a)
+{
+   return lp_build_intrinsic(builder, name, ret_type, &a, 1);
+}
+
+
+LLVMValueRef
+lp_build_intrinsic_binary(LLVMBuilderRef builder,
+                          const char *name,
+                          LLVMTypeRef ret_type,
+                          LLVMValueRef a,
+                          LLVMValueRef b)
+{
+   LLVMValueRef args[2];
+
+   args[0] = a;
+   args[1] = b;
+
+   return lp_build_intrinsic(builder, name, ret_type, args, 2);
+}
+
+
+LLVMValueRef
+lp_build_intrinsic_map(LLVMBuilderRef builder,
+                       const char *name,
+                       LLVMTypeRef ret_type,
+                       LLVMValueRef *args,
+                       unsigned num_args)
+{
+   LLVMTypeRef ret_elem_type = LLVMGetElementType(ret_type);
+   unsigned n = LLVMGetVectorSize(ret_type);
+   unsigned i, j;
+   LLVMValueRef res;
+
+   assert(num_args <= LP_MAX_FUNC_ARGS);
+
+   res = LLVMGetUndef(ret_type);
+   for(i = 0; i < n; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef arg_elems[LP_MAX_FUNC_ARGS];
+      LLVMValueRef res_elem;
+      for(j = 0; j < num_args; ++j)
+         arg_elems[j] = LLVMBuildExtractElement(builder, args[j], index, "");
+      res_elem = lp_build_intrinsic(builder, name, ret_elem_type, arg_elems, num_args);
+      res = LLVMBuildInsertElement(builder, res, res_elem, index, "");
+   }
+
+   return res;
+}
+
+
+LLVMValueRef
+lp_build_intrinsic_map_unary(LLVMBuilderRef builder,
+                             const char *name,
+                             LLVMTypeRef ret_type,
+                             LLVMValueRef a)
+{
+   return lp_build_intrinsic_map(builder, name, ret_type, &a, 1);
+}
+
+
+LLVMValueRef
+lp_build_intrinsic_map_binary(LLVMBuilderRef builder,
+                              const char *name,
+                              LLVMTypeRef ret_type,
+                              LLVMValueRef a,
+                              LLVMValueRef b)
+{
+   LLVMValueRef args[2];
+
+   args[0] = a;
+   args[1] = b;
+
+   return lp_build_intrinsic_map(builder, name, ret_type, args, 2);
+}
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_intr.h b/src/gallium/drivers/llvmpipe/lp_bld_intr.h
new file mode 100644
index 0000000000..f813f27074
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_intr.h
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for calling intrinsics.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_INTR_H
+#define LP_BLD_INTR_H
+
+
+#include <llvm-c/Core.h>  
+
+
+/**
+ * Max number of arguments in an intrinsic.
+ */
+#define LP_MAX_FUNC_ARGS 32
+
+
+LLVMValueRef
+lp_declare_intrinsic(LLVMModuleRef module,
+                     const char *name,
+                     LLVMTypeRef ret_type,
+                     LLVMTypeRef *arg_types,
+                     unsigned num_args);
+
+LLVMValueRef
+lp_build_intrinsic(LLVMBuilderRef builder,
+                   const char *name,
+                   LLVMTypeRef ret_type,
+                   LLVMValueRef *args,
+                   unsigned num_args);
+
+
+LLVMValueRef
+lp_build_intrinsic_unary(LLVMBuilderRef builder,
+                         const char *name,
+                         LLVMTypeRef ret_type,
+                         LLVMValueRef a);
+
+
+LLVMValueRef
+lp_build_intrinsic_binary(LLVMBuilderRef builder,
+                          const char *name,
+                          LLVMTypeRef ret_type,
+                          LLVMValueRef a,
+                          LLVMValueRef b);
+
+
+LLVMValueRef
+lp_build_intrinsic_map(LLVMBuilderRef builder,
+                       const char *name,
+                       LLVMTypeRef ret_type,
+                       LLVMValueRef *args,
+                       unsigned num_args);
+
+
+LLVMValueRef
+lp_build_intrinsic_map_unary(LLVMBuilderRef builder,
+                             const char *name,
+                             LLVMTypeRef ret_type,
+                             LLVMValueRef a);
+
+
+LLVMValueRef
+lp_build_intrinsic_map_binary(LLVMBuilderRef builder,
+                              const char *name,
+                              LLVMTypeRef ret_type,
+                              LLVMValueRef a,
+                              LLVMValueRef b);
+
+
+#endif /* !LP_BLD_INTR_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_logic.c b/src/gallium/drivers/llvmpipe/lp_bld_logic.c
new file mode 100644
index 0000000000..8631efd6c3
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_logic.c
@@ -0,0 +1,397 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for logical operations.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_logic.h"
+
+
+LLVMValueRef
+lp_build_cmp(struct lp_build_context *bld,
+             unsigned func,
+             LLVMValueRef a,
+             LLVMValueRef b)
+{
+   const union lp_type type = bld->type;
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+   LLVMValueRef zeros = LLVMConstNull(int_vec_type);
+   LLVMValueRef ones = LLVMConstAllOnes(int_vec_type);
+   LLVMValueRef cond;
+   LLVMValueRef res;
+   unsigned i;
+
+   if(func == PIPE_FUNC_NEVER)
+      return zeros;
+   if(func == PIPE_FUNC_ALWAYS)
+      return ones;
+
+   /* TODO: optimize the constant case */
+
+   /* XXX: It is not clear if we should use the ordered or unordered operators */
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   if(type.width * type.length == 128) {
+      if(type.floating) {
+         LLVMValueRef args[3];
+         unsigned cc;
+         boolean swap;
+
+         swap = FALSE;
+         switch(func) {
+         case PIPE_FUNC_EQUAL:
+            cc = 0;
+            break;
+         case PIPE_FUNC_NOTEQUAL:
+            cc = 4;
+            break;
+         case PIPE_FUNC_LESS:
+            cc = 1;
+            break;
+         case PIPE_FUNC_LEQUAL:
+            cc = 2;
+            break;
+         case PIPE_FUNC_GREATER:
+            cc = 1;
+            swap = TRUE;
+            break;
+         case PIPE_FUNC_GEQUAL:
+            cc = 2;
+            swap = TRUE;
+            break;
+         default:
+            assert(0);
+            return bld->undef;
+         }
+
+         if(swap) {
+            args[0] = b;
+            args[1] = a;
+         }
+         else {
+            args[0] = a;
+            args[1] = b;
+         }
+
+         args[2] = LLVMConstInt(LLVMInt8Type(), cc, 0);
+         res = lp_build_intrinsic(bld->builder,
+                                  "llvm.x86.sse.cmp.ps",
+                                  vec_type,
+                                  args, 3);
+         res = LLVMBuildBitCast(bld->builder, res, int_vec_type, "");
+         return res;
+      }
+      else {
+         static const struct {
+            unsigned swap:1;
+            unsigned eq:1;
+            unsigned gt:1;
+            unsigned not:1;
+         } table[] = {
+            {0, 0, 0, 1}, /* PIPE_FUNC_NEVER */
+            {1, 0, 1, 0}, /* PIPE_FUNC_LESS */
+            {0, 1, 0, 0}, /* PIPE_FUNC_EQUAL */
+            {0, 0, 1, 1}, /* PIPE_FUNC_LEQUAL */
+            {0, 0, 1, 0}, /* PIPE_FUNC_GREATER */
+            {0, 1, 0, 1}, /* PIPE_FUNC_NOTEQUAL */
+            {1, 0, 1, 1}, /* PIPE_FUNC_GEQUAL */
+            {0, 0, 0, 0}  /* PIPE_FUNC_ALWAYS */
+         };
+         const char *pcmpeq;
+         const char *pcmpgt;
+         LLVMValueRef args[2];
+         LLVMValueRef res;
+
+         switch (type.width) {
+         case 8:
+            pcmpeq = "llvm.x86.sse2.pcmpeq.b";
+            pcmpgt = "llvm.x86.sse2.pcmpgt.b";
+            break;
+         case 16:
+            pcmpeq = "llvm.x86.sse2.pcmpeq.w";
+            pcmpgt = "llvm.x86.sse2.pcmpgt.w";
+            break;
+         case 32:
+            pcmpeq = "llvm.x86.sse2.pcmpeq.d";
+            pcmpgt = "llvm.x86.sse2.pcmpgt.d";
+            break;
+         default:
+            assert(0);
+            return bld->undef;
+         }
+
+         /* There are no signed byte and unsigned word/dword comparison
+          * instructions. So flip the sign bit so that the results match.
+          */
+         if(table[func].gt &&
+            ((type.width == 8 && type.sign) ||
+             (type.width != 8 && !type.sign))) {
+            LLVMValueRef msb = lp_build_int_const_scalar(type, (unsigned long long)1 << (type.width - 1));
+            a = LLVMBuildXor(bld->builder, a, msb, "");
+            b = LLVMBuildXor(bld->builder, b, msb, "");
+         }
+
+         if(table[func].swap) {
+            args[0] = b;
+            args[1] = a;
+         }
+         else {
+            args[0] = a;
+            args[1] = b;
+         }
+
+         if(table[func].eq)
+            res = lp_build_intrinsic(bld->builder, pcmpeq, vec_type, args, 2);
+         else if (table[func].gt)
+            res = lp_build_intrinsic(bld->builder, pcmpgt, vec_type, args, 2);
+         else
+            res = LLVMConstNull(vec_type);
+
+         if(table[func].not)
+            res = LLVMBuildNot(bld->builder, res, "");
+
+         return res;
+      }
+   }
+#endif
+
+   if(type.floating) {
+      LLVMRealPredicate op;
+      switch(func) {
+      case PIPE_FUNC_NEVER:
+         op = LLVMRealPredicateFalse;
+         break;
+      case PIPE_FUNC_ALWAYS:
+         op = LLVMRealPredicateTrue;
+         break;
+      case PIPE_FUNC_EQUAL:
+         op = LLVMRealUEQ;
+         break;
+      case PIPE_FUNC_NOTEQUAL:
+         op = LLVMRealUNE;
+         break;
+      case PIPE_FUNC_LESS:
+         op = LLVMRealULT;
+         break;
+      case PIPE_FUNC_LEQUAL:
+         op = LLVMRealULE;
+         break;
+      case PIPE_FUNC_GREATER:
+         op = LLVMRealUGT;
+         break;
+      case PIPE_FUNC_GEQUAL:
+         op = LLVMRealUGE;
+         break;
+      default:
+         assert(0);
+         return bld->undef;
+      }
+
+#if 0
+      /* XXX: Although valid IR, no LLVM target currently support this */
+      cond = LLVMBuildFCmp(bld->builder, op, a, b, "");
+      res = LLVMBuildSelect(bld->builder, cond, ones, zeros, "");
+#else
+      debug_printf("%s: warning: using slow element-wise vector comparison\n",
+                   __FUNCTION__);
+      res = LLVMGetUndef(int_vec_type);
+      for(i = 0; i < type.length; ++i) {
+         LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+         cond = LLVMBuildFCmp(bld->builder, op,
+                              LLVMBuildExtractElement(bld->builder, a, index, ""),
+                              LLVMBuildExtractElement(bld->builder, b, index, ""),
+                              "");
+         cond = LLVMBuildSelect(bld->builder, cond,
+                                LLVMConstExtractElement(ones, index),
+                                LLVMConstExtractElement(zeros, index),
+                                "");
+         res = LLVMBuildInsertElement(bld->builder, res, cond, index, "");
+      }
+#endif
+   }
+   else {
+      LLVMIntPredicate op;
+      switch(func) {
+      case PIPE_FUNC_EQUAL:
+         op = LLVMIntEQ;
+         break;
+      case PIPE_FUNC_NOTEQUAL:
+         op = LLVMIntNE;
+         break;
+      case PIPE_FUNC_LESS:
+         op = type.sign ? LLVMIntSLT : LLVMIntULT;
+         break;
+      case PIPE_FUNC_LEQUAL:
+         op = type.sign ? LLVMIntSLE : LLVMIntULE;
+         break;
+      case PIPE_FUNC_GREATER:
+         op = type.sign ? LLVMIntSGT : LLVMIntUGT;
+         break;
+      case PIPE_FUNC_GEQUAL:
+         op = type.sign ? LLVMIntSGE : LLVMIntUGE;
+         break;
+      default:
+         assert(0);
+         return bld->undef;
+      }
+
+#if 0
+      /* XXX: Although valid IR, no LLVM target currently support this */
+      cond = LLVMBuildICmp(bld->builder, op, a, b, "");
+      res = LLVMBuildSelect(bld->builder, cond, ones, zeros, "");
+#else
+      debug_printf("%s: warning: using slow element-wise vector comparison\n",
+                   __FUNCTION__);
+      res = LLVMGetUndef(int_vec_type);
+      for(i = 0; i < type.length; ++i) {
+         LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+         cond = LLVMBuildICmp(bld->builder, op,
+                              LLVMBuildExtractElement(bld->builder, a, index, ""),
+                              LLVMBuildExtractElement(bld->builder, b, index, ""),
+                              "");
+         cond = LLVMBuildSelect(bld->builder, cond,
+                                LLVMConstExtractElement(ones, index),
+                                LLVMConstExtractElement(zeros, index),
+                                "");
+         res = LLVMBuildInsertElement(bld->builder, res, cond, index, "");
+      }
+#endif
+   }
+
+   return res;
+}
+
+
+LLVMValueRef
+lp_build_select(struct lp_build_context *bld,
+                LLVMValueRef mask,
+                LLVMValueRef a,
+                LLVMValueRef b)
+{
+   union lp_type type = bld->type;
+   LLVMValueRef res;
+
+   if(a == b)
+      return a;
+
+   if(type.floating) {
+      LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+      a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
+      b = LLVMBuildBitCast(bld->builder, b, int_vec_type, "");
+   }
+
+   /* TODO: On SSE4 we could do this with a single instruction -- PBLENDVB */
+
+   a = LLVMBuildAnd(bld->builder, a, mask, "");
+
+   /* This often gets translated to PANDN, but sometimes the NOT is
+    * pre-computed and stored in another constant. The best strategy depends
+    * on available registers, so it is not a big deal -- hopefully LLVM does
+    * the right decision attending the rest of the program.
+    */
+   b = LLVMBuildAnd(bld->builder, b, LLVMBuildNot(bld->builder, mask, ""), "");
+
+   res = LLVMBuildOr(bld->builder, a, b, "");
+
+   if(type.floating) {
+      LLVMTypeRef vec_type = lp_build_vec_type(type);
+      res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
+   }
+
+   return res;
+}
+
+
+LLVMValueRef
+lp_build_select_aos(struct lp_build_context *bld,
+                    LLVMValueRef a,
+                    LLVMValueRef b,
+                    boolean cond[4])
+{
+   const union lp_type type = bld->type;
+   const unsigned n = type.length;
+   unsigned i, j;
+
+   if(a == b)
+      return a;
+   if(cond[0] && cond[1] && cond[2] && cond[3])
+      return a;
+   if(!cond[0] && !cond[1] && !cond[2] && !cond[3])
+      return b;
+   if(a == bld->undef || b == bld->undef)
+      return bld->undef;
+
+   /*
+    * There are three major ways of accomplishing this:
+    * - with a shuffle,
+    * - with a select,
+    * - or with a bit mask.
+    *
+    * Select isn't supported for vector types yet.
+    * The flip between these is empirical and might need to be.
+    */
+   if (n <= 4) {
+      /*
+       * Shuffle.
+       */
+      LLVMTypeRef elem_type = LLVMInt32Type();
+      LLVMValueRef shuffles[LP_MAX_VECTOR_LENGTH];
+
+      for(j = 0; j < n; j += 4)
+         for(i = 0; i < 4; ++i)
+            shuffles[j + i] = LLVMConstInt(elem_type, (cond[i] ? 0 : n) + j + i, 0);
+
+      return LLVMBuildShuffleVector(bld->builder, a, b, LLVMConstVector(shuffles, n), "");
+   }
+#if 0
+   else if(0) {
+      /* FIXME: Unfortunately select of vectors do not work */
+      /* Use a select */
+      LLVMTypeRef elem_type = LLVMInt1Type();
+      LLVMValueRef cond[LP_MAX_VECTOR_LENGTH];
+
+      for(j = 0; j < n; j += 4)
+         for(i = 0; i < 4; ++i)
+            cond[j + i] = LLVMConstInt(elem_type, cond[i] ? 1 : 0, 0);
+
+      return LLVMBuildSelect(bld->builder, LLVMConstVector(cond, n), a, b, "");
+   }
+#endif
+   else {
+      LLVMValueRef mask = lp_build_const_mask_aos(type, cond);
+      return lp_build_select(bld, mask, a, b);
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_logic.h b/src/gallium/drivers/llvmpipe/lp_bld_logic.h
new file mode 100644
index 0000000000..29b9e1c45b
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_logic.h
@@ -0,0 +1,72 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for logical operations.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_LOGIC_H
+#define LP_BLD_LOGIC_H
+
+
+#include <llvm-c/Core.h>  
+
+#include "pipe/p_defines.h" /* For PIPE_FUNC_xxx */
+
+
+union lp_type type;
+struct lp_build_context;
+
+
+/**
+ * @param func is one of PIPE_FUNC_xxx
+ */
+LLVMValueRef
+lp_build_cmp(struct lp_build_context *bld,
+             unsigned func,
+             LLVMValueRef a,
+             LLVMValueRef b);
+
+
+LLVMValueRef
+lp_build_select(struct lp_build_context *bld,
+                LLVMValueRef mask,
+                LLVMValueRef a,
+                LLVMValueRef b);
+
+LLVMValueRef
+lp_build_select_aos(struct lp_build_context *bld,
+                    LLVMValueRef a,
+                    LLVMValueRef b,
+                    boolean cond[4]);
+
+
+#endif /* !LP_BLD_LOGIC_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_struct.c b/src/gallium/drivers/llvmpipe/lp_bld_struct.c
new file mode 100644
index 0000000000..14d2b10df9
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_struct.c
@@ -0,0 +1,59 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Helper functions for manipulation structures.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "util/u_debug.h"
+#include "util/u_memory.h"
+
+#include "lp_bld_debug.h"
+#include "lp_bld_struct.h"
+
+
+LLVMValueRef
+lp_build_struct_get(LLVMBuilderRef builder,
+                    LLVMValueRef ptr,
+                    unsigned member,
+                    const char *name)
+{
+   LLVMValueRef indices[2];
+   LLVMValueRef member_ptr;
+   LLVMValueRef res;
+   indices[0] = LLVMConstInt(LLVMInt32Type(), 0, 0);
+   indices[1] = LLVMConstInt(LLVMInt32Type(), member, 0);
+   member_ptr = LLVMBuildGEP(builder, ptr, indices, Elements(indices), "");
+   res = LLVMBuildLoad(builder, member_ptr, "");
+   lp_build_name(res, "%s.%s", LLVMGetValueName(ptr), name);
+   return res;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_struct.h b/src/gallium/drivers/llvmpipe/lp_bld_struct.h
new file mode 100644
index 0000000000..cbefdc9f81
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_struct.h
@@ -0,0 +1,63 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for type conversions.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_STRUCT_H
+#define LP_BLD_STRUCT_H
+
+
+#include <llvm-c/Core.h>  
+#include <llvm-c/Target.h>
+
+#include "util/u_debug.h"
+#include "util/u_memory.h"
+
+
+#define LP_CHECK_STRUCT_SIZE(_ctype, _ltarget, _ltype) \
+      assert(LLVMABISizeOfType(_ltarget, _ltype) == \
+             sizeof(_ctype))
+
+#define LP_CHECK_MEMBER_OFFSET(_ctype, _cmember, _ltarget, _ltype, _lindex) \
+      assert(LLVMOffsetOfElement(_ltarget, _ltype, _lindex) == \
+             offsetof(_ctype, _cmember))
+
+
+LLVMValueRef
+lp_build_struct_get(LLVMBuilderRef builder,
+                    LLVMValueRef ptr,
+                    unsigned member,
+                    const char *name);
+
+
+#endif /* !LP_BLD_STRUCT_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_swizzle.c b/src/gallium/drivers/llvmpipe/lp_bld_swizzle.c
new file mode 100644
index 0000000000..ac7eed9379
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_swizzle.c
@@ -0,0 +1,238 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for swizzling/shuffling.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "util/u_debug.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_swizzle.h"
+
+
+LLVMValueRef
+lp_build_broadcast(LLVMBuilderRef builder,
+                   LLVMTypeRef vec_type,
+                   LLVMValueRef scalar)
+{
+   const unsigned n = LLVMGetVectorSize(vec_type);
+   LLVMValueRef res;
+   unsigned i;
+
+   res = LLVMGetUndef(vec_type);
+   for(i = 0; i < n; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      res = LLVMBuildInsertElement(builder, res, scalar, index, "");
+   }
+
+   return res;
+}
+
+
+LLVMValueRef
+lp_build_broadcast_scalar(struct lp_build_context *bld,
+                          LLVMValueRef scalar)
+{
+   const union lp_type type = bld->type;
+   LLVMValueRef res;
+   unsigned i;
+
+   res = bld->undef;
+   for(i = 0; i < type.length; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      res = LLVMBuildInsertElement(bld->builder, res, scalar, index, "");
+   }
+
+   return res;
+}
+
+
+LLVMValueRef
+lp_build_broadcast_aos(struct lp_build_context *bld,
+                       LLVMValueRef a,
+                       unsigned channel)
+{
+   const union lp_type type = bld->type;
+   const unsigned n = type.length;
+   unsigned i, j;
+
+   if(a == bld->undef || a == bld->zero || a == bld->one)
+      return a;
+
+   /* XXX: SSE3 has PSHUFB which should be better than bitmasks, but forcing
+    * using shuffles here actually causes worst results. More investigation is
+    * needed. */
+   if (n <= 4) {
+      /*
+       * Shuffle.
+       */
+      LLVMTypeRef elem_type = LLVMInt32Type();
+      LLVMValueRef shuffles[LP_MAX_VECTOR_LENGTH];
+
+      for(j = 0; j < n; j += 4)
+         for(i = 0; i < 4; ++i)
+            shuffles[j + i] = LLVMConstInt(elem_type, j + channel, 0);
+
+      return LLVMBuildShuffleVector(bld->builder, a, bld->undef, LLVMConstVector(shuffles, n), "");
+   }
+   else {
+      /*
+       * Bit mask and recursive shifts
+       *
+       *   XYZW XYZW .... XYZW  <= input
+       *   0Y00 0Y00 .... 0Y00
+       *   YY00 YY00 .... YY00
+       *   YYYY YYYY .... YYYY  <= output
+       */
+      union lp_type type4 = type;
+      const char shifts[4][2] = {
+         { 1,  2},
+         {-1,  2},
+         { 1, -2},
+         {-1, -2}
+      };
+      boolean cond[4];
+      unsigned i;
+
+      memset(cond, 0, sizeof cond);
+      cond[channel] = 1;
+
+      a = LLVMBuildAnd(bld->builder, a, lp_build_const_mask_aos(type, cond), "");
+
+      type4.width *= 4;
+      type4.length /= 4;
+
+      a = LLVMBuildBitCast(bld->builder, a, lp_build_vec_type(type4), "");
+
+      for(i = 0; i < 2; ++i) {
+         LLVMValueRef tmp = NULL;
+         int shift = shifts[channel][i];
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+         shift = -shift;
+#endif
+
+         if(shift > 0)
+            tmp = LLVMBuildLShr(bld->builder, a, lp_build_int_const_scalar(type4, shift*type.width), "");
+         if(shift < 0)
+            tmp = LLVMBuildShl(bld->builder, a, lp_build_int_const_scalar(type4, -shift*type.width), "");
+
+         assert(tmp);
+         if(tmp)
+            a = LLVMBuildOr(bld->builder, a, tmp, "");
+      }
+
+      return LLVMBuildBitCast(bld->builder, a, lp_build_vec_type(type), "");
+   }
+}
+
+
+LLVMValueRef
+lp_build_swizzle1_aos(struct lp_build_context *bld,
+                      LLVMValueRef a,
+                      unsigned char swizzle[4])
+{
+   const unsigned n = bld->type.length;
+   unsigned i, j;
+
+   if(a == bld->undef || a == bld->zero || a == bld->one)
+      return a;
+
+   if(swizzle[0] == swizzle[1] && swizzle[1] == swizzle[2] && swizzle[2] == swizzle[3])
+      return lp_build_broadcast_aos(bld, a, swizzle[0]);
+
+   {
+      /*
+       * Shuffle.
+       */
+      LLVMTypeRef elem_type = LLVMInt32Type();
+      LLVMValueRef shuffles[LP_MAX_VECTOR_LENGTH];
+
+      for(j = 0; j < n; j += 4)
+         for(i = 0; i < 4; ++i)
+            shuffles[j + i] = LLVMConstInt(elem_type, j + swizzle[i], 0);
+
+      return LLVMBuildShuffleVector(bld->builder, a, bld->undef, LLVMConstVector(shuffles, n), "");
+   }
+}
+
+
+LLVMValueRef
+lp_build_swizzle2_aos(struct lp_build_context *bld,
+                      LLVMValueRef a,
+                      LLVMValueRef b,
+                      unsigned char swizzle[4])
+{
+   const unsigned n = bld->type.length;
+   unsigned i, j;
+
+   if(swizzle[0] < 4 && swizzle[1] < 4 && swizzle[2] < 4 && swizzle[3] < 4)
+      return lp_build_swizzle1_aos(bld, a, swizzle);
+
+   if(a == b) {
+      swizzle[0] %= 4;
+      swizzle[1] %= 4;
+      swizzle[2] %= 4;
+      swizzle[3] %= 4;
+      return lp_build_swizzle1_aos(bld, a, swizzle);
+   }
+
+   if(swizzle[0] % 4 == 0 &&
+      swizzle[1] % 4 == 1 &&
+      swizzle[2] % 4 == 2 &&
+      swizzle[3] % 4 == 3) {
+      boolean cond[4];
+      cond[0] = swizzle[0] / 4;
+      cond[1] = swizzle[1] / 4;
+      cond[2] = swizzle[2] / 4;
+      cond[3] = swizzle[3] / 4;
+      return lp_build_select_aos(bld, a, b, cond);
+   }
+
+   {
+      /*
+       * Shuffle.
+       */
+      LLVMTypeRef elem_type = LLVMInt32Type();
+      LLVMValueRef shuffles[LP_MAX_VECTOR_LENGTH];
+
+      for(j = 0; j < n; j += 4)
+         for(i = 0; i < 4; ++i)
+            shuffles[j + i] = LLVMConstInt(elem_type, j + (swizzle[i] % 4) + (swizzle[i] / 4 * n), 0);
+
+      return LLVMBuildShuffleVector(bld->builder, a, b, LLVMConstVector(shuffles, n), "");
+   }
+}
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_swizzle.h b/src/gallium/drivers/llvmpipe/lp_bld_swizzle.h
new file mode 100644
index 0000000000..d7dd6a8a60
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_swizzle.h
@@ -0,0 +1,91 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Helper functions for swizzling/shuffling.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_SWIZZLE_H
+#define LP_BLD_SWIZZLE_H
+
+
+#include <llvm-c/Core.h>  
+
+
+union lp_type type;
+struct lp_build_context;
+
+
+LLVMValueRef
+lp_build_broadcast(LLVMBuilderRef builder,
+                   LLVMTypeRef vec_type,
+                   LLVMValueRef scalar);
+
+
+LLVMValueRef
+lp_build_broadcast_scalar(struct lp_build_context *bld,
+                          LLVMValueRef scalar);
+
+
+/**
+ * Broadcast one channel of a vector composed of arrays of XYZW structures into
+ * all four channel.
+ */
+LLVMValueRef
+lp_build_broadcast_aos(struct lp_build_context *bld,
+                       LLVMValueRef a,
+                       unsigned channel);
+
+
+/**
+ * Swizzle a vector consisting of an array of XYZW structs.
+ *
+ * @param swizzle is the in [0,4[ range.
+ */
+LLVMValueRef
+lp_build_swizzle1_aos(struct lp_build_context *bld,
+                      LLVMValueRef a,
+                      unsigned char swizzle[4]);
+
+
+/**
+ * Swizzle two vector consisting of an array of XYZW structs.
+ *
+ * @param swizzle is the in [0,8[ range. Values in [4,8[ range refer to b.
+ */
+LLVMValueRef
+lp_build_swizzle2_aos(struct lp_build_context *bld,
+                      LLVMValueRef a,
+                      LLVMValueRef b,
+                      unsigned char swizzle[4]);
+
+
+#endif /* !LP_BLD_SWIZZLE_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi.h b/src/gallium/drivers/llvmpipe/lp_bld_tgsi.h
new file mode 100644
index 0000000000..912db24aec
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi.h
@@ -0,0 +1,69 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * TGSI to LLVM IR translation.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_BLD_TGSI_H
+#define LP_BLD_TGSI_H
+
+#include <llvm-c/Core.h>
+
+
+struct tgsi_token;
+union lp_type;
+struct lp_build_context;
+struct lp_build_mask_context;
+
+
+typedef void
+(*lp_emit_fetch_texel_soa_callback)( LLVMBuilderRef builder,
+                                void *context,
+                                unsigned unit,
+                                unsigned num_coords,
+                                const LLVMValueRef *coords,
+                                LLVMValueRef lodbias,
+                                LLVMValueRef *texel);
+
+void
+lp_build_tgsi_soa(LLVMBuilderRef builder,
+                  const struct tgsi_token *tokens,
+                  union lp_type type,
+                  struct lp_build_mask_context *mask,
+                  LLVMValueRef consts_ptr,
+                  const LLVMValueRef *pos,
+                  const LLVMValueRef (*inputs)[4],
+                  LLVMValueRef (*outputs)[4],
+                  lp_emit_fetch_texel_soa_callback emit_fetch_texel,
+                  void *emit_fetch_texel_context);
+
+
+#endif /* LP_BLD_TGSI_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
new file mode 100644
index 0000000000..d4d18febec
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
@@ -0,0 +1,1346 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * TGSI to LLVM IR translation -- SoA.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ *
+ * Based on tgsi_sse2.c code written by Michal Krol, Keith Whitwell,
+ * Brian Paul, and others.
+ */
+
+#include "pipe/p_config.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_debug.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_info.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_exec.h"
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_swizzle.h"
+#include "lp_bld_flow.h"
+#include "lp_bld_tgsi.h"
+#include "lp_bld_debug.h"
+
+
+#define LP_MAX_TEMPS 256
+#define LP_MAX_IMMEDIATES 256
+
+
+#define FOR_EACH_CHANNEL( CHAN )\
+   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
+
+#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   ((INST)->FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   if (IS_DST0_CHANNEL_ENABLED( INST, CHAN ))
+
+#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\
+   FOR_EACH_CHANNEL( CHAN )\
+      IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )
+
+#define CHAN_X 0
+#define CHAN_Y 1
+#define CHAN_Z 2
+#define CHAN_W 3
+
+
+struct lp_build_tgsi_soa_context
+{
+   struct lp_build_context base;
+
+   LLVMValueRef consts_ptr;
+   const LLVMValueRef *pos;
+   const LLVMValueRef (*inputs)[NUM_CHANNELS];
+   LLVMValueRef (*outputs)[NUM_CHANNELS];
+
+   lp_emit_fetch_texel_soa_callback emit_fetch_texel;
+   void *emit_fetch_texel_context;
+
+   LLVMValueRef immediates[LP_MAX_IMMEDIATES][NUM_CHANNELS];
+   LLVMValueRef temps[LP_MAX_TEMPS][NUM_CHANNELS];
+
+   struct lp_build_mask_context *mask;
+};
+
+
+/**
+ * Register fetch.
+ */
+static LLVMValueRef
+emit_fetch(
+   struct lp_build_tgsi_soa_context *bld,
+   const struct tgsi_full_instruction *inst,
+   unsigned index,
+   const unsigned chan_index )
+{
+   const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[index];
+   unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+   LLVMValueRef res;
+
+   switch (swizzle) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+
+      switch (reg->SrcRegister.File) {
+      case TGSI_FILE_CONSTANT: {
+         LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), reg->SrcRegister.Index*4 + swizzle, 0);
+         LLVMValueRef scalar_ptr = LLVMBuildGEP(bld->base.builder, bld->consts_ptr, &index, 1, "");
+         LLVMValueRef scalar = LLVMBuildLoad(bld->base.builder, scalar_ptr, "");
+         res = lp_build_broadcast_scalar(&bld->base, scalar);
+         break;
+      }
+
+      case TGSI_FILE_IMMEDIATE:
+         res = bld->immediates[reg->SrcRegister.Index][swizzle];
+         assert(res);
+         break;
+
+      case TGSI_FILE_INPUT:
+         res = bld->inputs[reg->SrcRegister.Index][swizzle];
+         assert(res);
+         break;
+
+      case TGSI_FILE_TEMPORARY:
+         res = bld->temps[reg->SrcRegister.Index][swizzle];
+         if(!res)
+            return bld->base.undef;
+         break;
+
+      default:
+         assert( 0 );
+         return bld->base.undef;
+      }
+      break;
+
+   case TGSI_EXTSWIZZLE_ZERO:
+      res = bld->base.zero;
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      res = bld->base.one;
+      break;
+
+   default:
+      assert( 0 );
+      return bld->base.undef;
+   }
+
+   switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) {
+   case TGSI_UTIL_SIGN_CLEAR:
+      res = lp_build_abs( &bld->base, res );
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      res = lp_build_abs( &bld->base, res );
+      res = LLVMBuildNeg( bld->base.builder, res, "" );
+      break;
+
+   case TGSI_UTIL_SIGN_TOGGLE:
+      res = LLVMBuildNeg( bld->base.builder, res, "" );
+      break;
+
+   case TGSI_UTIL_SIGN_KEEP:
+      break;
+   }
+
+   return res;
+}
+
+
+/**
+ * Register store.
+ */
+static void
+emit_store(
+   struct lp_build_tgsi_soa_context *bld,
+   const struct tgsi_full_instruction *inst,
+   unsigned index,
+   unsigned chan_index,
+   LLVMValueRef value)
+{
+   const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[index];
+
+   switch( inst->Instruction.Saturate ) {
+   case TGSI_SAT_NONE:
+      break;
+
+   case TGSI_SAT_ZERO_ONE:
+      value = lp_build_max(&bld->base, value, bld->base.zero);
+      value = lp_build_min(&bld->base, value, bld->base.one);
+      break;
+
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      value = lp_build_max(&bld->base, value, lp_build_const_scalar(bld->base.type, -1.0));
+      value = lp_build_min(&bld->base, value, bld->base.one);
+      break;
+
+   default:
+      assert(0);
+   }
+
+   switch( reg->DstRegister.File ) {
+   case TGSI_FILE_OUTPUT:
+      bld->outputs[reg->DstRegister.Index][chan_index] = value;
+      break;
+
+   case TGSI_FILE_TEMPORARY:
+      bld->temps[reg->DstRegister.Index][chan_index] = value;
+      break;
+
+   case TGSI_FILE_ADDRESS:
+      /* FIXME */
+      assert(0);
+      break;
+
+   default:
+      assert( 0 );
+   }
+}
+
+
+/**
+ * High-level instruction translators.
+ */
+
+static void
+emit_tex( struct lp_build_tgsi_soa_context *bld,
+          const struct tgsi_full_instruction *inst,
+          boolean apply_lodbias,
+          boolean projected)
+{
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   LLVMValueRef lodbias;
+   LLVMValueRef oow;
+   LLVMValueRef coords[3];
+   LLVMValueRef texel[4];
+   unsigned num_coords;
+   unsigned i;
+
+   switch (inst->InstructionExtTexture.Texture) {
+   case TGSI_TEXTURE_1D:
+      num_coords = 1;
+      break;
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+      num_coords = 2;
+      break;
+   case TGSI_TEXTURE_SHADOW1D:
+   case TGSI_TEXTURE_SHADOW2D:
+   case TGSI_TEXTURE_SHADOWRECT:
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+      num_coords = 3;
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+   if(apply_lodbias)
+      lodbias = emit_fetch( bld, inst, 0, 3 );
+   else
+      lodbias = bld->base.zero;
+
+   if (projected) {
+      oow = emit_fetch( bld, inst, 0, 3 );
+      oow = lp_build_rcp(&bld->base, oow);
+   }
+
+   for (i = 0; i < num_coords; i++) {
+      coords[i] = emit_fetch( bld, inst, 0, i );
+      if (projected)
+         coords[i] = lp_build_mul(&bld->base, coords[i], oow);
+   }
+
+   bld->emit_fetch_texel(bld->base.builder, bld->emit_fetch_texel_context,
+                         unit, num_coords, coords, lodbias, texel);
+
+   FOR_EACH_DST0_ENABLED_CHANNEL( inst, i ) {
+      emit_store( bld, inst, 0, i, texel[i] );
+   }
+}
+
+
+static void
+emit_kil(
+   struct lp_build_tgsi_soa_context *bld,
+   const struct tgsi_full_instruction *inst )
+{
+   const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[0];
+   LLVMValueRef terms[NUM_CHANNELS];
+   LLVMValueRef mask;
+   unsigned chan_index;
+
+   memset(&terms, 0, sizeof terms);
+
+   FOR_EACH_CHANNEL( chan_index ) {
+      unsigned swizzle;
+
+      /* Unswizzle channel */
+      swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+
+      /* Note that we test if the value is less than zero, so 1.0 and 0.0 need
+       * not to be tested. */
+      if(swizzle == TGSI_EXTSWIZZLE_ZERO || swizzle == TGSI_EXTSWIZZLE_ONE)
+         continue;
+
+      /* Check if the component has not been already tested. */
+      assert(swizzle < NUM_CHANNELS);
+      if( !terms[swizzle] )
+         /* TODO: change the comparison operator instead of setting the sign */
+         terms[swizzle] =  emit_fetch(bld, inst, 0, chan_index );
+   }
+
+   mask = NULL;
+   FOR_EACH_CHANNEL( chan_index ) {
+      if(terms[chan_index]) {
+         LLVMValueRef chan_mask;
+
+         chan_mask = lp_build_cmp(&bld->base, PIPE_FUNC_GEQUAL, terms[chan_index], bld->base.zero);
+
+         if(mask)
+            mask = LLVMBuildAnd(bld->base.builder, mask, chan_mask, "");
+         else
+            mask = chan_mask;
+      }
+   }
+
+   if(mask)
+      lp_build_mask_update(bld->mask, mask);
+}
+
+
+static void
+emit_kilp(
+   struct lp_build_tgsi_soa_context *bld )
+{
+   /* XXX todo / fix me */
+}
+
+
+/**
+ * Check if inst src/dest regs use indirect addressing into temporary
+ * register file.
+ */
+static boolean
+indirect_temp_reference(const struct tgsi_full_instruction *inst)
+{
+   uint i;
+   for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[i];
+      if (reg->SrcRegister.File == TGSI_FILE_TEMPORARY &&
+          reg->SrcRegister.Indirect)
+         return TRUE;
+   }
+   for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[i];
+      if (reg->DstRegister.File == TGSI_FILE_TEMPORARY &&
+          reg->DstRegister.Indirect)
+         return TRUE;
+   }
+   return FALSE;
+}
+
+
+static int
+emit_instruction(
+   struct lp_build_tgsi_soa_context *bld,
+   struct tgsi_full_instruction *inst )
+{
+   unsigned chan_index;
+   LLVMValueRef src0, src1, src2;
+   LLVMValueRef tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+   LLVMValueRef dst0;
+
+   /* we can't handle indirect addressing into temp register file yet */
+   if (indirect_temp_reference(inst))
+      return FALSE;
+
+   switch (inst->Instruction.Opcode) {
+#if 0
+   case TGSI_OPCODE_ARL:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_flr(bld, 0, 0);
+         emit_f2it( bld, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+#endif
+
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_LIT:
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ) {
+         emit_store( bld, inst, 0, CHAN_X, bld->base.one);
+      }
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ) {
+         src0 = emit_fetch( bld, inst, 0, CHAN_X );
+         dst0 = lp_build_max( &bld->base, src0, bld->base.zero);
+         emit_store( bld, inst, 0, CHAN_Y, dst0);
+      }
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) ) {
+         /* XMM[1] = SrcReg[0].yyyy */
+         tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );
+         /* XMM[1] = max(XMM[1], 0) */
+         tmp1 = lp_build_max( &bld->base, tmp1, bld->base.zero);
+         /* XMM[2] = SrcReg[0].wwww */
+         tmp2 = emit_fetch( bld, inst, 0, CHAN_W );
+         tmp1 = lp_build_pow( &bld->base, tmp1, tmp2);
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+         tmp2 = lp_build_cmp(&bld->base, PIPE_FUNC_GREATER, tmp0, bld->base.zero);
+         dst0 = lp_build_select(&bld->base, tmp2, tmp1, bld->base.zero);
+         emit_store( bld, inst, 0, CHAN_Z, dst0);
+      }
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_W ) ) {
+         emit_store( bld, inst, 0, CHAN_W, bld->base.one);
+      }
+      break;
+
+   case TGSI_OPCODE_RCP:
+   /* TGSI_OPCODE_RECIP */
+      src0 = emit_fetch( bld, inst, 0, CHAN_X );
+      dst0 = lp_build_rcp(&bld->base, src0);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, dst0 );
+      }
+      break;
+
+   case TGSI_OPCODE_RSQ:
+   /* TGSI_OPCODE_RECIPSQRT */
+      src0 = emit_fetch( bld, inst, 0, CHAN_X );
+      src0 = lp_build_abs(&bld->base, src0);
+      dst0 = lp_build_rsqrt(&bld->base, src0);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, dst0 );
+      }
+      break;
+
+   case TGSI_OPCODE_EXP:
+      if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z )) {
+         LLVMValueRef *p_exp2_int_part = NULL;
+         LLVMValueRef *p_frac_part = NULL;
+         LLVMValueRef *p_exp2 = NULL;
+
+         src0 = emit_fetch( bld, inst, 0, CHAN_X );
+
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ))
+            p_exp2_int_part = &tmp0;
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ))
+            p_frac_part = &tmp1;
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ))
+            p_exp2 = &tmp2;
+
+         lp_build_exp2_approx(&bld->base, src0, p_exp2_int_part, p_frac_part, p_exp2);
+
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ))
+            emit_store( bld, inst, 0, CHAN_X, tmp0);
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ))
+            emit_store( bld, inst, 0, CHAN_Y, tmp1);
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ))
+            emit_store( bld, inst, 0, CHAN_Z, tmp2);
+      }
+      /* dst.w = 1.0 */
+      if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_W )) {
+         tmp0 = bld->base.one;
+         emit_store( bld, inst, 0, CHAN_W, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_LOG:
+      if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z )) {
+         LLVMValueRef *p_floor_log2;
+         LLVMValueRef *p_exp;
+         LLVMValueRef *p_log2;
+
+         src0 = emit_fetch( bld, inst, 0, CHAN_X );
+         src0 = lp_build_abs( &bld->base, src0 );
+
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ))
+            p_floor_log2 = &tmp0;
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ))
+            p_exp = &tmp1;
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ))
+            p_log2 = &tmp2;
+
+         lp_build_log2_approx(&bld->base, src0, p_exp, p_floor_log2, p_log2);
+
+         /* dst.x = floor(lg2(abs(src.x))) */
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ))
+            emit_store( bld, inst, 0, CHAN_X, tmp0);
+         /* dst.y = abs(src)/ex2(floor(lg2(abs(src.x)))) */
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y )) {
+            tmp1 = lp_build_div( &bld->base, src0, tmp1);
+            emit_store( bld, inst, 0, CHAN_Y, tmp1);
+         }
+         /* dst.z = lg2(abs(src.x)) */
+         if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ))
+            emit_store( bld, inst, 0, CHAN_Z, tmp2);
+      }
+      /* dst.w = 1.0 */
+      if (IS_DST0_CHANNEL_ENABLED( inst, CHAN_W )) {
+         tmp0 = bld->base.one;
+         emit_store( bld, inst, 0, CHAN_W, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_MUL:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         dst0 = lp_build_mul(&bld->base, src0, src1);
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_ADD:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         dst0 = lp_build_add(&bld->base, src0, src1);
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_DP3:
+   /* TGSI_OPCODE_DOT3 */
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_X );
+      tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Y );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Z );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Z );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_DP4:
+   /* TGSI_OPCODE_DOT4 */
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_X );
+      tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Y );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Z );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Z );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_W );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_W );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_DST:
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) {
+         tmp0 = bld->base.one;
+         emit_store( bld, inst, 0, CHAN_X, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) {
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_Y );
+         tmp1 = emit_fetch( bld, inst, 1, CHAN_Y );
+         tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);
+         emit_store( bld, inst, 0, CHAN_Y, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) {
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_Z );
+         emit_store( bld, inst, 0, CHAN_Z, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_W ) {
+         tmp0 = emit_fetch( bld, inst, 1, CHAN_W );
+         emit_store( bld, inst, 0, CHAN_W, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_MIN:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         dst0 = lp_build_min( &bld->base, src0, src1 );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_MAX:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         dst0 = lp_build_max( &bld->base, src0, src1 );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SLT:
+   /* TGSI_OPCODE_SETLT */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_LESS, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SGE:
+   /* TGSI_OPCODE_SETGE */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_GEQUAL, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_MAD:
+   /* TGSI_OPCODE_MADD */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         tmp1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp2 = emit_fetch( bld, inst, 2, chan_index );
+         tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);
+         tmp0 = lp_build_add( &bld->base, tmp0, tmp2);
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_SUB:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         tmp1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_sub( &bld->base, tmp0, tmp1);
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_LRP:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         src2 = emit_fetch( bld, inst, 2, chan_index );
+         tmp0 = lp_build_sub( &bld->base, src1, src2 );
+         tmp0 = lp_build_mul( &bld->base, src0, tmp0 );
+         dst0 = lp_build_add( &bld->base, tmp0, src2 );
+         emit_store( bld, inst, 0, chan_index, dst0 );
+      }
+      break;
+
+   case TGSI_OPCODE_CND:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DP2A:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );  /* xmm0 = src[0].x */
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_X );  /* xmm1 = src[1].x */
+      tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);              /* xmm0 = xmm0 * xmm1 */
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );  /* xmm1 = src[0].y */
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Y );  /* xmm2 = src[1].y */
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);              /* xmm1 = xmm1 * xmm2 */
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);              /* xmm0 = xmm0 + xmm1 */
+      tmp1 = emit_fetch( bld, inst, 2, CHAN_X );  /* xmm1 = src[2].x */
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);              /* xmm0 = xmm0 + xmm1 */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);  /* dest[ch] = xmm0 */
+      }
+      break;
+
+#if 0
+   case TGSI_OPCODE_FRC:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_frc( bld, 0, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_CLAMP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_FLR:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_flr( bld, 0, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_ROUND:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_rnd( bld, 0, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+#endif
+
+   case TGSI_OPCODE_EX2: {
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp0 = lp_build_exp2( &bld->base, tmp0);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+   }
+
+   case TGSI_OPCODE_LG2:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp0 = lp_build_log2( &bld->base, tmp0);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_POW:
+      src0 = emit_fetch( bld, inst, 0, CHAN_X );
+      src1 = emit_fetch( bld, inst, 1, CHAN_X );
+      dst0 = lp_build_pow( &bld->base, src0, src1 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, dst0 );
+      }
+      break;
+
+   case TGSI_OPCODE_XPD:
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ) {
+         tmp1 = emit_fetch( bld, inst, 1, CHAN_Z );
+         tmp3 = emit_fetch( bld, inst, 0, CHAN_Z );
+      }
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) ) {
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_Y );
+         tmp4 = emit_fetch( bld, inst, 1, CHAN_Y );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) {
+         tmp2 = tmp0;
+         tmp2 = lp_build_mul( &bld->base, tmp2, tmp1);
+         tmp5 = tmp3;
+         tmp5 = lp_build_mul( &bld->base, tmp5, tmp4);
+         tmp2 = lp_build_sub( &bld->base, tmp2, tmp5);
+         emit_store( bld, inst, 0, CHAN_X, tmp2);
+      }
+      if( IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) ) {
+         tmp2 = emit_fetch( bld, inst, 1, CHAN_X );
+         tmp5 = emit_fetch( bld, inst, 0, CHAN_X );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) {
+         tmp3 = lp_build_mul( &bld->base, tmp3, tmp2);
+         tmp1 = lp_build_mul( &bld->base, tmp1, tmp5);
+         tmp3 = lp_build_sub( &bld->base, tmp3, tmp1);
+         emit_store( bld, inst, 0, CHAN_Y, tmp3);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) {
+         tmp5 = lp_build_mul( &bld->base, tmp5, tmp4);
+         tmp0 = lp_build_mul( &bld->base, tmp0, tmp2);
+         tmp5 = lp_build_sub( &bld->base, tmp5, tmp0);
+         emit_store( bld, inst, 0, CHAN_Z, tmp5);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_W ) {
+         tmp0 = bld->base.one;
+         emit_store( bld, inst, 0, CHAN_W, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_ABS:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         tmp0 = lp_build_abs( &bld->base, tmp0 ) ;
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_RCC:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DPH:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_X );
+      tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Y );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Z );
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Z );
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_W );
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_COS:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp0 = lp_build_cos( &bld->base, tmp0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_DDX:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DDY:
+      return 0;
+      break;
+
+#if 0
+   case TGSI_OPCODE_KILP:
+      /* predicated kill */
+      emit_kilp( bld );
+      return 0; /* XXX fix me */
+      break;
+#endif
+
+   case TGSI_OPCODE_KIL:
+      /* conditional kill */
+      emit_kil( bld, inst );
+      break;
+
+   case TGSI_OPCODE_PK2H:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK2US:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK4B:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK4UB:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_RFL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SEQ:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_EQUAL, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SFL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SGT:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_GREATER, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SIN:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+      tmp0 = lp_build_sin( &bld->base, tmp0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_SLE:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_LEQUAL, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SNE:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_NOTEQUAL, src0, src1 );
+         dst0 = lp_build_select( &bld->base, tmp0, bld->base.one, bld->base.zero );
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_STR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TEX:
+      emit_tex( bld, inst, FALSE, FALSE );
+      break;
+
+   case TGSI_OPCODE_TXD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP2H:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP2US:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP4B:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP4UB:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_X2D:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ARA:
+      return 0;
+      break;
+
+#if 0
+   case TGSI_OPCODE_ARR:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_rnd( bld, 0, 0 );
+         emit_f2it( bld, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+#endif
+
+   case TGSI_OPCODE_BRA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CAL:
+      return 0;
+      break;
+
+#if 0
+   case TGSI_OPCODE_RET:
+      emit_ret( bld );
+      break;
+#endif
+
+   case TGSI_OPCODE_END:
+      break;
+
+#if 0
+   case TGSI_OPCODE_SSG:
+   /* TGSI_OPCODE_SGN */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_sgn( bld, 0, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+#endif
+
+   case TGSI_OPCODE_CMP:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         src0 = emit_fetch( bld, inst, 0, chan_index );
+         src1 = emit_fetch( bld, inst, 1, chan_index );
+         src2 = emit_fetch( bld, inst, 2, chan_index );
+         tmp0 = lp_build_cmp( &bld->base, PIPE_FUNC_LESS, src0, bld->base.zero );
+         dst0 = lp_build_select( &bld->base, tmp0, src1, src2);
+         emit_store( bld, inst, 0, chan_index, dst0);
+      }
+      break;
+
+   case TGSI_OPCODE_SCS:
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_X ) {
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+         tmp0 = lp_build_cos( &bld->base, tmp0 );
+         emit_store( bld, inst, 0, CHAN_X, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Y ) {
+         tmp0 = emit_fetch( bld, inst, 0, CHAN_X );
+         tmp0 = lp_build_sin( &bld->base, tmp0 );
+         emit_store( bld, inst, 0, CHAN_Y, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_Z ) {
+         tmp0 = bld->base.zero;
+         emit_store( bld, inst, 0, CHAN_Z, tmp0);
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( inst, CHAN_W ) {
+         tmp0 = bld->base.one;
+         emit_store( bld, inst, 0, CHAN_W, tmp0);
+      }
+      break;
+
+   case TGSI_OPCODE_TXB:
+      emit_tex( bld, inst, TRUE, FALSE );
+      break;
+
+   case TGSI_OPCODE_NRM:
+      /* fall-through */
+   case TGSI_OPCODE_NRM4:
+      /* 3 or 4-component normalization */
+      {
+         uint dims = (inst->Instruction.Opcode == TGSI_OPCODE_NRM) ? 3 : 4;
+
+         if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_X) ||
+             IS_DST0_CHANNEL_ENABLED(inst, CHAN_Y) ||
+             IS_DST0_CHANNEL_ENABLED(inst, CHAN_Z) ||
+             (IS_DST0_CHANNEL_ENABLED(inst, CHAN_W) && dims == 4)) {
+
+            /* NOTE: Cannot use xmm regs 2/3 here (see emit_rsqrt() above). */
+
+            /* xmm4 = src.x */
+            /* xmm0 = src.x * src.x */
+            tmp0 = emit_fetch(bld, inst, 0, CHAN_X);
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_X)) {
+               tmp4 = tmp0;
+            }
+            tmp0 = lp_build_mul( &bld->base, tmp0, tmp0);
+
+            /* xmm5 = src.y */
+            /* xmm0 = xmm0 + src.y * src.y */
+            tmp1 = emit_fetch(bld, inst, 0, CHAN_Y);
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_Y)) {
+               tmp5 = tmp1;
+            }
+            tmp1 = lp_build_mul( &bld->base, tmp1, tmp1);
+            tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+
+            /* xmm6 = src.z */
+            /* xmm0 = xmm0 + src.z * src.z */
+            tmp1 = emit_fetch(bld, inst, 0, CHAN_Z);
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_Z)) {
+               tmp6 = tmp1;
+            }
+            tmp1 = lp_build_mul( &bld->base, tmp1, tmp1);
+            tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+
+            if (dims == 4) {
+               /* xmm7 = src.w */
+               /* xmm0 = xmm0 + src.w * src.w */
+               tmp1 = emit_fetch(bld, inst, 0, CHAN_W);
+               if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_W)) {
+                  tmp7 = tmp1;
+               }
+               tmp1 = lp_build_mul( &bld->base, tmp1, tmp1);
+               tmp0 = lp_build_add( &bld->base, tmp0, tmp1);
+            }
+
+            /* xmm1 = 1 / sqrt(xmm0) */
+            tmp1 = lp_build_rsqrt( &bld->base, tmp0);
+
+            /* dst.x = xmm1 * src.x */
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_X)) {
+               tmp4 = lp_build_mul( &bld->base, tmp4, tmp1);
+               emit_store(bld, inst, 0, CHAN_X, tmp4);
+            }
+
+            /* dst.y = xmm1 * src.y */
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_Y)) {
+               tmp5 = lp_build_mul( &bld->base, tmp5, tmp1);
+               emit_store(bld, inst, 0, CHAN_Y, tmp5);
+            }
+
+            /* dst.z = xmm1 * src.z */
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_Z)) {
+               tmp6 = lp_build_mul( &bld->base, tmp6, tmp1);
+               emit_store(bld, inst, 0, CHAN_Z, tmp6);
+            }
+
+            /* dst.w = xmm1 * src.w */
+            if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_X) && dims == 4) {
+               tmp7 = lp_build_mul( &bld->base, tmp7, tmp1);
+               emit_store(bld, inst, 0, CHAN_W, tmp7);
+            }
+         }
+
+         /* dst0.w = 1.0 */
+         if (IS_DST0_CHANNEL_ENABLED(inst, CHAN_W) && dims == 3) {
+            tmp0 = bld->base.one;
+            emit_store(bld, inst, 0, CHAN_W, tmp0);
+         }
+      }
+      break;
+
+   case TGSI_OPCODE_DIV:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DP2:
+      tmp0 = emit_fetch( bld, inst, 0, CHAN_X );  /* xmm0 = src[0].x */
+      tmp1 = emit_fetch( bld, inst, 1, CHAN_X );  /* xmm1 = src[1].x */
+      tmp0 = lp_build_mul( &bld->base, tmp0, tmp1);              /* xmm0 = xmm0 * xmm1 */
+      tmp1 = emit_fetch( bld, inst, 0, CHAN_Y );  /* xmm1 = src[0].y */
+      tmp2 = emit_fetch( bld, inst, 1, CHAN_Y );  /* xmm2 = src[1].y */
+      tmp1 = lp_build_mul( &bld->base, tmp1, tmp2);              /* xmm1 = xmm1 * xmm2 */
+      tmp0 = lp_build_add( &bld->base, tmp0, tmp1);              /* xmm0 = xmm0 + xmm1 */
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         emit_store( bld, inst, 0, chan_index, tmp0);  /* dest[ch] = xmm0 */
+      }
+      break;
+
+   case TGSI_OPCODE_TXL:
+      emit_tex( bld, inst, TRUE, FALSE );
+      break;
+
+   case TGSI_OPCODE_TXP:
+      emit_tex( bld, inst, FALSE, TRUE );
+      break;
+      
+   case TGSI_OPCODE_BRK:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_IF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_BGNFOR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_REP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ELSE:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDIF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDFOR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDREP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PUSHA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_POPA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CEIL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_I2F:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_NOT:
+      return 0;
+      break;
+
+#if 0
+   case TGSI_OPCODE_TRUNC:
+      FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+         tmp0 = emit_fetch( bld, inst, 0, chan_index );
+         emit_f2it( bld, 0 );
+         emit_i2f( bld, 0 );
+         emit_store( bld, inst, 0, chan_index, tmp0);
+      }
+      break;
+#endif
+
+   case TGSI_OPCODE_SHL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SHR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_AND:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_OR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_MOD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_XOR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SAD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TXF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TXQ:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CONT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_EMIT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDPRIM:
+      return 0;
+      break;
+
+   default:
+      return 0;
+   }
+   
+   return 1;
+}
+
+
+void
+lp_build_tgsi_soa(LLVMBuilderRef builder,
+                  const struct tgsi_token *tokens,
+                  union lp_type type,
+                  struct lp_build_mask_context *mask,
+                  LLVMValueRef consts_ptr,
+                  const LLVMValueRef *pos,
+                  const LLVMValueRef (*inputs)[NUM_CHANNELS],
+                  LLVMValueRef (*outputs)[NUM_CHANNELS],
+                  lp_emit_fetch_texel_soa_callback emit_fetch_texel,
+                  void *emit_fetch_texel_context)
+{
+   struct lp_build_tgsi_soa_context bld;
+   struct tgsi_parse_context parse;
+   uint num_immediates = 0;
+   unsigned i;
+
+   /* Setup build context */
+   memset(&bld, 0, sizeof bld);
+   lp_build_context_init(&bld.base, builder, type);
+   bld.mask = mask;
+   bld.pos = pos;
+   bld.inputs = inputs;
+   bld.outputs = outputs;
+   bld.consts_ptr = consts_ptr;
+   bld.emit_fetch_texel = emit_fetch_texel;
+   bld.emit_fetch_texel_context = emit_fetch_texel_context;
+
+   tgsi_parse_init( &parse, tokens );
+
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         /* Input already interpolated */
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (!emit_instruction( &bld, &parse.FullToken.FullInstruction )) {
+            unsigned opcode = parse.FullToken.FullInstruction.Instruction.Opcode;
+            const struct tgsi_opcode_info *info = tgsi_get_opcode_info(opcode);
+	    _debug_printf("warning: failed to translate tgsi opcode %s to LLVM\n",
+	                  info ? info->mnemonic : "<invalid>");
+	 }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         /* simply copy the immediate values into the next immediates[] slot */
+         {
+            const uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
+            assert(size <= 4);
+            assert(num_immediates < LP_MAX_IMMEDIATES);
+            for( i = 0; i < size; ++i )
+               bld.immediates[num_immediates][i] =
+                  lp_build_const_scalar(type, parse.FullToken.FullImmediate.u[i].Float);
+            for( i = size; i < 4; ++i )
+               bld.immediates[num_immediates][i] = bld.base.undef;
+            num_immediates++;
+         }
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   tgsi_parse_free( &parse );
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_type.c b/src/gallium/drivers/llvmpipe/lp_bld_type.c
new file mode 100644
index 0000000000..8e0026fd97
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_type.c
@@ -0,0 +1,170 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_debug.h"
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+
+
+LLVMTypeRef
+lp_build_elem_type(union lp_type type)
+{
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         return LLVMFloatType();
+         break;
+      case 64:
+         return LLVMDoubleType();
+         break;
+      default:
+         assert(0);
+         return LLVMFloatType();
+      }
+   }
+   else {
+      return LLVMIntType(type.width);
+   }
+}
+
+
+LLVMTypeRef
+lp_build_vec_type(union lp_type type)
+{
+   LLVMTypeRef elem_type = lp_build_elem_type(type);
+   return LLVMVectorType(elem_type, type.length);
+}
+
+
+/**
+ * This function is a mirror of lp_build_elem_type() above.
+ *
+ * XXX: I'm not sure if it wouldn't be easier/efficient to just recreate the
+ * type and check for identity.
+ */
+boolean
+lp_check_elem_type(union lp_type type, LLVMTypeRef elem_type) 
+{
+   LLVMTypeKind elem_kind;
+
+   assert(elem_type);
+   if(!elem_type)
+      return FALSE;
+
+   elem_kind = LLVMGetTypeKind(elem_type);
+
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         if(elem_kind != LLVMFloatTypeKind)
+            return FALSE;
+         break;
+      case 64:
+         if(elem_kind != LLVMDoubleTypeKind)
+            return FALSE;
+         break;
+      default:
+         assert(0);
+         return FALSE;
+      }
+   }
+   else {
+      if(elem_kind != LLVMIntegerTypeKind)
+         return FALSE;
+
+      if(LLVMGetIntTypeWidth(elem_type) != type.width)
+         return FALSE;
+   }
+
+   return TRUE; 
+}
+
+
+boolean
+lp_check_vec_type(union lp_type type, LLVMTypeRef vec_type) 
+{
+   LLVMTypeRef elem_type;
+
+   assert(vec_type);
+   if(!vec_type)
+      return FALSE;
+
+   if(LLVMGetTypeKind(vec_type) != LLVMVectorTypeKind)
+      return FALSE;
+
+   if(LLVMGetVectorSize(vec_type) != type.length)
+      return FALSE;
+
+   elem_type = LLVMGetElementType(vec_type);
+
+   return lp_check_elem_type(type, elem_type);
+}
+
+
+boolean
+lp_check_value(union lp_type type, LLVMValueRef val) 
+{
+   LLVMTypeRef vec_type;
+
+   assert(val);
+   if(!val)
+      return FALSE;
+
+   vec_type = LLVMTypeOf(val);
+
+   return lp_check_vec_type(type, vec_type);
+}
+
+
+LLVMTypeRef
+lp_build_int_elem_type(union lp_type type)
+{
+   return LLVMIntType(type.width);
+}
+
+
+LLVMTypeRef
+lp_build_int_vec_type(union lp_type type)
+{
+   LLVMTypeRef elem_type = lp_build_int_elem_type(type);
+   return LLVMVectorType(elem_type, type.length);
+}
+
+
+void
+lp_build_context_init(struct lp_build_context *bld,
+                      LLVMBuilderRef builder,
+                      union lp_type type)
+{
+   bld->builder = builder;
+   bld->type = type;
+   bld->undef = lp_build_undef(type);
+   bld->zero = lp_build_zero(type);
+   bld->one = lp_build_one(type);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_type.h b/src/gallium/drivers/llvmpipe/lp_bld_type.h
new file mode 100644
index 0000000000..3ce566be64
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_bld_type.h
@@ -0,0 +1,174 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Convenient representation of SIMD types.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_BLD_TYPE_H
+#define LP_BLD_TYPE_H
+
+
+#include <llvm-c/Core.h>  
+
+#include <pipe/p_compiler.h>
+
+
+/**
+ * Several functions can only cope with vectors of length up to this value.
+ * You may need to increase that value if you want to represent bigger vectors.
+ */
+#define LP_MAX_VECTOR_LENGTH 16
+
+#define LP_MAX_TYPE_WIDTH 64
+
+
+/**
+ * The LLVM type system can't conveniently express all the things we care about
+ * on the types used for intermediate computations, such as signed vs unsigned,
+ * normalized values, or fixed point.
+ */
+union lp_type {
+   struct {
+      /** 
+       * Floating-point. Cannot be used with fixed. Integer numbers are
+       * represented by this zero.
+       */
+      unsigned floating:1;
+
+      /** 
+       * Fixed-point. Cannot be used with floating. Integer numbers are
+       * represented by this zero.
+       */
+      unsigned fixed:1;
+      
+      /** 
+       * Whether it can represent negative values or not.
+       *
+       * If this is not set for floating point, it means that all values are
+       * assumed to be positive.
+       */
+      unsigned sign:1;
+
+      /**
+       * Whether values are normalized to fit [0, 1] interval, or [-1, 1]
+       * interval for signed types.
+       *
+       * For integer types it means the representable integer range should be
+       * interpreted as the interval above.
+       *
+       * For floating and fixed point formats it means the values should be
+       * clamped to the interval above.
+       */
+      unsigned norm:1;
+
+      /**
+       * Element width.
+       *
+       * For fixed point values, the fixed point is assumed to be at half the
+       * width.
+       */
+      unsigned width:14;
+
+      /** 
+       * Vector length.
+       *
+       * width*length should be a power of two greater or equal to eight.
+       *
+       * @sa LP_MAX_VECTOR_LENGTH
+       */
+      unsigned length:14;
+   };
+   uint32_t value;
+};
+
+
+/**
+ * We need most of the information here in order to correctly and efficiently
+ * translate an arithmetic operation into LLVM IR. Putting it here avoids the
+ * trouble of passing it as parameters.
+ */
+struct lp_build_context
+{
+   LLVMBuilderRef builder;
+
+   /**
+    * This not only describes the input/output LLVM types, but also whether
+    * to normalize/clamp the results.
+    */
+   union lp_type type;
+
+   /** Same as lp_build_undef(type) */
+   LLVMValueRef undef;
+
+   /** Same as lp_build_zero(type) */
+   LLVMValueRef zero;
+
+   /** Same as lp_build_one(type) */
+   LLVMValueRef one;
+};
+
+
+LLVMTypeRef
+lp_build_elem_type(union lp_type type);
+
+
+LLVMTypeRef
+lp_build_vec_type(union lp_type type);
+
+
+boolean
+lp_check_elem_type(union lp_type type, LLVMTypeRef elem_type);
+
+
+boolean
+lp_check_vec_type(union lp_type type, LLVMTypeRef vec_type);
+
+
+boolean
+lp_check_value(union lp_type type, LLVMValueRef val);
+
+
+LLVMTypeRef
+lp_build_int_elem_type(union lp_type type);
+
+
+LLVMTypeRef
+lp_build_int_vec_type(union lp_type type);
+
+
+void
+lp_build_context_init(struct lp_build_context *bld,
+                      LLVMBuilderRef builder,
+                      union lp_type type);
+
+
+#endif /* !LP_BLD_TYPE_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_buffer.c b/src/gallium/drivers/llvmpipe/lp_buffer.c
new file mode 100644
index 0000000000..66f1f8e138
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_buffer.c
@@ -0,0 +1,150 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "lp_winsys.h"
+#include "lp_screen.h"
+#include "lp_texture.h"
+#include "lp_buffer.h"
+
+
+static void *
+llvmpipe_buffer_map(struct pipe_screen *screen,
+                    struct pipe_buffer *buf,
+                    unsigned flags)
+{
+   struct llvmpipe_buffer *llvmpipe_buf = llvmpipe_buffer(buf);
+   return llvmpipe_buf->data;
+}
+
+
+static void
+llvmpipe_buffer_unmap(struct pipe_screen *screen,
+                      struct pipe_buffer *buf)
+{
+}
+
+
+static void
+llvmpipe_buffer_destroy(struct pipe_buffer *buf)
+{
+   struct llvmpipe_buffer *sbuf = llvmpipe_buffer(buf);
+
+   if (!sbuf->userBuffer)
+      align_free(sbuf->data);
+      
+   FREE(sbuf);
+}
+
+
+static struct pipe_buffer *
+llvmpipe_buffer_create(struct pipe_screen *screen,
+                       unsigned alignment,
+                       unsigned usage,
+                       unsigned size)
+{
+   struct llvmpipe_buffer *buffer = CALLOC_STRUCT(llvmpipe_buffer);
+
+   pipe_reference_init(&buffer->base.reference, 1);
+   buffer->base.screen = screen;
+   buffer->base.alignment = MAX2(alignment, 16);
+   buffer->base.usage = usage;
+   buffer->base.size = size;
+
+   buffer->data = align_malloc(size, alignment);
+
+   return &buffer->base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+llvmpipe_user_buffer_create(struct pipe_screen *screen,
+                            void *ptr,
+                            unsigned bytes)
+{
+   struct llvmpipe_buffer *buffer;
+
+   buffer = CALLOC_STRUCT(llvmpipe_buffer);
+   if(!buffer)
+      return NULL;
+
+   pipe_reference_init(&buffer->base.reference, 1);
+   buffer->base.screen = screen;
+   buffer->base.size = bytes;
+   buffer->userBuffer = TRUE;
+   buffer->data = ptr;
+
+   return &buffer->base;
+}
+
+
+static void
+llvmpipe_fence_reference(struct pipe_screen *screen,
+                         struct pipe_fence_handle **ptr,
+                         struct pipe_fence_handle *fence)
+{
+}
+
+
+static int
+llvmpipe_fence_signalled(struct pipe_screen *screen,
+                         struct pipe_fence_handle *fence,
+                         unsigned flag)
+{
+   return 0;
+}
+
+
+static int
+llvmpipe_fence_finish(struct pipe_screen *screen,
+                      struct pipe_fence_handle *fence,
+                      unsigned flag)
+{
+   return 0;
+}
+
+
+void
+llvmpipe_init_screen_buffer_funcs(struct pipe_screen *screen)
+{
+   screen->buffer_create = llvmpipe_buffer_create;
+   screen->user_buffer_create = llvmpipe_user_buffer_create;
+   screen->buffer_map = llvmpipe_buffer_map;
+   screen->buffer_unmap = llvmpipe_buffer_unmap;
+   screen->buffer_destroy = llvmpipe_buffer_destroy;
+
+   screen->fence_reference = llvmpipe_fence_reference;
+   screen->fence_signalled = llvmpipe_fence_signalled;
+   screen->fence_finish = llvmpipe_fence_finish;
+
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_buffer.h b/src/gallium/drivers/llvmpipe/lp_buffer.h
new file mode 100644
index 0000000000..d6b8184a0b
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_buffer.h
@@ -0,0 +1,55 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef LP_BUFFER_H
+#define LP_BUFFER_H
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+
+struct llvmpipe_buffer
+{
+   struct pipe_buffer base;
+   boolean userBuffer;  /** Is this a user-space buffer? */
+   void *data;
+};
+
+
+/** Cast wrapper */
+static INLINE struct llvmpipe_buffer *
+llvmpipe_buffer( struct pipe_buffer *buf )
+{
+   return (struct llvmpipe_buffer *)buf;
+}
+
+
+void
+llvmpipe_init_screen_buffer_funcs(struct pipe_screen *screen);
+
+
+#endif /* LP_BUFFER_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_clear.c b/src/gallium/drivers/llvmpipe/lp_clear.c
new file mode 100644
index 0000000000..580cca5b46
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_clear.c
@@ -0,0 +1,80 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ *    Michel Dänzer
+ */
+
+
+#include "pipe/p_defines.h"
+#include "util/u_pack_color.h"
+#include "lp_clear.h"
+#include "lp_context.h"
+#include "lp_surface.h"
+#include "lp_state.h"
+#include "lp_tile_cache.h"
+
+
+/**
+ * Clear the given buffers to the specified values.
+ * No masking, no scissor (clear entire buffer).
+ */
+void
+llvmpipe_clear(struct pipe_context *pipe, unsigned buffers, const float *rgba,
+               double depth, unsigned stencil)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   unsigned cv;
+   uint i;
+
+   if (llvmpipe->no_rast)
+      return;
+
+#if 0
+   llvmpipe_update_derived(llvmpipe); /* not needed?? */
+#endif
+
+   if (buffers & PIPE_CLEAR_COLOR) {
+      for (i = 0; i < llvmpipe->framebuffer.nr_cbufs; i++) {
+         struct pipe_surface *ps = llvmpipe->framebuffer.cbufs[i];
+
+         util_pack_color(rgba, ps->format, &cv);
+         lp_tile_cache_clear(llvmpipe->cbuf_cache[i], rgba, cv);
+      }
+   }
+
+   if (buffers & PIPE_CLEAR_DEPTHSTENCIL) {
+      struct pipe_surface *ps = llvmpipe->framebuffer.zsbuf;
+
+      cv = util_pack_z_stencil(ps->format, depth, stencil);
+
+      /* non-cached surface */
+      pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, cv);
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_clear.h b/src/gallium/drivers/llvmpipe/lp_clear.h
new file mode 100644
index 0000000000..6d4ffccdf4
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_clear.h
@@ -0,0 +1,43 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ */
+
+#ifndef LP_CLEAR_H
+#define LP_CLEAR_H
+
+#include "pipe/p_state.h"
+struct pipe_context;
+
+extern void
+llvmpipe_clear(struct pipe_context *pipe, unsigned buffers, const float *rgba,
+               double depth, unsigned stencil);
+
+
+#endif /* LP_CLEAR_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_context.c b/src/gallium/drivers/llvmpipe/lp_context.c
new file mode 100644
index 0000000000..233d1df0e1
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_context.c
@@ -0,0 +1,297 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "lp_clear.h"
+#include "lp_context.h"
+#include "lp_flush.h"
+#include "lp_prim_setup.h"
+#include "lp_prim_vbuf.h"
+#include "lp_state.h"
+#include "lp_surface.h"
+#include "lp_tile_cache.h"
+#include "lp_tex_cache.h"
+#include "lp_texture.h"
+#include "lp_winsys.h"
+#include "lp_query.h"
+
+
+
+/**
+ * Map any drawing surfaces which aren't already mapped
+ */
+void
+llvmpipe_map_transfers(struct llvmpipe_context *lp)
+{
+   struct pipe_screen *screen = lp->pipe.screen;
+   struct pipe_surface *zsbuf = lp->framebuffer.zsbuf;
+   unsigned i;
+
+   for (i = 0; i < lp->framebuffer.nr_cbufs; i++) {
+      lp_tile_cache_map_transfers(lp->cbuf_cache[i]);
+   }
+
+   if(zsbuf) {
+      if(!lp->zsbuf_transfer)
+         lp->zsbuf_transfer = screen->get_tex_transfer(screen, zsbuf->texture,
+                                                       zsbuf->face, zsbuf->level, zsbuf->zslice,
+                                                       PIPE_TRANSFER_READ_WRITE,
+                                                       0, 0, zsbuf->width, zsbuf->height);
+      if(lp->zsbuf_transfer && !lp->zsbuf_map)
+         lp->zsbuf_map = screen->transfer_map(screen, lp->zsbuf_transfer);
+
+   }
+}
+
+
+/**
+ * Unmap any mapped drawing surfaces
+ */
+void
+llvmpipe_unmap_transfers(struct llvmpipe_context *lp)
+{
+   uint i;
+
+   for (i = 0; i < lp->framebuffer.nr_cbufs; i++) {
+      lp_tile_cache_unmap_transfers(lp->cbuf_cache[i]);
+   }
+
+   if(lp->zsbuf_transfer) {
+      struct pipe_screen *screen = lp->pipe.screen;
+
+      if(lp->zsbuf_map) {
+         screen->transfer_unmap(screen, lp->zsbuf_transfer);
+         lp->zsbuf_map = NULL;
+      }
+   }
+}
+
+
+static void llvmpipe_destroy( struct pipe_context *pipe )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
+   uint i;
+
+   if (llvmpipe->draw)
+      draw_destroy( llvmpipe->draw );
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++)
+      lp_destroy_tile_cache(llvmpipe->cbuf_cache[i]);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      lp_destroy_tex_tile_cache(llvmpipe->tex_cache[i]);
+
+   for (i = 0; i < Elements(llvmpipe->constants); i++) {
+      if (llvmpipe->constants[i].buffer) {
+         pipe_buffer_reference(&llvmpipe->constants[i].buffer, NULL);
+      }
+   }
+
+   align_free( llvmpipe );
+}
+
+static unsigned int
+llvmpipe_is_texture_referenced( struct pipe_context *pipe,
+				struct pipe_texture *texture,
+				unsigned face, unsigned level)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
+   unsigned i;
+
+   if(llvmpipe->dirty_render_cache) {
+      for (i = 0; i < llvmpipe->framebuffer.nr_cbufs; i++) {
+         if(llvmpipe->framebuffer.cbufs[i] && 
+            llvmpipe->framebuffer.cbufs[i]->texture == texture)
+            return PIPE_REFERENCED_FOR_WRITE;
+      }
+      if(llvmpipe->framebuffer.zsbuf && 
+         llvmpipe->framebuffer.zsbuf->texture == texture)
+         return PIPE_REFERENCED_FOR_WRITE;
+   }
+   
+   /* FIXME: we also need to do the same for the texture cache */
+   
+   return PIPE_UNREFERENCED;
+}
+
+static unsigned int
+llvmpipe_is_buffer_referenced( struct pipe_context *pipe,
+			       struct pipe_buffer *buf)
+{
+   return PIPE_UNREFERENCED;
+}
+
+struct pipe_context *
+llvmpipe_create( struct pipe_screen *screen )
+{
+   struct llvmpipe_context *llvmpipe;
+   uint i;
+
+   llvmpipe = align_malloc(sizeof(struct llvmpipe_context), 16);
+   if (!llvmpipe)
+      return NULL;
+
+   util_init_math();
+
+   memset(llvmpipe, 0, sizeof *llvmpipe);
+
+   llvmpipe->pipe.winsys = screen->winsys;
+   llvmpipe->pipe.screen = screen;
+   llvmpipe->pipe.destroy = llvmpipe_destroy;
+
+   /* state setters */
+   llvmpipe->pipe.create_blend_state = llvmpipe_create_blend_state;
+   llvmpipe->pipe.bind_blend_state   = llvmpipe_bind_blend_state;
+   llvmpipe->pipe.delete_blend_state = llvmpipe_delete_blend_state;
+
+   llvmpipe->pipe.create_sampler_state = llvmpipe_create_sampler_state;
+   llvmpipe->pipe.bind_sampler_states  = llvmpipe_bind_sampler_states;
+   llvmpipe->pipe.delete_sampler_state = llvmpipe_delete_sampler_state;
+
+   llvmpipe->pipe.create_depth_stencil_alpha_state = llvmpipe_create_depth_stencil_state;
+   llvmpipe->pipe.bind_depth_stencil_alpha_state   = llvmpipe_bind_depth_stencil_state;
+   llvmpipe->pipe.delete_depth_stencil_alpha_state = llvmpipe_delete_depth_stencil_state;
+
+   llvmpipe->pipe.create_rasterizer_state = llvmpipe_create_rasterizer_state;
+   llvmpipe->pipe.bind_rasterizer_state   = llvmpipe_bind_rasterizer_state;
+   llvmpipe->pipe.delete_rasterizer_state = llvmpipe_delete_rasterizer_state;
+
+   llvmpipe->pipe.create_fs_state = llvmpipe_create_fs_state;
+   llvmpipe->pipe.bind_fs_state   = llvmpipe_bind_fs_state;
+   llvmpipe->pipe.delete_fs_state = llvmpipe_delete_fs_state;
+
+   llvmpipe->pipe.create_vs_state = llvmpipe_create_vs_state;
+   llvmpipe->pipe.bind_vs_state   = llvmpipe_bind_vs_state;
+   llvmpipe->pipe.delete_vs_state = llvmpipe_delete_vs_state;
+
+   llvmpipe->pipe.set_blend_color = llvmpipe_set_blend_color;
+   llvmpipe->pipe.set_clip_state = llvmpipe_set_clip_state;
+   llvmpipe->pipe.set_constant_buffer = llvmpipe_set_constant_buffer;
+   llvmpipe->pipe.set_framebuffer_state = llvmpipe_set_framebuffer_state;
+   llvmpipe->pipe.set_polygon_stipple = llvmpipe_set_polygon_stipple;
+   llvmpipe->pipe.set_scissor_state = llvmpipe_set_scissor_state;
+   llvmpipe->pipe.set_sampler_textures = llvmpipe_set_sampler_textures;
+   llvmpipe->pipe.set_viewport_state = llvmpipe_set_viewport_state;
+
+   llvmpipe->pipe.set_vertex_buffers = llvmpipe_set_vertex_buffers;
+   llvmpipe->pipe.set_vertex_elements = llvmpipe_set_vertex_elements;
+
+   llvmpipe->pipe.draw_arrays = llvmpipe_draw_arrays;
+   llvmpipe->pipe.draw_elements = llvmpipe_draw_elements;
+   llvmpipe->pipe.draw_range_elements = llvmpipe_draw_range_elements;
+   llvmpipe->pipe.set_edgeflags = llvmpipe_set_edgeflags;
+
+
+   llvmpipe->pipe.clear = llvmpipe_clear;
+   llvmpipe->pipe.flush = llvmpipe_flush;
+
+   llvmpipe->pipe.is_texture_referenced = llvmpipe_is_texture_referenced;
+   llvmpipe->pipe.is_buffer_referenced = llvmpipe_is_buffer_referenced;
+
+   llvmpipe_init_query_funcs( llvmpipe );
+   llvmpipe_init_texture_funcs( llvmpipe );
+
+   /*
+    * Alloc caches for accessing drawing surfaces and textures.
+    */
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++)
+      llvmpipe->cbuf_cache[i] = lp_create_tile_cache( screen );
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      llvmpipe->tex_cache[i] = lp_create_tex_tile_cache( screen );
+
+
+   /* vertex shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      llvmpipe->tgsi.vert_samplers[i].base.get_samples = lp_get_samples;
+      llvmpipe->tgsi.vert_samplers[i].processor = TGSI_PROCESSOR_VERTEX;
+      llvmpipe->tgsi.vert_samplers[i].cache = llvmpipe->tex_cache[i];
+      llvmpipe->tgsi.vert_samplers_list[i] = &llvmpipe->tgsi.vert_samplers[i];
+   }
+
+   /* fragment shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      llvmpipe->tgsi.frag_samplers[i].base.get_samples = lp_get_samples;
+      llvmpipe->tgsi.frag_samplers[i].processor = TGSI_PROCESSOR_FRAGMENT;
+      llvmpipe->tgsi.frag_samplers[i].cache = llvmpipe->tex_cache[i];
+      llvmpipe->tgsi.frag_samplers_list[i] = &llvmpipe->tgsi.frag_samplers[i];
+   }
+
+   /*
+    * Create drawing context and plug our rendering stage into it.
+    */
+   llvmpipe->draw = draw_create();
+   if (!llvmpipe->draw) 
+      goto fail;
+
+   draw_texture_samplers(llvmpipe->draw,
+                         PIPE_MAX_SAMPLERS,
+                         (struct tgsi_sampler **)
+                            llvmpipe->tgsi.vert_samplers_list);
+
+   llvmpipe->setup = lp_draw_render_stage(llvmpipe);
+   if (!llvmpipe->setup)
+      goto fail;
+
+   if (debug_get_bool_option( "LP_NO_RAST", FALSE ))
+      llvmpipe->no_rast = TRUE;
+
+   if (debug_get_bool_option( "LP_NO_VBUF", FALSE )) {
+      /* Deprecated path -- vbuf is the intended interface to the draw module:
+       */
+      draw_set_rasterize_stage(llvmpipe->draw, llvmpipe->setup);
+   }
+   else {
+      lp_init_vbuf(llvmpipe);
+   }
+
+   /* plug in AA line/point stages */
+   draw_install_aaline_stage(llvmpipe->draw, &llvmpipe->pipe);
+   draw_install_aapoint_stage(llvmpipe->draw, &llvmpipe->pipe);
+
+#if USE_DRAW_STAGE_PSTIPPLE
+   /* Do polygon stipple w/ texture map + frag prog? */
+   draw_install_pstipple_stage(llvmpipe->draw, &llvmpipe->pipe);
+#endif
+
+   lp_init_surface_functions(llvmpipe);
+
+   return &llvmpipe->pipe;
+
+ fail:
+   llvmpipe_destroy(&llvmpipe->pipe);
+   return NULL;
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_context.h b/src/gallium/drivers/llvmpipe/lp_context.h
new file mode 100644
index 0000000000..8d5a0d4f1f
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_context.h
@@ -0,0 +1,155 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef LP_CONTEXT_H
+#define LP_CONTEXT_H
+
+#include "pipe/p_context.h"
+
+#include "draw/draw_vertex.h"
+
+#include "lp_tex_sample.h"
+#include "lp_jit.h"
+
+
+struct llvmpipe_vbuf_render;
+struct draw_context;
+struct draw_stage;
+struct llvmpipe_tile_cache;
+struct llvmpipe_tex_tile_cache;
+struct lp_fragment_shader;
+struct lp_vertex_shader;
+struct lp_blend_state;
+
+
+struct llvmpipe_context {
+   struct pipe_context pipe;  /**< base class */
+
+   /** Constant state objects */
+   const struct pipe_blend_state *blend;
+   const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS];
+   const struct pipe_depth_stencil_alpha_state *depth_stencil;
+   const struct pipe_rasterizer_state *rasterizer;
+   struct lp_fragment_shader *fs;
+   const struct lp_vertex_shader *vs;
+
+   /** Other rendering state */
+   struct pipe_blend_color blend_color[4][16];
+   struct pipe_clip_state clip;
+   struct pipe_constant_buffer constants[PIPE_SHADER_TYPES];
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
+
+   unsigned num_samplers;
+   unsigned num_textures;
+   unsigned num_vertex_elements;
+   unsigned num_vertex_buffers;
+
+   unsigned dirty; /**< Mask of LP_NEW_x flags */
+
+   /* Counter for occlusion queries.  Note this supports overlapping
+    * queries.
+    */
+   uint64_t occlusion_count;
+   unsigned active_query_count;
+
+   /** Mapped vertex buffers */
+   ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS];
+   
+   /** Mapped constant buffers */
+   void *mapped_constants[PIPE_SHADER_TYPES];
+
+   /** Vertex format */
+   struct vertex_info vertex_info;
+   struct vertex_info vertex_info_vbuf;
+
+   /** Which vertex shader output slot contains point size */
+   int psize_slot;
+
+   /* The reduced version of the primitive supplied by the state
+    * tracker.
+    */
+   unsigned reduced_api_prim;
+
+   /* The reduced primitive after unfilled triangles, wide-line
+    * decomposition, etc, are taken into account.  This is the
+    * primitive actually rasterized.
+    */
+   unsigned reduced_prim;
+
+   /** Derived from scissor and surface bounds: */
+   struct pipe_scissor_state cliprect;
+
+   unsigned line_stipple_counter;
+
+   /** TGSI exec things */
+   struct {
+      struct lp_shader_sampler vert_samplers[PIPE_MAX_SAMPLERS];
+      struct lp_shader_sampler *vert_samplers_list[PIPE_MAX_SAMPLERS];
+      struct lp_shader_sampler frag_samplers[PIPE_MAX_SAMPLERS];
+      struct lp_shader_sampler *frag_samplers_list[PIPE_MAX_SAMPLERS];
+   } tgsi;
+
+   /** The primitive drawing context */
+   struct draw_context *draw;
+   struct draw_stage *setup;
+   struct draw_stage *vbuf;
+   struct llvmpipe_vbuf_render *vbuf_render;
+
+   boolean dirty_render_cache;
+   
+   struct llvmpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS];
+   
+   /* TODO: we shouldn't be using external interfaces internally like this */
+   struct pipe_transfer *zsbuf_transfer;
+   uint8_t *zsbuf_map;
+
+   unsigned tex_timestamp;
+   struct llvmpipe_tex_tile_cache *tex_cache[PIPE_MAX_SAMPLERS];
+
+   unsigned no_rast : 1;
+
+   struct lp_jit_context jit_context;
+};
+
+
+static INLINE struct llvmpipe_context *
+llvmpipe_context( struct pipe_context *pipe )
+{
+   return (struct llvmpipe_context *)pipe;
+}
+
+#endif /* LP_CONTEXT_H */
+
diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
new file mode 100644
index 0000000000..89772e62d3
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
@@ -0,0 +1,192 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ *    Keith Whitwell
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_prim.h"
+
+#include "lp_buffer.h"
+#include "lp_context.h"
+#include "lp_state.h"
+
+#include "draw/draw_context.h"
+
+
+
+static void
+llvmpipe_map_constant_buffers(struct llvmpipe_context *lp)
+{
+   struct pipe_screen *screen = lp->pipe.screen;
+   uint i, size;
+
+   for (i = 0; i < PIPE_SHADER_TYPES; i++) {
+      if (lp->constants[i].buffer && lp->constants[i].buffer->size)
+         lp->mapped_constants[i] = screen->buffer_map(screen, lp->constants[i].buffer,
+                                                      PIPE_BUFFER_USAGE_CPU_READ);
+   }
+
+   if (lp->constants[PIPE_SHADER_VERTEX].buffer)
+      size = lp->constants[PIPE_SHADER_VERTEX].buffer->size;
+   else
+      size = 0;
+
+   lp->jit_context.constants = lp->mapped_constants[PIPE_SHADER_FRAGMENT];
+
+   draw_set_mapped_constant_buffer(lp->draw,
+                                   lp->mapped_constants[PIPE_SHADER_VERTEX],
+                                   size);
+}
+
+
+static void
+llvmpipe_unmap_constant_buffers(struct llvmpipe_context *lp)
+{
+   struct pipe_screen *screen = lp->pipe.screen;
+   uint i;
+
+   /* really need to flush all prims since the vert/frag shaders const buffers
+    * are going away now.
+    */
+   draw_flush(lp->draw);
+
+   draw_set_mapped_constant_buffer(lp->draw, NULL, 0);
+
+   lp->jit_context.constants = NULL;
+
+   for (i = 0; i < 2; i++) {
+      if (lp->constants[i].buffer && lp->constants[i].buffer->size)
+         screen->buffer_unmap(screen, lp->constants[i].buffer);
+      lp->mapped_constants[i] = NULL;
+   }
+}
+
+
+boolean
+llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
+                     unsigned start, unsigned count)
+{
+   return llvmpipe_draw_elements(pipe, NULL, 0, mode, start, count);
+}
+
+
+/**
+ * Draw vertex arrays, with optional indexing.
+ * Basically, map the vertex buffers (and drawing surfaces), then hand off
+ * the drawing to the 'draw' module.
+ */
+boolean
+llvmpipe_draw_range_elements(struct pipe_context *pipe,
+                             struct pipe_buffer *indexBuffer,
+                             unsigned indexSize,
+                             unsigned min_index,
+                             unsigned max_index,
+                             unsigned mode, unsigned start, unsigned count)
+{
+   struct llvmpipe_context *lp = llvmpipe_context(pipe);
+   struct draw_context *draw = lp->draw;
+   unsigned i;
+
+   lp->reduced_api_prim = u_reduced_prim(mode);
+
+   if (lp->dirty)
+      llvmpipe_update_derived( lp );
+
+   llvmpipe_map_transfers(lp);
+   llvmpipe_map_constant_buffers(lp);
+
+   /*
+    * Map vertex buffers
+    */
+   for (i = 0; i < lp->num_vertex_buffers; i++) {
+      void *buf = llvmpipe_buffer(lp->vertex_buffer[i].buffer)->data;
+      draw_set_mapped_vertex_buffer(draw, i, buf);
+   }
+
+   /* Map index buffer, if present */
+   if (indexBuffer) {
+      void *mapped_indexes = llvmpipe_buffer(indexBuffer)->data;
+      draw_set_mapped_element_buffer_range(draw, indexSize,
+                                           min_index,
+                                           max_index,
+                                           mapped_indexes);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer_range(draw, 0, start,
+                                           start + count - 1, NULL);
+   }
+
+   /* draw! */
+   draw_arrays(draw, mode, start, count);
+
+   /*
+    * unmap vertex/index buffers - will cause draw module to flush
+    */
+   for (i = 0; i < lp->num_vertex_buffers; i++) {
+      draw_set_mapped_vertex_buffer(draw, i, NULL);
+   }
+   if (indexBuffer) {
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+
+
+   /* Note: leave drawing surfaces mapped */
+   llvmpipe_unmap_constant_buffers(lp);
+
+   lp->dirty_render_cache = TRUE;
+   
+   return TRUE;
+}
+
+
+boolean
+llvmpipe_draw_elements(struct pipe_context *pipe,
+                       struct pipe_buffer *indexBuffer,
+                       unsigned indexSize,
+                       unsigned mode, unsigned start, unsigned count)
+{
+   return llvmpipe_draw_range_elements( pipe, indexBuffer,
+                                        indexSize,
+                                        0, 0xffffffff,
+                                        mode, start, count );
+}
+
+
+void
+llvmpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
+{
+   struct llvmpipe_context *lp = llvmpipe_context(pipe);
+   draw_set_edgeflags(lp->draw, edgeflags);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.c b/src/gallium/drivers/llvmpipe/lp_flush.c
new file mode 100644
index 0000000000..b5c1c95bb7
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_flush.c
@@ -0,0 +1,98 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_defines.h"
+#include "draw/draw_context.h"
+#include "lp_flush.h"
+#include "lp_context.h"
+#include "lp_surface.h"
+#include "lp_state.h"
+#include "lp_tile_cache.h"
+#include "lp_tex_cache.h"
+#include "lp_winsys.h"
+
+
+void
+llvmpipe_flush( struct pipe_context *pipe,
+		unsigned flags,
+                struct pipe_fence_handle **fence )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   uint i;
+
+   draw_flush(llvmpipe->draw);
+
+   if (flags & PIPE_FLUSH_SWAPBUFFERS) {
+      /* If this is a swapbuffers, just flush color buffers.
+       *
+       * The zbuffer changes are not discarded, but held in the cache
+       * in the hope that a later clear will wipe them out.
+       */
+      for (i = 0; i < llvmpipe->framebuffer.nr_cbufs; i++)
+         if (llvmpipe->cbuf_cache[i])
+            lp_flush_tile_cache(llvmpipe->cbuf_cache[i]);
+
+      /* Need this call for hardware buffers before swapbuffers.
+       *
+       * there should probably be another/different flush-type function
+       * that's called before swapbuffers because we don't always want
+       * to unmap surfaces when flushing.
+       */
+      llvmpipe_unmap_transfers(llvmpipe);
+   }
+   else if (flags & PIPE_FLUSH_RENDER_CACHE) {
+      for (i = 0; i < llvmpipe->framebuffer.nr_cbufs; i++)
+         if (llvmpipe->cbuf_cache[i])
+            lp_flush_tile_cache(llvmpipe->cbuf_cache[i]);
+
+      /* FIXME: untile zsbuf! */
+     
+      llvmpipe->dirty_render_cache = FALSE;
+   }
+
+   /* Enable to dump BMPs of the color/depth buffers each frame */
+#if 0
+   if(flags & PIPE_FLUSH_FRAME) {
+      static unsigned frame_no = 1;
+      static char filename[256];
+      util_snprintf(filename, sizeof(filename), "cbuf_%u.bmp", frame_no);
+      debug_dump_surface_bmp(filename, llvmpipe->framebuffer.cbufs[0]);
+      util_snprintf(filename, sizeof(filename), "zsbuf_%u.bmp", frame_no);
+      debug_dump_surface_bmp(filename, llvmpipe->framebuffer.zsbuf);
+      ++frame_no;
+   }
+#endif
+   
+   if (fence)
+      *fence = NULL;
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.h b/src/gallium/drivers/llvmpipe/lp_flush.h
new file mode 100644
index 0000000000..10b2b52583
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_flush.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_FLUSH_H
+#define LP_FLUSH_H
+
+struct pipe_context;
+struct pipe_fence_handle;
+
+void llvmpipe_flush(struct pipe_context *pipe, unsigned flags,
+                    struct pipe_fence_handle **fence);
+
+#endif
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
new file mode 100644
index 0000000000..d288460a1b
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * C - JIT interfaces
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include <llvm-c/Transforms/Scalar.h>
+
+#include "util/u_memory.h"
+#include "lp_screen.h"
+#include "lp_bld_intr.h"
+#include "lp_jit.h"
+
+
+static void
+lp_jit_init_globals(struct llvmpipe_screen *screen)
+{
+   /* struct lp_jit_context */
+   {
+      LLVMTypeRef elem_types[4];
+      LLVMTypeRef context_type;
+
+      elem_types[0] = LLVMPointerType(LLVMFloatType(), 0); /* constants */
+      elem_types[1] = LLVMPointerType(LLVMInt8Type(), 0);  /* samplers */
+      elem_types[2] = LLVMFloatType();                     /* alpha_ref_value */
+      elem_types[3] = LLVMPointerType(LLVMInt8Type(), 0);  /* blend_color */
+
+      context_type = LLVMStructType(elem_types, Elements(elem_types), 0);
+
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, constants,
+                             screen->target, context_type, 0);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, samplers,
+                             screen->target, context_type, 1);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, alpha_ref_value,
+                             screen->target, context_type, 2);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_context, blend_color,
+                             screen->target, context_type, 3);
+      LP_CHECK_STRUCT_SIZE(struct lp_jit_context,
+                           screen->target, context_type);
+
+      LLVMAddTypeName(screen->module, "context", context_type);
+
+      screen->context_ptr_type = LLVMPointerType(context_type, 0);
+   }
+
+   /* fetch_texel
+    */
+   {
+      LLVMTypeRef ret_type;
+      LLVMTypeRef arg_types[3];
+      LLVMValueRef fetch_texel;
+
+      ret_type = LLVMVoidType();
+      arg_types[0] = LLVMPointerType(LLVMInt8Type(), 0);  /* samplers */
+      arg_types[1] = LLVMInt32Type();                     /* unit */
+      arg_types[2] = LLVMPointerType(LLVMVectorType(LLVMFloatType(), 4), 0); /* store */
+
+      fetch_texel = lp_declare_intrinsic(screen->module, "fetch_texel",
+                                         ret_type, arg_types, Elements(arg_types));
+
+      LLVMAddGlobalMapping(screen->engine, fetch_texel, lp_fetch_texel_soa);
+   }
+
+#ifdef DEBUG
+   LLVMDumpModule(screen->module);
+#endif
+}
+
+
+void
+lp_jit_screen_cleanup(struct llvmpipe_screen *screen)
+{
+   if(screen->engine)
+      LLVMDisposeExecutionEngine(screen->engine);
+
+   if(screen->pass)
+      LLVMDisposePassManager(screen->pass);
+}
+
+
+void
+lp_jit_screen_init(struct llvmpipe_screen *screen)
+{
+   char *error = NULL;
+
+   screen->module = LLVMModuleCreateWithName("llvmpipe");
+
+   screen->provider = LLVMCreateModuleProviderForExistingModule(screen->module);
+
+   if (LLVMCreateJITCompiler(&screen->engine, screen->provider, 1, &error)) {
+      fprintf(stderr, "%s\n", error);
+      LLVMDisposeMessage(error);
+      abort();
+   }
+
+   screen->target = LLVMGetExecutionEngineTargetData(screen->engine);
+
+   screen->pass = LLVMCreateFunctionPassManager(screen->provider);
+   LLVMAddTargetData(screen->target, screen->pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   LLVMAddConstantPropagationPass(screen->pass);
+   LLVMAddInstructionCombiningPass(screen->pass);
+   LLVMAddPromoteMemoryToRegisterPass(screen->pass);
+   LLVMAddGVNPass(screen->pass);
+   LLVMAddCFGSimplificationPass(screen->pass);
+
+   lp_jit_init_globals(screen);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.h b/src/gallium/drivers/llvmpipe/lp_jit.h
new file mode 100644
index 0000000000..a7fb60f9f5
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_jit.h
@@ -0,0 +1,109 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * C - JIT interfaces
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef LP_JIT_H
+#define LP_JIT_H
+
+
+#include "lp_bld_struct.h"
+
+
+struct tgsi_sampler;
+struct llvmpipe_screen;
+
+
+/**
+ * This structure is passed directly to the generated fragment shader.
+ *
+ * It contains the derived state.
+ *
+ * Changes here must be reflected in the lp_jit_context_* macros and
+ * lp_jit_init_types function. Changes to the ordering should be avoided.
+ *
+ * Only use types with a clear size and padding here, in particular prefer the
+ * stdint.h types to the basic integer types.
+ */
+struct lp_jit_context
+{
+   const float *constants;
+
+   struct tgsi_sampler **samplers;
+
+   /* TODO: alpha reference value */
+   float alpha_ref_value;
+
+   /* TODO: blend constant color */
+   uint8_t *blend_color;
+};
+
+
+#define lp_jit_context_constants(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 0, "constants")
+
+#define lp_jit_context_samplers(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 1, "samplers")
+
+#define lp_jit_context_alpha_ref_value(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 2, "alpha_ref_value")
+
+#define lp_jit_context_blend_color(_builder, _ptr) \
+   lp_build_struct_get(_builder, _ptr, 3, "blend_color")
+
+
+typedef void
+(*lp_jit_frag_func)(struct lp_jit_context *context,
+                    uint32_t x,
+                    uint32_t y,
+                    const void *a0,
+                    const void *dadx,
+                    const void *dady,
+                    uint32_t *mask,
+                    void *color,
+                    void *depth);
+
+void PIPE_CDECL
+lp_fetch_texel_soa( struct tgsi_sampler **samplers,
+                    uint32_t unit,
+                    float *store );
+
+
+void
+lp_jit_screen_cleanup(struct llvmpipe_screen *screen);
+
+
+void
+lp_jit_screen_init(struct llvmpipe_screen *screen);
+
+
+#endif /* LP_JIT_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_prim_setup.c b/src/gallium/drivers/llvmpipe/lp_prim_setup.c
new file mode 100644
index 0000000000..b14f8fb99d
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_prim_setup.c
@@ -0,0 +1,190 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief A draw stage that drives our triangle setup routines from
+ * within the draw pipeline.  One of two ways to drive setup, the
+ * other being in lp_prim_vbuf.c.
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ * \author  Brian Paul
+ */
+
+
+#include "lp_context.h"
+#include "lp_setup.h"
+#include "lp_state.h"
+#include "lp_prim_setup.h"
+#include "draw/draw_pipe.h"
+#include "draw/draw_vertex.h"
+#include "util/u_memory.h"
+
+/**
+ * Triangle setup info (derived from draw_stage).
+ * Also used for line drawing (taking some liberties).
+ */
+struct setup_stage {
+   struct draw_stage stage; /**< This must be first (base class) */
+
+   struct setup_context *setup;
+};
+
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct setup_stage *setup_stage( struct draw_stage *stage )
+{
+   return (struct setup_stage *)stage;
+}
+
+
+typedef const float (*cptrf4)[4];
+
+static void
+do_tri(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+   
+   llvmpipe_setup_tri( setup->setup,
+              (cptrf4)prim->v[0]->data,
+              (cptrf4)prim->v[1]->data,
+              (cptrf4)prim->v[2]->data );
+}
+
+static void
+do_line(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+
+   llvmpipe_setup_line( setup->setup,
+               (cptrf4)prim->v[0]->data,
+               (cptrf4)prim->v[1]->data );
+}
+
+static void
+do_point(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+
+   llvmpipe_setup_point( setup->setup,
+                (cptrf4)prim->v[0]->data );
+}
+
+
+
+
+static void setup_begin( struct draw_stage *stage )
+{
+   struct setup_stage *setup = setup_stage(stage);
+
+   llvmpipe_setup_prepare( setup->setup );
+
+   stage->point = do_point;
+   stage->line = do_line;
+   stage->tri = do_tri;
+}
+
+
+static void setup_first_point( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->point( stage, header );
+}
+
+static void setup_first_line( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->line( stage, header );
+}
+
+
+static void setup_first_tri( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->tri( stage, header );
+}
+
+
+
+static void setup_flush( struct draw_stage *stage,
+			 unsigned flags )
+{
+   stage->point = setup_first_point;
+   stage->line = setup_first_line;
+   stage->tri = setup_first_tri;
+}
+
+
+static void reset_stipple_counter( struct draw_stage *stage )
+{
+}
+
+
+static void render_destroy( struct draw_stage *stage )
+{
+   struct setup_stage *ssetup = setup_stage(stage);
+   llvmpipe_setup_destroy_context(ssetup->setup);
+   FREE( stage );
+}
+
+
+/**
+ * Create a new primitive setup/render stage.
+ */
+struct draw_stage *lp_draw_render_stage( struct llvmpipe_context *llvmpipe )
+{
+   struct setup_stage *sstage = CALLOC_STRUCT(setup_stage);
+
+   sstage->setup = llvmpipe_setup_create_context(llvmpipe);
+   sstage->stage.draw = llvmpipe->draw;
+   sstage->stage.point = setup_first_point;
+   sstage->stage.line = setup_first_line;
+   sstage->stage.tri = setup_first_tri;
+   sstage->stage.flush = setup_flush;
+   sstage->stage.reset_stipple_counter = reset_stipple_counter;
+   sstage->stage.destroy = render_destroy;
+
+   return (struct draw_stage *)sstage;
+}
+
+struct setup_context *
+lp_draw_setup_context( struct draw_stage *stage )
+{
+   struct setup_stage *ssetup = setup_stage(stage);
+   return ssetup->setup;
+}
+
+void
+lp_draw_flush( struct draw_stage *stage )
+{
+   stage->flush( stage, 0 );
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_prim_setup.h b/src/gallium/drivers/llvmpipe/lp_prim_setup.h
new file mode 100644
index 0000000000..da6cae6375
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_prim_setup.h
@@ -0,0 +1,85 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef LP_PRIM_SETUP_H
+#define LP_PRIM_SETUP_H
+
+
+/**
+ * vbuf is a special stage to gather the stream of triangles, lines, points
+ * together and reconstruct vertex buffers for hardware upload.
+ *
+ * First attempt, work in progress.
+ * 
+ * TODO:
+ *    - separate out vertex buffer building and primitive emit, ie >1 draw per vb.
+ *    - tell vbuf stage how to build hw vertices directly
+ *    - pass vbuf stage a buffer pointer for direct emit to agp/vram.
+ *
+ *
+ *
+ * Vertices are just an array of floats, with all the attributes
+ * packed.  We currently assume a layout like:
+ *
+ * attr[0][0..3] - window position
+ * attr[1..n][0..3] - remaining attributes.
+ *
+ * Attributes are assumed to be 4 floats wide but are packed so that
+ * all the enabled attributes run contiguously.
+ */
+
+
+struct draw_stage;
+struct llvmpipe_context;
+
+
+typedef void (*vbuf_draw_func)( struct pipe_context *pipe,
+                                unsigned prim,
+                                const ushort *elements,
+                                unsigned nr_elements,
+                                const void *vertex_buffer,
+                                unsigned nr_vertices );
+
+
+extern struct draw_stage *
+lp_draw_render_stage( struct llvmpipe_context *llvmpipe );
+
+extern struct setup_context *
+lp_draw_setup_context( struct draw_stage * );
+
+extern void
+lp_draw_flush( struct draw_stage * );
+
+
+extern struct draw_stage *
+lp_draw_vbuf_stage( struct draw_context *draw_context,
+                    struct pipe_context *pipe,
+                    vbuf_draw_func draw );
+
+
+#endif /* LP_PRIM_SETUP_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_prim_vbuf.c b/src/gallium/drivers/llvmpipe/lp_prim_vbuf.c
new file mode 100644
index 0000000000..c394dcb61d
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_prim_vbuf.c
@@ -0,0 +1,607 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Interface between 'draw' module's output and the llvmpipe rasterizer/setup
+ * code.  When the 'draw' module has finished filling a vertex buffer, the
+ * draw_arrays() functions below will be called.  Loop over the vertices and
+ * call the point/line/tri setup functions.
+ *
+ * Authors
+ *  Brian Paul
+ */
+
+
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_prim_vbuf.h"
+#include "lp_prim_setup.h"
+#include "lp_setup.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vbuf.h"
+#include "util/u_memory.h"
+#include "util/u_prim.h"
+
+
+#define LP_MAX_VBUF_INDEXES 1024
+#define LP_MAX_VBUF_SIZE    4096
+
+typedef const float (*cptrf4)[4];
+
+/**
+ * Subclass of vbuf_render.
+ */
+struct llvmpipe_vbuf_render
+{
+   struct vbuf_render base;
+   struct llvmpipe_context *llvmpipe;
+   uint prim;
+   uint vertex_size;
+   uint nr_vertices;
+   uint vertex_buffer_size;
+   void *vertex_buffer;
+};
+
+
+/** cast wrapper */
+static struct llvmpipe_vbuf_render *
+llvmpipe_vbuf_render(struct vbuf_render *vbr)
+{
+   return (struct llvmpipe_vbuf_render *) vbr;
+}
+
+
+static const struct vertex_info *
+lp_vbuf_get_vertex_info(struct vbuf_render *vbr)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   return llvmpipe_get_vbuf_vertex_info(cvbr->llvmpipe);
+}
+
+
+static boolean
+lp_vbuf_allocate_vertices(struct vbuf_render *vbr,
+                          ushort vertex_size, ushort nr_vertices)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   unsigned size = vertex_size * nr_vertices;
+
+   if (cvbr->vertex_buffer_size < size) {
+      align_free(cvbr->vertex_buffer);
+      cvbr->vertex_buffer = align_malloc(size, 16);
+      cvbr->vertex_buffer_size = size;
+   }
+
+   cvbr->vertex_size = vertex_size;
+   cvbr->nr_vertices = nr_vertices;
+   
+   return cvbr->vertex_buffer != NULL;
+}
+
+static void
+lp_vbuf_release_vertices(struct vbuf_render *vbr)
+{
+#if 0
+   {
+      struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+      const struct vertex_info *info = 
+         llvmpipe_get_vbuf_vertex_info(cvbr->llvmpipe);
+      const float *vtx = (const float *) cvbr->vertex_buffer;
+      uint i, j;
+      debug_printf("%s (vtx_size = %u,  vtx_used = %u)\n",
+             __FUNCTION__, cvbr->vertex_size, cvbr->nr_vertices);
+      for (i = 0; i < cvbr->nr_vertices; i++) {
+         for (j = 0; j < info->num_attribs; j++) {
+            uint k;
+            switch (info->attrib[j].emit) {
+            case EMIT_4F:  k = 4;   break;
+            case EMIT_3F:  k = 3;   break;
+            case EMIT_2F:  k = 2;   break;
+            case EMIT_1F:  k = 1;   break;
+            default: assert(0);
+            }
+            debug_printf("Vert %u attr %u: ", i, j);
+            while (k-- > 0) {
+               debug_printf("%g ", vtx[0]);
+               vtx++;
+            }
+            debug_printf("\n");
+         }
+      }
+   }
+#endif
+
+   /* keep the old allocation for next time */
+}
+
+static void *
+lp_vbuf_map_vertices(struct vbuf_render *vbr)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   return cvbr->vertex_buffer;
+}
+
+static void 
+lp_vbuf_unmap_vertices(struct vbuf_render *vbr, 
+                       ushort min_index,
+                       ushort max_index )
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   assert( cvbr->vertex_buffer_size >= (max_index+1) * cvbr->vertex_size );
+   /* do nothing */
+}
+
+
+static boolean
+lp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+
+   /* XXX: break this dependency - make setup_context live under
+    * llvmpipe, rename the old "setup" draw stage to something else.
+    */
+   struct setup_context *setup_ctx = lp_draw_setup_context(cvbr->llvmpipe->setup);
+   
+   llvmpipe_setup_prepare( setup_ctx );
+
+   cvbr->llvmpipe->reduced_prim = u_reduced_prim(prim);
+   cvbr->prim = prim;
+   return TRUE;
+
+}
+
+
+static INLINE cptrf4 get_vert( const void *vertex_buffer,
+                               int index,
+                               int stride )
+{
+   return (cptrf4)((char *)vertex_buffer + index * stride);
+}
+
+
+/**
+ * draw elements / indexed primitives
+ */
+static void
+lp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   struct llvmpipe_context *llvmpipe = cvbr->llvmpipe;
+   const unsigned stride = llvmpipe->vertex_info_vbuf.size * sizeof(float);
+   const void *vertex_buffer = cvbr->vertex_buffer;
+   unsigned i;
+
+   /* XXX: break this dependency - make setup_context live under
+    * llvmpipe, rename the old "setup" draw stage to something else.
+    */
+   struct draw_stage *setup = llvmpipe->setup;
+   struct setup_context *setup_ctx = lp_draw_setup_context(setup);
+
+   switch (cvbr->prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < nr; i++) {
+         llvmpipe_setup_point( setup_ctx,
+                      get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 1; i < nr; i += 2) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      for (i = 1; i < nr; i ++) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      for (i = 1; i < nr; i ++) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      if (nr) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[nr-1], stride),
+                     get_vert(vertex_buffer, indices[0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i += 3) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i += 3) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUADS:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 3; i < nr; i += 4) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+      }
+      else {
+         for (i = 3; i < nr; i += 4) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 3; i < nr; i += 2) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride));
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+      }
+      else {
+         for (i = 3; i < nr; i += 2) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      /* Almost same as tri fan but the _first_ vertex specifies the flat
+       * shading color.  Note that the first polygon vertex is passed as
+       * the last triangle vertex here.
+       * flatshade_first state makes no difference.
+       */
+      for (i = 2; i < nr; i += 1) {
+         llvmpipe_setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-0], stride),
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[0], stride) );
+      }
+      break;
+
+   default:
+      assert(0);
+   }
+
+   /* XXX: why are we calling this???  If we had to call something, it
+    * would be a function in lp_setup.c:
+    */
+   lp_draw_flush( setup );
+}
+
+
+/**
+ * This function is hit when the draw module is working in pass-through mode.
+ * It's up to us to convert the vertex array into point/line/tri prims.
+ */
+static void
+lp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   struct llvmpipe_context *llvmpipe = cvbr->llvmpipe;
+   const unsigned stride = llvmpipe->vertex_info_vbuf.size * sizeof(float);
+   const void *vertex_buffer =
+      (void *) get_vert(cvbr->vertex_buffer, start, stride);
+   unsigned i;
+
+   /* XXX: break this dependency - make setup_context live under
+    * llvmpipe, rename the old "setup" draw stage to something else.
+    */
+   struct draw_stage *setup = llvmpipe->setup;
+   struct setup_context *setup_ctx = lp_draw_setup_context(setup);
+
+   switch (cvbr->prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < nr; i++) {
+         llvmpipe_setup_point( setup_ctx,
+                      get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 1; i < nr; i += 2) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      for (i = 1; i < nr; i ++) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      for (i = 1; i < nr; i ++) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      if (nr) {
+         llvmpipe_setup_line( setup_ctx,
+                     get_vert(vertex_buffer, nr-1, stride),
+                     get_vert(vertex_buffer, 0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i += 3) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i += 3) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i++) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-(i&1), stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i++) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-2, stride),
+                       get_vert(vertex_buffer, i-(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride) );
+         }
+      }
+      else {
+         for (i = 2; i < nr; i += 1) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUADS:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 3; i < nr; i += 4) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+      }
+      else {
+         for (i = 3; i < nr; i += 4) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      if (llvmpipe->rasterizer->flatshade_first) {
+         for (i = 3; i < nr; i += 2) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+      }
+      else {
+         for (i = 3; i < nr; i += 2) {
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            llvmpipe_setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      /* Almost same as tri fan but the _first_ vertex specifies the flat
+       * shading color.  Note that the first polygon vertex is passed as
+       * the last triangle vertex here.
+       * flatshade_first state makes no difference.
+       */
+      for (i = 2; i < nr; i += 1) {
+         llvmpipe_setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-1, stride),
+                    get_vert(vertex_buffer, i-0, stride),
+                    get_vert(vertex_buffer, 0, stride) );
+      }
+      break;
+
+   default:
+      assert(0);
+   }
+}
+
+
+
+static void
+lp_vbuf_destroy(struct vbuf_render *vbr)
+{
+   struct llvmpipe_vbuf_render *cvbr = llvmpipe_vbuf_render(vbr);
+   cvbr->llvmpipe->vbuf_render = NULL;
+   FREE(cvbr);
+}
+
+
+/**
+ * Initialize the post-transform vertex buffer information for the given
+ * context.
+ */
+void
+lp_init_vbuf(struct llvmpipe_context *lp)
+{
+   assert(lp->draw);
+
+   lp->vbuf_render = CALLOC_STRUCT(llvmpipe_vbuf_render);
+
+   lp->vbuf_render->base.max_indices = LP_MAX_VBUF_INDEXES;
+   lp->vbuf_render->base.max_vertex_buffer_bytes = LP_MAX_VBUF_SIZE;
+
+   lp->vbuf_render->base.get_vertex_info = lp_vbuf_get_vertex_info;
+   lp->vbuf_render->base.allocate_vertices = lp_vbuf_allocate_vertices;
+   lp->vbuf_render->base.map_vertices = lp_vbuf_map_vertices;
+   lp->vbuf_render->base.unmap_vertices = lp_vbuf_unmap_vertices;
+   lp->vbuf_render->base.set_primitive = lp_vbuf_set_primitive;
+   lp->vbuf_render->base.draw = lp_vbuf_draw;
+   lp->vbuf_render->base.draw_arrays = lp_vbuf_draw_arrays;
+   lp->vbuf_render->base.release_vertices = lp_vbuf_release_vertices;
+   lp->vbuf_render->base.destroy = lp_vbuf_destroy;
+
+   lp->vbuf_render->llvmpipe = lp;
+
+   lp->vbuf = draw_vbuf_stage(lp->draw, &lp->vbuf_render->base);
+
+   draw_set_rasterize_stage(lp->draw, lp->vbuf);
+
+   draw_set_render(lp->draw, &lp->vbuf_render->base);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_prim_vbuf.h b/src/gallium/drivers/llvmpipe/lp_prim_vbuf.h
new file mode 100644
index 0000000000..6c4e6063e6
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_prim_vbuf.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_VBUF_H
+#define LP_VBUF_H
+
+
+struct llvmpipe_context;
+
+extern void
+lp_init_vbuf(struct llvmpipe_context *llvmpipe);
+
+
+#endif /* LP_VBUF_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_quad.h b/src/gallium/drivers/llvmpipe/lp_quad.h
new file mode 100644
index 0000000000..7eb05de77a
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_quad.h
@@ -0,0 +1,114 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef LP_QUAD_H
+#define LP_QUAD_H
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_exec.h"
+
+
+#define QUAD_PRIM_POINT 1
+#define QUAD_PRIM_LINE  2
+#define QUAD_PRIM_TRI   3
+
+
+/* The rasterizer generates 2x2 quads of fragment and feeds them to
+ * the current fp_machine (see below).
+ * Remember that Y=0=top with Y increasing down the window.
+ */
+#define QUAD_TOP_LEFT     0
+#define QUAD_TOP_RIGHT    1
+#define QUAD_BOTTOM_LEFT  2
+#define QUAD_BOTTOM_RIGHT 3
+
+#define MASK_TOP_LEFT     (1 << QUAD_TOP_LEFT)
+#define MASK_TOP_RIGHT    (1 << QUAD_TOP_RIGHT)
+#define MASK_BOTTOM_LEFT  (1 << QUAD_BOTTOM_LEFT)
+#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT)
+#define MASK_ALL          0xf
+
+
+/**
+ * Quad stage inputs (pos, coverage, front/back face, etc)
+ */
+struct quad_header_input
+{
+   int x0, y0;                /**< quad window pos, always even */
+   float coverage[QUAD_SIZE]; /**< fragment coverage for antialiasing */
+   unsigned facing:1;         /**< Front (0) or back (1) facing? */
+   unsigned prim:2;           /**< QUAD_PRIM_POINT, LINE, TRI */
+};
+
+
+/**
+ * Quad stage inputs/outputs.
+ */
+struct quad_header_inout
+{
+   unsigned mask:4;
+};
+
+
+/**
+ * Quad stage outputs (color & depth).
+ */
+struct quad_header_output
+{
+   /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */
+   float ALIGN16_ATTRIB color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE];
+};
+
+
+/**
+ * Input interpolation coefficients
+ */
+struct quad_interp_coef
+{
+   float ALIGN16_ATTRIB a0[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+   float ALIGN16_ATTRIB dadx[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+   float ALIGN16_ATTRIB dady[1 + PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
+};
+
+
+/**
+ * Encodes everything we need to know about a 2x2 pixel block.  Uses
+ * "Channel-Serial" or "SoA" layout.  
+ */
+struct quad_header {
+   struct quad_header_input input;
+   struct quad_header_inout inout;
+
+   /* Redundant/duplicated:
+    */
+   const struct quad_interp_coef *coef;
+};
+
+#endif /* LP_QUAD_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_query.c b/src/gallium/drivers/llvmpipe/lp_query.c
new file mode 100644
index 0000000000..5554285425
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_query.c
@@ -0,0 +1,111 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "lp_context.h"
+#include "lp_query.h"
+#include "lp_state.h"
+
+struct llvmpipe_query {
+   uint64_t start;
+   uint64_t end;
+};
+
+
+static struct llvmpipe_query *llvmpipe_query( struct pipe_query *p )
+{
+   return (struct llvmpipe_query *)p;
+}
+
+static struct pipe_query *
+llvmpipe_create_query(struct pipe_context *pipe, 
+		      unsigned type)
+{
+   assert(type == PIPE_QUERY_OCCLUSION_COUNTER);
+   return (struct pipe_query *)CALLOC_STRUCT( llvmpipe_query );
+}
+
+
+static void
+llvmpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   FREE(q);
+}
+
+
+static void
+llvmpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
+   struct llvmpipe_query *sq = llvmpipe_query(q);
+   
+   sq->start = llvmpipe->occlusion_count;
+   llvmpipe->active_query_count++;
+   llvmpipe->dirty |= LP_NEW_QUERY;
+}
+
+
+static void
+llvmpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
+   struct llvmpipe_query *sq = llvmpipe_query(q);
+
+   llvmpipe->active_query_count--;
+   sq->end = llvmpipe->occlusion_count;
+   llvmpipe->dirty |= LP_NEW_QUERY;
+}
+
+
+static boolean
+llvmpipe_get_query_result(struct pipe_context *pipe, 
+			  struct pipe_query *q,
+			  boolean wait,
+			  uint64_t *result )
+{
+   struct llvmpipe_query *sq = llvmpipe_query(q);
+   *result = sq->end - sq->start;
+   return TRUE;
+}
+
+
+void llvmpipe_init_query_funcs(struct llvmpipe_context *llvmpipe )
+{
+   llvmpipe->pipe.create_query = llvmpipe_create_query;
+   llvmpipe->pipe.destroy_query = llvmpipe_destroy_query;
+   llvmpipe->pipe.begin_query = llvmpipe_begin_query;
+   llvmpipe->pipe.end_query = llvmpipe_end_query;
+   llvmpipe->pipe.get_query_result = llvmpipe_get_query_result;
+}
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_query.h b/src/gallium/drivers/llvmpipe/lp_query.h
new file mode 100644
index 0000000000..fa9fcd8713
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_query.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell
+ */
+
+#ifndef LP_QUERY_H
+#define LP_QUERY_H
+
+struct llvmpipe_context;
+extern void llvmpipe_init_query_funcs(struct llvmpipe_context * );
+
+
+#endif /* LP_QUERY_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
new file mode 100644
index 0000000000..125035771e
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -0,0 +1,243 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_simple_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+
+#include "lp_texture.h"
+#include "lp_buffer.h"
+#include "lp_winsys.h"
+#include "lp_jit.h"
+#include "lp_screen.h"
+
+
+static const char *
+llvmpipe_get_vendor(struct pipe_screen *screen)
+{
+   return "VMware, Inc.";
+}
+
+
+static const char *
+llvmpipe_get_name(struct pipe_screen *screen)
+{
+   return "llvmpipe";
+}
+
+
+static int
+llvmpipe_get_param(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+      return PIPE_MAX_SAMPLERS;
+   case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+      return PIPE_MAX_SAMPLERS;
+   case PIPE_CAP_NPOT_TEXTURES:
+      return 1;
+   case PIPE_CAP_TWO_SIDED_STENCIL:
+      return 1;
+   case PIPE_CAP_GLSL:
+      return 1;
+   case PIPE_CAP_S3TC:
+      return 0;
+   case PIPE_CAP_ANISOTROPIC_FILTER:
+      return 0;
+   case PIPE_CAP_POINT_SPRITE:
+      return 1;
+   case PIPE_CAP_MAX_RENDER_TARGETS:
+      return PIPE_MAX_COLOR_BUFS;
+   case PIPE_CAP_OCCLUSION_QUERY:
+      return 1;
+   case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+      return 1;
+   case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+      return 1;
+   case PIPE_CAP_TEXTURE_SHADOW_MAP:
+      return 1;
+   case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+      return 13; /* max 4Kx4K */
+   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+      return 8;  /* max 128x128x128 */
+   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+      return 13; /* max 4Kx4K */
+   case PIPE_CAP_TGSI_CONT_SUPPORTED:
+      return 1;
+   case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+      return 1;
+   default:
+      return 0;
+   }
+}
+
+
+static float
+llvmpipe_get_paramf(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_LINE_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_LINE_WIDTH_AA:
+      return 255.0; /* arbitrary */
+   case PIPE_CAP_MAX_POINT_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_POINT_WIDTH_AA:
+      return 255.0; /* arbitrary */
+   case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+      return 16.0; /* not actually signficant at this time */
+   case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+      return 16.0; /* arbitrary */
+   default:
+      return 0;
+   }
+}
+
+
+/**
+ * Query format support for creating a texture, drawing surface, etc.
+ * \param format  the format to test
+ * \param type  one of PIPE_TEXTURE, PIPE_SURFACE
+ */
+static boolean
+llvmpipe_is_format_supported( struct pipe_screen *_screen,
+                              enum pipe_format format, 
+                              enum pipe_texture_target target,
+                              unsigned tex_usage, 
+                              unsigned geom_flags )
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+   struct llvmpipe_winsys *winsys = screen->winsys;
+
+   assert(target == PIPE_TEXTURE_1D ||
+          target == PIPE_TEXTURE_2D ||
+          target == PIPE_TEXTURE_3D ||
+          target == PIPE_TEXTURE_CUBE);
+
+   if(format == PIPE_FORMAT_Z16_UNORM)
+      return FALSE;
+   if(format == PIPE_FORMAT_S8_UNORM)
+      return FALSE;
+
+   switch(format) {
+   case PIPE_FORMAT_DXT1_RGB:
+   case PIPE_FORMAT_DXT1_RGBA:
+   case PIPE_FORMAT_DXT3_RGBA:
+   case PIPE_FORMAT_DXT5_RGBA:
+      return FALSE;
+   default:
+      break;
+   }
+
+   if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
+      return winsys->is_displaytarget_format_supported(winsys, format);
+
+   return TRUE;
+}
+
+
+static struct pipe_buffer *
+llvmpipe_surface_buffer_create(struct pipe_screen *screen,
+                               unsigned width, unsigned height,
+                               enum pipe_format format,
+                               unsigned tex_usage,
+                               unsigned usage,
+                               unsigned *stride)
+{
+   /* This function should never be used */
+   assert(0);
+   return NULL;
+}
+
+
+static void
+llvmpipe_flush_frontbuffer(struct pipe_screen *_screen,
+                           struct pipe_surface *surface,
+                           void *context_private)
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+   struct llvmpipe_winsys *winsys = screen->winsys;
+   struct llvmpipe_texture *texture = llvmpipe_texture(surface->texture);
+
+   assert(texture->dt);
+   if (texture->dt)
+      winsys->displaytarget_display(winsys, texture->dt, context_private);
+}
+
+
+static void
+llvmpipe_destroy_screen( struct pipe_screen *_screen )
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+
+   struct pipe_winsys *winsys = _screen->winsys;
+
+   lp_jit_screen_cleanup(screen);
+
+   if(winsys->destroy)
+      winsys->destroy(winsys);
+
+   FREE(screen);
+}
+
+
+
+/**
+ * Create a new pipe_screen object
+ * Note: we're not presently subclassing pipe_screen (no llvmpipe_screen).
+ */
+struct pipe_screen *
+llvmpipe_create_screen(struct llvmpipe_winsys *winsys)
+{
+   struct llvmpipe_screen *screen = CALLOC_STRUCT(llvmpipe_screen);
+
+   if (!screen)
+      return NULL;
+
+   screen->winsys = winsys;
+
+   screen->base.destroy = llvmpipe_destroy_screen;
+
+   screen->base.get_name = llvmpipe_get_name;
+   screen->base.get_vendor = llvmpipe_get_vendor;
+   screen->base.get_param = llvmpipe_get_param;
+   screen->base.get_paramf = llvmpipe_get_paramf;
+   screen->base.is_format_supported = llvmpipe_is_format_supported;
+
+   screen->base.surface_buffer_create = llvmpipe_surface_buffer_create;
+   screen->base.flush_frontbuffer = llvmpipe_flush_frontbuffer;
+
+   llvmpipe_init_screen_texture_funcs(&screen->base);
+   llvmpipe_init_screen_buffer_funcs(&screen->base);
+
+   lp_jit_screen_init(screen);
+
+   return &screen->base;
+}
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c b/src/gallium/drivers/llvmpipe/lp_screen.h
index 92f5bd09c9..4a1b4d6f3e 100644
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.h
@@ -1,8 +1,9 @@
 /**************************************************************************
  * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -25,49 +26,54 @@
  * 
  **************************************************************************/
 
-#include <pipe/p_thread.h>
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_drmif.h"
+/**
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ * @author Keith Whitwell <keith@tungstengraphics.com>
+ */
 
-pipe_static_mutex(lockMutex);
+#ifndef LP_SCREEN_H
+#define LP_SCREEN_H
 
-/* Lock the hardware and validate our state.
- */
-void
-LOCK_HARDWARE(struct nouveau_context *nv)
-{
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
-	char __ret=0;
-
-	assert(!nv->locked);
-	pipe_mutex_lock(lockMutex);
-
-	DRM_CAS(nvdev->lock, nvdev->ctx,
-		(DRM_LOCK_HELD | nvdev->ctx), __ret);
-
-	if (__ret) {
-		drmGetLock(nvdev->fd, nvdev->ctx, 0);
-		nouveau_contended_lock(nv);
-	}
-	nv->locked = 1;
-}
+#include <llvm-c/Core.h>
+#include <llvm-c/Analysis.h>
+#include <llvm-c/Target.h>
+#include <llvm-c/ExecutionEngine.h>
 
-/* Unlock the hardware using the global current context 
- */
-void
-UNLOCK_HARDWARE(struct nouveau_context *nv)
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+
+
+struct llvmpipe_winsys;
+
+
+struct llvmpipe_screen
 {
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
+   struct pipe_screen base;
 
-	assert(nv->locked);
-	nv->locked = 0;
+   struct llvmpipe_winsys *winsys;
+
+   LLVMModuleRef module;
+   LLVMExecutionEngineRef engine;
+   LLVMModuleProviderRef provider;
+   LLVMTargetDataRef target;
+   LLVMPassManagerRef pass;
+
+   LLVMTypeRef context_ptr_type;
+
+   /* Increments whenever textures are modified.  Contexts can track
+    * this.
+    */
+   unsigned timestamp;          
+};
+
+
+
+
+static INLINE struct llvmpipe_screen *
+llvmpipe_screen( struct pipe_screen *pipe )
+{
+   return (struct llvmpipe_screen *)pipe;
+}
 
-	DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx);
 
-	pipe_mutex_unlock(lockMutex);
-} 
+#endif /* LP_SCREEN_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
new file mode 100644
index 0000000000..d145f6d6bb
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -0,0 +1,1482 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \brief  Primitive rasterization/rendering (points, lines, triangles)
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ * \author  Brian Paul
+ */
+
+#include "lp_context.h"
+#include "lp_prim_setup.h"
+#include "lp_quad.h"
+#include "lp_setup.h"
+#include "lp_state.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vertex.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_thread.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "lp_tile_cache.h"
+#include "lp_tile_soa.h"
+
+
+#define DEBUG_VERTS 0
+#define DEBUG_FRAGS 0
+
+/**
+ * Triangle edge info
+ */
+struct edge {
+   float dx;		/**< X(v1) - X(v0), used only during setup */
+   float dy;		/**< Y(v1) - Y(v0), used only during setup */
+   float dxdy;		/**< dx/dy */
+   float sx, sy;	/**< first sample point coord */
+   int lines;		/**< number of lines on this edge */
+};
+
+
+#define MAX_QUADS 16
+
+
+/**
+ * Triangle setup info (derived from draw_stage).
+ * Also used for line drawing (taking some liberties).
+ */
+struct setup_context {
+   struct llvmpipe_context *llvmpipe;
+
+   /* Vertices are just an array of floats making up each attribute in
+    * turn.  Currently fixed at 4 floats, but should change in time.
+    * Codegen will help cope with this.
+    */
+   const float (*vmax)[4];
+   const float (*vmid)[4];
+   const float (*vmin)[4];
+   const float (*vprovoke)[4];
+
+   struct edge ebot;
+   struct edge etop;
+   struct edge emaj;
+
+   float oneoverarea;
+   int facing;
+
+   struct quad_header quad[MAX_QUADS];
+   struct quad_header *quad_ptrs[MAX_QUADS];
+   unsigned count;
+
+   struct quad_interp_coef coef;
+
+   struct {
+      int left[2];   /**< [0] = row0, [1] = row1 */
+      int right[2];
+      int y;
+   } span;
+
+#if DEBUG_FRAGS
+   uint numFragsEmitted;  /**< per primitive */
+   uint numFragsWritten;  /**< per primitive */
+#endif
+
+   unsigned winding;		/* which winding to cull */
+};
+
+
+
+/**
+ * Execute fragment shader for the four fragments in the quad.
+ */
+static void
+shade_quads(struct llvmpipe_context *llvmpipe,
+            struct quad_header *quads[],
+            unsigned nr)
+{
+   struct lp_fragment_shader *fs = llvmpipe->fs;
+   struct quad_header *quad = quads[0];
+   const unsigned x = quad->input.x0;
+   const unsigned y = quad->input.y0;
+   uint8_t *tile = lp_get_cached_tile(llvmpipe->cbuf_cache[0], x, y);
+   uint8_t *color;
+   void *depth;
+   uint32_t ALIGN16_ATTRIB mask[4][NUM_CHANNELS];
+   unsigned chan_index;
+   unsigned q;
+
+   assert(fs->current);
+   if(!fs->current)
+      return;
+
+   /* Sanity checks */
+   assert(nr * QUAD_SIZE == TILE_VECTOR_HEIGHT * TILE_VECTOR_WIDTH);
+   assert(x % TILE_VECTOR_WIDTH == 0);
+   assert(y % TILE_VECTOR_HEIGHT == 0);
+   for (q = 0; q < nr; ++q) {
+      assert(quads[q]->input.x0 == x + q*2);
+      assert(quads[q]->input.y0 == y);
+   }
+
+   /* mask */
+   for (q = 0; q < 4; ++q)
+      for (chan_index = 0; chan_index < NUM_CHANNELS; ++chan_index)
+         mask[q][chan_index] = quads[q]->inout.mask & (1 << chan_index) ? ~0 : 0;
+
+   /* color buffer */
+   color = &TILE_PIXEL(tile, x & (TILE_SIZE-1), y & (TILE_SIZE-1), 0);
+
+   /* depth buffer */
+   if(llvmpipe->zsbuf_map) {
+      assert((x % 2) == 0);
+      assert((y % 2) == 0);
+      depth = llvmpipe->zsbuf_map +
+              y*llvmpipe->zsbuf_transfer->stride +
+              2*x*llvmpipe->zsbuf_transfer->block.size;
+   }
+   else
+      depth = NULL;
+
+   /* TODO: blend color */
+
+   assert((((uintptr_t)mask) & 0xf) == 0);
+   assert((((uintptr_t)depth) & 0xf) == 0);
+   assert((((uintptr_t)color) & 0xf) == 0);
+   assert((((uintptr_t)llvmpipe->jit_context.blend_color) & 0xf) == 0);
+
+   /* run shader */
+   fs->current->jit_function( &llvmpipe->jit_context,
+                              x, y,
+                              quad->coef->a0,
+                              quad->coef->dadx,
+                              quad->coef->dady,
+                              &mask[0][0],
+                              color,
+                              depth);
+}
+
+
+
+
+/**
+ * Do triangle cull test using tri determinant (sign indicates orientation)
+ * \return true if triangle is to be culled.
+ */
+static INLINE boolean
+cull_tri(const struct setup_context *setup, float det)
+{
+   if (det != 0) {   
+      /* if (det < 0 then Z points toward camera and triangle is 
+       * counter-clockwise winding.
+       */
+      unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW;
+
+      if ((winding & setup->winding) == 0)
+	 return FALSE;
+   }
+
+   /* Culled:
+    */
+   return TRUE;
+}
+
+
+
+/**
+ * Clip setup->quad against the scissor/surface bounds.
+ */
+static INLINE void
+quad_clip( struct setup_context *setup, struct quad_header *quad )
+{
+   const struct pipe_scissor_state *cliprect = &setup->llvmpipe->cliprect;
+   const int minx = (int) cliprect->minx;
+   const int maxx = (int) cliprect->maxx;
+   const int miny = (int) cliprect->miny;
+   const int maxy = (int) cliprect->maxy;
+
+   if (quad->input.x0 >= maxx ||
+       quad->input.y0 >= maxy ||
+       quad->input.x0 + 1 < minx ||
+       quad->input.y0 + 1 < miny) {
+      /* totally clipped */
+      quad->inout.mask = 0x0;
+      return;
+   }
+   if (quad->input.x0 < minx)
+      quad->inout.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+   if (quad->input.y0 < miny)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+   if (quad->input.x0 == maxx - 1)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+   if (quad->input.y0 == maxy - 1)
+      quad->inout.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+}
+
+
+
+/**
+ * Given an X or Y coordinate, return the block/quad coordinate that it
+ * belongs to.
+ */
+static INLINE int block( int x )
+{
+   return x & ~(2-1);
+}
+
+static INLINE int block_x( int x )
+{
+   return x & ~(TILE_VECTOR_WIDTH - 1);
+}
+
+
+/**
+ * Emit a quad (pass to next stage) with clipping.
+ */
+static INLINE void
+clip_emit_quad( struct setup_context *setup, struct quad_header *quad )
+{
+   quad_clip( setup, quad );
+
+   if (quad->inout.mask) {
+      struct llvmpipe_context *lp = setup->llvmpipe;
+
+#if 1
+      /* XXX: The blender expects 4 quads. This is far from efficient, but
+       * until we codegenerate single-quad variants of the fragment pipeline
+       * we need this hack. */
+      const unsigned nr_quads = TILE_VECTOR_HEIGHT*TILE_VECTOR_WIDTH/QUAD_SIZE;
+      struct quad_header quads[nr_quads];
+      struct quad_header *quad_ptrs[nr_quads];
+      int x0 = block_x(quad->input.x0);
+      unsigned i;
+
+      for(i = 0; i < nr_quads; ++i) {
+         int x = x0 + 2*i;
+         if(x == quad->input.x0)
+            memcpy(&quads[i], quad, sizeof quads[i]);
+         else {
+            memset(&quads[i], 0, sizeof quads[i]);
+            quads[i].input.x0 = x;
+            quads[i].input.y0 = quad->input.y0;
+            quads[i].coef = quad->coef;
+         }
+         quad_ptrs[i] = &quads[i];
+      }
+
+      shade_quads( lp, quad_ptrs, nr_quads );
+#else
+      shade_quads( lp, &quad, 1 );
+#endif
+   }
+}
+
+
+/**
+ * Render a horizontal span of quads
+ */
+static void flush_spans( struct setup_context *setup )
+{
+   const int step = TILE_VECTOR_WIDTH;
+   const int xleft0 = setup->span.left[0];
+   const int xleft1 = setup->span.left[1];
+   const int xright0 = setup->span.right[0];
+   const int xright1 = setup->span.right[1];
+
+
+   int minleft = block_x(MIN2(xleft0, xleft1));
+   int maxright = MAX2(xright0, xright1);
+   int x;
+
+   for (x = minleft; x < maxright; x += step) {
+      unsigned skip_left0 = CLAMP(xleft0 - x, 0, step);
+      unsigned skip_left1 = CLAMP(xleft1 - x, 0, step);
+      unsigned skip_right0 = CLAMP(x + step - xright0, 0, step);
+      unsigned skip_right1 = CLAMP(x + step - xright1, 0, step);
+      unsigned lx = x;
+      const unsigned nr_quads = TILE_VECTOR_HEIGHT*TILE_VECTOR_WIDTH/QUAD_SIZE;
+      unsigned q = 0;
+
+      unsigned skipmask_left0 = (1U << skip_left0) - 1U;
+      unsigned skipmask_left1 = (1U << skip_left1) - 1U;
+
+      /* These calculations fail when step == 32 and skip_right == 0.
+       */
+      unsigned skipmask_right0 = ~0U << (unsigned)(step - skip_right0);
+      unsigned skipmask_right1 = ~0U << (unsigned)(step - skip_right1);
+
+      unsigned mask0 = ~skipmask_left0 & ~skipmask_right0;
+      unsigned mask1 = ~skipmask_left1 & ~skipmask_right1;
+
+      if (mask0 | mask1) {
+         for(q = 0; q < nr_quads; ++q) {
+            unsigned quadmask = (mask0 & 3) | ((mask1 & 3) << 2);
+            setup->quad[q].input.x0 = lx;
+            setup->quad[q].input.y0 = setup->span.y;
+            setup->quad[q].inout.mask = quadmask;
+            setup->quad_ptrs[q] = &setup->quad[q];
+            mask0 >>= 2;
+            mask1 >>= 2;
+            lx += 2;
+         }
+         assert(!(mask0 | mask1));
+
+         shade_quads(setup->llvmpipe, setup->quad_ptrs, nr_quads );
+      }
+   }
+
+
+   setup->span.y = 0;
+   setup->span.right[0] = 0;
+   setup->span.right[1] = 0;
+   setup->span.left[0] = 1000000;     /* greater than right[0] */
+   setup->span.left[1] = 1000000;     /* greater than right[1] */
+}
+
+
+#if DEBUG_VERTS
+static void print_vertex(const struct setup_context *setup,
+                         const float (*v)[4])
+{
+   int i;
+   debug_printf("   Vertex: (%p)\n", v);
+   for (i = 0; i < setup->quad[0].nr_attrs; i++) {
+      debug_printf("     %d: %f %f %f %f\n",  i,
+              v[i][0], v[i][1], v[i][2], v[i][3]);
+      if (util_is_inf_or_nan(v[i][0])) {
+         debug_printf("   NaN!\n");
+      }
+   }
+}
+#endif
+
+/**
+ * Sort the vertices from top to bottom order, setting up the triangle
+ * edge fields (ebot, emaj, etop).
+ * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
+ */
+static boolean setup_sort_vertices( struct setup_context *setup,
+                                    float det,
+                                    const float (*v0)[4],
+                                    const float (*v1)[4],
+                                    const float (*v2)[4] )
+{
+   setup->vprovoke = v2;
+
+   /* determine bottom to top order of vertices */
+   {
+      float y0 = v0[0][1];
+      float y1 = v1[0][1];
+      float y2 = v2[0][1];
+      if (y0 <= y1) {
+	 if (y1 <= y2) {
+	    /* y0<=y1<=y2 */
+	    setup->vmin = v0;
+	    setup->vmid = v1;
+	    setup->vmax = v2;
+	 }
+	 else if (y2 <= y0) {
+	    /* y2<=y0<=y1 */
+	    setup->vmin = v2;
+	    setup->vmid = v0;
+	    setup->vmax = v1;
+	 }
+	 else {
+	    /* y0<=y2<=y1 */
+	    setup->vmin = v0;
+	    setup->vmid = v2;
+	    setup->vmax = v1;
+	 }
+      }
+      else {
+	 if (y0 <= y2) {
+	    /* y1<=y0<=y2 */
+	    setup->vmin = v1;
+	    setup->vmid = v0;
+	    setup->vmax = v2;
+	 }
+	 else if (y2 <= y1) {
+	    /* y2<=y1<=y0 */
+	    setup->vmin = v2;
+	    setup->vmid = v1;
+	    setup->vmax = v0;
+	 }
+	 else {
+	    /* y1<=y2<=y0 */
+	    setup->vmin = v1;
+	    setup->vmid = v2;
+	    setup->vmax = v0;
+	 }
+      }
+   }
+
+   setup->ebot.dx = setup->vmid[0][0] - setup->vmin[0][0];
+   setup->ebot.dy = setup->vmid[0][1] - setup->vmin[0][1];
+   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
+   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
+   setup->etop.dx = setup->vmax[0][0] - setup->vmid[0][0];
+   setup->etop.dy = setup->vmax[0][1] - setup->vmid[0][1];
+
+   /*
+    * Compute triangle's area.  Use 1/area to compute partial
+    * derivatives of attributes later.
+    *
+    * The area will be the same as prim->det, but the sign may be
+    * different depending on how the vertices get sorted above.
+    *
+    * To determine whether the primitive is front or back facing we
+    * use the prim->det value because its sign is correct.
+    */
+   {
+      const float area = (setup->emaj.dx * setup->ebot.dy -
+			    setup->ebot.dx * setup->emaj.dy);
+
+      setup->oneoverarea = 1.0f / area;
+
+      /*
+      debug_printf("%s one-over-area %f  area %f  det %f\n",
+                   __FUNCTION__, setup->oneoverarea, area, det );
+      */
+      if (util_is_inf_or_nan(setup->oneoverarea))
+         return FALSE;
+   }
+
+   /* We need to know if this is a front or back-facing triangle for:
+    *  - the GLSL gl_FrontFacing fragment attribute (bool)
+    *  - two-sided stencil test
+    */
+   setup->facing = 
+      ((det > 0.0) ^ 
+       (setup->llvmpipe->rasterizer->front_winding == PIPE_WINDING_CW));
+
+   return TRUE;
+}
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a triangle.
+ */
+static void tri_pos_coeff( struct setup_context *setup,
+                           uint vertSlot, unsigned i)
+{
+   float botda = setup->vmid[vertSlot][i] - setup->vmin[vertSlot][i];
+   float majda = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
+   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
+   float dadx = a * setup->oneoverarea;
+   float dady = b * setup->oneoverarea;
+
+   assert(i <= 3);
+
+   setup->coef.dadx[0][i] = dadx;
+   setup->coef.dady[0][i] = dady;
+
+   /* calculate a0 as the value which would be sampled for the
+    * fragment at (0,0), taking into account that we want to sample at
+    * pixel centers, in other words (0.5, 0.5).
+    *
+    * this is neat but unfortunately not a good way to do things for
+    * triangles with very large values of dadx or dady as it will
+    * result in the subtraction and re-addition from a0 of a very
+    * large number, which means we'll end up loosing a lot of the
+    * fractional bits and precision from a0.  the way to fix this is
+    * to define a0 as the sample at a pixel center somewhere near vmin
+    * instead - i'll switch to this later.
+    */
+   setup->coef.a0[0][i] = (setup->vmin[vertSlot][i] -
+                           (dadx * (setup->vmin[0][0] - 0.5f) +
+                            dady * (setup->vmin[0][1] - 0.5f)));
+
+   /*
+   debug_printf("attr[%d].%c: %f dx:%f dy:%f\n",
+                slot, "xyzw"[i],
+                setup->coef[slot].a0[i],
+                setup->coef[slot].dadx[i],
+                setup->coef[slot].dady[i]);
+   */
+}
+
+
+/**
+ * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
+ * The value value comes from vertex[slot][i].
+ * The result will be put into setup->coef[slot].a0[i].
+ * \param slot  which attribute slot
+ * \param i  which component of the slot (0..3)
+ */
+static void const_pos_coeff( struct setup_context *setup,
+                             uint vertSlot, unsigned i)
+{
+   setup->coef.dadx[0][i] = 0;
+   setup->coef.dady[0][i] = 0;
+
+   /* need provoking vertex info!
+    */
+   setup->coef.a0[0][i] = setup->vprovoke[vertSlot][i];
+}
+
+
+/**
+ * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
+ * The value value comes from vertex[slot][i].
+ * The result will be put into setup->coef[slot].a0[i].
+ * \param slot  which attribute slot
+ * \param i  which component of the slot (0..3)
+ */
+static void const_coeff( struct setup_context *setup,
+                         unsigned attrib,
+                         uint vertSlot)
+{
+   unsigned i;
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      setup->coef.dadx[1 + attrib][i] = 0;
+      setup->coef.dady[1 + attrib][i] = 0;
+
+      /* need provoking vertex info!
+       */
+      setup->coef.a0[1 + attrib][i] = setup->vprovoke[vertSlot][i];
+   }
+}
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a triangle.
+ */
+static void tri_linear_coeff( struct setup_context *setup,
+                              unsigned attrib,
+                              uint vertSlot)
+{
+   unsigned i;
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      float botda = setup->vmid[vertSlot][i] - setup->vmin[vertSlot][i];
+      float majda = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+      float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
+      float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
+      float dadx = a * setup->oneoverarea;
+      float dady = b * setup->oneoverarea;
+
+      assert(i <= 3);
+
+      setup->coef.dadx[1 + attrib][i] = dadx;
+      setup->coef.dady[1 + attrib][i] = dady;
+
+      /* calculate a0 as the value which would be sampled for the
+       * fragment at (0,0), taking into account that we want to sample at
+       * pixel centers, in other words (0.5, 0.5).
+       *
+       * this is neat but unfortunately not a good way to do things for
+       * triangles with very large values of dadx or dady as it will
+       * result in the subtraction and re-addition from a0 of a very
+       * large number, which means we'll end up loosing a lot of the
+       * fractional bits and precision from a0.  the way to fix this is
+       * to define a0 as the sample at a pixel center somewhere near vmin
+       * instead - i'll switch to this later.
+       */
+      setup->coef.a0[1 + attrib][i] = (setup->vmin[vertSlot][i] -
+                     (dadx * (setup->vmin[0][0] - 0.5f) +
+                      dady * (setup->vmin[0][1] - 0.5f)));
+
+      /*
+      debug_printf("attr[%d].%c: %f dx:%f dy:%f\n",
+                   slot, "xyzw"[i],
+                   setup->coef[slot].a0[i],
+                   setup->coef[slot].dadx[i],
+                   setup->coef[slot].dady[i]);
+      */
+   }
+}
+
+
+/**
+ * Compute a0, dadx and dady for a perspective-corrected interpolant,
+ * for a triangle.
+ * We basically multiply the vertex value by 1/w before computing
+ * the plane coefficients (a0, dadx, dady).
+ * Later, when we compute the value at a particular fragment position we'll
+ * divide the interpolated value by the interpolated W at that fragment.
+ */
+static void tri_persp_coeff( struct setup_context *setup,
+                             unsigned attrib,
+                             uint vertSlot)
+{
+   unsigned i;
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      /* premultiply by 1/w  (v[0][3] is always W):
+       */
+      float mina = setup->vmin[vertSlot][i] * setup->vmin[0][3];
+      float mida = setup->vmid[vertSlot][i] * setup->vmid[0][3];
+      float maxa = setup->vmax[vertSlot][i] * setup->vmax[0][3];
+      float botda = mida - mina;
+      float majda = maxa - mina;
+      float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
+      float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
+      float dadx = a * setup->oneoverarea;
+      float dady = b * setup->oneoverarea;
+
+      /*
+      debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i,
+                   setup->vmin[vertSlot][i],
+                   setup->vmid[vertSlot][i],
+                   setup->vmax[vertSlot][i]
+             );
+      */
+      assert(i <= 3);
+
+      setup->coef.dadx[1 + attrib][i] = dadx;
+      setup->coef.dady[1 + attrib][i] = dady;
+      setup->coef.a0[1 + attrib][i] = (mina -
+                     (dadx * (setup->vmin[0][0] - 0.5f) +
+                      dady * (setup->vmin[0][1] - 0.5f)));
+   }
+}
+
+
+/**
+ * Special coefficient setup for gl_FragCoord.
+ * X and Y are trivial, though Y has to be inverted for OpenGL.
+ * Z and W are copied from posCoef which should have already been computed.
+ * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask.
+ */
+static void
+setup_fragcoord_coeff(struct setup_context *setup, uint slot)
+{
+   /*X*/
+   setup->coef.a0[1 + slot][0] = 0;
+   setup->coef.dadx[1 + slot][0] = 1.0;
+   setup->coef.dady[1 + slot][0] = 0.0;
+   /*Y*/
+   setup->coef.a0[1 + slot][1] = 0.0;
+   setup->coef.dadx[1 + slot][1] = 0.0;
+   setup->coef.dady[1 + slot][1] = 1.0;
+   /*Z*/
+   setup->coef.a0[1 + slot][2] = setup->coef.a0[0][2];
+   setup->coef.dadx[1 + slot][2] = setup->coef.dadx[0][2];
+   setup->coef.dady[1 + slot][2] = setup->coef.dady[0][2];
+   /*W*/
+   setup->coef.a0[1 + slot][3] = setup->coef.a0[0][3];
+   setup->coef.dadx[1 + slot][3] = setup->coef.dadx[0][3];
+   setup->coef.dady[1 + slot][3] = setup->coef.dady[0][3];
+}
+
+
+
+/**
+ * Compute the setup->coef[] array dadx, dady, a0 values.
+ * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized.
+ */
+static void setup_tri_coefficients( struct setup_context *setup )
+{
+   struct llvmpipe_context *llvmpipe = setup->llvmpipe;
+   const struct lp_fragment_shader *lpfs = llvmpipe->fs;
+   const struct vertex_info *vinfo = llvmpipe_get_vertex_info(llvmpipe);
+   uint fragSlot;
+
+   /* z and w are done by linear interpolation:
+    */
+   tri_pos_coeff(setup, 0, 2);
+   tri_pos_coeff(setup, 0, 3);
+
+   /* setup interpolation for all the remaining attributes:
+    */
+   for (fragSlot = 0; fragSlot < lpfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         const_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_LINEAR:
+         tri_linear_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_PERSPECTIVE:
+         tri_persp_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (lpfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
+         setup->coef.a0[1 + fragSlot][0] = 1.0f - setup->facing;
+         setup->coef.dadx[1 + fragSlot][0] = 0.0;
+         setup->coef.dady[1 + fragSlot][0] = 0.0;
+      }
+   }
+}
+
+
+
+static void setup_tri_edges( struct setup_context *setup )
+{
+   float vmin_x = setup->vmin[0][0] + 0.5f;
+   float vmid_x = setup->vmid[0][0] + 0.5f;
+
+   float vmin_y = setup->vmin[0][1] - 0.5f;
+   float vmid_y = setup->vmid[0][1] - 0.5f;
+   float vmax_y = setup->vmax[0][1] - 0.5f;
+
+   setup->emaj.sy = ceilf(vmin_y);
+   setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy);
+   setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy;
+   setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy;
+
+   setup->etop.sy = ceilf(vmid_y);
+   setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy);
+   setup->etop.dxdy = setup->etop.dx / setup->etop.dy;
+   setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy;
+
+   setup->ebot.sy = ceilf(vmin_y);
+   setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy);
+   setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy;
+   setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy;
+}
+
+
+/**
+ * Render the upper or lower half of a triangle.
+ * Scissoring/cliprect is applied here too.
+ */
+static void subtriangle( struct setup_context *setup,
+			 struct edge *eleft,
+			 struct edge *eright,
+			 unsigned lines )
+{
+   const struct pipe_scissor_state *cliprect = &setup->llvmpipe->cliprect;
+   const int minx = (int) cliprect->minx;
+   const int maxx = (int) cliprect->maxx;
+   const int miny = (int) cliprect->miny;
+   const int maxy = (int) cliprect->maxy;
+   int y, start_y, finish_y;
+   int sy = (int)eleft->sy;
+
+   assert((int)eleft->sy == (int) eright->sy);
+
+   /* clip top/bottom */
+   start_y = sy;
+   if (start_y < miny)
+      start_y = miny;
+
+   finish_y = sy + lines;
+   if (finish_y > maxy)
+      finish_y = maxy;
+
+   start_y -= sy;
+   finish_y -= sy;
+
+   /*
+   debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y);
+   */
+
+   for (y = start_y; y < finish_y; y++) {
+
+      /* avoid accumulating adds as floats don't have the precision to
+       * accurately iterate large triangle edges that way.  luckily we
+       * can just multiply these days.
+       *
+       * this is all drowned out by the attribute interpolation anyway.
+       */
+      int left = (int)(eleft->sx + y * eleft->dxdy);
+      int right = (int)(eright->sx + y * eright->dxdy);
+
+      /* clip left/right */
+      if (left < minx)
+         left = minx;
+      if (right > maxx)
+         right = maxx;
+
+      if (left < right) {
+         int _y = sy + y;
+         if (block(_y) != setup->span.y) {
+            flush_spans(setup);
+            setup->span.y = block(_y);
+         }
+
+         setup->span.left[_y&1] = left;
+         setup->span.right[_y&1] = right;
+      }
+   }
+
+
+   /* save the values so that emaj can be restarted:
+    */
+   eleft->sx += lines * eleft->dxdy;
+   eright->sx += lines * eright->dxdy;
+   eleft->sy += lines;
+   eright->sy += lines;
+}
+
+
+/**
+ * Recalculate prim's determinant.  This is needed as we don't have
+ * get this information through the vbuf_render interface & we must
+ * calculate it here.
+ */
+static float
+calc_det( const float (*v0)[4],
+          const float (*v1)[4],
+          const float (*v2)[4] )
+{
+   /* edge vectors e = v0 - v2, f = v1 - v2 */
+   const float ex = v0[0][0] - v2[0][0];
+   const float ey = v0[0][1] - v2[0][1];
+   const float fx = v1[0][0] - v2[0][0];
+   const float fy = v1[0][1] - v2[0][1];
+
+   /* det = cross(e,f).z */
+   return ex * fy - ey * fx;
+}
+
+
+/**
+ * Do setup for triangle rasterization, then render the triangle.
+ */
+void llvmpipe_setup_tri( struct setup_context *setup,
+                const float (*v0)[4],
+                const float (*v1)[4],
+                const float (*v2)[4] )
+{
+   float det;
+
+#if DEBUG_VERTS
+   debug_printf("Setup triangle:\n");
+   print_vertex(setup, v0);
+   print_vertex(setup, v1);
+   print_vertex(setup, v2);
+#endif
+
+   if (setup->llvmpipe->no_rast)
+      return;
+   
+   det = calc_det(v0, v1, v2);
+   /*
+   debug_printf("%s\n", __FUNCTION__ );
+   */
+
+#if DEBUG_FRAGS
+   setup->numFragsEmitted = 0;
+   setup->numFragsWritten = 0;
+#endif
+
+   if (cull_tri( setup, det ))
+      return;
+
+   if (!setup_sort_vertices( setup, det, v0, v1, v2 ))
+      return;
+   setup_tri_coefficients( setup );
+   setup_tri_edges( setup );
+
+   assert(setup->llvmpipe->reduced_prim == PIPE_PRIM_TRIANGLES);
+
+   setup->span.y = 0;
+   setup->span.right[0] = 0;
+   setup->span.right[1] = 0;
+   /*   setup->span.z_mode = tri_z_mode( setup->ctx ); */
+
+   /*   init_constant_attribs( setup ); */
+
+   if (setup->oneoverarea < 0.0) {
+      /* emaj on left:
+       */
+      subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines );
+      subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines );
+   }
+   else {
+      /* emaj on right:
+       */
+      subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines );
+      subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines );
+   }
+
+   flush_spans( setup );
+
+#if DEBUG_FRAGS
+   printf("Tri: %u frags emitted, %u written\n",
+          setup->numFragsEmitted,
+          setup->numFragsWritten);
+#endif
+}
+
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a line.
+ */
+static void
+linear_pos_coeff(struct setup_context *setup,
+                 uint vertSlot, uint i)
+{
+   const float da = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+   const float dadx = da * setup->emaj.dx * setup->oneoverarea;
+   const float dady = da * setup->emaj.dy * setup->oneoverarea;
+   setup->coef.dadx[0][i] = dadx;
+   setup->coef.dady[0][i] = dady;
+   setup->coef.a0[0][i] = (setup->vmin[vertSlot][i] -
+                           (dadx * (setup->vmin[0][0] - 0.5f) +
+                            dady * (setup->vmin[0][1] - 0.5f)));
+}
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a line.
+ */
+static void
+line_linear_coeff(struct setup_context *setup,
+                  unsigned attrib,
+                  uint vertSlot)
+{
+   unsigned i;
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      const float da = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+      const float dadx = da * setup->emaj.dx * setup->oneoverarea;
+      const float dady = da * setup->emaj.dy * setup->oneoverarea;
+      setup->coef.dadx[1 + attrib][i] = dadx;
+      setup->coef.dady[1 + attrib][i] = dady;
+      setup->coef.a0[1 + attrib][i] = (setup->vmin[vertSlot][i] -
+                     (dadx * (setup->vmin[0][0] - 0.5f) +
+                      dady * (setup->vmin[0][1] - 0.5f)));
+   }
+}
+
+
+/**
+ * Compute a0, dadx and dady for a perspective-corrected interpolant,
+ * for a line.
+ */
+static void
+line_persp_coeff(struct setup_context *setup,
+                 unsigned attrib,
+                 uint vertSlot)
+{
+   unsigned i;
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      /* XXX double-check/verify this arithmetic */
+      const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3];
+      const float a1 = setup->vmax[vertSlot][i] * setup->vmax[0][3];
+      const float da = a1 - a0;
+      const float dadx = da * setup->emaj.dx * setup->oneoverarea;
+      const float dady = da * setup->emaj.dy * setup->oneoverarea;
+      setup->coef.dadx[1 + attrib][i] = dadx;
+      setup->coef.dady[1 + attrib][i] = dady;
+      setup->coef.a0[1 + attrib][i] = (setup->vmin[vertSlot][i] -
+                     (dadx * (setup->vmin[0][0] - 0.5f) +
+                      dady * (setup->vmin[0][1] - 0.5f)));
+   }
+}
+
+
+/**
+ * Compute the setup->coef[] array dadx, dady, a0 values.
+ * Must be called after setup->vmin,vmax are initialized.
+ */
+static INLINE boolean
+setup_line_coefficients(struct setup_context *setup,
+                        const float (*v0)[4],
+                        const float (*v1)[4])
+{
+   struct llvmpipe_context *llvmpipe = setup->llvmpipe;
+   const struct lp_fragment_shader *lpfs = llvmpipe->fs;
+   const struct vertex_info *vinfo = llvmpipe_get_vertex_info(llvmpipe);
+   uint fragSlot;
+   float area;
+
+   /* use setup->vmin, vmax to point to vertices */
+   if (llvmpipe->rasterizer->flatshade_first)
+      setup->vprovoke = v0;
+   else
+      setup->vprovoke = v1;
+   setup->vmin = v0;
+   setup->vmax = v1;
+
+   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
+   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
+
+   /* NOTE: this is not really area but something proportional to it */
+   area = setup->emaj.dx * setup->emaj.dx + setup->emaj.dy * setup->emaj.dy;
+   if (area == 0.0f || util_is_inf_or_nan(area))
+      return FALSE;
+   setup->oneoverarea = 1.0f / area;
+
+   /* z and w are done by linear interpolation:
+    */
+   linear_pos_coeff(setup, 0, 2);
+   linear_pos_coeff(setup, 0, 3);
+
+   /* setup interpolation for all the remaining attributes:
+    */
+   for (fragSlot = 0; fragSlot < lpfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         const_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_LINEAR:
+         line_linear_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_PERSPECTIVE:
+         line_persp_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (lpfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
+         setup->coef.a0[1 + fragSlot][0] = 1.0f - setup->facing;
+         setup->coef.dadx[1 + fragSlot][0] = 0.0;
+         setup->coef.dady[1 + fragSlot][0] = 0.0;
+      }
+   }
+   return TRUE;
+}
+
+
+/**
+ * Plot a pixel in a line segment.
+ */
+static INLINE void
+plot(struct setup_context *setup, int x, int y)
+{
+   const int iy = y & 1;
+   const int ix = x & 1;
+   const int quadX = x - ix;
+   const int quadY = y - iy;
+   const int mask = (1 << ix) << (2 * iy);
+
+   if (quadX != setup->quad[0].input.x0 ||
+       quadY != setup->quad[0].input.y0)
+   {
+      /* flush prev quad, start new quad */
+
+      if (setup->quad[0].input.x0 != -1)
+         clip_emit_quad( setup, &setup->quad[0] );
+
+      setup->quad[0].input.x0 = quadX;
+      setup->quad[0].input.y0 = quadY;
+      setup->quad[0].inout.mask = 0x0;
+   }
+
+   setup->quad[0].inout.mask |= mask;
+}
+
+
+/**
+ * Do setup for line rasterization, then render the line.
+ * Single-pixel width, no stipple, etc.  We rely on the 'draw' module
+ * to handle stippling and wide lines.
+ */
+void
+llvmpipe_setup_line(struct setup_context *setup,
+           const float (*v0)[4],
+           const float (*v1)[4])
+{
+   int x0 = (int) v0[0][0];
+   int x1 = (int) v1[0][0];
+   int y0 = (int) v0[0][1];
+   int y1 = (int) v1[0][1];
+   int dx = x1 - x0;
+   int dy = y1 - y0;
+   int xstep, ystep;
+
+#if DEBUG_VERTS
+   debug_printf("Setup line:\n");
+   print_vertex(setup, v0);
+   print_vertex(setup, v1);
+#endif
+
+   if (setup->llvmpipe->no_rast)
+      return;
+
+   if (dx == 0 && dy == 0)
+      return;
+
+   if (!setup_line_coefficients(setup, v0, v1))
+      return;
+
+   assert(v0[0][0] < 1.0e9);
+   assert(v0[0][1] < 1.0e9);
+   assert(v1[0][0] < 1.0e9);
+   assert(v1[0][1] < 1.0e9);
+
+   if (dx < 0) {
+      dx = -dx;   /* make positive */
+      xstep = -1;
+   }
+   else {
+      xstep = 1;
+   }
+
+   if (dy < 0) {
+      dy = -dy;   /* make positive */
+      ystep = -1;
+   }
+   else {
+      ystep = 1;
+   }
+
+   assert(dx >= 0);
+   assert(dy >= 0);
+   assert(setup->llvmpipe->reduced_prim == PIPE_PRIM_LINES);
+
+   setup->quad[0].input.x0 = setup->quad[0].input.y0 = -1;
+   setup->quad[0].inout.mask = 0x0;
+
+   /* XXX temporary: set coverage to 1.0 so the line appears
+    * if AA mode happens to be enabled.
+    */
+   setup->quad[0].input.coverage[0] =
+   setup->quad[0].input.coverage[1] =
+   setup->quad[0].input.coverage[2] =
+   setup->quad[0].input.coverage[3] = 1.0;
+
+   if (dx > dy) {
+      /*** X-major line ***/
+      int i;
+      const int errorInc = dy + dy;
+      int error = errorInc - dx;
+      const int errorDec = error - dx;
+
+      for (i = 0; i < dx; i++) {
+         plot(setup, x0, y0);
+
+         x0 += xstep;
+         if (error < 0) {
+            error += errorInc;
+         }
+         else {
+            error += errorDec;
+            y0 += ystep;
+         }
+      }
+   }
+   else {
+      /*** Y-major line ***/
+      int i;
+      const int errorInc = dx + dx;
+      int error = errorInc - dy;
+      const int errorDec = error - dy;
+
+      for (i = 0; i < dy; i++) {
+         plot(setup, x0, y0);
+
+         y0 += ystep;
+         if (error < 0) {
+            error += errorInc;
+         }
+         else {
+            error += errorDec;
+            x0 += xstep;
+         }
+      }
+   }
+
+   /* draw final quad */
+   if (setup->quad[0].inout.mask) {
+      clip_emit_quad( setup, &setup->quad[0] );
+   }
+}
+
+
+static void
+point_persp_coeff(struct setup_context *setup,
+                  const float (*vert)[4],
+                  unsigned attrib,
+                  uint vertSlot)
+{
+   unsigned i;
+   for(i = 0; i < NUM_CHANNELS; ++i) {
+      setup->coef.dadx[1 + attrib][i] = 0.0F;
+      setup->coef.dady[1 + attrib][i] = 0.0F;
+      setup->coef.a0[1 + attrib][i] = vert[vertSlot][i] * vert[0][3];
+   }
+}
+
+
+/**
+ * Do setup for point rasterization, then render the point.
+ * Round or square points...
+ * XXX could optimize a lot for 1-pixel points.
+ */
+void
+llvmpipe_setup_point( struct setup_context *setup,
+             const float (*v0)[4] )
+{
+   struct llvmpipe_context *llvmpipe = setup->llvmpipe;
+   const struct lp_fragment_shader *lpfs = llvmpipe->fs;
+   const int sizeAttr = setup->llvmpipe->psize_slot;
+   const float size
+      = sizeAttr > 0 ? v0[sizeAttr][0]
+      : setup->llvmpipe->rasterizer->point_size;
+   const float halfSize = 0.5F * size;
+   const boolean round = (boolean) setup->llvmpipe->rasterizer->point_smooth;
+   const float x = v0[0][0];  /* Note: data[0] is always position */
+   const float y = v0[0][1];
+   const struct vertex_info *vinfo = llvmpipe_get_vertex_info(llvmpipe);
+   uint fragSlot;
+
+#if DEBUG_VERTS
+   debug_printf("Setup point:\n");
+   print_vertex(setup, v0);
+#endif
+
+   if (llvmpipe->no_rast)
+      return;
+
+   assert(setup->llvmpipe->reduced_prim == PIPE_PRIM_POINTS);
+
+   /* For points, all interpolants are constant-valued.
+    * However, for point sprites, we'll need to setup texcoords appropriately.
+    * XXX: which coefficients are the texcoords???
+    * We may do point sprites as textured quads...
+    *
+    * KW: We don't know which coefficients are texcoords - ultimately
+    * the choice of what interpolation mode to use for each attribute
+    * should be determined by the fragment program, using
+    * per-attribute declaration statements that include interpolation
+    * mode as a parameter.  So either the fragment program will have
+    * to be adjusted for pointsprite vs normal point behaviour, or
+    * otherwise a special interpolation mode will have to be defined
+    * which matches the required behaviour for point sprites.  But -
+    * the latter is not a feature of normal hardware, and as such
+    * probably should be ruled out on that basis.
+    */
+   setup->vprovoke = v0;
+
+   /* setup Z, W */
+   const_pos_coeff(setup, 0, 2);
+   const_pos_coeff(setup, 0, 3);
+
+   for (fragSlot = 0; fragSlot < lpfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         /* fall-through */
+      case INTERP_LINEAR:
+         const_coeff(setup, fragSlot, vertSlot);
+         break;
+      case INTERP_PERSPECTIVE:
+         point_persp_coeff(setup, setup->vprovoke, fragSlot, vertSlot);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (lpfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
+         setup->coef.a0[1 + fragSlot][0] = 1.0f - setup->facing;
+         setup->coef.dadx[1 + fragSlot][0] = 0.0;
+         setup->coef.dady[1 + fragSlot][0] = 0.0;
+      }
+   }
+
+
+   if (halfSize <= 0.5 && !round) {
+      /* special case for 1-pixel points */
+      const int ix = ((int) x) & 1;
+      const int iy = ((int) y) & 1;
+      setup->quad[0].input.x0 = (int) x - ix;
+      setup->quad[0].input.y0 = (int) y - iy;
+      setup->quad[0].inout.mask = (1 << ix) << (2 * iy);
+      clip_emit_quad( setup, &setup->quad[0] );
+   }
+   else {
+      if (round) {
+         /* rounded points */
+         const int ixmin = block((int) (x - halfSize));
+         const int ixmax = block((int) (x + halfSize));
+         const int iymin = block((int) (y - halfSize));
+         const int iymax = block((int) (y + halfSize));
+         const float rmin = halfSize - 0.7071F;  /* 0.7071 = sqrt(2)/2 */
+         const float rmax = halfSize + 0.7071F;
+         const float rmin2 = MAX2(0.0F, rmin * rmin);
+         const float rmax2 = rmax * rmax;
+         const float cscale = 1.0F / (rmax2 - rmin2);
+         int ix, iy;
+
+         for (iy = iymin; iy <= iymax; iy += 2) {
+            for (ix = ixmin; ix <= ixmax; ix += 2) {
+               float dx, dy, dist2, cover;
+
+               setup->quad[0].inout.mask = 0x0;
+
+               dx = (ix + 0.5f) - x;
+               dy = (iy + 0.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad[0].input.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad[0].inout.mask |= MASK_TOP_LEFT;
+               }
+
+               dx = (ix + 1.5f) - x;
+               dy = (iy + 0.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad[0].input.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad[0].inout.mask |= MASK_TOP_RIGHT;
+               }
+
+               dx = (ix + 0.5f) - x;
+               dy = (iy + 1.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad[0].input.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad[0].inout.mask |= MASK_BOTTOM_LEFT;
+               }
+
+               dx = (ix + 1.5f) - x;
+               dy = (iy + 1.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad[0].input.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad[0].inout.mask |= MASK_BOTTOM_RIGHT;
+               }
+
+               if (setup->quad[0].inout.mask) {
+                  setup->quad[0].input.x0 = ix;
+                  setup->quad[0].input.y0 = iy;
+                  clip_emit_quad( setup, &setup->quad[0] );
+               }
+            }
+         }
+      }
+      else {
+         /* square points */
+         const int xmin = (int) (x + 0.75 - halfSize);
+         const int ymin = (int) (y + 0.25 - halfSize);
+         const int xmax = xmin + (int) size;
+         const int ymax = ymin + (int) size;
+         /* XXX could apply scissor to xmin,ymin,xmax,ymax now */
+         const int ixmin = block(xmin);
+         const int ixmax = block(xmax - 1);
+         const int iymin = block(ymin);
+         const int iymax = block(ymax - 1);
+         int ix, iy;
+
+         /*
+         debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax);
+         */
+         for (iy = iymin; iy <= iymax; iy += 2) {
+            uint rowMask = 0xf;
+            if (iy < ymin) {
+               /* above the top edge */
+               rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+            }
+            if (iy + 1 >= ymax) {
+               /* below the bottom edge */
+               rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+            }
+
+            for (ix = ixmin; ix <= ixmax; ix += 2) {
+               uint mask = rowMask;
+
+               if (ix < xmin) {
+                  /* fragment is past left edge of point, turn off left bits */
+                  mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+               }
+               if (ix + 1 >= xmax) {
+                  /* past the right edge */
+                  mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+               }
+
+               setup->quad[0].inout.mask = mask;
+               setup->quad[0].input.x0 = ix;
+               setup->quad[0].input.y0 = iy;
+               clip_emit_quad( setup, &setup->quad[0] );
+            }
+         }
+      }
+   }
+}
+
+void llvmpipe_setup_prepare( struct setup_context *setup )
+{
+   struct llvmpipe_context *lp = setup->llvmpipe;
+
+   if (lp->dirty) {
+      llvmpipe_update_derived(lp);
+   }
+
+   if (lp->reduced_api_prim == PIPE_PRIM_TRIANGLES &&
+       lp->rasterizer->fill_cw == PIPE_POLYGON_MODE_FILL &&
+       lp->rasterizer->fill_ccw == PIPE_POLYGON_MODE_FILL) {
+      /* we'll do culling */
+      setup->winding = lp->rasterizer->cull_mode;
+   }
+   else {
+      /* 'draw' will do culling */
+      setup->winding = PIPE_WINDING_NONE;
+   }
+}
+
+
+
+void llvmpipe_setup_destroy_context( struct setup_context *setup )
+{
+   align_free( setup );
+}
+
+
+/**
+ * Create a new primitive setup/render stage.
+ */
+struct setup_context *llvmpipe_setup_create_context( struct llvmpipe_context *llvmpipe )
+{
+   struct setup_context *setup;
+   unsigned i;
+
+   setup = align_malloc(sizeof(struct setup_context), 16);
+   if (!setup)
+      return NULL;
+
+   memset(setup, 0, sizeof *setup);
+   setup->llvmpipe = llvmpipe;
+
+   for (i = 0; i < MAX_QUADS; i++) {
+      setup->quad[i].coef = &setup->coef;
+   }
+
+   setup->span.left[0] = 1000000;     /* greater than right[0] */
+   setup->span.left[1] = 1000000;     /* greater than right[1] */
+
+   return setup;
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.h b/src/gallium/drivers/llvmpipe/lp_setup.h
new file mode 100644
index 0000000000..89c43da046
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_setup.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+#ifndef LP_SETUP_H
+#define LP_SETUP_H
+
+struct setup_context;
+struct llvmpipe_context;
+
+void 
+llvmpipe_setup_tri( struct setup_context *setup,
+	   const float (*v0)[4],
+	   const float (*v1)[4],
+	   const float (*v2)[4] );
+
+void
+llvmpipe_setup_line(struct setup_context *setup,
+           const float (*v0)[4],
+           const float (*v1)[4]);
+
+void
+llvmpipe_setup_point( struct setup_context *setup,
+             const float (*v0)[4] );
+
+
+struct setup_context *llvmpipe_setup_create_context( struct llvmpipe_context *llvmpipe );
+void llvmpipe_setup_prepare( struct setup_context *setup );
+void llvmpipe_setup_destroy_context( struct setup_context *setup );
+
+#endif
diff --git a/src/gallium/drivers/llvmpipe/lp_state.h b/src/gallium/drivers/llvmpipe/lp_state.h
new file mode 100644
index 0000000000..fb10329887
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state.h
@@ -0,0 +1,225 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef LP_STATE_H
+#define LP_STATE_H
+
+#include <llvm-c/Core.h>
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+#include "lp_jit.h"
+
+
+#define LP_NEW_VIEWPORT      0x1
+#define LP_NEW_RASTERIZER    0x2
+#define LP_NEW_FS            0x4
+#define LP_NEW_BLEND         0x8
+#define LP_NEW_CLIP          0x10
+#define LP_NEW_SCISSOR       0x20
+#define LP_NEW_STIPPLE       0x40
+#define LP_NEW_FRAMEBUFFER   0x80
+#define LP_NEW_DEPTH_STENCIL_ALPHA 0x100
+#define LP_NEW_CONSTANTS     0x200
+#define LP_NEW_SAMPLER       0x400
+#define LP_NEW_TEXTURE       0x800
+#define LP_NEW_VERTEX        0x1000
+#define LP_NEW_VS            0x2000
+#define LP_NEW_QUERY         0x4000
+
+
+struct tgsi_sampler;
+struct vertex_info;
+
+
+struct lp_fragment_shader;
+
+
+struct lp_fragment_shader_variant_key
+{
+   struct pipe_depth_state depth;
+   struct pipe_alpha_state alpha;
+   struct pipe_blend_state blend;
+};
+
+
+struct lp_fragment_shader_variant
+{
+   struct lp_fragment_shader *shader;
+
+   struct lp_fragment_shader_variant_key key;
+
+   LLVMValueRef function;
+
+   lp_jit_frag_func jit_function;
+
+   struct lp_fragment_shader_variant *next;
+};
+
+
+/**
+ * Subclass of pipe_shader_state (though it doesn't really need to be).
+ *
+ * This is starting to look an awful lot like a quad pipeline stage...
+ */
+struct lp_fragment_shader
+{
+   struct pipe_shader_state base;
+
+   struct tgsi_shader_info info;
+
+   struct lp_fragment_shader_variant *variants;
+
+   struct lp_fragment_shader_variant *current;
+};
+
+
+/** Subclass of pipe_shader_state */
+struct lp_vertex_shader {
+   struct pipe_shader_state shader;
+   struct draw_vertex_shader *draw_data;
+};
+
+
+
+void *
+llvmpipe_create_blend_state(struct pipe_context *,
+                            const struct pipe_blend_state *);
+void llvmpipe_bind_blend_state(struct pipe_context *,
+                               void *);
+void llvmpipe_delete_blend_state(struct pipe_context *,
+                                 void *);
+
+void *
+llvmpipe_create_sampler_state(struct pipe_context *,
+                              const struct pipe_sampler_state *);
+void llvmpipe_bind_sampler_states(struct pipe_context *, unsigned, void **);
+void llvmpipe_delete_sampler_state(struct pipe_context *, void *);
+
+void *
+llvmpipe_create_depth_stencil_state(struct pipe_context *,
+                                    const struct pipe_depth_stencil_alpha_state *);
+void llvmpipe_bind_depth_stencil_state(struct pipe_context *, void *);
+void llvmpipe_delete_depth_stencil_state(struct pipe_context *, void *);
+
+void *
+llvmpipe_create_rasterizer_state(struct pipe_context *,
+                                 const struct pipe_rasterizer_state *);
+void llvmpipe_bind_rasterizer_state(struct pipe_context *, void *);
+void llvmpipe_delete_rasterizer_state(struct pipe_context *, void *);
+
+void llvmpipe_set_framebuffer_state( struct pipe_context *,
+			     const struct pipe_framebuffer_state * );
+
+void llvmpipe_set_blend_color( struct pipe_context *pipe,
+                               const struct pipe_blend_color *blend_color );
+
+void llvmpipe_set_clip_state( struct pipe_context *,
+			     const struct pipe_clip_state * );
+
+void llvmpipe_set_constant_buffer(struct pipe_context *,
+                                  uint shader, uint index,
+                                  const struct pipe_constant_buffer *buf);
+
+void *llvmpipe_create_fs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void llvmpipe_bind_fs_state(struct pipe_context *, void *);
+void llvmpipe_delete_fs_state(struct pipe_context *, void *);
+void *llvmpipe_create_vs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void llvmpipe_bind_vs_state(struct pipe_context *, void *);
+void llvmpipe_delete_vs_state(struct pipe_context *, void *);
+
+void llvmpipe_set_polygon_stipple( struct pipe_context *,
+				  const struct pipe_poly_stipple * );
+
+void llvmpipe_set_scissor_state( struct pipe_context *,
+                                 const struct pipe_scissor_state * );
+
+void llvmpipe_set_sampler_textures( struct pipe_context *,
+                                    unsigned num,
+                                    struct pipe_texture ** );
+
+void llvmpipe_set_viewport_state( struct pipe_context *,
+                                  const struct pipe_viewport_state * );
+
+void llvmpipe_set_vertex_elements(struct pipe_context *,
+                                  unsigned count,
+                                  const struct pipe_vertex_element *);
+
+void llvmpipe_set_vertex_buffers(struct pipe_context *,
+                                 unsigned count,
+                                 const struct pipe_vertex_buffer *);
+
+void llvmpipe_update_fs(struct llvmpipe_context *lp);
+
+void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe );
+
+
+boolean llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
+			     unsigned start, unsigned count);
+
+boolean llvmpipe_draw_elements(struct pipe_context *pipe,
+			       struct pipe_buffer *indexBuffer,
+			       unsigned indexSize,
+			       unsigned mode, unsigned start, unsigned count);
+boolean
+llvmpipe_draw_range_elements(struct pipe_context *pipe,
+                             struct pipe_buffer *indexBuffer,
+                             unsigned indexSize,
+                             unsigned min_index,
+                             unsigned max_index,
+                             unsigned mode, unsigned start, unsigned count);
+
+void
+llvmpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags);
+
+
+void
+llvmpipe_map_transfers(struct llvmpipe_context *lp);
+
+void
+llvmpipe_unmap_transfers(struct llvmpipe_context *lp);
+
+void
+llvmpipe_map_texture_surfaces(struct llvmpipe_context *lp);
+
+void
+llvmpipe_unmap_texture_surfaces(struct llvmpipe_context *lp);
+
+
+struct vertex_info *
+llvmpipe_get_vertex_info(struct llvmpipe_context *llvmpipe);
+
+struct vertex_info *
+llvmpipe_get_vbuf_vertex_info(struct llvmpipe_context *llvmpipe);
+
+
+#endif
diff --git a/src/gallium/drivers/llvmpipe/lp_state_blend.c b/src/gallium/drivers/llvmpipe/lp_state_blend.c
new file mode 100644
index 0000000000..3f03bd0057
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_blend.c
@@ -0,0 +1,114 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ * @author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_debug_dump.h"
+#include "lp_screen.h"
+#include "lp_context.h"
+#include "lp_state.h"
+
+
+void *
+llvmpipe_create_blend_state(struct pipe_context *pipe,
+                            const struct pipe_blend_state *blend)
+{
+   return mem_dup(blend, sizeof(*blend));
+}
+
+void llvmpipe_bind_blend_state( struct pipe_context *pipe,
+                                void *blend )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   llvmpipe->blend = blend;
+
+   llvmpipe->dirty |= LP_NEW_BLEND;
+}
+
+void llvmpipe_delete_blend_state(struct pipe_context *pipe,
+                                 void *blend)
+{
+   FREE( blend );
+}
+
+
+void llvmpipe_set_blend_color( struct pipe_context *pipe,
+			     const struct pipe_blend_color *blend_color )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   unsigned i, j;
+
+   memcpy(&llvmpipe->blend_color, blend_color, sizeof *blend_color);
+
+   if(!llvmpipe->jit_context.blend_color)
+      llvmpipe->jit_context.blend_color = align_malloc(4 * 16, 16);
+   for (i = 0; i < 4; ++i) {
+      uint8_t c = float_to_ubyte(blend_color->color[i]);
+      for (j = 0; j < 16; ++j)
+         llvmpipe->jit_context.blend_color[i*4 + j] = c;
+   }
+}
+
+
+/** XXX move someday?  Or consolidate all these simple state setters
+ * into one file.
+ */
+
+
+void *
+llvmpipe_create_depth_stencil_state(struct pipe_context *pipe,
+				    const struct pipe_depth_stencil_alpha_state *depth_stencil)
+{
+   return mem_dup(depth_stencil, sizeof(*depth_stencil));
+}
+
+void
+llvmpipe_bind_depth_stencil_state(struct pipe_context *pipe,
+                                  void *depth_stencil)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   llvmpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;
+
+   if(llvmpipe->depth_stencil)
+      llvmpipe->jit_context.alpha_ref_value = llvmpipe->depth_stencil->alpha.ref_value;
+
+   llvmpipe->dirty |= LP_NEW_DEPTH_STENCIL_ALPHA;
+}
+
+void
+llvmpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth)
+{
+   FREE( depth );
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_clip.c b/src/gallium/drivers/llvmpipe/lp_state_clip.c
new file mode 100644
index 0000000000..df68f27acc
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_clip.c
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+#include "lp_context.h"
+#include "lp_state.h"
+#include "draw/draw_context.h"
+
+
+void llvmpipe_set_clip_state( struct pipe_context *pipe,
+			     const struct pipe_clip_state *clip )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   /* pass the clip state to the draw module */
+   draw_set_clip_state(llvmpipe->draw, clip);
+}
+
+
+void llvmpipe_set_viewport_state( struct pipe_context *pipe,
+                                  const struct pipe_viewport_state *viewport )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   /* pass the viewport info to the draw module */
+   draw_set_viewport_state(llvmpipe->draw, viewport);
+
+   llvmpipe->viewport = *viewport; /* struct copy */
+   llvmpipe->dirty |= LP_NEW_VIEWPORT;
+}
+
+
+void llvmpipe_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   draw_flush(llvmpipe->draw);
+
+   llvmpipe->scissor = *scissor; /* struct copy */
+   llvmpipe->dirty |= LP_NEW_SCISSOR;
+}
+
+
+void llvmpipe_set_polygon_stipple( struct pipe_context *pipe,
+                                   const struct pipe_poly_stipple *stipple )
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   draw_flush(llvmpipe->draw);
+
+   llvmpipe->poly_stipple = *stipple; /* struct copy */
+   llvmpipe->dirty |= LP_NEW_STIPPLE;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_derived.c b/src/gallium/drivers/llvmpipe/lp_state_derived.c
new file mode 100644
index 0000000000..6fbb057937
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_derived.c
@@ -0,0 +1,258 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_private.h"
+#include "lp_context.h"
+#include "lp_screen.h"
+#include "lp_tex_cache.h"
+#include "lp_state.h"
+
+
+/**
+ * Mark the current vertex layout as "invalid".
+ * We'll validate the vertex layout later, when we start to actually
+ * render a point or line or tri.
+ */
+static void
+invalidate_vertex_layout(struct llvmpipe_context *llvmpipe)
+{
+   llvmpipe->vertex_info.num_attribs =  0;
+}
+
+
+/**
+ * The vertex info describes how to convert the post-transformed vertices
+ * (simple float[][4]) used by the 'draw' module into vertices for
+ * rasterization.
+ *
+ * This function validates the vertex layout and returns a pointer to a
+ * vertex_info object.
+ */
+struct vertex_info *
+llvmpipe_get_vertex_info(struct llvmpipe_context *llvmpipe)
+{
+   struct vertex_info *vinfo = &llvmpipe->vertex_info;
+
+   if (vinfo->num_attribs == 0) {
+      /* compute vertex layout now */
+      const struct lp_fragment_shader *lpfs = llvmpipe->fs;
+      const enum interp_mode colorInterp
+         = llvmpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
+      uint i;
+
+      if (llvmpipe->vbuf) {
+         /* if using the post-transform vertex buffer, tell draw_vbuf to
+          * simply emit the whole post-xform vertex as-is:
+          */
+         struct vertex_info *vinfo_vbuf = &llvmpipe->vertex_info_vbuf;
+         const uint num = draw_num_vs_outputs(llvmpipe->draw);
+         uint i;
+
+         /* No longer any need to try and emit draw vertex_header info.
+          */
+         vinfo_vbuf->num_attribs = 0;
+         for (i = 0; i < num; i++) {
+            draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i);
+         }
+         draw_compute_vertex_size(vinfo_vbuf);
+      }
+
+      /*
+       * Loop over fragment shader inputs, searching for the matching output
+       * from the vertex shader.
+       */
+      vinfo->num_attribs = 0;
+      for (i = 0; i < lpfs->info.num_inputs; i++) {
+         int src;
+         switch (lpfs->info.input_semantic_name[i]) {
+         case TGSI_SEMANTIC_POSITION:
+            src = draw_find_vs_output(llvmpipe->draw,
+                                      TGSI_SEMANTIC_POSITION, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
+            break;
+
+         case TGSI_SEMANTIC_COLOR:
+            src = draw_find_vs_output(llvmpipe->draw, TGSI_SEMANTIC_COLOR, 
+                                 lpfs->info.input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+            break;
+
+         case TGSI_SEMANTIC_FOG:
+            src = draw_find_vs_output(llvmpipe->draw, TGSI_SEMANTIC_FOG, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         case TGSI_SEMANTIC_GENERIC:
+         case TGSI_SEMANTIC_FACE:
+            /* this includes texcoords and varying vars */
+            src = draw_find_vs_output(llvmpipe->draw, TGSI_SEMANTIC_GENERIC,
+                                      lpfs->info.input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         default:
+            assert(0);
+         }
+      }
+
+      llvmpipe->psize_slot = draw_find_vs_output(llvmpipe->draw,
+                                                 TGSI_SEMANTIC_PSIZE, 0);
+      if (llvmpipe->psize_slot > 0) {
+         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
+                               llvmpipe->psize_slot);
+      }
+
+      draw_compute_vertex_size(vinfo);
+   }
+
+   return vinfo;
+}
+
+
+/**
+ * Called from vbuf module.
+ *
+ * Note that there's actually two different vertex layouts in llvmpipe.
+ *
+ * The normal one is computed in llvmpipe_get_vertex_info() above and is
+ * used by the point/line/tri "setup" code.
+ *
+ * The other one (this one) is only used by the vbuf module (which is
+ * not normally used by default but used in testing).  For the vbuf module,
+ * we basically want to pass-through the draw module's vertex layout as-is.
+ * When the llvmpipe vbuf code begins drawing, the normal vertex layout
+ * will come into play again.
+ */
+struct vertex_info *
+llvmpipe_get_vbuf_vertex_info(struct llvmpipe_context *llvmpipe)
+{
+   (void) llvmpipe_get_vertex_info(llvmpipe);
+   return &llvmpipe->vertex_info_vbuf;
+}
+
+
+/**
+ * Recompute cliprect from scissor bounds, scissor enable and surface size.
+ */
+static void
+compute_cliprect(struct llvmpipe_context *lp)
+{
+   /* LP_NEW_FRAMEBUFFER
+    */
+   uint surfWidth = lp->framebuffer.width;
+   uint surfHeight = lp->framebuffer.height;
+
+   /* LP_NEW_RASTERIZER
+    */
+   if (lp->rasterizer->scissor) {
+
+      /* LP_NEW_SCISSOR
+       *
+       * clip to scissor rect:
+       */
+      lp->cliprect.minx = MAX2(lp->scissor.minx, 0);
+      lp->cliprect.miny = MAX2(lp->scissor.miny, 0);
+      lp->cliprect.maxx = MIN2(lp->scissor.maxx, surfWidth);
+      lp->cliprect.maxy = MIN2(lp->scissor.maxy, surfHeight);
+   }
+   else {
+      /* clip to surface bounds */
+      lp->cliprect.minx = 0;
+      lp->cliprect.miny = 0;
+      lp->cliprect.maxx = surfWidth;
+      lp->cliprect.maxy = surfHeight;
+   }
+}
+
+
+static void
+update_tgsi_samplers( struct llvmpipe_context *llvmpipe )
+{
+   unsigned i;
+
+   /* vertex shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      llvmpipe->tgsi.vert_samplers[i].sampler = llvmpipe->sampler[i];
+      llvmpipe->tgsi.vert_samplers[i].texture = llvmpipe->texture[i];
+      llvmpipe->tgsi.frag_samplers[i].base.get_samples = lp_get_samples;
+   }
+
+   /* fragment shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      llvmpipe->tgsi.frag_samplers[i].sampler = llvmpipe->sampler[i];
+      llvmpipe->tgsi.frag_samplers[i].texture = llvmpipe->texture[i];
+      llvmpipe->tgsi.frag_samplers[i].base.get_samples = lp_get_samples;
+   }
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      lp_tex_tile_cache_validate_texture( llvmpipe->tex_cache[i] );
+   }
+
+   llvmpipe->jit_context.samplers = (struct tgsi_sampler **)llvmpipe->tgsi.frag_samplers_list;
+}
+
+/* Hopefully this will remain quite simple, otherwise need to pull in
+ * something like the state tracker mechanism.
+ */
+void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe )
+{
+   struct llvmpipe_screen *lp_screen = llvmpipe_screen(llvmpipe->pipe.screen);
+
+   /* Check for updated textures.
+    */
+   if (llvmpipe->tex_timestamp != lp_screen->timestamp) {
+      llvmpipe->tex_timestamp = lp_screen->timestamp;
+      llvmpipe->dirty |= LP_NEW_TEXTURE;
+   }
+      
+   if (llvmpipe->dirty & (LP_NEW_SAMPLER |
+                          LP_NEW_TEXTURE))
+      update_tgsi_samplers( llvmpipe );
+
+   if (llvmpipe->dirty & (LP_NEW_RASTERIZER |
+                          LP_NEW_FS |
+                          LP_NEW_VS))
+      invalidate_vertex_layout( llvmpipe );
+
+   if (llvmpipe->dirty & (LP_NEW_SCISSOR |
+                          LP_NEW_RASTERIZER |
+                          LP_NEW_FRAMEBUFFER))
+      compute_cliprect(llvmpipe);
+
+   if (llvmpipe->dirty & (LP_NEW_FS |
+                          LP_NEW_BLEND |
+                          LP_NEW_DEPTH_STENCIL_ALPHA))
+      llvmpipe_update_fs( llvmpipe );
+
+
+   llvmpipe->dirty = 0;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
new file mode 100644
index 0000000000..94170bd716
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -0,0 +1,804 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Code generate the whole fragment pipeline.
+ *
+ * The fragment pipeline consists of the following stages:
+ * - stipple (TBI)
+ * - early depth test
+ * - fragment shader
+ * - alpha test
+ * - depth/stencil test (stencil TBI)
+ * - blending
+ *
+ * This file has only the glue to assembly the fragment pipeline.  The actual
+ * plumbing of converting Gallium state into LLVM IR is done elsewhere, in the
+ * lp_bld_*.[ch] files, and in a complete generic and reusable way. Here we
+ * muster the LLVM JIT execution engine to create a function that follows an
+ * established binary interface and that can be called from C directly.
+ *
+ * A big source of complexity here is that we often want to run different
+ * stages with different precisions and data types and precisions. For example,
+ * the fragment shader needs typically to be done in floats, but the
+ * depth/stencil test and blending is better done in the type that most closely
+ * matches the depth/stencil and color buffer respectively.
+ *
+ * Since the width of a SIMD vector register stays the same regardless of the
+ * element type, different types imply different number of elements, so we must
+ * code generate more instances of the stages with larger types to be able to
+ * feed/consume the stages with smaller types.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "util/u_format.h"
+#include "util/u_debug_dump.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_scan.h"
+#include "tgsi/tgsi_parse.h"
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_conv.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_logic.h"
+#include "lp_bld_depth.h"
+#include "lp_bld_interp.h"
+#include "lp_bld_tgsi.h"
+#include "lp_bld_alpha.h"
+#include "lp_bld_blend.h"
+#include "lp_bld_swizzle.h"
+#include "lp_bld_flow.h"
+#include "lp_bld_debug.h"
+#include "lp_screen.h"
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_quad.h"
+
+
+static const unsigned char quad_offset_x[4] = {0, 1, 0, 1};
+static const unsigned char quad_offset_y[4] = {0, 0, 1, 1};
+
+
+/*
+ * Derive from the quad's upper left scalar coordinates the coordinates for
+ * all other quad pixels
+ */
+static void
+generate_pos0(LLVMBuilderRef builder,
+              LLVMValueRef x,
+              LLVMValueRef y,
+              LLVMValueRef *x0,
+              LLVMValueRef *y0)
+{
+   LLVMTypeRef int_elem_type = LLVMInt32Type();
+   LLVMTypeRef int_vec_type = LLVMVectorType(int_elem_type, QUAD_SIZE);
+   LLVMTypeRef elem_type = LLVMFloatType();
+   LLVMTypeRef vec_type = LLVMVectorType(elem_type, QUAD_SIZE);
+   LLVMValueRef x_offsets[QUAD_SIZE];
+   LLVMValueRef y_offsets[QUAD_SIZE];
+   unsigned i;
+
+   x = lp_build_broadcast(builder, int_vec_type, x);
+   y = lp_build_broadcast(builder, int_vec_type, y);
+
+   for(i = 0; i < QUAD_SIZE; ++i) {
+      x_offsets[i] = LLVMConstInt(int_elem_type, quad_offset_x[i], 0);
+      y_offsets[i] = LLVMConstInt(int_elem_type, quad_offset_y[i], 0);
+   }
+
+   x = LLVMBuildAdd(builder, x, LLVMConstVector(x_offsets, QUAD_SIZE), "");
+   y = LLVMBuildAdd(builder, y, LLVMConstVector(y_offsets, QUAD_SIZE), "");
+
+   *x0 = LLVMBuildSIToFP(builder, x, vec_type, "");
+   *y0 = LLVMBuildSIToFP(builder, y, vec_type, "");
+}
+
+
+/**
+ * Generate the depth test.
+ */
+static void
+generate_depth(struct llvmpipe_context *lp,
+               LLVMBuilderRef builder,
+               const struct pipe_depth_state *state,
+               union lp_type src_type,
+               struct lp_build_mask_context *mask,
+               LLVMValueRef src,
+               LLVMValueRef dst_ptr)
+{
+   const struct util_format_description *format_desc;
+   union lp_type dst_type;
+
+   if(!lp->framebuffer.zsbuf)
+      return;
+
+   format_desc = util_format_description(lp->framebuffer.zsbuf->format);
+   assert(format_desc);
+
+   /* Pick the depth type. */
+   dst_type = lp_depth_type(format_desc, src_type.width*src_type.length);
+
+   /* FIXME: Cope with a depth test type with a different bit width. */
+   assert(dst_type.width == src_type.width);
+   assert(dst_type.length == src_type.length);
+
+#if 1
+   src = lp_build_clamped_float_to_unsigned_norm(builder,
+                                                 src_type,
+                                                 dst_type.width,
+                                                 src);
+#else
+   lp_build_conv(builder, src_type, dst_type, &src, 1, &src, 1);
+#endif
+
+   lp_build_depth_test(builder,
+                       state,
+                       dst_type,
+                       format_desc,
+                       mask,
+                       src,
+                       dst_ptr);
+}
+
+
+struct build_fetch_texel_context
+{
+   LLVMValueRef context_ptr;
+
+   LLVMValueRef samplers_ptr;
+
+   /** Coords/texels store */
+   LLVMValueRef store_ptr;
+};
+
+
+void PIPE_CDECL
+lp_fetch_texel_soa( struct tgsi_sampler **samplers,
+                    uint32_t unit,
+                    float *store )
+{
+   struct tgsi_sampler *sampler = samplers[unit];
+
+#if 0
+   uint j;
+
+   debug_printf("%s sampler: %p (%p) store: %p\n",
+                __FUNCTION__,
+                sampler, *sampler,
+                store );
+
+   debug_printf("lodbias %f\n", store[12]);
+
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d texcoord %f %f\n",
+                   j,
+                   store[0+j],
+                   store[4+j]);
+#endif
+
+   {
+      float rgba[NUM_CHANNELS][QUAD_SIZE];
+      sampler->get_samples(sampler,
+                           &store[0],
+                           &store[4],
+                           &store[8],
+                           0.0f, /*store[12],  lodbias */
+                           rgba);
+      memcpy(store, rgba, sizeof rgba);
+   }
+
+#if 0
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d result %f %f %f %f\n",
+                   j,
+                   store[0+j],
+                   store[4+j],
+                   store[8+j],
+                   store[12+j]);
+#endif
+}
+
+
+static void
+emit_fetch_texel( LLVMBuilderRef builder,
+                  void *context,
+                  unsigned unit,
+                  unsigned num_coords,
+                  const LLVMValueRef *coords,
+                  LLVMValueRef lodbias,
+                  LLVMValueRef *texel)
+{
+   struct build_fetch_texel_context *bld = context;
+   LLVMTypeRef vec_type = LLVMTypeOf(coords[0]);
+   LLVMValueRef args[3];
+   unsigned i;
+
+   if(!bld->samplers_ptr)
+      bld->samplers_ptr = lp_jit_context_samplers(builder, bld->context_ptr);
+
+   if(!bld->store_ptr)
+      bld->store_ptr = LLVMBuildArrayAlloca(builder,
+                                            vec_type,
+                                            LLVMConstInt(LLVMInt32Type(), 4, 0),
+                                            "texel_store");
+
+   for (i = 0; i < num_coords; i++) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef coord_ptr = LLVMBuildGEP(builder, bld->store_ptr, &index, 1, "");
+      LLVMBuildStore(builder, coords[i], coord_ptr);
+   }
+
+   args[0] = bld->samplers_ptr;
+   args[1] = LLVMConstInt(LLVMInt32Type(), unit, 0);
+   args[2] = bld->store_ptr;
+
+   lp_build_intrinsic(builder, "fetch_texel", LLVMVoidType(), args, 3);
+
+   for (i = 0; i < NUM_CHANNELS; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef texel_ptr = LLVMBuildGEP(builder, bld->store_ptr, &index, 1, "");
+      texel[i] = LLVMBuildLoad(builder, texel_ptr, "");
+   }
+}
+
+
+/**
+ * Generate the fragment shader, depth/stencil test, and alpha tests.
+ */
+static void
+generate_fs(struct llvmpipe_context *lp,
+            struct lp_fragment_shader *shader,
+            const struct lp_fragment_shader_variant_key *key,
+            LLVMBuilderRef builder,
+            union lp_type type,
+            LLVMValueRef context_ptr,
+            unsigned i,
+            const struct lp_build_interp_soa_context *interp,
+            struct build_fetch_texel_context *sampler,
+            LLVMValueRef *pmask,
+            LLVMValueRef *color,
+            LLVMValueRef depth_ptr)
+{
+   const struct tgsi_token *tokens = shader->base.tokens;
+   LLVMTypeRef elem_type;
+   LLVMTypeRef vec_type;
+   LLVMTypeRef int_vec_type;
+   LLVMValueRef consts_ptr;
+   LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
+   LLVMValueRef z = interp->pos[2];
+   struct lp_build_mask_context mask;
+   boolean early_depth_test;
+   unsigned attrib;
+   unsigned chan;
+
+   elem_type = lp_build_elem_type(type);
+   vec_type = lp_build_vec_type(type);
+   int_vec_type = lp_build_int_vec_type(type);
+
+   consts_ptr = lp_jit_context_constants(builder, context_ptr);
+
+   lp_build_mask_begin(&mask, builder, type, *pmask);
+
+   early_depth_test =
+      lp->depth_stencil->depth.enabled &&
+      lp->framebuffer.zsbuf &&
+      !lp->depth_stencil->alpha.enabled &&
+      !lp->fs->info.uses_kill &&
+      !lp->fs->info.writes_z;
+
+   if(early_depth_test)
+      generate_depth(lp, builder, &key->depth,
+                     type, &mask,
+                     z, depth_ptr);
+
+   memset(outputs, 0, sizeof outputs);
+
+   lp_build_tgsi_soa(builder, tokens, type, &mask,
+                     consts_ptr, interp->pos, interp->inputs,
+                     outputs, emit_fetch_texel, sampler);
+
+   for (attrib = 0; attrib < shader->info.num_outputs; ++attrib) {
+      for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+         if(outputs[attrib][chan]) {
+            lp_build_name(outputs[attrib][chan], "output%u.%u.%c", i, attrib, "xyzw"[chan]);
+
+            switch (shader->info.output_semantic_name[attrib]) {
+            case TGSI_SEMANTIC_COLOR:
+               {
+                  unsigned cbuf = shader->info.output_semantic_index[attrib];
+
+                  lp_build_name(outputs[attrib][chan], "color%u.%u.%c", i, attrib, "rgba"[chan]);
+
+                  /* Alpha test */
+                  /* XXX: should the alpha reference value be passed separately? */
+                  if(cbuf == 0 && chan == 3) {
+                     LLVMValueRef alpha = outputs[attrib][chan];
+                     LLVMValueRef alpha_ref_value;
+                     alpha_ref_value = lp_jit_context_alpha_ref_value(builder, context_ptr);
+                     alpha_ref_value = lp_build_broadcast(builder, vec_type, alpha_ref_value);
+                     lp_build_alpha_test(builder, &key->alpha, type,
+                                         &mask, alpha, alpha_ref_value);
+                  }
+
+                  if(cbuf == 0)
+                     color[chan] = outputs[attrib][chan];
+
+                  break;
+               }
+
+            case TGSI_SEMANTIC_POSITION:
+               if(chan == 2)
+                  z = outputs[attrib][chan];
+               break;
+            }
+         }
+      }
+   }
+
+   if(!early_depth_test)
+      generate_depth(lp, builder, &key->depth,
+                     type, &mask,
+                     z, depth_ptr);
+
+   lp_build_mask_end(&mask);
+
+   *pmask = mask.value;
+
+}
+
+
+/**
+ * Generate color blending and color output.
+ */
+static void
+generate_blend(const struct pipe_blend_state *blend,
+               LLVMBuilderRef builder,
+               union lp_type type,
+               LLVMValueRef context_ptr,
+               LLVMValueRef mask,
+               LLVMValueRef *src,
+               LLVMValueRef dst_ptr)
+{
+   struct lp_build_context bld;
+   LLVMTypeRef vec_type;
+   LLVMTypeRef int_vec_type;
+   LLVMValueRef const_ptr;
+   LLVMValueRef con[4];
+   LLVMValueRef dst[4];
+   LLVMValueRef res[4];
+   unsigned chan;
+
+   vec_type = lp_build_vec_type(type);
+   int_vec_type = lp_build_int_vec_type(type);
+
+   lp_build_context_init(&bld, builder, type);
+
+   const_ptr = lp_jit_context_blend_color(builder, context_ptr);
+   const_ptr = LLVMBuildBitCast(builder, const_ptr,
+                                LLVMPointerType(vec_type, 0), "");
+
+   for(chan = 0; chan < 4; ++chan) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), chan, 0);
+      con[chan] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, const_ptr, &index, 1, ""), "");
+
+      dst[chan] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dst_ptr, &index, 1, ""), "");
+
+      lp_build_name(con[chan], "con.%c", "rgba"[chan]);
+      lp_build_name(dst[chan], "dst.%c", "rgba"[chan]);
+   }
+
+   lp_build_blend_soa(builder, blend, type, src, dst, con, res);
+
+   for(chan = 0; chan < 4; ++chan) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), chan, 0);
+      lp_build_name(res[chan], "res.%c", "rgba"[chan]);
+      res[chan] = lp_build_select(&bld, mask, res[chan], dst[chan]);
+      LLVMBuildStore(builder, res[chan], LLVMBuildGEP(builder, dst_ptr, &index, 1, ""));
+   }
+}
+
+
+/**
+ * Generate the runtime callable function for the whole fragment pipeline.
+ */
+static struct lp_fragment_shader_variant *
+generate_fragment(struct llvmpipe_context *lp,
+                  struct lp_fragment_shader *shader,
+                  const struct lp_fragment_shader_variant_key *key)
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(lp->pipe.screen);
+   struct lp_fragment_shader_variant *variant;
+   union lp_type fs_type;
+   union lp_type blend_type;
+   LLVMTypeRef fs_elem_type;
+   LLVMTypeRef fs_vec_type;
+   LLVMTypeRef fs_int_vec_type;
+   LLVMTypeRef blend_vec_type;
+   LLVMTypeRef blend_int_vec_type;
+   LLVMTypeRef arg_types[9];
+   LLVMTypeRef func_type;
+   LLVMValueRef context_ptr;
+   LLVMValueRef x;
+   LLVMValueRef y;
+   LLVMValueRef a0_ptr;
+   LLVMValueRef dadx_ptr;
+   LLVMValueRef dady_ptr;
+   LLVMValueRef mask_ptr;
+   LLVMValueRef color_ptr;
+   LLVMValueRef depth_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef x0;
+   LLVMValueRef y0;
+   struct build_fetch_texel_context sampler;
+   struct lp_build_interp_soa_context interp;
+   LLVMValueRef fs_mask[LP_MAX_VECTOR_LENGTH];
+   LLVMValueRef fs_out_color[NUM_CHANNELS][LP_MAX_VECTOR_LENGTH];
+   LLVMValueRef blend_mask;
+   LLVMValueRef blend_in_color[NUM_CHANNELS];
+   unsigned num_fs;
+   unsigned i;
+   unsigned chan;
+
+#ifdef DEBUG
+   tgsi_dump(shader->base.tokens, 0);
+   if(key->depth.enabled) {
+      debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE));
+      debug_printf("depth.writemask = %u\n", key->depth.writemask);
+      debug_printf("depth.occlusion_count = %u\n", key->depth.occlusion_count);
+   }
+   if(key->alpha.enabled) {
+      debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE));
+      debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value);
+   }
+   if(key->blend.logicop_enable) {
+      debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func);
+   }
+   else if(key->blend.blend_enable) {
+      debug_printf("blend.rgb_func = %s\n",   debug_dump_blend_func  (key->blend.rgb_func, TRUE));
+      debug_printf("rgb_src_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE));
+      debug_printf("rgb_dst_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE));
+      debug_printf("alpha_func = %s\n",       debug_dump_blend_func  (key->blend.alpha_func, TRUE));
+      debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE));
+      debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE));
+   }
+   debug_printf("blend.colormask = 0x%x\n", key->blend.colormask);
+#endif
+
+   variant = CALLOC_STRUCT(lp_fragment_shader_variant);
+   if(!variant)
+      return NULL;
+
+   variant->shader = shader;
+   memcpy(&variant->key, key, sizeof *key);
+
+   /* TODO: actually pick these based on the fs and color buffer
+    * characteristics. */
+
+   fs_type.value = 0;
+   fs_type.floating = TRUE; /* floating point values */
+   fs_type.sign = TRUE;     /* values are signed */
+   fs_type.norm = FALSE;    /* values are not limited to [0,1] or [-1,1] */
+   fs_type.width = 32;      /* 32-bit float */
+   fs_type.length = 4;      /* 4 element per vector */
+   num_fs = 4;
+
+   blend_type.value = 0;
+   blend_type.floating = FALSE; /* values are integers */
+   blend_type.sign = FALSE;     /* values are unsigned */
+   blend_type.norm = TRUE;      /* values are in [0,1] or [-1,1] */
+   blend_type.width = 8;        /* 8-bit ubyte values */
+   blend_type.length = 16;      /* 16 elements per vector */
+
+   /* 
+    * Generate the function prototype. Any change here must be reflected in
+    * lp_jit.h's lp_jit_frag_func function pointer type, and vice-versa.
+    */
+
+   fs_elem_type = lp_build_elem_type(fs_type);
+   fs_vec_type = lp_build_vec_type(fs_type);
+   fs_int_vec_type = lp_build_int_vec_type(fs_type);
+
+   blend_vec_type = lp_build_vec_type(blend_type);
+   blend_int_vec_type = lp_build_int_vec_type(blend_type);
+
+   arg_types[0] = screen->context_ptr_type;            /* context */
+   arg_types[1] = LLVMInt32Type();                     /* x */
+   arg_types[2] = LLVMInt32Type();                     /* y */
+   arg_types[3] = LLVMPointerType(fs_elem_type, 0);    /* a0 */
+   arg_types[4] = LLVMPointerType(fs_elem_type, 0);    /* dadx */
+   arg_types[5] = LLVMPointerType(fs_elem_type, 0);    /* dady */
+   arg_types[6] = LLVMPointerType(fs_int_vec_type, 0); /* mask */
+   arg_types[7] = LLVMPointerType(blend_vec_type, 0);  /* color */
+   arg_types[8] = LLVMPointerType(fs_int_vec_type, 0); /* depth */
+
+   func_type = LLVMFunctionType(LLVMVoidType(), arg_types, Elements(arg_types), 0);
+
+   variant->function = LLVMAddFunction(screen->module, "shader", func_type);
+   LLVMSetFunctionCallConv(variant->function, LLVMCCallConv);
+   for(i = 0; i < Elements(arg_types); ++i)
+      if(LLVMGetTypeKind(arg_types[i]) == LLVMPointerTypeKind)
+         LLVMAddAttribute(LLVMGetParam(variant->function, i), LLVMNoAliasAttribute);
+
+   context_ptr  = LLVMGetParam(variant->function, 0);
+   x            = LLVMGetParam(variant->function, 1);
+   y            = LLVMGetParam(variant->function, 2);
+   a0_ptr       = LLVMGetParam(variant->function, 3);
+   dadx_ptr     = LLVMGetParam(variant->function, 4);
+   dady_ptr     = LLVMGetParam(variant->function, 5);
+   mask_ptr     = LLVMGetParam(variant->function, 6);
+   color_ptr    = LLVMGetParam(variant->function, 7);
+   depth_ptr    = LLVMGetParam(variant->function, 8);
+
+   lp_build_name(context_ptr, "context");
+   lp_build_name(x, "x");
+   lp_build_name(y, "y");
+   lp_build_name(a0_ptr, "a0");
+   lp_build_name(dadx_ptr, "dadx");
+   lp_build_name(dady_ptr, "dady");
+   lp_build_name(mask_ptr, "mask");
+   lp_build_name(color_ptr, "color");
+   lp_build_name(depth_ptr, "depth");
+
+   /*
+    * Function body
+    */
+
+   block = LLVMAppendBasicBlock(variant->function, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   generate_pos0(builder, x, y, &x0, &y0);
+
+   lp_build_interp_soa_init(&interp, shader->base.tokens, builder, fs_type,
+                            a0_ptr, dadx_ptr, dady_ptr,
+                            x0, y0, 2, 0);
+
+   memset(&sampler, 0, sizeof sampler);
+   sampler.context_ptr = context_ptr;
+
+   for(i = 0; i < num_fs; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef out_color[NUM_CHANNELS];
+      LLVMValueRef depth_ptr_i;
+
+      if(i != 0)
+         lp_build_interp_soa_update(&interp);
+
+      fs_mask[i] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, mask_ptr, &index, 1, ""), "");
+      depth_ptr_i = LLVMBuildGEP(builder, depth_ptr, &index, 1, "");
+
+      generate_fs(lp, shader, key,
+                  builder,
+                  fs_type,
+                  context_ptr,
+                  i,
+                  &interp,
+                  &sampler,
+                  &fs_mask[i],
+                  out_color,
+                  depth_ptr_i);
+
+      for(chan = 0; chan < NUM_CHANNELS; ++chan)
+         fs_out_color[chan][i] = out_color[chan];
+   }
+
+   /* 
+    * Convert the fs's output color and mask to fit to the blending type. 
+    */
+
+   for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+      lp_build_conv(builder, fs_type, blend_type,
+                    fs_out_color[chan], num_fs,
+                    &blend_in_color[chan], 1);
+      lp_build_name(blend_in_color[chan], "color.%c", "rgba"[chan]);
+
+   }
+
+   lp_build_conv_mask(builder, fs_type, blend_type,
+                               fs_mask, num_fs,
+                               &blend_mask, 1);
+
+   /*
+    * Blending.
+    */
+
+   generate_blend(&key->blend,
+                  builder,
+                  blend_type,
+                  context_ptr,
+                  blend_mask,
+                  blend_in_color,
+                  color_ptr);
+
+   LLVMBuildRetVoid(builder);
+
+   LLVMDisposeBuilder(builder);
+
+   /*
+    * Translate the LLVM IR into machine code.
+    */
+
+   LLVMRunFunctionPassManager(screen->pass, variant->function);
+
+#ifdef DEBUG
+   LLVMDumpValue(variant->function);
+   debug_printf("\n");
+#endif
+
+   if(LLVMVerifyFunction(variant->function, LLVMPrintMessageAction)) {
+      LLVMDumpValue(variant->function);
+      abort();
+   }
+
+   variant->jit_function = (lp_jit_frag_func)LLVMGetPointerToGlobal(screen->engine, variant->function);
+
+#ifdef DEBUG
+   lp_disassemble(variant->jit_function);
+#endif
+
+   variant->next = shader->variants;
+   shader->variants = variant;
+
+   return variant;
+}
+
+
+void *
+llvmpipe_create_fs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct lp_fragment_shader *shader;
+
+   shader = CALLOC_STRUCT(lp_fragment_shader);
+   if (!shader)
+      return NULL;
+
+   /* get/save the summary info for this shader */
+   tgsi_scan_shader(templ->tokens, &shader->info);
+
+   /* we need to keep a local copy of the tokens */
+   shader->base.tokens = tgsi_dup_tokens(templ->tokens);
+
+   return shader;
+}
+
+
+void
+llvmpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   llvmpipe->fs = (struct lp_fragment_shader *) fs;
+
+   llvmpipe->dirty |= LP_NEW_FS;
+}
+
+
+void
+llvmpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   struct llvmpipe_screen *screen = llvmpipe_screen(pipe->screen);
+   struct lp_fragment_shader *shader = fs;
+   struct lp_fragment_shader_variant *variant;
+
+   assert(fs != llvmpipe->fs);
+
+   variant = shader->variants;
+   while(variant) {
+      struct lp_fragment_shader_variant *next = variant->next;
+
+      if(variant->function) {
+         if(variant->jit_function)
+            LLVMFreeMachineCodeForFunction(screen->engine, variant->function);
+         LLVMDeleteFunction(variant->function);
+      }
+
+      FREE(variant);
+
+      variant = next;
+   }
+
+   FREE((void *) shader->base.tokens);
+   FREE(shader);
+}
+
+
+
+void
+llvmpipe_set_constant_buffer(struct pipe_context *pipe,
+                             uint shader, uint index,
+                             const struct pipe_constant_buffer *buf)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   assert(shader < PIPE_SHADER_TYPES);
+   assert(index == 0);
+
+   /* note: reference counting */
+   pipe_buffer_reference(&llvmpipe->constants[shader].buffer,
+			 buf ? buf->buffer : NULL);
+
+   llvmpipe->dirty |= LP_NEW_CONSTANTS;
+}
+
+
+/**
+ * We need to generate several variants of the fragment pipeline to match
+ * all the combinations of the contributing state atoms.
+ *
+ * TODO: there is actually no reason to tie this to context state -- the
+ * generated code could be cached globally in the screen.
+ */
+static void
+make_variant_key(struct llvmpipe_context *lp,
+                 struct lp_fragment_shader_variant_key *key)
+{
+   memset(key, 0, sizeof *key);
+
+   memcpy(&key->depth, &lp->depth_stencil->depth, sizeof key->depth);
+
+   key->alpha.enabled = lp->depth_stencil->alpha.enabled;
+   if(key->alpha.enabled)
+      key->alpha.func = lp->depth_stencil->alpha.func;
+   /* alpha.ref_value is passed in jit_context */
+
+   memcpy(&key->blend, lp->blend, sizeof key->blend);
+}
+
+
+void 
+llvmpipe_update_fs(struct llvmpipe_context *lp)
+{
+   struct lp_fragment_shader *shader = lp->fs;
+   struct lp_fragment_shader_variant_key key;
+   struct lp_fragment_shader_variant *variant;
+
+   make_variant_key(lp, &key);
+
+   variant = shader->variants;
+   while(variant) {
+      if(memcmp(&variant->key, &key, sizeof key) == 0)
+         break;
+
+      variant = variant->next;
+   }
+
+   if(!variant)
+      variant = generate_fragment(lp, shader, &key);
+
+   shader->current = variant;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
new file mode 100644
index 0000000000..4561c6b845
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
@@ -0,0 +1,62 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "lp_context.h"
+#include "lp_state.h"
+#include "draw/draw_context.h"
+
+
+
+void *
+llvmpipe_create_rasterizer_state(struct pipe_context *pipe,
+                                 const struct pipe_rasterizer_state *rast)
+{
+   return mem_dup(rast, sizeof(*rast));
+}
+
+void llvmpipe_bind_rasterizer_state(struct pipe_context *pipe,
+                                    void *setup)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   /* pass-through to draw module */
+   draw_set_rasterizer_state(llvmpipe->draw, setup);
+
+   llvmpipe->rasterizer = (struct pipe_rasterizer_state *)setup;
+
+   llvmpipe->dirty |= LP_NEW_RASTERIZER;
+}
+
+void llvmpipe_delete_rasterizer_state(struct pipe_context *pipe,
+                                      void *rasterizer)
+{
+   FREE( rasterizer );
+}
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_state_sampler.c b/src/gallium/drivers/llvmpipe/lp_state_sampler.c
new file mode 100644
index 0000000000..4fef541b1e
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_sampler.c
@@ -0,0 +1,117 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:
+ *  Brian Paul
+ */
+
+#include "util/u_memory.h"
+
+#include "draw/draw_context.h"
+
+#include "lp_context.h"
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_texture.h"
+#include "lp_tex_cache.h"
+#include "draw/draw_context.h"
+
+
+
+void *
+llvmpipe_create_sampler_state(struct pipe_context *pipe,
+                              const struct pipe_sampler_state *sampler)
+{
+   return mem_dup(sampler, sizeof(*sampler));
+}
+
+
+void
+llvmpipe_bind_sampler_states(struct pipe_context *pipe,
+                             unsigned num, void **sampler)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   unsigned i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == llvmpipe->num_samplers &&
+       !memcmp(llvmpipe->sampler, sampler, num * sizeof(void *)))
+      return;
+
+   draw_flush(llvmpipe->draw);
+
+   for (i = 0; i < num; ++i)
+      llvmpipe->sampler[i] = sampler[i];
+   for (i = num; i < PIPE_MAX_SAMPLERS; ++i)
+      llvmpipe->sampler[i] = NULL;
+
+   llvmpipe->num_samplers = num;
+
+   llvmpipe->dirty |= LP_NEW_SAMPLER;
+}
+
+
+void
+llvmpipe_set_sampler_textures(struct pipe_context *pipe,
+                              unsigned num, struct pipe_texture **texture)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == llvmpipe->num_textures &&
+       !memcmp(llvmpipe->texture, texture, num * sizeof(struct pipe_texture *)))
+      return;
+
+   draw_flush(llvmpipe->draw);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      struct pipe_texture *tex = i < num ? texture[i] : NULL;
+
+      pipe_texture_reference(&llvmpipe->texture[i], tex);
+      lp_tex_tile_cache_set_texture(llvmpipe->tex_cache[i], tex);
+   }
+
+   llvmpipe->num_textures = num;
+
+   llvmpipe->dirty |= LP_NEW_TEXTURE;
+}
+
+
+void
+llvmpipe_delete_sampler_state(struct pipe_context *pipe,
+                              void *sampler)
+{
+   FREE( sampler );
+}
+
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c
new file mode 100644
index 0000000000..177a26b7b1
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c
@@ -0,0 +1,106 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_surface.h"
+#include "lp_tile_cache.h"
+
+#include "draw/draw_context.h"
+
+
+/**
+ * XXX this might get moved someday
+ * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer.
+ * Here, we flush the old surfaces and update the tile cache to point to the new
+ * surfaces.
+ */
+void
+llvmpipe_set_framebuffer_state(struct pipe_context *pipe,
+                               const struct pipe_framebuffer_state *fb)
+{
+   struct llvmpipe_context *lp = llvmpipe_context(pipe);
+   uint i;
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+      /* check if changing cbuf */
+      if (lp->framebuffer.cbufs[i] != fb->cbufs[i]) {
+         /* flush old */
+         lp_flush_tile_cache(lp->cbuf_cache[i]);
+
+         /* assign new */
+         lp->framebuffer.cbufs[i] = fb->cbufs[i];
+
+         /* update cache */
+         lp_tile_cache_set_surface(lp->cbuf_cache[i], fb->cbufs[i]);
+      }
+   }
+
+   lp->framebuffer.nr_cbufs = fb->nr_cbufs;
+
+   /* zbuf changing? */
+   if (lp->framebuffer.zsbuf != fb->zsbuf) {
+
+      if(lp->zsbuf_transfer) {
+         struct pipe_screen *screen = pipe->screen;
+
+         if(lp->zsbuf_map) {
+            screen->transfer_unmap(screen, lp->zsbuf_transfer);
+            lp->zsbuf_map = NULL;
+         }
+
+         screen->tex_transfer_destroy(lp->zsbuf_transfer);
+         lp->zsbuf_transfer = NULL;
+      }
+
+      /* assign new */
+      lp->framebuffer.zsbuf = fb->zsbuf;
+
+      /* Tell draw module how deep the Z/depth buffer is */
+      if (lp->framebuffer.zsbuf) {
+         int depth_bits;
+         double mrd;
+         depth_bits = pf_get_component_bits(lp->framebuffer.zsbuf->format,
+                                            PIPE_FORMAT_COMP_Z);
+         if (depth_bits > 16) {
+            mrd = 0.0000001;
+         }
+         else {
+            mrd = 0.00002;
+         }
+         draw_set_mrd(lp->draw, mrd);
+      }
+   }
+
+   lp->framebuffer.width = fb->width;
+   lp->framebuffer.height = fb->height;
+
+   lp->dirty |= LP_NEW_FRAMEBUFFER;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_vertex.c b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
new file mode 100644
index 0000000000..1a17631a4c
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_surface.h"
+
+#include "draw/draw_context.h"
+
+
+void
+llvmpipe_set_vertex_elements(struct pipe_context *pipe,
+                             unsigned count,
+                             const struct pipe_vertex_element *attribs)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(llvmpipe->vertex_element, attribs,
+          count * sizeof(struct pipe_vertex_element));
+   llvmpipe->num_vertex_elements = count;
+
+   llvmpipe->dirty |= LP_NEW_VERTEX;
+
+   draw_set_vertex_elements(llvmpipe->draw, count, attribs);
+}
+
+
+void
+llvmpipe_set_vertex_buffers(struct pipe_context *pipe,
+                            unsigned count,
+                            const struct pipe_vertex_buffer *buffers)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(llvmpipe->vertex_buffer, buffers, count * sizeof(buffers[0]));
+   llvmpipe->num_vertex_buffers = count;
+
+   llvmpipe->dirty |= LP_NEW_VERTEX;
+
+   draw_set_vertex_buffers(llvmpipe->draw, count, buffers);
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_state_vs.c b/src/gallium/drivers/llvmpipe/lp_state_vs.c
new file mode 100644
index 0000000000..15c3029614
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_state_vs.c
@@ -0,0 +1,96 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_defines.h"
+#include "tgsi/tgsi_parse.h"
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+
+#include "lp_context.h"
+#include "lp_state.h"
+
+
+void *
+llvmpipe_create_vs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   struct lp_vertex_shader *state;
+
+   state = CALLOC_STRUCT(lp_vertex_shader);
+   if (state == NULL ) 
+      goto fail;
+
+   /* copy shader tokens, the ones passed in will go away.
+    */
+   state->shader.tokens = tgsi_dup_tokens(templ->tokens);
+   if (state->shader.tokens == NULL)
+      goto fail;
+
+   state->draw_data = draw_create_vertex_shader(llvmpipe->draw, templ);
+   if (state->draw_data == NULL) 
+      goto fail;
+
+   return state;
+
+fail:
+   if (state) {
+      FREE( (void *)state->shader.tokens );
+      FREE( state->draw_data );
+      FREE( state );
+   }
+   return NULL;
+}
+
+
+void
+llvmpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   llvmpipe->vs = (const struct lp_vertex_shader *)vs;
+
+   draw_bind_vertex_shader(llvmpipe->draw,
+                           (llvmpipe->vs ? llvmpipe->vs->draw_data : NULL));
+
+   llvmpipe->dirty |= LP_NEW_VS;
+}
+
+
+void
+llvmpipe_delete_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   struct lp_vertex_shader *state =
+      (struct lp_vertex_shader *)vs;
+
+   draw_delete_vertex_shader(llvmpipe->draw, state->draw_data);
+   FREE( state );
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_surface.c b/src/gallium/drivers/llvmpipe/lp_surface.c
new file mode 100644
index 0000000000..6110b0a193
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_surface.c
@@ -0,0 +1,50 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_rect.h"
+#include "lp_context.h"
+#include "lp_surface.h"
+
+
+static void
+lp_surface_copy(struct pipe_context *pipe,
+                struct pipe_surface *dest, unsigned destx, unsigned desty,
+                struct pipe_surface *src, unsigned srcx, unsigned srcy,
+                unsigned width, unsigned height)
+{
+   util_surface_copy(pipe, FALSE,
+                     dest, destx, desty,
+                     src, srcx, srcy,
+                     width, height);
+}
+
+void
+lp_init_surface_functions(struct llvmpipe_context *lp)
+{
+   lp->pipe.surface_copy = lp_surface_copy;
+   lp->pipe.surface_fill = util_surface_fill;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_surface.h b/src/gallium/drivers/llvmpipe/lp_surface.h
new file mode 100644
index 0000000000..4d78a53c4f
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_surface.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef LP_SURFACE_H
+#define LP_SURFACE_H
+
+
+struct llvmpipe_context;
+
+
+extern void
+lp_init_surface_functions(struct llvmpipe_context *lp);
+
+
+#endif /* LP_SURFACE_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_test.h b/src/gallium/drivers/llvmpipe/lp_test.h
new file mode 100644
index 0000000000..69aaae26e0
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test.h
@@ -0,0 +1,128 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Shared testing code.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#ifndef LP_TEST_H
+#define LP_TEST_H
+
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <float.h>
+
+#include <llvm-c/Core.h>
+#include <llvm-c/Analysis.h>
+#include <llvm-c/ExecutionEngine.h>
+#include <llvm-c/Target.h>
+#include <llvm-c/BitWriter.h>
+#include <llvm-c/Transforms/Scalar.h>
+
+#include "pipe/p_state.h"
+#include "util/u_format.h"
+#include "util/u_math.h"
+#include "util/u_debug_dump.h"
+
+#include "lp_bld_type.h"
+
+
+void
+write_tsv_header(FILE *fp);
+
+
+boolean
+test_some(unsigned verbose, FILE *fp, unsigned long n);
+
+
+boolean
+test_all(unsigned verbose, FILE *fp);
+
+
+static INLINE uint64_t
+rdtsc(void)
+{
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+   uint32_t hi, lo;
+   __asm__ __volatile__ ("rdtsc" : "=a"(lo), "=d"(hi));
+   return ((uint64_t)lo) | (((uint64_t)hi) << 32);
+#else
+   return 0;
+#endif
+}
+
+
+float
+random_float(void);
+
+
+void
+dump_type(FILE *fp, union lp_type type);
+
+
+double
+read_elem(union lp_type type, const void *src, unsigned index);
+
+
+void
+write_elem(union lp_type type, void *dst, unsigned index, double src);
+
+
+void
+random_elem(union lp_type type, void *dst, unsigned index);
+
+
+void
+read_vec(union lp_type type, const void *src, double *dst);
+
+
+void
+write_vec(union lp_type type, void *dst, const double *src);
+
+
+void
+random_vec(union lp_type type, void *dst);
+
+
+boolean
+compare_vec_with_eps(union lp_type type, const void *res, const void *ref, double eps);
+
+
+boolean
+compare_vec(union lp_type type, const void *res, const void *ref);
+
+
+void
+dump_vec(FILE *fp, union lp_type type, const void *src);
+
+
+#endif /* !LP_TEST_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_test_blend.c b/src/gallium/drivers/llvmpipe/lp_test_blend.c
new file mode 100644
index 0000000000..8dfad468e3
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test_blend.c
@@ -0,0 +1,881 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Unit tests for blend LLVM IR generation
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ *
+ * Blend computation code derived from code written by
+ * @author Brian Paul <brian@vmware.com>
+ */
+
+
+#include "lp_bld_type.h"
+#include "lp_bld_arit.h"
+#include "lp_bld_blend.h"
+#include "lp_bld_debug.h"
+#include "lp_test.h"
+
+
+enum vector_mode
+{
+   AoS = 0,
+   SoA = 1
+};
+
+
+typedef void (*blend_test_ptr_t)(const void *src, const void *dst, const void *con, void *res);
+
+
+void
+write_tsv_header(FILE *fp)
+{
+   fprintf(fp,
+           "result\t"
+           "cycles_per_channel\t"
+           "mode\t"
+           "type\t"
+           "sep_func\t"
+           "sep_src_factor\t"
+           "sep_dst_factor\t"
+           "rgb_func\t"
+           "rgb_src_factor\t"
+           "rgb_dst_factor\t"
+           "alpha_func\t"
+           "alpha_src_factor\t"
+           "alpha_dst_factor\n");
+
+   fflush(fp);
+}
+
+
+static void
+write_tsv_row(FILE *fp,
+              const struct pipe_blend_state *blend,
+              enum vector_mode mode,
+              union lp_type type,
+              double cycles,
+              boolean success)
+{
+   fprintf(fp, "%s\t", success ? "pass" : "fail");
+
+   if (mode == AoS) {
+      fprintf(fp, "%.1f\t", cycles / type.length);
+      fprintf(fp, "aos\t");
+   }
+
+   if (mode == SoA) {
+      fprintf(fp, "%.1f\t", cycles / (4 * type.length));
+      fprintf(fp, "soa\t");
+   }
+
+   fprintf(fp, "%s%u%sx%u\t",
+           type.floating ? "f" : (type.fixed ? "h" : (type.sign ? "s" : "u")),
+           type.width,
+           type.norm ? "n" : "",
+           type.length);
+
+   fprintf(fp,
+           "%s\t%s\t%s\t",
+           blend->rgb_func != blend->alpha_func ? "true" : "false",
+           blend->rgb_src_factor != blend->alpha_src_factor ? "true" : "false",
+           blend->rgb_dst_factor != blend->alpha_dst_factor ? "true" : "false");
+
+   fprintf(fp,
+           "%s\t%s\t%s\t%s\t%s\t%s\n",
+           debug_dump_blend_func(blend->rgb_func, TRUE),
+           debug_dump_blend_factor(blend->rgb_src_factor, TRUE),
+           debug_dump_blend_factor(blend->rgb_dst_factor, TRUE),
+           debug_dump_blend_func(blend->alpha_func, TRUE),
+           debug_dump_blend_factor(blend->alpha_src_factor, TRUE),
+           debug_dump_blend_factor(blend->alpha_dst_factor, TRUE));
+
+   fflush(fp);
+}
+
+
+static void
+dump_blend_type(FILE *fp,
+                const struct pipe_blend_state *blend,
+                enum vector_mode mode,
+                union lp_type type)
+{
+   fprintf(fp, "%s", mode ? "soa" : "aos");
+
+   fprintf(fp, " type=%s%u%sx%u",
+           type.floating ? "f" : (type.fixed ? "h" : (type.sign ? "s" : "u")),
+           type.width,
+           type.norm ? "n" : "",
+           type.length);
+
+   fprintf(fp,
+           " %s=%s %s=%s %s=%s %s=%s %s=%s %s=%s",
+           "rgb_func",         debug_dump_blend_func(blend->rgb_func, TRUE),
+           "rgb_src_factor",   debug_dump_blend_factor(blend->rgb_src_factor, TRUE),
+           "rgb_dst_factor",   debug_dump_blend_factor(blend->rgb_dst_factor, TRUE),
+           "alpha_func",       debug_dump_blend_func(blend->alpha_func, TRUE),
+           "alpha_src_factor", debug_dump_blend_factor(blend->alpha_src_factor, TRUE),
+           "alpha_dst_factor", debug_dump_blend_factor(blend->alpha_dst_factor, TRUE));
+
+   fprintf(fp, " ...\n");
+   fflush(fp);
+}
+
+
+static LLVMValueRef
+add_blend_test(LLVMModuleRef module,
+               const struct pipe_blend_state *blend,
+               enum vector_mode mode,
+               union lp_type type)
+{
+   LLVMTypeRef ret_type;
+   LLVMTypeRef vec_type;
+   LLVMTypeRef args[4];
+   LLVMValueRef func;
+   LLVMValueRef src_ptr;
+   LLVMValueRef dst_ptr;
+   LLVMValueRef const_ptr;
+   LLVMValueRef res_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+
+   ret_type = LLVMInt64Type();
+   vec_type = lp_build_vec_type(type);
+
+   args[3] = args[2] = args[1] = args[0] = LLVMPointerType(vec_type, 0);
+   func = LLVMAddFunction(module, "test", LLVMFunctionType(LLVMVoidType(), args, 4, 0));
+   LLVMSetFunctionCallConv(func, LLVMCCallConv);
+   src_ptr = LLVMGetParam(func, 0);
+   dst_ptr = LLVMGetParam(func, 1);
+   const_ptr = LLVMGetParam(func, 2);
+   res_ptr = LLVMGetParam(func, 3);
+
+   block = LLVMAppendBasicBlock(func, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   if (mode == AoS) {
+      LLVMValueRef src;
+      LLVMValueRef dst;
+      LLVMValueRef con;
+      LLVMValueRef res;
+
+      src = LLVMBuildLoad(builder, src_ptr, "src");
+      dst = LLVMBuildLoad(builder, dst_ptr, "dst");
+      con = LLVMBuildLoad(builder, const_ptr, "const");
+
+      res = lp_build_blend_aos(builder, blend, type, src, dst, con, 3);
+
+      lp_build_name(res, "res");
+
+      LLVMBuildStore(builder, res, res_ptr);
+   }
+
+   if (mode == SoA) {
+      LLVMValueRef src[4];
+      LLVMValueRef dst[4];
+      LLVMValueRef con[4];
+      LLVMValueRef res[4];
+      unsigned i;
+
+      for(i = 0; i < 4; ++i) {
+         LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+         src[i] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, src_ptr, &index, 1, ""), "");
+         dst[i] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dst_ptr, &index, 1, ""), "");
+         con[i] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, const_ptr, &index, 1, ""), "");
+         lp_build_name(src[i], "src.%c", "rgba"[i]);
+         lp_build_name(con[i], "con.%c", "rgba"[i]);
+         lp_build_name(dst[i], "dst.%c", "rgba"[i]);
+      }
+
+      lp_build_blend_soa(builder, blend, type, src, dst, con, res);
+
+      for(i = 0; i < 4; ++i) {
+         LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+         lp_build_name(res[i], "res.%c", "rgba"[i]);
+         LLVMBuildStore(builder, res[i], LLVMBuildGEP(builder, res_ptr, &index, 1, ""));
+      }
+   }
+
+   LLVMBuildRetVoid(builder);;
+
+   LLVMDisposeBuilder(builder);
+   return func;
+}
+
+
+/** Add and limit result to ceiling of 1.0 */
+#define ADD_SAT(R, A, B) \
+do { \
+   R = (A) + (B);  if (R > 1.0f) R = 1.0f; \
+} while (0)
+
+/** Subtract and limit result to floor of 0.0 */
+#define SUB_SAT(R, A, B) \
+do { \
+   R = (A) - (B);  if (R < 0.0f) R = 0.0f; \
+} while (0)
+
+
+static void
+compute_blend_ref_term(unsigned rgb_factor,
+                       unsigned alpha_factor,
+                       const double *factor,
+                       const double *src,
+                       const double *dst,
+                       const double *con,
+                       double *term)
+{
+   double temp;
+
+   switch (rgb_factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      term[0] = factor[0]; /* R */
+      term[1] = factor[1]; /* G */
+      term[2] = factor[2]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      term[0] = factor[0] * src[0]; /* R */
+      term[1] = factor[1] * src[1]; /* G */
+      term[2] = factor[2] * src[2]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      term[0] = factor[0] * src[3]; /* R */
+      term[1] = factor[1] * src[3]; /* G */
+      term[2] = factor[2] * src[3]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      term[0] = factor[0] * dst[0]; /* R */
+      term[1] = factor[1] * dst[1]; /* G */
+      term[2] = factor[2] * dst[2]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      term[0] = factor[0] * dst[3]; /* R */
+      term[1] = factor[1] * dst[3]; /* G */
+      term[2] = factor[2] * dst[3]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      temp = MIN2(src[3], 1.0f - dst[3]);
+      term[0] = factor[0] * temp; /* R */
+      term[1] = factor[1] * temp; /* G */
+      term[2] = factor[2] * temp; /* B */
+      break;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      term[0] = factor[0] * con[0]; /* R */
+      term[1] = factor[1] * con[1]; /* G */
+      term[2] = factor[2] * con[2]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      term[0] = factor[0] * con[3]; /* R */
+      term[1] = factor[1] * con[3]; /* G */
+      term[2] = factor[2] * con[3]; /* B */
+      break;
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+      assert(0); /* to do */
+      break;
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+      assert(0); /* to do */
+      break;
+   case PIPE_BLENDFACTOR_ZERO:
+      term[0] = 0.0f; /* R */
+      term[1] = 0.0f; /* G */
+      term[2] = 0.0f; /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      term[0] = factor[0] * (1.0f - src[0]); /* R */
+      term[1] = factor[1] * (1.0f - src[1]); /* G */
+      term[2] = factor[2] * (1.0f - src[2]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      term[0] = factor[0] * (1.0f - src[3]); /* R */
+      term[1] = factor[1] * (1.0f - src[3]); /* G */
+      term[2] = factor[2] * (1.0f - src[3]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      term[0] = factor[0] * (1.0f - dst[3]); /* R */
+      term[1] = factor[1] * (1.0f - dst[3]); /* G */
+      term[2] = factor[2] * (1.0f - dst[3]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      term[0] = factor[0] * (1.0f - dst[0]); /* R */
+      term[1] = factor[1] * (1.0f - dst[1]); /* G */
+      term[2] = factor[2] * (1.0f - dst[2]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      term[0] = factor[0] * (1.0f - con[0]); /* R */
+      term[1] = factor[1] * (1.0f - con[1]); /* G */
+      term[2] = factor[2] * (1.0f - con[2]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      term[0] = factor[0] * (1.0f - con[3]); /* R */
+      term[1] = factor[1] * (1.0f - con[3]); /* G */
+      term[2] = factor[2] * (1.0f - con[3]); /* B */
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+      assert(0); /* to do */
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+      assert(0); /* to do */
+      break;
+   default:
+      assert(0);
+   }
+
+   /*
+    * Compute src/first term A
+    */
+   switch (alpha_factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      term[3] = factor[3]; /* A */
+      break;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      term[3] = factor[3] * src[3]; /* A */
+      break;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      term[3] = factor[3] * dst[3]; /* A */
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      term[3] = src[3]; /* A */
+      break;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      term[3] = factor[3] * con[3]; /* A */
+      break;
+   case PIPE_BLENDFACTOR_ZERO:
+      term[3] = 0.0f; /* A */
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      term[3] = factor[3] * (1.0f - src[3]); /* A */
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      term[3] = factor[3] * (1.0f - dst[3]); /* A */
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      term[3] = factor[3] * (1.0f - con[3]);
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static void
+compute_blend_ref(const struct pipe_blend_state *blend,
+                  const double *src,
+                  const double *dst,
+                  const double *con,
+                  double *res)
+{
+   double src_term[4];
+   double dst_term[4];
+
+   compute_blend_ref_term(blend->rgb_src_factor, blend->alpha_src_factor, src, src, dst, con, src_term);
+   compute_blend_ref_term(blend->rgb_dst_factor, blend->alpha_dst_factor, dst, src, dst, con, dst_term);
+
+   /*
+    * Combine RGB terms
+    */
+   switch (blend->rgb_func) {
+   case PIPE_BLEND_ADD:
+      ADD_SAT(res[0], src_term[0], dst_term[0]); /* R */
+      ADD_SAT(res[1], src_term[1], dst_term[1]); /* G */
+      ADD_SAT(res[2], src_term[2], dst_term[2]); /* B */
+      break;
+   case PIPE_BLEND_SUBTRACT:
+      SUB_SAT(res[0], src_term[0], dst_term[0]); /* R */
+      SUB_SAT(res[1], src_term[1], dst_term[1]); /* G */
+      SUB_SAT(res[2], src_term[2], dst_term[2]); /* B */
+      break;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      SUB_SAT(res[0], dst_term[0], src_term[0]); /* R */
+      SUB_SAT(res[1], dst_term[1], src_term[1]); /* G */
+      SUB_SAT(res[2], dst_term[2], src_term[2]); /* B */
+      break;
+   case PIPE_BLEND_MIN:
+      res[0] = MIN2(src_term[0], dst_term[0]); /* R */
+      res[1] = MIN2(src_term[1], dst_term[1]); /* G */
+      res[2] = MIN2(src_term[2], dst_term[2]); /* B */
+      break;
+   case PIPE_BLEND_MAX:
+      res[0] = MAX2(src_term[0], dst_term[0]); /* R */
+      res[1] = MAX2(src_term[1], dst_term[1]); /* G */
+      res[2] = MAX2(src_term[2], dst_term[2]); /* B */
+      break;
+   default:
+      assert(0);
+   }
+
+   /*
+    * Combine A terms
+    */
+   switch (blend->alpha_func) {
+   case PIPE_BLEND_ADD:
+      ADD_SAT(res[3], src_term[3], dst_term[3]); /* A */
+      break;
+   case PIPE_BLEND_SUBTRACT:
+      SUB_SAT(res[3], src_term[3], dst_term[3]); /* A */
+      break;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      SUB_SAT(res[3], dst_term[3], src_term[3]); /* A */
+      break;
+   case PIPE_BLEND_MIN:
+      res[3] = MIN2(src_term[3], dst_term[3]); /* A */
+      break;
+   case PIPE_BLEND_MAX:
+      res[3] = MAX2(src_term[3], dst_term[3]); /* A */
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static boolean
+test_one(unsigned verbose,
+         FILE *fp,
+         const struct pipe_blend_state *blend,
+         enum vector_mode mode,
+         union lp_type type)
+{
+   LLVMModuleRef module = NULL;
+   LLVMValueRef func = NULL;
+   LLVMExecutionEngineRef engine = NULL;
+   LLVMModuleProviderRef provider = NULL;
+   LLVMPassManagerRef pass = NULL;
+   char *error = NULL;
+   blend_test_ptr_t blend_test_ptr;
+   boolean success;
+   const unsigned n = 32;
+   int64_t cycles[n];
+   double cycles_avg = 0.0;
+   unsigned i, j;
+
+   if(verbose >= 1)
+      dump_blend_type(stdout, blend, mode, type);
+
+   module = LLVMModuleCreateWithName("test");
+
+   func = add_blend_test(module, blend, mode, type);
+
+   if(LLVMVerifyModule(module, LLVMPrintMessageAction, &error)) {
+      LLVMDumpModule(module);
+      abort();
+   }
+   LLVMDisposeMessage(error);
+
+   provider = LLVMCreateModuleProviderForExistingModule(module);
+   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
+      if(verbose < 1)
+         dump_blend_type(stderr, blend, mode, type);
+      fprintf(stderr, "%s\n", error);
+      LLVMDisposeMessage(error);
+      abort();
+   }
+
+#if 0
+   pass = LLVMCreatePassManager();
+   LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   LLVMAddConstantPropagationPass(pass);
+   LLVMAddInstructionCombiningPass(pass);
+   LLVMAddPromoteMemoryToRegisterPass(pass);
+   LLVMAddGVNPass(pass);
+   LLVMAddCFGSimplificationPass(pass);
+   LLVMRunPassManager(pass, module);
+#else
+   (void)pass;
+#endif
+
+   if(verbose >= 2)
+      LLVMDumpModule(module);
+
+   blend_test_ptr = (blend_test_ptr_t)LLVMGetPointerToGlobal(engine, func);
+
+   if(verbose >= 2)
+      lp_disassemble(blend_test_ptr);
+
+   success = TRUE;
+   for(i = 0; i < n && success; ++i) {
+      if(mode == AoS) {
+         uint8_t src[LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t dst[LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t con[LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t res[LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t ref[LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         int64_t start_counter = 0;
+         int64_t end_counter = 0;
+
+         random_vec(type, src);
+         random_vec(type, dst);
+         random_vec(type, con);
+
+         {
+            double fsrc[LP_MAX_VECTOR_LENGTH];
+            double fdst[LP_MAX_VECTOR_LENGTH];
+            double fcon[LP_MAX_VECTOR_LENGTH];
+            double fref[LP_MAX_VECTOR_LENGTH];
+
+            read_vec(type, src, fsrc);
+            read_vec(type, dst, fdst);
+            read_vec(type, con, fcon);
+
+            for(j = 0; j < type.length; j += 4)
+               compute_blend_ref(blend, fsrc + j, fdst + j, fcon + j, fref + j);
+
+            write_vec(type, ref, fref);
+         }
+
+         start_counter = rdtsc();
+         blend_test_ptr(src, dst, con, res);
+         end_counter = rdtsc();
+
+         cycles[i] = end_counter - start_counter;
+
+         if(!compare_vec(type, res, ref)) {
+            success = FALSE;
+
+            if(verbose < 1)
+               dump_blend_type(stderr, blend, mode, type);
+            fprintf(stderr, "MISMATCH\n");
+
+            fprintf(stderr, "  Src: ");
+            dump_vec(stderr, type, src);
+            fprintf(stderr, "\n");
+
+            fprintf(stderr, "  Dst: ");
+            dump_vec(stderr, type, dst);
+            fprintf(stderr, "\n");
+
+            fprintf(stderr, "  Con: ");
+            dump_vec(stderr, type, con);
+            fprintf(stderr, "\n");
+
+            fprintf(stderr, "  Res: ");
+            dump_vec(stderr, type, res);
+            fprintf(stderr, "\n");
+
+            fprintf(stderr, "  Ref: ");
+            dump_vec(stderr, type, ref);
+            fprintf(stderr, "\n");
+         }
+      }
+
+      if(mode == SoA) {
+         const unsigned stride = type.length*type.width/8;
+         uint8_t src[4*LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t dst[4*LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t con[4*LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t res[4*LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         uint8_t ref[4*LP_MAX_VECTOR_LENGTH*LP_MAX_TYPE_WIDTH/8];
+         int64_t start_counter = 0;
+         int64_t end_counter = 0;
+         boolean mismatch;
+
+         for(j = 0; j < 4; ++j) {
+            random_vec(type, src + j*stride);
+            random_vec(type, dst + j*stride);
+            random_vec(type, con + j*stride);
+         }
+
+         {
+            double fsrc[4];
+            double fdst[4];
+            double fcon[4];
+            double fref[4];
+            unsigned k;
+
+            for(k = 0; k < type.length; ++k) {
+               for(j = 0; j < 4; ++j) {
+                  fsrc[j] = read_elem(type, src + j*stride, k);
+                  fdst[j] = read_elem(type, dst + j*stride, k);
+                  fcon[j] = read_elem(type, con + j*stride, k);
+               }
+
+               compute_blend_ref(blend, fsrc, fdst, fcon, fref);
+
+               for(j = 0; j < 4; ++j)
+                  write_elem(type, ref + j*stride, k, fref[j]);
+            }
+         }
+
+         start_counter = rdtsc();
+         blend_test_ptr(src, dst, con, res);
+         end_counter = rdtsc();
+
+         cycles[i] = end_counter - start_counter;
+
+         mismatch = FALSE;
+         for (j = 0; j < 4; ++j)
+            if(!compare_vec(type, res + j*stride, ref + j*stride))
+               mismatch = TRUE;
+
+         if (mismatch) {
+            success = FALSE;
+
+            if(verbose < 1)
+               dump_blend_type(stderr, blend, mode, type);
+            fprintf(stderr, "MISMATCH\n");
+            for(j = 0; j < 4; ++j) {
+               char channel = "RGBA"[j];
+               fprintf(stderr, "  Src%c: ", channel);
+               dump_vec(stderr, type, src + j*stride);
+               fprintf(stderr, "\n");
+
+               fprintf(stderr, "  Dst%c: ", channel);
+               dump_vec(stderr, type, dst + j*stride);
+               fprintf(stderr, "\n");
+
+               fprintf(stderr, "  Con%c: ", channel);
+               dump_vec(stderr, type, con + j*stride);
+               fprintf(stderr, "\n");
+
+               fprintf(stderr, "  Res%c: ", channel);
+               dump_vec(stderr, type, res + j*stride);
+               fprintf(stderr, "\n");
+
+               fprintf(stderr, "  Ref%c: ", channel);
+               dump_vec(stderr, type, ref + j*stride);
+               fprintf(stderr, "\n");
+            }
+         }
+      }
+   }
+
+   /*
+    * Unfortunately the output of cycle counter is not very reliable as it comes
+    * -- sometimes we get outliers (due IRQs perhaps?) which are
+    * better removed to avoid random or biased data.
+    */
+   {
+      double sum = 0.0, sum2 = 0.0;
+      double avg, std;
+      unsigned m;
+
+      for(i = 0; i < n; ++i) {
+         sum += cycles[i];
+         sum2 += cycles[i]*cycles[i];
+      }
+
+      avg = sum/n;
+      std = sqrtf((sum2 - n*avg*avg)/n);
+
+      m = 0;
+      sum = 0.0;
+      for(i = 0; i < n; ++i) {
+         if(fabs(cycles[i] - avg) <= 4.0*std) {
+            sum += cycles[i];
+            ++m;
+         }
+      }
+
+      cycles_avg = sum/m;
+
+   }
+
+   if(fp)
+      write_tsv_row(fp, blend, mode, type, cycles_avg, success);
+
+   if (!success) {
+      if(verbose < 2)
+         LLVMDumpModule(module);
+      LLVMWriteBitcodeToFile(module, "blend.bc");
+      fprintf(stderr, "blend.bc written\n");
+      fprintf(stderr, "Invoke as \"llc -o - blend.bc\"\n");
+      abort();
+   }
+
+   LLVMFreeMachineCodeForFunction(engine, func);
+
+   LLVMDisposeExecutionEngine(engine);
+   if(pass)
+      LLVMDisposePassManager(pass);
+
+   return success;
+}
+
+
+const unsigned
+blend_factors[] = {
+   PIPE_BLENDFACTOR_ZERO,
+   PIPE_BLENDFACTOR_ONE,
+   PIPE_BLENDFACTOR_SRC_COLOR,
+   PIPE_BLENDFACTOR_SRC_ALPHA,
+   PIPE_BLENDFACTOR_DST_COLOR,
+   PIPE_BLENDFACTOR_DST_ALPHA,
+   PIPE_BLENDFACTOR_CONST_COLOR,
+   PIPE_BLENDFACTOR_CONST_ALPHA,
+#if 0
+   PIPE_BLENDFACTOR_SRC1_COLOR,
+   PIPE_BLENDFACTOR_SRC1_ALPHA,
+#endif
+   PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE,
+   PIPE_BLENDFACTOR_INV_SRC_COLOR,
+   PIPE_BLENDFACTOR_INV_SRC_ALPHA,
+   PIPE_BLENDFACTOR_INV_DST_COLOR,
+   PIPE_BLENDFACTOR_INV_DST_ALPHA,
+   PIPE_BLENDFACTOR_INV_CONST_COLOR,
+   PIPE_BLENDFACTOR_INV_CONST_ALPHA,
+#if 0
+   PIPE_BLENDFACTOR_INV_SRC1_COLOR,
+   PIPE_BLENDFACTOR_INV_SRC1_ALPHA,
+#endif
+};
+
+
+const unsigned
+blend_funcs[] = {
+   PIPE_BLEND_ADD,
+   PIPE_BLEND_SUBTRACT,
+   PIPE_BLEND_REVERSE_SUBTRACT,
+   PIPE_BLEND_MIN,
+   PIPE_BLEND_MAX
+};
+
+
+const union lp_type blend_types[] = {
+   /* float, fixed,  sign,  norm, width, len */
+   {{  TRUE, FALSE, FALSE,  TRUE,    32,   4 }}, /* f32 x 4 */
+   {{ FALSE, FALSE, FALSE,  TRUE,     8,  16 }}, /* u8n x 16 */
+};
+
+
+const unsigned num_funcs = sizeof(blend_funcs)/sizeof(blend_funcs[0]);
+const unsigned num_factors = sizeof(blend_factors)/sizeof(blend_factors[0]);
+const unsigned num_types = sizeof(blend_types)/sizeof(blend_types[0]);
+
+
+boolean
+test_all(unsigned verbose, FILE *fp)
+{
+   const unsigned *rgb_func;
+   const unsigned *rgb_src_factor;
+   const unsigned *rgb_dst_factor;
+   const unsigned *alpha_func;
+   const unsigned *alpha_src_factor;
+   const unsigned *alpha_dst_factor;
+   struct pipe_blend_state blend;
+   enum vector_mode mode;
+   const union lp_type *type;
+   bool success = TRUE;
+
+   for(rgb_func = blend_funcs; rgb_func < &blend_funcs[num_funcs]; ++rgb_func) {
+      for(alpha_func = blend_funcs; alpha_func < &blend_funcs[num_funcs]; ++alpha_func) {
+         for(rgb_src_factor = blend_factors; rgb_src_factor < &blend_factors[num_factors]; ++rgb_src_factor) {
+            for(rgb_dst_factor = blend_factors; rgb_dst_factor <= rgb_src_factor; ++rgb_dst_factor) {
+               for(alpha_src_factor = blend_factors; alpha_src_factor < &blend_factors[num_factors]; ++alpha_src_factor) {
+                  for(alpha_dst_factor = blend_factors; alpha_dst_factor <= alpha_src_factor; ++alpha_dst_factor) {
+                     for(mode = 0; mode < 2; ++mode) {
+                        for(type = blend_types; type < &blend_types[num_types]; ++type) {
+
+                           if(*rgb_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE ||
+                              *alpha_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE)
+                              continue;
+
+                           memset(&blend, 0, sizeof blend);
+                           blend.blend_enable      = 1;
+                           blend.rgb_func          = *rgb_func;
+                           blend.rgb_src_factor    = *rgb_src_factor;
+                           blend.rgb_dst_factor    = *rgb_dst_factor;
+                           blend.alpha_func        = *alpha_func;
+                           blend.alpha_src_factor  = *alpha_src_factor;
+                           blend.alpha_dst_factor  = *alpha_dst_factor;
+                           blend.colormask         = PIPE_MASK_RGBA;
+
+                           if(!test_one(verbose, fp, &blend, mode, *type))
+                             success = FALSE;
+
+                        }
+                     }
+                  }
+               }
+            }
+         }
+      }
+   }
+
+   return success;
+}
+
+
+boolean
+test_some(unsigned verbose, FILE *fp, unsigned long n)
+{
+   const unsigned *rgb_func;
+   const unsigned *rgb_src_factor;
+   const unsigned *rgb_dst_factor;
+   const unsigned *alpha_func;
+   const unsigned *alpha_src_factor;
+   const unsigned *alpha_dst_factor;
+   struct pipe_blend_state blend;
+   enum vector_mode mode;
+   const union lp_type *type;
+   unsigned long i;
+   bool success = TRUE;
+
+   for(i = 0; i < n; ++i) {
+      rgb_func = &blend_funcs[random() % num_funcs];
+      alpha_func = &blend_funcs[random() % num_funcs];
+      rgb_src_factor = &blend_factors[random() % num_factors];
+      alpha_src_factor = &blend_factors[random() % num_factors];
+      
+      do {
+         rgb_dst_factor = &blend_factors[random() % num_factors];
+      } while(*rgb_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE);
+
+      do {
+         alpha_dst_factor = &blend_factors[random() % num_factors];
+      } while(*alpha_dst_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE);
+
+      mode = random() & 1;
+
+      type = &blend_types[random() % num_types];
+
+      memset(&blend, 0, sizeof blend);
+      blend.blend_enable      = 1;
+      blend.rgb_func          = *rgb_func;
+      blend.rgb_src_factor    = *rgb_src_factor;
+      blend.rgb_dst_factor    = *rgb_dst_factor;
+      blend.alpha_func        = *alpha_func;
+      blend.alpha_src_factor  = *alpha_src_factor;
+      blend.alpha_dst_factor  = *alpha_dst_factor;
+      blend.colormask         = PIPE_MASK_RGBA;
+
+      if(!test_one(verbose, fp, &blend, mode, *type))
+        success = FALSE;
+   }
+
+   return success;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_test_conv.c b/src/gallium/drivers/llvmpipe/lp_test_conv.c
new file mode 100644
index 0000000000..e6489834af
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test_conv.c
@@ -0,0 +1,427 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Unit tests for type conversion.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "lp_bld_type.h"
+#include "lp_bld_const.h"
+#include "lp_bld_conv.h"
+#include "lp_bld_debug.h"
+#include "lp_test.h"
+
+
+typedef void (*conv_test_ptr_t)(const void *src, const void *dst);
+
+
+void
+write_tsv_header(FILE *fp)
+{
+   fprintf(fp,
+           "result\t"
+           "cycles_per_channel\t"
+           "src_type\t"
+           "dst_type\n");
+
+   fflush(fp);
+}
+
+
+static void
+write_tsv_row(FILE *fp,
+              union lp_type src_type,
+              union lp_type dst_type,
+              double cycles,
+              boolean success)
+{
+   fprintf(fp, "%s\t", success ? "pass" : "fail");
+
+   fprintf(fp, "%.1f\t", cycles / MAX2(src_type.length, dst_type.length));
+
+   dump_type(fp, src_type);
+   fprintf(fp, "\t");
+
+   dump_type(fp, dst_type);
+   fprintf(fp, "\n");
+
+   fflush(fp);
+}
+
+
+static void
+dump_conv_types(FILE *fp,
+               union lp_type src_type,
+               union lp_type dst_type)
+{
+   fprintf(fp, "src_type=");
+   dump_type(fp, src_type);
+
+   fprintf(fp, " dst_type=");
+   dump_type(fp, dst_type);
+
+   fprintf(fp, " ...\n");
+   fflush(fp);
+}
+
+
+static LLVMValueRef
+add_conv_test(LLVMModuleRef module,
+              union lp_type src_type, unsigned num_srcs,
+              union lp_type dst_type, unsigned num_dsts)
+{
+   LLVMTypeRef args[2];
+   LLVMValueRef func;
+   LLVMValueRef src_ptr;
+   LLVMValueRef dst_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef src[LP_MAX_VECTOR_LENGTH];
+   LLVMValueRef dst[LP_MAX_VECTOR_LENGTH];
+   unsigned i;
+
+   args[0] = LLVMPointerType(lp_build_vec_type(src_type), 0);
+   args[1] = LLVMPointerType(lp_build_vec_type(dst_type), 0);
+
+   func = LLVMAddFunction(module, "test", LLVMFunctionType(LLVMVoidType(), args, 2, 0));
+   LLVMSetFunctionCallConv(func, LLVMCCallConv);
+   src_ptr = LLVMGetParam(func, 0);
+   dst_ptr = LLVMGetParam(func, 1);
+
+   block = LLVMAppendBasicBlock(func, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   for(i = 0; i < num_srcs; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef ptr = LLVMBuildGEP(builder, src_ptr, &index, 1, "");
+      src[i] = LLVMBuildLoad(builder, ptr, "");
+   }
+
+   lp_build_conv(builder, src_type, dst_type, src, num_srcs, dst, num_dsts);
+
+   for(i = 0; i < num_dsts; ++i) {
+      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+      LLVMValueRef ptr = LLVMBuildGEP(builder, dst_ptr, &index, 1, "");
+      LLVMBuildStore(builder, dst[i], ptr);
+   }
+
+   LLVMBuildRetVoid(builder);;
+
+   LLVMDisposeBuilder(builder);
+   return func;
+}
+
+
+static boolean
+test_one(unsigned verbose,
+         FILE *fp,
+         union lp_type src_type,
+         union lp_type dst_type)
+{
+   LLVMModuleRef module = NULL;
+   LLVMValueRef func = NULL;
+   LLVMExecutionEngineRef engine = NULL;
+   LLVMModuleProviderRef provider = NULL;
+   LLVMPassManagerRef pass = NULL;
+   char *error = NULL;
+   conv_test_ptr_t conv_test_ptr;
+   boolean success;
+   const unsigned n = 32;
+   int64_t cycles[n];
+   double cycles_avg = 0.0;
+   unsigned num_srcs;
+   unsigned num_dsts;
+   double eps;
+   unsigned i, j;
+
+   if(verbose >= 1)
+      dump_conv_types(stdout, src_type, dst_type);
+
+   if(src_type.length > dst_type.length) {
+      num_srcs = 1;
+      num_dsts = src_type.length/dst_type.length;
+   }
+   else  {
+      num_dsts = 1;
+      num_srcs = dst_type.length/src_type.length;
+   }
+
+   assert(src_type.width * src_type.length == dst_type.width * dst_type.length);
+
+   /* We must not loose or gain channels. Only precision */
+   assert(src_type.length * num_srcs == dst_type.length * num_dsts);
+
+   eps = MAX2(lp_const_eps(src_type), lp_const_eps(dst_type));
+
+   module = LLVMModuleCreateWithName("test");
+
+   func = add_conv_test(module, src_type, num_srcs, dst_type, num_dsts);
+
+   if(LLVMVerifyModule(module, LLVMPrintMessageAction, &error)) {
+      LLVMDumpModule(module);
+      abort();
+   }
+   LLVMDisposeMessage(error);
+
+   provider = LLVMCreateModuleProviderForExistingModule(module);
+   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
+      if(verbose < 1)
+         dump_conv_types(stderr, src_type, dst_type);
+      fprintf(stderr, "%s\n", error);
+      LLVMDisposeMessage(error);
+      abort();
+   }
+
+#if 0
+   pass = LLVMCreatePassManager();
+   LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   LLVMAddConstantPropagationPass(pass);
+   LLVMAddInstructionCombiningPass(pass);
+   LLVMAddPromoteMemoryToRegisterPass(pass);
+   LLVMAddGVNPass(pass);
+   LLVMAddCFGSimplificationPass(pass);
+   LLVMRunPassManager(pass, module);
+#else
+   (void)pass;
+#endif
+
+   if(verbose >= 2)
+      LLVMDumpModule(module);
+
+   conv_test_ptr = (conv_test_ptr_t)LLVMGetPointerToGlobal(engine, func);
+
+   if(verbose >= 2)
+      lp_disassemble(conv_test_ptr);
+
+   success = TRUE;
+   for(i = 0; i < n && success; ++i) {
+      unsigned src_stride = src_type.length*src_type.width/8;
+      unsigned dst_stride = dst_type.length*dst_type.width/8;
+      uint8_t src[LP_MAX_VECTOR_LENGTH*LP_MAX_VECTOR_LENGTH];
+      uint8_t dst[LP_MAX_VECTOR_LENGTH*LP_MAX_VECTOR_LENGTH];
+      double fref[LP_MAX_VECTOR_LENGTH*LP_MAX_VECTOR_LENGTH];
+      uint8_t ref[LP_MAX_VECTOR_LENGTH*LP_MAX_VECTOR_LENGTH];
+      int64_t start_counter = 0;
+      int64_t end_counter = 0;
+
+      for(j = 0; j < num_srcs; ++j) {
+         random_vec(src_type, src + j*src_stride);
+         read_vec(src_type, src + j*src_stride, fref + j*src_type.length);
+      }
+
+      for(j = 0; j < num_dsts; ++j) {
+         write_vec(dst_type, ref + j*dst_stride, fref + j*dst_type.length);
+      }
+
+      start_counter = rdtsc();
+      conv_test_ptr(src, dst);
+      end_counter = rdtsc();
+
+      cycles[i] = end_counter - start_counter;
+
+      for(j = 0; j < num_dsts; ++j) {
+         if(!compare_vec_with_eps(dst_type, dst + j*dst_stride, ref + j*dst_stride, eps))
+            success = FALSE;
+      }
+
+      if (!success) {
+         if(verbose < 1)
+            dump_conv_types(stderr, src_type, dst_type);
+         fprintf(stderr, "MISMATCH\n");
+
+         for(j = 0; j < num_srcs; ++j) {
+            fprintf(stderr, "  Src%u: ", j);
+            dump_vec(stderr, src_type, src + j*src_stride);
+            fprintf(stderr, "\n");
+         }
+
+#if 1
+         fprintf(stderr, "  Ref: ");
+         for(j = 0; j < src_type.length*num_srcs; ++j)
+            fprintf(stderr, " %f", fref[j]);
+         fprintf(stderr, "\n");
+#endif
+
+         for(j = 0; j < num_dsts; ++j) {
+            fprintf(stderr, "  Dst%u: ", j);
+            dump_vec(stderr, dst_type, dst + j*dst_stride);
+            fprintf(stderr, "\n");
+
+            fprintf(stderr, "  Ref%u: ", j);
+            dump_vec(stderr, dst_type, ref + j*dst_stride);
+            fprintf(stderr, "\n");
+         }
+      }
+   }
+
+   /*
+    * Unfortunately the output of cycle counter is not very reliable as it comes
+    * -- sometimes we get outliers (due IRQs perhaps?) which are
+    * better removed to avoid random or biased data.
+    */
+   {
+      double sum = 0.0, sum2 = 0.0;
+      double avg, std;
+      unsigned m;
+
+      for(i = 0; i < n; ++i) {
+         sum += cycles[i];
+         sum2 += cycles[i]*cycles[i];
+      }
+
+      avg = sum/n;
+      std = sqrtf((sum2 - n*avg*avg)/n);
+
+      m = 0;
+      sum = 0.0;
+      for(i = 0; i < n; ++i) {
+         if(fabs(cycles[i] - avg) <= 4.0*std) {
+            sum += cycles[i];
+            ++m;
+         }
+      }
+
+      cycles_avg = sum/m;
+
+   }
+
+   if(fp)
+      write_tsv_row(fp, src_type, dst_type, cycles_avg, success);
+
+   if (!success) {
+      static boolean firsttime = TRUE;
+      if(firsttime) {
+         if(verbose < 2)
+            LLVMDumpModule(module);
+         LLVMWriteBitcodeToFile(module, "conv.bc");
+         fprintf(stderr, "conv.bc written\n");
+         fprintf(stderr, "Invoke as \"llc -o - conv.bc\"\n");
+         firsttime = FALSE;
+         //abort();
+      }
+   }
+
+   LLVMFreeMachineCodeForFunction(engine, func);
+
+   LLVMDisposeExecutionEngine(engine);
+   if(pass)
+      LLVMDisposePassManager(pass);
+
+   return success;
+}
+
+
+const union lp_type conv_types[] = {
+   /* float, fixed,  sign,  norm, width, len */
+
+   {{  TRUE, FALSE,  TRUE,  TRUE,    32,   4 }},
+   {{  TRUE, FALSE,  TRUE, FALSE,    32,   4 }},
+   {{  TRUE, FALSE, FALSE,  TRUE,    32,   4 }},
+   {{  TRUE, FALSE, FALSE, FALSE,    32,   4 }},
+
+   /* TODO: test fixed formats too */
+
+   {{ FALSE, FALSE,  TRUE,  TRUE,    16,   8 }},
+   {{ FALSE, FALSE,  TRUE, FALSE,    16,   8 }},
+   {{ FALSE, FALSE, FALSE,  TRUE,    16,   8 }},
+   {{ FALSE, FALSE, FALSE, FALSE,    16,   8 }},
+
+   {{ FALSE, FALSE,  TRUE,  TRUE,    32,   4 }},
+   {{ FALSE, FALSE,  TRUE, FALSE,    32,   4 }},
+   {{ FALSE, FALSE, FALSE,  TRUE,    32,   4 }},
+   {{ FALSE, FALSE, FALSE, FALSE,    32,   4 }},
+
+   {{ FALSE, FALSE,  TRUE,  TRUE,    16,   8 }},
+   {{ FALSE, FALSE,  TRUE, FALSE,    16,   8 }},
+   {{ FALSE, FALSE, FALSE,  TRUE,    16,   8 }},
+   {{ FALSE, FALSE, FALSE, FALSE,    16,   8 }},
+
+   {{ FALSE, FALSE,  TRUE,  TRUE,     8,  16 }},
+   {{ FALSE, FALSE,  TRUE, FALSE,     8,  16 }},
+   {{ FALSE, FALSE, FALSE,  TRUE,     8,  16 }},
+   {{ FALSE, FALSE, FALSE, FALSE,     8,  16 }},
+};
+
+
+const unsigned num_types = sizeof(conv_types)/sizeof(conv_types[0]);
+
+
+boolean
+test_all(unsigned verbose, FILE *fp)
+{
+   const union lp_type *src_type;
+   const union lp_type *dst_type;
+   bool success = TRUE;
+
+   for(src_type = conv_types; src_type < &conv_types[num_types]; ++src_type) {
+      for(dst_type = conv_types; dst_type < &conv_types[num_types]; ++dst_type) {
+
+         if(src_type == dst_type)
+            continue;
+
+         if(src_type->norm != dst_type->norm)
+            continue;
+
+         if(!test_one(verbose, fp, *src_type, *dst_type))
+           success = FALSE;
+
+      }
+   }
+
+   return success;
+}
+
+
+boolean
+test_some(unsigned verbose, FILE *fp, unsigned long n)
+{
+   const union lp_type *src_type;
+   const union lp_type *dst_type;
+   unsigned long i;
+   bool success = TRUE;
+
+   for(i = 0; i < n; ++i) {
+      src_type = &conv_types[random() % num_types];
+      
+      do {
+         dst_type = &conv_types[random() % num_types];
+      } while (src_type == dst_type || src_type->norm != dst_type->norm);
+
+      if(!test_one(verbose, fp, *src_type, *dst_type))
+        success = FALSE;
+   }
+
+   return success;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_test_format.c b/src/gallium/drivers/llvmpipe/lp_test_format.c
new file mode 100644
index 0000000000..1d192355ee
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test_format.c
@@ -0,0 +1,272 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include <stdlib.h>
+#include <stdio.h>
+
+#include <llvm-c/Core.h>
+#include <llvm-c/Analysis.h>
+#include <llvm-c/ExecutionEngine.h>
+#include <llvm-c/Target.h>
+#include <llvm-c/Transforms/Scalar.h>
+
+#include "util/u_format.h"
+
+#include "lp_bld_flow.h"
+#include "lp_bld_format.h"
+
+
+struct pixel_test_case
+{
+   enum pipe_format format;
+   uint32_t packed;
+   double unpacked[4];
+};
+
+
+struct pixel_test_case test_cases[] =
+{
+   {PIPE_FORMAT_R5G6B5_UNORM,   0x0000, {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5G6B5_UNORM,   0x001f, {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_R5G6B5_UNORM,   0x07e0, {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5G6B5_UNORM,   0xf800, {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5G6B5_UNORM,   0xffff, {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0x0000, {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0x001f, {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0x03e0, {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0x7c00, {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0x8000, {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_A1R5G5B5_UNORM, 0xffff, {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0x00000000, {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0x000000ff, {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0x0000ff00, {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0x00ff0000, {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0xff000000, {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, 0xffffffff, {1.0, 1.0, 1.0, 1.0}},
+
+#if 0
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0x00000000, {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0x000000ff, {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0x0000ff00, {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0x00ff0000, {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0xff000000, {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, 0xffffffff, {1.0, 1.0, 1.0, 1.0}},
+#endif
+
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0x00000000, {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0x000000ff, {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0x0000ff00, {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0x00ff0000, {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0xff000000, {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, 0xffffffff, {1.0, 1.0, 1.0, 1.0}},
+};
+
+
+typedef void (*load_ptr_t)(const void *, float *);
+
+
+static LLVMValueRef
+add_load_rgba_test(LLVMModuleRef module,
+                   enum pipe_format format)
+{
+   LLVMTypeRef args[2];
+   LLVMValueRef func;
+   LLVMValueRef ptr;
+   LLVMValueRef rgba_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef rgba;
+   struct lp_build_loop_state loop;
+
+   args[0] = LLVMPointerType(LLVMInt8Type(), 0);
+   args[1] = LLVMPointerType(LLVMVectorType(LLVMFloatType(), 4), 0);
+
+   func = LLVMAddFunction(module, "load", LLVMFunctionType(LLVMVoidType(), args, 2, 0));
+   LLVMSetFunctionCallConv(func, LLVMCCallConv);
+   ptr = LLVMGetParam(func, 0);
+   rgba_ptr = LLVMGetParam(func, 1);
+
+   block = LLVMAppendBasicBlock(func, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   lp_build_loop_begin(builder, LLVMConstInt(LLVMInt32Type(), 1, 0), &loop);
+
+   rgba = lp_build_load_rgba(builder, format, ptr);
+   LLVMBuildStore(builder, rgba, rgba_ptr);
+
+   lp_build_loop_end(builder, LLVMConstInt(LLVMInt32Type(), 4, 0), NULL, &loop);
+
+   LLVMBuildRetVoid(builder);
+
+   LLVMDisposeBuilder(builder);
+   return func;
+}
+
+
+typedef void (*store_ptr_t)(void *, const float *);
+
+
+static LLVMValueRef
+add_store_rgba_test(LLVMModuleRef module,
+                    enum pipe_format format)
+{
+   LLVMTypeRef args[2];
+   LLVMValueRef func;
+   LLVMValueRef ptr;
+   LLVMValueRef rgba_ptr;
+   LLVMBasicBlockRef block;
+   LLVMBuilderRef builder;
+   LLVMValueRef rgba;
+
+   args[0] = LLVMPointerType(LLVMInt8Type(), 0);
+   args[1] = LLVMPointerType(LLVMVectorType(LLVMFloatType(), 4), 0);
+
+   func = LLVMAddFunction(module, "store", LLVMFunctionType(LLVMVoidType(), args, 2, 0));
+   LLVMSetFunctionCallConv(func, LLVMCCallConv);
+   ptr = LLVMGetParam(func, 0);
+   rgba_ptr = LLVMGetParam(func, 1);
+
+   block = LLVMAppendBasicBlock(func, "entry");
+   builder = LLVMCreateBuilder();
+   LLVMPositionBuilderAtEnd(builder, block);
+
+   rgba = LLVMBuildLoad(builder, rgba_ptr, "");
+
+   lp_build_store_rgba(builder, format, ptr, rgba);
+
+   LLVMBuildRetVoid(builder);
+
+   LLVMDisposeBuilder(builder);
+   return func;
+}
+
+
+static boolean
+test_format(const struct pixel_test_case *test)
+{
+   LLVMModuleRef module = NULL;
+   LLVMValueRef load = NULL;
+   LLVMValueRef store = NULL;
+   LLVMExecutionEngineRef engine = NULL;
+   LLVMModuleProviderRef provider = NULL;
+   LLVMPassManagerRef pass = NULL;
+   char *error = NULL;
+   const struct util_format_description *desc;
+   load_ptr_t load_ptr;
+   store_ptr_t store_ptr;
+   float unpacked[4];
+   unsigned packed;
+   boolean success;
+   unsigned i;
+
+   desc = util_format_description(test->format);
+   fprintf(stderr, "%s\n", desc->name);
+
+   module = LLVMModuleCreateWithName("test");
+
+   load = add_load_rgba_test(module, test->format);
+   store = add_store_rgba_test(module, test->format);
+
+   if(LLVMVerifyModule(module, LLVMPrintMessageAction, &error)) {
+      LLVMDumpModule(module);
+      abort();
+   }
+   LLVMDisposeMessage(error);
+
+   provider = LLVMCreateModuleProviderForExistingModule(module);
+   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
+      fprintf(stderr, "%s\n", error);
+      LLVMDisposeMessage(error);
+      abort();
+   }
+
+#if 0
+   pass = LLVMCreatePassManager();
+   LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
+   /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
+    * but there are more on SVN. */
+   LLVMAddConstantPropagationPass(pass);
+   LLVMAddInstructionCombiningPass(pass);
+   LLVMAddPromoteMemoryToRegisterPass(pass);
+   LLVMAddGVNPass(pass);
+   LLVMAddCFGSimplificationPass(pass);
+   LLVMRunPassManager(pass, module);
+#else
+   (void)pass;
+#endif
+
+   load_ptr  = (load_ptr_t) LLVMGetPointerToGlobal(engine, load);
+   store_ptr = (store_ptr_t)LLVMGetPointerToGlobal(engine, store);
+
+   memset(unpacked, 0, sizeof unpacked);
+   packed = 0;
+
+   load_ptr(&test->packed, unpacked);
+   store_ptr(&packed, unpacked);
+
+   success = TRUE;
+   if(test->packed != packed)
+      success = FALSE;
+   for(i = 0; i < 4; ++i)
+      if(test->unpacked[i] != unpacked[i])
+         success = FALSE;
+
+   if (!success) {
+      printf("FAILED\n");
+      printf("  Packed: %08x\n", test->packed);
+      printf("          %08x\n", packed);
+      printf("  Unpacked: %f %f %f %f\n", unpacked[0], unpacked[1], unpacked[2], unpacked[3]);
+      printf("            %f %f %f %f\n", test->unpacked[0], test->unpacked[1], test->unpacked[2], test->unpacked[3]);
+      LLVMDumpModule(module);
+   }
+
+   LLVMFreeMachineCodeForFunction(engine, store);
+   LLVMFreeMachineCodeForFunction(engine, load);
+
+   LLVMDisposeExecutionEngine(engine);
+   if(pass)
+      LLVMDisposePassManager(pass);
+
+   return success;
+}
+
+
+int main(int argc, char **argv)
+{
+   unsigned i;
+   int ret;
+
+   for (i = 0; i < sizeof(test_cases)/sizeof(test_cases[0]); ++i)
+      if(!test_format(&test_cases[i]))
+        ret = 1;
+
+   return ret;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_test_main.c b/src/gallium/drivers/llvmpipe/lp_test_main.c
new file mode 100644
index 0000000000..49213fb4f0
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_test_main.c
@@ -0,0 +1,384 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Shared testing code.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "lp_bld_const.h"
+#include "lp_test.h"
+
+
+void
+dump_type(FILE *fp,
+          union lp_type type)
+{
+   fprintf(fp, "%s%s%u%sx%u",
+           type.sign ? (type.floating || type.fixed ? "" : "s") : "u",
+           type.floating ? "f" : (type.fixed ? "h" : "i"),
+           type.width,
+           type.norm ? "n" : "",
+           type.length);
+}
+
+
+double
+read_elem(union lp_type type, const void *src, unsigned index)
+{
+   double scale = lp_const_scale(type);
+   double value;
+   assert(index < type.length);
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         value = *((const float *)src + index);
+         break;
+      case 64:
+         value =  *((const double *)src + index);
+         break;
+      default:
+         assert(0);
+         return 0.0;
+      }
+   }
+   else {
+      if(type.sign) {
+         switch(type.width) {
+         case 8:
+            value = *((const int8_t *)src + index);
+            break;
+         case 16:
+            value = *((const int16_t *)src + index);
+            break;
+         case 32:
+            value = *((const int32_t *)src + index);
+            break;
+         case 64:
+            value = *((const int64_t *)src + index);
+            break;
+         default:
+            assert(0);
+            return 0.0;
+         }
+      }
+      else {
+         switch(type.width) {
+         case 8:
+            value = *((const uint8_t *)src + index);
+            break;
+         case 16:
+            value = *((const uint16_t *)src + index);
+            break;
+         case 32:
+            value = *((const uint32_t *)src + index);
+            break;
+         case 64:
+            value = *((const uint64_t *)src + index);
+            break;
+         default:
+            assert(0);
+            return 0.0;
+         }
+      }
+   }
+   return value/scale;
+}
+
+
+void
+write_elem(union lp_type type, void *dst, unsigned index, double value)
+{
+   assert(index < type.length);
+   if(!type.sign && value < 0.0)
+      value = 0.0;
+   if(type.norm && value < -1.0)
+      value = -1.0;
+   if(type.norm && value > 1.0)
+      value = 1.0;
+   if (type.floating) {
+      switch(type.width) {
+      case 32:
+         *((float *)dst + index) = (float)(value);
+         break;
+      case 64:
+          *((double *)dst + index) = value;
+         break;
+      default:
+         assert(0);
+      }
+   }
+   else {
+      double scale = lp_const_scale(type);
+      value = round(value*scale);
+      if(type.sign) {
+         long long lvalue = (long long)value;
+         lvalue = MIN2(lvalue, ((long long)1 << (type.width - 1)) - 1);
+         switch(type.width) {
+         case 8:
+            *((int8_t *)dst + index) = (int8_t)lvalue;
+            break;
+         case 16:
+            *((int16_t *)dst + index) = (int16_t)lvalue;
+            break;
+         case 32:
+            *((int32_t *)dst + index) = (int32_t)lvalue;
+            break;
+         case 64:
+            *((int64_t *)dst + index) = (int64_t)lvalue;
+            break;
+         default:
+            assert(0);
+         }
+      }
+      else {
+         unsigned long long lvalue = (long long)value;
+         lvalue = MIN2(lvalue, ((unsigned long long)1 << type.width) - 1);
+         switch(type.width) {
+         case 8:
+            *((uint8_t *)dst + index) = (uint8_t)lvalue;
+            break;
+         case 16:
+            *((uint16_t *)dst + index) = (uint16_t)lvalue;
+            break;
+         case 32:
+            *((uint32_t *)dst + index) = (uint32_t)lvalue;
+            break;
+         case 64:
+            *((uint64_t *)dst + index) = (uint64_t)lvalue;
+            break;
+         default:
+            assert(0);
+         }
+      }
+   }
+}
+
+
+void
+random_elem(union lp_type type, void *dst, unsigned index)
+{
+   double value;
+   assert(index < type.length);
+   value = (double)random()/(double)RAND_MAX;
+   if(!type.norm) {
+      unsigned long long mask;
+      if (type.floating)
+         mask = ~(unsigned long long)0;
+      else if (type.fixed)
+         mask = ((unsigned long long)1 << (type.width / 2)) - 1;
+      else if (type.sign)
+         mask = ((unsigned long long)1 << (type.width - 1)) - 1;
+      else
+         mask = ((unsigned long long)1 << type.width) - 1;
+      value += (double)(mask & random());
+   }
+   if(!type.sign)
+      if(random() & 1)
+         value = -value;
+   write_elem(type, dst, index, value);
+}
+
+
+void
+read_vec(union lp_type type, const void *src, double *dst)
+{
+   unsigned i;
+   for (i = 0; i < type.length; ++i)
+      dst[i] = read_elem(type, src, i);
+}
+
+
+void
+write_vec(union lp_type type, void *dst, const double *src)
+{
+   unsigned i;
+   for (i = 0; i < type.length; ++i)
+      write_elem(type, dst, i, src[i]);
+}
+
+
+float
+random_float(void)
+{
+    return (float)((double)random()/(double)RAND_MAX);
+}
+
+
+void
+random_vec(union lp_type type, void *dst)
+{
+   unsigned i;
+   for (i = 0; i < type.length; ++i)
+      random_elem(type, dst, i);
+}
+
+
+boolean
+compare_vec_with_eps(union lp_type type, const void *res, const void *ref, double eps)
+{
+   unsigned i;
+   for (i = 0; i < type.length; ++i) {
+      double res_elem = read_elem(type, res, i);
+      double ref_elem = read_elem(type, ref, i);
+      double delta = fabs(res_elem - ref_elem);
+      if(delta >= 2.0*eps)
+         return FALSE;
+   }
+
+   return TRUE;
+}
+
+
+boolean
+compare_vec(union lp_type type, const void *res, const void *ref)
+{
+   double eps = lp_const_eps(type);
+   return compare_vec_with_eps(type, res, ref, eps);
+}
+
+
+void
+dump_vec(FILE *fp, union lp_type type, const void *src)
+{
+   unsigned i;
+   for (i = 0; i < type.length; ++i) {
+      if(i)
+         fprintf(fp, " ");
+      if (type.floating) {
+         double value;
+         switch(type.width) {
+         case 32:
+            value = *((const float *)src + i);
+            break;
+         case 64:
+            value = *((const double *)src + i);
+            break;
+         default:
+            assert(0);
+            value = 0.0;
+         }
+         fprintf(fp, "%f", value);
+      }
+      else {
+         if(type.sign && !type.norm) {
+            long long value;
+            const char *format;
+            switch(type.width) {
+            case 8:
+               value = *((const int8_t *)src + i);
+               format = "%3lli";
+               break;
+            case 16:
+               value = *((const int16_t *)src + i);
+               format = "%5lli";
+               break;
+            case 32:
+               value = *((const int32_t *)src + i);
+               format = "%10lli";
+               break;
+            case 64:
+               value = *((const int64_t *)src + i);
+               format = "%20lli";
+               break;
+            default:
+               assert(0);
+               value = 0.0;
+               format = "?";
+            }
+            fprintf(fp, format, value);
+         }
+         else {
+            unsigned long long value;
+            const char *format;
+            switch(type.width) {
+            case 8:
+               value = *((const uint8_t *)src + i);
+               format = type.norm ? "%2x" : "%4llu";
+               break;
+            case 16:
+               value = *((const uint16_t *)src + i);
+               format = type.norm ? "%4x" : "%6llx";
+               break;
+            case 32:
+               value = *((const uint32_t *)src + i);
+               format = type.norm ? "%8x" : "%11llx";
+               break;
+            case 64:
+               value = *((const uint64_t *)src + i);
+               format = type.norm ? "%16x" : "%21llx";
+               break;
+            default:
+               assert(0);
+               value = 0.0;
+               format = "?";
+            }
+            fprintf(fp, format, value);
+         }
+      }
+   }
+}
+
+
+int main(int argc, char **argv)
+{
+   unsigned verbose = 0;
+   FILE *fp = NULL;
+   unsigned long n = 1000;
+   unsigned i;
+   boolean success;
+
+   for(i = 1; i < argc; ++i) {
+      if(strcmp(argv[i], "-v") == 0)
+         ++verbose;
+      else if(strcmp(argv[i], "-o") == 0)
+         fp = fopen(argv[++i], "wt");
+      else
+         n = atoi(argv[i]);
+   }
+
+   if(fp) {
+      /* Warm up the caches */
+      test_some(0, NULL, 100);
+
+      write_tsv_header(fp);
+   }
+      
+   if(n)
+      success = test_some(verbose, fp, n);
+   else
+      success = test_all(verbose, fp);
+
+   if(fp)
+      fclose(fp);
+
+   return success ? 0 : 1;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
new file mode 100644
index 0000000000..23a94b5b0d
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
@@ -0,0 +1,304 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Texture tile caching.
+ *
+ * Author:
+ *    Brian Paul
+ */
+
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "util/u_tile.h"
+#include "util/u_format.h"
+#include "lp_context.h"
+#include "lp_surface.h"
+#include "lp_texture.h"
+#include "lp_tex_cache.h"
+
+
+
+/**
+ * Return the position in the cache for the tile that contains win pos (x,y).
+ * We currently use a direct mapped cache so this is like a hack key.
+ * At some point we should investige something more sophisticated, like
+ * a LRU replacement policy.
+ */
+#define CACHE_POS(x, y) \
+   (((x) + (y) * 5) % NUM_ENTRIES)
+
+
+
+/**
+ * Is the tile at (x,y) in cleared state?
+ */
+static INLINE uint
+is_clear_flag_set(const uint *bitvec, union tex_tile_address addr)
+{
+   int pos, bit;
+   pos = addr.bits.y * (MAX_TEX_WIDTH / TEX_TILE_SIZE) + addr.bits.x;
+   assert(pos / 32 < (MAX_TEX_WIDTH / TEX_TILE_SIZE) * (MAX_TEX_HEIGHT / TEX_TILE_SIZE) / 32);
+   bit = bitvec[pos / 32] & (1 << (pos & 31));
+   return bit;
+}
+   
+
+/**
+ * Mark the tile at (x,y) as not cleared.
+ */
+static INLINE void
+clear_clear_flag(uint *bitvec, union tex_tile_address addr)
+{
+   int pos;
+   pos = addr.bits.y * (MAX_TEX_WIDTH / TEX_TILE_SIZE) + addr.bits.x;
+   assert(pos / 32 < (MAX_TEX_WIDTH / TEX_TILE_SIZE) * (MAX_TEX_HEIGHT / TEX_TILE_SIZE) / 32);
+   bitvec[pos / 32] &= ~(1 << (pos & 31));
+}
+   
+
+struct llvmpipe_tex_tile_cache *
+lp_create_tex_tile_cache( struct pipe_screen *screen )
+{
+   struct llvmpipe_tex_tile_cache *tc;
+   uint pos;
+
+   tc = CALLOC_STRUCT( llvmpipe_tex_tile_cache );
+   if (tc) {
+      tc->screen = screen;
+      for (pos = 0; pos < NUM_ENTRIES; pos++) {
+         tc->entries[pos].addr.bits.invalid = 1;
+      }
+      tc->last_tile = &tc->entries[0]; /* any tile */
+   }
+   return tc;
+}
+
+
+void
+lp_destroy_tex_tile_cache(struct llvmpipe_tex_tile_cache *tc)
+{
+   struct pipe_screen *screen;
+   uint pos;
+
+   for (pos = 0; pos < NUM_ENTRIES; pos++) {
+      /*assert(tc->entries[pos].x < 0);*/
+   }
+   if (tc->transfer) {
+      screen = tc->transfer->texture->screen;
+      screen->tex_transfer_destroy(tc->transfer);
+   }
+   if (tc->tex_trans) {
+      screen = tc->tex_trans->texture->screen;
+      screen->tex_transfer_destroy(tc->tex_trans);
+   }
+
+   FREE( tc );
+}
+
+
+void
+lp_tex_tile_cache_map_transfers(struct llvmpipe_tex_tile_cache *tc)
+{
+   if (tc->transfer && !tc->transfer_map)
+      tc->transfer_map = tc->screen->transfer_map(tc->screen, tc->transfer);
+
+   if (tc->tex_trans && !tc->tex_trans_map)
+      tc->tex_trans_map = tc->screen->transfer_map(tc->screen, tc->tex_trans);
+}
+
+
+void
+lp_tex_tile_cache_unmap_transfers(struct llvmpipe_tex_tile_cache *tc)
+{
+   if (tc->transfer_map) {
+      tc->screen->transfer_unmap(tc->screen, tc->transfer);
+      tc->transfer_map = NULL;
+   }
+
+   if (tc->tex_trans_map) {
+      tc->screen->transfer_unmap(tc->screen, tc->tex_trans);
+      tc->tex_trans_map = NULL;
+   }
+}
+
+void
+lp_tex_tile_cache_validate_texture(struct llvmpipe_tex_tile_cache *tc)
+{
+   if (tc->texture) {
+      struct llvmpipe_texture *lpt = llvmpipe_texture(tc->texture);
+      if (lpt->timestamp != tc->timestamp) {
+         /* texture was modified, invalidate all cached tiles */
+         uint i;
+         _debug_printf("INV %d %d\n", tc->timestamp, lpt->timestamp);
+         for (i = 0; i < NUM_ENTRIES; i++) {
+            tc->entries[i].addr.bits.invalid = 1;
+         }
+
+         tc->timestamp = lpt->timestamp;
+      }
+   }
+}
+
+/**
+ * Specify the texture to cache.
+ */
+void
+lp_tex_tile_cache_set_texture(struct llvmpipe_tex_tile_cache *tc,
+                          struct pipe_texture *texture)
+{
+   uint i;
+
+   assert(!tc->transfer);
+
+   if (tc->texture != texture) {
+      pipe_texture_reference(&tc->texture, texture);
+
+      if (tc->tex_trans) {
+         struct pipe_screen *screen = tc->tex_trans->texture->screen;
+         
+         if (tc->tex_trans_map) {
+            screen->transfer_unmap(screen, tc->tex_trans);
+            tc->tex_trans_map = NULL;
+         }
+
+         screen->tex_transfer_destroy(tc->tex_trans);
+         tc->tex_trans = NULL;
+      }
+
+      /* mark as entries as invalid/empty */
+      /* XXX we should try to avoid this when the teximage hasn't changed */
+      for (i = 0; i < NUM_ENTRIES; i++) {
+         tc->entries[i].addr.bits.invalid = 1;
+      }
+
+      tc->tex_face = -1; /* any invalid value here */
+   }
+}
+
+
+/**
+ * Given the texture face, level, zslice, x and y values, compute
+ * the cache entry position/index where we'd hope to find the
+ * cached texture tile.
+ * This is basically a direct-map cache.
+ * XXX There's probably lots of ways in which we can improve this.
+ */
+static INLINE uint
+tex_cache_pos( union tex_tile_address addr )
+{
+   uint entry = (addr.bits.x + 
+                 addr.bits.y * 9 + 
+                 addr.bits.z * 3 + 
+                 addr.bits.face + 
+                 addr.bits.level * 7);
+
+   return entry % NUM_ENTRIES;
+}
+
+/**
+ * Similar to lp_get_cached_tile() but for textures.
+ * Tiles are read-only and indexed with more params.
+ */
+const struct llvmpipe_cached_tex_tile *
+lp_find_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc,
+                        union tex_tile_address addr )
+{
+   struct pipe_screen *screen = tc->screen;
+   struct llvmpipe_cached_tex_tile *tile;
+   
+   tile = tc->entries + tex_cache_pos( addr );
+
+   if (addr.value != tile->addr.value) {
+
+      /* cache miss.  Most misses are because we've invaldiated the
+       * texture cache previously -- most commonly on binding a new
+       * texture.  Currently we effectively flush the cache on texture
+       * bind.
+       */
+#if 0
+      _debug_printf("miss at %u:  x=%d y=%d z=%d face=%d level=%d\n"
+                    "   tile %u:  x=%d y=%d z=%d face=%d level=%d\n",
+                    pos, x/TEX_TILE_SIZE, y/TEX_TILE_SIZE, z, face, level,
+                    pos, tile->addr.bits.x, tile->addr.bits.y, tile->z, tile->face, tile->level);
+#endif
+
+      /* check if we need to get a new transfer */
+      if (!tc->tex_trans ||
+          tc->tex_face != addr.bits.face ||
+          tc->tex_level != addr.bits.level ||
+          tc->tex_z != addr.bits.z) {
+         /* get new transfer (view into texture) */
+
+         if (tc->tex_trans) {
+            if (tc->tex_trans_map) {
+               tc->screen->transfer_unmap(tc->screen, tc->tex_trans);
+               tc->tex_trans_map = NULL;
+            }
+
+            screen->tex_transfer_destroy(tc->tex_trans);
+            tc->tex_trans = NULL;
+         }
+
+         tc->tex_trans = 
+            screen->get_tex_transfer(screen, tc->texture, 
+                                     addr.bits.face, 
+                                     addr.bits.level, 
+                                     addr.bits.z, 
+                                     PIPE_TRANSFER_READ, 0, 0,
+                                     tc->texture->width[addr.bits.level],
+                                     tc->texture->height[addr.bits.level]);
+
+         tc->tex_trans_map = screen->transfer_map(screen, tc->tex_trans);
+
+         tc->tex_face = addr.bits.face;
+         tc->tex_level = addr.bits.level;
+         tc->tex_z = addr.bits.z;
+      }
+
+      {
+         unsigned x = addr.bits.x * TEX_TILE_SIZE;
+         unsigned y = addr.bits.y * TEX_TILE_SIZE;
+         unsigned w = TEX_TILE_SIZE;
+         unsigned h = TEX_TILE_SIZE;
+
+         if (pipe_clip_tile(x, y, &w, &h, tc->tex_trans)) {
+            assert(0);
+         }
+
+         util_format_read_4ub(tc->tex_trans->format,
+                              (uint8_t *)tile->color, sizeof tile->color[0],
+                              tc->tex_trans_map, tc->tex_trans->stride,
+                              x, y, w, h);
+      }
+
+      tile->addr = addr;
+   }
+
+   tc->last_tile = tile;
+   return tile;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.h b/src/gallium/drivers/llvmpipe/lp_tex_cache.h
new file mode 100644
index 0000000000..9fa6c36812
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.h
@@ -0,0 +1,151 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_TEX_CACHE_H
+#define LP_TEX_CACHE_H
+
+
+#include "pipe/p_compiler.h"
+
+
+struct llvmpipe_context;
+struct llvmpipe_tex_tile_cache;
+
+
+/**
+ * Cache tile size (width and height). This needs to be a power of two.
+ */
+#define TEX_TILE_SIZE 64
+
+
+/* If we need to support > 4096, just expand this to be a 64 bit
+ * union, or consider tiling in Z as well.
+ */
+union tex_tile_address {
+   struct {
+      unsigned x:6;             /* 4096 / TEX_TILE_SIZE */
+      unsigned y:6;             /* 4096 / TEX_TILE_SIZE */
+      unsigned z:12;            /* 4096 -- z not tiled */
+      unsigned face:3;
+      unsigned level:4;
+      unsigned invalid:1;
+   } bits;
+   unsigned value;
+};
+
+
+struct llvmpipe_cached_tex_tile
+{
+   union tex_tile_address addr;
+   uint8_t color[TEX_TILE_SIZE][TEX_TILE_SIZE][4];
+};
+
+#define NUM_ENTRIES 50
+
+
+/** XXX move these */
+#define MAX_TEX_WIDTH 2048
+#define MAX_TEX_HEIGHT 2048
+
+
+struct llvmpipe_tex_tile_cache
+{
+   struct pipe_screen *screen;
+   struct pipe_surface *surface;  /**< the surface we're caching */
+   struct pipe_transfer *transfer;
+   void *transfer_map;
+
+   struct pipe_texture *texture;  /**< if caching a texture */
+   unsigned timestamp;
+
+   struct llvmpipe_cached_tex_tile entries[NUM_ENTRIES];
+
+   struct pipe_transfer *tex_trans;
+   void *tex_trans_map;
+   int tex_face, tex_level, tex_z;
+
+   struct llvmpipe_cached_tex_tile *last_tile;  /**< most recently retrieved tile */
+};
+
+
+extern struct llvmpipe_tex_tile_cache *
+lp_create_tex_tile_cache( struct pipe_screen *screen );
+
+extern void
+lp_destroy_tex_tile_cache(struct llvmpipe_tex_tile_cache *tc);
+
+extern void
+lp_tex_tile_cache_map_transfers(struct llvmpipe_tex_tile_cache *tc);
+
+extern void
+lp_tex_tile_cache_unmap_transfers(struct llvmpipe_tex_tile_cache *tc);
+
+extern void
+lp_tex_tile_cache_set_texture(struct llvmpipe_tex_tile_cache *tc,
+                          struct pipe_texture *texture);
+
+void
+lp_tex_tile_cache_validate_texture(struct llvmpipe_tex_tile_cache *tc);
+
+extern const struct llvmpipe_cached_tex_tile *
+lp_find_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc,
+                        union tex_tile_address addr );
+
+static INLINE const union tex_tile_address
+tex_tile_address( unsigned x,
+                  unsigned y,
+                  unsigned z,
+                  unsigned face,
+                  unsigned level )
+{
+   union tex_tile_address addr;
+
+   addr.value = 0;
+   addr.bits.x = x / TEX_TILE_SIZE;
+   addr.bits.y = y / TEX_TILE_SIZE;
+   addr.bits.z = z;
+   addr.bits.face = face;
+   addr.bits.level = level;
+      
+   return addr;
+}
+
+/* Quickly retrieve tile if it matches last lookup.
+ */
+static INLINE const struct llvmpipe_cached_tex_tile *
+lp_get_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc,
+                       union tex_tile_address addr )
+{
+   if (tc->last_tile->addr.value == addr.value)
+      return tc->last_tile;
+
+   return lp_find_cached_tex_tile( tc, addr );
+}
+
+
+#endif /* LP_TEX_CACHE_H */
+
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample.c b/src/gallium/drivers/llvmpipe/lp_tex_sample.c
new file mode 100644
index 0000000000..94eb6dad5a
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample.c
@@ -0,0 +1,1580 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Texture sampling
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "lp_context.h"
+#include "lp_quad.h"
+#include "lp_surface.h"
+#include "lp_texture.h"
+#include "lp_tex_sample.h"
+#include "lp_tex_cache.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+
+/*
+ * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
+ * see 1-pixel bands of improperly weighted linear-filtered textures.
+ * The tests/texwrap.c demo is a good test.
+ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
+ * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
+ */
+#define FRAC(f)  ((f) - util_ifloor(f))
+
+
+/**
+ * Linear interpolation macro
+ */
+static INLINE float
+lerp(float a, float v0, float v1)
+{
+   return v0 + a * (v1 - v0);
+}
+
+
+/**
+ * Do 2D/biliner interpolation of float values.
+ * v00, v10, v01 and v11 are typically four texture samples in a square/box.
+ * a and b are the horizontal and vertical interpolants.
+ * It's important that this function is inlined when compiled with
+ * optimization!  If we find that's not true on some systems, convert
+ * to a macro.
+ */
+static INLINE float
+lerp_2d(float a, float b,
+        float v00, float v10, float v01, float v11)
+{
+   const float temp0 = lerp(a, v00, v10);
+   const float temp1 = lerp(a, v01, v11);
+   return lerp(b, temp0, temp1);
+}
+
+
+/**
+ * As above, but 3D interpolation of 8 values.
+ */
+static INLINE float
+lerp_3d(float a, float b, float c,
+        float v000, float v100, float v010, float v110,
+        float v001, float v101, float v011, float v111)
+{
+   const float temp0 = lerp_2d(a, b, v000, v100, v010, v110);
+   const float temp1 = lerp_2d(a, b, v001, v101, v011, v111);
+   return lerp(c, temp0, temp1);
+}
+
+
+
+/**
+ * If A is a signed integer, A % B doesn't give the right value for A < 0
+ * (in terms of texture repeat).  Just casting to unsigned fixes that.
+ */
+#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B))
+
+
+/**
+ * Apply texture coord wrapping mode and return integer texture indexes
+ * for a vector of four texcoords (S or T or P).
+ * \param wrapMode  PIPE_TEX_WRAP_x
+ * \param s  the incoming texcoords
+ * \param size  the texture image size
+ * \param icoord  returns the integer texcoords
+ * \return  integer texture index
+ */
+static INLINE void
+nearest_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+                   int icoord[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      /* s limited to [0,1) */
+      /* i limited to [0,size-1] */
+      for (ch = 0; ch < 4; ch++) {
+         int i = util_ifloor(s[ch] * size);
+         icoord[ch] = REMAINDER(i, size);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP:
+      /* s limited to [0,1] */
+      /* i limited to [0,size-1] */
+      for (ch = 0; ch < 4; ch++) {
+         if (s[ch] <= 0.0F)
+            icoord[ch] = 0;
+         else if (s[ch] >= 1.0F)
+            icoord[ch] = size - 1;
+         else
+            icoord[ch] = util_ifloor(s[ch] * size);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            if (s[ch] < min)
+               icoord[ch] = 0;
+            else if (s[ch] > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(s[ch] * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [-1, size] */
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            if (s[ch] <= min)
+               icoord[ch] = -1;
+            else if (s[ch] >= max)
+               icoord[ch] = size;
+            else
+               icoord[ch] = util_ifloor(s[ch] * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      {
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const int flr = util_ifloor(s[ch]);
+            float u;
+            if (flr & 1)
+               u = 1.0F - (s[ch] - (float) flr);
+            else
+               u = s[ch] - (float) flr;
+            if (u < min)
+               icoord[ch] = 0;
+            else if (u > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         /* s limited to [0,1] */
+         /* i limited to [0,size-1] */
+         const float u = fabsf(s[ch]);
+         if (u <= 0.0F)
+            icoord[ch] = 0;
+         else if (u >= 1.0F)
+            icoord[ch] = size - 1;
+         else
+            icoord[ch] = util_ifloor(u * size);
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const float u = fabsf(s[ch]);
+            if (u < min)
+               icoord[ch] = 0;
+            else if (u > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const float u = fabsf(s[ch]);
+            if (u < min)
+               icoord[ch] = -1;
+            else if (u > max)
+               icoord[ch] = size;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Used to compute texel locations for linear sampling for four texcoords.
+ * \param wrapMode  PIPE_TEX_WRAP_x
+ * \param s  the texcoords
+ * \param size  the texture image size
+ * \param icoord0  returns first texture indexes
+ * \param icoord1  returns second texture indexes (usually icoord0 + 1)
+ * \param w  returns blend factor/weight between texture indexes
+ * \param icoord  returns the computed integer texture coords
+ */
+static INLINE void
+linear_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+                  int icoord0[4], int icoord1[4], float w[4])
+{
+   uint ch;
+
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      for (ch = 0; ch < 4; ch++) {
+         float u = s[ch] * size - 0.5F;
+         icoord0[ch] = REMAINDER(util_ifloor(u), size);
+         icoord1[ch] = REMAINDER(icoord0[ch] + 1, size);
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.0F, 1.0F);
+         u = u * size - 0.5f;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.0F, 1.0F);
+         u = u * size - 0.5f;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      {
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            float u = CLAMP(s[ch], min, max);
+            u = u * size - 0.5f;
+            icoord0[ch] = util_ifloor(u);
+            icoord1[ch] = icoord0[ch] + 1;
+            w[ch] = FRAC(u);
+         }
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      for (ch = 0; ch < 4; ch++) {
+         const int flr = util_ifloor(s[ch]);
+         float u;
+         if (flr & 1)
+            u = 1.0F - (s[ch] - (float) flr);
+         else
+            u = s[ch] - (float) flr;
+         u = u * size - 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         float u = fabsf(s[ch]);
+         if (u >= 1.0F)
+            u = (float) size;
+         else
+            u *= size;
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      for (ch = 0; ch < 4; ch++) {
+         float u = fabsf(s[ch]);
+         if (u >= 1.0F)
+            u = (float) size;
+         else
+            u *= size;
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            float u = fabsf(s[ch]);
+            if (u <= min)
+               u = min * size;
+            else if (u >= max)
+               u = max * size;
+            else
+               u *= size;
+            u -= 0.5F;
+            icoord0[ch] = util_ifloor(u);
+            icoord1[ch] = icoord0[ch] + 1;
+            w[ch] = FRAC(u);
+         }
+      }
+      break;;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+nearest_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+                          int icoord[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         int i = util_ifloor(s[ch]);
+         icoord[ch]= CLAMP(i, 0, (int) size-1);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      /* fall-through */
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      for (ch = 0; ch < 4; ch++) {
+         icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) );
+      }
+      return;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords.
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+linear_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+                         int icoord0[4], int icoord1[4], float w[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         /* Not exactly what the spec says, but it matches NVIDIA output */
+         float u = CLAMP(s[ch] - 0.5F, 0.0f, (float) size - 1.0f);
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      /* fall-through */
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F);
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord1[ch] > (int) size - 1)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static unsigned
+choose_cube_face(float rx, float ry, float rz, float *newS, float *newT)
+{
+   /*
+      major axis
+      direction     target                             sc     tc    ma
+      ----------    -------------------------------    ---    ---   ---
+       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
+       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
+       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
+       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
+       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
+       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
+   */
+   const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz);
+   unsigned face;
+   float sc, tc, ma;
+
+   if (arx > ary && arx > arz) {
+      if (rx >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_X;
+         sc = -rz;
+         tc = -ry;
+         ma = arx;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_X;
+         sc = rz;
+         tc = -ry;
+         ma = arx;
+      }
+   }
+   else if (ary > arx && ary > arz) {
+      if (ry >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_Y;
+         sc = rx;
+         tc = rz;
+         ma = ary;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Y;
+         sc = rx;
+         tc = -rz;
+         ma = ary;
+      }
+   }
+   else {
+      if (rz > 0.0F) {
+         face = PIPE_TEX_FACE_POS_Z;
+         sc = rx;
+         tc = -ry;
+         ma = arz;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Z;
+         sc = -rx;
+         tc = -ry;
+         ma = arz;
+      }
+   }
+
+   *newS = ( sc / ma + 1.0F ) * 0.5F;
+   *newT = ( tc / ma + 1.0F ) * 0.5F;
+
+   return face;
+}
+
+
+/**
+ * Examine the quad's texture coordinates to compute the partial
+ * derivatives w.r.t X and Y, then compute lambda (level of detail).
+ *
+ * This is only done for fragment shaders, not vertex shaders.
+ */
+static float
+compute_lambda(struct tgsi_sampler *tgsi_sampler,
+               const float s[QUAD_SIZE],
+               const float t[QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               float lodbias)
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+   float rho, lambda;
+
+   if (samp->processor == TGSI_PROCESSOR_VERTEX)
+      return lodbias;
+
+   assert(sampler->normalized_coords);
+
+   assert(s);
+   {
+      float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT];
+      float dsdy = s[QUAD_TOP_LEFT]     - s[QUAD_BOTTOM_LEFT];
+      dsdx = fabsf(dsdx);
+      dsdy = fabsf(dsdy);
+      rho = MAX2(dsdx, dsdy) * texture->width[0];
+   }
+   if (t) {
+      float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT];
+      float dtdy = t[QUAD_TOP_LEFT]     - t[QUAD_BOTTOM_LEFT];
+      float max;
+      dtdx = fabsf(dtdx);
+      dtdy = fabsf(dtdy);
+      max = MAX2(dtdx, dtdy) * texture->height[0];
+      rho = MAX2(rho, max);
+   }
+   if (p) {
+      float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT];
+      float dpdy = p[QUAD_TOP_LEFT]     - p[QUAD_BOTTOM_LEFT];
+      float max;
+      dpdx = fabsf(dpdx);
+      dpdy = fabsf(dpdy);
+      max = MAX2(dpdx, dpdy) * texture->depth[0];
+      rho = MAX2(rho, max);
+   }
+
+   lambda = util_fast_log2(rho);
+   lambda += lodbias + sampler->lod_bias;
+   lambda = CLAMP(lambda, sampler->min_lod, sampler->max_lod);
+
+   return lambda;
+}
+
+
+/**
+ * Do several things here:
+ * 1. Compute lambda from the texcoords, if needed
+ * 2. Determine if we're minifying or magnifying
+ * 3. If minifying, choose mipmap levels
+ * 4. Return image filter to use within mipmap images
+ * \param level0  Returns first mipmap level to sample from
+ * \param level1  Returns second mipmap level to sample from
+ * \param levelBlend  Returns blend factor between levels, in [0,1]
+ * \param imgFilter  Returns either the min or mag filter, depending on lambda
+ */
+static void
+choose_mipmap_levels(struct tgsi_sampler *tgsi_sampler,
+                     const float s[QUAD_SIZE],
+                     const float t[QUAD_SIZE],
+                     const float p[QUAD_SIZE],
+                     float lodbias,
+                     unsigned *level0, unsigned *level1, float *levelBlend,
+                     unsigned *imgFilter)
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+
+   if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
+      /* no mipmap selection needed */
+      *level0 = *level1 = CLAMP((int) sampler->min_lod,
+                                0, (int) texture->last_level);
+
+      if (sampler->min_img_filter != sampler->mag_img_filter) {
+         /* non-mipmapped texture, but still need to determine if doing
+          * minification or magnification.
+          */
+         float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+         if (lambda <= 0.0) {
+            *imgFilter = sampler->mag_img_filter;
+         }
+         else {
+            *imgFilter = sampler->min_img_filter;
+         }
+      }
+      else {
+         *imgFilter = sampler->mag_img_filter;
+      }
+   }
+   else {
+      float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+
+      if (lambda <= 0.0) { /* XXX threshold depends on the filter */
+         /* magnifying */
+         *imgFilter = sampler->mag_img_filter;
+         *level0 = *level1 = 0;
+      }
+      else {
+         /* minifying */
+         *imgFilter = sampler->min_img_filter;
+
+         /* choose mipmap level(s) and compute the blend factor between them */
+         if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) {
+            /* Nearest mipmap level */
+            const int lvl = (int) (lambda + 0.5);
+            *level0 =
+            *level1 = CLAMP(lvl, 0, (int) texture->last_level);
+         }
+         else {
+            /* Linear interpolation between mipmap levels */
+            const int lvl = (int) lambda;
+            *level0 = CLAMP(lvl,     0, (int) texture->last_level);
+            *level1 = CLAMP(lvl + 1, 0, (int) texture->last_level);
+            *levelBlend = FRAC(lambda);  /* blending weight between levels */
+         }
+      }
+   }
+}
+
+
+/**
+ * Get a texel from a texture, using the texture tile cache.
+ *
+ * \param face  the cube face in 0..5
+ * \param level  the mipmap level
+ * \param x  the x coord of texel within 2D image
+ * \param y  the y coord of texel within 2D image
+ * \param z  which slice of a 3D texture
+ * \param rgba  the quad to put the texel/color into
+ * \param j  which element of the rgba quad to write to
+ *
+ * XXX maybe move this into lp_tile_cache.c and merge with the
+ * lp_get_cached_tile_tex() function.  Also, get 4 texels instead of 1...
+ */
+static void
+get_texel_quad_2d(const struct tgsi_sampler *tgsi_sampler,
+                  unsigned face, unsigned level, int x, int y, 
+                  const uint8_t *out[4])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+
+   const struct llvmpipe_cached_tex_tile *tile
+      = lp_get_cached_tex_tile(samp->cache,
+                               tex_tile_address(x, y, 0, face, level));
+
+   y %= TEX_TILE_SIZE;
+   x %= TEX_TILE_SIZE;
+      
+   out[0] = &tile->color[y  ][x  ][0];
+   out[1] = &tile->color[y  ][x+1][0];
+   out[2] = &tile->color[y+1][x  ][0];
+   out[3] = &tile->color[y+1][x+1][0];
+}
+
+static INLINE const uint8_t *
+get_texel_2d_ptr(const struct tgsi_sampler *tgsi_sampler,
+                 unsigned face, unsigned level, int x, int y)
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+
+   const struct llvmpipe_cached_tex_tile *tile
+      = lp_get_cached_tex_tile(samp->cache,
+                               tex_tile_address(x, y, 0, face, level));
+
+   y %= TEX_TILE_SIZE;
+   x %= TEX_TILE_SIZE;
+
+   return &tile->color[y][x][0];
+}
+
+
+static void
+get_texel_quad_2d_mt(const struct tgsi_sampler *tgsi_sampler,
+                     unsigned face, unsigned level, 
+                     int x0, int y0, 
+                     int x1, int y1,
+                     const uint8_t *out[4])
+{
+   unsigned i;
+
+   for (i = 0; i < 4; i++) {
+      unsigned tx = (i & 1) ? x1 : x0;
+      unsigned ty = (i >> 1) ? y1 : y0;
+
+      out[i] = get_texel_2d_ptr( tgsi_sampler, face, level, tx, ty );
+   }
+}
+
+static void
+get_texel(const struct tgsi_sampler *tgsi_sampler,
+                 unsigned face, unsigned level, int x, int y, int z,
+                 float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j)
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+
+   if (x < 0 || x >= (int) texture->width[level] ||
+       y < 0 || y >= (int) texture->height[level] ||
+       z < 0 || z >= (int) texture->depth[level]) {
+      rgba[0][j] = sampler->border_color[0];
+      rgba[1][j] = sampler->border_color[1];
+      rgba[2][j] = sampler->border_color[2];
+      rgba[3][j] = sampler->border_color[3];
+   }
+   else {
+      const unsigned tx = x % TEX_TILE_SIZE;
+      const unsigned ty = y % TEX_TILE_SIZE;
+      const struct llvmpipe_cached_tex_tile *tile;
+
+      tile = lp_get_cached_tex_tile(samp->cache,
+                                    tex_tile_address(x, y, z, face, level));
+
+      rgba[0][j] = ubyte_to_float(tile->color[ty][tx][0]);
+      rgba[1][j] = ubyte_to_float(tile->color[ty][tx][1]);
+      rgba[2][j] = ubyte_to_float(tile->color[ty][tx][2]);
+      rgba[3][j] = ubyte_to_float(tile->color[ty][tx][3]);
+      if (0)
+      {
+         debug_printf("Get texel %f %f %f %f from %s\n",
+                      rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j],
+                      pf_name(texture->format));
+      }
+   }
+}
+
+
+/**
+ * Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
+ * When we sampled the depth texture, the depth value was put into all
+ * RGBA channels.  We look at the red channel here.
+ * \param rgba  quad of (depth) texel values
+ * \param p  texture 'P' components for four pixels in quad
+ * \param j  which pixel in the quad to test [0..3]
+ */
+static INLINE void
+shadow_compare(const struct pipe_sampler_state *sampler,
+               float rgba[NUM_CHANNELS][QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               uint j)
+{
+   int k;
+   switch (sampler->compare_func) {
+   case PIPE_FUNC_LESS:
+      k = p[j] < rgba[0][j];
+      break;
+   case PIPE_FUNC_LEQUAL:
+      k = p[j] <= rgba[0][j];
+      break;
+   case PIPE_FUNC_GREATER:
+      k = p[j] > rgba[0][j];
+      break;
+   case PIPE_FUNC_GEQUAL:
+      k = p[j] >= rgba[0][j];
+      break;
+   case PIPE_FUNC_EQUAL:
+      k = p[j] == rgba[0][j];
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      k = p[j] != rgba[0][j];
+      break;
+   case PIPE_FUNC_ALWAYS:
+      k = 1;
+      break;
+   case PIPE_FUNC_NEVER:
+      k = 0;
+      break;
+   default:
+      k = 0;
+      assert(0);
+      break;
+   }
+
+   /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+   rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k;
+   rgba[3][j] = 1.0F;
+}
+
+
+/**
+ * As above, but do four z/texture comparisons.
+ */
+static INLINE void
+shadow_compare4(const struct pipe_sampler_state *sampler,
+                float rgba[NUM_CHANNELS][QUAD_SIZE],
+                const float p[QUAD_SIZE])
+{
+   int j, k0, k1, k2, k3;
+   float val;
+
+   /* compare four texcoords vs. four texture samples */
+   switch (sampler->compare_func) {
+   case PIPE_FUNC_LESS:
+      k0 = p[0] < rgba[0][0];
+      k1 = p[1] < rgba[0][1];
+      k2 = p[2] < rgba[0][2];
+      k3 = p[3] < rgba[0][3];
+      break;
+   case PIPE_FUNC_LEQUAL:
+      k0 = p[0] <= rgba[0][0];
+      k1 = p[1] <= rgba[0][1];
+      k2 = p[2] <= rgba[0][2];
+      k3 = p[3] <= rgba[0][3];
+      break;
+   case PIPE_FUNC_GREATER:
+      k0 = p[0] > rgba[0][0];
+      k1 = p[1] > rgba[0][1];
+      k2 = p[2] > rgba[0][2];
+      k3 = p[3] > rgba[0][3];
+      break;
+   case PIPE_FUNC_GEQUAL:
+      k0 = p[0] >= rgba[0][0];
+      k1 = p[1] >= rgba[0][1];
+      k2 = p[2] >= rgba[0][2];
+      k3 = p[3] >= rgba[0][3];
+      break;
+   case PIPE_FUNC_EQUAL:
+      k0 = p[0] == rgba[0][0];
+      k1 = p[1] == rgba[0][1];
+      k2 = p[2] == rgba[0][2];
+      k3 = p[3] == rgba[0][3];
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      k0 = p[0] != rgba[0][0];
+      k1 = p[1] != rgba[0][1];
+      k2 = p[2] != rgba[0][2];
+      k3 = p[3] != rgba[0][3];
+      break;
+   case PIPE_FUNC_ALWAYS:
+      k0 = k1 = k2 = k3 = 1;
+      break;
+   case PIPE_FUNC_NEVER:
+      k0 = k1 = k2 = k3 = 0;
+      break;
+   default:
+      k0 = k1 = k2 = k3 = 0;
+      assert(0);
+      break;
+   }
+
+   /* convert four pass/fail values to an intensity in [0,1] */
+   val = 0.25F * (k0 + k1 + k2 + k3);
+
+   /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+   for (j = 0; j < 4; j++) {
+      rgba[0][j] = rgba[1][j] = rgba[2][j] = val;
+      rgba[3][j] = 1.0F;
+   }
+}
+
+
+
+static void
+lp_get_samples_2d_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+                                    const float s[QUAD_SIZE],
+                                    const float t[QUAD_SIZE],
+                                    const float p[QUAD_SIZE],
+                                    float lodbias,
+                                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   unsigned  j;
+   unsigned level = samp->level;
+   unsigned xpot = 1 << (samp->xpot - level);
+   unsigned ypot = 1 << (samp->ypot - level);
+   unsigned xmax = (xpot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, xpot) - 1; */
+   unsigned ymax = (ypot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, ypot) - 1; */
+      
+   for (j = 0; j < QUAD_SIZE; j++) {
+      int c;
+
+      float u = s[j] * xpot - 0.5F;
+      float v = t[j] * ypot - 0.5F;
+
+      int uflr = util_ifloor(u);
+      int vflr = util_ifloor(v);
+
+      float xw = u - (float)uflr;
+      float yw = v - (float)vflr;
+
+      int x0 = uflr & (xpot - 1);
+      int y0 = vflr & (ypot - 1);
+
+      const uint8_t *tx[4];
+      
+
+      /* Can we fetch all four at once:
+       */
+      if (x0 < xmax && y0 < ymax)
+      {
+         get_texel_quad_2d(tgsi_sampler, 0, level, x0, y0, tx);
+      }
+      else 
+      {
+         unsigned x1 = (x0 + 1) & (xpot - 1);
+         unsigned y1 = (y0 + 1) & (ypot - 1);
+         get_texel_quad_2d_mt(tgsi_sampler, 0, level, 
+                              x0, y0, x1, y1, tx);
+      }
+
+
+      /* interpolate R, G, B, A */
+      for (c = 0; c < 4; c++) {
+         rgba[c][j] = lerp_2d(xw, yw, 
+                              ubyte_to_float(tx[0][c]), ubyte_to_float(tx[1][c]),
+                              ubyte_to_float(tx[2][c]), ubyte_to_float(tx[3][c]));
+      }
+   }
+}
+
+
+static void
+lp_get_samples_2d_nearest_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+                                     const float s[QUAD_SIZE],
+                                     const float t[QUAD_SIZE],
+                                     const float p[QUAD_SIZE],
+                                     float lodbias,
+                                     float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   unsigned  j;
+   unsigned level = samp->level;
+   unsigned xpot = 1 << (samp->xpot - level);
+   unsigned ypot = 1 << (samp->ypot - level);
+
+   for (j = 0; j < QUAD_SIZE; j++) {
+      int c;
+
+      float u = s[j] * xpot;
+      float v = t[j] * ypot;
+
+      int uflr = util_ifloor(u);
+      int vflr = util_ifloor(v);
+
+      int x0 = uflr & (xpot - 1);
+      int y0 = vflr & (ypot - 1);
+
+      const uint8_t *out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0);
+
+      for (c = 0; c < 4; c++) {
+         rgba[c][j] = ubyte_to_float(out[c]);
+      }
+   }
+}
+
+
+static void
+lp_get_samples_2d_nearest_clamp_POT(struct tgsi_sampler *tgsi_sampler,
+                                     const float s[QUAD_SIZE],
+                                     const float t[QUAD_SIZE],
+                                     const float p[QUAD_SIZE],
+                                     float lodbias,
+                                     float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   unsigned  j;
+   unsigned level = samp->level;
+   unsigned xpot = 1 << (samp->xpot - level);
+   unsigned ypot = 1 << (samp->ypot - level);
+
+   for (j = 0; j < QUAD_SIZE; j++) {
+      int c;
+
+      float u = s[j] * xpot;
+      float v = t[j] * ypot;
+
+      int x0, y0;
+      const uint8_t *out;
+
+      x0 = util_ifloor(u);
+      if (x0 < 0) 
+         x0 = 0;
+      else if (x0 > xpot - 1)
+         x0 = xpot - 1;
+
+      y0 = util_ifloor(v);
+      if (y0 < 0) 
+         y0 = 0;
+      else if (y0 > ypot - 1)
+         y0 = ypot - 1;
+      
+      out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0);
+
+      for (c = 0; c < 4; c++) {
+         rgba[c][j] = ubyte_to_float(out[c]);
+      }
+   }
+}
+
+
+static void
+lp_get_samples_2d_linear_mip_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+                                               const float s[QUAD_SIZE],
+                                               const float t[QUAD_SIZE],
+                                               const float p[QUAD_SIZE],
+                                               float lodbias,
+                                               float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   int level0;
+   float lambda;
+
+   lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+   level0 = (int)lambda;
+
+   if (lambda < 0.0) { 
+      samp->level = 0;
+      lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+                                           s, t, p, 0, rgba );
+   }
+   else if (level0 >= texture->last_level) {
+      samp->level = texture->last_level;
+      lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+                                           s, t, p, 0, rgba );
+   }
+   else {
+      float levelBlend = lambda - level0;
+      float rgba0[4][4];
+      float rgba1[4][4];
+      int c,j;
+
+      samp->level = level0;
+      lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+                                           s, t, p, 0, rgba0 );
+
+      samp->level = level0+1;
+      lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+                                           s, t, p, 0, rgba1 );
+
+      for (j = 0; j < QUAD_SIZE; j++) {
+         for (c = 0; c < 4; c++) {
+            rgba[c][j] = lerp(levelBlend, rgba0[c][j], rgba1[c][j]);
+         }
+      }
+   }
+}
+
+/**
+ * Common code for sampling 1D/2D/cube textures.
+ * Could probably extend for 3D...
+ */
+static void
+lp_get_samples_2d_common(struct tgsi_sampler *tgsi_sampler,
+                         const float s[QUAD_SIZE],
+                         const float t[QUAD_SIZE],
+                         const float p[QUAD_SIZE],
+                         float lodbias,
+                         float rgba[NUM_CHANNELS][QUAD_SIZE],
+                         const unsigned faces[4])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+   unsigned level0, level1, j, imgFilter;
+   int width, height;
+   float levelBlend;
+
+   choose_mipmap_levels(tgsi_sampler, s, t, p, 
+                        lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   assert(sampler->normalized_coords);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+
+   assert(width > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4];
+         nearest_texcoord_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_4(sampler->wrap_t, t, height, y);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(sampler, rgba, p, j);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               unsigned c;
+               x[j] /= 2;
+               y[j] /= 2;
+               get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0,
+                         rgba2, j);
+               if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+                  shadow_compare(sampler, rgba2, p, j);
+               }
+
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], y0[4], x1[4], y1[4];
+         float xw[4], yw[4]; /* weights */
+
+         linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw);
+         linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            float tx[4][4]; /* texels */
+            int c;
+            get_texel(tgsi_sampler, faces[j], level0, x0[j], y0[j], 0, tx, 0);
+            get_texel(tgsi_sampler, faces[j], level0, x1[j], y0[j], 0, tx, 1);
+            get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2);
+            get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare4(sampler, tx, p);
+            }
+
+            /* interpolate R, G, B, A */
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_2d(xw[j], yw[j],
+                                    tx[c][0], tx[c][1],
+                                    tx[c][2], tx[c][3]);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+
+               /* XXX: This is incorrect -- will often end up with (x0
+                *  == x1 && y0 == y1), meaning that we fetch the same
+                *  texel four times and linearly interpolate between
+                *  identical values.  The correct approach would be to
+                *  call linear_texcoord again for the second level.
+                */
+               x0[j] /= 2;
+               y0[j] /= 2;
+               x1[j] /= 2;
+               y1[j] /= 2;
+               get_texel(tgsi_sampler, faces[j], level1, x0[j], y0[j], 0, tx, 0);
+               get_texel(tgsi_sampler, faces[j], level1, x1[j], y0[j], 0, tx, 1);
+               get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2);
+               get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3);
+               if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+                  shadow_compare4(sampler, tx, p);
+               }
+
+               /* interpolate R, G, B, A */
+               for (c = 0; c < 4; c++) {
+                  rgba2[c][j] = lerp_2d(xw[j], yw[j],
+                                        tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+               }
+
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static INLINE void
+lp_get_samples_1d(struct tgsi_sampler *sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   static const unsigned faces[4] = {0, 0, 0, 0};
+   static const float tzero[4] = {0, 0, 0, 0};
+   lp_get_samples_2d_common(sampler, s, tzero, NULL,
+                            lodbias, rgba, faces);
+}
+
+
+static INLINE void
+lp_get_samples_2d(struct tgsi_sampler *sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   static const unsigned faces[4] = {0, 0, 0, 0};
+   lp_get_samples_2d_common(sampler, s, t, p,
+                            lodbias, rgba, faces);
+}
+
+
+static INLINE void
+lp_get_samples_3d(struct tgsi_sampler *tgsi_sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+   /* get/map pipe_surfaces corresponding to 3D tex slices */
+   unsigned level0, level1, j, imgFilter;
+   int width, height, depth;
+   float levelBlend;
+   const uint face = 0;
+
+   choose_mipmap_levels(tgsi_sampler, s, t, p, 
+                        lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   assert(sampler->normalized_coords);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+   depth = texture->depth[level0];
+
+   assert(width > 0);
+   assert(height > 0);
+   assert(depth > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4], z[4];
+         nearest_texcoord_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_4(sampler->wrap_t, t, height, y);
+         nearest_texcoord_4(sampler->wrap_r, p, depth, z);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, face, level0, x[j], y[j], z[j], rgba, j);
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               unsigned c;
+               x[j] /= 2;
+               y[j] /= 2;
+               z[j] /= 2;
+               get_texel(tgsi_sampler, face, level1, x[j], y[j], z[j], rgba2, j);
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba2[c][j], rgba[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], x1[4], y0[4], y1[4], z0[4], z1[4];
+         float xw[4], yw[4], zw[4]; /* interpolation weights */
+         linear_texcoord_4(sampler->wrap_s, s, width,  x0, x1, xw);
+         linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+         linear_texcoord_4(sampler->wrap_r, p, depth,  z0, z1, zw);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int c;
+            float tx0[4][4], tx1[4][4];
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z0[j], tx0, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z0[j], tx0, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z0[j], tx0, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z0[j], tx0, 3);
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z1[j], tx1, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z1[j], tx1, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z1[j], tx1, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z1[j], tx1, 3);
+
+            /* interpolate R, G, B, A */
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+                                    tx0[c][0], tx0[c][1],
+                                    tx0[c][2], tx0[c][3],
+                                    tx1[c][0], tx1[c][1],
+                                    tx1[c][2], tx1[c][3]);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               x0[j] /= 2;
+               y0[j] /= 2;
+               z0[j] /= 2;
+               x1[j] /= 2;
+               y1[j] /= 2;
+               z1[j] /= 2;
+               get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z0[j], tx0, 0);
+               get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z0[j], tx0, 1);
+               get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z0[j], tx0, 2);
+               get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z0[j], tx0, 3);
+               get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z1[j], tx1, 0);
+               get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z1[j], tx1, 1);
+               get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z1[j], tx1, 2);
+               get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z1[j], tx1, 3);
+
+               /* interpolate R, G, B, A */
+               for (c = 0; c < 4; c++) {
+                  rgba2[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+                                        tx0[c][0], tx0[c][1],
+                                        tx0[c][2], tx0[c][3],
+                                        tx1[c][0], tx1[c][1],
+                                        tx1[c][2], tx1[c][3]);
+               }
+
+               /* blend mipmap levels */
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static void
+lp_get_samples_cube(struct tgsi_sampler *sampler,
+                    const float s[QUAD_SIZE],
+                    const float t[QUAD_SIZE],
+                    const float p[QUAD_SIZE],
+                    float lodbias,
+                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   unsigned faces[QUAD_SIZE], j;
+   float ssss[4], tttt[4];
+   for (j = 0; j < QUAD_SIZE; j++) {
+      faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j);
+   }
+   lp_get_samples_2d_common(sampler, ssss, tttt, NULL,
+                            lodbias, rgba, faces);
+}
+
+
+static void
+lp_get_samples_rect(struct tgsi_sampler *tgsi_sampler,
+                    const float s[QUAD_SIZE],
+                    const float t[QUAD_SIZE],
+                    const float p[QUAD_SIZE],
+                    float lodbias,
+                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+   const uint face = 0;
+   unsigned level0, level1, j, imgFilter;
+   int width, height;
+   float levelBlend;
+
+   choose_mipmap_levels(tgsi_sampler, s, t, p, 
+                        lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   /* texture RECTS cannot be mipmapped */
+   assert(level0 == level1);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+
+   assert(width > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4];
+         nearest_texcoord_unnorm_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_unnorm_4(sampler->wrap_t, t, height, y);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(sampler, rgba, p, j);
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], y0[4], x1[4], y1[4];
+         float xw[4], yw[4]; /* weights */
+         linear_texcoord_unnorm_4(sampler->wrap_s, s, width,  x0, x1, xw);
+         linear_texcoord_unnorm_4(sampler->wrap_t, t, height, y0, y1, yw);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            float tx[4][4]; /* texels */
+            int c;
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], 0, tx, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], 0, tx, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare4(sampler, tx, p);
+            }
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_2d(xw[j], yw[j],
+                                    tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Error condition handler
+ */
+static INLINE void
+lp_get_samples_null(struct tgsi_sampler *tgsi_sampler,
+                    const float s[QUAD_SIZE],
+                    const float t[QUAD_SIZE],
+                    const float p[QUAD_SIZE],
+                    float lodbias,
+                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   int i,j;
+
+   for (i = 0; i < 4; i++)
+      for (j = 0; j < 4; j++)
+         rgba[i][j] = 1.0;
+}
+
+/**
+ * Called via tgsi_sampler::get_samples() when using a sampler for the
+ * first time.  Determine the actual sampler function, link it in and
+ * call it.
+ */
+void
+lp_get_samples(struct tgsi_sampler *tgsi_sampler,
+               const float s[QUAD_SIZE],
+               const float t[QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               float lodbias,
+               float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+   const struct pipe_texture *texture = samp->texture;
+   const struct pipe_sampler_state *sampler = samp->sampler;
+
+   /* Default to the 'undefined' case:
+    */
+   tgsi_sampler->get_samples = lp_get_samples_null;
+
+   if (!texture) {
+      assert(0);                /* is this legal?? */
+      goto out;
+   }
+
+   if (!sampler->normalized_coords) {
+      assert (texture->target == PIPE_TEXTURE_2D);
+      tgsi_sampler->get_samples = lp_get_samples_rect;
+      goto out;
+   }
+
+   switch (texture->target) {
+   case PIPE_TEXTURE_1D:
+      tgsi_sampler->get_samples = lp_get_samples_1d;
+      break;
+   case PIPE_TEXTURE_2D:
+      tgsi_sampler->get_samples = lp_get_samples_2d;
+      break;
+   case PIPE_TEXTURE_3D:
+      tgsi_sampler->get_samples = lp_get_samples_3d;
+      break;
+   case PIPE_TEXTURE_CUBE:
+      tgsi_sampler->get_samples = lp_get_samples_cube;
+      break;
+   default:
+      assert(0);
+      break;
+   }
+
+   /* Do this elsewhere: 
+    */
+   samp->xpot = util_unsigned_logbase2( samp->texture->width[0] );
+   samp->ypot = util_unsigned_logbase2( samp->texture->height[0] );
+
+   /* Try to hook in a faster sampler.  Ultimately we'll have to
+    * code-generate these.  Luckily most of this looks like it is
+    * orthogonal state within the sampler.
+    */
+   if (texture->target == PIPE_TEXTURE_2D &&
+       sampler->min_img_filter == sampler->mag_img_filter &&
+       sampler->wrap_s == sampler->wrap_t &&
+       sampler->compare_mode == FALSE &&
+       sampler->normalized_coords) 
+   {
+      if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
+         samp->level = CLAMP((int) sampler->min_lod,
+                             0, (int) texture->last_level);
+
+         if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) {
+            switch (sampler->min_img_filter) {
+            case PIPE_TEX_FILTER_NEAREST:
+               tgsi_sampler->get_samples = lp_get_samples_2d_nearest_repeat_POT;
+               break;
+            case PIPE_TEX_FILTER_LINEAR:
+               tgsi_sampler->get_samples = lp_get_samples_2d_linear_repeat_POT;
+               break;
+            default:
+               break;
+            }
+         } 
+         else if (sampler->wrap_s == PIPE_TEX_WRAP_CLAMP) {
+            switch (sampler->min_img_filter) {
+            case PIPE_TEX_FILTER_NEAREST:
+               tgsi_sampler->get_samples = lp_get_samples_2d_nearest_clamp_POT;
+               break;
+            default:
+               break;
+            }
+         }
+      }
+      else if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_LINEAR) {
+         if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) {
+            switch (sampler->min_img_filter) {
+            case PIPE_TEX_FILTER_LINEAR:
+               tgsi_sampler->get_samples = lp_get_samples_2d_linear_mip_linear_repeat_POT;
+               break;
+            default:
+               break;
+            }
+         } 
+      }
+   }
+   else if (0) {
+      _debug_printf("target %d/%d min_mip %d/%d min_img %d/%d wrap %d/%d compare %d/%d norm %d/%d\n",
+                    texture->target, PIPE_TEXTURE_2D,
+                    sampler->min_mip_filter, PIPE_TEX_MIPFILTER_NONE,
+                    sampler->min_img_filter, sampler->mag_img_filter,
+                    sampler->wrap_s, sampler->wrap_t,
+                    sampler->compare_mode, FALSE,
+                    sampler->normalized_coords, TRUE);
+   }
+
+out:
+   tgsi_sampler->get_samples( tgsi_sampler, s, t, p, lodbias, rgba );
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample.h b/src/gallium/drivers/llvmpipe/lp_tex_sample.h
new file mode 100644
index 0000000000..628ec3f1ef
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample.h
@@ -0,0 +1,78 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_TEX_SAMPLE_H
+#define LP_TEX_SAMPLE_H
+
+
+#include "tgsi/tgsi_exec.h"
+
+
+struct llvmpipe_tex_tile_cache;
+
+
+/**
+ * Subclass of tgsi_sampler
+ */
+struct lp_shader_sampler
+{
+   struct tgsi_sampler base;  /**< base class */
+
+   unsigned processor;
+
+   /* For lp_get_samples_2d_linear_POT:
+    */
+   unsigned xpot;
+   unsigned ypot;
+   unsigned level;
+
+   const struct pipe_texture *texture;
+   const struct pipe_sampler_state *sampler;
+
+   struct llvmpipe_tex_tile_cache *cache;
+};
+
+
+
+static INLINE struct lp_shader_sampler *
+lp_shader_sampler(const struct tgsi_sampler *sampler)
+{
+   return (struct lp_shader_sampler *) sampler;
+}
+
+
+
+extern void
+lp_get_samples(struct tgsi_sampler *tgsi_sampler,
+               const float s[QUAD_SIZE],
+               const float t[QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               float lodbias,
+               float rgba[NUM_CHANNELS][QUAD_SIZE]);
+
+
+#endif /* LP_TEX_SAMPLE_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
new file mode 100644
index 0000000000..724d437833
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -0,0 +1,429 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "lp_context.h"
+#include "lp_state.h"
+#include "lp_texture.h"
+#include "lp_tex_cache.h"
+#include "lp_screen.h"
+#include "lp_winsys.h"
+
+
+/* Simple, maximally packed layout.
+ */
+
+
+/* Conventional allocation path for non-display textures:
+ */
+static boolean
+llvmpipe_texture_layout(struct llvmpipe_screen *screen,
+                        struct llvmpipe_texture * lpt)
+{
+   struct pipe_texture *pt = &lpt->base;
+   unsigned level;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned depth = pt->depth[0];
+
+   unsigned buffer_size = 0;
+
+   pf_get_block(lpt->base.format, &lpt->base.block);
+
+   for (level = 0; level <= pt->last_level; level++) {
+      pt->width[level] = width;
+      pt->height[level] = height;
+      pt->depth[level] = depth;
+      pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width);  
+      pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height);  
+      lpt->stride[level] = align(pt->nblocksx[level]*pt->block.size, 16);
+
+      lpt->level_offset[level] = buffer_size;
+
+      buffer_size += (pt->nblocksy[level] *
+                      ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
+                      lpt->stride[level]);
+
+      width  = minify(width);
+      height = minify(height);
+      depth = minify(depth);
+   }
+
+   lpt->data = align_malloc(buffer_size, 16);
+
+   return lpt->data != NULL;
+}
+
+static boolean
+llvmpipe_displaytarget_layout(struct llvmpipe_screen *screen,
+                              struct llvmpipe_texture * lpt)
+{
+   struct llvmpipe_winsys *winsys = screen->winsys;
+
+   pf_get_block(lpt->base.format, &lpt->base.block);
+   lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width[0]);  
+   lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height[0]);  
+
+   lpt->dt = winsys->displaytarget_create(winsys,
+                                          lpt->base.format,
+                                          lpt->base.width[0],
+                                          lpt->base.height[0],
+                                          16,
+                                          &lpt->stride[0] );
+
+   return lpt->dt != NULL;
+}
+
+
+
+
+
+static struct pipe_texture *
+llvmpipe_texture_create(struct pipe_screen *_screen,
+                        const struct pipe_texture *templat)
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+   struct llvmpipe_texture *lpt = CALLOC_STRUCT(llvmpipe_texture);
+   if (!lpt)
+      return NULL;
+
+   lpt->base = *templat;
+   pipe_reference_init(&lpt->base.reference, 1);
+   lpt->base.screen = &screen->base;
+
+   /* XXX: The xlib state tracker is brain-dead and will request
+    * PIPE_FORMAT_Z16_UNORM no matter how much we tell it we don't support it.
+    */
+   if(lpt->base.format == PIPE_FORMAT_Z16_UNORM)
+      lpt->base.format = PIPE_FORMAT_Z32_UNORM;
+
+   if (lpt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                              PIPE_TEXTURE_USAGE_PRIMARY)) {
+      if (!llvmpipe_displaytarget_layout(screen, lpt))
+         goto fail;
+   }
+   else {
+      if (!llvmpipe_texture_layout(screen, lpt))
+         goto fail;
+   }
+    
+   return &lpt->base;
+
+ fail:
+   FREE(lpt);
+   return NULL;
+}
+
+
+static struct pipe_texture *
+llvmpipe_texture_blanket(struct pipe_screen * screen,
+                         const struct pipe_texture *base,
+                         const unsigned *stride,
+                         struct pipe_buffer *buffer)
+{
+   /* FIXME */
+#if 0
+   struct llvmpipe_texture *lpt;
+   assert(screen);
+
+   /* Only supports one type */
+   if (base->target != PIPE_TEXTURE_2D ||
+       base->last_level != 0 ||
+       base->depth[0] != 1) {
+      return NULL;
+   }
+
+   lpt = CALLOC_STRUCT(llvmpipe_texture);
+   if (!lpt)
+      return NULL;
+
+   lpt->base = *base;
+   pipe_reference_init(&lpt->base.reference, 1);
+   lpt->base.screen = screen;
+   lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width[0]);  
+   lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height[0]);  
+   lpt->stride[0] = stride[0];
+
+   pipe_buffer_reference(&lpt->buffer, buffer);
+
+   return &lpt->base;
+#else
+   return NULL;
+#endif
+}
+
+
+static void
+llvmpipe_texture_destroy(struct pipe_texture *pt)
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(pt->screen);
+   struct llvmpipe_texture *lpt = llvmpipe_texture(pt);
+
+   if(lpt->dt) {
+      struct llvmpipe_winsys *winsys = screen->winsys;
+      winsys->displaytarget_destroy(winsys, lpt->dt);
+   }
+   else
+      align_free(lpt->data);
+
+   FREE(lpt);
+}
+
+
+static struct pipe_surface *
+llvmpipe_get_tex_surface(struct pipe_screen *screen,
+                         struct pipe_texture *pt,
+                         unsigned face, unsigned level, unsigned zslice,
+                         unsigned usage)
+{
+   struct llvmpipe_texture *lpt = llvmpipe_texture(pt);
+   struct pipe_surface *ps;
+
+   assert(level <= pt->last_level);
+
+   ps = CALLOC_STRUCT(pipe_surface);
+   if (ps) {
+      pipe_reference_init(&ps->reference, 1);
+      pipe_texture_reference(&ps->texture, pt);
+      ps->format = pt->format;
+      ps->width = pt->width[level];
+      ps->height = pt->height[level];
+      ps->offset = lpt->level_offset[level];
+      ps->usage = usage;
+
+      /* Because we are llvmpipe, anything that the state tracker
+       * thought was going to be done with the GPU will actually get
+       * done with the CPU.  Let's adjust the flags to take that into
+       * account.
+       */
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE) {
+         /* GPU_WRITE means "render" and that can involve reads (blending) */
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_CPU_READ;
+      }
+
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ)
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_READ;
+
+      if (ps->usage & (PIPE_BUFFER_USAGE_CPU_WRITE |
+                       PIPE_BUFFER_USAGE_GPU_WRITE)) {
+         /* Mark the surface as dirty.  The tile cache will look for this. */
+         lpt->timestamp++;
+         llvmpipe_screen(screen)->timestamp++;
+      }
+
+      ps->face = face;
+      ps->level = level;
+      ps->zslice = zslice;
+
+      if (pt->target == PIPE_TEXTURE_CUBE) {
+         ps->offset += face * pt->nblocksy[level] * lpt->stride[level];
+      }
+      else if (pt->target == PIPE_TEXTURE_3D) {
+         ps->offset += zslice * pt->nblocksy[level] * lpt->stride[level];
+      }
+      else {
+         assert(face == 0);
+         assert(zslice == 0);
+      }
+   }
+   return ps;
+}
+
+
+static void 
+llvmpipe_tex_surface_destroy(struct pipe_surface *surf)
+{
+   /* Effectively do the texture_update work here - if texture images
+    * needed post-processing to put them into hardware layout, this is
+    * where it would happen.  For llvmpipe, nothing to do.
+    */
+   assert(surf->texture);
+   pipe_texture_reference(&surf->texture, NULL);
+   FREE(surf);
+}
+
+
+static struct pipe_transfer *
+llvmpipe_get_tex_transfer(struct pipe_screen *screen,
+                          struct pipe_texture *texture,
+                          unsigned face, unsigned level, unsigned zslice,
+                          enum pipe_transfer_usage usage,
+                          unsigned x, unsigned y, unsigned w, unsigned h)
+{
+   struct llvmpipe_texture *lptex = llvmpipe_texture(texture);
+   struct llvmpipe_transfer *lpt;
+
+   assert(texture);
+   assert(level <= texture->last_level);
+
+   lpt = CALLOC_STRUCT(llvmpipe_transfer);
+   if (lpt) {
+      struct pipe_transfer *pt = &lpt->base;
+      pipe_texture_reference(&pt->texture, texture);
+      pt->format = texture->format;
+      pt->block = texture->block;
+      pt->x = x;
+      pt->y = y;
+      pt->width = w;
+      pt->height = h;
+      pt->nblocksx = texture->nblocksx[level];
+      pt->nblocksy = texture->nblocksy[level];
+      pt->stride = lptex->stride[level];
+      pt->usage = usage;
+      pt->face = face;
+      pt->level = level;
+      pt->zslice = zslice;
+
+      lpt->offset = lptex->level_offset[level];
+
+      if (texture->target == PIPE_TEXTURE_CUBE) {
+         lpt->offset += face * pt->nblocksy * pt->stride;
+      }
+      else if (texture->target == PIPE_TEXTURE_3D) {
+         lpt->offset += zslice * pt->nblocksy * pt->stride;
+      }
+      else {
+         assert(face == 0);
+         assert(zslice == 0);
+      }
+      return pt;
+   }
+   return NULL;
+}
+
+
+static void 
+llvmpipe_tex_transfer_destroy(struct pipe_transfer *transfer)
+{
+   /* Effectively do the texture_update work here - if texture images
+    * needed post-processing to put them into hardware layout, this is
+    * where it would happen.  For llvmpipe, nothing to do.
+    */
+   assert (transfer->texture);
+   pipe_texture_reference(&transfer->texture, NULL);
+   FREE(transfer);
+}
+
+
+static void *
+llvmpipe_transfer_map( struct pipe_screen *_screen,
+                       struct pipe_transfer *transfer )
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+   ubyte *map, *xfer_map;
+   struct llvmpipe_texture *lpt;
+
+   assert(transfer->texture);
+   lpt = llvmpipe_texture(transfer->texture);
+
+   if(lpt->dt) {
+      struct llvmpipe_winsys *winsys = screen->winsys;
+      unsigned flags = 0;
+
+      if (transfer->usage != PIPE_TRANSFER_READ) {
+         flags |= PIPE_BUFFER_USAGE_CPU_WRITE;
+      }
+
+      if (transfer->usage != PIPE_TRANSFER_WRITE) {
+         flags |= PIPE_BUFFER_USAGE_CPU_READ;
+      }
+
+      map = winsys->displaytarget_map(winsys, lpt->dt, flags);
+      if (map == NULL)
+         return NULL;
+   }
+   else
+      map = lpt->data;
+
+   /* May want to different things here depending on read/write nature
+    * of the map:
+    */
+   if (transfer->texture && transfer->usage != PIPE_TRANSFER_READ) 
+   {
+      /* Do something to notify sharing contexts of a texture change.
+       * In llvmpipe, that would mean flushing the texture cache.
+       */
+      screen->timestamp++;
+   }
+   
+   xfer_map = map + llvmpipe_transfer(transfer)->offset +
+      transfer->y / transfer->block.height * transfer->stride +
+      transfer->x / transfer->block.width * transfer->block.size;
+   /*printf("map = %p  xfer map = %p\n", map, xfer_map);*/
+   return xfer_map;
+}
+
+
+static void
+llvmpipe_transfer_unmap(struct pipe_screen *_screen,
+                       struct pipe_transfer *transfer)
+{
+   struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
+   struct llvmpipe_texture *lpt;
+
+   assert(transfer->texture);
+   lpt = llvmpipe_texture(transfer->texture);
+
+   if(lpt->dt) {
+      struct llvmpipe_winsys *winsys = screen->winsys;
+      winsys->displaytarget_unmap(winsys, lpt->dt);
+   }
+}
+
+
+void
+llvmpipe_init_texture_funcs(struct llvmpipe_context *lp)
+{
+}
+
+
+void
+llvmpipe_init_screen_texture_funcs(struct pipe_screen *screen)
+{
+   screen->texture_create = llvmpipe_texture_create;
+   screen->texture_blanket = llvmpipe_texture_blanket;
+   screen->texture_destroy = llvmpipe_texture_destroy;
+
+   screen->get_tex_surface = llvmpipe_get_tex_surface;
+   screen->tex_surface_destroy = llvmpipe_tex_surface_destroy;
+
+   screen->get_tex_transfer = llvmpipe_get_tex_transfer;
+   screen->tex_transfer_destroy = llvmpipe_tex_transfer_destroy;
+   screen->transfer_map = llvmpipe_transfer_map;
+   screen->transfer_unmap = llvmpipe_transfer_unmap;
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.h b/src/gallium/drivers/llvmpipe/lp_texture.h
new file mode 100644
index 0000000000..00a20763e4
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_texture.h
@@ -0,0 +1,90 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_TEXTURE_H
+#define LP_TEXTURE_H
+
+
+#include "pipe/p_state.h"
+
+
+struct pipe_context;
+struct pipe_screen;
+struct llvmpipe_context;
+struct llvmpipe_displaytarget;
+
+struct llvmpipe_texture
+{
+   struct pipe_texture base;
+
+   unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned stride[PIPE_MAX_TEXTURE_LEVELS];
+
+   /**
+    * Display target, for textures with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET
+    * usage.
+    */
+   struct llvmpipe_displaytarget *dt;
+
+   /**
+    * Malloc'ed data for regular textures, or a mapping to dt above.
+    */
+   void *data;
+
+   unsigned timestamp;
+};
+
+struct llvmpipe_transfer
+{
+   struct pipe_transfer base;
+
+   unsigned long offset;
+};
+
+
+/** cast wrappers */
+static INLINE struct llvmpipe_texture *
+llvmpipe_texture(struct pipe_texture *pt)
+{
+   return (struct llvmpipe_texture *) pt;
+}
+
+static INLINE struct llvmpipe_transfer *
+llvmpipe_transfer(struct pipe_transfer *pt)
+{
+   return (struct llvmpipe_transfer *) pt;
+}
+
+
+extern void
+llvmpipe_init_texture_funcs( struct llvmpipe_context *llvmpipe );
+
+extern void
+llvmpipe_init_screen_texture_funcs(struct pipe_screen *screen);
+
+
+#endif /* LP_TEXTURE */
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
new file mode 100644
index 0000000000..143afec3d3
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
@@ -0,0 +1,299 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Texture tile caching.
+ *
+ * Author:
+ *    Brian Paul
+ */
+
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_tile.h"
+#include "util/u_rect.h"
+#include "lp_context.h"
+#include "lp_surface.h"
+#include "lp_texture.h"
+#include "lp_tile_soa.h"
+#include "lp_tile_cache.h"
+
+
+struct llvmpipe_tile_cache *
+lp_create_tile_cache( struct pipe_screen *screen )
+{
+   struct llvmpipe_tile_cache *tc;
+
+   tc = CALLOC_STRUCT( llvmpipe_tile_cache );
+   if(!tc)
+      return NULL;
+
+   tc->screen = screen;
+
+   return tc;
+}
+
+
+void
+lp_destroy_tile_cache(struct llvmpipe_tile_cache *tc)
+{
+   struct pipe_screen *screen;
+   unsigned x, y;
+
+   for (y = 0; y < MAX_HEIGHT; y += TILE_SIZE) {
+      for (x = 0; x < MAX_WIDTH; x += TILE_SIZE) {
+         struct llvmpipe_cached_tile *tile = &tc->entries[y/TILE_SIZE][x/TILE_SIZE];
+
+         if(tile->color)
+            align_free(tile->color);
+      }
+   }
+
+   if (tc->transfer) {
+      screen = tc->transfer->texture->screen;
+      screen->tex_transfer_destroy(tc->transfer);
+   }
+
+   FREE( tc );
+}
+
+
+/**
+ * Specify the surface to cache.
+ */
+void
+lp_tile_cache_set_surface(struct llvmpipe_tile_cache *tc,
+                          struct pipe_surface *ps)
+{
+   if (tc->transfer) {
+      struct pipe_screen *screen = tc->transfer->texture->screen;
+
+      if (ps == tc->surface)
+         return;
+
+      if (tc->transfer_map) {
+         screen->transfer_unmap(screen, tc->transfer);
+         tc->transfer_map = NULL;
+      }
+
+      screen->tex_transfer_destroy(tc->transfer);
+      tc->transfer = NULL;
+   }
+
+   tc->surface = ps;
+
+   if (ps) {
+      struct pipe_screen *screen = ps->texture->screen;
+      unsigned x, y;
+
+      tc->transfer = screen->get_tex_transfer(screen, ps->texture, ps->face,
+                                              ps->level, ps->zslice,
+                                              PIPE_TRANSFER_READ_WRITE,
+                                              0, 0, ps->width, ps->height);
+
+      for (y = 0; y < ps->height; y += TILE_SIZE) {
+         for (x = 0; x < ps->width; x += TILE_SIZE) {
+            struct llvmpipe_cached_tile *tile = &tc->entries[y/TILE_SIZE][x/TILE_SIZE];
+
+            tile->status = LP_TILE_STATUS_UNDEFINED;
+
+            if(!tile->color)
+               tile->color = align_malloc( TILE_SIZE*TILE_SIZE*NUM_CHANNELS, 16 );
+         }
+      }
+   }
+}
+
+
+/**
+ * Return the transfer being cached.
+ */
+struct pipe_surface *
+lp_tile_cache_get_surface(struct llvmpipe_tile_cache *tc)
+{
+   return tc->surface;
+}
+
+
+void
+lp_tile_cache_map_transfers(struct llvmpipe_tile_cache *tc)
+{
+   if (tc->transfer && !tc->transfer_map)
+      tc->transfer_map = tc->screen->transfer_map(tc->screen, tc->transfer);
+}
+
+
+void
+lp_tile_cache_unmap_transfers(struct llvmpipe_tile_cache *tc)
+{
+   if (tc->transfer_map) {
+      tc->screen->transfer_unmap(tc->screen, tc->transfer);
+      tc->transfer_map = NULL;
+   }
+}
+
+
+/**
+ * Set a tile to a solid color.
+ */
+static void
+clear_tile(struct llvmpipe_cached_tile *tile,
+           uint8_t clear_color[4])
+{
+   if (clear_color[0] == clear_color[1] &&
+       clear_color[1] == clear_color[2] &&
+       clear_color[2] == clear_color[3]) {
+      memset(tile->color, clear_color[0], TILE_SIZE * TILE_SIZE * 4);
+   }
+   else {
+      uint x, y, chan;
+      for (y = 0; y < TILE_SIZE; y++)
+         for (x = 0; x < TILE_SIZE; x++)
+            for (chan = 0; chan < 4; ++chan)
+               TILE_PIXEL(tile->color, x, y, chan) = clear_color[chan];
+   }
+}
+
+
+/**
+ * Flush the tile cache: write all dirty tiles back to the transfer.
+ * any tiles "flagged" as cleared will be "really" cleared.
+ */
+void
+lp_flush_tile_cache(struct llvmpipe_tile_cache *tc)
+{
+   struct pipe_transfer *pt = tc->transfer;
+   unsigned x, y;
+
+   if(!pt)
+      return;
+
+   /* push the tile to all positions marked as clear */
+   for (y = 0; y < pt->height; y += TILE_SIZE) {
+      for (x = 0; x < pt->width; x += TILE_SIZE) {
+         struct llvmpipe_cached_tile *tile = &tc->entries[y/TILE_SIZE][x/TILE_SIZE];
+
+         switch(tile->status) {
+         case LP_TILE_STATUS_UNDEFINED:
+            break;
+
+         case LP_TILE_STATUS_CLEAR: {
+            /**
+             * Actually clear the tiles which were flagged as being in a clear state.
+             */
+
+            struct pipe_screen *screen = pt->texture->screen;
+            unsigned tw = TILE_SIZE;
+            unsigned th = TILE_SIZE;
+            void *dst;
+
+            if (pipe_clip_tile(x, y, &tw, &th, pt))
+               continue;
+
+            dst = screen->transfer_map(screen, pt);
+            assert(dst);
+            if(!dst)
+               continue;
+
+            util_fill_rect(dst, &pt->block, pt->stride,
+                           x, y, tw,  th,
+                           tc->clear_val);
+
+            screen->transfer_unmap(screen, pt);
+            break;
+         }
+
+         case LP_TILE_STATUS_DEFINED:
+            lp_put_tile_rgba_soa(pt, x, y, tile->color);
+            break;
+         }
+      }
+   }
+}
+
+
+/**
+ * Get a tile from the cache.
+ * \param x, y  position of tile, in pixels
+ */
+void *
+lp_get_cached_tile(struct llvmpipe_tile_cache *tc,
+                   unsigned x, unsigned y )
+{
+   struct llvmpipe_cached_tile *tile = &tc->entries[y/TILE_SIZE][x/TILE_SIZE];
+   struct pipe_transfer *pt = tc->transfer;
+   
+   switch(tile->status) {
+   case LP_TILE_STATUS_CLEAR:
+      /* don't get tile from framebuffer, just clear it */
+      clear_tile(tile, tc->clear_color);
+      tile->status = LP_TILE_STATUS_DEFINED;
+      break;
+
+   case LP_TILE_STATUS_UNDEFINED:
+      /* get new tile data from transfer */
+      lp_get_tile_rgba_soa(pt, x, y, tile->color);
+      tile->status = LP_TILE_STATUS_DEFINED;
+      break;
+
+   case LP_TILE_STATUS_DEFINED:
+      /* nothing to do */
+      break;
+   }
+
+   return tile->color;
+}
+
+
+/**
+ * When a whole surface is being cleared to a value we can avoid
+ * fetching tiles above.
+ * Save the color and set a 'clearflag' for each tile of the screen.
+ */
+void
+lp_tile_cache_clear(struct llvmpipe_tile_cache *tc, const float *rgba,
+                    uint clearValue)
+{
+   struct pipe_transfer *pt = tc->transfer;
+   const unsigned w = pt->width;
+   const unsigned h = pt->height;
+   unsigned x, y, chan;
+
+   for(chan = 0; chan < 4; ++chan)
+      tc->clear_color[chan] = float_to_ubyte(rgba[chan]);
+
+   tc->clear_val = clearValue;
+
+   /* push the tile to all positions marked as clear */
+   for (y = 0; y < h; y += TILE_SIZE) {
+      for (x = 0; x < w; x += TILE_SIZE) {
+         struct llvmpipe_cached_tile *tile = &tc->entries[y/TILE_SIZE][x/TILE_SIZE];
+         tile->status = LP_TILE_STATUS_CLEAR;
+      }
+   }
+}
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.h b/src/gallium/drivers/llvmpipe/lp_tile_cache.h
new file mode 100644
index 0000000000..6d8ba5ece7
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.h
@@ -0,0 +1,106 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_TILE_CACHE_H
+#define LP_TILE_CACHE_H
+
+
+#include "pipe/p_compiler.h"
+#include "lp_tile_soa.h"
+
+
+enum llvmpipe_tile_status
+{
+   LP_TILE_STATUS_UNDEFINED = 0,
+   LP_TILE_STATUS_CLEAR = 1,
+   LP_TILE_STATUS_DEFINED = 2
+};
+
+
+struct llvmpipe_cached_tile
+{
+   enum llvmpipe_tile_status status;
+
+   /** color in SOA format */
+   uint8_t *color;
+};
+
+
+/** XXX move these */
+#define MAX_WIDTH 2048
+#define MAX_HEIGHT 2048
+
+
+struct llvmpipe_tile_cache
+{
+   struct pipe_screen *screen;
+   struct pipe_surface *surface;  /**< the surface we're caching */
+   struct pipe_transfer *transfer;
+   void *transfer_map;
+
+   struct llvmpipe_cached_tile entries[MAX_WIDTH/TILE_SIZE][MAX_HEIGHT/TILE_SIZE];
+
+   uint8_t clear_color[4];  /**< for color bufs */
+   uint clear_val;        /**< for z+stencil, or packed color clear value */
+
+   struct llvmpipe_cached_tile *last_tile;  /**< most recently retrieved tile */
+};
+
+
+extern struct llvmpipe_tile_cache *
+lp_create_tile_cache( struct pipe_screen *screen );
+
+extern void
+lp_destroy_tile_cache(struct llvmpipe_tile_cache *tc);
+
+extern void
+lp_tile_cache_set_surface(struct llvmpipe_tile_cache *tc,
+                          struct pipe_surface *lps);
+
+extern struct pipe_surface *
+lp_tile_cache_get_surface(struct llvmpipe_tile_cache *tc);
+
+extern void
+lp_tile_cache_map_transfers(struct llvmpipe_tile_cache *tc);
+
+extern void
+lp_tile_cache_unmap_transfers(struct llvmpipe_tile_cache *tc);
+
+extern void
+lp_flush_tile_cache(struct llvmpipe_tile_cache *tc);
+
+extern void
+lp_tile_cache_clear(struct llvmpipe_tile_cache *tc, const float *rgba,
+                    uint clearValue);
+
+extern void *
+lp_get_cached_tile(struct llvmpipe_tile_cache *tc,
+                   unsigned x, unsigned y );
+
+
+#endif /* LP_TILE_CACHE_H */
+
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.c b/src/gallium/drivers/llvmpipe/lp_tile_soa.c
new file mode 100644
index 0000000000..4e4ccb31cc
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.c
@@ -0,0 +1,931 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * RGBA/float tile get/put functions.
+ * Usable both by drivers and state trackers.
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_rect.h"
+#include "util/u_tile.h"
+#include "lp_tile_cache.h"
+#include "lp_tile_soa.h"
+
+
+const unsigned char
+tile_offset[TILE_VECTOR_HEIGHT][TILE_VECTOR_WIDTH] = {
+   {  0,  1,  4,  5,  8,  9, 12, 13},
+   {  2,  3,  6,  7, 10, 11, 14, 15}
+};
+
+
+
+/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/
+
+static void
+a8r8g8b8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const unsigned pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
+         TILE_PIXEL(p, j, i, 1) = (pixel >>  8) & 0xff;
+         TILE_PIXEL(p, j, i, 2) = (pixel >>  0) & 0xff;
+         TILE_PIXEL(p, j, i, 3) = (pixel >> 24) & 0xff;
+      }
+   }
+}
+
+
+static void
+a8r8g8b8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, g, b, a;
+         r = TILE_PIXEL(p, j, i, 0);
+         g = TILE_PIXEL(p, j, i, 1);
+         b = TILE_PIXEL(p, j, i, 2);
+         a = TILE_PIXEL(p, j, i, 3);
+         *dst++ = (a << 24) | (r << 16) | (g << 8) | b;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/
+
+static void
+x8r8g8b8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const unsigned pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = (pixel >> 16) & 0xff;
+         TILE_PIXEL(p, j, i, 1) = (pixel >>  8) & 0xff;
+         TILE_PIXEL(p, j, i, 2) = (pixel >>  0) & 0xff;
+         TILE_PIXEL(p, j, i, 3) = 0xff;
+      }
+   }
+}
+
+
+static void
+x8r8g8b8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, g, b;
+         r = TILE_PIXEL(p, j, i, 0);
+         g = TILE_PIXEL(p, j, i, 1);
+         b = TILE_PIXEL(p, j, i, 2);
+         *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/
+
+static void
+b8g8r8a8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const unsigned pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = (pixel >>  8) & 0xff;
+         TILE_PIXEL(p, j, i, 1) = (pixel >> 16) & 0xff;
+         TILE_PIXEL(p, j, i, 2) = (pixel >> 24) & 0xff;
+         TILE_PIXEL(p, j, i, 3) = (pixel >>  0) & 0xff;
+      }
+   }
+}
+
+
+static void
+b8g8r8a8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, g, b, a;
+         r = TILE_PIXEL(p, j, i, 0);
+         g = TILE_PIXEL(p, j, i, 1);
+         b = TILE_PIXEL(p, j, i, 2);
+         a = TILE_PIXEL(p, j, i, 3);
+         *dst++ = (b << 24) | (g << 16) | (r << 8) | a;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/
+
+static void
+a1r5g5b5_get_tile_rgba(const ushort *src,
+                       unsigned w, unsigned h,
+                       uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const ushort pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = ((pixel >> 10) & 0x1f) * 255 / 31;
+         TILE_PIXEL(p, j, i, 1) = ((pixel >>  5) & 0x1f) * 255 / 31;
+         TILE_PIXEL(p, j, i, 2) = ((pixel      ) & 0x1f) * 255 / 31;
+         TILE_PIXEL(p, j, i, 3) = ((pixel >> 15)       ) * 255;
+      }
+   }
+}
+
+
+static void
+a1r5g5b5_put_tile_rgba(ushort *dst,
+                       unsigned w, unsigned h,
+                       const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, g, b, a;
+         r = TILE_PIXEL(p, j, i, 0);
+         g = TILE_PIXEL(p, j, i, 1);
+         b = TILE_PIXEL(p, j, i, 2);
+         a = TILE_PIXEL(p, j, i, 3);
+         r = r >> 3;  /* 5 bits */
+         g = g >> 3;  /* 5 bits */
+         b = b >> 3;  /* 5 bits */
+         a = a >> 7;  /* 1 bit */
+         *dst++ = (a << 15) | (r << 10) | (g << 5) | b;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/
+
+static void
+a4r4g4b4_get_tile_rgba(const ushort *src,
+                       unsigned w, unsigned h,
+                       uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const ushort pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = ((pixel >>  8) & 0xf) * 255 / 15;
+         TILE_PIXEL(p, j, i, 1) = ((pixel >>  4) & 0xf) * 255 / 15;
+         TILE_PIXEL(p, j, i, 2) = ((pixel      ) & 0xf) * 255 / 15;
+         TILE_PIXEL(p, j, i, 3) = ((pixel >> 12)      ) * 255 / 15;
+      }
+   }
+}
+
+
+static void
+a4r4g4b4_put_tile_rgba(ushort *dst,
+                       unsigned w, unsigned h,
+                       const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, g, b, a;
+         r = TILE_PIXEL(p, j, i, 0);
+         g = TILE_PIXEL(p, j, i, 1);
+         b = TILE_PIXEL(p, j, i, 2);
+         a = TILE_PIXEL(p, j, i, 3);
+         r >>= 4;
+         g >>= 4;
+         b >>= 4;
+         a >>= 4;
+         *dst++ = (a << 12) | (r << 16) | (g << 4) | b;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_R5G6B5_UNORM ***/
+
+static void
+r5g6b5_get_tile_rgba(const ushort *src,
+                     unsigned w, unsigned h,
+                     uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         const ushort pixel = *src++;
+         TILE_PIXEL(p, j, i, 0) = ((pixel >> 11) & 0x1f) * 255 / 31;
+         TILE_PIXEL(p, j, i, 1) = ((pixel >>  5) & 0x3f) * 255 / 63;
+         TILE_PIXEL(p, j, i, 2) = ((pixel      ) & 0x1f) * 255 / 31;
+         TILE_PIXEL(p, j, i, 3) = 255;
+      }
+   }
+}
+
+
+static void
+r5g6b5_put_tile_rgba(ushort *dst,
+                     unsigned w, unsigned h,
+                     const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         uint r = (uint) TILE_PIXEL(p, j, i, 0) * 31 / 255;
+         uint g = (uint) TILE_PIXEL(p, j, i, 1) * 63 / 255;
+         uint b = (uint) TILE_PIXEL(p, j, i, 2) * 31 / 255;
+         *dst++ = (r << 11) | (g << 5) | (b);
+      }
+   }
+}
+
+
+
+/*** PIPE_FORMAT_Z16_UNORM ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z16_get_tile_rgba(const ushort *src,
+                  unsigned w, unsigned h,
+                  uint8_t *p)
+{
+   const float scale = 1.0f / 65535.0f;
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = *src++ * scale;
+      }
+   }
+}
+
+
+
+
+/*** PIPE_FORMAT_L8_UNORM ***/
+
+static void
+l8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, src++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) = *src;
+         TILE_PIXEL(p, j, i, 3) = 255;
+      }
+   }
+}
+
+
+static void
+l8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r;
+         r = TILE_PIXEL(p, j, i, 0);
+         *dst++ = (ubyte) r;
+      }
+   }
+}
+
+
+
+/*** PIPE_FORMAT_A8_UNORM ***/
+
+static void
+a8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, src++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) = 0;
+         TILE_PIXEL(p, j, i, 3) = *src;
+      }
+   }
+}
+
+
+static void
+a8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned a;
+         a = TILE_PIXEL(p, j, i, 3);
+         *dst++ = (ubyte) a;
+      }
+   }
+}
+
+
+
+/*** PIPE_FORMAT_R16_SNORM ***/
+
+static void
+r16_get_tile_rgba(const short *src,
+                  unsigned w, unsigned h,
+                  uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, src++) {
+         TILE_PIXEL(p, j, i, 0) = MAX2(src[0] >> 7, 0);
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) = 0;
+         TILE_PIXEL(p, j, i, 3) = 255;
+      }
+   }
+}
+
+
+static void
+r16_put_tile_rgba(short *dst,
+                  unsigned w, unsigned h,
+                  const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, dst++) {
+         dst[0] = TILE_PIXEL(p, j, i, 0) << 7;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/
+
+static void
+r16g16b16a16_get_tile_rgba(const short *src,
+                           unsigned w, unsigned h,
+                           uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, src += 4) {
+         TILE_PIXEL(p, j, i, 0) = src[0] >> 8;
+         TILE_PIXEL(p, j, i, 1) = src[1] >> 8;
+         TILE_PIXEL(p, j, i, 2) = src[2] >> 8;
+         TILE_PIXEL(p, j, i, 3) = src[3] >> 8;
+      }
+   }
+}
+
+
+static void
+r16g16b16a16_put_tile_rgba(short *dst,
+                           unsigned w, unsigned h,
+                           const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, dst += 4) {
+         dst[0] = TILE_PIXEL(p, j, i, 0) << 8;
+         dst[1] = TILE_PIXEL(p, j, i, 1) << 8;
+         dst[2] = TILE_PIXEL(p, j, i, 2) << 8;
+         dst[3] = TILE_PIXEL(p, j, i, 3) << 8;
+      }
+   }
+}
+
+
+
+/*** PIPE_FORMAT_I8_UNORM ***/
+
+static void
+i8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++, src++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = *src;
+      }
+   }
+}
+
+
+static void
+i8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r;
+         r = TILE_PIXEL(p, j, i, 0);
+         *dst++ = (ubyte) r;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_A8L8_UNORM ***/
+
+static void
+a8l8_get_tile_rgba(const ushort *src,
+                   unsigned w, unsigned h,
+                   uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         ushort ra = *src++;
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) = ra & 0xff;
+         TILE_PIXEL(p, j, i, 3) = ra >> 8;
+      }
+   }
+}
+
+
+static void
+a8l8_put_tile_rgba(ushort *dst,
+                   unsigned w, unsigned h,
+                   const uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         unsigned r, a;
+         r = TILE_PIXEL(p, j, i, 0);
+         a = TILE_PIXEL(p, j, i, 3);
+         *dst++ = (a << 8) | r;
+      }
+   }
+}
+
+
+
+
+/*** PIPE_FORMAT_Z32_UNORM ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z32_get_tile_rgba(const unsigned *src,
+                  unsigned w, unsigned h,
+                  uint8_t *p)
+{
+   const double scale = 1.0 / (double) 0xffffffff;
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = (float) (*src++ * scale);
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_S8Z24_UNORM ***/
+
+/**
+ * Return Z component as four float in [0,1].  Stencil part ignored.
+ */
+static void
+s8z24_get_tile_rgba(const unsigned *src,
+                    unsigned w, unsigned h,
+                    uint8_t *p)
+{
+   const double scale = 1.0 / ((1 << 24) - 1);
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = (float) (scale * (*src++ & 0xffffff));
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_Z24S8_UNORM ***/
+
+/**
+ * Return Z component as four float in [0,1].  Stencil part ignored.
+ */
+static void
+z24s8_get_tile_rgba(const unsigned *src,
+                    unsigned w, unsigned h,
+                    uint8_t *p)
+{
+   const double scale = 1.0 / ((1 << 24) - 1);
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = (float) (scale * (*src++ >> 8));
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_Z32_FLOAT ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z32f_get_tile_rgba(const float *src,
+                   unsigned w, unsigned h,
+                   uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = *src++;
+      }
+   }
+}
+
+
+/*** PIPE_FORMAT_YCBCR / PIPE_FORMAT_YCBCR_REV ***/
+
+/**
+ * Convert YCbCr (or YCrCb) to RGBA.
+ */
+static void
+ycbcr_get_tile_rgba(const ushort *src,
+                    unsigned w, unsigned h,
+                    uint8_t *p,
+                    boolean rev)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      /* do two texels at a time */
+      for (j = 0; j < (w & ~1); j += 2, src += 2) {
+         const ushort t0 = src[0];
+         const ushort t1 = src[1];
+         const ubyte y0 = (t0 >> 8) & 0xff;  /* luminance */
+         const ubyte y1 = (t1 >> 8) & 0xff;  /* luminance */
+         ubyte cb, cr;
+         float r, g, b;
+
+         if (rev) {
+            cb = t1 & 0xff;         /* chroma U */
+            cr = t0 & 0xff;         /* chroma V */
+         }
+         else {
+            cb = t0 & 0xff;         /* chroma U */
+            cr = t1 & 0xff;         /* chroma V */
+         }
+
+         /* even pixel: y0,cr,cb */
+         r = 1.164f * (y0-16) + 1.596f * (cr-128);
+         g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y0-16) + 2.018f * (cb-128);
+         TILE_PIXEL(p, j, i, 0) = r;
+         TILE_PIXEL(p, j, i, 1) = g;
+         TILE_PIXEL(p, j, i, 2) = b;
+         TILE_PIXEL(p, j, i, 3) = 255;
+
+         /* odd pixel: use y1,cr,cb */
+         r = 1.164f * (y1-16) + 1.596f * (cr-128);
+         g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y1-16) + 2.018f * (cb-128);
+         TILE_PIXEL(p, j + 1, i, 0) = r;
+         TILE_PIXEL(p, j + 1, i, 1) = g;
+         TILE_PIXEL(p, j + 1, i, 2) = b;
+         TILE_PIXEL(p, j + 1, i, 3) = 255;
+      }
+      /* do the last texel */
+      if (w & 1) {
+         const ushort t0 = src[0];
+         const ushort t1 = src[1];
+         const ubyte y0 = (t0 >> 8) & 0xff;  /* luminance */
+         ubyte cb, cr;
+         float r, g, b;
+
+         if (rev) {
+            cb = t1 & 0xff;         /* chroma U */
+            cr = t0 & 0xff;         /* chroma V */
+         }
+         else {
+            cb = t0 & 0xff;         /* chroma U */
+            cr = t1 & 0xff;         /* chroma V */
+         }
+
+         /* even pixel: y0,cr,cb */
+         r = 1.164f * (y0-16) + 1.596f * (cr-128);
+         g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y0-16) + 2.018f * (cb-128);
+         TILE_PIXEL(p, j, i, 0) = r;
+         TILE_PIXEL(p, j, i, 1) = g;
+         TILE_PIXEL(p, j, i, 2) = b;
+         TILE_PIXEL(p, j, i, 3) = 255;
+      }
+   }
+}
+
+
+static void
+fake_get_tile_rgba(const ushort *src,
+                   unsigned w, unsigned h,
+                   uint8_t *p)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      for (j = 0; j < w; j++) {
+         TILE_PIXEL(p, j, i, 0) =
+         TILE_PIXEL(p, j, i, 1) =
+         TILE_PIXEL(p, j, i, 2) =
+         TILE_PIXEL(p, j, i, 3) = (i ^ j) & 1 ? 255 : 0;
+      }
+   }
+}
+
+
+static void
+lp_tile_raw_to_rgba_soa(enum pipe_format format,
+                        void *src,
+                        uint w, uint h,
+                        uint8_t *p)
+{
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      a1r5g5b5_get_tile_rgba((ushort *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      a4r4g4b4_get_tile_rgba((ushort *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      r5g6b5_get_tile_rgba((ushort *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_L8_UNORM:
+      l8_get_tile_rgba((ubyte *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_A8_UNORM:
+      a8_get_tile_rgba((ubyte *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_I8_UNORM:
+      i8_get_tile_rgba((ubyte *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_A8L8_UNORM:
+      a8l8_get_tile_rgba((ushort *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_R16_SNORM:
+      r16_get_tile_rgba((short *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      r16g16b16a16_get_tile_rgba((short *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      z16_get_tile_rgba((ushort *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      z32_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      s8z24_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      z24s8_get_tile_rgba((unsigned *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_Z32_FLOAT:
+      z32f_get_tile_rgba((float *) src, w, h, p);
+      break;
+   case PIPE_FORMAT_YCBCR:
+      ycbcr_get_tile_rgba((ushort *) src, w, h, p, FALSE);
+      break;
+   case PIPE_FORMAT_YCBCR_REV:
+      ycbcr_get_tile_rgba((ushort *) src, w, h, p, TRUE);
+      break;
+   default:
+      debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format));
+      fake_get_tile_rgba(src, w, h, p);
+   }
+}
+
+
+void
+lp_get_tile_rgba_soa(struct pipe_transfer *pt,
+                     uint x, uint y,
+                     uint8_t *p)
+{
+   uint w = TILE_SIZE, h = TILE_SIZE;
+   void *packed;
+
+   if (pipe_clip_tile(x, y, &w, &h, pt))
+      return;
+
+   packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size);
+
+   if (!packed)
+      return;
+
+   if(pt->format == PIPE_FORMAT_YCBCR || pt->format == PIPE_FORMAT_YCBCR_REV)
+      assert((x & 1) == 0);
+
+   pipe_get_tile_raw(pt, x, y, w, h, packed, 0);
+
+   lp_tile_raw_to_rgba_soa(pt->format, packed, w, h, p);
+
+   FREE(packed);
+}
+
+
+void
+lp_put_tile_rgba_soa(struct pipe_transfer *pt,
+                     uint x, uint y,
+                     const uint8_t *p)
+{
+   uint w = TILE_SIZE, h = TILE_SIZE;
+   void *packed;
+
+   if (pipe_clip_tile(x, y, &w, &h, pt))
+      return;
+
+   packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size);
+
+   if (!packed)
+      return;
+
+   switch (pt->format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      r5g6b5_put_tile_rgba((ushort *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      assert(0);
+      break;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_L8_UNORM:
+      l8_put_tile_rgba((ubyte *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_A8_UNORM:
+      a8_put_tile_rgba((ubyte *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_I8_UNORM:
+      i8_put_tile_rgba((ubyte *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_A8L8_UNORM:
+      a8l8_put_tile_rgba((ushort *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_R16_SNORM:
+      r16_put_tile_rgba((short *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      r16g16b16a16_put_tile_rgba((short *) packed, w, h, p);
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      /*z16_put_tile_rgba((ushort *) packed, w, h, p);*/
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      /*z32_put_tile_rgba((unsigned *) packed, w, h, p);*/
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p);*/
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p);*/
+      break;
+   default:
+      debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(pt->format));
+   }
+
+   pipe_put_tile_raw(pt, x, y, w, h, packed, 0);
+
+   FREE(packed);
+}
+
+
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.h b/src/gallium/drivers/llvmpipe/lp_tile_soa.h
new file mode 100644
index 0000000000..3d8c703b73
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.h
@@ -0,0 +1,81 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef LP_TILE_SOA_H
+#define LP_TILE_SOA_H
+
+#include "pipe/p_compiler.h"
+#include "tgsi/tgsi_exec.h" // for NUM_CHANNELS
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pipe_transfer;
+
+
+/**
+ * Cache tile size (width and height). This needs to be a power of two.
+ */
+#define TILE_SIZE 64
+
+
+#define TILE_VECTOR_HEIGHT 2
+#define TILE_VECTOR_WIDTH 8
+
+extern const unsigned char
+tile_offset[TILE_VECTOR_HEIGHT][TILE_VECTOR_WIDTH];
+
+#define TILE_C_STRIDE (TILE_VECTOR_HEIGHT*TILE_VECTOR_WIDTH)
+#define TILE_X_STRIDE (NUM_CHANNELS*TILE_C_STRIDE)
+#define TILE_Y_STRIDE (TILE_VECTOR_HEIGHT*TILE_SIZE*NUM_CHANNELS)
+
+#define TILE_PIXEL(_p, _x, _y, _c) \
+   ((_p)[((_y)/TILE_VECTOR_HEIGHT)*TILE_Y_STRIDE + \
+         ((_x)/TILE_VECTOR_WIDTH)*TILE_X_STRIDE + \
+         (_c)*TILE_C_STRIDE + \
+         tile_offset[(_y) % TILE_VECTOR_HEIGHT][(_x) % TILE_VECTOR_WIDTH]])
+
+
+void
+lp_get_tile_rgba_soa(struct pipe_transfer *pt,
+                     uint x, uint y,
+                     uint8_t *p);
+
+void
+lp_put_tile_rgba_soa(struct pipe_transfer *pt,
+                     uint x, uint y,
+                     const uint8_t *p);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/drivers/llvmpipe/lp_winsys.h b/src/gallium/drivers/llvmpipe/lp_winsys.h
new file mode 100644
index 0000000000..595481c2cb
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_winsys.h
@@ -0,0 +1,128 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * llvmpipe public interface.
+ */
+
+
+#ifndef LP_WINSYS_H
+#define LP_WINSYS_H
+
+
+#include "pipe/p_compiler.h" // for boolean
+#include "pipe/p_format.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pipe_screen;
+struct pipe_context;
+
+
+/**
+ * Opaque pointer.
+ */
+struct llvmpipe_displaytarget;
+
+
+/**
+ * This is the interface that llvmpipe expects any window system
+ * hosting it to implement.
+ * 
+ * llvmpipe is for the most part a self sufficient driver. The only thing it
+ * does not know is how to display a surface.
+ */
+struct llvmpipe_winsys
+{
+   void 
+   (*destroy)( struct llvmpipe_winsys *ws );
+
+   boolean
+   (*is_displaytarget_format_supported)( struct llvmpipe_winsys *ws,
+                                         enum pipe_format format );
+   
+   /**
+    * Allocate storage for a render target.
+    * 
+    * Often surfaces which are meant to be blitted to the front screen (i.e.,
+    * display targets) must be allocated with special characteristics, memory 
+    * pools, or obtained directly from the windowing system.
+    *  
+    * This callback is invoked by the pipe_screen when creating a texture marked
+    * with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET flag to get the underlying 
+    * storage.
+    */
+   struct llvmpipe_displaytarget *
+   (*displaytarget_create)( struct llvmpipe_winsys *ws,
+                            enum pipe_format format,
+                            unsigned width, unsigned height,
+                            unsigned alignment,
+                            unsigned *stride );
+
+   void *
+   (*displaytarget_map)( struct llvmpipe_winsys *ws, 
+                         struct llvmpipe_displaytarget *dt,
+                         unsigned flags );
+
+   void
+   (*displaytarget_unmap)( struct llvmpipe_winsys *ws,
+                           struct llvmpipe_displaytarget *dt );
+
+   /**
+    * @sa pipe_screen:flush_frontbuffer.
+    *
+    * This call will likely become asynchronous eventually.
+    */
+   void
+   (*displaytarget_display)( struct llvmpipe_winsys *ws, 
+                             struct llvmpipe_displaytarget *dt,
+                             void *context_private );
+
+   void 
+   (*displaytarget_destroy)( struct llvmpipe_winsys *ws, 
+                             struct llvmpipe_displaytarget *dt );
+};
+
+
+struct pipe_context *
+llvmpipe_create( struct pipe_screen * );
+
+
+struct pipe_screen *
+llvmpipe_create_screen( struct llvmpipe_winsys * );
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* LP_WINSYS_H */
diff --git a/src/gallium/drivers/llvmpipe/sp2lp.sh b/src/gallium/drivers/llvmpipe/sp2lp.sh
new file mode 100755
index 0000000000..c45a81ce3c
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/sp2lp.sh
@@ -0,0 +1,34 @@
+#!/bin/sh
+#
+# Port changes from softpipe to llvmpipe. Invoke as
+#
+#   sp2lp.sh <commit>
+#
+# Note that this will only affect llvmpipe -- you still need to actually
+# cherry-pick/merge the softpipe changes themselves if they affect directories
+# outside src/gallium/drivers/softpipe
+
+git format-patch \
+	--keep-subject \
+	--relative=src/gallium/drivers/softpipe \
+	--src-prefix=a/src/gallium/drivers/llvmpipe/ \
+	--dst-prefix=b/src/gallium/drivers/llvmpipe/ \
+	--stdout "$1^1..$1" \
+| sed \
+	-e 's/\<softpipe\>/llvmpipe/g' \
+	-e 's/\<sp\>/lp/g' \
+	-e 's/\<softpipe_/llvmpipe_/g' \
+	-e 's/\<sp_/lp_/g' \
+	-e 's/\<SP_/LP_/g' \
+	-e 's/\<SOFTPIPE_/LLVMPIPE_/g' \
+	-e 's/\<spt\>/lpt/g' \
+	-e 's/\<sps\>/lps/g' \
+	-e 's/\<spfs\>/lpfs/g' \
+	-e 's/\<sptex\>/lptex/g' \
+	-e 's/\<setup_\(point\|line\|tri\)\>/llvmpipe_\0/g' \
+	-e 's/\<llvmpipe_cached_tile\>/llvmpipe_cached_tex_tile/g' \
+	-e 's/_get_cached_tile_tex\>/_get_cached_tex_tile/g' \
+	-e 's/\<TILE_SIZE\>/TEX_TILE_SIZE/g' \
+	-e 's/\<tile_address\>/tex_tile_address/g' \
+	-e 's/\<tile->data\.color\>/tile->color/g' \
+| patch -p1
diff --git a/src/gallium/drivers/nouveau/Makefile b/src/gallium/drivers/nouveau/Makefile
new file mode 100644
index 0000000000..dbe8a6e7bf
--- /dev/null
+++ b/src/gallium/drivers/nouveau/Makefile
@@ -0,0 +1,8 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nouveau
+
+C_SOURCES = nouveau_screen.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/nouveau/nouveau_push.h b/src/gallium/drivers/nouveau/nouveau_push.h
index 54ef1c1291..9c235080a5 100644
--- a/src/gallium/drivers/nouveau/nouveau_push.h
+++ b/src/gallium/drivers/nouveau/nouveau_push.h
@@ -9,13 +9,13 @@
 
 #define OUT_RING(data) do {                                                    \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	(*pc->nvws->channel->pushbuf->cur++) = (data);                         \
+	(*pc->base.channel->pushbuf->cur++) = (data);                          \
 } while(0)
 
 #define OUT_RINGp(src,size) do {                                               \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	memcpy(pc->nvws->channel->pushbuf->cur, (src), (size) * 4);            \
-	pc->nvws->channel->pushbuf->cur += (size);                             \
+	memcpy(pc->base.channel->pushbuf->cur, (src), (size) * 4);             \
+	pc->base.channel->pushbuf->cur += (size);                              \
 } while(0)
 
 #define OUT_RINGf(data) do {                                                   \
@@ -26,25 +26,35 @@
 
 #define BEGIN_RING(obj,mthd,size) do {                                         \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
-		pc->nvws->push_flush(pc->nvws, ((size) + 1), NULL);            \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	if (chan->pushbuf->remaining < ((size) + 1))                           \
+		nouveau_pushbuf_flush(chan, ((size) + 1));                     \
 	OUT_RING((pc->obj->subc << 13) | ((size) << 18) | (mthd));             \
-	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+	chan->pushbuf->remaining -= ((size) + 1);                              \
 } while(0)
 
 #define BEGIN_RING_NI(obj,mthd,size) do {                                      \
 	BEGIN_RING(obj, (mthd) | 0x40000000, (size));                          \
 } while(0)
 
+static inline void
+DO_FIRE_RING(struct nouveau_channel *chan, struct pipe_fence_handle **fence)
+{
+	nouveau_pushbuf_flush(chan, 0);
+	if (fence)
+		*fence = NULL;
+}
+
 #define FIRE_RING(fence) do {                                                  \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	pc->nvws->push_flush(pc->nvws, 0, fence);                              \
+	DO_FIRE_RING(pc->base.channel, fence);                                 \
 } while(0)
 
 #define OUT_RELOC(bo,data,flags,vor,tor) do {                                  \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	pc->nvws->push_reloc(pc->nvws, pc->nvws->channel->pushbuf->cur++,      \
-			     (bo), (data), (flags), (vor), (tor));             \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	nouveau_pushbuf_emit_reloc(chan, chan->pushbuf->cur++, nouveau_bo(bo), \
+				   (data), 0, (flags), (vor), (tor));          \
 } while(0)
 
 /* Raw data + flags depending on FB/TT buffer */
@@ -55,8 +65,8 @@
 /* FB/TT object handle */
 #define OUT_RELOCo(bo,flags) do {                                              \
 	OUT_RELOC((bo), 0, (flags) | NOUVEAU_BO_OR,                            \
-		  pc->nvws->channel->vram->handle,                             \
-		  pc->nvws->channel->gart->handle);                            \
+		  pc->base.channel->vram->handle,                              \
+		  pc->base.channel->gart->handle);                             \
 } while(0)
 
 /* Low 32-bits of offset */
@@ -72,11 +82,12 @@
 /* A reloc which'll recombine into a NV_DMA_METHOD packet header */
 #define OUT_RELOCm(bo, flags, obj, mthd, size) do {                            \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
-		pc->nvws->push_flush(pc->nvws->channel, ((size) + 1), NULL);   \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	if (chan->pushbuf->remaining < ((size) + 1))                           \
+		nouveau_pushbuf_flush(chan, ((size) + 1));                     \
 	OUT_RELOCd((bo), (pc->obj->subc << 13) | ((size) << 18) | (mthd),      \
 		   (flags), 0, 0);                                             \
-	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+	chan->pushbuf->remaining -= ((size) + 1);                              \
 } while(0)
 
 #endif
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c
new file mode 100644
index 0000000000..e4cf91c005
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_screen.c
@@ -0,0 +1,243 @@
+#include <pipe/p_defines.h>
+#include <pipe/p_screen.h>
+#include <pipe/p_state.h>
+
+#include <util/u_memory.h>
+
+#include <errno.h>
+
+#include "nouveau/nouveau_bo.h"
+#include "nouveau_winsys.h"
+#include "nouveau_screen.h"
+
+static const char *
+nouveau_screen_get_name(struct pipe_screen *pscreen)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nouveau_screen_get_vendor(struct pipe_screen *pscreen)
+{
+	return "nouveau";
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_skel(struct pipe_screen *pscreen, struct nouveau_bo *bo,
+		       unsigned alignment, unsigned usage, unsigned size)
+{
+	struct pipe_buffer *pb;
+	
+	pb = CALLOC(1, sizeof(struct pipe_buffer)+sizeof(struct nouveau_bo *));
+	if (!pb) {
+		nouveau_bo_ref(NULL, &bo);
+		return NULL;
+	}
+
+	pipe_reference_init(&pb->reference, 1);
+	pb->screen = pscreen;
+	pb->alignment = alignment;
+	pb->usage = usage;
+	pb->size = size;
+	*(struct nouveau_bo **)(pb + 1) = bo;
+	return pb;
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_new(struct pipe_screen *pscreen, unsigned alignment,
+		      unsigned usage, unsigned size)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct nouveau_bo *bo = NULL;
+	uint32_t flags = NOUVEAU_BO_MAP;
+	int ret;
+
+	if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER)
+		flags |= NOUVEAU_BO_GART;
+	else
+	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF))
+			flags |= NOUVEAU_BO_GART;
+	} else
+	if (usage & PIPE_BUFFER_USAGE_INDEX) {
+		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF))
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
+			flags |= NOUVEAU_BO_GART;
+		if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE))
+			flags |= NOUVEAU_BO_VRAM;
+
+		if (dev->chipset == 0x50 || dev->chipset >= 0x80) {
+			flags |= NOUVEAU_BO_TILED;
+			if (usage & NOUVEAU_BUFFER_USAGE_ZETA)
+				flags |= NOUVEAU_BO_ZTILE;
+		}
+	}
+
+	ret = nouveau_bo_new(dev, flags, alignment, size, &bo);
+	if (ret)
+		return NULL;
+
+	return nouveau_screen_bo_skel(pscreen, bo, alignment, usage, size);
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_user(struct pipe_screen *pscreen, void *ptr, unsigned bytes)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct nouveau_bo *bo = NULL;
+	int ret;
+
+	ret = nouveau_bo_user(dev, ptr, bytes, &bo);
+	if (ret)
+		return NULL;
+
+	return nouveau_screen_bo_skel(pscreen, bo, 0, 0, bytes);
+}
+
+static inline uint32_t
+nouveau_screen_map_flags(unsigned pipe)
+{
+	uint32_t flags = 0;
+
+	if (pipe & PIPE_BUFFER_USAGE_CPU_READ)
+		flags |= NOUVEAU_BO_RD;
+	if (pipe & PIPE_BUFFER_USAGE_CPU_WRITE)
+		flags |= NOUVEAU_BO_WR;
+	if (pipe & PIPE_BUFFER_USAGE_DISCARD)
+		flags |= NOUVEAU_BO_INVAL;
+	if (pipe & PIPE_BUFFER_USAGE_DONTBLOCK)
+		flags |= NOUVEAU_BO_NOWAIT;
+	else
+	if (pipe & 0 /*PIPE_BUFFER_USAGE_UNSYNCHRONIZED*/)
+		flags |= NOUVEAU_BO_NOSYNC;
+
+	return flags;
+}
+
+static void *
+nouveau_screen_bo_map(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+		      unsigned usage)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+	int ret;
+
+	ret = nouveau_bo_map(bo, nouveau_screen_map_flags(usage));
+	if (ret) {
+		debug_printf("map failed: %d\n", ret);
+		return NULL;
+	}
+
+	return bo->map;
+}
+
+static void *
+nouveau_screen_bo_map_range(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+			    unsigned offset, unsigned length, unsigned usage)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+	uint32_t flags = nouveau_screen_map_flags(usage);
+	int ret;
+
+	ret = nouveau_bo_map_range(bo, offset, length, flags);
+	if (ret) {
+		if (!(flags & NOUVEAU_BO_NOWAIT) || ret != -EBUSY)
+			debug_printf("map_range failed: %d\n", ret);
+		return NULL;
+	}
+
+	return (char *)bo->map - offset; /* why gallium? why? */
+}
+
+static void
+nouveau_screen_bo_map_flush(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+			    unsigned offset, unsigned length)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_map_flush(bo, offset, length);
+}
+
+static void
+nouveau_screen_bo_unmap(struct pipe_screen *pscreen, struct pipe_buffer *pb)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_unmap(bo);
+}
+
+static void
+nouveau_screen_bo_del(struct pipe_buffer *pb)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_ref(NULL, &bo);
+	FREE(pb);
+}
+
+static void
+nouveau_screen_fence_ref(struct pipe_screen *pscreen,
+			 struct pipe_fence_handle **ptr,
+			 struct pipe_fence_handle *pfence)
+{
+	*ptr = pfence;
+}
+
+static int
+nouveau_screen_fence_signalled(struct pipe_screen *screen,
+			       struct pipe_fence_handle *pfence,
+			       unsigned flags)
+{
+	return 0;
+}
+
+static int
+nouveau_screen_fence_finish(struct pipe_screen *screen,
+			    struct pipe_fence_handle *pfence,
+			    unsigned flags)
+{
+	return 0;
+}
+
+int
+nouveau_screen_init(struct nouveau_screen *screen, struct nouveau_device *dev)
+{
+	struct pipe_screen *pscreen = &screen->base;
+	int ret;
+
+	ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202,
+				    &screen->channel);
+	if (ret)
+		return ret;
+	screen->device = dev;
+
+	pscreen->get_name = nouveau_screen_get_name;
+	pscreen->get_vendor = nouveau_screen_get_vendor;
+
+	pscreen->buffer_create = nouveau_screen_bo_new;
+	pscreen->user_buffer_create = nouveau_screen_bo_user;
+	pscreen->buffer_map = nouveau_screen_bo_map;
+	pscreen->buffer_map_range = nouveau_screen_bo_map_range;
+	pscreen->buffer_flush_mapped_range = nouveau_screen_bo_map_flush;
+	pscreen->buffer_unmap = nouveau_screen_bo_unmap;
+	pscreen->buffer_destroy = nouveau_screen_bo_del;
+
+	pscreen->fence_reference = nouveau_screen_fence_ref;
+	pscreen->fence_signalled = nouveau_screen_fence_signalled;
+	pscreen->fence_finish = nouveau_screen_fence_finish;
+
+	return 0;
+}
+
+void
+nouveau_screen_fini(struct nouveau_screen *screen)
+{
+}
+
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.h b/src/gallium/drivers/nouveau/nouveau_screen.h
new file mode 100644
index 0000000000..ebfc67ad1c
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_screen.h
@@ -0,0 +1,36 @@
+#ifndef __NOUVEAU_SCREEN_H__
+#define __NOUVEAU_SCREEN_H__
+
+struct nouveau_screen {
+	struct pipe_screen base;
+	struct nouveau_device *device;
+	struct nouveau_channel *channel;
+};
+
+static inline struct nouveau_screen *
+nouveau_screen(struct pipe_screen *pscreen)
+{
+	return (struct nouveau_screen *)pscreen;
+}
+
+static inline struct nouveau_bo *
+nouveau_bo(struct pipe_buffer *pb)
+{
+	return pb ? *(struct nouveau_bo **)(pb + 1) : NULL;
+}
+
+int nouveau_screen_init(struct nouveau_screen *, struct nouveau_device *);
+void nouveau_screen_fini(struct nouveau_screen *);
+
+struct nouveau_miptree {
+	struct pipe_texture base;
+	struct nouveau_bo *bo;
+};
+
+static inline struct nouveau_miptree *
+nouveau_miptree(struct pipe_texture *pt)
+{
+	return (struct nouveau_miptree *)pt;
+}
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h
index a54820e851..b595405357 100644
--- a/src/gallium/drivers/nouveau/nouveau_stateobj.h
+++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h
@@ -4,7 +4,7 @@
 #include "util/u_debug.h"
 
 struct nouveau_stateobj_reloc {
-	struct pipe_buffer *bo;
+	struct nouveau_bo *bo;
 
 	unsigned offset;
 	unsigned packet;
@@ -51,7 +51,7 @@ so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso)
         if (pipe_reference((struct pipe_reference**)pso, &ref->reference)) {
 		free(so->push);
 		for (i = 0; i < so->cur_reloc; i++)
-			pipe_buffer_reference(&so->reloc[i].bo, NULL);
+			nouveau_bo_ref(NULL, &so->reloc[i].bo);
 		free(so->reloc);
 		free(so);
 	}
@@ -81,13 +81,13 @@ so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr,
 }
 
 static INLINE void
-so_reloc(struct nouveau_stateobj *so, struct pipe_buffer *bo,
+so_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo,
 	 unsigned data, unsigned flags, unsigned vor, unsigned tor)
 {
 	struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++];
 	
 	r->bo = NULL;
-	pipe_buffer_reference(&r->bo, bo);
+	nouveau_bo_ref(bo, &r->bo);
 	r->offset = so->cur - so->push;
 	r->packet = so->cur_packet;
 	r->data = data;
@@ -107,50 +107,52 @@ so_dump(struct nouveau_stateobj *so)
 }
 
 static INLINE void
-so_emit(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so)
 {
-	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = chan->pushbuf;
 	unsigned nr, i;
 
 	nr = so->cur - so->push;
 	if (pb->remaining < nr)
-		nvws->push_flush(nvws, nr, NULL);
+		nouveau_pushbuf_flush(chan, nr);
 	pb->remaining -= nr;
 
 	memcpy(pb->cur, so->push, nr * 4);
 	for (i = 0; i < so->cur_reloc; i++) {
 		struct nouveau_stateobj_reloc *r = &so->reloc[i];
 
-		nvws->push_reloc(nvws, pb->cur + r->offset, r->bo,
-				 r->data, r->flags, r->vor, r->tor);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset,
+					   r->bo, r->data, 0, r->flags,
+					   r->vor, r->tor);
 	}
 	pb->cur += nr;
 }
 
 static INLINE void
-so_emit_reloc_markers(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+so_emit_reloc_markers(struct nouveau_channel *chan, struct nouveau_stateobj *so)
 {
-	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = chan->pushbuf;
 	unsigned i;
 
 	if (!so)
 		return;
 
 	i = so->cur_reloc << 1;
-	if (nvws->channel->pushbuf->remaining < i)
-		nvws->push_flush(nvws, i, NULL);
-	nvws->channel->pushbuf->remaining -= i;
+	if (pb->remaining < i)
+		nouveau_pushbuf_flush(chan, i);
+	pb->remaining -= i;
 
 	for (i = 0; i < so->cur_reloc; i++) {
 		struct nouveau_stateobj_reloc *r = &so->reloc[i];
 
-		nvws->push_reloc(nvws, pb->cur++, r->bo, r->packet,
-				 (r->flags & (NOUVEAU_BO_VRAM |
-					      NOUVEAU_BO_GART |
-					      NOUVEAU_BO_RDWR)) |
-				 NOUVEAU_BO_DUMMY, 0, 0);
-		nvws->push_reloc(nvws, pb->cur++, r->bo, r->data,
-				 r->flags | NOUVEAU_BO_DUMMY, r->vor, r->tor);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->packet, 0,
+					   (r->flags & (NOUVEAU_BO_VRAM |
+							NOUVEAU_BO_GART |
+							NOUVEAU_BO_RDWR)) |
+					   NOUVEAU_BO_DUMMY, 0, 0);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->data, 0,
+					   r->flags | NOUVEAU_BO_DUMMY,
+					   r->vor, r->tor);
 	}
 }
 
diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h
index ff7dd1c51c..42c77e5e77 100644
--- a/src/gallium/drivers/nouveau/nouveau_winsys.h
+++ b/src/gallium/drivers/nouveau/nouveau_winsys.h
@@ -23,77 +23,38 @@
 #define NOUVEAU_BUFFER_USAGE_ZETA     (1 << 17)
 #define NOUVEAU_BUFFER_USAGE_TRANSFER (1 << 18)
 
-struct nouveau_winsys {
-	struct pipe_winsys *ws;
-
-	struct nouveau_channel *channel;
-
-	int  (*res_init)(struct nouveau_resource **heap, unsigned start,
-			 unsigned size);
-	int  (*res_alloc)(struct nouveau_resource *heap, int size, void *priv,
-			  struct nouveau_resource **);
-	void (*res_free)(struct nouveau_resource **);
-
-	int  (*push_reloc)(struct nouveau_winsys *, void *ptr,
-			   struct pipe_buffer *, uint32_t data,
-			   uint32_t flags, uint32_t vor, uint32_t tor);
-	int  (*push_flush)(struct nouveau_winsys *, unsigned size,
-			   struct pipe_fence_handle **fence);
-			       
-	int       (*grobj_alloc)(struct nouveau_winsys *, int grclass,
-				 struct nouveau_grobj **);
-	void      (*grobj_free)(struct nouveau_grobj **);
-
-	int       (*notifier_alloc)(struct nouveau_winsys *, int count,
-				    struct nouveau_notifier **);
-	void      (*notifier_free)(struct nouveau_notifier **);
-	void      (*notifier_reset)(struct nouveau_notifier *, int id);
-	uint32_t  (*notifier_status)(struct nouveau_notifier *, int id);
-	uint32_t  (*notifier_retval)(struct nouveau_notifier *, int id);
-	int       (*notifier_wait)(struct nouveau_notifier *, int id,
-				   int status, double timeout);
-
-	int (*surface_copy)(struct nouveau_winsys *, struct pipe_surface *,
-			    unsigned, unsigned, struct pipe_surface *,
-			    unsigned, unsigned, unsigned, unsigned);
-	int (*surface_fill)(struct nouveau_winsys *, struct pipe_surface *,
-			    unsigned, unsigned, unsigned, unsigned, unsigned);
-
-	struct nouveau_bo *(*get_bo)(struct pipe_buffer *);
-};
-
 extern struct pipe_screen *
-nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv04_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv10_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv20_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv30_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv40_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv50_create(struct pipe_screen *, unsigned pctx_id);
diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c
index 4da833c25e..93f752faec 100644
--- a/src/gallium/drivers/nv04/nv04_miptree.c
+++ b/src/gallium/drivers/nv04/nv04_miptree.c
@@ -31,7 +31,8 @@ nv04_miptree_layout(struct nv04_miptree *nv04mt)
 
 	for (l = 0; l <= pt->last_level; l++) {
 
-		nv04mt->level[l].image_offset = offset;
+		nv04mt->level[l].image_offset = 
+			CALLOC(nr_faces, sizeof(unsigned));
 		offset += nv04mt->level[l].pitch * pt->height[l];
 	}
 
diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c
index f9f6d97426..ff2febb668 100644
--- a/src/gallium/drivers/nv04/nv04_screen.c
+++ b/src/gallium/drivers/nv04/nv04_screen.c
@@ -1,27 +1,9 @@
 #include "pipe/p_screen.h"
 #include "pipe/p_inlines.h"
-#include "util/u_simple_screen.h"
 
 #include "nv04_context.h"
 #include "nv04_screen.h"
 
-static const char *
-nv04_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv04_screen *nv04screen = nv04_screen(screen);
-	struct nouveau_device *dev = nv04screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv04_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv04_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -58,6 +40,10 @@ nv04_screen_get_param(struct pipe_screen *screen, int param)
 		return 0;
 	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
 		return 1;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -97,6 +83,13 @@ nv04_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -123,10 +116,9 @@ static void
 nv04_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv04_screen *screen = nv04_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->fahrenheit);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->fahrenheit);
 	nv04_surface_2d_takedown(&screen->eng2d);
 
 	FREE(pscreen);
@@ -141,21 +133,38 @@ nv04_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv04_screen *screen = CALLOC_STRUCT(nv04_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned fahrenheit_class = 0, sub3d_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
+
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv04_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv04_screen_destroy;
+	pscreen->get_param = nv04_screen_get_param;
+	pscreen->get_paramf = nv04_screen_get_paramf;
+	pscreen->is_format_supported = nv04_screen_is_format_supported;
 
-	if (chipset>=0x20) {
+	nv04_screen_init_miptree_functions(pscreen);
+	nv04_screen_init_transfer_functions(pscreen);
+
+	if (dev->chipset >= 0x20) {
 		fahrenheit_class = 0;
 		sub3d_class = 0;
-	} else if (chipset>=0x10) {
+	} else if (dev->chipset >= 0x10) {
 		fahrenheit_class = NV10_DX5_TEXTURED_TRIANGLE;
 		sub3d_class = NV10_CONTEXT_SURFACES_3D;
 	} else {
@@ -164,50 +173,40 @@ nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	}
 
 	if (!fahrenheit_class) {
-		NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv04_surface_buffer;
-
 	/* 3D object */
-	ret = nvws->grobj_alloc(nvws, fahrenheit_class, &screen->fahrenheit);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0001, fahrenheit_class,
+				  &screen->fahrenheit);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return NULL;
 	}
+	BIND_RING(chan, screen->fahrenheit, 7);
 
 	/* 3D surface object */
-	ret = nvws->grobj_alloc(nvws, sub3d_class, &screen->context_surfaces_3d);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0002, sub3d_class,
+				  &screen->context_surfaces_3d);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D surface object: %d\n", ret);
 		return NULL;
 	}
+	BIND_RING(chan, screen->context_surfaces_3d, 6);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv04_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv04_screen_destroy(&screen->pipe);
+		nv04_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv04_screen_destroy;
-
-	screen->pipe.get_name = nv04_screen_get_name;
-	screen->pipe.get_vendor = nv04_screen_get_vendor;
-	screen->pipe.get_param = nv04_screen_get_param;
-	screen->pipe.get_paramf = nv04_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv04_screen_is_format_supported;
-
-	nv04_screen_init_miptree_functions(&screen->pipe);
-	nv04_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h
index ee6fb6db44..11466b9442 100644
--- a/src/gallium/drivers/nv04/nv04_screen.h
+++ b/src/gallium/drivers/nv04/nv04_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV04_SCREEN_H__
 #define __NV04_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04_surface_2d.h"
 
 struct nv04_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 	unsigned chipset;
diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c
index 87c635f962..d356ebd8b3 100644
--- a/src/gallium/drivers/nv04/nv04_state.c
+++ b/src/gallium/drivers/nv04/nv04_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -334,7 +335,7 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv04_context *nv04 = nv04_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -342,12 +343,12 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv04->constbuf[shader], mapped, buf->buffer->size);
 			nv04->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h
index 0d51439e3f..399f750dbe 100644
--- a/src/gallium/drivers/nv04/nv04_state.h
+++ b/src/gallium/drivers/nv04/nv04_state.h
@@ -37,7 +37,7 @@ struct nv04_miptree {
 
 	struct {
 		uint pitch;
-		uint image_offset;
+		uint *image_offset;
 	} level[PIPE_MAX_TEXTURE_LEVELS];
 };
 
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c
index f3a8d7efee..f88e138c79 100644
--- a/src/gallium/drivers/nv04/nv04_surface_2d.c
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.c
@@ -4,6 +4,7 @@
 
 #include "nouveau/nouveau_winsys.h"
 #include "nouveau/nouveau_util.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04_surface_2d.h"
 
 static INLINE int
@@ -14,11 +15,13 @@ nv04_surface_format(enum pipe_format format)
 		return NV04_CONTEXT_SURFACES_2D_FORMAT_Y8;
 	case PIPE_FORMAT_R16_SNORM:
 	case PIPE_FORMAT_R5G6B5_UNORM:
+	case PIPE_FORMAT_Z16_UNORM:
 		return NV04_CONTEXT_SURFACES_2D_FORMAT_R5G6B5;
 	case PIPE_FORMAT_X8R8G8B8_UNORM:
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
 		return NV04_CONTEXT_SURFACES_2D_FORMAT_A8R8G8B8;
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 		return NV04_CONTEXT_SURFACES_2D_FORMAT_Y32;
 	default:
 		return -1;
@@ -32,9 +35,11 @@ nv04_rect_format(enum pipe_format format)
 	case PIPE_FORMAT_A8_UNORM:
 		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8;
 	case PIPE_FORMAT_R5G6B5_UNORM:
+	case PIPE_FORMAT_Z16_UNORM:
 		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A16R5G6B5;
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8;
 	default:
 		return -1;
@@ -96,11 +101,11 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 			  struct pipe_surface *src, int sx, int sy,
 			  int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->swzsurf->channel;
 	struct nouveau_grobj *swzsurf = ctx->swzsurf;
 	struct nouveau_grobj *sifm = ctx->sifm;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	const unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	const unsigned max_w = 1024;
 	const unsigned max_h = 1024;
@@ -109,10 +114,10 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 	unsigned cx;
 	unsigned cy;
 
-	/* POT or GTFO */
-	assert(!(w & (w - 1)) && !(h & (h - 1)));
+#if 0
 	/* That's the way she likes it */
 	assert(src_pitch == ((struct nv04_surface *)dst)->pitch);
+#endif
 
 	BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_DMA_IMAGE, 1);
 	OUT_RELOCo(chan, dst_bo,
@@ -132,7 +137,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 	for (cy = 0; cy < h; cy += sub_h) {
 	  for (cx = 0; cx < w; cx += sub_w) {
 	    BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1);
-	    OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx, cy) *
+	    OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx+dx, cy+dy) *
 			     dst->texture->block.size, NOUVEAU_BO_GART |
 			     NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 
@@ -152,8 +157,8 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 	    OUT_RING  (chan, src_pitch |
 			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER |
 			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE);
-	    OUT_RELOCl(chan, src_bo, src->offset + cy * src_pitch +
-			     cx * src->texture->block.size, NOUVEAU_BO_GART |
+	    OUT_RELOCl(chan, src_bo, src->offset + (cy+sy) * src_pitch +
+			     (cx+sx) * src->texture->block.size, NOUVEAU_BO_GART |
 			     NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
 	    OUT_RING  (chan, 0);
 	  }
@@ -167,10 +172,10 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
 		       struct pipe_surface *dst, int dx, int dy,
 		       struct pipe_surface *src, int sx, int sy, int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->m2mf->channel;
 	struct nouveau_grobj *m2mf = ctx->m2mf;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	unsigned dst_offset = dst->offset + dy * dst_pitch +
@@ -209,15 +214,52 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
 }
 
 static int
+nv04_surface_copy_m2mf_swizzle(struct nv04_surface_2d *ctx,
+			       struct pipe_surface *dst, int dx, int dy,
+			       struct pipe_surface *src, int sx, int sy)
+{
+	struct nouveau_channel *chan = ctx->m2mf->channel;
+	struct nouveau_grobj *m2mf = ctx->m2mf;
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
+	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
+	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
+	unsigned dst_offset = dst->offset + nv04_swizzle_bits(dx, dy) *
+	                      dst->texture->block.size;
+	unsigned src_offset = src->offset + sy * src_pitch +
+	                      sx * src->texture->block.size;
+
+	BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2);
+	OUT_RELOCo(chan, src_bo,
+		   NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCo(chan, dst_bo,
+		   NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
+	OUT_RELOCl(chan, src_bo, src_offset,
+		   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+	OUT_RELOCl(chan, dst_bo, dst_offset,
+		   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR);
+	OUT_RING  (chan, src_pitch);
+	OUT_RING  (chan, dst_pitch);
+	OUT_RING  (chan, 1 * src->texture->block.size);
+	OUT_RING  (chan, 1);
+	OUT_RING  (chan, 0x0101);
+	OUT_RING  (chan, 0);
+
+	return 0;
+}
+
+static int
 nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 		       int dx, int dy, struct pipe_surface *src, int sx, int sy,
 		       int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->surf2d->channel;
 	struct nouveau_grobj *surf2d = ctx->surf2d;
 	struct nouveau_grobj *blit = ctx->blit;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	int format;
@@ -257,8 +299,59 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 	assert(src->format == dst->format);
 
 	/* Setup transfer to swizzle the texture to vram if needed */
-	if (src_linear && !dst_linear && w > 1 && h > 1) {
-		nv04_surface_copy_swizzle(ctx, dst, dx, dy, src, sx, sy, w, h);
+	if (src_linear && !dst_linear) {
+		int x,y;
+
+		if ((w>1) && (h>1)) {
+			int potWidth = 1<<log2i(w);
+			int potHeight = 1<<log2i(h);
+			int remainWidth = w-potWidth;
+			int remainHeight = h-potHeight;
+			int squareDim = (potWidth>potHeight ? potHeight : potWidth);
+
+			/* top left is always POT, but we can only swizzle squares */
+			for (y=0; y<potHeight; y+=squareDim) {
+				for (x=0; x<potWidth; x+= squareDim) {
+					nv04_surface_copy_swizzle(ctx, dst, dx+x, dy+y,
+					                          src, sx+x, sy+y,
+					                          squareDim, squareDim);
+				}
+			}
+
+			/* top right */
+			if (remainWidth>0) {
+			nv04_surface_copy(ctx, dst, dx+potWidth, dy,
+				                  src, sx+potWidth, sy,
+				                  remainWidth, potHeight);
+			}
+
+			/* bottom left */
+			if (remainHeight>0) {
+				nv04_surface_copy(ctx, dst, dx, dy+potHeight,
+			                  src, sx, sy+potHeight,
+				                  potWidth, remainHeight);
+			}
+
+			/* bottom right */
+			if ((remainWidth>0) && (remainHeight>0)) {
+				nv04_surface_copy(ctx, dst, dx+potWidth, dy+potHeight,
+				                  src, sx+potWidth, sy+potHeight,
+				                  remainWidth, remainHeight);
+			}
+		} else if (w==1) {
+			/* We have a column to copy to a swizzled texture */
+			for (y=0; y<h; y++) {
+				nv04_surface_copy_m2mf_swizzle(ctx, dst, dx, dy+y,
+				                               src, sx, sy+y);
+			}
+		} else if (h==1) {
+			/* We have a row to copy to a swizzled texture */
+			for (x=0; x<w; x++) {
+				nv04_surface_copy_m2mf_swizzle(ctx, dst, dx+x, dy,
+				                               src, sx+x, sy);
+			}
+		}
+
 		return;
 	}
 
@@ -266,8 +359,7 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 	 * to NV_MEMORY_TO_MEMORY_FORMAT in this case.
 	 */
 	if ((src->offset & 63) || (dst->offset & 63) ||
-	    (src_pitch & 63) || (dst_pitch & 63) ||
-	    debug_get_bool_option("NOUVEAU_NO_COPYBLIT", FALSE)) {
+	    (src_pitch & 63) || (dst_pitch & 63)) {
 		nv04_surface_copy_m2mf(ctx, dst, dx, dy, src, sx, sy, w, h);
 		return;
 	}
@@ -279,10 +371,10 @@ static void
 nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 		  int dx, int dy, int w, int h, unsigned value)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->surf2d->channel;
 	struct nouveau_grobj *surf2d = ctx->surf2d;
 	struct nouveau_grobj *rect = ctx->rect;
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	int cs2d_format, gdirect_format;
 
@@ -334,10 +426,10 @@ nv04_surface_2d_takedown(struct nv04_surface_2d **pctx)
 }
 
 struct nv04_surface_2d *
-nv04_surface_2d_init(struct nouveau_winsys *nvws)
+nv04_surface_2d_init(struct nouveau_screen *screen)
 {
 	struct nv04_surface_2d *ctx = CALLOC_STRUCT(nv04_surface_2d);
-	struct nouveau_channel *chan = nvws->channel;
+	struct nouveau_channel *chan = screen->channel;
 	unsigned handle = 0x88000000, class;
 	int ret;
 
@@ -460,7 +552,6 @@ nv04_surface_2d_init(struct nouveau_winsys *nvws)
 		return NULL;
 	}
 
-	ctx->nvws = nvws;
 	ctx->copy = nv04_surface_copy;
 	ctx->fill = nv04_surface_fill;
 	return ctx;
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.h b/src/gallium/drivers/nv04/nv04_surface_2d.h
index 82ce7189c8..02b3f56ba8 100644
--- a/src/gallium/drivers/nv04/nv04_surface_2d.h
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.h
@@ -7,7 +7,6 @@ struct nv04_surface {
 };
 
 struct nv04_surface_2d {
-	struct nouveau_winsys *nvws;
 	struct nouveau_notifier *ntfy;
 	struct nouveau_grobj *surf2d;
 	struct nouveau_grobj *swzsurf;
@@ -26,7 +25,7 @@ struct nv04_surface_2d {
 };
 
 struct nv04_surface_2d *
-nv04_surface_2d_init(struct nouveau_winsys *nvws);
+nv04_surface_2d_init(struct nouveau_screen *screen);
 
 void
 nv04_surface_2d_takedown(struct nv04_surface_2d **);
diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c
index d21a0e34f7..e3167814f2 100644
--- a/src/gallium/drivers/nv04/nv04_vbo.c
+++ b/src/gallium/drivers/nv04/nv04_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv04_context.h"
 #include "nv04_state.h"
@@ -13,6 +14,7 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
                     unsigned indexSize,
                     unsigned prim, unsigned start, unsigned count)
 {
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nv04_context *nv04 = nv04_context( pipe );
 	struct draw_context *draw = nv04->draw;
 	unsigned i;
@@ -25,17 +27,17 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv04->vtxbuf[i].buffer) {
 			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv04->vtxbuf[i].buffer,
-						PIPE_BUFFER_USAGE_CPU_READ);
+				= pipe_buffer_map(pscreen,
+						  nv04->vtxbuf[i].buffer,
+						  PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
 	}
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +57,12 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv04->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv04->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv04->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv10/nv10_context.c b/src/gallium/drivers/nv10/nv10_context.c
index 3da8d2f568..a127b134ec 100644
--- a/src/gallium/drivers/nv10/nv10_context.c
+++ b/src/gallium/drivers/nv10/nv10_context.c
@@ -30,18 +30,18 @@ nv10_destroy(struct pipe_context *pipe)
 static void nv10_init_hwctx(struct nv10_context *nv10)
 {
 	struct nv10_screen *screen = nv10->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 	int i;
 	float projectionmatrix[16];
 
 	BEGIN_RING(celsius, NV10TCL_DMA_NOTIFY, 1);
 	OUT_RING  (screen->sync->handle);
 	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY0, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->gart->handle);
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->gart->handle);
 	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY2, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->vram->handle);
 
 	BEGIN_RING(celsius, NV10TCL_NOP, 1);
 	OUT_RING  (0);
diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
index 089c236302..1806d5f8cc 100644
--- a/src/gallium/drivers/nv10/nv10_prim_vbuf.c
+++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
@@ -40,7 +40,6 @@
 
 #include "util/u_debug.h"
 #include "pipe/p_inlines.h"
-#include "pipe/internal/p_winsys_screen.h"
 
 #include "nv10_context.h"
 #include "nv10_state.h"
@@ -124,11 +123,10 @@ nv10_vbuf_render_map_vertices( struct vbuf_render *render )
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
 	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	struct pipe_screen *pscreen = nv10->pipe.screen;
 
-	return winsys->buffer_map(winsys, 
-			nv10_render->buffer, 
-			PIPE_BUFFER_USAGE_CPU_WRITE);
+	return pipe_buffer_map(pscreen, nv10_render->buffer,
+			       PIPE_BUFFER_USAGE_CPU_WRITE);
 }
 
 static void
@@ -138,10 +136,10 @@ nv10_vbuf_render_unmap_vertices( struct vbuf_render *render,
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
 	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	struct pipe_screen *pscreen = nv10->pipe.screen;
 
 	assert(!nv10_render->buffer);
-	winsys->buffer_unmap(winsys, nv10_render->buffer);
+	pipe_buffer_unmap(pscreen, nv10_render->buffer);
 }
 
 static boolean
@@ -202,8 +200,6 @@ static void
 nv10_vbuf_render_release_vertices( struct vbuf_render *render )
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
-	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_screen *pscreen = &nv10->screen->pipe;
 
 	assert(nv10_render->buffer);
 	pipe_buffer_reference(&nv10_render->buffer, NULL);
diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c
index 6532a93c7b..4469b22d91 100644
--- a/src/gallium/drivers/nv10/nv10_screen.c
+++ b/src/gallium/drivers/nv10/nv10_screen.c
@@ -1,26 +1,8 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv10_context.h"
 #include "nv10_screen.h"
 
-static const char *
-nv10_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv10_screen *nv10screen = nv10_screen(screen);
-	struct nouveau_device *dev = nv10screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv10_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv10_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -53,6 +35,10 @@ nv10_screen_get_param(struct pipe_screen *screen, int param)
 		return 12;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -92,7 +78,15 @@ nv10_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			 return TRUE;
+		default:
+			 break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -120,10 +114,9 @@ static void
 nv10_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv10_screen *screen = nv10_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->celsius);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->celsius);
 
 	FREE(pscreen);
 }
@@ -137,64 +130,69 @@ nv10_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv10_screen *screen = CALLOC_STRUCT(nv10_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned celsius_class;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv10_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv10_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv10_screen_destroy;
+	pscreen->get_param = nv10_screen_get_param;
+	pscreen->get_paramf = nv10_screen_get_paramf;
+	pscreen->is_format_supported = nv10_screen_is_format_supported;
+
+	nv10_screen_init_miptree_functions(pscreen);
+	nv10_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	if (chipset>=0x20)
-		celsius_class=NV11TCL;
-	else if (chipset>=0x17)
-		celsius_class=NV17TCL;
-	else if (chipset>=0x11)
-		celsius_class=NV11TCL;
+	if (dev->chipset >= 0x20)
+		celsius_class = NV11TCL;
+	else if (dev->chipset >= 0x17)
+		celsius_class = NV17TCL;
+	else if (dev->chipset >= 0x11)
+		celsius_class = NV11TCL;
 	else
-		celsius_class=NV10TCL;
+		celsius_class = NV10TCL;
 
 	if (!celsius_class) {
-		NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, celsius_class, &screen->celsius);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0001, celsius_class,
+				  &screen->celsius);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->celsius, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv10_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv10_screen_destroy(&screen->pipe);
+		nv10_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv10_screen_destroy;
-
-	screen->pipe.get_name = nv10_screen_get_name;
-	screen->pipe.get_vendor = nv10_screen_get_vendor;
-	screen->pipe.get_param = nv10_screen_get_param;
-	screen->pipe.get_paramf = nv10_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv10_screen_is_format_supported;
-
-	nv10_screen_init_miptree_functions(&screen->pipe);
-	nv10_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h
index ad829ee3fd..86b6d8def5 100644
--- a/src/gallium/drivers/nv10/nv10_screen.h
+++ b/src/gallium/drivers/nv10/nv10_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV10_SCREEN_H__
 #define __NV10_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv10_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c
index 119af66dfd..9b38219b99 100644
--- a/src/gallium/drivers/nv10/nv10_state.c
+++ b/src/gallium/drivers/nv10/nv10_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -460,7 +461,7 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv10_context *nv10 = nv10_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -468,12 +469,12 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv10->constbuf[shader], mapped, buf->buffer->size);
 			nv10->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c
index d0e788ac03..441a4f75f3 100644
--- a/src/gallium/drivers/nv10/nv10_vbo.c
+++ b/src/gallium/drivers/nv10/nv10_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv10_context.h"
 #include "nv10_state.h"
@@ -15,6 +16,7 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 {
 	struct nv10_context *nv10 = nv10_context( pipe );
 	struct draw_context *draw = nv10->draw;
+	struct pipe_screen *pscreen = pipe->screen;
 	unsigned i;
 
 	nv10_emit_hw_state(nv10);
@@ -24,9 +26,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv10->vtxbuf[i].buffer) {
-			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv10->vtxbuf[i].buffer,
+			void *buf =
+				pipe_buffer_map(pscreen, nv10->vtxbuf[i].buffer,
 						PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
@@ -34,8 +35,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +56,12 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv10->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv10->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv10->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv20/nv20_context.c b/src/gallium/drivers/nv20/nv20_context.c
index cbc41707d5..b32d0d83ba 100644
--- a/src/gallium/drivers/nv20/nv20_context.c
+++ b/src/gallium/drivers/nv20/nv20_context.c
@@ -30,7 +30,7 @@ nv20_destroy(struct pipe_context *pipe)
 static void nv20_init_hwctx(struct nv20_context *nv20)
 {
 	struct nv20_screen *screen = nv20->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 	int i;
 	float projectionmatrix[16];
 	const boolean is_nv25tcl = (nv20->screen->kelvin->grclass == NV25TCL);
@@ -38,11 +38,11 @@ static void nv20_init_hwctx(struct nv20_context *nv20)
 	BEGIN_RING(kelvin, NV20TCL_DMA_NOTIFY, 1);
 	OUT_RING  (screen->sync->handle);
 	BEGIN_RING(kelvin, NV20TCL_DMA_TEXTURE0, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->gart->handle); /* TEXTURE1 */
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->gart->handle); /* TEXTURE1 */
 	BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->vram->handle); /* ZETA */
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->vram->handle); /* ZETA */
 
 	BEGIN_RING(kelvin, NV20TCL_DMA_QUERY, 1);
 	OUT_RING  (0); /* renouveau: beef0351, unique */
@@ -99,9 +99,9 @@ static void nv20_init_hwctx(struct nv20_context *nv20)
 		OUT_RING  (3);
 
 		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY9, 1);
-		OUT_RING  (nvws->channel->vram->handle);
+		OUT_RING  (chan->vram->handle);
 		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY8, 1);
-		OUT_RING  (nvws->channel->vram->handle);
+		OUT_RING  (chan->vram->handle);
 	}
 	BEGIN_RING(kelvin, NV20TCL_DMA_FENCE, 1);
 	OUT_RING  (0);	/* renouveau: beef1e10 */
diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
index 8aa342cd2d..ddfcdb8057 100644
--- a/src/gallium/drivers/nv20/nv20_prim_vbuf.c
+++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
@@ -152,12 +152,11 @@ static void *
 nv20_vbuf_render_map_vertices( struct vbuf_render *render )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
-	struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
+	struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen;
 
 	if (nv20_render->pbuffer) {
-		return winsys->buffer_map(winsys,
-				nv20_render->pbuffer,
-				PIPE_BUFFER_USAGE_CPU_WRITE);
+		return pipe_buffer_map(pscreen, nv20_render->pbuffer,
+				       PIPE_BUFFER_USAGE_CPU_WRITE);
 	} else if (nv20_render->mbuffer) {
 		return nv20_render->mbuffer;
 	} else
@@ -173,10 +172,10 @@ nv20_vbuf_render_unmap_vertices( struct vbuf_render *render,
 		ushort max_index )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
-	struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
+	struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen;
 
 	if (nv20_render->pbuffer)
-		winsys->buffer_unmap(winsys, nv20_render->pbuffer);
+		pipe_buffer_unmap(pscreen, nv20_render->pbuffer);
 }
 
 static boolean
@@ -358,7 +357,6 @@ nv20_vbuf_render_release_vertices( struct vbuf_render *render )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 	struct nv20_context *nv20 = nv20_render->nv20;
-	struct pipe_screen *pscreen = &nv20->screen->pipe;
 
 	if (nv20_render->pbuffer) {
 		pipe_buffer_reference(&nv20_render->pbuffer, NULL);
diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c
index 7760ae27c0..e6924ad71e 100644
--- a/src/gallium/drivers/nv20/nv20_screen.c
+++ b/src/gallium/drivers/nv20/nv20_screen.c
@@ -1,26 +1,8 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv20_context.h"
 #include "nv20_screen.h"
 
-static const char *
-nv20_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv20_screen *nv20screen = nv20_screen(screen);
-	struct nouveau_device *dev = nv20screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv20_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv20_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -53,6 +35,10 @@ nv20_screen_get_param(struct pipe_screen *screen, int param)
 		return 12;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -92,7 +78,15 @@ nv20_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			 break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -120,10 +114,9 @@ static void
 nv20_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv20_screen *screen = nv20_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->kelvin);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->kelvin);
 
 	FREE(pscreen);
 }
@@ -137,60 +130,65 @@ nv20_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv20_screen *screen = CALLOC_STRUCT(nv20_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned kelvin_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv20_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv20_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv20_screen_destroy;
+	pscreen->get_param = nv20_screen_get_param;
+	pscreen->get_paramf = nv20_screen_get_paramf;
+	pscreen->is_format_supported = nv20_screen_is_format_supported;
+
+	nv20_screen_init_miptree_functions(pscreen);
+	nv20_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	if (chipset >= 0x25)
+	if (dev->chipset >= 0x25)
 		kelvin_class = NV25TCL;
-	else if (chipset >= 0x20)
+	else if (dev->chipset >= 0x20)
 		kelvin_class = NV20TCL;
 
-	if (!kelvin_class || chipset >= 0x30) {
-		NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", chipset);
+	if (!kelvin_class || dev->chipset >= 0x30) {
+		NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, kelvin_class, &screen->kelvin);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0097, kelvin_class,
+				  &screen->kelvin);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->kelvin, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv20_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv20_screen_destroy(&screen->pipe);
+		nv20_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv20_screen_destroy;
-
-	screen->pipe.get_name = nv20_screen_get_name;
-	screen->pipe.get_vendor = nv20_screen_get_vendor;
-	screen->pipe.get_param = nv20_screen_get_param;
-	screen->pipe.get_paramf = nv20_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv20_screen_is_format_supported;
-
-	nv20_screen_init_miptree_functions(&screen->pipe);
-	nv20_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h
index d9fce2bced..fc7bb05033 100644
--- a/src/gallium/drivers/nv20/nv20_screen.h
+++ b/src/gallium/drivers/nv20/nv20_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV20_SCREEN_H__
 #define __NV20_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv20_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c
index ecec4f49a0..ed4084980f 100644
--- a/src/gallium/drivers/nv20/nv20_state.c
+++ b/src/gallium/drivers/nv20/nv20_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -453,7 +454,7 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv20_context *nv20 = nv20_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -461,12 +462,12 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv20->constbuf[shader], mapped, buf->buffer->size);
 			nv20->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c
index 24d8f4bef0..84d7db6c5e 100644
--- a/src/gallium/drivers/nv20/nv20_vbo.c
+++ b/src/gallium/drivers/nv20/nv20_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv20_context.h"
 #include "nv20_state.h"
@@ -13,6 +14,7 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
                     unsigned indexSize,
                     unsigned prim, unsigned start, unsigned count)
 {
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nv20_context *nv20 = nv20_context( pipe );
 	struct draw_context *draw = nv20->draw;
 	unsigned i;
@@ -25,17 +27,17 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv20->vtxbuf[i].buffer) {
 			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv20->vtxbuf[i].buffer,
-						PIPE_BUFFER_USAGE_CPU_READ);
+				= pipe_buffer_map(pscreen,
+						  nv20->vtxbuf[i].buffer,
+						  PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
 	}
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +57,12 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv20->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv20->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv20->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c
index 5db0e807ff..388245ecb0 100644
--- a/src/gallium/drivers/nv20/nv20_vertprog.c
+++ b/src/gallium/drivers/nv20/nv20_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -616,10 +617,10 @@ nv20_vertprog_translate(struct nv20_context *nv20,
 			assert(imm->Immediate.NrTokens == 4 + 1);
 			vpc->imm[vpc->nr_imm++] =
 				constant(vpc, -1,
-					 imm->u.ImmediateFloat32[0].Float,
-					 imm->u.ImmediateFloat32[1].Float,
-					 imm->u.ImmediateFloat32[2].Float,
-					 imm->u.ImmediateFloat32[3].Float);
+					 imm->u[0].Float,
+					 imm->u[1].Float,
+					 imm->u[2].Float,
+					 imm->u[3].Float);
 		}
 			break;
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
@@ -645,8 +646,8 @@ out_err:
 static boolean
 nv20_vertprog_validate(struct nv20_context *nv20)
 { 
+	struct pipe_screen *pscreen = nv20->pipe.screen;
 	struct nouveau_winsys *nvws = nv20->nvws;
-	struct pipe_winsys *ws = nv20->pipe.winsys;
 	struct nouveau_grobj *rankine = nv20->screen->rankine;
 	struct nv20_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -749,8 +750,8 @@ nv20_vertprog_validate(struct nv20_context *nv20)
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -770,9 +771,8 @@ nv20_vertprog_validate(struct nv20_context *nv20)
 			OUT_RINGp ((uint32_t *)vpd->value, 4);
 		}
 
-		if (constbuf) {
-			ws->buffer_unmap(ws, constbuf);
-		}
+		if (constbuf)
+			pipe_buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c
index bdfe1425d2..a48ba9782b 100644
--- a/src/gallium/drivers/nv30/nv30_fragprog.c
+++ b/src/gallium/drivers/nv30/nv30_fragprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -703,10 +704,10 @@ nv30_fragprog_prepare(struct nv30_fpc *fpc)
 			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
 			assert(fpc->nr_imm < MAX_IMM);
 
-			vals[0] = imm->u.ImmediateFloat32[0].Float;
-			vals[1] = imm->u.ImmediateFloat32[1].Float;
-			vals[2] = imm->u.ImmediateFloat32[2].Float;
-			vals[3] = imm->u.ImmediateFloat32[3].Float;
+			vals[0] = imm->u[0].Float;
+			vals[1] = imm->u[1].Float;
+			vals[2] = imm->u[2].Float;
+			vals[3] = imm->u[3].Float;
 			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
 		}
 			break;
@@ -798,12 +799,12 @@ static void
 nv30_fragprog_upload(struct nv30_context *nv30,
 		     struct nv30_fragment_program *fp)
 {
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	const uint32_t le = 1;
 	uint32_t *map;
 	int i;
 
-	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+	map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
 
 #if 0
 	for (i = 0; i < fp->insn_len; i++) {
@@ -825,7 +826,7 @@ nv30_fragprog_upload(struct nv30_context *nv30,
 		}
 	}
 
-	ws->buffer_unmap(ws, fp->buffer);
+	pipe_buffer_unmap(pscreen, fp->buffer);
 }
 
 static boolean
@@ -834,8 +835,7 @@ nv30_fragprog_validate(struct nv30_context *nv30)
 	struct nv30_fragment_program *fp = nv30->fragprog;
 	struct pipe_buffer *constbuf =
 		nv30->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *screen = nv30->pipe.screen;
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_stateobj *so;
 	boolean new_consts = FALSE;
 	int i;
@@ -850,14 +850,15 @@ nv30_fragprog_validate(struct nv30_context *nv30)
 		return FALSE;
 	}
 
-	fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4);
+	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
 	nv30_fragprog_upload(nv30, fp);
 
 	so = so_new(8, 1);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1);
-	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-		  NV34TCL_FP_ACTIVE_PROGRAM_DMA0, NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
+	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
+		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
+		      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
+		      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1);
 	so_data  (so, fp->fp_control);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1);
@@ -871,7 +872,8 @@ update_constants:
 	if (fp->nr_consts) {
 		float *map;
 		
-		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, constbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		for (i = 0; i < fp->nr_consts; i++) {
 			struct nv30_fragment_program_data *fpd = &fp->consts[i];
 			uint32_t *p = &fp->insn[fpd->offset];
@@ -882,7 +884,7 @@ update_constants:
 			memcpy(p, cb, 4 * sizeof(float));
 			new_consts = TRUE;
 		}
-		ws->buffer_unmap(ws, constbuf);
+		pipe_buffer_unmap(pscreen, constbuf);
 
 		if (new_consts)
 			nv30_fragprog_upload(nv30, fp);
diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c
index 8b6ab992d1..822e1d8def 100644
--- a/src/gallium/drivers/nv30/nv30_fragtex.c
+++ b/src/gallium/drivers/nv30/nv30_fragtex.c
@@ -61,6 +61,7 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 	struct nv30_sampler_state *ps = nv30->tex_sampler[unit];
 	struct nv30_miptree *nv30mt = nv30->tex_miptree[unit];
 	struct pipe_texture *pt = &nv30mt->base;
+	struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer);
 	struct nv30_texture_format *tf;
 	struct nouveau_stateobj *so;
 	uint32_t txf, txs , txp;
@@ -106,9 +107,9 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 
 	so = so_new(16, 2);
 	so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8);
-	so_reloc (so, nv30mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
-	so_reloc (so, nv30mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
-		  NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
+	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
+		      NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
 	so_data  (so, ps->wrap);
 	so_data  (so, NV34TCL_TX_ENABLE_ENABLE | ps->en);
 	so_data  (so, txs);
diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c
index 2f974cf5c4..1d1c8a484e 100644
--- a/src/gallium/drivers/nv30/nv30_query.c
+++ b/src/gallium/drivers/nv30/nv30_query.c
@@ -29,11 +29,10 @@ nv30_query_create(struct pipe_context *pipe, unsigned query_type)
 static void
 nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
 	struct nv30_query *q = nv30_query(pq);
 
 	if (q->object)
-		nv30->nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	FREE(q);
 }
 
@@ -54,9 +53,9 @@ nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 		pipe->get_query_result(pipe, pq, 1, &tmp);
 	}
 
-	if (nv30->nvws->res_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
+	if (nouveau_resource_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
 		assert(0);
-	nv30->nvws->notifier_reset(nv30->screen->query, q->object->start);
+	nouveau_notifier_reset(nv30->screen->query, q->object->start);
 
 	BEGIN_RING(rankine, NV34TCL_QUERY_RESET, 1);
 	OUT_RING  (1);
@@ -84,27 +83,27 @@ nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
 	struct nv30_query *q = nv30_query(pq);
-	struct nouveau_winsys *nvws = nv30->nvws;
 
 	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 
 	if (!q->ready) {
 		unsigned status;
 
-		status = nvws->notifier_status(nv30->screen->query,
-					       q->object->start);
+		status = nouveau_notifier_status(nv30->screen->query,
+						 q->object->start);
 		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
 			if (wait == FALSE)
 				return FALSE;
-			nvws->notifier_wait(nv30->screen->query, q->object->start,
-					    NV_NOTIFY_STATE_STATUS_COMPLETED,
-					    0);
+
+			nouveau_notifier_wait_status(nv30->screen->query,
+					q->object->start,
+					NV_NOTIFY_STATE_STATUS_COMPLETED, 0);
 		}
 
-		q->result = nvws->notifier_retval(nv30->screen->query,
-						  q->object->start);
+		q->result = nouveau_notifier_return_val(nv30->screen->query,
+							q->object->start);
 		q->ready = TRUE;
-		nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c
index d395c5e1b7..f8285e4455 100644
--- a/src/gallium/drivers/nv30/nv30_screen.c
+++ b/src/gallium/drivers/nv30/nv30_screen.c
@@ -1,5 +1,7 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
+#include "pipe/p_state.h"
+
+#include "nouveau/nouveau_screen.h"
 
 #include "nv30_context.h"
 #include "nv30_screen.h"
@@ -8,23 +10,6 @@
 #define NV34TCL_CHIPSET_3X_MASK 0x00000010
 #define NV35TCL_CHIPSET_3X_MASK 0x000001e0
 
-static const char *
-nv30_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv30_screen *screen = nv30_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv30_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -61,6 +46,10 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 1;
@@ -100,7 +89,15 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -139,45 +136,57 @@ static void
 nv30_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv30_screen *screen = nv30_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->res_free(&screen->vp_exec_heap);
-	nvws->res_free(&screen->vp_data_heap);
-	nvws->res_free(&screen->query_heap);
-	nvws->notifier_free(&screen->query);
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->rankine);
+	nouveau_resource_free(&screen->vp_exec_heap);
+	nouveau_resource_free(&screen->vp_data_heap);
+	nouveau_resource_free(&screen->query_heap);
+	nouveau_notifier_free(&screen->query);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->rankine);
 
 	FREE(pscreen);
 }
 
 struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
 	unsigned rankine_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret, i;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv30_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv30_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv30_screen_destroy;
+	pscreen->get_param = nv30_screen_get_param;
+	pscreen->get_paramf = nv30_screen_get_paramf;
+	pscreen->is_format_supported = nv30_screen_surface_format_supported;
+
+	nv30_screen_init_miptree_functions(pscreen);
+	nv30_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	switch (chipset & 0xf0) {
+	switch (dev->chipset & 0xf0) {
 	case 0x30:
-		if (NV30TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV30TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0397;
 		else
-		if (NV34TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV34TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0697;
 		else
-		if (NV35TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV35TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0497;
 		break;
 	default:
@@ -185,43 +194,49 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	}
 
 	if (!rankine_class) {
-		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, rankine_class, &screen->rankine);
+	ret = nouveau_grobj_alloc(chan, 0xbeef3097, rankine_class,
+				  &screen->rankine);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->rankine, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv30_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Query objects */
-	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	ret = nouveau_resource_init(&screen->query_heap, 0, 32);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Vtxprog resources */
-	if (nvws->res_init(&screen->vp_exec_heap, 0, 256) ||
-	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
-		nv30_screen_destroy(&screen->pipe);
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 256) ||
+	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
@@ -230,23 +245,23 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1);
 	so_data  (so, screen->sync->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 /*	so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2);
 	so_data  (so, 0);
 	so_data  (so, screen->query->handle);*/
 	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 
 	for (i=1; i<8; i++) {
 		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1);
@@ -301,23 +316,9 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->rankine, 0x1e94, 1);
 	so_data  (so, 0x13);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
-
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv30_screen_destroy;
-
-	screen->pipe.get_name = nv30_screen_get_name;
-	screen->pipe.get_vendor = nv30_screen_get_vendor;
-	screen->pipe.get_param = nv30_screen_get_param;
-	screen->pipe.get_paramf = nv30_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv30_screen_surface_format_supported;
-
-	nv30_screen_init_miptree_functions(&screen->pipe);
-	nv30_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h
index 8e36883975..5fbd998b53 100644
--- a/src/gallium/drivers/nv30/nv30_screen.h
+++ b/src/gallium/drivers/nv30/nv30_screen.h
@@ -1,11 +1,12 @@
 #ifndef __NV30_SCREEN_H__
 #define __NV30_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
+
 #include "nv04/nv04_surface_2d.h"
 
 struct nv30_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c
index c18be20a32..621b8846c8 100644
--- a/src/gallium/drivers/nv30/nv30_state_emit.c
+++ b/src/gallium/drivers/nv30/nv30_state_emit.c
@@ -38,6 +38,7 @@ nv30_state_do_validate(struct nv30_context *nv30,
 void
 nv30_state_emit(struct nv30_context *nv30)
 {
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	struct nv30_state *state = &nv30->state;
 	struct nv30_screen *screen = nv30->screen;
 	unsigned i, samplers;
@@ -57,23 +58,23 @@ nv30_state_emit(struct nv30_context *nv30)
 			continue;
 		so_ref (state->hw[i], &nv30->screen->state[i]);
 		if (state->hw[i])
-			so_emit(nv30->nvws, nv30->screen->state[i]);
+			so_emit(chan, nv30->screen->state[i]);
 		states &= ~(1ULL << i);
 	}
 
 	state->dirty = 0;
 
-	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FB]);
+	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FB]);
 	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
 		if (!(samplers & (1 << i)))
 			continue;
-		so_emit_reloc_markers(nv30->nvws,
+		so_emit_reloc_markers(chan,
 				      state->hw[NV30_STATE_FRAGTEX0+i]);
 		samplers &= ~(1ULL << i);
 	}
-	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FRAGPROG]);
+	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FRAGPROG]);
 	if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/)
-		so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_VTXBUF]);
+		so_emit_reloc_markers(chan, state->hw[NV30_STATE_VTXBUF]);
 }
 
 boolean
diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c
index fdc1cade90..44b6a74715 100644
--- a/src/gallium/drivers/nv30/nv30_state_fb.c
+++ b/src/gallium/drivers/nv30/nv30_state_fb.c
@@ -5,6 +5,8 @@ static boolean
 nv30_state_framebuffer_validate(struct nv30_context *nv30)
 {
 	struct pipe_framebuffer_state *fb = &nv30->framebuffer;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	struct nv04_surface *rt[2], *zeta = NULL;
 	uint32_t rt_enable, rt_format;
 	int i, colour_format = 0, zeta_format = 0;
@@ -63,6 +65,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30)
 		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
 		break;
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 	case 0:
 		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
 		break;
@@ -79,56 +82,53 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30)
 		}
 
 		nv30mt = (struct nv30_miptree *)rt[0]->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2);
+		so_method(so, rankine, NV34TCL_DMA_COLOR0, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_COLOR0_PITCH, 2);
 		so_data  (so, pitch);
-		so_reloc (so, nv30mt->buffer, rt[0]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), rt[0]->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 	}
 
 	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
 		nv30mt = (struct nv30_miptree *)rt[1]->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nv30mt->buffer, rt[1]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, rankine, NV34TCL_DMA_COLOR1, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), rt[1]->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 		so_data  (so, rt[1]->pitch);
 	}
 
 	if (zeta_format) {
 		nv30mt = (struct nv30_miptree *)zeta->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nv30mt->buffer, zeta->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, rankine, NV34TCL_DMA_ZETA, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), zeta->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 		/* TODO: allocate LMA depth buffer */
 	}
 
-	so_method(so, nv30->screen->rankine, NV34TCL_RT_ENABLE, 1);
+	so_method(so, rankine, NV34TCL_RT_ENABLE, 1);
 	so_data  (so, rt_enable);
-	so_method(so, nv30->screen->rankine, NV34TCL_RT_HORIZ, 3);
+	so_method(so, rankine, NV34TCL_RT_HORIZ, 3);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
 	so_data  (so, rt_format);
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_HORIZ, 2);
+	so_method(so, rankine, NV34TCL_VIEWPORT_HORIZ, 2);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_method(so, rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
 	so_data  (so, ((w - 1) << 16) | 0);
 	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, nv30->screen->rankine, 0x1d88, 1);
+	so_method(so, rankine, 0x1d88, 1);
 	so_data  (so, (1 << 12) | h);
 	/* Wonder why this is needed, context should all be set to zero on init */
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
+	so_method(so, rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
 	so_data  (so, 0);
 
 	so_ref(so, &nv30->state.hw[NV30_STATE_FB]);
diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c
index 990a876382..189656ec81 100644
--- a/src/gallium/drivers/nv30/nv30_vbo.c
+++ b/src/gallium/drivers/nv30/nv30_vbo.c
@@ -1,5 +1,6 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv30_context.h"
 #include "nv30_state.h"
@@ -70,7 +71,7 @@ static boolean
 nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib,
 		    unsigned ib_size)
 {
-	struct pipe_screen *pscreen = &nv30->screen->pipe;
+	struct pipe_screen *pscreen = &nv30->screen->base.base;
 	unsigned type;
 
 	if (!ib) {
@@ -108,7 +109,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 		       int attrib, struct pipe_vertex_element *ve,
 		       struct pipe_vertex_buffer *vb)
 {
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	unsigned type, ncomp;
 	void *map;
@@ -116,7 +117,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 	if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp))
 		return FALSE;
 
-	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
 	map += vb->buffer_offset + ve->src_offset;
 
 	switch (type) {
@@ -148,18 +149,17 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 			so_data  (so, fui(v[0]));
 			break;
 		default:
-			ws->buffer_unmap(ws, vb->buffer);
+			pipe_buffer_unmap(pscreen, vb->buffer);
 			return FALSE;
 		}
 	}
 		break;
 	default:
-		ws->buffer_unmap(ws, vb->buffer);
+		pipe_buffer_unmap(pscreen, vb->buffer);
 		return FALSE;
 	}
 
-	ws->buffer_unmap(ws, vb->buffer);
-
+	pipe_buffer_unmap(pscreen, vb->buffer);
 	return TRUE;
 }
 
@@ -168,7 +168,7 @@ nv30_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	unsigned restart = 0;
 
 	nv30_vbo_set_idxbuf(nv30, NULL, 0);
@@ -228,7 +228,7 @@ static INLINE void
 nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint8_t *elts = (uint8_t *)ib + start;
@@ -277,7 +277,7 @@ static INLINE void
 nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint16_t *elts = (uint16_t *)ib + start;
@@ -326,7 +326,7 @@ static INLINE void
 nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint32_t *elts = (uint32_t *)ib + start;
@@ -368,10 +368,10 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 			  unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	void *map;
 
-	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ);
 	if (!ib) {
 		NOUVEAU_ERR("failed mapping ib\n");
 		return FALSE;
@@ -392,7 +392,7 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 		break;
 	}
 
-	ws->buffer_unmap(ws, ib);
+	pipe_buffer_unmap(pscreen, ib);
 	return TRUE;
 }
 
@@ -401,7 +401,7 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
 		       unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	unsigned restart = 0;
 
 	while (count) {
@@ -521,18 +521,20 @@ nv30_vbo_validate(struct nv30_context *nv30)
 			return FALSE;
 		}
 
-		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
-			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-			 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
+		so_reloc(vtxbuf, nouveau_bo(vb->buffer), vb->buffer_offset +
+				 ve->src_offset, vb_flags | NOUVEAU_BO_LOW |
+				 NOUVEAU_BO_OR, 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
 		so_data (vtxfmt, ((vb->stride << NV34TCL_VTXFMT_STRIDE_SHIFT) |
 				  (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type));
 	}
 
 	if (ib) {
+		struct nouveau_bo *bo = nouveau_bo(ib);
+
 		so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2);
-		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
-			  0, NV34TCL_IDXBUF_FORMAT_DMA1);
+		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
+				  0, NV34TCL_IDXBUF_FORMAT_DMA1);
 	}
 
 	so_method(vtxbuf, rankine, 0x1710, 1);
diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c
index eaf543b8f7..14a5c0260d 100644
--- a/src/gallium/drivers/nv30/nv30_vertprog.c
+++ b/src/gallium/drivers/nv30/nv30_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -616,10 +617,10 @@ nv30_vertprog_translate(struct nv30_context *nv30,
 			assert(imm->Immediate.NrTokens == 4 + 1);
 			vpc->imm[vpc->nr_imm++] =
 				constant(vpc, -1,
-					 imm->u.ImmediateFloat32[0].Float,
-					 imm->u.ImmediateFloat32[1].Float,
-					 imm->u.ImmediateFloat32[2].Float,
-					 imm->u.ImmediateFloat32[3].Float);
+					 imm->u[0].Float,
+					 imm->u[1].Float,
+					 imm->u[2].Float,
+					 imm->u[3].Float);
 		}
 			break;
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
@@ -645,8 +646,7 @@ out_err:
 static boolean
 nv30_vertprog_validate(struct nv30_context *nv30)
 { 
-	struct nouveau_winsys *nvws = nv30->nvws;
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	struct nv30_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -669,15 +669,15 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 		struct nouveau_stateobj *so;
 		uint vplen = vp->nr_insns;
 
-		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
 			while (heap->next && heap->size < vplen) {
 				struct nv30_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->exec);
+				nouveau_resource_free(&evict->exec);
 			}
 
-			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+			if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
 				assert(0);
 		}
 
@@ -694,15 +694,16 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 	if (vp->nr_consts && !vp->data) {
 		struct nouveau_resource *heap = nv30->screen->vp_data_heap;
 
-		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
 			while (heap->next && heap->size < vp->nr_consts) {
 				struct nv30_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data);
 			}
 
-			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+			if (nouveau_resource_alloc(heap, vp->nr_consts, vp,
+						   &vp->data))
 				assert(0);
 		}
 
@@ -750,8 +751,8 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -771,9 +772,8 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 			OUT_RINGp ((uint32_t *)vpd->value, 4);
 		}
 
-		if (constbuf) {
-			ws->buffer_unmap(ws, constbuf);
-		}
+		if (constbuf)
+			pipe_buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
@@ -804,8 +804,6 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 void
 nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
 {
-	struct nouveau_winsys *nvws = nv30->screen->nvws;
-
 	vp->translated = FALSE;
 
 	if (vp->nr_insns) {
@@ -820,9 +818,9 @@ nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
 		vp->nr_consts = 0;
 	}
 
-	nvws->res_free(&vp->exec);
+	nouveau_resource_free(&vp->exec);
 	vp->exec_start = 0;
-	nvws->res_free(&vp->data);
+	nouveau_resource_free(&vp->data);
 	vp->data_start = 0;
 	vp->data_start_min = 0;
 
diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c
index c83ff91d7e..b2f19ecb69 100644
--- a/src/gallium/drivers/nv40/nv40_draw.c
+++ b/src/gallium/drivers/nv40/nv40_draw.c
@@ -1,4 +1,5 @@
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "util/u_pack_color.h"
 
@@ -81,7 +82,7 @@ nv40_render_prim(struct draw_stage *stage, struct prim_header *prim,
 {
 	struct nv40_render_stage *rs = nv40_render_stage(stage);
 	struct nv40_context *nv40 = rs->nv40;
-	struct nouveau_pushbuf *pb = nv40->nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = nv40->screen->base.channel->pushbuf;
 	unsigned i;
 
 	/* Ensure there's room for 4xfloat32 + potentially 3 begin/end */
@@ -231,7 +232,7 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 			 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	unsigned i;
 	void *map;
 
@@ -241,13 +242,14 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 	nv40_state_emit(nv40);
 
 	for (i = 0; i < nv40->vtxbuf_nr; i++) {
-		map = ws->buffer_map(ws, nv40->vtxbuf[i].buffer,
+		map = pipe_buffer_map(pscreen, nv40->vtxbuf[i].buffer,
                                       PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_vertex_buffer(nv40->draw, i, map);
 	}
 
 	if (idxbuf) {
-		map = ws->buffer_map(ws, idxbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, idxbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map);
 	} else {
 		draw_set_mapped_element_buffer(nv40->draw, 0, NULL);
@@ -256,21 +258,22 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 	if (nv40->constbuf[PIPE_SHADER_VERTEX]) {
 		const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX];
 
-		map = ws->buffer_map(ws, nv40->constbuf[PIPE_SHADER_VERTEX],
-				     PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen,
+				      nv40->constbuf[PIPE_SHADER_VERTEX],
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_constant_buffer(nv40->draw, map, nr);
 	}
 
 	draw_arrays(nv40->draw, mode, start, count);
 
 	for (i = 0; i < nv40->vtxbuf_nr; i++)
-		ws->buffer_unmap(ws, nv40->vtxbuf[i].buffer);
+		pipe_buffer_unmap(pscreen, nv40->vtxbuf[i].buffer);
 
 	if (idxbuf)
-		ws->buffer_unmap(ws, idxbuf);
+		pipe_buffer_unmap(pscreen, idxbuf);
 
 	if (nv40->constbuf[PIPE_SHADER_VERTEX])
-		ws->buffer_unmap(ws, nv40->constbuf[PIPE_SHADER_VERTEX]);
+		pipe_buffer_unmap(pscreen, nv40->constbuf[PIPE_SHADER_VERTEX]);
 
 	draw_flush(nv40->draw);
 	pipe->flush(pipe, 0, NULL);
diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c
index 16e40889ec..32d9ed1a7f 100644
--- a/src/gallium/drivers/nv40/nv40_fragprog.c
+++ b/src/gallium/drivers/nv40/nv40_fragprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -789,10 +790,10 @@ nv40_fragprog_prepare(struct nv40_fpc *fpc)
 			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
 			assert(fpc->nr_imm < MAX_IMM);
 
-			vals[0] = imm->u.ImmediateFloat32[0].Float;
-			vals[1] = imm->u.ImmediateFloat32[1].Float;
-			vals[2] = imm->u.ImmediateFloat32[2].Float;
-			vals[3] = imm->u.ImmediateFloat32[3].Float;
+			vals[0] = imm->u[0].Float;
+			vals[1] = imm->u[1].Float;
+			vals[2] = imm->u[2].Float;
+			vals[3] = imm->u[3].Float;
 			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
 		}
 			break;
@@ -881,12 +882,12 @@ static void
 nv40_fragprog_upload(struct nv40_context *nv40,
 		     struct nv40_fragment_program *fp)
 {
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	const uint32_t le = 1;
 	uint32_t *map;
 	int i;
 
-	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+	map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
 
 #if 0
 	for (i = 0; i < fp->insn_len; i++) {
@@ -908,7 +909,7 @@ nv40_fragprog_upload(struct nv40_context *nv40,
 		}
 	}
 
-	ws->buffer_unmap(ws, fp->buffer);
+	pipe_buffer_unmap(pscreen, fp->buffer);
 }
 
 static boolean
@@ -917,8 +918,7 @@ nv40_fragprog_validate(struct nv40_context *nv40)
 	struct nv40_fragment_program *fp = nv40->fragprog;
 	struct pipe_buffer *constbuf =
 		nv40->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *screen = nv40->pipe.screen;
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_stateobj *so;
 	boolean new_consts = FALSE;
 	int i;
@@ -933,14 +933,15 @@ nv40_fragprog_validate(struct nv40_context *nv40)
 		return FALSE;
 	}
 
-	fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4);
+	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
 	nv40_fragprog_upload(nv40, fp);
 
 	so = so_new(4, 1);
 	so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1);
-	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-		  NV40TCL_FP_ADDRESS_DMA0, NV40TCL_FP_ADDRESS_DMA1);
+	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
+		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
+		      NOUVEAU_BO_OR, NV40TCL_FP_ADDRESS_DMA0,
+		      NV40TCL_FP_ADDRESS_DMA1);
 	so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1);
 	so_data  (so, fp->fp_control);
 	so_ref(so, &fp->so);
@@ -950,7 +951,8 @@ update_constants:
 	if (fp->nr_consts) {
 		float *map;
 		
-		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, constbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		for (i = 0; i < fp->nr_consts; i++) {
 			struct nv40_fragment_program_data *fpd = &fp->consts[i];
 			uint32_t *p = &fp->insn[fpd->offset];
@@ -961,7 +963,7 @@ update_constants:
 			memcpy(p, cb, 4 * sizeof(float));
 			new_consts = TRUE;
 		}
-		ws->buffer_unmap(ws, constbuf);
+		pipe_buffer_unmap(pscreen, constbuf);
 
 		if (new_consts)
 			nv40_fragprog_upload(nv40, fp);
diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c
index eb3002dc05..f6cdf31dfe 100644
--- a/src/gallium/drivers/nv40/nv40_fragtex.c
+++ b/src/gallium/drivers/nv40/nv40_fragtex.c
@@ -62,6 +62,7 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 {
 	struct nv40_sampler_state *ps = nv40->tex_sampler[unit];
 	struct nv40_miptree *nv40mt = nv40->tex_miptree[unit];
+	struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer);
 	struct pipe_texture *pt = &nv40mt->base;
 	struct nv40_texture_format *tf;
 	struct nouveau_stateobj *so;
@@ -108,9 +109,9 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 
 	so = so_new(16, 2);
 	so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8);
-	so_reloc (so, nv40mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
-	so_reloc (so, nv40mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
-		  NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
+	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
+		      NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
 	so_data  (so, ps->wrap);
 	so_data  (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en);
 	so_data  (so, txs);
diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c
index 9b9a43f49d..7874aedd42 100644
--- a/src/gallium/drivers/nv40/nv40_query.c
+++ b/src/gallium/drivers/nv40/nv40_query.c
@@ -29,11 +29,10 @@ nv40_query_create(struct pipe_context *pipe, unsigned query_type)
 static void
 nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 {
-	struct nv40_context *nv40 = nv40_context(pipe);
 	struct nv40_query *q = nv40_query(pq);
 
 	if (q->object)
-		nv40->nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	FREE(q);
 }
 
@@ -54,9 +53,9 @@ nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 		pipe->get_query_result(pipe, pq, 1, &tmp);
 	}
 
-	if (nv40->nvws->res_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
+	if (nouveau_resource_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
 		assert(0);
-	nv40->nvws->notifier_reset(nv40->screen->query, q->object->start);
+	nouveau_notifier_reset(nv40->screen->query, q->object->start);
 
 	BEGIN_RING(curie, NV40TCL_QUERY_RESET, 1);
 	OUT_RING  (1);
@@ -84,27 +83,27 @@ nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
 	struct nv40_query *q = nv40_query(pq);
-	struct nouveau_winsys *nvws = nv40->nvws;
 
 	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 
 	if (!q->ready) {
 		unsigned status;
 
-		status = nvws->notifier_status(nv40->screen->query,
-					       q->object->start);
+		status = nouveau_notifier_status(nv40->screen->query,
+						 q->object->start);
 		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
 			if (wait == FALSE)
 				return FALSE;
-			nvws->notifier_wait(nv40->screen->query, q->object->start,
-					    NV_NOTIFY_STATE_STATUS_COMPLETED,
-					    0);
+			nouveau_notifier_wait_status(nv40->screen->query,
+					      q->object->start,
+					      NV_NOTIFY_STATE_STATUS_COMPLETED,
+					      0);
 		}
 
-		q->result = nvws->notifier_retval(nv40->screen->query,
-						  q->object->start);
+		q->result = nouveau_notifier_return_val(nv40->screen->query,
+							q->object->start);
 		q->ready = TRUE;
-		nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c
index 0d4baefaea..5d2a4216c5 100644
--- a/src/gallium/drivers/nv40/nv40_screen.c
+++ b/src/gallium/drivers/nv40/nv40_screen.c
@@ -1,5 +1,4 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv40_context.h"
 #include "nv40_screen.h"
@@ -8,23 +7,6 @@
 #define NV4X_GRCLASS4497_CHIPSETS 0x00005450
 #define NV6X_GRCLASS4497_CHIPSETS 0x00000088
 
-static const char *
-nv40_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv40_screen *screen = nv40_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv40_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv40_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -62,6 +44,10 @@ nv40_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0; /* We have 4 - but unsupported currently */
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 1;
 	case NOUVEAU_CAP_HW_VTXBUF:
 		return 1;
 	case NOUVEAU_CAP_HW_IDXBUF:
@@ -104,7 +90,15 @@ nv40_screen_surface_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -148,88 +142,107 @@ static void
 nv40_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv40_screen *screen = nv40_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->res_free(&screen->vp_exec_heap);
-	nvws->res_free(&screen->vp_data_heap);
-	nvws->res_free(&screen->query_heap);
-	nvws->notifier_free(&screen->query);
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->curie);
+	nouveau_resource_free(&screen->vp_exec_heap);
+	nouveau_resource_free(&screen->vp_data_heap);
+	nouveau_resource_free(&screen->query_heap);
+	nouveau_notifier_free(&screen->query);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->curie);
+
+	nouveau_screen_fini(&screen->base);
 
 	FREE(pscreen);
 }
 
 struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
 	unsigned curie_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv40_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv40_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv40_screen_destroy;
+	pscreen->get_param = nv40_screen_get_param;
+	pscreen->get_paramf = nv40_screen_get_paramf;
+	pscreen->is_format_supported = nv40_screen_surface_format_supported;
+
+	nv40_screen_init_miptree_functions(pscreen);
+	nv40_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	switch (chipset & 0xf0) {
+	switch (dev->chipset & 0xf0) {
 	case 0x40:
-		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV40TCL;
 		else
-		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV44TCL;
 		break;
 	case 0x60:
-		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV44TCL;
 		break;
 	}
 
 	if (!curie_class) {
-		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, curie_class, &screen->curie);
+	ret = nouveau_grobj_alloc(chan, 0xbeef3097, curie_class, &screen->curie);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->curie, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv40_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Query objects */
-	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	nouveau_resource_init(&screen->query_heap, 0, 32);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Vtxprog resources */
-	if (nvws->res_init(&screen->vp_exec_heap, 0, 512) ||
-	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
-		nv40_screen_destroy(&screen->pipe);
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 512) ||
+	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
@@ -238,25 +251,25 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1);
 	so_data  (so, screen->sync->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2);
 	so_data  (so, 0);
 	so_data  (so, screen->query->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 
 	so_method(so, screen->curie, 0x1ea4, 3);
 	so_data  (so, 0x00000010);
@@ -281,24 +294,10 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->curie, 0x1e94, 1);
 	so_data  (so, 0x00000001);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
-
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv40_screen_destroy;
-
-	screen->pipe.get_name = nv40_screen_get_name;
-	screen->pipe.get_vendor = nv40_screen_get_vendor;
-	screen->pipe.get_param = nv40_screen_get_param;
-	screen->pipe.get_paramf = nv40_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv40_screen_surface_format_supported;
-
-	nv40_screen_init_miptree_functions(&screen->pipe);
-	nv40_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h
index 7b503bd207..57b4c8fc46 100644
--- a/src/gallium/drivers/nv40/nv40_screen.h
+++ b/src/gallium/drivers/nv40/nv40_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV40_SCREEN_H__
 #define __NV40_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv40_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c
index 10aae29832..198692965d 100644
--- a/src/gallium/drivers/nv40/nv40_state_emit.c
+++ b/src/gallium/drivers/nv40/nv40_state_emit.c
@@ -54,6 +54,7 @@ nv40_state_do_validate(struct nv40_context *nv40,
 void
 nv40_state_emit(struct nv40_context *nv40)
 {
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	struct nv40_state *state = &nv40->state;
 	struct nv40_screen *screen = nv40->screen;
 	unsigned i, samplers;
@@ -73,7 +74,7 @@ nv40_state_emit(struct nv40_context *nv40)
 			continue;
 		so_ref (state->hw[i], &nv40->screen->state[i]);
 		if (state->hw[i])
-			so_emit(nv40->nvws, nv40->screen->state[i]);
+			so_emit(chan, nv40->screen->state[i]);
 		states &= ~(1ULL << i);
 	}
 
@@ -87,17 +88,17 @@ nv40_state_emit(struct nv40_context *nv40)
 
 	state->dirty = 0;
 
-	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FB]);
+	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FB]);
 	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
 		if (!(samplers & (1 << i)))
 			continue;
-		so_emit_reloc_markers(nv40->nvws,
+		so_emit_reloc_markers(chan,
 				      state->hw[NV40_STATE_FRAGTEX0+i]);
 		samplers &= ~(1ULL << i);
 	}
-	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FRAGPROG]);
+	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FRAGPROG]);
 	if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW)
-		so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_VTXBUF]);
+		so_emit_reloc_markers(chan, state->hw[NV40_STATE_VTXBUF]);
 }
 
 boolean
diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c
index be618a306b..c2f739157a 100644
--- a/src/gallium/drivers/nv40/nv40_state_fb.c
+++ b/src/gallium/drivers/nv40/nv40_state_fb.c
@@ -2,15 +2,19 @@
 #include "nouveau/nouveau_util.h"
 
 static struct pipe_buffer *
-nv40_surface_buffer(struct pipe_surface *surface)
+nv40_do_surface_buffer(struct pipe_surface *surface)
 {
 	struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
 	return mt->buffer;
 }
 
+#define nv40_surface_buffer(ps) nouveau_bo(nv40_do_surface_buffer(ps))
+
 static boolean
 nv40_state_framebuffer_validate(struct nv40_context *nv40)
 {
+	struct nouveau_channel *chan = nv40->screen->base.channel;
+	struct nouveau_grobj *curie = nv40->screen->curie;
 	struct pipe_framebuffer_state *fb = &nv40->framebuffer;
 	struct nv04_surface *rt[4], *zeta;
 	uint32_t rt_enable, rt_format;
@@ -69,6 +73,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40)
 		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16;
 		break;
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 	case 0:
 		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8;
 		break;
@@ -77,76 +82,80 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40)
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2);
+		so_method(so, curie, NV40TCL_DMA_COLOR0, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR0_PITCH, 2);
 		so_data  (so, rt[0]->pitch);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base), rt[0]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (so, nv40_surface_buffer(&rt[0]->base),
+			      rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base), rt[1]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, curie, NV40TCL_DMA_COLOR1, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nv40_surface_buffer(&rt[1]->base),
+			      rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
 		so_data  (so, rt[1]->pitch);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base), rt[2]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_COLOR2, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR2_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[2]->base),
+			      rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+		so_method(so, curie, NV40TCL_COLOR2_PITCH, 1);
 		so_data  (so, rt[2]->pitch);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base), rt[3]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_COLOR3, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR3_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[3]->base),
+			      rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+		so_method(so, curie, NV40TCL_COLOR3_PITCH, 1);
 		so_data  (so, rt[3]->pitch);
 	}
 
 	if (zeta_format) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base), zeta->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_ZETA, 1);
+		so_reloc (so, nv40_surface_buffer(&zeta->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&zeta->base),
+			      zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, curie, NV40TCL_ZETA_PITCH, 1);
 		so_data  (so, zeta->pitch);
 	}
 
-	so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1);
+	so_method(so, curie, NV40TCL_RT_ENABLE, 1);
 	so_data  (so, rt_enable);
-	so_method(so, nv40->screen->curie, NV40TCL_RT_HORIZ, 3);
+	so_method(so, curie, NV40TCL_RT_HORIZ, 3);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
 	so_data  (so, rt_format);
-	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_HORIZ, 2);
+	so_method(so, curie, NV40TCL_VIEWPORT_HORIZ, 2);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
-	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_method(so, curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
 	so_data  (so, ((w - 1) << 16) | 0);
 	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, nv40->screen->curie, 0x1d88, 1);
+	so_method(so, curie, 0x1d88, 1);
 	so_data  (so, (1 << 12) | h);
 
 	so_ref(so, &nv40->state.hw[NV40_STATE_FB]);
diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c
index 1a849da32d..a596547974 100644
--- a/src/gallium/drivers/nv40/nv40_surface.c
+++ b/src/gallium/drivers/nv40/nv40_surface.c
@@ -26,12 +26,13 @@
  * 
  **************************************************************************/
 
-#include "nv40_context.h"
 #include "pipe/p_defines.h"
-#include "pipe/internal/p_winsys_screen.h"
 #include "pipe/p_inlines.h"
+
 #include "util/u_tile.h"
 
+#include "nv40_context.h"
+
 static void
 nv40_surface_copy(struct pipe_context *pipe,
 		  struct pipe_surface *dest, unsigned destx, unsigned desty,
diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c
index ce45055fe8..6d92ac3db9 100644
--- a/src/gallium/drivers/nv40/nv40_transfer.c
+++ b/src/gallium/drivers/nv40/nv40_transfer.c
@@ -136,7 +136,7 @@ nv40_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 }
 
 static void
-nv40_transfer_del(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
+nv40_transfer_del(struct pipe_transfer *ptx)
 {
 	struct nv40_transfer *tx = (struct nv40_transfer *)ptx;
 
diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c
index f3518b2e4f..b2753b8e2e 100644
--- a/src/gallium/drivers/nv40/nv40_vbo.c
+++ b/src/gallium/drivers/nv40/nv40_vbo.c
@@ -1,5 +1,6 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv40_context.h"
 #include "nv40_state.h"
@@ -70,7 +71,7 @@ static boolean
 nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib,
 		    unsigned ib_size)
 {
-	struct pipe_screen *pscreen = &nv40->screen->pipe;
+	struct pipe_screen *pscreen = &nv40->screen->base.base;
 	unsigned type;
 
 	if (!ib) {
@@ -108,7 +109,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 		       int attrib, struct pipe_vertex_element *ve,
 		       struct pipe_vertex_buffer *vb)
 {
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_grobj *curie = nv40->screen->curie;
 	unsigned type, ncomp;
 	void *map;
@@ -116,7 +117,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 	if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp))
 		return FALSE;
 
-	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
 	map += vb->buffer_offset + ve->src_offset;
 
 	switch (type) {
@@ -148,17 +149,17 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 			so_data  (so, fui(v[0]));
 			break;
 		default:
-			ws->buffer_unmap(ws, vb->buffer);
+			pipe_buffer_unmap(pscreen, vb->buffer);
 			return FALSE;
 		}
 	}
 		break;
 	default:
-		ws->buffer_unmap(ws, vb->buffer);
+		pipe_buffer_unmap(pscreen, vb->buffer);
 		return FALSE;
 	}
 
-	ws->buffer_unmap(ws, vb->buffer);
+	pipe_buffer_unmap(pscreen, vb->buffer);
 
 	return TRUE;
 }
@@ -168,7 +169,7 @@ nv40_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	unsigned restart;
 
 	nv40_vbo_set_idxbuf(nv40, NULL, 0);
@@ -227,7 +228,7 @@ static INLINE void
 nv40_draw_elements_u08(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint8_t *elts = (uint8_t *)ib + start;
@@ -276,7 +277,7 @@ static INLINE void
 nv40_draw_elements_u16(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint16_t *elts = (uint16_t *)ib + start;
@@ -325,7 +326,7 @@ static INLINE void
 nv40_draw_elements_u32(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint32_t *elts = (uint32_t *)ib + start;
@@ -367,10 +368,10 @@ nv40_draw_elements_inline(struct pipe_context *pipe,
 			  unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	void *map;
 
-	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ);
 	if (!ib) {
 		NOUVEAU_ERR("failed mapping ib\n");
 		return FALSE;
@@ -391,7 +392,7 @@ nv40_draw_elements_inline(struct pipe_context *pipe,
 		break;
 	}
 
-	ws->buffer_unmap(ws, ib);
+	pipe_buffer_unmap(pscreen, ib);
 	return TRUE;
 }
 
@@ -400,7 +401,7 @@ nv40_draw_elements_vbo(struct pipe_context *pipe,
 		       unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	unsigned restart;
 
 	while (count) {
@@ -519,17 +520,20 @@ nv40_vbo_validate(struct nv40_context *nv40)
 			return FALSE;
 		}
 
-		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
-			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-			 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
+		so_reloc(vtxbuf, nouveau_bo(vb->buffer),
+				 vb->buffer_offset + ve->src_offset,
+				 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+				 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
 		so_data (vtxfmt, ((vb->stride << NV40TCL_VTXFMT_STRIDE_SHIFT) |
 				  (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type));
 	}
 
 	if (ib) {
+		struct nouveau_bo *bo = nouveau_bo(ib);
+
 		so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2);
-		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
+		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
 			  0, NV40TCL_IDXBUF_FORMAT_DMA1);
 	}
 
diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c
index 7df9a4d326..0382dbba8f 100644
--- a/src/gallium/drivers/nv40/nv40_vertprog.c
+++ b/src/gallium/drivers/nv40/nv40_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -787,10 +788,10 @@ nv40_vertprog_translate(struct nv40_context *nv40,
 			assert(imm->Immediate.NrTokens == 4 + 1);
 			vpc->imm[vpc->nr_imm++] =
 				constant(vpc, -1,
-					 imm->u.ImmediateFloat32[0].Float,
-					 imm->u.ImmediateFloat32[1].Float,
-					 imm->u.ImmediateFloat32[2].Float,
-					 imm->u.ImmediateFloat32[3].Float);
+					 imm->u[0].Float,
+					 imm->u[1].Float,
+					 imm->u[2].Float,
+					 imm->u[3].Float);
 		}
 			break;
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
@@ -855,8 +856,7 @@ out_err:
 static boolean
 nv40_vertprog_validate(struct nv40_context *nv40)
 { 
-	struct nouveau_winsys *nvws = nv40->nvws;
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_grobj *curie = nv40->screen->curie;
 	struct nv40_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -895,15 +895,15 @@ check_gpu_resources:
 		struct nouveau_stateobj *so;
 		uint vplen = vp->nr_insns;
 
-		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
 			while (heap->next && heap->size < vplen) {
 				struct nv40_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->exec);
+				nouveau_resource_free(&evict->exec);
 			}
 
-			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+			if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
 				assert(0);
 		}
 
@@ -925,15 +925,15 @@ check_gpu_resources:
 	if (vp->nr_consts && !vp->data) {
 		struct nouveau_resource *heap = nv40->screen->vp_data_heap;
 
-		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
 			while (heap->next && heap->size < vp->nr_consts) {
 				struct nv40_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data);
 			}
 
-			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+			if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data))
 				assert(0);
 		}
 
@@ -981,8 +981,8 @@ check_gpu_resources:
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -1003,7 +1003,7 @@ check_gpu_resources:
 		}
 
 		if (constbuf)
-			ws->buffer_unmap(ws, constbuf);
+			pscreen->buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
@@ -1035,8 +1035,6 @@ check_gpu_resources:
 void
 nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
 {
-	struct nouveau_winsys *nvws = nv40->screen->nvws;
-
 	vp->translated = FALSE;
 
 	if (vp->nr_insns) {
@@ -1051,9 +1049,9 @@ nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
 		vp->nr_consts = 0;
 	}
 
-	nvws->res_free(&vp->exec);
+	nouveau_resource_free(&vp->exec);
 	vp->exec_start = 0;
-	nvws->res_free(&vp->data);
+	nouveau_resource_free(&vp->data);
 	vp->data_start = 0;
 	vp->data_start_min = 0;
 
diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c
index 33427a15a5..e0b2d2880b 100644
--- a/src/gallium/drivers/nv50/nv50_clear.c
+++ b/src/gallium/drivers/nv50/nv50_clear.c
@@ -31,7 +31,7 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers,
 	   const float *rgba, double depth, unsigned stencil)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct pipe_framebuffer_state *fb = &nv50->framebuffer;
 	unsigned mode = 0, i;
diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c
index a511f655c1..6e8f4f9750 100644
--- a/src/gallium/drivers/nv50/nv50_context.c
+++ b/src/gallium/drivers/nv50/nv50_context.c
@@ -31,15 +31,23 @@ static void
 nv50_flush(struct pipe_context *pipe, unsigned flags,
 	   struct pipe_fence_handle **fence)
 {
-	struct nv50_context *nv50 = (struct nv50_context *)pipe;
-	
-	FIRE_RING(nv50->screen->nvws->channel);
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->base.channel;
+	struct nouveau_grobj *eng2d = nv50->screen->eng2d;
+
+	/* We need this in the ddx for reliable composite, not sure what we're
+	 * actually flushing. We generate all our own flushes with flags = 0. */
+	WAIT_RING(chan, 3);
+	BEGIN_RING(chan, eng2d, 0x0110, 1);
+	OUT_RING  (chan, 0);
+
+	FIRE_RING(chan);
 }
 
 static void
 nv50_destroy(struct pipe_context *pipe)
 {
-	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	struct nv50_context *nv50 = nv50_context(pipe);
 
 	draw_destroy(nv50->draw);
 	FREE(nv50);
@@ -112,5 +120,3 @@ nv50_create(struct pipe_screen *pscreen, unsigned pctx_id)
 
 	return &nv50->pipe;
 }
-
-		
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index 7b67a75439..1e9e8e49bf 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -63,14 +63,19 @@ struct nv50_rasterizer_stateobj {
 	struct nouveau_stateobj *so;
 };
 
+struct nv50_sampler_stateobj {
+	bool normalized;
+	unsigned tsc[8];
+};
+
 struct nv50_miptree_level {
 	int *image_offset;
 	unsigned pitch;
+	unsigned tile_mode;
 };
 
 struct nv50_miptree {
-	struct pipe_texture base;
-	struct pipe_buffer *buffer;
+	struct nouveau_miptree base;
 
 	struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS];
 	int image_nr;
@@ -93,13 +98,6 @@ nv50_surface(struct pipe_surface *pt)
 	return (struct nv50_surface *)pt;
 }
 
-static INLINE struct pipe_buffer *
-nv50_surface_buffer(struct pipe_surface *surface)
-{
-	struct nv50_miptree *mt = (struct nv50_miptree *)surface->texture;
-	return mt->buffer;
-}
-
 struct nv50_state {
 	unsigned dirty;
 
@@ -115,10 +113,12 @@ struct nv50_state {
 	unsigned viewport_bypass;
 	struct nouveau_stateobj *tsc_upload;
 	struct nouveau_stateobj *tic_upload;
+	unsigned miptree_nr;
 	struct nouveau_stateobj *vertprog;
 	struct nouveau_stateobj *fragprog;
 	struct nouveau_stateobj *vtxfmt;
 	struct nouveau_stateobj *vtxbuf;
+	struct nouveau_stateobj *vtxattr;
 };
 
 struct nv50_context {
@@ -147,7 +147,7 @@ struct nv50_context {
 	unsigned vtxbuf_nr;
 	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
 	unsigned vtxelt_nr;
-	unsigned *sampler[PIPE_MAX_SAMPLERS];
+	struct nv50_sampler_stateobj *sampler[PIPE_MAX_SAMPLERS];
 	unsigned sampler_nr;
 	struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS];
 	unsigned miptree_nr;
diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c
index f79a7ca86c..03b9243b82 100644
--- a/src/gallium/drivers/nv50/nv50_miptree.c
+++ b/src/gallium/drivers/nv50/nv50_miptree.c
@@ -29,26 +29,41 @@
 static struct pipe_texture *
 nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 {
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
 	struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
-	struct pipe_texture *pt = &mt->base;
-	unsigned usage, width = tmp->width[0], height = tmp->height[0];
+	struct pipe_texture *pt = &mt->base.base;
+	unsigned width = tmp->width[0], height = tmp->height[0];
 	unsigned depth = tmp->depth[0];
-	int i, l;
+	uint32_t tile_mode, tile_flags, tile_h;
+	int ret, i, l;
 
-	mt->base = *tmp;
-	pipe_reference_init(&mt->base.reference, 1);
-	mt->base.screen = pscreen;
+	*pt = *tmp;
+	pipe_reference_init(&pt->reference, 1);
+	pt->screen = pscreen;
 
-	usage = PIPE_BUFFER_USAGE_PIXEL;
 	switch (pt->format) {
+	case PIPE_FORMAT_Z32_FLOAT:
+		tile_flags = 0x4800;
+		break;
 	case PIPE_FORMAT_Z24S8_UNORM:
-	case PIPE_FORMAT_Z16_UNORM:
-		usage |= NOUVEAU_BUFFER_USAGE_ZETA;
+		tile_flags = 0x1800;
+		break;
+	case PIPE_FORMAT_X8Z24_UNORM:
+	case PIPE_FORMAT_S8Z24_UNORM:
+		tile_flags = 0x2800;
 		break;
 	default:
+		tile_flags = 0x7000;
 		break;
 	}
 
+	if      (pt->height[0] > 32) tile_mode = 4;
+	else if (pt->height[0] > 16) tile_mode = 3;
+	else if (pt->height[0] >  8) tile_mode = 2;
+	else if (pt->height[0] >  4) tile_mode = 1;
+	else                         tile_mode = 0;
+	tile_h = 1 << (tile_mode + 2);
+
 	switch (pt->target) {
 	case PIPE_TEXTURE_3D:
 		mt->image_nr = pt->depth[0];
@@ -72,20 +87,27 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 
 		lvl->image_offset = CALLOC(mt->image_nr, sizeof(int));
 		lvl->pitch = align(pt->width[l] * pt->block.size, 64);
+		lvl->tile_mode = tile_mode;
 
 		width = MAX2(1, width >> 1);
 		height = MAX2(1, height >> 1);
 		depth = MAX2(1, depth >> 1);
+
+		if (tile_mode && height <= (tile_h >> 1)) {
+			tile_mode--;
+			tile_h >>= 1;
+		}
 	}
 
 	for (i = 0; i < mt->image_nr; i++) {
 		for (l = 0; l <= pt->last_level; l++) {
 			struct nv50_miptree_level *lvl = &mt->level[l];
 			int size;
+			tile_h = 1 << (lvl->tile_mode + 2);
 
 			size  = align(pt->width[l], 8) * pt->block.size;
 			size  = align(size, 64);
-			size *= align(pt->height[l], 8) * pt->block.size;
+			size *= align(pt->height[l], tile_h);
 
 			lvl->image_offset[i] = mt->total_size;
 
@@ -93,19 +115,22 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 		}
 	}
 
-	mt->buffer = pscreen->buffer_create(pscreen, 256, usage, mt->total_size);
-	if (!mt->buffer) {
+	ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 256, mt->total_size,
+				  mt->level[0].tile_mode, tile_flags,
+				  &mt->base.bo);
+	if (ret) {
 		FREE(mt);
 		return NULL;
 	}
 
-	return &mt->base;
+	return pt;
 }
 
 static struct pipe_texture *
 nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 		     const unsigned *stride, struct pipe_buffer *pb)
 {
+	struct nouveau_bo *bo = nouveau_bo(pb);
 	struct nv50_miptree *mt;
 
 	/* Only supports 2D, non-mipmapped textures for the moment */
@@ -117,15 +142,15 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 	if (!mt)
 		return NULL;
 
-	mt->base = *pt;
-	pipe_reference_init(&mt->base.reference, 1);
-	mt->base.screen = pscreen;
+	mt->base.base = *pt;
+	pipe_reference_init(&mt->base.base.reference, 1);
+	mt->base.base.screen = pscreen;
 	mt->image_nr = 1;
 	mt->level[0].pitch = *stride;
 	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
 
-	pipe_buffer_reference(&mt->buffer, pb);
-	return &mt->base;
+	nouveau_bo_ref(bo, &mt->base.bo);
+	return &mt->base.base;
 }
 
 static void
@@ -133,8 +158,8 @@ nv50_miptree_destroy(struct pipe_texture *pt)
 {
 	struct nv50_miptree *mt = nv50_miptree(pt);
 
-        pipe_buffer_reference(&mt->buffer, NULL);
-        FREE(mt);
+	nouveau_bo_ref(NULL, &mt->base.bo);
+	FREE(mt);
 }
 
 static struct pipe_surface *
@@ -177,8 +202,8 @@ nv50_miptree_surface_del(struct pipe_surface *ps)
 {
 	struct nv50_surface *s = nv50_surface(ps);
 
-        pipe_texture_reference(&ps->texture, NULL);
-        FREE(s);
+	pipe_texture_reference(&ps->texture, NULL);
+	FREE(s);
 }
 
 void
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index 2d15868ae8..4a838529de 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -85,6 +85,9 @@ struct nv50_reg {
 
 	int hw;
 	int neg;
+
+	int rhw; /* result hw for FP outputs, or interpolant index */
+	int acc; /* instruction where this reg is last read (first insn == 1) */
 };
 
 struct nv50_pc {
@@ -108,12 +111,23 @@ struct nv50_pc {
 
 	struct nv50_reg *temp_temp[16];
 	unsigned temp_temp_nr;
+
+	unsigned interp_mode[32];
+	/* perspective interpolation registers */
+	struct nv50_reg *iv_p;
+	struct nv50_reg *iv_c;
+
+	/* current instruction and total number of insns */
+	unsigned insn_cur;
+	unsigned insn_nr;
+
+	boolean allow32;
 };
 
 static void
 alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg)
 {
-	int i;
+	int i = 0;
 
 	if (reg->type == P_RESULT) {
 		if (pc->p->cfg.high_result < (reg->hw + 1))
@@ -131,7 +145,22 @@ alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg)
 		return;
 	}
 
-	for (i = 0; i < NV50_SU_MAX_TEMP; i++) {
+	if (reg->rhw != -1) {
+		/* try to allocate temporary with index rhw first */
+		if (!(pc->r_temp[reg->rhw])) {
+			pc->r_temp[reg->rhw] = reg;
+			reg->hw = reg->rhw;
+			if (pc->p->cfg.high_temp < (reg->rhw + 1))
+				pc->p->cfg.high_temp = reg->rhw + 1;
+			return;
+		}
+		/* make sure we don't get things like $r0 needs to go
+		 * in $r1 and $r1 in $r0
+		 */
+		i = pc->result_nr * 4;
+	}
+
+	for (; i < NV50_SU_MAX_TEMP; i++) {
 		if (!(pc->r_temp[i])) {
 			pc->r_temp[i] = reg;
 			reg->hw = i;
@@ -159,6 +188,7 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst)
 			r->type = P_TEMP;
 			r->index = -1;
 			r->hw = i;
+			r->rhw = -1;
 			pc->r_temp[i] = r;
 			return r;
 		}
@@ -168,6 +198,38 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst)
 	return NULL;
 }
 
+/* Assign the hw of the discarded temporary register src
+ * to the tgsi register dst and free src.
+ */
+static void
+assimilate_temp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	assert(src->index == -1 && src->hw != -1);
+
+	if (dst->hw != -1)
+		pc->r_temp[dst->hw] = NULL;
+	pc->r_temp[src->hw] = dst;
+	dst->hw = src->hw;
+
+	FREE(src);
+}
+
+/* release the hardware resource held by r */
+static void
+release_hw(struct nv50_pc *pc, struct nv50_reg *r)
+{
+	assert(r->type == P_TEMP);
+	if (r->hw == -1)
+		return;
+
+	assert(pc->r_temp[r->hw] == r);
+	pc->r_temp[r->hw] = NULL;
+
+	r->acc = 0;
+	if (r->index == -1)
+		FREE(r);
+}
+
 static void
 free_temp(struct nv50_pc *pc, struct nv50_reg *r)
 {
@@ -189,7 +251,7 @@ alloc_temp4(struct nv50_pc *pc, struct nv50_reg *dst[4], int idx)
 
 	if (pc->r_temp[idx] || pc->r_temp[idx + 1] ||
 	    pc->r_temp[idx + 2] || pc->r_temp[idx + 3])
-		return alloc_temp4(pc, dst, idx + 1);
+		return alloc_temp4(pc, dst, idx + 4);
 
 	for (i = 0; i < 4; i++) {
 		dst[i] = CALLOC_STRUCT(nv50_reg);
@@ -234,7 +296,7 @@ kill_temp_temp(struct nv50_pc *pc)
 static int
 ctor_immd(struct nv50_pc *pc, float x, float y, float z, float w)
 {
-	pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * r * sizeof(float)),
+	pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * 4 * sizeof(float)),
 			       (pc->immd_nr + 1) * 4 * sizeof(float));
 	pc->immd_buf[(pc->immd_nr * 4) + 0] = x;
 	pc->immd_buf[(pc->immd_nr * 4) + 1] = y;
@@ -250,7 +312,13 @@ alloc_immd(struct nv50_pc *pc, float f)
 	struct nv50_reg *r = CALLOC_STRUCT(nv50_reg);
 	unsigned hw;
 
-	hw = ctor_immd(pc, f, 0, 0, 0) * 4;
+	for (hw = 0; hw < pc->immd_nr * 4; hw++)
+		if (pc->immd_buf[hw] == f)
+			break;
+
+	if (hw == pc->immd_nr * 4)
+		hw = ctor_immd(pc, f, -f, 0.5 * f, 0) * 4;
+
 	r->type = P_IMMD;
 	r->hw = hw;
 	r->index = -1;
@@ -341,7 +409,8 @@ set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e)
 static INLINE void
 set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e)
 {
-	unsigned val = fui(pc->immd_buf[imm->hw]); /* XXX */
+	float f = pc->immd_buf[imm->hw];
+	unsigned val = fui(imm->neg ? -f : f);
 
 	set_long(pc, e);
 	/*XXX: can't be predicated - bits overlap.. catch cases where both
@@ -354,20 +423,35 @@ set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e)
 	e->inst[1] |= (val >> 6) << 2;
 }
 
+
+#define INTERP_LINEAR		0
+#define INTERP_FLAT			1
+#define INTERP_PERSPECTIVE	2
+#define INTERP_CENTROID		4
+
+/* interpolant index has been stored in dst->rhw */
 static void
-emit_interp(struct nv50_pc *pc, struct nv50_reg *dst,
-	    struct nv50_reg *src, struct nv50_reg *iv)
+emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *iv,
+		unsigned mode)
 {
+	assert(dst->rhw != -1);
 	struct nv50_program_exec *e = exec(pc);
 
 	e->inst[0] |= 0x80000000;
 	set_dst(pc, dst, e);
-	alloc_reg(pc, src);
-	e->inst[0] |= (src->hw << 16);
-	if (iv) {
-		e->inst[0] |= (1 << 25);
-		alloc_reg(pc, iv);
-		e->inst[0] |= (iv->hw << 9);
+	e->inst[0] |= (dst->rhw << 16);
+
+	if (mode & INTERP_FLAT) {
+		e->inst[0] |= (1 << 8);
+	} else {
+		if (mode & INTERP_PERSPECTIVE) {
+			e->inst[0] |= (1 << 25);
+			alloc_reg(pc, iv);
+			e->inst[0] |= (iv->hw << 9);
+		}
+
+		if (mode & INTERP_CENTROID)
+			e->inst[0] |= (1 << 24);
 	}
 
 	emit(pc, e);
@@ -378,22 +462,12 @@ set_data(struct nv50_pc *pc, struct nv50_reg *src, unsigned m, unsigned s,
 	 struct nv50_program_exec *e)
 {
 	set_long(pc, e);
-#if 1
-	e->inst[1] |= (1 << 22);
-#else
-	if (src->type == P_IMMD) {
-		e->inst[1] |= (NV50_CB_PMISC << 22);
-	} else {
-		if (pc->p->type == PIPE_SHADER_VERTEX)
-			e->inst[1] |= (NV50_CB_PVP << 22);
-		else
-			e->inst[1] |= (NV50_CB_PFP << 22);
-	}
-#endif
 
 	e->param.index = src->hw;
 	e->param.shift = s;
 	e->param.mask = m << (s % 32);
+
+	e->inst[1] |= (((src->type == P_IMMD) ? 0 : 1) << 22);
 }
 
 static void
@@ -405,12 +479,11 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 
 	set_dst(pc, dst, e);
 
-	if (0 && dst->type != P_RESULT && src->type == P_IMMD) {
+	if (pc->allow32 && dst->type != P_RESULT && src->type == P_IMMD) {
 		set_immd(pc, src, e);
 		/*XXX: 32-bit, but steals part of "half" reg space - need to
 		 *     catch and handle this case if/when we do half-regs
 		 */
-		e->inst[0] |= 0x00008000;
 	} else
 	if (src->type == P_IMMD || src->type == P_CONST) {
 		set_long(pc, e);
@@ -426,18 +499,25 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 		e->inst[0] |= (src->hw << 9);
 	}
 
-	/* We really should support "half" instructions here at some point,
-	 * but I don't feel confident enough about them yet.
-	 */
-	set_long(pc, e);
 	if (is_long(e) && !is_immd(e)) {
 		e->inst[1] |= 0x04000000; /* 32-bit */
-		e->inst[1] |= 0x0003c000; /* "subsubop" 0xf == mov */
-	}
+		e->inst[1] |= 0x0000c000; /* "subsubop" 0x3 */
+		if (!(e->inst[1] & 0x20000000))
+			e->inst[1] |= 0x00030000; /* "subsubop" 0xf */
+	} else
+		e->inst[0] |= 0x00008000;
 
 	emit(pc, e);
 }
 
+static INLINE void
+emit_mov_immdval(struct nv50_pc *pc, struct nv50_reg *dst, float f)
+{
+	struct nv50_reg *imm = alloc_immd(pc, f);
+	emit_mov(pc, dst, imm);
+	FREE(imm);
+}
+
 static boolean
 check_swap_src_0_1(struct nv50_pc *pc,
 		   struct nv50_reg **s0, struct nv50_reg **s1)
@@ -541,12 +621,26 @@ emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	struct nv50_program_exec *e = exec(pc);
 
 	e->inst[0] |= 0xc0000000;
-	set_long(pc, e);
+
+	if (!pc->allow32)
+		set_long(pc, e);
 
 	check_swap_src_0_1(pc, &src0, &src1);
 	set_dst(pc, dst, e);
 	set_src_0(pc, src0, e);
-	set_src_1(pc, src1, e);
+	if (src1->type == P_IMMD && !is_long(e)) {
+		if (src0->neg)
+			e->inst[0] |= 0x00008000;
+		set_immd(pc, src1, e);
+	} else {
+		set_src_1(pc, src1, e);
+		if (src0->neg ^ src1->neg) {
+			if (is_long(e))
+				e->inst[1] |= 0x08000000;
+			else
+				e->inst[0] |= 0x00008000;
+		}
+	}
 
 	emit(pc, e);
 }
@@ -560,11 +654,20 @@ emit_add(struct nv50_pc *pc, struct nv50_reg *dst,
 	e->inst[0] |= 0xb0000000;
 
 	check_swap_src_0_1(pc, &src0, &src1);
+
+	if (!pc->allow32 || src0->neg || src1->neg) {
+		set_long(pc, e);
+		e->inst[1] |= (src0->neg << 26) | (src1->neg << 27);
+	}
+
 	set_dst(pc, dst, e);
 	set_src_0(pc, src0, e);
-	if (is_long(e))
+	if (src1->type == P_CONST || src1->type == P_ATTR || is_long(e))
 		set_src_2(pc, src1, e);
 	else
+	if (src1->type == P_IMMD)
+		set_immd(pc, src1, e);
+	else
 		set_src_1(pc, src1, e);
 
 	emit(pc, e);
@@ -588,25 +691,13 @@ emit_minmax(struct nv50_pc *pc, unsigned sub, struct nv50_reg *dst,
 	emit(pc, e);
 }
 
-static void
+static INLINE void
 emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	 struct nv50_reg *src1)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] |= 0xb0000000;
-
-	set_long(pc, e);
-	if (check_swap_src_0_1(pc, &src0, &src1))
-		e->inst[1] |= 0x04000000;
-	else
-		e->inst[1] |= 0x08000000;
-
-	set_dst(pc, dst, e);
-	set_src_0(pc, src0, e);
-	set_src_2(pc, src1, e);
-
-	emit(pc, e);
+	src1->neg ^= 1;
+	emit_add(pc, dst, src0, src1);
+	src1->neg ^= 1;
 }
 
 static void
@@ -623,26 +714,21 @@ emit_mad(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	set_src_1(pc, src1, e);
 	set_src_2(pc, src2, e);
 
+	if (src0->neg ^ src1->neg)
+		e->inst[1] |= 0x04000000;
+	if (src2->neg)
+		e->inst[1] |= 0x08000000;
+
 	emit(pc, e);
 }
 
-static void
+static INLINE void
 emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	 struct nv50_reg *src1, struct nv50_reg *src2)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] |= 0xe0000000;
-	set_long(pc, e);
-	e->inst[1] |= 0x08000000; /* src0 * src1 - src2 */
-
-	check_swap_src_0_1(pc, &src0, &src1);
-	set_dst(pc, dst, e);
-	set_src_0(pc, src0, e);
-	set_src_1(pc, src1, e);
-	set_src_2(pc, src2, e);
-
-	emit(pc, e);
+	src2->neg ^= 1;
+	emit_mad(pc, dst, src0, src1, src2);
+	src2->neg ^= 1;
 }
 
 static void
@@ -693,6 +779,48 @@ emit_precossin(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 	emit(pc, e);
 }
 
+#define CVTOP_RN	0x01
+#define CVTOP_FLOOR	0x03
+#define CVTOP_CEIL	0x05
+#define CVTOP_TRUNC	0x07
+#define CVTOP_SAT	0x08
+#define CVTOP_ABS	0x10
+
+#define CVT_F32_F32 0xc4
+#define CVT_F32_S32 0x44
+#define CVT_F32_U32 0x64
+#define CVT_S32_F32 0x8c
+#define CVT_S32_S32 0x0c
+#define CVT_F32_F32_ROP 0xcc
+
+static void
+emit_cvt(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src,
+	 int wp, unsigned cop, unsigned fmt)
+{
+	struct nv50_program_exec *e;
+
+	e = exec(pc);
+	set_long(pc, e);
+
+	e->inst[0] |= 0xa0000000;
+	e->inst[1] |= 0x00004000;
+	e->inst[1] |= (cop << 16);
+	e->inst[1] |= (fmt << 24);
+	set_src_0(pc, src, e);
+
+	if (wp >= 0)
+		set_pred_wr(pc, 1, wp, e);
+
+	if (dst)
+		set_dst(pc, dst, e);
+	else {
+		e->inst[0] |= 0x000001fc;
+		e->inst[1] |= 0x00000008;
+	}
+
+	emit(pc, e);
+}
+
 static void
 emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst,
 	 struct nv50_reg *src0, struct nv50_reg *src1)
@@ -736,22 +864,10 @@ emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst,
 		free_temp(pc, dst);
 }
 
-static void
+static INLINE void
 emit_flr(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] = 0xa0000000; /* cvt */
-	set_long(pc, e);
-	e->inst[1] |= (6 << 29); /* cvt */
-	e->inst[1] |= 0x08000000; /* integer mode */
-	e->inst[1] |= 0x04000000; /* 32 bit */
-	e->inst[1] |= ((0x1 << 3)) << 14; /* .rn */
-	e->inst[1] |= (1 << 14); /* src .f32 */
-	set_dst(pc, dst, e);
-	set_src_0(pc, src, e);
-
-	emit(pc, e);
+	emit_cvt(pc, dst, src, -1, CVTOP_FLOOR, CVT_F32_F32_ROP);
 }
 
 static void
@@ -768,21 +884,10 @@ emit_pow(struct nv50_pc *pc, struct nv50_reg *dst,
 	free_temp(pc, temp);
 }
 
-static void
+static INLINE void
 emit_abs(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] = 0xa0000000; /* cvt */
-	set_long(pc, e);
-	e->inst[1] |= (6 << 29); /* cvt */
-	e->inst[1] |= 0x04000000; /* 32 bit */
-	e->inst[1] |= (1 << 14); /* src .f32 */
-	e->inst[1] |= ((1 << 6) << 14); /* .abs */
-	set_dst(pc, dst, e);
-	set_src_0(pc, src, e);
-
-	emit(pc, e);
+	emit_cvt(pc, dst, src, -1, CVTOP_ABS, CVT_F32_F32);
 }
 
 static void
@@ -794,18 +899,12 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
 	struct nv50_reg *neg128 = alloc_immd(pc, -127.999999);
 	struct nv50_reg *pos128 = alloc_immd(pc,  127.999999);
 	struct nv50_reg *tmp[4];
+	boolean allow32 = pc->allow32;
 
-	if (mask & (1 << 0))
-		emit_mov(pc, dst[0], one);
-
-	if (mask & (1 << 3))
-		emit_mov(pc, dst[3], one);
+	pc->allow32 = FALSE;
 
 	if (mask & (3 << 1)) {
-		if (mask & (1 << 1))
-			tmp[0] = dst[1];
-		else
-			tmp[0] = temp_temp(pc);
+		tmp[0] = alloc_temp(pc, NULL);
 		emit_minmax(pc, 4, tmp[0], src[0], zero);
 	}
 
@@ -823,6 +922,26 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
 		emit_mov(pc, dst[2], zero);
 		set_pred(pc, 3, 0, pc->p->exec_tail);
 	}
+
+	if (mask & (1 << 1))
+		assimilate_temp(pc, dst[1], tmp[0]);
+	else
+	if (mask & (1 << 2))
+		free_temp(pc, tmp[0]);
+
+	pc->allow32 = allow32;
+
+	/* do this last, in case src[i,j] == dst[0,3] */
+	if (mask & (1 << 0))
+		emit_mov(pc, dst[0], one);
+
+	if (mask & (1 << 3))
+		emit_mov(pc, dst[3], one);
+
+	FREE(pos128);
+	FREE(neg128);
+	FREE(zero);
+	FREE(one);
 }
 
 static void
@@ -853,6 +972,8 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src)
 	e->inst[1] = 0xc4014788;
 	set_src_0(pc, src, e);
 	set_pred_wr(pc, 1, r_pred, e);
+	if (src->neg)
+		e->inst[1] |= 0x20000000;
 	emit(pc, e);
 
 	/* This is probably KILP */
@@ -863,6 +984,181 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src)
 	emit(pc, e);
 }
 
+static void
+emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
+	 struct nv50_reg **src, unsigned unit, unsigned type, boolean proj)
+{
+	struct nv50_reg *temp, *t[4];
+	struct nv50_program_exec *e;
+
+	unsigned c, mode, dim;
+
+	switch (type) {
+	case TGSI_TEXTURE_1D:
+		dim = 1;
+		break;
+	case TGSI_TEXTURE_UNKNOWN:
+	case TGSI_TEXTURE_2D:
+	case TGSI_TEXTURE_SHADOW1D: /* XXX: x, z */
+	case TGSI_TEXTURE_RECT:
+		dim = 2;
+		break;
+	case TGSI_TEXTURE_3D:
+	case TGSI_TEXTURE_CUBE:
+	case TGSI_TEXTURE_SHADOW2D:
+	case TGSI_TEXTURE_SHADOWRECT: /* XXX */
+		dim = 3;
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	/* some cards need t[0]'s hw index to be a multiple of 4 */
+	alloc_temp4(pc, t, 0);
+
+	if (proj) {
+		if (src[0]->type == P_TEMP && src[0]->rhw != -1) {
+			mode = pc->interp_mode[src[0]->index];
+
+			t[3]->rhw = src[3]->rhw;
+			emit_interp(pc, t[3], NULL, (mode & INTERP_CENTROID));
+			emit_flop(pc, 0, t[3], t[3]);
+
+			for (c = 0; c < dim; c++) {
+				t[c]->rhw = src[c]->rhw;
+				emit_interp(pc, t[c], t[3],
+					    (mode | INTERP_PERSPECTIVE));
+			}
+		} else {
+			emit_flop(pc, 0, t[3], src[3]);
+			for (c = 0; c < dim; c++)
+				emit_mul(pc, t[c], src[c], t[3]);
+
+			/* XXX: for some reason the blob sometimes uses MAD:
+			 * emit_mad(pc, t[c], src[0][c], t[3], t[3])
+			 * pc->p->exec_tail->inst[1] |= 0x080fc000;
+			 */
+		}
+	} else {
+		if (type == TGSI_TEXTURE_CUBE) {
+			temp = temp_temp(pc);
+			emit_minmax(pc, 4, temp, src[0], src[1]);
+			emit_minmax(pc, 4, temp, temp, src[2]);
+			emit_flop(pc, 0, temp, temp);
+			for (c = 0; c < 3; c++)
+				emit_mul(pc, t[c], src[c], temp);
+		} else {
+			for (c = 0; c < dim; c++)
+				emit_mov(pc, t[c], src[c]);
+		}
+	}
+
+	e = exec(pc);
+	set_long(pc, e);
+	e->inst[0] |= 0xf0000000;
+	e->inst[1] |= 0x00000004;
+	set_dst(pc, t[0], e);
+	e->inst[0] |= (unit << 9);
+
+	if (dim == 2)
+		e->inst[0] |= 0x00400000;
+	else
+	if (dim == 3)
+		e->inst[0] |= 0x00800000;
+
+	e->inst[0] |= (mask & 0x3) << 25;
+	e->inst[1] |= (mask & 0xc) << 12;
+
+	emit(pc, e);
+
+#if 1
+	if (mask & 1) emit_mov(pc, dst[0], t[0]);
+	if (mask & 2) emit_mov(pc, dst[1], t[1]);
+	if (mask & 4) emit_mov(pc, dst[2], t[2]);
+	if (mask & 8) emit_mov(pc, dst[3], t[3]);
+
+	free_temp4(pc, t);
+#else
+	/* XXX: if p.e. MUL is used directly after TEX, it would still use
+	 * the texture coordinates, not the fetched values: latency ? */
+
+	for (c = 0; c < 4; c++) {
+		if (mask & (1 << c))
+			assimilate_temp(pc, dst[c], t[c]);
+		else
+			free_temp(pc, t[c]);
+	}
+#endif
+}
+
+static void
+convert_to_long(struct nv50_pc *pc, struct nv50_program_exec *e)
+{
+	unsigned q = 0, m = ~0;
+
+	assert(!is_long(e));
+
+	switch (e->inst[0] >> 28) {
+	case 0x1:
+		/* MOV */
+		q = 0x0403c000;
+		m = 0xffff7fff;
+		break;
+	case 0x8:
+		/* INTERP (move centroid, perspective and flat bits) */
+		m = ~0x03000100;
+		q = (e->inst[0] & (3 << 24)) >> (24 - 16);
+		q |= (e->inst[0] & (1 << 8)) << (18 - 8);
+		break;
+	case 0x9:
+		/* RCP */
+		break;
+	case 0xB:
+		/* ADD */
+		m = ~(127 << 16);
+		q = ((e->inst[0] & (~m)) >> 2);
+		break;
+	case 0xC:
+		/* MUL */
+		m = ~0x00008000;
+		q = ((e->inst[0] & (~m)) << 12);
+		break;
+	case 0xE:
+		/* MAD (if src2 == dst) */
+		q = ((e->inst[0] & 0x1fc) << 12);
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	set_long(pc, e);
+	pc->p->exec_size++;
+
+	e->inst[0] &= m;
+	e->inst[1] |= q;
+}
+
+static boolean
+negate_supported(const struct tgsi_full_instruction *insn, int i)
+{
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_DP3:
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_MUL:
+	case TGSI_OPCODE_KIL:
+	case TGSI_OPCODE_ADD:
+	case TGSI_OPCODE_SUB:
+	case TGSI_OPCODE_MAD:
+		return TRUE;
+	case TGSI_OPCODE_POW:
+		return (i == 1) ? TRUE : FALSE;
+	default:
+		return FALSE;
+	}
+}
+
 static struct nv50_reg *
 tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst)
 {
@@ -881,11 +1177,14 @@ tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst)
 }
 
 static struct nv50_reg *
-tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
+tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src,
+	 boolean neg)
 {
 	struct nv50_reg *r = NULL;
 	struct nv50_reg *temp;
-	unsigned c;
+	unsigned sgn, c;
+
+	sgn = tgsi_util_get_full_src_register_sign_mode(src, chan);
 
 	c = tgsi_util_get_full_src_register_extswizzle(src, chan);
 	switch (c) {
@@ -915,16 +1214,17 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 		break;
 	case TGSI_EXTSWIZZLE_ZERO:
 		r = alloc_immd(pc, 0.0);
-		break;
+		return r;
 	case TGSI_EXTSWIZZLE_ONE:
-		r = alloc_immd(pc, 1.0);
-		break;
+		if (sgn == TGSI_UTIL_SIGN_TOGGLE || sgn == TGSI_UTIL_SIGN_SET)
+			return alloc_immd(pc, -1.0);
+		return alloc_immd(pc, 1.0);
 	default:
 		assert(0);
 		break;
 	}
 
-	switch (tgsi_util_get_full_src_register_sign_mode(src, chan)) {
+	switch (sgn) {
 	case TGSI_UTIL_SIGN_KEEP:
 		break;
 	case TGSI_UTIL_SIGN_CLEAR:
@@ -933,14 +1233,21 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 		r = temp;
 		break;
 	case TGSI_UTIL_SIGN_TOGGLE:
-		temp = temp_temp(pc);
-		emit_neg(pc, temp, r);
-		r = temp;
+		if (neg)
+			r->neg = 1;
+		else {
+			temp = temp_temp(pc);
+			emit_neg(pc, temp, r);
+			r = temp;
+		}
 		break;
 	case TGSI_UTIL_SIGN_SET:
 		temp = temp_temp(pc);
 		emit_abs(pc, temp, r);
-		emit_neg(pc, temp, r);
+		if (neg)
+			temp->neg = 1;
+		else
+			emit_neg(pc, temp, temp);
 		r = temp;
 		break;
 	default:
@@ -951,12 +1258,40 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 	return r;
 }
 
+/* returns TRUE if instruction can overwrite sources before they're read */
+static boolean
+direct2dest_op(const struct tgsi_full_instruction *insn)
+{
+	if (insn->Instruction.Saturate)
+		return FALSE;
+
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_COS:
+	case TGSI_OPCODE_DP3:
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_DPH:
+	case TGSI_OPCODE_KIL:
+	case TGSI_OPCODE_LIT:
+	case TGSI_OPCODE_POW:
+	case TGSI_OPCODE_RCP:
+	case TGSI_OPCODE_RSQ:
+	case TGSI_OPCODE_SCS:
+	case TGSI_OPCODE_SIN:
+	case TGSI_OPCODE_TEX:
+	case TGSI_OPCODE_TXP:
+		return FALSE;
+	default:
+		return TRUE;
+	}
+}
+
 static boolean
 nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 {
 	const struct tgsi_full_instruction *inst = &tok->FullInstruction;
 	struct nv50_reg *rdst[4], *dst[4], *src[3][4], *temp;
 	unsigned mask, sat, unit;
+	boolean assimilate = FALSE;
 	int i, c;
 
 	mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
@@ -967,6 +1302,10 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			dst[c] = tgsi_dst(pc, c, &inst->FullDstRegisters[0]);
 		else
 			dst[c] = NULL;
+		rdst[c] = NULL;
+		src[0][c] = NULL;
+		src[1][c] = NULL;
+		src[2][c] = NULL;
 	}
 
 	for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
@@ -976,7 +1315,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			unit = fs->SrcRegister.Index;
 
 		for (c = 0; c < 4; c++)
-			src[i][c] = tgsi_src(pc, c, fs);
+			src[i][c] = tgsi_src(pc, c, fs,
+					     negate_supported(inst, i));
 	}
 
 	if (sat) {
@@ -984,6 +1324,25 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			rdst[c] = dst[c];
 			dst[c] = temp_temp(pc);
 		}
+	} else
+	if (direct2dest_op(inst)) {
+		for (c = 0; c < 4; c++) {
+			if (!dst[c] || dst[c]->type != P_TEMP)
+				continue;
+
+			for (i = c + 1; i < 4; i++) {
+				if (dst[c] == src[0][i] ||
+				    dst[c] == src[1][i] ||
+				    dst[c] == src[2][i])
+					break;
+			}
+			if (i == 4)
+				continue;
+
+			assimilate = TRUE;
+			rdst[c] = dst[c];
+			dst[c] = alloc_temp(pc, NULL);
+		}
 	}
 
 	switch (inst->Instruction.Opcode) {
@@ -1002,7 +1361,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_COS:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		emit_flop(pc, 5, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1012,7 +1371,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_DP3:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1021,10 +1380,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DP4:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1034,10 +1392,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DPH:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1047,7 +1404,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DST:
 	{
@@ -1064,7 +1420,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 	}
 		break;
 	case TGSI_OPCODE_EX2:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_preex2(pc, temp, src[0][0]);
 		emit_flop(pc, 6, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1072,7 +1428,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_FLR:
 		for (c = 0; c < 4; c++) {
@@ -1082,26 +1437,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_FRC:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flr(pc, temp, src[0][c]);
 			emit_sub(pc, dst[c], src[0][c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_KIL:
 		emit_kil(pc, src[0][0]);
 		emit_kil(pc, src[0][1]);
 		emit_kil(pc, src[0][2]);
 		emit_kil(pc, src[0][3]);
+		pc->p->cfg.fp.regs[2] |= 0x00100000;
 		break;
 	case TGSI_OPCODE_LIT:
 		emit_lit(pc, &dst[0], mask, &src[0][0]);
 		break;
 	case TGSI_OPCODE_LG2:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_flop(pc, 3, temp, src[0][0]);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
@@ -1110,15 +1465,12 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_LRP:
+		temp = temp_temp(pc);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
-			/*XXX: we can do better than this */
-			temp = alloc_temp(pc, NULL);
-			emit_neg(pc, temp, src[0][c]);
-			emit_mad(pc, temp, temp, src[2][c], src[2][c]);
-			emit_mad(pc, dst[c], src[0][c], src[1][c], temp);
-			free_temp(pc, temp);
+			emit_sub(pc, temp, src[1][c], src[2][c]);
+			emit_mad(pc, dst[c], temp, src[0][c], src[2][c]);
 		}
 		break;
 	case TGSI_OPCODE_MAD:
@@ -1143,6 +1495,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_MOV:
+	case TGSI_OPCODE_SWZ:
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
@@ -1157,36 +1510,39 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_POW:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_pow(pc, temp, src[0][0], src[1][0]);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_RCP:
-		for (c = 0; c < 4; c++) {
+		for (c = 3; c >= 0; c--) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flop(pc, 0, dst[c], src[0][0]);
 		}
 		break;
 	case TGSI_OPCODE_RSQ:
-		for (c = 0; c < 4; c++) {
+		for (c = 3; c >= 0; c--) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flop(pc, 2, dst[c], src[0][0]);
 		}
 		break;
 	case TGSI_OPCODE_SCS:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		if (mask & (1 << 0))
 			emit_flop(pc, 5, dst[0], temp);
 		if (mask & (1 << 1))
 			emit_flop(pc, 4, dst[1], temp);
+		if (mask & (1 << 2))
+			emit_mov_immdval(pc, dst[2], 0.0);
+		if (mask & (1 << 3))
+			emit_mov_immdval(pc, dst[3], 1.0);
 		break;
 	case TGSI_OPCODE_SGE:
 		for (c = 0; c < 4; c++) {
@@ -1196,7 +1552,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_SIN:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		emit_flop(pc, 4, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1220,33 +1576,15 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_TEX:
+		emit_tex(pc, dst, mask, src[0], unit,
+			 inst->InstructionExtTexture.Texture, FALSE);
+		break;
 	case TGSI_OPCODE_TXP:
-	{
-		struct nv50_reg *t[4];
-		struct nv50_program_exec *e;
-
-		alloc_temp4(pc, t, 0);
-		emit_mov(pc, t[0], src[0][0]);
-		emit_mov(pc, t[1], src[0][1]);
-
-		e = exec(pc);
-		e->inst[0] = 0xf6400000;
-		e->inst[0] |= (unit << 9);
-		set_long(pc, e);
-		e->inst[1] |= 0x0000c004;
-		set_dst(pc, t[0], e);
-		emit(pc, e);
-
-		if (mask & (1 << 0)) emit_mov(pc, dst[0], t[0]);
-		if (mask & (1 << 1)) emit_mov(pc, dst[1], t[1]);
-		if (mask & (1 << 2)) emit_mov(pc, dst[2], t[2]);
-		if (mask & (1 << 3)) emit_mov(pc, dst[3], t[3]);
-
-		free_temp4(pc, t);
-	}
+		emit_tex(pc, dst, mask, src[0], unit,
+			 inst->InstructionExtTexture.Texture, TRUE);
 		break;
 	case TGSI_OPCODE_XPD:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		if (mask & (1 << 0)) {
 			emit_mul(pc, temp, src[0][2], src[1][1]);
 			emit_msb(pc, dst[0], src[0][1], src[1][2], temp);
@@ -1259,7 +1597,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			emit_mul(pc, temp, src[0][1], src[1][0]);
 			emit_msb(pc, dst[2], src[0][0], src[1][1], temp);
 		}
-		free_temp(pc, temp);
+		if (mask & (1 << 3))
+			emit_mov_immdval(pc, dst[3], 1.0);
 		break;
 	case TGSI_OPCODE_END:
 		break;
@@ -1270,21 +1609,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 
 	if (sat) {
 		for (c = 0; c < 4; c++) {
-			struct nv50_program_exec *e;
-
 			if (!(mask & (1 << c)))
 				continue;
-			e = exec(pc);
+			emit_cvt(pc, rdst[c], dst[c], -1, CVTOP_SAT,
+				 CVT_F32_F32);
+		}
+	} else if (assimilate) {
+		for (c = 0; c < 4; c++)
+			if (rdst[c])
+				assimilate_temp(pc, rdst[c], dst[c]);
+	}
 
-			e->inst[0] = 0xa0000000; /* cvt */
-			set_long(pc, e);
-			e->inst[1] |= (6 << 29); /* cvt */
-			e->inst[1] |= 0x04000000; /* 32 bit */
-			e->inst[1] |= (1 << 14); /* src .f32 */
-			e->inst[1] |= ((1 << 5) << 14); /* .sat */
-			set_dst(pc, rdst[c], e);
-			set_src_0(pc, dst[c], e);
-			emit(pc, e);
+	for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+		for (c = 0; c < 4; c++) {
+			if (!src[i][c])
+				continue;
+			if (src[i][c]->index == -1 && src[i][c]->type == P_IMMD)
+				FREE(src[i][c]);
+			else
+			if (src[i][c]->acc == pc->insn_cur)
+				release_hw(pc, src[i][c]);
 		}
 	}
 
@@ -1292,12 +1636,169 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 	return TRUE;
 }
 
+/* Adjust a bitmask that indicates what components of a source are used,
+ * we use this in tx_prep so we only load interpolants that are needed.
+ */
+static void
+insn_adjust_mask(const struct tgsi_full_instruction *insn, unsigned *mask)
+{
+	const struct tgsi_instruction_ext_texture *tex;
+
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_DP3:
+		*mask = 0x7;
+		break;
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_DPH:
+		*mask = 0xF;
+		break;
+	case TGSI_OPCODE_LIT:
+		*mask = 0xB;
+		break;
+	case TGSI_OPCODE_RCP:
+	case TGSI_OPCODE_RSQ:
+		*mask = 0x1;
+		break;
+	case TGSI_OPCODE_TEX:
+	case TGSI_OPCODE_TXP:
+		assert(insn->Instruction.Extended);
+		tex = &insn->InstructionExtTexture;
+
+		*mask = 0x7;
+		if (tex->Texture == TGSI_TEXTURE_1D)
+			*mask = 0x1;
+		else
+		if (tex->Texture == TGSI_TEXTURE_2D)
+			*mask = 0x3;
+
+		if (insn->Instruction.Opcode == TGSI_OPCODE_TXP)
+			*mask |= 0x8;
+		break;
+	default:
+		break;
+	}
+}
+
+static void
+prep_inspect_insn(struct nv50_pc *pc, const union tgsi_full_token *tok,
+		  unsigned *r_usage[2])
+{
+	const struct tgsi_full_instruction *insn;
+	const struct tgsi_full_src_register *src;
+	const struct tgsi_dst_register *dst;
+
+	unsigned i, c, k, n, mask, *acc_p;
+
+	insn = &tok->FullInstruction;
+	dst = &insn->FullDstRegisters[0].DstRegister;
+	mask = dst->WriteMask;
+
+	if (!r_usage[0])
+		r_usage[0] = CALLOC(pc->temp_nr * 4, sizeof(unsigned));
+	if (!r_usage[1])
+		r_usage[1] = CALLOC(pc->attr_nr * 4, sizeof(unsigned));
+
+	if (dst->File == TGSI_FILE_TEMPORARY) {
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			r_usage[0][dst->Index * 4 + c] = pc->insn_nr;
+		}
+	}
+
+	for (i = 0; i < insn->Instruction.NumSrcRegs; i++) {
+		src = &insn->FullSrcRegisters[i];
+
+		switch (src->SrcRegister.File) {
+		case TGSI_FILE_TEMPORARY:
+			acc_p = r_usage[0];
+			break;
+		case TGSI_FILE_INPUT:
+			acc_p = r_usage[1];
+			break;
+		default:
+			continue;
+		}
+
+		insn_adjust_mask(insn, &mask);
+
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+
+			k = tgsi_util_get_full_src_register_extswizzle(src, c);
+			switch (k) {
+			case TGSI_EXTSWIZZLE_X:
+			case TGSI_EXTSWIZZLE_Y:
+			case TGSI_EXTSWIZZLE_Z:
+			case TGSI_EXTSWIZZLE_W:
+				n = src->SrcRegister.Index * 4 + k;
+				acc_p[n] = pc->insn_nr;
+				break;
+			default:
+				break;
+			}
+		}
+	}
+}
+
+static unsigned
+load_fp_attrib(struct nv50_pc *pc, int i, unsigned *acc, int *mid,
+	       int *aid, int *p_oid)
+{
+	struct nv50_reg *iv;
+	int oid, c, n;
+	unsigned mask = 0;
+
+	iv = (pc->interp_mode[i] & INTERP_CENTROID) ? pc->iv_c : pc->iv_p;
+
+	for (c = 0, n = i * 4; c < 4; c++, n++) {
+		oid = (*p_oid)++;
+		pc->attr[n].type = P_TEMP;
+		pc->attr[n].index = i;
+
+		if (pc->attr[n].acc == acc[n])
+			continue;
+		mask |= (1 << c);
+
+		pc->attr[n].acc = acc[n];
+		pc->attr[n].rhw = pc->attr[n].hw = -1;
+		alloc_reg(pc, &pc->attr[n]);
+
+		pc->attr[n].rhw = (*aid)++;
+		emit_interp(pc, &pc->attr[n], iv, pc->interp_mode[i]);
+
+		pc->p->cfg.fp.map[(*mid) / 4] |= oid << (8 * ((*mid) % 4));
+		(*mid)++;
+		pc->p->cfg.fp.regs[1] += 0x00010001;
+	}
+
+	return mask;
+}
+
 static boolean
 nv50_program_tx_prep(struct nv50_pc *pc)
 {
 	struct tgsi_parse_context p;
 	boolean ret = FALSE;
 	unsigned i, c;
+	unsigned fcol, bcol, fcrd, depr;
+
+	/* count (centroid) perspective interpolations */
+	unsigned centroid_loads = 0;
+	unsigned perspect_loads = 0;
+
+	/* track register access for temps and attrs */
+	unsigned *r_usage[2];
+	r_usage[0] = NULL;
+	r_usage[1] = NULL;
+
+	depr = fcol = bcol = fcrd = 0xffff;
+
+	if (pc->p->type == PIPE_SHADER_FRAGMENT) {
+		pc->p->cfg.fp.regs[0] = 0x01000404;
+		pc->p->cfg.fp.regs[1] = 0x00000400;
+	}
 
 	tgsi_parse_init(&p, pc->p->pipe.tokens);
 	while (!tgsi_parse_end_of_tokens(&p)) {
@@ -1310,18 +1811,19 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			const struct tgsi_full_immediate *imm =
 				&p.FullToken.FullImmediate;
 
-			ctor_immd(pc, imm->u.ImmediateFloat32[0].Float,
-				      imm->u.ImmediateFloat32[1].Float,
-				      imm->u.ImmediateFloat32[2].Float,
-				      imm->u.ImmediateFloat32[3].Float);
+			ctor_immd(pc, imm->u[0].Float,
+				      imm->u[1].Float,
+				      imm->u[2].Float,
+				      imm->u[3].Float);
 		}
 			break;
 		case TGSI_TOKEN_TYPE_DECLARATION:
 		{
 			const struct tgsi_full_declaration *d;
-			unsigned last;
+			unsigned last, first, mode;
 
 			d = &p.FullToken.FullDeclaration;
+			first = d->DeclarationRange.First;
 			last = d->DeclarationRange.Last;
 
 			switch (d->Declaration.File) {
@@ -1332,10 +1834,69 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			case TGSI_FILE_OUTPUT:
 				if (pc->result_nr < (last + 1))
 					pc->result_nr = last + 1;
+
+				if (!d->Declaration.Semantic)
+					break;
+
+				switch (d->Semantic.SemanticName) {
+				case TGSI_SEMANTIC_POSITION:
+					depr = first;
+					pc->p->cfg.fp.regs[2] |= 0x00000100;
+					pc->p->cfg.fp.regs[3] |= 0x00000011;
+					break;
+				default:
+					break;
+				}
+
 				break;
 			case TGSI_FILE_INPUT:
+			{
 				if (pc->attr_nr < (last + 1))
 					pc->attr_nr = last + 1;
+
+				if (pc->p->type != PIPE_SHADER_FRAGMENT)
+					break;
+
+				switch (d->Declaration.Interpolate) {
+				case TGSI_INTERPOLATE_CONSTANT:
+					mode = INTERP_FLAT;
+					break;
+				case TGSI_INTERPOLATE_PERSPECTIVE:
+					mode = INTERP_PERSPECTIVE;
+					break;
+				default:
+					mode = INTERP_LINEAR;
+					break;
+				}
+
+				if (d->Declaration.Semantic) {
+					switch (d->Semantic.SemanticName) {
+					case TGSI_SEMANTIC_POSITION:
+						fcrd = first;
+						break;
+					case TGSI_SEMANTIC_COLOR:
+						fcol = first;
+						mode = INTERP_PERSPECTIVE;
+						break;
+					case TGSI_SEMANTIC_BCOLOR:
+						bcol = first;
+						mode = INTERP_PERSPECTIVE;
+						break;
+					}
+				}
+
+				if (d->Declaration.Centroid) {
+					mode |= INTERP_CENTROID;
+					if (mode & INTERP_PERSPECTIVE)
+						centroid_loads++;
+				} else
+				if (mode & INTERP_PERSPECTIVE)
+					perspect_loads++;
+
+				assert(last < 32);
+				for (i = first; i <= last; i++)
+					pc->interp_mode[i] = mode;
+			}
 				break;
 			case TGSI_FILE_CONSTANT:
 				if (pc->param_nr < (last + 1))
@@ -1351,6 +1912,8 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 		}
 			break;
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			pc->insn_nr++;
+			prep_inspect_insn(pc, tok, r_usage);
 			break;
 		default:
 			break;
@@ -1366,56 +1929,95 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			for (c = 0; c < 4; c++) {
 				pc->temp[i*4+c].type = P_TEMP;
 				pc->temp[i*4+c].hw = -1;
+				pc->temp[i*4+c].rhw = -1;
 				pc->temp[i*4+c].index = i;
+				pc->temp[i*4+c].acc = r_usage[0][i*4+c];
 			}
 		}
 	}
 
 	if (pc->attr_nr) {
-		struct nv50_reg *iv = NULL;
-		int aid = 0;
+		int oid = 4, mid = 4, aid = 0;
+		/* oid = VP output id
+		 * aid = FP attribute/interpolant id
+		 * mid = VP output mapping field ID
+		 */
 
 		pc->attr = CALLOC(pc->attr_nr * 4, sizeof(struct nv50_reg));
 		if (!pc->attr)
 			goto out_err;
 
 		if (pc->p->type == PIPE_SHADER_FRAGMENT) {
-			iv = alloc_temp(pc, NULL);
-			emit_interp(pc, iv, iv, NULL);
-			emit_flop(pc, 0, iv, iv);
-			aid++;
-		}
+			/* position should be loaded first */
+			if (fcrd != 0xffff) {
+				unsigned mask;
+				mid = 0;
+				mask = load_fp_attrib(pc, fcrd, r_usage[1],
+						      &mid, &aid, &oid);
+				oid = 0;
+				pc->p->cfg.fp.regs[1] |= (mask << 24);
+				pc->p->cfg.fp.map[0] = 0x04040404 * fcrd;
+			}
+			pc->p->cfg.fp.map[0] += 0x03020100;
 
-		for (i = 0; i < pc->attr_nr; i++) {
-			struct nv50_reg *a = &pc->attr[i*4];
+			/* should do MAD fcrd.xy, fcrd, SOME_CONST, fcrd */
 
-			for (c = 0; c < 4; c++) {
-				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
-					struct nv50_reg *at =
-						alloc_temp(pc, NULL);
-					pc->attr[i*4+c].type = at->type;
-					pc->attr[i*4+c].hw = at->hw;
-					pc->attr[i*4+c].index = at->index;
+			if (perspect_loads) {
+				pc->iv_p = alloc_temp(pc, NULL);
+
+				if (!(pc->p->cfg.fp.regs[1] & 0x08000000)) {
+					pc->p->cfg.fp.regs[1] |= 0x08000000;
+					pc->iv_p->rhw = aid++;
+					emit_interp(pc, pc->iv_p, NULL,
+						    INTERP_LINEAR);
+					emit_flop(pc, 0, pc->iv_p, pc->iv_p);
 				} else {
-					pc->p->cfg.vp.attr[aid/32] |=
-						(1 << (aid % 32));
-					pc->attr[i*4+c].type = P_ATTR;
-					pc->attr[i*4+c].hw = aid++;
-					pc->attr[i*4+c].index = i;
+					pc->iv_p->rhw = aid - 1;
+					emit_flop(pc, 0, pc->iv_p,
+						  &pc->attr[fcrd * 4 + 3]);
 				}
 			}
 
-			if (pc->p->type != PIPE_SHADER_FRAGMENT)
-				continue;
+			if (centroid_loads) {
+				pc->iv_c = alloc_temp(pc, NULL);
+				pc->iv_c->rhw = pc->iv_p ? aid - 1 : aid++;
+				emit_interp(pc, pc->iv_c, NULL,
+					    INTERP_CENTROID);
+				emit_flop(pc, 0, pc->iv_c, pc->iv_c);
+				pc->p->cfg.fp.regs[1] |= 0x08000000;
+			}
 
-			emit_interp(pc, &a[0], &a[0], iv);
-			emit_interp(pc, &a[1], &a[1], iv);
-			emit_interp(pc, &a[2], &a[2], iv);
-			emit_interp(pc, &a[3], &a[3], iv);
-		}
+			for (c = 0; c < 4; c++) {
+				/* I don't know what these values do, but
+				 * let's set them like the blob does:
+				 */
+				if (fcol != 0xffff && r_usage[1][fcol * 4 + c])
+					pc->p->cfg.fp.regs[0] += 0x00010000;
+				if (bcol != 0xffff && r_usage[1][bcol * 4 + c])
+					pc->p->cfg.fp.regs[0] += 0x00010000;
+			}
+
+			for (i = 0; i < pc->attr_nr; i++)
+				load_fp_attrib(pc, i, r_usage[1],
+					       &mid, &aid, &oid);
 
-		if (iv)
-			free_temp(pc, iv);
+			if (pc->iv_p)
+				free_temp(pc, pc->iv_p);
+			if (pc->iv_c)
+				free_temp(pc, pc->iv_c);
+
+			pc->p->cfg.fp.high_map = (mid / 4);
+			pc->p->cfg.fp.high_map += ((mid % 4) ? 1 : 0);
+		} else {
+			/* vertex program */
+			for (i = 0; i < pc->attr_nr * 4; i++) {
+				pc->p->cfg.vp.attr[aid / 32] |=
+					(1 << (aid % 32));
+				pc->attr[i].type = P_ATTR;
+				pc->attr[i].hw = aid++;
+				pc->attr[i].index = i / 4;
+			}
+		}
 	}
 
 	if (pc->result_nr) {
@@ -1430,12 +2032,20 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
 					pc->result[i*4+c].type = P_TEMP;
 					pc->result[i*4+c].hw = -1;
+					pc->result[i*4+c].rhw = (i == depr) ?
+						-1 : rid++;
 				} else {
 					pc->result[i*4+c].type = P_RESULT;
 					pc->result[i*4+c].hw = rid++;
 				}
 				pc->result[i*4+c].index = i;
 			}
+
+			if (pc->p->type == PIPE_SHADER_FRAGMENT &&
+			    depr != 0xffff) {
+				pc->result[depr * 4 + 2].rhw =
+					(pc->result_nr - 1) * 4;
+			}
 		}
 	}
 
@@ -1456,7 +2066,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 	}
 
 	if (pc->immd_nr) {
-		int rid = pc->param_nr * 4;
+		int rid = 0;
 
 		pc->immd = CALLOC(pc->immd_nr * 4, sizeof(struct nv50_reg));
 		if (!pc->immd)
@@ -1473,15 +2083,38 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 
 	ret = TRUE;
 out_err:
+	if (r_usage[0])
+		FREE(r_usage[0]);
+	if (r_usage[1])
+		FREE(r_usage[1]);
+
 	tgsi_parse_free(&p);
 	return ret;
 }
 
+static void
+free_nv50_pc(struct nv50_pc *pc)
+{
+	if (pc->immd)
+		FREE(pc->immd);
+	if (pc->param)
+		FREE(pc->param);
+	if (pc->result)
+		FREE(pc->result);
+	if (pc->attr)
+		FREE(pc->attr);
+	if (pc->temp)
+		FREE(pc->temp);
+
+	FREE(pc);
+}
+
 static boolean
 nv50_program_tx(struct nv50_program *p)
 {
 	struct tgsi_parse_context parse;
 	struct nv50_pc *pc;
+	unsigned k;
 	boolean ret;
 
 	pc = CALLOC_STRUCT(nv50_pc);
@@ -1498,10 +2131,16 @@ nv50_program_tx(struct nv50_program *p)
 	while (!tgsi_parse_end_of_tokens(&parse)) {
 		const union tgsi_full_token *tok = &parse.FullToken;
 
+		/* don't allow half insn/immd on first and last instruction */
+		pc->allow32 = TRUE;
+		if (pc->insn_cur == 0 || pc->insn_cur + 2 == pc->insn_nr)
+			pc->allow32 = FALSE;
+
 		tgsi_parse_token(&parse);
 
 		switch (tok->Token.Type) {
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			++pc->insn_cur;
 			ret = nv50_program_tx_insn(pc, tok);
 			if (ret == FALSE)
 				goto out_err;
@@ -1515,8 +2154,40 @@ nv50_program_tx(struct nv50_program *p)
 		struct nv50_reg out;
 
 		out.type = P_TEMP;
-		for (out.hw = 0; out.hw < pc->result_nr * 4; out.hw++)
-			emit_mov(pc, &out, &pc->result[out.hw]);
+		for (k = 0; k < pc->result_nr * 4; k++) {
+			if (pc->result[k].rhw == -1)
+				continue;
+			if (pc->result[k].hw != pc->result[k].rhw) {
+				out.hw = pc->result[k].rhw;
+				emit_mov(pc, &out, &pc->result[k]);
+			}
+			if (pc->p->cfg.high_result < (pc->result[k].rhw + 1))
+				pc->p->cfg.high_result = pc->result[k].rhw + 1;
+		}
+	}
+
+	/* look for single half instructions and make them long */
+	struct nv50_program_exec *e, *e_prev;
+
+	for (k = 0, e = pc->p->exec_head, e_prev = NULL; e; e = e->next) {
+		if (!is_long(e))
+			k++;
+
+		if (!e->next || is_long(e->next)) {
+			if (k & 1)
+				convert_to_long(pc, e);
+			k = 0;
+		}
+
+		if (e->next)
+			e_prev = e;
+	}
+
+	if (!is_long(pc->p->exec_tail)) {
+		/* this may occur if moving FP results */
+		assert(e_prev && !is_long(e_prev));
+		convert_to_long(pc, e_prev);
+		convert_to_long(pc, pc->p->exec_tail);
 	}
 
 	assert(is_long(pc->p->exec_tail) && !is_immd(pc->p->exec_head));
@@ -1530,6 +2201,7 @@ out_err:
 	tgsi_parse_free(&parse);
 
 out_cleanup:
+	free_nv50_pc(pc);
 	return ret;
 }
 
@@ -1543,17 +2215,17 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p)
 
 static void
 nv50_program_upload_data(struct nv50_context *nv50, float *map,
-			 unsigned start, unsigned count)
+			unsigned start, unsigned count, unsigned cbuf)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	while (count) {
 		unsigned nr = count > 2047 ? 2047 : count;
 
-		BEGIN_RING(chan, tesla, 0x00000f00, 1);
-		OUT_RING  (chan, (NV50_CB_PMISC << 0) | (start << 8));
-		BEGIN_RING(chan, tesla, 0x40000f04, nr);
+		BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1);
+		OUT_RING  (chan, (cbuf << 0) | (start << 8));
+		BEGIN_RING(chan, tesla, NV50TCL_CB_DATA(0) | 0x40000000, nr);
 		OUT_RINGp (chan, map, nr);
 
 		map += nr;
@@ -1565,70 +2237,93 @@ nv50_program_upload_data(struct nv50_context *nv50, float *map,
 static void
 nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct nouveau_winsys *nvws = nv50->screen->nvws;
-	struct pipe_winsys *ws = nv50->pipe.winsys;
-	unsigned nr = p->param_nr + p->immd_nr;
+	struct pipe_screen *pscreen = nv50->pipe.screen;
 
-	if (!p->data && nr) {
-		struct nouveau_resource *heap = nv50->screen->vp_data_heap;
+	if (!p->data[0] && p->immd_nr) {
+		struct nouveau_resource *heap = nv50->screen->immd_heap[0];
 
-		if (nvws->res_alloc(heap, nr, p, &p->data)) {
-			while (heap->next && heap->size < nr) {
+		if (nouveau_resource_alloc(heap, p->immd_nr, p, &p->data[0])) {
+			while (heap->next && heap->size < p->immd_nr) {
 				struct nv50_program *evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data[0]);
 			}
 
-			if (nvws->res_alloc(heap, nr, p, &p->data))
+			if (nouveau_resource_alloc(heap, p->immd_nr, p,
+						   &p->data[0]))
 				assert(0);
 		}
+
+		/* immediates only need to be uploaded again when freed */
+		nv50_program_upload_data(nv50, p->immd, p->data[0]->start,
+					 p->immd_nr, NV50_CB_PMISC);
 	}
 
-	if (p->param_nr) {
-		float *map = ws->buffer_map(ws, nv50->constbuf[p->type],
-					    PIPE_BUFFER_USAGE_CPU_READ);
-		nv50_program_upload_data(nv50, map, p->data->start,
-					 p->param_nr);
-		ws->buffer_unmap(ws, nv50->constbuf[p->type]);
+	if (!p->data[1] && p->param_nr) {
+		struct nouveau_resource *heap =
+			nv50->screen->parm_heap[p->type];
+
+		if (nouveau_resource_alloc(heap, p->param_nr, p, &p->data[1])) {
+			while (heap->next && heap->size < p->param_nr) {
+				struct nv50_program *evict = heap->next->priv;
+				nouveau_resource_free(&evict->data[1]);
+			}
+
+			if (nouveau_resource_alloc(heap, p->param_nr, p,
+						   &p->data[1]))
+				assert(0);
+		}
 	}
 
-	if (p->immd_nr) {
-		nv50_program_upload_data(nv50, p->immd,
-					 p->data->start + p->param_nr,
-					 p->immd_nr);
+	if (p->param_nr) {
+		unsigned cbuf = NV50_CB_PVP;
+		float *map = pipe_buffer_map(pscreen, nv50->constbuf[p->type],
+					     PIPE_BUFFER_USAGE_CPU_READ);
+		if (p->type == PIPE_SHADER_FRAGMENT)
+			cbuf = NV50_CB_PFP;
+		nv50_program_upload_data(nv50, map, p->data[1]->start,
+					 p->param_nr, cbuf);
+		pipe_buffer_unmap(pscreen, nv50->constbuf[p->type]);
 	}
 }
 
 static void
 nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	struct pipe_screen *screen = nv50->pipe.screen;
 	struct nv50_program_exec *e;
 	struct nouveau_stateobj *so;
 	const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR;
 	unsigned start, count, *up, *ptr;
 	boolean upload = FALSE;
 
-	if (!p->buffer) {
-		p->buffer = screen->buffer_create(screen, 0x100, 0, p->exec_size * 4);
+	if (!p->bo) {
+		nouveau_bo_new(chan->device, NOUVEAU_BO_VRAM, 0x100,
+			       p->exec_size * 4, &p->bo);
 		upload = TRUE;
 	}
 
-	if (p->data && p->data->start != p->data_start) {
+	if ((p->data[0] && p->data[0]->start != p->data_start[0]) ||
+		(p->data[1] && p->data[1]->start != p->data_start[1])) {
 		for (e = p->exec_head; e; e = e->next) {
-			unsigned ei, ci;
+			unsigned ei, ci, bs;
 
 			if (e->param.index < 0)
 				continue;
+			bs = (e->inst[1] >> 22) & 0x07;
+			assert(bs < 2);
 			ei = e->param.shift >> 5;
-			ci = e->param.index + p->data->start;
+			ci = e->param.index + p->data[bs]->start;
 
 			e->inst[ei] &= ~e->param.mask;
 			e->inst[ei] |= (ci << e->param.shift);
 		}
 
-		p->data_start = p->data->start;
+		if (p->data[0])
+			p->data_start[0] = p->data[0]->start;
+		if (p->data[1])
+			p->data_start[1] = p->data[1]->start;
+
 		upload = TRUE;
 	}
 
@@ -1637,13 +2332,11 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 
 #ifdef NV50_PROGRAM_DUMP
 	NOUVEAU_ERR("-------\n");
-	up = ptr = MALLOC(p->exec_size * 4);
 	for (e = p->exec_head; e; e = e->next) {
 		NOUVEAU_ERR("0x%08x\n", e->inst[0]);
 		if (is_long(e))
 			NOUVEAU_ERR("0x%08x\n", e->inst[1]);
 	}
-
 #endif
 
 	up = ptr = MALLOC(p->exec_size * 4);
@@ -1654,28 +2347,28 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 	}
 
 	so = so_new(4,2);
-	so_method(so, nv50->screen->tesla, 0x1280, 3);
-	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_LOW, 0, 0);
+	so_method(so, nv50->screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4));
 
 	start = 0; count = p->exec_size;
 	while (count) {
-		struct nouveau_winsys *nvws = nv50->screen->nvws;
+		struct nouveau_channel *chan = nv50->screen->base.channel;
 		unsigned nr;
 
-		so_emit(nvws, so);
+		so_emit(chan, so);
 
 		nr = MIN2(count, 2047);
-		nr = MIN2(nvws->channel->pushbuf->remaining, nr);
-		if (nvws->channel->pushbuf->remaining < (nr + 3)) {
+		nr = MIN2(chan->pushbuf->remaining, nr);
+		if (chan->pushbuf->remaining < (nr + 3)) {
 			FIRE_RING(chan);
 			continue;
 		}
 
-		BEGIN_RING(chan, tesla, 0x0f00, 1);
+		BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1);
 		OUT_RING  (chan, (start << 8) | NV50_CB_PUPLOAD);
-		BEGIN_RING(chan, tesla, 0x40000f04, nr);	
+		BEGIN_RING(chan, tesla, NV50TCL_CB_DATA(0) | 0x40000000, nr);
 		OUT_RINGp (chan, up + start, nr);
 
 		start += nr;
@@ -1704,19 +2397,19 @@ nv50_vertprog_validate(struct nv50_context *nv50)
 
 	so = so_new(13, 2);
 	so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_LOW, 0, 0);
-	so_method(so, tesla, 0x1650, 2);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_LOW, 0, 0);
+	so_method(so, tesla, NV50TCL_VP_ATTR_EN_0, 2);
 	so_data  (so, p->cfg.vp.attr[0]);
 	so_data  (so, p->cfg.vp.attr[1]);
-	so_method(so, tesla, 0x16b8, 1);
+	so_method(so, tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1);
 	so_data  (so, p->cfg.high_result);
-	so_method(so, tesla, 0x16ac, 2);
+	so_method(so, tesla, NV50TCL_VP_RESULT_MAP_SIZE, 2);
 	so_data  (so, p->cfg.high_result); //8);
 	so_data  (so, p->cfg.high_temp);
-	so_method(so, tesla, 0x140c, 1);
+	so_method(so, tesla, NV50TCL_VP_START_ID, 1);
 	so_data  (so, 0); /* program start offset */
 	so_ref(so, &nv50->state.vertprog);
 	so_ref(NULL, &so);
@@ -1728,6 +2421,7 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_program *p = nv50->fragprog;
 	struct nouveau_stateobj *so;
+	unsigned i;
 
 	if (!p->translated) {
 		nv50_program_validate(nv50, p);
@@ -1740,23 +2434,28 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 
 	so = so_new(64, 2);
 	so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_LOW, 0, 0);
-	so_method(so, tesla, 0x1904, 4);
-	so_data  (so, 0x00040404); /* p: 0x01000404 */
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_LOW, 0, 0);
+	so_method(so, tesla, NV50TCL_MAP_SEMANTIC_0, 4);
+	so_data  (so, p->cfg.fp.regs[0]); /* 0x01000404 / 0x00040404 */
 	so_data  (so, 0x00000004);
 	so_data  (so, 0x00000000);
 	so_data  (so, 0x00000000);
-	so_method(so, tesla, 0x16bc, 3); /*XXX: fixme */
-	so_data  (so, 0x03020100);
-	so_data  (so, 0x07060504);
-	so_data  (so, 0x0b0a0908);
-	so_method(so, tesla, 0x1988, 2);
-	so_data  (so, 0x08080408); //0x08040404); /* p: 0x0f000401 */
+	so_method(so, tesla, NV50TCL_VP_RESULT_MAP(0), p->cfg.fp.high_map);
+	for (i = 0; i < p->cfg.fp.high_map; i++)
+		so_data(so, p->cfg.fp.map[i]);
+	so_method(so, tesla, NV50TCL_FP_INTERPOLANT_CTRL, 2);
+	so_data  (so, p->cfg.fp.regs[1]); /* 0x08040404 / 0x0f000401 */
 	so_data  (so, p->cfg.high_temp);
-	so_method(so, tesla, 0x1414, 1);
+	so_method(so, tesla, NV50TCL_FP_RESULT_COUNT, 1);
+	so_data  (so, p->cfg.high_result);
+	so_method(so, tesla, NV50TCL_FP_CTRL_UNK19A8, 1);
+	so_data  (so, p->cfg.fp.regs[2]);
+	so_method(so, tesla, NV50TCL_FP_CTRL_UNK196C, 1);
+	so_data  (so, p->cfg.fp.regs[3]);
+	so_method(so, tesla, NV50TCL_FP_START_ID, 1);
 	so_data  (so, 0); /* program start offset */
 	so_ref(so, &nv50->state.fragprog);
 	so_ref(NULL, &so);
@@ -1765,8 +2464,6 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 void
 nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct pipe_screen *pscreen = nv50->pipe.screen;
-
 	while (p->exec_head) {
 		struct nv50_program_exec *e = p->exec_head;
 
@@ -1776,11 +2473,10 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p)
 	p->exec_tail = NULL;
 	p->exec_size = 0;
 
-	if (p->buffer)
-		pipe_buffer_reference(&p->buffer, NULL);
+	nouveau_bo_ref(NULL, &p->bo);
 
-	nv50->screen->nvws->res_free(&p->data);
+	nouveau_resource_free(&p->data[0]);
+	nouveau_resource_free(&p->data[1]);
 
 	p->translated = 0;
 }
-
diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h
index 78deed6a38..096e0476aa 100644
--- a/src/gallium/drivers/nv50/nv50_program.h
+++ b/src/gallium/drivers/nv50/nv50_program.h
@@ -24,10 +24,10 @@ struct nv50_program {
 	struct nv50_program_exec *exec_head;
 	struct nv50_program_exec *exec_tail;
 	unsigned exec_size;
-	struct nouveau_resource *data;
-	unsigned data_start;
+	struct nouveau_resource *data[2];
+	unsigned data_start[2];
 
-	struct pipe_buffer *buffer;
+	struct nouveau_bo *bo;
 
 	float *immd;
 	unsigned immd_nr;
@@ -39,6 +39,11 @@ struct nv50_program {
 		struct {
 			unsigned attr[2];
 		} vp;
+		struct {
+			unsigned regs[4];
+			unsigned map[5];
+			unsigned high_map;
+		} fp;
 	} cfg;
 };
 
diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c
index 35cebdbdc3..5305c93d59 100644
--- a/src/gallium/drivers/nv50/nv50_query.c
+++ b/src/gallium/drivers/nv50/nv50_query.c
@@ -26,7 +26,7 @@
 #include "nv50_context.h"
 
 struct nv50_query {
-	struct pipe_buffer *buffer;
+	struct nouveau_bo *bo;
 	unsigned type;
 	boolean ready;
 	uint64_t result;
@@ -41,14 +41,16 @@ nv50_query(struct pipe_query *pipe)
 static struct pipe_query *
 nv50_query_create(struct pipe_context *pipe, unsigned type)
 {
-	struct pipe_screen *screen = pipe->screen;
+	struct nouveau_device *dev = nouveau_screen(pipe->screen)->device;
 	struct nv50_query *q = CALLOC_STRUCT(nv50_query);
+	int ret;
 
 	assert (q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 	q->type = type;
 
-	q->buffer = screen->buffer_create(screen, 256, 0, 16);
-	if (!q->buffer) {
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM | NOUVEAU_BO_MAP, 256,
+			     16, &q->bo);
+	if (ret) {
 		FREE(q);
 		return NULL;
 	}
@@ -62,7 +64,7 @@ nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 	struct nv50_query *q = nv50_query(pq);
 
 	if (q) {
-		pipe_buffer_reference(&q->buffer, NULL);
+		nouveau_bo_ref(NULL, &q->bo);
 		FREE(q);
 	}
 }
@@ -71,7 +73,7 @@ static void
 nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_query *q = nv50_query(pq);
 
@@ -87,15 +89,14 @@ static void
 nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_query *q = nv50_query(pq);
-	struct nouveau_bo *bo = nv50->screen->nvws->get_bo(q->buffer);
 
 	WAIT_RING (chan, 5);
-	BEGIN_RING(chan, tesla, 0x1b00, 4);
-	OUT_RELOCh(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
-	OUT_RELOCl(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(chan, tesla, NV50TCL_QUERY_ADDRESS_HIGH, 4);
+	OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 	OUT_RING  (chan, 0x00000000);
 	OUT_RING  (chan, 0x0100f002);
 	FIRE_RING (chan);
@@ -105,19 +106,17 @@ static boolean
 nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 		  boolean wait, uint64_t *result)
 {
-	struct pipe_winsys *ws = pipe->winsys;
 	struct nv50_query *q = nv50_query(pq);
-
-	/*XXX: Want to be able to return FALSE here instead of blocking
-	 *     until the result is available..
-	 */
+	int ret;
 
 	if (!q->ready) {
-		uint32_t *map = ws->buffer_map(ws, q->buffer,
-					       PIPE_BUFFER_USAGE_CPU_READ);
-		q->result = map[1];
+		ret = nouveau_bo_map(q->bo, NOUVEAU_BO_RD |
+				     wait ? 0 : NOUVEAU_BO_NOWAIT);
+		if (ret)
+			return false;
+		q->result = ((uint32_t *)q->bo->map)[1];
 		q->ready = TRUE;
-		ws->buffer_unmap(ws, q->buffer);
+		nouveau_bo_unmap(q->bo);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
index 2980564594..c7f80a2203 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -22,8 +22,6 @@
 
 #include "pipe/p_screen.h"
 
-#include "util/u_simple_screen.h"
-
 #include "nv50_context.h"
 #include "nv50_screen.h"
 
@@ -39,8 +37,17 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
+		case PIPE_FORMAT_Z32_FLOAT:
 		case PIPE_FORMAT_Z24S8_UNORM:
-		case PIPE_FORMAT_Z16_UNORM:
+		case PIPE_FORMAT_X8Z24_UNORM:
+		case PIPE_FORMAT_S8Z24_UNORM:
 			return TRUE;
 		default:
 			break;
@@ -68,23 +75,6 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 	return FALSE;
 }
 
-static const char *
-nv50_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv50_screen *screen = nv50_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv50_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -120,6 +110,10 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 1;
 	case NOUVEAU_CAP_HW_VTXBUF:
 		return 1;
 	case NOUVEAU_CAP_HW_IDXBUF:
@@ -153,116 +147,131 @@ nv50_screen_get_paramf(struct pipe_screen *pscreen, int param)
 static void
 nv50_screen_destroy(struct pipe_screen *pscreen)
 {
-	FREE(pscreen);
+	struct nv50_screen *screen = nv50_screen(pscreen);
+
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->tesla);
+	nouveau_grobj_free(&screen->eng2d);
+	nouveau_grobj_free(&screen->m2mf);
+	nouveau_screen_fini(&screen->base);
+	FREE(screen);
 }
 
 struct pipe_screen *
-nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
-	unsigned tesla_class = 0, ret;
-	unsigned chipset = nvws->channel->device->chipset;
-	int i;
+	unsigned chipset = dev->chipset;
+	unsigned tesla_class = 0;
+	int ret, i;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
+
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv50_screen_destroy;
+	pscreen->get_param = nv50_screen_get_param;
+	pscreen->get_paramf = nv50_screen_get_paramf;
+	pscreen->is_format_supported = nv50_screen_is_format_supported;
+
+	nv50_screen_init_miptree_functions(pscreen);
+	nv50_transfer_init_screen_functions(pscreen);
 
 	/* DMA engine object */
-	ret = nvws->grobj_alloc(nvws, 0x5039, &screen->m2mf);
+	ret = nouveau_grobj_alloc(chan, 0xbeef5039,
+		NV50_MEMORY_TO_MEMORY_FORMAT, &screen->m2mf);
 	if (ret) {
 		NOUVEAU_ERR("Error creating M2MF object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->m2mf, 1);
 
 	/* 2D object */
-	ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d);
+	ret = nouveau_grobj_alloc(chan, 0xbeef502d, NV50_2D, &screen->eng2d);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 2D object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->eng2d, 2);
 
 	/* 3D object */
 	switch (chipset & 0xf0) {
 	case 0x50:
-		tesla_class = 0x5097;
+		tesla_class = NV50TCL;
 		break;
 	case 0x80:
 	case 0x90:
-		tesla_class = 0x8297;
+		/* this stupid name should be corrected. */
+		tesla_class = NV54TCL;
 		break;
 	case 0xa0:
-		tesla_class = 0x8397;
+		tesla_class = NVA0TCL;
 		break;
 	default:
 		NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", chipset);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	if (tesla_class == 0) {
 		NOUVEAU_ERR("Unknown G8x chipset: NV%02x\n", chipset);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, tesla_class, &screen->tesla);
+	ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class,
+		&screen->tesla);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->tesla, 3);
 
 	/* Sync notifier */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
-        /* Setup the pipe */
-	screen->pipe.winsys = ws;
-
-	screen->pipe.destroy = nv50_screen_destroy;
-
-	screen->pipe.get_name = nv50_screen_get_name;
-	screen->pipe.get_vendor = nv50_screen_get_vendor;
-	screen->pipe.get_param = nv50_screen_get_param;
-	screen->pipe.get_paramf = nv50_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv50_screen_is_format_supported;
-
-	nv50_screen_init_miptree_functions(&screen->pipe);
-	nv50_transfer_init_screen_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
 	/* Static M2MF init */
 	so = so_new(32, 0);
-	so_method(so, screen->m2mf, 0x0180, 3);
+	so_method(so, screen->m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 3);
 	so_data  (so, screen->sync->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_emit(nvws, so);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_emit(chan, so);
 	so_ref (NULL, &so);
 
 	/* Static 2D init */
 	so = so_new(64, 0);
 	so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4);
 	so_data  (so, screen->sync->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->eng2d, NV50_2D_OPERATION, 1);
 	so_data  (so, NV50_2D_OPERATION_SRCCOPY);
 	so_method(so, screen->eng2d, 0x0290, 1);
 	so_data  (so, 0);
 	so_method(so, screen->eng2d, 0x0888, 1);
 	so_data  (so, 1);
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
 
 	/* Static tesla init */
@@ -275,63 +284,123 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0),
 				     NV50TCL_DMA_UNK0__SIZE);
 	for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++)
-		so_data(so, nvws->channel->vram->handle);
+		so_data(so, chan->vram->handle);
 	so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0),
 				     NV50TCL_DMA_UNK1__SIZE);
 	for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++)
-		so_data(so, nvws->channel->vram->handle);
+		so_data(so, chan->vram->handle);
 	so_method(so, screen->tesla, 0x121c, 1);
 	so_data  (so, 1);
 
 	so_method(so, screen->tesla, 0x13bc, 1);
 	so_data  (so, 0x54);
+	/* origin is top left (set to 1 for bottom left) */
 	so_method(so, screen->tesla, 0x13ac, 1);
-	so_data  (so, 1);
-	so_method(so, screen->tesla, 0x16b8, 1);
+	so_data  (so, 0);
+	so_method(so, screen->tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1);
 	so_data  (so, 8);
 
-	/* Shared constant buffer */
-	screen->constbuf = screen->pipe.buffer_create(&screen->pipe, 0, 0, 128 * 4 * 4);
-	if (nvws->res_init(&screen->vp_data_heap, 0, 128)) {
-		NOUVEAU_ERR("Error initialising constant buffer\n");
-		nv50_screen_destroy(&screen->pipe);
+	/* constant buffers for immediates and VP/FP parameters */
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4,
+			     &screen->constbuf_misc[0]);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	so_method(so, screen->tesla, 0x1280, 3);
-	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+	for (i = 0; i < 2; i++) {
+		ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4,
+				     &screen->constbuf_parm[i]);
+		if (ret) {
+			nv50_screen_destroy(pscreen);
+			return NULL;
+		}
+	}
+
+	if (nouveau_resource_init(&screen->immd_heap[0], 0, 128) ||
+		nouveau_resource_init(&screen->parm_heap[0], 0, 128) ||
+		nouveau_resource_init(&screen->parm_heap[1], 0, 128))
+	{
+		NOUVEAU_ERR("Error initialising constant buffers.\n");
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
+	/*
+	// map constant buffers:
+	//  B = buffer ID (maybe more than 1 byte)
+	//  N = CB index used in shader instruction
+	//  P = program type (0 = VP, 2 = GP, 3 = FP)
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x000BBNP1);
+	*/
+
+	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PMISC << 16) | 0x00000800);
+	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
+	so_data  (so, 0x00000001 | (NV50_CB_PMISC << 12));
+	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
+	so_data  (so, 0x00000031 | (NV50_CB_PMISC << 12));
+
+	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PVP << 16) | 0x00000800);
+	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
+	so_data  (so, 0x00000101 | (NV50_CB_PVP << 12));
+
+	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
+	so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+	so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PMISC << 16) | 0x00001000);
+	so_data  (so, (NV50_CB_PFP << 16) | 0x00000800);
+	so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1);
+	so_data  (so, 0x00000131 | (NV50_CB_PFP << 12));
 
 	/* Texture sampler/image unit setup - we abuse the constant buffer
 	 * upload mechanism for the moment to upload data to the tex config
 	 * blocks.  At some point we *may* want to go the NVIDIA way of doing
 	 * things?
 	 */
-	screen->tic = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4);
-	so_method(so, screen->tesla, 0x1280, 3);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tic);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
+	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
 	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
 	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, (NV50_CB_TIC << 16) | 0x0800);
-	so_method(so, screen->tesla, 0x1574, 3);
+	so_method(so, screen->tesla, NV50TCL_TIC_ADDRESS_HIGH, 3);
 	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
 	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, 0x00000800);
 
-	screen->tsc = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4);
-	so_method(so, screen->tesla, 0x1280, 3);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tsc);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
+	so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
 	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
 	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, (NV50_CB_TSC << 16) | 0x0800);
-	so_method(so, screen->tesla, 0x155c, 3);
+	so_method(so, screen->tesla, NV50TCL_TSC_ADDRESS_HIGH, 3);
 	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
 	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
@@ -341,7 +410,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 
 	/* Vertex array limits - max them out */
 	for (i = 0; i < 16; i++) {
-		so_method(so, screen->tesla, 0x1080 + (i * 8), 2);
+		so_method(so, screen->tesla, NV50TCL_UNK1080_OFFSET_HIGH(i), 2);
 		so_data  (so, 0x000000ff);
 		so_data  (so, 0xffffffff);
 	}
@@ -352,14 +421,16 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 
 	so_method(so, screen->tesla, 0x1234, 1);
 	so_data  (so, 1);
-	so_method(so, screen->tesla, 0x1458, 1);
+
+	/* activate first scissor rectangle */
+	so_method(so, screen->tesla, NV50TCL_SCISSOR_ENABLE, 1);
 	so_data  (so, 1);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref (so, &screen->static_init);
 	so_ref (NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h
index db567aaac8..61e24a5b57 100644
--- a/src/gallium/drivers/nv50/nv50_screen.h
+++ b/src/gallium/drivers/nv50/nv50_screen.h
@@ -1,10 +1,10 @@
 #ifndef __NV50_SCREEN_H__
 #define __NV50_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 
 struct nv50_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
@@ -15,11 +15,14 @@ struct nv50_screen {
 	struct nouveau_grobj *m2mf;
 	struct nouveau_notifier *sync;
 
-	struct pipe_buffer *constbuf;
-	struct nouveau_resource *vp_data_heap;
+	struct nouveau_bo *constbuf_misc[1];
+	struct nouveau_bo *constbuf_parm[2];
 
-	struct pipe_buffer *tic;
-	struct pipe_buffer *tsc;
+	struct nouveau_resource *immd_heap[1];
+	struct nouveau_resource *parm_heap[2];
+
+	struct nouveau_bo *tic;
+	struct nouveau_bo *tsc;
 
 	struct nouveau_stateobj *static_init;
 };
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
index ba852194cd..4283808ed9 100644
--- a/src/gallium/drivers/nv50/nv50_state.c
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -136,9 +136,11 @@ static void *
 nv50_sampler_state_create(struct pipe_context *pipe,
 			  const struct pipe_sampler_state *cso)
 {
-	unsigned *tsc = CALLOC(8, sizeof(unsigned));
+	struct nv50_sampler_stateobj *sso = CALLOC(1, sizeof(*sso));
+	unsigned *tsc = sso->tsc;
+	float limit;
 
-	tsc[0] = (0x00024000 |
+	tsc[0] = (0x00026000 |
 		  (wrap_mode(cso->wrap_s) << 0) |
 		  (wrap_mode(cso->wrap_t) << 3) |
 		  (wrap_mode(cso->wrap_r) << 6));
@@ -202,7 +204,19 @@ nv50_sampler_state_create(struct pipe_context *pipe,
 		tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7);
 	}
 
-	return (void *)tsc;
+	limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+	tsc[1] |= ((int)(limit * 256.0) & 0x1fff) << 12;
+
+	tsc[2] |= ((int)CLAMP(cso->max_lod, 0.0, 15.0) << 20) |
+		  ((int)CLAMP(cso->min_lod, 0.0, 15.0) << 8);
+
+	tsc[4] = fui(cso->border_color[0]);
+	tsc[5] = fui(cso->border_color[1]);
+	tsc[6] = fui(cso->border_color[2]);
+	tsc[7] = fui(cso->border_color[3]);
+
+	sso->normalized = cso->normalized_coords;
+	return (void *)sso;
 }
 
 static void
@@ -259,6 +273,8 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
 	so_method(so, tesla, NV50TCL_SHADE_MODEL, 1);
 	so_data  (so, cso->flatshade ? NV50TCL_SHADE_MODEL_FLAT :
 				       NV50TCL_SHADE_MODEL_SMOOTH);
+	so_method(so, tesla, 0x1684, 1);
+	so_data  (so, cso->flatshade_first ? 0 : 1);
 
 	so_method(so, tesla, NV50TCL_LINE_WIDTH, 1);
 	so_data  (so, fui(cso->line_width));
@@ -395,35 +411,35 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe,
 		so_data  (so, 0);
 	}
 
-	/*XXX: yes, I know they're backwards.. header needs fixing */
+	/* XXX: keep hex values until header is updated (names reversed) */
 	if (cso->stencil[0].enabled) {
-		so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5);
+		so_method(so, tesla, 0x1380, 8);
 		so_data  (so, 1);
 		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
 		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
 		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
 		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
-		so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 3);
 		so_data  (so, cso->stencil[0].ref_value);
 		so_data  (so, cso->stencil[0].writemask);
 		so_data  (so, cso->stencil[0].valuemask);
 	} else {
-		so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1);
+		so_method(so, tesla, 0x1380, 1);
 		so_data  (so, 0);
 	}
 
 	if (cso->stencil[1].enabled) {
-		so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 8);
+		so_method(so, tesla, 0x1594, 5);
 		so_data  (so, 1);
 		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
 		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
 		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
 		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
+		so_method(so, tesla, 0x0f54, 3);
 		so_data  (so, cso->stencil[1].ref_value);
 		so_data  (so, cso->stencil[1].writemask);
 		so_data  (so, cso->stencil[1].valuemask);
 	} else {
-		so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1);
+		so_method(so, tesla, 0x1594, 1);
 		so_data  (so, 0);
 	}
 
diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c
index c13d3de1cb..344c2cf6dd 100644
--- a/src/gallium/drivers/nv50/nv50_state_validate.c
+++ b/src/gallium/drivers/nv50/nv50_state_validate.c
@@ -32,6 +32,9 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	unsigned i, w, h, gw = 0;
 
 	for (i = 0; i < fb->nr_cbufs; i++) {
+		struct pipe_texture *pt = fb->cbufs[i]->texture;
+		struct nouveau_bo *bo = nv50_miptree(pt)->base.bo;
+
 		if (!gw) {
 			w = fb->cbufs[i]->width;
 			h = fb->cbufs[i]->height;
@@ -46,26 +49,24 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 		so_data  (so, fb->cbufs[i]->height);
 
 		so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5);
-		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-			  NOUVEAU_BO_RDWR, 0, 0);
-		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
-			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0);
 		switch (fb->cbufs[i]->format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
-			so_data(so, 0xcf);
+			so_data(so, NV50TCL_RT_FORMAT_A8R8G8B8_UNORM);
 			break;
 		case PIPE_FORMAT_R5G6B5_UNORM:
-			so_data(so, 0xe8);
+			so_data(so, NV50TCL_RT_FORMAT_R5G6B5_UNORM);
 			break;
 		default:
 			NOUVEAU_ERR("AIIII unknown format %s\n",
 				    pf_name(fb->cbufs[i]->format));
-			so_data(so, 0xe6);
+			so_data(so, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM);
 			break;
 		}
-		so_data(so, 0x00000000);
+		so_data(so, bo->tile_mode << 4);
 		so_data(so, 0x00000000);
 
 		so_method(so, tesla, 0x1224, 1);
@@ -73,6 +74,9 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	}
 
 	if (fb->zsbuf) {
+		struct pipe_texture *pt = fb->zsbuf->texture;
+		struct nouveau_bo *bo = nv50_miptree(pt)->base.bo;
+
 		if (!gw) {
 			w = fb->zsbuf->width;
 			h = fb->zsbuf->height;
@@ -83,31 +87,35 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 		}
 
 		so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5);
-		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-			  NOUVEAU_BO_RDWR, 0, 0);
-		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
-			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0);
 		switch (fb->zsbuf->format) {
+		case PIPE_FORMAT_Z32_FLOAT:
+			so_data(so, NV50TCL_ZETA_FORMAT_Z32_FLOAT);
+			break;
 		case PIPE_FORMAT_Z24S8_UNORM:
-			so_data(so, 0x16);
+			so_data(so, NV50TCL_ZETA_FORMAT_Z24S8_UNORM);
+			break;
+		case PIPE_FORMAT_X8Z24_UNORM:
+			so_data(so, NV50TCL_ZETA_FORMAT_X8Z24_UNORM);
 			break;
-		case PIPE_FORMAT_Z16_UNORM:
-			so_data(so, 0x15);
+		case PIPE_FORMAT_S8Z24_UNORM:
+			so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM);
 			break;
 		default:
 			NOUVEAU_ERR("AIIII unknown format %s\n",
 				    pf_name(fb->zsbuf->format));
-			so_data(so, 0x16);
+			so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM);
 			break;
 		}
-		so_data(so, 0x00000000);
+		so_data(so, bo->tile_mode << 4);
 		so_data(so, 0x00000000);
 
 		so_method(so, tesla, 0x1538, 1);
 		so_data  (so, 1);
-		so_method(so, tesla, 0x1228, 3);
+		so_method(so, tesla, NV50TCL_ZETA_HORIZ, 3);
 		so_data  (so, fb->zsbuf->width);
 		so_data  (so, fb->zsbuf->height);
 		so_data  (so, 0x00010001);
@@ -116,12 +124,18 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ, 2);
 	so_data  (so, w << 16);
 	so_data  (so, h << 16);
-	so_method(so, tesla, 0x0e04, 2);
+	/* set window lower left corner */
+	so_method(so, tesla, NV50TCL_WINDOW_LEFT, 2);
+	so_data  (so, 0);
+	so_data  (so, 0);
+	/* set screen scissor rectangle */
+	so_method(so, tesla, NV50TCL_SCREEN_SCISSOR_HORIZ, 2);
 	so_data  (so, w << 16);
 	so_data  (so, h << 16);
-	so_method(so, tesla, 0xdf8, 2);
-	so_data  (so, 0);
-	so_data  (so, h);
+
+	/* we set scissors to framebuffer size when they're 'turned off' */
+	nv50->dirty |= NV50_NEW_SCISSOR;
+	so_ref(NULL, &nv50->state.scissor);
 
 	so_ref(so, &nv50->state.fb);
 	so_ref(NULL, &so);
@@ -131,48 +145,75 @@ static void
 nv50_state_emit(struct nv50_context *nv50)
 {
 	struct nv50_screen *screen = nv50->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 
 	if (nv50->pctx_id != screen->cur_pctx) {
-		nv50->state.dirty |= 0xffffffff;
+		if (nv50->state.fb)
+			nv50->state.dirty |= NV50_NEW_FRAMEBUFFER;
+		if (nv50->state.blend)
+			nv50->state.dirty |= NV50_NEW_BLEND;
+		if (nv50->state.zsa)
+			nv50->state.dirty |= NV50_NEW_ZSA;
+		if (nv50->state.vertprog)
+			nv50->state.dirty |= NV50_NEW_VERTPROG;
+		if (nv50->state.fragprog)
+			nv50->state.dirty |= NV50_NEW_FRAGPROG;
+		if (nv50->state.rast)
+			nv50->state.dirty |= NV50_NEW_RASTERIZER;
+		if (nv50->state.blend_colour)
+			nv50->state.dirty |= NV50_NEW_BLEND_COLOUR;
+		if (nv50->state.stipple)
+			nv50->state.dirty |= NV50_NEW_STIPPLE;
+		if (nv50->state.scissor)
+			nv50->state.dirty |= NV50_NEW_SCISSOR;
+		if (nv50->state.viewport)
+			nv50->state.dirty |= NV50_NEW_VIEWPORT;
+		if (nv50->state.tsc_upload)
+			nv50->state.dirty |= NV50_NEW_SAMPLER;
+		if (nv50->state.tic_upload)
+			nv50->state.dirty |= NV50_NEW_TEXTURE;
+		if (nv50->state.vtxfmt && nv50->state.vtxbuf)
+			nv50->state.dirty |= NV50_NEW_ARRAYS;
 		screen->cur_pctx = nv50->pctx_id;
 	}
 
 	if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER)
-		so_emit(nvws, nv50->state.fb);
+		so_emit(chan, nv50->state.fb);
 	if (nv50->state.dirty & NV50_NEW_BLEND)
-		so_emit(nvws, nv50->state.blend);
+		so_emit(chan, nv50->state.blend);
 	if (nv50->state.dirty & NV50_NEW_ZSA)
-		so_emit(nvws, nv50->state.zsa);
+		so_emit(chan, nv50->state.zsa);
 	if (nv50->state.dirty & NV50_NEW_VERTPROG)
-		so_emit(nvws, nv50->state.vertprog);
+		so_emit(chan, nv50->state.vertprog);
 	if (nv50->state.dirty & NV50_NEW_FRAGPROG)
-		so_emit(nvws, nv50->state.fragprog);
+		so_emit(chan, nv50->state.fragprog);
 	if (nv50->state.dirty & NV50_NEW_RASTERIZER)
-		so_emit(nvws, nv50->state.rast);
+		so_emit(chan, nv50->state.rast);
 	if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR)
-		so_emit(nvws, nv50->state.blend_colour);
+		so_emit(chan, nv50->state.blend_colour);
 	if (nv50->state.dirty & NV50_NEW_STIPPLE)
-		so_emit(nvws, nv50->state.stipple);
+		so_emit(chan, nv50->state.stipple);
 	if (nv50->state.dirty & NV50_NEW_SCISSOR)
-		so_emit(nvws, nv50->state.scissor);
+		so_emit(chan, nv50->state.scissor);
 	if (nv50->state.dirty & NV50_NEW_VIEWPORT)
-		so_emit(nvws, nv50->state.viewport);
+		so_emit(chan, nv50->state.viewport);
 	if (nv50->state.dirty & NV50_NEW_SAMPLER)
-		so_emit(nvws, nv50->state.tsc_upload);
+		so_emit(chan, nv50->state.tsc_upload);
 	if (nv50->state.dirty & NV50_NEW_TEXTURE)
-		so_emit(nvws, nv50->state.tic_upload);
+		so_emit(chan, nv50->state.tic_upload);
 	if (nv50->state.dirty & NV50_NEW_ARRAYS) {
-		so_emit(nvws, nv50->state.vtxfmt);
-		so_emit(nvws, nv50->state.vtxbuf);
+		so_emit(chan, nv50->state.vtxfmt);
+		so_emit(chan, nv50->state.vtxbuf);
+		if (nv50->state.vtxattr)
+			so_emit(chan, nv50->state.vtxattr);
 	}
 	nv50->state.dirty = 0;
 
-	so_emit_reloc_markers(nvws, nv50->state.fb);
-	so_emit_reloc_markers(nvws, nv50->state.vertprog);
-	so_emit_reloc_markers(nvws, nv50->state.fragprog);
-	so_emit_reloc_markers(nvws, nv50->state.vtxbuf);
-	so_emit_reloc_markers(nvws, nv50->screen->static_init);
+	so_emit_reloc_markers(chan, nv50->state.fb);
+	so_emit_reloc_markers(chan, nv50->state.vertprog);
+	so_emit_reloc_markers(chan, nv50->state.fragprog);
+	so_emit_reloc_markers(chan, nv50->state.vtxbuf);
+	so_emit_reloc_markers(chan, nv50->screen->static_init);
 }
 
 boolean
@@ -230,13 +271,13 @@ nv50_state_validate(struct nv50_context *nv50)
 		nv50->state.scissor_enabled = rast->scissor;
 
 		so = so_new(3, 0);
-		so_method(so, tesla, 0x0ff4, 2);
+		so_method(so, tesla, NV50TCL_SCISSOR_HORIZ, 2);
 		if (nv50->state.scissor_enabled) {
-			so_data(so, ((s->maxx - s->minx) << 16) | s->minx);
-			so_data(so, ((s->maxy - s->miny) << 16) | s->miny);
+			so_data(so, (s->maxx << 16) | s->minx);
+			so_data(so, (s->maxy << 16) | s->miny);
 		} else {
-			so_data(so, (8192 << 16));
-			so_data(so, (8192 << 16));
+			so_data(so, (nv50->framebuffer.width << 16));
+			so_data(so, (nv50->framebuffer.height << 16));
 		}
 		so_ref(so, &nv50->state.scissor);
 		so_ref(NULL, &so);
@@ -260,20 +301,22 @@ scissor_uptodate:
 
 		so = so_new(12, 0);
 		if (!bypass) {
-			so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3);
+			so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE(0), 3);
 			so_data  (so, fui(nv50->viewport.translate[0]));
 			so_data  (so, fui(nv50->viewport.translate[1]));
 			so_data  (so, fui(nv50->viewport.translate[2]));
-			so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3);
+			so_method(so, tesla, NV50TCL_VIEWPORT_SCALE(0), 3);
 			so_data  (so, fui(nv50->viewport.scale[0]));
-			so_data  (so, fui(-nv50->viewport.scale[1]));
+			so_data  (so, fui(nv50->viewport.scale[1]));
 			so_data  (so, fui(nv50->viewport.scale[2]));
-			so_method(so, tesla, 0x192c, 1);
+
+			so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1);
 			so_data  (so, 1);
+			/* no idea what 0f90 does */
 			so_method(so, tesla, 0x0f90, 1);
 			so_data  (so, 0);
 		} else {
-			so_method(so, tesla, 0x192c, 1);
+			so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1);
 			so_data  (so, 0);
 			so_method(so, tesla, 0x0f90, 1);
 			so_data  (so, 1);
@@ -289,16 +332,17 @@ viewport_uptodate:
 		int i;
 
 		so = so_new(nv50->sampler_nr * 8 + 3, 0);
-		so_method(so, tesla, 0x0f00, 1);
+		so_method(so, tesla, NV50TCL_CB_ADDR, 1);
 		so_data  (so, NV50_CB_TSC);
-		so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8);
+		so_method(so, tesla, NV50TCL_CB_DATA(0) | 0x40000000,
+			nv50->sampler_nr * 8);
 		for (i = 0; i < nv50->sampler_nr; i++)
-			so_datap (so, nv50->sampler[i], 8);
+			so_datap (so, nv50->sampler[i]->tsc, 8);
 		so_ref(so, &nv50->state.tsc_upload);
 		so_ref(NULL, &so);
 	}
 
-	if (nv50->dirty & NV50_NEW_TEXTURE)
+	if (nv50->dirty & (NV50_NEW_TEXTURE | NV50_NEW_SAMPLER))
 		nv50_tex_validate(nv50);
 
 	if (nv50->dirty & NV50_NEW_ARRAYS)
diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c
index 0cc5168144..b266324f58 100644
--- a/src/gallium/drivers/nv50/nv50_surface.c
+++ b/src/gallium/drivers/nv50/nv50_surface.c
@@ -35,14 +35,13 @@ nv50_format(enum pipe_format format)
 {
 	switch (format) {
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
-	case PIPE_FORMAT_Z24S8_UNORM:
-		return NV50_2D_DST_FORMAT_32BPP;
+		return NV50_2D_DST_FORMAT_A8R8G8B8_UNORM;
 	case PIPE_FORMAT_X8R8G8B8_UNORM:
-		return NV50_2D_DST_FORMAT_24BPP;
+		return NV50_2D_DST_FORMAT_X8R8G8B8_UNORM;
 	case PIPE_FORMAT_R5G6B5_UNORM:
-		return NV50_2D_DST_FORMAT_16BPP;
+		return NV50_2D_DST_FORMAT_R5G6B5_UNORM;
 	case PIPE_FORMAT_A8_UNORM:
-		return NV50_2D_DST_FORMAT_8BPP;
+		return NV50_2D_DST_FORMAT_R8_UNORM;
 	default:
 		return -1;
 	}
@@ -52,21 +51,17 @@ static int
 nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
 {
 	struct nv50_miptree *mt = nv50_miptree(ps->texture);
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
-	struct nouveau_bo *bo;
+	struct nouveau_bo *bo = nv50_miptree(ps->texture)->base.bo;
  	int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT;
  	int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
- 
-	bo = screen->nvws->get_bo(nv50_miptree(ps->texture)->buffer);
-	if (!bo)
-		return 1;
 
  	format = nv50_format(ps->format);
  	if (format < 0)
  		return 1;
   
- 	if (!bo->tiled) {
+ 	if (!bo->tile_flags) {
  		BEGIN_RING(chan, eng2d, mthd, 2);
  		OUT_RING  (chan, format);
  		OUT_RING  (chan, 1);
@@ -80,7 +75,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
  		BEGIN_RING(chan, eng2d, mthd, 5);
  		OUT_RING  (chan, format);
  		OUT_RING  (chan, 0);
- 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, bo->tile_mode << 4);
  		OUT_RING  (chan, 1);
  		OUT_RING  (chan, 0);
  		BEGIN_RING(chan, eng2d, mthd + 0x18, 4);
@@ -108,7 +103,7 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst,
 		     int dx, int dy, struct pipe_surface *src, int sx, int sy,
 		     int w, int h)
 {
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
 	int ret;
 
@@ -149,7 +144,7 @@ nv50_surface_copy(struct pipe_context *pipe,
 		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
 		  unsigned width, unsigned height)
 {
-	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	struct nv50_context *nv50 = nv50_context(pipe);
 	struct nv50_screen *screen = nv50->screen;
 
 	assert(src->format == dest->format);
@@ -163,9 +158,9 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
 		  unsigned destx, unsigned desty, unsigned width,
 		  unsigned height, unsigned value)
 {
-	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	struct nv50_context *nv50 = nv50_context(pipe);
 	struct nv50_screen *screen = nv50->screen;
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
 	int format, ret;
 
diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c
index 223c8a3a45..033cb50c11 100644
--- a/src/gallium/drivers/nv50/nv50_tex.c
+++ b/src/gallium/drivers/nv50/nv50_tex.c
@@ -26,35 +26,36 @@
 #include "nouveau/nouveau_stateobj.h"
 
 static int
-nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt)
+nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so,
+		   struct nv50_miptree *mt, int unit)
 {
-	switch (mt->base.format) {
+	switch (mt->base.base.format) {
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
 		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
-			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM |
 			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
-			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
 			    NV50TIC_0_0_FMT_8_8_8_8);
 		break;
 	case PIPE_FORMAT_A1R5G5B5_UNORM:
 		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
-			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM |
 			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
-			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
 			    NV50TIC_0_0_FMT_1_5_5_5);
 		break;
 	case PIPE_FORMAT_A4R4G4B4_UNORM:
 		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
-			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM |
 			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
-			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
 			    NV50TIC_0_0_FMT_4_4_4_4);
 		break;
 	case PIPE_FORMAT_R5G6B5_UNORM:
 		so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM |
-			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM |
 			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
-			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
 			    NV50TIC_0_0_FMT_5_6_5);
 		break;
 	case PIPE_FORMAT_L8_UNORM:
@@ -117,15 +118,18 @@ nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt)
 		return 1;
 	}
 
-	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
+	so_reloc(so, mt->base.bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
 		     NOUVEAU_BO_RD, 0, 0);
-	so_data (so, 0xd0005000);
+	if (nv50->sampler[unit]->normalized)
+		so_data (so, 0xd0005000 | mt->base.bo->tile_mode << 22);
+	else
+		so_data (so, 0x5001d000 | mt->base.bo->tile_mode << 22);
 	so_data (so, 0x00300000);
-	so_data (so, mt->base.width[0]);
-	so_data (so, (mt->base.depth[0] << 16) | mt->base.height[0]);
+	so_data (so, mt->base.base.width[0]);
+	so_data (so, (mt->base.base.last_level << 28) |
+		     (mt->base.base.depth[0] << 16) | mt->base.base.height[0]);
 	so_data (so, 0x03000000);
-	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-		     NOUVEAU_BO_RD, 0, 0);
+	so_data (so, mt->base.base.last_level << 4);
 
 	return 0;
 }
@@ -135,23 +139,38 @@ nv50_tex_validate(struct nv50_context *nv50)
 {
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nouveau_stateobj *so;
-	int unit;
+	int unit, push;
 
-	so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2);
-	so_method(so, tesla, 0x0f00, 1);
+	push  = nv50->miptree_nr * 9 + 2;
+	push += MAX2(nv50->miptree_nr, nv50->state.miptree_nr) * 2;
+
+	so = so_new(push, nv50->miptree_nr * 2);
+	so_method(so, tesla, NV50TCL_CB_ADDR, 1);
 	so_data  (so, NV50_CB_TIC);
-	so_method(so, tesla, 0x40000f04, nv50->miptree_nr * 8);
 	for (unit = 0; unit < nv50->miptree_nr; unit++) {
 		struct nv50_miptree *mt = nv50->miptree[unit];
 
-		if (nv50_tex_construct(so, mt)) {
+		so_method(so, tesla, NV50TCL_CB_DATA(0) | 0x40000000, 8);
+		if (nv50_tex_construct(nv50, so, mt, unit)) {
 			NOUVEAU_ERR("failed tex validate\n");
 			so_ref(NULL, &so);
 			return;
 		}
+
+		so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1);
+		so_data  (so, (unit << NV50TCL_SET_SAMPLER_TEX_TIC_SHIFT) |
+			(unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) |
+			NV50TCL_SET_SAMPLER_TEX_VALID);
+	}
+
+	for (; unit < nv50->state.miptree_nr; unit++) {
+		so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1);
+		so_data  (so,
+			(unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) | 0);
 	}
 
 	so_ref(so, &nv50->state.tic_upload);
 	so_ref(NULL, &so);
+	nv50->state.miptree_nr = nv50->miptree_nr;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h
index aca622c73b..207fb039f7 100644
--- a/src/gallium/drivers/nv50/nv50_texture.h
+++ b/src/gallium/drivers/nv50/nv50_texture.h
@@ -14,13 +14,13 @@
 #define NV50TIC_0_0_MAPA_C2                                       0x20000000
 #define NV50TIC_0_0_MAPA_C3                                       0x28000000
 #define NV50TIC_0_0_MAPA_ONE                                      0x38000000
-#define NV50TIC_0_0_MAPR_MASK                                     0x07000000
-#define NV50TIC_0_0_MAPR_ZERO                                     0x00000000
-#define NV50TIC_0_0_MAPR_C0                                       0x02000000
-#define NV50TIC_0_0_MAPR_C1                                       0x03000000
-#define NV50TIC_0_0_MAPR_C2                                       0x04000000
-#define NV50TIC_0_0_MAPR_C3                                       0x05000000
-#define NV50TIC_0_0_MAPR_ONE                                      0x07000000
+#define NV50TIC_0_0_MAPB_MASK                                     0x07000000
+#define NV50TIC_0_0_MAPB_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPB_C0                                       0x02000000
+#define NV50TIC_0_0_MAPB_C1                                       0x03000000
+#define NV50TIC_0_0_MAPB_C2                                       0x04000000
+#define NV50TIC_0_0_MAPB_C3                                       0x05000000
+#define NV50TIC_0_0_MAPB_ONE                                      0x07000000
 #define NV50TIC_0_0_MAPG_MASK                                     0x00e00000
 #define NV50TIC_0_0_MAPG_ZERO                                     0x00000000
 #define NV50TIC_0_0_MAPG_C0                                       0x00400000
@@ -28,31 +28,49 @@
 #define NV50TIC_0_0_MAPG_C2                                       0x00800000
 #define NV50TIC_0_0_MAPG_C3                                       0x00a00000
 #define NV50TIC_0_0_MAPG_ONE                                      0x00e00000
-#define NV50TIC_0_0_MAPB_MASK                                     0x001c0000
-#define NV50TIC_0_0_MAPB_ZERO                                     0x00000000
-#define NV50TIC_0_0_MAPB_C0                                       0x00080000
-#define NV50TIC_0_0_MAPB_C1                                       0x000c0000
-#define NV50TIC_0_0_MAPB_C2                                       0x00100000
-#define NV50TIC_0_0_MAPB_C3                                       0x00140000
-#define NV50TIC_0_0_MAPB_ONE                                      0x001c0000
+#define NV50TIC_0_0_MAPR_MASK                                     0x001c0000
+#define NV50TIC_0_0_MAPR_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPR_C0                                       0x00080000
+#define NV50TIC_0_0_MAPR_C1                                       0x000c0000
+#define NV50TIC_0_0_MAPR_C2                                       0x00100000
+#define NV50TIC_0_0_MAPR_C3                                       0x00140000
+#define NV50TIC_0_0_MAPR_ONE                                      0x001c0000
 #define NV50TIC_0_0_TYPEA_MASK                                    0x00038000
 #define NV50TIC_0_0_TYPEA_UNORM                                   0x00010000
-#define NV50TIC_0_0_TYPER_MASK                                    0x00007000
-#define NV50TIC_0_0_TYPER_UNORM                                   0x00002000
+#define NV50TIC_0_0_TYPEA_SNORM                                   0x00008000
+#define NV50TIC_0_0_TYPEA_FLOAT                                   0x00038000
+#define NV50TIC_0_0_TYPEB_MASK                                    0x00007000
+#define NV50TIC_0_0_TYPEB_UNORM                                   0x00002000
+#define NV50TIC_0_0_TYPEB_SNORM                                   0x00001000
+#define NV50TIC_0_0_TYPEB_FLOAT                                   0x00007000
 #define NV50TIC_0_0_TYPEG_MASK                                    0x00000e00
 #define NV50TIC_0_0_TYPEG_UNORM                                   0x00000400
-#define NV50TIC_0_0_TYPEB_MASK                                    0x000001c0
-#define NV50TIC_0_0_TYPEB_UNORM                                   0x00000080
-#define NV50TIC_0_0_FMT_MASK                                      0x0000003c
+#define NV50TIC_0_0_TYPEG_SNORM                                   0x00000200
+#define NV50TIC_0_0_TYPEG_FLOAT                                   0x00000e00
+#define NV50TIC_0_0_TYPER_MASK                                    0x000001c0
+#define NV50TIC_0_0_TYPER_UNORM                                   0x00000080
+#define NV50TIC_0_0_TYPER_SNORM                                   0x00000040
+#define NV50TIC_0_0_TYPER_FLOAT                                   0x000001c0
+#define NV50TIC_0_0_FMT_MASK                                      0x0000003f
+#define NV50TIC_0_0_FMT_32_32_32_32                               0x00000001
+#define NV50TIC_0_0_FMT_16_16_16_16                               0x00000003
+#define NV50TIC_0_0_FMT_32_32                                     0x00000004
 #define NV50TIC_0_0_FMT_8_8_8_8                                   0x00000008
+#define NV50TIC_0_0_FMT_2_10_10_10                                0x00000009
+#define NV50TIC_0_0_FMT_32                                        0x0000000f
 #define NV50TIC_0_0_FMT_4_4_4_4                                   0x00000012
-#define NV50TIC_0_0_FMT_1_5_5_5                                   0x00000013
+/* #define NV50TIC_0_0_FMT_1_5_5_5                                0x00000013 */
+#define NV50TIC_0_0_FMT_1_5_5_5                                   0x00000014
 #define NV50TIC_0_0_FMT_5_6_5                                     0x00000015
 #define NV50TIC_0_0_FMT_8_8                                       0x00000018
+#define NV50TIC_0_0_FMT_16                                        0x0000001b
 #define NV50TIC_0_0_FMT_8                                         0x0000001d
+#define NV50TIC_0_0_FMT_10_11_11                                  0x00000021
 #define NV50TIC_0_0_FMT_DXT1                                      0x00000024
 #define NV50TIC_0_0_FMT_DXT3                                      0x00000025
 #define NV50TIC_0_0_FMT_DXT5                                      0x00000026
+#define NV50TIC_0_0_FMT_RGTC1                                     0x00000027
+#define NV50TIC_0_0_FMT_RGTC2                                     0x00000028
 
 #define NV50TIC_0_1_OFFSET_LOW_MASK                               0xffffffff
 #define NV50TIC_0_1_OFFSET_LOW_SHIFT                                       0
@@ -102,6 +120,7 @@
 #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_EDGE                   0x00000140
 #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_BORDER                 0x00000180
 #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP                           0x000001c0
+#define NV50TSC_1_0_MAX_ANISOTROPY_MASK                          0x00700000
 
 #define NV50TSC_1_1_MAGF_MASK                                    0x00000003
 #define NV50TSC_1_1_MAGF_NEAREST                                 0x00000001
@@ -113,17 +132,19 @@
 #define NV50TSC_1_1_MIPF_NONE                                    0x00000040
 #define NV50TSC_1_1_MIPF_NEAREST                                 0x00000080
 #define NV50TSC_1_1_MIPF_LINEAR                                  0x000000c0
+#define NV50TSC_1_1_LOD_BIAS_MASK                                0x01fff000
 
-#define NV50TSC_1_2_UNKNOWN_MASK                                 0xffffffff
+#define NV50TSC_1_2_MIN_LOD_MASK                                 0x00000f00
+#define NV50TSC_1_2_MAX_LOD_MASK                                 0x00f00000
 
 #define NV50TSC_1_3_UNKNOWN_MASK                                 0xffffffff
 
-#define NV50TSC_1_4_UNKNOWN_MASK                                 0xffffffff
+#define NV50TSC_1_4_BORDER_COLOR_RED_MASK                        0xffffffff
 
-#define NV50TSC_1_5_UNKNOWN_MASK                                 0xffffffff
+#define NV50TSC_1_5_BORDER_COLOR_GREEN_MASK                      0xffffffff
 
-#define NV50TSC_1_6_UNKNOWN_MASK                                 0xffffffff
+#define NV50TSC_1_6_BORDER_COLOR_BLUE_MASK                       0xffffffff
 
-#define NV50TSC_1_7_UNKNOWN_MASK                                 0xffffffff
+#define NV50TSC_1_7_BORDER_COLOR_ALPHA_MASK                      0xffffffff
 
 #endif
diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c
index 747195b4f6..e9c3562194 100644
--- a/src/gallium/drivers/nv50/nv50_transfer.c
+++ b/src/gallium/drivers/nv50/nv50_transfer.c
@@ -6,8 +6,9 @@
 
 struct nv50_transfer {
 	struct pipe_transfer base;
-	struct pipe_buffer *buffer;
-	struct nv50_miptree_level *level;
+	struct nouveau_bo *bo;
+	unsigned level_offset;
+	unsigned level_tiling;
 	int level_pitch;
 	int level_width;
 	int level_height;
@@ -16,51 +17,57 @@ struct nv50_transfer {
 };
 
 static void
-nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src,
-			int src_pitch, int sx, int sy, int sw, int sh,
-			struct pipe_buffer *dst, int dst_pitch, int dx, int dy,
-			int dw, int dh, int cpp, int width, int height,
+nv50_transfer_rect_m2mf(struct pipe_screen *pscreen,
+			struct nouveau_bo *src_bo, unsigned src_offset,
+			int src_pitch, unsigned src_tile_mode,
+			int sx, int sy, int sw, int sh,
+			struct nouveau_bo *dst_bo, unsigned dst_offset,
+			int dst_pitch, unsigned dst_tile_mode,
+			int dx, int dy, int dw, int dh,
+			int cpp, int width, int height,
 			unsigned src_reloc, unsigned dst_reloc)
 {
 	struct nv50_screen *screen = nv50_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
-	struct nouveau_channel *chan = nvws->channel;
+	struct nouveau_channel *chan = screen->m2mf->channel;
 	struct nouveau_grobj *m2mf = screen->m2mf;
-	struct nouveau_bo *src_bo = nvws->get_bo(src);
-	struct nouveau_bo *dst_bo = nvws->get_bo(dst);
-	unsigned src_offset = 0, dst_offset = 0;
 
 	src_reloc |= NOUVEAU_BO_RD;
 	dst_reloc |= NOUVEAU_BO_WR;
 
 	WAIT_RING (chan, 14);
 
-	if (!src_bo->tiled) {
-		BEGIN_RING(chan, m2mf, 0x0200, 1);
+	if (!src_bo->tile_flags) {
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 1);
 		OUT_RING  (chan, 1);
-		BEGIN_RING(chan, m2mf, 0x0314, 1);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_IN, 1);
 		OUT_RING  (chan, src_pitch);
-		src_offset = (sy * src_pitch) + (sx * cpp);
+		src_offset += (sy * src_pitch) + (sx * cpp);
 	} else {
-		BEGIN_RING(chan, m2mf, 0x0200, 6);
-		OUT_RING  (chan, 0);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 6);
 		OUT_RING  (chan, 0);
+		OUT_RING  (chan, src_tile_mode << 4);
 		OUT_RING  (chan, sw * cpp);
 		OUT_RING  (chan, sh);
 		OUT_RING  (chan, 1);
 		OUT_RING  (chan, 0);
 	}
 
-	if (!dst_bo->tiled) {
-		BEGIN_RING(chan, m2mf, 0x021c, 1);
+	if (!dst_bo->tile_flags) {
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 1);
 		OUT_RING  (chan, 1);
-		BEGIN_RING(chan, m2mf, 0x0318, 1);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT, 1);
 		OUT_RING  (chan, dst_pitch);
-		dst_offset = (dy * dst_pitch) + (dx * cpp);
+		dst_offset += (dy * dst_pitch) + (dx * cpp);
 	} else {
-		BEGIN_RING(chan, m2mf, 0x021c, 6);
-		OUT_RING  (chan, 0);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 6);
 		OUT_RING  (chan, 0);
+		OUT_RING  (chan, dst_tile_mode << 4);
 		OUT_RING  (chan, dw * cpp);
 		OUT_RING  (chan, dh);
 		OUT_RING  (chan, 1);
@@ -71,25 +78,30 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src,
 		int line_count = height > 2047 ? 2047 : height;
 
 		WAIT_RING (chan, 15);
-		BEGIN_RING(chan, m2mf, 0x0238, 2);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH, 2);
 		OUT_RELOCh(chan, src_bo, src_offset, src_reloc);
 		OUT_RELOCh(chan, dst_bo, dst_offset, dst_reloc);
-		BEGIN_RING(chan, m2mf, 0x030c, 2);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 2);
 		OUT_RELOCl(chan, src_bo, src_offset, src_reloc);
 		OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc);
-		if (src_bo->tiled) {
-			BEGIN_RING(chan, m2mf, 0x0218, 1);
-			OUT_RING  (chan, (dy << 16) | sx);
+		if (src_bo->tile_flags) {
+			BEGIN_RING(chan, m2mf,
+				NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_IN, 1);
+			OUT_RING  (chan, (sy << 16) | sx);
 		} else {
 			src_offset += (line_count * src_pitch);
 		}
-		if (dst_bo->tiled) {
-			BEGIN_RING(chan, m2mf, 0x0234, 1);
-			OUT_RING  (chan, (sy << 16) | dx);
+		if (dst_bo->tile_flags) {
+			BEGIN_RING(chan, m2mf,
+				NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_OUT, 1);
+			OUT_RING  (chan, (dy << 16) | dx);
 		} else {
 			dst_offset += (line_count * dst_pitch);
 		}
-		BEGIN_RING(chan, m2mf, 0x031c, 4);
+		BEGIN_RING(chan, m2mf,
+			NV50_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN, 4);
 		OUT_RING  (chan, width * cpp);
 		OUT_RING  (chan, line_count);
 		OUT_RING  (chan, 0x00000101);
@@ -108,10 +120,12 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 		  enum pipe_transfer_usage usage,
 		  unsigned x, unsigned y, unsigned w, unsigned h)
 {
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
 	struct nv50_miptree *mt = nv50_miptree(pt);
 	struct nv50_miptree_level *lvl = &mt->level[level];
 	struct nv50_transfer *tx;
 	unsigned image = 0;
+	int ret;
 
 	if (pt->target == PIPE_TEXTURE_CUBE)
 		image = face;
@@ -133,20 +147,27 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	tx->base.stride = (w * pt->block.size);
 	tx->base.usage = usage;
 
-	tx->level = lvl;
 	tx->level_pitch = lvl->pitch;
-	tx->level_width = mt->base.width[level];
-	tx->level_height = mt->base.height[level];
+	tx->level_width = mt->base.base.width[level];
+	tx->level_height = mt->base.base.height[level];
+	tx->level_offset = lvl->image_offset[image];
+	tx->level_tiling = lvl->tile_mode;
 	tx->level_x = x;
 	tx->level_y = y;
-	tx->buffer =
-		pipe_buffer_create(pscreen, 0, NOUVEAU_BUFFER_USAGE_TRANSFER,
-				   w * tx->base.block.size * h);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0,
+			     w * pt->block.size * h, &tx->bo);
+	if (ret) {
+		FREE(tx);
+		return NULL;
+	}
 
 	if (usage != PIPE_TRANSFER_WRITE) {
-		nv50_transfer_rect_m2mf(pscreen, mt->buffer, tx->level_pitch,
-					x, y, tx->level_width, tx->level_height,
-					tx->buffer, tx->base.stride, 0, 0,
+		nv50_transfer_rect_m2mf(pscreen, mt->base.bo, tx->level_offset,
+					tx->level_pitch, tx->level_tiling,
+					x, y,
+					tx->level_width, tx->level_height,
+					tx->bo, 0, tx->base.stride,
+					tx->bo->tile_mode, 0, 0,
 					tx->base.width, tx->base.height,
 					tx->base.block.size, w, h,
 					NOUVEAU_BO_VRAM | NOUVEAU_BO_GART,
@@ -164,17 +185,20 @@ nv50_transfer_del(struct pipe_transfer *ptx)
 
 	if (ptx->usage != PIPE_TRANSFER_READ) {
 		struct pipe_screen *pscreen = ptx->texture->screen;
-		nv50_transfer_rect_m2mf(pscreen, tx->buffer, tx->base.stride,
-					0, 0, tx->base.width, tx->base.height,
-					mt->buffer, tx->level_pitch,
+		nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride,
+					tx->bo->tile_mode, 0, 0,
+					tx->base.width, tx->base.height,
+					mt->base.bo, tx->level_offset,
+					tx->level_pitch, tx->level_tiling,
 					tx->level_x, tx->level_y,
 					tx->level_width, tx->level_height,
 					tx->base.block.size, tx->base.width,
-					tx->base.height, NOUVEAU_BO_GART,
-					NOUVEAU_BO_VRAM | NOUVEAU_BO_GART);
+					tx->base.height,
+					NOUVEAU_BO_GART, NOUVEAU_BO_VRAM |
+					NOUVEAU_BO_GART);
 	}
 
-	pipe_buffer_reference(&tx->buffer, NULL);
+	nouveau_bo_ref(NULL, &tx->bo);
 	pipe_texture_reference(&ptx->texture, NULL);
 	FREE(ptx);
 }
@@ -184,13 +208,17 @@ nv50_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 {
 	struct nv50_transfer *tx = (struct nv50_transfer *)ptx;
 	unsigned flags = 0;
+	int ret;
 
 	if (ptx->usage & PIPE_TRANSFER_WRITE)
-		flags |= PIPE_BUFFER_USAGE_CPU_WRITE;
+		flags |= NOUVEAU_BO_WR;
 	if (ptx->usage & PIPE_TRANSFER_READ)
-		flags |= PIPE_BUFFER_USAGE_CPU_READ;
+		flags |= NOUVEAU_BO_RD;
 
-	return pipe_buffer_map(pscreen, tx->buffer, flags);
+	ret = nouveau_bo_map(tx->bo, flags);
+	if (ret)
+		return NULL;
+	return tx->bo->map;
 }
 
 static void
@@ -198,7 +226,7 @@ nv50_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 {
 	struct nv50_transfer *tx = (struct nv50_transfer *)ptx;
 
-	pipe_buffer_unmap(pscreen, tx->buffer);
+	nouveau_bo_unmap(tx->bo);
 }
 
 void
diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c
index 0749c90691..eeed148c7b 100644
--- a/src/gallium/drivers/nv50/nv50_vbo.c
+++ b/src/gallium/drivers/nv50/nv50_vbo.c
@@ -22,6 +22,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv50_context.h"
 
@@ -48,12 +49,88 @@ nv50_prim(unsigned mode)
 	return NV50TCL_VERTEX_BEGIN_POINTS;
 }
 
+static INLINE uint32_t
+nv50_vbo_type_to_hw(unsigned type)
+{
+	switch (type) {
+	case PIPE_FORMAT_TYPE_FLOAT:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT;
+	case PIPE_FORMAT_TYPE_UNORM:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM;
+	case PIPE_FORMAT_TYPE_SNORM:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM;
+	case PIPE_FORMAT_TYPE_USCALED:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED;
+	case PIPE_FORMAT_TYPE_SSCALED:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED;
+	/*
+	case PIPE_FORMAT_TYPE_UINT:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UINT;
+	case PIPE_FORMAT_TYPE_SINT:
+		return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SINT; */
+	default:
+		return 0;
+	}
+}
+
+static INLINE uint32_t
+nv50_vbo_size_to_hw(unsigned size, unsigned nr_c)
+{
+	static const uint32_t hw_values[] = {
+		0, 0, 0, 0,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8_8,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16_16,
+		0, 0, 0, 0,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32,
+		NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32_32 };
+
+	/* we'd also have R11G11B10 and R10G10B10A2 */
+
+	assert(nr_c > 0 && nr_c <= 4);
+
+	if (size > 32)
+		return 0;
+	size >>= (3 - 2);
+
+	return hw_values[size + (nr_c - 1)];
+}
+
+static INLINE uint32_t
+nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve)
+{
+	uint32_t hw_type, hw_size;
+	enum pipe_format pf = ve->src_format;
+	unsigned size = pf_size_x(pf) << pf_exp2(pf);
+
+	hw_type = nv50_vbo_type_to_hw(pf_type(pf));
+	hw_size = nv50_vbo_size_to_hw(size, ve->nr_components);
+
+	if (!hw_type || !hw_size) {
+		NOUVEAU_ERR("unsupported vbo format: %s\n", pf_name(pf));
+		abort();
+		return 0x24e80000;
+	}
+
+	if (pf_swizzle_x(pf) == 2) /* BGRA */
+		hw_size |= (1 << 31); /* no real swizzle bits :-( */
+
+	return (hw_type | hw_size);
+}
+
 boolean
 nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start,
 		 unsigned count)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	nv50_state_validate(nv50);
@@ -83,7 +160,7 @@ static INLINE void
 nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -101,7 +178,7 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map,
 
 		BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1);
 		for (i = 0; i < nr; i += 2)
-			OUT_RING  (chan, (map[1] << 16) | map[0]);
+			OUT_RING  (chan, (map[i + 1] << 16) | map[i]);
 
 		count -= nr;
 		map += nr;
@@ -112,7 +189,7 @@ static INLINE void
 nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -130,7 +207,7 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map,
 
 		BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1);
 		for (i = 0; i < nr; i += 2)
-			OUT_RING  (chan, (map[1] << 16) | map[0]);
+			OUT_RING  (chan, (map[i + 1] << 16) | map[i]);
 
 		count -= nr;
 		map += nr;
@@ -138,10 +215,10 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map,
 }
 
 static INLINE void
-nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map,
+nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -163,10 +240,12 @@ nv50_draw_elements(struct pipe_context *pipe,
 		   unsigned mode, unsigned start, unsigned count)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	struct pipe_winsys *ws = pipe->winsys;
-	void *map = ws->buffer_map(ws, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ);
+	struct pipe_screen *pscreen = pipe->screen;
+	void *map;
+	
+	map = pipe_buffer_map(pscreen, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ);
 
 	nv50_state_validate(nv50);
 
@@ -193,64 +272,129 @@ nv50_draw_elements(struct pipe_context *pipe,
 	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
 	OUT_RING  (chan, 0);
 
+	pipe_buffer_unmap(pscreen, indexBuffer);
 	pipe->flush(pipe, 0, NULL);
 	return TRUE;
 }
 
+static INLINE boolean
+nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib,
+		       struct nouveau_stateobj **pso,
+		       struct pipe_vertex_element *ve,
+		       struct pipe_vertex_buffer *vb)
+
+{
+	struct nouveau_stateobj *so;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nouveau_bo *bo = nouveau_bo(vb->buffer);
+	float *v;
+	int ret;
+	enum pipe_format pf = ve->src_format;
+
+	if ((pf_type(pf) != PIPE_FORMAT_TYPE_FLOAT) ||
+	    (pf_size_x(pf) << pf_exp2(pf)) != 32)
+		return FALSE;
+
+	ret = nouveau_bo_map(bo, NOUVEAU_BO_RD);
+	if (ret)
+		return FALSE;
+	v = (float *)(bo->map + (vb->buffer_offset + ve->src_offset));
+
+	so = *pso;
+	if (!so)
+		*pso = so = so_new(nv50->vtxelt_nr * 5, 0);
+
+	switch (ve->nr_components) {
+	case 4:
+		so_method(so, tesla, NV50TCL_VTX_ATTR_4F_X(attrib), 4);
+		so_data  (so, fui(v[0]));
+		so_data  (so, fui(v[1]));
+		so_data  (so, fui(v[2]));
+		so_data  (so, fui(v[3]));
+		break;
+	case 3:
+		so_method(so, tesla, NV50TCL_VTX_ATTR_3F_X(attrib), 3);
+		so_data  (so, fui(v[0]));
+		so_data  (so, fui(v[1]));
+		so_data  (so, fui(v[2]));
+		break;
+	case 2:
+		so_method(so, tesla, NV50TCL_VTX_ATTR_2F_X(attrib), 2);
+		so_data  (so, fui(v[0]));
+		so_data  (so, fui(v[1]));
+		break;
+	case 1:
+		so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 1);
+		so_data  (so, fui(v[0]));
+		break;
+	default:
+		nouveau_bo_unmap(bo);
+		return FALSE;
+	}
+
+	nouveau_bo_unmap(bo);
+	return TRUE;
+}
+
 void
 nv50_vbo_validate(struct nv50_context *nv50)
 {
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	struct nouveau_stateobj *vtxbuf, *vtxfmt;
-	int i;
+	struct nouveau_stateobj *vtxbuf, *vtxfmt, *vtxattr;
+	unsigned i;
+
+	/* don't validate if Gallium took away our buffers */
+	if (nv50->vtxbuf_nr == 0)
+		return;
 
-	vtxbuf = so_new(nv50->vtxelt_nr * 4, nv50->vtxelt_nr * 2);
+	vtxattr = NULL;
+	vtxbuf = so_new(nv50->vtxelt_nr * 7, nv50->vtxelt_nr * 4);
 	vtxfmt = so_new(nv50->vtxelt_nr + 1, 0);
-	so_method(vtxfmt, tesla, 0x1ac0, nv50->vtxelt_nr);
+	so_method(vtxfmt, tesla, NV50TCL_VERTEX_ARRAY_ATTRIB(0),
+		nv50->vtxelt_nr);
 
 	for (i = 0; i < nv50->vtxelt_nr; i++) {
 		struct pipe_vertex_element *ve = &nv50->vtxelt[i];
 		struct pipe_vertex_buffer *vb =
 			&nv50->vtxbuf[ve->vertex_buffer_index];
+		struct nouveau_bo *bo = nouveau_bo(vb->buffer);
+		uint32_t hw = nv50_vbo_vtxelt_to_hw(ve);
 
-		switch (ve->src_format) {
-		case PIPE_FORMAT_R32G32B32A32_FLOAT:
-			so_data(vtxfmt, 0x7e080000 | i);
-			break;
-		case PIPE_FORMAT_R32G32B32_FLOAT:
-			so_data(vtxfmt, 0x7e100000 | i);
-			break;
-		case PIPE_FORMAT_R32G32_FLOAT:
-			so_data(vtxfmt, 0x7e200000 | i);
-			break;
-		case PIPE_FORMAT_R32_FLOAT:
-			so_data(vtxfmt, 0x7e900000 | i);
-			break;
-		case PIPE_FORMAT_R8G8B8A8_UNORM:
-			so_data(vtxfmt, 0x24500000 | i);
-			break;
-		default:
-		{
-			NOUVEAU_ERR("invalid vbo format %s\n",
-				    pf_name(ve->src_format));
-			assert(0);
-			return;
-		}
+		if (!vb->stride &&
+		    nv50_vbo_static_attrib(nv50, i, &vtxattr, ve, vb)) {
+			so_data(vtxfmt, hw | (1 << 4));
+
+			so_method(vtxbuf, tesla,
+				  NV50TCL_VERTEX_ARRAY_FORMAT(i), 1);
+			so_data  (vtxbuf, 0);
+			continue;
 		}
+		so_data(vtxfmt, hw | i);
 
-		so_method(vtxbuf, tesla, 0x900 + (i * 16), 3);
+		so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_FORMAT(i), 3);
 		so_data  (vtxbuf, 0x20000000 | vb->stride);
-		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+		so_reloc (vtxbuf, bo, vb->buffer_offset +
 			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
 			  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+		so_reloc (vtxbuf, bo, vb->buffer_offset +
 			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
 			  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+
+		/* vertex array limits */
+		so_method(vtxbuf, tesla, 0x1080 + (i * 8), 2);
+		so_reloc (vtxbuf, bo, vb->buffer->size - 1,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD |
+			  NOUVEAU_BO_HIGH, 0, 0);
+		so_reloc (vtxbuf, bo, vb->buffer->size - 1,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD |
+			  NOUVEAU_BO_LOW, 0, 0);
 	}
 
 	so_ref (vtxfmt, &nv50->state.vtxfmt);
 	so_ref (vtxbuf, &nv50->state.vtxbuf);
+	so_ref (vtxattr, &nv50->state.vtxattr);
 	so_ref (NULL, &vtxbuf);
 	so_ref (NULL, &vtxfmt);
+	so_ref (NULL, &vtxattr);
 }
 
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index e44f9b9dfc..d7a2c8c462 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -4,21 +4,38 @@ include $(TOP)/configs/current
 LIBNAME = r300
 
 C_SOURCES = \
+	r3xx_fs.c \
+	r5xx_fs.c \
 	r300_chipset.c \
 	r300_clear.c \
 	r300_context.c \
-	r300_debug.c \
 	r300_emit.c \
 	r300_flush.c \
+	r300_fs.c \
 	r300_query.c \
 	r300_render.c \
 	r300_screen.c \
 	r300_state.c \
 	r300_state_derived.c \
 	r300_state_invariant.c \
-	r300_state_shader.c \
-	r300_state_tcl.c \
+	r300_vs.c \
 	r300_surface.c \
-	r300_texture.c
+	r300_texture.c \
+	r300_tgsi_to_rc.c
+
+LIBRARY_INCLUDES = \
+	-I$(TOP)/src/mesa/drivers/dri/r300/compiler \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include
+
+COMPILER_ARCHIVE = $(TOP)/src/mesa/drivers/dri/r300/compiler/libr300compiler.a
+
+EXTRA_OBJECTS = \
+	$(COMPILER_ARCHIVE)
 
 include ../../Makefile.template
+
+.PHONY : $(COMPILER_ARCHIVE)
+
+$(COMPILER_ARCHIVE):
+	cd $(TOP)/src/mesa/drivers/dri/r300/compiler; make
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index 182ed2d459..493d7b28bc 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -5,20 +5,22 @@ env = env.Clone()
 r300 = env.ConvenienceLibrary(
     target = 'r300',
     source = [
+        'r3xx_fs.c',
+        'r5xx_fs.c',
         'r300_chipset.c',
         'r300_clear.c',
         'r300_context.c',
         'r300_debug.c',
         'r300_emit.c',
         'r300_flush.c',
+        'r300_fs.c',
         'r300_query.c',
         'r300_render.c',
         'r300_screen.c',
         'r300_state.c',
         'r300_state_derived.c',
         'r300_state_invariant.c',
-        'r300_state_shader.c',
-        'r300_state_tcl.c',
+        'r300_vs.c',
         'r300_surface.c',
         'r300_texture.c',
     ])
diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c
index 9d95ad918c..d138866d33 100644
--- a/src/gallium/drivers/r300/r300_chipset.c
+++ b/src/gallium/drivers/r300/r300_chipset.c
@@ -30,9 +30,10 @@
 void r300_parse_chipset(struct r300_capabilities* caps)
 {
     /* Reasonable defaults */
+    caps->num_vert_fpus = 4;
     caps->has_tcl = getenv("RADEON_NO_TCL") ? FALSE : TRUE;
     caps->is_r500 = FALSE;
-    caps->num_vert_fpus = 4;
+    caps->high_second_pipe = FALSE;
 
 
     /* Note: These are not ordered by PCI ID. I leave that task to GCC,
@@ -41,6 +42,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
     switch (caps->pci_id) {
         case 0x4144:
             caps->family = CHIP_FAMILY_R300;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x4145:
@@ -51,6 +53,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x4E46:
         case 0x4E47:
             caps->family = CHIP_FAMILY_R300;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x4150:
@@ -67,6 +70,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x4E54:
         case 0x4E56:
             caps->family = CHIP_FAMILY_RV350;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x4148:
@@ -77,10 +81,12 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x4E49:
         case 0x4E4B:
             caps->family = CHIP_FAMILY_R350;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x4E4A:
             caps->family = CHIP_FAMILY_R360;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x5460:
@@ -92,6 +98,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x5B64:
         case 0x5B65:
             caps->family = CHIP_FAMILY_RV370;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x3150:
@@ -100,6 +107,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x3E50:
         case 0x3E54:
             caps->family = CHIP_FAMILY_RV380;
+            caps->high_second_pipe = TRUE;
             break;
 
         case 0x4A48:
@@ -150,6 +158,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->num_vert_fpus = 6;
             break;
 
+        case 0x4B48:
         case 0x4B49:
         case 0x4B4A:
         case 0x4B4B:
@@ -349,7 +358,4 @@ void r300_parse_chipset(struct r300_capabilities* caps)
                 caps->pci_id);
             break;
     }
-
-    /* XXX SW TCL is broken so no forcing it off right now
-    caps->has_tcl = FALSE; */
 }
diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h
index 21eebeae60..322d4a57e4 100644
--- a/src/gallium/drivers/r300/r300_chipset.h
+++ b/src/gallium/drivers/r300/r300_chipset.h
@@ -44,6 +44,8 @@ struct r300_capabilities {
      * - Blend color is split across two registers
      * - Universal Shader (US) block used for fragment shaders */
     boolean is_r500;
+    /* Whether or not the second pixel pipe is accessed with the high bit */
+    boolean high_second_pipe;
 };
 
 /* Enumerations for legibility and telling which card we're running on. */
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 6bdf544a05..da67bc29b8 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -34,10 +34,6 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     int i;
 
-    if (r300->dirty_state) {
-        r300_emit_dirty_state(r300);
-    }
-
     for (i = 0; i < r300->vertex_buffer_count; i++) {
         void* buf = pipe_buffer_map(pipe->screen,
                                     r300->vertex_buffers[i].buffer,
@@ -56,7 +52,7 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe,
 
     draw_set_mapped_constant_buffer(r300->draw,
             r300->shader_constants[PIPE_SHADER_VERTEX].constants,
-            r300->shader_constants[PIPE_SHADER_VERTEX].user_count *
+            r300->shader_constants[PIPE_SHADER_VERTEX].count *
                 (sizeof(float) * 4));
 
     draw_arrays(r300->draw, mode, start, count);
@@ -92,9 +88,21 @@ static boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
 
 static void r300_destroy_context(struct pipe_context* context) {
     struct r300_context* r300 = r300_context(context);
+    struct r300_query* query, * temp;
 
     draw_destroy(r300->draw);
 
+    /* Free the OQ BO. */
+    context->screen->buffer_destroy(r300->oqbo);
+
+    /* If there are any queries pending or not destroyed, remove them now. */
+    if (r300->query_list) {
+        foreach_s(query, temp, r300->query_list) {
+            remove_from_list(query);
+            FREE(query);
+        }
+    }
+
     FREE(r300->blend_color_state);
     FREE(r300->rs_block);
     FREE(r300->scissor_state);
@@ -133,7 +141,6 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     if (!r300)
         return NULL;
 
-    /* XXX this could be refactored now? */
     r300->winsys = r300_winsys;
 
     r300->context.winsys = (struct pipe_winsys*)r300_winsys;
@@ -150,14 +157,25 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->context.is_texture_referenced = r300_is_texture_referenced;
     r300->context.is_buffer_referenced = r300_is_buffer_referenced;
 
-    r300->draw = draw_create();
-    draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300));
-
     r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
     r300->viewport_state = CALLOC_STRUCT(r300_viewport_state);
 
+    /* Create a Draw. This is used for vert collation and SW TCL. */
+    r300->draw = draw_create();
+    /* Enable our renderer. */
+    draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300));
+    /* Disable Draw's clipping if TCL is present. */
+    draw_set_driver_clipping(r300->draw, r300_screen(screen)->caps->has_tcl);
+    /* Force Draw to never do viewport transform, since (again) we can do
+     * transform in hardware, always. */
+    draw_set_viewport_state(r300->draw, &r300_viewport_identity);
+
+    /* Open up the OQ BO. */
+    r300->oqbo = screen->buffer_create(screen, 4096,
+            PIPE_BUFFER_USAGE_VERTEX, 4096);
+
     r300_init_flush_functions(r300);
 
     r300_init_query_functions(r300);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 6f62998b35..f78492d4aa 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -25,15 +25,22 @@
 
 #include "draw/draw_context.h"
 #include "draw/draw_vertex.h"
+
 #include "pipe/p_context.h"
+
 #include "tgsi/tgsi_scan.h"
+
 #include "util/u_memory.h"
+#include "util/u_simple_list.h"
 
 #include "r300_clear.h"
 #include "r300_query.h"
 #include "r300_screen.h"
 #include "r300_winsys.h"
 
+struct r300_fragment_shader;
+struct r300_vertex_shader;
+
 struct r300_blend_state {
     uint32_t blend_control;       /* R300_RB3D_CBLEND: 0x4e04 */
     uint32_t alpha_blend_control; /* R300_RB3D_ABLEND: 0x4e08 */
@@ -63,6 +70,11 @@ struct r300_rs_state {
     /* Draw-specific rasterizer state */
     struct pipe_rasterizer_state rs;
 
+    /* Whether or not to enable the VTE. This is referenced at the very
+     * last moment during emission of VTE state, to decide whether or not
+     * the VTE should be used for transformation. */
+    boolean enable_vte;
+
     uint32_t vap_control_status;    /* R300_VAP_CNTL_STATUS: 0x2140 */
     uint32_t point_size;            /* R300_GA_POINT_SIZE: 0x421c */
     uint32_t point_minmax;          /* R300_GA_POINT_MINMAX: 0x4230 */
@@ -112,23 +124,24 @@ struct r300_viewport_state {
     uint32_t vte_control; /* R300_VAP_VTE_CNTL:      0x20b0 */
 };
 
-#define R300_NEW_BLEND           0x0000001
-#define R300_NEW_BLEND_COLOR     0x0000002
-#define R300_NEW_CONSTANTS       0x0000004
-#define R300_NEW_DSA             0x0000008
-#define R300_NEW_FRAMEBUFFERS    0x0000010
-#define R300_NEW_FRAGMENT_SHADER 0x0000020
-#define R300_NEW_RASTERIZER      0x0000040
-#define R300_NEW_RS_BLOCK        0x0000080
-#define R300_NEW_SAMPLER         0x0000100
-#define R300_ANY_NEW_SAMPLERS    0x000ff00
-#define R300_NEW_SCISSOR         0x0010000
-#define R300_NEW_TEXTURE         0x0020000
-#define R300_ANY_NEW_TEXTURES    0x1fe0000
-#define R300_NEW_VERTEX_FORMAT   0x2000000
-#define R300_NEW_VERTEX_SHADER   0x4000000
-#define R300_NEW_VIEWPORT        0x8000000
-#define R300_NEW_KITCHEN_SINK    0xfffffff
+#define R300_NEW_BLEND           0x00000001
+#define R300_NEW_BLEND_COLOR     0x00000002
+#define R300_NEW_CLIP            0x00000004
+#define R300_NEW_CONSTANTS       0x00000008
+#define R300_NEW_DSA             0x00000010
+#define R300_NEW_FRAMEBUFFERS    0x00000020
+#define R300_NEW_FRAGMENT_SHADER 0x00000040
+#define R300_NEW_RASTERIZER      0x00000080
+#define R300_NEW_RS_BLOCK        0x00000100
+#define R300_NEW_SAMPLER         0x00000200
+#define R300_ANY_NEW_SAMPLERS    0x0001fe00
+#define R300_NEW_SCISSOR         0x00020000
+#define R300_NEW_TEXTURE         0x00040000
+#define R300_ANY_NEW_TEXTURES    0x03fc0000
+#define R300_NEW_VERTEX_FORMAT   0x04000000
+#define R300_NEW_VERTEX_SHADER   0x08000000
+#define R300_NEW_VIEWPORT        0x10000000
+#define R300_NEW_KITCHEN_SINK    0x1fffffff
 
 /* The next several objects are not pure Radeon state; they inherit from
  * various Gallium classes. */
@@ -136,66 +149,32 @@ struct r300_viewport_state {
 struct r300_constant_buffer {
     /* Buffer of constants */
     /* XXX first number should be raised */
-    float constants[8][4];
-    /* Number of user-defined constants */
-    int user_count;
+    float constants[32][4];
     /* Total number of constants */
-    int count;
-};
-
-struct r3xx_fragment_shader {
-    /* Parent class */
-    struct pipe_shader_state state;
-    struct tgsi_shader_info info;
-
-    /* Has this shader been translated yet? */
-    boolean translated;
-
-    /* Pixel stack size */
-    int stack_size;
-};
-
-struct r300_fragment_shader {
-    /* Parent class */
-    struct r3xx_fragment_shader shader;
-
-    /* Number of ALU instructions */
-    int alu_instruction_count;
-
-    /* Number of texture instructions */
-    int tex_instruction_count;
-
-    /* Number of texture indirections */
-    int indirections;
-
-    /* Indirection node offsets */
-    int alu_offset[4];
-
-    /* Machine instructions */
-    struct {
-        uint32_t alu_rgb_inst;
-        uint32_t alu_rgb_addr;
-        uint32_t alu_alpha_inst;
-        uint32_t alu_alpha_addr;
-    } instructions[64]; /* XXX magic num */
+    unsigned count;
 };
 
-struct r500_fragment_shader {
-    /* Parent class */
-    struct r3xx_fragment_shader shader;
-
-    /* Number of used instructions */
-    int instruction_count;
-
-    /* Machine instructions */
-    struct {
-        uint32_t inst0;
-        uint32_t inst1;
-        uint32_t inst2;
-        uint32_t inst3;
-        uint32_t inst4;
-        uint32_t inst5;
-    } instructions[256]; /*< XXX magic number */
+/* Query object.
+ *
+ * This is not a subclass of pipe_query because pipe_query is never
+ * actually fully defined. So, rather than have it as a member, and do
+ * subclass-style casting, we treat pipe_query as an opaque, and just
+ * trust that our state tracker does not ever mess up query objects.
+ */
+struct r300_query {
+    /* The kind of query. Currently only OQ is supported. */
+    unsigned type;
+    /* Whether this query is currently active. Only active queries will
+     * get emitted into the command stream, and only active queries get
+     * tallied. */
+    boolean active;
+    /* The current count of this query. Required to be at least 32 bits. */
+    unsigned int count;
+    /* The offset of this query into the query buffer, in bytes. */
+    unsigned offset;
+    /* Linked list members. */
+    struct r300_query* prev;
+    struct r300_query* next;
 };
 
 struct r300_texture {
@@ -207,7 +186,7 @@ struct r300_texture {
 
     /* Stride (pitch?) of this texture in bytes */
     unsigned stride;
-    
+
     /* Total size of this texture, in bytes. */
     unsigned size;
 
@@ -232,27 +211,9 @@ struct r300_vertex_format {
     int fs_tab[16];
 };
 
-struct r300_vertex_shader {
-    /* Parent class */
-    struct pipe_shader_state state;
-    struct tgsi_shader_info info;
-
-    /* Fallback shader, because Draw has issues */
-    struct draw_vertex_shader* draw;
-
-    /* Has this shader been translated yet? */
-    boolean translated;
-
-    /* Number of used instructions */
-    int instruction_count;
-
-    /* Machine instructions */
-    struct {
-        uint32_t inst0;
-        uint32_t inst1;
-        uint32_t inst2;
-        uint32_t inst3;
-    } instructions[128]; /*< XXX magic number */
+static struct pipe_viewport_state r300_viewport_identity = {
+    .scale = {1.0, 1.0, 1.0, 1.0},
+    .translate = {0.0, 0.0, 0.0, 0.0},
 };
 
 struct r300_context {
@@ -264,17 +225,29 @@ struct r300_context {
     /* Draw module. Used mostly for SW TCL. */
     struct draw_context* draw;
 
+    /* Vertex buffer for rendering. */
+    struct pipe_buffer* vbo;
+    /* Offset into the VBO. */
+    size_t vbo_offset;
+
+    /* Occlusion query buffer. */
+    struct pipe_buffer* oqbo;
+    /* Query list. */
+    struct r300_query* query_list;
+
     /* Various CSO state objects. */
     /* Blend state. */
     struct r300_blend_state* blend_state;
     /* Blend color state. */
     struct r300_blend_color_state* blend_color_state;
+    /* User clip planes. */
+    struct pipe_clip_state clip_state;
     /* Shader constants. */
     struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES];
     /* Depth, stencil, and alpha state. */
     struct r300_dsa_state* dsa_state;
     /* Fragment shader. */
-    struct r3xx_fragment_shader* fs;
+    struct r300_fragment_shader* fs;
     /* Framebuffer state. We currently don't need our own version of this. */
     struct pipe_framebuffer_state framebuffer_state;
     /* Rasterizer state. */
@@ -289,7 +262,7 @@ struct r300_context {
     /* Texture states. */
     struct r300_texture* textures[8];
     int texture_count;
-    /* Vertex buffers. */
+    /* Vertex buffers for Gallium. */
     struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
     int vertex_buffer_count;
     /* Vertex information. */
@@ -305,7 +278,8 @@ struct r300_context {
 };
 
 /* Convenience cast wrapper. */
-static struct r300_context* r300_context(struct pipe_context* context) {
+static INLINE struct r300_context* r300_context(struct pipe_context* context)
+{
     return (struct r300_context*)context;
 }
 
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 82a3942248..71b142c0db 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -34,6 +34,7 @@
 
 #define MAX_CS_SIZE 64 * 1024 / 4
 
+#define VERY_VERBOSE_CS 0
 #define VERY_VERBOSE_REGISTERS 0
 
 /* XXX stolen from radeon_drm.h */
@@ -56,8 +57,10 @@
 
 #define BEGIN_CS(size) do { \
     CHECK_CS(size); \
-    debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
-        size, __FUNCTION__, __FILE__, __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
+                size, __FUNCTION__, __FILE__, __LINE__); \
+    } \
     cs_winsys->begin_cs(cs_winsys, (size), \
             __FILE__, __FUNCTION__, __LINE__); \
     cs_count = size; \
@@ -93,8 +96,9 @@
 } while (0)
 
 #define OUT_CS_RELOC(bo, offset, rd, wd, flags) do { \
-    debug_printf("r300: writing relocation for buffer %p, offset %d\n", \
-        bo, offset); \
+    debug_printf("r300: writing relocation for buffer %p, offset %d, " \
+            "domains (%d, %d, %d)\n", \
+        bo, offset, rd, wd, flags); \
     assert(bo); \
     OUT_CS(offset); \
     cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
@@ -102,16 +106,20 @@
 } while (0)
 
 #define END_CS do { \
-    debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \
-        __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \
+                __FILE__, __LINE__); \
+    } \
     if (cs_count != 0) \
         debug_printf("r300: Warning: cs_count off by %d\n", cs_count); \
     cs_winsys->end_cs(cs_winsys, __FILE__, __FUNCTION__, __LINE__); \
 } while (0)
 
 #define FLUSH_CS do { \
-    debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, __FILE__, \
-        __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, \
+                __FILE__, __LINE__); \
+    } \
     cs_winsys->flush_cs(cs_winsys); \
 } while (0)
 
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
deleted file mode 100644
index dd63136c9d..0000000000
--- a/src/gallium/drivers/r300/r300_debug.c
+++ /dev/null
@@ -1,238 +0,0 @@
-/*
- * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE. */
-
-#include "r300_debug.h"
-
-static void r300_dump_fs(struct r300_fragment_shader* fs)
-{
-    int i;
-
-    for (i = 0; i < fs->alu_instruction_count; i++) {
-    }
-}
-
-static char* r500_fs_swiz[] = {
-    " R",
-    " G",
-    " B",
-    " A",
-    " 0",
-    ".5",
-    " 1",
-    " U",
-};
-
-static char* r500_fs_op_rgb[] = {
-    "MAD",
-    "DP3",
-    "DP4",
-    "D2A",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "SOP",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_op_alpha[] = {
-    "MAD",
-    " DP",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "EX2",
-    "LN2",
-    "RCP",
-    "RSQ",
-    "SIN",
-    "COS",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_mask[] = {
-    "NONE",
-    "R   ",
-    " G  ",
-    "RG  ",
-    "  B ",
-    "R B ",
-    " GB ",
-    "RGB ",
-    "   A",
-    "R  A",
-    " G A",
-    "RG A",
-    "  BA",
-    "R BA",
-    " GBA",
-    "RGBA",
-};
-
-static char* r500_fs_tex[] = {
-    "    NOP",
-    "     LD",
-    "TEXKILL",
-    "   PROJ",
-    "LODBIAS",
-    "    LOD",
-    "   DXDY",
-};
-
-void r500_fs_dump(struct r500_fragment_shader* fs)
-{
-    int i;
-    uint32_t inst;
-
-    for (i = 0; i < fs->instruction_count; i++) {
-        inst = fs->instructions[i].inst0;
-        debug_printf("%d:  0: CMN_INST   0x%08x:", i, inst);
-        switch (inst & 0x3) {
-            case R500_INST_TYPE_ALU:
-                debug_printf("ALU ");
-                break;
-            case R500_INST_TYPE_OUT:
-                debug_printf("OUT ");
-                break;
-            case R500_INST_TYPE_FC:
-                debug_printf("FC  ");
-                break;
-            case R500_INST_TYPE_TEX:
-                debug_printf("TEX ");
-                break;
-        }
-        debug_printf("%s %s %s %s ",
-                inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "",
-                inst & R500_INST_LAST ? "LAST" : "",
-                inst & R500_INST_NOP ? "NOP" : "",
-                inst & R500_INST_ALU_WAIT ? "ALU_WAIT" : "");
-        debug_printf("wmask: %s omask: %s\n",
-                r500_fs_mask[(inst >> 11) & 0xf],
-                r500_fs_mask[(inst >> 15) & 0xf]);
-        switch (inst & 0x3) {
-            case R500_INST_TYPE_ALU:
-            case R500_INST_TYPE_OUT:
-                inst = fs->instructions[i].inst1;
-                debug_printf("    1: RGB_ADDR   0x%08x:", inst);
-                debug_printf("Addr0: %d%c, Addr1: %d%c, "
-                        "Addr2: %d%c, srcp:%d\n",
-                        inst & 0xff, (inst & (1 << 8)) ? 'c' : 't',
-                        (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't',
-                        (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't',
-                        (inst >> 30));
-
-                inst = fs->instructions[i].inst2;
-                debug_printf("    2: ALPHA_ADDR 0x%08x:", inst);
-                debug_printf("Addr0: %d%c, Addr1: %d%c, "
-                        "Addr2: %d%c, srcp:%d\n",
-                        inst & 0xff, (inst & (1 << 8)) ? 'c' : 't',
-                        (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't',
-                        (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't',
-                        (inst >> 30));
-
-                inst = fs->instructions[i].inst3;
-                debug_printf("    3: RGB_INST   0x%08x:", inst);
-                debug_printf("rgb_A_src:%d %s/%s/%s %d "
-                        "rgb_B_src:%d %s/%s/%s %d\n",
-                        inst & 0x3, r500_fs_swiz[(inst >> 2) & 0x7],
-                        r500_fs_swiz[(inst >> 5) & 0x7],
-                        r500_fs_swiz[(inst >> 8) & 0x7],
-                        (inst >> 11) & 0x3, (inst >> 13) & 0x3,
-                        r500_fs_swiz[(inst >> 15) & 0x7],
-                        r500_fs_swiz[(inst >> 18) & 0x7],
-                        r500_fs_swiz[(inst >> 21) & 0x7],
-                        (inst >> 24) & 0x3);
-
-                inst = fs->instructions[i].inst4;
-                debug_printf("    4: ALPHA_INST 0x%08x:", inst);
-                debug_printf("%s dest:%d%s alp_A_src:%d %s %d "
-                        "alp_B_src:%d %s %d w:%d\n",
-                        r500_fs_op_alpha[inst & 0xf], (inst >> 4) & 0x7f,
-                        inst & (1<<11) ? "(rel)":"", (inst >> 12) & 0x3,
-                        r500_fs_swiz[(inst >> 14) & 0x7], (inst >> 17) & 0x3,
-                        (inst >> 19) & 0x3, r500_fs_swiz[(inst >> 21) & 0x7],
-                        (inst >> 24) & 0x3, (inst >> 31) & 0x1);
-
-                inst = fs->instructions[i].inst5;
-                debug_printf("    5: RGBA_INST  0x%08x:", inst);
-                debug_printf("%s dest:%d%s rgb_C_src:%d %s/%s/%s %d "
-                        "alp_C_src:%d %s %d\n",
-                        r500_fs_op_rgb[inst & 0xf], (inst >> 4) & 0x7f,
-                        inst & (1 << 11) ? "(rel)":"", (inst >> 12) & 0x3,
-                        r500_fs_swiz[(inst >> 14) & 0x7],
-                        r500_fs_swiz[(inst >> 17) & 0x7],
-                        r500_fs_swiz[(inst >> 20) & 0x7],
-                        (inst >> 23) & 0x3, (inst >> 25) & 0x3,
-                        r500_fs_swiz[(inst >> 27) & 0x7], (inst >> 30) & 0x3);
-                break;
-            case R500_INST_TYPE_FC:
-                /* XXX don't even bother yet */
-                break;
-            case R500_INST_TYPE_TEX:
-                inst = fs->instructions[i].inst1;
-                debug_printf("    1: TEX_INST   0x%08x: id: %d "
-                        "op:%s, %s, %s %s\n",
-                        inst, (inst >> 16) & 0xf,
-                        r500_fs_tex[(inst >> 22) & 0x7],
-                        (inst & (1 << 25)) ? "ACQ" : "",
-                        (inst & (1 << 26)) ? "IGNUNC" : "",
-                        (inst & (1 << 27)) ? "UNSCALED" : "SCALED");
-
-                inst = fs->instructions[i].inst2;
-                debug_printf("    2: TEX_ADDR   0x%08x: "
-                        "src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n",
-                        inst, inst & 0x7f, inst & (1 << 7) ? "(rel)" : "",
-                        r500_fs_swiz[(inst >> 8) & 0x3],
-                        r500_fs_swiz[(inst >> 10) & 0x3],
-                        r500_fs_swiz[(inst >> 12) & 0x3],
-                        r500_fs_swiz[(inst >> 14) & 0x3],
-                        (inst >> 16) & 0x7f, inst & (1 << 23) ? "(rel)" : "",
-                        r500_fs_swiz[(inst >> 24) & 0x3],
-                        r500_fs_swiz[(inst >> 26) & 0x3],
-                        r500_fs_swiz[(inst >> 28) & 0x3],
-                        r500_fs_swiz[(inst >> 30) & 0x3]);
-                
-                inst = fs->instructions[i].inst3;
-                debug_printf("    3: TEX_DXDY   0x%08x\n", inst);
-                break;
-        }
-    }
-}
-
-void r300_vs_dump(struct r300_vertex_shader* vs)
-{
-    int i;
-
-    for (i = 0; i < vs->instruction_count; i++) {
-        debug_printf("inst0: 0x%x\n", vs->instructions[i].inst0);
-        debug_printf("inst1: 0x%x\n", vs->instructions[i].inst1);
-        debug_printf("inst2: 0x%x\n", vs->instructions[i].inst2);
-        debug_printf("inst3: 0x%x\n", vs->instructions[i].inst3);
-    }
-}
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 01bac5f759..bd4d59e6f1 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -24,6 +24,9 @@
 
 #include "r300_emit.h"
 
+#include "r300_fs.h"
+#include "r300_vs.h"
+
 void r300_emit_blend_state(struct r300_context* r300,
                            struct r300_blend_state* blend)
 {
@@ -56,6 +59,36 @@ void r300_emit_blend_color_state(struct r300_context* r300,
     }
 }
 
+void r300_emit_clip_state(struct r300_context* r300,
+                          struct pipe_clip_state* clip)
+{
+    int i;
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    CS_LOCALS(r300);
+
+    if (r300screen->caps->has_tcl) {
+        BEGIN_CS(5 + (6 * 4));
+        OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
+                (r300screen->caps->is_r500 ?
+                 R500_PVS_UCP_START : R300_PVS_UCP_START));
+        OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, 6 * 4);
+        for (i = 0; i < 6; i++) {
+            OUT_CS_32F(clip->ucp[i][0]);
+            OUT_CS_32F(clip->ucp[i][1]);
+            OUT_CS_32F(clip->ucp[i][2]);
+            OUT_CS_32F(clip->ucp[i][3]);
+        }
+        OUT_CS_REG(R300_VAP_CLIP_CNTL, ((1 << clip->nr) - 1) |
+                R300_PS_UCP_MODE_CLIP_AS_TRIFAN);
+        END_CS;
+    } else {
+        BEGIN_CS(2);
+        OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE);
+        END_CS;
+    }
+
+}
+
 void r300_emit_dsa_state(struct r300_context* r300,
                            struct r300_dsa_state* dsa)
 {
@@ -79,73 +112,158 @@ void r300_emit_dsa_state(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_fragment_shader(struct r300_context* r300,
-                               struct r300_fragment_shader* fs)
+static const float * get_shader_constant(
+    struct r300_context * r300,
+    struct rc_constant * constant,
+    struct r300_constant_buffer * externals)
+{
+    static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 };
+    switch(constant->Type) {
+        case RC_CONSTANT_EXTERNAL:
+            return externals->constants[constant->u.External];
+
+        case RC_CONSTANT_IMMEDIATE:
+            return constant->u.Immediate;
+
+        default:
+            debug_printf("r300: Implementation error: Unhandled constant type %i\n",
+                constant->Type);
+            return zero;
+    }
+}
+
+/* Convert a normal single-precision float into the 7.16 format
+ * used by the R300 fragment shader.
+ */
+static uint32_t pack_float24(float f)
+{
+    union {
+        float fl;
+        uint32_t u;
+    } u;
+    float mantissa;
+    int exponent;
+    uint32_t float24 = 0;
+
+    if (f == 0.0)
+        return 0;
+
+    u.fl = f;
+
+    mantissa = frexpf(f, &exponent);
+
+    /* Handle -ve */
+    if (mantissa < 0) {
+        float24 |= (1 << 23);
+        mantissa = mantissa * -1.0;
+    }
+    /* Handle exponent, bias of 63 */
+    exponent += 62;
+    float24 |= (exponent << 16);
+    /* Kill 7 LSB of mantissa */
+    float24 |= (u.u & 0x7FFFFF) >> 7;
+
+    return float24;
+}
+
+void r300_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code,
+                                     struct r300_constant_buffer* externals)
 {
+    struct r300_fragment_program_code * code = &generic_code->code.r300;
+    struct rc_constant_list * constants = &generic_code->constants;
     int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS(22);
-
-    OUT_CS_REG(R300_US_CONFIG, fs->indirections);
-    OUT_CS_REG(R300_US_PIXSIZE, fs->shader.stack_size);
-    /* XXX figure out exactly how big the sizes are on this reg */
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x40);
-    /* XXX figure these ones out a bit better kthnx */
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x40 | R300_RGBA_OUT);
-
-    for (i = 0; i < fs->alu_instruction_count; i++) {
-        OUT_CS_REG(R300_US_ALU_RGB_INST_0 + (4 * i),
-            fs->instructions[i].alu_rgb_inst);
-        OUT_CS_REG(R300_US_ALU_RGB_ADDR_0 + (4 * i),
-            fs->instructions[i].alu_rgb_addr);
-        OUT_CS_REG(R300_US_ALU_ALPHA_INST_0 + (4 * i),
-            fs->instructions[i].alu_alpha_inst);
-        OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0 + (4 * i),
-            fs->instructions[i].alu_alpha_addr);
+    BEGIN_CS(15 +
+             code->alu.length * 4 +
+             (code->tex.length ? (1 + code->tex.length) : 0) +
+             (constants->Count ? (1 + constants->Count * 4) : 0));
+
+    OUT_CS_REG(R300_US_CONFIG, code->config);
+    OUT_CS_REG(R300_US_PIXSIZE, code->pixsize);
+    OUT_CS_REG(R300_US_CODE_OFFSET, code->code_offset);
+
+    OUT_CS_REG_SEQ(R300_US_CODE_ADDR_0, 4);
+    for(i = 0; i < 4; ++i)
+        OUT_CS(code->code_addr[i]);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_RGB_INST_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].rgb_inst);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].rgb_addr);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].alpha_inst);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].alpha_addr);
+
+    if (code->tex.length) {
+        OUT_CS_REG_SEQ(R300_US_TEX_INST_0, code->tex.length);
+        for(i = 0; i < code->tex.length; ++i)
+            OUT_CS(code->tex.inst[i]);
+    }
+
+    if (constants->Count) {
+        OUT_CS_ONE_REG(R300_PFS_PARAM_0_X, constants->Count * 4);
+        for(i = 0; i < constants->Count; ++i) {
+            const float * data = get_shader_constant(r300, &constants->Constants[i], externals);
+            OUT_CS(pack_float24(data[0]));
+            OUT_CS(pack_float24(data[1]));
+            OUT_CS(pack_float24(data[2]));
+            OUT_CS(pack_float24(data[3]));
+        }
     }
 
     END_CS;
 }
 
-void r500_emit_fragment_shader(struct r300_context* r300,
-                               struct r500_fragment_shader* fs)
+void r500_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code,
+                                     struct r300_constant_buffer* externals)
 {
+    struct r500_fragment_program_code * code = &generic_code->code.r500;
+    struct rc_constant_list * constants = &generic_code->constants;
     int i;
-    struct r300_constant_buffer* constants =
-        &r300->shader_constants[PIPE_SHADER_FRAGMENT];
     CS_LOCALS(r300);
 
-    BEGIN_CS(9 + (fs->instruction_count * 6) + (constants->count ? 3 : 0) +
-            (constants->count * 4));
-    OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
-    OUT_CS_REG(R500_US_PIXSIZE, fs->shader.stack_size);
-    OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) |
-            R500_US_CODE_END_ADDR(fs->instruction_count));
+    BEGIN_CS(13 +
+             ((code->inst_end + 1) * 6) +
+             (constants->Count ? (3 + (constants->Count * 4)) : 0));
+    OUT_CS_REG(R500_US_CONFIG, 0);
+    OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx);
+    OUT_CS_REG(R500_US_CODE_RANGE,
+               R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end));
+    OUT_CS_REG(R500_US_CODE_OFFSET, 0);
+    OUT_CS_REG(R500_US_CODE_ADDR,
+               R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(code->inst_end));
 
     OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR);
-    OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, fs->instruction_count * 6);
-    for (i = 0; i < fs->instruction_count; i++) {
-        OUT_CS(fs->instructions[i].inst0);
-        OUT_CS(fs->instructions[i].inst1);
-        OUT_CS(fs->instructions[i].inst2);
-        OUT_CS(fs->instructions[i].inst3);
-        OUT_CS(fs->instructions[i].inst4);
-        OUT_CS(fs->instructions[i].inst5);
-    }
-
-    if (constants->count) {
-        OUT_CS_REG(R500_GA_US_VECTOR_INDEX,
-                R500_GA_US_VECTOR_INDEX_TYPE_CONST);
-        OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->count * 4);
-        for (i = 0; i < constants->count; i++) {
-            OUT_CS_32F(constants->constants[i][0]);
-            OUT_CS_32F(constants->constants[i][1]);
-            OUT_CS_32F(constants->constants[i][2]);
-            OUT_CS_32F(constants->constants[i][3]);
+    OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, (code->inst_end + 1) * 6);
+    for (i = 0; i <= code->inst_end; i++) {
+        OUT_CS(code->inst[i].inst0);
+        OUT_CS(code->inst[i].inst1);
+        OUT_CS(code->inst[i].inst2);
+        OUT_CS(code->inst[i].inst3);
+        OUT_CS(code->inst[i].inst4);
+        OUT_CS(code->inst[i].inst5);
+    }
+
+    if (constants->Count) {
+        OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST);
+        OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4);
+        for (i = 0; i < constants->Count; i++) {
+            const float * data = get_shader_constant(r300, &constants->Constants[i], externals);
+            OUT_CS_32F(data[0]);
+            OUT_CS_32F(data[1]);
+            OUT_CS_32F(data[2]);
+            OUT_CS_32F(data[3]);
         }
     }
 
@@ -160,16 +278,19 @@ void r300_emit_fb_state(struct r300_context* r300,
     int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4);
+    BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4);
     for (i = 0; i < fb->nr_cbufs; i++) {
         tex = (struct r300_texture*)fb->cbufs[i]->texture;
+        assert(tex && tex->buffer && "cbuf is marked, but NULL!");
         pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
         OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
-        OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), pixpitch |
-            r300_translate_colorformat(tex->tex.format));
+        OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1);
+        OUT_CS_RELOC(tex->buffer, pixpitch |
+                     r300_translate_colorformat(tex->tex.format), 0,
+                     RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i),
             r300_translate_out_fmt(fb->cbufs[i]->format));
@@ -177,14 +298,16 @@ void r300_emit_fb_state(struct r300_context* r300,
 
     if (fb->zsbuf) {
         tex = (struct r300_texture*)fb->zsbuf->texture;
-        pixpitch = (tex->stride / tex->tex.block.size);
+        assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
+        pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1);
         OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format));
 
-        OUT_CS_REG(R300_ZB_DEPTHPITCH, pixpitch);
+        OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1);
+        OUT_CS_RELOC(tex->buffer, pixpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
     }
 
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
@@ -196,6 +319,79 @@ void r300_emit_fb_state(struct r300_context* r300,
     END_CS;
 }
 
+void r300_emit_query_begin(struct r300_context* r300,
+                           struct r300_query* query)
+{
+    CS_LOCALS(r300);
+
+    /* XXX This will almost certainly not return good results
+     * for overlapping queries. */
+    BEGIN_CS(2);
+    OUT_CS_REG(R300_ZB_ZPASS_DATA, 0);
+    END_CS;
+}
+
+void r300_emit_query_end(struct r300_context* r300,
+                         struct r300_query* query)
+{
+    struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
+    CS_LOCALS(r300);
+
+    if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo,
+                0, RADEON_GEM_DOMAIN_GTT)) {
+        debug_printf("r300: There wasn't room for the OQ buffer!?"
+                " Oh noes!\n");
+    }
+
+    assert(caps->num_frag_pipes);
+    BEGIN_CS(6 * caps->num_frag_pipes + 2);
+    /* I'm not so sure I like this switch, but it's hard to be elegant
+     * when there's so many special cases...
+     *
+     * So here's the basic idea. For each pipe, enable writes to it only,
+     * then put out the relocation for ZPASS_ADDR, taking into account a
+     * 4-byte offset for each pipe. RV380 and older are special; they have
+     * only two pipes, and the second pipe's enable is on bit 3, not bit 1,
+     * so there's a chipset cap for that. */
+    switch (caps->num_frag_pipes) {
+        case 4:
+            /* pipe 3 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 3);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 3),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 3:
+            /* pipe 2 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 2);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 2),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 2:
+            /* pipe 1 only */
+            /* As mentioned above, accomodate RV380 and older. */
+            OUT_CS_REG(R300_SU_REG_DEST,
+                    1 << (caps->high_second_pipe ? 3 : 1));
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 1),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 1:
+            /* pipe 0 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 0);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        default:
+            debug_printf("r300: Implementation error: Chipset reports %d"
+                    " pixel pipes!\n", caps->num_frag_pipes);
+            assert(0);
+    }
+
+    /* And, finally, reset it to normal... */
+    OUT_CS_REG(R300_SU_REG_DEST, 0xF);
+    END_CS;
+
+}
+
 void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
 {
     CS_LOCALS(r300);
@@ -234,7 +430,7 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->ip[i]);
-        debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]);
+        /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */
     }
 
     OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
@@ -248,27 +444,15 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->inst[i]);
-        debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]);
+        /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */
     }
 
-    debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
-            rs->inst_count);
+    /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
+     *        rs->inst_count); */
 
     END_CS;
 }
 
-void r300_emit_sampler(struct r300_context* r300,
-                       struct r300_sampler_state* sampler, unsigned offset)
-{
-    CS_LOCALS(r300);
-
-    BEGIN_CS(6);
-    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
-    OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
-    OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
-    END_CS;
-}
-
 void r300_emit_scissor_state(struct r300_context* r300,
                              struct r300_scissor_state* scissor)
 {
@@ -282,11 +466,17 @@ void r300_emit_scissor_state(struct r300_context* r300,
 }
 
 void r300_emit_texture(struct r300_context* r300,
-                       struct r300_texture* tex, unsigned offset)
+                       struct r300_sampler_state* sampler,
+                       struct r300_texture* tex,
+                       unsigned offset)
 {
     CS_LOCALS(r300);
 
-    BEGIN_CS(10);
+    BEGIN_CS(16);
+    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
+    OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
+    OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
+
     OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0);
     OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1);
     OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2);
@@ -296,6 +486,30 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
+void r300_emit_vertex_buffer(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    debug_printf("r300: Preparing vertex buffer %p for render, "
+            "vertex size %d\n", r300->vbo,
+            r300->vertex_info.vinfo.size);
+    /* Set the pointer to our vertex buffer. The emitted values are this:
+     * PACKET3 [3D_LOAD_VBPNTR]
+     * COUNT   [1]
+     * FORMAT  [size | stride << 8]
+     * OFFSET  [offset into BO]
+     * VBPNTR  [relocated BO]
+     */
+    BEGIN_CS(7);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
+    OUT_CS(1);
+    OUT_CS(r300->vertex_info.vinfo.size |
+            (r300->vertex_info.vinfo.size << 8));
+    OUT_CS(r300->vbo_offset);
+    OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
+    END_CS;
+}
+
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
     int i;
@@ -310,33 +524,33 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
     OUT_CS(r300->vertex_info.vinfo.hwfmt[2]);
     OUT_CS(r300->vertex_info.vinfo.hwfmt[3]);
-    for (i = 0; i < 4; i++) {
-        debug_printf("hwfmt%d: 0x%08x\n", i,
-                r300->vertex_info.vinfo.hwfmt[i]);
-    }
+    /* for (i = 0; i < 4; i++) {
+     *    debug_printf("hwfmt%d: 0x%08x\n", i,
+     *            r300->vertex_info.vinfo.hwfmt[i]);
+     * } */
 
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]);
-        debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
-                r300->vertex_info.vap_prog_stream_cntl[i]);
+        /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
+         *        r300->vertex_info.vap_prog_stream_cntl[i]); */
     }
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]);
-        debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
-                r300->vertex_info.vap_prog_stream_cntl_ext[i]);
+        /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
+         *        r300->vertex_info.vap_prog_stream_cntl_ext[i]); */
     }
     END_CS;
 }
 
-void r300_emit_vertex_shader(struct r300_context* r300,
-                             struct r300_vertex_shader* vs)
+void r300_emit_vertex_program_code(struct r300_context* r300,
+                                   struct r300_vertex_program_code* code,
+                                   struct r300_constant_buffer* constants)
 {
     int i;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
-    struct r300_constant_buffer* constants =
-        &r300->shader_constants[PIPE_SHADER_VERTEX];
+    unsigned instruction_count = code->length / 4;
     CS_LOCALS(r300);
 
     if (!r300screen->caps->has_tcl) {
@@ -345,38 +559,40 @@ void r300_emit_vertex_shader(struct r300_context* r300,
         return;
     }
 
-    if (constants->count) {
-        BEGIN_CS(16 + (vs->instruction_count * 4) + (constants->count * 4));
+    if (code->constants.Count) {
+        BEGIN_CS(14 + code->length + (code->constants.Count * 4));
     } else {
-        BEGIN_CS(13 + (vs->instruction_count * 4) + (constants->count * 4));
+        BEGIN_CS(11 + code->length);
     }
 
-    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) |
-            R300_PVS_LAST_INST(vs->instruction_count - 1));
-    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, vs->instruction_count - 1);
-
-    /* XXX */
-    OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x0);
+    /* R300_VAP_PVS_CODE_CNTL_0
+     * R300_VAP_PVS_CONST_CNTL
+     * R300_VAP_PVS_CODE_CNTL_1
+     * See the r5xx docs for instructions on how to use these.
+     * XXX these could be optimized to select better values... */
+    OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
+    OUT_CS(R300_PVS_FIRST_INST(0) |
+            R300_PVS_XYZW_VALID_INST(instruction_count - 1) |
+            R300_PVS_LAST_INST(instruction_count - 1));
+    OUT_CS(R300_PVS_MAX_CONST_ADDR(code->constants.Count - 1));
+    OUT_CS(instruction_count - 1);
 
     OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0);
-    OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4);
-    for (i = 0; i < vs->instruction_count; i++) {
-        OUT_CS(vs->instructions[i].inst0);
-        OUT_CS(vs->instructions[i].inst1);
-        OUT_CS(vs->instructions[i].inst2);
-        OUT_CS(vs->instructions[i].inst3);
-    }
+    OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->length);
+    for (i = 0; i < code->length; i++)
+        OUT_CS(code->body.d[i]);
 
-    if (constants->count) {
+    if (code->constants.Count) {
         OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
                 (r300screen->caps->is_r500 ?
                  R500_PVS_CONST_START : R300_PVS_CONST_START));
-        OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->count * 4);
-        for (i = 0; i < constants->count; i++) {
-            OUT_CS_32F(constants->constants[i][0]);
-            OUT_CS_32F(constants->constants[i][1]);
-            OUT_CS_32F(constants->constants[i][2]);
-            OUT_CS_32F(constants->constants[i][3]);
+        OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->constants.Count * 4);
+        for (i = 0; i < code->constants.Count; i++) {
+            const float * data = get_shader_constant(r300, &code->constants.Constants[i], constants);
+            OUT_CS_32F(data[0]);
+            OUT_CS_32F(data[1]);
+            OUT_CS_32F(data[2]);
+            OUT_CS_32F(data[3]);
         }
     }
 
@@ -386,7 +602,12 @@ void r300_emit_vertex_shader(struct r300_context* r300,
             R300_PVS_VF_MAX_VTX_NUM(12));
     OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
     END_CS;
+}
 
+void r300_emit_vertex_shader(struct r300_context* r300,
+                             struct r300_vertex_shader* vs)
+{
+    r300_emit_vertex_program_code(r300, &vs->code, &r300->shader_constants[PIPE_SHADER_VERTEX]);
 }
 
 void r300_emit_viewport_state(struct r300_context* r300,
@@ -403,7 +624,11 @@ void r300_emit_viewport_state(struct r300_context* r300,
     OUT_CS_32F(viewport->zscale);
     OUT_CS_32F(viewport->zoffset);
 
-    OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+    if (r300->rs_state->enable_vte) {
+        OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+    } else {
+        OUT_CS_REG(R300_VAP_VTE_CNTL, 0);
+    }
     END_CS;
 }
 
@@ -421,23 +646,73 @@ void r300_flush_textures(struct r300_context* r300)
 void r300_emit_dirty_state(struct r300_context* r300)
 {
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
-    int i;
-    int dirty_tex = 0;
+    struct r300_texture* tex;
+    int i, dirty_tex = 0;
+    boolean invalid = FALSE;
 
-    if (!(r300->dirty_hw)) {
+    if (!(r300->dirty_state)) {
         return;
     }
 
     r300_update_derived_state(r300);
 
     /* XXX check size */
-    struct r300_texture* fb_tex =
-        (struct r300_texture*)r300->framebuffer_state.cbufs[0];
-    r300->winsys->add_buffer(r300->winsys, fb_tex->buffer,
-            0, RADEON_GEM_DOMAIN_VRAM);
-    if (r300->winsys->validate(r300->winsys)) {
-        /* XXX */
+validate:
+    /* Color buffers... */
+    for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) {
+        tex = (struct r300_texture*)r300->framebuffer_state.cbufs[i]->texture;
+        assert(tex && tex->buffer && "cbuf is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    0, RADEON_GEM_DOMAIN_VRAM)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...depth buffer... */
+    if (r300->framebuffer_state.zsbuf) {
+        tex = (struct r300_texture*)r300->framebuffer_state.zsbuf->texture;
+        assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    0, RADEON_GEM_DOMAIN_VRAM)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...textures... */
+    for (i = 0; i < r300->texture_count; i++) {
+        tex = r300->textures[i];
+        assert(tex && tex->buffer && "texture is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...occlusion query buffer... */
+    if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo,
+                0, RADEON_GEM_DOMAIN_GTT)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    /* ...and vertex buffer. */
+    if (r300->vbo) {
+        if (!r300->winsys->add_buffer(r300->winsys, r300->vbo,
+                    RADEON_GEM_DOMAIN_GTT, 0)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    } else {
+        debug_printf("No VBO while emitting dirty state!\n");
+    }
+    if (!r300->winsys->validate(r300->winsys)) {
         r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            /* Well, hell. */
+            debug_printf("r300: Stuck in validation loop, gonna quit now.");
+            exit(1);
+        }
+        invalid = TRUE;
+        goto validate;
     }
 
     if (r300->dirty_state & R300_NEW_BLEND) {
@@ -450,6 +725,11 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300->dirty_state &= ~R300_NEW_BLEND_COLOR;
     }
 
+    if (r300->dirty_state & R300_NEW_CLIP) {
+        r300_emit_clip_state(r300, &r300->clip_state);
+        r300->dirty_state &= ~R300_NEW_CLIP;
+    }
+
     if (r300->dirty_state & R300_NEW_DSA) {
         r300_emit_dsa_state(r300, r300->dsa_state);
         r300->dirty_state &= ~R300_NEW_DSA;
@@ -457,11 +737,9 @@ void r300_emit_dirty_state(struct r300_context* r300)
 
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
         if (r300screen->caps->is_r500) {
-            r500_emit_fragment_shader(r300,
-                (struct r500_fragment_shader*)r300->fs);
+            r500_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]);
         } else {
-            r300_emit_fragment_shader(r300,
-                (struct r300_fragment_shader*)r300->fs);
+            r300_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]);
         }
         r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER;
     }
@@ -481,29 +759,27 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300->dirty_state &= ~R300_NEW_RS_BLOCK;
     }
 
-    if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) {
-        for (i = 0; i < r300->sampler_count; i++) {
-            if (r300->dirty_state & (R300_NEW_SAMPLER << i)) {
-                r300_emit_sampler(r300, r300->sampler_states[i], i);
-                r300->dirty_state &= ~(R300_NEW_SAMPLER << i);
-                dirty_tex++;
-            }
-        }
-    }
-
     if (r300->dirty_state & R300_NEW_SCISSOR) {
         r300_emit_scissor_state(r300, r300->scissor_state);
         r300->dirty_state &= ~R300_NEW_SCISSOR;
     }
 
-    if (r300->dirty_state & R300_ANY_NEW_TEXTURES) {
-        for (i = 0; i < r300->texture_count; i++) {
-            if (r300->dirty_state & (R300_NEW_TEXTURE << i)) {
-                r300_emit_texture(r300, r300->textures[i], i);
-                r300->dirty_state &= ~(R300_NEW_TEXTURE << i);
+    /* Samplers and textures are tracked separately but emitted together. */
+    if (r300->dirty_state &
+            (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) {
+        for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) {
+            if (r300->dirty_state &
+                    ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) {
+                r300_emit_texture(r300,
+                        r300->sampler_states[i],
+                        r300->textures[i],
+                        i);
+                r300->dirty_state &=
+                    ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i));
                 dirty_tex++;
             }
         }
+        r300->dirty_state &= ~(R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES);
     }
 
     if (r300->dirty_state & R300_NEW_VIEWPORT) {
@@ -519,4 +795,18 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300_emit_vertex_format_state(r300);
         r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT;
     }
+
+    if (r300->dirty_state & R300_NEW_VERTEX_SHADER) {
+        r300_emit_vertex_shader(r300, r300->vs);
+        r300->dirty_state &= ~R300_NEW_VERTEX_SHADER;
+    }
+
+    /* XXX
+    assert(r300->dirty_state == 0);
+    */
+
+    /* Finally, emit the VBO. */
+    r300_emit_vertex_buffer(r300);
+
+    r300->dirty_hw++;
 }
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 31dbc7ab85..350691d592 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -30,20 +30,28 @@
 #include "r300_screen.h"
 #include "r300_state_inlines.h"
 
+struct rX00_fragment_program_code;
+struct r300_vertex_program_code;
+
 void r300_emit_blend_state(struct r300_context* r300,
                            struct r300_blend_state* blend);
 
 void r300_emit_blend_color_state(struct r300_context* r300,
                                  struct r300_blend_color_state* bc);
 
+void r300_emit_clip_state(struct r300_context* r300,
+                          struct pipe_clip_state* clip);
+
 void r300_emit_dsa_state(struct r300_context* r300,
                          struct r300_dsa_state* dsa);
 
-void r300_emit_fragment_shader(struct r300_context* r300,
-                               struct r300_fragment_shader* fs);
+void r300_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code,
+                                     struct r300_constant_buffer* externals);
 
-void r500_emit_fragment_shader(struct r300_context* r300,
-                               struct r500_fragment_shader* fs);
+void r500_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code,
+                                     struct r300_constant_buffer* externals);
 
 void r300_emit_fb_state(struct r300_context* r300,
                         struct pipe_framebuffer_state* fb);
@@ -53,17 +61,22 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs);
 void r300_emit_rs_block_state(struct r300_context* r300,
                               struct r300_rs_block* rs);
 
-void r300_emit_sampler(struct r300_context* r300,
-                       struct r300_sampler_state* sampler, unsigned offset);
-
 void r300_emit_scissor_state(struct r300_context* r300,
                              struct r300_scissor_state* scissor);
 
 void r300_emit_texture(struct r300_context* r300,
-                       struct r300_texture* tex, unsigned offset);
+                       struct r300_sampler_state* sampler,
+                       struct r300_texture* tex,
+                       unsigned offset);
+
+void r300_emit_vertex_buffer(struct r300_context* r300);
 
 void r300_emit_vertex_format_state(struct r300_context* r300);
 
+void r300_emit_vertex_program_code(struct r300_context* r300,
+                                   struct r300_vertex_program_code* code,
+                                   struct r300_constant_buffer* constants);
+
 void r300_emit_vertex_shader(struct r300_context* r300,
                              struct r300_vertex_shader* vs);
 
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index 89a5f2b20c..0dff1c6f4f 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -29,7 +29,11 @@ static void r300_flush(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     CS_LOCALS(r300);
 
-    draw_flush(r300->draw);
+    /* We probably need to flush Draw, but we may have been called from
+     * within Draw. This feels kludgy, but it might be the best thing. */
+    if (!r300->draw->flushing) {
+        draw_flush(r300->draw);
+    }
 
     if (r300->dirty_hw) {
         FLUSH_CS;
diff --git a/src/gallium/drivers/r300/r300_flush.h b/src/gallium/drivers/r300/r300_flush.h
index a1b224b39c..9a83d89daa 100644
--- a/src/gallium/drivers/r300/r300_flush.h
+++ b/src/gallium/drivers/r300/r300_flush.h
@@ -23,6 +23,8 @@
 #ifndef R300_FLUSH_H
 #define R300_FLUSH_H
 
+#include "draw/draw_private.h"
+
 #include "pipe/p_context.h"
 
 #include "r300_context.h"
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
new file mode 100644
index 0000000000..36463b9a2e
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -0,0 +1,137 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_fs.h"
+
+#include "r300_tgsi_to_rc.h"
+
+#include "radeon_compiler.h"
+
+static void find_output_registers(struct r300_fragment_program_compiler * compiler,
+                                  struct r300_fragment_shader * fs)
+{
+    unsigned i;
+
+    /* Mark the outputs as not present initially */
+    compiler->OutputColor = fs->info.num_outputs;
+    compiler->OutputDepth = fs->info.num_outputs;
+
+    /* Now see where they really are. */
+    for(i = 0; i < fs->info.num_outputs; ++i) {
+        switch(fs->info.output_semantic_name[i]) {
+            case TGSI_SEMANTIC_COLOR:
+                compiler->OutputColor = i;
+                break;
+            case TGSI_SEMANTIC_POSITION:
+                compiler->OutputDepth = i;
+                break;
+        }
+    }
+}
+
+static void allocate_hardware_inputs(
+    struct r300_fragment_program_compiler * c,
+    void (*allocate)(void * data, unsigned input, unsigned hwreg),
+    void * mydata)
+{
+    struct tgsi_shader_info* info = &((struct r300_fragment_shader*)c->UserData)->info;
+    int total_colors = 0;
+    int colors = 0;
+    int total_generic = 0;
+    int generic = 0;
+    int i;
+
+    for (i = 0; i < info->num_inputs; i++) {
+        switch (info->input_semantic_name[i]) {
+            case TGSI_SEMANTIC_COLOR:
+                total_colors++;
+                break;
+            case TGSI_SEMANTIC_FOG:
+            case TGSI_SEMANTIC_GENERIC:
+                total_generic++;
+                break;
+        }
+    }
+
+    for(i = 0; i < info->num_inputs; i++) {
+        switch (info->input_semantic_name[i]) {
+            case TGSI_SEMANTIC_COLOR:
+                allocate(mydata, i, colors);
+                colors++;
+                break;
+            case TGSI_SEMANTIC_FOG:
+            case TGSI_SEMANTIC_GENERIC:
+                allocate(mydata, i, total_colors + generic);
+                generic++;
+                break;
+        }
+    }
+}
+
+void r300_translate_fragment_shader(struct r300_context* r300,
+                                    struct r300_fragment_shader* fs)
+{
+    struct r300_fragment_program_compiler compiler;
+    struct tgsi_to_rc ttr;
+
+    memset(&compiler, 0, sizeof(compiler));
+    rc_init(&compiler.Base);
+    compiler.Base.Debug = 1;
+
+    compiler.code = &fs->code;
+    compiler.is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
+    compiler.AllocateHwInputs = &allocate_hardware_inputs;
+    compiler.UserData = fs;
+
+    /* TODO: Program compilation depends on texture compare modes,
+     * which are sampler state. Therefore, programs need to be recompiled
+     * depending on this state as in the classic Mesa driver.
+     *
+     * This is not yet handled correctly.
+     */
+
+    find_output_registers(&compiler, fs);
+
+    if (compiler.Base.Debug) {
+        debug_printf("r300: Initial fragment program\n");
+        tgsi_dump(fs->state.tokens, 0);
+    }
+
+    /* Translate TGSI to our internal representation */
+    ttr.compiler = &compiler.Base;
+    ttr.info = &fs->info;
+
+    r300_tgsi_to_rc(&ttr, fs->state.tokens);
+
+    /* Invoke the compiler */
+    r3xx_compile_fragment_program(&compiler);
+    if (compiler.Base.Error) {
+        /* Todo: Fail gracefully */
+        fprintf(stderr, "r300 FP: Compiler error\n");
+        abort();
+    }
+
+    /* And, finally... */
+    rc_destroy(&compiler.Base);
+    fs->translated = TRUE;
+}
diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h
new file mode 100644
index 0000000000..9fab789402
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_fs.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_FS_H
+#define R300_FS_H
+
+#include "tgsi/tgsi_dump.h"
+
+#include "r300_context.h"
+#include "r3xx_fs.h"
+#include "r5xx_fs.h"
+
+#include "radeon_code.h"
+
+struct r300_fragment_shader {
+    /* Parent class */
+    struct pipe_shader_state state;
+    struct tgsi_shader_info info;
+
+    /* Has this shader been translated yet? */
+    boolean translated;
+
+    /* Compiled code */
+    struct rX00_fragment_program_code code;
+};
+
+
+void r300_translate_fragment_shader(struct r300_context* r300,
+                                    struct r300_fragment_shader* fs);
+
+    #endif /* R300_FS_H */
diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c
index 8fc61c2dec..1d5185b417 100644
--- a/src/gallium/drivers/r300/r300_query.c
+++ b/src/gallium/drivers/r300/r300_query.c
@@ -25,14 +25,30 @@
 static struct pipe_query* r300_create_query(struct pipe_context* pipe,
                                             unsigned query_type)
 {
-    struct r300_query* q = CALLOC_STRUCT(r300_query);
+    struct r300_context* r300 = r300_context(pipe);
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    unsigned query_size = r300screen->caps->num_frag_pipes * 4;
+    struct r300_query* q, * qptr;
+
+    q = CALLOC_STRUCT(r300_query);
 
     q->type = query_type;
     assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 
-    /* XXX this is to force winsys to give us a GTT buffer */
-    q->buf = pipe->screen->buffer_create(pipe->screen, 64,
-            PIPE_BUFFER_USAGE_VERTEX, 64);
+    q->active = FALSE;
+
+    if (!r300->query_list) {
+        r300->query_list = q;
+    } else if (!is_empty_list(r300->query_list)) {
+        qptr = last_elem(r300->query_list);
+        q->offset = qptr->offset + query_size;
+        insert_at_tail(r300->query_list, q);
+    }
+
+    /* XXX */
+    if (q->offset >= 4096) {
+        q->offset = 0;
+    }
 
     return (struct pipe_query*)q;
 }
@@ -40,6 +56,9 @@ static struct pipe_query* r300_create_query(struct pipe_context* pipe,
 static void r300_destroy_query(struct pipe_context* pipe,
                                struct pipe_query* query)
 {
+    struct r300_query* q = (struct r300_query*)query;
+
+    remove_from_list(q);
     FREE(query);
 }
 
@@ -49,15 +68,15 @@ static void r300_begin_query(struct pipe_context* pipe,
     uint32_t* map;
     struct r300_context* r300 = r300_context(pipe);
     struct r300_query* q = (struct r300_query*)query;
-    CS_LOCALS(r300);
 
-    map = pipe_buffer_map(pipe->screen, q->buf, PIPE_BUFFER_USAGE_CPU_WRITE);
+    map = pipe->screen->buffer_map(pipe->screen, r300->oqbo,
+            PIPE_BUFFER_USAGE_CPU_WRITE);
+    map += q->offset / 4;
     *map = ~0;
-    pipe_buffer_unmap(pipe->screen, q->buf);
+    pipe->screen->buffer_unmap(pipe->screen, r300->oqbo);
 
-    BEGIN_CS(2);
-    OUT_CS_REG(R300_ZB_ZPASS_DATA, 0);
-    END_CS;
+    r300_emit_dirty_state(r300);
+    r300_emit_query_begin(r300, q);
 }
 
 static void r300_end_query(struct pipe_context* pipe,
@@ -65,12 +84,9 @@ static void r300_end_query(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_query* q = (struct r300_query*)query;
-    CS_LOCALS(r300);
 
-    BEGIN_CS(4);
-    OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
-    OUT_CS_RELOC(q->buf, 0, 0, RADEON_GEM_DOMAIN_GTT, 0);
-    END_CS;
+    r300_emit_dirty_state(r300);
+    r300_emit_query_end(r300, q);
 }
 
 static boolean r300_get_query_result(struct pipe_context* pipe,
@@ -78,22 +94,38 @@ static boolean r300_get_query_result(struct pipe_context* pipe,
                                      boolean wait,
                                      uint64_t* result)
 {
+    struct r300_context* r300 = r300_context(pipe);
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct r300_query* q = (struct r300_query*)query;
+    unsigned flags = PIPE_BUFFER_USAGE_CPU_READ;
     uint32_t* map;
     uint32_t temp;
+    unsigned i;
 
     if (wait) {
-        /* Well, we're expected to just sit here and spin, so let's go ahead
-         * and flush so we can be sure that the card's spinning... */
-        /* XXX double-check these params */
         pipe->flush(pipe, 0, NULL);
+    } else {
+        flags |= PIPE_BUFFER_USAGE_DONTBLOCK;
     }
 
-    map = pipe_buffer_map(pipe->screen, q->buf, PIPE_BUFFER_USAGE_CPU_READ);
-    temp = *map;
-    pipe_buffer_unmap(pipe->screen, q->buf);
+    map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, flags);
+    map += q->offset / 4;
+    for (i = 0; i < r300screen->caps->num_frag_pipes; i++) {
+        if (*map == ~0) {
+            /* Looks like our results aren't ready yet. */
+            if (wait) {
+                debug_printf("r300: Despite waiting, OQ results haven't"
+                        " come in yet.\n");
+            }
+            temp = ~0;
+            break;
+        }
+        temp += *map;
+        map++;
+    }
+    pipe->screen->buffer_unmap(pipe->screen, r300->oqbo);
 
-    if (temp < 0) {
+    if (temp == ~0) {
         /* Our results haven't been written yet... */
         return FALSE;
     }
diff --git a/src/gallium/drivers/r300/r300_query.h b/src/gallium/drivers/r300/r300_query.h
index 4f447ea45b..4f50e8f844 100644
--- a/src/gallium/drivers/r300/r300_query.h
+++ b/src/gallium/drivers/r300/r300_query.h
@@ -27,12 +27,7 @@
 #include "r300_cs.h"
 #include "r300_reg.h"
 
-struct r300_query {
-    /* The kind of query. Currently only OQ is supported. */
-    unsigned type;
-    /* Buffer object where we want our results to reside. */
-    struct pipe_buffer* buf;
-};
+struct r300_context;
 
 static INLINE struct r300_query* r300_query(struct pipe_query* q)
 {
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index 660816e1da..03cd219cde 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -511,11 +511,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_PVS_XYZW_VALID_INST_SHIFT    10
 #       define R300_PVS_LAST_INST_SHIFT          20
 #       define R300_PVS_FIRST_INST(x)            ((x) << 0)
+#       define R300_PVS_XYZW_VALID_INST(x)       ((x) << 10)
 #       define R300_PVS_LAST_INST(x)             ((x) << 20)
 /* Addresses are relative the the vertex program parameters area. */
 #define R300_VAP_PVS_CONST_CNTL             0x22D4
 #       define R300_PVS_CONST_BASE_OFFSET_SHIFT  0
 #       define R300_PVS_MAX_CONST_ADDR_SHIFT     16
+#       define R300_PVS_MAX_CONST_ADDR(x)        ((x) << 16)
 #define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
 #       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
 #define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
@@ -1062,8 +1064,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
         R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | \
         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \
         R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | \
-        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT | \
-        R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT )
 
 #       define R300_SHADE_MODEL_SMOOTH ( \
         R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | \
@@ -1073,8 +1074,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
         R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | \
         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \
         R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | \
-        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \
-        R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD )
 
 /* Specifies red & green components of fill color -- S312 format -- Backwards comp. */
 #define R300_GA_SOLID_RG                         0x427c
@@ -1478,6 +1478,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_TX_PITCH_EN                  (1 << 31)
 #       define R300_TX_WIDTH(x)                  ((x) << 0)
 #       define R300_TX_HEIGHT(x)                 ((x) << 11)
+#       define R300_TX_NUM_LEVELS(x)             ((x) << 26)
 
 #define R300_TX_FORMAT1_0                   0x44C0
 	/* The interpretation of the format word by Wladimir van der Laan */
@@ -1485,9 +1486,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 	   They are given meanings as R, G, B and Alpha by the swizzle
 	   specification */
 #	define R300_TX_FORMAT_X8		    0x0
-#	define R500_TX_FORMAT_X1		    0x0 // bit set in format 2
 #	define R300_TX_FORMAT_X16		    0x1
-#	define R500_TX_FORMAT_X1_REV		    0x0 // bit set in format 2
 #	define R300_TX_FORMAT_Y4X4		    0x2
 #	define R300_TX_FORMAT_Y8X8		    0x3
 #	define R300_TX_FORMAT_Y16X16		    0x4
@@ -1504,31 +1503,29 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_DXT1	    	    0xF
 #	define R300_TX_FORMAT_DXT3	    	    0x10
 #	define R300_TX_FORMAT_DXT5	    	    0x11
-#	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
-#	define R300_TX_FORMAT_A8R8G8B8	    	    0x13     /* no swizzle */
-#	define R300_TX_FORMAT_B8G8_B8G8	    	    0x14     /* no swizzle */
-#	define R300_TX_FORMAT_G8R8_G8B8	    	    0x15     /* no swizzle */
-
-	/* These two values are wrong, but they're the only values that
-	 * produce any even vaguely correct results.  Can r300 only do 16-bit
-	 * depth textures?
-	 */
-#	define R300_TX_FORMAT_X24_Y8	    	    0x1e
-#	define R300_TX_FORMAT_X32	    	    0x1e
-
-	/* 0x16 - some 16 bit green format.. ?? */
+#	define R300_TX_FORMAT_Y8           	    0x12
+#	define R300_TX_FORMAT_AVYU444 	    	    0x13
+#	define R300_TX_FORMAT_VYUY422  	    	    0x14
+#	define R300_TX_FORMAT_YVYU422  	    	    0x15
+#	define R300_TX_FORMAT_16_MPEG  	    	    0x16
+#	define R300_TX_FORMAT_16_16_MPEG    	    0x17
+#	define R300_TX_FORMAT_16F     	    	    0x18
+#	define R300_TX_FORMAT_16F_16F 	    	    0x19
+#	define R300_TX_FORMAT_16F_16F_16F_16F  	    0x1A
+#	define R300_TX_FORMAT_32F     	    	    0x1B
+#	define R300_TX_FORMAT_32F_32F 	    	    0x1C
+#	define R300_TX_FORMAT_32F_32F_32F_32F  	    0x1D
+#       define R300_TX_FORMAT_W24_FP                0x1E
+
+#       define R300_TX_FORMAT_SIGNED_W             (1 << 5)
+#       define R300_TX_FORMAT_SIGNED_Z             (1 << 6)
+#       define R300_TX_FORMAT_SIGNED_Y             (1 << 7)
+#       define R300_TX_FORMAT_SIGNED_X             (1 << 8)
+#       define R300_TX_FORMAT_SIGNED               (0xf << 5)
+
 #	define R300_TX_FORMAT_3D		   (1 << 25)
 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
 
-	/* gap */
-	/* Floating point formats */
-	/* Note - hardware supports both 16 and 32 bit floating point */
-#	define R300_TX_FORMAT_FL_I16	    	    0x18
-#	define R300_TX_FORMAT_FL_I16A16	    	    0x19
-#	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
-#	define R300_TX_FORMAT_FL_I32	    	    0x1B
-#	define R300_TX_FORMAT_FL_I32A32	    	    0x1C
-#	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
 	/* alpha modes, convenience mostly */
 	/* if you have alpha, pick constant appropriate to the
 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
@@ -1569,7 +1566,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_CONST_Z		(4<<5)
 #	define R300_TX_FORMAT_CONST_W		(8<<5)
 
-#	define R300_TX_FORMAT_YUV_MODE		0x00800000
+#       define R300_TX_FORMAT_GAMMA               (1 << 21)
+#       define R300_TX_FORMAT_YUV_TO_RGB          (1 << 22)
 
 #define R300_TX_FORMAT2_0		    0x4500 /* obvious missing in gap */
 #       define R300_TX_PITCHMASK_SHIFT           0
@@ -3040,6 +3038,7 @@ enum {
 #   define R500_INST_RGB_WMASK_R			(1 << 11)
 #   define R500_INST_RGB_WMASK_G			(1 << 12)
 #   define R500_INST_RGB_WMASK_B			(1 << 13)
+#   define R500_INST_RGB_WMASK_RGB			(7 << 11)
 #   define R500_INST_ALPHA_WMASK			(1 << 14)
 #   define R500_INST_RGB_OMASK_R			(1 << 15)
 #   define R500_INST_RGB_OMASK_G			(1 << 16)
@@ -3313,6 +3312,10 @@ enum {
 
 #define R200_3D_DRAW_IMMD_2      0xC0003500
 
+/* XXX Oh look, stuff not brought over from docs yet */
+
+#define R300_SU_REG_DEST                    0x42C8
+
 #endif /* _R300_REG_H */
 
 /* *INDENT-ON* */
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index cbd84d7c56..cd458d019a 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -45,11 +45,7 @@ struct r300_render {
 
     /* VBO */
     struct pipe_buffer* vbo;
-    size_t vbo_size;
-    size_t vbo_offset;
-    void* vbo_map;
     size_t vbo_alloc_size;
-    size_t vbo_max_used;
 };
 
 static INLINE struct r300_render*
@@ -78,24 +74,21 @@ static boolean r300_render_allocate_vertices(struct vbuf_render* render,
     struct pipe_screen* screen = r300->context.screen;
     size_t size = (size_t)vertex_size * (size_t)count;
 
-    if (r300render->vbo) {
+    if (r300render->vbo && (size > r300render->vbo_alloc_size)) {
         pipe_buffer_reference(&r300render->vbo, NULL);
     }
+    
+    if (!r300render->vbo) {
+        r300render->vbo = pipe_buffer_create(screen,
+                                             64,
+                                             PIPE_BUFFER_USAGE_VERTEX,
+                                             size);
+    }
 
-    r300render->vbo_size = MAX2(size, r300render->vbo_alloc_size);
-    r300render->vbo_offset = 0;
-    r300render->vbo = pipe_buffer_create(screen,
-                                         64,
-                                         PIPE_BUFFER_USAGE_VERTEX,
-                                         r300render->vbo_size);
-
+    r300render->vbo_alloc_size = MAX2(size, r300render->vbo_alloc_size);
     r300render->vertex_size = vertex_size;
 
-    if (r300render->vbo) {
-        return TRUE;
-    } else {
-        return FALSE;
-    }
+    return (r300render->vbo) ? TRUE : FALSE;
 }
 
 static void* r300_render_map_vertices(struct vbuf_render* render)
@@ -103,10 +96,8 @@ static void* r300_render_map_vertices(struct vbuf_render* render)
     struct r300_render* r300render = r300_render(render);
     struct pipe_screen* screen = r300render->r300->context.screen;
 
-    r300render->vbo_map = pipe_buffer_map(screen, r300render->vbo,
-                                          PIPE_BUFFER_USAGE_CPU_WRITE);
-
-    return (unsigned char*)r300render->vbo_map + r300render->vbo_offset;
+    return (unsigned char*)pipe_buffer_map(screen, r300render->vbo,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE);
 }
 
 static void r300_render_unmap_vertices(struct vbuf_render* render,
@@ -116,9 +107,6 @@ static void r300_render_unmap_vertices(struct vbuf_render* render,
     struct r300_render* r300render = r300_render(render);
     struct pipe_screen* screen = r300render->r300->context.screen;
 
-    r300render->vbo_max_used = MAX2(r300render->vbo_max_used,
-             r300render->vertex_size * (max + 1));
-
     pipe_buffer_unmap(screen, r300render->vbo);
 }
 
@@ -180,27 +168,9 @@ static void prepare_render(struct r300_render* render, unsigned count)
 
     CS_LOCALS(r300);
 
-    /* Make sure that all possible state is emitted. */
-    r300_emit_dirty_state(r300);
+    r300->vbo = render->vbo;
 
-    debug_printf("r300: Preparing vertex buffer %p for render, "
-            "vertex size %d, vertex count %d\n", render->vbo,
-            r300->vertex_info.vinfo.size, count);
-    /* Set the pointer to our vertex buffer. The emitted values are this:
-     * PACKET3 [3D_LOAD_VBPNTR]
-     * COUNT   [1]
-     * FORMAT  [size | stride << 8]
-     * OFFSET  [0]
-     * VBPNTR  [relocated BO]
-     */
-    BEGIN_CS(7);
-    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
-    OUT_CS(1);
-    OUT_CS(r300->vertex_info.vinfo.size |
-            (r300->vertex_info.vinfo.size << 8));
-    OUT_CS(render->vbo_offset);
-    OUT_CS_RELOC(render->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
-    END_CS;
+    r300_emit_dirty_state(r300);
 }
 
 static void r300_render_draw_arrays(struct vbuf_render* render,
@@ -212,8 +182,6 @@ static void r300_render_draw_arrays(struct vbuf_render* render,
 
     CS_LOCALS(r300);
 
-    r300render->vbo_offset = start;
-
     prepare_render(r300render, count);
 
     debug_printf("r300: Doing vbuf render, count %d\n", count);
@@ -248,13 +216,14 @@ static void r300_render_draw(struct vbuf_render* render,
         return;
     }
 
+/*
     index_map = pipe_buffer_map(screen, index_buffer,
                                 PIPE_BUFFER_USAGE_CPU_WRITE);
     memcpy(index_map, indices, count);
     pipe_buffer_unmap(screen, index_buffer);
 
     debug_printf("r300: Doing indexbuf render, count %d\n", count);
-/*
+
     BEGIN_CS(8);
     OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0);
     OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
@@ -264,13 +233,15 @@ static void r300_render_draw(struct vbuf_render* render,
     OUT_CS_INDEX_RELOC(index_buffer, 0, count, RADEON_GEM_DOMAIN_GTT, 0, 0);
     END_CS; */
 
-    BEGIN_CS(2 + count);
-    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, count);
+    BEGIN_CS(2 + (count+1)/2);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, (count+1)/2);
     OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
-           r300render->hwprim | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-    for (i = 0; i < count; i++) {
-        index = indices[i];
-        OUT_CS(index);
+           r300render->hwprim);
+    for (i = 0; i < count-1; i += 2) {
+        OUT_CS(indices[i+1] << 16 | indices[i]);
+    }
+    if (count % 2) {
+        OUT_CS(indices[count-1]);
     }
     END_CS;
 }
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index d2c5998c26..15740f6125 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -87,23 +87,24 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
             } else {
                 return 0;
             }
-            return 0;
         case PIPE_CAP_GLSL:
-            /* IN THEORY */
-            return 0;
+            if (r300screen->caps->is_r500) {
+                return 1;
+            } else {
+                return 0;
+            }
         case PIPE_CAP_S3TC:
-            /* IN THEORY */
-            return 0;
+            return 1;
         case PIPE_CAP_ANISOTROPIC_FILTER:
-            /* IN THEORY */
-            return 0;
+            return 1;
         case PIPE_CAP_POINT_SPRITE:
             /* IN THEORY */
             return 0;
         case PIPE_CAP_MAX_RENDER_TARGETS:
             return 4;
         case PIPE_CAP_OCCLUSION_QUERY:
-            return 1;
+            /* IN THEORY */
+            return 0;
         case PIPE_CAP_TEXTURE_SHADOW_MAP:
             /* IN THEORY */
             return 0;
@@ -143,6 +144,11 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
         case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
             /* XXX guessing (what a terrible guess) */
             return 2;
+        case PIPE_CAP_TGSI_CONT_SUPPORTED:
+            /* XXX */
+            return 0;
+        case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+            return 1;
         default:
             debug_printf("r300: Implementation error: Bad param %d\n",
                 param);
@@ -152,17 +158,20 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
 
 static float r300_get_paramf(struct pipe_screen* pscreen, int param)
 {
+    struct r300_screen* r300screen = r300_screen(pscreen);
+
     switch (param) {
         case PIPE_CAP_MAX_LINE_WIDTH:
         case PIPE_CAP_MAX_LINE_WIDTH_AA:
-            /* XXX this is the biggest thing that will fit in that register.
-            * Perhaps the actual rendering limits are less? */
-            return 10922.0f;
         case PIPE_CAP_MAX_POINT_WIDTH:
         case PIPE_CAP_MAX_POINT_WIDTH_AA:
-            /* XXX this is the biggest thing that will fit in that register.
-             * Perhaps the actual rendering limits are less? */
-            return 10922.0f;
+            /* The maximum dimensions of the colorbuffer are our practical
+             * rendering limits. 2048 pixels should be enough for anybody. */
+            if (r300screen->caps->is_r500) {
+                return 4096.0f;
+            } else {
+                return 2048.0f;
+            }
         case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
             return 16.0f;
         case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
@@ -174,21 +183,58 @@ static float r300_get_paramf(struct pipe_screen* pscreen, int param)
     }
 }
 
-static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500)
+static boolean check_tex_2d_format(enum pipe_format format, uint32_t usage,
+                                   boolean is_r500)
 {
     switch (format) {
+        /* Supported formats. */
         /* Colorbuffer */
         case PIPE_FORMAT_A4R4G4B4_UNORM:
         case PIPE_FORMAT_R5G6B5_UNORM:
         case PIPE_FORMAT_A1R5G5B5_UNORM:
-        case PIPE_FORMAT_A8R8G8B8_UNORM:
+            return usage &
+                (PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                 PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                 PIPE_TEXTURE_USAGE_PRIMARY);
+
+        /* Texture */
+        case PIPE_FORMAT_A8R8G8B8_SRGB:
+        case PIPE_FORMAT_R8G8B8A8_SRGB:
+        case PIPE_FORMAT_DXT1_RGB:
+        case PIPE_FORMAT_DXT1_RGBA:
+        case PIPE_FORMAT_DXT3_RGBA:
+        case PIPE_FORMAT_DXT5_RGBA:
+        case PIPE_FORMAT_YCBCR:
+            return usage & PIPE_TEXTURE_USAGE_SAMPLER;
+
         /* Colorbuffer or texture */
+        case PIPE_FORMAT_A8R8G8B8_UNORM:
+        case PIPE_FORMAT_R8G8B8A8_UNORM:
         case PIPE_FORMAT_I8_UNORM:
+            return usage &
+                (PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                 PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                 PIPE_TEXTURE_USAGE_PRIMARY |
+                 PIPE_TEXTURE_USAGE_SAMPLER);
+
         /* Z buffer */
         case PIPE_FORMAT_Z16_UNORM:
-        /* Z buffer with stencil */
+            return usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+
+        /* Z buffer with stencil or texture */
         case PIPE_FORMAT_Z24S8_UNORM:
-            return TRUE;
+            return usage &
+                (PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
+                 PIPE_TEXTURE_USAGE_SAMPLER);
+
+        /* Definitely unsupported formats. */
+        /* Non-usable Z buffer/stencil formats. */
+        case PIPE_FORMAT_Z24X8_UNORM:
+        case PIPE_FORMAT_S8Z24_UNORM:
+        case PIPE_FORMAT_X8Z24_UNORM:
+            debug_printf("r300: Note: Got unsupported format: %s in %s\n",
+                pf_name(format), __FUNCTION__);
+            return FALSE;
 
         /* XXX These don't even exist
         case PIPE_FORMAT_A32R32G32B32:
@@ -211,7 +257,8 @@ static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500)
             return FALSE; */
 
         default:
-            debug_printf("r300: Warning: Got unsupported format: %s in %s\n",
+            /* Unknown format... */
+            debug_printf("r300: Warning: Got unknown format: %s in %s\n",
                 pf_name(format), __FUNCTION__);
             break;
     }
@@ -228,11 +275,18 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen,
 {
     switch (target) {
         case PIPE_TEXTURE_2D:
-            return check_tex_2d_format(format,
+            return check_tex_2d_format(format, tex_usage,
                 r300_screen(pscreen)->caps->is_r500);
+        case PIPE_TEXTURE_1D:
+        case PIPE_TEXTURE_3D:
+        case PIPE_TEXTURE_CUBE:
+            debug_printf("r300: Implementation error: Unsupported format "
+                    "target: %d\n", target);
+            break;
         default:
-            debug_printf("r300: Warning: Got unknown format target: %d\n",
-                format);
+            debug_printf("r300: Fatal: This is not a format target: %d\n",
+                target);
+            assert(0);
             break;
     }
 
@@ -268,13 +322,14 @@ r300_get_tex_transfer(struct pipe_screen *screen,
     trans = CALLOC_STRUCT(r300_transfer);
     if (trans) {
         pipe_texture_reference(&trans->transfer.texture, texture);
-        trans->transfer.format = trans->transfer.format;
+        trans->transfer.format = texture->format;
         trans->transfer.width = w;
         trans->transfer.height = h;
         trans->transfer.block = texture->block;
         trans->transfer.nblocksx = texture->nblocksx[level];
         trans->transfer.nblocksy = texture->nblocksy[level];
-        trans->transfer.stride = tex->stride;
+        trans->transfer.stride = align(pf_get_stride(&trans->transfer.block,
+                                                     texture->width[level]), 32);
         trans->transfer.usage = usage;
         trans->offset = offset;
     }
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 3f52dbc3be..2a0e41fbc3 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -49,7 +49,7 @@ struct r300_transfer {
 };
 
 /* Convenience cast wrapper. */
-static struct r300_screen* r300_screen(struct pipe_screen* screen) {
+static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) {
     return (struct r300_screen*)screen;
 }
 
diff --git a/src/gallium/drivers/r300/r300_shader_inlines.h b/src/gallium/drivers/r300/r300_shader_inlines.h
new file mode 100644
index 0000000000..a04f45b03e
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_shader_inlines.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_SHADER_INLINES_H
+#define R300_SHADER_INLINES_H
+
+/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're
+ * not using enough of it. */
+static const struct tgsi_full_src_register r300_constant_zero = {
+    .SrcRegister.Extended = TRUE,
+    .SrcRegister.File = TGSI_FILE_NULL,
+    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
+};
+
+static const struct tgsi_full_src_register r300_constant_one = {
+    .SrcRegister.Extended = TRUE,
+    .SrcRegister.File = TGSI_FILE_NULL,
+    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE,
+};
+
+#endif /* R300_SHADER_INLINES_H */
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 184a23c9e6..c16cadd040 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -24,12 +24,15 @@
 #include "util/u_pack_color.h"
 
 #include "util/u_debug.h"
+
+#include "pipe/p_config.h"
 #include "pipe/internal/p_winsys_screen.h"
 
 #include "r300_context.h"
 #include "r300_reg.h"
 #include "r300_state_inlines.h"
-#include "r300_state_shader.h"
+#include "r300_fs.h"
+#include "r300_vs.h"
 
 /* r300_state: Functions used to intialize state context by translating
  * Gallium state objects into semi-native r300 state objects. */
@@ -62,8 +65,6 @@ static void* r300_create_blend_state(struct pipe_context* pipe,
     }
 
     /* PIPE_LOGICOP_* don't need to be translated, fortunately. */
-    /* XXX are logicops still allowed if blending's disabled?
-     * Does Gallium take care of it for us? */
     if (state->logicop_enable) {
         blend->rop = R300_RB3D_ROPCNTL_ROP_ENABLE |
                 (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT;
@@ -121,9 +122,14 @@ static void r300_set_clip_state(struct pipe_context* pipe,
                                 const struct pipe_clip_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
-    /* XXX Draw */
-    draw_flush(r300->draw);
-    draw_set_clip_state(r300->draw, state);
+
+    if (r300_screen(pipe->screen)->caps->has_tcl) {
+        r300->clip_state = *state;
+        r300->dirty_state |= R300_NEW_CLIP;
+    } else {
+        draw_flush(r300->draw);
+        draw_set_clip_state(r300->draw, state);
+    }
 }
 
 static void
@@ -132,7 +138,6 @@ static void
                              const struct pipe_constant_buffer* buffer)
 {
     struct r300_context* r300 = r300_context(pipe);
-    int i = r300->shader_constants[shader].user_count;
 
     /* This entire chunk of code seems ever-so-slightly baked.
      * It's as if I've got pipe_buffer* matryoshkas... */
@@ -143,24 +148,13 @@ static void
             map, buffer->buffer->size);
         pipe->winsys->buffer_unmap(pipe->winsys, buffer->buffer);
 
-        r300->shader_constants[shader].user_count =
+        r300->shader_constants[shader].count =
             buffer->buffer->size / (sizeof(float) * 4);
     } else {
-        r300->shader_constants[shader].user_count = 0;
+        r300->shader_constants[shader].count = 0;
     }
 
     r300->dirty_state |= R300_NEW_CONSTANTS;
-
-    /* If the number of constants have changed, invalidate the shader. */
-    if (r300->shader_constants[shader].user_count != i) {
-        if (shader == PIPE_SHADER_FRAGMENT && r300->fs) {
-            r300->fs->translated = FALSE;
-            r300_translate_fragment_shader(r300, r300->fs);
-        } else if (shader == PIPE_SHADER_VERTEX && r300->vs) {
-            r300->vs->translated = FALSE;
-            r300_translate_vertex_shader(r300, r300->vs);
-        }
-    }
 }
 
 /* Create a new depth, stencil, and alpha state based on the CSO dsa state.
@@ -230,7 +224,8 @@ static void*
         dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f,
                                      0, 1023);
     } else {
-        dsa->z_buffer_top = R300_ZTOP_ENABLE;
+        /* XXX need to fix this to be dynamically set
+        dsa->z_buffer_top = R300_ZTOP_ENABLE; */
     }
 
     return (void*)dsa;
@@ -257,6 +252,7 @@ static void r300_set_edgeflags(struct pipe_context* pipe,
                                const unsigned* bitfield)
 {
     /* XXX you know it's bad when i915 has this blank too */
+    /* XXX and even worse, I have no idea WTF the bitfield is */
 }
 
 static void
@@ -276,19 +272,13 @@ static void
 static void* r300_create_fs_state(struct pipe_context* pipe,
                                   const struct pipe_shader_state* shader)
 {
-    struct r300_context* r300 = r300_context(pipe);
-    struct r3xx_fragment_shader* fs = NULL;
+    struct r300_fragment_shader* fs = NULL;
 
-    if (r300_screen(r300->context.screen)->caps->is_r500) {
-        fs =
-            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r500_fragment_shader);
-    } else {
-        fs =
-            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader);
-    }
+    fs = (struct r300_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader);
 
     /* Copy state directly into shader. */
     fs->state = *shader;
+    fs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
     tgsi_scan_shader(shader->tokens, &fs->info);
 
@@ -299,7 +289,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe,
 static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
-    struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader;
+    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
 
     if (fs == NULL) {
         r300->fs = NULL;
@@ -308,7 +298,6 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
         r300_translate_fragment_shader(r300, fs);
     }
 
-    fs->translated = TRUE;
     r300->fs = fs;
 
     r300->dirty_state |= R300_NEW_FRAGMENT_SHADER;
@@ -317,13 +306,16 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
 /* Delete fragment shader state. */
 static void r300_delete_fs_state(struct pipe_context* pipe, void* shader)
 {
+    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
+    rc_constants_destroy(&fs->code.constants);
+    FREE(fs->state.tokens);
     FREE(shader);
 }
 
 static void r300_set_polygon_stipple(struct pipe_context* pipe,
                                      const struct pipe_poly_stipple* state)
 {
-    /* XXX */
+    /* XXX no idea how to set this up, but not terribly important */
 }
 
 /* Create a new rasterizer state based on the CSO rasterizer state.
@@ -341,14 +333,21 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
     /* Copy rasterizer state for Draw. */
     rs->rs = *state;
 
+    rs->enable_vte = !state->bypass_vs_clip_and_viewport;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+    rs->vap_control_status = R300_VC_NO_SWAP;
+#else
+    rs->vap_control_status = R300_VC_32BIT_SWAP;
+#endif
+
     /* If bypassing TCL, or if no TCL engine is present, turn off the HW TCL.
      * Else, enable HW TCL and force Draw's TCL off. */
     if (state->bypass_vs_clip_and_viewport ||
             !r300_screen(pipe->screen)->caps->has_tcl) {
-        rs->vap_control_status = R300_VAP_TCL_BYPASS;
+        rs->vap_control_status |= R300_VAP_TCL_BYPASS;
     } else {
         rs->rs.bypass_vs_clip_and_viewport = TRUE;
-        rs->vap_control_status = 0;
     }
 
     rs->point_size = pack_float_16_6x(state->point_size) |
@@ -412,6 +411,10 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
         rs->color_control = R300_SHADE_MODEL_SMOOTH;
     }
 
+    if (!state->flatshade_first) {
+        rs->color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST;
+    }
+
     return (void*)rs;
 }
 
@@ -535,16 +538,16 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
             (state->minx << R300_SCISSORS_X_SHIFT) |
             (state->miny << R300_SCISSORS_Y_SHIFT);
         r300->scissor_state->scissor_bottom_right =
-            (state->maxx << R300_SCISSORS_X_SHIFT) |
-            (state->maxy << R300_SCISSORS_Y_SHIFT);
+            ((state->maxx - 1) << R300_SCISSORS_X_SHIFT) |
+            ((state->maxy - 1) << R300_SCISSORS_Y_SHIFT);
     } else {
         /* Offset of 1440 in non-R500 chipsets. */
         r300->scissor_state->scissor_top_left =
             ((state->minx + 1440) << R300_SCISSORS_X_SHIFT) |
             ((state->miny + 1440) << R300_SCISSORS_Y_SHIFT);
         r300->scissor_state->scissor_bottom_right =
-            ((state->maxx + 1440) << R300_SCISSORS_X_SHIFT) |
-            ((state->maxy + 1440) << R300_SCISSORS_Y_SHIFT);
+            (((state->maxx - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
+            (((state->maxy - 1) + 1440) << R300_SCISSORS_Y_SHIFT);
     }
 
     r300->dirty_state |= R300_NEW_SCISSOR;
@@ -555,40 +558,35 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    draw_flush(r300->draw);
-
-    if (r300_screen(r300->context.screen)->caps->has_tcl) {
-        /* Do the transform in HW. */
-        r300->viewport_state->vte_control = R300_VTX_W0_FMT;
+    /* Do the transform in HW. */
+    r300->viewport_state->vte_control = R300_VTX_W0_FMT;
 
-        if (state->scale[0] != 1.0f) {
-            r300->viewport_state->xscale = state->scale[0];
-            r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
-        }
-        if (state->scale[1] != 1.0f) {
-            r300->viewport_state->yscale = state->scale[1];
-            r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
-        }
-        if (state->scale[2] != 1.0f) {
-            r300->viewport_state->zscale = state->scale[2];
-            r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
-        }
-        if (state->translate[0] != 0.0f) {
-            r300->viewport_state->xoffset = state->translate[0];
-            r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA;
-        }
-        if (state->translate[1] != 0.0f) {
-            r300->viewport_state->yoffset = state->translate[1];
-            r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA;
-        }
-        if (state->translate[2] != 0.0f) {
-            r300->viewport_state->zoffset = state->translate[2];
-            r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA;
-        }
-    } else {
-        r300->viewport_state->vte_control = 0;
-        /* Have Draw do the actual transform. */
-        draw_set_viewport_state(r300->draw, state);
+    if (state->scale[0] != 1.0f) {
+        assert(state->scale[0] != 0.0f);
+        r300->viewport_state->xscale = state->scale[0];
+        r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
+    }
+    if (state->scale[1] != 1.0f) {
+        assert(state->scale[1] != 0.0f);
+        r300->viewport_state->yscale = state->scale[1];
+        r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
+    }
+    if (state->scale[2] != 1.0f) {
+        assert(state->scale[2] != 0.0f);
+        r300->viewport_state->zscale = state->scale[2];
+        r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
+    }
+    if (state->translate[0] != 0.0f) {
+        r300->viewport_state->xoffset = state->translate[0];
+        r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA;
+    }
+    if (state->translate[1] != 0.0f) {
+        r300->viewport_state->yoffset = state->translate[1];
+        r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA;
+    }
+    if (state->translate[2] != 0.0f) {
+        r300->viewport_state->zoffset = state->translate[2];
+        r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA;
     }
 
     r300->dirty_state |= R300_NEW_VIEWPORT;
@@ -628,6 +626,7 @@ static void* r300_create_vs_state(struct pipe_context* pipe,
         struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader);
         /* Copy state directly into shader. */
         vs->state = *shader;
+        vs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
         tgsi_scan_shader(shader->tokens, &vs->info);
 
@@ -672,7 +671,9 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader)
     if (r300_screen(pipe->screen)->caps->has_tcl) {
         struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
 
+        rc_constants_destroy(&vs->code.constants);
         draw_delete_vertex_shader(r300->draw, vs->draw);
+        FREE(vs->state.tokens);
         FREE(shader);
     } else {
         draw_delete_vertex_shader(r300->draw,
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index c4c9784a00..c01e61a9b1 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -22,6 +22,9 @@
 
 #include "r300_state_derived.h"
 
+#include "r300_fs.h"
+#include "r300_vs.h"
+
 /* r300_state_derived: Various bits of state which are dependent upon
  * currently bound CSO data. */
 
@@ -46,63 +49,71 @@ static void r300_vs_tab_routes(struct r300_context* r300,
 
     assert(info->num_inputs <= 16);
 
-    if (r300screen->caps->has_tcl) {
-        /* Just copy vert attribs over as-is. */
+    if (!r300screen->caps->has_tcl || !r300->rs_state->enable_vte)
+    {
         for (i = 0; i < info->num_inputs; i++) {
-            tab[i] = i;
-        }
-        for (i = 0; i < info->num_outputs; i++) {
-            switch (info->output_semantic_name[i]) {
+            switch (info->input_semantic_name[i]) {
                 case TGSI_SEMANTIC_POSITION:
                     pos = TRUE;
+                    tab[i] = 0;
                     break;
                 case TGSI_SEMANTIC_COLOR:
+                    tab[i] = 2 + cols;
                     cols++;
                     break;
                 case TGSI_SEMANTIC_PSIZE:
                     psize = TRUE;
+                    tab[i] = 15;
                     break;
                 case TGSI_SEMANTIC_FOG:
                     fog = TRUE;
+                    /* Fall through */
                 case TGSI_SEMANTIC_GENERIC:
+                    tab[i] = 6 + texs;
                     texs++;
                     break;
                 default:
-                    debug_printf("r300: Unknown vertex output %d\n",
-                        info->output_semantic_name[i]);
+                    debug_printf("r300: Unknown vertex input %d\n",
+                        info->input_semantic_name[i]);
                     break;
             }
         }
-    } else {
+    }
+    else
+    {
+        /* Just copy vert attribs over as-is. */
         for (i = 0; i < info->num_inputs; i++) {
-            switch (info->input_semantic_name[i]) {
+            tab[i] = i;
+        }
+
+        for (i = 0; i < info->num_outputs; i++) {
+            switch (info->output_semantic_name[i]) {
                 case TGSI_SEMANTIC_POSITION:
                     pos = TRUE;
-                    tab[i] = 0;
                     break;
                 case TGSI_SEMANTIC_COLOR:
-                    tab[i] = 2 + cols;
                     cols++;
                     break;
                 case TGSI_SEMANTIC_PSIZE:
                     psize = TRUE;
-                    tab[i] = 15;
                     break;
                 case TGSI_SEMANTIC_FOG:
                     fog = TRUE;
                     /* Fall through */
                 case TGSI_SEMANTIC_GENERIC:
-                    tab[i] = 6 + texs;
                     texs++;
                     break;
                 default:
-                    debug_printf("r300: Unknown vertex input %d\n",
-                        info->input_semantic_name[i]);
+                    debug_printf("r300: Unknown vertex output %d\n",
+                        info->output_semantic_name[i]);
                     break;
             }
         }
     }
 
+    /* XXX magic */
+    assert(texs <= 8);
+
     /* Do the actual vertex_info setup.
      *
      * vertex_info has four uints of hardware-specific data in it.
@@ -140,21 +151,32 @@ static void r300_vs_tab_routes(struct r300_context* r300,
         vinfo->hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i);
     }
 
-    for (i = 0; i < texs; i++) {
+    /* Init i right here, increment it if fog is enabled.
+     * This gets around a double-increment problem. */
+    i = 0;
+
+    if (fog) {
+        i++;
         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE,
-            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i));
+            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0));
         vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i);
         vinfo->hwfmt[3] |= (4 << (3 * i));
     }
 
-    if (fog) {
-        i++;
+    for (i; i < texs; i++) {
         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE,
-            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0));
+            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i));
         vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i);
         vinfo->hwfmt[3] |= (4 << (3 * i));
     }
 
+    /* Handle the case where the vertex shader will be generating some of
+     * the attribs based on its inputs. */
+    if (r300screen->caps->has_tcl &&
+            info->num_inputs < info->num_outputs) {
+        vinfo->num_attribs = info->num_inputs;
+    }
+
     draw_compute_vertex_size(vinfo);
 }
 
@@ -162,26 +184,40 @@ static void r300_vs_tab_routes(struct r300_context* r300,
 static void r300_vertex_psc(struct r300_context* r300,
                             struct r300_vertex_format* vformat)
 {
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct vertex_info* vinfo = &vformat->vinfo;
     int* tab = vformat->vs_tab;
     uint32_t temp;
-    int i;
+    int i, attrib_count;
 
-    debug_printf("r300: attrib count: %d\n", vinfo->num_attribs);
-    for (i = 0; i < vinfo->num_attribs; i++) {
-        debug_printf("r300: attrib: offset %d, interp %d, size %d,"
-               " tab %d\n", vinfo->attrib[i].src_index,
-               vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
-               tab[i]);
+    /* Vertex shaders have no semantics on their inputs,
+     * so PSC should just route stuff based on their info,
+     * and not on attrib information. */
+    if (r300screen->caps->has_tcl) {
+        attrib_count = r300->vs->info.num_inputs;
+        debug_printf("r300: routing %d attribs in psc for vs\n",
+                attrib_count);
+    } else {
+        attrib_count = vinfo->num_attribs;
+        debug_printf("r300: attrib count: %d\n", attrib_count);
+        for (i = 0; i < attrib_count; i++) {
+            debug_printf("r300: attrib: offset %d, interp %d, size %d,"
+                   " tab %d\n", vinfo->attrib[i].src_index,
+                   vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
+                   tab[i]);
+        }
     }
 
-    for (i = 0; i < vinfo->num_attribs; i++) {
+    for (i = 0; i < attrib_count; i++) {
         /* Make sure we have a proper destination for our attribute */
         assert(tab[i] != -1);
 
         /* Add the attribute to the PSC table. */
-        temp = translate_vertex_data_type(vinfo->attrib[i].emit) |
-            (tab[i] << R300_DST_VEC_LOC_SHIFT);
+        temp = r300screen->caps->has_tcl ?
+            R300_DATA_TYPE_FLOAT_4 :
+            translate_vertex_data_type(vinfo->attrib[i].emit);
+        temp |= tab[i] << R300_DST_VEC_LOC_SHIFT;
+
         if (i & 1) {
             vformat->vap_prog_stream_cntl[i >> 1] &= 0x0000ffff;
             vformat->vap_prog_stream_cntl[i >> 1] |= temp << 16;
@@ -206,7 +242,6 @@ static void r300_vertex_psc(struct r300_context* r300,
 /* Update the vertex format. */
 static void r300_update_vertex_format(struct r300_context* r300)
 {
-    struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct r300_vertex_format vformat;
     int i;
 
diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c
index 8bd9b41bd7..7d822fec48 100644
--- a/src/gallium/drivers/r300/r300_state_invariant.c
+++ b/src/gallium/drivers/r300/r300_state_invariant.c
@@ -34,42 +34,31 @@ void r300_emit_invariant_state(struct r300_context* r300)
     struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
     CS_LOCALS(r300);
 
-    BEGIN_CS(30 + (caps->has_tcl ? 2: 0));
+    BEGIN_CS(24 + (caps->has_tcl ? 2: 0));
 
     /*** Graphics Backend (GB) ***/
     /* Various GB enables */
-    OUT_CS_REG(R300_GB_ENABLE, 0x0);
-    /* Subpixel multisampling for AA */
-    OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
-    OUT_CS_REG(R300_GB_MSPOS1, 0x66666666);
-    /* GB tile config and pipe setup */
-    OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_DISABLE |
-        r300_translate_gb_pipes(caps->num_frag_pipes));
+    OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE |
+                               R300_GB_LINE_STUFF_ENABLE  |
+                               R300_GB_TRIANGLE_STUFF_ENABLE);
+    /* Subpixel multisampling for AA
+     * These are commented out because glisse's CS checker doesn't like them.
+     * I presume these will be re-enabled later.
+     * OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
+     * OUT_CS_REG(R300_GB_MSPOS1, 0x6666666);
+     */
     /* Source of fog depth */
     OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W);
     /* AA enable */
     OUT_CS_REG(R300_GB_AA_CONFIG, 0x0);
 
-    /*** Geometry Assembly (GA) ***/
-    /* GA errata fixes. */
-    if (caps->is_r500) {
-        OUT_CS_REG(R300_GA_ENHANCE,
-                R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL |
-                R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE |
-                R500_GA_ENHANCE_REG_READWRITE_ENABLE |
-                R500_GA_ENHANCE_REG_NOSTALL_ENABLE);
-    } else {
-        OUT_CS_REG(R300_GA_ENHANCE,
-                R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL |
-                R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE);
-    }
-
     /*** Fog (FG) ***/
     OUT_CS_REG(R300_FG_FOG_BLEND, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x0);
     OUT_CS_REG(R300_FG_DEPTH_SRC, 0x0);
+    OUT_CS_REG(R300_US_W_FMT, 0x0);
 
     /*** VAP ***/
     /* Max and min vertex index clamp. */
@@ -86,7 +75,7 @@ void r300_emit_invariant_state(struct r300_context* r300)
     END_CS;
 
     /* XXX unsorted stuff from surface_fill */
-    BEGIN_CS(79 + (caps->has_tcl ? 7 : 0));
+    BEGIN_CS(64 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0));
     /* Flush PVS. */
     OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
 
@@ -94,30 +83,26 @@ void r300_emit_invariant_state(struct r300_context* r300)
         R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
         R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
         R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT);
-    /* XXX endian */
     if (caps->has_tcl) {
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP);
-        OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE |
-            R300_PS_UCP_MODE_CLIP_AS_TRIFAN);
         OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
-    } else {
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP |
-                R300_VAP_TCL_BYPASS);
     }
     /* XXX point tex stuffing */
     OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1);
     OUT_CS_32F(0.0);
     OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1);
     OUT_CS_32F(1.0);
+    /* XXX line tex stuffing */
+    OUT_CS_REG_SEQ(R300_GA_LINE_S0, 1);
+    OUT_CS_32F(0.0);
+    OUT_CS_REG_SEQ(R300_GA_LINE_S1, 1);
+    OUT_CS_32F(1.0);
     OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 |
         (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT));
     /* XXX this big chunk should be refactored into rs_state */
-    OUT_CS_REG(R300_GA_LINE_S0, 0x00000000);
-    OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000);
     OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000);
     OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000);
     OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000);
@@ -133,8 +118,10 @@ void r300_emit_invariant_state(struct r300_context* r300)
     OUT_CS_REG(R300_RB3D_CCTL, 0x00000000);
     OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F);
     OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000);
-    OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000);
-    OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF);
+    if (caps->is_r500) {
+        OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000);
+        OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF);
+    }
     OUT_CS_REG(R300_ZB_FORMAT, 0x00000002);
     OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003);
     OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000);
@@ -144,17 +131,9 @@ void r300_emit_invariant_state(struct r300_context* r300)
     OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1);
     OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405);
     OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F);
-    /* Vertex size. */
-    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
 
     /* XXX */
     OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa);
 
-    OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4);
-    OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A);
-    OUT_CS(R300_US_OUT_FMT_UNUSED);
-    OUT_CS(R300_US_OUT_FMT_UNUSED);
-    OUT_CS(R300_US_OUT_FMT_UNUSED);
-    OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0);
     END_CS;
 }
diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r300_state_shader.c
deleted file mode 100644
index 1b02239ee7..0000000000
--- a/src/gallium/drivers/r300/r300_state_shader.c
+++ /dev/null
@@ -1,649 +0,0 @@
-/*
- * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE. */
-
-#include "r300_state_shader.h"
-
-static void r300_copy_passthrough_shader(struct r300_fragment_shader* fs)
-{
-    struct r300_fragment_shader* pt = &r300_passthrough_fragment_shader;
-    fs->shader.stack_size = pt->shader.stack_size;
-    fs->alu_instruction_count = pt->alu_instruction_count;
-    fs->tex_instruction_count = pt->tex_instruction_count;
-    fs->indirections = pt->indirections;
-    fs->instructions[0] = pt->instructions[0];
-}
-
-static void r500_copy_passthrough_shader(struct r500_fragment_shader* fs)
-{
-    struct r500_fragment_shader* pt = &r500_passthrough_fragment_shader;
-    fs->shader.stack_size = pt->shader.stack_size;
-    fs->instruction_count = pt->instruction_count;
-    fs->instructions[0] = pt->instructions[0];
-}
-
-static void r300_fs_declare(struct r300_fs_asm* assembler,
-                            struct tgsi_full_declaration* decl)
-{
-    switch (decl->Declaration.File) {
-        case TGSI_FILE_INPUT:
-            switch (decl->Semantic.SemanticName) {
-                case TGSI_SEMANTIC_COLOR:
-                    assembler->color_count++;
-                    break;
-                case TGSI_SEMANTIC_GENERIC:
-                    assembler->tex_count++;
-                    break;
-                default:
-                    debug_printf("r300: fs: Bad semantic declaration %d\n",
-                        decl->Semantic.SemanticName);
-                    break;
-            }
-            break;
-        case TGSI_FILE_OUTPUT:
-        case TGSI_FILE_CONSTANT:
-            break;
-        case TGSI_FILE_TEMPORARY:
-            assembler->temp_count++;
-            break;
-        default:
-            debug_printf("r300: fs: Bad file %d\n", decl->Declaration.File);
-            break;
-    }
-
-    assembler->temp_offset = assembler->color_count + assembler->tex_count;
-}
-
-static INLINE unsigned r300_fs_src(struct r300_fs_asm* assembler,
-                                   struct tgsi_src_register* src)
-{
-    switch (src->File) {
-        case TGSI_FILE_NULL:
-            return 0;
-        case TGSI_FILE_INPUT:
-            /* XXX may be wrong */
-            return src->Index;
-            break;
-        case TGSI_FILE_TEMPORARY:
-            return src->Index + assembler->temp_offset;
-            break;
-        case TGSI_FILE_IMMEDIATE:
-            return (src->Index + assembler->imm_offset) | (1 << 8);
-            break;
-        case TGSI_FILE_CONSTANT:
-            /* XXX magic */
-            return src->Index | (1 << 8);
-            break;
-        default:
-            debug_printf("r300: fs: Unimplemented src %d\n", src->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler,
-                                   struct tgsi_dst_register* dst)
-{
-    switch (dst->File) {
-        case TGSI_FILE_NULL:
-            /* This happens during KIL instructions. */
-            return 0;
-            break;
-        case TGSI_FILE_OUTPUT:
-            return 0;
-            break;
-        case TGSI_FILE_TEMPORARY:
-            return dst->Index + assembler->temp_offset;
-            break;
-        default:
-            debug_printf("r300: fs: Unimplemented dst %d\n", dst->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r500_fix_swiz(unsigned s)
-{
-    /* For historical reasons, the swizzle values x, y, z, w, and 0 are
-     * equivalent to the actual machine code, but 1 is not. Thus, we just
-     * adjust it a bit... */
-    if (s == TGSI_EXTSWIZZLE_ONE) {
-        return R500_SWIZZLE_ONE;
-    } else {
-        return s;
-    }
-}
-
-static uint32_t r500_rgba_swiz(struct tgsi_full_src_register* reg)
-{
-    if (reg->SrcRegister.Extended) {
-        return r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleX) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleY) << 3) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleZ) << 6) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleW) << 9);
-    } else {
-        return reg->SrcRegister.SwizzleX |
-            (reg->SrcRegister.SwizzleY << 3) |
-            (reg->SrcRegister.SwizzleZ << 6) |
-            (reg->SrcRegister.SwizzleW << 9);
-    }
-}
-
-static uint32_t r500_strq_swiz(struct tgsi_full_src_register* reg)
-{
-    return reg->SrcRegister.SwizzleX |
-        (reg->SrcRegister.SwizzleY << 2) |
-        (reg->SrcRegister.SwizzleZ << 4) |
-        (reg->SrcRegister.SwizzleW << 6);
-}
-
-static INLINE uint32_t r500_rgb_swiz(struct tgsi_full_src_register* reg)
-{
-    /* Only the first 9 bits... */
-    return (r500_rgba_swiz(reg) & 0x1ff) |
-        (reg->SrcRegister.Negate ? (1 << 9) : 0) |
-        (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0);
-}
-
-static INLINE uint32_t r500_alpha_swiz(struct tgsi_full_src_register* reg)
-{
-    /* Only the last 3 bits... */
-    return (r500_rgba_swiz(reg) >> 9) |
-        (reg->SrcRegister.Negate ? (1 << 9) : 0) |
-        (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0);
-}
-
-static INLINE uint32_t r300_rgb_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_MOV:
-            return R300_ALU_OUTC_CMP;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r300_alpha_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_MOV:
-            return R300_ALU_OUTA_CMP;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r500_rgba_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_EX2:
-        case TGSI_OPCODE_LG2:
-        case TGSI_OPCODE_RCP:
-        case TGSI_OPCODE_RSQ:
-            return R500_ALU_RGBA_OP_SOP;
-        case TGSI_OPCODE_FRC:
-            return R500_ALU_RGBA_OP_FRC;
-        case TGSI_OPCODE_DP3:
-            return R500_ALU_RGBA_OP_DP3;
-        case TGSI_OPCODE_DP4:
-        case TGSI_OPCODE_DPH:
-            return R500_ALU_RGBA_OP_DP4;
-        case TGSI_OPCODE_ABS:
-        case TGSI_OPCODE_CMP:
-        case TGSI_OPCODE_MOV:
-        case TGSI_OPCODE_SWZ:
-            return R500_ALU_RGBA_OP_CMP;
-        case TGSI_OPCODE_ADD:
-        case TGSI_OPCODE_MAD:
-        case TGSI_OPCODE_MUL:
-        case TGSI_OPCODE_SUB:
-            return R500_ALU_RGBA_OP_MAD;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r500_alpha_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_EX2:
-            return R500_ALPHA_OP_EX2;
-        case TGSI_OPCODE_LG2:
-            return R500_ALPHA_OP_LN2;
-        case TGSI_OPCODE_RCP:
-            return R500_ALPHA_OP_RCP;
-        case TGSI_OPCODE_RSQ:
-            return R500_ALPHA_OP_RSQ;
-        case TGSI_OPCODE_FRC:
-            return R500_ALPHA_OP_FRC;
-        case TGSI_OPCODE_DP3:
-        case TGSI_OPCODE_DP4:
-        case TGSI_OPCODE_DPH:
-            return R500_ALPHA_OP_DP;
-        case TGSI_OPCODE_ABS:
-        case TGSI_OPCODE_CMP:
-        case TGSI_OPCODE_MOV:
-        case TGSI_OPCODE_SWZ:
-            return R500_ALPHA_OP_CMP;
-        case TGSI_OPCODE_ADD:
-        case TGSI_OPCODE_MAD:
-        case TGSI_OPCODE_MUL:
-        case TGSI_OPCODE_SUB:
-            return R500_ALPHA_OP_MAD;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r500_tex_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_KIL:
-            return R500_TEX_INST_TEXKILL;
-        case TGSI_OPCODE_TEX:
-            return R500_TEX_INST_LD;
-        case TGSI_OPCODE_TXB:
-            return R500_TEX_INST_LODBIAS;
-        case TGSI_OPCODE_TXP:
-            return R500_TEX_INST_PROJ;
-        default:
-            return 0;
-    }
-}
-
-static INLINE void r300_emit_maths(struct r300_fragment_shader* fs,
-                                   struct r300_fs_asm* assembler,
-                                   struct tgsi_full_src_register* src,
-                                   struct tgsi_full_dst_register* dst,
-                                   unsigned op,
-                                   unsigned count)
-{
-    int i = fs->alu_instruction_count;
-
-    fs->instructions[i].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        r300_rgb_op(op);
-    fs->instructions[i].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ;
-    fs->instructions[i].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        r300_alpha_op(op);
-    fs->instructions[i].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT;
-
-    fs->alu_instruction_count++;
-}
-
-/* Setup an ALU operation. */
-static INLINE void r500_emit_alu(struct r500_fragment_shader* fs,
-                                 struct r300_fs_asm* assembler,
-                                 struct tgsi_full_dst_register* dst)
-{
-    int i = fs->instruction_count;
-
-    if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
-        fs->instructions[i].inst0 = R500_INST_TYPE_OUT |
-        R500_ALU_OMASK(dst->DstRegister.WriteMask);
-    } else {
-        fs->instructions[i].inst0 = R500_INST_TYPE_ALU |
-        R500_ALU_WMASK(dst->DstRegister.WriteMask);
-    }
-
-    fs->instructions[i].inst0 |= R500_INST_TEX_SEM_WAIT;
-
-    fs->instructions[i].inst4 =
-        R500_ALPHA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister));
-    fs->instructions[i].inst5 =
-        R500_ALU_RGBA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister));
-}
-
-static INLINE void r500_emit_maths(struct r500_fragment_shader* fs,
-                                   struct r300_fs_asm* assembler,
-                                   struct tgsi_full_src_register* src,
-                                   struct tgsi_full_dst_register* dst,
-                                   unsigned op,
-                                   unsigned count)
-{
-    int i = fs->instruction_count;
-
-    r500_emit_alu(fs, assembler, dst);
-
-    switch (count) {
-        case 3:
-            fs->instructions[i].inst1 =
-                R500_RGB_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister));
-            fs->instructions[i].inst2 =
-                R500_ALPHA_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister));
-            fs->instructions[i].inst5 |=
-                R500_ALU_RGBA_SEL_C_SRC2 |
-                R500_SWIZ_RGBA_C(r500_rgb_swiz(&src[2])) |
-                R500_ALU_RGBA_ALPHA_SEL_C_SRC2 |
-                R500_SWIZ_ALPHA_C(r500_alpha_swiz(&src[2]));
-        case 2:
-            fs->instructions[i].inst1 |=
-                R500_RGB_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister));
-            fs->instructions[i].inst2 |=
-                R500_ALPHA_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister));
-            fs->instructions[i].inst3 =
-                R500_ALU_RGB_SEL_B_SRC1 |
-                R500_SWIZ_RGB_B(r500_rgb_swiz(&src[1]));
-            fs->instructions[i].inst4 |=
-                R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])) |
-                R500_ALPHA_SEL_B_SRC1;
-        case 1:
-        case 0:
-        default:
-            fs->instructions[i].inst1 |=
-                R500_RGB_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister));
-            fs->instructions[i].inst2 |=
-                R500_ALPHA_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister));
-            fs->instructions[i].inst3 |=
-                R500_ALU_RGB_SEL_A_SRC0 |
-                R500_SWIZ_RGB_A(r500_rgb_swiz(&src[0]));
-            fs->instructions[i].inst4 |=
-                R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])) |
-                R500_ALPHA_SEL_A_SRC0;
-            break;
-    }
-
-    fs->instructions[i].inst4 |= r500_alpha_op(op);
-    fs->instructions[i].inst5 |= r500_rgba_op(op);
-
-    fs->instruction_count++;
-}
-
-static INLINE void r500_emit_tex(struct r500_fragment_shader* fs,
-                                 struct r300_fs_asm* assembler,
-                                 struct tgsi_full_src_register* src,
-                                 struct tgsi_full_dst_register* dst,
-                                 uint32_t op)
-{
-    int i = fs->instruction_count;
-
-    fs->instructions[i].inst0 = R500_INST_TYPE_TEX |
-        R500_TEX_WMASK(dst->DstRegister.WriteMask) |
-        R500_INST_TEX_SEM_WAIT;
-    fs->instructions[i].inst1 = R500_TEX_ID(0) |
-        R500_TEX_SEM_ACQUIRE | //R500_TEX_IGNORE_UNCOVERED |
-        r500_tex_op(op);
-    fs->instructions[i].inst2 =
-        R500_TEX_SRC_ADDR(r300_fs_src(assembler, &src->SrcRegister)) |
-        R500_SWIZ_TEX_STRQ(r500_strq_swiz(src)) |
-        R500_TEX_DST_ADDR(r300_fs_dst(assembler, &dst->DstRegister)) |
-        R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G |
-        R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A;
-
-    if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
-        fs->instructions[i].inst2 |=
-            R500_TEX_DST_ADDR(assembler->temp_count +
-                    assembler->temp_offset);
-
-        fs->instruction_count++;
-
-        /* Setup and emit a MOV. */
-        src[0].SrcRegister.Index = assembler->temp_count;
-        src[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-
-        src[1] = src[0];
-        src[2] = r500_constant_zero;
-        r500_emit_maths(fs, assembler, src, dst, TGSI_OPCODE_MOV, 3);
-    } else {
-        fs->instruction_count++;
-    }
-}
-
-static void r300_fs_instruction(struct r300_fragment_shader* fs,
-                                struct r300_fs_asm* assembler,
-                                struct tgsi_full_instruction* inst)
-{
-    switch (inst->Instruction.Opcode) {
-        case TGSI_OPCODE_MOV:
-            /* src0 -> src1 and src2 forced to zero */
-            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r300_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-        case TGSI_OPCODE_END:
-            break;
-        default:
-            debug_printf("r300: fs: Bad opcode %d\n",
-                    inst->Instruction.Opcode);
-            break;
-    }
-}
-
-static void r500_fs_instruction(struct r500_fragment_shader* fs,
-                                struct r300_fs_asm* assembler,
-                                struct tgsi_full_instruction* inst)
-{
-    /* Switch between opcodes. When possible, prefer using the official
-     * AMD/ATI names for opcodes, please, as it facilitates using the
-     * documentation. */
-    switch (inst->Instruction.Opcode) {
-        /* The simple scalar ops. */
-        case TGSI_OPCODE_EX2:
-        case TGSI_OPCODE_LG2:
-        case TGSI_OPCODE_RCP:
-        case TGSI_OPCODE_RSQ:
-            /* Copy red swizzle to alpha for src0 */
-            inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW =
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX;
-            inst->FullSrcRegisters[0].SrcRegister.SwizzleW =
-                inst->FullSrcRegisters[0].SrcRegister.SwizzleX;
-            /* Fall through */
-        case TGSI_OPCODE_FRC:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 1);
-            break;
-
-        /* The dot products. */
-        case TGSI_OPCODE_DPH:
-            /* Set alpha swizzle to one for src0 */
-            if (!inst->FullSrcRegisters[0].SrcRegister.Extended) {
-                inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleX;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleY;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleZ;
-            }
-            inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW =
-                TGSI_EXTSWIZZLE_ONE;
-            /* Fall through */
-        case TGSI_OPCODE_DP3:
-        case TGSI_OPCODE_DP4:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 2);
-            break;
-
-        /* Simple three-source operations. */
-        case TGSI_OPCODE_CMP:
-            /* Swap src0 and src2 */
-            inst->FullSrcRegisters[3] = inst->FullSrcRegisters[2];
-            inst->FullSrcRegisters[2] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[0] = inst->FullSrcRegisters[3];
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-
-        /* The MAD variants. */
-        case TGSI_OPCODE_SUB:
-            /* Just like ADD, but flip the negation on src1 first */
-            inst->FullSrcRegisters[1].SrcRegister.Negate =
-                !inst->FullSrcRegisters[1].SrcRegister.Negate;
-            /* Fall through */
-        case TGSI_OPCODE_ADD:
-            /* Force src0 to one, move all registers over */
-            inst->FullSrcRegisters[2] = inst->FullSrcRegisters[1];
-            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[0] = r500_constant_one;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-        case TGSI_OPCODE_MUL:
-            /* Force our src2 to zero */
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-        case TGSI_OPCODE_MAD:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-
-        /* The MOV variants. */
-        case TGSI_OPCODE_ABS:
-            /* Set absolute value modifiers. */
-            inst->FullSrcRegisters[0].SrcRegisterExtMod.Absolute = TRUE;
-            /* Fall through */
-        case TGSI_OPCODE_MOV:
-        case TGSI_OPCODE_SWZ:
-            /* src0 -> src1 and src2 forced to zero */
-            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-
-        /* The texture instruction set. */
-        case TGSI_OPCODE_KIL:
-        case TGSI_OPCODE_TEX:
-        case TGSI_OPCODE_TXB:
-        case TGSI_OPCODE_TXP:
-            r500_emit_tex(fs, assembler, &inst->FullSrcRegisters[0],
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode);
-            break;
-
-        /* This is the end. My only friend, the end. */
-        case TGSI_OPCODE_END:
-            break;
-        default:
-            debug_printf("r300: fs: Bad opcode %d\n",
-                    inst->Instruction.Opcode);
-            break;
-    }
-
-    /* Clamp, if saturation flags are set. */
-    if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) {
-        fs->instructions[fs->instruction_count - 1].inst0 |=
-            R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP;
-    }
-}
-
-static void r300_fs_finalize(struct r3xx_fragment_shader* fs,
-                             struct r300_fs_asm* assembler)
-{
-    fs->stack_size = assembler->temp_count + assembler->temp_offset;
-}
-
-static void r500_fs_finalize(struct r500_fragment_shader* fs,
-                             struct r300_fs_asm* assembler)
-{
-    /* XXX should this just go with OPCODE_END? */
-    fs->instructions[fs->instruction_count - 1].inst0 |=
-        R500_INST_LAST;
-}
-
-void r300_translate_fragment_shader(struct r300_context* r300,
-                                    struct r3xx_fragment_shader* fs)
-{
-    struct tgsi_parse_context parser;
-    int i;
-    boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
-    struct r300_constant_buffer* consts =
-        &r300->shader_constants[PIPE_SHADER_FRAGMENT];
-
-    struct r300_fs_asm* assembler = CALLOC_STRUCT(r300_fs_asm);
-    if (assembler == NULL) {
-        return;
-    }
-    /* Setup starting offset for immediates. */
-    assembler->imm_offset = consts->user_count;
-
-    /* Make sure we start at the beginning of the shader. */
-    if (is_r500) {
-        ((struct r500_fragment_shader*)fs)->instruction_count = 0;
-    }
-
-    tgsi_parse_init(&parser, fs->state.tokens);
-
-    while (!tgsi_parse_end_of_tokens(&parser)) {
-        tgsi_parse_token(&parser);
-
-        /* This is seriously the lamest way to create fragment programs ever.
-         * I blame TGSI. */
-        switch (parser.FullToken.Token.Type) {
-            case TGSI_TOKEN_TYPE_DECLARATION:
-                /* Allocated registers sitting at the beginning
-                 * of the program. */
-                r300_fs_declare(assembler, &parser.FullToken.FullDeclaration);
-                break;
-            case TGSI_TOKEN_TYPE_IMMEDIATE:
-                debug_printf("r300: Emitting immediate to constant buffer, "
-                        "position %d\n",
-                        assembler->imm_offset + assembler->imm_count);
-                /* I am not amused by the length of these. */
-                for (i = 0; i < 4; i++) {
-                    consts->constants[assembler->imm_offset +
-                        assembler->imm_count][i] =
-                        parser.FullToken.FullImmediate.u.ImmediateFloat32[i]
-                        .Float;
-                }
-                assembler->imm_count++;
-                break;
-            case TGSI_TOKEN_TYPE_INSTRUCTION:
-                if (is_r500) {
-                    r500_fs_instruction((struct r500_fragment_shader*)fs,
-                            assembler, &parser.FullToken.FullInstruction);
-                } else {
-                    r300_fs_instruction((struct r300_fragment_shader*)fs,
-                            assembler, &parser.FullToken.FullInstruction);
-                }
-                break;
-        }
-    }
-
-    debug_printf("r300: fs: %d texs and %d colors, first free reg is %d\n",
-            assembler->tex_count, assembler->color_count,
-            assembler->tex_count + assembler->color_count);
-
-    consts->count = consts->user_count + assembler->imm_count;
-    debug_printf("r300: fs: %d total constants, "
-            "%d from user and %d from immediates\n", consts->count,
-            consts->user_count, assembler->imm_count);
-    r300_fs_finalize(fs, assembler);
-    if (is_r500) {
-        r500_fs_finalize((struct r500_fragment_shader*)fs, assembler);
-    }
-
-    tgsi_dump(fs->state.tokens);
-    /* XXX finish r300 dumper too */
-    if (is_r500) {
-        r500_fs_dump((struct r500_fragment_shader*)fs);
-    }
-
-    tgsi_parse_free(&parser);
-    FREE(assembler);
-}
diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r300_state_shader.h
deleted file mode 100644
index 185fdd90f0..0000000000
--- a/src/gallium/drivers/r300/r300_state_shader.h
+++ /dev/null
@@ -1,230 +0,0 @@
-/*
- * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE. */
-
-#ifndef R300_STATE_SHADER_H
-#define R300_STATE_SHADER_H
-
-#include "tgsi/tgsi_parse.h"
-
-#include "r300_context.h"
-#include "r300_debug.h"
-#include "r300_reg.h"
-#include "r300_screen.h"
-
-/* XXX this all should find its way back to r300_reg */
-/* Swizzle tools */
-#define R500_SWIZZLE_ZERO 4
-#define R500_SWIZZLE_HALF 5
-#define R500_SWIZZLE_ONE 6
-#define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6))
-#define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6))
-#define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6))
-#define R500_SWIZ_MOD_NEG 1
-#define R500_SWIZ_MOD_ABS 2
-#define R500_SWIZ_MOD_NEG_ABS 3
-/* Swizzles for inst2 */
-#define R500_SWIZ_TEX_STRQ(x) ((x) << 8)
-#define R500_SWIZ_TEX_RGBA(x) ((x) << 24)
-/* Swizzles for inst3 */
-#define R500_SWIZ_RGB_A(x) ((x) << 2)
-#define R500_SWIZ_RGB_B(x) ((x) << 15)
-/* Swizzles for inst4 */
-#define R500_SWIZ_ALPHA_A(x) ((x) << 14)
-#define R500_SWIZ_ALPHA_B(x) ((x) << 21)
-/* Swizzle for inst5 */
-#define R500_SWIZ_RGBA_C(x) ((x) << 14)
-#define R500_SWIZ_ALPHA_C(x) ((x) << 27)
-/* Writemasks */
-#define R500_TEX_WMASK(x) ((x) << 11)
-#define R500_ALU_WMASK(x) ((x) << 11)
-#define R500_ALU_OMASK(x) ((x) << 15)
-
-/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're
- * not using enough of it. */
-static const struct tgsi_full_src_register r500_constant_zero = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
-};
-
-static const struct tgsi_full_src_register r500_constant_one = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE,
-};
-
-/* Temporary struct used to hold assembly state while putting together
- * fragment programs. */
-struct r300_fs_asm {
-    /* Pipe context. */
-    struct r300_context* r300;
-    /* Number of colors. */
-    unsigned color_count;
-    /* Number of texcoords. */
-    unsigned tex_count;
-    /* Offset for temporary registers. Inputs and temporaries have no
-     * distinguishing markings, so inputs start at 0 and the first usable
-     * temporary register is after all inputs. */
-    unsigned temp_offset;
-    /* Number of requested temporary registers. */
-    unsigned temp_count;
-    /* Offset for immediate constants. Neither R300 nor R500 can do four
-     * inline constants per source, so instead we copy immediates into the
-     * constant buffer. */
-    unsigned imm_offset;
-    /* Number of immediate constants. */
-    unsigned imm_count;
-};
-
-void r300_translate_fragment_shader(struct r300_context* r300,
-                           struct r3xx_fragment_shader* fs);
-
-static struct r300_fragment_shader r300_passthrough_fragment_shader = {
-    /* XXX This is the emission code. TODO: decode
-    OUT_CS_REG(R300_US_CONFIG, 0);
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
-*/
-    .alu_instruction_count = 1,
-    .tex_instruction_count = 0,
-    .indirections = 0,
-    .shader.stack_size = 1,
-
-    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        R300_ALU_OUTC_CMP,
-    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
-    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        R300_ALU_OUTA_CMP,
-    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
-};
-
-static struct r500_fragment_shader r500_passthrough_fragment_shader = {
-    .shader.stack_size = 0,
-    .instruction_count = 1,
-    .instructions[0].inst0 = R500_INST_TYPE_OUT |
-        R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
-        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
-        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
-    .instructions[0].inst1 =
-        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
-        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST,
-    .instructions[0].inst2 =
-        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
-        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST,
-    .instructions[0].inst3 =
-        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
-        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
-        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
-        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B,
-    .instructions[0].inst4 =
-        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A,
-    .instructions[0].inst5 =
-        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
-        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
-        R500_ALU_RGBA_A_SWIZ_0,
-};
-
-static struct r300_fragment_shader r300_texture_fragment_shader = {
-    /* XXX This is the emission code. TODO: decode
-    OUT_CS_REG(R300_US_CONFIG, 0);
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
-*/
-    .alu_instruction_count = 1,
-    .tex_instruction_count = 0,
-    .indirections = 0,
-    .shader.stack_size = 1,
-
-    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        R300_ALU_OUTC_CMP,
-    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
-    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        R300_ALU_OUTA_CMP,
-    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
-};
-
-static struct r500_fragment_shader r500_texture_fragment_shader = {
-    .shader.stack_size = 1,
-    .instruction_count = 2,
-    .instructions[0].inst0 = R500_INST_TYPE_TEX |
-        R500_INST_TEX_SEM_WAIT |
-        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
-        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
-    .instructions[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD |
-        R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED,
-    .instructions[0].inst2 = R500_TEX_SRC_ADDR(0) |
-        R500_TEX_SRC_S_SWIZ_R | R500_TEX_SRC_T_SWIZ_G |
-        R500_TEX_SRC_R_SWIZ_B | R500_TEX_SRC_Q_SWIZ_A |
-        R500_TEX_DST_ADDR(0) |
-        R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G |
-        R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A,
-    .instructions[0].inst3 = 0x0,
-    .instructions[0].inst4 = 0x0,
-    .instructions[0].inst5 = 0x0,
-    .instructions[1].inst0 = R500_INST_TYPE_OUT |
-        R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
-        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
-        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
-    .instructions[1].inst1 =
-        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
-        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST,
-    .instructions[1].inst2 =
-        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
-        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST,
-    .instructions[1].inst3 =
-        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
-        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
-        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
-        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B,
-    .instructions[1].inst4 =
-        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A,
-    .instructions[1].inst5 =
-        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
-        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
-        R500_ALU_RGBA_A_SWIZ_0,
-};
-
-#endif /* R300_STATE_SHADER_H */
diff --git a/src/gallium/drivers/r300/r300_state_tcl.c b/src/gallium/drivers/r300/r300_state_tcl.c
deleted file mode 100644
index d84912de48..0000000000
--- a/src/gallium/drivers/r300/r300_state_tcl.c
+++ /dev/null
@@ -1,353 +0,0 @@
-/*
- * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE. */
-
-#include "r300_state_tcl.h"
-
-static void r300_vs_declare(struct r300_vs_asm* assembler,
-                            struct tgsi_full_declaration* decl)
-{
-    switch (decl->Declaration.File) {
-        case TGSI_FILE_INPUT:
-            break;
-        case TGSI_FILE_OUTPUT:
-            switch (decl->Semantic.SemanticName) {
-                case TGSI_SEMANTIC_POSITION:
-                    assembler->tab[decl->DeclarationRange.First] = 0;
-                    break;
-                case TGSI_SEMANTIC_COLOR:
-                    assembler->tab[decl->DeclarationRange.First] =
-                        (assembler->point_size ? 1 : 0) +
-                        assembler->out_colors;
-                    break;
-                case TGSI_SEMANTIC_FOG:
-                case TGSI_SEMANTIC_GENERIC:
-                    /* XXX multiple? */
-                    assembler->tab[decl->DeclarationRange.First] =
-                        (assembler->point_size ? 1 : 0) +
-                        assembler->out_colors +
-                        assembler->out_texcoords;
-                    break;
-                case TGSI_SEMANTIC_PSIZE:
-                    assembler->tab[decl->DeclarationRange.First] = 1;
-                    break;
-                default:
-                    debug_printf("r300: vs: Bad semantic declaration %d\n",
-                        decl->Semantic.SemanticName);
-                    break;
-            }
-            break;
-        case TGSI_FILE_CONSTANT:
-            break;
-        case TGSI_FILE_TEMPORARY:
-            assembler->temp_count++;
-            break;
-        default:
-            debug_printf("r300: vs: Bad file %d\n", decl->Declaration.File);
-            break;
-    }
-}
-
-static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler,
-                                        struct tgsi_src_register* src)
-{
-    switch (src->File) {
-        case TGSI_FILE_NULL:
-            /* Probably a zero or one swizzle */
-            return R300_PVS_SRC_REG_INPUT;
-            break;
-        case TGSI_FILE_INPUT:
-            return R300_PVS_SRC_REG_INPUT;
-            break;
-        case TGSI_FILE_TEMPORARY:
-            return R300_PVS_SRC_REG_TEMPORARY;
-            break;
-        case TGSI_FILE_CONSTANT:
-            return R300_PVS_SRC_REG_CONSTANT;
-        default:
-            debug_printf("r300: vs: Unimplemented src type %d\n", src->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r300_vs_dst_type(struct r300_vs_asm* assembler,
-                                        struct tgsi_dst_register* dst)
-{
-    switch (dst->File) {
-        case TGSI_FILE_TEMPORARY:
-            return R300_PVS_DST_REG_TEMPORARY;
-            break;
-        case TGSI_FILE_OUTPUT:
-            return R300_PVS_DST_REG_OUT;
-            break;
-        default:
-            debug_printf("r300: vs: Unimplemented dst type %d\n", dst->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler,
-                                   struct tgsi_dst_register* dst)
-{
-    switch (dst->File) {
-        case TGSI_FILE_TEMPORARY:
-            return dst->Index;
-            break;
-        case TGSI_FILE_OUTPUT:
-            return assembler->tab[dst->Index];
-            break;
-        default:
-            debug_printf("r300: vs: Unimplemented dst %d\n", dst->File);
-            break;
-    }
-    return 0;
-}
-
-static uint32_t r300_vs_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_DP3:
-        case TGSI_OPCODE_DP4:
-            return R300_VE_DOT_PRODUCT;
-        case TGSI_OPCODE_MUL:
-            return R300_VE_MULTIPLY;
-        case TGSI_OPCODE_ADD:
-        case TGSI_OPCODE_MOV:
-        case TGSI_OPCODE_SWZ:
-            return R300_VE_ADD;
-        case TGSI_OPCODE_MAD:
-            return R300_PVS_DST_MACRO_INST | R300_PVS_MACRO_OP_2CLK_MADD;
-        default:
-            break;
-    }
-    return 0;
-}
-
-static uint32_t r300_vs_swiz(struct tgsi_full_src_register* reg)
-{
-    if (reg->SrcRegister.Extended) {
-        return reg->SrcRegisterExtSwz.ExtSwizzleX |
-            (reg->SrcRegisterExtSwz.ExtSwizzleY << 3) |
-            (reg->SrcRegisterExtSwz.ExtSwizzleZ << 6) |
-            (reg->SrcRegisterExtSwz.ExtSwizzleW << 9);
-    } else {
-        return reg->SrcRegister.SwizzleX |
-            (reg->SrcRegister.SwizzleY << 3) |
-            (reg->SrcRegister.SwizzleZ << 6) |
-            (reg->SrcRegister.SwizzleW << 9);
-    }
-}
-
-static void r300_vs_emit_inst(struct r300_vertex_shader* vs,
-                              struct r300_vs_asm* assembler,
-                              struct tgsi_full_src_register* src,
-                              struct tgsi_full_dst_register* dst,
-                              unsigned op,
-                              unsigned count)
-{
-    int i = vs->instruction_count;
-    vs->instructions[i].inst0 = R300_PVS_DST_OPCODE(r300_vs_op(op)) |
-        R300_PVS_DST_REG_TYPE(r300_vs_dst_type(assembler, &dst->DstRegister)) |
-        R300_PVS_DST_OFFSET(r300_vs_dst(assembler, &dst->DstRegister)) |
-        R300_PVS_DST_WE_XYZW;
-    switch (count) {
-        case 3:
-            vs->instructions[i].inst3 =
-                R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
-                            &src[2].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[2].SrcRegister.Index) |
-                R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[2]));
-            /* Fall through */
-        case 2:
-            vs->instructions[i].inst2 =
-                R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
-                            &src[1].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[1].SrcRegister.Index) |
-                R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[1]));
-            /* Fall through */
-        case 1:
-            vs->instructions[i].inst1 =
-                R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
-                            &src[0].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[0].SrcRegister.Index) |
-                R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[0]));
-            break;
-    }
-    vs->instruction_count++;
-}
-
-static void r300_vs_instruction(struct r300_vertex_shader* vs,
-                                struct r300_vs_asm* assembler,
-                                struct tgsi_full_instruction* inst)
-{
-    switch (inst->Instruction.Opcode) {
-        case TGSI_OPCODE_ADD:
-        case TGSI_OPCODE_MUL:
-            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
-            break;
-        case TGSI_OPCODE_DP3:
-            /* Set alpha swizzle to zero for src0 and src1 */
-            if (!inst->FullSrcRegisters[0].SrcRegister.Extended) {
-                inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleX;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleY;
-                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ =
-                    inst->FullSrcRegisters[0].SrcRegister.SwizzleZ;
-            }
-            inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW =
-                TGSI_EXTSWIZZLE_ZERO;
-            if (!inst->FullSrcRegisters[1].SrcRegister.Extended) {
-                inst->FullSrcRegisters[1].SrcRegister.Extended = TRUE;
-                inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleX =
-                    inst->FullSrcRegisters[1].SrcRegister.SwizzleX;
-                inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleY =
-                    inst->FullSrcRegisters[1].SrcRegister.SwizzleY;
-                inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleZ =
-                    inst->FullSrcRegisters[1].SrcRegister.SwizzleZ;
-            }
-            inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleW =
-                TGSI_EXTSWIZZLE_ZERO;
-            /* Fall through */
-        case TGSI_OPCODE_DP4:
-            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
-            break;
-        case TGSI_OPCODE_MOV:
-        case TGSI_OPCODE_SWZ:
-            inst->FullSrcRegisters[1] = r300_constant_zero;
-            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
-            break;
-        case TGSI_OPCODE_MAD:
-            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    3);
-            break;
-        case TGSI_OPCODE_END:
-            break;
-        default:
-            debug_printf("r300: vs: Bad opcode %d\n",
-                    inst->Instruction.Opcode);
-            break;
-    }
-}
-
-static void r300_vs_init(struct r300_vertex_shader* vs,
-                         struct r300_vs_asm* assembler)
-{
-    struct tgsi_shader_info* info = &vs->info;
-    int i;
-
-    for (i = 0; i < info->num_outputs; i++) {
-        switch (info->output_semantic_name[i]) {
-            case TGSI_SEMANTIC_PSIZE:
-                assembler->point_size = TRUE;
-                break;
-            case TGSI_SEMANTIC_COLOR:
-                assembler->out_colors++;
-                break;
-            case TGSI_SEMANTIC_FOG:
-            case TGSI_SEMANTIC_GENERIC:
-                assembler->out_texcoords++;
-                break;
-        }
-    }
-}
-
-void r300_translate_vertex_shader(struct r300_context* r300,
-                                  struct r300_vertex_shader* vs)
-{
-    struct tgsi_parse_context parser;
-    int i;
-    struct r300_constant_buffer* consts =
-        &r300->shader_constants[PIPE_SHADER_VERTEX];
-
-    struct r300_vs_asm* assembler = CALLOC_STRUCT(r300_vs_asm);
-    if (assembler == NULL) {
-        return;
-    }
-
-    /* Init assembler. */
-    r300_vs_init(vs, assembler);
-
-    /* Setup starting offset for immediates. */
-    assembler->imm_offset = consts->user_count;
-
-    tgsi_parse_init(&parser, vs->state.tokens);
-
-    while (!tgsi_parse_end_of_tokens(&parser)) {
-        tgsi_parse_token(&parser);
-
-        /* This is seriously the lamest way to create fragment programs ever.
-         * I blame TGSI. */
-        switch (parser.FullToken.Token.Type) {
-            case TGSI_TOKEN_TYPE_DECLARATION:
-                /* Allocated registers sitting at the beginning
-                 * of the program. */
-                r300_vs_declare(assembler, &parser.FullToken.FullDeclaration);
-                break;
-            case TGSI_TOKEN_TYPE_IMMEDIATE:
-                debug_printf("r300: Emitting immediate to constant buffer, "
-                        "position %d\n",
-                        assembler->imm_offset + assembler->imm_count);
-                /* I am not amused by the length of these. */
-                for (i = 0; i < 4; i++) {
-                    consts->constants[assembler->imm_offset +
-                        assembler->imm_count][i] =
-                        parser.FullToken.FullImmediate.u.ImmediateFloat32[i]
-                        .Float;
-                }
-                assembler->imm_count++;
-                break;
-            case TGSI_TOKEN_TYPE_INSTRUCTION:
-                r300_vs_instruction(vs, assembler,
-                        &parser.FullToken.FullInstruction);
-                break;
-        }
-    }
-
-    debug_printf("r300: vs: %d texs and %d colors, first free reg is %d\n",
-            assembler->tex_count, assembler->color_count,
-            assembler->tex_count + assembler->color_count);
-
-    consts->count = consts->user_count + assembler->imm_count;
-    debug_printf("r300: vs: %d total constants, "
-            "%d from user and %d from immediates\n", consts->count,
-            consts->user_count, assembler->imm_count);
-
-    debug_printf("r300: vs: tab: %d %d %d %d\n", assembler->tab[0],
-            assembler->tab[1], assembler->tab[2], assembler->tab[3]);
-
-    tgsi_dump(vs->state.tokens);
-    /* XXX finish r300 vertex shader dumper */
-    r300_vs_dump(vs);
-
-    tgsi_parse_free(&parser);
-    FREE(assembler);
-}
diff --git a/src/gallium/drivers/r300/r300_state_tcl.h b/src/gallium/drivers/r300/r300_state_tcl.h
deleted file mode 100644
index e2e1357d43..0000000000
--- a/src/gallium/drivers/r300/r300_state_tcl.h
+++ /dev/null
@@ -1,153 +0,0 @@
-/*
- * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE. */
-
-#ifndef R300_STATE_TCL_H
-#define R300_STATE_TCL_H
-
-#include "tgsi/tgsi_parse.h"
-
-#include "r300_context.h"
-#include "r300_debug.h"
-#include "r300_reg.h"
-#include "r300_screen.h"
-
-/* XXX get these to r300_reg */
-#define R300_PVS_DST_OPCODE(x)   ((x) << 0)
-#   define R300_VE_DOT_PRODUCT            1
-#   define R300_VE_MULTIPLY               2
-#   define R300_VE_ADD                    3
-#define R300_PVS_DST_MACRO_INST    (1 << 7)
-#   define R300_PVS_MACRO_OP_2CLK_MADD    0
-#define R300_PVS_DST_REG_TYPE(x) ((x) << 8)
-#   define R300_PVS_DST_REG_TEMPORARY     0
-#   define R300_PVS_DST_REG_A0            1
-#   define R300_PVS_DST_REG_OUT           2
-#   define R300_PVS_DST_REG_OUT_REPL_X    3
-#   define R300_PVS_DST_REG_ALT_TEMPORARY 4
-#   define R300_PVS_DST_REG_INPUT         5
-#define R300_PVS_DST_OFFSET(x)   ((x) << 13)
-#define R300_PVS_DST_WE(x)       ((x) << 20)
-#define R300_PVS_DST_WE_XYZW     (0xf << 20)
-
-#define R300_PVS_SRC_REG_TYPE(x) ((x) << 0)
-#   define R300_PVS_SRC_REG_TEMPORARY     0
-#   define R300_PVS_SRC_REG_INPUT         1
-#   define R300_PVS_SRC_REG_CONSTANT      2
-#   define R300_PVS_SRC_REG_ALT_TEMPORARY 3
-#define R300_PVS_SRC_OFFSET(x)   ((x) << 5)
-#define R300_PVS_SRC_SWIZZLE(x)  ((x) << 13)
-#   define R300_PVS_SRC_SELECT_X          0
-#   define R300_PVS_SRC_SELECT_Y          1
-#   define R300_PVS_SRC_SELECT_Z          2
-#   define R300_PVS_SRC_SELECT_W          3
-#   define R300_PVS_SRC_SELECT_FORCE_0    4
-#   define R300_PVS_SRC_SELECT_FORCE_1    5
-#   define R300_PVS_SRC_SWIZZLE_XYZW \
-    ((R300_PVS_SRC_SELECT_X | (R300_PVS_SRC_SELECT_Y << 3) | \
-     (R300_PVS_SRC_SELECT_Z << 6) | (R300_PVS_SRC_SELECT_W << 9)) << 13)
-#   define R300_PVS_SRC_SWIZZLE_ZERO \
-    ((R300_PVS_SRC_SELECT_FORCE_0 | (R300_PVS_SRC_SELECT_FORCE_0 << 3) | \
-     (R300_PVS_SRC_SELECT_FORCE_0 << 6) | \
-      (R300_PVS_SRC_SELECT_FORCE_0 << 9)) << 13)
-#   define R300_PVS_SRC_SWIZZLE_ONE \
-    ((R300_PVS_SRC_SELECT_FORCE_1 | (R300_PVS_SRC_SELECT_FORCE_1 << 3) | \
-     (R300_PVS_SRC_SELECT_FORCE_1 << 6) | \
-      (R300_PVS_SRC_SELECT_FORCE_1 << 9)) << 13)
-
-static const struct tgsi_full_src_register r300_constant_zero = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
-};
-
-/* Temporary struct used to hold assembly state while putting together
- * fragment programs. */
-struct r300_vs_asm {
-    /* Pipe context. */
-    struct r300_context* r300;
-    /* Number of colors. */
-    unsigned color_count;
-    /* Number of texcoords. */
-    unsigned tex_count;
-    /* Number of requested temporary registers. */
-    unsigned temp_count;
-    /* Offset for immediate constants. Neither R300 nor R500 can do four
-     * inline constants per source, so instead we copy immediates into the
-     * constant buffer. */
-    unsigned imm_offset;
-    /* Number of immediate constants. */
-    unsigned imm_count;
-    /* Number of colors to write. */
-    unsigned out_colors;
-    /* Number of texcoords to write. */
-    unsigned out_texcoords;
-    /* Whether to emit point size. */
-    boolean point_size;
-    /* Tab of declared outputs to OVM outputs. */
-    unsigned tab[16];
-};
-
-static struct r300_vertex_shader r300_passthrough_vertex_shader = {
-        /* XXX translate these back into normal instructions */
-    .instruction_count = 2,
-    .instructions[0].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
-        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(0) | R300_PVS_DST_WE_XYZW,
-    .instructions[0].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
-        R300_PVS_SRC_OFFSET(0) | R300_PVS_SRC_SWIZZLE_XYZW,
-    .instructions[0].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
-    .instructions[0].inst3 = 0x0,
-    .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
-        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW,
-    .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
-        R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW,
-    .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
-    .instructions[1].inst3 = 0x0,
-};
-
-static struct r300_vertex_shader r300_texture_vertex_shader = {
-        /* XXX translate these back into normal instructions */
-    .instruction_count = 2,
-    .instructions[0].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
-        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(0) | R300_PVS_DST_WE_XYZW,
-    .instructions[0].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
-        R300_PVS_SRC_OFFSET(0) | R300_PVS_SRC_SWIZZLE_XYZW,
-    .instructions[0].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
-    .instructions[0].inst3 = 0x0,
-    .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
-        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW,
-    .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
-        R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW,
-    .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
-    .instructions[1].inst3 = 0x0,
-};
-
-void r300_translate_vertex_shader(struct r300_context* r300,
-                                  struct r300_vertex_shader* vs);
-
-#endif /* R300_STATE_TCL_H */
diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c
index 4dd5b8af99..96e6e4a77d 100644
--- a/src/gallium/drivers/r300/r300_surface.c
+++ b/src/gallium/drivers/r300/r300_surface.c
@@ -23,42 +23,48 @@
 
 #include "r300_surface.h"
 
-static void r300_surface_setup(struct pipe_context* pipe,
-                               struct pipe_surface* dest,
+static void r300_surface_setup(struct r300_context* r300,
+                               struct r300_texture* dest,
                                unsigned x, unsigned y,
                                unsigned w, unsigned h)
 {
-    struct r300_context* r300 = r300_context(pipe);
-    struct r300_capabilities* caps = r300_screen(pipe->screen)->caps;
-    struct r300_texture* tex = (struct r300_texture*)dest->texture;
-    unsigned pixpitch = tex->stride / tex->tex.block.size;
+    struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
+    unsigned pixpitch = dest->stride / dest->tex.block.size;
     CS_LOCALS(r300);
 
-    /* Make sure our target BO is okay. */
-    r300->winsys->add_buffer(r300->winsys, tex->buffer,
-            0, RADEON_GEM_DOMAIN_VRAM);
-    if (r300->winsys->validate(r300->winsys)) {
-        r300->context.flush(&r300->context, 0, NULL);
-    }
-
     r300_emit_blend_state(r300, &blend_clear_state);
     r300_emit_blend_color_state(r300, &blend_color_clear_state);
     r300_emit_dsa_state(r300, &dsa_clear_state);
     r300_emit_rs_state(r300, &rs_clear_state);
 
-    BEGIN_CS(15);
+    BEGIN_CS(26);
+
+    /* Viewport setup */
+    OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+    OUT_CS_32F((float)w);
+    OUT_CS_32F((float)x);
+    OUT_CS_32F((float)h);
+    OUT_CS_32F((float)y);
+    OUT_CS_32F(1.0);
+    OUT_CS_32F(0.0);
+
+    OUT_CS_REG(R300_VAP_VTE_CNTL, R300_VPORT_X_SCALE_ENA |
+            R300_VPORT_X_OFFSET_ENA |
+            R300_VPORT_Y_SCALE_ENA |
+            R300_VPORT_Y_OFFSET_ENA |
+            R300_VTX_XY_FMT | R300_VTX_Z_FMT);
 
     /* Pixel scissors. */
     OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2);
     if (caps->is_r500) {
         OUT_CS((x << R300_SCISSORS_X_SHIFT) | (y << R300_SCISSORS_Y_SHIFT));
-        OUT_CS((w << R300_SCISSORS_X_SHIFT) | (h << R300_SCISSORS_Y_SHIFT));
+        OUT_CS(((w - 1) << R300_SCISSORS_X_SHIFT) | ((h - 1) << R300_SCISSORS_Y_SHIFT));
     } else {
         /* Non-R500 chipsets have an offset of 1440 in their scissors. */
         OUT_CS(((x + 1440) << R300_SCISSORS_X_SHIFT) |
                 ((y + 1440) << R300_SCISSORS_Y_SHIFT));
-        OUT_CS(((w + 1440) << R300_SCISSORS_X_SHIFT) |
-                ((h + 1440) << R300_SCISSORS_Y_SHIFT));
+        OUT_CS((((w - 1) + 1440) << R300_SCISSORS_X_SHIFT) |
+                (((h - 1) + 1440) << R300_SCISSORS_Y_SHIFT));
     }
 
     /* Flush colorbuffer and blend caches. */
@@ -71,9 +77,11 @@ static void r300_surface_setup(struct pipe_context* pipe,
 
     /* Setup colorbuffer. */
     OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1);
-    OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
-    OUT_CS_REG(R300_RB3D_COLORPITCH0, pixpitch |
-        r300_translate_colorformat(tex->tex.format));
+    OUT_CS_RELOC(dest->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+    OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0, 1);
+    OUT_CS_RELOC(dest->buffer, pixpitch |
+                 r300_translate_colorformat(dest->tex.format), 0,
+                 RADEON_GEM_DOMAIN_VRAM, 0);
     OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0xf);
 
     END_CS;
@@ -93,6 +101,7 @@ static void r300_surface_fill(struct pipe_context* pipe,
     struct r300_capabilities* caps = r300_screen(pipe->screen)->caps;
     struct r300_texture* tex = (struct r300_texture*)dest->texture;
     unsigned pixpitch = tex->stride / tex->tex.block.size;
+    boolean invalid = FALSE;
     CS_LOCALS(r300);
 
     a = (float)((color >> 24) & 0xff) / 255.0f;
@@ -105,19 +114,41 @@ static void r300_surface_fill(struct pipe_context* pipe,
 
     /* Fallback? */
     if (FALSE) {
+fallback:
         debug_printf("r300: Falling back on surface clear...");
         util_surface_fill(pipe, dest, x, y, w, h, color);
         return;
     }
 
-    r300_surface_setup(r300, dest, x, y, w, h);
+    /* Make sure our target BO is okay. */
+validate:
+    if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (!r300->winsys->validate(r300->winsys)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            debug_printf("r300: Stuck in validation loop, gonna fallback.");
+            goto fallback;
+        }
+        invalid = TRUE;
+        goto validate;
+    }
+
+    r300_surface_setup(r300, tex, x, y, w, h);
 
     /* Vertex shader setup */
     if (caps->has_tcl) {
-        r300_emit_vertex_shader(r300, &r300_passthrough_vertex_shader);
+        r300_emit_vertex_program_code(r300, &r300_passthrough_vertex_shader, 0);
     } else {
         BEGIN_CS(4);
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS);
+        OUT_CS_REG(R300_VAP_CNTL_STATUS,
+#ifdef PIPE_ARCH_BIG_ENDIAN
+                   R300_VC_32BIT_SWAP |
+#endif
+                   R300_VAP_TCL_BYPASS);
         OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(5) |
                 R300_PVS_NUM_CNTLRS(5) |
                 R300_PVS_NUM_FPUS(caps->num_vert_fpus) |
@@ -127,14 +158,14 @@ static void r300_surface_fill(struct pipe_context* pipe,
 
     /* Fragment shader setup */
     if (caps->is_r500) {
-        r500_emit_fragment_shader(r300, &r500_passthrough_fragment_shader);
-        r300_emit_rs_block_state(r300, &r500_rs_block_clear_state);
+        r500_emit_fragment_program_code(r300, &r5xx_passthrough_fragment_shader, 0);
+        r300_emit_rs_block_state(r300, &r5xx_rs_block_clear_state);
     } else {
-        r300_emit_fragment_shader(r300, &r300_passthrough_fragment_shader);
-        r300_emit_rs_block_state(r300, &r300_rs_block_clear_state);
+        r300_emit_fragment_program_code(r300, &r3xx_passthrough_fragment_shader, 0);
+        r300_emit_rs_block_state(r300, &r3xx_rs_block_clear_state);
     }
 
-    BEGIN_CS(31);
+    BEGIN_CS(26);
 
     /* VAP stream control, mapping from input memory to PVS/RS memory */
     if (caps->has_tcl) {
@@ -161,27 +192,21 @@ static void r300_surface_fill(struct pipe_context* pipe,
     /* Disable textures */
     OUT_CS_REG(R300_TX_ENABLE, 0x0);
 
-    /* Viewport setup */
-    OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F((float)x);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F((float)y);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F(0.0);
-
     /* The size of the point we're about to draw, in sixths of pixels */
     OUT_CS_REG(R300_GA_POINT_SIZE,
-        ((h * 6) & R300_POINTSIZE_Y_MASK) |
+        ((h * 6)  & R300_POINTSIZE_Y_MASK) |
         ((w * 6) << R300_POINTSIZE_X_SHIFT));
 
+    /* Vertex size. */
+    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
+
     /* Packet3 with our point vertex */
     OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8);
     OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
             (1 << R300_PRIM_NUM_VERTICES_SHIFT));
     /* Position */
-    OUT_CS_32F(w / 2.0);
-    OUT_CS_32F(h / 2.0);
+    OUT_CS_32F(0.5);
+    OUT_CS_32F(0.5);
     OUT_CS_32F(1.0);
     OUT_CS_32F(1.0);
     /* Color */
@@ -190,11 +215,7 @@ static void r300_surface_fill(struct pipe_context* pipe,
     OUT_CS_32F(b);
     OUT_CS_32F(a);
 
-    /* XXX figure out why this is 0xA and not 0x2 */
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
-    /* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */
 
     END_CS;
 
@@ -213,30 +234,63 @@ static void r300_surface_copy(struct pipe_context* pipe,
     struct r300_texture* srctex = (struct r300_texture*)src->texture;
     struct r300_texture* desttex = (struct r300_texture*)dest->texture;
     unsigned pixpitch = srctex->stride / srctex->tex.block.size;
+    boolean invalid = FALSE;
+    float fsrcx = srcx, fsrcy = srcy, fdestx = destx, fdesty = desty;
     CS_LOCALS(r300);
 
     debug_printf("r300: Copying surface %p at (%d,%d) to %p at (%d, %d),"
         " dimensions %dx%d (pixel pitch %d)\n",
         src, srcx, srcy, dest, destx, desty, w, h, pixpitch);
 
-    if ((srctex == desttex) &&
+    if ((srctex->buffer == desttex->buffer) &&
             ((destx < srcx + w) || (srcx < destx + w)) &&
             ((desty < srcy + h) || (srcy < desty + h))) {
+fallback:
         debug_printf("r300: Falling back on surface_copy\n");
         util_surface_copy(pipe, FALSE, dest, destx, desty, src,
                 srcx, srcy, w, h);
     }
 
-    r300_emit_sampler(r300, &r300_sampler_copy_state, 0);
-    r300_emit_texture(r300, srctex, 0);
+    /* Add our target BOs to the list. */
+validate:
+    if (!r300->winsys->add_buffer(r300->winsys, srctex->buffer,
+                RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (!r300->winsys->add_buffer(r300->winsys, desttex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (!r300->winsys->validate(r300->winsys)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            debug_printf("r300: Stuck in validation loop, gonna fallback.");
+            goto fallback;
+        }
+        invalid = TRUE;
+        goto validate;
+    }
+
+    r300_surface_setup(r300, desttex, destx, desty, w, h);
+
+    /* Setup the texture. */
+    r300_emit_texture(r300, &r300_sampler_copy_state, srctex, 0);
+
+    /* Flush and enable. */
     r300_flush_textures(r300);
 
     /* Vertex shader setup */
     if (caps->has_tcl) {
-        r300_emit_vertex_shader(r300, &r300_texture_vertex_shader);
+        r300_emit_vertex_program_code(r300, &r300_passthrough_vertex_shader, 0);
     } else {
         BEGIN_CS(4);
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS);
+        OUT_CS_REG(R300_VAP_CNTL_STATUS,
+#ifdef PIPE_ARCH_BIG_ENDIAN
+                   R300_VC_32BIT_SWAP |
+#endif
+                   R300_VAP_TCL_BYPASS);
         OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(5) |
                 R300_PVS_NUM_CNTLRS(5) |
                 R300_PVS_NUM_FPUS(caps->num_vert_fpus) |
@@ -246,13 +300,14 @@ static void r300_surface_copy(struct pipe_context* pipe,
 
     /* Fragment shader setup */
     if (caps->is_r500) {
-        r500_emit_fragment_shader(r300, &r500_texture_fragment_shader);
-        r300_emit_rs_block_state(r300, &r500_rs_block_copy_state);
+        r500_emit_fragment_program_code(r300, &r5xx_texture_fragment_shader, 0);
+        r300_emit_rs_block_state(r300, &r5xx_rs_block_copy_state);
     } else {
-        r300_emit_fragment_shader(r300, &r300_texture_fragment_shader);
-        r300_emit_rs_block_state(r300, &r300_rs_block_copy_state);
+        r300_emit_fragment_program_code(r300, &r3xx_texture_fragment_shader, 0);
+        r300_emit_rs_block_state(r300, &r3xx_rs_block_copy_state);
     }
 
+    BEGIN_CS(30);
     /* VAP stream control, mapping from input memory to PVS/RS memory */
     if (caps->has_tcl) {
         OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0,
@@ -275,33 +330,38 @@ static void r300_surface_copy(struct pipe_context* pipe,
     /* Two components of texture 0 */
     OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x2);
 
+    /* Vertex size. */
+    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x4);
+
     /* Packet3 with our texcoords */
-    OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8);
+    OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 16);
     OUT_CS(R300_PRIM_TYPE_QUADS | R300_PRIM_WALK_RING |
             (4 << R300_PRIM_NUM_VERTICES_SHIFT));
     /* (x    , y    ) */
-    OUT_CS_32F((float)destx);
-    OUT_CS_32F((float)desty);
-    OUT_CS_32F((float)srcx);
-    OUT_CS_32F((float)srcy);
+    OUT_CS_32F(fdestx / dest->width);
+    OUT_CS_32F(fdesty / dest->height);
+    OUT_CS_32F(fsrcx  / src->width);
+    OUT_CS_32F(fsrcy  / src->height);
     /* (x    , y + h) */
-    OUT_CS_32F((float)destx);
-    OUT_CS_32F((float)(desty + h));
-    OUT_CS_32F((float)srcx);
-    OUT_CS_32F((float)(srcy + h));
+    OUT_CS_32F(fdestx / dest->width);
+    OUT_CS_32F((fdesty + h) / dest->height);
+    OUT_CS_32F(fsrcx  / src->width);
+    OUT_CS_32F((fsrcy  + h) / src->height);
     /* (x + w, y + h) */
-    OUT_CS_32F((float)(destx + w));
-    OUT_CS_32F((float)(desty + h));
-    OUT_CS_32F((float)(srcx + w));
-    OUT_CS_32F((float)(srcy + h));
+    OUT_CS_32F((fdestx + w) / dest->width);
+    OUT_CS_32F((fdesty + h) / dest->height);
+    OUT_CS_32F((fsrcx  + w) / src->width);
+    OUT_CS_32F((fsrcy  + h) / src->height);
     /* (x + w, y    ) */
-    OUT_CS_32F((float)(destx + w));
-    OUT_CS_32F((float)desty);
-    OUT_CS_32F((float)(srcx + w));
-    OUT_CS_32F((float)srcy);
+    OUT_CS_32F((fdestx + w) / dest->width);
+    OUT_CS_32F(fdesty / dest->height);
+    OUT_CS_32F((fsrcx  + w) / src->width);
+    OUT_CS_32F(fsrcy  / src->height);
 
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
 
+    END_CS;
+
     r300->dirty_hw++;
 }
 
diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h
index 894def07aa..f9e98b2ec9 100644
--- a/src/gallium/drivers/r300/r300_surface.h
+++ b/src/gallium/drivers/r300/r300_surface.h
@@ -31,8 +31,8 @@
 #include "r300_context.h"
 #include "r300_cs.h"
 #include "r300_emit.h"
-#include "r300_state_shader.h"
-#include "r300_state_tcl.h"
+#include "r300_fs.h"
+#include "r300_vs.h"
 #include "r300_state_inlines.h"
 
 static struct r300_blend_state blend_clear_state = {
@@ -72,17 +72,17 @@ static struct r300_rs_state rs_clear_state = {
     .color_control = R300_SHADE_MODEL_FLAT,
 };
 
-static struct r300_rs_block r300_rs_block_clear_state = {
-    .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) |
-        R500_RS_SEL_T(R300_RS_SEL_K0) |
-        R500_RS_SEL_R(R300_RS_SEL_K0) |
+static struct r300_rs_block r3xx_rs_block_clear_state = {
+    .ip[0] = R500_RS_SEL_S(R300_RS_SEL_C0) |
+        R500_RS_SEL_T(R300_RS_SEL_C0) |
+        R500_RS_SEL_R(R300_RS_SEL_C0) |
         R500_RS_SEL_Q(R300_RS_SEL_K1),
     .inst[0] = R300_RS_INST_COL_CN_WRITE,
     .count = R300_IT_COUNT(0) | R300_IC_COUNT(1) | R300_HIRES_EN,
     .inst_count = 0,
 };
 
-static struct r300_rs_block r500_rs_block_clear_state = {
+static struct r300_rs_block r5xx_rs_block_clear_state = {
     .ip[0] = R500_RS_SEL_S(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_T(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
@@ -94,24 +94,24 @@ static struct r300_rs_block r500_rs_block_clear_state = {
 
 /* The following state is used for surface_copy only. */
 
-static struct r300_rs_block r300_rs_block_copy_state = {
+static struct r300_rs_block r3xx_rs_block_copy_state = {
     .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) |
         R500_RS_SEL_T(R300_RS_SEL_K0) |
         R500_RS_SEL_R(R300_RS_SEL_K0) |
         R500_RS_SEL_Q(R300_RS_SEL_K1),
     .inst[0] = R300_RS_INST_COL_CN_WRITE,
     .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN,
-    .inst_count = R300_RS_TX_OFFSET(6),
+    .inst_count = R300_RS_TX_OFFSET(0),
 };
 
-static struct r300_rs_block r500_rs_block_copy_state = {
+static struct r300_rs_block r5xx_rs_block_copy_state = {
     .ip[0] = R500_RS_SEL_S(0) |
         R500_RS_SEL_T(1) |
         R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_Q(R500_RS_IP_PTR_K1),
     .inst[0] = R500_RS_INST_TEX_CN_WRITE,
     .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN,
-    .inst_count = R300_RS_TX_OFFSET(6),
+    .inst_count = R300_RS_TX_OFFSET(0),
 };
 
 static struct r300_sampler_state r300_sampler_copy_state = {
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index fe91f4e184..590052509c 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -22,33 +22,35 @@
 
 #include "r300_texture.h"
 
-static int minify(int i)
-{
-    return MAX2(1, i >> 1);
-}
-
 static void r300_setup_texture_state(struct r300_texture* tex,
                                      unsigned width,
                                      unsigned height,
-                                     unsigned pitch)
+                                     unsigned pitch,
+                                     unsigned levels)
 {
     struct r300_texture_state* state = &tex->state;
 
     state->format0 = R300_TX_WIDTH((width - 1) & 0x7ff) |
-        R300_TX_HEIGHT((height - 1) & 0x7ff) | R300_TX_PITCH_EN;
+        R300_TX_HEIGHT((height - 1) & 0x7ff) |
+        R300_TX_NUM_LEVELS(levels) |
+        R300_TX_PITCH_EN;
 
     /* XXX */
-    state->format1 = R300_TX_FORMAT_A8R8G8B8;
+    state->format1 = r300_translate_texformat(tex->tex.format);
 
     state->format2 = pitch - 1;
 
-    /* XXX
+    /* Assume (somewhat foolishly) that oversized textures will
+     * not be permitted by the state tracker. */
     if (width > 2048) {
-        state->pitch |= R300_TXWIDTH_11;
+        state->format2 |= R500_TXWIDTH_BIT11;
     }
     if (height > 2048) {
-        state->pitch |= R300_TXHEIGHT_11;
-    } */
+        state->format2 |= R500_TXHEIGHT_BIT11;
+    }
+
+    debug_printf("r300: Set texture state (%dx%d, pitch %d, %d levels)\n",
+            width, height, pitch, levels);
 }
 
 static void r300_setup_miptree(struct r300_texture* tex)
@@ -65,16 +67,24 @@ static void r300_setup_miptree(struct r300_texture* tex)
         }
 
         base->nblocksx[i] = pf_get_nblocksx(&base->block, base->width[i]);
-        base->nblocksy[i] = pf_get_nblocksy(&base->block, base->width[i]);
-
-        /* Radeons enjoy things in multiples of 32. */
-        /* XXX this can be 32 when POT */
-        stride = (base->nblocksx[i] * base->block.size + 63) & ~63;
+        base->nblocksy[i] = pf_get_nblocksy(&base->block, base->height[i]);
+
+        /* Radeons enjoy things in multiples of 64.
+         *
+         * XXX
+         * POT, uncompressed, unmippmapped textures can be aligned to 32,
+         * instead of 64. */
+        stride = align(pf_get_stride(&base->block, base->width[i]), 32);
         size = stride * base->nblocksy[i] * base->depth[i];
 
-        tex->offset[i] = (tex->size + 63) & ~63;
+        tex->offset[i] = align(tex->size, 32);
         tex->size = tex->offset[i] + size;
 
+        debug_printf("r300: Texture miptree: Level %d "
+                "(%dx%dx%d px, pitch %d bytes)\n",
+                i, base->width[i], base->height[i], base->depth[i],
+                stride);
+        /* Save stride of first level to the texture. */
         if (i == 0) {
             tex->stride = stride;
         }
@@ -86,8 +96,6 @@ static struct pipe_texture*
     r300_texture_create(struct pipe_screen* screen,
                         const struct pipe_texture* template)
 {
-    /* XXX struct r300_screen* r300screen = r300_screen(screen); */
-
     struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
 
     if (!tex) {
@@ -100,11 +108,10 @@ static struct pipe_texture*
 
     r300_setup_miptree(tex);
 
-    /* XXX */
-    r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0],
-            tex->tex.width[0]);
+    r300_setup_texture_state(tex, template->width[0], template->height[0],
+            template->width[0], template->last_level);
 
-    tex->buffer = screen->buffer_create(screen, 64,
+    tex->buffer = screen->buffer_create(screen, 1024,
                                         PIPE_BUFFER_USAGE_PIXEL,
                                         tex->size);
 
@@ -166,6 +173,7 @@ static struct pipe_texture*
 {
     struct r300_texture* tex;
 
+    /* XXX we should start doing mips now... */
     if (base->target != PIPE_TEXTURE_2D ||
         base->last_level != 0 ||
         base->depth[0] != 1) {
@@ -183,8 +191,9 @@ static struct pipe_texture*
 
     tex->stride = *stride;
 
+    /* XXX */
     r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0],
-            tex->stride);
+            tex->stride, 0);
 
     pipe_buffer_reference(&tex->buffer, buffer);
 
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
index 98fb5c9a08..3b56f0307c 100644
--- a/src/gallium/drivers/r300/r300_texture.h
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -32,6 +32,52 @@
 
 void r300_init_screen_texture_functions(struct pipe_screen* screen);
 
+/* Note the signature of R300_EASY_TX_FORMAT(A, R, G, B, FORMAT)... */
+static INLINE uint32_t r300_translate_texformat(enum pipe_format format)
+{
+    switch (format) {
+        /* X8 */
+        case PIPE_FORMAT_I8_UNORM:
+            return R300_EASY_TX_FORMAT(X, X, X, X, X8);
+        /* W8Z8Y8X8 */
+        case PIPE_FORMAT_A8R8G8B8_UNORM:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+        case PIPE_FORMAT_R8G8B8A8_UNORM:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8);
+        case PIPE_FORMAT_A8R8G8B8_SRGB:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8) |
+                R300_TX_FORMAT_GAMMA;
+        case PIPE_FORMAT_R8G8B8A8_SRGB:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) |
+                R300_TX_FORMAT_GAMMA;
+        /* DXT1 */
+        case PIPE_FORMAT_DXT1_RGB:
+            return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1);
+        case PIPE_FORMAT_DXT1_RGBA:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1);
+        /* DXT3 */
+        case PIPE_FORMAT_DXT3_RGBA:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3);
+        /* DXT5 */
+        case PIPE_FORMAT_DXT5_RGBA:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5);
+        /* YVYU422 */
+        case PIPE_FORMAT_YCBCR:
+            return R300_EASY_TX_FORMAT(X, Y, Z, ONE, YVYU422) |
+                R300_TX_FORMAT_YUV_TO_RGB;
+        /* W24_FP */
+        case PIPE_FORMAT_Z24S8_UNORM:
+            return R300_EASY_TX_FORMAT(X, X, X, X, W24_FP);
+        default:
+            debug_printf("r300: Implementation error: "
+                "Got unsupported texture format %s in %s\n",
+                pf_name(format), __FUNCTION__);
+            assert(0);
+            break;
+    }
+    return 0;
+}
+
 #ifndef R300_WINSYS_H
 
 boolean r300_get_texture_buffer(struct pipe_texture* texture,
diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
new file mode 100644
index 0000000000..d68a104106
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
@@ -0,0 +1,336 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_tgsi_to_rc.h"
+
+#include "radeon_compiler.h"
+#include "radeon_program.h"
+
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_scan.h"
+#include "tgsi/tgsi_util.h"
+
+
+static unsigned translate_opcode(unsigned opcode)
+{
+    switch(opcode) {
+        case TGSI_OPCODE_ARL: return OPCODE_ARL;
+        case TGSI_OPCODE_MOV: return OPCODE_MOV;
+        case TGSI_OPCODE_LIT: return OPCODE_LIT;
+        case TGSI_OPCODE_RCP: return OPCODE_RCP;
+        case TGSI_OPCODE_RSQ: return OPCODE_RSQ;
+        case TGSI_OPCODE_EXP: return OPCODE_EXP;
+        case TGSI_OPCODE_LOG: return OPCODE_LOG;
+        case TGSI_OPCODE_MUL: return OPCODE_MUL;
+        case TGSI_OPCODE_ADD: return OPCODE_ADD;
+        case TGSI_OPCODE_DP3: return OPCODE_DP3;
+        case TGSI_OPCODE_DP4: return OPCODE_DP4;
+        case TGSI_OPCODE_DST: return OPCODE_DST;
+        case TGSI_OPCODE_MIN: return OPCODE_MIN;
+        case TGSI_OPCODE_MAX: return OPCODE_MAX;
+        case TGSI_OPCODE_SLT: return OPCODE_SLT;
+        case TGSI_OPCODE_SGE: return OPCODE_SGE;
+        case TGSI_OPCODE_MAD: return OPCODE_MAD;
+        case TGSI_OPCODE_SUB: return OPCODE_SUB;
+        case TGSI_OPCODE_LRP: return OPCODE_LRP;
+     /* case TGSI_OPCODE_CND: return OPCODE_CND; */
+        case TGSI_OPCODE_DP2A: return OPCODE_DP2A;
+                                        /* gap */
+        case TGSI_OPCODE_FRC: return OPCODE_FRC;
+     /* case TGSI_OPCODE_CLAMP: return OPCODE_CLAMP; */
+        case TGSI_OPCODE_FLR: return OPCODE_FLR;
+     /* case TGSI_OPCODE_ROUND: return OPCODE_ROUND; */
+        case TGSI_OPCODE_EX2: return OPCODE_EX2;
+        case TGSI_OPCODE_LG2: return OPCODE_LG2;
+        case TGSI_OPCODE_POW: return OPCODE_POW;
+        case TGSI_OPCODE_XPD: return OPCODE_XPD;
+                                        /* gap */
+        case TGSI_OPCODE_ABS: return OPCODE_ABS;
+        case TGSI_OPCODE_RCC: return OPCODE_RCC;
+        case TGSI_OPCODE_DPH: return OPCODE_DPH;
+        case TGSI_OPCODE_COS: return OPCODE_COS;
+        case TGSI_OPCODE_DDX: return OPCODE_DDX;
+        case TGSI_OPCODE_DDY: return OPCODE_DDY;
+     /* case TGSI_OPCODE_KILP: return OPCODE_KILP; */
+        case TGSI_OPCODE_PK2H: return OPCODE_PK2H;
+        case TGSI_OPCODE_PK2US: return OPCODE_PK2US;
+        case TGSI_OPCODE_PK4B: return OPCODE_PK4B;
+        case TGSI_OPCODE_PK4UB: return OPCODE_PK4UB;
+        case TGSI_OPCODE_RFL: return OPCODE_RFL;
+        case TGSI_OPCODE_SEQ: return OPCODE_SEQ;
+        case TGSI_OPCODE_SFL: return OPCODE_SFL;
+        case TGSI_OPCODE_SGT: return OPCODE_SGT;
+        case TGSI_OPCODE_SIN: return OPCODE_SIN;
+        case TGSI_OPCODE_SLE: return OPCODE_SLE;
+        case TGSI_OPCODE_SNE: return OPCODE_SNE;
+        case TGSI_OPCODE_STR: return OPCODE_STR;
+        case TGSI_OPCODE_TEX: return OPCODE_TEX;
+        case TGSI_OPCODE_TXD: return OPCODE_TXD;
+        case TGSI_OPCODE_TXP: return OPCODE_TXP;
+        case TGSI_OPCODE_UP2H: return OPCODE_UP2H;
+        case TGSI_OPCODE_UP2US: return OPCODE_UP2US;
+        case TGSI_OPCODE_UP4B: return OPCODE_UP4B;
+        case TGSI_OPCODE_UP4UB: return OPCODE_UP4UB;
+        case TGSI_OPCODE_X2D: return OPCODE_X2D;
+        case TGSI_OPCODE_ARA: return OPCODE_ARA;
+        case TGSI_OPCODE_ARR: return OPCODE_ARR;
+        case TGSI_OPCODE_BRA: return OPCODE_BRA;
+        case TGSI_OPCODE_CAL: return OPCODE_CAL;
+        case TGSI_OPCODE_RET: return OPCODE_RET;
+        case TGSI_OPCODE_SSG: return OPCODE_SSG;
+        case TGSI_OPCODE_CMP: return OPCODE_CMP;
+        case TGSI_OPCODE_SCS: return OPCODE_SCS;
+        case TGSI_OPCODE_TXB: return OPCODE_TXB;
+     /* case TGSI_OPCODE_NRM: return OPCODE_NRM; */
+     /* case TGSI_OPCODE_DIV: return OPCODE_DIV; */
+        case TGSI_OPCODE_DP2: return OPCODE_DP2;
+        case TGSI_OPCODE_TXL: return OPCODE_TXL;
+        case TGSI_OPCODE_BRK: return OPCODE_BRK;
+        case TGSI_OPCODE_IF: return OPCODE_IF;
+     /* case TGSI_OPCODE_LOOP: return OPCODE_LOOP; */
+     /* case TGSI_OPCODE_REP: return OPCODE_REP; */
+        case TGSI_OPCODE_ELSE: return OPCODE_ELSE;
+        case TGSI_OPCODE_ENDIF: return OPCODE_ENDIF;
+        case TGSI_OPCODE_ENDLOOP: return OPCODE_ENDLOOP;
+     /* case TGSI_OPCODE_ENDREP: return OPCODE_ENDREP; */
+        case TGSI_OPCODE_PUSHA: return OPCODE_PUSHA;
+        case TGSI_OPCODE_POPA: return OPCODE_POPA;
+     /* case TGSI_OPCODE_CEIL: return OPCODE_CEIL; */
+     /* case TGSI_OPCODE_I2F: return OPCODE_I2F; */
+        case TGSI_OPCODE_NOT: return OPCODE_NOT;
+        case TGSI_OPCODE_TRUNC: return OPCODE_TRUNC;
+     /* case TGSI_OPCODE_SHL: return OPCODE_SHL; */
+     /* case TGSI_OPCODE_SHR: return OPCODE_SHR; */
+        case TGSI_OPCODE_AND: return OPCODE_AND;
+        case TGSI_OPCODE_OR: return OPCODE_OR;
+     /* case TGSI_OPCODE_MOD: return OPCODE_MOD; */
+        case TGSI_OPCODE_XOR: return OPCODE_XOR;
+     /* case TGSI_OPCODE_SAD: return OPCODE_SAD; */
+     /* case TGSI_OPCODE_TXF: return OPCODE_TXF; */
+     /* case TGSI_OPCODE_TXQ: return OPCODE_TXQ; */
+        case TGSI_OPCODE_CONT: return OPCODE_CONT;
+     /* case TGSI_OPCODE_EMIT: return OPCODE_EMIT; */
+     /* case TGSI_OPCODE_ENDPRIM: return OPCODE_ENDPRIM; */
+     /* case TGSI_OPCODE_BGNLOOP2: return OPCODE_BGNLOOP2; */
+        case TGSI_OPCODE_BGNSUB: return OPCODE_BGNSUB;
+     /* case TGSI_OPCODE_ENDLOOP2: return OPCODE_ENDLOOP2; */
+        case TGSI_OPCODE_ENDSUB: return OPCODE_ENDSUB;
+        case TGSI_OPCODE_NOISE1: return OPCODE_NOISE1;
+        case TGSI_OPCODE_NOISE2: return OPCODE_NOISE2;
+        case TGSI_OPCODE_NOISE3: return OPCODE_NOISE3;
+        case TGSI_OPCODE_NOISE4: return OPCODE_NOISE4;
+        case TGSI_OPCODE_NOP: return OPCODE_NOP;
+                                        /* gap */
+        case TGSI_OPCODE_NRM4: return OPCODE_NRM4;
+     /* case TGSI_OPCODE_CALLNZ: return OPCODE_CALLNZ; */
+     /* case TGSI_OPCODE_IFC: return OPCODE_IFC; */
+     /* case TGSI_OPCODE_BREAKC: return OPCODE_BREAKC; */
+        case TGSI_OPCODE_KIL: return OPCODE_KIL;
+        case TGSI_OPCODE_END: return OPCODE_END;
+        case TGSI_OPCODE_SWZ: return OPCODE_SWZ;
+    }
+
+    fprintf(stderr, "Unknown opcode: %i\n", opcode);
+    abort();
+}
+
+static unsigned translate_saturate(unsigned saturate)
+{
+    switch(saturate) {
+        case TGSI_SAT_NONE: return SATURATE_OFF;
+        case TGSI_SAT_ZERO_ONE: return SATURATE_ZERO_ONE;
+        case TGSI_SAT_MINUS_PLUS_ONE: return SATURATE_PLUS_MINUS_ONE;
+    }
+
+    fprintf(stderr, "Unknown saturate mode: %i\n", saturate);
+    abort();
+}
+
+static unsigned translate_register_file(unsigned file)
+{
+    switch(file) {
+        case TGSI_FILE_CONSTANT: return PROGRAM_CONSTANT;
+        case TGSI_FILE_IMMEDIATE: return PROGRAM_CONSTANT;
+        case TGSI_FILE_INPUT: return PROGRAM_INPUT;
+        case TGSI_FILE_OUTPUT: return PROGRAM_OUTPUT;
+        case TGSI_FILE_TEMPORARY: return PROGRAM_TEMPORARY;
+        case TGSI_FILE_ADDRESS: return PROGRAM_ADDRESS;
+    }
+
+    fprintf(stderr, "Unhandled register file: %i\n", file);
+    abort();
+}
+
+static int translate_register_index(
+    struct tgsi_to_rc * ttr,
+    unsigned file,
+    int index)
+{
+    if (file == TGSI_FILE_IMMEDIATE)
+        return ttr->immediate_offset + index;
+
+    return index;
+}
+
+static void transform_dstreg(
+    struct tgsi_to_rc * ttr,
+    struct prog_dst_register * dst,
+    struct tgsi_full_dst_register * src)
+{
+    dst->File = translate_register_file(src->DstRegister.File);
+    dst->Index = translate_register_index(ttr, src->DstRegister.File, src->DstRegister.Index);
+    dst->WriteMask = src->DstRegister.WriteMask;
+    dst->RelAddr = src->DstRegister.Indirect;
+}
+
+static void transform_srcreg(
+    struct tgsi_to_rc * ttr,
+    struct prog_src_register * dst,
+    struct tgsi_full_src_register * src)
+{
+    dst->File = translate_register_file(src->SrcRegister.File);
+    dst->Index = translate_register_index(ttr, src->SrcRegister.File, src->SrcRegister.Index);
+    dst->RelAddr = src->SrcRegister.Indirect;
+    dst->Swizzle = tgsi_util_get_full_src_register_extswizzle(src, 0);
+    dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 1) << 3;
+    dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 2) << 6;
+    dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 3) << 9;
+    dst->Abs = src->SrcRegisterExtMod.Absolute;
+    dst->Negate =
+        src->SrcRegisterExtSwz.NegateX |
+        (src->SrcRegisterExtSwz.NegateY << 1) |
+        (src->SrcRegisterExtSwz.NegateZ << 2) |
+        (src->SrcRegisterExtSwz.NegateW << 3);
+    dst->Negate ^= src->SrcRegister.Negate ? NEGATE_XYZW : 0;
+}
+
+static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_ext_texture src)
+{
+    switch(src.Texture) {
+        case TGSI_TEXTURE_1D:
+            dst->I.TexSrcTarget = TEXTURE_1D_INDEX;
+            break;
+        case TGSI_TEXTURE_2D:
+            dst->I.TexSrcTarget = TEXTURE_2D_INDEX;
+            break;
+        case TGSI_TEXTURE_3D:
+            dst->I.TexSrcTarget = TEXTURE_3D_INDEX;
+            break;
+        case TGSI_TEXTURE_CUBE:
+            dst->I.TexSrcTarget = TEXTURE_CUBE_INDEX;
+            break;
+        case TGSI_TEXTURE_RECT:
+            dst->I.TexSrcTarget = TEXTURE_RECT_INDEX;
+            break;
+        case TGSI_TEXTURE_SHADOW1D:
+            dst->I.TexSrcTarget = TEXTURE_1D_INDEX;
+            dst->I.TexShadow = 1;
+            break;
+        case TGSI_TEXTURE_SHADOW2D:
+            dst->I.TexSrcTarget = TEXTURE_2D_INDEX;
+            dst->I.TexShadow = 1;
+            break;
+        case TGSI_TEXTURE_SHADOWRECT:
+            dst->I.TexSrcTarget = TEXTURE_RECT_INDEX;
+            dst->I.TexShadow = 1;
+            break;
+    }
+}
+
+static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_instruction * src)
+{
+    if (src->Instruction.Opcode == TGSI_OPCODE_END)
+        return;
+
+    struct rc_instruction * dst = rc_insert_new_instruction(ttr->compiler, ttr->compiler->Program.Instructions.Prev);
+    int i;
+
+    dst->I.Opcode = translate_opcode(src->Instruction.Opcode);
+    dst->I.SaturateMode = translate_saturate(src->Instruction.Saturate);
+
+    if (src->Instruction.NumDstRegs)
+        transform_dstreg(ttr, &dst->I.DstReg, &src->FullDstRegisters[0]);
+
+    for(i = 0; i < src->Instruction.NumSrcRegs; ++i) {
+        if (src->FullSrcRegisters[i].SrcRegister.File == TGSI_FILE_SAMPLER)
+            dst->I.TexSrcUnit = src->FullSrcRegisters[i].SrcRegister.Index;
+        else
+            transform_srcreg(ttr, &dst->I.SrcReg[i], &src->FullSrcRegisters[i]);
+    }
+
+    /* Texturing. */
+    transform_texture(dst, src->InstructionExtTexture);
+}
+
+static void handle_immediate(struct tgsi_to_rc * ttr, struct tgsi_full_immediate * imm)
+{
+    struct rc_constant constant;
+    int i;
+
+    constant.Type = RC_CONSTANT_IMMEDIATE;
+    constant.Size = 4;
+    for(i = 0; i < 4; ++i)
+        constant.u.Immediate[i] = imm->u[i].Float;
+    rc_constants_add(&ttr->compiler->Program.Constants, &constant);
+}
+
+void r300_tgsi_to_rc(struct tgsi_to_rc * ttr, const struct tgsi_token * tokens)
+{
+    struct tgsi_parse_context parser;
+    int i;
+
+    /* Allocate constants placeholders.
+     *
+     * Note: What if declared constants are not contiguous? */
+    for(i = 0; i <= ttr->info->file_max[TGSI_FILE_CONSTANT]; ++i) {
+        struct rc_constant constant;
+        memset(&constant, 0, sizeof(constant));
+        constant.Type = RC_CONSTANT_EXTERNAL;
+        constant.Size = 4;
+        constant.u.External = i;
+        rc_constants_add(&ttr->compiler->Program.Constants, &constant);
+    }
+
+    ttr->immediate_offset = ttr->compiler->Program.Constants.Count;
+
+    tgsi_parse_init(&parser, tokens);
+
+    while (!tgsi_parse_end_of_tokens(&parser)) {
+        tgsi_parse_token(&parser);
+
+        switch (parser.FullToken.Token.Type) {
+            case TGSI_TOKEN_TYPE_DECLARATION:
+                break;
+            case TGSI_TOKEN_TYPE_IMMEDIATE:
+                handle_immediate(ttr, &parser.FullToken.FullImmediate);
+                break;
+            case TGSI_TOKEN_TYPE_INSTRUCTION:
+                transform_instruction(ttr, &parser.FullToken.FullInstruction);
+                break;
+        }
+    }
+
+    tgsi_parse_free(&parser);
+
+    rc_calculate_inputs_outputs(ttr->compiler);
+}
+
diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.h b/src/gallium/drivers/r300/r300_tgsi_to_rc.h
new file mode 100644
index 0000000000..93e90ec6d2
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.h
@@ -0,0 +1,41 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_TGSI_TO_RC_H
+#define R300_TGSI_TO_RC_H
+
+struct radeon_compiler;
+
+struct tgsi_full_declaration;
+struct tgsi_shader_info;
+struct tgsi_token;
+
+struct tgsi_to_rc {
+    struct radeon_compiler * compiler;
+    const struct tgsi_shader_info * info;
+
+    int immediate_offset;
+};
+
+void r300_tgsi_to_rc(struct tgsi_to_rc * ttr, const struct tgsi_token * tokens);
+
+#endif /* R300_TGSI_TO_RC_H */
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
new file mode 100644
index 0000000000..2cb903bba2
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -0,0 +1,234 @@
+/*
+ * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_vs.h"
+
+#include "r300_context.h"
+#include "r300_tgsi_to_rc.h"
+
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "radeon_compiler.h"
+
+
+static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
+{
+    struct r300_vertex_shader * vs = c->UserData;
+    struct tgsi_shader_info* info = &vs->info;
+    boolean pointsize = false;
+    int out_colors = 0;
+    int colors = 0;
+    int out_generic = 0;
+    int generic = 0;
+    int i;
+
+    /* Fill in the input mapping */
+    for (i = 0; i < info->num_inputs; i++)
+        c->code->inputs[i] = i;
+
+    /* Fill in the output mapping */
+    for (i = 0; i < info->num_outputs; i++) {
+        switch (info->output_semantic_name[i]) {
+            case TGSI_SEMANTIC_PSIZE:
+                pointsize = true;
+                break;
+            case TGSI_SEMANTIC_COLOR:
+                out_colors++;
+                break;
+            case TGSI_SEMANTIC_FOG:
+            case TGSI_SEMANTIC_GENERIC:
+                out_generic++;
+                break;
+        }
+    }
+
+    struct tgsi_parse_context parser;
+
+    tgsi_parse_init(&parser, vs->state.tokens);
+
+    while (!tgsi_parse_end_of_tokens(&parser)) {
+        tgsi_parse_token(&parser);
+
+        if (parser.FullToken.Token.Type != TGSI_TOKEN_TYPE_DECLARATION)
+            continue;
+
+        struct tgsi_full_declaration * decl = &parser.FullToken.FullDeclaration;
+
+        if (decl->Declaration.File != TGSI_FILE_OUTPUT)
+            continue;
+
+        switch (decl->Semantic.SemanticName) {
+            case TGSI_SEMANTIC_POSITION:
+                c->code->outputs[decl->DeclarationRange.First] = 0;
+                break;
+            case TGSI_SEMANTIC_PSIZE:
+                c->code->outputs[decl->DeclarationRange.First] = 1;
+                break;
+            case TGSI_SEMANTIC_COLOR:
+                c->code->outputs[decl->DeclarationRange.First] = 1 +
+                    (pointsize ? 1 : 0) +
+                    colors++;
+                break;
+            case TGSI_SEMANTIC_FOG:
+            case TGSI_SEMANTIC_GENERIC:
+                c->code->outputs[decl->DeclarationRange.First] = 1 +
+                    (pointsize ? 1 : 0) +
+                    out_colors +
+                    generic++;
+                break;
+            default:
+                debug_printf("r300: vs: Bad semantic declaration %d\n",
+                    decl->Semantic.SemanticName);
+                break;
+        }
+    }
+
+    tgsi_parse_free(&parser);
+}
+
+
+void r300_translate_vertex_shader(struct r300_context* r300,
+                                  struct r300_vertex_shader* vs)
+{
+    struct r300_vertex_program_compiler compiler;
+    struct tgsi_to_rc ttr;
+
+    /* Setup the compiler */
+    rc_init(&compiler.Base);
+
+    compiler.Base.Debug = 1;
+    compiler.code = &vs->code;
+    compiler.UserData = vs;
+
+    if (compiler.Base.Debug) {
+        debug_printf("r300: Initial vertex program\n");
+        tgsi_dump(vs->state.tokens, 0);
+    }
+
+    /* Translate TGSI to our internal representation */
+    ttr.compiler = &compiler.Base;
+    ttr.info = &vs->info;
+
+    r300_tgsi_to_rc(&ttr, vs->state.tokens);
+
+    compiler.RequiredOutputs = ~(~0 << vs->info.num_outputs);
+    compiler.SetHwInputOutput = &set_vertex_inputs_outputs;
+
+    /* Invoke the compiler */
+    r3xx_compile_vertex_program(&compiler);
+    if (compiler.Base.Error) {
+        /* Todo: Fail gracefully */
+        fprintf(stderr, "r300 VP: Compiler error\n");
+        abort();
+    }
+
+    /* And, finally... */
+    rc_destroy(&compiler.Base);
+    vs->translated = TRUE;
+}
+
+
+/* XXX get these to r300_reg */
+#define R300_PVS_DST_OPCODE(x)   ((x) << 0)
+#   define R300_VE_DOT_PRODUCT            1
+#   define R300_VE_MULTIPLY               2
+#   define R300_VE_ADD                    3
+#   define R300_VE_MAXIMUM                7
+#   define R300_VE_SET_LESS_THAN          10
+#define R300_PVS_DST_MATH_INST     (1 << 6)
+#   define R300_ME_RECIP_DX               6
+#define R300_PVS_DST_MACRO_INST    (1 << 7)
+#   define R300_PVS_MACRO_OP_2CLK_MADD    0
+#define R300_PVS_DST_REG_TYPE(x) ((x) << 8)
+#   define R300_PVS_DST_REG_TEMPORARY     0
+#   define R300_PVS_DST_REG_A0            1
+#   define R300_PVS_DST_REG_OUT           2
+#   define R300_PVS_DST_REG_OUT_REPL_X    3
+#   define R300_PVS_DST_REG_ALT_TEMPORARY 4
+#   define R300_PVS_DST_REG_INPUT         5
+#define R300_PVS_DST_OFFSET(x)   ((x) << 13)
+#define R300_PVS_DST_WE(x)       ((x) << 20)
+#define R300_PVS_DST_WE_XYZW     (0xf << 20)
+
+#define R300_PVS_SRC_REG_TYPE(x) ((x) << 0)
+#   define R300_PVS_SRC_REG_TEMPORARY     0
+#   define R300_PVS_SRC_REG_INPUT         1
+#   define R300_PVS_SRC_REG_CONSTANT      2
+#   define R300_PVS_SRC_REG_ALT_TEMPORARY 3
+#define R300_PVS_SRC_OFFSET(x)   ((x) << 5)
+#define R300_PVS_SRC_SWIZZLE(x)  ((x) << 13)
+#   define R300_PVS_SRC_SELECT_X          0
+#   define R300_PVS_SRC_SELECT_Y          1
+#   define R300_PVS_SRC_SELECT_Z          2
+#   define R300_PVS_SRC_SELECT_W          3
+#   define R300_PVS_SRC_SELECT_FORCE_0    4
+#   define R300_PVS_SRC_SELECT_FORCE_1    5
+#   define R300_PVS_SRC_SWIZZLE_XYZW \
+    ((R300_PVS_SRC_SELECT_X | (R300_PVS_SRC_SELECT_Y << 3) | \
+     (R300_PVS_SRC_SELECT_Z << 6) | (R300_PVS_SRC_SELECT_W << 9)) << 13)
+#   define R300_PVS_SRC_SWIZZLE_ZERO \
+    ((R300_PVS_SRC_SELECT_FORCE_0 | (R300_PVS_SRC_SELECT_FORCE_0 << 3) | \
+     (R300_PVS_SRC_SELECT_FORCE_0 << 6) | \
+      (R300_PVS_SRC_SELECT_FORCE_0 << 9)) << 13)
+#   define R300_PVS_SRC_SWIZZLE_ONE \
+    ((R300_PVS_SRC_SELECT_FORCE_1 | (R300_PVS_SRC_SELECT_FORCE_1 << 3) | \
+     (R300_PVS_SRC_SELECT_FORCE_1 << 6) | \
+      (R300_PVS_SRC_SELECT_FORCE_1 << 9)) << 13)
+#define R300_PVS_MODIFIER_X        (1 << 25)
+#define R300_PVS_MODIFIER_Y        (1 << 26)
+#define R300_PVS_MODIFIER_Z        (1 << 27)
+#define R300_PVS_MODIFIER_W        (1 << 28)
+#define R300_PVS_NEGATE_XYZW \
+    (R300_PVS_MODIFIER_X | R300_PVS_MODIFIER_Y | \
+     R300_PVS_MODIFIER_Z | R300_PVS_MODIFIER_W)
+
+struct r300_vertex_program_code r300_passthrough_vertex_shader = {
+    .length = 8, /* two instructions */
+
+    /* MOV out[0], in[0] */
+    .body.d[0] = R300_PVS_DST_OPCODE(R300_VE_ADD) |
+        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
+        R300_PVS_DST_OFFSET(0) | R300_PVS_DST_WE_XYZW,
+    .body.d[1] = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
+        R300_PVS_SRC_OFFSET(0) | R300_PVS_SRC_SWIZZLE_XYZW,
+    .body.d[2] = R300_PVS_SRC_SWIZZLE_ZERO,
+    .body.d[3] = 0x0,
+
+    /* MOV out[1], in[1] */
+    .body.d[4] = R300_PVS_DST_OPCODE(R300_VE_ADD) |
+        R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
+        R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW,
+    .body.d[5] = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
+        R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW,
+    .body.d[6] = R300_PVS_SRC_SWIZZLE_ZERO,
+    .body.d[7] = 0x0,
+
+    .inputs[0] = 0,
+    .inputs[1] = 1,
+    .outputs[0] = 0,
+    .outputs[1] = 1,
+
+    .InputsRead = 3,
+    .OutputsWritten = 3
+};
+
diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h
new file mode 100644
index 0000000000..2a4ce315e3
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_vs.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_VS_H
+#define R300_VS_H
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+#include "radeon_code.h"
+
+struct r300_context;
+
+struct r300_vertex_shader {
+    /* Parent class */
+    struct pipe_shader_state state;
+    struct tgsi_shader_info info;
+
+    /* Fallback shader, because Draw has issues */
+    struct draw_vertex_shader* draw;
+
+    /* Has this shader been translated yet? */
+    boolean translated;
+
+    /* Machine code (if translated) */
+    struct r300_vertex_program_code code;
+};
+
+
+extern struct r300_vertex_program_code r300_passthrough_vertex_shader;
+
+void r300_translate_vertex_shader(struct r300_context* r300,
+                                  struct r300_vertex_shader* vs);
+
+#endif /* R300_VS_H */
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index a833bb0399..f18ad75a47 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -55,10 +55,10 @@ struct r300_winsys {
     uint32_t vram_size;
 
     /* Add a pipe_buffer to the list of buffer objects to validate. */
-    void (*add_buffer)(struct r300_winsys* winsys,
-                       struct pipe_buffer* pbuffer,
-                       uint32_t rd,
-                       uint32_t wd);
+    boolean (*add_buffer)(struct r300_winsys* winsys,
+                          struct pipe_buffer* pbuffer,
+                          uint32_t rd,
+                          uint32_t wd);
 
     /* Revalidate all currently setup pipe_buffers.
      * Returns TRUE if a flush is required. */
diff --git a/src/gallium/drivers/r300/r3xx_fs.c b/src/gallium/drivers/r300/r3xx_fs.c
new file mode 100644
index 0000000000..c1c1194d58
--- /dev/null
+++ b/src/gallium/drivers/r300/r3xx_fs.c
@@ -0,0 +1,74 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r3xx_fs.h"
+
+#include "r300_reg.h"
+
+struct rX00_fragment_program_code r3xx_passthrough_fragment_shader = {
+    .code.r300.alu.length = 1,
+    .code.r300.tex.length = 0,
+
+    .code.r300.config = 0,
+    .code.r300.pixsize = 0,
+    .code.r300.code_offset = 0,
+    .code.r300.code_addr[3] = R300_RGBA_OUT,
+
+    .code.r300.alu.inst[0].rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
+        R300_ALU_OUTC_CMP,
+    .code.r300.alu.inst[0].rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
+        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
+    .code.r300.alu.inst[0].alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
+        R300_ALU_OUTA_CMP,
+    .code.r300.alu.inst[0].alpha_addr = R300_ALPHA_ADDR0(0) |
+        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
+};
+
+struct rX00_fragment_program_code r3xx_texture_fragment_shader = {
+    .code.r300.alu.length = 1,
+    .code.r300.tex.length = 1,
+
+    .code.r300.config = R300_PFS_CNTL_FIRST_NODE_HAS_TEX,
+    .code.r300.pixsize = 0,
+    .code.r300.code_offset = 0,
+    .code.r300.code_addr[3] = R300_RGBA_OUT,
+
+    .code.r300.tex.inst[0] = R300_TEX_OP_LD << R300_TEX_INST_SHIFT,
+
+    .code.r300.alu.inst[0].rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
+        R300_ALU_OUTC_CMP,
+    .code.r300.alu.inst[0].rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
+        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
+    .code.r300.alu.inst[0].alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
+        R300_ALU_OUTA_CMP,
+    .code.r300.alu.inst[0].alpha_addr = R300_ALPHA_ADDR0(0) |
+        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
+};
diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r3xx_fs.h
index a1f873656d..51cd245724 100644
--- a/src/gallium/drivers/r300/r300_debug.h
+++ b/src/gallium/drivers/r300/r3xx_fs.h
@@ -1,5 +1,6 @@
 /*
- * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,15 +21,12 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#ifndef R300_DEBUG_H
-#define R300_DEBUG_H
+#ifndef R3XX_FS_H
+#define R3XX_FS_H
 
-#include "r300_reg.h"
-#include "r300_state_shader.h"
-#include "r300_state_tcl.h"
+#include "radeon_code.h"
 
-void r500_fs_dump(struct r500_fragment_shader* fs);
+struct rX00_fragment_program_code r3xx_passthrough_fragment_shader;
+struct rX00_fragment_program_code r3xx_texture_fragment_shader;
 
-void r300_vs_dump(struct r300_vertex_shader* vs);
-
-#endif /* R300_DEBUG_H */
+#endif /* R3XX_FS_H */
diff --git a/src/gallium/drivers/r300/r5xx_fs.c b/src/gallium/drivers/r300/r5xx_fs.c
new file mode 100644
index 0000000000..f072deab0d
--- /dev/null
+++ b/src/gallium/drivers/r300/r5xx_fs.c
@@ -0,0 +1,125 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r5xx_fs.h"
+
+#include "r300_reg.h"
+
+/* XXX this all should find its way back to r300_reg */
+/* Swizzle tools */
+#define R500_SWIZZLE_ZERO 4
+#define R500_SWIZZLE_HALF 5
+#define R500_SWIZZLE_ONE 6
+#define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6))
+#define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6))
+#define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6))
+#define R500_SWIZ_MOD_NEG 1
+#define R500_SWIZ_MOD_ABS 2
+#define R500_SWIZ_MOD_NEG_ABS 3
+/* Swizzles for inst2 */
+#define R500_SWIZ_TEX_STRQ(x) ((x) << 8)
+#define R500_SWIZ_TEX_RGBA(x) ((x) << 24)
+/* Swizzles for inst3 */
+#define R500_SWIZ_RGB_A(x) ((x) << 2)
+#define R500_SWIZ_RGB_B(x) ((x) << 15)
+/* Swizzles for inst4 */
+#define R500_SWIZ_ALPHA_A(x) ((x) << 14)
+#define R500_SWIZ_ALPHA_B(x) ((x) << 21)
+/* Swizzle for inst5 */
+#define R500_SWIZ_RGBA_C(x) ((x) << 14)
+#define R500_SWIZ_ALPHA_C(x) ((x) << 27)
+/* Writemasks */
+#define R500_TEX_WMASK(x) ((x) << 11)
+#define R500_ALU_WMASK(x) ((x) << 11)
+#define R500_ALU_OMASK(x) ((x) << 15)
+#define R500_W_OMASK (1 << 31)
+
+struct rX00_fragment_program_code r5xx_passthrough_fragment_shader = {
+    .code.r500.max_temp_idx = 0,
+    .code.r500.inst_end = 0,
+
+    .code.r500.inst[0].inst0 = R500_INST_TYPE_OUT |
+        R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
+        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
+        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
+    .code.r500.inst[0].inst1 =
+        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
+        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST,
+    .code.r500.inst[0].inst2 =
+        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
+        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST,
+    .code.r500.inst[0].inst3 =
+        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
+        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
+        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
+        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B,
+    .code.r500.inst[0].inst4 =
+        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A,
+    .code.r500.inst[0].inst5 =
+        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
+        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
+        R500_ALU_RGBA_A_SWIZ_0,
+};
+
+struct rX00_fragment_program_code r5xx_texture_fragment_shader = {
+    .code.r500.max_temp_idx = 0,
+    .code.r500.inst_end = 1,
+
+    .code.r500.inst[0].inst0 = R500_INST_TYPE_TEX |
+        R500_INST_TEX_SEM_WAIT |
+        R500_INST_RGB_WMASK_RGB | R500_INST_ALPHA_WMASK |
+        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
+    .code.r500.inst[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD |
+        R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED,
+    .code.r500.inst[0].inst2 = R500_TEX_SRC_ADDR(0) |
+        R500_TEX_SRC_S_SWIZ_R | R500_TEX_SRC_T_SWIZ_G |
+        R500_TEX_SRC_R_SWIZ_B | R500_TEX_SRC_Q_SWIZ_A |
+        R500_TEX_DST_ADDR(0) |
+        R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G |
+        R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A,
+    .code.r500.inst[0].inst3 = 0x0,
+    .code.r500.inst[0].inst4 = 0x0,
+    .code.r500.inst[0].inst5 = 0x0,
+
+    .code.r500.inst[1].inst0 = R500_INST_TYPE_OUT |
+        R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
+        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
+        R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
+    .code.r500.inst[1].inst1 =
+        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
+        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST,
+    .code.r500.inst[1].inst2 =
+        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
+        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST,
+    .code.r500.inst[1].inst3 =
+        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
+        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
+        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
+        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B,
+    .code.r500.inst[1].inst4 =
+        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A,
+    .code.r500.inst[1].inst5 =
+        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
+        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
+        R500_ALU_RGBA_A_SWIZ_0,
+};
diff --git a/src/gallium/drivers/r300/r5xx_fs.h b/src/gallium/drivers/r300/r5xx_fs.h
new file mode 100644
index 0000000000..a4addde32b
--- /dev/null
+++ b/src/gallium/drivers/r300/r5xx_fs.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R5XX_FS_H
+#define R5XX_FS_H
+
+#include "radeon_code.h"
+
+struct rX00_fragment_program_code r5xx_passthrough_fragment_shader;
+struct rX00_fragment_program_code r5xx_texture_fragment_shader;
+
+#endif /* R5XX_FS_H */
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index dffc15a4f1..7888c2f644 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -92,6 +92,7 @@ struct softpipe_context {
     * queries.
     */
    uint64_t occlusion_count;
+   unsigned active_query_count;
 
    /** Mapped vertex buffers */
    ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS];
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index ba2766ff13..d4045816d0 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -35,6 +35,7 @@
 #include "pipe/p_context.h"
 #include "pipe/internal/p_winsys_screen.h"
 #include "pipe/p_inlines.h"
+#include "util/u_prim.h"
 
 #include "sp_context.h"
 #include "sp_state.h"
@@ -65,6 +66,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp)
                                    size);
 }
 
+
 static void
 softpipe_unmap_constant_buffers(struct softpipe_context *sp)
 {
@@ -86,20 +88,6 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp)
 }
 
 
-static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = {
-   PIPE_PRIM_POINTS,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES
-};
-
-
 boolean
 softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
                      unsigned start, unsigned count)
@@ -108,15 +96,11 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
 }
 
 
-
 /**
  * Draw vertex arrays, with optional indexing.
  * Basically, map the vertex buffers (and drawing surfaces), then hand off
  * the drawing to the 'draw' module.
- *
- * XXX should the element buffer be specified/bound with a separate function?
  */
-
 boolean
 softpipe_draw_range_elements(struct pipe_context *pipe,
                              struct pipe_buffer *indexBuffer,
@@ -129,7 +113,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    struct draw_context *draw = sp->draw;
    unsigned i;
 
-   sp->reduced_api_prim = reduced_prim[mode];
+   sp->reduced_api_prim = u_reduced_prim(mode);
 
    if (sp->dirty)
       softpipe_update_derived( sp );
@@ -147,6 +131,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
                                     PIPE_BUFFER_USAGE_CPU_READ);
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
+
    /* Map index buffer, if present */
    if (indexBuffer) {
       void *mapped_indexes
@@ -159,10 +144,10 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    }
    else {
       /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
+      draw_set_mapped_element_buffer_range(draw, 0, start,
+                                           start + count - 1, NULL);
    }
 
-
    /* draw! */
    draw_arrays(draw, mode, start, count);
 
@@ -187,6 +172,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    return TRUE;
 }
 
+
 boolean
 softpipe_draw_elements(struct pipe_context *pipe,
                        struct pipe_buffer *indexBuffer,
@@ -200,11 +186,9 @@ softpipe_draw_elements(struct pipe_context *pipe,
 }
 
 
-
 void
 softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
 {
    struct softpipe_context *sp = softpipe_context(pipe);
    draw_set_edgeflags(sp->draw, edgeflags);
 }
-
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
index 31c3ca21c5..f4fa0905d7 100644
--- a/src/gallium/drivers/softpipe/sp_fs_sse.c
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -45,17 +45,6 @@
 
 #include "rtasm/rtasm_x86sse.h"
 
-/* Surely this should be defined somewhere in a tgsi header:
- */
-typedef void (PIPE_CDECL *codegen_function)(
-   const struct tgsi_exec_vector *input,
-   struct tgsi_exec_vector *output,
-   const float (*constant)[4],
-   struct tgsi_exec_vector *temporary,
-   const struct tgsi_interp_coef *coef,
-   float (*immediates)[4]
-   //, const struct tgsi_exec_vector *quadPos
- );
 
 
 /**
@@ -65,7 +54,7 @@ struct sp_sse_fragment_shader
 {
    struct sp_fragment_shader base;
    struct x86_function sse2_program;
-   codegen_function func;
+   tgsi_sse2_fs_function func;
    float immediates[TGSI_EXEC_NUM_IMMEDIATES][4];
 };
 
@@ -83,6 +72,7 @@ fs_sse_prepare( const struct sp_fragment_shader *base,
 		struct tgsi_exec_machine *machine,
 		struct tgsi_sampler **samplers )
 {
+   machine->Samplers = samplers;
 }
 
 
@@ -107,12 +97,10 @@ fs_sse_run( const struct sp_fragment_shader *base,
    tgsi_set_kill_mask(machine, 0x0);
    tgsi_set_exec_mask(machine, 1, 1, 1, 1);
 
-   shader->func( machine->Inputs,
-		 machine->Outputs,
+   shader->func( machine,
 		 machine->Consts,
-		 machine->Temps,
-		 machine->InterpCoefs,
-                 shader->immediates
+                 (const float (*)[4])shader->immediates,
+		 machine->InterpCoefs
 		 //	 , &machine->QuadPos
       );
 
@@ -151,7 +139,7 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe,
       return NULL;
    }
 
-   shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
+   shader->func = (tgsi_sse2_fs_function) x86_get_func( &shader->sse2_program );
    if (!shader->func) {
       x86_release_func( &shader->sse2_program );
       FREE(shader);
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index 06725fd09b..42021789ea 100644
--- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -238,65 +238,117 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
 
    case PIPE_PRIM_TRIANGLES:
       for (i = 2; i < nr; i += 3) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_STRIP:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
-                    get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_FAN:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[0], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUADS:
       for (i = 3; i < nr; i += 4) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 3; i < nr; i += 2) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride));
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_POLYGON:
+      /* Almost same as tri fan but the _first_ vertex specifies the flat
+       * shading color.  Note that the first polygon vertex is passed as
+       * the last triangle vertex here.
+       * flatshade_first state makes no difference.
+       */
       for (i = 2; i < nr; i += 1) {
          setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[0-1], stride),
                     get_vert(vertex_buffer, indices[i-0], stride),
-                    get_vert(vertex_buffer, indices[0], stride));
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[0], stride) );
       }
       break;
 
@@ -368,58 +420,104 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       }
       break;
 
-
    case PIPE_PRIM_TRIANGLES:
       for (i = 2; i < nr; i += 3) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_STRIP:
-      for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i+(i&1)-2, stride),
-                    get_vert(vertex_buffer, i-(i&1)-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+      for (i = 2; i < nr; i++) {
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-(i&1), stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-2, stride),
+                       get_vert(vertex_buffer, i-(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_FAN:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, 0, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUADS:
       for (i = 3; i < nr; i += 4) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-0, stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
+
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 3; i < nr; i += 2) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-0, stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
@@ -427,12 +525,13 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       /* Almost same as tri fan but the _first_ vertex specifies the flat
        * shading color.  Note that the first polygon vertex is passed as
        * the last triangle vertex here.
+       * flatshade_first state makes no difference.
        */
       for (i = 2; i < nr; i += 1) {
          setup_tri( setup_ctx,
                     get_vert(vertex_buffer, i-1, stride),
                     get_vert(vertex_buffer, i-0, stride),
-                    get_vert(vertex_buffer, 0, stride));
+                    get_vert(vertex_buffer, 0, stride) );
       }
       break;
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index ca637a1d6a..28f8d1a60e 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -52,7 +52,7 @@
 struct quad_shade_stage
 {
    struct quad_stage stage;  /**< base class */
-   struct tgsi_exec_machine machine;
+   struct tgsi_exec_machine *machine;
    struct tgsi_exec_vector *inputs, *outputs;
 };
 
@@ -73,7 +73,7 @@ shade_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    struct quad_shade_stage *qss = quad_shade_stage( qs );
    struct softpipe_context *softpipe = qs->softpipe;
-   struct tgsi_exec_machine *machine = &qss->machine;
+   struct tgsi_exec_machine *machine = qss->machine;
    boolean z_written;
    
    /* Consts do not require 16 byte alignment. */
@@ -146,7 +146,7 @@ shade_begin(struct quad_stage *qs)
    struct softpipe_context *softpipe = qs->softpipe;
 
    softpipe->fs->prepare( softpipe->fs, 
-			  &qss->machine,
+			  qss->machine,
 			  (struct tgsi_sampler **)
                              softpipe->tgsi.frag_samplers_list );
 
@@ -159,9 +159,8 @@ shade_destroy(struct quad_stage *qs)
 {
    struct quad_shade_stage *qss = (struct quad_shade_stage *) qs;
 
-   tgsi_exec_machine_free_data(&qss->machine);
-   FREE( qss->inputs );
-   FREE( qss->outputs );
+   tgsi_exec_machine_destroy(qss->machine);
+
    FREE( qs );
 }
 
@@ -170,19 +169,24 @@ struct quad_stage *
 sp_quad_shade_stage( struct softpipe_context *softpipe )
 {
    struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage);
-
-   /* allocate storage for program inputs/outputs, aligned to 16 bytes */
-   qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16);
-   qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16);
-   qss->machine.Inputs = align16(qss->inputs);
-   qss->machine.Outputs = align16(qss->outputs);
+   if (!qss)
+      goto fail;
 
    qss->stage.softpipe = softpipe;
    qss->stage.begin = shade_begin;
    qss->stage.run = shade_quad;
    qss->stage.destroy = shade_destroy;
 
-   tgsi_exec_machine_init( &qss->machine );
+   qss->machine = tgsi_exec_machine_create();
+   if (!qss->machine)
+      goto fail;
 
    return &qss->stage;
+
+fail:
+   if (qss && qss->machine)
+      tgsi_exec_machine_destroy(qss->machine);
+
+   FREE(qss);
+   return NULL;
 }
diff --git a/src/gallium/drivers/softpipe/sp_quad_pipe.c b/src/gallium/drivers/softpipe/sp_quad_pipe.c
index 892ef87ee9..b5f69b7426 100644
--- a/src/gallium/drivers/softpipe/sp_quad_pipe.c
+++ b/src/gallium/drivers/softpipe/sp_quad_pipe.c
@@ -80,7 +80,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp)
          sp_push_quad_first( sp, sp->quad[i].blend, i );
       }
 
-      if (sp->depth_stencil->depth.occlusion_count) {
+      if (sp->active_query_count) {
          sp_push_quad_first( sp, sp->quad[i].occlusion, i );
       }
 
diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
index 93dab236d6..379cf4ad06 100644
--- a/src/gallium/drivers/softpipe/sp_query.c
+++ b/src/gallium/drivers/softpipe/sp_query.c
@@ -34,6 +34,7 @@
 #include "util/u_memory.h"
 #include "sp_context.h"
 #include "sp_query.h"
+#include "sp_state.h"
 
 struct softpipe_query {
    uint64_t start;
@@ -69,6 +70,8 @@ softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
    struct softpipe_query *sq = softpipe_query(q);
    
    sq->start = softpipe->occlusion_count;
+   softpipe->active_query_count++;
+   softpipe->dirty |= SP_NEW_QUERY;
 }
 
 
@@ -78,7 +81,9 @@ softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
    struct softpipe_context *softpipe = softpipe_context( pipe );
    struct softpipe_query *sq = softpipe_query(q);
 
+   softpipe->active_query_count--;
    sq->end = softpipe->occlusion_count;
+   softpipe->dirty |= SP_NEW_QUERY;
 }
 
 
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index cc39d33ede..6178c4ac7e 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -82,13 +82,15 @@ softpipe_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 12; /* max 2Kx2K */
+      return 13; /* max 4Kx4K */
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
       return 8;  /* max 128x128x128 */
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 12; /* max 2Kx2K */
+      return 13; /* max 4Kx4K */
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
+   case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+      return 1;
    default:
       return 0;
    }
@@ -108,7 +110,7 @@ softpipe_get_paramf(struct pipe_screen *screen, int param)
    case PIPE_CAP_MAX_POINT_WIDTH_AA:
       return 255.0; /* arbitrary */
    case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
-      return 0.0;
+      return 16.0; /* not actually signficant at this time */
    case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
       return 16.0; /* arbitrary */
    default:
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 1eb23cd6b6..de3ae3c369 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -507,6 +507,8 @@ static void print_vertex(const struct setup_context *setup,
 #endif
 
 /**
+ * Sort the vertices from top to bottom order, setting up the triangle
+ * edge fields (ebot, emaj, etop).
  * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
  */
 static boolean setup_sort_vertices( struct setup_context *setup,
@@ -1051,7 +1053,10 @@ setup_line_coefficients(struct setup_context *setup,
    float area;
 
    /* use setup->vmin, vmax to point to vertices */
-   setup->vprovoke = v1;
+   if (softpipe->rasterizer->flatshade_first)
+      setup->vprovoke = v0;
+   else
+      setup->vprovoke = v1;
    setup->vmin = v0;
    setup->vmax = v1;
 
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
index 5de358dae9..a1d3bade27 100644
--- a/src/gallium/drivers/softpipe/sp_tex_sample.c
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -705,15 +705,18 @@ get_texel(const struct tgsi_sampler *tgsi_sampler,
  * Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
  * When we sampled the depth texture, the depth value was put into all
  * RGBA channels.  We look at the red channel here.
+ * \param rgba  quad of (depth) texel values
+ * \param p  texture 'P' components for four pixels in quad
+ * \param j  which pixel in the quad to test [0..3]
  */
 static INLINE void
-shadow_compare(uint compare_func,
+shadow_compare(const struct pipe_sampler_state *sampler,
                float rgba[NUM_CHANNELS][QUAD_SIZE],
                const float p[QUAD_SIZE],
                uint j)
 {
    int k;
-   switch (compare_func) {
+   switch (sampler->compare_func) {
    case PIPE_FUNC_LESS:
       k = p[j] < rgba[0][j];
       break;
@@ -751,6 +754,78 @@ shadow_compare(uint compare_func,
 
 
 /**
+ * As above, but do four z/texture comparisons.
+ */
+static INLINE void
+shadow_compare4(const struct pipe_sampler_state *sampler,
+                float rgba[NUM_CHANNELS][QUAD_SIZE],
+                const float p[QUAD_SIZE])
+{
+   int j, k0, k1, k2, k3;
+   float val;
+
+   /* compare four texcoords vs. four texture samples */
+   switch (sampler->compare_func) {
+   case PIPE_FUNC_LESS:
+      k0 = p[0] < rgba[0][0];
+      k1 = p[1] < rgba[0][1];
+      k2 = p[2] < rgba[0][2];
+      k3 = p[3] < rgba[0][3];
+      break;
+   case PIPE_FUNC_LEQUAL:
+      k0 = p[0] <= rgba[0][0];
+      k1 = p[1] <= rgba[0][1];
+      k2 = p[2] <= rgba[0][2];
+      k3 = p[3] <= rgba[0][3];
+      break;
+   case PIPE_FUNC_GREATER:
+      k0 = p[0] > rgba[0][0];
+      k1 = p[1] > rgba[0][1];
+      k2 = p[2] > rgba[0][2];
+      k3 = p[3] > rgba[0][3];
+      break;
+   case PIPE_FUNC_GEQUAL:
+      k0 = p[0] >= rgba[0][0];
+      k1 = p[1] >= rgba[0][1];
+      k2 = p[2] >= rgba[0][2];
+      k3 = p[3] >= rgba[0][3];
+      break;
+   case PIPE_FUNC_EQUAL:
+      k0 = p[0] == rgba[0][0];
+      k1 = p[1] == rgba[0][1];
+      k2 = p[2] == rgba[0][2];
+      k3 = p[3] == rgba[0][3];
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      k0 = p[0] != rgba[0][0];
+      k1 = p[1] != rgba[0][1];
+      k2 = p[2] != rgba[0][2];
+      k3 = p[3] != rgba[0][3];
+      break;
+   case PIPE_FUNC_ALWAYS:
+      k0 = k1 = k2 = k3 = 1;
+      break;
+   case PIPE_FUNC_NEVER:
+      k0 = k1 = k2 = k3 = 0;
+      break;
+   default:
+      k0 = k1 = k2 = k3 = 0;
+      assert(0);
+      break;
+   }
+
+   /* convert four pass/fail values to an intensity in [0,1] */
+   val = 0.25F * (k0 + k1 + k2 + k3);
+
+   /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+   for (j = 0; j < 4; j++) {
+      rgba[0][j] = rgba[1][j] = rgba[2][j] = val;
+      rgba[3][j] = 1.0F;
+   }
+}
+
+
+/**
  * Common code for sampling 1D/2D/cube textures.
  * Could probably extend for 3D...
  */
@@ -769,7 +844,6 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
    const uint unit = samp->unit;
    const struct pipe_texture *texture = sp->texture[unit];
    const struct pipe_sampler_state *sampler = sp->sampler[unit];
-   const uint compare_func = sampler->compare_func;
    unsigned level0, level1, j, imgFilter;
    int width, height;
    float levelBlend;
@@ -794,7 +868,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
          for (j = 0; j < QUAD_SIZE; j++) {
             get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, rgba, p, j);
+               shadow_compare(sampler, rgba, p, j);
             }
 
             if (level0 != level1) {
@@ -806,7 +880,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
                get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0,
                          rgba2, j);
                if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
-                  shadow_compare(compare_func, rgba2, p, j);
+                  shadow_compare(sampler, rgba2, p, j);
                }
 
                for (c = 0; c < NUM_CHANNELS; c++) {
@@ -833,10 +907,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
             get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2);
             get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, tx, p, 0);
-               shadow_compare(compare_func, tx, p, 1);
-               shadow_compare(compare_func, tx, p, 2);
-               shadow_compare(compare_func, tx, p, 3);
+               shadow_compare4(sampler, tx, p);
             }
 
             /* interpolate R, G, B, A */
@@ -858,10 +929,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
                get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2);
                get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3);
                if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
-                  shadow_compare(compare_func, tx, p, 0);
-                  shadow_compare(compare_func, tx, p, 1);
-                  shadow_compare(compare_func, tx, p, 2);
-                  shadow_compare(compare_func, tx, p, 3);
+                  shadow_compare4(sampler, tx, p);
                }
 
                /* interpolate R, G, B, A */
@@ -1076,7 +1144,6 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
    const struct pipe_texture *texture = sp->texture[unit];
    const struct pipe_sampler_state *sampler = sp->sampler[unit];
    const uint face = 0;
-   const uint compare_func = sampler->compare_func;
    unsigned level0, level1, j, imgFilter;
    int width, height;
    float levelBlend;
@@ -1101,7 +1168,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
          for (j = 0; j < QUAD_SIZE; j++) {
             get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, rgba, p, j);
+               shadow_compare(sampler, rgba, p, j);
             }
          }
       }
@@ -1121,10 +1188,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
             get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2);
             get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, tx, p, 0);
-               shadow_compare(compare_func, tx, p, 1);
-               shadow_compare(compare_func, tx, p, 2);
-               shadow_compare(compare_func, tx, p, 3);
+               shadow_compare4(sampler, tx, p);
             }
             for (c = 0; c < 4; c++) {
                rgba[c][j] = lerp_2d(xw[j], yw[j],
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 7a533dad9f..70f0932431 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -48,11 +48,6 @@
 /* Simple, maximally packed layout.
  */
 
-static unsigned minify( unsigned d )
-{
-   return MAX2(1, d>>1);
-}
-
 
 /* Conventional allocation path for non-display textures:
  */
@@ -100,6 +95,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen,
 {
    unsigned usage = (PIPE_BUFFER_USAGE_CPU_READ_WRITE |
                      PIPE_BUFFER_USAGE_GPU_READ_WRITE);
+   unsigned tex_usage = spt->base.tex_usage;
 
    spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]);  
    spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]);  
@@ -109,6 +105,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen,
                                                 spt->base.height[0],
                                                 spt->base.format,
                                                 usage,
+                                                tex_usage,
                                                 &spt->stride[0]);
 
    return spt->buffer != NULL;
@@ -130,7 +127,8 @@ softpipe_texture_create(struct pipe_screen *screen,
    pipe_reference_init(&spt->base.reference, 1);
    spt->base.screen = screen;
 
-   if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+   if (spt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                              PIPE_TEXTURE_USAGE_PRIMARY)) {
       if (!softpipe_displaytarget_layout(screen, spt))
          goto fail;
    }
@@ -224,12 +222,6 @@ softpipe_get_tex_surface(struct pipe_screen *screen,
       if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ)
          ps->usage |= PIPE_BUFFER_USAGE_CPU_READ;
 
-      if (ps->usage & (PIPE_BUFFER_USAGE_CPU_WRITE |
-                       PIPE_BUFFER_USAGE_GPU_WRITE)) {
-         /* Mark the surface as dirty.  The tile cache will look for this. */
-         spt->modified = TRUE;
-      }
-
       ps->face = face;
       ps->level = level;
       ps->zslice = zslice;
@@ -376,6 +368,11 @@ softpipe_transfer_unmap(struct pipe_screen *screen,
    spt = softpipe_texture(transfer->texture);
 
    pipe_buffer_unmap( screen, spt->buffer );
+
+   if (transfer->usage != PIPE_TRANSFER_READ) {
+      /* Mark the texture as dirty to expire the tile caches. */
+      spt->modified = TRUE;
+   }
 }
 
 
diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile
index e087db169a..dd6831c70a 100644
--- a/src/gallium/drivers/trace/Makefile
+++ b/src/gallium/drivers/trace/Makefile
@@ -7,8 +7,11 @@ C_SOURCES = \
 	tr_buffer.c \
 	tr_context.c \
 	tr_dump.c \
+	tr_dump_state.c \
 	tr_screen.c \
 	tr_state.c \
+	tr_rbug.c \
+	tr_drm.c \
 	tr_texture.c
 
 include ../../Makefile.template
diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README
index 73dce20372..1000c31e49 100644
--- a/src/gallium/drivers/trace/README
+++ b/src/gallium/drivers/trace/README
@@ -3,7 +3,8 @@
 
 = About =
 
-This directory contains a Gallium3D pipe driver which traces all incoming calls.
+This directory contains a Gallium3D debugger pipe driver.
+It can traces all incoming calls and/or provide remote debugging functionality.
 
 
 = Build Instructions =
@@ -23,7 +24,9 @@ ensure the right libGL.so is being picked by doing
 
  ldd progs/trivial/tri 
 
-and then try running
+== Traceing ==
+
+For traceing then do
 
  export XMESA_TRACE=y
  GALLIUM_TRACE=tri.trace progs/trivial/tri
@@ -32,6 +35,16 @@ which should create a tri.trace file, which is an XML file. You can view copying
 trace.xsl to the same directory, and opening with a XSLT capable browser such as 
 Firefox or Internet Explorer.
 
+== Remote debugging ==
+
+For remote debugging
+
+ export XMESA_TRACE=y
+ GALLIUM_RBUG=true progs/trivial/tri
+
+which should open gallium remote debugging session. While the program is running
+you can launch the small remote debugging application from progs/rbug. More
+information is in that directory.
 
 = Integrating =
 
@@ -62,3 +75,4 @@ trace_screen with real_screen when creating them.
 
 --
 Jose Fonseca <jrfonseca@tungstengraphics.com>
+Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript
index 4215215d1a..c1675d1c16 100644
--- a/src/gallium/drivers/trace/SConscript
+++ b/src/gallium/drivers/trace/SConscript
@@ -7,9 +7,12 @@ trace = env.ConvenienceLibrary(
     source = [
         'tr_buffer.c',
         'tr_context.c',
+        'tr_drm.c',
         'tr_dump.c',
+        'tr_dump_state.c',
         'tr_screen.c',
         'tr_state.c',
+        'tr_rbug.c',
         'tr_texture.c',
     ])
 
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
index 47280459a7..ae0af4d055 100644
--- a/src/gallium/drivers/trace/tr_context.c
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -31,11 +31,11 @@
 #include "pipe/p_screen.h"
 
 #include "tr_dump.h"
+#include "tr_dump_state.h"
 #include "tr_state.h"
 #include "tr_buffer.h"
 #include "tr_screen.h"
 #include "tr_texture.h"
-#include "tr_context.h"
 
 
 static INLINE struct pipe_buffer *
@@ -113,6 +113,65 @@ trace_context_set_edgeflags(struct pipe_context *_pipe,
 }
 
 
+static INLINE void
+trace_context_draw_block(struct trace_context *tr_ctx, int flag)
+{
+   int k;
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+
+   if (tr_ctx->draw_blocker & flag) {
+      tr_ctx->draw_blocked |= flag;
+   } else if ((tr_ctx->draw_rule.blocker & flag) &&
+              (tr_ctx->draw_blocker & 4)) {
+      boolean block = FALSE;
+      debug_printf("%s (%lu %lu) (%lu %lu) (%lu %u) (%lu %u)\n", __FUNCTION__,
+					tr_ctx->draw_rule.fs, tr_ctx->curr.fs,
+					tr_ctx->draw_rule.vs, tr_ctx->curr.vs,
+					tr_ctx->draw_rule.surf, 0,
+					tr_ctx->draw_rule.tex, 0);
+      if (tr_ctx->draw_rule.fs &&
+          tr_ctx->draw_rule.fs == tr_ctx->curr.fs)
+         block = TRUE;
+      if (tr_ctx->draw_rule.vs &&
+          tr_ctx->draw_rule.vs == tr_ctx->curr.vs)
+         block = TRUE;
+      if (tr_ctx->draw_rule.surf &&
+          tr_ctx->draw_rule.surf == tr_ctx->curr.zsbuf)
+            block = TRUE;
+      if (tr_ctx->draw_rule.surf)
+         for (k = 0; k < tr_ctx->curr.nr_cbufs; k++)
+            if (tr_ctx->draw_rule.surf == tr_ctx->curr.cbufs[k])
+               block = TRUE;
+      if (tr_ctx->draw_rule.tex)
+         for (k = 0; k < tr_ctx->curr.num_texs; k++)
+            if (tr_ctx->draw_rule.tex == tr_ctx->curr.tex[k])
+               block = TRUE;
+
+      if (block)
+         tr_ctx->draw_blocked |= (flag | 4);
+   }
+
+   if (tr_ctx->draw_blocked)
+      trace_rbug_notify_draw_blocked(tr_ctx);
+
+   /* wait for rbug to clear the blocked flag */
+   while (tr_ctx->draw_blocked & flag) {
+      tr_ctx->draw_blocked |= flag;
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+      pipe_condvar_wait(tr_ctx->draw_cond, tr_ctx->draw_mutex);
+#else
+      pipe_mutex_unlock(tr_ctx->draw_mutex);
+#ifdef PIPE_SUBSYSTEM_WINDOWS_USER
+      Sleep(1);
+#endif
+      pipe_mutex_lock(tr_ctx->draw_mutex);
+#endif
+   }
+
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+}
+
 static INLINE boolean
 trace_context_draw_arrays(struct pipe_context *_pipe,
                           unsigned mode, unsigned start, unsigned count)
@@ -121,6 +180,11 @@ trace_context_draw_arrays(struct pipe_context *_pipe,
    struct pipe_context *pipe = tr_ctx->pipe;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_dump_call_begin("pipe_context", "draw_arrays");
 
    trace_dump_arg(ptr, pipe);
@@ -134,6 +198,8 @@ trace_context_draw_arrays(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -150,6 +216,11 @@ trace_context_draw_elements(struct pipe_context *_pipe,
    struct pipe_buffer *indexBuffer = tr_buf->buffer;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_screen_user_buffer_update(_pipe->screen, indexBuffer);
 
    trace_dump_call_begin("pipe_context", "draw_elements");
@@ -167,6 +238,8 @@ trace_context_draw_elements(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -187,6 +260,11 @@ trace_context_draw_range_elements(struct pipe_context *_pipe,
    struct pipe_buffer *indexBuffer = tr_buf->buffer;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_screen_user_buffer_update(_pipe->screen, indexBuffer);
 
    trace_dump_call_begin("pipe_context", "draw_range_elements");
@@ -209,6 +287,8 @@ trace_context_draw_range_elements(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -573,23 +653,32 @@ trace_context_create_fs_state(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_FRAGMENT);
+
    return result;
 }
 
 
 static INLINE void
 trace_context_bind_fs_state(struct pipe_context *_pipe,
-                            void *state)
+                            void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr ? tr_shdr->state : NULL;
 
    trace_dump_call_begin("pipe_context", "bind_fs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
-   pipe->bind_fs_state(pipe, state);;
+   tr_ctx->curr.fs = tr_shdr;
+
+   if (tr_shdr && tr_shdr->replaced)
+      state = tr_shdr->replaced;
+
+   pipe->bind_fs_state(pipe, state);
 
    trace_dump_call_end();
 }
@@ -597,19 +686,23 @@ trace_context_bind_fs_state(struct pipe_context *_pipe,
 
 static INLINE void
 trace_context_delete_fs_state(struct pipe_context *_pipe,
-                              void *state)
+                              void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr->state;
 
    trace_dump_call_begin("pipe_context", "delete_fs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
-   pipe->delete_fs_state(pipe, state);;
+   pipe->delete_fs_state(pipe, state);
 
    trace_dump_call_end();
+
+   trace_shader_destroy(tr_ctx, tr_shdr);
 }
 
 
@@ -626,28 +719,37 @@ trace_context_create_vs_state(struct pipe_context *_pipe,
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(shader_state, state);
 
-   result = pipe->create_vs_state(pipe, state);;
+   result = pipe->create_vs_state(pipe, state);
 
    trace_dump_ret(ptr, result);
 
    trace_dump_call_end();
 
+   result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_VERTEX);
+
    return result;
 }
 
 
 static INLINE void
 trace_context_bind_vs_state(struct pipe_context *_pipe,
-                            void *state)
+                            void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr ? tr_shdr->state : NULL;
 
    trace_dump_call_begin("pipe_context", "bind_vs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
+   tr_ctx->curr.vs = tr_shdr;
+
+   if (tr_shdr && tr_shdr->replaced)
+      state = tr_shdr->replaced;
+
    pipe->bind_vs_state(pipe, state);;
 
    trace_dump_call_end();
@@ -656,10 +758,12 @@ trace_context_bind_vs_state(struct pipe_context *_pipe,
 
 static INLINE void
 trace_context_delete_vs_state(struct pipe_context *_pipe,
-                              void *state)
+                              void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr->state;
 
    trace_dump_call_begin("pipe_context", "delete_vs_state");
 
@@ -669,6 +773,8 @@ trace_context_delete_vs_state(struct pipe_context *_pipe,
    pipe->delete_vs_state(pipe, state);;
 
    trace_dump_call_end();
+
+   trace_shader_destroy(tr_ctx, tr_shdr);
 }
 
 
@@ -747,6 +853,19 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe,
    struct pipe_framebuffer_state unwrapped_state;
    unsigned i;
 
+   {
+      tr_ctx->curr.nr_cbufs = state->nr_cbufs;
+      for (i = 0; i < state->nr_cbufs; i++)
+         if (state->cbufs[i])
+            tr_ctx->curr.cbufs[i] = trace_texture(state->cbufs[i]->texture);
+         else
+            tr_ctx->curr.cbufs[i] = NULL;
+      if (state->zsbuf)
+         tr_ctx->curr.zsbuf = trace_texture(state->zsbuf->texture);
+      else
+         tr_ctx->curr.zsbuf = NULL;
+   }
+
    /* Unwrap the input state */
    memcpy(&unwrapped_state, state, sizeof(unwrapped_state));
    for(i = 0; i < state->nr_cbufs; ++i)
@@ -827,12 +946,17 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe,
                                    struct pipe_texture **textures)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_texture *tr_tex;
    struct pipe_context *pipe = tr_ctx->pipe;
    struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
    unsigned i;
 
-   for(i = 0; i < num_textures; ++i)
-      unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]);
+   tr_ctx->curr.num_texs = num_textures;
+   for(i = 0; i < num_textures; ++i) {
+      tr_tex = trace_texture(textures[i]);
+      tr_ctx->curr.tex[i] = tr_tex;
+      unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL;
+   }
    textures = unwrapped_textures;
 
    trace_dump_call_begin("pipe_context", "set_sampler_textures");
@@ -1078,6 +1202,12 @@ trace_is_buffer_referenced( struct pipe_context *_pipe,
    return referenced;
 }
 
+static const struct debug_named_value rbug_blocker_flags[] = {
+   {"before", 1},
+   {"after", 2},
+   {NULL, 0},
+};
+
 struct pipe_context *
 trace_context_create(struct pipe_screen *_screen,
                      struct pipe_context *pipe)
@@ -1089,7 +1219,7 @@ trace_context_create(struct pipe_screen *_screen,
    if(!pipe)
       goto error1;
 
-   if(!trace_dump_trace_enabled())
+   if(!trace_enabled())
       goto error1;
 
    tr_scr = trace_screen(_screen);
@@ -1099,6 +1229,14 @@ trace_context_create(struct pipe_screen *_screen,
    if(!tr_ctx)
       goto error1;
 
+   tr_ctx->draw_blocker = debug_get_flags_option("RBUG_BLOCK",
+                                                 rbug_blocker_flags,
+                                                 0);
+   pipe_mutex_init(tr_ctx->draw_mutex);
+   pipe_condvar_init(tr_ctx->draw_cond);
+   pipe_mutex_init(tr_ctx->list_mutex);
+   make_empty_list(&tr_ctx->shaders);
+
    tr_ctx->base.winsys = _screen->winsys;
    tr_ctx->base.screen = _screen;
    tr_ctx->base.destroy = trace_context_destroy;
@@ -1139,8 +1277,10 @@ trace_context_create(struct pipe_screen *_screen,
    tr_ctx->base.set_sampler_textures = trace_context_set_sampler_textures;
    tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers;
    tr_ctx->base.set_vertex_elements = trace_context_set_vertex_elements;
-   tr_ctx->base.surface_copy = trace_context_surface_copy;
-   tr_ctx->base.surface_fill = trace_context_surface_fill;
+   if (pipe->surface_copy)
+      tr_ctx->base.surface_copy = trace_context_surface_copy;
+   if (pipe->surface_fill)
+      tr_ctx->base.surface_fill = trace_context_surface_fill;
    tr_ctx->base.clear = trace_context_clear;
    tr_ctx->base.flush = trace_context_flush;
    tr_ctx->base.is_texture_referenced = trace_is_texture_referenced;
diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h
index 2512a0a232..6febe4b411 100644
--- a/src/gallium/drivers/trace/tr_context.h
+++ b/src/gallium/drivers/trace/tr_context.h
@@ -33,6 +33,7 @@
 #include "util/u_debug.h"
 #include "pipe/p_context.h"
 
+#include "tr_screen.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -45,7 +46,41 @@ struct trace_context
 
    struct pipe_context *pipe;
 
+   /* current state */
+   struct {
+      struct trace_shader *fs;
+      struct trace_shader *vs;
+
+      struct trace_texture *tex[PIPE_MAX_SAMPLERS];
+      unsigned num_texs;
+
+      unsigned nr_cbufs;
+      struct trace_texture *cbufs[PIPE_MAX_COLOR_BUFS];
+      struct trace_texture *zsbuf;
+   } curr;
+
+   struct {
+      struct trace_shader *fs;
+      struct trace_shader *vs;
+
+      struct trace_texture *tex;
+      struct trace_texture *surf;
+
+      int blocker;
+   } draw_rule;
+   unsigned draw_num_rules;
+   pipe_condvar draw_cond;
+   pipe_mutex draw_mutex;
+   int draw_blocker;
+   int draw_blocked;
+
+   /* for list on screen */
    struct tr_list list;
+
+   /* list of state objects */
+   pipe_mutex list_mutex;
+   unsigned num_shaders;
+   struct tr_list shaders;
 };
 
 
@@ -62,6 +97,9 @@ struct pipe_context *
 trace_context_create(struct pipe_screen *screen,
                      struct pipe_context *pipe);
 
+void
+trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx);
+
 
 #ifdef __cplusplus
 }
diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c
new file mode 100644
index 0000000000..781ca5d3bc
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_drm.c
@@ -0,0 +1,186 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "state_tracker/drm_api.h"
+
+#include "util/u_memory.h"
+#include "trace/tr_drm.h"
+#include "trace/tr_screen.h"
+#include "trace/tr_context.h"
+#include "trace/tr_buffer.h"
+#include "trace/tr_texture.h"
+
+struct trace_drm_api
+{
+   struct drm_api base;
+
+   struct drm_api *api;
+};
+
+static INLINE struct trace_drm_api *
+trace_drm_api(struct drm_api *_api)
+{
+   return (struct trace_drm_api *)_api;
+}
+
+static struct pipe_screen *
+trace_drm_create_screen(struct drm_api *_api, int fd,
+                        struct drm_create_screen_arg *arg)
+{
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct drm_api *api = tr_api->api;
+   struct pipe_screen *screen;
+
+   /* TODO trace call */
+
+   if (arg && arg->mode != DRM_CREATE_NORMAL)
+      return NULL;
+
+   screen = api->create_screen(api, fd, arg);
+
+   return trace_screen_create(screen);
+}
+
+static struct pipe_context *
+trace_drm_create_context(struct drm_api *_api,
+                         struct pipe_screen *_screen)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct drm_api *api = tr_api->api;
+   struct pipe_context *pipe;
+
+   /* TODO trace call */
+
+   pipe = api->create_context(api, screen);
+
+   pipe = trace_context_create(_screen, pipe);
+
+   return pipe;
+}
+
+static struct pipe_texture *
+trace_drm_texture_from_shared_handle(struct drm_api *_api,
+                                     struct pipe_screen *_screen,
+                                     struct pipe_texture *templ,
+                                     const char *name,
+                                     unsigned stride,
+                                     unsigned handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct drm_api *api = tr_api->api;
+   struct pipe_texture *result;
+
+   /* TODO trace call */
+
+   result = api->texture_from_shared_handle(api, screen, templ, name, stride, handle);
+
+   result = trace_texture_create(trace_screen(_screen), result);
+
+   return result;
+}
+
+static boolean
+trace_drm_shared_handle_from_texture(struct drm_api *_api,
+                                     struct pipe_screen *_screen,
+                                     struct pipe_texture *_texture,
+                                     unsigned *stride,
+                                     unsigned *handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_texture *tr_texture = trace_texture(_texture);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct pipe_texture *texture = tr_texture->texture;
+   struct drm_api *api = tr_api->api;
+
+   /* TODO trace call */
+
+   return api->shared_handle_from_texture(api, screen, texture, stride, handle);
+}
+
+static boolean
+trace_drm_local_handle_from_texture(struct drm_api *_api,
+                                    struct pipe_screen *_screen,
+                                    struct pipe_texture *_texture,
+                                    unsigned *stride,
+                                    unsigned *handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_texture *tr_texture = trace_texture(_texture);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct pipe_texture *texture = tr_texture->texture;
+   struct drm_api *api = tr_api->api;
+
+   /* TODO trace call */
+
+   return api->local_handle_from_texture(api, screen, texture, stride, handle);
+}
+
+static void
+trace_drm_destroy(struct drm_api *_api)
+{
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct drm_api *api = tr_api->api;
+   api->destroy(api);
+
+   free(tr_api);
+}
+
+struct drm_api *
+trace_drm_create(struct drm_api *api)
+{
+   struct trace_drm_api *tr_api;
+
+   if (!api)
+      goto error;
+
+   if (!trace_enabled())
+      goto error;
+
+   tr_api = CALLOC_STRUCT(trace_drm_api);
+
+   if (!tr_api)
+      goto error;
+
+   tr_api->base.create_screen = trace_drm_create_screen;
+   tr_api->base.create_context = trace_drm_create_context;
+   tr_api->base.texture_from_shared_handle = trace_drm_texture_from_shared_handle;
+   tr_api->base.shared_handle_from_texture = trace_drm_shared_handle_from_texture;
+   tr_api->base.local_handle_from_texture = trace_drm_local_handle_from_texture;
+   tr_api->base.destroy = trace_drm_destroy;
+   tr_api->api = api;
+
+   return &tr_api->base;
+
+error:
+   return api;
+}
diff --git a/src/gallium/drivers/trace/tr_drm.h b/src/gallium/drivers/trace/tr_drm.h
new file mode 100644
index 0000000000..845c66a32a
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_drm.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_DRM_H
+#define TR_DRM_H
+
+struct drm_api;
+
+struct drm_api* trace_drm_create(struct drm_api *api);
+
+#endif /* ID_DRM_H */
diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c
new file mode 100644
index 0000000000..bcf6751af4
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump_state.c
@@ -0,0 +1,549 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "tr_dump.h"
+#include "tr_dump_state.h"
+
+
+void trace_dump_format(enum pipe_format format)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_enum(pf_name(format) );
+}
+
+
+void trace_dump_block(const struct pipe_format_block *block)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_format_block");
+   trace_dump_member(uint, block, size);
+   trace_dump_member(uint, block, width);
+   trace_dump_member(uint, block, height);
+   trace_dump_struct_end();
+}
+
+
+static void trace_dump_reference(const struct pipe_reference *reference)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_reference");
+   trace_dump_member(int, &reference->count, count);
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_template(const struct pipe_texture *templat)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!templat) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_texture");
+
+   trace_dump_member(int, templat, target);
+   trace_dump_member(format, templat, format);
+
+   trace_dump_member_begin("width");
+   trace_dump_array(uint, templat->width, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("height");
+   trace_dump_array(uint, templat->height, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("depth");
+   trace_dump_array(uint, templat->depth, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&templat->block);
+   trace_dump_member_end();
+
+   trace_dump_member(uint, templat, last_level);
+   trace_dump_member(uint, templat, tex_usage);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_rasterizer_state");
+
+   trace_dump_member(bool, state, flatshade);
+   trace_dump_member(bool, state, light_twoside);
+   trace_dump_member(uint, state, front_winding);
+   trace_dump_member(uint, state, cull_mode);
+   trace_dump_member(uint, state, fill_cw);
+   trace_dump_member(uint, state, fill_ccw);
+   trace_dump_member(bool, state, offset_cw);
+   trace_dump_member(bool, state, offset_ccw);
+   trace_dump_member(bool, state, scissor);
+   trace_dump_member(bool, state, poly_smooth);
+   trace_dump_member(bool, state, poly_stipple_enable);
+   trace_dump_member(bool, state, point_smooth);
+   trace_dump_member(bool, state, point_sprite);
+   trace_dump_member(bool, state, point_size_per_vertex);
+   trace_dump_member(bool, state, multisample);
+   trace_dump_member(bool, state, line_smooth);
+   trace_dump_member(bool, state, line_stipple_enable);
+   trace_dump_member(uint, state, line_stipple_factor);
+   trace_dump_member(uint, state, line_stipple_pattern);
+   trace_dump_member(bool, state, line_last_pixel);
+   trace_dump_member(bool, state, bypass_vs_clip_and_viewport);
+   trace_dump_member(bool, state, flatshade_first);
+   trace_dump_member(bool, state, gl_rasterization_rules);
+
+   trace_dump_member(float, state, line_width);
+   trace_dump_member(float, state, point_size);
+   trace_dump_member(float, state, point_size_min);
+   trace_dump_member(float, state, point_size_max);
+   trace_dump_member(float, state, offset_units);
+   trace_dump_member(float, state, offset_scale);
+
+   trace_dump_member_array(uint, state, sprite_coord_mode);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_poly_stipple");
+
+   trace_dump_member_begin("stipple");
+   trace_dump_array(uint,
+                    state->stipple,
+                    Elements(state->stipple));
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_viewport_state");
+
+   trace_dump_member_array(float, state, scale);
+   trace_dump_member_array(float, state, translate);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_scissor_state");
+
+   trace_dump_member(uint, state, minx);
+   trace_dump_member(uint, state, miny);
+   trace_dump_member(uint, state, maxx);
+   trace_dump_member(uint, state, maxy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_clip_state(const struct pipe_clip_state *state)
+{
+   unsigned i;
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_clip_state");
+
+   trace_dump_member_begin("ucp");
+   trace_dump_array_begin();
+   for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) {
+      trace_dump_elem_begin();
+      trace_dump_array(float, state->ucp[i], 4);
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member(uint, state, nr);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_constant_buffer");
+
+   trace_dump_member(buffer_ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_shader_state(const struct pipe_shader_state *state)
+{
+   static char str[8192];
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   tgsi_dump_str(state->tokens, 0, str, sizeof(str));
+
+   trace_dump_struct_begin("pipe_shader_state");
+
+   trace_dump_member_begin("tokens");
+   trace_dump_string(str);
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state)
+{
+   unsigned i;
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_depth_stencil_alpha_state");
+
+   trace_dump_member_begin("depth");
+   trace_dump_struct_begin("pipe_depth_state");
+   trace_dump_member(bool, &state->depth, enabled);
+   trace_dump_member(bool, &state->depth, writemask);
+   trace_dump_member(uint, &state->depth, func);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+
+   trace_dump_member_begin("stencil");
+   trace_dump_array_begin();
+   for(i = 0; i < Elements(state->stencil); ++i) {
+      trace_dump_elem_begin();
+      trace_dump_struct_begin("pipe_stencil_state");
+      trace_dump_member(bool, &state->stencil[i], enabled);
+      trace_dump_member(uint, &state->stencil[i], func);
+      trace_dump_member(uint, &state->stencil[i], fail_op);
+      trace_dump_member(uint, &state->stencil[i], zpass_op);
+      trace_dump_member(uint, &state->stencil[i], zfail_op);
+      trace_dump_member(uint, &state->stencil[i], ref_value);
+      trace_dump_member(uint, &state->stencil[i], valuemask);
+      trace_dump_member(uint, &state->stencil[i], writemask);
+      trace_dump_struct_end();
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member_begin("alpha");
+   trace_dump_struct_begin("pipe_alpha_state");
+   trace_dump_member(bool, &state->alpha, enabled);
+   trace_dump_member(uint, &state->alpha, func);
+   trace_dump_member(float, &state->alpha, ref_value);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_state(const struct pipe_blend_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_state");
+
+   trace_dump_member(bool, state, blend_enable);
+
+   trace_dump_member(uint, state, rgb_func);
+   trace_dump_member(uint, state, rgb_src_factor);
+   trace_dump_member(uint, state, rgb_dst_factor);
+
+   trace_dump_member(uint, state, alpha_func);
+   trace_dump_member(uint, state, alpha_src_factor);
+   trace_dump_member(uint, state, alpha_dst_factor);
+
+   trace_dump_member(bool, state, logicop_enable);
+   trace_dump_member(uint, state, logicop_func);
+
+   trace_dump_member(uint, state, colormask);
+   trace_dump_member(bool, state, dither);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_color(const struct pipe_blend_color *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_color");
+
+   trace_dump_member_array(float, state, color);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_framebuffer_state");
+
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+   trace_dump_member(uint, state, nr_cbufs);
+   trace_dump_member_array(ptr, state, cbufs);
+   trace_dump_member(ptr, state, zsbuf);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_sampler_state");
+
+   trace_dump_member(uint, state, wrap_s);
+   trace_dump_member(uint, state, wrap_t);
+   trace_dump_member(uint, state, wrap_r);
+   trace_dump_member(uint, state, min_img_filter);
+   trace_dump_member(uint, state, min_mip_filter);
+   trace_dump_member(uint, state, mag_img_filter);
+   trace_dump_member(bool, state, compare_mode);
+   trace_dump_member(uint, state, compare_func);
+   trace_dump_member(bool, state, normalized_coords);
+   trace_dump_member(uint, state, prefilter);
+   trace_dump_member(float, state, lod_bias);
+   trace_dump_member(float, state, min_lod);
+   trace_dump_member(float, state, max_lod);
+   trace_dump_member_array(float, state, border_color);
+   trace_dump_member(float, state, max_anisotropy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_surface(const struct pipe_surface *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_surface");
+
+   trace_dump_reference(&state->reference);
+
+   trace_dump_member(format, state, format);
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+
+   trace_dump_member(uint, state, layout);
+   trace_dump_member(uint, state, offset);
+   trace_dump_member(uint, state, usage);
+
+   trace_dump_member(ptr, state, texture);
+   trace_dump_member(uint, state, face);
+   trace_dump_member(uint, state, level);
+   trace_dump_member(uint, state, zslice);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_transfer(const struct pipe_transfer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_transfer");
+
+   trace_dump_member(format, state, format);
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&state->block);
+   trace_dump_member_end();
+
+   trace_dump_member(uint, state, nblocksx);
+   trace_dump_member(uint, state, nblocksy);
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, usage);
+
+   trace_dump_member(ptr, state, texture);
+   trace_dump_member(uint, state, face);
+   trace_dump_member(uint, state, level);
+   trace_dump_member(uint, state, zslice);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_buffer");
+
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, max_index);
+   trace_dump_member(uint, state, buffer_offset);
+   trace_dump_member(buffer_ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_element");
+
+   trace_dump_member(uint, state, src_offset);
+
+   trace_dump_member(uint, state, vertex_buffer_index);
+   trace_dump_member(uint, state, nr_components);
+
+   trace_dump_member(format, state, src_format);
+
+   trace_dump_struct_end();
+}
diff --git a/src/gallium/drivers/trace/tr_dump_state.h b/src/gallium/drivers/trace/tr_dump_state.h
new file mode 100644
index 0000000000..05b821adb6
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump_state.h
@@ -0,0 +1,78 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_DUMP_STATE_H_
+#define TR_DUMP_STATE_H_
+
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+
+
+void trace_dump_format(enum pipe_format format);
+
+void trace_dump_block(const struct pipe_format_block *block);
+
+void trace_dump_template(const struct pipe_texture *templat);
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state);
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state);
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state);
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state);
+
+void trace_dump_clip_state(const struct pipe_clip_state *state);
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state);
+
+void trace_dump_token(const struct tgsi_token *token);
+
+void trace_dump_shader_state(const struct pipe_shader_state *state);
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state);
+
+void trace_dump_blend_state(const struct pipe_blend_state *state);
+
+void trace_dump_blend_color(const struct pipe_blend_color *state);
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state);
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state);
+
+void trace_dump_surface(const struct pipe_surface *state);
+
+void trace_dump_transfer(const struct pipe_transfer *state);
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state);
+
+
+#endif /* TR_STATE_H */
diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c
new file mode 100644
index 0000000000..e85ac15edc
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_rbug.c
@@ -0,0 +1,864 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_string.h"
+#include "util/u_memory.h"
+#include "util/u_simple_list.h"
+#include "util/u_network.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+#include "tr_buffer.h"
+#include "tr_texture.h"
+
+#include "rbug/rbug.h"
+
+#include <errno.h>
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  define sleep Sleep
+#elif defined(PIPE_OS_LINUX)
+void usleep(int);
+#  define sleep usleep
+#else
+#  warning "No socket implementation"
+#endif
+
+#define U642VOID(x) ((void *)(unsigned long)(x))
+#define VOID2U64(x) ((uint64_t)(unsigned long)(x))
+
+struct trace_rbug
+{
+   struct trace_screen *tr_scr;
+   struct rbug_connection *con;
+   pipe_thread thread;
+   boolean running;
+};
+
+PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug);
+
+
+/**********************************************************
+ * Helper functions
+ */
+
+
+static struct trace_context *
+trace_rbug_get_context_locked(struct trace_screen *tr_scr, rbug_context_t ctx)
+{
+   struct trace_context *tr_ctx = NULL;
+   struct tr_list *ptr;
+
+   foreach(ptr, &tr_scr->contexts) {
+      tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list));
+      if (ctx == VOID2U64(tr_ctx))
+         break;
+      tr_ctx = NULL;
+   }
+
+   return tr_ctx;
+}
+
+static struct trace_shader *
+trace_rbug_get_shader_locked(struct trace_context *tr_ctx, rbug_shader_t shdr)
+{
+   struct trace_shader *tr_shdr = NULL;
+   struct tr_list *ptr;
+
+   foreach(ptr, &tr_ctx->shaders) {
+      tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list));
+      if (shdr == VOID2U64(tr_shdr))
+         break;
+      tr_shdr = NULL;
+   }
+
+   return tr_shdr;
+}
+
+static void *
+trace_shader_create_locked(struct pipe_context *pipe,
+                           struct trace_shader *tr_shdr,
+                           struct tgsi_token *tokens)
+{
+   void *state = NULL;
+   struct pipe_shader_state pss = { 0 };
+   pss.tokens = tokens;
+
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      state = pipe->create_fs_state(pipe, &pss);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      state = pipe->create_vs_state(pipe, &pss);
+   } else
+      assert(0);
+
+   return state;
+}
+
+static void
+trace_shader_bind_locked(struct pipe_context *pipe,
+                         struct trace_shader *tr_shdr,
+                         void *state)
+{
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      pipe->bind_fs_state(pipe, state);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      pipe->bind_vs_state(pipe, state);
+   } else
+      assert(0);
+}
+
+static void
+trace_shader_delete_locked(struct pipe_context *pipe,
+                           struct trace_shader *tr_shdr,
+                           void *state)
+{
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      pipe->delete_fs_state(pipe, state);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      pipe->delete_vs_state(pipe, state);
+   } else
+      assert(0);
+}
+
+/************************************************
+ * Request handler functions
+ */
+
+
+static int
+trace_rbug_texture_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex = NULL;
+   struct tr_list *ptr;
+   rbug_texture_t *texs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   texs = MALLOC(tr_scr->num_textures * sizeof(rbug_texture_t));
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      texs[i++] = VOID2U64(tr_tex);
+   }
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_texture_list_reply(tr_rbug->con, serial, texs, i, NULL);
+   FREE(texs);
+
+   return 0;
+}
+
+static int
+trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex;
+   struct rbug_proto_texture_info *gpti = (struct rbug_proto_texture_info *)header;
+   struct tr_list *ptr;
+   struct pipe_texture *t;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      if (gpti->texture == VOID2U64(tr_tex))
+         break;
+      tr_tex = NULL;
+   }
+
+   if (!tr_tex) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   t = tr_tex->texture;
+   rbug_send_texture_info_reply(tr_rbug->con, serial,
+                               t->target, t->format,
+                               t->width, t->last_level + 1,
+                               t->height, t->last_level + 1,
+                               t->depth, t->last_level + 1,
+                               t->block.width, t->block.height, t->block.size,
+                               t->last_level,
+                               t->nr_samples,
+                               t->tex_usage,
+                               NULL);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_texture_read *gptr = (struct rbug_proto_texture_read *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex;
+   struct tr_list *ptr;
+
+   struct pipe_screen *screen = tr_scr->screen;
+   struct pipe_texture *tex;
+   struct pipe_transfer *t;
+
+   void *map;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      if (gptr->texture == VOID2U64(tr_tex))
+         break;
+      tr_tex = NULL;
+   }
+
+   if (!tr_tex) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   tex = tr_tex->texture;
+   t = screen->get_tex_transfer(tr_scr->screen, tex,
+                                gptr->face, gptr->level, gptr->zslice,
+                                PIPE_TRANSFER_READ,
+                                gptr->x, gptr->y, gptr->w, gptr->h);
+
+   map = screen->transfer_map(screen, t);
+
+   rbug_send_texture_read_reply(tr_rbug->con, serial,
+                                t->format,
+                                t->block.width, t->block.height, t->block.size,
+                                (uint8_t*)map, t->stride * t->nblocksy,
+                                t->stride,
+                                NULL);
+
+   screen->transfer_unmap(screen, t);
+   screen->tex_transfer_destroy(t);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct tr_list *ptr;
+   struct trace_context *tr_ctx = NULL;
+   rbug_context_t *ctxs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   ctxs = MALLOC(tr_scr->num_contexts * sizeof(rbug_context_t));
+   foreach(ptr, &tr_scr->contexts) {
+      tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list));
+      ctxs[i++] = VOID2U64(tr_ctx);
+   }
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_context_list_reply(tr_rbug->con, serial, ctxs, i, NULL);
+   FREE(ctxs);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_info *info = (struct rbug_proto_context_info *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   rbug_texture_t cbufs[PIPE_MAX_COLOR_BUFS];
+   rbug_texture_t texs[PIPE_MAX_SAMPLERS];
+   int i;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   trace_dump_call_lock();
+
+   for (i = 0; i < tr_ctx->curr.nr_cbufs; i++)
+      cbufs[i] = VOID2U64(tr_ctx->curr.cbufs[i]);
+
+   for (i = 0; i < tr_ctx->curr.num_texs; i++)
+      texs[i] = VOID2U64(tr_ctx->curr.tex[i]);
+
+   rbug_send_context_info_reply(tr_rbug->con, serial,
+                                VOID2U64(tr_ctx->curr.vs), VOID2U64(tr_ctx->curr.fs),
+                                texs, tr_ctx->curr.num_texs,
+                                cbufs, tr_ctx->curr.nr_cbufs,
+                                VOID2U64(tr_ctx->curr.zsbuf),
+                                tr_ctx->draw_blocker, tr_ctx->draw_blocked, NULL);
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_block(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_block *block = (struct rbug_proto_context_draw_block *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, block->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   tr_ctx->draw_blocker |= block->block;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_step(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_step *step = (struct rbug_proto_context_draw_step *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, step->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) {
+      if (step->step & RBUG_BLOCK_RULE)
+         tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK;
+   } else {
+      tr_ctx->draw_blocked &= ~step->step;
+   }
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_unblock(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_unblock *unblock = (struct rbug_proto_context_draw_unblock *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, unblock->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) {
+      if (unblock->unblock & RBUG_BLOCK_RULE)
+         tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK;
+   } else {
+      tr_ctx->draw_blocked &= ~unblock->unblock;
+   }
+   tr_ctx->draw_blocker &= ~unblock->unblock;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_rule(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_rule *rule = (struct rbug_proto_context_draw_rule *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, rule->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   tr_ctx->draw_rule.vs = U642VOID(rule->vertex);
+   tr_ctx->draw_rule.fs = U642VOID(rule->fragment);
+   tr_ctx->draw_rule.tex = U642VOID(rule->texture);
+   tr_ctx->draw_rule.surf = U642VOID(rule->surface);
+   tr_ctx->draw_rule.blocker = rule->block;
+   tr_ctx->draw_blocker |= RBUG_BLOCK_RULE;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_flush(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_flush *flush = (struct rbug_proto_context_flush *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, flush->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   trace_dump_call_lock();
+
+   tr_ctx->pipe->flush(tr_ctx->pipe, flush->flags, NULL);
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_shader_list *list = (struct rbug_proto_shader_list *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   struct tr_list *ptr;
+   rbug_shader_t *shdrs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, list->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+   shdrs = MALLOC(tr_ctx->num_shaders * sizeof(rbug_shader_t));
+   foreach(ptr, &tr_ctx->shaders) {
+      tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list));
+      shdrs[i++] = VOID2U64(tr_shdr);
+   }
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_shader_list_reply(tr_rbug->con, serial, shdrs, i, NULL);
+   FREE(shdrs);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_shader_info *info = (struct rbug_proto_shader_info *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   unsigned original_len;
+   unsigned replaced_len;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, info->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* just in case */
+   assert(sizeof(struct tgsi_token) == 4);
+
+   original_len = tgsi_num_tokens(tr_shdr->tokens);
+   if (tr_shdr->replaced_tokens)
+      replaced_len = tgsi_num_tokens(tr_shdr->replaced_tokens);
+   else
+      replaced_len = 0;
+
+   rbug_send_shader_info_reply(tr_rbug->con, serial,
+                               (uint32_t*)tr_shdr->tokens, original_len,
+                               (uint32_t*)tr_shdr->replaced_tokens, replaced_len,
+                               tr_shdr->disabled,
+                               NULL);
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_disable(struct trace_rbug *tr_rbug, struct rbug_header *header)
+{
+   struct rbug_proto_shader_disable *dis = (struct rbug_proto_shader_disable *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, dis->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, dis->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   tr_shdr->disabled = dis->disable;
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_replace(struct trace_rbug *tr_rbug, struct rbug_header *header)
+{
+   struct rbug_proto_shader_replace *rep = (struct rbug_proto_shader_replace *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   struct pipe_context *pipe = NULL;
+   void *state;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, rep->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, rep->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   trace_dump_call_lock();
+
+   pipe = tr_ctx->pipe;
+
+   /* remove old replaced shader */
+   if (tr_shdr->replaced) {
+      if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr)
+         trace_shader_bind_locked(pipe, tr_shdr, tr_shdr->state);
+
+      FREE(tr_shdr->replaced_tokens);
+      trace_shader_delete_locked(pipe, tr_shdr, tr_shdr->replaced);
+      tr_shdr->replaced = NULL;
+      tr_shdr->replaced_tokens = NULL;
+   }
+
+   /* empty inputs means restore old which we did above */
+   if (rep->tokens_len == 0)
+      goto out;
+
+   tr_shdr->replaced_tokens = tgsi_dup_tokens((struct tgsi_token *)rep->tokens);
+   if (!tr_shdr->replaced_tokens)
+      goto err;
+
+   state = trace_shader_create_locked(pipe, tr_shdr, tr_shdr->replaced_tokens);
+   if (!state)
+      goto err;
+
+   /* bind new shader if the shader is currently a bound */
+   if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr)
+      trace_shader_bind_locked(pipe, tr_shdr, state);
+
+   /* save state */
+   tr_shdr->replaced = state;
+
+out:
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+
+err:
+   FREE(tr_shdr->replaced_tokens);
+   tr_shdr->replaced = NULL;
+   tr_shdr->replaced_tokens = NULL;
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+   return -EINVAL;
+}
+
+static boolean
+trace_rbug_header(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   int ret = 0;
+
+   switch(header->opcode) {
+      case RBUG_OP_PING:
+         rbug_send_ping_reply(tr_rbug->con, serial, NULL);
+         break;
+      case RBUG_OP_TEXTURE_LIST:
+         ret = trace_rbug_texture_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_TEXTURE_INFO:
+         ret = trace_rbug_texture_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_TEXTURE_READ:
+         ret = trace_rbug_texture_read(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_LIST:
+         ret = trace_rbug_context_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_INFO:
+         ret = trace_rbug_context_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_BLOCK:
+         ret = trace_rbug_context_draw_block(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_STEP:
+         ret = trace_rbug_context_draw_step(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_UNBLOCK:
+         ret = trace_rbug_context_draw_unblock(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_RULE:
+         ret = trace_rbug_context_draw_rule(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_FLUSH:
+         ret = trace_rbug_context_flush(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_LIST:
+         ret = trace_rbug_shader_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_INFO:
+         ret = trace_rbug_shader_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_DISABLE:
+         ret = trace_rbug_shader_disable(tr_rbug, header);
+         break;
+      case RBUG_OP_SHADER_REPLACE:
+         ret = trace_rbug_shader_replace(tr_rbug, header);
+         break;
+      default:
+         debug_printf("%s - unsupported opcode %u\n", __FUNCTION__, header->opcode);
+         ret = -ENOSYS;
+         break;
+   }
+   rbug_free_header(header);
+
+   if (ret)
+      rbug_send_error_reply(tr_rbug->con, serial, ret, NULL);
+
+   return TRUE;
+}
+
+static void
+trace_rbug_con(struct trace_rbug *tr_rbug)
+{
+   struct rbug_header *header;
+   uint32_t serial;
+
+   debug_printf("%s - connection received\n", __FUNCTION__);
+
+   while(tr_rbug->running) {
+      header = rbug_get_message(tr_rbug->con, &serial);
+      if (!header)
+         break;
+
+      if (!trace_rbug_header(tr_rbug, header, serial))
+         break;
+   }
+
+   debug_printf("%s - connection closed\n", __FUNCTION__);
+
+   rbug_disconnect(tr_rbug->con);
+   tr_rbug->con = NULL;
+}
+
+PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug)
+{
+   struct trace_rbug *tr_rbug = void_tr_rbug;
+   uint16_t port = 13370;
+   int s = -1;
+   int c;
+
+   u_socket_init();
+
+   for (;port <= 13379 && s < 0; port++)
+      s = u_socket_listen_on_port(port);
+
+   if (s < 0) {
+      debug_printf("trace_rbug - failed to listen\n");
+      return NULL;
+   }
+
+   u_socket_block(s, false);
+
+   debug_printf("trace_rbug - remote debugging listening on port %u\n", --port);
+
+   while(tr_rbug->running) {
+      sleep(1);
+
+      c = u_socket_accept(s);
+      if (c < 0)
+         continue;
+
+      u_socket_block(c, true);
+      tr_rbug->con = rbug_from_socket(c);
+
+      trace_rbug_con(tr_rbug);
+
+      u_socket_close(c);
+   }
+
+   u_socket_close(s);
+
+   u_socket_stop();
+
+   return NULL;
+}
+
+/**********************************************************
+ *
+ */
+
+struct trace_rbug *
+trace_rbug_start(struct trace_screen *tr_scr)
+{
+   struct trace_rbug *tr_rbug = CALLOC_STRUCT(trace_rbug);
+   if (!tr_rbug)
+      return NULL;
+
+   tr_rbug->tr_scr = tr_scr;
+   tr_rbug->running = TRUE;
+   tr_rbug->thread = pipe_thread_create(trace_rbug_thread, tr_rbug);
+
+   return tr_rbug;
+}
+
+void
+trace_rbug_stop(struct trace_rbug *tr_rbug)
+{
+   if (!tr_rbug)
+      return;
+
+   tr_rbug->running = false;
+   pipe_thread_wait(tr_rbug->thread);
+
+   FREE(tr_rbug);
+
+   return;
+}
+
+void
+trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx)
+{
+   struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen);
+   struct trace_rbug *tr_rbug = tr_scr->rbug;
+
+   if (tr_rbug && tr_rbug->con)
+      rbug_send_context_draw_blocked(tr_rbug->con,
+                                     VOID2U64(tr_ctx), tr_ctx->draw_blocked, NULL);
+}
diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c
index 12a8535342..26f1c04594 100644
--- a/src/gallium/drivers/trace/tr_screen.c
+++ b/src/gallium/drivers/trace/tr_screen.c
@@ -30,13 +30,16 @@
 
 #include "tr_buffer.h"
 #include "tr_dump.h"
-#include "tr_state.h"
+#include "tr_dump_state.h"
 #include "tr_texture.h"
 #include "tr_screen.h"
 
 #include "pipe/p_inlines.h"
 
 
+static boolean trace = FALSE;
+static boolean rbug = FALSE;
+
 static const char *
 trace_screen_get_name(struct pipe_screen *_screen)
 {
@@ -212,10 +215,12 @@ static struct pipe_texture *
 trace_screen_texture_blanket(struct pipe_screen *_screen,
                              const struct pipe_texture *templat,
                              const unsigned *ppitch,
-                             struct pipe_buffer *buffer)
+                             struct pipe_buffer *_buffer)
 {
    struct trace_screen *tr_scr = trace_screen(_screen);
+   struct trace_buffer *tr_buf = trace_buffer(_buffer);
    struct pipe_screen *screen = tr_scr->screen;
+   struct pipe_buffer *buffer = tr_buf->buffer;
    unsigned pitch = *ppitch;
    struct pipe_texture *result;
 
@@ -457,6 +462,7 @@ trace_screen_surface_buffer_create(struct pipe_screen *_screen,
                                    unsigned width, unsigned height,
                                    enum pipe_format format,
                                    unsigned usage,
+                                   unsigned tex_usage,
                                    unsigned *pstride)
 {
    struct trace_screen *tr_scr = trace_screen(_screen);
@@ -471,11 +477,13 @@ trace_screen_surface_buffer_create(struct pipe_screen *_screen,
    trace_dump_arg(uint, height);
    trace_dump_arg(format, format);
    trace_dump_arg(uint, usage);
+   trace_dump_arg(uint, tex_usage);
 
    result = screen->surface_buffer_create(screen,
                                           width, height,
                                           format,
                                           usage,
+                                          tex_usage,
                                           pstride);
 
    stride = *pstride;
@@ -818,18 +826,41 @@ trace_screen_destroy(struct pipe_screen *_screen)
    struct pipe_screen *screen = tr_scr->screen;
 
    trace_dump_call_begin("pipe_screen", "destroy");
-
    trace_dump_arg(ptr, screen);
-
-   screen->destroy(screen);
-
    trace_dump_call_end();
-
    trace_dump_trace_end();
 
+   if (tr_scr->rbug)
+      trace_rbug_stop(tr_scr->rbug);
+
+   screen->destroy(screen);
+
    FREE(tr_scr);
 }
 
+boolean
+trace_enabled(void)
+{
+   static boolean firstrun = TRUE;
+
+   if (!firstrun)
+      return trace;
+   firstrun = FALSE;
+
+   trace_dump_init();
+
+   if(trace_dump_trace_begin()) {
+      trace_dumping_start();
+      trace = TRUE;
+   }
+
+   if (debug_get_bool_option("GALLIUM_RBUG", FALSE)) {
+      trace = TRUE;
+      rbug = TRUE;
+   }
+
+   return trace;
+}
 
 struct pipe_screen *
 trace_screen_create(struct pipe_screen *screen)
@@ -840,13 +871,9 @@ trace_screen_create(struct pipe_screen *screen)
    if(!screen)
       goto error1;
 
-   trace_dump_init();
-
-   if(!trace_dump_trace_begin())
+   if (!trace_enabled())
       goto error1;
 
-   trace_dumping_start();
-
    trace_dump_call_begin("", "pipe_screen_create");
 
    tr_scr = CALLOC_STRUCT(trace_screen);
@@ -904,6 +931,9 @@ trace_screen_create(struct pipe_screen *screen)
    trace_dump_ret(ptr, screen);
    trace_dump_call_end();
 
+   if (rbug)
+      tr_scr->rbug = trace_rbug_start(tr_scr);
+
    return &tr_scr->base;
 
 #if 0
diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h
index 59f254166d..dba8cd7c65 100644
--- a/src/gallium/drivers/trace/tr_screen.h
+++ b/src/gallium/drivers/trace/tr_screen.h
@@ -57,6 +57,9 @@ struct trace_screen
 
    struct pipe_screen *screen;
 
+   /* remote debugger */
+   struct trace_rbug *rbug;
+
    pipe_mutex list_mutex;
    int num_buffers;
    int num_contexts;
@@ -71,14 +74,34 @@ struct trace_screen
 };
 
 
+/*
+ * tr_rbug.c
+ */
+
+
+struct trace_rbug;
+
+struct trace_rbug *
+trace_rbug_start(struct trace_screen *tr_scr);
+
+void
+trace_rbug_stop(struct trace_rbug *tr_rbug);
+
+
+/*
+ * tr_screen.c
+ */
+
+
+boolean
+trace_enabled(void);
+
 struct trace_screen *
 trace_screen(struct pipe_screen *screen);
 
-
 struct pipe_screen *
 trace_screen_create(struct pipe_screen *screen);
 
-
 void
 trace_screen_user_buffer_update(struct pipe_screen *screen,
                                 struct pipe_buffer *buffer);
@@ -99,6 +122,7 @@ trace_screen_user_buffer_update(struct pipe_screen *screen,
       pipe_mutex_unlock(tr_scr->list_mutex);             \
    } while (0)
 
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c
index a9570c1aeb..d8c11640bf 100644
--- a/src/gallium/drivers/trace/tr_state.c
+++ b/src/gallium/drivers/trace/tr_state.c
@@ -1,491 +1,66 @@
-/**************************************************************************
- *
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+/*
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
  *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
  *
- **************************************************************************/
-
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
 
-#include "pipe/p_compiler.h"
-#include "util/u_memory.h"
-#include "tgsi/tgsi_dump.h"
-
-#include "tr_dump.h"
 #include "tr_state.h"
 
+#include "util/u_memory.h"
+#include "util/u_simple_list.h"
 
-void trace_dump_format(enum pipe_format format)
-{
-   trace_dump_enum(pf_name(format) );
-}
-
-
-void trace_dump_block(const struct pipe_format_block *block)
-{
-   trace_dump_struct_begin("pipe_format_block");
-   trace_dump_member(uint, block, size);
-   trace_dump_member(uint, block, width);
-   trace_dump_member(uint, block, height);
-   trace_dump_struct_end();
-}
-
-
-static void trace_dump_reference(const struct pipe_reference *reference)
-{
-   trace_dump_struct_begin("pipe_reference");
-   trace_dump_member(int, &reference->count, count);
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_template(const struct pipe_texture *templat)
-{
-   if(!templat) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_texture");
-
-   trace_dump_member(int, templat, target);
-   trace_dump_member(format, templat, format);
-
-   trace_dump_member_begin("width");
-   trace_dump_array(uint, templat->width, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("height");
-   trace_dump_array(uint, templat->height, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("depth");
-   trace_dump_array(uint, templat->depth, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("block");
-   trace_dump_block(&templat->block);
-   trace_dump_member_end();
-
-   trace_dump_member(uint, templat, last_level);
-   trace_dump_member(uint, templat, tex_usage);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_rasterizer_state");
-
-   trace_dump_member(bool, state, flatshade);
-   trace_dump_member(bool, state, light_twoside);
-   trace_dump_member(uint, state, front_winding);
-   trace_dump_member(uint, state, cull_mode);
-   trace_dump_member(uint, state, fill_cw);
-   trace_dump_member(uint, state, fill_ccw);
-   trace_dump_member(bool, state, offset_cw);
-   trace_dump_member(bool, state, offset_ccw);
-   trace_dump_member(bool, state, scissor);
-   trace_dump_member(bool, state, poly_smooth);
-   trace_dump_member(bool, state, poly_stipple_enable);
-   trace_dump_member(bool, state, point_smooth);
-   trace_dump_member(bool, state, point_sprite);
-   trace_dump_member(bool, state, point_size_per_vertex);
-   trace_dump_member(bool, state, multisample);
-   trace_dump_member(bool, state, line_smooth);
-   trace_dump_member(bool, state, line_stipple_enable);
-   trace_dump_member(uint, state, line_stipple_factor);
-   trace_dump_member(uint, state, line_stipple_pattern);
-   trace_dump_member(bool, state, line_last_pixel);
-   trace_dump_member(bool, state, bypass_vs_clip_and_viewport);
-   trace_dump_member(bool, state, flatshade_first);
-   trace_dump_member(bool, state, gl_rasterization_rules);
-
-   trace_dump_member(float, state, line_width);
-   trace_dump_member(float, state, point_size);
-   trace_dump_member(float, state, point_size_min);
-   trace_dump_member(float, state, point_size_max);
-   trace_dump_member(float, state, offset_units);
-   trace_dump_member(float, state, offset_scale);
-
-   trace_dump_member_array(uint, state, sprite_coord_mode);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_poly_stipple(const struct pipe_poly_stipple *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_poly_stipple");
-
-   trace_dump_member_begin("stipple");
-   trace_dump_array(uint,
-                    state->stipple,
-                    Elements(state->stipple));
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_viewport_state(const struct pipe_viewport_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_viewport_state");
-
-   trace_dump_member_array(float, state, scale);
-   trace_dump_member_array(float, state, translate);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_scissor_state(const struct pipe_scissor_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_scissor_state");
-
-   trace_dump_member(uint, state, minx);
-   trace_dump_member(uint, state, miny);
-   trace_dump_member(uint, state, maxx);
-   trace_dump_member(uint, state, maxy);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_clip_state(const struct pipe_clip_state *state)
-{
-   unsigned i;
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_clip_state");
-
-   trace_dump_member_begin("ucp");
-   trace_dump_array_begin();
-   for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) {
-      trace_dump_elem_begin();
-      trace_dump_array(float, state->ucp[i], 4);
-      trace_dump_elem_end();
-   }
-   trace_dump_array_end();
-   trace_dump_member_end();
-
-   trace_dump_member(uint, state, nr);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_constant_buffer(const struct pipe_constant_buffer *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_constant_buffer");
-
-   trace_dump_member(buffer_ptr, state, buffer);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_shader_state(const struct pipe_shader_state *state)
-{
-   static char str[8192];
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   tgsi_dump_str(state->tokens, 0, str, sizeof(str));
-
-   trace_dump_struct_begin("pipe_shader_state");
-
-   trace_dump_member_begin("tokens");
-   trace_dump_string(str);
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state)
-{
-   unsigned i;
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_depth_stencil_alpha_state");
-
-   trace_dump_member_begin("depth");
-   trace_dump_struct_begin("pipe_depth_state");
-   trace_dump_member(bool, &state->depth, enabled);
-   trace_dump_member(bool, &state->depth, writemask);
-   trace_dump_member(uint, &state->depth, func);
-   trace_dump_member(bool, &state->depth, occlusion_count);
-   trace_dump_struct_end();
-   trace_dump_member_end();
-
-   trace_dump_member_begin("stencil");
-   trace_dump_array_begin();
-   for(i = 0; i < Elements(state->stencil); ++i) {
-      trace_dump_elem_begin();
-      trace_dump_struct_begin("pipe_stencil_state");
-      trace_dump_member(bool, &state->stencil[i], enabled);
-      trace_dump_member(uint, &state->stencil[i], func);
-      trace_dump_member(uint, &state->stencil[i], fail_op);
-      trace_dump_member(uint, &state->stencil[i], zpass_op);
-      trace_dump_member(uint, &state->stencil[i], zfail_op);
-      trace_dump_member(uint, &state->stencil[i], ref_value);
-      trace_dump_member(uint, &state->stencil[i], valuemask);
-      trace_dump_member(uint, &state->stencil[i], writemask);
-      trace_dump_struct_end();
-      trace_dump_elem_end();
-   }
-   trace_dump_array_end();
-   trace_dump_member_end();
-
-   trace_dump_member_begin("alpha");
-   trace_dump_struct_begin("pipe_alpha_state");
-   trace_dump_member(bool, &state->alpha, enabled);
-   trace_dump_member(uint, &state->alpha, func);
-   trace_dump_member(float, &state->alpha, ref_value);
-   trace_dump_struct_end();
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_blend_state(const struct pipe_blend_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_blend_state");
-
-   trace_dump_member(bool, state, blend_enable);
-
-   trace_dump_member(uint, state, rgb_func);
-   trace_dump_member(uint, state, rgb_src_factor);
-   trace_dump_member(uint, state, rgb_dst_factor);
-
-   trace_dump_member(uint, state, alpha_func);
-   trace_dump_member(uint, state, alpha_src_factor);
-   trace_dump_member(uint, state, alpha_dst_factor);
-
-   trace_dump_member(bool, state, logicop_enable);
-   trace_dump_member(uint, state, logicop_func);
-
-   trace_dump_member(uint, state, colormask);
-   trace_dump_member(bool, state, dither);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_blend_color(const struct pipe_blend_color *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_blend_color");
-
-   trace_dump_member_array(float, state, color);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state)
-{
-   trace_dump_struct_begin("pipe_framebuffer_state");
-
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
-   trace_dump_member(uint, state, nr_cbufs);
-   trace_dump_member_array(ptr, state, cbufs);
-   trace_dump_member(ptr, state, zsbuf);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_sampler_state(const struct pipe_sampler_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_sampler_state");
-
-   trace_dump_member(uint, state, wrap_s);
-   trace_dump_member(uint, state, wrap_t);
-   trace_dump_member(uint, state, wrap_r);
-   trace_dump_member(uint, state, min_img_filter);
-   trace_dump_member(uint, state, min_mip_filter);
-   trace_dump_member(uint, state, mag_img_filter);
-   trace_dump_member(bool, state, compare_mode);
-   trace_dump_member(uint, state, compare_func);
-   trace_dump_member(bool, state, normalized_coords);
-   trace_dump_member(uint, state, prefilter);
-   trace_dump_member(float, state, shadow_ambient);
-   trace_dump_member(float, state, lod_bias);
-   trace_dump_member(float, state, min_lod);
-   trace_dump_member(float, state, max_lod);
-   trace_dump_member_array(float, state, border_color);
-   trace_dump_member(float, state, max_anisotropy);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_surface(const struct pipe_surface *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_surface");
-
-   trace_dump_reference(&state->reference);
-
-   trace_dump_member(format, state, format);
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
-
-   trace_dump_member(uint, state, layout);
-   trace_dump_member(uint, state, offset);
-   trace_dump_member(uint, state, usage);
-
-   trace_dump_member(ptr, state, texture);
-   trace_dump_member(uint, state, face);
-   trace_dump_member(uint, state, level);
-   trace_dump_member(uint, state, zslice);
-
-   trace_dump_struct_end();
-}
-
+#include "tgsi/tgsi_parse.h"
 
-void trace_dump_transfer(const struct pipe_transfer *state)
+struct trace_shader * trace_shader_create(struct trace_context *tr_ctx,
+                                          const struct pipe_shader_state *state,
+                                          void *result,
+                                          enum trace_shader_type type)
 {
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
+   struct trace_shader *tr_shdr = CALLOC_STRUCT(trace_shader);
 
-   trace_dump_struct_begin("pipe_transfer");
+   tr_shdr->state = result;
+   tr_shdr->type = type;
+   tr_shdr->tokens = tgsi_dup_tokens(state->tokens);
 
-   trace_dump_member(format, state, format);
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
+   /* works on context as well */
+   trace_screen_add_to_list(tr_ctx, shaders, tr_shdr);
 
-   trace_dump_member_begin("block");
-   trace_dump_block(&state->block);
-   trace_dump_member_end();
-
-   trace_dump_member(uint, state, nblocksx);
-   trace_dump_member(uint, state, nblocksy);
-   trace_dump_member(uint, state, stride);
-   trace_dump_member(uint, state, usage);
-
-   trace_dump_member(ptr, state, texture);
-   trace_dump_member(uint, state, face);
-   trace_dump_member(uint, state, level);
-   trace_dump_member(uint, state, zslice);
-
-   trace_dump_struct_end();
+   return tr_shdr;
 }
 
-
-void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
+void trace_shader_destroy(struct trace_context *tr_ctx,
+                          struct trace_shader *tr_shdr)
 {
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
+   trace_screen_remove_from_list(tr_ctx, shaders, tr_shdr);
 
-   trace_dump_struct_begin("pipe_vertex_buffer");
-
-   trace_dump_member(uint, state, stride);
-   trace_dump_member(uint, state, max_index);
-   trace_dump_member(uint, state, buffer_offset);
-   trace_dump_member(buffer_ptr, state, buffer);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_vertex_element(const struct pipe_vertex_element *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
+   if (tr_shdr->replaced) {
+      if (tr_shdr->type == TRACE_SHADER_FRAGMENT)
+         tr_ctx->pipe->delete_fs_state(tr_ctx->pipe, tr_shdr->replaced);
+      else if (tr_shdr->type == TRACE_SHADER_VERTEX)
+         tr_ctx->pipe->delete_vs_state(tr_ctx->pipe, tr_shdr->replaced);
+      else
+         assert(0);
    }
 
-   trace_dump_struct_begin("pipe_vertex_element");
-
-   trace_dump_member(uint, state, src_offset);
-
-   trace_dump_member(uint, state, vertex_buffer_index);
-   trace_dump_member(uint, state, nr_components);
-
-   trace_dump_member(format, state, src_format);
-
-   trace_dump_struct_end();
+   FREE(tr_shdr->replaced_tokens);
+   FREE(tr_shdr->tokens);
+   FREE(tr_shdr);
 }
diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h
index 513ed0ac98..1c16042ee5 100644
--- a/src/gallium/drivers/trace/tr_state.h
+++ b/src/gallium/drivers/trace/tr_state.h
@@ -1,78 +1,68 @@
-/**************************************************************************
- *
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+/*
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
  *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
  *
- **************************************************************************/
-
-#ifndef TR_STATE_H
-#define TR_STATE_H
-
-#include "pipe/p_format.h"
-#include "pipe/p_state.h"
-#include "pipe/p_shader_tokens.h"
-
-
-void trace_dump_format(enum pipe_format format);
-
-void trace_dump_block(const struct pipe_format_block *block);
-
-void trace_dump_template(const struct pipe_texture *templat);
-
-
-void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state);
-
-void trace_dump_poly_stipple(const struct pipe_poly_stipple *state);
-
-void trace_dump_viewport_state(const struct pipe_viewport_state *state);
-
-void trace_dump_scissor_state(const struct pipe_scissor_state *state);
-
-void trace_dump_clip_state(const struct pipe_clip_state *state);
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
 
-void trace_dump_constant_buffer(const struct pipe_constant_buffer *state);
+#ifndef TR_STATE_H_
+#define TR_STATE_H_
 
-void trace_dump_token(const struct tgsi_token *token);
+#include "tr_context.h"
 
-void trace_dump_shader_state(const struct pipe_shader_state *state);
+struct tgsi_token;
 
-void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state);
+enum trace_shader_type {
+   TRACE_SHADER_FRAGMENT = 0,
+   TRACE_SHADER_VERTEX   = 1,
+   TRACE_SHADER_GEOMETRY = 2,
+};
 
-void trace_dump_blend_state(const struct pipe_blend_state *state);
+struct trace_shader
+{
+   struct tr_list list;
 
-void trace_dump_blend_color(const struct pipe_blend_color *state);
+   enum trace_shader_type type;
 
-void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state);
+   void *state;
+   void *replaced;
 
-void trace_dump_sampler_state(const struct pipe_sampler_state *state);
+   struct tgsi_token *tokens;
+   struct tgsi_token *replaced_tokens;
 
-void trace_dump_surface(const struct pipe_surface *state);
+   boolean disabled;
+};
 
-void trace_dump_transfer(const struct pipe_transfer *state);
 
-void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
+static INLINE struct trace_shader *
+trace_shader(void *state)
+{
+   return (struct trace_shader *)state;
+}
 
-void trace_dump_vertex_element(const struct pipe_vertex_element *state);
+struct trace_shader * trace_shader_create(struct trace_context *tr_ctx,
+                                          const struct pipe_shader_state *state,
+                                          void *result,
+                                          enum trace_shader_type type);
 
+void trace_shader_destroy(struct trace_context *tr_ctx,
+                          struct trace_shader *tr_shdr);
 
-#endif /* TR_STATE_H */
+#endif
diff --git a/src/gallium/include/pipe/internal/p_winsys_screen.h b/src/gallium/include/pipe/internal/p_winsys_screen.h
index f4a29e63c7..a1542dada7 100644
--- a/src/gallium/include/pipe/internal/p_winsys_screen.h
+++ b/src/gallium/include/pipe/internal/p_winsys_screen.h
@@ -140,6 +140,7 @@ struct pipe_winsys
 						unsigned width, unsigned height,
 						enum pipe_format format,
 						unsigned usage,
+						unsigned tex_usage,
 						unsigned *stride);
 
 
diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h
index e6a67f8c2f..c13cffceb0 100644
--- a/src/gallium/include/pipe/p_compiler.h
+++ b/src/gallium/include/pipe/p_compiler.h
@@ -93,9 +93,11 @@ typedef int     _Bool;
 #endif
 
 
+#ifndef __HAIKU__
 typedef unsigned int       uint;
-typedef unsigned char      ubyte;
 typedef unsigned short     ushort;
+#endif
+typedef unsigned char      ubyte;
 
 #if 0
 #define boolean bool
diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h
index 63238ea46e..de99957d9d 100644
--- a/src/gallium/include/pipe/p_config.h
+++ b/src/gallium/include/pipe/p_config.h
@@ -102,6 +102,19 @@
 
 
 /*
+ * Endian detection.
+ */
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+#define PIPE_ARCH_LITTLE_ENDIAN
+#elif defined(PIPE_ARCH_PPC) || defined(PIPE_ARCH_PPC_64)
+#define PIPE_ARCH_BIG_ENDIAN
+#else
+#define PIPE_ARCH_UNKNOWN_ENDIAN
+#endif
+
+
+/*
  * Operating system family.
  * 
  * See subsystem below for a more fine-grained distinction.
@@ -119,10 +132,17 @@
 #define PIPE_OS_SOLARIS
 #endif
 
+#if defined(__APPLE__)
+#define PIPE_OS_APPLE
+#endif
+
 #if defined(_WIN32) || defined(WIN32)
 #define PIPE_OS_WINDOWS
 #endif
 
+#if defined(__HAIKU__)
+#define PIPE_OS_HAIKU
+#endif
 
 /*
  * Subsystem.
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index 57e966ac3b..39620a7198 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -98,7 +98,7 @@ struct pipe_context {
     */
    /*@{*/
    struct pipe_query *(*create_query)( struct pipe_context *pipe,
-                                              unsigned query_type );
+                                       unsigned query_type );
 
    void (*destroy_query)(struct pipe_context *pipe,
                          struct pipe_query *q);
@@ -215,10 +215,12 @@ struct pipe_context {
 
    /**
     * Clear the specified set of currently bound buffers to specified values.
+    * The entire buffers are cleared (no scissor, no colormask, etc).
     *
-    * buffers is a bitfield of PIPE_CLEAR_* values.
-    *
-    * rgba is a pointer to an array of one float for each of r, g, b, a.
+    * \param buffers  bitfield of PIPE_CLEAR_* values.
+    * \param rgba  pointer to an array of one float for each of r, g, b, a.
+    * \param depth  depth clear value in [0,1].
+    * \param stencil  stencil clear value
     */
    void (*clear)(struct pipe_context *pipe,
                  unsigned buffers,
@@ -226,7 +228,9 @@ struct pipe_context {
                  double depth,
 		 unsigned stencil);
 
-   /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */
+   /** Flush rendering
+    * \param flags  bitmask of PIPE_FLUSH_x tokens)
+    */
    void (*flush)( struct pipe_context *pipe,
                   unsigned flags,
                   struct pipe_fence_handle **fence );
@@ -242,10 +246,10 @@ struct pipe_context {
     * \param texture  texture to check.
     * \param face  cubemap face. Use 0 for non-cubemap texture.
     */
-
    unsigned int (*is_texture_referenced) (struct pipe_context *pipe,
 					  struct pipe_texture *texture,
 					  unsigned face, unsigned level);
+
    /**
     * Check whether a buffer is referenced by an unflushed hw command.
     * The state-tracker uses this function to optimize away unnecessary
@@ -255,7 +259,6 @@ struct pipe_context {
     *              checked.
     * \param buf  Buffer to check.
     */
-
    unsigned int (*is_buffer_referenced) (struct pipe_context *pipe,
 					 struct pipe_buffer *buf);
 };
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 5c16884c16..b01ab6d137 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -132,6 +132,7 @@ enum pipe_texture_target {
 #define PIPE_TEX_FACE_NEG_Y 3
 #define PIPE_TEX_FACE_POS_Z 4
 #define PIPE_TEX_FACE_NEG_Z 5
+#define PIPE_TEX_FACE_MAX   6
 
 #define PIPE_TEX_WRAP_REPEAT                   0
 #define PIPE_TEX_WRAP_CLAMP                    1
@@ -158,14 +159,6 @@ enum pipe_texture_target {
 #define PIPE_TEX_COMPARE_NONE          0
 #define PIPE_TEX_COMPARE_R_TO_TEXTURE  1
 
-#define PIPE_TEX_FACE_POS_X   0
-#define PIPE_TEX_FACE_NEG_X   1
-#define PIPE_TEX_FACE_POS_Y   2
-#define PIPE_TEX_FACE_NEG_Y   3
-#define PIPE_TEX_FACE_POS_Z   4
-#define PIPE_TEX_FACE_NEG_Z   5
-#define PIPE_TEX_FACE_MAX     6
-
 #define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
 #define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
@@ -198,9 +191,9 @@ enum pipe_texture_target {
  * Transfer object usage flags
  */
 enum pipe_transfer_usage {
-   PIPE_TRANSFER_READ,
-   PIPE_TRANSFER_WRITE,
-   PIPE_TRANSFER_READ_WRITE  /**< Read/modify/write */
+   PIPE_TRANSFER_READ = (1 << 0),
+   PIPE_TRANSFER_WRITE = (1 << 1),
+   PIPE_TRANSFER_READ_WRITE = PIPE_TRANSFER_READ | PIPE_TRANSFER_WRITE /**< Read/modify/write */
 };
 
 
@@ -281,9 +274,8 @@ enum pipe_transfer_usage {
 
 
 /**
- * Implementation capabilities/limits
- * Passed to pipe->get_param()
- * XXX this will need some fine tuning...
+ * Implementation capabilities/limits which are queried through
+ * pipe_screen::get_param() and pipe_screen::get_paramf().
  */
 #define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1
 #define PIPE_CAP_NPOT_TEXTURES           2
@@ -312,6 +304,7 @@ enum pipe_transfer_usage {
 #define PIPE_CAP_TEXTURE_MIRROR_REPEAT   25
 #define PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS 26
 #define PIPE_CAP_TGSI_CONT_SUPPORTED     27
+#define PIPE_CAP_BLEND_EQUATION_SEPARATE 28
 
 
 /**
diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h
index 6cbdd75943..3f30c52a16 100644
--- a/src/gallium/include/pipe/p_screen.h
+++ b/src/gallium/include/pipe/p_screen.h
@@ -194,6 +194,7 @@ struct pipe_screen {
 						unsigned width, unsigned height,
 						enum pipe_format format,
 						unsigned usage,
+						unsigned tex_usage,
 						unsigned *stride);
 
 
diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
index b00cfe3423..5fa6c9af30 100644
--- a/src/gallium/include/pipe/p_shader_tokens.h
+++ b/src/gallium/include/pipe/p_shader_tokens.h
@@ -78,6 +78,7 @@ enum tgsi_file_type {
    TGSI_FILE_SAMPLER     =5,
    TGSI_FILE_ADDRESS     =6,
    TGSI_FILE_IMMEDIATE   =7,
+   TGSI_FILE_LOOP        =8,
    TGSI_FILE_COUNT      /**< how many TGSI_FILE_ types */
 };
 
@@ -152,13 +153,16 @@ struct tgsi_immediate
    unsigned Extended   : 1;  /**< BOOL */
 };
 
-struct tgsi_immediate_float32
+union tgsi_immediate_data
 {
    float Float;
 };
 
-/*
- * GL_NV_vertex_program
+/* TGSI opcodes.  
+ * 
+ * For more information on semantics of opcodes and
+ * which APIs are known to use which opcodes, see
+ * auxiliary/tgsi/tgsi-instruction-set.txt
  */
 #define TGSI_OPCODE_ARL                 0
 #define TGSI_OPCODE_MOV                 1
@@ -177,62 +181,32 @@ struct tgsi_immediate_float32
 #define TGSI_OPCODE_SLT                 14
 #define TGSI_OPCODE_SGE                 15
 #define TGSI_OPCODE_MAD                 16
-
-/*
- * GL_ATI_fragment_shader
- */
 #define TGSI_OPCODE_SUB                 17
-#define TGSI_OPCODE_DOT3                TGSI_OPCODE_DP3
-#define TGSI_OPCODE_DOT4                TGSI_OPCODE_DP4
-#define TGSI_OPCODE_LERP                18
+#define TGSI_OPCODE_LRP                 18
 #define TGSI_OPCODE_CND                 19
-#define TGSI_OPCODE_CND0                20
-#define TGSI_OPCODE_DOT2ADD             21
-
-/*
- * GL_EXT_vertex_shader
- */
-#define TGSI_OPCODE_INDEX               22   /* considered for removal */
-#define TGSI_OPCODE_NEGATE              23   /* considered for removal */
-#define TGSI_OPCODE_MADD                TGSI_OPCODE_MAD
-#define TGSI_OPCODE_FRAC                24
-#define TGSI_OPCODE_SETGE               TGSI_OPCODE_SGE
-#define TGSI_OPCODE_SETLT               TGSI_OPCODE_SLT
+                                /* gap */
+#define TGSI_OPCODE_DP2A                21
+                                /* gap */
+#define TGSI_OPCODE_FRC                 24
 #define TGSI_OPCODE_CLAMP               25
-#define TGSI_OPCODE_FLOOR               26
+#define TGSI_OPCODE_FLR                 26
 #define TGSI_OPCODE_ROUND               27
-#define TGSI_OPCODE_EXPBASE2            28
-#define TGSI_OPCODE_LOGBASE2            29
-#define TGSI_OPCODE_POWER               30
-#define TGSI_OPCODE_RECIP               TGSI_OPCODE_RCP
-#define TGSI_OPCODE_RECIPSQRT           TGSI_OPCODE_RSQ
-#define TGSI_OPCODE_CROSSPRODUCT        31
-#define TGSI_OPCODE_MULTIPLYMATRIX      32   /* considered for removal */
-
-/*
- * GL_NV_vertex_program1_1
- */
+#define TGSI_OPCODE_EX2                 28
+#define TGSI_OPCODE_LG2                 29
+#define TGSI_OPCODE_POW                 30
+#define TGSI_OPCODE_XPD                 31
+                                /* gap */
 #define TGSI_OPCODE_ABS                 33
 #define TGSI_OPCODE_RCC                 34
 #define TGSI_OPCODE_DPH                 35
-
-/*
- * GL_NV_fragment_program
- */
 #define TGSI_OPCODE_COS                 36
 #define TGSI_OPCODE_DDX                 37
 #define TGSI_OPCODE_DDY                 38
-#define TGSI_OPCODE_EX2                 TGSI_OPCODE_EXPBASE2
-#define TGSI_OPCODE_FLR                 TGSI_OPCODE_FLOOR
-#define TGSI_OPCODE_FRC                 TGSI_OPCODE_FRAC
 #define TGSI_OPCODE_KILP                39  /* predicated kill */
-#define TGSI_OPCODE_LG2                 TGSI_OPCODE_LOGBASE2
-#define TGSI_OPCODE_LRP                 TGSI_OPCODE_LERP
 #define TGSI_OPCODE_PK2H                40
 #define TGSI_OPCODE_PK2US               41
 #define TGSI_OPCODE_PK4B                42
 #define TGSI_OPCODE_PK4UB               43
-#define TGSI_OPCODE_POW                 TGSI_OPCODE_POWER
 #define TGSI_OPCODE_RFL                 44
 #define TGSI_OPCODE_SEQ                 45
 #define TGSI_OPCODE_SFL                 46
@@ -249,66 +223,29 @@ struct tgsi_immediate_float32
 #define TGSI_OPCODE_UP4B                57
 #define TGSI_OPCODE_UP4UB               58
 #define TGSI_OPCODE_X2D                 59
-
-/*
- * GL_NV_vertex_program2
- */
 #define TGSI_OPCODE_ARA                 60
 #define TGSI_OPCODE_ARR                 61
 #define TGSI_OPCODE_BRA                 62
 #define TGSI_OPCODE_CAL                 63
 #define TGSI_OPCODE_RET                 64
-#define TGSI_OPCODE_SSG                 65
-
-/*
- * GL_ARB_vertex_program
- */
-#define TGSI_OPCODE_SWZ                 118
-#define TGSI_OPCODE_XPD                 TGSI_OPCODE_CROSSPRODUCT
-
-/*
- * GL_ARB_fragment_program
- */
+#define TGSI_OPCODE_SSG                 65 /* SGN */
 #define TGSI_OPCODE_CMP                 66
-#define TGSI_OPCODE_KIL                 116  /* conditional kill */
 #define TGSI_OPCODE_SCS                 67
 #define TGSI_OPCODE_TXB                 68
-
-/*
- * GL_NV_fragment_program_option
- */
-/* No new opcode */
-
-/*
- * GL_NV_fragment_program2
- */
 #define TGSI_OPCODE_NRM                 69
 #define TGSI_OPCODE_DIV                 70
 #define TGSI_OPCODE_DP2                 71
-#define TGSI_OPCODE_DP2A                TGSI_OPCODE_DOT2ADD
 #define TGSI_OPCODE_TXL                 72
 #define TGSI_OPCODE_BRK                 73
 #define TGSI_OPCODE_IF                  74
-#define TGSI_OPCODE_LOOP                75
+#define TGSI_OPCODE_BGNFOR              75
 #define TGSI_OPCODE_REP                 76
 #define TGSI_OPCODE_ELSE                77
 #define TGSI_OPCODE_ENDIF               78
-#define TGSI_OPCODE_ENDLOOP             79
+#define TGSI_OPCODE_ENDFOR              79
 #define TGSI_OPCODE_ENDREP              80
-
-/*
- * GL_NV_vertex_program2_option
- */
-
-/*
- * GL_NV_vertex_program3
- */
 #define TGSI_OPCODE_PUSHA               81
 #define TGSI_OPCODE_POPA                82
-
-/*
- * GL_NV_gpu_program4
- */
 #define TGSI_OPCODE_CEIL                83
 #define TGSI_OPCODE_I2F                 84
 #define TGSI_OPCODE_NOT                 85
@@ -323,103 +260,25 @@ struct tgsi_immediate_float32
 #define TGSI_OPCODE_TXF                 94
 #define TGSI_OPCODE_TXQ                 95
 #define TGSI_OPCODE_CONT                96
-
-/*
- * GL_NV_vertex_program4
- */
-/* Same as GL_NV_gpu_program4 */
-
-/*
- * GL_NV_fragment_program4
- */
-/* Same as GL_NV_gpu_program4 */
-
-/*
- * GL_NV_geometry_program4
- */
-/* Same as GL_NV_gpu_program4 */
 #define TGSI_OPCODE_EMIT                97
 #define TGSI_OPCODE_ENDPRIM             98
-
-/*
- * GLSL
- */
-#define TGSI_OPCODE_BGNLOOP2            99
+#define TGSI_OPCODE_BGNLOOP             99
 #define TGSI_OPCODE_BGNSUB              100
-#define TGSI_OPCODE_ENDLOOP2            101
+#define TGSI_OPCODE_ENDLOOP             101
 #define TGSI_OPCODE_ENDSUB              102
-#define TGSI_OPCODE_INT                 TGSI_OPCODE_TRUNC
 #define TGSI_OPCODE_NOISE1              103
 #define TGSI_OPCODE_NOISE2              104
 #define TGSI_OPCODE_NOISE3              105
 #define TGSI_OPCODE_NOISE4              106
 #define TGSI_OPCODE_NOP                 107
-
-/*
- * ps_1_1
- */
-#define TGSI_OPCODE_TEXKILL             TGSI_OPCODE_KIL
-
-/*
- * ps_1_2
- */
-/* CMP - use TGSI_OPCODE_CND0 */
-
-/*
- * ps_1_3
- */
-/* CMP - use TGSI_OPCODE_CND0 */
-
-/*
- * ps_1_4
- */
-#define TGSI_OPCODE_TEXLD               TGSI_OPCODE_TEX
-
-/*
- * ps_2_0
- */
-#define TGSI_OPCODE_M4X4                TGSI_OPCODE_MULTIPLYMATRIX
-#define TGSI_OPCODE_M4X3                108
-#define TGSI_OPCODE_M3X4                109
-#define TGSI_OPCODE_M3X3                110
-#define TGSI_OPCODE_M3X2                111
-#define TGSI_OPCODE_CRS                 TGSI_OPCODE_XPD
+                                /* gap */
 #define TGSI_OPCODE_NRM4                112
-#define TGSI_OPCODE_SINCOS              TGSI_OPCODE_SCS
-#define TGSI_OPCODE_TEXLDB              TGSI_OPCODE_TXB
-#define TGSI_OPCODE_DP2ADD              TGSI_OPCODE_DP2A
-
-/*
- * ps_2_x
- */
-#define TGSI_OPCODE_CALL                TGSI_OPCODE_CAL
 #define TGSI_OPCODE_CALLNZ              113
 #define TGSI_OPCODE_IFC                 114
-#define TGSI_OPCODE_BREAK               TGSI_OPCODE_BRK
 #define TGSI_OPCODE_BREAKC              115
-#define TGSI_OPCODE_DSX                 TGSI_OPCODE_DDX
-#define TGSI_OPCODE_DSY                 TGSI_OPCODE_DDY
-#define TGSI_OPCODE_TEXLDD              TGSI_OPCODE_TXD
-
-/*
- * vs_1_1
- */
-#define TGSI_OPCODE_EXPP                TGSI_OPCODE_EXP
-#define TGSI_OPCODE_LOGP                TGSI_OPCODE_LG2
-
-/*
- * vs_2_0
- */
-#define TGSI_OPCODE_SGN                 TGSI_OPCODE_SSG
-#define TGSI_OPCODE_MOVA                TGSI_OPCODE_ARR
-/* EXPP - use TGSI_OPCODE_EX2 */
-
-/*
- * vs_2_x
- */
-
+#define TGSI_OPCODE_KIL                 116  /* conditional kill */
 #define TGSI_OPCODE_END                 117  /* aka HALT */
-
+#define TGSI_OPCODE_SWZ                 118
 #define TGSI_OPCODE_LAST                119
 
 #define TGSI_SAT_NONE            0  /* do not saturate */
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 4b590bdc90..626bedb35a 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -75,11 +75,11 @@ struct pipe_surface;
  */
 struct pipe_buffer
 {
-   struct pipe_reference reference;
-   struct pipe_screen *screen;
-   unsigned alignment;
-   unsigned usage;
-   unsigned size;
+   struct pipe_reference  reference;
+   unsigned               size;
+   struct pipe_screen    *screen;
+   unsigned               alignment;
+   unsigned               usage;
 };
 
 
@@ -265,7 +265,6 @@ struct pipe_sampler_state
    unsigned compare_func:3;      /**< PIPE_FUNC_x */
    unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */
    unsigned prefilter:4;         /**< Wierd sampling state exposed by some api's */
-   float shadow_ambient;         /**< shadow test fail color/intensity */
    float lod_bias;               /**< LOD/lambda bias */
    float min_lod, max_lod;       /**< LOD clamp range, after bias */
    float border_color[4];
@@ -287,10 +286,10 @@ struct pipe_surface
    unsigned offset;              /**< offset from start of buffer, in bytes */
    unsigned usage;               /**< PIPE_BUFFER_USAGE_*  */
 
+   unsigned zslice;
    struct pipe_texture *texture; /**< texture into which this is a view  */
    unsigned face;
    unsigned level;
-   unsigned zslice;
 };
 
 
diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h
index ce8d79549a..25e4148232 100644
--- a/src/gallium/include/pipe/p_thread.h
+++ b/src/gallium/include/pipe/p_thread.h
@@ -39,11 +39,13 @@
 #include "util/u_debug.h" /* for assert */
 
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
 
 #include <pthread.h> /* POSIX threads headers */
 #include <stdio.h> /* for perror() */
 
+#define PIPE_THREAD_HAVE_CONDVAR
+
 typedef pthread_t pipe_thread;
 
 #define PIPE_THREAD_ROUTINE( name, param ) \
@@ -211,7 +213,7 @@ typedef unsigned pipe_condvar;
  */
 
 typedef struct {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    pthread_key_t key;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    DWORD key;
@@ -226,7 +228,7 @@ typedef struct {
 static INLINE void
 pipe_tsd_init(pipe_tsd *tsd)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    if (pthread_key_create(&tsd->key, NULL/*free*/) != 0) {
       perror("pthread_key_create(): failed to allocate key for thread specific data");
       exit(-1);
@@ -243,7 +245,7 @@ pipe_tsd_get(pipe_tsd *tsd)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    return pthread_getspecific(tsd->key);
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    assert(0);
@@ -260,7 +262,7 @@ pipe_tsd_set(pipe_tsd *tsd, void *value)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
    if (pthread_setspecific(tsd->key, value) != 0) {
       perror("pthread_set_specific() failed");
       exit(-1);
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
index 5790b2f6c7..4d1259e1ee 100644
--- a/src/gallium/include/state_tracker/drm_api.h
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -32,33 +32,38 @@ struct drm_api
 	 * Special buffer functions
 	 */
 	/*@{*/
-	struct pipe_screen*  (*create_screen)(int drm_fd,
-					      struct drm_create_screen_arg *arg);
-	struct pipe_context* (*create_context)(struct pipe_screen *screen);
+	struct pipe_screen*  (*create_screen)(struct drm_api *api, int drm_fd,
+	                                      struct drm_create_screen_arg *arg);
+	struct pipe_context* (*create_context)(struct drm_api *api,
+	                                       struct pipe_screen *screen);
 	/*@}*/
 
 	/**
 	 * Special buffer functions
 	 */
 	/*@{*/
-	boolean (*buffer_from_texture)(struct pipe_texture *texture,
-                                       struct pipe_buffer **buffer,
-                                       unsigned *stride);
-	struct pipe_buffer* (*buffer_from_handle)(struct pipe_screen *screen,
-                                                  const char *name,
-                                                  unsigned handle);
-	boolean (*handle_from_buffer)(struct pipe_screen *screen,
-                                      struct pipe_buffer *buffer,
-                                      unsigned *handle);
-	boolean (*global_handle_from_buffer)(struct pipe_screen *screen,
-                                             struct pipe_buffer *buffer,
-                                             unsigned *handle);
+	struct pipe_texture*
+	    (*texture_from_shared_handle)(struct drm_api *api,
+	                                  struct pipe_screen *screen,
+	                                  struct pipe_texture *templ,
+	                                  const char *name,
+	                                  unsigned stride,
+	                                  unsigned handle);
+	boolean (*shared_handle_from_texture)(struct drm_api *api,
+	                                      struct pipe_screen *screen,
+	                                      struct pipe_texture *texture,
+	                                      unsigned *stride,
+	                                      unsigned *handle);
+	boolean (*local_handle_from_texture)(struct drm_api *api,
+	                                     struct pipe_screen *screen,
+	                                     struct pipe_texture *texture,
+	                                     unsigned *stride,
+	                                     unsigned *handle);
 	/*@}*/
+
+	void (*destroy)(struct drm_api *api);
 };
 
-/**
- * A driver needs to export this symbol
- */
-extern struct drm_api drm_api_hooks;
+extern struct drm_api * drm_api_create(void);
 
 #endif
diff --git a/src/gallium/state_trackers/dri/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c
index 45eaec4ed3..8819936fca 100644
--- a/src/gallium/state_trackers/dri/dri_context.c
+++ b/src/gallium/state_trackers/dri/dri_context.c
@@ -69,7 +69,7 @@ dri_create_context(const __GLcontextModes * visual,
    driParseConfigFiles(&ctx->optionCache,
 		       &screen->optionCache, sPriv->myNum, "dri");
 
-   ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen);
+   ctx->pipe = screen->api->create_context(screen->api, screen->pipe_screen);
 
    if (ctx->pipe == NULL)
       goto fail;
@@ -100,7 +100,6 @@ void
 dri_destroy_context(__DRIcontextPrivate * cPriv)
 {
    struct dri_context *ctx = dri_context(cPriv);
-   struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
 
    /* No particular reason to wait for command completion before
     * destroying a context, but it is probably worthwhile flushing it
@@ -109,9 +108,6 @@ dri_destroy_context(__DRIcontextPrivate * cPriv)
     */
    st_flush(ctx->st, 0, NULL);
 
-   if (screen->dummyContext == ctx)
-      screen->dummyContext = NULL;
-
    /* Also frees ctx->pipe?
     */
    st_destroy_context(ctx->st);
@@ -143,7 +139,6 @@ dri_make_current(__DRIcontextPrivate * cPriv,
 {
    if (cPriv) {
       struct dri_context *ctx = dri_context(cPriv);
-      struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
       struct dri_drawable *draw = dri_drawable(driDrawPriv);
       struct dri_drawable *read = dri_drawable(driReadPriv);
       struct st_context *old_st = st_get_current();
@@ -153,11 +148,6 @@ dri_make_current(__DRIcontextPrivate * cPriv,
 
       ++ctx->bind_count;
 
-      /* This is for situations in which we need a rendering context but
-       * there may not be any currently bound.
-       */
-      screen->dummyContext = ctx;
-
       if (ctx->dPriv != driDrawPriv) {
 	 ctx->dPriv = driDrawPriv;
 	 ctx->d_stamp = driDrawPriv->lastStamp - 1;
diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c
index 5846390d02..5cec9e329d 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.c
+++ b/src/gallium/state_trackers/dri/dri_drawable.c
@@ -36,6 +36,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
 #include "pipe/p_inlines.h"
+#include "main/mtypes.h"
+#include "main/renderbuffer.h"
 #include "state_tracker/drm_api.h"
 #include "state_tracker/dri1_api.h"
 #include "state_tracker/st_public.h"
@@ -44,16 +46,9 @@
 
 #include "util/u_memory.h"
 
-static void
-dri_copy_to_front(__DRIdrawablePrivate * dPriv,
-		  struct pipe_surface *from,
-		  int x, int y, unsigned w, unsigned h)
-{
-   /* TODO send a message to the Xserver to copy to the real front buffer */
-}
-
 static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
+dri_surface_from_handle(struct drm_api *api,
+			struct pipe_screen *screen,
 			unsigned handle,
 			enum pipe_format format,
 			unsigned width, unsigned height, unsigned pitch)
@@ -61,11 +56,6 @@ dri_surface_from_handle(struct pipe_screen *screen,
    struct pipe_surface *surface = NULL;
    struct pipe_texture *texture = NULL;
    struct pipe_texture templat;
-   struct pipe_buffer *buf = NULL;
-
-   buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle);
-   if (!buf)
-      return NULL;
 
    memset(&templat, 0, sizeof(templat));
    templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
@@ -77,13 +67,13 @@ dri_surface_from_handle(struct pipe_screen *screen,
    templat.height[0] = height;
    pf_get_block(templat.format, &templat.block);
 
-   texture = screen->texture_blanket(screen, &templat, &pitch, buf);
-
-   /* we don't need the buffer from this point on */
-   pipe_buffer_reference(&buf, NULL);
+   texture = api->texture_from_shared_handle(api, screen, &templat,
+                                             "dri2 buffer", pitch, handle);
 
-   if (!texture)
+   if (!texture) {
+      debug_printf("%s: Failed to blanket the buffer with a texture\n", __func__);
       return NULL;
+   }
 
    surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
 				     PIPE_BUFFER_USAGE_GPU_READ |
@@ -95,17 +85,48 @@ dri_surface_from_handle(struct pipe_screen *screen,
 }
 
 /**
+ * Pixmaps have will have the same name of fake front and front.
+ */
+static boolean
+dri2_check_if_pixmap(__DRIbuffer *buffers, int count)
+{
+   boolean found = FALSE;
+   boolean is_pixmap = FALSE;
+   unsigned name;
+   int i;
+
+   for (i = 0; i < count; i++) {
+      switch (buffers[i].attachment) {
+      case __DRI_BUFFER_FRONT_LEFT:
+      case __DRI_BUFFER_FAKE_FRONT_LEFT:
+         if (found) {
+            is_pixmap = buffers[i].name == name;
+         } else {
+            name = buffers[i].name;
+            found = TRUE;
+         }
+      default:
+	 continue;
+      }
+   }
+
+   return is_pixmap;
+}
+
+/**
  * This will be called a drawable is known to have been resized.
  */
 void
 dri_get_buffers(__DRIdrawablePrivate * dPriv)
 {
+
    struct dri_drawable *drawable = dri_drawable(dPriv);
    struct pipe_surface *surface = NULL;
    struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
    __DRIbuffer *buffers = NULL;
    __DRIscreen *dri_screen = drawable->sPriv;
    __DRIdrawable *dri_drawable = drawable->dPriv;
+   struct drm_api *api = ((struct dri_screen*)(dri_screen->private))->api;
    boolean have_depth = FALSE;
    int i, count;
 
@@ -138,36 +159,45 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
    dri_drawable->pBackClipRects[0].x2 = dri_drawable->w;
    dri_drawable->pBackClipRects[0].y2 = dri_drawable->h;
 
+   if (drawable->old_num == count &&
+       drawable->old_w == dri_drawable->w &&
+       drawable->old_h == dri_drawable->h &&
+       memcmp(drawable->old, buffers, sizeof(__DRIbuffer) * count) == 0) {
+       return;
+   } else {
+      drawable->old_num = count;
+      drawable->old_w = dri_drawable->w;
+      drawable->old_h = dri_drawable->h;
+      memcpy(drawable->old, buffers, sizeof(__DRIbuffer) * count);
+   }
+
+   drawable->is_pixmap = dri2_check_if_pixmap(buffers, count);
+
    for (i = 0; i < count; i++) {
       enum pipe_format format = 0;
       int index = 0;
 
       switch (buffers[i].attachment) {
       case __DRI_BUFFER_FRONT_LEFT:
-	 index = ST_SURFACE_FRONT_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
-	 break;
+	 continue;
       case __DRI_BUFFER_FAKE_FRONT_LEFT:
 	 index = ST_SURFACE_FRONT_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 format = drawable->color_format;
 	 break;
       case __DRI_BUFFER_BACK_LEFT:
 	 index = ST_SURFACE_BACK_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 format = drawable->color_format;
 	 break;
       case __DRI_BUFFER_DEPTH:
-	 index = ST_SURFACE_DEPTH;
-	 format = PIPE_FORMAT_Z24S8_UNORM;
-	 break;
+      case __DRI_BUFFER_DEPTH_STENCIL:
       case __DRI_BUFFER_STENCIL:
 	 index = ST_SURFACE_DEPTH;
-	 format = PIPE_FORMAT_Z24S8_UNORM;
+	 format = drawable->depth_stencil_format;
 	 break;
       case __DRI_BUFFER_ACCUM:
       default:
 	 assert(0);
       }
-      assert(buffers[i].cpp == 4);
 
       if (index == ST_SURFACE_DEPTH) {
 	 if (have_depth)
@@ -176,12 +206,25 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
 	    have_depth = TRUE;
       }
 
-      surface = dri_surface_from_handle(screen,
+      surface = dri_surface_from_handle(api,
+					screen,
 					buffers[i].name,
 					format,
 					dri_drawable->w,
 					dri_drawable->h, buffers[i].pitch);
 
+      switch (buffers[i].attachment) {
+      case __DRI_BUFFER_FAKE_FRONT_LEFT:
+      case __DRI_BUFFER_BACK_LEFT:
+	 drawable->color_format = surface->format;
+	 break;
+      case __DRI_BUFFER_DEPTH:
+      case __DRI_BUFFER_DEPTH_STENCIL:
+      case __DRI_BUFFER_STENCIL:
+	 drawable->depth_stencil_format = surface->format;
+	 break;
+      }
+
       st_set_framebuffer_surface(drawable->stfb, index, surface);
       pipe_surface_reference(&surface, NULL);
    }
@@ -189,13 +232,60 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
    st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h);
 }
 
+/**
+ * These are used for GLX_EXT_texture_from_pixmap
+ */
+void dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target,
+                          GLint format, __DRIdrawable *dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_surface *ps;
+
+   if (!drawable->stfb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer) {
+      struct gl_renderbuffer *rb =
+         st_new_renderbuffer_fb(drawable->color_format, 0 /*XXX*/, FALSE);
+      _mesa_add_renderbuffer(&drawable->stfb->Base, BUFFER_FRONT_LEFT, rb);
+   }
+
+   dri_get_buffers(drawable->dPriv);
+   st_get_framebuffer_surface(drawable->stfb, ST_SURFACE_FRONT_LEFT, &ps);
+
+   st_bind_texture_surface(ps, target == GL_TEXTURE_2D ? ST_TEXTURE_2D :
+                           ST_TEXTURE_RECT, 0, drawable->color_format);
+}
+
+void dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target,
+                         __DRIdrawable *dPriv)
+{
+   dri2_set_tex_buffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
 void
 dri_flush_frontbuffer(struct pipe_screen *screen,
 		      struct pipe_surface *surf, void *context_private)
 {
    struct dri_context *ctx = (struct dri_context *)context_private;
+   struct dri_drawable *drawable = dri_drawable(ctx->dPriv);
+   __DRIdrawable *dri_drawable = ctx->dPriv;
+   __DRIscreen *dri_screen = ctx->sPriv;
+
+   /* XXX Does this function get called with DRI1? */
 
-   dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height);
+   if (ctx->dPriv == NULL) {
+      debug_printf("%s: no drawable bound to context\n", __func__);
+      return;
+   }
+
+#if 0
+   /* TODO if rendering to pixmaps is slow enable this code. */
+   if (drawable->is_pixmap)
+      return;
+#else
+   (void)drawable;
+#endif
+
+   (*dri_screen->dri2.loader->flushFrontBuffer)(dri_drawable,
+						dri_drawable->loaderPrivate);
 }
 
 /**
@@ -206,10 +296,8 @@ dri_create_buffer(__DRIscreenPrivate * sPriv,
 		  __DRIdrawablePrivate * dPriv,
 		  const __GLcontextModes * visual, boolean isPixmap)
 {
-   enum pipe_format colorFormat, depthFormat, stencilFormat;
    struct dri_screen *screen = sPriv->private;
    struct dri_drawable *drawable = NULL;
-   struct pipe_screen *pscreen = screen->pipe_screen;
    int i;
 
    if (isPixmap)
@@ -219,39 +307,43 @@ dri_create_buffer(__DRIscreenPrivate * sPriv,
    if (drawable == NULL)
       goto fail;
 
-   /* XXX: todo: use the pipe_screen queries to figure out which
-    * render targets are supportable.
-    */
-   assert(visual->redBits == 8);
-   assert(visual->depthBits == 24 || visual->depthBits == 0);
-   assert(visual->stencilBits == 8 || visual->stencilBits == 0);
-
-   colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-   if (visual->depthBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-				       PIPE_TEXTURE_2D,
-				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-	 depthFormat = PIPE_FORMAT_Z24S8_UNORM;
-      else
-	 depthFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      depthFormat = PIPE_FORMAT_NONE;
-
-   if (visual->stencilBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-				       PIPE_TEXTURE_2D,
-				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-	 stencilFormat = PIPE_FORMAT_Z24S8_UNORM;
+   if (visual->redBits == 8) {
+      if (visual->alphaBits == 8)
+         drawable->color_format = PIPE_FORMAT_A8R8G8B8_UNORM;
       else
-	 stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      stencilFormat = PIPE_FORMAT_NONE;
+         drawable->color_format = PIPE_FORMAT_X8R8G8B8_UNORM;
+   } else {
+      drawable->color_format = PIPE_FORMAT_R5G6B5_UNORM;
+   }
+
+   switch(visual->depthBits) {
+   default:
+   case 0:
+      drawable->depth_stencil_format = PIPE_FORMAT_NONE;
+      break;
+   case 16:
+      drawable->depth_stencil_format = PIPE_FORMAT_Z16_UNORM;
+      break;
+   case 24:
+      if (visual->stencilBits == 0) {
+	 drawable->depth_stencil_format = (screen->d_depth_bits_last) ?
+	    PIPE_FORMAT_X8Z24_UNORM:
+	    PIPE_FORMAT_Z24X8_UNORM;
+      } else {
+	 drawable->depth_stencil_format = (screen->sd_depth_bits_last) ?
+	    PIPE_FORMAT_S8Z24_UNORM:
+	    PIPE_FORMAT_Z24S8_UNORM;
+      }
+      break;
+   case 32:
+      drawable->depth_stencil_format = PIPE_FORMAT_Z32_UNORM;
+      break;
+   }
 
    drawable->stfb = st_create_framebuffer(visual,
-					  colorFormat,
-					  depthFormat,
-					  stencilFormat,
+					  drawable->color_format,
+					  drawable->depth_stencil_format,
+					  drawable->depth_stencil_format,
 					  dPriv->w,
 					  dPriv->h, (void *)drawable);
    if (drawable->stfb == NULL)
@@ -262,17 +354,19 @@ dri_create_buffer(__DRIscreenPrivate * sPriv,
    dPriv->driverPrivate = (void *)drawable;
 
    /* setup dri2 buffers information */
+   /* TODO incase of double buffer visual, delay fake creation */
    i = 0;
    drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
-#if 0
-   /* TODO incase of double buffer visual, delay fake creation */
-   drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
-#endif
+
    if (visual->doubleBufferMode)
       drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT;
-   if (visual->depthBits)
+   else
+      drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
+   if (visual->depthBits && visual->stencilBits)
+      drawable->attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL;
+   else if (visual->depthBits)
       drawable->attachments[i++] = __DRI_BUFFER_DEPTH;
-   if (visual->stencilBits)
+   else if (visual->stencilBits)
       drawable->attachments[i++] = __DRI_BUFFER_STENCIL;
    drawable->num_attachments = i;
 
diff --git a/src/gallium/state_trackers/dri/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h
index 78a66624aa..b910930db4 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.h
+++ b/src/gallium/state_trackers/dri/dri_drawable.h
@@ -33,6 +33,7 @@
 struct pipe_surface;
 struct pipe_fence_handle;
 struct st_framebuffer;
+struct dri_context;
 
 #define DRI_SWAP_FENCES_MAX  8
 #define DRI_SWAP_FENCES_MASK 7
@@ -46,6 +47,13 @@ struct dri_drawable
    unsigned attachments[8];
    unsigned num_attachments;
 
+   boolean is_pixmap;
+
+   __DRIbuffer old[8];
+   unsigned old_num;
+   unsigned old_w;
+   unsigned old_h;
+
    /* gallium */
    struct st_framebuffer *stfb;
    struct pipe_fence_handle *swap_fences[DRI_SWAP_FENCES_MAX];
@@ -53,6 +61,9 @@ struct dri_drawable
    unsigned int tail;
    unsigned int desired_fences;
    unsigned int cur_fences;
+
+   enum pipe_format color_format;
+   enum pipe_format depth_stencil_format;
 };
 
 static INLINE struct dri_drawable *
@@ -82,6 +93,12 @@ void dri_get_buffers(__DRIdrawablePrivate * dPriv);
 
 void dri_destroy_buffer(__DRIdrawablePrivate * dPriv);
 
+void dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target,
+                          GLint glx_texture_format, __DRIdrawable *dPriv);
+
+void dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target,
+                         __DRIdrawable *dPriv);
+
 void
 dri1_update_drawables(struct dri_context *ctx,
 		      struct dri_drawable *draw, struct dri_drawable *read);
diff --git a/src/gallium/state_trackers/dri/dri_extensions.c b/src/gallium/state_trackers/dri/dri_extensions.c
index 0c59d42d5c..4349a4d1d2 100644
--- a/src/gallium/state_trackers/dri/dri_extensions.c
+++ b/src/gallium/state_trackers/dri/dri_extensions.c
@@ -33,12 +33,15 @@
 #include "dri_context.h"
 #include "state_tracker/st_context.h"
 
+#define need_GL_ARB_map_buffer_range
 #define need_GL_ARB_multisample
 #define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
 #define need_GL_ARB_texture_compression
 #define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
 #define need_GL_ARB_window_pos
 #define need_GL_EXT_blend_color
 #define need_GL_EXT_blend_equation_separate
@@ -50,16 +53,24 @@
 #define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_secondary_color
 #define need_GL_NV_vertex_program
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
 #include "extension_helper.h"
 
 /**
  * Extension strings exported by the driver.
  */
 const struct dri_extension card_extensions[] = {
+   {"GL_ARB_fragment_shader", NULL},
+   {"GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions},
    {"GL_ARB_multisample", GL_ARB_multisample_functions},
    {"GL_ARB_multitexture", NULL},
    {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions},
+   {"GL_ARB_pixel_buffer_object", NULL},
    {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+   {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
+   {"GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
+   {"GL_ARB_shader_objects", GL_ARB_shader_objects_functions},
    {"GL_ARB_texture_border_clamp", NULL},
    {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
    {"GL_ARB_texture_cube_map", NULL},
@@ -69,12 +80,11 @@ const struct dri_extension card_extensions[] = {
    {"GL_ARB_texture_mirrored_repeat", NULL},
    {"GL_ARB_texture_rectangle", NULL},
    {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
-   {"GL_ARB_pixel_buffer_object", NULL},
+   {"GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions},
    {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
    {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
    {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
-   {"GL_EXT_blend_equation_separate",
-    GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
    {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
    {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
    {"GL_EXT_blend_subtract", NULL},
diff --git a/src/gallium/state_trackers/dri/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c
index d3392ee690..884b6d5011 100644
--- a/src/gallium/state_trackers/dri/dri_screen.c
+++ b/src/gallium/state_trackers/dri/dri_screen.c
@@ -57,82 +57,144 @@ PUBLIC const char __driConfigOptions[] =
 
    const uint __driNConfigOptions = 3;
 
+static const __DRItexBufferExtension dri2TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   dri2_set_tex_buffer,
+   dri2_set_tex_buffer2,
+};
+
    static const __DRIextension *dri_screen_extensions[] = {
       &driReadDrawableExtension,
       &driCopySubBufferExtension.base,
       &driSwapControlExtension.base,
       &driFrameTrackingExtension.base,
       &driMediaStreamCounterExtension.base,
+      &dri2TexBufferExtension.base,
       NULL
    };
 
 struct dri1_api *__dri1_api_hooks = NULL;
 
 static const __DRIconfig **
-dri_fill_in_modes(__DRIscreenPrivate * psp,
-		  unsigned pixel_bits, unsigned depth_bits,
-		  unsigned stencil_bits, GLboolean have_back_buffer)
+dri_fill_in_modes(struct dri_screen *screen,
+		  unsigned pixel_bits)
 {
-   __DRIconfig **configs;
-   __GLcontextModes *m;
+   __DRIconfig **configs = NULL;
    unsigned num_modes;
-   uint8_t depth_bits_array[3];
-   uint8_t stencil_bits_array[3];
+   uint8_t depth_bits_array[5];
+   uint8_t stencil_bits_array[5];
    uint8_t msaa_samples_array[1];
    unsigned depth_buffer_factor;
    unsigned back_buffer_factor;
    unsigned msaa_samples_factor;
-   GLenum fb_format;
-   GLenum fb_type;
-   int i;
+   struct pipe_screen *p_screen = screen->pipe_screen;
+   boolean pf_r5g6b5, pf_a8r8g8b8, pf_x8r8g8b8;
+   boolean pf_z16, pf_x8z24, pf_z24x8, pf_s8z24, pf_z24s8, pf_z32;
 
    static const GLenum back_buffer_modes[] = {
       GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
    };
 
-   /* TODO probe the hardware of what is supports */
    depth_bits_array[0] = 0;
-   depth_bits_array[1] = 24;
-   depth_bits_array[2] = 24;
-
-   stencil_bits_array[0] = 0;	       /* no depth or stencil */
-   stencil_bits_array[1] = 0;	       /* z24x8 */
-   stencil_bits_array[2] = 8;	       /* z24s8 */
+   stencil_bits_array[0] = 0;
+   depth_buffer_factor = 1;
+
+   pf_z16 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z16_UNORM,
+					  PIPE_TEXTURE_2D,
+					  PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_z32 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z32_UNORM,
+					  PIPE_TEXTURE_2D,
+					  PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_x8z24 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_X8Z24_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_z24x8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24X8_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_s8z24 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_S8Z24_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_z24s8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24S8_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+   pf_r5g6b5 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_R5G6B5_UNORM,
+					     PIPE_TEXTURE_2D,
+					     PIPE_TEXTURE_USAGE_RENDER_TARGET, 0);
+   pf_a8r8g8b8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_A8R8G8B8_UNORM,
+					       PIPE_TEXTURE_2D,
+					       PIPE_TEXTURE_USAGE_RENDER_TARGET, 0);
+   pf_x8r8g8b8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_X8R8G8B8_UNORM,
+					       PIPE_TEXTURE_2D,
+					       PIPE_TEXTURE_USAGE_RENDER_TARGET, 0);
+
+   if (pf_z16) {
+      depth_bits_array[depth_buffer_factor] = 16;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+   }
+   if (pf_x8z24 || pf_z24x8) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+      screen->d_depth_bits_last = pf_x8z24;
+   }
+   if (pf_s8z24 || pf_z24s8) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 8;
+      screen->sd_depth_bits_last = pf_s8z24;
+   }
+   if (pf_z32) {
+      depth_bits_array[depth_buffer_factor] = 32;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+   }
 
    msaa_samples_array[0] = 0;
-
-   depth_buffer_factor = 3;
    back_buffer_factor = 3;
    msaa_samples_factor = 1;
 
    num_modes =
       depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4;
 
-   if (pixel_bits == 16) {
-      fb_format = GL_RGB;
-      fb_type = GL_UNSIGNED_SHORT_5_6_5;
+   if (pixel_bits == 16 && pf_r5g6b5) {
+      configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5,
+				 depth_bits_array, stencil_bits_array,
+				 depth_buffer_factor, back_buffer_modes,
+				 back_buffer_factor,
+				 msaa_samples_array, 1);
    } else {
-      fb_format = GL_BGRA;
-      fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+      __DRIconfig **configs_a8r8g8b8 = NULL;
+      __DRIconfig **configs_x8r8g8b8 = NULL;
+
+      if (pf_a8r8g8b8)
+	 configs_a8r8g8b8 = driCreateConfigs(GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV,
+					     depth_bits_array,
+					     stencil_bits_array,
+					     depth_buffer_factor,
+					     back_buffer_modes,
+					     back_buffer_factor,
+					     msaa_samples_array, 1);
+      if (pf_x8r8g8b8)
+	 configs_x8r8g8b8 = driCreateConfigs(GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV,
+					     depth_bits_array,
+					     stencil_bits_array,
+					     depth_buffer_factor,
+					     back_buffer_modes,
+					     back_buffer_factor,
+					     msaa_samples_array, 1);
+
+      if (configs_a8r8g8b8 && configs_x8r8g8b8)
+	 configs = driConcatConfigs(configs_x8r8g8b8, configs_a8r8g8b8);
+      else if (configs_a8r8g8b8)
+	 configs = configs_a8r8g8b8;
+      else if (configs_x8r8g8b8)
+	 configs = configs_x8r8g8b8;
+      else
+	 configs = NULL;
    }
 
-   configs = driCreateConfigs(fb_format, fb_type,
-			      depth_bits_array,
-			      stencil_bits_array, depth_buffer_factor,
-			      back_buffer_modes, back_buffer_factor,
-			      msaa_samples_array, msaa_samples_factor);
    if (configs == NULL) {
       debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__);
       return NULL;
    }
 
-   for (i = 0; configs[i]; i++) {
-      m = &configs[i]->modes;
-      if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
-	 m->visualRating = GLX_SLOW_CONFIG;
-      }
-   }
-
    return (const const __DRIconfig **)configs;
 }
 
@@ -170,6 +232,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    if (!screen)
       return NULL;
 
+   screen->api = drm_api_create();
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
    screen->drmLock = (drmLock *) & sPriv->pSAREA->lock;
@@ -187,7 +250,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    dri_copy_version(&arg.drm_version, &sPriv->drm_version);
    arg.api = NULL;
 
-   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg.base);
+   screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg.base);
 
    if (!screen->pipe_screen || !arg.api) {
       debug_printf("%s: failed to create dri1 screen\n", __FUNCTION__);
@@ -200,7 +263,13 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    driParseOptionInfo(&screen->optionCache,
 		      __driConfigOptions, __driNConfigOptions);
 
-   configs = dri_fill_in_modes(sPriv, sPriv->fbBPP, 24, 8, 1);
+   /**
+    * FIXME: If the driver supports format conversion swapbuffer blits, we might
+    * want to support other color bit depths than the server is currently
+    * using.
+    */
+
+   configs = dri_fill_in_modes(screen, sPriv->fbBPP);
    if (!configs)
       goto out_no_configs;
 
@@ -230,13 +299,14 @@ dri_init_screen2(__DRIscreenPrivate * sPriv)
    if (!screen)
       goto fail;
 
+   screen->api = drm_api_create();
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
    sPriv->private = (void *)screen;
    sPriv->extensions = dri_screen_extensions;
    arg.mode = DRM_CREATE_NORMAL;
 
-   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg);
+   screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg);
    if (!screen->pipe_screen) {
       debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__);
       goto fail;
@@ -248,7 +318,7 @@ dri_init_screen2(__DRIscreenPrivate * sPriv)
    driParseOptionInfo(&screen->optionCache,
 		      __driConfigOptions, __driNConfigOptions);
 
-   return dri_fill_in_modes(sPriv, 4 * 8, 24, 8, 1);
+   return dri_fill_in_modes(screen, 32);
  fail:
    return NULL;
 }
diff --git a/src/gallium/state_trackers/dri/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h
index 100d9e50e0..f6c56d0f0c 100644
--- a/src/gallium/state_trackers/dri/dri_screen.h
+++ b/src/gallium/state_trackers/dri/dri_screen.h
@@ -49,19 +49,16 @@ struct dri_screen
     */
    driOptionCache optionCache;
 
-   /**
-    * Temporary(?) context to use for SwapBuffers or other situations in
-    * which we need a rendering context, but none is currently bound.
-    */
-   struct dri_context *dummyContext;
-
    /* drm */
    int fd;
    drmLock *drmLock;
 
    /* gallium */
+   struct drm_api *api;
    struct pipe_winsys *pipe_winsys;
    struct pipe_screen *pipe_screen;
+   boolean d_depth_bits_last;
+   boolean sd_depth_bits_last;
 };
 
 /** cast wrapper */
diff --git a/src/gallium/state_trackers/egl/Makefile b/src/gallium/state_trackers/egl/Makefile
index 692a3c8b76..e825aa718b 100644
--- a/src/gallium/state_trackers/egl/Makefile
+++ b/src/gallium/state_trackers/egl/Makefile
@@ -1,29 +1,19 @@
-TARGET     = libegldrm.a
-CFILES     = $(wildcard ./*.c)
-OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../..
-TOP        = ../../../..
+TOP = ../../../..
+include $(TOP)/configs/current
 
-include ${TOP}/configs/current
+LIBNAME = egldrm
 
-CFLAGS := \
-          -I${GALLIUMDIR}/include \
-          -I${GALLIUMDIR}/auxiliary \
-          -I${TOP}/src/mesa/drivers/dri/common \
-          -I${TOP}/src/mesa \
-          -I$(TOP)/include \
-          -I$(TOP)/src/egl/main \
-          ${LIBDRM_CFLAGS} \
-          ${CFLAGS}
+LIBRARY_INCLUDES = \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include \
+	-I$(TOP)/src/egl/main \
+	$(shell pkg-config --cflags-only-I libdrm)
 
-#############################################
 
-.PHONY	= all clean
+C_SOURCES = $(wildcard ./*.c)
 
-all: ${TARGET}
 
-${TARGET}: ${OBJECTS}
-	ar rcs $@ $^
-
-clean:
-	rm -rf ${OBJECTS} ${TARGET}
+include ../../Makefile.template
diff --git a/src/gallium/state_trackers/egl/egl_context.c b/src/gallium/state_trackers/egl/egl_context.c
index 36548fae26..c4f7361ca0 100644
--- a/src/gallium/state_trackers/egl/egl_context.c
+++ b/src/gallium/state_trackers/egl/egl_context.c
@@ -83,23 +83,16 @@ const struct dri_extension card_extensions[] = {
 	{NULL, NULL}
 };
 
-EGLContext
-drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list)
+_EGLContext *
+drm_create_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list)
 {
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev = lookup_drm_device(dpy);
 	struct drm_context *ctx;
 	struct drm_context *share = NULL;
 	struct st_context *st_share = NULL;
-	_EGLConfig *conf;
 	int i;
 	__GLcontextModes *visual;
 
-	conf = _eglLookupConfig(drv, dpy, config);
-	if (!conf) {
-		_eglError(EGL_BAD_CONFIG, "eglCreateContext");
-		return EGL_NO_CONTEXT;
-	}
-
 	for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
 		switch (attrib_list[i]) {
 			/* no attribs defined for now */
@@ -113,9 +106,9 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
 	if (!ctx)
 		goto err_c;
 
-	_eglInitContext(drv, dpy, &ctx->base, config, attrib_list);
+	_eglInitContext(drv, &ctx->base, conf, attrib_list);
 
-	ctx->pipe = drm_api_hooks.create_context(dev->screen);
+	ctx->pipe = dev->api->create_context(dev->api, dev->screen);
 	if (!ctx->pipe)
 		goto err_pipe;
 
@@ -129,28 +122,21 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
 	if (!ctx->st)
 		goto err_gl;
 
-	/* generate handle and insert into hash table */
-	_eglSaveContext(&ctx->base);
-	assert(_eglGetContextHandle(&ctx->base));
-
-	return _eglGetContextHandle(&ctx->base);
+	return &ctx->base;
 
 err_gl:
 	ctx->pipe->destroy(ctx->pipe);
 err_pipe:
 	free(ctx);
 err_c:
-	return EGL_NO_CONTEXT;
+	return NULL;
 }
 
 EGLBoolean
-drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
+drm_destroy_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *context)
 {
 	struct drm_context *c = lookup_drm_context(context);
-	_eglRemoveContext(&c->base);
-	if (c->base.IsBound) {
-		c->base.DeletePending = EGL_TRUE;
-	} else {
+	if (!_eglIsContextBound(&c->base)) {
 		st_destroy_context(c->st);
 		c->pipe->destroy(c->pipe);
 		free(c);
@@ -159,7 +145,7 @@ drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
 }
 
 EGLBoolean
-drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context)
+drm_make_current(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *context)
 {
 	struct drm_surface *readSurf = lookup_drm_surface(read);
 	struct drm_surface *drawSurf = lookup_drm_surface(draw);
diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c
index 489aa8d9af..69e2d6b708 100644
--- a/src/gallium/state_trackers/egl/egl_surface.c
+++ b/src/gallium/state_trackers/egl/egl_surface.c
@@ -67,39 +67,21 @@ drm_create_framebuffer(const __GLcontextModes *visual,
 }
 
 static void
-drm_create_texture(_EGLDriver *drv,
+drm_create_texture(_EGLDisplay *dpy,
                    struct drm_screen *scrn,
                    unsigned w, unsigned h)
 {
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev = lookup_drm_device(dpy);
 	struct pipe_screen *screen = dev->screen;
 	struct pipe_surface *surface;
 	struct pipe_texture *texture;
 	struct pipe_texture templat;
-	struct pipe_buffer *buf;
-	unsigned stride = 1024;
+	struct pipe_buffer *buf = NULL;
 	unsigned pitch = 0;
-	unsigned size = 0;
-
-	/* ugly */
-	if (stride < w)
-		stride = 2048;
-
-	pitch = stride * 4;
-	size = h * 2 * pitch;
-
-	buf = pipe_buffer_create(screen,
-	                         0, /* alignment */
-	                         PIPE_BUFFER_USAGE_GPU_READ_WRITE |
-	                         PIPE_BUFFER_USAGE_CPU_READ_WRITE,
-	                         size);
-
-	if (!buf)
-		goto err_buf;
 
 	memset(&templat, 0, sizeof(templat));
-	templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
-	templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
+	templat.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY;
 	templat.target = PIPE_TEXTURE_2D;
 	templat.last_level = 0;
 	templat.depth[0] = 1;
@@ -108,10 +90,9 @@ drm_create_texture(_EGLDriver *drv,
 	templat.height[0] = h;
 	pf_get_block(templat.format, &templat.block);
 
-	texture = screen->texture_blanket(dev->screen,
-	                                  &templat,
-	                                  &pitch,
-	                                  buf);
+	texture = screen->texture_create(dev->screen,
+	                                 &templat);
+
 	if (!texture)
 		goto err_tex;
 
@@ -125,14 +106,13 @@ drm_create_texture(_EGLDriver *drv,
 	if (!surface)
 		goto err_surf;
 
-
 	scrn->tex = texture;
 	scrn->surface = surface;
-	scrn->buffer = buf;
 	scrn->front.width = w;
 	scrn->front.height = h;
 	scrn->front.pitch = pitch;
-	drm_api_hooks.handle_from_buffer(screen, scrn->buffer, &scrn->front.handle);
+	dev->api->local_handle_from_texture(dev->api, screen, texture,
+	                                    &scrn->front.pitch, &scrn->front.handle);
 	if (0)
 		goto err_handle;
 
@@ -144,7 +124,6 @@ err_surf:
 	pipe_texture_reference(&texture, NULL);
 err_tex:
 	pipe_buffer_reference(&buf, NULL);
-err_buf:
 	return;
 }
 
@@ -153,9 +132,9 @@ err_buf:
  */
 
 void
-drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen)
+drm_takedown_shown_screen(_EGLDisplay *dpy, struct drm_screen *screen)
 {
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev = lookup_drm_device(dpy);
 
 	screen->surf = NULL;
 
@@ -178,22 +157,22 @@ drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen)
 	screen->shown = 0;
 }
 
-EGLSurface
-drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list)
+_EGLSurface *
+drm_create_window_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list)
 {
-	return EGL_NO_SURFACE;
+	return NULL;
 }
 
 
-EGLSurface
-drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list)
+_EGLSurface *
+drm_create_pixmap_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list)
 {
-	return EGL_NO_SURFACE;
+	return NULL;
 }
 
 
-EGLSurface
-drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+_EGLSurface *
+drm_create_pbuffer_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
                            const EGLint *attrib_list)
 {
 	int i;
@@ -201,13 +180,6 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 	int height = -1;
 	struct drm_surface *surf = NULL;
 	__GLcontextModes *visual;
-	_EGLConfig *conf;
-
-	conf = _eglLookupConfig(drv, dpy, config);
-	if (!conf) {
-		_eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface");
-		return EGL_NO_CONTEXT;
-	}
 
 	for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
 		switch (attrib_list[i]) {
@@ -225,14 +197,14 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
 	if (width < 1 || height < 1) {
 		_eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface");
-		return EGL_NO_SURFACE;
+		return NULL;
 	}
 
 	surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface));
 	if (!surf)
 		goto err;
 
-	if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list))
+	if (!_eglInitSurface(drv, &surf->base, EGL_PBUFFER_BIT, conf, attrib_list))
 		goto err_surf;
 
 	surf->w = width;
@@ -245,17 +217,16 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 	                                    (void*)surf);
 	drm_visual_modes_destroy(visual);
 
-	_eglSaveSurface(&surf->base);
-	return surf->base.Handle;
+	return &surf->base;
 
 err_surf:
 	free(surf);
 err:
-	return EGL_NO_SURFACE;
+	return NULL;
 }
 
-EGLSurface
-drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
+_EGLSurface *
+drm_create_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *cfg,
                                const EGLint *attrib_list)
 {
 	EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list);
@@ -264,22 +235,21 @@ drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
 }
 
 EGLBoolean
-drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy,
-                             EGLScreenMESA screen,
-                             EGLSurface surface, EGLModeMESA m)
+drm_show_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy,
+                             _EGLScreen *screen,
+                             _EGLSurface *surface, _EGLMode *mode)
 {
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev = lookup_drm_device(dpy);
 	struct drm_surface *surf = lookup_drm_surface(surface);
-	struct drm_screen *scrn = lookup_drm_screen(dpy, screen);
-	_EGLMode *mode = _eglLookupMode(dpy, m);
+	struct drm_screen *scrn = lookup_drm_screen(screen);
 	int ret;
 	unsigned int i, k;
 
 	if (scrn->shown)
-		drm_takedown_shown_screen(drv, scrn);
+		drm_takedown_shown_screen(dpy, scrn);
 
 
-	drm_create_texture(drv, scrn, mode->Width, mode->Height);
+	drm_create_texture(dpy, scrn, mode->Width, mode->Height);
 	if (!scrn->buffer)
 		return EGL_FALSE;
 
@@ -361,15 +331,12 @@ err_bo:
 }
 
 EGLBoolean
-drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+drm_destroy_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface)
 {
 	struct drm_surface *surf = lookup_drm_surface(surface);
-	_eglRemoveSurface(&surf->base);
-	if (surf->base.IsBound) {
-		surf->base.DeletePending = EGL_TRUE;
-	} else {
+	if (!_eglIsSurfaceBound(&surf->base)) {
 		if (surf->screen)
-			drm_takedown_shown_screen(drv, surf->screen);
+			drm_takedown_shown_screen(dpy, surf->screen);
 		st_unreference_framebuffer(surf->stfb);
 		free(surf);
 	}
@@ -377,8 +344,9 @@ drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 }
 
 EGLBoolean
-drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+drm_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw)
 {
+	struct drm_device *dev = lookup_drm_device(dpy);
 	struct drm_surface *surf = lookup_drm_surface(draw);
 	struct pipe_surface *back_surf;
 
@@ -396,7 +364,6 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
 		st_notify_swapbuffers(surf->stfb);
 
 		if (surf->screen) {
-			surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL);
 			surf->user->pipe->surface_copy(surf->user->pipe,
 				surf->screen->surface,
 				0, 0,
@@ -404,7 +371,15 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
 				0, 0,
 				surf->w, surf->h);
 			surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL);
-			/* TODO stuff here */
+
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+			/* TODO query connector property to see if this is needed */
+			drmModeDirtyFB(dev->drmFD, surf->screen->fbID, NULL, 0);
+#else
+			(void)dev;
+#endif
+
+			/* TODO more stuff here */
 		}
 	}
 
diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c
index 8e62008461..5140755001 100644
--- a/src/gallium/state_trackers/egl/egl_tracker.c
+++ b/src/gallium/state_trackers/egl/egl_tracker.c
@@ -6,6 +6,8 @@
 #include <string.h>
 #include "egl_tracker.h"
 
+#include <fcntl.h>
+
 #include "egllog.h"
 #include "state_tracker/drm_api.h"
 
@@ -21,44 +23,46 @@ extern const struct dri_extension card_extensions[];
  * Exported functions
  */
 
+static void
+drm_unload(_EGLDriver *drv)
+{
+	free(drv);
+}
+
 /**
  * The bootstrap function.  Return a new drm_driver object and
  * plug in API functions.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *dpy, const char *args)
+_eglMain(const char *args)
 {
-	struct drm_device *drm;
+	_EGLDriver *drv;
 
-	drm = (struct drm_device *) calloc(1, sizeof(struct drm_device));
-	if (!drm) {
+	drv = (_EGLDriver *) calloc(1, sizeof(_EGLDriver));
+	if (!drv) {
 		return NULL;
 	}
 
 	/* First fill in the dispatch table with defaults */
-	_eglInitDriverFallbacks(&drm->base);
+	_eglInitDriverFallbacks(drv);
 	/* then plug in our Drm-specific functions */
-	drm->base.API.Initialize = drm_initialize;
-	drm->base.API.Terminate = drm_terminate;
-	drm->base.API.CreateContext = drm_create_context;
-	drm->base.API.MakeCurrent = drm_make_current;
-	drm->base.API.CreateWindowSurface = drm_create_window_surface;
-	drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface;
-	drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface;
-	drm->base.API.DestroySurface = drm_destroy_surface;
-	drm->base.API.DestroyContext = drm_destroy_context;
-	drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa;
-	drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa;
-	drm->base.API.SwapBuffers = drm_swap_buffers;
-
-	drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/;
-	drm->base.Name = "DRM/Gallium/Win";
-
-	/* enable supported extensions */
-	drm->base.Extensions.MESA_screen_surface = EGL_TRUE;
-	drm->base.Extensions.MESA_copy_context = EGL_TRUE;
-
-	return &drm->base;
+	drv->API.Initialize = drm_initialize;
+	drv->API.Terminate = drm_terminate;
+	drv->API.CreateContext = drm_create_context;
+	drv->API.MakeCurrent = drm_make_current;
+	drv->API.CreateWindowSurface = drm_create_window_surface;
+	drv->API.CreatePixmapSurface = drm_create_pixmap_surface;
+	drv->API.CreatePbufferSurface = drm_create_pbuffer_surface;
+	drv->API.DestroySurface = drm_destroy_surface;
+	drv->API.DestroyContext = drm_destroy_context;
+	drv->API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa;
+	drv->API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa;
+	drv->API.SwapBuffers = drm_swap_buffers;
+
+	drv->Name = "DRM/Gallium/Win";
+	drv->Unload = drm_unload;
+
+	return drv;
 }
 
 static void
@@ -126,11 +130,18 @@ drm_find_dpms(struct drm_device *dev, struct drm_screen *screen)
 	screen->dpms = p;
 }
 
+static int drm_open_minor(int minor)
+{
+	char buf[64];
+
+	sprintf(buf, DRM_DEV_NAME, DRM_DIR_NAME, minor);
+	return open(buf, O_RDWR, 0);
+}
+
 EGLBoolean
-drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
+drm_initialize(_EGLDriver *drv, _EGLDisplay *disp, EGLint *major, EGLint *minor)
 {
-	_EGLDisplay *disp = _eglLookupDisplay(dpy);
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev;
 	struct drm_screen *screen = NULL;
 	drmModeConnectorPtr connector = NULL;
 	drmModeResPtr res = NULL;
@@ -139,14 +150,20 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 	EGLint i;
 	int fd;
 
-	fd = drmOpen("i915", NULL);
+	dev = (struct drm_device *) calloc(1, sizeof(struct drm_device));
+	if (!dev)
+		return EGL_FALSE;
+	dev->api = drm_api_create();
+
+	/* try the first node */
+	fd = drm_open_minor(0);
 	if (fd < 0)
 		goto err_fd;
 
 	dev->drmFD = fd;
 	drm_get_device_id(dev);
 
-	dev->screen = drm_api_hooks.create_screen(dev->drmFD, NULL);
+	dev->screen = dev->api->create_screen(dev->api, dev->drmFD, NULL);
 	if (!dev->screen)
 		goto err_screen;
 	dev->winsys = dev->screen->winsys;
@@ -184,6 +201,8 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 	}
 	dev->count_screens = num_screens;
 
+	disp->DriverData = dev;
+
 	/* for now we only have one config */
 	_EGLConfig *config = calloc(1, sizeof(*config));
 	memset(config, 1, sizeof(*config));
@@ -198,7 +217,10 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 	_eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT);
 	_eglAddConfig(disp, config);
 
-	drv->Initialized = EGL_TRUE;
+	disp->ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/;
+	/* enable supported extensions */
+	disp->Extensions.MESA_screen_surface = EGL_TRUE;
+	disp->Extensions.MESA_copy_context = EGL_TRUE;
 
 	*major = 1;
 	*minor = 4;
@@ -208,23 +230,27 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 err_screen:
 	drmClose(fd);
 err_fd:
+	free(dev);
 	return EGL_FALSE;
 }
 
 EGLBoolean
-drm_terminate(_EGLDriver *drv, EGLDisplay dpy)
+drm_terminate(_EGLDriver *drv, _EGLDisplay *dpy)
 {
-	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_device *dev = lookup_drm_device(dpy);
 	struct drm_screen *screen;
 	int i = 0;
 
+	_eglReleaseDisplayResources(drv, dpy);
+	_eglCleanupDisplay(dpy);
+
 	drmFreeVersion(dev->version);
 
 	for (i = 0; i < dev->count_screens; i++) {
 		screen = dev->screens[i];
 
 		if (screen->shown)
-			drm_takedown_shown_screen(drv, screen);
+			drm_takedown_shown_screen(dpy, screen);
 
 		drmModeFreeProperty(screen->dpms);
 		drmModeFreeConnector(screen->connector);
@@ -237,8 +263,9 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy)
 
 	drmClose(dev->drmFD);
 
-	_eglCleanupDisplay(_eglLookupDisplay(dpy));
+	dev->api->destroy(dev->api);
 	free(dev);
+	dpy->DriverData = NULL;
 
 	return EGL_TRUE;
 }
diff --git a/src/gallium/state_trackers/egl/egl_tracker.h b/src/gallium/state_trackers/egl/egl_tracker.h
index ce2717de63..dd4730f957 100644
--- a/src/gallium/state_trackers/egl/egl_tracker.h
+++ b/src/gallium/state_trackers/egl/egl_tracker.h
@@ -32,12 +32,11 @@ struct drm_context;
 
 struct drm_device
 {
-	_EGLDriver base; /* base class/object */
-
 	/*
 	 * pipe
 	 */
 
+	struct drm_api *api;
 	struct pipe_winsys *winsys;
 	struct pipe_screen *screen;
 
@@ -135,25 +134,29 @@ struct drm_screen
 };
 
 
+static INLINE struct drm_device *
+lookup_drm_device(_EGLDisplay *d)
+{
+	return (struct drm_device *) d->DriverData;
+}
+
+
 static INLINE struct drm_context *
-lookup_drm_context(EGLContext context)
+lookup_drm_context(_EGLContext *c)
 {
-	_EGLContext *c = _eglLookupContext(context);
 	return (struct drm_context *) c;
 }
 
 
 static INLINE struct drm_surface *
-lookup_drm_surface(EGLSurface surface)
+lookup_drm_surface(_EGLSurface *s)
 {
-	_EGLSurface *s = _eglLookupSurface(surface);
 	return (struct drm_surface *) s;
 }
 
 static INLINE struct drm_screen *
-lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen)
+lookup_drm_screen(_EGLScreen *s)
 {
-	_EGLScreen *s = _eglLookupScreen(dpy, screen);
 	return (struct drm_screen *) s;
 }
 
@@ -170,25 +173,25 @@ __GLcontextModes* drm_visual_from_config(_EGLConfig *conf);
  * egl_surface.h
  */
 /*@{*/
-void drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen);
+void drm_takedown_shown_screen(_EGLDisplay *dpy, struct drm_screen *screen);
 /*@}*/
 
 /**
  * All function exported to the egl side.
  */
 /*@{*/
-EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor);
-EGLBoolean drm_terminate(_EGLDriver *drv, EGLDisplay dpy);
-EGLContext drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list);
-EGLBoolean drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context);
-EGLSurface drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list);
-EGLSurface drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list);
-EGLSurface drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
-EGLSurface drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, const EGLint *attrib_list);
-EGLBoolean drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface surface, EGLModeMESA m);
-EGLBoolean drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface);
-EGLBoolean drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context);
-EGLBoolean drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw);
+EGLBoolean drm_initialize(_EGLDriver *drv, _EGLDisplay *dpy, EGLint *major, EGLint *minor);
+EGLBoolean drm_terminate(_EGLDriver *drv, _EGLDisplay *dpy);
+_EGLContext *drm_create_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list);
+EGLBoolean drm_destroy_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *context);
+_EGLSurface *drm_create_window_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list);
+_EGLSurface *drm_create_pixmap_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list);
+_EGLSurface *drm_create_pbuffer_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list);
+_EGLSurface *drm_create_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list);
+EGLBoolean drm_show_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, _EGLSurface *surface, _EGLMode *mode);
+EGLBoolean drm_destroy_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface);
+EGLBoolean drm_make_current(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *context);
+EGLBoolean drm_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw);
 /*@}*/
 
 #endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
index 16d4f1e32c..20d682de3f 100644
--- a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
@@ -82,7 +82,6 @@ static int vlResizeFrameBuffer
 	template.width[0] = width;
 	template.height[0] = height;
 	template.depth[0] = 1;
-	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
 	template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
 
@@ -669,7 +668,6 @@ static int vlInit
 	sampler.compare_func = PIPE_FUNC_ALWAYS;
 	sampler.normalized_coords = 1;
 	/*sampler.prefilter = ;*/
-	/*sampler.shadow_ambient = ;*/
 	/*sampler.lod_bias = ;*/
 	/*sampler.min_lod = ;*/
 	/*sampler.max_lod = ;*/
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
index eb8270ecad..23631adb69 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
@@ -1074,7 +1074,6 @@ static int vlInit
 		sampler.compare_func = PIPE_FUNC_ALWAYS;
 		sampler.normalized_coords = 1;
 		/*sampler.prefilter = ;*/
-		/*sampler.shadow_ambient = ;*/
 		/*sampler.lod_bias = ;*/
 		sampler.min_lod = 0;
 		/*sampler.max_lod = ;*/
@@ -1090,7 +1089,6 @@ static int vlInit
 	template.width[0] = vlRoundUpPOT(mc->picture_width);
 	template.height[0] = vlRoundUpPOT(mc->picture_height);
 	template.depth[0] = 1;
-	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
 	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_DYNAMIC;
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 92388f7978..7f60852cae 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -45,7 +45,6 @@ int vlCreateSurface
 	template.width[0] = vlRoundUpPOT(sfc->width);
 	template.height[0] = vlRoundUpPOT(sfc->height);
 	template.depth[0] = 1;
-	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
 	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET;
 
diff --git a/src/gallium/state_trackers/glx/xlib/Makefile b/src/gallium/state_trackers/glx/xlib/Makefile
index 6d10b090aa..7b2adc62c3 100644
--- a/src/gallium/state_trackers/glx/xlib/Makefile
+++ b/src/gallium/state_trackers/glx/xlib/Makefile
@@ -5,13 +5,12 @@ LIBNAME = xlib
 
 LIBRARY_INCLUDES = \
 	-I$(TOP)/include \
-	-I$(TOP)/src/mesa \
-	-I$(TOP)/src/mesa/main
+	-I$(TOP)/src/mesa
 
 C_SOURCES = \
-	glxapi.c \
-	fakeglx.c \
-	fakeglx_fonts.c \
+	glx_api.c \
+	glx_getproc.c \
+	glx_usefont.c \
 	xm_api.c
 
 include ../../../Makefile.template
diff --git a/src/gallium/state_trackers/glx/xlib/SConscript b/src/gallium/state_trackers/glx/xlib/SConscript
index 0dbe341397..fa96df357d 100644
--- a/src/gallium/state_trackers/glx/xlib/SConscript
+++ b/src/gallium/state_trackers/glx/xlib/SConscript
@@ -4,8 +4,7 @@
 Import('*')
 
 if env['platform'] == 'linux' \
-        and 'mesa' in env['statetrackers'] \
-        and ('softpipe' or 'i915simple' or 'trace') in env['drivers']:
+        and 'mesa' in env['statetrackers']:
 
     env = env.Clone()
 
@@ -18,9 +17,10 @@ if env['platform'] == 'linux' \
 
     st_xlib = env.ConvenienceLibrary(
 	target = 'st_xlib',
-	source = [ 'glxapi.c',
-		'fakeglx.c',
-		'fakeglx_fonts.c',
+	source = [
+		'glx_api.c',
+		'glx_getproc.c',
+		'glx_usefont.c',
 		'xm_api.c',
 		]
     )
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/glx_api.c
index 6df4c7d693..d1a98f8991 100644
--- a/src/gallium/state_trackers/glx/xlib/fakeglx.c
+++ b/src/gallium/state_trackers/glx/xlib/glx_api.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.6
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -23,31 +24,21 @@
  */
 
 
-/*
- * This is an emulation of the GLX API which allows Mesa/GLX-based programs
- * to run on X servers which do not have the real GLX extension.
- *
- * Thanks to the contributors:
- *
- * Initial version:  Philip Brown (phil@bolthole.com)
- * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu)
- * Further visual-handling refinements: Wolfram Gloger
- *    (wmglo@Dent.MED.Uni-Muenchen.DE).
- *
- * Notes:
- *   Don't be fooled, stereo isn't supported yet.
+/**
+ * "Fake" GLX API implemented in terms of the XMesa*() functions.
  */
 
 
 
-#include "glxapi.h"
+#define GLX_GLXEXT_PROTOTYPES
+#include "GL/glx.h"
+
 #include "xm_api.h"
-#include "context.h"
-#include "config.h"
-#include "macros.h"
-#include "imports.h"
-#include "version.h"
-#include "fakeglx.h"
+#include "main/context.h"
+#include "main/config.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/version.h"
 #include "state_tracker/st_context.h"
 #include "state_tracker/st_public.h"
 
@@ -80,23 +71,44 @@
    "GLX_SGIX_fbconfig " \
    "GLX_SGIX_pbuffer "
 
-/*
- * Our fake GLX context will contain a "real" GLX context and an XMesa context.
- *
- * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context,
- * and vice versa.
- *
- * We really just need this structure in order to make the libGL functions
- * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay()
- * work correctly.
+#define DEFAULT_DIRECT GL_TRUE
+
+
+
+/**
+ * The GLXContext typedef is defined as a pointer to this structure.
  */
-struct fake_glx_context {
-   __GLXcontext glxContext;   /* this MUST be first! */
+struct __GLXcontextRec
+{
+   Display *currentDpy;
+   GLboolean isDirect;
+   GLXDrawable currentDrawable;
+   GLXDrawable currentReadable;
+   XID xid;
+
    XMesaContext xmesaContext;
 };
 
 
 
+static pipe_tsd ContextTSD;
+
+/** Set current context for calling thread */
+static void
+SetCurrentContext(GLXContext c)
+{
+   pipe_tsd_set(&ContextTSD, c);
+}
+
+/** Get current context for calling thread */
+static GLXContext
+GetCurrentContext(void)
+{
+   return pipe_tsd_get(&ContextTSD);
+}
+
+
+
 /**********************************************************************/
 /***                       GLX Visual Code                          ***/
 /**********************************************************************/
@@ -363,10 +375,6 @@ find_glx_visual( Display *dpy, XVisualInfo *vinfo )
 }
 
 
-
-
-
-
 /**
  * Try to get an X visual which matches the given arguments.
  */
@@ -415,7 +423,6 @@ get_visual( Display *dpy, int scr, unsigned int depth, int xclass )
 }
 
 
-
 /*
  * Retrieve the value of the given environment variable and find
  * the X visual which matches it.
@@ -999,8 +1006,8 @@ choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig )
 }
 
 
-static XVisualInfo *
-Fake_glXChooseVisual( Display *dpy, int screen, int *list )
+XVisualInfo *
+glXChooseVisual( Display *dpy, int screen, int *list )
 {
    XMesaVisual xmvis;
 
@@ -1021,18 +1028,18 @@ Fake_glXChooseVisual( Display *dpy, int screen, int *list )
 }
 
 
-static GLXContext
-Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
-                       GLXContext share_list, Bool direct )
+GLXContext
+glXCreateContext( Display *dpy, XVisualInfo *visinfo,
+                  GLXContext share_list, Bool direct )
 {
    XMesaVisual xmvis;
-   struct fake_glx_context *glxCtx;
-   struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list;
+   GLXContext glxCtx;
+   GLXContext shareCtx = share_list;
 
    if (!dpy || !visinfo)
       return 0;
 
-   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   glxCtx = CALLOC_STRUCT(__GLXcontextRec);
    if (!glxCtx)
       return 0;
 
@@ -1059,13 +1066,11 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   glxCtx->isDirect = DEFAULT_DIRECT;
+   glxCtx->currentDpy = dpy;
+   glxCtx->xid = (XID) glxCtx;  /* self pointer */
 
-   return (GLXContext) glxCtx;
+   return glxCtx;
 }
 
 
@@ -1078,11 +1083,11 @@ static XMesaBuffer MakeCurrent_PrevReadBuffer = 0;
 
 
 /* GLX 1.3 and later */
-static Bool
-Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
-                            GLXDrawable read, GLXContext ctx )
+Bool
+glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
+                       GLXDrawable read, GLXContext ctx )
 {
-   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   GLXContext glxCtx = ctx;
    static boolean firsttime = 1, no_rast = 0;
 
    if (firsttime) {
@@ -1090,7 +1095,6 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
       firsttime = 0;
    }
 
-
    if (ctx && draw && read) {
       XMesaBuffer drawBuffer, readBuffer;
       XMesaContext xmctx = glxCtx->xmesaContext;
@@ -1145,9 +1149,10 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
 
       /* Now make current! */
       if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) {
-         ((__GLXcontext *) ctx)->currentDpy = dpy;
-         ((__GLXcontext *) ctx)->currentDrawable = draw;
-         ((__GLXcontext *) ctx)->currentReadable = read;
+         ctx->currentDpy = dpy;
+         ctx->currentDrawable = draw;
+         ctx->currentReadable = read;
+         SetCurrentContext(ctx);
          return True;
       }
       else {
@@ -1162,6 +1167,7 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
       MakeCurrent_PrevReadable = 0;
       MakeCurrent_PrevDrawBuffer = 0;
       MakeCurrent_PrevReadBuffer = 0;
+      SetCurrentContext(NULL);
       return True;
    }
    else {
@@ -1173,15 +1179,61 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
 }
 
 
-static Bool
-Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx )
+Bool
+glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx )
+{
+   return glXMakeContextCurrent( dpy, drawable, drawable, ctx );
+}
+
+
+GLXContext
+glXGetCurrentContext(void)
+{
+   return GetCurrentContext();
+}
+
+
+Display *
+glXGetCurrentDisplay(void)
+{
+   GLXContext glxCtx = glXGetCurrentContext();
+
+   return glxCtx ? glxCtx->currentDpy : NULL;
+}
+
+
+Display *
+glXGetCurrentDisplayEXT(void)
+{
+   return glXGetCurrentDisplay();
+}
+
+
+GLXDrawable
+glXGetCurrentDrawable(void)
+{
+   GLXContext gc = glXGetCurrentContext();
+   return gc ? gc->currentDrawable : 0;
+}
+
+
+GLXDrawable
+glXGetCurrentReadDrawable(void)
+{
+   GLXContext gc = glXGetCurrentContext();
+   return gc ? gc->currentReadable : 0;
+}
+
+
+GLXDrawable
+glXGetCurrentReadDrawableSGI(void)
 {
-   return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx );
+   return glXGetCurrentReadDrawable();
 }
 
 
-static GLXPixmap
-Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap )
+GLXPixmap
+glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap )
 {
    XMesaVisual v;
    XMesaBuffer b;
@@ -1205,9 +1257,9 @@ Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap )
 
 /*** GLX_MESA_pixmap_colormap ***/
 
-static GLXPixmap
-Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo,
-                             Pixmap pixmap, Colormap cmap )
+GLXPixmap
+glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo,
+                        Pixmap pixmap, Colormap cmap )
 {
    XMesaVisual v;
    XMesaBuffer b;
@@ -1229,8 +1281,8 @@ Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo,
 }
 
 
-static void
-Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap )
+void
+glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap )
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, pixmap);
    if (b) {
@@ -1242,14 +1294,12 @@ Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap )
 }
 
 
-static void
-Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst,
-                     unsigned long mask )
+void
+glXCopyContext( Display *dpy, GLXContext src, GLXContext dst,
+                unsigned long mask )
 {
-   struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src;
-   struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst;
-   XMesaContext xm_src = fakeSrc->xmesaContext;
-   XMesaContext xm_dst = fakeDst->xmesaContext;
+   XMesaContext xm_src = src->xmesaContext;
+   XMesaContext xm_dst = dst->xmesaContext;
    (void) dpy;
    if (MakeCurrent_PrevContext == src) {
       _mesa_Flush();
@@ -1258,8 +1308,8 @@ Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst,
 }
 
 
-static Bool
-Fake_glXQueryExtension( Display *dpy, int *errorb, int *event )
+Bool
+glXQueryExtension( Display *dpy, int *errorb, int *event )
 {
    /* Mesa's GLX isn't really an X extension but we try to act like one. */
    (void) dpy;
@@ -1269,18 +1319,10 @@ Fake_glXQueryExtension( Display *dpy, int *errorb, int *event )
 }
 
 
-extern void _kw_ungrab_all( Display *dpy );
-void _kw_ungrab_all( Display *dpy )
-{
-   XUngrabPointer( dpy, CurrentTime );
-   XUngrabKeyboard( dpy, CurrentTime );
-}
-
-
-static void
-Fake_glXDestroyContext( Display *dpy, GLXContext ctx )
+void
+glXDestroyContext( Display *dpy, GLXContext ctx )
 {
-   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   GLXContext glxCtx = ctx;
    (void) dpy;
    MakeCurrent_PrevContext = 0;
    MakeCurrent_PrevDrawable = 0;
@@ -1293,18 +1335,18 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx )
 }
 
 
-static Bool
-Fake_glXIsDirect( Display *dpy, GLXContext ctx )
+Bool
+glXIsDirect( Display *dpy, GLXContext ctx )
 {
-   (void) dpy;
+   GLXContext glxCtx = ctx;
    (void) ctx;
-   return False;
+   return glxCtx->isDirect;
 }
 
 
 
-static void
-Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable )
+void
+glXSwapBuffers( Display *dpy, GLXDrawable drawable )
 {
    XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable );
    static boolean firsttime = 1, no_rast = 0;
@@ -1330,8 +1372,8 @@ Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable )
 
 /*** GLX_MESA_copy_sub_buffer ***/
 
-static void
-Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable,
+void
+glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable,
                            int x, int y, int width, int height )
 {
    XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable );
@@ -1344,8 +1386,8 @@ Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable,
 }
 
 
-static Bool
-Fake_glXQueryVersion( Display *dpy, int *maj, int *min )
+Bool
+glXQueryVersion( Display *dpy, int *maj, int *min )
 {
    (void) dpy;
    /* Return GLX version, not Mesa version */
@@ -1561,8 +1603,8 @@ get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig )
 }
 
 
-static int
-Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo,
+int
+glXGetConfig( Display *dpy, XVisualInfo *visinfo,
                    int attrib, int *value )
 {
    XMesaVisual xmvis;
@@ -1591,8 +1633,8 @@ Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo,
 }
 
 
-static void
-Fake_glXWaitGL( void )
+void
+glXWaitGL( void )
 {
    XMesaContext xmesa = XMesaGetCurrentContext();
    XMesaFlush( xmesa );
@@ -1600,8 +1642,8 @@ Fake_glXWaitGL( void )
 
 
 
-static void
-Fake_glXWaitX( void )
+void
+glXWaitX( void )
 {
    XMesaContext xmesa = XMesaGetCurrentContext();
    XMesaFlush( xmesa );
@@ -1617,8 +1659,8 @@ get_extensions( void )
 
 
 /* GLX 1.1 and later */
-static const char *
-Fake_glXQueryExtensionsString( Display *dpy, int screen )
+const char *
+glXQueryExtensionsString( Display *dpy, int screen )
 {
    (void) dpy;
    (void) screen;
@@ -1628,8 +1670,8 @@ Fake_glXQueryExtensionsString( Display *dpy, int screen )
 
 
 /* GLX 1.1 and later */
-static const char *
-Fake_glXQueryServerString( Display *dpy, int screen, int name )
+const char *
+glXQueryServerString( Display *dpy, int screen, int name )
 {
    static char version[1000];
    _mesa_sprintf(version, "%d.%d %s",
@@ -1653,8 +1695,8 @@ Fake_glXQueryServerString( Display *dpy, int screen, int name )
 
 
 /* GLX 1.1 and later */
-static const char *
-Fake_glXGetClientString( Display *dpy, int name )
+const char *
+glXGetClientString( Display *dpy, int name )
 {
    static char version[1000];
    _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION,
@@ -1681,8 +1723,8 @@ Fake_glXGetClientString( Display *dpy, int name )
  */
 
 
-static int
-Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config,
+int
+glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config,
                            int attribute, int *value )
 {
    XMesaVisual v = (XMesaVisual) config;
@@ -1696,8 +1738,8 @@ Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config,
 }
 
 
-static GLXFBConfig *
-Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements )
+GLXFBConfig *
+glXGetFBConfigs( Display *dpy, int screen, int *nelements )
 {
    XVisualInfo *visuals, visTemplate;
    const long visMask = VisualScreenMask;
@@ -1722,15 +1764,15 @@ Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements )
 }
 
 
-static GLXFBConfig *
-Fake_glXChooseFBConfig( Display *dpy, int screen,
+GLXFBConfig *
+glXChooseFBConfig( Display *dpy, int screen,
                         const int *attribList, int *nitems )
 {
    XMesaVisual xmvis;
 
    if (!attribList || !attribList[0]) {
       /* return list of all configs (per GLX_SGIX_fbconfig spec) */
-      return Fake_glXGetFBConfigs(dpy, screen, nitems);
+      return glXGetFBConfigs(dpy, screen, nitems);
    }
 
    xmvis = choose_visual(dpy, screen, attribList, GL_TRUE);
@@ -1751,8 +1793,8 @@ Fake_glXChooseFBConfig( Display *dpy, int screen,
 }
 
 
-static XVisualInfo *
-Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config )
+XVisualInfo *
+glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config )
 {
    if (dpy && config) {
       XMesaVisual xmvis = (XMesaVisual) config;
@@ -1773,8 +1815,8 @@ Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config )
 }
 
 
-static GLXWindow
-Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win,
+GLXWindow
+glXCreateWindow( Display *dpy, GLXFBConfig config, Window win,
                       const int *attribList )
 {
    XMesaVisual xmvis = (XMesaVisual) config;
@@ -1793,8 +1835,8 @@ Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win,
 }
 
 
-static void
-Fake_glXDestroyWindow( Display *dpy, GLXWindow window )
+void
+glXDestroyWindow( Display *dpy, GLXWindow window )
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable) window);
    if (b)
@@ -1804,8 +1846,8 @@ Fake_glXDestroyWindow( Display *dpy, GLXWindow window )
 
 
 /* XXX untested */
-static GLXPixmap
-Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap,
+GLXPixmap
+glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap,
                       const int *attribList )
 {
    XMesaVisual v = (XMesaVisual) config;
@@ -1914,8 +1956,8 @@ Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap,
 }
 
 
-static void
-Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap )
+void
+glXDestroyPixmap( Display *dpy, GLXPixmap pixmap )
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable)pixmap);
    if (b)
@@ -1924,8 +1966,8 @@ Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap )
 }
 
 
-static GLXPbuffer
-Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config,
+GLXPbuffer
+glXCreatePbuffer( Display *dpy, GLXFBConfig config,
                        const int *attribList )
 {
    XMesaVisual xmvis = (XMesaVisual) config;
@@ -1977,8 +2019,8 @@ Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config,
 }
 
 
-static void
-Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf )
+void
+glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf )
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, pbuf);
    if (b) {
@@ -1987,8 +2029,8 @@ Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf )
 }
 
 
-static void
-Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute,
+void
+glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute,
                        unsigned int *value )
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw);
@@ -2029,19 +2071,19 @@ Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute,
 }
 
 
-static GLXContext
-Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
+GLXContext
+glXCreateNewContext( Display *dpy, GLXFBConfig config,
                           int renderType, GLXContext shareList, Bool direct )
 {
-   struct fake_glx_context *glxCtx;
-   struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList;
+   GLXContext glxCtx;
+   GLXContext shareCtx = shareList;
    XMesaVisual xmvis = (XMesaVisual) config;
 
    if (!dpy || !config ||
        (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE))
       return 0;
 
-   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   glxCtx = CALLOC_STRUCT(__GLXcontextRec);
    if (!glxCtx)
       return 0;
 
@@ -2055,20 +2097,18 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   glxCtx->isDirect = DEFAULT_DIRECT;
+   glxCtx->currentDpy = dpy;
+   glxCtx->xid = (XID) glxCtx;  /* self pointer */
 
-   return (GLXContext) glxCtx;
+   return glxCtx;
 }
 
 
-static int
-Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value )
+int
+glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value )
 {
-   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   GLXContext glxCtx = ctx;
    XMesaContext xmctx = glxCtx->xmesaContext;
 
    (void) dpy;
@@ -2094,8 +2134,8 @@ Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value )
 }
 
 
-static void
-Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask )
+void
+glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask )
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
    if (xmbuf)
@@ -2103,8 +2143,8 @@ Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask )
 }
 
 
-static void
-Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable,
+void
+glXGetSelectedEvent( Display *dpy, GLXDrawable drawable,
                           unsigned long *mask )
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
@@ -2118,8 +2158,8 @@ Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable,
 
 /*** GLX_SGI_swap_control ***/
 
-static int
-Fake_glXSwapIntervalSGI(int interval)
+int
+glXSwapIntervalSGI(int interval)
 {
    (void) interval;
    return 0;
@@ -2131,16 +2171,16 @@ Fake_glXSwapIntervalSGI(int interval)
 
 static unsigned int FrameCounter = 0;
 
-static int
-Fake_glXGetVideoSyncSGI(unsigned int *count)
+int
+glXGetVideoSyncSGI(unsigned int *count)
 {
    /* this is a bogus implementation */
    *count = FrameCounter++;
    return 0;
 }
 
-static int
-Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
+int
+glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
 {
    if (divisor <= 0 || remainder < 0)
       return GLX_BAD_VALUE;
@@ -2156,15 +2196,15 @@ Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
 
 /*** GLX_SGI_make_current_read ***/
 
-static Bool
-Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
+Bool
+glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
 {
-   return Fake_glXMakeContextCurrent( dpy, draw, read, ctx );
+   return glXMakeContextCurrent( dpy, draw, read, ctx );
 }
 
 /* not used
 static GLXDrawable
-Fake_glXGetCurrentReadDrawableSGI(void)
+glXGetCurrentReadDrawableSGI(void)
 {
    return 0;
 }
@@ -2174,8 +2214,8 @@ Fake_glXGetCurrentReadDrawableSGI(void)
 /*** GLX_SGIX_video_source ***/
 #if defined(_VL_H)
 
-static GLXVideoSourceSGIX
-Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode)
+GLXVideoSourceSGIX
+glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode)
 {
    (void) dpy;
    (void) screen;
@@ -2186,8 +2226,8 @@ Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPa
    return 0;
 }
 
-static void
-Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
+void
+glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
 {
    (void) dpy;
    (void) src;
@@ -2198,30 +2238,30 @@ Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
 
 /*** GLX_EXT_import_context ***/
 
-static void
-Fake_glXFreeContextEXT(Display *dpy, GLXContext context)
+void
+glXFreeContextEXT(Display *dpy, GLXContext context)
 {
    (void) dpy;
    (void) context;
 }
 
-static GLXContextID
-Fake_glXGetContextIDEXT(const GLXContext context)
+GLXContextID
+glXGetContextIDEXT(const GLXContext context)
 {
    (void) context;
    return 0;
 }
 
-static GLXContext
-Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID)
+GLXContext
+glXImportContextEXT(Display *dpy, GLXContextID contextID)
 {
    (void) dpy;
    (void) contextID;
    return 0;
 }
 
-static int
-Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value)
+int
+glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value)
 {
    (void) dpy;
    (void) context;
@@ -2234,21 +2274,21 @@ Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int
 
 /*** GLX_SGIX_fbconfig ***/
 
-static int
-Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value)
+int
+glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value)
 {
-   return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value);
+   return glXGetFBConfigAttrib(dpy, config, attribute, value);
 }
 
-static GLXFBConfigSGIX *
-Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements)
+GLXFBConfigSGIX *
+glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements)
 {
-   return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements);
+   return (GLXFBConfig *) glXChooseFBConfig(dpy, screen, attrib_list, nelements);
 }
 
 
-static GLXPixmap
-Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap)
+GLXPixmap
+glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap)
 {
    XMesaVisual xmvis = (XMesaVisual) config;
    XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0);
@@ -2256,14 +2296,14 @@ Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixm
 }
 
 
-static GLXContext
-Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct)
+GLXContext
+glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct)
 {
    XMesaVisual xmvis = (XMesaVisual) config;
-   struct fake_glx_context *glxCtx;
-   struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list;
+   GLXContext glxCtx;
+   GLXContext shareCtx = share_list;
 
-   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   glxCtx = CALLOC_STRUCT(__GLXcontextRec);
    if (!glxCtx)
       return 0;
 
@@ -2277,25 +2317,23 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   glxCtx->isDirect = DEFAULT_DIRECT;
+   glxCtx->currentDpy = dpy;
+   glxCtx->xid = (XID) glxCtx;  /* self pointer */
 
-   return (GLXContext) glxCtx;
+   return glxCtx;
 }
 
 
-static XVisualInfo *
-Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config)
+XVisualInfo *
+glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config)
 {
-   return Fake_glXGetVisualFromFBConfig(dpy, config);
+   return glXGetVisualFromFBConfig(dpy, config);
 }
 
 
-static GLXFBConfigSGIX
-Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
+GLXFBConfigSGIX
+glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
 {
    XMesaVisual xmvis = find_glx_visual(dpy, vis);
    if (!xmvis) {
@@ -2310,8 +2348,8 @@ Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
 
 /*** GLX_SGIX_pbuffer ***/
 
-static GLXPbufferSGIX
-Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config,
+GLXPbufferSGIX
+glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config,
                              unsigned int width, unsigned int height,
                              int *attribList)
 {
@@ -2349,8 +2387,8 @@ Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config,
 }
 
 
-static void
-Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
+void
+glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf);
    if (xmbuf) {
@@ -2359,8 +2397,8 @@ Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
 }
 
 
-static int
-Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value)
+int
+glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value)
 {
    const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf);
 
@@ -2392,8 +2430,8 @@ Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, un
 }
 
 
-static void
-Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
+void
+glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
    if (xmbuf) {
@@ -2403,8 +2441,8 @@ Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
 }
 
 
-static void
-Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask)
+void
+glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask)
 {
    XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
    if (xmbuf) {
@@ -2419,8 +2457,8 @@ Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *
 
 /*** GLX_SGI_cushion ***/
 
-static void
-Fake_glXCushionSGI(Display *dpy, Window win, float cushion)
+void
+glXCushionSGI(Display *dpy, Window win, float cushion)
 {
    (void) dpy;
    (void) win;
@@ -2431,8 +2469,8 @@ Fake_glXCushionSGI(Display *dpy, Window win, float cushion)
 
 /*** GLX_SGIX_video_resize ***/
 
-static int
-Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window)
+int
+glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window)
 {
    (void) dpy;
    (void) screen;
@@ -2441,8 +2479,8 @@ Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window w
    return 0;
 }
 
-static int
-Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h)
+int
+glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h)
 {
    (void) dpy;
    (void) screen;
@@ -2454,8 +2492,8 @@ Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int
    return 0;
 }
 
-static int
-Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h)
+int
+glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h)
 {
    (void) dpy;
    (void) screen;
@@ -2467,8 +2505,8 @@ Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int
    return 0;
 }
 
-static int
-Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh)
+int
+glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh)
 {
    (void) dpy;
    (void) screen;
@@ -2480,8 +2518,8 @@ Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, i
    return 0;
 }
 
-static int
-Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype)
+int
+glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype)
 {
    (void) dpy;
    (void) screen;
@@ -2495,8 +2533,8 @@ Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum syncty
 /*** GLX_SGIX_dmbuffer **/
 
 #if defined(_DM_BUFFER_H_)
-static Bool
-Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer)
+Bool
+glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer)
 {
    (void) dpy;
    (void) pbuffer;
@@ -2509,8 +2547,8 @@ Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *p
 
 /*** GLX_SGIX_swap_group ***/
 
-static void
-Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member)
+void
+glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member)
 {
    (void) dpy;
    (void) drawable;
@@ -2521,16 +2559,16 @@ Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member
 
 /*** GLX_SGIX_swap_barrier ***/
 
-static void
-Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier)
+void
+glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier)
 {
    (void) dpy;
    (void) drawable;
    (void) barrier;
 }
 
-static Bool
-Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
+Bool
+glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
 {
    (void) dpy;
    (void) screen;
@@ -2542,8 +2580,8 @@ Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
 
 /*** GLX_SUN_get_transparent_index ***/
 
-static Status
-Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent)
+Status
+glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent)
 {
    (void) dpy;
    (void) overlay;
@@ -2560,8 +2598,8 @@ Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, lo
  * Release the depth, stencil, accum buffers attached to a GLXDrawable
  * (a window or pixmap) prior to destroying the GLXDrawable.
  */
-static Bool
-Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
+Bool
+glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, d);
    if (b) {
@@ -2573,8 +2611,8 @@ Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
 
 /*** GLX_EXT_texture_from_pixmap ***/
 
-static void
-Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
+void
+glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
                         const int *attrib_list)
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, drawable);
@@ -2582,162 +2620,10 @@ Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
       XMesaBindTexImage(dpy, b, buffer, attrib_list);
 }
 
-static void
-Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
+void
+glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
 {
    XMesaBuffer b = XMesaFindBuffer(dpy, drawable);
    if (b)
       XMesaReleaseTexImage(dpy, b, buffer);
 }
-
-
-
-/**
- * Create a new GLX API dispatch table with its function pointers
- * initialized to point to Mesa's "fake" GLX API functions.
- *
- * Note: there used to be a similar function
- * (_real_GetGLXDispatchTable) that returns a new dispatch table with
- * all pointers initalized to point to "real" GLX functions (which
- * understand GLX wire protocol, etc).
- */
-struct _glxapi_table *
-_mesa_GetGLXDispatchTable(void)
-{
-   static struct _glxapi_table glx;
-
-   /* be sure our dispatch table size <= libGL's table */
-   {
-      GLuint size = sizeof(struct _glxapi_table) / sizeof(void *);
-      (void) size;
-      assert(_glxapi_get_dispatch_table_size() >= size);
-   }
-
-   /* initialize the whole table to no-ops */
-   _glxapi_set_no_op_table(&glx);
-
-   /* now initialize the table with the functions I implement */
-   glx.ChooseVisual = Fake_glXChooseVisual;
-   glx.CopyContext = Fake_glXCopyContext;
-   glx.CreateContext = Fake_glXCreateContext;
-   glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap;
-   glx.DestroyContext = Fake_glXDestroyContext;
-   glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap;
-   glx.GetConfig = Fake_glXGetConfig;
-   /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/
-   /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/
-   glx.IsDirect = Fake_glXIsDirect;
-   glx.MakeCurrent = Fake_glXMakeCurrent;
-   glx.QueryExtension = Fake_glXQueryExtension;
-   glx.QueryVersion = Fake_glXQueryVersion;
-   glx.SwapBuffers = Fake_glXSwapBuffers;
-   glx.UseXFont = Fake_glXUseXFont;
-   glx.WaitGL = Fake_glXWaitGL;
-   glx.WaitX = Fake_glXWaitX;
-
-   /*** GLX_VERSION_1_1 ***/
-   glx.GetClientString = Fake_glXGetClientString;
-   glx.QueryExtensionsString = Fake_glXQueryExtensionsString;
-   glx.QueryServerString = Fake_glXQueryServerString;
-
-   /*** GLX_VERSION_1_2 ***/
-   /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/
-
-   /*** GLX_VERSION_1_3 ***/
-   glx.ChooseFBConfig = Fake_glXChooseFBConfig;
-   glx.CreateNewContext = Fake_glXCreateNewContext;
-   glx.CreatePbuffer = Fake_glXCreatePbuffer;
-   glx.CreatePixmap = Fake_glXCreatePixmap;
-   glx.CreateWindow = Fake_glXCreateWindow;
-   glx.DestroyPbuffer = Fake_glXDestroyPbuffer;
-   glx.DestroyPixmap = Fake_glXDestroyPixmap;
-   glx.DestroyWindow = Fake_glXDestroyWindow;
-   /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/
-   glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib;
-   glx.GetFBConfigs = Fake_glXGetFBConfigs;
-   glx.GetSelectedEvent = Fake_glXGetSelectedEvent;
-   glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig;
-   glx.MakeContextCurrent = Fake_glXMakeContextCurrent;
-   glx.QueryContext = Fake_glXQueryContext;
-   glx.QueryDrawable = Fake_glXQueryDrawable;
-   glx.SelectEvent = Fake_glXSelectEvent;
-
-   /*** GLX_SGI_swap_control ***/
-   glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI;
-
-   /*** GLX_SGI_video_sync ***/
-   glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI;
-   glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI;
-
-   /*** GLX_SGI_make_current_read ***/
-   glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI;
-   /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/
-
-/*** GLX_SGIX_video_source ***/
-#if defined(_VL_H)
-   glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX;
-   glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX;
-#endif
-
-   /*** GLX_EXT_import_context ***/
-   glx.FreeContextEXT = Fake_glXFreeContextEXT;
-   glx.GetContextIDEXT = Fake_glXGetContextIDEXT;
-   /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/
-   glx.ImportContextEXT = Fake_glXImportContextEXT;
-   glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT;
-
-   /*** GLX_SGIX_fbconfig ***/
-   glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX;
-   glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX;
-   glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX;
-   glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX;
-   glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX;
-   glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX;
-
-   /*** GLX_SGIX_pbuffer ***/
-   glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX;
-   glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX;
-   glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX;
-   glx.SelectEventSGIX = Fake_glXSelectEventSGIX;
-   glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX;
-
-   /*** GLX_SGI_cushion ***/
-   glx.CushionSGI = Fake_glXCushionSGI;
-
-   /*** GLX_SGIX_video_resize ***/
-   glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX;
-   glx.ChannelRectSGIX = Fake_glXChannelRectSGIX;
-   glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX;
-   glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX;
-   glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX;
-
-   /*** GLX_SGIX_dmbuffer **/
-#if defined(_DM_BUFFER_H_)
-   glx.AssociateDMPbufferSGIX = NULL;
-#endif
-
-   /*** GLX_SGIX_swap_group ***/
-   glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX;
-
-   /*** GLX_SGIX_swap_barrier ***/
-   glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX;
-   glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX;
-
-   /*** GLX_SUN_get_transparent_index ***/
-   glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN;
-
-   /*** GLX_MESA_copy_sub_buffer ***/
-   glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA;
-
-   /*** GLX_MESA_release_buffers ***/
-   glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA;
-
-   /*** GLX_MESA_pixmap_colormap ***/
-   glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA;
-
-   /*** GLX_EXT_texture_from_pixmap ***/
-   glx.BindTexImageEXT = Fake_glXBindTexImageEXT;
-   glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT;
-
-   return &glx;
-}
diff --git a/src/gallium/state_trackers/glx/xlib/glx_getproc.c b/src/gallium/state_trackers/glx/xlib/glx_getproc.c
new file mode 100644
index 0000000000..ca7d88c922
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/glx_getproc.c
@@ -0,0 +1,214 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.6
+ * 
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * glXGetProcAddress()
+ */
+
+
+#define GLX_GLXEXT_PROTOTYPES
+
+#include <string.h>
+#include "GL/glx.h"
+#include "glapi/glapi.h"
+
+
+struct name_address_pair {
+   const char *Name;
+   __GLXextFuncPtr Address;
+};
+
+
+static struct name_address_pair GLX_functions[] = {
+   /*** GLX_VERSION_1_0 ***/
+   { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual },
+   { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext },
+   { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext },
+   { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap },
+   { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext },
+   { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap },
+   { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig },
+   { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext },
+   { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable },
+   { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect },
+   { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent },
+   { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension },
+   { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion },
+   { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers },
+   { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont },
+   { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL },
+   { "glXWaitX", (__GLXextFuncPtr) glXWaitX },
+
+   /*** GLX_VERSION_1_1 ***/
+   { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString },
+   { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString },
+   { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString },
+
+   /*** GLX_VERSION_1_2 ***/
+   { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay },
+
+   /*** GLX_VERSION_1_3 ***/
+   { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig },
+   { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext },
+   { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer },
+   { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap },
+   { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow },
+   { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer },
+   { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap },
+   { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow },
+   { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable },
+   { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib },
+   { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs },
+   { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent },
+   { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig },
+   { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent },
+   { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext },
+   { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable },
+   { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent },
+
+   /*** GLX_VERSION_1_4 ***/
+   { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress },
+
+   /*** GLX_SGI_swap_control ***/
+   { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI },
+
+   /*** GLX_SGI_video_sync ***/
+   { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI },
+   { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI },
+
+   /*** GLX_SGI_make_current_read ***/
+   { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI },
+   { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI },
+
+   /*** GLX_SGIX_video_source ***/
+#if defined(_VL_H)
+   { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX },
+   { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX },
+#endif
+
+   /*** GLX_EXT_import_context ***/
+   { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT },
+   { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT },
+   { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT },
+   { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT },
+   { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT },
+
+   /*** GLX_SGIX_fbconfig ***/
+   { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX },
+   { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX },
+   { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX },
+   { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX },
+   { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX },
+   { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX },
+
+   /*** GLX_SGIX_pbuffer ***/
+   { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX },
+   { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX },
+   { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX },
+   { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX },
+   { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX },
+
+   /*** GLX_SGI_cushion ***/
+   { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI },
+
+   /*** GLX_SGIX_video_resize ***/
+   { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX },
+   { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX },
+   { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX },
+   { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX },
+   { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX },
+
+   /*** GLX_SGIX_dmbuffer **/
+#if defined(_DM_BUFFER_H_)
+   { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX },
+#endif
+
+   /*** GLX_SGIX_swap_group ***/
+   { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX },
+
+   /*** GLX_SGIX_swap_barrier ***/
+   { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX },
+   { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX },
+
+   /*** GLX_SUN_get_transparent_index ***/
+   { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN },
+
+   /*** GLX_MESA_copy_sub_buffer ***/
+   { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA },
+
+   /*** GLX_MESA_pixmap_colormap ***/
+   { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA },
+
+   /*** GLX_MESA_release_buffers ***/
+   { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA },
+
+   /*** GLX_ARB_get_proc_address ***/
+   { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB },
+
+   /*** GLX_EXT_texture_from_pixmap ***/
+   { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT },
+   { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT },
+
+   { NULL, NULL }   /* end of list */
+};
+
+
+
+/**
+ * Return address of named glX function, or NULL if not found.
+ */
+static __GLXextFuncPtr
+_glxapi_get_proc_address(const char *funcName)
+{
+   GLuint i;
+   for (i = 0; GLX_functions[i].Name; i++) {
+      if (strcmp(GLX_functions[i].Name, funcName) == 0)
+         return GLX_functions[i].Address;
+   }
+   return NULL;
+}
+
+
+__GLXextFuncPtr
+glXGetProcAddressARB(const GLubyte *procName)
+{
+   __GLXextFuncPtr f;
+
+   f = _glxapi_get_proc_address((const char *) procName);
+   if (f) {
+      return f;
+   }
+
+   f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName);
+   return f;
+}
+
+
+/* GLX 1.4 */
+void (*glXGetProcAddress(const GLubyte *procName))()
+{
+   return glXGetProcAddressARB(procName);
+}
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c b/src/gallium/state_trackers/glx/xlib/glx_usefont.c
index e359046756..acc64df62b 100644
--- a/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c
+++ b/src/gallium/state_trackers/glx/xlib/glx_usefont.c
@@ -1,9 +1,10 @@
-
 /*
  * Mesa 3-D graphics library
- * Version:  3.5
+ * Version:  7.6
  *
- * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -24,13 +25,13 @@
  */
 
 
-/* xfonts.c -- glXUseXFont() for Mesa written by
- * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de
+/**
+ * Fake implementation of glXUseXFont().
  */
 
-#include "context.h"
-#include "imports.h"
-#include "fakeglx.h"
+
+#include "main/context.h"
+#include "main/imports.h"
 #include <GL/glx.h>
 
 
@@ -210,7 +211,7 @@ isvalid(XFontStruct * fs, unsigned int which)
 
 
 void
-Fake_glXUseXFont(Font font, int first, int count, int listbase)
+glXUseXFont(Font font, int first, int count, int listbase)
 {
    Display *dpy;
    Window win;
@@ -228,7 +229,8 @@ Fake_glXUseXFont(Font font, int first, int count, int listbase)
    dpy = glXGetCurrentDisplay();
    if (!dpy)
       return;			/* I guess glXMakeCurrent wasn't called */
-   win = RootWindow(dpy, DefaultScreen(dpy));
+   i = DefaultScreen(dpy);
+   win = RootWindow(dpy, i);
 
    fs = XQueryFont(dpy, font);
    if (!fs) {
diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.c b/src/gallium/state_trackers/glx/xlib/glxapi.c
deleted file mode 100644
index c2cb34d7cf..0000000000
--- a/src/gallium/state_trackers/glx/xlib/glxapi.c
+++ /dev/null
@@ -1,1254 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.1
- * 
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- * 
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-/*
- * This is the GLX API dispatcher.  Calls to the glX* functions are
- * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions.
- * See the glxapi.h file for more details.
- */
-
-
-#include <assert.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include "glapi/glapi.h"
-#include "glxapi.h"
-#include "fakeglx.h"
-#include "pipe/p_thread.h"
-
-
-#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 303
-#  define PUBLIC __attribute__((visibility("default")))
-#  define USED __attribute__((used))
-#else
-#  define PUBLIC
-#  define USED
-#endif
-
-
-struct display_dispatch {
-   Display *Dpy;
-   struct _glxapi_table *Table;
-   struct display_dispatch *Next;
-};
-
-static struct display_dispatch *DispatchList = NULL;
-
-
-/* Display -> Dispatch caching */
-static Display *prevDisplay = NULL;
-static struct _glxapi_table *prevTable = NULL;
-
-
-static struct _glxapi_table *
-get_dispatch(Display *dpy)
-{
-   if (!dpy)
-      return NULL;
-
-   /* search list of display/dispatch pairs for this display */
-   {
-      const struct display_dispatch *d = DispatchList;
-      while (d) {
-         if (d->Dpy == dpy) {
-            prevDisplay = dpy;
-            prevTable = d->Table;
-            return d->Table;  /* done! */
-         }
-         d = d->Next;
-      }
-   }
-
-   /* A new display, determine if we should use real GLX
-    * or Mesa's pseudo-GLX.
-    */
-   {
-      struct _glxapi_table *t = _mesa_GetGLXDispatchTable();
-
-      if (t) {
-         struct display_dispatch *d;
-         d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch));
-         if (d) {
-            d->Dpy = dpy;
-            d->Table = t;
-            /* insert at head of list */
-            d->Next = DispatchList;
-            DispatchList = d;
-            /* update cache */
-            prevDisplay = dpy;
-            prevTable = t;
-            return t;
-         }
-      }
-   }
-
-   /* If we get here that means we can't use real GLX on this display
-    * and the Mesa pseudo-GLX software renderer wasn't compiled in.
-    * Or, we ran out of memory!
-    */
-   return NULL;
-}
-
-
-/* Don't use the GET_DISPATCH defined in glthread.h */
-#undef GET_DISPATCH
-
-#define GET_DISPATCH(DPY, TABLE)	\
-   if (DPY == prevDisplay) {		\
-      TABLE = prevTable;		\
-   }					\
-   else if (!DPY) {			\
-      TABLE = NULL;			\
-   }					\
-   else {				\
-      TABLE = get_dispatch(DPY);	\
-   }
-
-   
-
-
-/**
- * GLX API current context.
- */
-pipe_tsd ContextTSD;
-
-
-static void
-SetCurrentContext(GLXContext c)
-{
-   pipe_tsd_set(&ContextTSD, c);
-}
-
-
-/*
- * GLX API entrypoints
- */
-
-/*** GLX_VERSION_1_0 ***/
-
-XVisualInfo PUBLIC *
-glXChooseVisual(Display *dpy, int screen, int *list)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return NULL;
-   return (t->ChooseVisual)(dpy, screen, list);
-}
-
-
-void PUBLIC
-glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->CopyContext)(dpy, src, dst, mask);
-}
-
-
-GLXContext PUBLIC
-glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateContext)(dpy, visinfo, shareList, direct);
-}
-
-
-GLXPixmap PUBLIC
-glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateGLXPixmap)(dpy, visinfo, pixmap);
-}
-
-
-void PUBLIC
-glXDestroyContext(Display *dpy, GLXContext ctx)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   if (glXGetCurrentContext() == ctx)
-      SetCurrentContext(NULL);
-   (t->DestroyContext)(dpy, ctx);
-}
-
-
-void PUBLIC
-glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->DestroyGLXPixmap)(dpy, pixmap);
-}
-
-
-int PUBLIC
-glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return GLX_NO_EXTENSION;
-   return (t->GetConfig)(dpy, visinfo, attrib, value);
-}
-
-
-GLXContext PUBLIC
-glXGetCurrentContext(void)
-{
-   return (GLXContext) pipe_tsd_get(&ContextTSD);
-}
-
-
-GLXDrawable PUBLIC
-glXGetCurrentDrawable(void)
-{
-   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
-   return gc ? gc->currentDrawable : 0;
-}
-
-
-Bool PUBLIC
-glXIsDirect(Display *dpy, GLXContext ctx)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->IsDirect)(dpy, ctx);
-}
-
-
-Bool PUBLIC
-glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx)
-{
-   Bool b;
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t) {
-      return False;
-   }
-   b = (*t->MakeCurrent)(dpy, drawable, ctx);
-   if (b) {
-      SetCurrentContext(ctx);
-   }
-   return b;
-}
-
-
-Bool PUBLIC
-glXQueryExtension(Display *dpy, int *errorb, int *event)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->QueryExtension)(dpy, errorb, event);
-}
-
-
-Bool PUBLIC
-glXQueryVersion(Display *dpy, int *maj, int *min)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->QueryVersion)(dpy, maj, min);
-}
-
-
-void PUBLIC
-glXSwapBuffers(Display *dpy, GLXDrawable drawable)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->SwapBuffers)(dpy, drawable);
-}
-
-
-void PUBLIC
-glXUseXFont(Font font, int first, int count, int listBase)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->UseXFont)(font, first, count, listBase);
-}
-
-
-void PUBLIC
-glXWaitGL(void)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->WaitGL)();
-}
-
-
-void PUBLIC
-glXWaitX(void)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->WaitX)();
-}
-
-
-
-/*** GLX_VERSION_1_1 ***/
-
-const char PUBLIC *
-glXGetClientString(Display *dpy, int name)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return NULL;
-   return (t->GetClientString)(dpy, name);
-}
-
-
-const char PUBLIC *
-glXQueryExtensionsString(Display *dpy, int screen)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return NULL;
-   return (t->QueryExtensionsString)(dpy, screen);
-}
-
-
-const char PUBLIC *
-glXQueryServerString(Display *dpy, int screen, int name)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return NULL;
-   return (t->QueryServerString)(dpy, screen, name);
-}
-
-
-/*** GLX_VERSION_1_2 ***/
-
-Display PUBLIC *
-glXGetCurrentDisplay(void)
-{
-   /* Same code as in libGL's glxext.c */
-   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
-   if (NULL == gc) return NULL;
-   return gc->currentDpy;
-}
-
-
-
-/*** GLX_VERSION_1_3 ***/
-
-GLXFBConfig PUBLIC *
-glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->ChooseFBConfig)(dpy, screen, attribList, nitems);
-}
-
-
-GLXContext PUBLIC
-glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateNewContext)(dpy, config, renderType, shareList, direct);
-}
-
-
-GLXPbuffer PUBLIC
-glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreatePbuffer)(dpy, config, attribList);
-}
-
-
-GLXPixmap PUBLIC
-glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreatePixmap)(dpy, config, pixmap, attribList);
-}
-
-
-GLXWindow PUBLIC
-glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateWindow)(dpy, config, win, attribList);
-}
-
-
-void PUBLIC
-glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->DestroyPbuffer)(dpy, pbuf);
-}
-
-
-void PUBLIC
-glXDestroyPixmap(Display *dpy, GLXPixmap pixmap)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->DestroyPixmap)(dpy, pixmap);
-}
-
-
-void PUBLIC
-glXDestroyWindow(Display *dpy, GLXWindow window)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->DestroyWindow)(dpy, window);
-}
-
-
-GLXDrawable PUBLIC
-glXGetCurrentReadDrawable(void)
-{
-   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
-   return gc ? gc->currentReadable : 0;
-}
-
-
-int PUBLIC
-glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return GLX_NO_EXTENSION;
-   return (t->GetFBConfigAttrib)(dpy, config, attribute, value);
-}
-
-
-GLXFBConfig PUBLIC *
-glXGetFBConfigs(Display *dpy, int screen, int *nelements)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->GetFBConfigs)(dpy, screen, nelements);
-}
-
-void PUBLIC
-glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->GetSelectedEvent)(dpy, drawable, mask);
-}
-
-
-XVisualInfo PUBLIC *
-glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return NULL;
-   return (t->GetVisualFromFBConfig)(dpy, config);
-}
-
-
-Bool PUBLIC
-glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
-{
-   Bool b;
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   b = (t->MakeContextCurrent)(dpy, draw, read, ctx);
-   if (b) {
-      SetCurrentContext(ctx);
-   }
-   return b;
-}
-
-
-int PUBLIC
-glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   assert(t);
-   if (!t)
-      return 0; /* XXX correct? */
-   return (t->QueryContext)(dpy, ctx, attribute, value);
-}
-
-
-void PUBLIC
-glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->QueryDrawable)(dpy, draw, attribute, value);
-}
-
-
-void PUBLIC
-glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->SelectEvent)(dpy, drawable, mask);
-}
-
-
-
-/*** GLX_SGI_swap_control ***/
-
-int PUBLIC
-glXSwapIntervalSGI(int interval)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->SwapIntervalSGI)(interval);
-}
-
-
-
-/*** GLX_SGI_video_sync ***/
-
-int PUBLIC
-glXGetVideoSyncSGI(unsigned int *count)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t || !glXGetCurrentContext())
-      return GLX_BAD_CONTEXT;
-   return (t->GetVideoSyncSGI)(count);
-}
-
-int PUBLIC
-glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
-{
-   struct _glxapi_table *t;
-   Display *dpy = glXGetCurrentDisplay();
-   GET_DISPATCH(dpy, t);
-   if (!t || !glXGetCurrentContext())
-      return GLX_BAD_CONTEXT;
-   return (t->WaitVideoSyncSGI)(divisor, remainder, count);
-}
-
-
-
-/*** GLX_SGI_make_current_read ***/
-
-Bool PUBLIC
-glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx);
-}
-
-GLXDrawable PUBLIC
-glXGetCurrentReadDrawableSGI(void)
-{
-   return glXGetCurrentReadDrawable();
-}
-
-
-#if defined(_VL_H)
-
-GLXVideoSourceSGIX PUBLIC
-glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode);
-}
-
-void PUBLIC
-glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->DestroyGLXVideoSourceSGIX)(dpy, src);
-}
-
-#endif
-
-
-/*** GLX_EXT_import_context ***/
-
-void PUBLIC
-glXFreeContextEXT(Display *dpy, GLXContext context)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->FreeContextEXT)(dpy, context);
-}
-
-GLXContextID PUBLIC
-glXGetContextIDEXT(const GLXContext context)
-{
-   return ((__GLXcontext *) context)->xid;
-}
-
-Display PUBLIC *
-glXGetCurrentDisplayEXT(void)
-{
-   return glXGetCurrentDisplay();
-}
-
-GLXContext PUBLIC
-glXImportContextEXT(Display *dpy, GLXContextID contextID)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->ImportContextEXT)(dpy, contextID);
-}
-
-int PUBLIC
-glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;  /* XXX ok? */
-   return (t->QueryContextInfoEXT)(dpy, context, attribute, value);
-}
-
-
-
-/*** GLX_SGIX_fbconfig ***/
-
-int PUBLIC
-glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value);
-}
-
-GLXFBConfigSGIX PUBLIC *
-glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements);
-}
-
-GLXPixmap PUBLIC
-glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap);
-}
-
-GLXContext PUBLIC
-glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct);
-}
-
-XVisualInfo PUBLIC *
-glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->GetVisualFromFBConfigSGIX)(dpy, config);
-}
-
-GLXFBConfigSGIX PUBLIC
-glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->GetFBConfigFromVisualSGIX)(dpy, vis);
-}
-
-
-
-/*** GLX_SGIX_pbuffer ***/
-
-GLXPbufferSGIX PUBLIC
-glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list);
-}
-
-void PUBLIC
-glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->DestroyGLXPbufferSGIX)(dpy, pbuf);
-}
-
-int PUBLIC
-glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value);
-}
-
-void PUBLIC
-glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->SelectEventSGIX)(dpy, drawable, mask);
-}
-
-void PUBLIC
-glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->GetSelectedEventSGIX)(dpy, drawable, mask);
-}
-
-
-
-/*** GLX_SGI_cushion ***/
-
-void PUBLIC
-glXCushionSGI(Display *dpy, Window win, float cushion)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->CushionSGI)(dpy, win, cushion);
-}
-
-
-
-/*** GLX_SGIX_video_resize ***/
-
-int PUBLIC
-glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window);
-}
-
-int PUBLIC
-glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h);
-}
-
-int PUBLIC
-glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h);
-}
-
-int PUBLIC
-glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh);
-}
-
-int PUBLIC
-glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype);
-}
-
-
-
-#if defined(_DM_BUFFER_H_)
-
-Bool PUBLIC
-glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer);
-}
-
-#endif
-
-
-/*** GLX_SGIX_swap_group ***/
-
-void PUBLIC
-glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (*t->JoinSwapGroupSGIX)(dpy, drawable, member);
-}
-
-
-/*** GLX_SGIX_swap_barrier ***/
-
-void PUBLIC
-glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier);
-}
-
-Bool PUBLIC
-glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max);
-}
-
-
-
-/*** GLX_SUN_get_transparent_index ***/
-
-Status PUBLIC
-glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent);
-}
-
-
-
-/*** GLX_MESA_copy_sub_buffer ***/
-
-void PUBLIC
-glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return;
-   (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height);
-}
-
-
-
-/*** GLX_MESA_release_buffers ***/
-
-Bool PUBLIC
-glXReleaseBuffersMESA(Display *dpy, Window w)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return False;
-   return (t->ReleaseBuffersMESA)(dpy, w);
-}
-
-
-
-/*** GLX_MESA_pixmap_colormap ***/
-
-GLXPixmap PUBLIC
-glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (!t)
-      return 0;
-   return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap);
-}
-
-/*** GLX_EXT_texture_from_pixmap */
-
-void
-glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
-                   const int *attrib_list)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (t)
-      t->BindTexImageEXT(dpy, drawable, buffer, attrib_list);
-}
-
-void
-glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
-{
-   struct _glxapi_table *t;
-   GET_DISPATCH(dpy, t);
-   if (t)
-      t->ReleaseTexImageEXT(dpy, drawable, buffer);
-}
-
-
-/**********************************************************************/
-/* GLX API management functions                                       */
-/**********************************************************************/
-
-
-const char *
-_glxapi_get_version(void)
-{
-   return "1.3";
-}
-
-
-
-/*
- * Return size of the GLX dispatch table, in entries, not bytes.
- */
-GLuint
-_glxapi_get_dispatch_table_size(void)
-{
-   return sizeof(struct _glxapi_table) / sizeof(void *);
-}
-
-
-static int
-generic_no_op_func(void)
-{
-   return 0;
-}
-
-
-/*
- * Initialize all functions in given dispatch table to be no-ops
- */
-void
-_glxapi_set_no_op_table(struct _glxapi_table *t)
-{
-   typedef int (*nop_func)(void);
-   nop_func *dispatch = (nop_func *) t;
-   GLuint n = _glxapi_get_dispatch_table_size();
-   GLuint i;
-   for (i = 0; i < n; i++) {
-      dispatch[i] = generic_no_op_func;
-   }
-}
-
-
-struct name_address_pair {
-   const char *Name;
-   __GLXextFuncPtr Address;
-};
-
-static struct name_address_pair GLX_functions[] = {
-   /*** GLX_VERSION_1_0 ***/
-   { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual },
-   { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext },
-   { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext },
-   { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap },
-   { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext },
-   { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap },
-   { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig },
-   { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext },
-   { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable },
-   { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect },
-   { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent },
-   { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension },
-   { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion },
-   { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers },
-   { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont },
-   { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL },
-   { "glXWaitX", (__GLXextFuncPtr) glXWaitX },
-
-   /*** GLX_VERSION_1_1 ***/
-   { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString },
-   { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString },
-   { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString },
-
-   /*** GLX_VERSION_1_2 ***/
-   { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay },
-
-   /*** GLX_VERSION_1_3 ***/
-   { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig },
-   { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext },
-   { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer },
-   { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap },
-   { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow },
-   { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer },
-   { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap },
-   { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow },
-   { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable },
-   { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib },
-   { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs },
-   { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent },
-   { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig },
-   { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent },
-   { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext },
-   { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable },
-   { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent },
-
-   /*** GLX_VERSION_1_4 ***/
-   { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress },
-
-   /*** GLX_SGI_swap_control ***/
-   { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI },
-
-   /*** GLX_SGI_video_sync ***/
-   { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI },
-   { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI },
-
-   /*** GLX_SGI_make_current_read ***/
-   { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI },
-   { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI },
-
-   /*** GLX_SGIX_video_source ***/
-#if defined(_VL_H)
-   { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX },
-   { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX },
-#endif
-
-   /*** GLX_EXT_import_context ***/
-   { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT },
-   { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT },
-   { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT },
-   { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT },
-   { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT },
-
-   /*** GLX_SGIX_fbconfig ***/
-   { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX },
-   { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX },
-   { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX },
-   { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX },
-   { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX },
-   { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX },
-
-   /*** GLX_SGIX_pbuffer ***/
-   { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX },
-   { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX },
-   { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX },
-   { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX },
-   { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX },
-
-   /*** GLX_SGI_cushion ***/
-   { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI },
-
-   /*** GLX_SGIX_video_resize ***/
-   { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX },
-   { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX },
-   { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX },
-   { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX },
-   { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX },
-
-   /*** GLX_SGIX_dmbuffer **/
-#if defined(_DM_BUFFER_H_)
-   { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX },
-#endif
-
-   /*** GLX_SGIX_swap_group ***/
-   { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX },
-
-   /*** GLX_SGIX_swap_barrier ***/
-   { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX },
-   { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX },
-
-   /*** GLX_SUN_get_transparent_index ***/
-   { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN },
-
-   /*** GLX_MESA_copy_sub_buffer ***/
-   { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA },
-
-   /*** GLX_MESA_pixmap_colormap ***/
-   { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA },
-
-   /*** GLX_MESA_release_buffers ***/
-   { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA },
-
-   /*** GLX_ARB_get_proc_address ***/
-   { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB },
-
-   /*** GLX_EXT_texture_from_pixmap ***/
-   { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT },
-   { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT },
-
-   { NULL, NULL }   /* end of list */
-};
-
-
-
-/*
- * Return address of named glX function, or NULL if not found.
- */
-__GLXextFuncPtr
-_glxapi_get_proc_address(const char *funcName)
-{
-   GLuint i;
-   for (i = 0; GLX_functions[i].Name; i++) {
-      if (strcmp(GLX_functions[i].Name, funcName) == 0)
-         return GLX_functions[i].Address;
-   }
-   return NULL;
-}
-
-
-
-/*
- * This function does not get dispatched through the dispatch table
- * since it's really a "meta" function.
- */
-__GLXextFuncPtr
-glXGetProcAddressARB(const GLubyte *procName)
-{
-   __GLXextFuncPtr f;
-
-   f = _glxapi_get_proc_address((const char *) procName);
-   if (f) {
-      return f;
-   }
-
-   f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName);
-   return f;
-}
-
-
-/* GLX 1.4 */
-void (*glXGetProcAddress(const GLubyte *procName))()
-{
-   return glXGetProcAddressARB(procName);
-}
diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.h b/src/gallium/state_trackers/glx/xlib/glxapi.h
deleted file mode 100644
index b4e12b4162..0000000000
--- a/src/gallium/state_trackers/glx/xlib/glxapi.h
+++ /dev/null
@@ -1,213 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- * 
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- * 
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-#ifndef _glxapi_h_
-#define _glxapi_h_
-
-
-#define GLX_GLXEXT_PROTOTYPES
-#include "GL/glx.h"
-
-
-/* The GLX API dispatcher (i.e. this code) is being built into stand-alone
- * Mesa.  We don't know anything about XFree86 or real GLX so we define a
- * minimal __GLXContextRec here so some of the functions in this file can
- * work properly.
- */
-typedef struct __GLXcontextRec {
-   Display *currentDpy;
-   GLboolean isDirect;
-   GLXDrawable currentDrawable;
-   GLXDrawable currentReadable;
-   XID xid;
-} __GLXcontext;
-
-
-/*
- * Almost all the GLX API functions get routed through this dispatch table.
- * The exceptions are the glXGetCurrentXXX() functions.
- *
- * This dispatch table allows multiple GLX client-side modules to coexist.
- * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's
- * pseudo-GLX can be present at the same time.  The former being used on
- * GLX-enabled X servers and the later on non-GLX X servers.
- *
- * Red Hat has been using this since Red Hat Linux 7.0 (I think).
- * This'll be a standard feature in XFree86 4.3.  It basically allows one
- * libGL to do both DRI-rendering and "fake GLX" rendering to X displays
- * that lack the GLX extension.
- */
-struct _glxapi_table {
-   /*** GLX_VERSION_1_0 ***/
-   XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list);
-   void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask);
-   GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct);
-   GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap);
-   void (*DestroyContext)(Display *dpy, GLXContext ctx);
-   void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap);
-   int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value);
-   /*GLXContext (*GetCurrentContext)(void);*/
-   /*GLXDrawable (*GetCurrentDrawable)(void);*/
-   Bool (*IsDirect)(Display *dpy, GLXContext ctx);
-   Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx);
-   Bool (*QueryExtension)(Display *dpy, int *errorb, int *event);
-   Bool (*QueryVersion)(Display *dpy, int *maj, int *min);
-   void (*SwapBuffers)(Display *dpy, GLXDrawable drawable);
-   void (*UseXFont)(Font font, int first, int count, int listBase);
-   void (*WaitGL)(void);
-   void (*WaitX)(void);
-
-   /*** GLX_VERSION_1_1 ***/
-   const char *(*GetClientString)(Display *dpy, int name);
-   const char *(*QueryExtensionsString)(Display *dpy, int screen);
-   const char *(*QueryServerString)(Display *dpy, int screen, int name);
-
-   /*** GLX_VERSION_1_2 ***/
-   /*Display *(*GetCurrentDisplay)(void);*/
-
-   /*** GLX_VERSION_1_3 ***/
-   GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems);
-   GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct);
-   GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList);
-   GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList);
-   GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList);
-   void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf);
-   void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap);
-   void (*DestroyWindow)(Display *dpy, GLXWindow window);
-   /*GLXDrawable (*GetCurrentReadDrawable)(void);*/
-   int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value);
-   GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements);
-   void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask);
-   XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config);
-   Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx);
-   int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value);
-   void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value);
-   void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask);
-
-   /*** GLX_SGI_swap_control ***/
-   int (*SwapIntervalSGI)(int);
-
-   /*** GLX_SGI_video_sync ***/
-   int (*GetVideoSyncSGI)(unsigned int *count);
-   int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count);
-
-   /*** GLX_SGI_make_current_read ***/
-   Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext);
-   /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/
-
-   /*** GLX_SGIX_video_source (needs video library) ***/
-#if defined(_VL_H_)
-   GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode);
-   void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX);
-#else
-   void *CreateGLXVideoSourceSGIX;
-   void *DestroyGLXVideoSourceSGIX;
-#endif
-
-   /*** GLX_EXT_import_context ***/
-   void (*FreeContextEXT)(Display *dpy, GLXContext context);
-   GLXContextID (*GetContextIDEXT)(const GLXContext context);
-   /*Display *(*GetCurrentDisplayEXT)(void);*/
-   GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID);
-   int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value);
-
-   /*** GLX_SGIX_fbconfig ***/
-   int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *);
-   GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *);
-   GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap);
-   GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool);
-   XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX);
-   GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *);
-
-   /*** GLX_SGIX_pbuffer ***/
-   GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *);
-   void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX);
-   int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *);
-   void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long);
-   void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *);
-
-   /*** GLX_SGI_cushion ***/
-   void (*CushionSGI)(Display *, Window, float);
-
-   /*** GLX_SGIX_video_resize ***/
-   int (*BindChannelToWindowSGIX)(Display *, int, int, Window);
-   int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int);
-   int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *);
-   int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *);
-   int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum);
-
-   /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/
-#if defined (_DM_BUFFER_H_)
-   Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer);
-#else
-   void *AssociciateDMPbufferSGIX;
-#endif
-
-   /*** GLX_SGIX_swap_group ***/
-   void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable);
-
-   /*** GLX_SGIX_swap_barrier ***/
-   void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int);
-   Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *);
-
-   /*** GLX_SUN_get_transparent_index ***/
-   Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *);
-
-   /*** GLX_MESA_copy_sub_buffer ***/
-   void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height);
-
-   /*** GLX_MESA_release_buffers ***/
-   Bool (*ReleaseBuffersMESA)(Display *dpy, Window w);
-
-   /*** GLX_MESA_pixmap_colormap ***/
-   GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap);
-
-   /*** GLX_EXT_texture_from_pixmap ***/
-   void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer,
-                           const int *attrib_list);
-   void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer);
-};
-
-
-
-extern const char *
-_glxapi_get_version(void);
-
-
-
-
-extern GLuint
-_glxapi_get_dispatch_table_size(void);
-
-
-extern void
-_glxapi_set_no_op_table(struct _glxapi_table *t);
-
-
-extern __GLXextFuncPtr
-_glxapi_get_proc_address(const char *funcName);
-
-
-#endif
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c
index 45ad694e23..957002ddd5 100644
--- a/src/gallium/state_trackers/glx/xlib/xm_api.c
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.c
@@ -286,6 +286,18 @@ choose_pixel_format(XMesaVisual v)
          return PIPE_FORMAT_B8G8R8A8_UNORM;
       }
    }
+   else if (   GET_REDMASK(v)   == 0x0000ff00
+            && GET_GREENMASK(v) == 0x00ff0000
+            && GET_BLUEMASK(v)  == 0xff000000
+            && v->BitsPerPixel == 32) {
+      if (native_byte_order) {
+         /* no byteswapping needed */
+         return PIPE_FORMAT_B8G8R8A8_UNORM;
+      }
+      else {
+         return PIPE_FORMAT_A8R8G8B8_UNORM;
+      }
+   }
    else if (   GET_REDMASK(v)   == 0xf800
             && GET_GREENMASK(v) == 0x07e0
             && GET_BLUEMASK(v)  == 0x001f
@@ -656,7 +668,7 @@ XMesaVisual XMesaCreateVisual( Display *display,
     * at a later time.
     */
    v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo));
-   if(!v->visinfo) {
+   if (!v->visinfo) {
       _mesa_free(v);
       return NULL;
    }
@@ -742,7 +754,7 @@ PUBLIC
 XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
 {
    static GLboolean firstTime = GL_TRUE;
-   struct pipe_screen *screen;
+   static struct pipe_screen *screen = NULL;
    struct pipe_context *pipe = NULL;
    XMesaContext c;
    GLcontext *mesaCtx;
@@ -750,6 +762,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
 
    if (firstTime) {
       pipe_mutex_init(_xmesa_lock);
+      screen = driver.create_pipe_screen();
       firstTime = GL_FALSE;
    }
 
@@ -765,14 +778,10 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    c->xm_buffer = NULL;   /* set later by XMesaMakeCurrent */
    c->xm_read_buffer = NULL;
 
-   /* XXX: create once per Xlib Display.
-    */
-   screen = driver.create_pipe_screen();
    if (screen == NULL)
       goto fail;
 
-   pipe = driver.create_pipe_context( screen,
-                                     (void *)c );
+   pipe = driver.create_pipe_context(screen, (void *) c);
    if (pipe == NULL)
       goto fail;
 
@@ -785,26 +794,15 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    mesaCtx = c->st->ctx;
    c->st->ctx->DriverCtx = c;
 
-#if 00
-   _mesa_enable_sw_extensions(mesaCtx);
-   _mesa_enable_1_3_extensions(mesaCtx);
-   _mesa_enable_1_4_extensions(mesaCtx);
-   _mesa_enable_1_5_extensions(mesaCtx);
-   _mesa_enable_2_0_extensions(mesaCtx);
-#endif
-
    return c;
 
- fail:
+fail:
    if (c->st)
       st_destroy_context(c->st);
    else if (pipe)
       pipe->destroy(pipe);
 
-   if (screen)
-      screen->destroy( screen );
-
-   FREE(c);
+   _mesa_free(c);
    return NULL;
 }
 
@@ -1100,26 +1098,19 @@ XMesaContext XMesaGetCurrentContext( void )
 
 
 
-
-
-
-/*
- * Copy the back buffer to the front buffer.  If there's no back buffer
- * this is a no-op.
+/**
+ * Swap front and back color buffers and have winsys display front buffer.
+ * If there's no front color buffer no swap actually occurs.
  */
 PUBLIC
 void XMesaSwapBuffers( XMesaBuffer b )
 {
-   struct pipe_surface *surf;
+   struct pipe_surface *frontLeftSurf;
 
-   /* If we're swapping the buffer associated with the current context
-    * we have to flush any pending rendering commands first.
-    */
-   st_notify_swapbuffers(b->stfb);
+   st_swapbuffers(b->stfb, &frontLeftSurf, NULL);
 
-   st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf);
-   if (surf) {
-      driver.display_surface(b, surf);
+   if (frontLeftSurf) {
+      driver.display_surface(b, frontLeftSurf);
    }
 
    xmesa_check_and_update_buffer_size(NULL, b);
@@ -1165,7 +1156,7 @@ void XMesaFlush( XMesaContext c )
 XMesaBuffer XMesaFindBuffer( Display *dpy, Drawable d )
 {
    XMesaBuffer b;
-   for (b=XMesaBufferList; b; b=b->Next) {
+   for (b = XMesaBufferList; b; b = b->Next) {
       if (b->drawable == d && b->xm_visual->display == dpy) {
          return b;
       }
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.h b/src/gallium/state_trackers/glx/xlib/xm_api.h
index bdd434cd36..ce97a3ec76 100644
--- a/src/gallium/state_trackers/glx/xlib/xm_api.h
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.h
@@ -57,7 +57,7 @@ and create a window, you must do the following to use the X/Mesa interface:
 #define XMESA_H
 
 
-#include "mtypes.h"
+#include "main/mtypes.h"
 #include "state_tracker/st_context.h"
 #include "state_tracker/st_public.h"
 #include "pipe/p_thread.h"
diff --git a/src/gallium/state_trackers/vega/Makefile b/src/gallium/state_trackers/vega/Makefile
new file mode 100644
index 0000000000..b8c805b06c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/Makefile
@@ -0,0 +1,128 @@
+# src/mesa/Makefile
+
+TOP = ../../../..
+include $(TOP)/configs/current
+GALLIUM = $(TOP)
+
+### Lists of source files, included by Makefiles
+
+VG_SOURCES = \
+           api_context.c   \
+           api_filters.c   \
+           api_images.c    \
+           api_masks.c     \
+           api_misc.c      \
+           api_paint.c     \
+           api_params.c    \
+           api_path.c      \
+           api_text.c      \
+           api_transform.c \
+           vgu.c        \
+           vg_context.c \
+           vg_state.c   \
+           vg_tracker.c \
+           vg_translate.c \
+           polygon.c    \
+           bezier.c     \
+           path.c       \
+           paint.c      \
+           arc.c \
+           image.c \
+           renderer.c \
+           stroker.c \
+           mask.c \
+           shader.c \
+           shaders_cache.c
+
+
+### All the core C sources
+
+ALL_SOURCES = \
+        $(VG_SOURCES)
+
+
+### Object files
+VG_OBJECTS = \
+	$(VG_SOURCES:.c=.o)
+
+### Include directories
+
+INCLUDE_DIRS = \
+	-I$(TOP)/include \
+	-I$(GALLIUM)/include \
+	-I$(GALLIUM)/src/gallium/include \
+	-I$(GALLIUM)/src/gallium/auxiliary
+
+VG_LIB = OpenVG
+VG_LIB_NAME = lib$(VG_LIB).so
+
+VG_MAJOR = 1
+VG_MINOR = 0
+VG_TINY = 0
+
+GALLIUM_LIBS = \
+	$(GALLIUM)/src/gallium/auxiliary/pipebuffer/libpipebuffer.a \
+	$(GALLIUM)/src/gallium/auxiliary/sct/libsct.a \
+	$(GALLIUM)/src/gallium/auxiliary/draw/libdraw.a \
+	$(GALLIUM)/src/gallium/auxiliary/rtasm/librtasm.a \
+	$(GALLIUM)/src/gallium/auxiliary/translate/libtranslate.a \
+	$(GALLIUM)/src/gallium/auxiliary/cso_cache/libcso_cache.a \
+	$(GALLIUM)/src/gallium/auxiliary/util/libutil.a \
+	$(GALLIUM)/src/gallium/auxiliary/tgsi/libtgsi.a
+
+.SUFFIXES : .cpp
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+default: depend subdirs $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME)
+
+# Make the OpenVG library
+$(TOP)/$(LIB_DIR)/$(VG_LIB_NAME): $(VG_OBJECTS) $(GALLIUM_LIBS)
+	$(TOP)/bin/mklib -o $(VG_LIB) \
+		-major $(VG_MAJOR) \
+		-minor $(VG_MINOR) \
+		-patch $(VG_TINY) \
+		-install $(TOP)/$(LIB_DIR) \
+		$(VG_OBJECTS) $(GALLIUM_LIBS) \
+		-Wl,--whole-archive $(LIBS) -Wl,--no-whole-archive $(SYS_LIBS)
+
+######################################################################
+# Generic stuff
+
+depend: $(ALL_SOURCES)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend  # workaround oops on gutsy?!?
+	@ touch depend
+	@ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \
+		> /dev/null 2>/dev/null
+
+
+subdirs:
+
+install: default
+	$(INSTALL) -d $(INSTALL_DIR)/include/VG
+	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
+	$(INSTALL) -m 644 $(TOP)/include/VG/*.h $(INSTALL_DIR)/include/VG
+	@if [ -e $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME) ]; then \
+		$(INSTALL) $(TOP)/$(LIB_DIR)/libOpenVG* $(INSTALL_DIR)/$(LIB_DIR); \
+	fi
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` $(TOP)/include/VG/*.h
+
+clean:
+	-rm -f *.o
+	-rm -f */*.o
+	-rm -f */*/*.o
+	-rm -f depend depend.bak
+
+include depend
diff --git a/src/gallium/state_trackers/vega/api_consts.h b/src/gallium/state_trackers/vega/api_consts.h
new file mode 100644
index 0000000000..e1b48d4a46
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_consts.h
@@ -0,0 +1,56 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef API_CONSTS_H
+#define API_CONSTS_H
+
+/*must be at least 32*/
+#define VEGA_MAX_SCISSOR_RECTS 32
+
+/*must be at least 16*/
+#define VEGA_MAX_DASH_COUNT 32
+
+/*must be at least 7*/
+#define VEGA_MAX_KERNEL_SIZE 7
+
+/*must be at least 15*/
+#define VEGA_MAX_SEPARABLE_KERNEL_SIZE 15
+
+/*must be at least 32*/
+#define VEGA_MAX_COLOR_RAMP_STOPS 256
+
+#define VEGA_MAX_IMAGE_WIDTH 2048
+
+#define VEGA_MAX_IMAGE_HEIGHT 2048
+
+#define VEGA_MAX_IMAGE_PIXELS (2048*2048)
+
+#define VEGA_MAX_IMAGE_BYTES (2048*2048 * 4)
+
+/*must be at least 128*/
+#define VEGA_MAX_GAUSSIAN_STD_DEVIATION 128
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_context.c b/src/gallium/state_trackers/vega/api_context.c
new file mode 100644
index 0000000000..47db102dd2
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_context.c
@@ -0,0 +1,75 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+
+VGErrorCode vgGetError(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGErrorCode error = VG_NO_CONTEXT_ERROR;
+
+   if (!ctx)
+      return error;
+
+   error = ctx->_error;
+   ctx->_error = VG_NO_ERROR;
+
+   return error;
+}
+
+void vgFlush(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe;
+
+   if (!ctx)
+      return;
+
+   pipe = ctx->pipe;
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+}
+
+void vgFinish(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_fence_handle *fence = NULL;
+   struct pipe_context *pipe;
+
+   if (!ctx)
+      return;
+
+   pipe = ctx->pipe;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence);
+
+   pipe->screen->fence_finish(pipe->screen, fence, 0);
+   pipe->screen->fence_reference(pipe->screen, &fence, NULL);
+}
diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c
new file mode 100644
index 0000000000..862cbb03c4
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_filters.c
@@ -0,0 +1,805 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "image.h"
+#include "renderer.h"
+#include "shaders_cache.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+
+
+#include "asm_filters.h"
+
+
+struct filter_info {
+   struct vg_image *dst;
+   struct vg_image *src;
+   struct vg_shader * (*setup_shader)(struct vg_context *, void *);
+   void *user_data;
+   const void *const_buffer;
+   VGint const_buffer_len;
+   VGTilingMode tiling_mode;
+   struct pipe_texture *extra_texture;
+};
+
+static INLINE struct pipe_texture *create_texture_1d(struct vg_context *ctx,
+                                                     const VGuint *color_data,
+                                                     const VGint color_data_len)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *tex = 0;
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_1D;
+   templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templ.last_level = 0;
+   templ.width[0] = color_data_len;
+   templ.height[0] = 1;
+   templ.depth[0] = 1;
+   pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   tex = screen->texture_create(screen, &templ);
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer =
+         screen->get_tex_transfer(screen, tex,
+                                  0, 0, 0,
+                                  PIPE_TRANSFER_READ_WRITE ,
+                                  0, 0, tex->width[0], tex->height[0]);
+      void *map = screen->transfer_map(screen, transfer);
+      memcpy(map, color_data, sizeof(VGint)*color_data_len);
+      screen->transfer_unmap(screen, transfer);
+      screen->tex_transfer_destroy(transfer);
+   }
+
+   return tex;
+}
+
+static INLINE struct pipe_surface * setup_framebuffer(struct vg_image *dst)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_framebuffer_state fb;
+   struct pipe_surface *dst_surf = pipe->screen->get_tex_surface(
+      pipe->screen, dst->texture, 0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width  = dst->x + dst_surf->width;
+   fb.height = dst->y + dst_surf->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst_surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso_context, &fb);
+
+   return dst_surf;
+}
+
+static void setup_viewport(struct vg_image *dst)
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_set_viewport(ctx, VEGA_Y0_TOP);
+}
+
+static void setup_blend()
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_blend_state blend;
+   memset(&blend, 0, sizeof(blend));
+   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   if (ctx->state.vg.filter_channel_mask & VG_RED)
+      blend.colormask |= PIPE_MASK_R;
+   if (ctx->state.vg.filter_channel_mask & VG_GREEN)
+      blend.colormask |= PIPE_MASK_G;
+   if (ctx->state.vg.filter_channel_mask & VG_BLUE)
+      blend.colormask |= PIPE_MASK_B;
+   if (ctx->state.vg.filter_channel_mask & VG_ALPHA)
+      blend.colormask |= PIPE_MASK_A;
+   blend.blend_enable = 1;
+   cso_set_blend(ctx->cso_context, &blend);
+}
+
+static void setup_constant_buffer(struct vg_context *ctx, const void *buffer,
+                                  VGint param_bytes)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_constant_buffer *cbuf = &ctx->filter.buffer;
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization. */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                    0, param_bytes, buffer);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static void setup_samplers(struct vg_context *ctx, struct filter_info *info)
+{
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_state sampler[3];
+   int num_samplers = 0;
+   int num_textures = 0;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   samplers[3] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+   textures[3] = NULL;
+
+   memset(&sampler[0], 0, sizeof(struct pipe_sampler_state));
+   sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].min_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
+   sampler[0].mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
+   sampler[0].normalized_coords = 1;
+
+   switch(info->tiling_mode) {
+   case VG_TILE_FILL:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      memcpy(sampler[0].border_color,
+             ctx->state.vg.tile_fill_color,
+             sizeof(VGfloat) * 4);
+      break;
+   case VG_TILE_PAD:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_TILE_REPEAT:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_REPEAT;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_TILE_REFLECT:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   default:
+      debug_assert(!"Unknown tiling mode");
+   }
+
+   samplers[0] = &sampler[0];
+   textures[0] = info->src->texture;
+   ++num_samplers;
+   ++num_textures;
+
+   if (info->extra_texture) {
+      memcpy(&sampler[1], &sampler[0], sizeof(struct pipe_sampler_state));
+      samplers[1] = &sampler[1];
+      textures[1] = info->extra_texture;
+      ++num_samplers;
+      ++num_textures;
+   }
+
+
+   cso_set_samplers(ctx->cso_context, num_samplers, (const struct pipe_sampler_state **)samplers);
+   cso_set_sampler_textures(ctx->cso_context, num_textures, textures);
+}
+
+static struct vg_shader * setup_color_matrix(struct vg_context *ctx, void *user_data)
+{
+   struct vg_shader *shader =
+      shader_create_from_text(ctx->pipe, color_matrix_asm, 200,
+         PIPE_SHADER_FRAGMENT);
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static struct vg_shader * setup_convolution(struct vg_context *ctx, void *user_data)
+{
+   char buffer[1024];
+   VGint num_consts = (VGint)(long)(user_data);
+   struct vg_shader *shader;
+
+   snprintf(buffer, 1023, convolution_asm, num_consts, num_consts / 2 + 1);
+
+   shader = shader_create_from_text(ctx->pipe, buffer, 200,
+                                    PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static struct vg_shader * setup_lookup(struct vg_context *ctx, void *user_data)
+{
+   struct vg_shader *shader =
+      shader_create_from_text(ctx->pipe, lookup_asm,
+                              200, PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+
+static struct vg_shader * setup_lookup_single(struct vg_context *ctx, void *user_data)
+{
+   char buffer[1024];
+   VGImageChannel channel = (VGImageChannel)(user_data);
+   struct vg_shader *shader;
+
+   switch(channel) {
+   case VG_RED:
+      snprintf(buffer, 1023, lookup_single_asm, "xxxx");
+      break;
+   case VG_GREEN:
+      snprintf(buffer, 1023, lookup_single_asm, "yyyy");
+      break;
+   case VG_BLUE:
+      snprintf(buffer, 1023, lookup_single_asm, "zzzz");
+      break;
+   case VG_ALPHA:
+      snprintf(buffer, 1023, lookup_single_asm, "wwww");
+      break;
+   default:
+      debug_assert(!"Unknown color channel");
+   }
+
+   shader = shader_create_from_text(ctx->pipe, buffer, 200,
+                                    PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static void execute_filter(struct vg_context *ctx,
+                           struct filter_info *info)
+{
+   struct pipe_surface *dst_surf;
+   struct vg_shader *shader;
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_samplers(ctx->cso_context);
+   cso_save_sampler_textures(ctx->cso_context);
+
+   dst_surf = setup_framebuffer(info->dst);
+   setup_viewport(info->dst);
+   setup_blend();
+   setup_constant_buffer(ctx, info->const_buffer, info->const_buffer_len);
+   shader = info->setup_shader(ctx, info->user_data);
+   setup_samplers(ctx, info);
+
+   renderer_draw_texture(ctx->renderer,
+                         info->src->texture,
+                         info->dst->x, info->dst->y,
+                         info->dst->x + info->dst->width,
+                         info->dst->y + info->dst->height,
+                         info->dst->x, info->dst->y,
+                         info->dst->x + info->dst->width,
+                         info->dst->y + info->dst->height);
+
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_samplers(ctx->cso_context);
+   cso_restore_sampler_textures(ctx->cso_context);
+
+   vg_shader_destroy(ctx, shader);
+
+   pipe_surface_reference(&dst_surf, NULL);
+}
+
+void vgColorMatrix(VGImage dst, VGImage src,
+                   const VGfloat * matrix)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!matrix || !is_aligned(matrix)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_color_matrix;
+   info.user_data = NULL;
+   info.const_buffer = matrix;
+   info.const_buffer_len = 20 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+}
+
+static VGfloat texture_offset(VGfloat width, VGint kernelSize, VGint current, VGint shift)
+{
+   VGfloat diff = current - shift;
+
+   return diff / width;
+}
+
+void vgConvolve(VGImage dst, VGImage src,
+                VGint kernelWidth, VGint kernelHeight,
+                VGint shiftX, VGint shiftY,
+                const VGshort * kernel,
+                VGfloat scale,
+                VGfloat bias,
+                VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat *buffer;
+   VGint buffer_len;
+   VGint i, j;
+   VGint idx = 0;
+   struct vg_image *d, *s;
+   VGint kernel_size = kernelWidth * kernelHeight;
+   struct filter_info info;
+   const VGint max_kernel_size = vgGeti(VG_MAX_KERNEL_SIZE);
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (kernelWidth <= 0 || kernelHeight <= 0 ||
+      kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!kernel || !is_aligned_to(kernel, 2)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+
+   buffer_len = 8 + 2 * 4 * kernel_size;
+   buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat));
+
+   buffer[0] = 0.f;
+   buffer[1] = 1.f;
+   buffer[2] = 2.f; /*unused*/
+   buffer[3] = 4.f; /*unused*/
+
+   buffer[4] = kernelWidth * kernelHeight;
+   buffer[5] = scale;
+   buffer[6] = bias;
+   buffer[7] = 0.f;
+
+   idx = 8;
+   for (j = 0; j < kernelHeight; ++j) {
+      for (i = 0; i < kernelWidth; ++i) {
+         VGint index = j * kernelWidth + i;
+         VGfloat x, y;
+
+         x = texture_offset(s->width, kernelWidth, i, shiftX);
+         y = texture_offset(s->height, kernelHeight, j, shiftY);
+
+         buffer[idx + index*4 + 0] = x;
+         buffer[idx + index*4 + 1] = y;
+         buffer[idx + index*4 + 2] = 0.f;
+         buffer[idx + index*4 + 3] = 0.f;
+      }
+   }
+   idx += kernel_size * 4;
+
+   for (j = 0; j < kernelHeight; ++j) {
+      for (i = 0; i < kernelWidth; ++i) {
+         /* transpose the kernel */
+         VGint index = j * kernelWidth + i;
+         VGint kindex = (kernelWidth - i - 1) * kernelHeight + (kernelHeight - j - 1);
+         buffer[idx + index*4 + 0] = kernel[kindex];
+         buffer[idx + index*4 + 1] = kernel[kindex];
+         buffer[idx + index*4 + 2] = kernel[kindex];
+         buffer[idx + index*4 + 3] = kernel[kindex];
+      }
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_convolution;
+   info.user_data = (void*)(long)(buffer_len/4);
+   info.const_buffer = buffer;
+   info.const_buffer_len = buffer_len * sizeof(VGfloat);
+   info.tiling_mode = tilingMode;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+
+   free(buffer);
+}
+
+void vgSeparableConvolve(VGImage dst, VGImage src,
+                         VGint kernelWidth,
+                         VGint kernelHeight,
+                         VGint shiftX, VGint shiftY,
+                         const VGshort * kernelX,
+                         const VGshort * kernelY,
+                         VGfloat scale,
+                         VGfloat bias,
+                         VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGshort *kernel;
+   VGint i, j, idx = 0;
+   const VGint max_kernel_size = vgGeti(VG_MAX_SEPARABLE_KERNEL_SIZE);
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (kernelWidth <= 0 || kernelHeight <= 0 ||
+       kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!kernelX || !kernelY ||
+       !is_aligned_to(kernelX, 2) || !is_aligned_to(kernelY, 2)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   kernel = malloc(sizeof(VGshort)*kernelWidth*kernelHeight);
+   for (i = 0; i < kernelWidth; ++i) {
+      for (j = 0; j < kernelHeight; ++j) {
+         kernel[idx] = kernelX[i] * kernelY[j];
+         ++idx;
+      }
+   }
+   vgConvolve(dst, src, kernelWidth, kernelHeight, shiftX, shiftY,
+              kernel, scale, bias, tilingMode);
+   free(kernel);
+}
+
+static INLINE VGfloat compute_gaussian_componenet(VGfloat x, VGfloat y,
+                                                  VGfloat stdDeviationX,
+                                                  VGfloat stdDeviationY)
+{
+   VGfloat mult = 1 / ( 2 * M_PI * stdDeviationX * stdDeviationY);
+   VGfloat e = exp( - ( pow(x, 2)/(2*pow(stdDeviationX, 2)) +
+                        pow(y, 2)/(2*pow(stdDeviationY, 2)) ) );
+   return mult * e;
+}
+
+static INLINE VGint compute_kernel_size(VGfloat deviation)
+{
+   VGint size = ceil(2.146 * deviation);
+   if (size > 11)
+      return 11;
+   return size;
+}
+
+static void compute_gaussian_kernel(VGfloat *kernel,
+                                    VGint width, VGint height,
+                                    VGfloat stdDeviationX,
+                                    VGfloat stdDeviationY)
+{
+   VGint i, j;
+   VGfloat scale = 0.0f;
+
+   for (j = 0; j < height; ++j) {
+      for (i = 0; i < width; ++i) {
+         VGint idx =  (height - j -1) * width + (width - i -1);
+         kernel[idx] = compute_gaussian_componenet(i-(ceil(width/2))-1,
+                                                   j-ceil(height/2)-1,
+                                                   stdDeviationX, stdDeviationY);
+         scale += kernel[idx];
+      }
+   }
+
+   for (j = 0; j < height; ++j) {
+      for (i = 0; i < width; ++i) {
+         VGint idx = j * width + i;
+         kernel[idx] /= scale;
+      }
+   }
+}
+
+void vgGaussianBlur(VGImage dst, VGImage src,
+                    VGfloat stdDeviationX,
+                    VGfloat stdDeviationY,
+                    VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   VGfloat *buffer, *kernel;
+   VGint kernel_width, kernel_height, kernel_size;
+   VGint buffer_len;
+   VGint idx, i, j;
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (stdDeviationX <= 0 || stdDeviationY <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   kernel_width = compute_kernel_size(stdDeviationX);
+   kernel_height = compute_kernel_size(stdDeviationY);
+   kernel_size = kernel_width * kernel_height;
+   kernel = malloc(sizeof(VGfloat)*kernel_size);
+   compute_gaussian_kernel(kernel, kernel_width, kernel_height,
+                           stdDeviationX, stdDeviationY);
+
+   buffer_len = 8 + 2 * 4 * kernel_size;
+   buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat));
+
+   buffer[0] = 0.f;
+   buffer[1] = 1.f;
+   buffer[2] = 2.f; /*unused*/
+   buffer[3] = 4.f; /*unused*/
+
+   buffer[4] = kernel_width * kernel_height;
+   buffer[5] = 1.f;/*scale*/
+   buffer[6] = 0.f;/*bias*/
+   buffer[7] = 0.f;
+
+   idx = 8;
+   for (j = 0; j < kernel_height; ++j) {
+      for (i = 0; i < kernel_width; ++i) {
+         VGint index = j * kernel_width + i;
+         VGfloat x, y;
+
+         x = texture_offset(s->width, kernel_width, i, kernel_width/2);
+         y = texture_offset(s->height, kernel_height, j, kernel_height/2);
+
+         buffer[idx + index*4 + 0] = x;
+         buffer[idx + index*4 + 1] = y;
+         buffer[idx + index*4 + 2] = 0.f;
+         buffer[idx + index*4 + 3] = 0.f;
+      }
+   }
+   idx += kernel_size * 4;
+
+   for (j = 0; j < kernel_height; ++j) {
+      for (i = 0; i < kernel_width; ++i) {
+         /* transpose the kernel */
+         VGint index = j * kernel_width + i;
+         VGint kindex = (kernel_width - i - 1) * kernel_height + (kernel_height - j - 1);
+         buffer[idx + index*4 + 0] = kernel[kindex];
+         buffer[idx + index*4 + 1] = kernel[kindex];
+         buffer[idx + index*4 + 2] = kernel[kindex];
+         buffer[idx + index*4 + 3] = kernel[kindex];
+      }
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_convolution;
+   info.user_data = (void*)(long)(buffer_len/4);
+   info.const_buffer = buffer;
+   info.const_buffer_len = buffer_len * sizeof(VGfloat);
+   info.tiling_mode = tilingMode;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+
+   free(buffer);
+   free(kernel);
+}
+
+void vgLookup(VGImage dst, VGImage src,
+              const VGubyte * redLUT,
+              const VGubyte * greenLUT,
+              const VGubyte * blueLUT,
+              const VGubyte * alphaLUT,
+              VGboolean outputLinear,
+              VGboolean outputPremultiplied)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   VGuint color_data[256];
+   VGint i;
+   struct pipe_texture *lut_texture;
+   VGfloat buffer[4];
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!redLUT || !greenLUT || !blueLUT || !alphaLUT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < 256; ++i) {
+      color_data[i] = blueLUT[i] << 24 | greenLUT[i] << 16 |
+                      redLUT[i]  <<  8 | alphaLUT[i];
+   }
+   lut_texture = create_texture_1d(ctx, color_data, 255);
+
+   buffer[0] = 0.f;
+   buffer[1] = 0.f;
+   buffer[2] = 1.f;
+   buffer[3] = 1.f;
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_lookup;
+   info.user_data = NULL;
+   info.const_buffer = buffer;
+   info.const_buffer_len = 4 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = lut_texture;
+
+   execute_filter(ctx, &info);
+
+   pipe_texture_reference(&lut_texture, NULL);
+}
+
+void vgLookupSingle(VGImage dst, VGImage src,
+                    const VGuint * lookupTable,
+                    VGImageChannel sourceChannel,
+                    VGboolean outputLinear,
+                    VGboolean outputPremultiplied)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   struct pipe_texture *lut_texture;
+   VGfloat buffer[4];
+   struct filter_info info;
+   VGuint color_data[256];
+   VGint i;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!lookupTable || !is_aligned(lookupTable)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (sourceChannel != VG_RED && sourceChannel != VG_GREEN &&
+       sourceChannel != VG_BLUE && sourceChannel != VG_ALPHA) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < 256; ++i) {
+      VGuint rgba = lookupTable[i];
+      VGubyte blue, green, red, alpha;
+      red   = (rgba & 0xff000000)>>24;
+      green = (rgba & 0x00ff0000)>>16;
+      blue  = (rgba & 0x0000ff00)>> 8;
+      alpha = (rgba & 0x000000ff)>> 0;
+      color_data[i] = blue << 24 | green << 16 |
+                      red  <<  8 | alpha;
+   }
+   lut_texture = create_texture_1d(ctx, color_data, 256);
+
+   buffer[0] = 0.f;
+   buffer[1] = 0.f;
+   buffer[2] = 1.f;
+   buffer[3] = 1.f;
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_lookup_single;
+   info.user_data = (void*)sourceChannel;
+   info.const_buffer = buffer;
+   info.const_buffer_len = 4 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = lut_texture;
+
+   execute_filter(ctx, &info);
+
+   pipe_texture_reference(&lut_texture, NULL);
+}
diff --git a/src/gallium/state_trackers/vega/api_images.c b/src/gallium/state_trackers/vega/api_images.c
new file mode 100644
index 0000000000..c437553bc2
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_images.c
@@ -0,0 +1,489 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "image.h"
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "vg_translate.h"
+#include "api_consts.h"
+#include "image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_blit.h"
+#include "util/u_tile.h"
+#include "util/u_memory.h"
+
+static INLINE VGboolean supported_image_format(VGImageFormat format)
+{
+   switch(format) {
+   case VG_sRGBX_8888:
+   case VG_sRGBA_8888:
+   case VG_sRGBA_8888_PRE:
+   case VG_sRGB_565:
+   case VG_sRGBA_5551:
+   case VG_sRGBA_4444:
+   case VG_sL_8:
+   case VG_lRGBX_8888:
+   case VG_lRGBA_8888:
+   case VG_lRGBA_8888_PRE:
+   case VG_lL_8:
+   case VG_A_8:
+   case VG_BW_1:
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+   case VG_A_4:
+#endif
+   case VG_sXRGB_8888:
+   case VG_sARGB_8888:
+   case VG_sARGB_8888_PRE:
+   case VG_sARGB_1555:
+   case VG_sARGB_4444:
+   case VG_lXRGB_8888:
+   case VG_lARGB_8888:
+   case VG_lARGB_8888_PRE:
+   case VG_sBGRX_8888:
+   case VG_sBGRA_8888:
+   case VG_sBGRA_8888_PRE:
+   case VG_sBGR_565:
+   case VG_sBGRA_5551:
+   case VG_sBGRA_4444:
+   case VG_lBGRX_8888:
+   case VG_lBGRA_8888:
+   case VG_lBGRA_8888_PRE:
+   case VG_sXBGR_8888:
+   case VG_sABGR_8888:
+   case VG_sABGR_8888_PRE:
+   case VG_sABGR_1555:
+   case VG_sABGR_4444:
+   case VG_lXBGR_8888:
+   case VG_lABGR_8888:
+   case VG_lABGR_8888_PRE:
+      return VG_TRUE;
+   default:
+      return VG_FALSE;
+   }
+   return VG_FALSE;
+}
+
+VGImage vgCreateImage(VGImageFormat format,
+                      VGint width, VGint height,
+                      VGbitfield allowedQuality)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (!supported_image_format(format)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width > vgGeti(VG_MAX_IMAGE_WIDTH) ||
+       height > vgGeti(VG_MAX_IMAGE_HEIGHT)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width * height > vgGeti(VG_MAX_IMAGE_PIXELS)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   if (!(allowedQuality & ((VG_IMAGE_QUALITY_NONANTIALIASED |
+                           VG_IMAGE_QUALITY_FASTER |
+                           VG_IMAGE_QUALITY_BETTER)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGImage)image_create(format, width, height);
+}
+
+void vgDestroyImage(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img = (struct vg_image *)image;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)image, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   image_destroy(img);
+}
+
+void vgClearImage(VGImage image,
+                  VGint x, VGint y,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)image;
+
+   if (x + width < 0 || y + height < 0)
+      return;
+
+   image_clear(img, x, y, width, height);
+
+}
+
+void vgImageSubData(VGImage image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0 || !data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)(image);
+   image_sub_data(img, data, dataStride, dataFormat,
+                  x, y, width, height);
+}
+
+void vgGetImageSubData(VGImage image,
+                       void * data,
+                       VGint dataStride,
+                       VGImageFormat dataFormat,
+                       VGint x, VGint y,
+                       VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0 || !data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   img = (struct vg_image*)image;
+   image_get_sub_data(img, data, dataStride, dataFormat,
+                      x, y, width, height);
+}
+
+VGImage vgChildImage(VGImage parent,
+                     VGint x, VGint y,
+                     VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *p;
+
+   if (parent == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void*)parent) ||
+       !vg_object_is_valid((void*)parent, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width <= 0 || height <= 0 || x < 0 || y < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   p = (struct vg_image *)parent;
+   if (x > p->width  || y > p->height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (x + width > p->width  || y + height > p->height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGImage)image_child_image(p, x, y, width, height);
+}
+
+VGImage vgGetParent(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   img = (struct vg_image*)image;
+   if (img->parent)
+      return (VGImage)img->parent;
+   else
+      return image;
+}
+
+void vgCopyImage(VGImage dst, VGint dx, VGint dy,
+                 VGImage src, VGint sx, VGint sy,
+                 VGint width, VGint height,
+                 VGboolean dither)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (src == VG_INVALID_HANDLE || dst == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   vg_validate_state(ctx);
+   image_copy((struct vg_image*)dst, dx, dy,
+              (struct vg_image*)src, sx, sy,
+              width, height, dither);
+}
+
+void vgDrawImage(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (!ctx)
+      return;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+   image_draw((struct vg_image*)image);
+}
+
+void vgSetPixels(VGint dx, VGint dy,
+                 VGImage src, VGint sx, VGint sy,
+                 VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   vg_validate_state(ctx);
+
+   if (src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   image_set_pixels(dx, dy, (struct vg_image*)src, sx, sy, width,
+                    height);
+}
+
+void vgGetPixels(VGImage dst, VGint dx, VGint dy,
+                 VGint sx, VGint sy,
+                 VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (dst == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)dst;
+
+   image_get_pixels(img, dx, dy,
+                    sx, sy, width, height);
+}
+
+void vgWritePixels(const void * data, VGint dataStride,
+                   VGImageFormat dataFormat,
+                   VGint dx, VGint dy,
+                   VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (!data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+   {
+      struct vg_image *img = image_create(dataFormat, width, height);
+      image_sub_data(img, data, dataStride, dataFormat, 0, 0,
+                     width, height);
+#if 0
+      struct matrix *matrix = &ctx->state.vg.image_user_to_surface_matrix;
+      matrix_translate(matrix, dx, dy);
+      image_draw(img);
+      matrix_translate(matrix, -dx, -dy);
+#else
+      /* this looks like a better approach */
+      image_set_pixels(dx, dy, img, 0, 0, width, height);
+#endif
+      image_destroy(img);
+   }
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+}
+
+void vgReadPixels(void * data, VGint dataStride,
+                  VGImageFormat dataFormat,
+                  VGint sx, VGint sy,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = (fb->height - sy) - 1, yStep = -1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+   VGint xoffset = 0, yoffset = 0;
+
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (!data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   if (sx < 0) {
+      xoffset = -sx;
+      xoffset *= _vega_size_for_format(dataFormat);
+      width += sx;
+      sx = 0;
+   }
+   if (sy < 0) {
+      yoffset = -sy;
+      height += sy;
+      sy = 0;
+      y = (fb->height - sy) - 1;
+      yoffset *= dataStride;
+   }
+
+   {
+      struct pipe_transfer *transfer;
+
+      transfer = screen->get_tex_transfer(screen, strb->texture,  0, 0, 0,
+                                          PIPE_TRANSFER_READ,
+                                          0, 0, width, height);
+
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(transfer, sx, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat,
+                                    dst + yoffset + xoffset);
+         dst += dataStride;
+      }
+
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+void vgCopyPixels(VGint dx, VGint dy,
+                  VGint sx, VGint sy,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   /* do nothing if we copy from outside the fb */
+   if (dx >= (VGint)fb->width || dy >= (VGint)fb->height ||
+       sx >= (VGint)fb->width || sy >= (VGint)fb->height)
+      return;
+
+   vg_validate_state(ctx);
+   /* make sure rendering has completed */
+   vgFinish();
+
+   vg_copy_surface(ctx, strb->surface, dx, dy,
+                   strb->surface, sx, sy, width, height);
+}
diff --git a/src/gallium/state_trackers/vega/api_masks.c b/src/gallium/state_trackers/vega/api_masks.c
new file mode 100644
index 0000000000..4f9f3dae17
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_masks.c
@@ -0,0 +1,373 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "mask.h"
+#include "renderer.h"
+
+#include "vg_context.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h" /* for winsys->update_buffer */
+
+#include "util/u_pack_color.h"
+#include "util/u_draw_quad.h"
+#include "util/u_memory.h"
+
+
+#define DISABLE_1_1_MASKING 1
+
+/**
+ * Draw a screen-aligned quadrilateral.
+ * Coords are window coords with y=0=bottom.  These coords will be transformed
+ * by the vertex shader and viewport transform.
+ */
+static void
+draw_clear_quad(struct vg_context *st,
+                float x0, float y0, float x1, float y1, float z,
+                const VGfloat color[4])
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_buffer *buf;
+   VGuint i;
+
+   /* positions */
+   st->clear.vertices[0][0][0] = x0;
+   st->clear.vertices[0][0][1] = y0;
+
+   st->clear.vertices[1][0][0] = x1;
+   st->clear.vertices[1][0][1] = y0;
+
+   st->clear.vertices[2][0][0] = x1;
+   st->clear.vertices[2][0][1] = y1;
+
+   st->clear.vertices[3][0][0] = x0;
+   st->clear.vertices[3][0][1] = y1;
+
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->clear.vertices[i][0][2] = z;
+      st->clear.vertices[i][0][3] = 1.0;
+      st->clear.vertices[i][1][0] = color[0];
+      st->clear.vertices[i][1][1] = color[1];
+      st->clear.vertices[i][1][2] = color[2];
+      st->clear.vertices[i][1][3] = color[3];
+   }
+
+
+   /* put vertex data into vbuf */
+   buf =  pipe_user_buffer_create(pipe->screen,
+                                  st->clear.vertices,
+                                  sizeof(st->clear.vertices));
+
+
+   /* draw */
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+}
+
+/**
+ * Do vgClear by drawing a quadrilateral.
+ */
+static void
+clear_with_quad(struct vg_context *st, float x0, float y0,
+                float width, float height, const VGfloat clear_color[4])
+{
+   VGfloat x1, y1;
+
+   vg_validate_state(st);
+
+   x1 = x0 + width;
+   y1 = y0 + height;
+
+   /*
+     printf("%s %f,%f %f,%f\n", __FUNCTION__,
+     x0, y0,
+     x1, y1);
+   */
+
+   if (st->pipe->winsys && st->pipe->winsys->update_buffer)
+      st->pipe->winsys->update_buffer( st->pipe->winsys,
+                                       st->pipe->priv );
+
+   cso_save_blend(st->cso_context);
+   cso_save_rasterizer(st->cso_context);
+   cso_save_fragment_shader(st->cso_context);
+   cso_save_vertex_shader(st->cso_context);
+
+   /* blend state: RGBA masking */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask |= PIPE_MASK_R;
+      blend.colormask |= PIPE_MASK_G;
+      blend.colormask |= PIPE_MASK_B;
+      blend.colormask |= PIPE_MASK_A;
+      cso_set_blend(st->cso_context, &blend);
+   }
+
+   cso_set_rasterizer(st->cso_context, &st->clear.raster);
+
+   cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
+   cso_set_vertex_shader_handle(st->cso_context, vg_clear_vs(st));
+
+   /* draw quad matching scissor rect (XXX verify coord round-off) */
+   draw_clear_quad(st, x0, y0, x1, y1, 0, clear_color);
+
+   /* Restore pipe state */
+   cso_restore_blend(st->cso_context);
+   cso_restore_rasterizer(st->cso_context);
+   cso_restore_fragment_shader(st->cso_context);
+   cso_restore_vertex_shader(st->cso_context);
+}
+
+
+void vgMask(VGHandle mask, VGMaskOperation operation,
+            VGint x, VGint y,
+            VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <=0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (operation < VG_CLEAR_MASK || operation > VG_SUBTRACT_MASK) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+
+   vg_validate_state(ctx);
+
+   if (operation == VG_CLEAR_MASK) {
+      mask_fill(x, y, width, height, 0.f);
+   } else if (operation == VG_FILL_MASK) {
+      mask_fill(x, y, width, height, 1.f);
+   } else if (vg_object_is_valid((void*)mask, VG_OBJECT_IMAGE)) {
+      struct vg_image *image = (struct vg_image *)mask;
+      mask_using_image(image, operation, x, y, width, height);
+   } else if (vg_object_is_valid((void*)mask, VG_OBJECT_MASK)) {
+#if DISABLE_1_1_MASKING
+      return;
+#else
+      struct vg_mask_layer *layer = (struct vg_mask_layer *)mask;
+      mask_using_layer(layer, operation, x, y, width, height);
+#endif
+   } else {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+   }
+}
+
+void vgClear(VGint x, VGint y,
+             VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb;
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+#if 0
+   debug_printf("Clear [%d, %d, %d, %d] with [%f, %f, %f, %f]\n",
+                x, y, width, height,
+                ctx->state.vg.clear_color[0],
+                ctx->state.vg.clear_color[1],
+                ctx->state.vg.clear_color[2],
+                ctx->state.vg.clear_color[3]);
+#endif
+
+   fb = &ctx->state.g3d.fb;
+   /* check for a whole surface clear */
+   if (!ctx->state.vg.scissoring &&
+       (x == 0 && y == 0 && width == fb->width && height == fb->height)) {
+      ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_COLOR | PIPE_CLEAR_DEPTHSTENCIL,
+                       ctx->state.vg.clear_color, 1., 0);
+   } else {
+      clear_with_quad(ctx, x, y, width, height, ctx->state.vg.clear_color);
+   }
+}
+
+
+#ifdef OPENVG_VERSION_1_1
+
+
+void vgRenderToMask(VGPath path,
+                    VGbitfield paintModes,
+                    VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!paintModes || (paintModes&(~(VG_STROKE_PATH|VG_FILL_PATH)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (operation < VG_CLEAR_MASK ||
+       operation > VG_SUBTRACT_MASK) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)path, VG_OBJECT_PATH)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+
+   vg_validate_state(ctx);
+
+   mask_render_to((struct path *)path, paintModes, operation);
+}
+
+VGMaskLayer vgCreateMaskLayer(VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <= 0 || height <= 0 ||
+       width > vgGeti(VG_MAX_IMAGE_WIDTH) ||
+       height > vgGeti(VG_MAX_IMAGE_HEIGHT)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGMaskLayer)mask_layer_create(width, height);
+}
+
+void vgDestroyMaskLayer(VGMaskLayer maskLayer)
+{
+   struct vg_mask_layer *mask = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   mask = (struct vg_mask_layer *)maskLayer;
+   mask_layer_destroy(mask);
+}
+
+void vgFillMaskLayer(VGMaskLayer maskLayer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value)
+{
+   struct vg_mask_layer *mask = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (value < 0 || value > 1) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (x < 0 || y < 0 || (x + width) < 0 || (y + height) < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   mask = (struct vg_mask_layer*)maskLayer;
+
+   if (x + width > mask_layer_width(mask) ||
+       y + height > mask_layer_height(mask)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+   mask_layer_fill(mask, x, y, width, height, value);
+}
+
+void vgCopyMask(VGMaskLayer maskLayer,
+                VGint sx, VGint sy,
+                VGint dx, VGint dy,
+                VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_mask_layer *mask = 0;
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+
+   mask = (struct vg_mask_layer*)maskLayer;
+   mask_copy(mask, sx, sy, dx, dy, width, height);
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_misc.c b/src/gallium/state_trackers/vega/api_misc.c
new file mode 100644
index 0000000000..78ba0bc110
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_misc.c
@@ -0,0 +1,83 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+/* Hardware Queries */
+VGHardwareQueryResult vgHardwareQuery(VGHardwareQueryType key,
+                                      VGint setting)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (key < VG_IMAGE_FORMAT_QUERY ||
+       key > VG_PATH_DATATYPE_QUERY) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_HARDWARE_UNACCELERATED;
+   }
+
+   if (key == VG_IMAGE_FORMAT_QUERY) {
+      if (setting < VG_sRGBX_8888 ||
+          setting > VG_lABGR_8888_PRE) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return VG_HARDWARE_UNACCELERATED;
+      }
+   } else if (key == VG_PATH_DATATYPE_QUERY) {
+      if (setting < VG_PATH_DATATYPE_S_8 ||
+          setting > VG_PATH_DATATYPE_F) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return VG_HARDWARE_UNACCELERATED;
+      }
+   }
+   /* we're supposed to accelerate everything */
+   return VG_HARDWARE_ACCELERATED;
+}
+
+/* Renderer and Extension Information */
+const VGubyte *vgGetString(VGStringID name)
+{
+   struct vg_context *ctx = vg_current_context();
+   static const VGubyte *vendor = (VGubyte *)"Tungsten Graphics, Inc";
+   static const VGubyte *renderer = (VGubyte *)"Vega OpenVG 1.0";
+   static const VGubyte *version = (VGubyte *)"1.0";
+
+   if (!ctx)
+      return NULL;
+
+   switch(name) {
+   case VG_VENDOR:
+      return vendor;
+   case VG_RENDERER:
+      return renderer;
+   case VG_VERSION:
+      return version;
+   case VG_EXTENSIONS:
+      return NULL;
+   default:
+      return NULL;
+   }
+}
diff --git a/src/gallium/state_trackers/vega/api_paint.c b/src/gallium/state_trackers/vega/api_paint.c
new file mode 100644
index 0000000000..dd3ac5bdb0
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_paint.c
@@ -0,0 +1,166 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "paint.h"
+#include "image.h"
+
+VGPaint vgCreatePaint(void)
+{
+   return (VGPaint) paint_create(vg_current_context());
+}
+
+void vgDestroyPaint(VGPaint p)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_paint *paint;
+
+   if (p == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   paint = (struct vg_paint *)p;
+   paint_destroy(paint);
+}
+
+void vgSetPaint(VGPaint paint, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE) {
+      /* restore the default */
+      paint = (VGPaint)ctx->default_paint;
+   } else if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!(paintModes & ((VG_FILL_PATH|VG_STROKE_PATH)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (paintModes & VG_FILL_PATH) {
+      ctx->state.vg.fill_paint = (struct vg_paint *)paint;
+   }
+   if (paintModes & VG_STROKE_PATH) {
+      ctx->state.vg.stroke_paint = (struct vg_paint *)paint;
+   }
+}
+
+VGPaint vgGetPaint(VGPaintMode paintMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGPaint paint = VG_INVALID_HANDLE;
+
+   if (paintMode < VG_STROKE_PATH || paintMode > VG_FILL_PATH) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   if (paintMode == VG_FILL_PATH)
+      paint = (VGPaint)ctx->state.vg.fill_paint;
+   else if (paintMode == VG_STROKE_PATH)
+      paint = (VGPaint)ctx->state.vg.stroke_paint;
+
+   if (paint == (VGPaint)ctx->default_paint)
+      paint = VG_INVALID_HANDLE;
+
+   return paint;
+}
+
+void vgSetColor(VGPaint paint, VGuint rgba)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   {
+      struct vg_paint *p = (struct vg_paint *)paint;
+      paint_set_colori(p, rgba);
+   }
+}
+
+VGuint vgGetColor(VGPaint paint)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_paint *p;
+   VGuint rgba = 0;
+
+   if (paint == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return rgba;
+   }
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return rgba;
+   }
+   p = (struct vg_paint *)paint;
+
+   return paint_colori(p);
+}
+
+void vgPaintPattern(VGPaint paint, VGImage pattern)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_PAINT, (void *)paint)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (pattern == VG_INVALID_HANDLE) {
+      paint_set_type((struct vg_paint*)paint, VG_PAINT_TYPE_COLOR);
+      return;
+   }
+
+   if (!vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)pattern)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT) ||
+       !vg_object_is_valid((void*)pattern, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   paint_set_pattern((struct vg_paint*)paint,
+                     (struct vg_image*)pattern);
+}
+
diff --git a/src/gallium/state_trackers/vega/api_params.c b/src/gallium/state_trackers/vega/api_params.c
new file mode 100644
index 0000000000..db77fd9cb0
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_params.c
@@ -0,0 +1,1673 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "paint.h"
+#include "path.h"
+#include "image.h"
+#include "matrix.h"
+#include "api_consts.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_pointer.h"
+#include "util/u_math.h"
+
+#include <math.h>
+
+static INLINE struct vg_state *current_state()
+{
+   struct vg_context *ctx = vg_current_context();
+   if (!ctx)
+      return 0;
+   else
+      return &ctx->state.vg;
+}
+
+static INLINE VGboolean count_in_bounds(VGParamType type, VGint count)
+{
+   if (count < 0)
+      return VG_FALSE;
+
+   if (type == VG_SCISSOR_RECTS)
+      return (!(count % 4) && (count >= 0 || count <= VEGA_MAX_SCISSOR_RECTS * 4));
+   else if (type == VG_STROKE_DASH_PATTERN) {
+      return count <= VEGA_MAX_DASH_COUNT;
+   } else {
+      VGint real_count = vgGetVectorSize(type);
+      return count == real_count;
+   }
+}
+
+void vgSetf (VGParamType type, VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      vgSeti(type, floor(value));
+      return;
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = value;
+      state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = value;
+      state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = value;
+      state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSeti (VGParamType type, VGint value)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+      if (value < VG_MATRIX_PATH_USER_TO_SURFACE ||
+#ifdef OPENVG_VERSION_1_1
+          value > VG_MATRIX_GLYPH_USER_TO_SURFACE)
+#else
+          value > VG_MATRIX_STROKE_PAINT_TO_USER)
+#endif
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->matrix_mode = value;
+      break;
+   case VG_FILL_RULE:
+      if (value < VG_EVEN_ODD ||
+          value > VG_NON_ZERO)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->fill_rule = value;
+      break;
+   case VG_IMAGE_QUALITY:
+      state->image_quality = value;
+      break;
+   case VG_RENDERING_QUALITY:
+      if (value < VG_RENDERING_QUALITY_NONANTIALIASED ||
+          value > VG_RENDERING_QUALITY_BETTER)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->rendering_quality = value;
+      break;
+   case VG_BLEND_MODE:
+      if (value < VG_BLEND_SRC ||
+          value > VG_BLEND_ADDITIVE)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else {
+         ctx->state.dirty |= BLEND_DIRTY;
+         state->blend_mode = value;
+      }
+      break;
+   case VG_IMAGE_MODE:
+      if (value < VG_DRAW_IMAGE_NORMAL ||
+          value > VG_DRAW_IMAGE_STENCIL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->image_mode = value;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      state->color_transform = value;
+#endif
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = value;
+      state->stroke.line_width.i = value;
+      break;
+   case VG_STROKE_CAP_STYLE:
+      if (value < VG_CAP_BUTT ||
+          value > VG_CAP_SQUARE)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->stroke.cap_style = value;
+      break;
+   case VG_STROKE_JOIN_STYLE:
+      if (value < VG_JOIN_MITER ||
+          value > VG_JOIN_BEVEL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->stroke.join_style = value;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = value;
+      state->stroke.miter_limit.i = value;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = value;
+      state->stroke.dash_phase.i = value;
+      break;
+   case VG_STROKE_DASH_PHASE_RESET:
+      state->stroke.dash_phase_reset = value;
+      break;
+   case VG_MASKING:
+      state->masking = value;
+      break;
+   case VG_SCISSORING:
+      state->scissoring = value;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+      break;
+   case VG_PIXEL_LAYOUT:
+      if (value < VG_PIXEL_LAYOUT_UNKNOWN ||
+          value > VG_PIXEL_LAYOUT_BGR_HORIZONTAL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->pixel_layout = value;
+      break;
+   case VG_SCREEN_LAYOUT:
+      /* read only ignore */
+      break;
+   case VG_FILTER_FORMAT_LINEAR:
+      state->filter_format_linear = value;
+      break;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      state->filter_format_premultiplied = value;
+      break;
+   case VG_FILTER_CHANNEL_MASK:
+      state->filter_channel_mask = value;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      /* read only ignore */
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSetfv(VGParamType type, VGint count,
+             const VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      vgSeti(type, floor(values[0]));
+      return;
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      VGuint *x = (VGuint*)values;
+      for (i = 0; i < count; ++i) {
+         state->scissor_rects[i].f =  values[i];
+         state->scissor_rects[i].i =  float_to_int_floor(x[i]);
+      }
+      state->scissor_rects_num = count / 4;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->color_transform_values[i] =  values[i];
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = values[0];
+      state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = values[0];
+      state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      int i;
+      for (i = 0; i < count; ++i) {
+         state->stroke.dash_pattern[i].f = values[i];
+         state->stroke.dash_pattern[i].i =
+            float_to_int_floor(*((VGuint*)(values + i)));
+      }
+      state->stroke.dash_pattern_num = count;
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = values[0];
+      state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_TILE_FILL_COLOR:
+      state->tile_fill_color[0] = values[0];
+      state->tile_fill_color[1] = values[1];
+      state->tile_fill_color[2] = values[2];
+      state->tile_fill_color[3] = values[3];
+
+      state->tile_fill_colori[0] = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->tile_fill_colori[1] = float_to_int_floor(*((VGuint*)(values + 1)));
+      state->tile_fill_colori[2] = float_to_int_floor(*((VGuint*)(values + 2)));
+      state->tile_fill_colori[3] = float_to_int_floor(*((VGuint*)(values + 3)));
+      break;
+   case VG_CLEAR_COLOR:
+      state->clear_color[0] = values[0];
+      state->clear_color[1] = values[1];
+      state->clear_color[2] = values[2];
+      state->clear_color[3] = values[3];
+
+      state->clear_colori[0] = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->clear_colori[1] = float_to_int_floor(*((VGuint*)(values + 1)));
+      state->clear_colori[2] = float_to_int_floor(*((VGuint*)(values + 2)));
+      state->clear_colori[3] = float_to_int_floor(*((VGuint*)(values + 3)));
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      state->glyph_origin[0].f = values[0];
+      state->glyph_origin[1].f = values[1];
+
+      state->glyph_origin[0].i = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->glyph_origin[1].i = float_to_int_floor(*((VGuint*)(values + 1)));
+      break;
+#endif
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSetiv(VGParamType type, VGint count,
+             const VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+
+   if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      vgSeti(type, values[0]);
+      return;
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->scissor_rects[i].i =  values[i];
+         state->scissor_rects[i].f =  values[i];
+      }
+      state->scissor_rects_num = count / 4;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->color_transform_values[i] =  values[i];
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = values[0];
+      state->stroke.line_width.i = values[0];
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = values[0];
+      state->stroke.miter_limit.i = values[0];
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      int i;
+      for (i = 0; i < count; ++i) {
+         state->stroke.dash_pattern[i].f = values[i];
+         state->stroke.dash_pattern[i].i = values[i];
+      }
+      state->stroke.dash_pattern_num = count;
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = values[0];
+      state->stroke.dash_phase.i = values[0];
+      break;
+   case VG_TILE_FILL_COLOR:
+      state->tile_fill_color[0] = values[0];
+      state->tile_fill_color[1] = values[1];
+      state->tile_fill_color[2] = values[2];
+      state->tile_fill_color[3] = values[3];
+
+      state->tile_fill_colori[0] = values[0];
+      state->tile_fill_colori[1] = values[1];
+      state->tile_fill_colori[2] = values[2];
+      state->tile_fill_colori[3] = values[3];
+      break;
+   case VG_CLEAR_COLOR:
+      state->clear_color[0] = values[0];
+      state->clear_color[1] = values[1];
+      state->clear_color[2] = values[2];
+      state->clear_color[3] = values[3];
+
+      state->clear_colori[0] = values[0];
+      state->clear_colori[1] = values[1];
+      state->clear_colori[2] = values[2];
+      state->clear_colori[3] = values[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      state->glyph_origin[0].f = values[0];
+      state->glyph_origin[1].f = values[1];
+      state->glyph_origin[0].i = values[0];
+      state->glyph_origin[1].i = values[1];
+      break;
+#endif
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      break;
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+VGfloat vgGetf(VGParamType type)
+{
+   struct vg_context *ctx = vg_current_context();
+   const struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+   VGfloat value = 0.0f;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      return vgGeti(type);
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      value = state->stroke.line_width.f;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      value = state->stroke.miter_limit.f;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      value = state->stroke.dash_phase.f;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      return vgGeti(type);
+      break;
+   case VG_MAX_FLOAT:
+      value = 1e+10;/*must be at least 1e+10*/
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+   return value;
+}
+
+VGint vgGeti(VGParamType type)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGErrorCode error = VG_NO_ERROR;
+   VGint value = 0;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+      value = state->matrix_mode;
+      break;
+   case VG_FILL_RULE:
+      value = state->fill_rule;
+      break;
+   case VG_IMAGE_QUALITY:
+      value = state->image_quality;
+      break;
+   case VG_RENDERING_QUALITY:
+      value = state->rendering_quality;
+      break;
+   case VG_BLEND_MODE:
+      value = state->blend_mode;
+      break;
+   case VG_IMAGE_MODE:
+      value = state->image_mode;
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      value = state->color_transform;
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      value = state->stroke.line_width.i;
+      break;
+   case VG_STROKE_CAP_STYLE:
+      value = state->stroke.cap_style;
+      break;
+   case VG_STROKE_JOIN_STYLE:
+      value = state->stroke.join_style;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      value = state->stroke.miter_limit.i;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      value = state->stroke.dash_phase.i;
+      break;
+   case VG_STROKE_DASH_PHASE_RESET:
+      value = state->stroke.dash_phase_reset;
+      break;
+   case VG_MASKING:
+      value = state->masking;
+      break;
+   case VG_SCISSORING:
+      value = state->scissoring;
+      break;
+   case VG_PIXEL_LAYOUT:
+      value = state->pixel_layout;
+      break;
+   case VG_SCREEN_LAYOUT:
+      value = state->screen_layout;
+      break;
+   case VG_FILTER_FORMAT_LINEAR:
+      value = state->filter_format_linear;
+      break;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      value = state->filter_format_premultiplied;
+      break;
+   case VG_FILTER_CHANNEL_MASK:
+      value = state->filter_channel_mask;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+      value = 32; /*must be at least 32*/
+      break;
+   case VG_MAX_DASH_COUNT:
+      value = 16; /*must be at least 16*/
+      break;
+   case VG_MAX_KERNEL_SIZE:
+      value = 7; /*must be at least 7*/
+      break;
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+      value = 15; /*must be at least 15*/
+      break;
+   case VG_MAX_COLOR_RAMP_STOPS:
+      value = 256; /*must be at least 32*/
+      break;
+   case VG_MAX_IMAGE_WIDTH:
+      value = 2048;
+      break;
+   case VG_MAX_IMAGE_HEIGHT:
+      value = 2048;
+      break;
+   case VG_MAX_IMAGE_PIXELS:
+      value = 2048*2048;
+      break;
+   case VG_MAX_IMAGE_BYTES:
+      value = 2048*2048 * 4;
+      break;
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      value = 128; /*must be at least 128*/
+      break;
+
+   case VG_MAX_FLOAT: {
+      VGfloat val = vgGetf(type);
+      value = float_to_int_floor(*((VGuint*)&val));
+   }
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+   return value;
+}
+
+VGint vgGetVectorSize(VGParamType type)
+{
+   struct vg_context *ctx = vg_current_context();
+   const struct vg_state *state = current_state();
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+      return 1;
+   case VG_SCISSOR_RECTS:
+      return state->scissor_rects_num * 4;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      return 1;
+   case VG_COLOR_TRANSFORM_VALUES:
+      return 8;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_MITER_LIMIT:
+      return 1;
+   case VG_STROKE_DASH_PATTERN:
+      return state->stroke.dash_pattern_num;
+   case VG_STROKE_DASH_PHASE:
+      return 1;
+   case VG_STROKE_DASH_PHASE_RESET:
+      return 1;
+   case VG_TILE_FILL_COLOR:
+      return 4;
+   case VG_CLEAR_COLOR:
+      return 4;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      return 2;
+#endif
+   case VG_MASKING:
+      return 1;
+   case VG_SCISSORING:
+      return 1;
+   case VG_PIXEL_LAYOUT:
+      return 1;
+   case VG_SCREEN_LAYOUT:
+      return 1;
+   case VG_FILTER_FORMAT_LINEAR:
+      return 1;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      return 1;
+   case VG_FILTER_CHANNEL_MASK:
+      return 1;
+
+   case VG_MAX_COLOR_RAMP_STOPS:
+      return 1;
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_FLOAT:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      return 1;
+   default:
+      if (ctx)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return 0;
+   }
+}
+
+void vgGetfv(VGParamType type, VGint count,
+             VGfloat * values)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGint real_count = vgGetVectorSize(type);
+
+   if (!values || count <= 0 || count > real_count || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      values[0] = vgGeti(type);
+      break;
+   case VG_MAX_FLOAT:
+      values[0] = vgGetf(type);
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->scissor_rects[i].f;
+      }
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      memcpy(values, state->color_transform_values,
+             sizeof(VGfloat) * count);
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      values[0] = state->stroke.line_width.f;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      values[0] = state->stroke.miter_limit.f;
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->stroke.dash_pattern[i].f;
+      }
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      values[0] = state->stroke.dash_phase.f;
+      break;
+   case VG_TILE_FILL_COLOR:
+      values[0] = state->tile_fill_color[0];
+      values[1] = state->tile_fill_color[1];
+      values[2] = state->tile_fill_color[2];
+      values[3] = state->tile_fill_color[3];
+      break;
+   case VG_CLEAR_COLOR:
+      values[0] = state->clear_color[0];
+      values[1] = state->clear_color[1];
+      values[2] = state->clear_color[2];
+      values[3] = state->clear_color[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      values[0] = state->glyph_origin[0].f;
+      values[1] = state->glyph_origin[1].f;
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgGetiv(VGParamType type, VGint count,
+             VGint * values)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGint real_count = vgGetVectorSize(type);
+
+   if (!values || count <= 0 || count > real_count || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      values[0] = vgGeti(type);
+      break;
+   case VG_MAX_FLOAT: {
+      VGfloat val = vgGetf(type);
+      values[0] = float_to_int_floor(*((VGuint*)&val));
+   }
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->scissor_rects[i].i;
+      }
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      VGuint *x = (VGuint*)state->color_transform_values;
+      for (i = 0; i < count; ++i) {
+         values[i] = float_to_int_floor(x[i]);
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      values[0] = state->stroke.line_width.i;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      values[0] = state->stroke.miter_limit.i;
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->stroke.dash_pattern[i].i;
+      }
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      values[0] = state->stroke.dash_phase.i;
+      break;
+   case VG_TILE_FILL_COLOR:
+      values[0] = state->tile_fill_colori[0];
+      values[1] = state->tile_fill_colori[1];
+      values[2] = state->tile_fill_colori[2];
+      values[3] = state->tile_fill_colori[3];
+      break;
+   case VG_CLEAR_COLOR:
+      values[0] = state->clear_colori[0];
+      values[1] = state->clear_colori[1];
+      values[2] = state->clear_colori[2];
+      values[3] = state->clear_colori[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      values[0] = state->glyph_origin[0].i;
+      values[1] = state->glyph_origin[1].i;
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgSetParameterf(VGHandle object,
+                     VGint paramType,
+                     VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      vgSetParameteri(object, paramType, floor(value));
+      return;
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      /* it's an error if paramType refers to a vector parameter */
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      paint_set_color_ramp_premultiplied(p, value);
+   }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgSetParameteri(VGHandle object,
+                     VGint paramType,
+                     VGint value)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+      if (value < VG_PAINT_TYPE_COLOR ||
+          value > VG_PAINT_TYPE_PATTERN)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_type(paint, value);
+      }
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      /* it's an error if paramType refers to a vector parameter */
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+      if (value < VG_COLOR_RAMP_SPREAD_PAD ||
+          value > VG_COLOR_RAMP_SPREAD_REFLECT)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_spread_mode(paint, value);
+      }
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      paint_set_color_ramp_premultiplied(p, value);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (value < VG_TILE_FILL ||
+          value > VG_TILE_REFLECT)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_pattern_tiling(paint, value);
+      }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgSetParameterfv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      const VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (count < 0 || count < real_count ||
+       (values == NULL && count != 0) ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (count != 1)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else
+         vgSetParameterf(object, paramType, values[0]);
+      return;
+      break;
+   case VG_PAINT_COLOR: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_color(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      if (count && count < 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         count = MIN2(count, VEGA_MAX_COLOR_RAMP_STOPS);
+         paint_set_ramp_stops(paint, values, count);
+         {
+            VGint stopsi[VEGA_MAX_COLOR_RAMP_STOPS];
+            int i = 0;
+            for (i = 0; i < count; ++i) {
+               stopsi[i] = float_to_int_floor(*((VGuint*)(values + i)));
+            }
+            paint_set_ramp_stopsi(paint, stopsi, count);
+         }
+      }
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_linear_gradient(paint, values);
+         {
+            VGint vals[4];
+            vals[0] = FLT_TO_INT(values[0]);
+            vals[1] = FLT_TO_INT(values[1]);
+            vals[2] = FLT_TO_INT(values[2]);
+            vals[3] = FLT_TO_INT(values[3]);
+            paint_set_linear_gradienti(paint, vals);
+         }
+      }
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      if (count != 5)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_radial_gradient(paint, values);
+         {
+            VGint vals[5];
+            vals[0] = FLT_TO_INT(values[0]);
+            vals[1] = FLT_TO_INT(values[1]);
+            vals[2] = FLT_TO_INT(values[2]);
+            vals[3] = FLT_TO_INT(values[3]);
+            vals[4] = FLT_TO_INT(values[4]);
+            paint_set_radial_gradienti(paint, vals);
+         }
+      }
+   }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgSetParameteriv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      const VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (count < 0 || count < real_count ||
+       (values == NULL && count != 0) ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (count != 1)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else
+         vgSetParameteri(object, paramType, values[0]);
+      return;
+      break;
+   case VG_PAINT_COLOR: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_coloriv(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      if ((count % 5))
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat *vals = 0;
+         int i;
+         struct vg_paint *paint = (struct vg_paint *)object;
+         if (count) {
+            vals = malloc(sizeof(VGfloat)*count);
+            for (i = 0; i < count; ++i)
+               vals[i] = values[i];
+         }
+
+         paint_set_ramp_stopsi(paint, values, count);
+         paint_set_ramp_stops(paint, vals, count);
+         free(vals);
+      }
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat vals[4];
+         struct vg_paint *paint = (struct vg_paint *)object;
+         vals[0] = values[0];
+         vals[1] = values[1];
+         vals[2] = values[2];
+         vals[3] = values[3];
+         paint_set_linear_gradient(paint, vals);
+         paint_set_linear_gradienti(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      if (count != 5)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat vals[5];
+         struct vg_paint *paint = (struct vg_paint *)object;
+         vals[0] = values[0];
+         vals[1] = values[1];
+         vals[2] = values[2];
+         vals[3] = values[3];
+         vals[4] = values[4];
+         paint_set_radial_gradient(paint, vals);
+         paint_set_radial_gradienti(paint, values);
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      /* read only don't produce an error */
+      break;
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+VGint vgGetParameterVectorSize(VGHandle object,
+                               VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      return 1;
+   case VG_PAINT_COLOR:
+      return 4;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_num_ramp_stops(p);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT:
+      return 4;
+   case VG_PAINT_RADIAL_GRADIENT:
+      return 5;
+
+
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      return 1;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+      return 1;
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      return 1;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+
+VGfloat vgGetParameterf(VGHandle object,
+                        VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      return vgGetParameteri(object, paramType);
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+
+   case VG_PATH_FORMAT:
+      return VG_PATH_FORMAT_STANDARD;
+   case VG_PATH_SCALE: {
+      struct path *p = (struct path*)object;
+      return path_scale(p);
+   }
+   case VG_PATH_BIAS: {
+      struct path *p = (struct path*)object;
+      return path_bias(p);
+   }
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      return vgGetParameteri(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS: 
+      return vgGetParameteri(object, paramType);
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+VGint vgGetParameteri(VGHandle object,
+                      VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE: {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         return paint_type(paint);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_spread_mode(p);
+   }
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_color_ramp_premultiplied(p);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_pattern_tiling(p);
+   }
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+
+   case VG_PATH_FORMAT:
+      return VG_PATH_FORMAT_STANDARD;
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      return vgGetParameterf(object, paramType);
+   case VG_PATH_DATATYPE: {
+      struct path *p = (struct path*)object;
+      return path_datatype(p);
+   }
+   case VG_PATH_NUM_SEGMENTS: {
+      struct path *p = (struct path*)object;
+      return path_num_segments(p);
+   }
+   case VG_PATH_NUM_COORDS: {
+      struct path *p = (struct path*)object;
+      return path_num_coords(p);
+   }
+      break;
+
+   case VG_IMAGE_FORMAT: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->format;
+   }
+      break;
+   case VG_IMAGE_WIDTH: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->width;
+   }
+      break;
+   case VG_IMAGE_HEIGHT: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->height;
+   }
+      break;
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS: {
+      return 1;
+   }
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+void vgGetParameterfv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!values || count <= 0 || count > real_count ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_type(p);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_spread_mode(p);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_color_ramp_premultiplied(p);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE: {
+      values[0] = vgGetParameterf(object, paramType);
+   }
+      break;
+   case VG_PAINT_COLOR: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_get_color(paint, values);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_ramp_stops(paint, values, count);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_linear_gradient(paint, values);
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_radial_gradient(paint, values);
+   }
+      break;
+
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      values[0] = vgGetParameterf(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgGetParameteriv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!values || count <= 0 || count > real_count ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+#endif
+   case VG_PAINT_COLOR: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_get_coloriv(paint, values);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_ramp_stopsi(paint, values, count);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_linear_gradienti(paint, values);
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_radial_gradienti(paint, values);
+   }
+      break;
+
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      values[0] = vgGetParameterf(object, paramType);
+      break;
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
diff --git a/src/gallium/state_trackers/vega/api_path.c b/src/gallium/state_trackers/vega/api_path.c
new file mode 100644
index 0000000000..a6b7a2bb93
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_path.c
@@ -0,0 +1,488 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "path.h"
+#include "polygon.h"
+#include "paint.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_draw_quad.h"
+
+VGPath vgCreatePath(VGint pathFormat,
+                    VGPathDatatype datatype,
+                    VGfloat scale, VGfloat bias,
+                    VGint segmentCapacityHint,
+                    VGint coordCapacityHint,
+                    VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (pathFormat != VG_PATH_FORMAT_STANDARD) {
+      vg_set_error(ctx, VG_UNSUPPORTED_PATH_FORMAT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (datatype < VG_PATH_DATATYPE_S_8 ||
+       datatype > VG_PATH_DATATYPE_F) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (!scale) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGPath)path_create(datatype, scale, bias,
+                              segmentCapacityHint, coordCapacityHint,
+                              capabilities);
+}
+
+void vgClearPath(VGPath path, VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   p = (struct path *)path;
+   path_clear(p, capabilities);
+}
+
+void vgDestroyPath(VGPath p)
+{
+   struct path *path = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (p == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   path = (struct path *)p;
+   path_destroy(path);
+}
+
+void vgRemovePathCapabilities(VGPath path,
+                              VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGbitfield current;
+   struct path *p;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+   current = path_capabilities(p);
+   path_set_capabilities(p, (current &
+                             (~(capabilities & VG_PATH_CAPABILITY_ALL))));
+}
+
+VGbitfield vgGetPathCapabilities(VGPath path)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+   p = (struct path*)path;
+   return path_capabilities(p);
+}
+
+void vgAppendPath(VGPath dstPath, VGPath srcPath)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *src, *dst;
+
+   if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   src = (struct path *)srcPath;
+   dst = (struct path *)dstPath;
+
+   if (!(path_capabilities(src) & VG_PATH_CAPABILITY_APPEND_FROM) ||
+       !(path_capabilities(dst) & VG_PATH_CAPABILITY_APPEND_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_append_path(dst, src);
+}
+
+void vgAppendPathData(VGPath dstPath,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGint i;
+
+   if (dstPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!pathSegments) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   for (i = 0; i < numSegments; ++i) {
+      if (pathSegments[i] < VG_CLOSE_PATH ||
+          pathSegments[i] > VG_LCWARC_TO_REL) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return;
+      }
+   }
+
+   p = (struct path*)dstPath;
+
+   if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!(path_capabilities(p)&VG_PATH_CAPABILITY_APPEND_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   path_append_data(p, numSegments, pathSegments, pathData);
+}
+
+void vgModifyPathCoords(VGPath dstPath,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (dstPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (startIndex < 0 || numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path *)dstPath;
+
+   if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (startIndex + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!(path_capabilities(p)&VG_PATH_CAPABILITY_MODIFY)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_modify_coords(p, startIndex, numSegments, pathData);
+}
+
+void vgTransformPath(VGPath dstPath, VGPath srcPath)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *src = 0, *dst = 0;
+
+   if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   src = (struct path *)srcPath;
+   dst = (struct path *)dstPath;
+
+   if (!(path_capabilities(src) & VG_PATH_CAPABILITY_TRANSFORM_FROM) ||
+       !(path_capabilities(dst) & VG_PATH_CAPABILITY_TRANSFORM_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_transform(dst, src);
+}
+
+VGboolean vgInterpolatePath(VGPath dstPath,
+                            VGPath startPath,
+                            VGPath endPath,
+                            VGfloat amount)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *start = 0, *dst = 0, *end = 0;
+
+   if (dstPath == VG_INVALID_HANDLE ||
+       startPath == VG_INVALID_HANDLE ||
+       endPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_FALSE;
+   }
+   dst = (struct path *)dstPath;
+   start = (struct path *)startPath;
+   end = (struct path *)endPath;
+
+   if (!(path_capabilities(dst) & VG_PATH_CAPABILITY_INTERPOLATE_TO) ||
+       !(path_capabilities(start) & VG_PATH_CAPABILITY_INTERPOLATE_FROM) ||
+       !(path_capabilities(end) & VG_PATH_CAPABILITY_INTERPOLATE_FROM)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return VG_FALSE;
+   }
+
+   return path_interpolate(dst,
+                           start, end, amount);
+}
+
+VGfloat vgPathLength(VGPath path,
+                     VGint startSegment,
+                     VGint numSegments)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return -1;
+   }
+   if (startSegment < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+   p = (struct path*)path;
+
+   if (!(path_capabilities(p) & VG_PATH_CAPABILITY_PATH_LENGTH)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return -1;
+   }
+   if (startSegment + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+
+   return path_length(p, startSegment, numSegments);
+}
+
+void vgPointAlongPath(VGPath path,
+                      VGint startSegment,
+                      VGint numSegments,
+                      VGfloat distance,
+                      VGfloat * x, VGfloat * y,
+                      VGfloat * tangentX,
+                      VGfloat * tangentY)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (startSegment < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(x) || !is_aligned(y) ||
+       !is_aligned(tangentX) || !is_aligned(tangentY)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_POINT_ALONG_PATH) ||
+       !(caps & VG_PATH_CAPABILITY_TANGENT_ALONG_PATH)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   if (startSegment + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   {
+      VGfloat point[2], normal[2];
+      path_point(p, startSegment, numSegments, distance,
+                 point, normal);
+      if (x)
+         *x = point[0];
+      if (y)
+         *y = point[1];
+      if (tangentX)
+         *tangentX = -normal[1];
+      if (tangentY)
+         *tangentY = normal[0];
+   }
+}
+
+void vgPathBounds(VGPath path,
+                  VGfloat * minX,
+                  VGfloat * minY,
+                  VGfloat * width,
+                  VGfloat * height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!minX || !minY || !width || !height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(minX) || !is_aligned(minY) ||
+       !is_aligned(width) || !is_aligned(height)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_PATH_BOUNDS)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   path_bounding_rect(p, minX, minY, width, height);
+}
+
+void vgPathTransformedBounds(VGPath path,
+                             VGfloat * minX,
+                             VGfloat * minY,
+                             VGfloat * width,
+                             VGfloat * height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!minX || !minY || !width || !height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(minX) || !is_aligned(minY) ||
+       !is_aligned(width) || !is_aligned(height)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_PATH_TRANSFORMED_BOUNDS)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+#if 0
+   /* faster, but seems to have precision problems... */
+   path_bounding_rect(p, minX, minY, width, height);
+   if (*width > 0 && *height > 0) {
+      VGfloat pts[] = {*minX,          *minY,
+                       *minX + *width, *minY,
+                       *minX + *width, *minY + *height,
+                       *minX,          *minY + *height};
+      struct matrix *matrix = &ctx->state.vg.path_user_to_surface_matrix;
+      VGfloat maxX, maxY;
+      matrix_map_point(matrix, pts[0], pts[1], pts + 0, pts + 1);
+      matrix_map_point(matrix, pts[2], pts[3], pts + 2, pts + 3);
+      matrix_map_point(matrix, pts[4], pts[5], pts + 4, pts + 5);
+      matrix_map_point(matrix, pts[6], pts[7], pts + 6, pts + 7);
+      *minX = MIN2(pts[0], MIN2(pts[2], MIN2(pts[4], pts[6])));
+      *minY = MIN2(pts[1], MIN2(pts[3], MIN2(pts[5], pts[7])));
+      maxX = MAX2(pts[0], MAX2(pts[2], MAX2(pts[4], pts[6])));
+      maxY = MAX2(pts[1], MAX2(pts[3], MAX2(pts[5], pts[7])));
+      *width  = maxX - *minX;
+      *height = maxY - *minY;
+   }
+#else
+   {
+      struct path *dst = path_create(VG_PATH_DATATYPE_F, 1.0, 0,
+                                     0, 0, VG_PATH_CAPABILITY_ALL);
+      path_transform(dst, p);
+      path_bounding_rect(dst, minX, minY, width, height);
+      path_destroy(dst);
+   }
+#endif
+}
+
+
+void vgDrawPath(VGPath path, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!(paintModes & (VG_STROKE_PATH | VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (path_is_empty((struct path*)path))
+      return;
+   path_render((struct path*)path, paintModes);
+}
+
diff --git a/src/gallium/state_trackers/vega/api_text.c b/src/gallium/state_trackers/vega/api_text.c
new file mode 100644
index 0000000000..d8411cf3e8
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_text.c
@@ -0,0 +1,258 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "util/u_memory.h"
+
+#ifdef OPENVG_VERSION_1_1
+
+struct vg_font {
+   struct vg_object base;
+
+   VGint glyph_indices[200];
+   VGint num_glyphs;
+};
+
+VGFont vgCreateFont(VGint glyphCapacityHint)
+{
+   struct vg_font *font = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (glyphCapacityHint < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   font = CALLOC_STRUCT(vg_font);
+   vg_init_object(&font->base, ctx, VG_OBJECT_FONT);
+   vg_context_add_object(ctx, VG_OBJECT_FONT, font);
+   return (VGFont)font;
+}
+
+void vgDestroyFont(VGFont f)
+{
+   struct vg_font *font = (struct vg_font *)f;
+   struct vg_context *ctx = vg_current_context();
+
+   if (f == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   vg_context_remove_object(ctx, VG_OBJECT_FONT, font);
+   /*free(font);*/
+}
+
+void vgSetGlyphToPath(VGFont font,
+                      VGuint glyphIndex,
+                      VGPath path,
+                      VGboolean isHinted,
+                      VGfloat glyphOrigin [2],
+                      VGfloat escapement[2])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_object *pathObj;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!glyphOrigin || !escapement ||
+       !is_aligned(glyphOrigin) || !is_aligned(escapement)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (path != VG_INVALID_HANDLE &&
+       !vg_context_is_object_valid(ctx, VG_OBJECT_PATH, (void *)path)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   pathObj = (struct vg_object*)path;
+   if (pathObj && pathObj->type != VG_OBJECT_PATH) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   f = (struct vg_font*)font;
+   f->glyph_indices[f->num_glyphs] = glyphIndex;
+   ++f->num_glyphs;
+}
+
+void vgSetGlyphToImage(VGFont font,
+                       VGuint glyphIndex,
+                       VGImage image,
+                       VGfloat glyphOrigin [2],
+                       VGfloat escapement[2])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_object *img_obj;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!glyphOrigin || !escapement ||
+       !is_aligned(glyphOrigin) || !is_aligned(escapement)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (image != VG_INVALID_HANDLE &&
+       !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)image)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   img_obj = (struct vg_object*)image;
+   if (img_obj && img_obj->type != VG_OBJECT_IMAGE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   f->glyph_indices[f->num_glyphs] = glyphIndex;
+   ++f->num_glyphs;
+}
+
+static INLINE VGboolean font_contains_glyph(struct vg_font *font,
+                                            VGuint glyph_index)
+{
+   VGint i;
+   for (i = 0; i < font->num_glyphs; ++i) {
+      if (font->glyph_indices[i] == glyph_index) {
+         return VG_TRUE;
+      }
+   }
+   return VG_FALSE;
+}
+
+void vgClearGlyph(VGFont font,
+                  VGuint glyphIndex)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_font *f;
+   VGint i;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphIndex <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   if (!font_contains_glyph(f, glyphIndex)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < f->num_glyphs; ++i) {
+      if (f->glyph_indices[i] == glyphIndex) {
+         /*FIXME*/
+         f->glyph_indices[f->num_glyphs] = 0;
+         --f->num_glyphs;
+         return;
+      }
+   }
+}
+
+void vgDrawGlyph(VGFont font,
+                 VGuint glyphIndex,
+                 VGbitfield paintModes,
+                 VGboolean allowAutoHinting)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphIndex <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   if (!font_contains_glyph(f, glyphIndex)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgDrawGlyphs(VGFont font,
+                  VGint glyphCount,
+                  VGuint *glyphIndices,
+                  VGfloat *adjustments_x,
+                  VGfloat *adjustments_y,
+                  VGbitfield paintModes,
+                  VGboolean allowAutoHinting)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGint i;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphCount <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!glyphIndices || !is_aligned(glyphIndices)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!adjustments_x || !is_aligned(adjustments_x) ||
+       !adjustments_y || !is_aligned(adjustments_y)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   f = (struct vg_font*)font;
+   for (i = 0; i < glyphCount; ++i) {
+      VGuint glyph_index = glyphIndices[i];
+      if (!font_contains_glyph(f, glyph_index)) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return;
+      }
+   }
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_transform.c b/src/gallium/state_trackers/vega/api_transform.c
new file mode 100644
index 0000000000..763a5ec415
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_transform.c
@@ -0,0 +1,128 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "matrix.h"
+
+void vgLoadIdentity(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct  matrix *mat = vg_state_matrix(&ctx->state.vg);
+   matrix_load_identity(mat);
+}
+
+void vgLoadMatrix(const VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct  matrix *mat;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   mat = vg_state_matrix(&ctx->state.vg);
+   matrix_init(mat, m);
+   if (!matrix_is_affine(mat)) {
+      if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) {
+         matrix_make_affine(mat);
+      }
+   }
+}
+
+void vgGetMatrix(VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   mat = vg_state_matrix(&ctx->state.vg);
+   memcpy(m, mat->m, sizeof(VGfloat)*9);
+}
+
+void vgMultMatrix(const VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst, src;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   matrix_init(&src, m);
+   dst = vg_state_matrix(&ctx->state.vg);
+   if (!matrix_is_affine(&src)) {
+      if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) {
+         matrix_make_affine(&src);
+      }
+   }
+   matrix_mult(dst, &src);
+
+}
+
+void vgTranslate(VGfloat tx, VGfloat ty)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_translate(dst, tx, ty);
+}
+
+void vgScale(VGfloat sx, VGfloat sy)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_scale(dst, sx, sy);
+}
+
+void vgShear(VGfloat shx, VGfloat shy)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_shear(dst, shx, shy);
+}
+
+void vgRotate(VGfloat angle)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_rotate(dst, angle);
+}
diff --git a/src/gallium/state_trackers/vega/arc.c b/src/gallium/state_trackers/vega/arc.c
new file mode 100644
index 0000000000..e74c7f0334
--- /dev/null
+++ b/src/gallium/state_trackers/vega/arc.c
@@ -0,0 +1,708 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "arc.h"
+
+#include "matrix.h"
+#include "bezier.h"
+#include "polygon.h"
+#include "stroker.h"
+#include "path.h"
+
+#include "util/u_debug.h"
+
+#include <math.h>
+
+#ifndef M_PI
+#define M_PI 3.14159265358979323846
+#endif
+
+#define DEBUG_ARCS 0
+
+static const VGfloat two_pi = M_PI * 2;
+
+
+static const double coeffs3Low[2][4][4] = {
+   {
+      {  3.85268,   -21.229,      -0.330434,    0.0127842  },
+      { -1.61486,     0.706564,    0.225945,    0.263682   },
+      { -0.910164,    0.388383,    0.00551445,  0.00671814 },
+      { -0.630184,    0.192402,    0.0098871,   0.0102527  }
+   },
+   {
+      { -0.162211,    9.94329,     0.13723,     0.0124084  },
+      { -0.253135,    0.00187735,  0.0230286,   0.01264    },
+      { -0.0695069,  -0.0437594,   0.0120636,   0.0163087  },
+      { -0.0328856,  -0.00926032, -0.00173573,  0.00527385 }
+   }
+};
+
+/* coefficients for error estimation
+   while using cubic Bézier curves for approximation
+   1/4 <= b/a <= 1 */
+static const double coeffs3High[2][4][4] = {
+   {
+      {  0.0899116, -19.2349,     -4.11711,     0.183362   },
+      {  0.138148,   -1.45804,     1.32044,     1.38474    },
+      {  0.230903,   -0.450262,    0.219963,    0.414038   },
+      {  0.0590565,  -0.101062,    0.0430592,   0.0204699  }
+   },
+   {
+      {  0.0164649,   9.89394,     0.0919496,   0.00760802 },
+      {  0.0191603,  -0.0322058,   0.0134667,  -0.0825018  },
+      {  0.0156192,  -0.017535,    0.00326508, -0.228157   },
+      { -0.0236752,   0.0405821,  -0.0173086,   0.176187   }
+   }
+};
+
+/* safety factor to convert the "best" error approximation
+   into a "max bound" error */
+static const double safety3[] = {
+   0.001, 4.98, 0.207, 0.0067
+};
+
+/* The code below is from the OpenVG 1.1 Spec
+ * Section 18.4 */
+
+/* Given: Points (x0, y0) and (x1, y1)
+ * Return: TRUE if a solution exists, FALSE otherwise
+ *         Circle centers are written to (cx0, cy0) and (cx1, cy1)
+ */
+static VGboolean
+find_unit_circles(double x0, double y0, double x1, double y1,
+                  double *cx0, double *cy0,
+                  double *cx1, double *cy1)
+{
+   /* Compute differences and averages */
+   double dx = x0 - x1;
+   double dy = y0 - y1;
+   double xm = (x0 + x1)/2;
+   double ym = (y0 + y1)/2;
+   double dsq, disc, s, sdx, sdy;
+
+   /* Solve for intersecting unit circles */
+   dsq = dx*dx + dy*dy;
+   if (dsq == 0.0) return VG_FALSE; /* Points are coincident */
+   disc = 1.0/dsq - 1.0/4.0;
+
+   /* the precision we care about here is around float so if we're
+    * around the float defined zero then make it official to avoid
+    * precision problems later on */
+   if (floatIsZero(disc))
+      disc = 0.0;
+
+   if (disc < 0.0) return VG_FALSE; /* Points are too far apart */
+   s = sqrt(disc);
+   sdx = s*dx;
+   sdy = s*dy;
+   *cx0 = xm + sdy;
+   *cy0 = ym - sdx;
+   *cx1 = xm - sdy;
+   *cy1 = ym + sdx;
+   return VG_TRUE;
+}
+
+
+/* Given:  Ellipse parameters rh, rv, rot (in degrees),
+ *         endpoints (x0, y0) and (x1, y1)
+ * Return: TRUE if a solution exists, FALSE otherwise
+ *         Ellipse centers are written to (cx0, cy0) and (cx1, cy1)
+ */
+static VGboolean
+find_ellipses(double rh, double rv, double rot,
+              double x0, double y0, double x1, double y1,
+              double *cx0, double *cy0, double *cx1, double *cy1)
+{
+   double COS, SIN, x0p, y0p, x1p, y1p, pcx0, pcy0, pcx1, pcy1;
+   /* Convert rotation angle from degrees to radians */
+   rot *= M_PI/180.0;
+   /* Pre-compute rotation matrix entries */
+   COS = cos(rot); SIN = sin(rot);
+   /* Transform (x0, y0) and (x1, y1) into unit space */
+   /* using (inverse) rotate, followed by (inverse) scale   */
+   x0p = (x0*COS + y0*SIN)/rh;
+   y0p = (-x0*SIN + y0*COS)/rv;
+   x1p = (x1*COS + y1*SIN)/rh;
+   y1p = (-x1*SIN + y1*COS)/rv;
+   if (!find_unit_circles(x0p, y0p, x1p, y1p,
+                          &pcx0, &pcy0, &pcx1, &pcy1)) {
+      return VG_FALSE;
+   }
+   /* Transform back to original coordinate space */
+   /* using (forward) scale followed by (forward) rotate */
+   pcx0 *= rh; pcy0 *= rv;
+   pcx1 *= rh; pcy1 *= rv;
+   *cx0 = pcx0*COS - pcy0*SIN;
+   *cy0 = pcx0*SIN + pcy0*COS;
+   *cx1 = pcx1*COS - pcy1*SIN;
+   *cy1 = pcx1*SIN + pcy1*COS;
+   return VG_TRUE;
+}
+
+static INLINE VGboolean
+try_to_fix_radii(struct arc *arc)
+{
+   double COS, SIN, rot, x0p, y0p, x1p, y1p;
+   double dx, dy, dsq, scale;
+
+   /* Convert rotation angle from degrees to radians */
+   rot = DEGREES_TO_RADIANS(arc->theta);
+
+   /* Pre-compute rotation matrix entries */
+   COS = cos(rot); SIN = sin(rot);
+
+   /* Transform (x0, y0) and (x1, y1) into unit space */
+   /* using (inverse) rotate, followed by (inverse) scale   */
+   x0p = (arc->x1*COS + arc->y1*SIN)/arc->a;
+   y0p = (-arc->x1*SIN + arc->y1*COS)/arc->b;
+   x1p = (arc->x2*COS + arc->y2*SIN)/arc->a;
+   y1p = (-arc->x2*SIN + arc->y2*COS)/arc->b;
+   /* Compute differences and averages */
+   dx = x0p - x1p;
+   dy = y0p - y1p;
+
+   dsq = dx*dx + dy*dy;
+#if 0
+   if (dsq <= 0.001) {
+      debug_printf("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAaaaaa\n");
+   }
+#endif
+   scale = 1/(2/sqrt(dsq));
+   arc->a *= scale;
+   arc->b *= scale;
+   return VG_TRUE;
+}
+
+static INLINE double vector_normalize(double *v)
+{
+   double sq = v[0] * v[0] + v[1] * v[1];
+   return sqrt(sq);
+}
+static INLINE double vector_orientation(double *v)
+{
+   double norm = vector_normalize(v);
+   double cosa = v[0] / norm;
+   double sina = v[1] / norm;
+   return (sina>=0 ? acos(cosa) : 2*M_PI - acos(cosa));
+}
+static INLINE double vector_dot(double *v0,
+                                double *v1)
+{
+   return v0[0] * v1[0] + v0[1] * v1[1];
+}
+
+static INLINE double vector_angles(double *v0,
+                                   double *v1)
+{
+   double dot = vector_dot(v0, v1);
+   double norm0 = vector_normalize(v0);
+   double norm1 = vector_normalize(v1);
+
+   return acos(dot / (norm0 * norm1));
+}
+
+static VGboolean find_angles(struct arc *arc)
+{
+   double vec0[2], vec1[2];
+   double lambda1, lambda2;
+   double angle;
+   struct matrix matrix;
+
+   if (floatIsZero(arc->a) || floatIsZero(arc->b)) {
+      return VG_FALSE;
+   }
+   /* map the points to an identity circle */
+   matrix_load_identity(&matrix);
+   matrix_scale(&matrix, 1.f, arc->a/arc->b);
+   matrix_rotate(&matrix, -arc->theta);
+   matrix_map_point(&matrix,
+                    arc->x1, arc->y1,
+                    &arc->x1, &arc->y1);
+   matrix_map_point(&matrix,
+                    arc->x2, arc->y2,
+                    &arc->x2, &arc->y2);
+   matrix_map_point(&matrix,
+                    arc->cx, arc->cy,
+                    &arc->cx, &arc->cy);
+
+#if DEBUG_ARCS
+   debug_printf("Matrix 3 [%f, %f, %f| %f, %f, %f| %f, %f, %f]\n",
+                matrix.m[0], matrix.m[1], matrix.m[2],
+                matrix.m[3], matrix.m[4], matrix.m[5],
+                matrix.m[6], matrix.m[7], matrix.m[8]);
+   debug_printf("Endpoints [%f, %f], [%f, %f]\n",
+                arc->x1, arc->y1, arc->x2, arc->y2);
+#endif
+
+   vec0[0] = arc->x1 - arc->cx;
+   vec0[1] = arc->y1 - arc->cy;
+   vec1[0] = arc->x2 - arc->cx;
+   vec1[1] = arc->y2 - arc->cy;
+
+#if DEBUG_ARCS
+   debug_printf("Vec is [%f, %f], [%f, %f], [%f, %f]\n",
+                vec0[0], vec0[1], vec1[0], vec1[1], arc->cx, arc->cy);
+#endif
+
+   lambda1 = vector_orientation(vec0);
+
+   if (isnan(lambda1))
+      lambda1 = 0.f;
+
+   if (arc->type == VG_SCWARC_TO ||
+       arc->type == VG_SCCWARC_TO)
+      angle = vector_angles(vec0, vec1);
+   else if (arc->type == VG_LCWARC_TO ||
+            arc->type == VG_LCCWARC_TO) {
+      angle = 2*M_PI - vector_angles(vec0, vec1);
+   } else
+      abort();
+
+   if (isnan(angle))
+      angle = M_PI;
+
+
+   if (arc->type == VG_SCWARC_TO ||
+       arc->type == VG_LCWARC_TO)
+      lambda2 = lambda1 - angle;
+   else
+      lambda2 = lambda1 + angle;
+
+#if DEBUG_ARCS
+   debug_printf("Angle is %f and (%f, %f)\n", angle, lambda1, lambda2);
+#endif
+
+#if 0
+   arc->eta1 = atan2(sin(lambda1) / arc->b,
+                     cos(lambda1) / arc->a);
+   arc->eta2 = atan2(sin(lambda2) / arc->b,
+                     cos(lambda2) / arc->a);
+
+   /* make sure we have eta1 <= eta2 <= eta1 + 2 PI */
+   arc->eta2 -= two_pi * floor((arc->eta2 - arc->eta1) / two_pi);
+
+   /* the preceding correction fails if we have exactly et2 - eta1 = 2 PI
+      it reduces the interval to zero length */
+   if ((lambda2 - lambda1 > M_PI) && (arc->eta2 - arc->eta1 < M_PI)) {
+      arc->eta2 += 2 * M_PI;
+   }
+#else
+   arc->eta1 = lambda1;
+   arc->eta2 = lambda2;
+#endif
+
+   return VG_TRUE;
+}
+
+#if DEBUG_ARCS
+static void check_endpoints(struct arc *arc)
+{
+   double x1, y1, x2, y2;
+
+   double a_cos_eta1 = arc->a * cos(arc->eta1);
+   double b_sin_eta1 = arc->b * sin(arc->eta1);
+   x1 = arc->cx + a_cos_eta1 * arc->cos_theta -
+        b_sin_eta1 * arc->sin_theta;
+   y1 = arc->cy + a_cos_eta1 * arc->sin_theta +
+        b_sin_eta1 * arc->cos_theta;
+
+   double a_cos_eta2 = arc->a * cos(arc->eta2);
+   double b_sin_eta2 = arc->b * sin(arc->eta2);
+   x2 = arc->cx + a_cos_eta2 * arc->cos_theta -
+        b_sin_eta2 * arc->sin_theta;
+   y2 = arc->cy + a_cos_eta2 * arc->sin_theta +
+        b_sin_eta2 * arc->cos_theta;
+
+   debug_printf("Computed (%f, %f), (%f, %f)\n",
+                x1, y1, x2, y2);
+   debug_printf("Real     (%f, %f), (%f, %f)\n",
+                arc->x1, arc->y1,
+                arc->x2, arc->y2);
+}
+#endif
+
+void arc_init(struct arc *arc,
+              VGPathSegment type,
+              VGfloat x1, VGfloat y1,
+              VGfloat x2, VGfloat y2,
+              VGfloat rh, VGfloat rv,
+              VGfloat rot)
+{
+   assert(type == VG_SCCWARC_TO ||
+          type == VG_SCWARC_TO ||
+          type == VG_LCCWARC_TO ||
+          type == VG_LCWARC_TO);
+   arc->type = type;
+   arc->x1  = x1;
+   arc->y1  = y1;
+   arc->x2  = x2;
+   arc->y2  = y2;
+   arc->a   = rh;
+   arc->b   = rv;
+   arc->theta = rot;
+   arc->cos_theta = cos(arc->theta);
+   arc->sin_theta = sin(arc->theta);
+   {
+      double cx0, cy0, cx1, cy1;
+      double cx, cy;
+      arc->is_valid =  find_ellipses(rh, rv, rot, x1, y1, x2, y2,
+                                     &cx0, &cy0, &cx1, &cy1);
+
+      if (!arc->is_valid && try_to_fix_radii(arc)) {
+         rh = arc->a;
+         rv = arc->b;
+         arc->is_valid =
+            find_ellipses(rh, rv, rot, x1, y1, x2, y2,
+                          &cx0, &cy0, &cx1, &cy1);
+      }
+
+      if (type == VG_SCWARC_TO ||
+          type == VG_LCCWARC_TO) {
+         cx = cx1;
+         cy = cy1;
+      } else {
+         cx = cx0;
+         cy = cy0;
+      }
+#if DEBUG_ARCS
+      debug_printf("Centers are : (%f, %f) , (%f, %f). Real (%f, %f)\n",
+                   cx0, cy0, cx1, cy1, cx, cy);
+#endif
+      arc->cx = cx;
+      arc->cy = cy;
+      if (arc->is_valid) {
+         arc->is_valid = find_angles(arc);
+#if DEBUG_ARCS
+         check_endpoints(arc);
+#endif
+         /* remap a few points. find_angles requires
+          * rot in angles, the rest of the code
+          * will need them in radians. and find_angles
+          * modifies the center to match an identity
+          * circle so lets reset it */
+         arc->theta = DEGREES_TO_RADIANS(rot);
+         arc->cos_theta = cos(arc->theta);
+         arc->sin_theta = sin(arc->theta);
+         arc->cx = cx;
+         arc->cy = cy;
+      }
+   }
+}
+
+static INLINE double rational_function(double x, const double *c)
+{
+   return (x * (x * c[0] + c[1]) + c[2]) / (x + c[3]);
+}
+
+static double estimate_error(struct arc *arc,
+                             double etaA, double etaB)
+{
+   double eta  = 0.5 * (etaA + etaB);
+
+   double x    = arc->b / arc->a;
+   double dEta = etaB - etaA;
+   double cos2 = cos(2 * eta);
+   double cos4 = cos(4 * eta);
+   double cos6 = cos(6 * eta);
+   double c0, c1;
+
+   /* select the right coeficients set according to degree and b/a */
+   const double (*coeffs)[4][4];
+   const double *safety;
+   coeffs = (x < 0.25) ? coeffs3Low : coeffs3High;
+   safety = safety3;
+
+   c0 = rational_function(x, coeffs[0][0])
+        + cos2 * rational_function(x, coeffs[0][1])
+        + cos4 * rational_function(x, coeffs[0][2])
+        + cos6 * rational_function(x, coeffs[0][3]);
+
+   c1 = rational_function(x, coeffs[1][0])
+        + cos2 * rational_function(x, coeffs[1][1])
+        + cos4 * rational_function(x, coeffs[1][2])
+        + cos6 * rational_function(x, coeffs[1][3]);
+
+   return rational_function(x, safety) * arc->a * exp(c0 + c1 * dEta);
+}
+
+struct arc_cb {
+   void (*move)(struct arc_cb *cb, VGfloat x, VGfloat y);
+   void (*point)(struct arc_cb *cb, VGfloat x, VGfloat y);
+   void (*bezier)(struct arc_cb *cb, struct bezier *bezier);
+
+   void *user_data;
+};
+
+static void cb_null_move(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+}
+
+static void polygon_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct polygon *poly = (struct polygon*)cb->user_data;
+   polygon_vertex_append(poly, x, y);
+}
+
+static void polygon_bezier(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct polygon *poly = (struct polygon*)cb->user_data;
+   bezier_add_to_polygon(bezier, poly);
+}
+
+static void stroke_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_line_to(stroker, x, y);
+}
+
+static void stroke_curve(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_curve_to(stroker,
+                    bezier->x2, bezier->y2,
+                    bezier->x3, bezier->y3,
+                    bezier->x4, bezier->y4);
+}
+
+static void stroke_emit_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_emit_line_to(stroker, x, y);
+}
+
+static void stroke_emit_curve(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_emit_curve_to(stroker,
+                         bezier->x2, bezier->y2,
+                         bezier->x3, bezier->y3,
+                         bezier->x4, bezier->y4);
+}
+
+static void arc_path_move(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_move_to(path, x, y);
+}
+
+static void arc_path_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_line_to(path, x, y);
+}
+
+static void arc_path_bezier(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_cubic_to(path,
+                 bezier->x2, bezier->y2,
+                 bezier->x3, bezier->y3,
+                 bezier->x4, bezier->y4);
+}
+
+static INLINE int num_beziers_needed(struct arc *arc)
+{
+   double threshold = 0.05;
+   VGboolean found = VG_FALSE;
+   int n = 1;
+   double min_eta, max_eta;
+
+   min_eta = MIN2(arc->eta1, arc->eta2);
+   max_eta = MAX2(arc->eta1, arc->eta2);
+
+   while ((! found) && (n < 1024)) {
+      double d_eta = (max_eta - min_eta) / n;
+      if (d_eta <= 0.5 * M_PI) {
+         double eta_b = min_eta;
+         found = VG_TRUE;
+         for (int i = 0; found && (i < n); ++i) {
+            double etaA = eta_b;
+            eta_b += d_eta;
+            found = (estimate_error(arc, etaA, eta_b) <= threshold);
+         }
+      }
+      n = n << 1;
+   }
+
+   return n;
+}
+
+static void arc_to_beziers(struct arc *arc,
+                           struct arc_cb cb,
+                           struct matrix *matrix)
+{
+   int n = 1;
+   double d_eta, eta_b, cos_eta_b,
+      sin_eta_b, a_cos_eta_b, b_sin_eta_b, a_sin_eta_b,
+      b_cos_eta_b, x_b, y_b, x_b_dot, y_b_dot, lx, ly;
+   double t, alpha;
+
+   { /* always move to the start of the arc */
+      VGfloat x = arc->x1;
+      VGfloat y = arc->y1;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.move(&cb, x, y);
+   }
+
+   if (!arc->is_valid) {
+      VGfloat x = arc->x2;
+      VGfloat y = arc->y2;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.point(&cb, x, y);
+      return;
+   }
+
+   /* find the number of Bézier curves needed */
+   n = num_beziers_needed(arc);
+
+   d_eta = (arc->eta2 - arc->eta1) / n;
+   eta_b = arc->eta1;
+
+   cos_eta_b  = cos(eta_b);
+   sin_eta_b  = sin(eta_b);
+   a_cos_eta_b = arc->a * cos_eta_b;
+   b_sin_eta_b = arc->b * sin_eta_b;
+   a_sin_eta_b = arc->a * sin_eta_b;
+   b_cos_eta_b = arc->b * cos_eta_b;
+   x_b       = arc->cx + a_cos_eta_b * arc->cos_theta -
+               b_sin_eta_b * arc->sin_theta;
+   y_b       = arc->cy + a_cos_eta_b * arc->sin_theta +
+               b_sin_eta_b * arc->cos_theta;
+   x_b_dot    = -a_sin_eta_b * arc->cos_theta -
+                b_cos_eta_b * arc->sin_theta;
+   y_b_dot    = -a_sin_eta_b * arc->sin_theta +
+                b_cos_eta_b * arc->cos_theta;
+
+   {
+      VGfloat x = x_b, y = y_b;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.point(&cb, x, y);
+   }
+   lx = x_b;
+   ly = y_b;
+
+   t     = tan(0.5 * d_eta);
+   alpha = sin(d_eta) * (sqrt(4 + 3 * t * t) - 1) / 3;
+
+   for (int i = 0; i < n; ++i) {
+      struct bezier bezier;
+      double xA    = x_b;
+      double yA    = y_b;
+      double xADot = x_b_dot;
+      double yADot = y_b_dot;
+
+      eta_b    += d_eta;
+      cos_eta_b  = cos(eta_b);
+      sin_eta_b  = sin(eta_b);
+      a_cos_eta_b = arc->a * cos_eta_b;
+      b_sin_eta_b = arc->b * sin_eta_b;
+      a_sin_eta_b = arc->a * sin_eta_b;
+      b_cos_eta_b = arc->b * cos_eta_b;
+      x_b       = arc->cx + a_cos_eta_b * arc->cos_theta -
+                  b_sin_eta_b * arc->sin_theta;
+      y_b       = arc->cy + a_cos_eta_b * arc->sin_theta +
+                  b_sin_eta_b * arc->cos_theta;
+      x_b_dot    = -a_sin_eta_b * arc->cos_theta -
+                   b_cos_eta_b * arc->sin_theta;
+      y_b_dot    = -a_sin_eta_b * arc->sin_theta +
+                   b_cos_eta_b * arc->cos_theta;
+
+      bezier_init(&bezier,
+                  lx, ly,
+                  (float) (xA + alpha * xADot), (float) (yA + alpha * yADot),
+                  (float) (x_b - alpha * x_b_dot), (float) (y_b - alpha * y_b_dot),
+                  (float) x_b,                   (float) y_b);
+#if 0
+      debug_printf("%d) Bezier (%f, %f), (%f, %f), (%f, %f), (%f, %f)\n",
+                   i,
+                   bezier.x1, bezier.y1,
+                   bezier.x2, bezier.y2,
+                   bezier.x3, bezier.y3,
+                   bezier.x4, bezier.y4);
+#endif
+      bezier_transform(&bezier, matrix);
+      cb.bezier(&cb, &bezier);
+      lx = x_b;
+      ly = y_b;
+   }
+}
+
+
+void arc_add_to_polygon(struct arc *arc,
+                        struct polygon *poly,
+                        struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = polygon_point;
+   cb.bezier = polygon_bezier;
+   cb.user_data = poly;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_stroke_cb(struct arc *arc,
+                   struct stroker *stroke,
+                   struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = stroke_point;
+   cb.bezier = stroke_curve;
+   cb.user_data = stroke;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_stroker_emit(struct arc *arc,
+                      struct stroker *stroker,
+                      struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = stroke_emit_point;
+   cb.bezier = stroke_emit_curve;
+   cb.user_data = stroker;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_to_path(struct arc *arc,
+                 struct path *path,
+                 struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = arc_path_move;
+   cb.point = arc_path_point;
+   cb.bezier = arc_path_bezier;
+   cb.user_data = path;
+
+   arc_to_beziers(arc, cb, matrix);
+}
diff --git a/src/gallium/state_trackers/vega/arc.h b/src/gallium/state_trackers/vega/arc.h
new file mode 100644
index 0000000000..3205cd5021
--- /dev/null
+++ b/src/gallium/state_trackers/vega/arc.h
@@ -0,0 +1,80 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ARC_H
+#define ARC_H
+
+#include "VG/openvg.h"
+
+struct polygon;
+struct matrix;
+struct stroker;
+struct path;
+
+struct arc {
+   VGPathSegment type;
+
+   VGfloat cx, cy;
+
+   VGfloat a, b;
+
+   VGfloat theta;
+   VGfloat cos_theta, sin_theta;
+
+   VGfloat eta1;
+   VGfloat eta2;
+
+   VGfloat x1, y1, x2, y2;
+
+   VGboolean is_valid;
+};
+
+void arc_init(struct arc *arc,
+              VGPathSegment type,
+              VGfloat x1, VGfloat y1,
+              VGfloat x2, VGfloat y2,
+              VGfloat rh, VGfloat rv,
+              VGfloat rot);
+
+void arc_add_to_polygon(struct arc *arc,
+                        struct polygon *poly,
+                        struct matrix *matrix);
+
+
+void arc_to_path(struct arc *arc,
+                 struct path *p,
+                 struct matrix *matrix);
+
+void arc_stroke_cb(struct arc *arc,
+                   struct stroker *stroke,
+                   struct matrix *matrix);
+
+void arc_stroker_emit(struct arc *arc,
+                      struct stroker *stroke,
+                      struct matrix *matrix);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_fill.h b/src/gallium/state_trackers/vega/asm_fill.h
new file mode 100644
index 0000000000..2f394ad6c5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_fill.h
@@ -0,0 +1,246 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_FILL_H
+#define ASM_FILL_H
+
+static const char solid_fill_asm[] =
+   "MOV %s, CONST[0]\n";
+
+
+static const char linear_grad_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[4].x, TEMP[1]\n"
+   "MOV TEMP[4].y, TEMP[2]\n"
+   "MUL TEMP[0], CONST[0].yyyy, TEMP[4].yyyy\n"
+   "MAD TEMP[1], CONST[0].xxxx, TEMP[4].xxxx, TEMP[0]\n"
+   "MUL TEMP[2], TEMP[1], CONST[0].zzzz\n"
+   "TEX %s, TEMP[2], SAMP[0], 1D\n";
+
+static const char radial_grad_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[5].x, TEMP[1]\n"
+   "MOV TEMP[5].y, TEMP[2]\n"
+   "MUL TEMP[0], CONST[0].yyyy, TEMP[5].yyyy\n"
+   "MAD TEMP[1], CONST[0].xxxx, TEMP[5].xxxx, TEMP[0]\n"
+   "ADD TEMP[1], TEMP[1], TEMP[1]\n"
+   "MUL TEMP[3], TEMP[5].yyyy, TEMP[5].yyyy\n"
+   "MAD TEMP[4], TEMP[5].xxxx, TEMP[5].xxxx, TEMP[3]\n"
+   "MOV TEMP[4], -TEMP[4]\n"
+   "MUL TEMP[2], CONST[0].zzzz, TEMP[4]\n"
+   "MUL TEMP[0], CONST[1].wwww, TEMP[2]\n"
+   "MUL TEMP[3], TEMP[1], TEMP[1]\n"
+   "SUB TEMP[2], TEMP[3], TEMP[0]\n"
+   "RSQ TEMP[2], |TEMP[2]|\n"
+   "RCP TEMP[2], TEMP[2]\n"
+   "SUB TEMP[1], TEMP[2], TEMP[1]\n"
+   "ADD TEMP[0], CONST[0].zzzz, CONST[0].zzzz\n"
+   "RCP TEMP[0], TEMP[0]\n"
+   "MUL TEMP[2], TEMP[1], TEMP[0]\n"
+   "TEX %s, TEMP[2], SAMP[0], 1D\n";
+
+static const char pattern_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[4].x, TEMP[1]\n"
+   "MOV TEMP[4].y, TEMP[2]\n"
+   "RCP TEMP[0], CONST[1].zwzw\n"
+   "MOV TEMP[1], TEMP[4]\n"
+   "MUL TEMP[1].x, TEMP[1], TEMP[0]\n"
+   "MUL TEMP[1].y, TEMP[1], TEMP[0]\n"
+   "TEX %s, TEMP[1], SAMP[0], 2D\n";
+
+
+static const char mask_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[1], 2D\n"
+   "MUL TEMP[0].w, TEMP[0].wwww, TEMP[1].wwww\n"
+   "MOV %s, TEMP[0]\n";
+
+
+static const char image_normal_asm[] =
+   "TEX %s, IN[1], SAMP[3], 2D\n";
+
+static const char image_multiply_asm[] =
+   "TEX TEMP[1], IN[1], SAMP[3], 2D\n"
+   "MUL %s, TEMP[0], TEMP[1]\n";
+
+static const char image_stencil_asm[] =
+   "TEX TEMP[1], IN[1], SAMP[3], 2D\n"
+   "MUL %s, TEMP[0], TEMP[1]\n";
+
+
+#define EXTENDED_BLEND_OVER                     \
+   "SUB TEMP[3], CONST[1].yyyy, TEMP[1].wwww\n" \
+   "SUB TEMP[4], CONST[1].yyyy, TEMP[0].wwww\n" \
+   "MUL TEMP[3], TEMP[0], TEMP[3]\n"            \
+   "MUL TEMP[4], TEMP[1], TEMP[4]\n"            \
+   "ADD TEMP[3], TEMP[3], TEMP[4]\n"
+
+static const char blend_multiply_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1]\n"
+   "ADD TEMP[1], TEMP[4], TEMP[3]\n"/*result.rgb*/
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+#if 1
+static const char blend_screen_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   "ADD TEMP[3], TEMP[0], TEMP[1]\n"
+   "MUL TEMP[2], TEMP[0], TEMP[1]\n"
+   "SUB %s, TEMP[3], TEMP[2]\n";
+#else
+static const char blend_screen_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   "MOV %s, TEMP[1]\n";
+#endif
+
+static const char blend_darken_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n"
+   "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n"
+   "MIN TEMP[4], TEMP[4], TEMP[5]\n"
+   "ADD TEMP[1], TEMP[3], TEMP[4]\n"
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+
+static const char blend_lighten_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n"
+   "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n"
+   "MAX TEMP[4], TEMP[4], TEMP[5]\n"
+   "ADD TEMP[1], TEMP[3], TEMP[4]\n"
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+
+
+static const char premultiply_asm[] =
+   "MUL TEMP[0].xyz, TEMP[0], TEMP[0].wwww\n";
+
+static const char unpremultiply_asm[] =
+   "TEX TEMP[0], IN[0], SAMP[1], 2D\n";
+
+
+static const char color_bw_asm[] =
+   "ADD TEMP[1], CONST[1].yyyy, CONST[1].yyyy\n"
+   "RCP TEMP[2], TEMP[1]\n"
+   "ADD TEMP[1], CONST[1].yyyy, TEMP[2]\n"
+   "ADD TEMP[2].x, TEMP[0].xxxx, TEMP[0].yyyy\n"
+   "ADD TEMP[2].x, TEMP[0].zzzz, TEMP[0].xxxx\n"
+   "SGE TEMP[0].xyz, TEMP[2].xxxx, TEMP[1]\n"
+   "SGE TEMP[0].w, TEMP[0].wwww, TEMP[2].yyyy\n"
+   "MOV %s, TEMP[0]\n";
+
+
+struct shader_asm_info {
+   VGint id;
+   VGint num_tokens;
+   const char * txt;
+
+   VGboolean needs_position;
+
+   VGint start_const;
+   VGint num_consts;
+
+   VGint start_sampler;
+   VGint num_samplers;
+
+   VGint start_temp;
+   VGint num_temps;
+};
+
+
+static const struct shader_asm_info shaders_asm[] = {
+   /* fills */
+   {VEGA_SOLID_FILL_SHADER,       40,  solid_fill_asm,
+    VG_FALSE, 0, 1, 0, 0, 0, 0},
+   {VEGA_LINEAR_GRADIENT_SHADER, 200,  linear_grad_asm,
+    VG_TRUE,  0, 5, 0, 1, 0, 5},
+   {VEGA_RADIAL_GRADIENT_SHADER, 200,  radial_grad_asm,
+    VG_TRUE,  0, 5, 0, 1, 0, 6},
+   {VEGA_PATTERN_SHADER,         100,      pattern_asm,
+    VG_TRUE,  1, 4, 0, 1, 0, 5},
+
+   /* image draw modes */
+   {VEGA_IMAGE_NORMAL_SHADER,    200, image_normal_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 0},
+   {VEGA_IMAGE_MULTIPLY_SHADER,  200, image_multiply_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 2},
+   {VEGA_IMAGE_STENCIL_SHADER,   200, image_stencil_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 2},
+
+   {VEGA_MASK_SHADER,            100,         mask_asm,
+    VG_TRUE,  0, 0, 1, 1, 0, 2},
+
+   /* extra blend modes */
+   {VEGA_BLEND_MULTIPLY_SHADER,  200, blend_multiply_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 5},
+   {VEGA_BLEND_SCREEN_SHADER,    200, blend_screen_asm,
+    VG_TRUE,  0, 0, 2, 1, 0, 4},
+   {VEGA_BLEND_DARKEN_SHADER,    200, blend_darken_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 6},
+   {VEGA_BLEND_LIGHTEN_SHADER,   200, blend_lighten_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 6},
+
+   /* premultiply */
+   {VEGA_PREMULTIPLY_SHADER,   100, premultiply_asm,
+    VG_FALSE,  0, 0, 0, 0, 0, 1},
+   {VEGA_UNPREMULTIPLY_SHADER,   100, unpremultiply_asm,
+    VG_FALSE,  0, 0, 0, 0, 0, 1},
+
+   /* color transform to black and white */
+   {VEGA_BW_SHADER,   150, color_bw_asm,
+    VG_FALSE,  1, 1, 0, 0, 0, 3},
+};
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_filters.h b/src/gallium/state_trackers/vega/asm_filters.h
new file mode 100644
index 0000000000..9a49f2e12d
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_filters.h
@@ -0,0 +1,117 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_FILTERS_H
+#define ASM_FILTERS_H
+
+static const char color_matrix_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL CONST[0..4], CONSTANT\n"
+   "DCL TEMP[0..4], CONSTANT\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "MOV TEMP[1], TEMP[0].xxxx\n"
+   "MOV TEMP[2], TEMP[0].yyyy\n"
+   "MOV TEMP[3], TEMP[0].zzzz\n"
+   "MOV TEMP[4], TEMP[0].wwww\n"
+   "MUL TEMP[1], TEMP[1], CONST[0]\n"
+   "MUL TEMP[2], TEMP[2], CONST[1]\n"
+   "MUL TEMP[3], TEMP[3], CONST[2]\n"
+   "MUL TEMP[4], TEMP[4], CONST[3]\n"
+   "ADD TEMP[0], TEMP[1], CONST[4]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[2]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[3]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[4]\n"
+   "MOV OUT[0], TEMP[0]\n"
+   "END\n";
+
+static const char convolution_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..4], CONSTANT\n"
+   "DCL ADDR[0], CONSTANT\n"
+   "DCL CONST[0..%d], CONSTANT\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "0: MOV TEMP[0], CONST[0].xxxx\n"
+   "1: MOV TEMP[1], CONST[0].xxxx\n"
+   "2: BGNLOOP :14\n"
+   "3: SGE TEMP[0].z, TEMP[0].yyyy, CONST[1].xxxx\n"
+   "4: IF TEMP[0].zzzz :7\n"
+   "5: BRK\n"
+   "6: ENDIF\n"
+   "7: ARL ADDR[0].x, TEMP[0].yyyy\n"
+   "8: MOV TEMP[3], CONST[ADDR[0]+2]\n"
+   "9: ADD TEMP[4].xy, IN[0], TEMP[3]\n"
+   "10: TEX TEMP[2], TEMP[4], SAMP[0], 2D\n"
+   "11: MOV TEMP[3], CONST[ADDR[0]+%d]\n"
+   "12: MAD TEMP[1], TEMP[2], TEMP[3], TEMP[1]\n"
+   "13: ADD TEMP[0].y, TEMP[0].yyyy, CONST[0].yyyy\n"
+   "14: ENDLOOP :2\n"
+   "15: MAD OUT[0], TEMP[1], CONST[1].yyyy, CONST[1].zzzz\n"
+   "16: END\n";
+
+
+static const char lookup_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "MOV TEMP[1], TEMP[0]\n"
+   /* do red */
+   "TEX TEMP[2], TEMP[1].xxxx, SAMP[1], 1D\n"
+   "MOV TEMP[0].x, TEMP[2].xxxx\n"
+   /* do blue */
+   "TEX TEMP[2], TEMP[1].yyyy, SAMP[1], 1D\n"
+   "MOV TEMP[0].y, TEMP[2].yyyy\n"
+   /* do green */
+   "TEX TEMP[2], TEMP[1].zzzz, SAMP[1], 1D\n"
+   "MOV TEMP[0].z, TEMP[2].zzzz\n"
+   /* do alpha */
+   "TEX TEMP[2], TEMP[1].wwww, SAMP[1], 1D\n"
+   "MOV TEMP[0].w, TEMP[2].wwww\n"
+   "MOV OUT[0], TEMP[0]\n"
+   "END\n";
+
+
+static const char lookup_single_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "TEX TEMP[1], TEMP[0].%s, SAMP[1], 1D\n"
+   "MOV OUT[0], TEMP[1]\n"
+   "END\n";
+
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_util.h b/src/gallium/state_trackers/vega/asm_util.h
new file mode 100644
index 0000000000..218e1d166d
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_util.h
@@ -0,0 +1,136 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_UTIL_H
+#define ASM_UTIL_H
+
+
+static const char pass_through_depth_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], POSITION, LINEAR\n"
+   "DCL OUT[0].z, POSITION, CONSTANT\n"
+   "0: MOV OUT[0].z, IN[0].zzzz\n"
+   "1: END\n";
+
+
+
+/* μnew = μmask */
+static const char set_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX OUT[0], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: END\n";
+
+/* μnew = 1 – (1 – μmask)*(1 – μprev) */
+static const char union_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..3], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "2: SUB TEMP[2], CONST[0], TEMP[0]\n"
+   "3: SUB TEMP[3], CONST[0], TEMP[1]\n"
+   "4: MUL TEMP[0].w, TEMP[2].wwww, TEMP[3].wwww\n"
+   "5: SUB OUT[0], CONST[0], TEMP[0]\n"
+   "6: END\n";
+
+/* μnew = μmask *μprev */
+static const char intersect_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..1], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "1: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "2: MUL OUT[0], TEMP[0].wwww, TEMP[1].wwww\n"
+   "3: END\n";
+
+/* μnew = μprev*(1 – μmask) */
+static const char subtract_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "2: SUB TEMP[2], CONST[0], TEMP[1]\n"
+   "3: MUL OUT[0], TEMP[2].wwww, TEMP[0].wwww\n"
+   "4: END\n";
+
+
+static const char vs_plain_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: END\n";
+
+static const char vs_clear_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL IN[1]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL OUT[1], COLOR\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: MOV OUT[1], IN[1]\n"
+   "4: END\n";
+
+
+static const char vs_texture_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL IN[1]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL OUT[1], GENERIC\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: MOV OUT[1], IN[1]\n"
+   "4: END\n";
+
+#endif
diff --git a/src/gallium/state_trackers/vega/bezier.c b/src/gallium/state_trackers/vega/bezier.c
new file mode 100644
index 0000000000..39a7ade016
--- /dev/null
+++ b/src/gallium/state_trackers/vega/bezier.c
@@ -0,0 +1,704 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "bezier.h"
+
+#include "matrix.h"
+#include "polygon.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_debug.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <assert.h>
+#include <math.h>
+
+static const float flatness = 0.5;
+
+
+static INLINE void split_left(struct bezier *bez, VGfloat t, struct bezier* left)
+{
+    left->x1 = bez->x1;
+    left->y1 = bez->y1;
+
+    left->x2 = bez->x1 + t * (bez->x2 - bez->x1);
+    left->y2 = bez->y1 + t * (bez->y2 - bez->y1);
+
+    left->x3 = bez->x2 + t * (bez->x3 - bez->x2);
+    left->y3 = bez->y2 + t * (bez->y3 - bez->y2);
+
+    bez->x3 = bez->x3 + t * (bez->x4 - bez->x3);
+    bez->y3 = bez->y3 + t * (bez->y4 - bez->y3);
+
+    bez->x2 = left->x3 + t * (bez->x3 - left->x3);
+    bez->y2 = left->y3 + t * (bez->y3 - left->y3);
+
+    left->x3 = left->x2 + t * (left->x3 - left->x2);
+    left->y3 = left->y2 + t * (left->y3 - left->y2);
+
+    left->x4 = bez->x1 = left->x3 + t * (bez->x2 - left->x3);
+    left->y4 = bez->y1 = left->y3 + t * (bez->y2 - left->y3);
+}
+
+static INLINE void split(struct bezier *bez,
+                         struct bezier *first_half,
+                         struct bezier *second_half)
+{
+   double c         = (bez->x2 + bez->x3) * 0.5;
+   first_half->x2  = (bez->x1 + bez->x2) * 0.5;
+   second_half->x3 = (bez->x3 + bez->x4) * 0.5;
+   first_half->x1  = bez->x1;
+   second_half->x4 = bez->x4;
+   first_half->x3  = (first_half->x2 + c) * 0.5;
+   second_half->x2 = (second_half->x3 + c) * 0.5;
+   first_half->x4  = second_half->x1 =
+                     (first_half->x3 + second_half->x2) * 0.5;
+
+   c = (bez->y2 + bez->y3) / 2;
+   first_half->y2  = (bez->y1 + bez->y2) * 0.5;
+   second_half->y3 = (bez->y3 + bez->y4) * 0.5;
+   first_half->y1  = bez->y1;
+   second_half->y4 = bez->y4;
+   first_half->y3  = (first_half->y2 + c) * 0.5;
+   second_half->y2 = (second_half->y3 + c) * 0.5;
+   first_half->y4  = second_half->y1 =
+                     (first_half->y3 + second_half->y2) * 0.5;
+}
+
+struct polygon * bezier_to_polygon(struct bezier *bez)
+{
+   struct polygon *poly = polygon_create(64);
+   polygon_vertex_append(poly, bez->x1, bez->y1);
+   bezier_add_to_polygon(bez, poly);
+   return poly;
+}
+
+void bezier_add_to_polygon(const struct bezier *bez,
+                           struct polygon *poly)
+{
+   struct bezier beziers[32];
+   struct bezier *b;
+
+   beziers[0] = *bez;
+   b = beziers;
+
+   while (b >= beziers) {
+      double y4y1 = b->y4 - b->y1;
+      double x4x1 = b->x4 - b->x1;
+      double l = ABS(x4x1) + ABS(y4y1);
+      double d;
+      if (l > 1.f) {
+         d = ABS((x4x1)*(b->y1 - b->y2) - (y4y1)*(b->x1 - b->x2))
+             + ABS((x4x1)*(b->y1 - b->y3) - (y4y1)*(b->x1 - b->x3));
+      } else {
+         d = ABS(b->x1 - b->x2) + ABS(b->y1 - b->y2) +
+             ABS(b->x1 - b->x3) + ABS(b->y1 - b->y3);
+         l = 1.;
+      }
+      if (d < flatness*l || b == beziers + 31) {
+         /* good enough, we pop it off and add the endpoint */
+         polygon_vertex_append(poly, b->x4, b->y4);
+         --b;
+      } else {
+         /* split, second half of the bezier goes lower into the stack */
+         split(b, b+1, b);
+         ++b;
+      }
+   }
+}
+
+static void add_if_close(struct bezier *bez, VGfloat *length, VGfloat error)
+{
+   struct bezier left, right;     /* bez poly splits */
+   VGfloat len = 0.0;        /* arc length */
+   VGfloat chord;            /* chord length */
+
+   len = len + line_length(bez->x1, bez->y1, bez->x2, bez->y2);
+   len = len + line_length(bez->x2, bez->y2, bez->x3, bez->y3);
+   len = len + line_length(bez->x3, bez->y3, bez->x4, bez->y4);
+
+   chord = line_length(bez->x1, bez->y1, bez->x4, bez->y4);
+
+   if ((len-chord) > error) {
+      split(bez, &left, &right);                 /* split in two */
+      add_if_close(&left, length, error);       /* try left side */
+      add_if_close(&right, length, error);      /* try right side */
+      return;
+   }
+
+   *length = *length + len;
+
+   return;
+}
+
+float bezier_length(struct bezier *bez, float error)
+{
+   VGfloat length = 0.f;
+
+   add_if_close(bez, &length, error);
+   return length;
+}
+
+void bezier_init(struct bezier *bez,
+                 float x1, float y1,
+                 float x2, float y2,
+                 float x3, float y3,
+                 float x4, float y4)
+{
+   bez->x1 = x1;
+   bez->y1 = y1;
+   bez->x2 = x2;
+   bez->y2 = y2;
+   bez->x3 = x3;
+   bez->y3 = y3;
+   bez->x4 = x4;
+   bez->y4 = y4;
+#if 0
+   debug_printf("bezier in [%f, %f, %f, %f, %f, %f]\n",
+                x1, y1, x2, y2, x3, y3, x4, y4);
+#endif
+}
+
+
+static INLINE void bezier_init2v(struct bezier *bez,
+                                 float *pt1,
+                                 float *pt2,
+                                 float *pt3,
+                                 float *pt4)
+{
+   bez->x1 = pt1[0];
+   bez->y1 = pt1[1];
+
+   bez->x2 = pt2[0];
+   bez->y2 = pt2[1];
+
+   bez->x3 = pt3[0];
+   bez->y3 = pt3[1];
+
+   bez->x4 = pt4[0];
+   bez->y4 = pt4[1];
+}
+
+
+void bezier_transform(struct bezier *bez,
+                      struct matrix *matrix)
+{
+   assert(matrix_is_affine(matrix));
+   matrix_map_point(matrix, bez->x1, bez->y1, &bez->x1, &bez->y1);
+   matrix_map_point(matrix, bez->x2, bez->y2, &bez->x2, &bez->y2);
+   matrix_map_point(matrix, bez->x3, bez->y3, &bez->x3, &bez->y3);
+   matrix_map_point(matrix, bez->x4, bez->y4, &bez->x4, &bez->y4);
+}
+
+static INLINE void bezier_point_at(const struct bezier *bez, float t, float *pt)
+{
+   float a, b, c, d;
+   float m_t;
+   m_t = 1. - t;
+   b = m_t * m_t;
+   c = t * t;
+   d = c * t;
+   a = b * m_t;
+   b *= 3. * t;
+   c *= 3. * m_t;
+   pt[0] = a*bez->x1 + b*bez->x2 + c*bez->x3 + d*bez->x4;
+   pt[1] = a*bez->y1 + b*bez->y2 + c*bez->y3 + d*bez->y4;
+}
+
+static INLINE void bezier_normal_at(const struct bezier *bez, float t, float *norm)
+{
+   float m_t = 1. - t;
+   float a = m_t * m_t;
+   float b = t * m_t;
+   float c = t * t;
+
+   norm[0] =  (bez->y2-bez->y1) * a + (bez->y3-bez->y2) * b + (bez->y4-bez->y3) * c;
+   norm[1] = -(bez->x2-bez->x1) * a - (bez->x3-bez->x2) * b - (bez->x4-bez->x3) * c;
+}
+
+enum shift_result {
+   Ok,
+   Discard,
+   Split,
+   Circle
+};
+
+static enum shift_result good_offset(const struct bezier *b1,
+                                     const struct bezier *b2,
+                                     float offset, float threshold)
+{
+   const float o2 = offset*offset;
+   const float max_dist_line = threshold*offset*offset;
+   const float max_dist_normal = threshold*offset;
+   const float spacing = 0.25;
+   for (float i = spacing; i < 0.99; i += spacing) {
+      float p1[2],p2[2], d, l;
+      float normal[2];
+      bezier_point_at(b1, i, p1);
+      bezier_point_at(b2, i, p2);
+      d = (p1[0] - p2[0])*(p1[0] - p2[0]) + (p1[1] - p2[1])*(p1[1] - p2[1]);
+      if (ABS(d - o2) > max_dist_line)
+         return Split;
+
+      bezier_normal_at(b1, i, normal);
+      l = ABS(normal[0]) + ABS(normal[1]);
+      if (l != 0.) {
+         d = ABS(normal[0]*(p1[1] - p2[1]) - normal[1]*(p1[0] - p2[0]) ) / l;
+         if (d > max_dist_normal)
+            return Split;
+      }
+   }
+   return Ok;
+}
+
+static INLINE void shift_line_by_normal(float *l, float offset)
+{
+   float norm[4];
+   float tx, ty;
+
+   line_normal(l, norm);
+   line_normalize(norm);
+
+   tx = (norm[2] - norm[0]) * offset;
+   ty = (norm[3] - norm[1]) * offset;
+   l[0] += tx; l[1] += ty;
+   l[2] += tx; l[3] += ty;
+}
+
+static INLINE VGboolean is_bezier_line(float (*points)[2], int count)
+{
+   float dx13 = points[2][0] - points[0][0];
+   float dy13 = points[2][1] - points[0][1];
+
+   float dx12 = points[1][0] - points[0][0];
+   float dy12 = points[1][1] - points[0][1];
+
+   debug_assert(count > 2);
+
+   if (count == 3) {
+      return floatsEqual(dx12 * dy13, dx13 * dy12);
+   } else if (count == 4) {
+      float dx14 = points[3][0] - points[0][0];
+      float dy14 = points[3][1] - points[0][1];
+
+      return (floatsEqual(dx12 * dy13, dx13 * dy12) &&
+              floatsEqual(dx12 * dy14, dx14 * dy12));
+   }
+
+   return VG_FALSE;
+}
+
+static INLINE void compute_pt_normal(float *pt1, float *pt2, float *res)
+{
+   float line[4];
+   float normal[4];
+   line[0] = 0.f; line[1] = 0.f;
+   line[2] = pt2[0] - pt1[0];
+   line[3] = pt2[1] - pt1[1];
+   line_normal(line, normal);
+   line_normalize(normal);
+
+   res[0] = normal[2];
+   res[1] = normal[3];
+}
+
+static enum shift_result shift(const struct bezier *orig,
+                               struct bezier *shifted,
+                               float offset, float threshold)
+{
+   int map[4];
+   VGboolean p1_p2_equal = (orig->x1 == orig->x2 && orig->y1 == orig->y2);
+   VGboolean p2_p3_equal = (orig->x2 == orig->x3 && orig->y2 == orig->y3);
+   VGboolean p3_p4_equal = (orig->x3 == orig->x4 && orig->y3 == orig->y4);
+
+   float points[4][2];
+   int np = 0;
+   float bounds[4];
+   float points_shifted[4][2];
+   float prev_normal[2];
+
+   points[np][0] = orig->x1;
+   points[np][1] = orig->y1;
+   map[0] = 0;
+   ++np;
+   if (!p1_p2_equal) {
+      points[np][0] = orig->x2;
+      points[np][1] = orig->y2;
+      ++np;
+   }
+   map[1] = np - 1;
+   if (!p2_p3_equal) {
+      points[np][0] = orig->x3;
+      points[np][1] = orig->y3;
+      ++np;
+   }
+   map[2] = np - 1;
+   if (!p3_p4_equal) {
+      points[np][0] = orig->x4;
+      points[np][1] = orig->y4;
+      ++np;
+   }
+   map[3] = np - 1;
+   if (np == 1)
+      return Discard;
+
+   /* We need to specialcase lines of 3 or 4 points due to numerical
+      instability in intersection code below */
+   if (np > 2 && is_bezier_line(points, np)) {
+      float l[4] = { points[0][0], points[0][1],
+                     points[np-1][0], points[np-1][1] };
+      float ctrl1[2], ctrl2[2];
+      if (floatsEqual(points[0][0], points[np-1][0]) &&
+          floatsEqual(points[0][1], points[np-1][1]))
+         return Discard;
+
+      shift_line_by_normal(l, offset);
+      line_point_at(l, 0.33, ctrl1);
+      line_point_at(l, 0.66, ctrl2);
+      bezier_init(shifted, l[0], l[1],
+                  ctrl1[0], ctrl1[1], ctrl2[0], ctrl2[1],
+                  l[2], l[3]);
+      return Ok;
+   }
+
+   bezier_bounds(orig, bounds);
+   if (np == 4 && bounds[2] < .1*offset && bounds[3] < .1*offset) {
+      float l = (orig->x1 - orig->x2)*(orig->x1 - orig->x2) +
+                (orig->y1 - orig->y2)*(orig->y1 - orig->y1) *
+                (orig->x3 - orig->x4)*(orig->x3 - orig->x4) +
+                (orig->y3 - orig->y4)*(orig->y3 - orig->y4);
+      float dot = (orig->x1 - orig->x2)*(orig->x3 - orig->x4) +
+                  (orig->y1 - orig->y2)*(orig->y3 - orig->y4);
+      if (dot < 0 && dot*dot < 0.8*l)
+         /* the points are close and reverse dirction. Approximate the whole
+            thing by a semi circle */
+         return Circle;
+   }
+
+   compute_pt_normal(points[0], points[1], prev_normal);
+
+   points_shifted[0][0] = points[0][0] + offset * prev_normal[0];
+   points_shifted[0][1] = points[0][1] + offset * prev_normal[1];
+
+   for (int i = 1; i < np - 1; ++i) {
+      float normal_sum[2], r;
+      float next_normal[2];
+      compute_pt_normal(points[i], points[i + 1], next_normal);
+
+      normal_sum[0] = prev_normal[0] + next_normal[0];
+      normal_sum[1] = prev_normal[1] + next_normal[1];
+
+      r = 1.0 + prev_normal[0] * next_normal[0]
+          + prev_normal[1] * next_normal[1];
+
+      if (floatsEqual(r + 1, 1)) {
+         points_shifted[i][0] = points[i][0] + offset * prev_normal[0];
+         points_shifted[i][1] = points[i][1] + offset * prev_normal[1];
+      } else {
+         float k = offset / r;
+         points_shifted[i][0] = points[i][0] + k * normal_sum[0];
+         points_shifted[i][1] = points[i][1] + k * normal_sum[1];
+      }
+
+      prev_normal[0] = next_normal[0];
+      prev_normal[1] = next_normal[1];
+   }
+
+   points_shifted[np - 1][0] = points[np - 1][0] + offset * prev_normal[0];
+   points_shifted[np - 1][1] = points[np - 1][1] + offset * prev_normal[1];
+
+   bezier_init2v(shifted,
+                 points_shifted[map[0]], points_shifted[map[1]],
+                 points_shifted[map[2]], points_shifted[map[3]]);
+
+   return good_offset(orig, shifted, offset, threshold);
+}
+
+static VGboolean make_circle(const struct bezier *b, float offset, struct bezier *o)
+{
+   float normals[3][2];
+   float dist;
+   float angles[2];
+   float sign = 1.f;
+   int i;
+   float circle[3][2];
+
+   normals[0][0] = b->y2 - b->y1;
+   normals[0][1] = b->x1 - b->x2;
+   dist = sqrt(normals[0][0]*normals[0][0] + normals[0][1]*normals[0][1]);
+   if (floatsEqual(dist + 1, 1.f))
+      return VG_FALSE;
+   normals[0][0] /= dist;
+   normals[0][1] /= dist;
+
+   normals[2][0] = b->y4 - b->y3;
+   normals[2][1] = b->x3 - b->x4;
+   dist = sqrt(normals[2][0]*normals[2][0] + normals[2][1]*normals[2][1]);
+   if (floatsEqual(dist + 1, 1.f))
+      return VG_FALSE;
+   normals[2][0] /= dist;
+   normals[2][1] /= dist;
+
+   normals[1][0] = b->x1 - b->x2 - b->x3 + b->x4;
+   normals[1][1] = b->y1 - b->y2 - b->y3 + b->y4;
+   dist = -1*sqrt(normals[1][0]*normals[1][0] + normals[1][1]*normals[1][1]);
+   normals[1][0] /= dist;
+   normals[1][1] /= dist;
+
+   for (i = 0; i < 2; ++i) {
+      float cos_a = normals[i][0]*normals[i+1][0] + normals[i][1]*normals[i+1][1];
+      if (cos_a > 1.)
+         cos_a = 1.;
+      if (cos_a < -1.)
+         cos_a = -1;
+      angles[i] = acos(cos_a)/M_PI;
+   }
+
+   if (angles[0] + angles[1] > 1.) {
+      /* more than 180 degrees */
+      normals[1][0] = -normals[1][0];
+      normals[1][1] = -normals[1][1];
+      angles[0] = 1. - angles[0];
+      angles[1] = 1. - angles[1];
+      sign = -1.;
+   }
+
+   circle[0][0] = b->x1 + normals[0][0]*offset;
+   circle[0][1] = b->y1 + normals[0][1]*offset;
+
+   circle[1][0] = 0.5*(b->x1 + b->x4) + normals[1][0]*offset;
+   circle[1][1] = 0.5*(b->y1 + b->y4) + normals[1][1]*offset;
+
+   circle[2][0] = b->x4 + normals[2][0]*offset;
+   circle[2][1] = b->y4 + normals[2][1]*offset;
+
+   for (i = 0; i < 2; ++i) {
+      float kappa = 2.*KAPPA * sign * offset * angles[i];
+
+      o->x1 = circle[i][0];
+      o->y1 = circle[i][1];
+      o->x2 = circle[i][0] - normals[i][1]*kappa;
+      o->y2 = circle[i][1] + normals[i][0]*kappa;
+      o->x3 = circle[i+1][0] + normals[i+1][1]*kappa;
+      o->y3 = circle[i+1][1] - normals[i+1][0]*kappa;
+      o->x4 = circle[i+1][0];
+      o->y4 = circle[i+1][1];
+
+      ++o;
+   }
+   return VG_TRUE;
+}
+
+int bezier_translate_by_normal(struct bezier *bez,
+                               struct bezier *curves,
+                               int max_curves,
+                               float normal_len,
+                               float threshold)
+{
+   struct bezier beziers[10];
+   struct bezier *b, *o;
+
+   /* fixme: this should really be floatsEqual */
+   if (bez->x1 == bez->x2 && bez->x1 == bez->x3 && bez->x1 == bez->x4 &&
+       bez->y1 == bez->y2 && bez->y1 == bez->y3 && bez->y1 == bez->y4)
+      return 0;
+
+   --max_curves;
+redo:
+   beziers[0] = *bez;
+   b = beziers;
+   o = curves;
+
+   while (b >= beziers) {
+      int stack_segments = b - beziers + 1;
+      enum shift_result res;
+      if ((stack_segments == 10) || (o - curves == max_curves - stack_segments)) {
+         threshold *= 1.5;
+         if (threshold > 2.)
+            goto give_up;
+         goto redo;
+      }
+      res = shift(b, o, normal_len, threshold);
+      if (res == Discard) {
+         --b;
+      } else if (res == Ok) {
+         ++o;
+         --b;
+         continue;
+      } else if (res == Circle && max_curves - (o - curves) >= 2) {
+         /* add semi circle */
+         if (make_circle(b, normal_len, o))
+            o += 2;
+         --b;
+      } else {
+         split(b, b+1, b);
+         ++b;
+      }
+   }
+
+give_up:
+   while (b >= beziers) {
+      enum shift_result res = shift(b, o, normal_len, threshold);
+
+      /* if res isn't Ok or Split then *o is undefined */
+      if (res == Ok || res == Split)
+         ++o;
+
+      --b;
+   }
+
+   debug_assert(o - curves <= max_curves);
+   return o - curves;
+}
+
+void bezier_bounds(const struct bezier *bez,
+                   float *bounds/*x/y/width/height*/)
+{
+   float xmin = bez->x1;
+   float xmax = bez->x1;
+   float ymin = bez->y1;
+   float ymax = bez->y1;
+
+   if (bez->x2 < xmin)
+      xmin = bez->x2;
+   else if (bez->x2 > xmax)
+      xmax = bez->x2;
+   if (bez->x3 < xmin)
+      xmin = bez->x3;
+   else if (bez->x3 > xmax)
+      xmax = bez->x3;
+   if (bez->x4 < xmin)
+      xmin = bez->x4;
+   else if (bez->x4 > xmax)
+      xmax = bez->x4;
+
+   if (bez->y2 < ymin)
+      ymin = bez->y2;
+   else if (bez->y2 > ymax)
+      ymax = bez->y2;
+   if (bez->y3 < ymin)
+      ymin = bez->y3;
+   else if (bez->y3 > ymax)
+      ymax = bez->y3;
+   if (bez->y4 < ymin)
+      ymin = bez->y4;
+   else if (bez->y4 > ymax)
+      ymax = bez->y4;
+
+   bounds[0] = xmin; /* x */
+   bounds[1] = ymin; /* y */
+   bounds[2] = xmax - xmin; /* width */
+   bounds[3] = ymax - ymin; /* height */
+}
+
+void bezier_start_tangent(const struct bezier *bez,
+                          float *tangent)
+{
+   tangent[0] = bez->x1;
+   tangent[1] = bez->y1;
+   tangent[2] = bez->x2;
+   tangent[3] = bez->y2;
+
+   if (null_line(tangent)) {
+      tangent[0] = bez->x1;
+      tangent[1] = bez->y1;
+      tangent[2] = bez->x3;
+      tangent[3] = bez->y3;
+   }
+   if (null_line(tangent)) {
+      tangent[0] = bez->x1;
+      tangent[1] = bez->y1;
+      tangent[2] = bez->x4;
+      tangent[3] = bez->y4;
+   }
+}
+
+
+static INLINE VGfloat bezier_t_at_length(struct bezier *bez,
+                                         VGfloat at_length,
+                                         VGfloat error)
+{
+   VGfloat len = bezier_length(bez, error);
+   VGfloat t   = 1.0;
+   VGfloat last_bigger = 1.;
+
+   if (at_length > len || floatsEqual(at_length, len))
+      return t;
+
+   if (floatIsZero(at_length))
+      return 0.f;
+
+   t *= 0.5;
+   while (1) {
+      struct bezier right = *bez;
+      struct bezier left;
+      VGfloat tmp_len;
+      split_left(&right, t, &left);
+      tmp_len = bezier_length(&left, error);
+      if (ABS(tmp_len - at_length) < error)
+         break;
+
+      if (tmp_len < at_length) {
+         t += (last_bigger - t)*.5;
+      } else {
+         last_bigger = t;
+         t -= t*.5;
+      }
+   }
+   return t;
+}
+
+void bezier_point_at_length(struct bezier *bez,
+                            float length,
+                            float *point,
+                            float *normal)
+{
+   /* ~0.000001 seems to be required to pass G2080x tests */
+   VGfloat t = bezier_t_at_length(bez, length, 0.000001);
+   bezier_point_at(bez, t, point);
+   bezier_normal_at(bez, t, normal);
+   vector_unit(normal);
+}
+
+void bezier_point_at_t(struct bezier *bez, float t,
+                       float *point, float *normal)
+{
+   bezier_point_at(bez, t, point);
+   bezier_normal_at(bez, t, normal);
+   vector_unit(normal);
+}
+
+void bezier_exact_bounds(const struct bezier *bez,
+                         float *bounds/*x/y/width/height*/)
+{
+   struct polygon *poly = polygon_create(64);
+   polygon_vertex_append(poly, bez->x1, bez->y1);
+   bezier_add_to_polygon(bez, poly);
+   polygon_bounding_rect(poly, bounds);
+   polygon_destroy(poly);
+}
+
diff --git a/src/gallium/state_trackers/vega/bezier.h b/src/gallium/state_trackers/vega/bezier.h
new file mode 100644
index 0000000000..e06666051c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/bezier.h
@@ -0,0 +1,81 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef BEZIER_H
+#define BEZIER_H
+
+struct polygon;
+struct matrix;
+
+struct bezier {
+    float x1, y1;
+    float x2, y2;
+    float x3, y3;
+    float x4, y4;
+};
+
+
+#define BEZIER_DEFAULT_ERROR 0.01
+
+/* kappa as being l of a circle with r = 1, we can emulate any
+ * circle of radius r by using the formula
+ * l = r . kappa
+ * More at:
+ * http://www.whizkidtech.redprince.net/bezier/circle/ */
+#define KAPPA 0.5522847498
+
+void bezier_init(struct bezier *bez,
+                 float x1, float y1,
+                 float x2, float y2,
+                 float x3, float y3,
+                 float x4, float y4);
+
+struct polygon *bezier_to_polygon(struct bezier *bez);
+void bezier_add_to_polygon(const struct bezier *bez,
+                           struct polygon *poly);
+float bezier_length(struct bezier *bez, float error);
+void bezier_transform(struct bezier *bez,
+                      struct matrix *mat);
+
+int bezier_translate_by_normal(struct bezier *b,
+                               struct bezier *curves,
+                               int max_curves,
+                               float normal_len,
+                               float threshold);
+void bezier_bounds(const struct bezier *bez,
+                   float *bounds/*x/y/width/height*/);
+void bezier_exact_bounds(const struct bezier *bez,
+                         float *bounds/*x/y/width/height*/);
+
+void bezier_start_tangent(const struct bezier *bez,
+                          float *tangent);
+
+void bezier_point_at_length(struct bezier *bez, float length,
+                            float *point, float *normal);
+void bezier_point_at_t(struct bezier *bez, float t,
+                       float *point, float *normal);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c
new file mode 100644
index 0000000000..9a722980d5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/image.c
@@ -0,0 +1,654 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "image.h"
+
+#include "vg_translate.h"
+#include "vg_context.h"
+#include "matrix.h"
+#include "renderer.h"
+#include "util_array.h"
+#include "api_consts.h"
+#include "shaders_cache.h"
+#include "shader.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_blit.h"
+#include "util/u_tile.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+static enum pipe_format vg_format_to_pipe(VGImageFormat format)
+{
+   switch(format) {
+   case VG_sRGB_565:
+      return PIPE_FORMAT_R5G6B5_UNORM;
+   case VG_sRGBA_5551:
+      return PIPE_FORMAT_A1R5G5B5_UNORM;
+   case VG_sRGBA_4444:
+      return PIPE_FORMAT_A4R4G4B4_UNORM;
+   case VG_sL_8:
+   case VG_lL_8:
+      return PIPE_FORMAT_L8_UNORM;
+   case VG_BW_1:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   case VG_A_8:
+      return PIPE_FORMAT_A8_UNORM;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+   case VG_A_4:
+      return PIPE_FORMAT_A8_UNORM;
+#endif
+   default:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   }
+}
+
+static INLINE void vg_sync_size(VGfloat *src_loc, VGfloat *dst_loc)
+{
+   src_loc[2] = MIN2(src_loc[2], dst_loc[2]);
+   src_loc[3] = MIN2(src_loc[3], dst_loc[3]);
+   dst_loc[2] = src_loc[2];
+   dst_loc[3] = src_loc[3];
+}
+
+
+static void vg_copy_texture(struct vg_context *ctx,
+                            struct pipe_texture *dst, VGint dx, VGint dy,
+                            struct pipe_texture *src, VGint sx, VGint sy,
+                            VGint width, VGint height)
+{
+   VGfloat dst_loc[4], src_loc[4];
+   VGfloat dst_bounds[4], src_bounds[4];
+   VGfloat src_shift[4], dst_shift[4], shift[4];
+
+   dst_loc[0] = dx;
+   dst_loc[1] = dy;
+   dst_loc[2] = width;
+   dst_loc[3] = height;
+   dst_bounds[0] = 0.f;
+   dst_bounds[1] = 0.f;
+   dst_bounds[2] = dst->width[0];
+   dst_bounds[3] = dst->height[0];
+
+   src_loc[0] = sx;
+   src_loc[1] = sy;
+   src_loc[2] = width;
+   src_loc[3] = height;
+   src_bounds[0] = 0.f;
+   src_bounds[1] = 0.f;
+   src_bounds[2] = src->width[0];
+   src_bounds[3] = src->height[0];
+
+   vg_bound_rect(src_loc, src_bounds, src_shift);
+   vg_bound_rect(dst_loc, dst_bounds, dst_shift);
+   shift[0] = src_shift[0] - dst_shift[0];
+   shift[1] = src_shift[1] - dst_shift[1];
+
+   if (shift[0] < 0)
+      vg_shift_rectx(src_loc, src_bounds, -shift[0]);
+   else
+      vg_shift_rectx(dst_loc, dst_bounds, shift[0]);
+
+   if (shift[1] < 0)
+      vg_shift_recty(src_loc, src_bounds, -shift[1]);
+   else
+      vg_shift_recty(dst_loc, dst_bounds, shift[1]);
+
+   vg_sync_size(src_loc, dst_loc);
+
+   if (src_loc[2] >= 0 && src_loc[3] >= 0 &&
+       dst_loc[2] >= 0 && dst_loc[3] >= 0) {
+      renderer_copy_texture(ctx->renderer,
+                            src,
+                            src_loc[0],
+                            src_loc[1] + src_loc[3],
+                            src_loc[0] + src_loc[2],
+                            src_loc[1],
+                            dst,
+                            dst_loc[0],
+                            dst_loc[1] + dst_loc[3],
+                            dst_loc[0] + dst_loc[2],
+                            dst_loc[1]);
+   }
+
+}
+
+void vg_copy_surface(struct vg_context *ctx,
+                     struct pipe_surface *dst, VGint dx, VGint dy,
+                     struct pipe_surface *src, VGint sx, VGint sy,
+                     VGint width, VGint height)
+{
+   VGfloat dst_loc[4], src_loc[4];
+   VGfloat dst_bounds[4], src_bounds[4];
+   VGfloat src_shift[4], dst_shift[4], shift[4];
+
+   dst_loc[0] = dx;
+   dst_loc[1] = dy;
+   dst_loc[2] = width;
+   dst_loc[3] = height;
+   dst_bounds[0] = 0.f;
+   dst_bounds[1] = 0.f;
+   dst_bounds[2] = dst->width;
+   dst_bounds[3] = dst->height;
+
+   src_loc[0] = sx;
+   src_loc[1] = sy;
+   src_loc[2] = width;
+   src_loc[3] = height;
+   src_bounds[0] = 0.f;
+   src_bounds[1] = 0.f;
+   src_bounds[2] = src->width;
+   src_bounds[3] = src->height;
+
+   vg_bound_rect(src_loc, src_bounds, src_shift);
+   vg_bound_rect(dst_loc, dst_bounds, dst_shift);
+   shift[0] = src_shift[0] - dst_shift[0];
+   shift[1] = src_shift[1] - dst_shift[1];
+
+   if (shift[0] < 0)
+      vg_shift_rectx(src_loc, src_bounds, -shift[0]);
+   else
+      vg_shift_rectx(dst_loc, dst_bounds, shift[0]);
+
+   if (shift[1] < 0)
+      vg_shift_recty(src_loc, src_bounds, -shift[1]);
+   else
+      vg_shift_recty(dst_loc, dst_bounds, shift[1]);
+
+   vg_sync_size(src_loc, dst_loc);
+
+   if (src_loc[2] > 0 && src_loc[3] > 0 &&
+       dst_loc[2] > 0 && dst_loc[3] > 0) {
+      if (src == dst)
+         renderer_copy_surface(ctx->renderer,
+                               src,
+                               src_loc[0],
+                               src->height - (src_loc[1] + src_loc[3]),
+                               src_loc[0] + src_loc[2],
+                               src->height - src_loc[1],
+                               dst,
+                               dst_loc[0],
+                               dst->height - (dst_loc[1] + dst_loc[3]),
+                               dst_loc[0] + dst_loc[2],
+                               dst->height - dst_loc[1],
+                               0, 0);
+      else
+         renderer_copy_surface(ctx->renderer,
+                               src,
+                               src_loc[0],
+                               src->height - src_loc[1],
+                               src_loc[0] + src_loc[2],
+                               src->height - (src_loc[1] + src_loc[3]),
+                               dst,
+                               dst_loc[0],
+                               dst->height - (dst_loc[1] + dst_loc[3]),
+                               dst_loc[0] + dst_loc[2],
+                               dst->height - dst_loc[1],
+                               0, 0);
+   }
+
+}
+
+static struct pipe_texture *image_texture(struct vg_image *img)
+{
+   struct pipe_texture *tex = img->texture;
+   return tex;
+}
+
+
+static void image_cleari(struct vg_image *img, VGint clear_colori,
+                         VGint x, VGint y, VGint width, VGint height)
+{
+   VGint *clearbuf;
+   VGint i;
+   VGfloat dwidth, dheight;
+
+   clearbuf = malloc(sizeof(VGint)*width*height);
+   for (i = 0; i < width*height; ++i)
+      clearbuf[i] = clear_colori;
+
+   dwidth = MIN2(width, img->width);
+   dheight = MIN2(height, img->height);
+
+   image_sub_data(img, clearbuf, width * sizeof(VGint),
+                  VG_sRGBA_8888,
+                  x, y, dwidth, dheight);
+   free(clearbuf);
+}
+
+struct vg_image * image_create(VGImageFormat format,
+                               VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *image = CALLOC_STRUCT(vg_image);
+   enum pipe_format pformat = vg_format_to_pipe(format);
+   struct pipe_texture pt, *newtex;
+   struct pipe_screen *screen = ctx->pipe->screen;
+
+   vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE);
+
+   image->format = format;
+   image->width = width;
+   image->height = height;
+
+   image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.normalized_coords = 1;
+
+   assert(screen->is_format_supported(screen, pformat, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   memset(&pt, 0, sizeof(pt));
+   pt.target = PIPE_TEXTURE_2D;
+   pt.format = pformat;
+   pf_get_block(pformat, &pt.block);
+   pt.last_level = 0;
+   pt.width[0] = width;
+   pt.height[0] = height;
+   pt.depth[0] = 1;
+   pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   newtex = screen->texture_create(screen, &pt);
+
+   debug_assert(newtex);
+
+   image->texture = newtex;
+
+   vg_context_add_object(ctx, VG_OBJECT_IMAGE, image);
+
+   image_cleari(image, 0, 0, 0, image->width, image->height);
+   return image;
+}
+
+void image_destroy(struct vg_image *img)
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_context_remove_object(ctx, VG_OBJECT_IMAGE, img);
+
+
+   if (img->parent) {
+      /* remove img from the parent child array */
+      int idx;
+      struct vg_image **array =
+         (struct vg_image **)img->parent->children_array->data;
+
+      for (idx = 0; idx < img->parent->children_array->num_elements; ++idx) {
+         struct vg_image *child = array[idx];
+         if (child == img) {
+            break;
+         }
+      }
+      debug_assert(idx < img->parent->children_array->num_elements);
+      array_remove_element(img->parent->children_array, idx);
+   }
+
+   if (img->children_array && img->children_array->num_elements) {
+      /* reparent the children */
+      VGint i;
+      struct vg_image *parent = img->parent;
+      struct vg_image **children =
+         (struct vg_image **)img->children_array->data;
+      if (!parent) {
+         VGint min_x = children[0]->x;
+         parent = children[0];
+
+         for (i = 1; i < img->children_array->num_elements; ++i) {
+            struct vg_image *child = children[i];
+            if (child->x < min_x) {
+               parent = child;
+            }
+         }
+      }
+
+      for (i = 0; i < img->children_array->num_elements; ++i) {
+         struct vg_image *child = children[i];
+         if (child != parent) {
+            child->parent = parent;
+            if (!parent->children_array) {
+               parent->children_array = array_create(
+                  sizeof(struct vg_image*));
+            }
+            array_append_data(parent->children_array,
+                              &child, 1);
+         } else
+            child->parent = NULL;
+      }
+      array_destroy(img->children_array);
+   }
+
+   pipe_texture_reference(&img->texture, NULL);
+   free(img);
+}
+
+void image_clear(struct vg_image *img,
+                 VGint x, VGint y, VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat *clear_colorf = ctx->state.vg.clear_color;
+   VGubyte r, g, b ,a;
+   VGint clear_colori;
+   /* FIXME: this is very nasty */
+   r = float_to_ubyte(clear_colorf[0]);
+   g = float_to_ubyte(clear_colorf[1]);
+   b = float_to_ubyte(clear_colorf[2]);
+   a = float_to_ubyte(clear_colorf[3]);
+   clear_colori = r << 24 | g << 16 | b << 8 | a;
+   image_cleari(img, clear_colori, x, y, width, height);
+}
+
+void image_sub_data(struct vg_image *image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height)
+{
+   const VGint yStep = 1;
+   VGubyte *src = (VGubyte *)data;
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint i;
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct pipe_texture *texture = image_texture(image);
+   VGint xoffset = 0, yoffset = 0;
+
+   if (x < 0) {
+      xoffset -= x;
+      width += x;
+      x = 0;
+   }
+   if (y < 0) {
+      yoffset -= y;
+      height += y;
+      y = 0;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (x > image->width || y > image->width) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (x + width > image->width) {
+      width = image->width - x;
+   }
+
+   if (y + height > image->height) {
+      height = image->height - y;
+   }
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer = screen->get_tex_transfer(
+         screen, texture, 0, 0, 0,
+         PIPE_TRANSFER_WRITE, 0, 0, texture->width[0], texture->height[0]);
+      src += (dataStride * yoffset);
+      for (i = 0; i < height; i++) {
+         _vega_unpack_float_span_rgba(ctx, width, xoffset, src, dataFormat, temp);
+         pipe_put_tile_rgba(transfer, x+image->x, y+image->y, width, 1, df);
+         y += yStep;
+         src += dataStride;
+      }
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+void image_get_sub_data(struct vg_image * image,
+                        void * data,
+                        VGint dataStride,
+                        VGImageFormat dataFormat,
+                        VGint sx, VGint sy,
+                        VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = 0, yStep = 1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+
+   {
+      struct pipe_transfer *transfer =
+         screen->get_tex_transfer(screen,
+                                  image->texture,  0, 0, 0,
+                                  PIPE_TRANSFER_READ,
+                                  0, 0,
+                                  image->x + image->width,
+                                  image->y + image->height);
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(transfer, sx+image->x, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat, dst);
+         dst += dataStride;
+      }
+
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+struct vg_image * image_child_image(struct vg_image *parent,
+                                    VGint x, VGint y,
+                                    VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *image = CALLOC_STRUCT(vg_image);
+
+   vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE);
+
+   image->x = parent->x + x;
+   image->y = parent->y + y;
+   image->width = width;
+   image->height = height;
+   image->parent = parent;
+   image->texture = 0;
+   pipe_texture_reference(&image->texture,
+                          parent->texture);
+
+   image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.normalized_coords = 1;
+
+   if (!parent->children_array)
+      parent->children_array = array_create(
+         sizeof(struct vg_image*));
+
+   array_append_data(parent->children_array,
+                     &image, 1);
+
+   vg_context_add_object(ctx, VG_OBJECT_IMAGE, image);
+
+   return image;
+}
+
+void image_copy(struct vg_image *dst, VGint dx, VGint dy,
+                struct vg_image *src, VGint sx, VGint sy,
+                VGint width, VGint height,
+                VGboolean dither)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   vg_copy_texture(ctx, dst->texture, dst->x + dx, dst->y + dy,
+                   src->texture, src->x + sx, src->y + sy, width, height);
+}
+
+void image_draw(struct vg_image *img)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat x1, y1;
+   VGfloat x2, y2;
+   VGfloat x3, y3;
+   VGfloat x4, y4;
+   struct matrix *matrix;
+
+   x1 = 0;
+   y1 = 0;
+   x2 = img->width;
+   y2 = 0;
+   x3 = img->width;
+   y3 = img->height;
+   x4 = 0;
+   y4 = img->height;
+
+   matrix = &ctx->state.vg.image_user_to_surface_matrix;
+
+   matrix_map_point(matrix, x1, y1, &x1, &y1);
+   matrix_map_point(matrix, x2, y2, &x2, &y2);
+   matrix_map_point(matrix, x3, y3, &x3, &y3);
+   matrix_map_point(matrix, x4, y4, &x4, &y4);
+
+   shader_set_drawing_image(ctx->shader, VG_TRUE);
+   shader_set_paint(ctx->shader, ctx->state.vg.fill_paint);
+   shader_set_image(ctx->shader, img);
+   shader_bind(ctx->shader);
+
+   renderer_texture_quad(ctx->renderer, image_texture(img),
+                         img->x, img->y, img->x + img->width, img->y + img->height,
+                         x1, y1, x2, y2, x3, y3, x4, y4);
+}
+
+void image_set_pixels(VGint dx, VGint dy,
+                      struct vg_image *src, VGint sx, VGint sy,
+                      VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surf;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   surf = screen->get_tex_surface(screen, image_texture(src),  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_READ);
+
+   vg_copy_surface(ctx, strb->surface, dx, dy,
+                   surf, sx+src->x, sy+src->y, width, height);
+
+   screen->tex_surface_destroy(surf);
+}
+
+void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy,
+                      VGint sx, VGint sy,
+                      VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surf;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   /* flip the y coordinates */
+   /*dy = dst->height - dy - height;*/
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   surf = screen->get_tex_surface(screen, image_texture(dst),  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_WRITE |
+                                  PIPE_BUFFER_USAGE_GPU_READ);
+   vg_copy_surface(ctx, surf, dst->x + dx, dst->y + dy,
+                   strb->surface, sx, sy, width, height);
+
+   pipe_surface_reference(&surf, NULL);
+}
+
+
+VGboolean vg_image_overlaps(struct vg_image *dst,
+                            struct vg_image *src)
+{
+   if (dst == src || dst->parent == src ||
+       dst == src->parent)
+      return VG_TRUE;
+   if (dst->parent && dst->parent == src->parent) {
+      VGfloat left1 = dst->x;
+      VGfloat left2 = src->x;
+      VGfloat right1 = dst->x + dst->width;
+      VGfloat right2 = src->x + src->width;
+      VGfloat bottom1 = dst->y;
+      VGfloat bottom2 = src->y;
+      VGfloat top1 = dst->y + dst->height;
+      VGfloat top2 = src->y + src->height;
+
+      return !(left2 > right1 || right2 < left1 ||
+               top2 > bottom1 || bottom2 < top1);
+   }
+   return VG_FALSE;
+}
+
+VGint image_bind_samplers(struct vg_image *img, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures)
+{
+   img->sampler.min_img_filter = image_sampler_filter(img->base.ctx);
+   img->sampler.mag_img_filter = image_sampler_filter(img->base.ctx);
+   samplers[3] = &img->sampler;
+   textures[3] = img->texture;
+   return 1;
+}
+
+VGint image_sampler_filter(struct vg_context *ctx)
+{
+    switch(ctx->state.vg.image_quality) {
+    case VG_IMAGE_QUALITY_NONANTIALIASED:
+       return PIPE_TEX_FILTER_NEAREST;
+       break;
+    case VG_IMAGE_QUALITY_FASTER:
+       return PIPE_TEX_FILTER_NEAREST;
+       break;
+    case VG_IMAGE_QUALITY_BETTER:
+       /*return PIPE_TEX_FILTER_ANISO;*/
+       return PIPE_TEX_FILTER_LINEAR;
+       break;
+    default:
+       debug_printf("Unknown image quality");
+    }
+    return PIPE_TEX_FILTER_NEAREST;
+}
diff --git a/src/gallium/state_trackers/vega/image.h b/src/gallium/state_trackers/vega/image.h
new file mode 100644
index 0000000000..78e17cffa6
--- /dev/null
+++ b/src/gallium/state_trackers/vega/image.h
@@ -0,0 +1,104 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef IMAGES_H
+#define IMAGES_H
+
+#include "vg_context.h"
+#include "pipe/p_state.h"
+
+struct pipe_texture;
+struct array;
+struct vg_context;
+struct pipe_surface;
+
+struct vg_image {
+   struct vg_object base;
+   VGImageFormat format;
+   VGint x, y;
+   VGint width, height;
+
+   struct vg_image *parent;
+
+   struct pipe_texture *texture;
+   struct pipe_sampler_state sampler;
+
+   struct array *children_array;
+};
+
+struct vg_image *image_create(VGImageFormat format,
+                              VGint width, VGint height);
+void image_destroy(struct vg_image *img);
+
+void image_clear(struct vg_image *img,
+                 VGint x, VGint y, VGint width, VGint height);
+
+void image_sub_data(struct vg_image *image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height);
+
+void image_get_sub_data(struct vg_image * image,
+                        void * data,
+                        VGint dataStride,
+                        VGImageFormat dataFormat,
+                        VGint x, VGint y,
+                        VGint width, VGint height);
+
+struct vg_image *image_child_image(struct vg_image *parent,
+                                   VGint x, VGint y,
+                                   VGint width, VGint height);
+
+void image_copy(struct vg_image *dst, VGint dx, VGint dy,
+                struct vg_image *src, VGint sx, VGint sy,
+                VGint width, VGint height,
+                VGboolean dither);
+
+void image_draw(struct vg_image *img);
+
+void image_set_pixels(VGint dx, VGint dy,
+                      struct vg_image *src, VGint sx, VGint sy,
+                      VGint width, VGint height);
+void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy,
+                      VGint sx, VGint sy,
+                      VGint width, VGint height);
+
+VGint image_bind_samplers(struct vg_image *dst, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures);
+
+VGboolean vg_image_overlaps(struct vg_image *dst,
+                            struct vg_image *src);
+
+VGint image_sampler_filter(struct vg_context *ctx);
+
+void vg_copy_surface(struct vg_context *ctx,
+                     struct pipe_surface *dst, VGint dx, VGint dy,
+                     struct pipe_surface *src, VGint sx, VGint sy,
+                     VGint width, VGint height);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/mask.c b/src/gallium/state_trackers/vega/mask.c
new file mode 100644
index 0000000000..24650a37d5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/mask.c
@@ -0,0 +1,690 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "mask.h"
+
+#include "path.h"
+#include "image.h"
+#include "shaders_cache.h"
+#include "renderer.h"
+#include "asm_util.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+struct vg_mask_layer {
+   struct vg_object base;
+
+   VGint width;
+   VGint height;
+
+   struct pipe_texture *texture;
+};
+
+static INLINE struct pipe_surface *
+alpha_mask_surface(struct vg_context *ctx, int usage)
+{
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   return screen->get_tex_surface(screen,
+                                  stfb->alpha_mask,
+                                  0, 0, 0,
+                                  usage);
+}
+
+static INLINE VGboolean
+intersect_rectangles(VGint dwidth, VGint dheight,
+                     VGint swidth, VGint sheight,
+                     VGint tx, VGint ty,
+                     VGint twidth, VGint theight,
+                     VGint *offsets,
+                     VGint *location)
+{
+   if (tx + twidth <= 0 || tx >= dwidth)
+      return VG_FALSE;
+   if (ty + theight <= 0 || ty >= dheight)
+      return VG_FALSE;
+
+   offsets[0] = 0;
+   offsets[1] = 0;
+   location[0] = tx;
+   location[1] = ty;
+
+   if (tx < 0) {
+      offsets[0] -= tx;
+      location[0] = 0;
+
+      location[2] = MIN2(tx + swidth, MIN2(dwidth, tx + twidth));
+      offsets[2] = location[2];
+   } else {
+      offsets[2] = MIN2(twidth, MIN2(dwidth - tx, swidth ));
+      location[2] = offsets[2];
+   }
+
+   if (ty < 0) {
+      offsets[1] -= ty;
+      location[1] = 0;
+
+      location[3] = MIN2(ty + sheight, MIN2(dheight, ty + theight));
+      offsets[3] = location[3];
+   } else {
+      offsets[3] = MIN2(theight, MIN2(dheight - ty, sheight));
+      location[3] = offsets[3];
+   }
+
+   return VG_TRUE;
+}
+
+#if DEBUG_MASKS
+static void read_alpha_mask(void * data, VGint dataStride,
+                            VGImageFormat dataFormat,
+                            VGint sx, VGint sy,
+                            VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->alpha_mask;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = (fb->height - sy) - 1, yStep = -1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+   VGint xoffset = 0, yoffset = 0;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   if (sx < 0) {
+      xoffset = -sx;
+      xoffset *= _vega_size_for_format(dataFormat);
+      width += sx;
+      sx = 0;
+   }
+   if (sy < 0) {
+      yoffset = -sy;
+      height += sy;
+      sy = 0;
+      y = (fb->height - sy) - 1;
+      yoffset *= dataStride;
+   }
+
+   {
+      struct pipe_surface *surf;
+
+      surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_READ);
+
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(surf, sx, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat,
+                                    dst + yoffset + xoffset);
+         dst += dataStride;
+      }
+
+      pipe_surface_reference(&surf, NULL);
+   }
+}
+
+void save_alpha_to_file(const char *filename)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   VGint *data;
+   int i, j;
+
+   data = malloc(sizeof(int) * fb->width * fb->height);
+   read_alpha_mask(data, fb->width * sizeof(int),
+                   VG_sRGBA_8888,
+                   0, 0, fb->width, fb->height);
+   fprintf(stderr, "/*---------- start */\n");
+   fprintf(stderr, "const int image_width = %d;\n",
+           fb->width);
+   fprintf(stderr, "const int image_height = %d;\n",
+           fb->height);
+   fprintf(stderr, "const int image_data = {\n");
+   for (i = 0; i < fb->height; ++i) {
+      for (j = 0; j < fb->width; ++j) {
+         int rgba = data[i * fb->height + j];
+         int argb = 0;
+         argb = (rgba >> 8);
+         argb |= ((rgba & 0xff) << 24);
+         fprintf(stderr, "0x%x, ", argb);
+      }
+      fprintf(stderr, "\n");
+   }
+   fprintf(stderr, "};\n");
+   fprintf(stderr, "/*---------- end */\n");
+}
+#endif
+
+static void setup_mask_framebuffer(struct pipe_surface *surf,
+                                   VGint surf_width, VGint surf_height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state fb;
+
+   memset(&fb, 0, sizeof(fb));
+   fb.width = surf_width;
+   fb.height = surf_height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso_context, &fb);
+}
+
+
+/* setup shader constants */
+static void setup_mask_operation(VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   const VGint param_bytes = 4 * sizeof(VGfloat);
+   const VGfloat ones[4] = {1.f, 1.f, 1.f, 1.f};
+   void *shader = 0;
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization.
+    */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                    0, param_bytes, ones);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+   switch (operation) {
+   case VG_UNION_MASK: {
+      if (!ctx->mask.union_fs) {
+         ctx->mask.union_fs = shader_create_from_text(ctx->pipe,
+                                                      union_mask_asm,
+                                                      200,
+                                                      PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.union_fs->driver;
+   }
+      break;
+   case VG_INTERSECT_MASK: {
+      if (!ctx->mask.intersect_fs) {
+         ctx->mask.intersect_fs = shader_create_from_text(ctx->pipe,
+                                                          intersect_mask_asm,
+                                                          200,
+                                                          PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.intersect_fs->driver;
+   }
+      break;
+   case VG_SUBTRACT_MASK: {
+      if (!ctx->mask.subtract_fs) {
+         ctx->mask.subtract_fs = shader_create_from_text(ctx->pipe,
+                                                         subtract_mask_asm,
+                                                         200,
+                                                         PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.subtract_fs->driver;
+   }
+      break;
+   case VG_SET_MASK: {
+      if (!ctx->mask.set_fs) {
+         ctx->mask.set_fs = shader_create_from_text(ctx->pipe,
+                                                    set_mask_asm,
+                                                    200,
+                                                    PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.set_fs->driver;
+   }
+      break;
+   default:
+         assert(0);
+      break;
+   }
+   cso_set_fragment_shader_handle(ctx->cso_context, shader);
+}
+
+static void setup_mask_samplers(struct pipe_texture *umask)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+   struct pipe_texture *uprev = NULL;
+   struct pipe_sampler_state sampler;
+
+   uprev = fb_buffers->blend_texture;
+   sampler = ctx->mask.sampler;
+   sampler.normalized_coords = 1;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+
+   samplers[0] = &sampler;
+   samplers[1] = &ctx->mask.sampler;
+
+   textures[0] = umask;
+   textures[1] = uprev;
+
+   cso_set_samplers(ctx->cso_context, 2,
+                    (const struct pipe_sampler_state **)samplers);
+   cso_set_sampler_textures(ctx->cso_context, 2, textures);
+}
+
+
+/* setup shader constants */
+static void setup_mask_fill(const VGfloat color[4])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   const VGint param_bytes = 4 * sizeof(VGfloat);
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization.
+    */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, 0, param_bytes, color);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+   cso_set_fragment_shader_handle(ctx->cso_context,
+                                  shaders_cache_fill(ctx->sc,
+                                                     VEGA_SOLID_FILL_SHADER));
+}
+
+static void setup_mask_viewport()
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_set_viewport(ctx, VEGA_Y0_TOP);
+}
+
+static void setup_mask_blend()
+{
+   struct vg_context *ctx = vg_current_context();
+
+   struct pipe_blend_state blend;
+
+   memset(&blend, 0, sizeof(struct pipe_blend_state));
+   blend.blend_enable = 1;
+   blend.colormask |= PIPE_MASK_R;
+   blend.colormask |= PIPE_MASK_G;
+   blend.colormask |= PIPE_MASK_B;
+   blend.colormask |= PIPE_MASK_A;
+   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+
+   cso_set_blend(ctx->cso_context, &blend);
+}
+
+
+static void surface_fill(struct pipe_surface *surf,
+                         int surf_width, int surf_height,
+                         int x, int y, int width, int height,
+                         const VGfloat color[4])
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (x < 0) {
+      width += x;
+      x = 0;
+   }
+   if (y < 0) {
+      height += y;
+      y = 0;
+   }
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_blend();
+   setup_mask_fill(color);
+   setup_mask_framebuffer(surf, surf_width, surf_height);
+   setup_mask_viewport();
+
+   renderer_draw_quad(ctx->renderer, x, y,
+                      x + width, y + height, 0.0f/*depth should be disabled*/);
+
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe,
+                    PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME,
+                    NULL);
+
+#if DEBUG_MASKS
+   save_alpha_to_file(0);
+#endif
+
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+}
+
+
+static void mask_using_texture(struct pipe_texture *texture,
+                               VGMaskOperation operation,
+                               VGint x, VGint y,
+                               VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_surface *surface =
+      alpha_mask_surface(ctx, PIPE_BUFFER_USAGE_GPU_WRITE);
+   VGint offsets[4], loc[4];
+
+   if (!surface)
+      return;
+   if (!intersect_rectangles(surface->width, surface->height,
+                             texture->width[0], texture->height[0],
+                             x, y, width, height,
+                             offsets, loc))
+      return;
+#if 0
+   debug_printf("Offset = [%d, %d, %d, %d]\n", offsets[0],
+                offsets[1], offsets[2], offsets[3]);
+   debug_printf("Locati = [%d, %d, %d, %d]\n", loc[0],
+                loc[1], loc[2], loc[3]);
+#endif
+
+   /* prepare our blend surface */
+   vg_prepare_blend_surface_from_mask(ctx);
+
+   cso_save_samplers(ctx->cso_context);
+   cso_save_sampler_textures(ctx->cso_context);
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_samplers(texture);
+   setup_mask_blend();
+   setup_mask_operation(operation);
+   setup_mask_framebuffer(surface, surface->width, surface->height);
+   setup_mask_viewport();
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_texture(ctx->renderer, texture,
+                         offsets[0], offsets[1],
+                         offsets[0] + offsets[2], offsets[1] + offsets[3],
+                         loc[0], loc[1], loc[0] + loc[2], loc[1] + loc[3]);
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_samplers(ctx->cso_context);
+   cso_restore_sampler_textures(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+
+   pipe_surface_reference(&surface, NULL);
+}
+
+
+#ifdef OPENVG_VERSION_1_1
+
+struct vg_mask_layer * mask_layer_create(VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_mask_layer *mask = 0;
+
+   mask = CALLOC_STRUCT(vg_mask_layer);
+   vg_init_object(&mask->base, ctx, VG_OBJECT_MASK);
+   mask->width = width;
+   mask->height = height;
+
+   {
+      struct pipe_texture pt;
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      memset(&pt, 0, sizeof(pt));
+      pt.target = PIPE_TEXTURE_2D;
+      pt.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &pt.block);
+      pt.last_level = 0;
+      pt.width[0] = width;
+      pt.height[0] = height;
+      pt.depth[0] = 1;
+      pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+      pt.compressed = 0;
+
+      mask->texture = screen->texture_create(screen, &pt);
+   }
+
+   vg_context_add_object(ctx, VG_OBJECT_MASK, mask);
+
+   return mask;
+}
+
+void mask_layer_destroy(struct vg_mask_layer *layer)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   vg_context_remove_object(ctx, VG_OBJECT_MASK, layer);
+   pipe_texture_release(&layer->texture);
+   free(layer);
+}
+
+void mask_layer_fill(struct vg_mask_layer *layer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat alpha_color[4] = {0, 0, 0, 0};
+   struct pipe_surface *surface;
+
+   alpha_color[3] = value;
+
+   surface = ctx->pipe->screen->get_tex_surface(
+      ctx->pipe->screen, layer->texture,
+      0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   surface_fill(surface,
+                layer->width, layer->height,
+                x, y, width, height, alpha_color);
+
+   ctx->pipe->screen->tex_surface_release(ctx->pipe->screen, &surface);
+}
+
+void mask_copy(struct vg_mask_layer *layer,
+               VGint sx, VGint sy,
+               VGint dx, VGint dy,
+               VGint width, VGint height)
+{
+    struct vg_context *ctx = vg_current_context();
+    struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+
+    renderer_copy_texture(ctx->renderer,
+                          layer->texture,
+                          sx, sy,
+                          sx + width, sy + height,
+                          fb_buffers->alpha_mask,
+                          dx, dy,
+                          dx + width, dy + height);
+}
+
+static void mask_layer_render_to(struct vg_mask_layer *layer,
+                                 struct path *path,
+                                 VGbitfield paint_modes)
+{
+   struct vg_context *ctx = vg_current_context();
+   const VGfloat fill_color[4] = {1.f, 1.f, 1.f, 1.f};
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct pipe_surface *surface;
+
+   surface = screen->get_tex_surface(screen, layer->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_blend();
+   setup_mask_fill(fill_color);
+   setup_mask_framebuffer(surface, layer->width, layer->height);
+   setup_mask_viewport();
+
+   if (paint_modes & VG_FILL_PATH) {
+      struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+      path_fill(path, mat);
+   }
+
+   if (paint_modes & VG_STROKE_PATH){
+      path_stroke(path);
+   }
+
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+   ctx->state.dirty |= BLEND_DIRTY;
+
+   screen->tex_surface_release(ctx->pipe->screen, &surface);
+}
+
+void mask_render_to(struct path *path,
+                    VGbitfield paint_modes,
+                    VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+   struct vg_mask_layer *temp_layer;
+   VGint width, height;
+
+   width = fb_buffers->alpha_mask->width[0];
+   height = fb_buffers->alpha_mask->width[0];
+
+   temp_layer = mask_layer_create(width, height);
+
+   mask_layer_render_to(temp_layer, path, paint_modes);
+
+   mask_using_layer(temp_layer, 0, 0, width, height,
+                    operation);
+
+   mask_layer_destroy(temp_layer);
+}
+
+void mask_using_layer(struct vg_mask_layer *layer,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height)
+{
+   mask_using_texture(layer->texture, operation,
+                      x, y, width, height);
+}
+
+VGint mask_layer_width(struct vg_mask_layer *layer)
+{
+   return layer->width;
+}
+
+VGint mask_layer_height(struct vg_mask_layer *layer)
+{
+   return layer->height;
+}
+
+
+#endif
+
+void mask_using_image(struct vg_image *image,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height)
+{
+   mask_using_texture(image->texture, operation,
+                      x, y, width, height);
+}
+
+void mask_fill(VGint x, VGint y, VGint width, VGint height,
+               VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat alpha_color[4] = {.0f, .0f, .0f, value};
+   struct pipe_surface *surf = alpha_mask_surface(
+      ctx, PIPE_BUFFER_USAGE_GPU_WRITE);
+
+#if DEBUG_MASKS
+   debug_printf("mask_fill(%d, %d, %d, %d) with  rgba(%f, %f, %f, %f)\n",
+                x, y, width, height,
+                alpha_color[0], alpha_color[1],
+                alpha_color[2], alpha_color[3]);
+   debug_printf("XXX %f  === %f \n",
+                alpha_color[3], value);
+#endif
+
+   surface_fill(surf, surf->width, surf->height,
+                x, y, width, height, alpha_color);
+
+   pipe_surface_reference(&surf, NULL);
+}
+
+VGint mask_bind_samplers(struct pipe_sampler_state **samplers,
+                         struct pipe_texture **textures)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (ctx->state.vg.masking) {
+      struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+
+      samplers[1] = &ctx->mask.sampler;
+      textures[1] = fb_buffers->alpha_mask;
+      return 1;
+   } else
+      return 0;
+}
diff --git a/src/gallium/state_trackers/vega/mask.h b/src/gallium/state_trackers/vega/mask.h
new file mode 100644
index 0000000000..5eaaede0e3
--- /dev/null
+++ b/src/gallium/state_trackers/vega/mask.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef MASK_H
+#define MASK_H
+
+#include "vg_context.h"
+
+struct path;
+struct vg_image;
+struct pipe_texture;
+
+struct vg_mask_layer *mask_layer_create(VGint width, VGint height);
+void mask_layer_destroy(struct vg_mask_layer *layer);
+void mask_layer_fill(struct vg_mask_layer *layer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value);
+VGint mask_layer_width(struct vg_mask_layer *layer);
+VGint mask_layer_height(struct vg_mask_layer *layer);
+void mask_copy(struct vg_mask_layer *layer,
+               VGint sx, VGint sy,
+               VGint dx, VGint dy,
+               VGint width, VGint height);
+
+void mask_render_to(struct path *path,
+                    VGbitfield paint_modes,
+                    VGMaskOperation operation);
+
+void mask_using_layer(struct vg_mask_layer *layer,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height);
+void mask_using_image(struct vg_image *image,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height);
+void mask_fill(VGint x, VGint y,
+               VGint width, VGint height,
+               VGfloat value);
+
+VGint mask_bind_samplers(struct pipe_sampler_state **samplers,
+                         struct pipe_texture **textures);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/matrix.h b/src/gallium/state_trackers/vega/matrix.h
new file mode 100644
index 0000000000..4c207f912a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/matrix.h
@@ -0,0 +1,462 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef MATRIX_H
+#define MATRIX_H
+
+#include "VG/openvg.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_math.h"
+
+#include <stdio.h>
+#include <math.h>
+
+#define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y)))
+#define floatIsZero(x) (floatsEqual((x) + 1, 1))
+#define ABS(x) (fabsf(x))
+
+#define DEGREES_TO_RADIANS(d) (0.0174532925199 * (d))
+#define FLT_TO_INT(flt) float_to_int_floor(((VGuint*)&(flt))[0])
+
+static INLINE VGint float_to_int_floor(VGuint bits)
+{
+   int sign = (bits >> 31) ? -1 : 1;
+   int exp  = ((bits >> 23) & 255) - 127;
+   int mant = bits & 0x007fffff;
+   int sh   = 23 - exp;
+
+   /* abs(value) >= 2^31 -> clamp. */
+
+   if (exp >= 31)
+      return (VGint)((sign < 0) ? 0x80000000u : 0x7fffffffu);
+
+   /* abs(value) < 1 -> return -1 or 0. */
+
+   if (exp < 0)
+      return (sign < 0 && (exp > -127 || mant != 0)) ? -1 : 0;
+
+   /* abs(value) >= 2^23 -> shift left. */
+
+   mant |= 0x00800000;
+   if (sh <= 0)
+      return sign * (mant << -sh);
+
+   /* Negative -> add a rounding term. */
+
+   if (sign < 0)
+      mant += (1 << sh) - 1;
+
+   /* Shift right to obtain the result. */
+
+   return sign * (mant >> sh);
+}
+
+
+struct matrix {
+   VGfloat m[9];
+};
+
+static INLINE void matrix_init(struct matrix *mat,
+                               const VGfloat *val)
+{
+   memcpy(mat->m, val, sizeof(VGfloat) * 9);
+}
+
+static INLINE void matrix_inits(struct matrix *mat,
+                                VGfloat m11, VGfloat m12, VGfloat m13,
+                                VGfloat m21, VGfloat m22, VGfloat m23,
+                                VGfloat m31, VGfloat m32, VGfloat m33)
+{
+   mat->m[0] = m11; mat->m[1] = m12; mat->m[2] = m13;
+   mat->m[3] = m21; mat->m[4] = m22; mat->m[5] = m23;
+   mat->m[6] = m31; mat->m[7] = m32; mat->m[8] = m33;
+}
+
+static INLINE void matrix_load_identity(struct matrix *matrix)
+{
+   static const VGfloat identity[9] = {1.f, 0.f, 0.f,
+                                       0.f, 1.f, 0.f,
+                                       0.f, 0.f, 1.f};
+   memcpy(matrix->m, identity, sizeof(identity));
+}
+
+static INLINE VGboolean matrix_is_identity(struct matrix *matrix)
+{
+   return floatsEqual(matrix->m[0], 1) && floatIsZero(matrix->m[1]) &&
+      floatIsZero(matrix->m[2]) &&
+      floatIsZero(matrix->m[3]) && floatsEqual(matrix->m[4], 1) &&
+      floatIsZero(matrix->m[5]) &&
+      floatIsZero(matrix->m[6]) && floatIsZero(matrix->m[7]) &&
+      floatIsZero(matrix->m[8]);
+}
+
+static INLINE VGboolean matrix_is_affine(struct matrix *matrix)
+{
+   return floatIsZero(matrix->m[2]) && floatIsZero(matrix->m[5])
+      && floatsEqual(matrix->m[8], 1);
+}
+
+
+static INLINE void matrix_make_affine(struct matrix *matrix)
+{
+   matrix->m[2] = 0.f;
+   matrix->m[5] = 0.f;
+   matrix->m[8] = 1.f;
+}
+
+static INLINE void matrix_mult(struct matrix *dst,
+                               struct matrix *src)
+{
+   VGfloat m11 = dst->m[0]*src->m[0] + dst->m[3]*src->m[1] + dst->m[6]*src->m[2];
+   VGfloat m12 = dst->m[0]*src->m[3] + dst->m[3]*src->m[4] + dst->m[6]*src->m[5];
+   VGfloat m13 = dst->m[0]*src->m[6] + dst->m[3]*src->m[7] + dst->m[6]*src->m[8];
+
+   VGfloat m21 = dst->m[1]*src->m[0] + dst->m[4]*src->m[1] + dst->m[7]*src->m[2];
+   VGfloat m22 = dst->m[1]*src->m[3] + dst->m[4]*src->m[4] + dst->m[7]*src->m[5];
+   VGfloat m23 = dst->m[1]*src->m[6] + dst->m[4]*src->m[7] + dst->m[7]*src->m[8];
+
+   VGfloat m31 = dst->m[2]*src->m[0] + dst->m[5]*src->m[1] + dst->m[8]*src->m[2];
+   VGfloat m32 = dst->m[2]*src->m[3] + dst->m[5]*src->m[4] + dst->m[8]*src->m[5];
+   VGfloat m33 = dst->m[2]*src->m[6] + dst->m[5]*src->m[7] + dst->m[8]*src->m[8];
+
+   dst->m[0] = m11; dst->m[1] = m21; dst->m[2] = m31;
+   dst->m[3] = m12; dst->m[4] = m22; dst->m[5] = m32;
+   dst->m[6] = m13; dst->m[7] = m23; dst->m[8] = m33;
+}
+
+
+static INLINE void matrix_map_point(struct matrix *mat,
+                                    VGfloat x, VGfloat y,
+                                    VGfloat *out_x, VGfloat *out_y)
+{
+   /* to be able to do matrix_map_point(m, x, y, &x, &y) use
+    * temporaries */
+   VGfloat tmp_x = x, tmp_y = y;
+
+   *out_x = mat->m[0]*tmp_x + mat->m[3]*tmp_y + mat->m[6];
+   *out_y = mat->m[1]*tmp_x + mat->m[4]*tmp_y + mat->m[7];
+   if (!matrix_is_affine(mat)) {
+      VGfloat w = 1/(mat->m[2]*tmp_x + mat->m[5]*tmp_y + mat->m[8]);
+      *out_x *= w;
+      *out_y *= w;
+   }
+}
+
+static INLINE void matrix_translate(struct matrix *dst,
+                                    VGfloat tx, VGfloat ty)
+{
+   if (!matrix_is_affine(dst)) {
+      struct matrix trans_matrix;
+      matrix_load_identity(&trans_matrix);
+      trans_matrix.m[6] = tx;
+      trans_matrix.m[7] = ty;
+      matrix_mult(dst, &trans_matrix);
+   } else {
+      dst->m[6] += tx*dst->m[0] + ty*dst->m[3];
+      dst->m[7] += ty*dst->m[4] + tx*dst->m[1];
+   }
+}
+
+static INLINE void matrix_scale(struct matrix *dst,
+                                VGfloat sx, VGfloat sy)
+{
+   if (!matrix_is_affine(dst)) {
+      struct matrix scale_matrix;
+      matrix_load_identity(&scale_matrix);
+      scale_matrix.m[0] = sx;
+      scale_matrix.m[4] = sy;
+      matrix_mult(dst, &scale_matrix);
+   } else {
+      dst->m[0] *= sx; dst->m[1] *= sx;
+      dst->m[3] *= sy; dst->m[4] *= sy;
+   }
+}
+
+static INLINE void matrix_shear(struct matrix *dst,
+                                VGfloat shx, VGfloat shy)
+{
+   struct matrix shear_matrix;
+   matrix_load_identity(&shear_matrix);
+   shear_matrix.m[1] = shy;
+   shear_matrix.m[3] = shx;
+   matrix_mult(dst, &shear_matrix);
+}
+
+static INLINE void matrix_rotate(struct matrix *dst,
+                                 VGfloat angle)
+{
+   struct matrix mat;
+   float sin_val = 0;
+   float cos_val = 0;
+
+
+   if (floatsEqual(angle, 90) || floatsEqual(angle, -270))
+      sin_val = 1.f;
+   else if (floatsEqual(angle, 270) || floatsEqual(angle, -90))
+      sin_val = -1.f;
+   else if (floatsEqual(angle, 180))
+      cos_val = -1.f;
+   else {
+      float radians = DEGREES_TO_RADIANS(angle);
+      sin_val = sin(radians);
+      cos_val = cos(radians);
+   }
+
+   if (!matrix_is_affine(dst)) {
+      matrix_load_identity(&mat);
+      mat.m[0] =  cos_val;   mat.m[1] =  sin_val;
+      mat.m[3] = -sin_val;   mat.m[4] =  cos_val;
+
+      matrix_mult(dst, &mat);
+   } else  {
+      VGfloat m11 =  cos_val*dst->m[0] + sin_val*dst->m[3];
+      VGfloat m12 =  cos_val*dst->m[1] + sin_val*dst->m[4];
+      VGfloat m21 = -sin_val*dst->m[0] + cos_val*dst->m[3];
+      VGfloat m22 = -sin_val*dst->m[1] + cos_val*dst->m[4];
+      dst->m[0] = m11; dst->m[1] = m12;
+      dst->m[3] = m21; dst->m[4] = m22;
+   }
+}
+
+
+static INLINE VGfloat matrix_determinant(struct matrix *mat)
+{
+   return mat->m[0]*(mat->m[8]*mat->m[4]-mat->m[7]*mat->m[5]) -
+      mat->m[3]*(mat->m[8]*mat->m[1]-mat->m[7]*mat->m[2])+
+      mat->m[6]*(mat->m[5]*mat->m[1]-mat->m[4]*mat->m[2]);
+}
+
+
+static INLINE void matrix_adjoint(struct matrix *mat)
+{
+    VGfloat h[9];
+    h[0] = mat->m[4]*mat->m[8] - mat->m[5]*mat->m[7];
+    h[3] = mat->m[5]*mat->m[6] - mat->m[3]*mat->m[8];
+    h[6] = mat->m[3]*mat->m[7] - mat->m[4]*mat->m[6];
+    h[1] = mat->m[2]*mat->m[7] - mat->m[1]*mat->m[8];
+    h[4] = mat->m[0]*mat->m[8] - mat->m[2]*mat->m[6];
+    h[7] = mat->m[1]*mat->m[6] - mat->m[0]*mat->m[7];
+    h[2] = mat->m[1]*mat->m[5] - mat->m[2]*mat->m[4];
+    h[5] = mat->m[2]*mat->m[3] - mat->m[0]*mat->m[5];
+    h[8] = mat->m[0]*mat->m[4] - mat->m[1]*mat->m[3];
+
+
+    memcpy(mat->m, h, sizeof(VGfloat) * 9);
+}
+
+static INLINE void matrix_divs(struct matrix *mat,
+                               VGfloat s)
+{
+   mat->m[0] /= s;
+   mat->m[1] /= s;
+   mat->m[2] /= s;
+   mat->m[3] /= s;
+   mat->m[4] /= s;
+   mat->m[5] /= s;
+   mat->m[6] /= s;
+   mat->m[7] /= s;
+   mat->m[8] /= s;
+}
+
+static INLINE VGboolean matrix_invert(struct matrix *mat)
+{
+   VGfloat det = matrix_determinant(mat);
+
+   if (floatIsZero(det))
+      return VG_FALSE;
+
+   matrix_adjoint(mat);
+   matrix_divs(mat, det);
+   return VG_TRUE;
+}
+
+static INLINE VGboolean matrix_is_invertible(struct matrix *mat)
+{
+   return !floatIsZero(matrix_determinant(mat));
+}
+
+
+static INLINE VGboolean matrix_square_to_quad(VGfloat dx0, VGfloat dy0,
+                                              VGfloat dx1, VGfloat dy1,
+                                              VGfloat dx3, VGfloat dy3,
+                                              VGfloat dx2, VGfloat dy2,
+                                              struct matrix *mat)
+{
+   VGfloat ax  = dx0 - dx1 + dx2 - dx3;
+   VGfloat ay  = dy0 - dy1 + dy2 - dy3;
+
+   if (floatIsZero(ax) && floatIsZero(ay)) {
+      /* affine case */
+      matrix_inits(mat,
+                   dx1 - dx0, dy1 - dy0,  0,
+                   dx2 - dx1, dy2 - dy1,  0,
+                         dx0,       dy0,  1);
+   } else {
+      VGfloat a, b, c, d, e, f, g, h;
+      VGfloat ax1 = dx1 - dx2;
+      VGfloat ax2 = dx3 - dx2;
+      VGfloat ay1 = dy1 - dy2;
+      VGfloat ay2 = dy3 - dy2;
+
+      /* determinants */
+      VGfloat gtop    =  ax  * ay2 - ax2 * ay;
+      VGfloat htop    =  ax1 * ay  - ax  * ay1;
+      VGfloat bottom  =  ax1 * ay2 - ax2 * ay1;
+
+      if (!bottom)
+         return VG_FALSE;
+
+      g = gtop / bottom;
+      h = htop / bottom;
+
+      a = dx1 - dx0 + g * dx1;
+      b = dx3 - dx0 + h * dx3;
+      c = dx0;
+      d = dy1 - dy0 + g * dy1;
+      e = dy3 - dy0 + h * dy3;
+      f = dy0;
+
+      matrix_inits(mat,
+                   a,  d,   g,
+                   b,  e,   h,
+                   c,  f, 1.f);
+   }
+
+   return VG_TRUE;
+}
+
+static INLINE VGboolean matrix_quad_to_square(VGfloat sx0, VGfloat sy0,
+                                              VGfloat sx1, VGfloat sy1,
+                                              VGfloat sx2, VGfloat sy2,
+                                              VGfloat sx3, VGfloat sy3,
+                                              struct matrix *mat)
+{
+   if (!matrix_square_to_quad(sx0, sy0, sx1, sy1,
+                              sx2, sy2, sx3, sy3,
+                              mat))
+      return VG_FALSE;
+
+    return matrix_invert(mat);
+}
+
+
+static INLINE VGboolean matrix_quad_to_quad(VGfloat dx0, VGfloat dy0,
+                                            VGfloat dx1, VGfloat dy1,
+                                            VGfloat dx2, VGfloat dy2,
+                                            VGfloat dx3, VGfloat dy3,
+                                            VGfloat sx0, VGfloat sy0,
+                                            VGfloat sx1, VGfloat sy1,
+                                            VGfloat sx2, VGfloat sy2,
+                                            VGfloat sx3, VGfloat sy3,
+                                            struct matrix *mat)
+{
+   struct matrix sqr_to_qd;
+
+   if (!matrix_square_to_quad(dx0, dy0, dx1, dy1,
+                              dx2, dy2, dx3, dy3,
+                              mat))
+      return VG_FALSE;
+
+   if (!matrix_quad_to_square(sx0, sy0, sx1, sy1,
+                              sx2, sy2, sx3, sy3,
+                              &sqr_to_qd))
+      return VG_FALSE;
+
+   matrix_mult(mat, &sqr_to_qd);
+
+   return VG_TRUE;
+}
+
+
+static INLINE VGboolean null_line(const VGfloat *l)
+{
+   return floatsEqual(l[0], l[2]) && floatsEqual(l[1], l[3]);
+}
+
+static INLINE void line_normal(float *l, float *norm)
+{
+   norm[0] = l[0];
+   norm[1] = l[1];
+
+   norm[2] = l[0] + (l[3] - l[1]);
+   norm[3] = l[1] - (l[2] - l[0]);
+}
+
+static INLINE void line_normalize(float *l)
+{
+   float x = l[2] - l[0];
+   float y = l[3] - l[1];
+   float len = sqrt(x*x + y*y);
+   l[2] = l[0] + x/len;
+   l[3] = l[1] + y/len;
+}
+
+static INLINE VGfloat line_length(VGfloat x1, VGfloat y1,
+                                  VGfloat x2, VGfloat y2)
+{
+   VGfloat x = x2 - x1;
+   VGfloat y = y2 - y1;
+   return sqrt(x*x + y*y);
+}
+
+static INLINE VGfloat line_lengthv(const VGfloat *l)
+{
+   VGfloat x = l[2] - l[0];
+   VGfloat y = l[3] - l[1];
+   return sqrt(x*x + y*y);
+}
+
+
+static INLINE void line_point_at(float *l, float t, float *pt)
+{
+   float dx = l[2] - l[0];
+   float dy = l[3] - l[1];
+
+   pt[0] = l[0] + dx * t;
+   pt[1] = l[1] + dy * t;
+}
+
+static INLINE void vector_unit(float *vec)
+{
+   float len = sqrt(vec[0] * vec[0] + vec[1] * vec[1]);
+   vec[0] /= len;
+   vec[1] /= len;
+}
+
+static INLINE void line_normal_vector(float *line, float *vec)
+{
+   VGfloat normal[4];
+
+   line_normal(line, normal);
+
+   vec[0] = normal[2] - normal[0];
+   vec[1] = normal[3] - normal[1];
+
+   vector_unit(vec);
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c
new file mode 100644
index 0000000000..04a6ba9cdc
--- /dev/null
+++ b/src/gallium/state_trackers/vega/paint.c
@@ -0,0 +1,699 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "paint.h"
+
+#include "shaders_cache.h"
+#include "matrix.h"
+#include "image.h"
+#include "st_inlines.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_inlines.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "cso_cache/cso_context.h"
+
+struct vg_paint {
+   struct vg_object base;
+
+   VGPaintType type;
+
+   struct {
+      VGfloat color[4];
+      VGint colori[4];
+   } solid;
+
+   struct {
+      VGColorRampSpreadMode spread;
+      VGuint color_data[1024];
+      struct {
+         VGfloat  coords[4];
+         VGint  coordsi[4];
+      } linear;
+      struct {
+         VGfloat vals[5];
+         VGint valsi[5];
+      } radial;
+      struct pipe_texture *texture;
+      struct pipe_sampler_state sampler;
+
+      VGfloat *ramp_stops;
+      VGint *ramp_stopsi;
+      VGint    num_stops;
+
+      VGboolean color_ramps_premultiplied;
+   } gradient;
+
+   struct {
+      struct pipe_texture *texture;
+      VGTilingMode tiling_mode;
+      struct pipe_sampler_state sampler;
+   } pattern;
+
+   struct pipe_constant_buffer cbuf;
+   struct pipe_shader_state fs_state;
+   void *fs;
+};
+
+static INLINE VGuint mix_pixels(VGuint p1, VGuint a, VGuint p2, VGuint b)
+{
+   VGuint t = (p1 & 0xff00ff) * a + (p2 & 0xff00ff) * b;
+   t >>= 8; t &= 0xff00ff;
+
+   p1 = ((p1 >> 8) & 0xff00ff) * a + ((p2 >> 8) & 0xff00ff) * b;
+   p1 &= 0xff00ff00; p1 |= t;
+
+   return p1;
+}
+
+static INLINE VGuint float4_to_argb(const VGfloat *clr)
+{
+   return float_to_ubyte(clr[3]) << 24 |
+      float_to_ubyte(clr[0]) << 16 |
+      float_to_ubyte(clr[1]) << 8 |
+      float_to_ubyte(clr[2]) << 0;
+}
+
+static INLINE void create_gradient_data(const VGfloat *ramp_stops,
+                                        VGint num,
+                                        VGuint *data,
+                                        VGint size)
+{
+   VGint i;
+   VGint pos = 0;
+   VGfloat fpos = 0, incr = 1.f / size;
+   VGuint last_color;
+
+   while (fpos < ramp_stops[0]) {
+      data[pos] = float4_to_argb(ramp_stops + 1);
+      fpos += incr;
+      ++pos;
+   }
+
+   for (i = 0; i < num - 1; ++i) {
+      VGint rcur  = 5 * i;
+      VGint rnext = 5 * (i + 1);
+      VGfloat delta = 1.f/(ramp_stops[rnext] - ramp_stops[rcur]);
+      while (fpos < ramp_stops[rnext] && pos < size) {
+         VGint dist = 256 * ((fpos - ramp_stops[rcur]) * delta);
+         VGint idist = 256 - dist;
+         VGuint current_color = float4_to_argb(ramp_stops + rcur + 1);
+         VGuint next_color = float4_to_argb(ramp_stops + rnext + 1);
+         data[pos] = mix_pixels(current_color, idist,
+                                next_color, dist);
+         fpos += incr;
+         ++pos;
+      }
+   }
+
+   last_color = float4_to_argb(ramp_stops + ((num - 1) * 5 + 1));
+   while (pos < size) {
+      data[pos] = last_color;
+      ++pos;
+   }
+   data[size-1] = last_color;
+}
+
+static INLINE struct pipe_texture *create_gradient_texture(struct vg_paint *p)
+{
+   struct pipe_context *pipe = p->base.ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *tex = 0;
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_1D;
+   templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templ.last_level = 0;
+   templ.width[0] = 1024;
+   templ.height[0] = 1;
+   templ.depth[0] = 1;
+   pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   tex = screen->texture_create(screen, &templ);
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer =
+         st_no_flush_get_tex_transfer(p->base.ctx, tex, 0, 0, 0,
+                                      PIPE_TRANSFER_WRITE, 0, 0, 1024, 1);
+      void *map = screen->transfer_map(screen, transfer);
+      memcpy(map, p->gradient.color_data, sizeof(VGint)*1024);
+      screen->transfer_unmap(screen, transfer);
+      screen->tex_transfer_destroy(transfer);
+   }
+
+   return tex;
+}
+
+struct vg_paint * paint_create(struct vg_context *ctx)
+{
+   struct vg_paint *paint = CALLOC_STRUCT(vg_paint);
+   const VGfloat default_color[] = {0.0f, 0.0f, 0.0f, 1.0f};
+   const VGfloat def_ling[] = {0.0f, 0.0f, 1.0f, 0.0f};
+   const VGfloat def_radg[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
+   vg_init_object(&paint->base, ctx, VG_OBJECT_PAINT);
+   vg_context_add_object(ctx, VG_OBJECT_PAINT, paint);
+
+   paint->type = VG_PAINT_TYPE_COLOR;
+   memcpy(paint->solid.color, default_color,
+          4 * sizeof(VGfloat));
+   paint->gradient.spread = VG_COLOR_RAMP_SPREAD_PAD;
+   memcpy(paint->gradient.linear.coords, def_ling,
+          4 * sizeof(VGfloat));
+   memcpy(paint->gradient.radial.vals, def_radg,
+          5 * sizeof(VGfloat));
+
+   paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   paint->gradient.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   paint->gradient.sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   paint->gradient.sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   paint->gradient.sampler.normalized_coords = 1;
+
+   memcpy(&paint->pattern.sampler,
+          &paint->gradient.sampler,
+          sizeof(struct pipe_sampler_state));
+
+   return paint;
+}
+
+void paint_destroy(struct vg_paint *paint)
+{
+   struct vg_context *ctx = paint->base.ctx;
+   if (paint->pattern.texture)
+      pipe_texture_reference(&paint->pattern.texture, NULL);
+   if (ctx)
+      vg_context_remove_object(ctx, VG_OBJECT_PAINT, paint);
+
+   free(paint->gradient.ramp_stopsi);
+   free(paint->gradient.ramp_stops);
+   free(paint);
+}
+
+void paint_set_color(struct vg_paint *paint,
+                     const VGfloat *color)
+{
+   paint->solid.color[0] = color[0];
+   paint->solid.color[1] = color[1];
+   paint->solid.color[2] = color[2];
+   paint->solid.color[3] = color[3];
+
+   paint->solid.colori[0] = FLT_TO_INT(color[0]);
+   paint->solid.colori[1] = FLT_TO_INT(color[1]);
+   paint->solid.colori[2] = FLT_TO_INT(color[2]);
+   paint->solid.colori[3] = FLT_TO_INT(color[3]);
+}
+
+static INLINE void paint_color_buffer(struct vg_paint *paint, void *buffer)
+{
+   VGfloat *map = (VGfloat*)buffer;
+   memcpy(buffer, paint->solid.color, 4 * sizeof(VGfloat));
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+}
+
+static INLINE void paint_linear_gradient_buffer(struct vg_paint *paint, void *buffer)
+{
+   struct vg_context *ctx = paint->base.ctx;
+   VGfloat *map = (VGfloat*)buffer;
+
+   map[0] = paint->gradient.linear.coords[2] - paint->gradient.linear.coords[0];
+   map[1] = paint->gradient.linear.coords[3] - paint->gradient.linear.coords[1];
+   map[2] = 1.f / (map[0] * map[0] + map[1] * map[1]);
+   map[3] = 1.f;
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+   {
+      struct matrix mat;
+      struct matrix inv;
+      matrix_load_identity(&mat);
+      matrix_translate(&mat, -paint->gradient.linear.coords[0], -paint->gradient.linear.coords[1]);
+      memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&inv);
+      matrix_mult(&inv, &mat);
+      memcpy(&mat, &inv,
+             sizeof(struct matrix));
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+#if 0
+   debug_printf("Coords  (%f, %f, %f, %f)\n",
+                map[0], map[1], map[2], map[3]);
+#endif
+}
+
+
+static INLINE void paint_radial_gradient_buffer(struct vg_paint *paint, void *buffer)
+{
+   VGfloat *radialCoords = paint->gradient.radial.vals;
+   struct vg_context *ctx = paint->base.ctx;
+
+   VGfloat *map = (VGfloat*)buffer;
+
+   map[0] = radialCoords[0] - radialCoords[2];
+   map[1] = radialCoords[1] - radialCoords[3];
+   map[2] = -map[0] * map[0] - map[1] * map[1] +
+            radialCoords[4] * radialCoords[4];
+   map[3] = 1.f;
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+
+   {
+      struct matrix mat;
+      struct matrix inv;
+      matrix_load_identity(&mat);
+      matrix_translate(&mat, -radialCoords[2], -radialCoords[3]);
+      memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&inv);
+      matrix_mult(&inv, &mat);
+      memcpy(&mat, &inv,
+             sizeof(struct matrix));
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+
+#if 0
+   debug_printf("Coords  (%f, %f, %f, %f)\n",
+                map[0], map[1], map[2], map[3]);
+#endif
+}
+
+
+static INLINE void  paint_pattern_buffer(struct vg_paint *paint, void *buffer)
+{
+   struct vg_context *ctx = paint->base.ctx;
+
+   VGfloat *map = (VGfloat *)buffer;
+   memcpy(map, paint->solid.color, 4 * sizeof(VGfloat));
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = paint->pattern.texture->width[0];
+   map[7] = paint->pattern.texture->height[0];
+   {
+      struct matrix mat;
+      memcpy(&mat, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&mat);
+      {
+         struct matrix pm;
+         memcpy(&pm, &ctx->state.vg.path_user_to_surface_matrix,
+                sizeof(struct matrix));
+         matrix_invert(&pm);
+         matrix_mult(&pm, &mat);
+         memcpy(&mat, &pm, sizeof(struct matrix));
+      }
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+}
+
+void paint_set_type(struct vg_paint *paint, VGPaintType type)
+{
+   paint->type = type;
+}
+
+void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops,
+                          int num)
+{
+   const VGfloat default_stops[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f,
+                                    1.0f, 1.0f, 1.0f, 1.0f, 1.0f};
+   VGint i;
+   const VGint num_stops = num / 5;
+   VGfloat last_coord;
+
+   paint->gradient.num_stops = num;
+   if (num) {
+      free(paint->gradient.ramp_stops);
+      paint->gradient.ramp_stops = malloc(sizeof(VGfloat)*num);
+      memcpy(paint->gradient.ramp_stops, stops, sizeof(VGfloat)*num);
+   } else
+      return;
+
+   /* stops must be in increasing order. the last stop is 1.0. if the
+    * first one is bigger than 1 then the whole sequence is invalid*/
+   if (stops[0] > 1) {
+      stops = default_stops;
+      num = 10;
+   }
+   last_coord = stops[0];
+   for (i = 1; i < num_stops; ++i) {
+      VGint idx = 5 * i;
+      VGfloat coord = stops[idx];
+      if (!floatsEqual(last_coord, coord) && coord < last_coord) {
+         stops = default_stops;
+         num = 10;
+         break;
+      }
+      last_coord = coord;
+   }
+
+   create_gradient_data(stops, num / 5, paint->gradient.color_data,
+                        1024);
+
+   if (paint->gradient.texture) {
+      pipe_texture_reference(&paint->gradient.texture, NULL);
+      paint->gradient.texture = 0;
+   }
+
+   paint->gradient.texture = create_gradient_texture(paint);
+}
+
+void paint_set_colori(struct vg_paint *p,
+                      VGuint rgba)
+{
+   p->solid.color[0] = ((rgba >> 24) & 0xff) / 255.f;
+   p->solid.color[1] = ((rgba >> 16) & 0xff) / 255.f;
+   p->solid.color[2] = ((rgba >>  8) & 0xff) / 255.f;
+   p->solid.color[3] = ((rgba >>  0) & 0xff) / 255.f;
+}
+
+VGuint paint_colori(struct vg_paint *p)
+{
+#define F2B(f) (float_to_ubyte(f))
+
+   return ((F2B(p->solid.color[0]) << 24) |
+           (F2B(p->solid.color[1]) << 16) |
+           (F2B(p->solid.color[2]) << 8)  |
+           (F2B(p->solid.color[3]) << 0));
+#undef F2B
+}
+
+void paint_set_linear_gradient(struct vg_paint *paint,
+                               const VGfloat *coords)
+{
+   memcpy(paint->gradient.linear.coords, coords, sizeof(VGfloat) * 4);
+}
+
+void paint_set_spread_mode(struct vg_paint *paint,
+                           VGint mode)
+{
+   paint->gradient.spread = mode;
+   switch(mode) {
+   case VG_COLOR_RAMP_SPREAD_PAD:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_COLOR_RAMP_SPREAD_REPEAT:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_COLOR_RAMP_SPREAD_REFLECT:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   }
+}
+
+VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint)
+{
+   return paint->gradient.spread;
+}
+
+void paint_set_radial_gradient(struct vg_paint *paint,
+                               const VGfloat *values)
+{
+   memcpy(paint->gradient.radial.vals, values, sizeof(VGfloat) * 5);
+}
+
+void paint_set_pattern(struct vg_paint *paint,
+                       struct vg_image *img)
+{
+   if (paint->pattern.texture)
+      pipe_texture_reference(&paint->pattern.texture, NULL);
+
+   paint->pattern.texture = 0;
+   pipe_texture_reference(&paint->pattern.texture,
+                          img->texture);
+}
+
+void paint_set_pattern_tiling(struct vg_paint *paint,
+                              VGTilingMode mode)
+{
+   paint->pattern.tiling_mode = mode;
+
+   switch(mode) {
+   case VG_TILE_FILL:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      break;
+   case VG_TILE_PAD:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_TILE_REPEAT:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_TILE_REFLECT:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   default:
+      debug_assert("!Unknown tiling mode");
+   }
+}
+
+void paint_get_color(struct vg_paint *paint,
+                     VGfloat *color)
+{
+   color[0] = paint->solid.color[0];
+   color[1] = paint->solid.color[1];
+   color[2] = paint->solid.color[2];
+   color[3] = paint->solid.color[3];
+}
+
+void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops,
+                      int num)
+{
+   memcpy(stops, paint->gradient.ramp_stops, sizeof(VGfloat)*num);
+}
+
+void paint_linear_gradient(struct vg_paint *paint,
+                           VGfloat *coords)
+{
+   memcpy(coords, paint->gradient.linear.coords, sizeof(VGfloat)*4);
+}
+
+void paint_radial_gradient(struct vg_paint *paint,
+                           VGfloat *coords)
+{
+   memcpy(coords, paint->gradient.radial.vals, sizeof(VGfloat)*5);
+}
+
+int paint_num_ramp_stops(struct vg_paint *paint)
+{
+   return paint->gradient.num_stops;
+}
+
+VGPaintType paint_type(struct vg_paint *paint)
+{
+   return paint->type;
+}
+
+void paint_set_coloriv(struct vg_paint *paint,
+                      const VGint *color)
+{
+   paint->solid.color[0] = color[0];
+   paint->solid.color[1] = color[1];
+   paint->solid.color[2] = color[2];
+   paint->solid.color[3] = color[3];
+
+   paint->solid.colori[0] = color[0];
+   paint->solid.colori[1] = color[1];
+   paint->solid.colori[2] = color[2];
+   paint->solid.colori[3] = color[3];
+}
+
+void paint_get_coloriv(struct vg_paint *paint,
+                      VGint *color)
+{
+   color[0] = paint->solid.colori[0];
+   color[1] = paint->solid.colori[1];
+   color[2] = paint->solid.colori[2];
+   color[3] = paint->solid.colori[3];
+}
+
+void paint_set_color_ramp_premultiplied(struct vg_paint *paint,
+                                        VGboolean set)
+{
+   paint->gradient.color_ramps_premultiplied = set;
+}
+
+VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint)
+{
+   return paint->gradient.color_ramps_premultiplied;
+}
+
+void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops,
+                           int num)
+{
+   if (num) {
+      free(paint->gradient.ramp_stopsi);
+      paint->gradient.ramp_stopsi = malloc(sizeof(VGint)*num);
+      memcpy(paint->gradient.ramp_stopsi, stops, sizeof(VGint)*num);
+   }
+}
+
+void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops,
+                       int num)
+{
+   memcpy(stops, paint->gradient.ramp_stopsi, sizeof(VGint)*num);
+}
+
+void paint_set_linear_gradienti(struct vg_paint *paint,
+                                const VGint *coords)
+{
+   memcpy(paint->gradient.linear.coordsi, coords, sizeof(VGint) * 4);
+}
+
+void paint_linear_gradienti(struct vg_paint *paint,
+                            VGint *coords)
+{
+   memcpy(coords, paint->gradient.linear.coordsi, sizeof(VGint)*4);
+}
+
+void paint_set_radial_gradienti(struct vg_paint *paint,
+                                const VGint *values)
+{
+   memcpy(paint->gradient.radial.valsi, values, sizeof(VGint) * 5);
+}
+
+void paint_radial_gradienti(struct vg_paint *paint,
+                            VGint *coords)
+{
+   memcpy(coords, paint->gradient.radial.valsi, sizeof(VGint)*5);
+}
+
+VGTilingMode paint_pattern_tiling(struct vg_paint *paint)
+{
+   return paint->pattern.tiling_mode;
+}
+
+VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   switch(paint->type) {
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+   case VG_PAINT_TYPE_RADIAL_GRADIENT: {
+      if (paint->gradient.texture) {
+         paint->gradient.sampler.min_img_filter = image_sampler_filter(ctx);
+         paint->gradient.sampler.mag_img_filter = image_sampler_filter(ctx);
+         samplers[0] = &paint->gradient.sampler;
+         textures[0] = paint->gradient.texture;
+         return 1;
+      }
+   }
+      break;
+   case VG_PAINT_TYPE_PATTERN: {
+      memcpy(paint->pattern.sampler.border_color,
+             ctx->state.vg.tile_fill_color,
+             sizeof(VGfloat) * 4);
+      paint->pattern.sampler.min_img_filter = image_sampler_filter(ctx);
+      paint->pattern.sampler.mag_img_filter = image_sampler_filter(ctx);
+      samplers[0] = &paint->pattern.sampler;
+      textures[0] = paint->pattern.texture;
+      return 1;
+   }
+      break;
+   default:
+      samplers[0] = &paint->pattern.sampler; /* dummy */
+      textures[0] = 0;
+      return 0;
+      break;
+   }
+   return 0;
+}
+
+void paint_resolve_type(struct vg_paint *paint)
+{
+   if (paint->type == VG_PAINT_TYPE_PATTERN &&
+       !paint->pattern.texture) {
+      paint->type = VG_PAINT_TYPE_COLOR;
+   }
+}
+
+VGint paint_constant_buffer_size(struct vg_paint *paint)
+{
+   switch(paint->type) {
+   case VG_PAINT_TYPE_COLOR:
+      return 8 * sizeof(VGfloat);/*4 color + 4 constants (0.f,1.f,2.f,4.f)*/
+      break;
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+      return 20 * sizeof(VGfloat);
+      break;
+   case VG_PAINT_TYPE_RADIAL_GRADIENT:
+      return 20 * sizeof(VGfloat);
+      break;
+   case VG_PAINT_TYPE_PATTERN:
+      return 20 * sizeof(VGfloat);
+      break;
+   default:
+      debug_printf("Uknown paint type: %d\n", paint->type);
+   }
+
+   return 0;
+}
+
+void paint_fill_constant_buffer(struct vg_paint *paint,
+                                void *buffer)
+{
+   switch(paint->type) {
+   case VG_PAINT_TYPE_COLOR:
+      paint_color_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+      paint_linear_gradient_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_RADIAL_GRADIENT:
+      paint_radial_gradient_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_PATTERN:
+      paint_pattern_buffer(paint, buffer);
+      break;
+
+   default:
+      abort();
+   }
+}
diff --git a/src/gallium/state_trackers/vega/paint.h b/src/gallium/state_trackers/vega/paint.h
new file mode 100644
index 0000000000..999b5c167c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/paint.h
@@ -0,0 +1,118 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef PAINT_H
+#define PAINT_H
+
+#include "vg_context.h"
+
+#include "VG/openvg.h"
+#include "pipe/p_state.h"
+
+struct vg_paint;
+struct vg_image;
+struct pipe_sampler_state;
+struct pipe_texture;
+
+struct vg_paint *paint_create(struct vg_context *ctx);
+void paint_destroy(struct vg_paint *paint);
+
+void paint_set_color(struct vg_paint *paint,
+                     const VGfloat *color);
+void paint_get_color(struct vg_paint *paint,
+                     VGfloat *color);
+
+void paint_set_coloriv(struct vg_paint *paint,
+                      const VGint *color);
+void paint_get_coloriv(struct vg_paint *paint,
+                      VGint *color);
+
+void paint_set_colori(struct vg_paint *paint,
+                      VGuint rgba);
+
+VGuint paint_colori(struct vg_paint *paint);
+
+void paint_set_type(struct vg_paint *paint, VGPaintType type);
+VGPaintType paint_type(struct vg_paint *paint);
+void paint_resolve_type(struct vg_paint *paint);
+
+void paint_set_linear_gradient(struct vg_paint *paint,
+                               const VGfloat *coords);
+void paint_linear_gradient(struct vg_paint *paint,
+                           VGfloat *coords);
+void paint_set_linear_gradienti(struct vg_paint *paint,
+                               const VGint *coords);
+void paint_linear_gradienti(struct vg_paint *paint,
+                           VGint *coords);
+
+
+void paint_set_radial_gradient(struct vg_paint *paint,
+                               const VGfloat *values);
+void paint_radial_gradient(struct vg_paint *paint,
+                           VGfloat *coords);
+void paint_set_radial_gradienti(struct vg_paint *paint,
+                                const VGint *values);
+void paint_radial_gradienti(struct vg_paint *paint,
+                            VGint *coords);
+
+
+void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops,
+                          int num);
+void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops,
+                      int num);
+
+void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops,
+                           int num);
+void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops,
+                       int num);
+
+int paint_num_ramp_stops(struct vg_paint *paint);
+
+void paint_set_spread_mode(struct vg_paint *paint,
+                           VGint mode);
+VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint);
+
+
+void paint_set_pattern(struct vg_paint *paint,
+                       struct vg_image *img);
+void paint_set_pattern_tiling(struct vg_paint *paint,
+                              VGTilingMode mode);
+VGTilingMode paint_pattern_tiling(struct vg_paint *paint);
+
+void paint_set_color_ramp_premultiplied(struct vg_paint *paint,
+                                        VGboolean set);
+VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint);
+
+
+VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures);
+
+VGint paint_constant_buffer_size(struct vg_paint *paint);
+void paint_fill_constant_buffer(struct vg_paint *paint,
+                                void *buffer);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/path.c b/src/gallium/state_trackers/vega/path.c
new file mode 100644
index 0000000000..4fc23a7a27
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path.c
@@ -0,0 +1,2034 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "path.h"
+
+#include "stroker.h"
+#include "polygon.h"
+#include "bezier.h"
+#include "matrix.h"
+#include "vg_context.h"
+#include "util_array.h"
+#include "arc.h"
+#include "path_utils.h"
+#include "paint.h"
+#include "shader.h"
+
+#include "util/u_memory.h"
+
+#include <assert.h>
+
+#define DEBUG_PATH 0
+
+struct path {
+   struct vg_object base;
+   VGbitfield caps;
+   VGboolean dirty;
+   VGboolean dirty_stroke;
+
+   VGPathDatatype datatype;
+
+   VGfloat scale;
+   VGfloat bias;
+
+   VGint num_segments;
+
+   struct array * segments;
+   struct array * control_points;
+
+   struct {
+      struct polygon_array polygon_array;
+      struct matrix matrix;
+   } fill_polys;
+
+   struct {
+      struct path *path;
+      struct matrix matrix;
+      VGfloat stroke_width;
+      VGfloat miter_limit;
+      VGCapStyle cap_style;
+      VGJoinStyle join_style;
+   } stroked;
+};
+
+
+static INLINE void data_at(void **data,
+                           struct path *p,
+                           VGint start, VGint count,
+                           VGfloat *out)
+{
+   VGPathDatatype dt = p->datatype;
+   VGint i;
+   VGint end = start + count;
+   VGfloat *itr = out;
+
+   switch(dt) {
+   case VG_PATH_DATATYPE_S_8: {
+      VGbyte **bdata = (VGbyte **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_S_16: {
+      VGshort **bdata = (VGshort **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_S_32: {
+      VGint **bdata = (VGint **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_F: {
+      VGfloat **fdata = (VGfloat **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*fdata)[i];
+         ++itr;
+      }
+      *fdata += count;
+   }
+      break;
+   default:
+      debug_assert(!"Unknown path datatype!");
+   }
+}
+
+
+void vg_float_to_datatype(VGPathDatatype datatype,
+                          VGubyte *common_data,
+                          const VGfloat *data,
+                          VGint num_coords)
+{
+   VGint i;
+   switch(datatype) {
+   case VG_PATH_DATATYPE_S_8: {
+      for (i = 0; i < num_coords; ++i) {
+         common_data[i] = (VGubyte)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_S_16: {
+      VGshort *buf = (VGshort*)common_data;
+      for (i = 0; i < num_coords; ++i) {
+         buf[i] = (VGshort)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_S_32: {
+      VGint *buf = (VGint*)common_data;
+      for (i = 0; i < num_coords; ++i) {
+         buf[i] = (VGint)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_F: {
+      memcpy(common_data, data, sizeof(VGfloat) * num_coords);
+   }
+      break;
+   default:
+      debug_assert(!"Unknown path datatype!");
+   }
+}
+
+static void coords_adjust_by_scale_bias(struct path *p,
+                                        void *pdata, VGint num_coords,
+                                        VGfloat scale, VGfloat bias,
+                                        VGPathDatatype datatype)
+{
+   VGfloat data[8];
+   void *coords = (VGfloat *)pdata;
+   VGubyte *common_data = (VGubyte *)pdata;
+   VGint size_dst = size_for_datatype(datatype);
+   VGint i;
+
+   for (i = 0; i < num_coords; ++i) {
+      data_at(&coords, p, 0, 1, data);
+      data[0] = data[0] * scale + bias;
+      vg_float_to_datatype(datatype, common_data, data, 1);
+      common_data += size_dst;
+   }
+}
+
+struct path * path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias,
+                          VGint segmentCapacityHint,
+                          VGint coordCapacityHint,
+                          VGbitfield capabilities)
+{
+   struct path *path = CALLOC_STRUCT(path);
+
+   vg_init_object(&path->base, vg_current_context(), VG_OBJECT_PATH);
+   path->caps = capabilities & VG_PATH_CAPABILITY_ALL;
+   vg_context_add_object(vg_current_context(), VG_OBJECT_PATH, path);
+
+   path->datatype = dt;
+   path->scale = scale;
+   path->bias = bias;
+
+   path->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8));
+   path->control_points = array_create(size_for_datatype(dt));
+
+   path->dirty = VG_TRUE;
+   path->dirty_stroke = VG_TRUE;
+
+   return path;
+}
+
+void path_destroy(struct path *p)
+{
+   vg_context_remove_object(vg_current_context(), VG_OBJECT_PATH, p);
+
+   array_destroy(p->segments);
+   array_destroy(p->control_points);
+   array_destroy(p->fill_polys.polygon_array.array);
+
+   if (p->stroked.path)
+      path_destroy(p->stroked.path);
+
+   free(p);
+}
+
+VGbitfield path_capabilities(struct path *p)
+{
+   return p->caps;
+}
+
+void path_set_capabilities(struct path *p, VGbitfield bf)
+{
+   p->caps = (bf & VG_PATH_CAPABILITY_ALL);
+}
+
+void path_append_data(struct path *p,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData)
+{
+   VGint old_segments = p->num_segments;
+   VGint num_new_coords = num_elements_for_segments(pathSegments, numSegments);
+   array_append_data(p->segments, pathSegments, numSegments);
+   array_append_data(p->control_points, pathData, num_new_coords);
+
+   p->num_segments += numSegments;
+   if (!floatsEqual(p->scale, 1.f) || !floatsEqual(p->bias, 0.f)) {
+      VGubyte *coords = (VGubyte*)p->control_points->data;
+      coords_adjust_by_scale_bias(p,
+                                  coords + old_segments * p->control_points->datatype_size,
+                                  num_new_coords,
+                                  p->scale, p->bias, p->datatype);
+   }
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+VGint path_num_segments(struct path *p)
+{
+   return p->num_segments;
+}
+
+static INLINE void map_if_relative(VGfloat ox, VGfloat oy,
+                                   VGboolean relative,
+                                   VGfloat *x, VGfloat *y)
+{
+   if (relative) {
+      if (x)
+         *x += ox;
+      if (y)
+         *y += oy;
+   }
+}
+
+static INLINE void close_polygon(struct polygon *current,
+                                 VGfloat sx, VGfloat sy,
+                                 VGfloat ox, VGfloat oy,
+                                 struct  matrix *matrix)
+{
+   if (!floatsEqual(sx, ox) ||
+       !floatsEqual(sy, oy)) {
+      VGfloat x0 = sx;
+      VGfloat y0 = sy;
+      matrix_map_point(matrix, x0, y0, &x0, &y0);
+      polygon_vertex_append(current, x0, y0);
+   }
+}
+
+static void convert_path(struct path *p,
+                          VGPathDatatype to,
+                          void *dst,
+                          VGint num_coords)
+{
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   VGubyte *common_data = (VGubyte *)dst;
+   VGint size_dst = size_for_datatype(to);
+   VGint i;
+
+   for (i = 0; i < num_coords; ++i) {
+      data_at(&coords, p, 0, 1, data);
+      vg_float_to_datatype(to, common_data, data, 1);
+      common_data += size_dst;
+   }
+}
+
+
+static void polygon_array_calculate_bounds( struct polygon_array *polyarray )
+{
+   struct array *polys = polyarray->array;
+   VGfloat min_x, max_x;
+   VGfloat min_y, max_y;
+   VGfloat bounds[4];
+   unsigned i;
+
+   assert(polys);
+   assert(polys->num_elements);
+   polygon_bounding_rect((((struct polygon**)polys->data)[0]), bounds);
+   min_x = bounds[0];
+   min_y = bounds[1];
+   max_x = bounds[0] + bounds[2];
+   max_y = bounds[1] + bounds[3];
+   for (i = 1; i < polys->num_elements; ++i) {
+      struct polygon *p = (((struct polygon**)polys->data)[i]);
+      polygon_bounding_rect(p, bounds);
+      min_x = MIN2(min_x, bounds[0]);
+      min_y = MIN2(min_y, bounds[1]);
+      max_x = MAX2(max_x, bounds[0] + bounds[2]);
+      max_y = MAX2(max_y, bounds[1] + bounds[3]);
+   }
+
+   polyarray->min_x = min_x;
+   polyarray->min_y = min_y;
+   polyarray->max_x = max_x;
+   polyarray->max_y = max_y;
+}
+
+
+static struct polygon_array * path_get_fill_polygons(struct path *p, struct matrix *matrix)
+{
+   VGint i;
+   struct polygon *current = 0;
+   VGfloat sx, sy, px, py, ox, oy;
+   VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   struct array *array;
+
+   if (p->fill_polys.polygon_array.array)
+   {
+      if (memcmp( &p->fill_polys.matrix,
+                  matrix,
+                  sizeof *matrix ) == 0 && p->dirty == VG_FALSE)
+      {
+         return &p->fill_polys.polygon_array;
+      }
+      else {
+         array_destroy( p->fill_polys.polygon_array.array );
+         p->fill_polys.polygon_array.array = NULL;
+      }
+   }
+
+   array = array_create(sizeof(struct array*));
+
+   sx = sy = px = py = ox = oy = 0.f;
+
+   current = polygon_create(32);
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment = ((VGubyte*)(p->segments->data))[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      VGboolean relative = SEGMENT_ABS_REL(segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH:
+         close_polygon(current, sx, sy, ox, oy, matrix);
+         ox = sx;
+         oy = sy;
+         break;
+      case VG_MOVE_TO:
+         if (current && polygon_vertex_count(current) > 0) {
+            /* add polygon */
+            close_polygon(current, sx, sy, ox, oy, matrix);
+            array_append_data(array, &current, 1);
+            current = polygon_create(32);
+         }
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         sx = x0;
+         sy = y0;
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = data[0];
+         y0 = oy;
+         map_if_relative(ox, oy, relative, &x0, 0);
+         ox = x0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = ox;
+         y0 = data[0];
+         map_if_relative(ox, oy, relative, 0, &y0);
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_CUBIC_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 6, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x2 = data[2];
+         y2 = data[3];
+         x3 = data[4];
+         y3 = data[5];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                       x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_QUAD_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                       x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SQUAD_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 2, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x3 = data[0];
+         y3 = data[1];
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                     x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x2 = data[0];
+         y2 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                              x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         VGfloat rh, rv, rot;
+         struct arc arc;
+
+         data_at(&coords, p, 0, 5, data);
+         x0  = ox;
+         y0  = oy;
+         rh  = data[0];
+         rv  = data[1];
+         rot = data[2];
+         x1  = data[3];
+         y1  = data[4];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+#if 0
+         debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n",
+                      x0, y0, x1, y1, rh, rv, rot);
+#endif
+         arc_init(&arc, command, x0, y0, x1, y1,
+                  rh, rv, rot);
+         arc_add_to_polygon(&arc, current,
+                            matrix);
+         ox = x1;
+         oy = y1;
+         px = x1;
+         py = y1;
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+   if (current) {
+      if (polygon_vertex_count(current) > 0) {
+         close_polygon(current, sx, sy, ox, oy, matrix);
+         array_append_data(array, &current, 1);
+      } else
+         polygon_destroy(current);
+   }
+
+   p->fill_polys.polygon_array.array = array;
+   p->fill_polys.matrix = *matrix;
+
+   polygon_array_calculate_bounds( &p->fill_polys.polygon_array );
+
+   p->dirty = VG_FALSE;
+
+   return &p->fill_polys.polygon_array;
+}
+
+VGbyte path_datatype_size(struct path *p)
+{
+   return size_for_datatype(p->datatype);
+}
+
+VGPathDatatype path_datatype(struct path *p)
+{
+   return p->datatype;
+}
+
+VGfloat path_scale(struct path *p)
+{
+   return p->scale;
+}
+
+VGfloat path_bias(struct path *p)
+{
+   return p->bias;
+}
+
+VGint path_num_coords(struct path *p)
+{
+   return num_elements_for_segments((VGubyte*)p->segments->data,
+                                    p->num_segments);
+}
+
+void path_modify_coords(struct path *p,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData)
+{
+   VGubyte *segments = (VGubyte*)(p->segments->data);
+   VGint count = num_elements_for_segments(&segments[startIndex], numSegments);
+   VGint start_cp = num_elements_for_segments(segments, startIndex);
+
+   array_change_data(p->control_points, pathData, start_cp, count);
+   coords_adjust_by_scale_bias(p,
+                               ((VGubyte*)p->control_points->data) +
+                               (startIndex * p->control_points->datatype_size),
+                               path_num_coords(p),
+                               p->scale, p->bias, p->datatype);
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+void path_for_each_segment(struct path *path,
+                           path_for_each_cb cb,
+                           void *user_data)
+{
+   VGint i;
+   struct path_for_each_data p;
+   VGfloat data[8];
+   void *coords = (VGfloat *)path->control_points->data;
+
+   p.coords = data;
+   p.sx = p.sy = p.px = p.py = p.ox = p.oy = 0.f;
+   p.user_data = user_data;
+
+   for (i = 0; i < path->num_segments; ++i) {
+      VGint command;
+      VGboolean relative;
+
+      p.segment = ((VGubyte*)(path->segments->data))[i];
+      command = SEGMENT_COMMAND(p.segment);
+      relative = SEGMENT_ABS_REL(p.segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH:
+         cb(path, &p);
+         break;
+      case VG_MOVE_TO:
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.sx = data[0];
+         p.sy = data[1];
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, path, 0, 1, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], 0);
+         p.segment = VG_LINE_TO;
+         data[1] = p.oy;
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, path, 0, 1, data);
+         map_if_relative(p.ox, p.oy, relative, 0, &data[0]);
+         p.segment = VG_LINE_TO;
+         data[1] = data[0];
+         data[0] = p.ox;
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_CUBIC_TO: {
+         data_at(&coords, path, 0, 6, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         map_if_relative(p.ox, p.oy, relative, &data[4], &data[5]);
+         cb(path, &p);
+         p.px = data[2];
+         p.py = data[3];
+         p.ox = data[4];
+         p.oy = data[5];
+      }
+         break;
+      case VG_QUAD_TO: {
+         data_at(&coords, path, 0, 4, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         cb(path, &p);
+         p.px = data[0];
+         p.py = data[1];
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SQUAD_TO: {
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.px = 2*p.ox-p.px;
+         p.py = 2*p.oy-p.py;
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         data_at(&coords, path, 0, 4, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         cb(path, &p);
+         p.px = data[0];
+         p.py = data[1];
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         data_at(&coords, path, 0, 5, data);
+         map_if_relative(p.ox, p.oy, relative, &data[3], &data[4]);
+#if 0
+         debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n",
+                      p.ox, p.oy, data[3], data[4], data[0], data[1], data[2]);
+#endif
+         cb(path, &p);
+         p.ox = data[3];
+         p.oy = data[4];
+         p.px = data[3];
+         p.py = data[4];
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+}
+
+struct transform_data {
+   struct array *segments;
+   struct array *coords;
+
+   struct matrix *matrix;
+
+   VGPathDatatype datatype;
+};
+
+static VGboolean transform_cb(struct path *p,
+                              struct path_for_each_data *pd)
+{
+   struct transform_data *td = (struct transform_data *)pd->user_data;
+   VGint num_coords = num_elements_for_segments(&pd->segment, 1);
+   VGubyte segment = SEGMENT_COMMAND(pd->segment);/* abs bit is 0 */
+   VGfloat data[8];
+   VGubyte common_data[sizeof(VGfloat)*8];
+
+   memcpy(data, pd->coords, sizeof(VGfloat) * num_coords);
+
+   switch(segment) {
+   case VG_CLOSE_PATH:
+      break;
+   case VG_MOVE_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_LINE_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_HLINE_TO:
+   case VG_VLINE_TO:
+      assert(0);
+      break;
+   case VG_QUAD_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      break;
+   case VG_CUBIC_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      matrix_map_point(td->matrix,
+                       data[4], data[5], &data[4], &data[5]);
+      break;
+   case VG_SQUAD_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_SCUBIC_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      struct arc arc;
+      struct path *path = path_create(td->datatype,
+                                      1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+      arc_init(&arc, segment,
+               pd->ox, pd->oy, data[3], data[4],
+               data[0], data[1], data[2]);
+
+      arc_to_path(&arc, path, td->matrix);
+
+      num_coords = path_num_coords(path);
+
+      array_append_data(td->segments, path->segments->data,
+                        path->num_segments);
+      array_append_data(td->coords, path->control_points->data,
+                        num_coords);
+      path_destroy(path);
+
+      return VG_TRUE;
+   }
+      break;
+   default:
+      break;
+   }
+
+   vg_float_to_datatype(td->datatype, common_data, data, num_coords);
+
+   array_append_data(td->segments, &pd->segment, 1);
+   array_append_data(td->coords, common_data, num_coords);
+   return VG_TRUE;
+}
+
+void path_transform(struct path *dst, struct path *src)
+{
+   struct transform_data data;
+   struct vg_context *ctx = dst->base.ctx;
+
+   data.segments =  dst->segments;
+   data.coords   =  dst->control_points;
+   data.matrix   = &ctx->state.vg.path_user_to_surface_matrix;
+   data.datatype = dst->datatype;
+
+   path_for_each_segment(src, transform_cb, (void*)&data);
+
+   dst->num_segments = dst->segments->num_elements;
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+}
+
+void path_append_path(struct path *dst,
+                      struct path *src)
+{
+   VGint num_coords = path_num_coords(src);
+   void *dst_data = malloc(size_for_datatype(dst->datatype) * num_coords);
+   array_append_data(dst->segments,
+                     src->segments->data,
+                     src->num_segments);
+   convert_path(src, dst->datatype,
+                dst_data, num_coords);
+   array_append_data(dst->control_points,
+                     dst_data,
+                     num_coords);
+   free(dst_data);
+
+   dst->num_segments += src->num_segments;
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+}
+
+static INLINE VGboolean is_segment_arc(VGubyte segment)
+{
+   VGubyte scommand = SEGMENT_COMMAND(segment);
+   return (scommand == VG_SCCWARC_TO ||
+           scommand == VG_SCWARC_TO ||
+           scommand == VG_LCCWARC_TO ||
+           scommand == VG_LCWARC_TO);
+}
+
+struct path_iter_data {
+   struct path *path;
+   VGubyte segment;
+   void *coords;
+   VGfloat px, py, ox, oy, sx, sy;
+};
+static INLINE VGubyte normalize_coords(struct path_iter_data *pd,
+                                       VGint *num_coords,
+                                       VGfloat *data)
+{
+   VGint command = SEGMENT_COMMAND(pd->segment);
+   VGboolean relative = SEGMENT_ABS_REL(pd->segment);
+
+   switch(command) {
+   case VG_CLOSE_PATH:
+      *num_coords = 0;
+      pd->ox = pd->sx;
+      pd->oy = pd->sy;
+      return VG_CLOSE_PATH;
+      break;
+   case VG_MOVE_TO:
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      pd->sx = data[0];
+      pd->sy = data[1];
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_MOVE_TO_ABS;
+      break;
+   case VG_LINE_TO:
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_HLINE_TO:
+      data_at(&pd->coords, pd->path, 0, 1, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], 0);
+      data[1] = pd->oy;
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_VLINE_TO:
+      data_at(&pd->coords, pd->path, 0, 1, data);
+      map_if_relative(pd->ox, pd->oy, relative, 0, &data[0]);
+      data[1] = data[0];
+      data[0] = pd->ox;
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_CUBIC_TO: {
+      data_at(&pd->coords, pd->path, 0, 6, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      map_if_relative(pd->ox, pd->oy, relative, &data[2], &data[3]);
+      map_if_relative(pd->ox, pd->oy, relative, &data[4], &data[5]);
+      pd->px = data[2];
+      pd->py = data[3];
+      pd->ox = data[4];
+      pd->oy = data[5];
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_QUAD_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 4, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = data[0];
+      y1 = data[1];
+      x3 = data[2];
+      y3 = data[3];
+      map_if_relative(pd->ox, pd->oy, relative, &x1, &y1);
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->px = x1;
+      pd->py = y1;
+      { /* form a cubic out of it */
+         x2 = (x3 + 2*x1) / 3.f;
+         y2 = (y3 + 2*y1) / 3.f;
+         x1 = (x0 + 2*x1) / 3.f;
+         y1 = (y0 + 2*y1) / 3.f;
+      }
+      pd->ox = x3;
+      pd->oy = y3;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SQUAD_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = 2 * pd->ox - pd->px;
+      y1 = 2 * pd->oy - pd->py;
+      x3 = data[0];
+      y3 = data[1];
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->px = x1;
+      pd->py = y1;
+      { /* form a cubic out of it */
+         x2 = (x3 + 2*x1) / 3.f;
+         y2 = (y3 + 2*y1) / 3.f;
+         x1 = (x0 + 2*x1) / 3.f;
+         y1 = (y0 + 2*y1) / 3.f;
+      }
+      pd->ox = x3;
+      pd->oy = y3;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SCUBIC_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 4, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = 2*pd->ox-pd->px;
+      y1 = 2*pd->oy-pd->py;
+      x2 = data[0];
+      y2 = data[1];
+      x3 = data[2];
+      y3 = data[3];
+      map_if_relative(pd->ox, pd->oy, relative, &x2, &y2);
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->ox = x3;
+      pd->oy = y3;
+      pd->px = x2;
+      pd->py = y2;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      data_at(&pd->coords, pd->path, 0, 5, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[3], &data[4]);
+      pd->ox = data[3];
+      pd->oy = data[4];
+      pd->px = data[3];
+      pd->py = data[4];
+      *num_coords = 5;
+      return command | VG_ABSOLUTE;
+   }
+      break;
+   default:
+      abort();
+      assert(!"Unknown segment!");
+   }
+}
+
+static void linearly_interpolate(VGfloat *result,
+                                 const VGfloat *start,
+                                 const VGfloat *end,
+                                 VGfloat amount,
+                                 VGint number)
+{
+   VGint i;
+   for (i = 0; i < number; ++i) {
+      result[i] = start[i] + (end[i] - start[i]) * amount;
+   }
+}
+
+VGboolean path_interpolate(struct path *dst,
+                           struct path *start, struct path *end,
+                           VGfloat amount)
+{
+   /* temporary path that we can discard if it will turn
+    * out that start is not compatible with end */
+   struct path *res_path = path_create(dst->datatype,
+                                       1.0, 0.0,
+                                       0, 0, dst->caps);
+   VGint i;
+   VGfloat start_coords[8];
+   VGfloat end_coords[8];
+   VGfloat results[8];
+   VGubyte common_data[sizeof(VGfloat)*8];
+   struct path_iter_data start_iter, end_iter;
+
+   memset(&start_iter, 0, sizeof(struct path_iter_data));
+   memset(&end_iter, 0, sizeof(struct path_iter_data));
+
+   start_iter.path = start;
+   start_iter.coords = start->control_points->data;
+   end_iter.path = end;
+   end_iter.coords = end->control_points->data;
+
+   for (i = 0; i < start->num_segments; ++i) {
+      VGubyte segment;
+      VGubyte ssegment, esegment;
+      VGint snum_coords, enum_coords;
+      start_iter.segment = ((VGubyte*)(start->segments->data))[i];
+      end_iter.segment = ((VGubyte*)(end->segments->data))[i];
+
+      ssegment = normalize_coords(&start_iter, &snum_coords,
+                                  start_coords);
+      esegment = normalize_coords(&end_iter, &enum_coords,
+                                  end_coords);
+
+      if (is_segment_arc(ssegment)) {
+         if (!is_segment_arc(esegment)) {
+            path_destroy(res_path);
+            return VG_FALSE;
+         }
+         if (amount > 0.5)
+            segment = esegment;
+         else
+            segment = ssegment;
+      } else if (is_segment_arc(esegment)) {
+         path_destroy(res_path);
+         return VG_FALSE;
+      }
+      else if (ssegment != esegment) {
+         path_destroy(res_path);
+         return VG_FALSE;
+      }
+      else
+         segment = ssegment;
+
+      linearly_interpolate(results, start_coords, end_coords,
+                           amount, snum_coords);
+      vg_float_to_datatype(dst->datatype, common_data, results, snum_coords);
+      path_append_data(res_path, 1, &segment, common_data);
+   }
+
+   path_append_path(dst, res_path);
+   path_destroy(res_path);
+
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+
+   return VG_TRUE;
+}
+
+void path_clear(struct path *p, VGbitfield capabilities)
+{
+   path_set_capabilities(p, capabilities);
+   array_destroy(p->segments);
+   array_destroy(p->control_points);
+   p->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8));
+   p->control_points = array_create(size_for_datatype(p->datatype));
+   p->num_segments = 0;
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+struct path * path_create_stroke(struct path *p,
+                                 struct matrix *matrix)
+{
+   VGint i;
+   VGfloat sx, sy, px, py, ox, oy;
+   VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   int dashed = (p->base.ctx->state.vg.stroke.dash_pattern_num ? 1 : 0);
+   struct dash_stroker stroker;
+   struct vg_state *vg_state = &p->base.ctx->state.vg;
+
+   if (p->stroked.path)
+   {
+      /* ### compare the dash patterns to see if we can cache them.
+       *     for now we simply always bail out if the path is dashed.
+       */
+      if (memcmp( &p->stroked.matrix,
+                  matrix,
+                  sizeof *matrix ) == 0 &&
+          !dashed && !p->dirty_stroke &&
+          floatsEqual(p->stroked.stroke_width, vg_state->stroke.line_width.f) &&
+          floatsEqual(p->stroked.miter_limit, vg_state->stroke.miter_limit.f) &&
+          p->stroked.cap_style == vg_state->stroke.cap_style &&
+          p->stroked.join_style == vg_state->stroke.join_style)
+      {
+         return p->stroked.path;
+      }
+      else {
+         path_destroy( p->stroked.path );
+         p->stroked.path = NULL;
+      }
+   }
+
+
+   sx = sy = px = py = ox = oy = 0.f;
+
+   if (dashed)
+      dash_stroker_init((struct stroker *)&stroker, vg_state);
+   else
+      stroker_init((struct stroker *)&stroker, vg_state);
+
+   stroker_begin((struct stroker *)&stroker);
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment = ((VGubyte*)(p->segments->data))[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      VGboolean relative = SEGMENT_ABS_REL(segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH: {
+            VGfloat x0 = sx;
+            VGfloat y0 = sy;
+            matrix_map_point(matrix, x0, y0, &x0, &y0);
+            stroker_line_to((struct stroker *)&stroker, x0, y0);
+      }
+         break;
+      case VG_MOVE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         sx = x0;
+         sy = y0;
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_move_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = data[0];
+         y0 = oy;
+         map_if_relative(ox, oy, relative, &x0, 0);
+         ox = x0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = ox;
+         y0 = data[0];
+         map_if_relative(ox, oy, relative, 0, &y0);
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_CUBIC_TO: {
+         data_at(&coords, p, 0, 6, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x2 = data[2];
+         y2 = data[3];
+         x3 = data[4];
+         y3 = data[5];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_QUAD_TO: {
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SQUAD_TO: {
+         data_at(&coords, p, 0, 2, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x3 = data[0];
+         y3 = data[1];
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x2 = data[0];
+         y2 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         VGfloat rh, rv, rot;
+         struct arc arc;
+
+         data_at(&coords, p, 0, 5, data);
+         x0  = ox;
+         y0  = oy;
+         rh  = data[0];
+         rv  = data[1];
+         rot = data[2];
+         x1  = data[3];
+         y1  = data[4];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x1;
+            oy = y1;
+            matrix_map_point(matrix, x1, y1, &x1, &y1);
+            stroker_line_to((struct stroker *)&stroker, x1, y1);
+            continue;
+         }
+         arc_init(&arc, command, x0, y0, x1, y1,
+                  rh, rv, rot);
+         arc_stroke_cb(&arc, (struct stroker *)&stroker,
+                       matrix);
+         ox = x1;
+         oy = y1;
+         px = x1;
+         py = y1;
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+
+   stroker_end((struct stroker *)&stroker);
+
+   if (dashed)
+      dash_stroker_cleanup((struct dash_stroker *)&stroker);
+   else
+      stroker_cleanup((struct stroker *)&stroker);
+
+   p->stroked.path = stroker.base.path;
+   p->stroked.matrix = *matrix;
+   p->dirty_stroke = VG_FALSE;
+   p->stroked.stroke_width = vg_state->stroke.line_width.f;
+   p->stroked.miter_limit = vg_state->stroke.miter_limit.f;
+   p->stroked.cap_style = vg_state->stroke.cap_style;
+   p->stroked.join_style = vg_state->stroke.join_style;
+
+   return stroker.base.path;
+}
+
+void path_render(struct path *p, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+
+   vg_validate_state(ctx);
+
+   shader_set_drawing_image(ctx->shader, VG_FALSE);
+   shader_set_image(ctx->shader, 0);
+#if 0
+   fprintf(stderr, "Matrix(11=%f 12=%f 13=%f 21=%f 22=%f 23=%f 31=%f 32=%f 33=%f)\n",
+           mat->m[0], mat->m[1], mat->m[2],
+           mat->m[3], mat->m[4], mat->m[5],
+           mat->m[6], mat->m[7], mat->m[8]);
+#endif
+   if (paintModes & VG_FILL_PATH) {
+      /* First the fill */
+      shader_set_paint(ctx->shader, ctx->state.vg.fill_paint);
+      shader_bind(ctx->shader);
+      path_fill(p, mat);
+   }
+
+   if (paintModes & VG_STROKE_PATH){
+      /* 8.7.5: "line width less than or equal to 0 prevents stroking from
+       *  taking place."*/
+      if (ctx->state.vg.stroke.line_width.f <= 0)
+         return;
+      shader_set_paint(ctx->shader, ctx->state.vg.stroke_paint);
+      shader_bind(ctx->shader);
+      path_stroke(p);
+   }
+}
+
+void path_fill(struct path *p, struct matrix *mat)
+{
+   struct vg_context *ctx = vg_current_context();
+   {
+      struct polygon_array *polygon_array = path_get_fill_polygons(p, mat);
+      struct array *polys = polygon_array->array;
+
+      if (!polygon_array || !polys || !polys->num_elements) {
+         return;
+      }
+      polygon_array_fill(polygon_array, ctx);
+   }
+}
+
+void path_stroke(struct path *p)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+   VGFillRule old_fill = ctx->state.vg.fill_rule;
+   struct matrix identity;
+   struct path *stroke;
+
+   matrix_load_identity(&identity);
+   stroke = path_create_stroke(p, &identity);
+   if (stroke && !path_is_empty(stroke)) {
+      ctx->state.vg.fill_rule = VG_NON_ZERO;
+
+      path_fill(stroke, mat);
+
+      ctx->state.vg.fill_rule = old_fill;
+   }
+}
+
+void path_move_to(struct path *p, float x, float y)
+{
+   VGubyte segment = VG_MOVE_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 2];
+   VGfloat data[2] = {x, y};
+
+   vg_float_to_datatype(p->datatype, common_data, data, 2);
+   path_append_data(p, 1, &segment, common_data);
+}
+
+void path_line_to(struct path *p, float x, float y)
+{
+   VGubyte segment = VG_LINE_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 2];
+   VGfloat data[2] = {x, y};
+
+   vg_float_to_datatype(p->datatype, common_data, data, 2);
+
+   path_append_data(p, 1, &segment, common_data);
+}
+
+void path_cubic_to(struct path *p, float px1, float py1,
+                   float px2, float py2,
+                   float x, float y)
+{
+   VGubyte segment = VG_CUBIC_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 6];
+   VGfloat data[6];
+
+   data[0] = px1; data[1] = py1;
+   data[2] = px2; data[3] = py2;
+   data[4] = x;   data[5] = y;
+
+   vg_float_to_datatype(p->datatype, common_data, data, 6);
+
+   path_append_data(p, 1, &segment, common_data);
+}
+
+static INLINE void line_bounds(VGfloat *line /*x1,y1,x2,y2*/,
+                               VGfloat *bounds)
+{
+   bounds[0] = MIN2(line[0], line[2]);
+   bounds[1] = MIN2(line[1], line[3]);
+   bounds[2] = MAX2(line[0], line[2]) - bounds[0];
+   bounds[3] = MAX2(line[1], line[3]) - bounds[1];
+}
+
+static INLINE void unite_bounds(VGfloat *bounds,
+                                VGfloat *el)
+{
+   VGfloat cx1, cy1, cx2, cy2;
+   VGfloat nx1, ny1, nx2, ny2;
+
+   cx1 = bounds[0];
+   cy1 = bounds[1];
+   cx2 = bounds[0] + bounds[2];
+   cy2 = bounds[1] + bounds[3];
+
+   nx1 = el[0];
+   ny1 = el[1];
+   nx2 = el[0] + el[2];
+   ny2 = el[1] + el[3];
+
+   bounds[0] = MIN2(cx1, nx1);
+   bounds[1] = MIN2(cy1, ny1);
+   bounds[2] = MAX2(cx2, nx2) - bounds[0];
+   bounds[3] = MAX2(cy2, ny2) - bounds[1];
+}
+
+static INLINE void set_bounds(VGfloat *bounds,
+                              VGfloat *element_bounds,
+                              VGboolean *initialized)
+{
+   if (!(*initialized)) {
+      memcpy(bounds, element_bounds, 4 * sizeof(VGfloat));
+      *initialized = VG_TRUE;
+   } else
+      unite_bounds(bounds, element_bounds);
+}
+
+void path_bounding_rect(struct path *p, float *x, float *y,
+                        float *w, float *h)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter;
+   VGint num_coords;
+   VGfloat bounds[4];
+   VGfloat element_bounds[4];
+   VGfloat ox, oy;
+   VGboolean bounds_inited = VG_FALSE;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+   memset(&bounds, 0, sizeof(bounds));
+
+   if (!p->num_segments) {
+      bounds[2] = -1;
+      bounds[3] = -1;
+   }
+
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment;
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+
+      ox = iter.ox;
+      oy = iter.oy;
+
+      segment = normalize_coords(&iter, &num_coords, coords);
+
+      switch(segment) {
+      case VG_CLOSE_PATH:
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {ox, oy, coords[0], coords[1]};
+         line_bounds(line, element_bounds);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, ox, oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         bezier_exact_bounds(&bezier, element_bounds);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, segment,
+                  ox, oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         path_bounding_rect(path, element_bounds + 0, element_bounds + 1,
+                            element_bounds + 2, element_bounds + 3);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   *x = bounds[0];
+   *y = bounds[1];
+   *w = bounds[2];
+   *h = bounds[3];
+}
+
+float path_length(struct path *p, int start_segment, int num_segments)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter;
+   VGint num_coords;
+   VGfloat length = 0;
+   VGfloat ox, oy;
+   VGboolean in_range = VG_FALSE;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+
+   for (i = 0; i < (start_segment + num_segments); ++i) {
+      VGubyte segment;
+
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+
+      ox = iter.ox;
+      oy = iter.oy;
+
+      segment = normalize_coords(&iter, &num_coords, coords);
+
+      in_range = (i >= start_segment) && i <= (start_segment + num_segments);
+      if (!in_range)
+         continue;
+
+      switch(segment) {
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_CLOSE_PATH: {
+         VGfloat line[4] = {ox, oy, iter.sx, iter.sy};
+         length += line_lengthv(line);
+      }
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {ox, oy, coords[0], coords[1]};
+         length += line_lengthv(line);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, ox, oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         length += bezier_length(&bezier, BEZIER_DEFAULT_ERROR);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, segment,
+                  ox, oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         length += path_length(path, 0, path_num_segments(path));
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   return length;
+}
+
+static INLINE VGboolean point_on_current_segment(VGfloat distance,
+                                                 VGfloat length,
+                                                 VGfloat segment_length)
+{
+   return
+      (((floatIsZero(distance) || distance < 0) && floatIsZero(length)) ||
+       ((distance > length || floatsEqual(distance, length)) &&
+        (floatsEqual(distance, length + segment_length) ||
+         distance < (length + segment_length))));
+}
+
+static VGboolean path_point_segment(struct path_iter_data iter,
+                                    struct path_iter_data prev_iter,
+                                    VGfloat coords[8],
+                                    VGfloat distance,
+                                    VGfloat length, VGfloat *current_length,
+                                    VGfloat *point, VGfloat *normal)
+{
+   switch (iter.segment) {
+   case VG_MOVE_TO_ABS:
+      break;
+   case VG_CLOSE_PATH: {
+      VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy};
+      VGboolean on_current_segment = VG_FALSE;
+      *current_length = line_lengthv(line);
+      on_current_segment = point_on_current_segment(distance,
+                                                    length,
+                                                    *current_length);
+      if (on_current_segment) {
+         VGfloat at = (distance - length) / line_lengthv(line);
+         line_normal_vector(line, normal);
+         line_point_at(line, at, point);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_LINE_TO_ABS: {
+      VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]};
+      VGboolean on_current_segment = VG_FALSE;
+      *current_length = line_lengthv(line);
+      on_current_segment = point_on_current_segment(distance,
+                                                    length,
+                                                    *current_length);
+      if (on_current_segment) {
+         VGfloat at = (distance - length) / line_lengthv(line);
+         line_normal_vector(line, normal);
+         line_point_at(line, at, point);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_CUBIC_TO_ABS: {
+      struct bezier bezier;
+      bezier_init(&bezier, prev_iter.ox, prev_iter.oy,
+                  coords[0], coords[1],
+                  coords[2], coords[3],
+                  coords[4], coords[5]);
+      *current_length = bezier_length(&bezier, BEZIER_DEFAULT_ERROR);
+      if (point_on_current_segment(distance, length, *current_length)) {
+         bezier_point_at_length(&bezier, distance - length,
+                                point, normal);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      struct arc arc;
+      struct matrix identity;
+      struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                      1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+      matrix_load_identity(&identity);
+      arc_init(&arc, iter.segment,
+               prev_iter.ox, prev_iter.oy, coords[3], coords[4],
+               coords[0], coords[1], coords[2]);
+
+      arc_to_path(&arc, path, &identity);
+
+      *current_length = path_length(path, 0, path_num_segments(path));
+      if (point_on_current_segment(distance, length, *current_length)) {
+         path_point(path, 0, path_num_segments(path),
+                    distance - length, point, normal);
+         return VG_TRUE;
+      }
+   }
+      break;
+   default:
+      assert(0);
+   }
+   return VG_FALSE;
+}
+
+void path_point(struct path *p, VGint start_segment, VGint num_segments,
+                VGfloat distance, VGfloat *point, VGfloat *normal)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter, prev_iter;
+   VGint num_coords;
+   VGfloat length = 0;
+   VGfloat current_length = 0;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+   memset(&prev_iter, 0, sizeof(struct path_iter_data));
+
+   point[0] = 0;
+   point[1] = 0;
+
+   normal[0] = 0;
+   normal[1] = -1;
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+   if (distance < 0)
+      distance = 0;
+
+   for (i = 0; i < (start_segment + num_segments); ++i) {
+      VGboolean outside_range = (i < start_segment ||
+                                 i >= (start_segment + num_segments));
+
+      prev_iter = iter;
+
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+      iter.segment = normalize_coords(&iter, &num_coords, coords);
+
+      if (outside_range)
+         continue;
+
+      if (path_point_segment(iter, prev_iter, coords,
+                             distance, length, &current_length,
+                             point, normal))
+         return;
+
+      length += current_length;
+   }
+
+   /*
+    *OpenVG 1.0 - 8.6.11 vgPointAlongPath
+    *
+    * If distance is greater than or equal to the path length
+    *(i.e., the value returned by vgPathLength when called with the same
+    *startSegment and numSegments parameters), the visual ending point of
+    *the path is used.
+    */
+   {
+      switch (iter.segment) {
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_CLOSE_PATH: {
+         VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy};
+         line_normal_vector(line, normal);
+         line_point_at(line, 1.f, point);
+      }
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]};
+         line_normal_vector(line, normal);
+         line_point_at(line, 1.f, point);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, prev_iter.ox, prev_iter.oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         bezier_point_at_t(&bezier, 1.f, point, normal);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, iter.segment,
+                  prev_iter.ox, prev_iter.oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         path_point(path, 0, path_num_segments(path),
+                    /* to make sure we're bigger than len * 2 it */
+                    2 * path_length(path, 0, path_num_segments(path)),
+                    point, normal);
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+}
+
+VGboolean path_is_empty(struct path *p)
+{
+   return p->segments->num_elements == 0;
+}
diff --git a/src/gallium/state_trackers/vega/path.h b/src/gallium/state_trackers/vega/path.h
new file mode 100644
index 0000000000..e34538b736
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path.h
@@ -0,0 +1,126 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef _PATH_H
+#define _PATH_H
+
+#include "VG/openvg.h"
+
+struct path;
+struct polygon;
+struct matrix;
+
+enum fill_rule {
+   ODD_EVEN_FILL,
+   WINDING_FILL
+};
+
+
+struct path_for_each_data {
+   VGubyte segment;
+   /* all coords are absolute, even if segment is relative */
+   const VGfloat *coords;
+   VGfloat sx, sy, ox, oy, px, py;
+   void *user_data;
+};
+
+typedef VGboolean (*path_for_each_cb)(struct path *p,
+                                      struct path_for_each_data *data);
+
+
+struct path *path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias,
+                         VGint segmentCapacityHint,
+                         VGint coordCapacityHint,
+                         VGbitfield capabilities);
+void path_destroy(struct path *p);
+
+VGbitfield path_capabilities(struct path *p);
+void path_set_capabilities(struct path *p, VGbitfield bf);
+
+void path_append_data(struct path *p,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData);
+
+void path_append_path(struct path *dst,
+                      struct path *src);
+
+VGint path_num_segments(struct path *p);
+
+void path_bounding_rect(struct path *p, float *x, float *y,
+                        float *w, float *h);
+float path_length(struct path *p, int start_segment, int num_segments);
+
+void path_set_fill_rule(enum fill_rule fill);
+enum fill_rule path_fill_rule(enum fill_rule fill);
+
+VGboolean path_is_empty(struct path *p);
+
+VGbyte path_datatype_size(struct path *p);
+
+VGPathDatatype path_datatype(struct path *p);
+VGfloat path_scale(struct path *p);
+VGfloat path_bias(struct path *p);
+VGint path_num_coords(struct path *p);
+
+void path_modify_coords(struct path *p,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData);
+
+struct path *path_create_stroke(struct path *p,
+                                struct matrix *m);
+
+void path_for_each_segment(struct path *path,
+                           path_for_each_cb cb,
+                           void *user_data);
+
+void path_transform(struct path *dst, struct path *src);
+VGboolean path_interpolate(struct path *dst,
+                           struct path *start, struct path *end,
+                           VGfloat amount);
+
+void path_clear(struct path *p, VGbitfield capabilities);
+void path_render(struct path *p, VGbitfield paintModes);
+void path_fill(struct path *p, struct matrix *mat);
+void path_stroke(struct path *p);
+
+void path_move_to(struct path *p, float x, float y);
+void path_line_to(struct path *p, float x, float y);
+void path_cubic_to(struct path *p, float px1, float py1,
+                   float px2, float py2,
+                   float x, float y);
+
+void path_point(struct path *p, VGint startSegment, VGint numSegments,
+                VGfloat distance, VGfloat *point, VGfloat *normal);
+
+
+
+void vg_float_to_datatype(VGPathDatatype datatype,
+                          VGubyte *common_data,
+                          const VGfloat *data,
+                          VGint num_coords);
+#endif
diff --git a/src/gallium/state_trackers/vega/path_utils.h b/src/gallium/state_trackers/vega/path_utils.h
new file mode 100644
index 0000000000..c2b3221dc5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path_utils.h
@@ -0,0 +1,109 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef PATH_UTILS_H
+#define PATH_UTILS_H
+
+#include "VG/openvg.h"
+
+#define SEGMENT_COMMAND(command) /* Extract segment type */     \
+    ((command) & 0x1e)
+#define SEGMENT_ABS_REL(command) /* Extract absolute/relative bit */ \
+    ((command) & 0x1)
+
+static INLINE VGint size_for_datatype(VGPathDatatype datatype)
+{
+   switch(datatype) {
+   case VG_PATH_DATATYPE_S_8:
+      return 1;
+   case VG_PATH_DATATYPE_S_16:
+      return 2;
+   case VG_PATH_DATATYPE_S_32:
+      return 4;
+   case VG_PATH_DATATYPE_F:
+      return 4;
+   default:
+      assert(!"unknown datatype");
+   }
+   return 0;
+}
+
+static INLINE VGint num_elements_for_segments(const VGubyte *segments,
+                                              VGint num_segments)
+{
+   VGint i;
+   VGint count = 0;
+
+   for (i = 0; i < num_segments; ++i) {
+      VGubyte segment = segments[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      switch(command) {
+      case VG_CLOSE_PATH:
+         break;
+      case VG_MOVE_TO:
+         count += 2;
+         break;
+      case VG_LINE_TO:
+         count += 2;
+         break;
+      case VG_HLINE_TO:
+         count += 1;
+         break;
+      case VG_VLINE_TO:
+         count += 1;
+         break;
+      case VG_QUAD_TO:
+         count += 4;
+         break;
+      case VG_CUBIC_TO:
+         count += 6;
+         break;
+      case VG_SQUAD_TO:
+         count += 2;
+         break;
+      case VG_SCUBIC_TO:
+         count += 4;
+         break;
+      case VG_SCCWARC_TO:
+         count += 5;
+         break;
+      case VG_SCWARC_TO:
+         count += 5;
+         break;
+      case VG_LCCWARC_TO:
+         count += 5;
+         break;
+      case VG_LCWARC_TO:
+         count += 5;
+         break;
+      default:
+         assert(!"Unknown segment!");
+      }
+   }
+   return count;
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/polygon.c b/src/gallium/state_trackers/vega/polygon.c
new file mode 100644
index 0000000000..b6d282d803
--- /dev/null
+++ b/src/gallium/state_trackers/vega/polygon.c
@@ -0,0 +1,550 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "polygon.h"
+
+#include "matrix.h" /*for floatsEqual*/
+#include "vg_context.h"
+#include "vg_state.h"
+#include "paint.h"
+#include "renderer.h"
+#include "util_array.h"
+#include "VG/openvg.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+
+#include "util/u_draw_quad.h"
+#include "util/u_math.h"
+
+#include <string.h>
+#include <stdlib.h>
+
+#define DEBUG_POLYGON 0
+
+#define COMPONENTS 2
+
+struct polygon
+{
+   VGfloat *data;
+   VGint    size;
+
+   VGint    num_verts;
+
+   VGboolean dirty;
+   struct pipe_buffer *vbuf;
+   struct pipe_screen *screen;
+};
+
+static float *ptr_to_vertex(float *data, int idx)
+{
+   return data + (idx * COMPONENTS);
+}
+
+#if 0
+static void polygon_print(struct polygon *poly)
+{
+   int i;
+   float *vert;
+   debug_printf("Polygon %p, size = %d\n", poly, poly->num_verts);
+   for (i = 0; i < poly->num_verts; ++i) {
+      vert = ptr_to_vertex(poly->data, i);
+      debug_printf("%f, %f,  ", vert[0], vert[1]);
+   }
+   debug_printf("\nend\n");
+}
+#endif
+
+
+struct polygon * polygon_create(int size)
+{
+   struct polygon *poly = (struct polygon*)malloc(sizeof(struct polygon));
+
+   poly->data = malloc(sizeof(float) * COMPONENTS * size);
+   poly->size = size;
+   poly->num_verts = 0;
+   poly->dirty = VG_TRUE;
+   poly->vbuf = NULL;
+
+   return poly;
+}
+
+struct polygon * polygon_create_from_data(float *data, int size)
+{
+   struct polygon *poly = polygon_create(size);
+
+   memcpy(poly->data, data, sizeof(float) * COMPONENTS * size);
+   poly->num_verts = size;
+   poly->dirty = VG_TRUE;
+   poly->vbuf = NULL;
+
+   return poly;
+}
+
+void polygon_destroy(struct polygon *poly)
+{
+   if (poly->vbuf)
+      pipe_buffer_reference(&poly->vbuf, NULL);
+
+   free(poly->data);
+   free(poly);
+}
+
+void polygon_resize(struct polygon *poly, int new_size)
+{
+   float *data = (float*)malloc(sizeof(float) * COMPONENTS * new_size);
+   int size = MIN2(sizeof(float) * COMPONENTS * new_size,
+                   sizeof(float) * COMPONENTS * poly->size);
+   memcpy(data, poly->data, size);
+   free(poly->data);
+   poly->data = data;
+   poly->size = new_size;
+   poly->dirty = VG_TRUE;
+}
+
+int polygon_size(struct polygon *poly)
+{
+   return poly->size;
+}
+
+int polygon_vertex_count(struct polygon *poly)
+{
+   return poly->num_verts;
+}
+
+float * polygon_data(struct polygon *poly)
+{
+   return poly->data;
+}
+
+void polygon_vertex_append(struct polygon *p,
+                           float x, float y)
+{
+   float *vert;
+#if DEBUG_POLYGON
+   debug_printf("Append vertex [%f, %f]\n", x, y);
+#endif
+   if (p->num_verts >= p->size) {
+      polygon_resize(p, p->size * 2);
+   }
+
+   vert = ptr_to_vertex(p->data, p->num_verts);
+   vert[0] = x;
+   vert[1] = y;
+   ++p->num_verts;
+   p->dirty = VG_TRUE;
+}
+
+void polygon_set_vertex(struct polygon *p, int idx,
+                        float x, float y)
+{
+   float *vert;
+   if (idx >= p->num_verts) {
+      /*fixme: error reporting*/
+      abort();
+      return;
+   }
+
+   vert = ptr_to_vertex(p->data, idx);
+   vert[0] = x;
+   vert[1] = y;
+   p->dirty = VG_TRUE;
+}
+
+void polygon_vertex(struct polygon *p, int idx,
+                    float *vertex)
+{
+   float *vert;
+   if (idx >= p->num_verts) {
+      /*fixme: error reporting*/
+      abort();
+      return;
+   }
+
+   vert = ptr_to_vertex(p->data, idx);
+   vertex[0] = vert[0];
+   vertex[1] = vert[1];
+}
+
+void polygon_bounding_rect(struct polygon *p,
+                           float *rect)
+{
+   int i;
+   float minx, miny, maxx, maxy;
+   float *vert = ptr_to_vertex(p->data, 0);
+   minx = vert[0];
+   maxx = vert[0];
+   miny = vert[1];
+   maxy = vert[1];
+
+   for (i = 1; i < p->num_verts; ++i) {
+      vert = ptr_to_vertex(p->data, i);
+      minx = MIN2(vert[0], minx);
+      miny = MIN2(vert[1], miny);
+
+      maxx = MAX2(vert[0], maxx);
+      maxy = MAX2(vert[1], maxy);
+   }
+
+   rect[0] = minx;
+   rect[1] = miny;
+   rect[2] = maxx - minx;
+   rect[3] = maxy - miny;
+}
+
+int polygon_contains_point(struct polygon *p,
+                           float x, float y)
+{
+   return 0;
+}
+
+void polygon_append_polygon(struct polygon *dst,
+                            struct polygon *src)
+{
+   if (dst->num_verts + src->num_verts >= dst->size) {
+      polygon_resize(dst, dst->num_verts + src->num_verts * 1.5);
+   }
+   memcpy(ptr_to_vertex(dst->data, dst->num_verts),
+          src->data, src->num_verts * COMPONENTS * sizeof(VGfloat));
+   dst->num_verts += src->num_verts;
+}
+
+VGboolean polygon_is_closed(struct polygon *p)
+{
+   VGfloat start[2], end[2];
+
+   polygon_vertex(p, 0, start);
+   polygon_vertex(p, p->num_verts - 1, end);
+
+   return floatsEqual(start[0], end[0]) && floatsEqual(start[1], end[1]);
+}
+
+static void set_blend_for_fill(struct pipe_blend_state *blend)
+{
+   memset(blend, 0, sizeof(struct pipe_blend_state));
+   blend->colormask = 0; /*disable colorwrites*/
+
+   blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend->rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+   blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+}
+
+static void draw_polygon(struct vg_context *ctx,
+                         struct polygon *poly)
+{
+   int vert_size;
+   struct pipe_context *pipe;
+   struct pipe_vertex_buffer vbuffer;
+   struct pipe_vertex_element velement;
+
+   vert_size = poly->num_verts * COMPONENTS * sizeof(float);
+
+   /*polygon_print(poly);*/
+
+   pipe = ctx->pipe;
+
+   if (poly->vbuf == NULL || poly->dirty) {
+      if (poly->vbuf) {
+         pipe_buffer_reference(&poly->vbuf,
+                               NULL);
+      }
+      poly->screen = pipe->screen;
+      poly->vbuf= pipe_user_buffer_create(poly->screen,
+                                          poly->data,
+                                          vert_size);
+      poly->dirty = VG_FALSE;
+   }
+
+
+   /* tell pipe about the vertex buffer */
+   memset(&vbuffer, 0, sizeof(vbuffer));
+   vbuffer.buffer = poly->vbuf;
+   vbuffer.stride = COMPONENTS * sizeof(float);  /* vertex size */
+   vbuffer.buffer_offset = 0;
+   vbuffer.max_index = poly->num_verts - 1;
+   pipe->set_vertex_buffers(pipe, 1, &vbuffer);
+
+   /* tell pipe about the vertex attributes */
+   velement.src_offset = 0;
+   velement.vertex_buffer_index = 0;
+   velement.src_format = PIPE_FORMAT_R32G32_FLOAT;
+   velement.nr_components = COMPONENTS;
+   pipe->set_vertex_elements(pipe, 1, &velement);
+
+   /* draw */
+   pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 
+                     0, poly->num_verts);
+}
+
+void polygon_fill(struct polygon *poly, struct vg_context *ctx)
+{
+   struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_blend_state blend;
+   VGfloat bounds[4];
+   VGfloat min_x, min_y, max_x, max_y;
+   assert(poly);
+   polygon_bounding_rect(poly, bounds);
+   min_x = bounds[0];
+   min_y = bounds[1];
+   max_x = bounds[0] + bounds[2];
+   max_y = bounds[1] + bounds[3];
+
+#if DEBUG_POLYGON
+   debug_printf("Poly bounds are [%f, %f], [%f, %f]\n",
+                min_x, min_y, max_x, max_y);
+#endif
+
+   set_blend_for_fill(&blend);
+
+   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   cso_save_blend(ctx->cso_context);
+   cso_save_depth_stencil_alpha(ctx->cso_context);
+
+   dsa.stencil[0].enabled = 1;
+   if (ctx->state.vg.fill_rule == VG_EVEN_ODD) {
+      dsa.stencil[0].writemask = 1;
+      dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].ref_value = 0;
+      dsa.stencil[0].valuemask = ~0;
+
+      cso_set_blend(ctx->cso_context, &blend);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+      draw_polygon(ctx, poly);
+   } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) {
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+         /* front */
+         dsa.stencil[0].writemask = ~0;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         /* back */
+         dsa.stencil[1].enabled = 1;
+         dsa.stencil[1].writemask = ~0;
+         dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         dsa.stencil[1].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[1].ref_value = 0;
+         dsa.stencil[1].valuemask = ~0;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         draw_polygon(ctx, poly);
+      } else {
+         struct pipe_rasterizer_state raster;
+
+         memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state));
+
+         cso_save_rasterizer(ctx->cso_context);
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         draw_polygon(ctx, poly);
+
+         raster.cull_mode = raster.front_winding;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         draw_polygon(ctx, poly);
+
+         cso_restore_rasterizer(ctx->cso_context);
+      }
+   }
+
+   /* restore color writes */
+   cso_restore_blend(ctx->cso_context);
+   /* setup stencil ops */
+   dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL;
+   dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].ref_value = 0;
+   dsa.stencil[0].valuemask = dsa.stencil[0].writemask;
+   dsa.stencil[1].enabled = 0;
+   memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth,
+          sizeof(struct pipe_depth_state));
+   cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_quad(ctx->renderer, min_x, min_y,
+                      max_x, max_y, 0.0f/*depth should be disabled*/);
+
+   cso_restore_depth_stencil_alpha(ctx->cso_context);
+}
+
+void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx)
+{
+   struct array *polys = polyarray->array;
+   struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_blend_state blend;
+   VGfloat min_x = polyarray->min_x;
+   VGfloat min_y = polyarray->min_y;
+   VGfloat max_x = polyarray->max_x;
+   VGfloat max_y = polyarray->max_y;
+   VGint i;
+
+
+#if DEBUG_POLYGON
+   debug_printf("%s: Poly bounds are [%f, %f], [%f, %f]\n",
+                __FUNCTION__,
+                min_x, min_y, max_x, max_y);
+#endif
+
+   set_blend_for_fill(&blend);
+
+   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   cso_save_blend(ctx->cso_context);
+   cso_save_depth_stencil_alpha(ctx->cso_context);
+
+   dsa.stencil[0].enabled = 1;
+   if (ctx->state.vg.fill_rule == VG_EVEN_ODD) {
+      dsa.stencil[0].writemask = 1;
+      dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].ref_value = 0;
+      dsa.stencil[0].valuemask = ~0;
+
+      cso_set_blend(ctx->cso_context, &blend);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+      for (i = 0; i < polys->num_elements; ++i) {
+         struct polygon *poly = (((struct polygon**)polys->data)[i]);
+         draw_polygon(ctx, poly);
+      }
+   } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) {
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+         /* front */
+         dsa.stencil[0].writemask = ~0;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         /* back */
+         dsa.stencil[1].enabled = 1;
+         dsa.stencil[1].writemask = ~0;
+         dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         dsa.stencil[1].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[1].ref_value = 0;
+         dsa.stencil[1].valuemask = ~0;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+      } else {
+         struct pipe_rasterizer_state raster;
+
+         memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state));
+
+         cso_save_rasterizer(ctx->cso_context);
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+
+         raster.cull_mode = raster.front_winding;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+
+         cso_restore_rasterizer(ctx->cso_context);
+      }
+   }
+
+   /* restore color writes */
+   cso_restore_blend(ctx->cso_context);
+   /* setup stencil ops */
+   dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL;
+   dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].ref_value = 0;
+   dsa.stencil[0].valuemask = dsa.stencil[0].writemask;
+   dsa.stencil[1].enabled = 0;
+   memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth,
+          sizeof(struct pipe_depth_state));
+   cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_quad(ctx->renderer, min_x, min_y,
+                      max_x, max_y, 0.0f/*depth should be disabled*/);
+
+   cso_restore_depth_stencil_alpha(ctx->cso_context);
+}
diff --git a/src/gallium/state_trackers/vega/polygon.h b/src/gallium/state_trackers/vega/polygon.h
new file mode 100644
index 0000000000..22672b728e
--- /dev/null
+++ b/src/gallium/state_trackers/vega/polygon.h
@@ -0,0 +1,75 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef POLYGON_H
+#define POLYGON_H
+
+#include "VG/openvg.h"
+
+struct polygon;
+struct vg_context;
+struct vg_paint;
+struct array;
+
+struct polygon *polygon_create(int size);
+struct polygon *polygon_create_from_data(float *data, int size);
+void            polygon_destroy(struct polygon *poly);
+
+void            polygon_resize(struct polygon *poly, int new_size);
+int             polygon_size(struct polygon *poly);
+
+int             polygon_vertex_count(struct polygon *poly);
+float *         polygon_data(struct polygon *poly);
+
+void            polygon_vertex_append(struct polygon *p,
+                                      float x, float y);
+void            polygon_append_polygon(struct polygon *dst,
+                                       struct polygon *src);
+void            polygon_set_vertex(struct polygon *p, int idx,
+                                   float x, float y);
+void            polygon_vertex(struct polygon *p, int idx,
+                               float *vertex);
+
+void            polygon_bounding_rect(struct polygon *p,
+                                      float *rect);
+int             polygon_contains_point(struct polygon *p,
+                                       float x, float y);
+
+VGboolean       polygon_is_closed(struct polygon *p);
+
+void polygon_fill(struct polygon *p, struct vg_context *pipe);
+
+/* TODO: make a file/module around this struct
+ */
+struct polygon_array {
+   struct array *array;
+   VGfloat min_x, max_x;
+   VGfloat min_y, max_y;
+};
+
+void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c
new file mode 100644
index 0000000000..f7c5f2f0cd
--- /dev/null
+++ b/src/gallium/state_trackers/vega/renderer.c
@@ -0,0 +1,592 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "renderer.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_draw_quad.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_memory.h"
+
+#include "cso_cache/cso_context.h"
+
+struct renderer {
+   struct pipe_context *pipe;
+   struct vg_context *owner;
+
+   struct cso_context *cso;
+
+   void *fs;
+
+   VGfloat vertices[4][2][4];
+};
+
+static void setup_shaders(struct renderer *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   /* fragment shader */
+   ctx->fs = util_make_fragment_tex_shader(pipe);
+}
+
+static struct pipe_buffer *
+setup_vertex_data(struct renderer *ctx,
+                  float x0, float y0, float x1, float y1, float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = 1.0f; /*s*/
+   ctx->vertices[1][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+static struct pipe_buffer *
+setup_vertex_data_tex(struct renderer *ctx,
+                      float x0, float y0, float x1, float y1,
+                      float s0, float t0, float s1, float t1,
+                      float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = s0; /*s*/
+   ctx->vertices[0][1][1] = t0; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = s1; /*s*/
+   ctx->vertices[1][1][1] = t0; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = s1;
+   ctx->vertices[2][1][1] = t1;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = s0;
+   ctx->vertices[3][1][1] = t1;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+
+static struct pipe_buffer *
+setup_vertex_data_qtex(struct renderer *ctx,
+                       float x0, float y0, float x1, float y1,
+                       float x2, float y2, float x3, float y3,
+                       float s0, float t0, float s1, float t1,
+                       float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = s0; /*s*/
+   ctx->vertices[0][1][1] = t0; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y1;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = s1; /*s*/
+   ctx->vertices[1][1][1] = t0; /*t*/
+
+   ctx->vertices[2][0][0] = x2;
+   ctx->vertices[2][0][1] = y2;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = s1;
+   ctx->vertices[2][1][1] = t1;
+
+   ctx->vertices[3][0][0] = x3;
+   ctx->vertices[3][0][1] = y3;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = s0;
+   ctx->vertices[3][1][1] = t1;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+struct renderer * renderer_create(struct vg_context *owner)
+{
+   VGint i;
+   struct renderer *renderer = CALLOC_STRUCT(renderer);
+
+   if (!renderer)
+      return NULL;
+
+   renderer->owner = owner;
+   renderer->pipe = owner->pipe;
+   renderer->cso = owner->cso_context;
+
+   setup_shaders(renderer);
+
+   /* init vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      renderer->vertices[i][0][3] = 1.0f; /* w */
+      renderer->vertices[i][1][2] = 0.0f; /* r */
+      renderer->vertices[i][1][3] = 1.0f; /* q */
+   }
+
+   return renderer;
+}
+
+void renderer_destroy(struct renderer *ctx)
+{
+#if 0
+   if (ctx->fs) {
+      cso_delete_fragment_shader(ctx->cso, ctx->fs);
+      ctx->fs = NULL;
+   }
+#endif
+   free(ctx);
+}
+
+void renderer_draw_quad(struct renderer *r,
+                        VGfloat x1, VGfloat y1,
+                        VGfloat x2, VGfloat y2,
+                        VGfloat depth)
+{
+   struct pipe_buffer *buf;
+
+   buf = setup_vertex_data(r, x1, y1, x2, y2, depth);
+
+   if (buf) {
+      util_draw_vertex_buffer(r->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+}
+
+void renderer_draw_texture(struct renderer *r,
+                           struct pipe_texture *tex,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf;
+   VGfloat s0, t0, s1, t1;
+
+   assert(tex->width[0] != 0);
+   assert(tex->height[0] != 0);
+
+   s0 = x1offset / tex->width[0];
+   s1 = x2offset / tex->width[0];
+   t0 = y1offset / tex->height[0];
+   t1 = y2offset / tex->height[0];
+
+   cso_save_vertex_shader(r->cso);
+   /* shaders */
+   cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner));
+
+   /* draw quad */
+   buf = setup_vertex_data_tex(r, x1, y1, x2, y2,
+                               s0, t0, s1, t1, 0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+   cso_restore_vertex_shader(r->cso);
+}
+
+void renderer_copy_texture(struct renderer *ctx,
+                           struct pipe_texture *src,
+                           VGfloat sx1, VGfloat sy1,
+                           VGfloat sx2, VGfloat sy2,
+                           struct pipe_texture *dst,
+                           VGfloat dx1, VGfloat dy1,
+                           VGfloat dx2, VGfloat dy2)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_buffer *buf;
+   struct pipe_surface *dst_surf = screen->get_tex_surface(
+      screen, dst, 0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+   struct pipe_framebuffer_state fb;
+   float s0, t0, s1, t1;
+
+   assert(src->width[0] != 0);
+   assert(src->height[0] != 0);
+   assert(dst->width[0] != 0);
+   assert(dst->height[0] != 0);
+
+#if 0
+   debug_printf("copy texture [%f, %f, %f, %f], [%f, %f, %f, %f]\n",
+                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2);
+#endif
+
+#if 1
+   s0 = sx1 / src->width[0];
+   s1 = sx2 / src->width[0];
+   t0 = sy1 / src->height[0];
+   t1 = sy2 / src->height[0];
+#else
+   s0 = 0;
+   s1 = 1;
+   t0 = 0;
+   t1 = 1;
+#endif
+
+   assert(screen->is_format_supported(screen, dst_surf->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+
+   cso_save_viewport(ctx->cso);
+
+
+   /* set misc state we care about */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(ctx->cso, &blend);
+   }
+
+   /* sampler */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      cso_single_sampler(ctx->cso, 0, &sampler);
+      cso_single_sampler_done(ctx->cso);
+   }
+
+   vg_set_viewport(ctx->owner, VEGA_Y0_TOP);
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &src);
+
+   /* shaders */
+   cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner));
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst_surf->width;
+   fb.height = dst_surf->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst_surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso, &fb);
+
+   /* draw quad */
+   buf = setup_vertex_data_tex(ctx,
+                         dx1, dy1,
+                         dx2, dy2,
+                         s0, t0, s1, t1,
+                         0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(ctx->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+
+   pipe_surface_reference(&dst_surf, NULL);
+}
+
+void renderer_copy_surface(struct renderer *ctx,
+                           struct pipe_surface *src,
+                           int srcX0, int srcY0,
+                           int srcX1, int srcY1,
+                           struct pipe_surface *dst,
+                           int dstX0, int dstY0,
+                           int dstX1, int dstY1,
+                           float z, unsigned filter)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_buffer *buf;
+   struct pipe_texture texTemp, *tex;
+   struct pipe_surface *texSurf;
+   struct pipe_framebuffer_state fb;
+   struct st_framebuffer *stfb = ctx->owner->draw_buffer;
+   const int srcW = abs(srcX1 - srcX0);
+   const int srcH = abs(srcY1 - srcY0);
+   const int srcLeft = MIN2(srcX0, srcX1);
+   const int srcTop = MIN2(srcY0, srcY1);
+
+   assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
+          filter == PIPE_TEX_MIPFILTER_LINEAR);
+
+   if (srcLeft != srcX0) {
+      /* left-right flip */
+      int tmp = dstX0;
+      dstX0 = dstX1;
+      dstX1 = tmp;
+   }
+
+   if (srcTop != srcY0) {
+      /* up-down flip */
+      int tmp = dstY0;
+      dstY0 = dstY1;
+      dstY1 = tmp;
+   }
+
+   assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /*
+    * XXX for now we're always creating a temporary texture.
+    * Strictly speaking that's not always needed.
+    */
+
+   /* create temp texture */
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = src->format;
+   texTemp.last_level = 0;
+   texTemp.width[0] = srcW;
+   texTemp.height[0] = srcH;
+   texTemp.depth[0] = 1;
+   pf_get_block(src->format, &texTemp.block);
+
+   tex = screen->texture_create(screen, &texTemp);
+   if (!tex)
+      return;
+
+   texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* load temp texture */
+   pipe->surface_copy(pipe,
+                      texSurf, 0, 0,   /* dest */
+                      src, srcLeft, srcTop, /* src */
+                      srcW, srcH);     /* size */
+
+   /* free the surface, update the texture if necessary.*/
+   screen->tex_surface_destroy(texSurf);
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+   cso_save_viewport(ctx->cso);
+
+   /* set misc state we care about */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(ctx->cso, &blend);
+   }
+
+   vg_set_viewport(ctx->owner, VEGA_Y0_TOP);
+
+   /* sampler */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      cso_single_sampler(ctx->cso, 0, &sampler);
+      cso_single_sampler_done(ctx->cso);
+   }
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &tex);
+
+   /* shaders */
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner));
+
+   /* drawing dest */
+   if (stfb->strb->surface != dst) {
+      memset(&fb, 0, sizeof(fb));
+      fb.width = dst->width;
+      fb.height = dst->height;
+      fb.nr_cbufs = 1;
+      fb.cbufs[0] = dst;
+      fb.zsbuf = stfb->dsrb->surface;
+      cso_set_framebuffer(ctx->cso, &fb);
+   }
+
+   /* draw quad */
+   buf = setup_vertex_data(ctx,
+                           (float) dstX0, (float) dstY0,
+                           (float) dstX1, (float) dstY1, z);
+
+   if (buf) {
+      util_draw_vertex_buffer(ctx->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+
+   pipe_texture_reference(&tex, NULL);
+}
+
+void renderer_texture_quad(struct renderer *r,
+                           struct pipe_texture *tex,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2,
+                           VGfloat x3, VGfloat y3,
+                           VGfloat x4, VGfloat y4)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf;
+   VGfloat s0, t0, s1, t1;
+
+   assert(tex->width[0] != 0);
+   assert(tex->height[0] != 0);
+
+   s0 = x1offset / tex->width[0];
+   s1 = x2offset / tex->width[0];
+   t0 = y1offset / tex->height[0];
+   t1 = y2offset / tex->height[0];
+
+   cso_save_vertex_shader(r->cso);
+   /* shaders */
+   cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner));
+
+   /* draw quad */
+   buf = setup_vertex_data_qtex(r, x1, y1, x2, y2, x3, y3, x4, y4,
+                          s0, t0, s1, t1, 0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf,
+                            NULL);
+   }
+
+   cso_restore_vertex_shader(r->cso);
+}
diff --git a/src/gallium/state_trackers/vega/renderer.h b/src/gallium/state_trackers/vega/renderer.h
new file mode 100644
index 0000000000..990cd32c31
--- /dev/null
+++ b/src/gallium/state_trackers/vega/renderer.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef RENDERER_H
+#define RENDERER_H
+
+#include "VG/openvg.h"
+
+struct renderer;
+
+struct vg_context;
+struct pipe_texture;
+struct pipe_surface;
+
+struct renderer *renderer_create(struct vg_context *owner);
+void renderer_destroy(struct renderer *);
+
+void renderer_draw_quad(struct renderer *,
+                        VGfloat x1, VGfloat y1,
+                        VGfloat x2, VGfloat y2,
+                        VGfloat depth);
+void renderer_draw_texture(struct renderer *,
+                           struct pipe_texture *texture,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2);
+void renderer_texture_quad(struct renderer *,
+                           struct pipe_texture *texture,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2,
+                           VGfloat x3, VGfloat y3,
+                           VGfloat x4, VGfloat y4);
+void renderer_copy_texture(struct renderer *r,
+                           struct pipe_texture *src,
+                           VGfloat sx1, VGfloat sy1,
+                           VGfloat sx2, VGfloat sy2,
+                           struct pipe_texture *dst,
+                           VGfloat dx1, VGfloat dy1,
+                           VGfloat dx2, VGfloat dy2);
+void renderer_copy_surface(struct renderer *r,
+                           struct pipe_surface *src,
+                           int sx1, int sy1,
+                           int sx2, int sy2,
+                           struct pipe_surface *dst,
+                           int dx1, int dy1,
+                           int dx2, int dy2,
+                           float z, unsigned filter);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/shader.c b/src/gallium/state_trackers/vega/shader.c
new file mode 100644
index 0000000000..d9074a377b
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shader.c
@@ -0,0 +1,310 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "shader.h"
+
+#include "vg_context.h"
+#include "shaders_cache.h"
+#include "paint.h"
+#include "mask.h"
+#include "image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+#define MAX_CONSTANTS 20
+
+struct shader {
+   struct vg_context *context;
+
+   VGboolean masking;
+   struct vg_paint *paint;
+   struct vg_image *image;
+
+   VGboolean drawing_image;
+   VGImageMode image_mode;
+
+   float constants[MAX_CONSTANTS];
+   struct pipe_constant_buffer cbuf;
+   struct pipe_shader_state fs_state;
+   void *fs;
+};
+
+struct shader * shader_create(struct vg_context *ctx)
+{
+   struct shader *shader = 0;
+
+   shader = CALLOC_STRUCT(shader);
+   shader->context = ctx;
+
+   return shader;
+}
+
+void shader_destroy(struct shader *shader)
+{
+   free(shader);
+}
+
+void shader_set_masking(struct shader *shader, VGboolean set)
+{
+   shader->masking = set;
+}
+
+VGboolean shader_is_masking(struct shader *shader)
+{
+   return shader->masking;
+}
+
+void shader_set_paint(struct shader *shader, struct vg_paint *paint)
+{
+   shader->paint = paint;
+}
+
+struct vg_paint * shader_paint(struct shader *shader)
+{
+   return shader->paint;
+}
+
+
+static void setup_constant_buffer(struct shader *shader)
+{
+   struct vg_context *ctx = shader->context;
+   struct pipe_context *pipe = shader->context->pipe;
+   struct pipe_constant_buffer *cbuf = &shader->cbuf;
+   VGint param_bytes = paint_constant_buffer_size(shader->paint);
+   float temp_buf[MAX_CONSTANTS];
+
+   assert(param_bytes <= sizeof(temp_buf));
+   paint_fill_constant_buffer(shader->paint, temp_buf);
+
+   if (cbuf->buffer == NULL ||
+       memcmp(temp_buf, shader->constants, param_bytes) != 0)
+   {
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+
+      memcpy(shader->constants, temp_buf, param_bytes);
+      cbuf->buffer = pipe_user_buffer_create(pipe->screen,
+                                             &shader->constants,
+                                             sizeof(shader->constants));
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static VGint blend_bind_samplers(struct vg_context *ctx,
+                                 struct pipe_sampler_state **samplers,
+                                 struct pipe_texture **textures)
+{
+   VGBlendMode bmode = ctx->state.vg.blend_mode;
+
+   if (bmode == VG_BLEND_MULTIPLY ||
+       bmode == VG_BLEND_SCREEN ||
+       bmode == VG_BLEND_DARKEN ||
+       bmode == VG_BLEND_LIGHTEN) {
+      struct st_framebuffer *stfb = ctx->draw_buffer;
+
+      vg_prepare_blend_surface(ctx);
+
+      samplers[2] = &ctx->blend_sampler;
+      textures[2] = stfb->blend_texture;
+
+      if (!samplers[0] || !textures[0]) {
+         samplers[1] = samplers[2];
+         textures[1] = textures[2];
+      }
+      if (!samplers[1] || !textures[1]) {
+         samplers[1] = samplers[0];
+         textures[1] = textures[0];
+      }
+
+      return 1;
+   }
+   return 0;
+}
+
+static void setup_samplers(struct shader *shader)
+{
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct vg_context *ctx = shader->context;
+   /* a little wonky: we use the num as a boolean that just says
+    * whether any sampler/textures have been set. the actual numbering
+    * for samplers is always the same:
+    * 0 - paint sampler/texture for gradient/pattern
+    * 1 - mask sampler/texture
+    * 2 - blend sampler/texture
+    * 3 - image sampler/texture
+    * */
+   VGint num = 0;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   samplers[3] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+   textures[3] = NULL;
+
+   num += paint_bind_samplers(shader->paint, samplers, textures);
+   num += mask_bind_samplers(samplers, textures);
+   num += blend_bind_samplers(ctx, samplers, textures);
+   if (shader->drawing_image && shader->image)
+      num += image_bind_samplers(shader->image, samplers, textures);
+
+   if (num) {
+      cso_set_samplers(ctx->cso_context, 4, (const struct pipe_sampler_state **)samplers);
+      cso_set_sampler_textures(ctx->cso_context, 4, textures);
+   }
+}
+
+static INLINE VGboolean is_format_bw(struct shader *shader)
+{
+#if 0
+   struct vg_context *ctx = shader->context;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+#endif
+
+   if (shader->drawing_image && shader->image) {
+      if (shader->image->format == VG_BW_1)
+         return VG_TRUE;
+   }
+
+   return VG_FALSE;
+}
+
+static void setup_shader_program(struct shader *shader)
+{
+   struct vg_context *ctx = shader->context;
+   VGint shader_id = 0;
+   VGBlendMode blend_mode = ctx->state.vg.blend_mode;
+   VGboolean black_white = is_format_bw(shader);
+
+   /* 1st stage: fill */
+   if (!shader->drawing_image ||
+       (shader->image_mode == VG_DRAW_IMAGE_MULTIPLY || shader->image_mode == VG_DRAW_IMAGE_STENCIL)) {
+      switch(paint_type(shader->paint)) {
+      case VG_PAINT_TYPE_COLOR:
+         shader_id |= VEGA_SOLID_FILL_SHADER;
+         break;
+      case VG_PAINT_TYPE_LINEAR_GRADIENT:
+         shader_id |= VEGA_LINEAR_GRADIENT_SHADER;
+         break;
+      case VG_PAINT_TYPE_RADIAL_GRADIENT:
+         shader_id |= VEGA_RADIAL_GRADIENT_SHADER;
+         break;
+      case VG_PAINT_TYPE_PATTERN:
+         shader_id |= VEGA_PATTERN_SHADER;
+         break;
+
+      default:
+         abort();
+      }
+   }
+
+   /* second stage image */
+   if (shader->drawing_image) {
+      switch(shader->image_mode) {
+      case VG_DRAW_IMAGE_NORMAL:
+         shader_id |= VEGA_IMAGE_NORMAL_SHADER;
+         break;
+      case VG_DRAW_IMAGE_MULTIPLY:
+         shader_id |= VEGA_IMAGE_MULTIPLY_SHADER;
+         break;
+      case VG_DRAW_IMAGE_STENCIL:
+         shader_id |= VEGA_IMAGE_STENCIL_SHADER;
+         break;
+      default:
+         debug_printf("Unknown image mode!");
+      }
+   }
+
+   if (shader->masking)
+      shader_id |= VEGA_MASK_SHADER;
+
+   switch(blend_mode) {
+   case VG_BLEND_MULTIPLY:
+      shader_id |= VEGA_BLEND_MULTIPLY_SHADER;
+      break;
+   case VG_BLEND_SCREEN:
+      shader_id |= VEGA_BLEND_SCREEN_SHADER;
+      break;
+   case VG_BLEND_DARKEN:
+      shader_id |= VEGA_BLEND_DARKEN_SHADER;
+      break;
+   case VG_BLEND_LIGHTEN:
+      shader_id |= VEGA_BLEND_LIGHTEN_SHADER;
+      break;
+   default:
+      /* handled by pipe_blend_state */
+      break;
+   }
+
+   if (black_white)
+      shader_id |= VEGA_BW_SHADER;
+
+   shader->fs = shaders_cache_fill(ctx->sc, shader_id);
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->fs);
+}
+
+
+void shader_bind(struct shader *shader)
+{
+   /* first resolve the real paint type */
+   paint_resolve_type(shader->paint);
+
+   setup_constant_buffer(shader);
+   setup_samplers(shader);
+   setup_shader_program(shader);
+}
+
+void shader_set_image_mode(struct shader *shader, VGImageMode image_mode)
+{
+   shader->image_mode = image_mode;
+}
+
+VGImageMode shader_image_mode(struct shader *shader)
+{
+   return shader->image_mode;
+}
+
+void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image)
+{
+   shader->drawing_image = drawing_image;
+}
+
+VGboolean shader_drawing_image(struct shader *shader)
+{
+   return shader->drawing_image;
+}
+
+void shader_set_image(struct shader *shader, struct vg_image *img)
+{
+   shader->image = img;
+}
diff --git a/src/gallium/state_trackers/vega/shader.h b/src/gallium/state_trackers/vega/shader.h
new file mode 100644
index 0000000000..847eee6a31
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shader.h
@@ -0,0 +1,56 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef SHADER_H
+#define SHADER_H
+
+#include "VG/openvg.h"
+
+struct shader;
+struct vg_paint;
+struct vg_context;
+struct vg_image;
+
+struct shader *shader_create(struct vg_context *context);
+void shader_destroy(struct shader *shader);
+
+void shader_set_masking(struct shader *shader, VGboolean set);
+VGboolean shader_is_masking(struct shader *shader);
+
+void shader_set_paint(struct shader *shader, struct vg_paint *paint);
+struct vg_paint *shader_paint(struct shader *shader);
+
+void shader_set_image_mode(struct shader *shader, VGImageMode image_mode);
+VGImageMode shader_image_mode(struct shader *shader);
+
+void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image);
+VGboolean shader_drawing_image(struct shader *shader);
+
+void shader_set_image(struct shader *shader, struct vg_image *img);
+
+void shader_bind(struct shader *shader);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/shaders_cache.c b/src/gallium/state_trackers/vega/shaders_cache.c
new file mode 100644
index 0000000000..fd0831fab1
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shaders_cache.c
@@ -0,0 +1,439 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "shaders_cache.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_text.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_debug.h"
+#include "cso_cache/cso_hash.h"
+#include "cso_cache/cso_context.h"
+
+#include "VG/openvg.h"
+
+#include "asm_fill.h"
+
+/* Essentially we construct an ubber-shader based on the state
+ * of the pipeline. The stages are:
+ * 1) Fill (mandatory, solid color/gradient/pattern/image draw)
+ * 2) Image composition (image mode multiply and stencil)
+ * 3) Mask
+ * 4) Extended blend (multiply/screen/darken/lighten)
+ * 5) Premultiply/Unpremultiply
+ * 6) Color transform (to black and white)
+ */
+#define SHADER_STAGES 6
+
+struct cached_shader {
+   void *driver_shader;
+   struct pipe_shader_state state;
+};
+
+struct shaders_cache {
+   struct vg_context *pipe;
+
+   struct cso_hash *hash;
+};
+
+
+static INLINE struct tgsi_token *tokens_from_assembly(const char *txt, int num_tokens)
+{
+   struct tgsi_token *tokens;
+
+   tokens = (struct tgsi_token *) MALLOC(num_tokens * sizeof(tokens[0]));
+
+   tgsi_text_translate(txt, tokens, num_tokens);
+
+#if DEBUG_SHADERS
+   tgsi_dump(tokens, 0);
+#endif
+
+   return tokens;
+}
+
+#define ALL_FILLS (VEGA_SOLID_FILL_SHADER | \
+   VEGA_LINEAR_GRADIENT_SHADER | \
+   VEGA_RADIAL_GRADIENT_SHADER | \
+   VEGA_PATTERN_SHADER         | \
+   VEGA_IMAGE_NORMAL_SHADER)
+
+
+/*
+static const char max_shader_preamble[] =
+   "FRAG1.1\n"
+   "DCL IN[0], POSITION, LINEAR\n"
+   "DCL IN[1], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL CONST[0..9], CONSTANT\n"
+   "DCL TEMP[0..9], CONSTANT\n"
+   "DCL SAMP[0..9], CONSTANT\n";
+
+   max_shader_preamble strlen == 175
+*/
+#define MAX_PREAMBLE 175
+
+static INLINE VGint range_min(VGint min, VGint current)
+{
+   if (min < 0)
+      min = current;
+   else
+      min = MIN2(min, current);
+   return min;
+}
+
+static INLINE VGint range_max(VGint max, VGint current)
+{
+   return MAX2(max, current);
+}
+
+static void
+create_preamble(char *txt,
+                const struct shader_asm_info *shaders[SHADER_STAGES],
+                int num_shaders)
+{
+   VGboolean declare_input = VG_FALSE;
+   VGint start_const   = -1, end_const   = 0;
+   VGint start_temp    = -1, end_temp    = 0;
+   VGint start_sampler = -1, end_sampler = 0;
+   VGint i;
+   VGint num_consts, num_temps, num_samplers;
+
+   for (i = 0; i < num_shaders; ++i) {
+      if (shaders[i]->num_consts)
+         start_const = range_min(start_const, shaders[i]->start_const);
+      if (shaders[i]->num_temps)
+         start_temp = range_min(start_temp, shaders[i]->start_temp);
+      if (shaders[i]->num_samplers)
+         start_sampler = range_min(start_sampler, shaders[i]->start_sampler);
+
+      end_const = range_max(end_const, shaders[i]->start_const +
+                            shaders[i]->num_consts);
+      end_temp = range_max(end_temp, shaders[i]->start_temp +
+                            shaders[i]->num_temps);
+      end_sampler = range_max(end_sampler, shaders[i]->start_sampler +
+                            shaders[i]->num_samplers);
+      if (shaders[i]->needs_position)
+         declare_input = VG_TRUE;
+   }
+   /* if they're still unitialized, initialize them */
+   if (start_const < 0)
+      start_const = 0;
+   if (start_temp < 0)
+      start_temp = 0;
+   if (start_sampler < 0)
+      start_sampler = 0;
+
+   num_consts   = end_const   - start_const;
+   num_temps    = end_temp    - start_temp;
+   num_samplers = end_sampler - start_sampler;
+   /* end exclusive */
+   --end_const;
+   --end_temp;
+   --end_sampler;
+
+   sprintf(txt, "FRAG1.1\n");
+
+   if (declare_input) {
+      sprintf(txt + strlen(txt), "DCL IN[0], POSITION, LINEAR\n");
+      sprintf(txt + strlen(txt), "DCL IN[1], GENERIC[0], PERSPECTIVE\n");
+   }
+
+   /* we always have a color output */
+   sprintf(txt + strlen(txt), "DCL OUT[0], COLOR, CONSTANT\n");
+
+   if (num_consts > 1)
+      sprintf(txt + strlen(txt), "DCL CONST[%d..%d], CONSTANT\n", start_const, end_const);
+   else if (num_consts == 1)
+      sprintf(txt + strlen(txt), "DCL CONST[%d], CONSTANT\n", start_const);
+
+   if (num_temps > 1)
+      sprintf(txt + strlen(txt), "DCL TEMP[%d..%d], CONSTANT\n", start_temp, end_temp);
+   else if (num_temps > 1)
+      sprintf(txt + strlen(txt), "DCL TEMP[%d], CONSTANT\n", start_temp);
+
+   if (num_samplers > 1)
+      sprintf(txt + strlen(txt), "DCL SAMP[%d..%d], CONSTANT\n", start_sampler, end_sampler);
+   else if (num_samplers == 1)
+      sprintf(txt + strlen(txt), "DCL SAMP[%d], CONSTANT\n", start_sampler);
+}
+
+static void *
+combine_shaders(const struct shader_asm_info *shaders[SHADER_STAGES], int num_shaders,
+                struct pipe_context *pipe,
+                struct pipe_shader_state *shader)
+{
+   char *combined_txt;
+   int combined_len = MAX_PREAMBLE;
+   int combined_tokens = 0;
+   int i = 0;
+   int current_shader = 0;
+   int current_len;
+
+   for (i = 0; i < num_shaders; ++i) {
+      combined_len += strlen(shaders[i]->txt);
+      combined_tokens += shaders[i]->num_tokens;
+   }
+   /* add for the %s->TEMP[0] substitutions */
+   combined_len += num_shaders * 7 /*TEMP[0]*/ + 4 /*"END\n"*/;
+
+   combined_txt = (char*)malloc(combined_len);
+   combined_txt[0] = '\0';
+
+   create_preamble(combined_txt, shaders, num_shaders);
+
+   while (current_shader < num_shaders) {
+      const char temp[] = "TEMP[0]";
+      const char out[] = "OUT[0]";
+      const char *subst = temp;
+
+      current_len = strlen(combined_txt);
+
+      /* if the last shader then output */
+      if (current_shader + 1 == num_shaders)
+         subst = out;
+
+      snprintf(combined_txt + current_len,
+               combined_len - current_len,
+               shaders[current_shader]->txt,
+               subst);
+      ++current_shader;
+   }
+
+
+   current_len = strlen(combined_txt);
+   snprintf(combined_txt + current_len,
+            combined_len - current_len,
+            "END\n");
+
+   debug_printf("Combined shader is : \n%s\n",
+                 combined_txt);
+
+   shader->tokens = tokens_from_assembly(
+            combined_txt, combined_tokens);
+
+   free(combined_txt);
+
+   return pipe->create_fs_state(pipe, shader);
+}
+
+static void *
+create_shader(struct pipe_context *pipe,
+              int id,
+              struct pipe_shader_state *shader)
+{
+   int idx = 0;
+   const struct shader_asm_info * shaders[SHADER_STAGES];
+
+   /* the shader has to have a fill */
+   debug_assert(id & ALL_FILLS);
+
+   /* first stage */
+   if (id & VEGA_SOLID_FILL_SHADER) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[0];
+      debug_assert(shaders_asm[0].id == VEGA_SOLID_FILL_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_LINEAR_GRADIENT_SHADER)) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[1];
+      debug_assert(shaders_asm[1].id == VEGA_LINEAR_GRADIENT_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_RADIAL_GRADIENT_SHADER)) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[2];
+      debug_assert(shaders_asm[2].id == VEGA_RADIAL_GRADIENT_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_PATTERN_SHADER)) {
+      debug_assert(idx == 0);
+      debug_assert(shaders_asm[3].id == VEGA_PATTERN_SHADER);
+      shaders[idx] = &shaders_asm[3];
+      ++idx;
+   }
+   if ((id & VEGA_IMAGE_NORMAL_SHADER)) {
+      debug_assert(idx == 0);
+      debug_assert(shaders_asm[4].id == VEGA_IMAGE_NORMAL_SHADER);
+      shaders[idx] = &shaders_asm[4];
+      ++idx;
+   }
+
+   /* second stage */
+   if ((id & VEGA_IMAGE_MULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[5].id == VEGA_IMAGE_MULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[5];
+      ++idx;
+   } else if ((id & VEGA_IMAGE_STENCIL_SHADER)) {
+      debug_assert(shaders_asm[6].id == VEGA_IMAGE_STENCIL_SHADER);
+      shaders[idx] = &shaders_asm[6];
+      ++idx;
+   }
+
+   /* third stage */
+   if ((id & VEGA_MASK_SHADER)) {
+      debug_assert(idx == 1);
+      debug_assert(shaders_asm[7].id == VEGA_MASK_SHADER);
+      shaders[idx] = &shaders_asm[7];
+      ++idx;
+   }
+
+   /* fourth stage */
+   if ((id & VEGA_BLEND_MULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[8].id == VEGA_BLEND_MULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[8];
+      ++idx;
+   } else if ((id & VEGA_BLEND_SCREEN_SHADER)) {
+      debug_assert(shaders_asm[9].id == VEGA_BLEND_SCREEN_SHADER);
+      shaders[idx] = &shaders_asm[9];
+      ++idx;
+   } else if ((id & VEGA_BLEND_DARKEN_SHADER)) {
+      debug_assert(shaders_asm[10].id == VEGA_BLEND_DARKEN_SHADER);
+      shaders[idx] = &shaders_asm[10];
+      ++idx;
+   } else if ((id & VEGA_BLEND_LIGHTEN_SHADER)) {
+      debug_assert(shaders_asm[11].id == VEGA_BLEND_LIGHTEN_SHADER);
+      shaders[idx] = &shaders_asm[11];
+      ++idx;
+   }
+
+   /* fifth stage */
+   if ((id & VEGA_PREMULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[12].id == VEGA_PREMULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[12];
+      ++idx;
+   } else if ((id & VEGA_UNPREMULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[13].id == VEGA_UNPREMULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[13];
+      ++idx;
+   }
+
+   /* sixth stage */
+   if ((id & VEGA_BW_SHADER)) {
+      debug_assert(shaders_asm[14].id == VEGA_BW_SHADER);
+      shaders[idx] = &shaders_asm[14];
+      ++idx;
+   }
+
+   return combine_shaders(shaders, idx, pipe, shader);
+}
+
+/*************************************************/
+
+struct shaders_cache * shaders_cache_create(struct vg_context *vg)
+{
+   struct shaders_cache *sc = CALLOC_STRUCT(shaders_cache);
+
+   sc->pipe = vg;
+   sc->hash = cso_hash_create();
+
+   return sc;
+}
+
+void shaders_cache_destroy(struct shaders_cache *sc)
+{
+   struct cso_hash_iter iter = cso_hash_first_node(sc->hash);
+
+   while (!cso_hash_iter_is_null(iter)) {
+      struct cached_shader *cached =
+         (struct cached_shader *)cso_hash_iter_data(iter);
+      cso_delete_fragment_shader(sc->pipe->cso_context,
+                                 cached->driver_shader);
+      iter = cso_hash_erase(sc->hash, iter);
+   }
+
+   cso_hash_delete(sc->hash);
+   free(sc);
+}
+
+void * shaders_cache_fill(struct shaders_cache *sc,
+                          int shader_key)
+{
+   VGint key = shader_key;
+   struct cached_shader *cached;
+   struct cso_hash_iter iter = cso_hash_find(sc->hash, key);
+
+   if (cso_hash_iter_is_null(iter)) {
+      cached = CALLOC_STRUCT(cached_shader);
+      cached->driver_shader = create_shader(sc->pipe->pipe, key, &cached->state);
+
+      cso_hash_insert(sc->hash, key, cached);
+
+      return cached->driver_shader;
+   }
+
+   cached = (struct cached_shader *)cso_hash_iter_data(iter);
+
+   assert(cached->driver_shader);
+   return cached->driver_shader;
+}
+
+struct vg_shader * shader_create_from_text(struct pipe_context *pipe,
+                                           const char *txt, int num_tokens,
+                                           int type)
+{
+   struct vg_shader *shader = (struct vg_shader *)malloc(
+      sizeof(struct vg_shader));
+   struct tgsi_token *tokens = tokens_from_assembly(txt, num_tokens);
+   struct pipe_shader_state state;
+
+   debug_assert(type == PIPE_SHADER_VERTEX ||
+                type == PIPE_SHADER_FRAGMENT);
+
+   state.tokens = tokens;
+   shader->type = type;
+   shader->tokens = tokens;
+
+   if (type == PIPE_SHADER_FRAGMENT)
+      shader->driver = pipe->create_fs_state(pipe, &state);
+   else
+      shader->driver = pipe->create_vs_state(pipe, &state);
+   return shader;
+}
+
+void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader)
+{
+   if (shader->type == PIPE_SHADER_FRAGMENT)
+      cso_delete_fragment_shader(ctx->cso_context, shader->driver);
+   else
+      cso_delete_vertex_shader(ctx->cso_context, shader->driver);
+   free(shader->tokens);
+   free(shader);
+}
diff --git a/src/gallium/state_trackers/vega/shaders_cache.h b/src/gallium/state_trackers/vega/shaders_cache.h
new file mode 100644
index 0000000000..feca58b61a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shaders_cache.h
@@ -0,0 +1,77 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef SHADERS_CACHE_H
+#define SHADERS_CACHE_H
+
+
+struct vg_context;
+struct pipe_context;
+struct tgsi_token;
+struct shaders_cache;
+
+enum VegaShaderType {
+   VEGA_SOLID_FILL_SHADER         = 1 <<  0,
+   VEGA_LINEAR_GRADIENT_SHADER    = 1 <<  1,
+   VEGA_RADIAL_GRADIENT_SHADER    = 1 <<  2,
+   VEGA_PATTERN_SHADER            = 1 <<  3,
+   VEGA_IMAGE_NORMAL_SHADER       = 1 <<  4,
+   VEGA_IMAGE_MULTIPLY_SHADER     = 1 <<  5,
+   VEGA_IMAGE_STENCIL_SHADER      = 1 <<  6,
+
+   VEGA_MASK_SHADER               = 1 <<  7,
+
+   VEGA_BLEND_MULTIPLY_SHADER     = 1 <<  8,
+   VEGA_BLEND_SCREEN_SHADER       = 1 <<  9,
+   VEGA_BLEND_DARKEN_SHADER       = 1 << 10,
+   VEGA_BLEND_LIGHTEN_SHADER      = 1 << 11,
+
+   VEGA_PREMULTIPLY_SHADER        = 1 << 12,
+   VEGA_UNPREMULTIPLY_SHADER      = 1 << 13,
+
+   VEGA_BW_SHADER                 = 1 << 14
+};
+
+struct vg_shader {
+   void *driver;
+   struct tgsi_token *tokens;
+   int type;/* PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT */
+};
+
+struct shaders_cache *shaders_cache_create(struct vg_context *pipe);
+void shaders_cache_destroy(struct shaders_cache *sc);
+void *shaders_cache_fill(struct shaders_cache *sc,
+                         int shader_key);
+
+struct vg_shader *shader_create_from_text(struct pipe_context *pipe,
+                                          const char *txt, int num_tokens,
+                                          int type);
+
+void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader);
+
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/st_inlines.h b/src/gallium/state_trackers/vega/st_inlines.h
new file mode 100644
index 0000000000..1f331dfcdb
--- /dev/null
+++ b/src/gallium/state_trackers/vega/st_inlines.h
@@ -0,0 +1,159 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for checking if buffers/textures are referenced when we need
+ * to read/write from/to them.  Flush when needed.
+ */
+
+#ifndef ST_INLINES_H
+#define ST_INLINES_H
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_state.h"
+
+static INLINE struct pipe_transfer *
+st_cond_flush_get_tex_transfer(struct vg_context *st,
+			       struct pipe_texture *pt,
+			       unsigned int face,
+			       unsigned int level,
+			       unsigned int zslice,
+			       enum pipe_transfer_usage usage,
+			       unsigned int x, unsigned int y,
+			       unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_context *pipe = st->pipe;
+   unsigned referenced =
+      pipe->is_texture_referenced(pipe, pt, face, level);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      usage == PIPE_TRANSFER_WRITE ||
+		      usage == PIPE_TRANSFER_READ_WRITE))
+      vgFlush();
+
+   return screen->get_tex_transfer(screen, pt, face, level, zslice, usage,
+				   x, y, w, h);
+}
+
+static INLINE struct pipe_transfer *
+st_no_flush_get_tex_transfer(struct vg_context *st,
+			     struct pipe_texture *pt,
+			     unsigned int face,
+			     unsigned int level,
+			     unsigned int zslice,
+			     enum pipe_transfer_usage usage,
+			     unsigned int x, unsigned int y,
+			     unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   return screen->get_tex_transfer(screen, pt, face, level,
+				   zslice, usage, x, y, w, h);
+}
+
+static INLINE void *
+st_cond_flush_pipe_buffer_map(struct vg_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int map_flags)
+{
+   struct pipe_context *pipe = st->pipe;
+   unsigned int referenced = pipe->is_buffer_referenced(pipe, buf);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      (map_flags & PIPE_BUFFER_USAGE_CPU_WRITE)))
+      vgFlush();
+
+   return pipe_buffer_map(pipe->screen, buf, map_flags);
+}
+
+static INLINE void *
+st_no_flush_pipe_buffer_map(struct vg_context *st,
+			    struct pipe_buffer *buf,
+			    unsigned int map_flags)
+{
+   return pipe_buffer_map(st->pipe->screen, buf, map_flags);
+}
+
+
+static INLINE void
+st_cond_flush_pipe_buffer_write(struct vg_context *st,
+				struct pipe_buffer *buf,
+				unsigned int offset,
+				unsigned int size,
+				const void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf))
+      vgFlush();
+
+   pipe_buffer_write(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_write(struct vg_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int offset,
+			      unsigned int size,
+			      const void * data)
+{
+   pipe_buffer_write(st->pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_cond_flush_pipe_buffer_read(struct vg_context *st,
+			       struct pipe_buffer *buf,
+			       unsigned int offset,
+			       unsigned int size,
+			       void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf) & PIPE_REFERENCED_FOR_WRITE)
+      vgFlush();
+
+   pipe_buffer_read(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_read(struct vg_context *st,
+			     struct pipe_buffer *buf,
+			     unsigned int offset,
+			     unsigned int size,
+			     void * data)
+{
+   pipe_buffer_read(st->pipe->screen, buf, offset, size, data);
+}
+
+#endif
+
diff --git a/src/gallium/state_trackers/vega/stroker.c b/src/gallium/state_trackers/vega/stroker.c
new file mode 100644
index 0000000000..1b92d2b5c6
--- /dev/null
+++ b/src/gallium/state_trackers/vega/stroker.c
@@ -0,0 +1,1349 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "stroker.h"
+
+#include "path.h"
+#include "vg_state.h"
+#include "util_array.h"
+#include "arc.h"
+#include "bezier.h"
+#include "matrix.h"
+#include "path_utils.h"
+#include "polygon.h"
+
+#include "math.h"
+
+#ifndef M_2PI
+#define M_2PI 6.28318530717958647692528676655900576
+#endif
+
+#define STROKE_SEGMENTS 0
+#define STROKE_DEBUG 0
+#define DEBUG_EMITS 0
+
+static const VGfloat curve_threshold = 0.25f;
+
+static const VGfloat zero_coords[] = {0.f, 0.f};
+
+enum intersection_type {
+   NoIntersections,
+   BoundedIntersection,
+   UnboundedIntersection,
+};
+
+enum line_join_mode {
+   FlatJoin,
+   SquareJoin,
+   MiterJoin,
+   RoundJoin,
+   RoundCap
+};
+
+struct stroke_iterator {
+   void (*next)(struct stroke_iterator *);
+   VGboolean (*has_next)(struct stroke_iterator *);
+
+   VGPathCommand (*current_command)(struct stroke_iterator *it);
+   void (*current_coords)(struct stroke_iterator *it, VGfloat *coords);
+
+   VGint position;
+   VGint coord_position;
+
+   const VGubyte *cmds;
+   const VGfloat *coords;
+   VGint num_commands;
+   VGint num_coords;
+
+   struct polygon *curve_poly;
+   VGint curve_index;
+};
+
+static VGPathCommand stroke_itr_command(struct stroke_iterator *itr)
+{
+   return itr->current_command(itr);
+}
+
+static void stroke_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   itr->current_coords(itr, coords);
+}
+
+static void stroke_fw_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->position >= itr->num_commands)
+      return;
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      coords[2] = itr->coords[itr->coord_position + 2];
+      coords[3] = itr->coords[itr->coord_position + 3];
+      coords[4] = itr->coords[itr->coord_position + 4];
+      coords[5] = itr->coords[itr->coord_position + 5];
+      break;
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+
+static void stroke_bw_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->position >= itr->num_commands)
+      return;
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position + 4];
+      coords[1] = itr->coords[itr->coord_position + 5];
+      coords[2] = itr->coords[itr->coord_position + 2];
+      coords[3] = itr->coords[itr->coord_position + 3];
+      coords[4] = itr->coords[itr->coord_position + 0];
+      coords[5] = itr->coords[itr->coord_position + 1];
+      break;
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+
+static VGPathCommand stroke_fw_current_command(struct stroke_iterator *it)
+{
+   return it->cmds[it->position];
+}
+
+static VGPathCommand stroke_bw_current_command(struct stroke_iterator *it)
+{
+   VGPathCommand prev_cmd;
+   if (it->position == it->num_commands  -1)
+      return VG_MOVE_TO_ABS;
+
+   prev_cmd = it->cmds[it->position + 1];
+   return prev_cmd;
+}
+
+static VGboolean stroke_fw_has_next(struct stroke_iterator *itr)
+{
+   return itr->position < (itr->num_commands - 1);
+}
+
+static VGboolean stroke_bw_has_next(struct stroke_iterator *itr)
+{
+   return itr->position > 0;
+}
+
+static void stroke_fw_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+   debug_assert(stroke_fw_has_next(itr));
+
+   cmd = stroke_itr_command(itr);
+
+   itr->coord_position += num_elements_for_segments(&cmd, 1);
+   ++itr->position;
+}
+
+static void stroke_bw_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+   debug_assert(stroke_bw_has_next(itr));
+
+   --itr->position;
+   cmd = stroke_itr_command(itr);
+
+   itr->coord_position -= num_elements_for_segments(&cmd, 1);
+}
+
+static void stroke_itr_common_init(struct stroke_iterator *itr,
+                                   struct array *cmds,
+                                   struct array *coords)
+{
+   itr->cmds = (VGubyte*)cmds->data;
+   itr->num_commands = cmds->num_elements;
+
+   itr->coords = (VGfloat*)coords->data;
+   itr->num_coords = coords->num_elements;
+}
+
+static void stroke_forward_iterator(struct stroke_iterator *itr,
+                                    struct array *cmds,
+                                    struct array *coords)
+{
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = 0;
+   itr->coord_position = 0;
+
+   itr->next = stroke_fw_next;
+   itr->has_next = stroke_fw_has_next;
+   itr->current_command = stroke_fw_current_command;
+   itr->current_coords = stroke_fw_itr_coords;
+}
+
+static void stroke_backward_iterator(struct stroke_iterator *itr,
+                                     struct array *cmds,
+                                     struct array *coords)
+{
+   VGubyte cmd;
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = itr->num_commands - 1;
+
+   cmd = stroke_bw_current_command(itr);
+   itr->coord_position = itr->num_coords -
+                         num_elements_for_segments(&cmd, 1);
+
+   itr->next = stroke_bw_next;
+   itr->has_next = stroke_bw_has_next;
+   itr->current_command = stroke_bw_current_command;
+   itr->current_coords = stroke_bw_itr_coords;
+}
+
+
+
+static void stroke_flat_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+
+   if (itr->curve_index >= 0) {
+      ++itr->curve_index;
+      if (itr->curve_index >= polygon_vertex_count(itr->curve_poly)) {
+         itr->curve_index = -1;
+         polygon_destroy(itr->curve_poly);
+         itr->curve_poly = 0;
+      } else
+         return;
+   }
+   debug_assert(stroke_fw_has_next(itr));
+
+   cmd = itr->cmds[itr->position];
+   itr->coord_position += num_elements_for_segments(&cmd, 1);
+   ++itr->position;
+
+   cmd = itr->cmds[itr->position];
+
+   if (cmd == VG_CUBIC_TO_ABS) {
+      struct bezier bezier;
+      VGfloat bez[8];
+
+      bez[0] = itr->coords[itr->coord_position - 2];
+      bez[1] = itr->coords[itr->coord_position - 1];
+      bez[2] = itr->coords[itr->coord_position];
+      bez[3] = itr->coords[itr->coord_position + 1];
+      bez[4] = itr->coords[itr->coord_position + 2];
+      bez[5] = itr->coords[itr->coord_position + 3];
+      bez[6] = itr->coords[itr->coord_position + 4];
+      bez[7] = itr->coords[itr->coord_position + 5];
+
+      bezier_init(&bezier,
+                  bez[0], bez[1],
+                  bez[2], bez[3],
+                  bez[4], bez[5],
+                  bez[6], bez[7]);
+      /* skip the first one, it's the same as the prev point */
+      itr->curve_index = 1;
+      if (itr->curve_poly) {
+         polygon_destroy(itr->curve_poly);
+         itr->curve_poly = 0;
+      }
+      itr->curve_poly = bezier_to_polygon(&bezier);
+   }
+}
+
+static VGboolean stroke_flat_has_next(struct stroke_iterator *itr)
+{
+   return  (itr->curve_index >= 0 &&
+            itr->curve_index < (polygon_vertex_count(itr->curve_poly)-1))
+            || itr->position < (itr->num_commands - 1);
+}
+
+static VGPathCommand stroke_flat_current_command(struct stroke_iterator *it)
+{
+   if (it->cmds[it->position] == VG_CUBIC_TO_ABS) {
+      return VG_LINE_TO_ABS;
+   }
+   return it->cmds[it->position];
+}
+
+static void stroke_flat_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->curve_index <= -1 && itr->position >= itr->num_commands)
+      return;
+
+   if (itr->curve_index >= 0) {
+      polygon_vertex(itr->curve_poly, itr->curve_index,
+                     coords);
+      return;
+   }
+
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+static void stroke_flat_iterator(struct stroke_iterator *itr,
+                                 struct array *cmds,
+                                 struct array *coords)
+{
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = 0;
+   itr->coord_position = 0;
+
+   itr->next = stroke_flat_next;
+   itr->has_next = stroke_flat_has_next;
+   itr->current_command = stroke_flat_current_command;
+   itr->current_coords = stroke_flat_itr_coords;
+   itr->curve_index = -1;
+   itr->curve_poly = 0;
+}
+
+
+static INLINE VGboolean finite_coords4(const VGfloat *c)
+{
+   return
+      isfinite(c[0]) && isfinite(c[1]) &&
+      isfinite(c[2]) && isfinite(c[3]);
+}
+
+/* from Graphics Gems II */
+#define SAME_SIGNS(a, b) ((a) * (b) >= 0)
+static VGboolean do_lines_intersect(VGfloat x1, VGfloat y1, VGfloat x2, VGfloat y2,
+                                    VGfloat x3, VGfloat y3, VGfloat x4, VGfloat y4)
+{
+   VGfloat a1, a2, b1, b2, c1, c2; /* Coefficients of line eqns */
+   VGfloat r1, r2, r3, r4;         /* 'sign' values */
+
+   a1 = y2 - y1;
+   b1 = x1 - x2;
+   c1 = x2 * y1 - x1 * y2;
+
+   r3 = a1 * x3 + b1 * y3 + c1;
+   r4 = a1 * x4 + b1 * y4 + c1;
+
+   if (r3 != 0 && r4 != 0 && SAME_SIGNS(r3, r4))
+      return VG_FALSE;
+
+   a2 = y4 - y3;
+   b2 = x3 - x4;
+   c2 = x4 * y3 - x3 * y4;
+
+   r1 = a2 * x1 + b2 * y1 + c2;
+   r2 = a2 * x2 + b2 * y2 + c2;
+
+   if (r1 != 0 && r2 != 0 && SAME_SIGNS(r1, r2))
+      return VG_FALSE;
+
+   return VG_TRUE;
+}
+
+static INLINE VGfloat line_dx(const VGfloat *l)
+{
+   return l[2] - l[0];
+}
+
+static INLINE VGfloat line_dy(const VGfloat *l)
+{
+   return l[3] - l[1];
+}
+
+static INLINE VGfloat line_angle(const VGfloat *l)
+{
+   const VGfloat dx = line_dx(l);
+   const VGfloat dy = line_dy(l);
+
+   const VGfloat theta = atan2(-dy, dx) * 360.0 / M_2PI;
+
+   const VGfloat theta_normalized = theta < 0 ? theta + 360 : theta;
+
+   if (floatsEqual(theta_normalized, 360.f))
+      return 0;
+   else
+      return theta_normalized;
+}
+
+static INLINE void line_set_length(VGfloat *l, VGfloat len)
+{
+   VGfloat uv[] = {l[0], l[1], l[2], l[3]};
+   if (null_line(l))
+      return;
+   line_normalize(uv);
+   l[2] = l[0] + line_dx(uv) * len;
+   l[3] = l[1] + line_dy(uv) * len;
+}
+
+static INLINE void line_translate(VGfloat *l, VGfloat x, VGfloat y)
+{
+   l[0] += x;
+   l[1] += y;
+   l[2] += x;
+   l[3] += y;
+}
+
+static INLINE VGfloat line_angle_to(const VGfloat *l1,
+                                    const VGfloat *l2)
+{
+   VGfloat a1, a2, delta, delta_normalized;
+   if (null_line(l1) || null_line(l1))
+      return 0;
+
+   a1 = line_angle(l1);
+   a2 = line_angle(l2);
+
+   delta = a2 - a1;
+   delta_normalized = delta < 0 ? delta + 360 : delta;
+
+   if (floatsEqual(delta, 360.f))
+      return 0;
+   else
+      return delta_normalized;
+}
+
+static INLINE VGfloat line_angles(const VGfloat *l1,
+                                  const VGfloat *l2)
+{
+   VGfloat cos_line, rad = 0;
+
+   if (null_line(l1) || null_line(l2))
+      return 0;
+
+   cos_line = (line_dx(l1)*line_dx(l2) + line_dy(l1)*line_dy(l2)) /
+              (line_lengthv(l1)*line_lengthv(l2));
+   rad = 0;
+
+   if (cos_line >= -1.0 && cos_line <= 1.0)
+      rad = acos(cos_line);
+   return rad * 360 / M_2PI;
+}
+
+
+static INLINE VGfloat adapted_angle_on_x(const VGfloat *line)
+{
+   const VGfloat identity[] = {0, 0, 1, 0};
+   VGfloat angle = line_angles(line, identity);
+   if (line_dy(line) > 0)
+      angle = 360 - angle;
+   return angle;
+}
+
+static enum intersection_type line_intersect(const VGfloat *l1,
+                                             const VGfloat *l2,
+                                             float *intersection_point)
+{
+   VGfloat isect[2];
+   enum intersection_type type;
+   VGboolean dx_zero, ldx_zero;
+
+   if (null_line(l1) || null_line(l2) ||
+       !finite_coords4(l1) || !finite_coords4(l2))
+      return NoIntersections;
+
+   type = do_lines_intersect(l1[0], l1[1], l1[2], l1[3], l2[0], l2[1], l2[2], l2[3])
+          ? BoundedIntersection : UnboundedIntersection;
+
+   dx_zero  = floatsEqual(line_dx(l1) + 1, 1);
+   ldx_zero = floatsEqual(line_dx(l2) + 1, 1);
+
+   /* one of the lines is vertical */
+   if (dx_zero && ldx_zero) {
+      type = NoIntersections;
+   } else if (dx_zero) {
+      VGfloat la = line_dy(l2) / line_dx(l2);
+      isect[0] = l1[0];
+      isect[1] = la * l1[0] + l2[1] - la * l2[0];
+   } else if (ldx_zero) {
+      VGfloat ta = line_dy(l1) / line_dx(l1);
+      isect[0] = l2[0];
+      isect[1] = ta * l2[0] + l1[1] - ta*l1[0];
+   } else {
+      VGfloat x;
+      VGfloat ta = line_dy(l1) / line_dx(l1);
+      VGfloat la = line_dy(l2) / line_dx(l2);
+      if (ta == la)
+         return NoIntersections;
+
+      x = ( - l2[1] + la * l2[0] + l1[1] - ta * l1[0] ) / (la - ta);
+      isect[0] = x;
+      isect[1] = ta*(x - l1[0]) + l1[1];
+   }
+   if (intersection_point) {
+      intersection_point[0] = isect[0];
+      intersection_point[1] = isect[1];
+   }
+   return type;
+}
+
+static INLINE enum line_join_mode stroker_join_mode(struct stroker *s)
+{
+   switch(s->join_style) {
+   case VG_JOIN_MITER:
+      return MiterJoin;
+   case VG_JOIN_ROUND:
+      return RoundJoin;
+   case VG_JOIN_BEVEL:
+      return FlatJoin;
+   default:
+      return FlatJoin;
+   }
+}
+
+static INLINE enum line_join_mode stroker_cap_mode(struct stroker *s)
+{
+   switch(s->cap_style) {
+   case VG_CAP_BUTT:
+      return FlatJoin;
+   case VG_CAP_ROUND:
+      return RoundCap;
+   case VG_CAP_SQUARE:
+      return SquareJoin;
+   default:
+      return FlatJoin;
+   }
+}
+
+void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_MOVE_TO_ABS;
+   VGfloat coords[2] = {x, y};
+#if DEBUG_EMITS
+   debug_printf("emit move %f, %f\n", x, y);
+#endif
+   stroker->back2_x = stroker->back1_x;
+   stroker->back2_y = stroker->back1_y;
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_LINE_TO_ABS;
+   VGfloat coords[2] = {x, y};
+#if DEBUG_EMITS
+   debug_printf("emit line %f, %f\n", x, y);
+#endif
+   stroker->back2_x = stroker->back1_x;
+   stroker->back2_y = stroker->back1_y;
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                                  VGfloat px2, VGfloat py2,
+                                  VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_CUBIC_TO_ABS;
+   VGfloat coords[6] = {px1, py1, px2, py2, x, y};
+#if DEBUG_EMITS
+   debug_printf("emit curve %f, %f, %f, %f, %f, %f\n", px1, py1,
+                px2, py2, x, y);
+#endif
+
+   if (px2 == x && py2 == y) {
+      if (px1 == x && py1 == y) {
+         stroker->back2_x = stroker->back1_x;
+         stroker->back2_y = stroker->back1_y;
+      } else {
+         stroker->back2_x = px1;
+         stroker->back2_y = py1;
+      }
+   } else {
+      stroker->back2_x = px2;
+      stroker->back2_y = py2;
+   }
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+static INLINE void create_round_join(struct stroker *stroker,
+                                     VGfloat x1, VGfloat y1,
+                                     VGfloat x2, VGfloat y2,
+                                     VGfloat width, VGfloat height)
+{
+   struct arc arc;
+   struct matrix matrix;
+
+   matrix_load_identity(&matrix);
+
+   /*stroker_emit_line_to(stroker, nx, ny);*/
+
+   arc_init(&arc, VG_SCCWARC_TO_ABS,
+            x1, y1, x2, y2, width/2, height/2, 0);
+   arc_stroker_emit(&arc, stroker, &matrix);
+}
+
+
+static void create_joins(struct stroker *stroker,
+                         VGfloat focal_x, VGfloat focal_y,
+                         const VGfloat *next_line, enum line_join_mode join)
+{
+#if DEBUG_EMITS
+   debug_printf("create_joins: focal=[%f, %f], next_line=[%f, %f,%f, %f]\n",
+                focal_x, focal_y,
+                next_line[0], next_line[1], next_line[2], next_line[3]);
+#endif
+   /* if we're alredy connected do nothing */
+   if (floatsEqual(stroker->back1_x, next_line[0]) &&
+       floatsEqual(stroker->back1_y, next_line[1]))
+      return;
+
+   if (join == FlatJoin) {
+      stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+   } else {
+      VGfloat prev_line[] = {stroker->back2_x, stroker->back2_y,
+                             stroker->back1_x, stroker->back1_y};
+
+      VGfloat isect[2];
+      enum intersection_type type = line_intersect(prev_line, next_line, isect);
+
+      if (join == SquareJoin) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat l1[4] = {prev_line[0],
+                          prev_line[1],
+                          prev_line[2],
+                          prev_line[3]};
+         VGfloat l2[4] = {next_line[2],
+                          next_line[3],
+                          next_line[0],
+                          next_line[1]};
+
+         line_translate(l1, line_dx(l1), line_dy(l1));
+         line_set_length(l1, offset);
+
+         line_translate(l2, line_dx(l2), line_dy(l2));
+         line_set_length(l2, offset);
+
+         stroker_emit_line_to(stroker, l1[2], l1[3]);
+         stroker_emit_line_to(stroker, l2[2], l2[3]);
+         stroker_emit_line_to(stroker, l2[0], l2[1]);
+      } else if (join == RoundJoin) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat short_cut[4] = {prev_line[2], prev_line[3],
+                                 next_line[0], next_line[1]};
+         VGfloat angle = line_angles(prev_line, short_cut);
+
+         if (type == BoundedIntersection ||
+             (angle > 90 && !floatsEqual(angle, 90.f))) {
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+            return;
+         }
+         create_round_join(stroker, prev_line[2], prev_line[3],
+                           next_line[0], next_line[1],
+                           offset * 2, offset * 2);
+
+         stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+      } else if (join == RoundCap) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat l1[4] = { prev_line[0], prev_line[1],
+                           prev_line[2], prev_line[3] };
+         VGfloat l2[4] = {focal_x, focal_y,
+                          prev_line[2], prev_line[3]};
+
+         line_translate(l1, line_dx(l1), line_dy(l1));
+         line_set_length(l1, KAPPA * offset);
+
+         /* normal between prev_line and focal */
+         line_translate(l2, -line_dy(l2), line_dx(l2));
+         line_set_length(l2, KAPPA * offset);
+
+         stroker_emit_curve_to(stroker, l1[2], l1[3],
+                               l2[2], l2[3],
+                               l2[0], l2[1]);
+
+         l2[0] = l2[0];
+         l2[1] = l2[1];
+         l2[2] = l2[0] - line_dx(l2);
+         l2[3] = l2[1] - line_dy(l2);
+
+         line_translate(l1, next_line[0] - l1[0], next_line[1] - l1[1]);
+
+         stroker_emit_curve_to(stroker,
+                               l2[2], l2[3],
+                               l1[2], l1[3],
+                               l1[0], l1[1]);
+      } else if (join == MiterJoin) {
+         VGfloat miter_line[4] = {stroker->back1_x, stroker->back1_y,
+                                  isect[0], isect[1]};
+         VGfloat sl = (stroker->stroke_width * stroker->miter_limit);
+         VGfloat inside_line[4] = {prev_line[2], prev_line[3],
+                                   next_line[0], next_line[1]};
+         VGfloat angle = line_angle_to(inside_line, prev_line);
+
+         if (type == BoundedIntersection ||
+             (angle > 90 && !floatsEqual(angle, 90.f))) {
+            /*
+            debug_printf("f = %f, nl = %f, pl = %f, is = %f\n",
+                         focal_x, next_line[0],
+                         prev_line[2], isect[0]);*/
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+            return;
+         }
+
+         if (type == NoIntersections || line_lengthv(miter_line) > sl) {
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+         } else {
+            stroker_emit_line_to(stroker, isect[0], isect[1]);
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+         }
+      } else {
+         debug_assert(!"create_joins bad join style");
+      }
+   }
+}
+
+static void stroker_add_segment(struct stroker *stroker,
+                                VGPathCommand cmd,
+                                const VGfloat *coords,
+                                VGint num_coords)
+{
+   /* skip duplicated points */
+   if (stroker->segments->num_elements &&
+       stroker->last_cmd == cmd) {
+      VGfloat *data = stroker->control_points->data;
+      data += stroker->control_points->num_elements;
+      data -= num_coords;
+      switch (cmd) {
+      case VG_MOVE_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]))
+            return;
+         break;
+      case VG_LINE_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]))
+            return;
+         break;
+      case VG_CUBIC_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]) &&
+             floatsEqual(coords[2], data[2]) &&
+             floatsEqual(coords[3], data[3]) &&
+             floatsEqual(coords[4], data[4]) &&
+             floatsEqual(coords[5], data[5]))
+            return;
+         break;
+      default:
+         debug_assert(!"Invalid stroke segment");
+      }
+   } else if (stroker->last_cmd == VG_CUBIC_TO_ABS &&
+              cmd == VG_LINE_TO_ABS) {
+      VGfloat *data = stroker->control_points->data;
+      data += stroker->control_points->num_elements;
+      data -= 2;
+      if (floatsEqual(coords[0], data[0]) &&
+          floatsEqual(coords[1], data[1]))
+         return;
+   }
+   stroker->last_cmd = cmd;
+   array_append_data(stroker->segments, &cmd, 1);
+   array_append_data(stroker->control_points, coords, num_coords);
+}
+
+void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGfloat coords[2] = {x, y};
+#if STROKE_SEGMENTS
+   debug_printf("stroker_move_to(%f, %f)\n", x, y);
+#endif
+
+   if (stroker->segments->num_elements > 0)
+      stroker->process_subpath(stroker);
+
+   array_reset(stroker->segments);
+   array_reset(stroker->control_points);
+
+   stroker_add_segment(stroker, VG_MOVE_TO_ABS, coords, 2);
+}
+
+void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGfloat coords[] = {x, y};
+
+#if STROKE_SEGMENTS
+   debug_printf("stroker_line_to(%f, %f)\n", x, y);
+#endif
+   if (!stroker->segments->num_elements)
+      stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2);
+
+   stroker_add_segment(stroker, VG_LINE_TO_ABS, coords, 2);
+}
+
+void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                      VGfloat px2, VGfloat py2,
+                      VGfloat x, VGfloat y)
+{
+   VGfloat coords[] = {px1, py1,
+                       px2, py2,
+                       x, y};
+#if STROKE_SEGMENTS
+   debug_printf("stroker_curve_to(%f, %f, %f, %f, %f, %f)\n",
+                px1, py1, px2, py2, x, y);
+#endif
+   if (!stroker->segments->num_elements)
+      stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2);
+
+   stroker_add_segment(stroker, VG_CUBIC_TO_ABS, coords, 6);
+}
+
+static INLINE VGboolean is_segment_null(VGPathCommand cmd,
+                                        VGfloat *coords,
+                                        VGfloat *res)
+{
+   switch(cmd) {
+   case VG_MOVE_TO_ABS:
+   case VG_LINE_TO_ABS:
+      return floatsEqual(coords[0], res[0]) &&
+         floatsEqual(coords[1], res[1]);
+      break;
+   case VG_CUBIC_TO_ABS:
+      return floatsEqual(coords[0], res[0]) &&
+         floatsEqual(coords[1], res[1]) &&
+         floatsEqual(coords[2], res[0]) &&
+         floatsEqual(coords[3], res[1]) &&
+         floatsEqual(coords[4], res[0]) &&
+         floatsEqual(coords[5], res[1]);
+      break;
+   default:
+      assert(0);
+   }
+   return VG_FALSE;
+}
+
+static VGboolean vg_stroke_outline(struct stroke_iterator *it,
+                                struct stroker *stroker,
+                                VGboolean cap_first,
+                                VGfloat *start_tangent)
+{
+   const int MAX_OFFSET = 16;
+   struct bezier offset_curves[MAX_OFFSET];
+   VGPathCommand first_element;
+   VGfloat start[2], prev[2];
+   VGboolean first = VG_TRUE;
+   VGfloat offset;
+
+   first_element = stroke_itr_command(it);
+   if (first_element != VG_MOVE_TO_ABS) {
+      stroker_emit_move_to(stroker, 0.f, 0.f);
+      prev[0] = 0.f;
+      prev[1] = 0.f;
+   }
+   stroke_itr_coords(it, start);
+#if STROKE_DEBUG
+   debug_printf(" -> (side) [%.2f, %.2f]\n",
+                start[0],
+                start[1]);
+#endif
+
+   prev[0] = start[0];
+   prev[1] = start[1];
+
+   offset = stroker->stroke_width / 2;
+
+   if (!it->has_next(it)) {
+      /* single point */
+
+      return VG_TRUE;
+   }
+
+   while (it->has_next(it)) {
+      VGPathCommand cmd;
+      VGfloat coords[8];
+
+      it->next(it);
+      cmd = stroke_itr_command(it);
+      stroke_itr_coords(it, coords);
+
+      if (cmd == VG_LINE_TO_ABS) {
+         VGfloat line[4] = {prev[0], prev[1], coords[0], coords[1]};
+         VGfloat normal[4];
+         line_normal(line, normal);
+
+#if STROKE_DEBUG
+         debug_printf("\n ---> (side) lineto [%.2f, %.2f]\n", coords[0], coords[1]);
+#endif
+         line_set_length(normal, offset);
+         line_translate(line, line_dx(normal), line_dy(normal));
+
+         /* if we are starting a new subpath, move to correct starting point */
+         if (first) {
+            if (cap_first)
+               create_joins(stroker, prev[0], prev[1], line,
+                            stroker_cap_mode(stroker));
+            else
+               stroker_emit_move_to(stroker, line[0], line[1]);
+            memcpy(start_tangent, line,
+                   sizeof(VGfloat) * 4);
+            first = VG_FALSE;
+         } else {
+            create_joins(stroker, prev[0], prev[1], line,
+                         stroker_join_mode(stroker));
+         }
+
+         /* add the stroke for this line */
+         stroker_emit_line_to(stroker, line[2], line[3]);
+         prev[0] = coords[0];
+         prev[1] = coords[1];
+      } else if (cmd == VG_CUBIC_TO_ABS) {
+#if STROKE_DEBUG
+         debug_printf("\n ---> (side) cubicTo [%.2f, %.2f]\n",
+                coords[4],
+                coords[5]);
+#endif
+         struct bezier bezier;
+         int count;
+
+         bezier_init(&bezier,
+                     prev[0], prev[1], coords[0], coords[1],
+                     coords[2], coords[3], coords[4], coords[5]);
+
+         count = bezier_translate_by_normal(&bezier,
+                                            offset_curves,
+                                            MAX_OFFSET,
+                                            offset,
+                                            curve_threshold);
+
+         if (count) {
+            /* if we are starting a new subpath, move to correct starting point */
+            VGfloat tangent[4];
+            VGint i;
+
+            bezier_start_tangent(&bezier, tangent);
+            line_translate(tangent,
+                           offset_curves[0].x1 - bezier.x1,
+                           offset_curves[0].y1 - bezier.y1);
+            if (first) {
+               VGfloat pt[2] = {offset_curves[0].x1,
+                                offset_curves[0].y1};
+
+               if (cap_first) {
+                  create_joins(stroker, prev[0], prev[1], tangent,
+                               stroker_cap_mode(stroker));
+               } else {
+                  stroker_emit_move_to(stroker, pt[0], pt[1]);
+               }
+               start_tangent[0] = tangent[0];
+               start_tangent[1] = tangent[1];
+               start_tangent[2] = tangent[2];
+               start_tangent[3] = tangent[3];
+               first = VG_FALSE;
+            } else {
+               create_joins(stroker, prev[0], prev[1], tangent,
+                            stroker_join_mode(stroker));
+            }
+
+            /* add these beziers */
+            for (i = 0; i < count; ++i) {
+               struct bezier *bez = &offset_curves[i];
+               stroker_emit_curve_to(stroker,
+                                     bez->x2, bez->y2,
+                                     bez->x3, bez->y3,
+                                     bez->x4, bez->y4);
+            }
+         }
+
+         prev[0] = coords[4];
+         prev[1] = coords[5];
+      }
+   }
+
+   if (floatsEqual(start[0], prev[0]) &&
+       floatsEqual(start[1], prev[1])) {
+      /* closed subpath, join first and last point */
+#if STROKE_DEBUG
+      debug_printf("\n stroker: closed subpath\n");
+#endif
+      create_joins(stroker, prev[0], prev[1], start_tangent,
+                   stroker_join_mode(stroker));
+      return VG_TRUE;
+   } else {
+#if STROKE_DEBUG
+      debug_printf("\n stroker: open subpath\n");
+#endif
+      return VG_FALSE;
+   }
+}
+
+static void stroker_process_subpath(struct stroker *stroker)
+{
+   VGboolean fwclosed, bwclosed;
+   VGfloat fw_start_tangent[4], bw_start_tangent[4];
+   struct stroke_iterator fwit;
+   struct stroke_iterator bwit;
+   debug_assert(stroker->segments->num_elements > 0);
+
+   memset(fw_start_tangent, 0,
+          sizeof(VGfloat)*4);
+   memset(bw_start_tangent, 0,
+          sizeof(VGfloat)*4);
+
+   stroke_forward_iterator(&fwit, stroker->segments,
+                           stroker->control_points);
+   stroke_backward_iterator(&bwit, stroker->segments,
+                            stroker->control_points);
+
+   debug_assert(fwit.cmds[0] == VG_MOVE_TO_ABS);
+
+   fwclosed = vg_stroke_outline(&fwit, stroker, VG_FALSE, fw_start_tangent);
+   bwclosed = vg_stroke_outline(&bwit, stroker, !fwclosed, bw_start_tangent);
+
+   if (!bwclosed)
+      create_joins(stroker,
+                   fwit.coords[0], fwit.coords[1], fw_start_tangent,
+                   stroker_cap_mode(stroker));
+   else {
+      /* hack to handle the requirement of the VG spec that says that strokes
+       * of len==0 that have butt cap or round cap still need
+       * to be rendered. (8.7.4 Stroke Generation) */
+      if (stroker->segments->num_elements <= 3) {
+         VGPathCommand cmd;
+         VGfloat data[8], coords[8];
+         struct stroke_iterator *it = &fwit;
+
+         stroke_forward_iterator(it, stroker->segments,
+                                 stroker->control_points);
+         cmd = stroke_itr_command(it);
+         stroke_itr_coords(it, coords);
+         if (cmd != VG_MOVE_TO_ABS) {
+            memset(data, 0, sizeof(VGfloat) * 8);
+            if (!is_segment_null(cmd, coords, data))
+               return;
+         } else {
+            data[0] = coords[0];
+            data[1] = coords[1];
+         }
+         while (it->has_next(it)) {
+            it->next(it);
+            cmd = stroke_itr_command(it);
+            stroke_itr_coords(it, coords);
+            if (!is_segment_null(cmd, coords, data))
+               return;
+         }
+         /* generate the square/round cap */
+         if (stroker->cap_style == VG_CAP_SQUARE) {
+            VGfloat offset = stroker->stroke_width / 2;
+            stroker_emit_move_to(stroker, data[0] - offset,
+                                 data[1] - offset);
+            stroker_emit_line_to(stroker, data[0] + offset,
+                                 data[1] - offset);
+            stroker_emit_line_to(stroker, data[0] + offset,
+                                 data[1] + offset);
+            stroker_emit_line_to(stroker, data[0] - offset,
+                                 data[1] + offset);
+            stroker_emit_line_to(stroker, data[0] - offset,
+                                 data[1] - offset);
+         } else if (stroker->cap_style == VG_CAP_ROUND) {
+            VGfloat offset = stroker->stroke_width / 2;
+            VGfloat cx = data[0], cy = data[1];
+            { /*circle */
+               struct arc arc;
+               struct matrix matrix;
+               matrix_load_identity(&matrix);
+
+               stroker_emit_move_to(stroker, cx + offset, cy);
+               arc_init(&arc, VG_SCCWARC_TO_ABS,
+                        cx + offset, cy,
+                        cx - offset, cy,
+                        offset, offset, 0);
+               arc_stroker_emit(&arc, stroker, &matrix);
+               arc_init(&arc, VG_SCCWARC_TO_ABS,
+                         cx - offset, cy,
+                         cx + offset, cy,
+                         offset, offset, 0);
+               arc_stroker_emit(&arc, stroker, &matrix);
+            }
+         }
+      }
+   }
+}
+
+static INLINE VGfloat dash_pattern(struct dash_stroker *stroker,
+                                   VGint idx)
+{
+   if (stroker->dash_pattern[idx] < 0)
+      return 0.f;
+   return stroker->dash_pattern[idx];
+}
+
+static void dash_stroker_process_subpath(struct stroker *str)
+{
+   struct dash_stroker *stroker = (struct dash_stroker *)str;
+   VGfloat sum_length = 0;
+   VGint i;
+   VGint idash = 0;
+   VGfloat pos = 0;
+   VGfloat elen = 0;
+   VGfloat doffset = stroker->dash_phase;
+   VGfloat estart = 0;
+   VGfloat estop = 0;
+   VGfloat cline[4];
+   struct stroke_iterator it;
+   VGfloat prev[2];
+   VGfloat move_to_pos[2];
+   VGfloat line_to_pos[2];
+
+   VGboolean has_move_to = VG_FALSE;
+
+   stroke_flat_iterator(&it, stroker->base.segments,
+                        stroker->base.control_points);
+
+   stroke_itr_coords(&it, prev);
+   move_to_pos[0] = prev[0];
+   move_to_pos[1] = prev[1];
+
+   debug_assert(stroker->dash_pattern_num > 0);
+
+   for (i = 0; i < stroker->dash_pattern_num; ++i) {
+      sum_length += dash_pattern(stroker, i);
+   }
+
+   if (floatIsZero(sum_length)) {
+      return;
+   }
+
+   doffset -= floorf(doffset / sum_length) * sum_length;
+
+   while (!floatIsZero(doffset) && doffset >= dash_pattern(stroker, idash)) {
+      doffset -= dash_pattern(stroker, idash);
+      idash = (idash + 1) % stroker->dash_pattern_num;
+   }
+
+   while (it.has_next(&it)) {
+      VGPathCommand cmd;
+      VGfloat coords[8];
+      VGboolean done;
+
+      it.next(&it);
+      cmd = stroke_itr_command(&it);
+      stroke_itr_coords(&it, coords);
+
+      debug_assert(cmd == VG_LINE_TO_ABS);
+      cline[0] = prev[0];
+      cline[1] = prev[1];
+      cline[2] = coords[0];
+      cline[3] = coords[1];
+
+      elen = line_lengthv(cline);
+
+      estop = estart + elen;
+
+      done = pos >= estop;
+      while (!done) {
+         VGfloat p2[2];
+
+         VGint idash_incr = 0;
+         VGboolean has_offset = doffset > 0;
+         VGfloat dpos = pos + dash_pattern(stroker, idash) - doffset - estart;
+
+         debug_assert(dpos >= 0);
+
+         if (dpos > elen) { /* dash extends this line */
+            doffset = dash_pattern(stroker, idash) - (dpos - elen);
+            pos = estop;
+            done = VG_TRUE;
+            p2[0] = cline[2];
+            p2[1] = cline[3];
+         } else { /* Dash is on this line */
+            line_point_at(cline, dpos/elen, p2);
+            pos = dpos + estart;
+            done = pos >= estop;
+            idash_incr = 1;
+            doffset = 0;
+         }
+
+         if (idash % 2 == 0) {
+            line_to_pos[0] = p2[0];
+            line_to_pos[1] = p2[1];
+
+            if (!has_offset || !has_move_to) {
+               stroker_move_to(&stroker->stroker, move_to_pos[0], move_to_pos[1]);
+               has_move_to = VG_TRUE;
+            }
+            stroker_line_to(&stroker->stroker, line_to_pos[0], line_to_pos[1]);
+         } else {
+            move_to_pos[0] = p2[0];
+            move_to_pos[1] = p2[1];
+         }
+
+         idash = (idash + idash_incr) % stroker->dash_pattern_num;
+      }
+
+      estart = estop;
+      prev[0] = coords[0];
+      prev[1] = coords[1];
+   }
+
+   if (it.curve_poly) {
+      polygon_destroy(it.curve_poly);
+      it.curve_poly = 0;
+   }
+
+   stroker->base.path = stroker->stroker.path;
+}
+
+static void default_begin(struct stroker *stroker)
+{
+   array_reset(stroker->segments);
+   array_reset(stroker->control_points);
+}
+
+static void default_end(struct stroker *stroker)
+{
+   if (stroker->segments->num_elements > 0)
+      stroker->process_subpath(stroker);
+}
+
+
+static void dash_stroker_begin(struct stroker *stroker)
+{
+   struct dash_stroker *dasher =
+      (struct dash_stroker *)stroker;
+
+   default_begin(&dasher->stroker);
+   default_begin(stroker);
+}
+
+static void dash_stroker_end(struct stroker *stroker)
+{
+   struct dash_stroker *dasher =
+      (struct dash_stroker *)stroker;
+
+   default_end(stroker);
+   default_end(&dasher->stroker);
+}
+
+void stroker_init(struct stroker *stroker,
+                  struct vg_state *state)
+{
+   stroker->stroke_width = state->stroke.line_width.f;
+   stroker->miter_limit = state->stroke.miter_limit.f;
+   stroker->cap_style = state->stroke.cap_style;
+   stroker->join_style = state->stroke.join_style;
+
+   stroker->begin = default_begin;
+   stroker->process_subpath = stroker_process_subpath;
+   stroker->end = default_end;
+
+   stroker->segments = array_create(sizeof(VGubyte));
+   stroker->control_points = array_create(sizeof(VGfloat));
+
+   stroker->back1_x = 0;
+   stroker->back1_y = 0;
+   stroker->back2_x = 0;
+   stroker->back2_y = 0;
+
+   stroker->path = path_create(VG_PATH_DATATYPE_F, 1.0f, 0.0f,
+                               0, 0, VG_PATH_CAPABILITY_ALL);
+
+   stroker->last_cmd = VG_CLOSE_PATH;
+}
+
+void dash_stroker_init(struct stroker *str,
+                       struct vg_state *state)
+{
+   struct dash_stroker *stroker = (struct dash_stroker *)str;
+   int i;
+
+   stroker_init(str, state);
+   stroker_init(&stroker->stroker, state);
+
+   {
+      int real_num = state->stroke.dash_pattern_num;
+      if (real_num % 2)/* if odd, ignore the last one */
+         --real_num;
+      for (i = 0; i < real_num; ++i)
+         stroker->dash_pattern[i] = state->stroke.dash_pattern[i].f;
+      stroker->dash_pattern_num = real_num;
+   }
+
+   stroker->dash_phase = state->stroke.dash_phase.f;
+   stroker->dash_phase_reset = state->stroke.dash_phase_reset;
+
+   stroker->base.begin = dash_stroker_begin;
+   stroker->base.process_subpath = dash_stroker_process_subpath;
+   stroker->base.end = dash_stroker_end;
+   path_destroy(stroker->base.path);
+   stroker->base.path = NULL;
+}
+
+void stroker_begin(struct stroker *stroker)
+{
+   stroker->begin(stroker);
+}
+
+void stroker_end(struct stroker *stroker)
+{
+   stroker->end(stroker);
+}
+
+void stroker_cleanup(struct stroker *stroker)
+{
+   array_destroy(stroker->segments);
+   array_destroy(stroker->control_points);
+}
+
+void dash_stroker_cleanup(struct dash_stroker *stroker)
+{
+   /* if stroker->base.path is null means we never
+    * processed a valid path so delete the temp one
+    * we already created */
+   if (!stroker->base.path)
+      path_destroy(stroker->stroker.path);
+   stroker_cleanup(&stroker->stroker);
+   stroker_cleanup((struct stroker*)stroker);
+}
diff --git a/src/gallium/state_trackers/vega/stroker.h b/src/gallium/state_trackers/vega/stroker.h
new file mode 100644
index 0000000000..36543cd923
--- /dev/null
+++ b/src/gallium/state_trackers/vega/stroker.h
@@ -0,0 +1,89 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STROKER_H
+#define STROKER_H
+
+#include "VG/openvg.h"
+#include "api_consts.h"
+
+struct path;
+struct vg_state;
+struct array;
+
+struct stroker {
+   void (*begin)(struct stroker *stroker);
+   void (*process_subpath)(struct stroker *stroker);
+   void (*end)(struct stroker *stroker);
+
+   struct array *segments;
+   struct array *control_points;
+   struct path *path;
+
+   VGfloat back1_x, back1_y;
+   VGfloat back2_x, back2_y;
+
+   VGfloat     stroke_width;
+   VGfloat     miter_limit;
+   VGCapStyle  cap_style;
+   VGJoinStyle join_style;
+
+   VGPathCommand last_cmd;
+};
+
+struct dash_stroker {
+   struct stroker base;
+
+   struct stroker stroker;
+
+   VGfloat dash_pattern[VEGA_MAX_DASH_COUNT];
+   VGint dash_pattern_num;
+   VGfloat dash_phase;
+   VGboolean dash_phase_reset;
+};
+
+void stroker_init(struct stroker *stroker,
+                  struct vg_state *state);
+void dash_stroker_init(struct stroker *stroker,
+                       struct vg_state *state);
+void dash_stroker_cleanup(struct dash_stroker *stroker);
+void stroker_cleanup(struct stroker *stroker);
+
+void stroker_begin(struct stroker *stroker);
+void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                      VGfloat px2, VGfloat py2,
+                      VGfloat x, VGfloat y);
+void stroker_end(struct stroker *stroker);
+
+void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                           VGfloat px2, VGfloat py2,
+                           VGfloat x, VGfloat y);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/util_array.h b/src/gallium/state_trackers/vega/util_array.h
new file mode 100644
index 0000000000..4343dfe36e
--- /dev/null
+++ b/src/gallium/state_trackers/vega/util_array.h
@@ -0,0 +1,122 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef UTIL_ARRAY_H
+#define UTIL_ARRAY_H
+
+#include "util/u_memory.h"
+
+#define DEFAULT_ARRAY_SIZE 256
+
+struct array {
+   VGint          datatype_size;
+   void          *data;
+   VGint          size;
+   VGint          num_elements;
+};
+
+static INLINE struct array *array_create(VGint datatype_size)
+{
+   struct array *array = CALLOC_STRUCT(array);
+   array->datatype_size = datatype_size;
+
+   array->size = DEFAULT_ARRAY_SIZE;
+   array->data = malloc(array->size * array->datatype_size);
+
+   return array;
+}
+
+
+static INLINE struct array *array_create_size(VGint datatype_size, VGint size)
+{
+   struct array *array = CALLOC_STRUCT(array);
+   array->datatype_size = datatype_size;
+
+   array->size = size;
+   array->data = malloc(array->size * array->datatype_size);
+
+   return array;
+}
+
+static INLINE void array_destroy(struct array *array)
+{
+   if (array)
+      free(array->data);
+   free(array);
+}
+
+static INLINE void array_resize(struct array *array, int num)
+{
+   VGint size = array->datatype_size * num;
+   void *data = malloc(size);
+   memcpy(data, array->data, array->size * array->datatype_size);
+   free(array->data);
+   array->data = data;
+   array->size = num;
+   array->num_elements = (array->num_elements > num) ? num :
+                         array->num_elements;
+}
+
+static INLINE void array_append_data(struct array *array,
+                              const void *data, int num_elements)
+{
+   VGbyte *adata;
+
+   while (array->num_elements + num_elements > array->size) {
+      array_resize(array, (array->num_elements + num_elements) * 1.5);
+   }
+   adata = (VGbyte *)array->data;
+   memcpy(adata + (array->num_elements * array->datatype_size), data,
+          num_elements * array->datatype_size);
+   array->num_elements += num_elements;
+}
+
+static INLINE void array_change_data(struct array *array,
+                              const void *data,
+                              int start_idx,
+                              int num_elements)
+{
+   VGbyte *adata = (VGbyte *)array->data;
+   memcpy(adata + (start_idx * array->datatype_size), data,
+          num_elements * array->datatype_size);
+}
+
+static INLINE void array_remove_element(struct array *array,
+                                        int idx)
+{
+   VGbyte *adata = (VGbyte *)array->data;
+   memmove(adata + (idx * array->datatype_size),
+           adata + ((idx + 1) * array->datatype_size),
+           (array->num_elements - idx - 1) * array->datatype_size);
+   --array->num_elements;
+}
+
+static INLINE void array_reset(struct array *array)
+{
+   array->num_elements = 0;
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_context.c b/src/gallium/state_trackers/vega/vg_context.c
new file mode 100644
index 0000000000..e0ff02f3a9
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_context.c
@@ -0,0 +1,543 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_context.h"
+
+#include "paint.h"
+#include "renderer.h"
+#include "shaders_cache.h"
+#include "shader.h"
+#include "asm_util.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "cso_cache/cso_context.h"
+
+#include "util/u_simple_shaders.h"
+#include "util/u_memory.h"
+#include "util/u_blit.h"
+
+struct vg_context *_vg_context = 0;
+
+struct vg_context * vg_current_context(void)
+{
+   return _vg_context;
+}
+
+static void init_clear(struct vg_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   /* rasterizer state: bypass clipping */
+   memset(&st->clear.raster, 0, sizeof(st->clear.raster));
+   st->clear.raster.gl_rasterization_rules = 1;
+
+   /* fragment shader state: color pass-through program */
+   st->clear.fs =
+      util_make_fragment_passthrough_shader(pipe);
+}
+void vg_set_current_context(struct vg_context *ctx)
+{
+   _vg_context = ctx;
+}
+
+struct vg_context * vg_create_context(struct pipe_context *pipe,
+                                      const void *visual,
+                                      struct vg_context *share)
+{
+   struct vg_context *ctx;
+
+   ctx = CALLOC_STRUCT(vg_context);
+
+   ctx->pipe = pipe;
+
+   vg_init_state(&ctx->state.vg);
+   ctx->state.dirty = ALL_DIRTY;
+
+   ctx->cso_context = cso_create_context(pipe);
+
+   init_clear(ctx);
+
+   ctx->default_paint = paint_create(ctx);
+   ctx->state.vg.stroke_paint = ctx->default_paint;
+   ctx->state.vg.fill_paint = ctx->default_paint;
+
+
+   ctx->mask.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->mask.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->mask.sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->mask.sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->mask.sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->mask.sampler.normalized_coords = 0;
+
+   ctx->blend_sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->blend_sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->blend_sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->blend_sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->blend_sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->blend_sampler.normalized_coords = 0;
+
+   vg_set_error(ctx, VG_NO_ERROR);
+
+   ctx->owned_objects[VG_OBJECT_PAINT] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_IMAGE] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_MASK] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_FONT] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_PATH] = cso_hash_create();
+
+   ctx->renderer = renderer_create(ctx);
+   ctx->sc = shaders_cache_create(ctx);
+   ctx->shader = shader_create(ctx);
+
+   ctx->blit = util_create_blit(ctx->pipe, ctx->cso_context);
+
+   return ctx;
+}
+
+void vg_destroy_context(struct vg_context *ctx)
+{
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   struct pipe_constant_buffer *vsbuf = &ctx->vs_const_buffer;
+
+   util_destroy_blit(ctx->blit);
+   renderer_destroy(ctx->renderer);
+   shaders_cache_destroy(ctx->sc);
+   shader_destroy(ctx->shader);
+   paint_destroy(ctx->default_paint);
+
+   if (cbuf && cbuf->buffer)
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   if (vsbuf && vsbuf->buffer)
+      pipe_buffer_reference(&vsbuf->buffer, NULL);
+
+   if (ctx->clear.fs) {
+      cso_delete_fragment_shader(ctx->cso_context, ctx->clear.fs);
+      ctx->clear.fs = NULL;
+   }
+
+   if (ctx->plain_vs) {
+      vg_shader_destroy(ctx, ctx->plain_vs);
+      ctx->plain_vs = NULL;
+   }
+   if (ctx->clear_vs) {
+      vg_shader_destroy(ctx, ctx->clear_vs);
+      ctx->clear_vs = NULL;
+   }
+   if (ctx->texture_vs) {
+      vg_shader_destroy(ctx, ctx->texture_vs);
+      ctx->texture_vs = NULL;
+   }
+
+   if (ctx->pass_through_depth_fs)
+      vg_shader_destroy(ctx, ctx->pass_through_depth_fs);
+   if (ctx->mask.union_fs)
+      vg_shader_destroy(ctx, ctx->mask.union_fs);
+   if (ctx->mask.intersect_fs)
+      vg_shader_destroy(ctx, ctx->mask.intersect_fs);
+   if (ctx->mask.subtract_fs)
+      vg_shader_destroy(ctx, ctx->mask.subtract_fs);
+   if (ctx->mask.set_fs)
+      vg_shader_destroy(ctx, ctx->mask.set_fs);
+
+   cso_release_all(ctx->cso_context);
+   cso_destroy_context(ctx->cso_context);
+
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_PAINT]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_IMAGE]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_MASK]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_FONT]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_PATH]);
+
+   free(ctx);
+}
+
+void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type)
+{
+   obj->type = type;
+   obj->ctx = ctx;
+}
+
+VGboolean vg_context_is_object_valid(struct vg_context *ctx,
+                                enum vg_object_type type,
+                                void *ptr)
+{
+    if (ctx) {
+       struct cso_hash *hash = ctx->owned_objects[type];
+       if (!hash)
+          return VG_FALSE;
+       return cso_hash_contains(hash, (unsigned)(long)ptr);
+    }
+    return VG_FALSE;
+}
+
+void vg_context_add_object(struct vg_context *ctx,
+                           enum vg_object_type type,
+                           void *ptr)
+{
+    if (ctx) {
+       struct cso_hash *hash = ctx->owned_objects[type];
+       if (!hash)
+          return;
+       cso_hash_insert(hash, (unsigned)(long)ptr, ptr);
+    }
+}
+
+void vg_context_remove_object(struct vg_context *ctx,
+                              enum vg_object_type type,
+                              void *ptr)
+{
+   if (ctx) {
+      struct cso_hash *hash = ctx->owned_objects[type];
+      if (!hash)
+         return;
+      cso_hash_take(hash, (unsigned)(long)ptr);
+   }
+}
+
+static void update_clip_state(struct vg_context *ctx)
+{
+   struct pipe_depth_stencil_alpha_state *dsa = &ctx->state.g3d.dsa;
+   struct vg_state *state =  &ctx->state.vg;
+
+   memset(dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   if (state->scissoring) {
+      struct pipe_blend_state *blend = &ctx->state.g3d.blend;
+      struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+      dsa->depth.writemask = 1;/*glDepthMask(TRUE);*/
+      dsa->depth.func = PIPE_FUNC_ALWAYS;
+      dsa->depth.enabled = 1;
+
+      cso_save_blend(ctx->cso_context);
+      cso_save_fragment_shader(ctx->cso_context);
+      /* set a passthrough shader */
+      if (!ctx->pass_through_depth_fs)
+         ctx->pass_through_depth_fs = shader_create_from_text(ctx->pipe,
+                                                              pass_through_depth_asm,
+                                                              40,
+                                                              PIPE_SHADER_FRAGMENT);
+      cso_set_fragment_shader_handle(ctx->cso_context,
+                                     ctx->pass_through_depth_fs->driver);
+      cso_set_depth_stencil_alpha(ctx->cso_context, dsa);
+
+      ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL, NULL, 1.0, 0);
+
+      /* disable color writes */
+      blend->colormask = 0; /*disable colorwrites*/
+      cso_set_blend(ctx->cso_context, blend);
+
+      /* enable scissoring */
+      for (int i = 0; i < state->scissor_rects_num; ++i) {
+         const float x      = state->scissor_rects[i * 4 + 0].f;
+         const float y      = state->scissor_rects[i * 4 + 1].f;
+         const float width  = state->scissor_rects[i * 4 + 2].f;
+         const float height = state->scissor_rects[i * 4 + 3].f;
+         VGfloat minx, miny, maxx, maxy;
+
+         minx = 0;
+         miny = 0;
+         maxx = fb->width;
+         maxy = fb->height;
+
+         if (x > minx)
+            minx = x;
+         if (y > miny)
+            miny = y;
+
+         if (x + width < maxx)
+            maxx = x + width;
+         if (y + height < maxy)
+            maxy = y + height;
+
+         /* check for null space */
+         if (minx >= maxx || miny >= maxy)
+            minx = miny = maxx = maxy = 0;
+
+         /*glClear(GL_DEPTH_BUFFER_BIT);*/
+         renderer_draw_quad(ctx->renderer, minx, miny, maxx, maxy, 0.0f);
+      }
+
+      blend->colormask = 1; /*enable colorwrites*/
+      cso_restore_blend(ctx->cso_context);
+      cso_restore_fragment_shader(ctx->cso_context);
+
+      dsa->depth.enabled = 1; /* glEnable(GL_DEPTH_TEST); */
+      dsa->depth.writemask = 0;/*glDepthMask(FALSE);*/
+      dsa->depth.func = PIPE_FUNC_GEQUAL;
+   }
+}
+
+void vg_validate_state(struct vg_context *ctx)
+{
+   if ((ctx->state.dirty & BLEND_DIRTY)) {
+      struct pipe_blend_state *blend = &ctx->state.g3d.blend;
+      memset(blend, 0, sizeof(struct pipe_blend_state));
+      blend->blend_enable = 1;
+      blend->colormask |= PIPE_MASK_R;
+      blend->colormask |= PIPE_MASK_G;
+      blend->colormask |= PIPE_MASK_B;
+      blend->colormask |= PIPE_MASK_A;
+
+      switch (ctx->state.vg.blend_mode) {
+      case VG_BLEND_SRC:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_SRC_OVER:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_SRC_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+         break;
+      case VG_BLEND_DST_OVER:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_INV_DST_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_INV_DST_ALPHA;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_DST_ALPHA;
+         break;
+      case VG_BLEND_SRC_IN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_DST_IN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ZERO;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_SRC_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_SRC_ALPHA;
+         break;
+      case VG_BLEND_MULTIPLY:
+      case VG_BLEND_SCREEN:
+      case VG_BLEND_DARKEN:
+      case VG_BLEND_LIGHTEN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_ADDITIVE:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
+         break;
+      default:
+         assert(!"not implemented blend mode");
+      }
+      cso_set_blend(ctx->cso_context, &ctx->state.g3d.blend);
+   }
+   if ((ctx->state.dirty & RASTERIZER_DIRTY)) {
+      struct pipe_rasterizer_state *raster = &ctx->state.g3d.rasterizer;
+      memset(raster, 0, sizeof(struct pipe_rasterizer_state));
+      raster->gl_rasterization_rules = 1;
+      cso_set_rasterizer(ctx->cso_context, &ctx->state.g3d.rasterizer);
+   }
+   if ((ctx->state.dirty & VIEWPORT_DIRTY)) {
+      struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+      const VGint param_bytes = 8 * sizeof(VGfloat);
+      VGfloat vs_consts[8] = {
+         2.f/fb->width, 2.f/fb->height, 1, 1,
+         -1, -1, 0, 0
+      };
+      struct pipe_constant_buffer *cbuf = &ctx->vs_const_buffer;
+
+      vg_set_viewport(ctx, VEGA_Y0_BOTTOM);
+
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+      cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 16,
+                                        PIPE_BUFFER_USAGE_CONSTANT,
+                                        param_bytes);
+
+      if (cbuf->buffer) {
+         st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                       0, param_bytes, vs_consts);
+      }
+      ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_VERTEX, 0, cbuf);
+   }
+   if ((ctx->state.dirty & VS_DIRTY)) {
+      cso_set_vertex_shader_handle(ctx->cso_context,
+                                   vg_plain_vs(ctx));
+   }
+
+   /* must be last because it renders to the depth buffer*/
+   if ((ctx->state.dirty & DEPTH_STENCIL_DIRTY)) {
+      update_clip_state(ctx);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &ctx->state.g3d.dsa);
+   }
+
+   shader_set_masking(ctx->shader, ctx->state.vg.masking);
+   shader_set_image_mode(ctx->shader, ctx->state.vg.image_mode);
+
+   ctx->state.dirty = NONE_DIRTY;
+}
+
+VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type)
+{
+   struct vg_object *obj = ptr;
+   if (ptr && is_aligned(obj) && obj->type == type)
+      return VG_TRUE;
+   else
+      return VG_FALSE;
+}
+
+void vg_set_error(struct vg_context *ctx,
+                  VGErrorCode code)
+{
+   /*vgGetError returns the oldest error code provided by
+    * an API call on the current context since the previous
+    * call to vgGetError on that context (or since the creation
+    of the context).*/
+   if (ctx->_error == VG_NO_ERROR)
+      ctx->_error = code;
+}
+
+void vg_prepare_blend_surface(struct vg_context *ctx)
+{
+   struct pipe_surface *dest_surface = NULL;
+   struct pipe_context *pipe = ctx->pipe;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+
+   /* first finish all pending rendering */
+   vgFinish();
+
+   dest_surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                stfb->blend_texture,
+                                                0, 0, 0,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+   /* flip it, because we want to use it as a sampler */
+   util_blit_pixels_tex(ctx->blit,
+                        strb->texture,
+                        0, strb->height,
+                        strb->width, 0,
+                        dest_surface,
+                        0, 0,
+                        strb->width, strb->height,
+                        0.0, PIPE_TEX_MIPFILTER_NEAREST);
+
+   if (dest_surface)
+      pipe_surface_reference(&dest_surface, NULL);
+
+   /* make sure it's complete */
+   vgFinish();
+}
+
+
+void vg_prepare_blend_surface_from_mask(struct vg_context *ctx)
+{
+   struct pipe_surface *dest_surface = NULL;
+   struct pipe_context *pipe = ctx->pipe;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+
+   vg_validate_state(ctx);
+
+   /* first finish all pending rendering */
+   vgFinish();
+
+   dest_surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                stfb->blend_texture,
+                                                0, 0, 0,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* flip it, because we want to use it as a sampler */
+   util_blit_pixels_tex(ctx->blit,
+                        stfb->alpha_mask,
+                        0, strb->height,
+                        strb->width, 0,
+                        dest_surface,
+                        0, 0,
+                        strb->width, strb->height,
+                        0.0, PIPE_TEX_MIPFILTER_NEAREST);
+
+   /* make sure it's complete */
+   vgFinish();
+
+   if (dest_surface)
+      pipe_surface_reference(&dest_surface, NULL);
+}
+
+void * vg_plain_vs(struct vg_context *ctx)
+{
+   if (!ctx->plain_vs) {
+      ctx->plain_vs = shader_create_from_text(ctx->pipe,
+                                              vs_plain_asm,
+                                              200,
+                                              PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->plain_vs->driver;
+}
+
+
+void * vg_clear_vs(struct vg_context *ctx)
+{
+   if (!ctx->clear_vs) {
+      ctx->clear_vs = shader_create_from_text(ctx->pipe,
+                                              vs_clear_asm,
+                                              200,
+                                              PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->clear_vs->driver;
+}
+
+void * vg_texture_vs(struct vg_context *ctx)
+{
+   if (!ctx->texture_vs) {
+      ctx->texture_vs = shader_create_from_text(ctx->pipe,
+                                                vs_texture_asm,
+                                                200,
+                                                PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->texture_vs->driver;
+}
+
+void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation)
+{
+   struct pipe_viewport_state viewport;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   VGfloat y_scale = (orientation == VEGA_Y0_BOTTOM) ? -2.f : 2.f;
+
+   viewport.scale[0] =  fb->width / 2.f;
+   viewport.scale[1] =  fb->height / y_scale;
+   viewport.scale[2] =  1.0;
+   viewport.scale[3] =  1.0;
+   viewport.translate[0] = fb->width / 2.f;
+   viewport.translate[1] = fb->height / 2.f;
+   viewport.translate[2] = 0.0;
+   viewport.translate[3] = 0.0;
+
+   cso_set_viewport(ctx->cso_context, &viewport);
+}
diff --git a/src/gallium/state_trackers/vega/vg_context.h b/src/gallium/state_trackers/vega/vg_context.h
new file mode 100644
index 0000000000..ccc8889c8c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_context.h
@@ -0,0 +1,292 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_CONTEXT_H
+#define VG_CONTEXT_H
+
+#include "vg_state.h"
+
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "util/u_pointer.h"
+#include "util/u_math.h"
+
+#include "cso_cache/cso_hash.h"
+#include "cso_cache/cso_context.h"
+
+struct renderer;
+struct shaders_cache;
+struct shader;
+struct vg_shader;
+
+struct st_renderbuffer {
+   enum pipe_format   format;
+   struct pipe_surface *surface;
+   struct pipe_texture *texture;
+   VGint width, height;
+};
+
+struct st_framebuffer {
+   VGint init_width, init_height;
+   struct st_renderbuffer *strb;
+   struct st_renderbuffer *dsrb;
+
+   struct pipe_texture *alpha_mask;
+
+   struct pipe_texture *blend_texture;
+
+   void *privateData;
+};
+
+enum vg_object_type {
+   VG_OBJECT_UNKNOWN = 0,
+   VG_OBJECT_PAINT,
+   VG_OBJECT_IMAGE,
+   VG_OBJECT_MASK,
+   VG_OBJECT_FONT,
+   VG_OBJECT_PATH,
+
+   VG_OBJECT_LAST
+};
+enum dirty_state {
+   NONE_DIRTY          = 0<<0,
+   BLEND_DIRTY         = 1<<1,
+   RASTERIZER_DIRTY    = 1<<2,
+   VIEWPORT_DIRTY      = 1<<3,
+   VS_DIRTY            = 1<<4,
+   DEPTH_STENCIL_DIRTY = 1<<5,
+   ALL_DIRTY           = BLEND_DIRTY | RASTERIZER_DIRTY |
+   VIEWPORT_DIRTY | VS_DIRTY | DEPTH_STENCIL_DIRTY
+};
+
+struct vg_context
+{
+   struct pipe_context *pipe;
+
+   struct {
+      struct vg_state vg;
+      struct {
+         struct pipe_blend_state blend;
+         struct pipe_rasterizer_state rasterizer;
+         struct pipe_shader_state vs_state;
+         struct pipe_depth_stencil_alpha_state dsa;
+         struct pipe_framebuffer_state fb;
+      } g3d;
+      VGbitfield dirty;
+   } state;
+
+   VGErrorCode _error;
+
+   struct st_framebuffer *draw_buffer;
+
+   struct cso_hash *owned_objects[VG_OBJECT_LAST];
+
+   struct {
+      struct pipe_shader_state vert_shader;
+      struct pipe_shader_state frag_shader;
+      struct pipe_rasterizer_state raster;
+      void *fs;
+      float vertices[4][2][4];  /**< vertex pos + color */
+   } clear;
+
+   struct {
+      struct pipe_constant_buffer cbuf;
+      struct pipe_sampler_state sampler;
+
+      struct vg_shader *union_fs;
+      struct vg_shader *intersect_fs;
+      struct vg_shader *subtract_fs;
+      struct vg_shader *set_fs;
+   } mask;
+
+   struct vg_shader *pass_through_depth_fs;
+
+   struct cso_context *cso_context;
+
+   struct pipe_buffer *stencil_quad;
+   VGfloat stencil_vertices[4][2][4];
+
+   struct renderer *renderer;
+   struct shaders_cache *sc;
+   struct shader *shader;
+
+   struct pipe_sampler_state blend_sampler;
+   struct {
+      struct pipe_constant_buffer buffer;
+      void *color_matrix_fs;
+   } filter;
+   struct vg_paint *default_paint;
+
+   struct blit_state *blit;
+
+   struct vg_shader *plain_vs;
+   struct vg_shader *clear_vs;
+   struct vg_shader *texture_vs;
+   struct pipe_constant_buffer vs_const_buffer;
+};
+
+struct vg_object {
+   enum vg_object_type type;
+   struct vg_context *ctx;
+};
+void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type);
+VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type);
+
+struct vg_context *vg_create_context(struct pipe_context *pipe,
+                                     const void *visual,
+                                     struct vg_context *share);
+void vg_destroy_context(struct vg_context *ctx);
+struct vg_context *vg_current_context(void);
+void vg_set_current_context(struct vg_context *ctx);
+
+VGboolean vg_context_is_object_valid(struct vg_context *ctx,
+                                     enum vg_object_type type,
+                                     void *ptr);
+void vg_context_add_object(struct vg_context *ctx,
+                           enum vg_object_type type,
+                           void *ptr);
+void vg_context_remove_object(struct vg_context *ctx,
+                              enum vg_object_type type,
+                              void *ptr);
+
+void vg_validate_state(struct vg_context *ctx);
+
+void vg_set_error(struct vg_context *ctx,
+                  VGErrorCode code);
+
+void vg_prepare_blend_surface(struct vg_context *ctx);
+void vg_prepare_blend_surface_from_mask(struct vg_context *ctx);
+
+
+static INLINE VGboolean is_aligned_to(const void *ptr, VGbyte alignment)
+{
+   void *aligned = align_pointer(ptr, alignment);
+   return (ptr == aligned) ? VG_TRUE : VG_FALSE;
+}
+
+static INLINE VGboolean is_aligned(const void *ptr)
+{
+   return is_aligned_to(ptr, 4);
+}
+
+static INLINE void vg_shift_rectx(VGfloat coords[4],
+                                 const VGfloat *bounds,
+                                 const VGfloat shift)
+{
+   coords[0] += shift;
+   coords[2] -= shift;
+   if (bounds) {
+      coords[2] = MIN2(coords[2], bounds[2]);
+      /* bound x/y + width/height */
+      if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+         coords[2] = (bounds[0] + bounds[2]) - coords[0];
+      }
+   }
+}
+
+static INLINE void vg_shift_recty(VGfloat coords[4],
+                                 const VGfloat *bounds,
+                                 const VGfloat shift)
+{
+   coords[1] += shift;
+   coords[3] -= shift;
+   if (bounds) {
+      coords[3] = MIN2(coords[3], bounds[3]);
+      if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+         coords[3] = (bounds[1] + bounds[3]) - coords[1];
+      }
+   }
+}
+
+static INLINE void vg_bound_rect(VGfloat coords[4],
+                                 const VGfloat bounds[4],
+                                 VGfloat shift[4])
+{
+   /* if outside the bounds */
+   if (coords[0] > (bounds[0] + bounds[2]) ||
+       coords[1] > (bounds[1] + bounds[3]) ||
+       (coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      shift[0] = 0.f;
+      shift[1] = 0.f;
+      return;
+   }
+
+   /* bound x */
+   if (coords[0] < bounds[0]) {
+      shift[0] = bounds[0] - coords[0];
+      coords[2] -= shift[0];
+      coords[0] = bounds[0];
+   } else
+      shift[0] = 0.f;
+
+   /* bound y */
+   if (coords[1] < bounds[1]) {
+      shift[1] = bounds[1] - coords[1];
+      coords[3] -= shift[1];
+      coords[1] = bounds[1];
+   } else
+      shift[1] = 0.f;
+
+   shift[2] = bounds[2] - coords[2];
+   shift[3] = bounds[3] - coords[3];
+   /* bound width/height */
+   coords[2] = MIN2(coords[2], bounds[2]);
+   coords[3] = MIN2(coords[3], bounds[3]);
+
+   /* bound x/y + width/height */
+   if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+      coords[2] = (bounds[0] + bounds[2]) - coords[0];
+   }
+   if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+      coords[3] = (bounds[1] + bounds[3]) - coords[1];
+   }
+
+   /* if outside the bounds */
+   if ((coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      return;
+   }
+}
+
+void *vg_plain_vs(struct vg_context *ctx);
+void *vg_clear_vs(struct vg_context *ctx);
+void *vg_texture_vs(struct vg_context *ctx);
+typedef enum {
+   VEGA_Y0_TOP,
+   VEGA_Y0_BOTTOM
+} VegaOrientation;
+void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_state.c b/src/gallium/state_trackers/vega/vg_state.c
new file mode 100644
index 0000000000..6f6bfdaf7a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_state.c
@@ -0,0 +1,124 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_state.h"
+
+#include <string.h>
+
+void vg_init_state(struct vg_state *state)
+{
+   state->matrix_mode = VG_MATRIX_PATH_USER_TO_SURFACE;
+   state->fill_rule = VG_EVEN_ODD;
+   state->image_quality = VG_IMAGE_QUALITY_FASTER;
+   state->rendering_quality = VG_RENDERING_QUALITY_BETTER;
+   state->blend_mode = VG_BLEND_SRC_OVER;
+   state->image_mode = VG_DRAW_IMAGE_NORMAL;
+
+   memset(state->scissor_rects, 0, sizeof(state->scissor_rects));
+   state->scissor_rects_num = 0;
+
+   state->color_transform = VG_FALSE;
+   state->color_transform_values[0] = 1.0f;
+   state->color_transform_values[1] = 1.0f;
+   state->color_transform_values[2] = 1.0f;
+   state->color_transform_values[3] = 1.0f;
+   state->color_transform_values[4] = 0.0f;
+   state->color_transform_values[5] = 0.0f;
+   state->color_transform_values[6] = 0.0f;
+   state->color_transform_values[7] = 0.0f;
+
+   /* Stroke parameters */
+   state->stroke.line_width.f       = 1.0f;
+   state->stroke.line_width.i       = 1;
+   state->stroke.cap_style        = VG_CAP_BUTT;
+   state->stroke.join_style       = VG_JOIN_MITER;
+   state->stroke.miter_limit.f      = 4.0f;
+   state->stroke.miter_limit.i      = 4;
+   state->stroke.dash_pattern_num = 0;
+   state->stroke.dash_phase.f       = 0.0f;
+   state->stroke.dash_phase.i       = 0;
+   state->stroke.dash_phase_reset = VG_FALSE;
+
+   /* Edge fill color for VG_TILE_FILL tiling mode */
+   state->tile_fill_color[0] = 0.0f;
+   state->tile_fill_color[1] = 0.0f;
+   state->tile_fill_color[2] = 0.0f;
+   state->tile_fill_color[3] = 0.0f;
+
+   /* Color for vgClear */
+   state->clear_color[0] = 0.0f;
+   state->clear_color[1] = 0.0f;
+   state->clear_color[2] = 0.0f;
+   state->clear_color[3] = 0.0f;
+
+   /* Glyph origin */
+   state->glyph_origin[0].f = 0.0f;
+   state->glyph_origin[1].f = 0.0f;
+   state->glyph_origin[0].i = 0;
+   state->glyph_origin[1].i = 0;
+
+   /* Enable/disable alpha masking and scissoring */
+   state->masking = VG_FALSE;
+   state->scissoring = VG_FALSE;
+
+   /* Pixel layout information */
+   state->pixel_layout = VG_PIXEL_LAYOUT_UNKNOWN;
+   state->screen_layout = VG_PIXEL_LAYOUT_UNKNOWN;
+
+   /* Source format selection for image filters */
+   state->filter_format_linear = VG_FALSE;
+   state->filter_format_premultiplied = VG_FALSE;
+
+   /* Destination write enable mask for image filters */
+   state->filter_channel_mask = (VG_RED | VG_GREEN | VG_BLUE | VG_ALPHA);
+
+   matrix_load_identity(&state->path_user_to_surface_matrix);
+   matrix_load_identity(&state->image_user_to_surface_matrix);
+   matrix_load_identity(&state->fill_paint_to_user_matrix);
+   matrix_load_identity(&state->stroke_paint_to_user_matrix);
+   matrix_load_identity(&state->glyph_user_to_surface_matrix);
+}
+
+struct matrix *vg_state_matrix(struct vg_state *state)
+{
+    switch(state->matrix_mode) {
+    case VG_MATRIX_PATH_USER_TO_SURFACE:
+       return &state->path_user_to_surface_matrix;
+    case VG_MATRIX_IMAGE_USER_TO_SURFACE:
+       return &state->image_user_to_surface_matrix;
+    case VG_MATRIX_FILL_PAINT_TO_USER:
+       return &state->fill_paint_to_user_matrix;
+    case VG_MATRIX_STROKE_PAINT_TO_USER:
+       return &state->stroke_paint_to_user_matrix;
+#ifdef OPENVG_VERSION_1_1
+    case VG_MATRIX_GLYPH_USER_TO_SURFACE:
+       return &state->glyph_user_to_surface_matrix;
+#endif
+    default:
+       break;
+    }
+    return NULL;
+}
diff --git a/src/gallium/state_trackers/vega/vg_state.h b/src/gallium/state_trackers/vega/vg_state.h
new file mode 100644
index 0000000000..ed90689f91
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_state.h
@@ -0,0 +1,109 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_STATE_H
+#define VG_STATE_H
+
+#include "VG/openvg.h"
+
+#include "api_consts.h"
+#include "matrix.h"
+
+struct vg_value
+{
+   VGfloat f;
+   VGint   i;
+};
+
+struct vg_state {
+   /* Mode settings */
+   VGMatrixMode matrix_mode;
+   VGFillRule fill_rule;
+   VGImageQuality image_quality;
+   VGRenderingQuality rendering_quality;
+   VGBlendMode blend_mode;
+   VGImageMode image_mode;
+
+   /* Scissoring rectangles */
+   struct vg_value  scissor_rects[32*4];
+   VGint  scissor_rects_num;
+
+   /* Color Transformation */
+   VGboolean color_transform;
+   VGfloat color_transform_values[8];
+
+   /* Stroke parameters */
+   struct {
+      struct vg_value line_width;
+      VGCapStyle cap_style;
+      VGJoinStyle join_style;
+      struct vg_value miter_limit;
+      struct vg_value dash_pattern[VEGA_MAX_DASH_COUNT];
+      VGint   dash_pattern_num;
+      struct vg_value dash_phase;
+      VGboolean dash_phase_reset;
+   } stroke;
+
+   /* Edge fill color for VG_TILE_FILL tiling mode */
+   VGfloat tile_fill_color[4];
+   VGint tile_fill_colori[4];
+
+   /* Color for vgClear */
+   VGfloat clear_color[4];
+   VGint clear_colori[4];
+
+   /* Glyph origin */
+   struct vg_value glyph_origin[2];
+
+   /* Enable/disable alpha masking and scissoring */
+   VGboolean masking;
+   VGboolean scissoring;
+
+   /* Pixel layout information */
+   VGPixelLayout pixel_layout;
+   VGPixelLayout screen_layout;
+
+   /* Source format selection for image filters */
+   VGboolean filter_format_linear;
+   VGboolean filter_format_premultiplied;
+
+   /* Destination write enable mask for image filters */
+   VGbitfield filter_channel_mask;
+
+   struct matrix path_user_to_surface_matrix;
+   struct matrix image_user_to_surface_matrix;
+   struct matrix fill_paint_to_user_matrix;
+   struct matrix stroke_paint_to_user_matrix;
+   struct matrix glyph_user_to_surface_matrix;
+
+   struct vg_paint *stroke_paint;
+   struct vg_paint *fill_paint;
+};
+
+void vg_init_state(struct vg_state *state);
+struct matrix * vg_state_matrix(struct vg_state *state);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_tracker.c b/src/gallium/state_trackers/vega/vg_tracker.c
new file mode 100644
index 0000000000..56cc60aebe
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_tracker.c
@@ -0,0 +1,416 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_context.h"
+#include "vg_tracker.h"
+#include "mask.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+static struct pipe_texture *
+create_texture(struct pipe_context *pipe, enum pipe_format format,
+               VGint width, VGint height)
+{
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+
+   if (format != PIPE_FORMAT_NONE) {
+      templ.format = format;
+   }
+   else {
+      templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   }
+
+   templ.target = PIPE_TEXTURE_2D;
+   pf_get_block(templ.format, &templ.block);
+   templ.width[0] = width;
+   templ.height[0] = height;
+   templ.depth[0] = 1;
+   templ.last_level = 0;
+
+   if (pf_get_component_bits(format, PIPE_FORMAT_COMP_S)) {
+      templ.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   } else {
+      templ.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                         PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                         PIPE_TEXTURE_USAGE_SAMPLER);
+   }
+
+   return pipe->screen->texture_create(pipe->screen, &templ);
+}
+
+/**
+ * Allocate a renderbuffer for a an on-screen window (not a user-created
+ * renderbuffer).  The window system code determines the format.
+ */
+static struct st_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format)
+{
+   struct st_renderbuffer *strb;
+
+   strb = CALLOC_STRUCT(st_renderbuffer);
+   if (!strb) {
+      /*_vega_error(NULL, VG_OUT_OF_MEMORY, "creating renderbuffer");*/
+      return NULL;
+   }
+
+   strb->format = format;
+
+   return strb;
+}
+
+
+/**
+ * This is called to allocate the original drawing surface, and
+ * during window resize.
+ */
+static VGboolean
+st_renderbuffer_alloc_storage(struct vg_context * ctx,
+                              struct st_renderbuffer *strb,
+                              VGuint width, VGuint height)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   unsigned surface_usage;
+
+   /* Free the old surface and texture
+    */
+   pipe_surface_reference(&strb->surface, NULL);
+   pipe_texture_reference(&strb->texture, NULL);
+
+
+   /* Probably need dedicated flags for surface usage too:
+    */
+   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ  |
+                    PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   strb->texture = create_texture(pipe, strb->format,
+                                  width, height);
+
+   if (!strb->texture)
+      return FALSE;
+
+   strb->surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                 strb->texture,
+                                                 0, 0, 0,
+                                                 surface_usage);
+   strb->width = width;
+   strb->height = height;
+
+   assert(strb->surface->width == width);
+   assert(strb->surface->height == height);
+
+   return strb->surface != NULL;
+}
+
+struct vg_context * st_create_context(struct pipe_context *pipe,
+                                      const void *visual,
+                                      struct vg_context *share)
+{
+   struct vg_context *ctx = vg_create_context(pipe, visual, share);
+   /*debug_printf("--------- CREATE CONTEXT %p\n", ctx);*/
+   return ctx;
+}
+
+void st_destroy_context(struct vg_context *st)
+{
+   /*debug_printf("--------- DESTROY CONTEXT %p\n", st);*/
+   vg_destroy_context(st);
+}
+
+void st_copy_context_state(struct vg_context *dst, struct vg_context *src,
+                           uint mask)
+{
+   fprintf(stderr, "FIXME: %s\n", __FUNCTION__);
+}
+
+void st_get_framebuffer_dimensions(struct st_framebuffer *stfb,
+				   uint *width,
+				   uint *height)
+{
+   *width = stfb->strb->width;
+   *height = stfb->strb->height;
+}
+
+struct st_framebuffer * st_create_framebuffer(const void *visual,
+                                              enum pipe_format colorFormat,
+                                              enum pipe_format depthFormat,
+                                              enum pipe_format stencilFormat,
+                                              uint width, uint height,
+                                              void *privateData)
+{
+   struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer);
+   if (stfb) {
+      struct st_renderbuffer *rb =
+         st_new_renderbuffer_fb(colorFormat);
+      stfb->strb = rb;
+#if 0
+      if (doubleBuffer) {
+         struct st_renderbuffer *rb =
+            st_new_renderbuffer_fb(colorFormat);
+      }
+#endif
+
+      /* we want to combine the depth/stencil */
+      if (stencilFormat == depthFormat)
+         stfb->dsrb = st_new_renderbuffer_fb(stencilFormat);
+      else
+         stfb->dsrb = st_new_renderbuffer_fb(PIPE_FORMAT_S8Z24_UNORM);
+
+      /*### currently we always allocate it but it's possible it's
+        not necessary if EGL_ALPHA_MASK_SIZE was 0
+      */
+      stfb->alpha_mask = 0;
+
+      stfb->init_width = width;
+      stfb->init_height = height;
+      stfb->privateData = privateData;
+   }
+
+   return stfb;
+}
+
+static void setup_new_alpha_mask(struct vg_context *ctx,
+                                 struct st_framebuffer *stfb,
+                                 uint width, uint height)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_texture *old_texture = stfb->alpha_mask;
+
+   /*
+     we use PIPE_FORMAT_A8R8G8B8_UNORM because we want to render to
+     this texture and use it as a sampler, so while this wastes some
+     space it makes both of those a lot simpler
+   */
+   stfb->alpha_mask =
+      create_texture(pipe, PIPE_FORMAT_A8R8G8B8_UNORM, width, height);
+
+   if (!stfb->alpha_mask) {
+      if (old_texture)
+         pipe_texture_reference(&old_texture, NULL);
+      return;
+   }
+
+   vg_validate_state(ctx);
+
+   /* alpha mask starts with 1.f alpha */
+   mask_fill(0, 0, width, height, 1.f);
+
+   /* if we had an old surface copy it over */
+   if (old_texture) {
+      struct pipe_surface *surface = pipe->screen->get_tex_surface(
+         pipe->screen,
+         stfb->alpha_mask,
+         0, 0, 0,
+         PIPE_BUFFER_USAGE_GPU_WRITE);
+      struct pipe_surface *old_surface = pipe->screen->get_tex_surface(
+         pipe->screen,
+         old_texture,
+         0, 0, 0,
+         PIPE_BUFFER_USAGE_GPU_READ);
+      pipe->surface_copy(pipe,
+                         surface,
+                         0, 0,
+                         old_surface,
+                         0, 0,
+                         MIN2(old_surface->width, width),
+                         MIN2(old_surface->height, height));
+      if (surface)
+         pipe_surface_reference(&surface, NULL);
+      if (old_surface)
+         pipe_surface_reference(&old_surface, NULL);
+   }
+
+   /* Free the old texture
+    */
+   if (old_texture)
+      pipe_texture_reference(&old_texture, NULL);
+}
+
+void st_resize_framebuffer(struct st_framebuffer *stfb,
+                           uint width, uint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct st_renderbuffer *strb = stfb->strb;
+   struct pipe_framebuffer_state *state;
+
+   if (!ctx)
+      return;
+
+   state = &ctx->state.g3d.fb;
+
+   /* If this is a noop, exit early and don't do the clear, etc below.
+    */
+   if (strb->width == width &&
+       strb->height == height &&
+       state->zsbuf)
+      return;
+
+   if (strb->width != width || strb->height != height)
+      st_renderbuffer_alloc_storage(ctx, strb,
+                                 width, height);
+
+   if (stfb->dsrb->width != width || stfb->dsrb->height != height)
+      st_renderbuffer_alloc_storage(ctx, stfb->dsrb,
+                                 width, height);
+
+   {
+      VGuint i;
+
+      memset(state, 0, sizeof(struct pipe_framebuffer_state));
+
+      state->width  = width;
+      state->height = height;
+
+      state->nr_cbufs = 1;
+      state->cbufs[0] = strb->surface;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         state->cbufs[i] = 0;
+
+      state->zsbuf = stfb->dsrb->surface;
+
+      cso_set_framebuffer(ctx->cso_context, state);
+   }
+
+   ctx->state.dirty |= VIEWPORT_DIRTY;
+   ctx->state.dirty |= DEPTH_STENCIL_DIRTY;/*to reset the scissors*/
+
+   ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL,
+                    NULL, 0.0, 0);
+
+   /* we need all the other state already set */
+
+   setup_new_alpha_mask(ctx, stfb, width, height);
+
+   pipe_texture_reference( &stfb->blend_texture, NULL );
+   stfb->blend_texture = create_texture(ctx->pipe, PIPE_FORMAT_A8R8G8B8_UNORM,
+                                        width, height);
+}
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+
+   /* unreference existing surfaces */
+   pipe_surface_reference( &rb->surface, NULL );
+   pipe_texture_reference( &rb->texture, NULL );
+
+   /* reference new ones */
+   pipe_surface_reference( &rb->surface, surf );
+   pipe_texture_reference( &rb->texture, surf->texture );
+
+   rb->width  = surf->width;
+   rb->height = surf->height;
+}
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surf)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+   *surf = rb->surface;
+   return VG_TRUE;
+}
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **tex)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+   *tex = rb->texture;
+   return VG_TRUE;
+}
+
+void * st_framebuffer_private(struct st_framebuffer *stfb)
+{
+   return stfb->privateData;
+}
+
+void st_unreference_framebuffer(struct st_framebuffer *stfb)
+{
+   /* FIXME */
+}
+
+void st_make_current(struct vg_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read)
+{
+   vg_set_current_context(st);
+   if (st) {
+      st->draw_buffer = draw;
+   }
+}
+
+struct vg_context *st_get_current(void)
+{
+   return vg_current_context();
+}
+
+void st_flush(struct vg_context *st, uint pipeFlushFlags,
+              struct pipe_fence_handle **fence)
+{
+   st->pipe->flush(st->pipe, pipeFlushFlags, fence);
+}
+
+void st_finish(struct vg_context *st)
+{
+   struct pipe_fence_handle *fence = NULL;
+
+   st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence);
+
+   st->pipe->screen->fence_finish(st->pipe->screen, fence, 0);
+   st->pipe->screen->fence_reference(st->pipe->screen, &fence, NULL);
+}
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb)
+{
+   struct vg_context *ctx = vg_current_context();
+   if (ctx && ctx->draw_buffer == stfb) {
+      st_flush(ctx,
+	       PIPE_FLUSH_RENDER_CACHE | 
+	       PIPE_FLUSH_SWAPBUFFERS |
+	       PIPE_FLUSH_FRAME,
+               NULL);
+   }
+}
+
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb)
+{
+}
+
+int st_bind_texture_surface(struct pipe_surface *ps, int target, int level,
+                            enum pipe_format format)
+{
+   return 0;
+}
+
+int st_unbind_texture_surface(struct pipe_surface *ps, int target, int level)
+{
+   return 0;
+}
diff --git a/src/gallium/state_trackers/vega/vg_tracker.h b/src/gallium/state_trackers/vega/vg_tracker.h
new file mode 100644
index 0000000000..5457631106
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_tracker.h
@@ -0,0 +1,107 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_TRACKER_H
+#define VG_TRACKER_H
+
+#include "VG/openvg.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+#define ST_SURFACE_FRONT_LEFT   0
+#define ST_SURFACE_BACK_LEFT    1
+#define ST_SURFACE_FRONT_RIGHT  2
+#define ST_SURFACE_BACK_RIGHT   3
+#define ST_SURFACE_DEPTH        8
+
+struct vg_context;
+struct st_framebuffer;
+struct pipe_context;
+struct pipe_fence_handle;
+struct pipe_surface;
+
+
+struct vg_context *st_create_context(struct pipe_context *pipe,
+                                     const void *visual,
+                                     struct vg_context *share);
+
+void st_destroy_context( struct vg_context *st );
+
+void st_copy_context_state(struct vg_context *dst, struct vg_context *src,
+                           uint mask);
+
+struct st_framebuffer *st_create_framebuffer(const void *visual,
+                                             enum pipe_format colorFormat,
+                                             enum pipe_format depthFormat,
+                                             enum pipe_format stencilFormat,
+                                             uint width, uint height,
+                                             void *privateData);
+
+void st_resize_framebuffer(struct st_framebuffer *stfb,
+                           uint width, uint height);
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf);
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width, uint *height);
+
+int st_bind_texture_surface(struct pipe_surface *ps, int target, int level,
+                            enum pipe_format format);
+
+int st_unbind_texture_surface(struct pipe_surface *ps, int target, int level);
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surf);
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **tex);
+
+void *st_framebuffer_private(struct st_framebuffer *stfb);
+
+void st_unreference_framebuffer(struct st_framebuffer *stfb);
+
+void st_make_current(struct vg_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read);
+
+struct vg_context *st_get_current(void);
+
+void st_flush(struct vg_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence);
+void st_finish(struct vg_context *st);
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb);
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb);
+
+
+/** Generic function type */
+typedef void (*st_proc)();
+
+st_proc st_get_proc_address(const char *procname);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_translate.c b/src/gallium/state_trackers/vega/vg_translate.c
new file mode 100644
index 0000000000..00e0764706
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_translate.c
@@ -0,0 +1,1030 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_translate.h"
+
+#include "pipe/p_format.h"
+#include "util/u_pack_color.h"
+
+void _vega_pack_rgba_span_float(struct vg_context *ctx,
+                                VGuint n, VGfloat rgba[][4],
+                                VGImageFormat dstFormat,
+                                void *dstAddr)
+{
+   VGint i;
+
+   switch (dstFormat) {
+   case VG_sRGBX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 255;
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGB_565: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         r = (r / 255.0) * 32;
+         g = (g / 255.0) * 32;
+         b = (b / 255.0) * 32;
+
+         dst[i] = b | g << 5 | r << 11;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_5551: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         r = (r / 255.0) * 32;
+         g = (g / 255.0) * 32;
+         b = (b / 255.0) * 32;
+         a = (a / 255.0);
+
+         dst[i] =  a | b << 1 | g << 6 | r << 11;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_4444: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         r = (r / 255.0) * 16;
+         g = (g / 255.0) * 16;
+         b = (b / 255.0) * 16;
+         a = (a / 255.0) * 16;
+
+         dst[i] =  a | b << 4 | g << 8 | r << 12;
+      }
+      return;
+   }
+      break;
+   case VG_sL_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] =  a;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 255;
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+   case VG_lRGBA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lL_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a;
+      }
+      return;
+   }
+      break;
+   case VG_A_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] = a;
+      }
+      return;
+   }
+      break;
+   case VG_BW_1: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         VGubyte res;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         res = (r + g + b + a)/4;
+         dst[i] =   (res & (128));
+      }
+      return;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] =   (a & (128));
+      }
+      return;
+   }
+      break;
+   case VG_A_4: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         VGubyte res;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         res = a/4;
+         dst[i] =   (res & (128));
+      }
+      return;
+   }
+      break;
+#endif
+   case VG_sXRGB_8888:
+      break;
+   case VG_sARGB_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_8888_PRE:  {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_1555:
+      break;
+   case VG_sARGB_4444:
+      break;
+   case VG_lXRGB_8888:
+      break;
+   case VG_lARGB_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_lARGB_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 0xff;
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGR_565:
+      break;
+   case VG_sBGRA_5551:
+      break;
+   case VG_sBGRA_4444:
+      break;
+   case VG_lBGRX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 0xff;
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sXBGR_8888:
+      break;
+   case VG_sABGR_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_1555:
+      break;
+   case VG_sABGR_4444:
+      break;
+   case VG_lXBGR_8888:
+      break;
+   case VG_lABGR_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_lABGR_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+}
+
+void _vega_unpack_float_span_rgba(struct vg_context *ctx,
+                                  VGuint n,
+                                  VGuint offset,
+                                  const void * data,
+                                  VGImageFormat dataFormat,
+                                  VGfloat rgba[][4])
+{
+   VGint i;
+
+   switch (dataFormat) {
+   case VG_sRGBX_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sRGB_565: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 10) & 31)/31.;
+         clr[1] = ((*src >>  5) & 95)/95.;
+         clr[2] = ((*src >>  0) & 31)/31.;
+         clr[3] = 1.f;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_5551: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 10) & 31)/31.;
+         clr[1] = ((*src >>  5) & 31)/31.;
+         clr[2] = ((*src >>  1) & 31)/31.;
+         clr[3] = ((*src >>  0) & 1)/1.;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_4444: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 12) & 15)/15.;
+         clr[1] = ((*src >>  8) & 15)/15.;
+         clr[2] = ((*src >>  4) & 15)/15.;
+         clr[3] = ((*src >>  0) & 15)/15.;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sL_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_lRGBX_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_lRGBA_8888: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBA_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lL_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_A_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_BW_1: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; i += 8) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < 8 && j < n ; ++j) {
+            VGint shift = j;
+            clr[0] = (((*src) & (1<<shift)) >> shift);
+            clr[1] = clr[0];
+            clr[2] = clr[0];
+            clr[3] = 1.f;
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i+j]);
+         }
+         ++src;
+      }
+   }
+      return;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; i += 8) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < 8 && j < n ; ++j) {
+            VGint shift = j;
+            clr[0] = 0.f;
+            clr[1] = 0.f;
+            clr[2] = 0.f;
+            clr[3] = (((*src) & (1<<shift)) >> shift);
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i+j]);
+         }
+         ++src;
+      }
+   }
+      return;
+   case VG_A_4: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset/2;
+      for (i = 0; i < n; i += 2) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < n && j < 2; ++j) {
+            VGint bitter, shift;
+            if (j == 0) {
+               bitter = 0x0f;
+               shift = 0;
+            } else {
+               bitter = 0xf0;
+               shift = 4;
+            }
+            clr[0] = 0.f;
+            clr[1] = 0.f;
+            clr[2] = 0.f;
+            clr[3] = ((*src) & (bitter)) >> shift;
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i +j]);
+         }
+         ++src;
+      }
+   }
+      return;
+#endif
+   case VG_sXRGB_8888:
+      break;
+   case VG_sARGB_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_1555:
+      break;
+   case VG_sARGB_4444:
+      break;
+   case VG_lXRGB_8888:
+      break;
+   case VG_lARGB_8888: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lARGB_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRX_8888:
+      break;
+   case VG_sBGRA_8888:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888_PRE:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGR_565:
+      break;
+   case VG_sBGRA_5551:
+      break;
+   case VG_sBGRA_4444:
+      break;
+   case VG_lBGRX_8888:
+      break;
+   case VG_lBGRA_8888:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888_PRE:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sXBGR_8888:
+      break;
+   case VG_sABGR_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_1555:
+      break;
+   case VG_sABGR_4444:
+      break;
+   case VG_lXBGR_8888:
+      break;
+   case VG_lABGR_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lABGR_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+}
+
+VGint _vega_size_for_format(VGImageFormat dataFormat)
+{
+   switch (dataFormat) {
+   case VG_sRGBX_8888:
+   case VG_sRGBA_8888:
+   case VG_sRGBA_8888_PRE:
+      return 4;
+   case VG_sRGB_565:
+   case VG_sRGBA_5551:
+   case VG_sRGBA_4444:
+      return 2;
+   case VG_sL_8:
+      return 1;
+   case VG_lRGBX_8888:
+   case VG_lRGBA_8888:
+   case VG_lRGBA_8888_PRE:
+      return 4;
+   case VG_lL_8:
+      return 1;
+   case VG_A_8:
+      return 1;
+   case VG_BW_1:
+      return 1;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+      break;
+   case VG_A_4:
+      break;
+#endif
+   case VG_sXRGB_8888:
+   case VG_sARGB_8888:
+   case VG_sARGB_8888_PRE:
+      return 4;
+   case VG_sARGB_1555:
+   case VG_sARGB_4444:
+      return 2;
+   case VG_lXRGB_8888:
+   case VG_lARGB_8888:
+   case VG_lARGB_8888_PRE:
+   case VG_sBGRX_8888:
+   case VG_sBGRA_8888:
+   case VG_sBGRA_8888_PRE:
+      return 4;
+   case VG_sBGR_565:
+   case VG_sBGRA_5551:
+   case VG_sBGRA_4444:
+      return 2;
+   case VG_lBGRX_8888:
+   case VG_lBGRA_8888:
+   case VG_lBGRA_8888_PRE:
+   case VG_sXBGR_8888:
+   case VG_sABGR_8888:
+   case VG_sABGR_8888_PRE:
+      return 4;
+   case VG_sABGR_1555:
+   case VG_sABGR_4444:
+      return 2;
+   case VG_lXBGR_8888:
+   case VG_lABGR_8888:
+   case VG_lABGR_8888_PRE:
+      return 4;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+   return 0;
+}
diff --git a/src/gallium/state_trackers/vega/vg_translate.h b/src/gallium/state_trackers/vega/vg_translate.h
new file mode 100644
index 0000000000..70815bacbc
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_translate.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_TRANSLATE_H
+#define VG_TRANSLATE_H
+
+#include "VG/openvg.h"
+#include "vg_context.h"
+
+/*FIXME: we really should be using translate module
+ * but pipe_format can't express some of the VG formats
+ * (the premultiplied ones) so currently it won't work */
+
+void _vega_pack_rgba_span_float(struct vg_context *ctx,
+                                VGuint n, VGfloat rgba[][4],
+                                VGImageFormat dstFormat,
+                                void *dstAddr);
+void _vega_unpack_float_span_rgba(struct vg_context *ctx,
+                                  VGuint n,
+                                  VGuint offset,
+                                  const void * data,
+                                  VGImageFormat dataFormat,
+                                  VGfloat rgba[][4]);
+VGint _vega_size_for_format(VGImageFormat format);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vgu.c b/src/gallium/state_trackers/vega/vgu.c
new file mode 100644
index 0000000000..7dc51c5599
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vgu.c
@@ -0,0 +1,450 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+#include "VG/vgu.h"
+
+#include "matrix.h"
+#include "path.h"
+
+#include "util/u_debug.h"
+#include "util/u_pointer.h"
+
+#include <math.h>
+#include <assert.h>
+
+static VGboolean is_aligned_to(const void *ptr, VGbyte alignment)
+{
+   void *aligned = align_pointer(ptr, alignment);
+   return (ptr == aligned) ? VG_TRUE : VG_FALSE;
+}
+
+static VGboolean is_aligned(const void *ptr)
+{
+   return is_aligned_to(ptr, 4);
+}
+
+static void vgu_append_float_coords(VGPath path,
+                                    const VGubyte *cmds,
+                                    VGint num_cmds,
+                                    const VGfloat *coords,
+                                    VGint num_coords)
+{
+   VGubyte common_data[40 * sizeof(VGfloat)];
+   struct path *p = (struct path *)path;
+
+   vg_float_to_datatype(path_datatype(p), common_data, coords, num_coords);
+   vgAppendPathData(path, num_cmds, cmds, common_data);
+}
+
+VGUErrorCode vguLine(VGPath path,
+                     VGfloat x0, VGfloat y0,
+                     VGfloat x1, VGfloat y1)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS};
+   VGfloat coords[4];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+
+   coords[0] = x0;
+   coords[1] = y0;
+   coords[2] = x1;
+   coords[3] = y1;
+
+   vgu_append_float_coords(path, cmds, 2, coords, 4);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguPolygon(VGPath path,
+                        const VGfloat * points,
+                        VGint count,
+                        VGboolean closed)
+{
+   VGubyte *cmds;
+   VGfloat *coords;
+   VGbitfield caps;
+   VGint i;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+
+   if (!points || count <= 0 || !is_aligned(points)) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+
+   cmds   = malloc(sizeof(VGubyte) * count + 1);
+   coords = malloc(sizeof(VGfloat) * count * 2);
+
+   cmds[0] = VG_MOVE_TO_ABS;
+   coords[0] = points[0];
+   coords[1] = points[1];
+   for (i = 1; i < count; ++i) {
+      cmds[i] = VG_LINE_TO_ABS;
+      coords[2*i + 0] = points[2*i + 0];
+      coords[2*i + 1] = points[2*i + 1];
+   }
+
+   if (closed) {
+      cmds[i] = VG_CLOSE_PATH;
+      ++i;
+   }
+
+   vgu_append_float_coords(path, cmds, i, coords, 2*i);
+
+   free(cmds);
+   free(coords);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode  vguRect(VGPath path,
+                      VGfloat x, VGfloat y,
+                      VGfloat width, VGfloat height)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_HLINE_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_HLINE_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat coords[5];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   coords[0] =  x;
+   coords[1] =  y;
+   coords[2] =  width;
+   coords[3] =  height;
+   coords[4] = -width;
+
+   vgu_append_float_coords(path, cmds, 5, coords, 5);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguRoundRect(VGPath path,
+                          VGfloat x, VGfloat y,
+                          VGfloat width,
+                          VGfloat height,
+                          VGfloat arcWidth,
+                          VGfloat arcHeight)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_HLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_HLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat c[26];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   c[0] =  x + arcWidth/2; c[1] =  y;
+
+   c[2] = width - arcWidth;
+
+   c[3] = arcWidth/2; c[4] = arcHeight/2; c[5] = 0;
+   c[6] = arcWidth/2; c[7] = arcHeight/2;
+
+   c[8] = height - arcHeight;
+
+   c[9] = arcWidth/2; c[10] = arcHeight/2; c[11] = 0;
+   c[12] = -arcWidth/2; c[13] = arcHeight/2;
+
+   c[14] = -(width - arcWidth);
+
+   c[15] = arcWidth/2; c[16] = arcHeight/2; c[17] = 0;
+   c[18] = -arcWidth/2; c[19] = -arcHeight/2;
+
+   c[20] = -(height - arcHeight);
+
+   c[21] = arcWidth/2; c[22] = arcHeight/2; c[23] = 0;
+   c[24] = arcWidth/2; c[25] = -arcHeight/2;
+
+   vgu_append_float_coords(path, cmds, 10, c, 26);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguEllipse(VGPath path,
+                        VGfloat cx, VGfloat cy,
+                        VGfloat width,
+                        VGfloat height)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat coords[12];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   coords[0] = cx + width/2; coords[1] = cy;
+
+   coords[2] = width/2; coords[3] = height/2; coords[4] = 0;
+   coords[5] = -width; coords[6] = 0;
+
+   coords[7] = width/2; coords[8] = height/2; coords[9] = 0;
+   coords[10] = width; coords[11] = 0;
+
+   vgu_append_float_coords(path, cmds, 4, coords, 11);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguArc(VGPath path,
+                    VGfloat x, VGfloat y,
+                    VGfloat width, VGfloat height,
+                    VGfloat startAngle,
+                    VGfloat angleExtent,
+                    VGUArcType arcType)
+{
+   VGubyte cmds[11];
+   VGfloat coords[40];
+   VGbitfield caps;
+   VGfloat last = startAngle + angleExtent;
+   VGint i, c = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+   if (arcType != VGU_ARC_OPEN &&
+       arcType != VGU_ARC_CHORD &&
+       arcType != VGU_ARC_PIE) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   cmds[c] = VG_MOVE_TO_ABS; ++c;
+   coords[0] = x+cos(DEGREES_TO_RADIANS(startAngle))*width/2;
+   coords[1] = y+sin(DEGREES_TO_RADIANS(startAngle))*height/2;
+#ifdef DEBUG_VGUARC
+   debug_printf("start [%f, %f]\n", coords[0], coords[1]);
+#endif
+   i = 2;
+   if (angleExtent > 0) {
+      VGfloat angle = startAngle + 180;
+      while (angle < last) {
+         cmds[c] = VG_SCCWARC_TO_ABS; ++c;
+         coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+         coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle))*width/2;
+         coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle))*height/2;
+#ifdef DEBUG_VGUARC
+         debug_printf("1 [%f, %f]\n", coords[i+3],
+                      coords[i+4]);
+#endif
+         i += 5;
+         angle += 180;
+      }
+      cmds[c] = VG_SCCWARC_TO_ABS; ++c;
+      coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+      coords[i+3] = x+cos(DEGREES_TO_RADIANS(last))*width/2;
+      coords[i+4] = y+sin(DEGREES_TO_RADIANS(last))*height/2;
+#ifdef DEBUG_VGUARC
+      debug_printf("2 [%f, %f]\n", coords[i+3],
+                   coords[i+4]);
+#endif
+      i += 5;
+   } else {
+      VGfloat angle = startAngle - 180;
+      while (angle > last) {
+         cmds[c] = VG_SCWARC_TO_ABS; ++c;
+         coords[i] =  width/2; coords[i+1] = height/2; coords[i+2] = 0;
+         coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle)) * width/2;
+         coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle)) * height/2;
+#ifdef DEBUG_VGUARC
+         debug_printf("3 [%f, %f]\n", coords[i+3],
+                      coords[i+4]);
+#endif
+         angle -= 180;
+         i += 5;
+      }
+      cmds[c] = VG_SCWARC_TO_ABS; ++c;
+      coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+      coords[i+3] = x + cos(DEGREES_TO_RADIANS(last)) * width/2;
+      coords[i+4] = y + sin(DEGREES_TO_RADIANS(last)) * height/2;
+#ifdef DEBUG_VGUARC
+      debug_printf("4 [%f, %f]\n", coords[i+3],
+                   coords[i+4]);
+#endif
+      i += 5;
+   }
+
+   if (arcType == VGU_ARC_PIE) {
+      cmds[c] = VG_LINE_TO_ABS; ++c;
+      coords[i] = x; coords[i + 1] = y;
+      i += 2;
+   }
+   if (arcType == VGU_ARC_PIE || arcType == VGU_ARC_CHORD) {
+      cmds[c] = VG_CLOSE_PATH;
+      ++c;
+   }
+
+   assert(c < 11);
+
+   vgu_append_float_coords(path, cmds, c, coords, i);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpQuadToSquare(VGfloat sx0, VGfloat sy0,
+                                        VGfloat sx1, VGfloat sy1,
+                                        VGfloat sx2, VGfloat sy2,
+                                        VGfloat sx3, VGfloat sy3,
+                                        VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_quad_to_square(sx0, sy0,
+                              sx1, sy1,
+                              sx2, sy2,
+                              sx3, sy3,
+                              &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   if (!matrix_is_invertible(&mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpSquareToQuad(VGfloat dx0, VGfloat dy0,
+                                        VGfloat dx1, VGfloat dy1,
+                                        VGfloat dx2, VGfloat dy2,
+                                        VGfloat dx3, VGfloat dy3,
+                                        VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_square_to_quad(dx0, dy0,
+                              dx1, dy1,
+                              dx2, dy2,
+                              dx3, dy3,
+                              &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   if (!matrix_is_invertible(&mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpQuadToQuad(VGfloat dx0, VGfloat dy0,
+                                      VGfloat dx1, VGfloat dy1,
+                                      VGfloat dx2, VGfloat dy2,
+                                      VGfloat dx3, VGfloat dy3,
+                                      VGfloat sx0, VGfloat sy0,
+                                      VGfloat sx1, VGfloat sy1,
+                                      VGfloat sx2, VGfloat sy2,
+                                      VGfloat sx3, VGfloat sy3,
+                                      VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_quad_to_quad(dx0, dy0,
+                            dx1, dy1,
+                            dx2, dy2,
+                            dx3, dy3,
+                            sx0, sy0,
+                            sx1, sy1,
+                            sx2, sy2,
+                            sx3, sy3,
+                            &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
diff --git a/src/gallium/state_trackers/xorg/Makefile b/src/gallium/state_trackers/xorg/Makefile
index a00ea3e2a4..27a1990724 100644
--- a/src/gallium/state_trackers/xorg/Makefile
+++ b/src/gallium/state_trackers/xorg/Makefile
@@ -1,29 +1,18 @@
-TARGET     = libxorgtracker.a
-CFILES     = $(wildcard ./*.c)
-OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../..
-TOP        = ../../../..
-
+TOP = ../../../..
 include $(TOP)/configs/current
 
-CFLAGS = -DHAVE_CONFIG_H \
-         -g -Wall -Wimplicit-function-declaration -fPIC \
-         $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \
-         -I$(GALLIUMDIR)/include \
-         -I$(GALLIUMDIR)/auxiliary \
-         -I$(TOP)/src/mesa/drivers/dri/common \
-         -I$(TOP)/src/mesa \
-         -I$(TOP)/include \
-         -I$(TOP)/src/egl/main
-
-#############################################
-
-.PHONY	= all clean
+LIBNAME = xorgtracker
 
-all: $(TARGET)
+LIBRARY_INCLUDES = \
+	-DHAVE_CONFIG_H \
+	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/include \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/mesa/main
 
-$(TARGET): $(OBJECTS)
-	ar rcs $(TARGET) $(OBJECTS)
+C_SOURCES = $(wildcard ./*.c)
 
-clean:
-	rm -rf $(OBJECTS) $(TARGET)
+include ../../Makefile.template
diff --git a/src/gallium/state_trackers/xorg/SConscript b/src/gallium/state_trackers/xorg/SConscript
new file mode 100644
index 0000000000..65f55ea378
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/SConscript
@@ -0,0 +1,27 @@
+#######################################################################
+# SConscript for xorg state_tracker
+
+Import('*')
+
+if 'xorg' in env['statetrackers']:
+
+    env = env.Clone()
+
+    env.Append(CPPPATH = [
+	'#/src/mesa',
+    ])
+
+    env.ParseConfig('pkg-config --cflags --libs xorg-server')
+
+    st_xorg = env.ConvenienceLibrary(
+	target = 'st_xorg',
+	source = [ 'xorg_composite.c',
+		'xorg_crtc.c',
+		'xorg_dri2.c',
+		'xorg_driver.c',
+		'xorg_exa.c',
+		'xorg_exa_tgsi.c',
+		'xorg_output.c',
+		]
+    )
+    Export('st_xorg')
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c
new file mode 100644
index 0000000000..c708ac3170
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_composite.c
@@ -0,0 +1,584 @@
+#include "xorg_composite.h"
+
+#include "xorg_exa_tgsi.h"
+
+#include "cso_cache/cso_context.h"
+#include "util/u_draw_quad.h"
+
+#include "pipe/p_inlines.h"
+
+struct xorg_composite_blend {
+   int op:8;
+
+   unsigned rgb_src_factor:5;    /**< PIPE_BLENDFACTOR_x */
+   unsigned rgb_dst_factor:5;    /**< PIPE_BLENDFACTOR_x */
+
+   unsigned alpha_src_factor:5;  /**< PIPE_BLENDFACTOR_x */
+   unsigned alpha_dst_factor:5;  /**< PIPE_BLENDFACTOR_x */
+};
+
+#define BLEND_OP_OVER 3
+static const struct xorg_composite_blend xorg_blends[] = {
+   { PictOpClear,
+     PIPE_BLENDFACTOR_CONST_COLOR, PIPE_BLENDFACTOR_CONST_ALPHA,
+     PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO },
+
+   { PictOpSrc,
+     PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ONE,
+     PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO },
+
+   { PictOpDst,
+     PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO,
+     PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ONE },
+
+   { PictOpOver,
+     PIPE_BLENDFACTOR_SRC_ALPHA, PIPE_BLENDFACTOR_ONE,
+     PIPE_BLENDFACTOR_INV_SRC_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA },
+
+   { PictOpOverReverse,
+     PIPE_BLENDFACTOR_SRC_ALPHA, PIPE_BLENDFACTOR_ONE,
+     PIPE_BLENDFACTOR_INV_SRC_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA },
+};
+
+static INLINE void
+pixel_to_float4(PictFormatPtr format,
+                CARD32 pixel, float *color)
+{
+   CARD32	    r, g, b, a;
+
+   debug_assert(format->type == PictTypeDirect);
+
+   r = (pixel >> format->direct.red) & format->direct.redMask;
+   g = (pixel >> format->direct.green) & format->direct.greenMask;
+   b = (pixel >> format->direct.blue) & format->direct.blueMask;
+   a = (pixel >> format->direct.alpha) & format->direct.alphaMask;
+   color[0] = ((float)r) / ((float)format->direct.redMask);
+   color[1] = ((float)g) / ((float)format->direct.greenMask);
+   color[2] = ((float)b) / ((float)format->direct.blueMask);
+   color[3] = ((float)a) / ((float)format->direct.alphaMask);
+}
+
+struct acceleration_info {
+   int op : 16;
+   int with_mask : 1;
+   int component_alpha : 1;
+};
+static const struct acceleration_info accelerated_ops[] = {
+   {PictOpClear,       1, 0},
+   {PictOpSrc,         1, 0},
+   {PictOpDst,         1, 0},
+   {PictOpOver,        1, 0},
+   {PictOpOverReverse, 1, 0},
+   {PictOpIn,          1, 0},
+   {PictOpInReverse,   1, 0},
+   {PictOpOut,         1, 0},
+   {PictOpOutReverse,  1, 0},
+   {PictOpAtop,        1, 0},
+   {PictOpAtopReverse, 1, 0},
+   {PictOpXor,         1, 0},
+   {PictOpAdd,         1, 0},
+   {PictOpSaturate,    1, 0},
+};
+
+static struct xorg_composite_blend
+blend_for_op(int op)
+{
+   const int num_blends =
+      sizeof(xorg_blends)/sizeof(struct xorg_composite_blend);
+   int i;
+
+   for (i = 0; i < num_blends; ++i) {
+      if (xorg_blends[i].op == op)
+         return xorg_blends[i];
+   }
+   return xorg_blends[BLEND_OP_OVER];
+}
+
+static INLINE int
+render_repeat_to_gallium(int mode)
+{
+   switch(mode) {
+   case RepeatNone:
+      return PIPE_TEX_WRAP_CLAMP;
+   case RepeatNormal:
+      return PIPE_TEX_WRAP_REPEAT;
+   case RepeatReflect:
+      return PIPE_TEX_WRAP_MIRROR_REPEAT;
+   case RepeatPad:
+      return PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   default:
+      debug_printf("Unsupported repeat mode\n");
+   }
+   return PIPE_TEX_WRAP_REPEAT;
+}
+
+
+static INLINE void
+setup_vertex0(float vertex[2][4], float x, float y,
+              float color[4])
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = color[0]; /*r*/
+   vertex[1][1] = color[1]; /*g*/
+   vertex[1][2] = color[2]; /*b*/
+   vertex[1][3] = color[3]; /*a*/
+}
+
+static struct pipe_buffer *
+setup_vertex_data0(struct exa_context *ctx,
+                   int srcX, int srcY, int maskX, int maskY,
+                   int dstX, int dstY, int width, int height)
+{
+   float vertices[4][2][4];
+
+   /* 1st vertex */
+   setup_vertex0(vertices[0], dstX, dstY,
+                 ctx->solid_color);
+   /* 2nd vertex */
+   setup_vertex0(vertices[1], dstX + width, dstY,
+                 ctx->solid_color);
+   /* 3rd vertex */
+   setup_vertex0(vertices[2], dstX + width, dstY + height,
+                 ctx->solid_color);
+   /* 4th vertex */
+   setup_vertex0(vertices[3], dstX, dstY + height,
+                 ctx->solid_color);
+
+   return pipe_user_buffer_create(ctx->ctx->screen,
+                                  vertices,
+                                  sizeof(vertices));
+}
+
+static INLINE void
+setup_vertex1(float vertex[2][4], float x, float y, float s, float t)
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = s;   /*s*/
+   vertex[1][1] = t;   /*t*/
+   vertex[1][2] = 0.f; /*r*/
+   vertex[1][3] = 1.f; /*q*/
+}
+
+static struct pipe_buffer *
+setup_vertex_data1(struct exa_context *ctx,
+                   int srcX, int srcY, int maskX, int maskY,
+                   int dstX, int dstY, int width, int height)
+{
+   float vertices[4][2][4];
+   float s0, t0, s1, t1;
+   struct pipe_texture *src = ctx->bound_textures[0];
+
+   s0 = srcX / src->width[0];
+   s1 = srcX + width / src->width[0];
+   t0 = srcY / src->height[0];
+   t1 = srcY + height / src->height[0];
+
+   /* 1st vertex */
+   setup_vertex1(vertices[0], dstX, dstY,
+                 s0, t0);
+   /* 2nd vertex */
+   setup_vertex1(vertices[1], dstX + width, dstY,
+                 s1, t0);
+   /* 3rd vertex */
+   setup_vertex1(vertices[2], dstX + width, dstY + height,
+                 s1, t1);
+   /* 4th vertex */
+   setup_vertex1(vertices[3], dstX, dstY + height,
+                 s0, t1);
+
+   return pipe_user_buffer_create(ctx->ctx->screen,
+                                  vertices,
+                                  sizeof(vertices));
+}
+
+
+static INLINE void
+setup_vertex2(float vertex[3][4], float x, float y,
+              float s0, float t0, float s1, float t1)
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = s0;  /*s*/
+   vertex[1][1] = t0;  /*t*/
+   vertex[1][2] = 0.f; /*r*/
+   vertex[1][3] = 1.f; /*q*/
+
+   vertex[2][0] = s1;  /*s*/
+   vertex[2][1] = t1;  /*t*/
+   vertex[2][2] = 0.f; /*r*/
+   vertex[2][3] = 1.f; /*q*/
+}
+
+static struct pipe_buffer *
+setup_vertex_data2(struct exa_context *ctx,
+                   int srcX, int srcY, int maskX, int maskY,
+                   int dstX, int dstY, int width, int height)
+{
+   float vertices[4][3][4];
+   float st0[4], st1[4];
+   struct pipe_texture *src = ctx->bound_textures[0];
+   struct pipe_texture *mask = ctx->bound_textures[0];
+
+   st0[0] = srcX / src->width[0];
+   st0[1] = srcY / src->height[0];
+   st0[2] = srcX + width / src->width[0];
+   st0[3] = srcY + height / src->height[0];
+
+   st1[0] = maskX / mask->width[0];
+   st1[1] = maskY / mask->height[0];
+   st1[2] = maskX + width / mask->width[0];
+   st1[3] = maskY + height / mask->height[0];
+
+   /* 1st vertex */
+   setup_vertex2(vertices[0], dstX, dstY,
+                 st0[0], st0[1], st1[0], st1[1]);
+   /* 2nd vertex */
+   setup_vertex2(vertices[1], dstX + width, dstY,
+                 st0[2], st0[1], st1[2], st1[1]);
+   /* 3rd vertex */
+   setup_vertex2(vertices[2], dstX + width, dstY + height,
+                 st0[2], st0[3], st1[2], st1[3]);
+   /* 4th vertex */
+   setup_vertex2(vertices[3], dstX, dstY + height,
+                 st0[0], st0[3], st1[0], st1[3]);
+
+   return pipe_user_buffer_create(ctx->ctx->screen,
+                                  vertices,
+                                  sizeof(vertices));
+}
+
+boolean xorg_composite_accelerated(int op,
+                                   PicturePtr pSrcPicture,
+                                   PicturePtr pMaskPicture,
+                                   PicturePtr pDstPicture)
+{
+   unsigned i;
+   unsigned accel_ops_count =
+      sizeof(accelerated_ops)/sizeof(struct acceleration_info);
+
+
+   /*FIXME: currently accel is disabled */
+   return FALSE;
+
+   if (pSrcPicture) {
+      /* component alpha not supported */
+      if (pSrcPicture->componentAlpha)
+         return FALSE;
+      /* fills not supported */
+      if (pSrcPicture->pSourcePict)
+         return FALSE;
+   }
+
+   for (i = 0; i < accel_ops_count; ++i) {
+      if (op == accelerated_ops[i].op) {
+         if (pMaskPicture && !accelerated_ops[i].with_mask)
+            return FALSE;
+         return TRUE;
+      }
+   }
+   return FALSE;
+}
+
+static void
+bind_framebuffer_state(struct exa_context *exa, PicturePtr pDstPicture,
+                       struct exa_pixmap_priv *pDst)
+{
+   unsigned i;
+   struct pipe_framebuffer_state state;
+   struct pipe_surface *surface = exa_gpu_surface(exa, pDst);
+   memset(&state, 0, sizeof(struct pipe_framebuffer_state));
+
+   state.width  = pDstPicture->pDrawable->width;
+   state.height = pDstPicture->pDrawable->height;
+
+   state.nr_cbufs = 1;
+   state.cbufs[0] = surface;
+   for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+      state.cbufs[i] = 0;
+
+   /* currently we don't use depth/stencil */
+   state.zsbuf = 0;
+
+   cso_set_framebuffer(exa->cso, &state);
+}
+
+enum AxisOrientation {
+   Y0_BOTTOM,
+   Y0_TOP
+};
+
+static void
+set_viewport(struct exa_context *exa, int width, int height,
+             enum AxisOrientation orientation)
+{
+   struct pipe_viewport_state viewport;
+   float y_scale = (orientation == Y0_BOTTOM) ? -2.f : 2.f;
+
+   viewport.scale[0] =  width / 2.f;
+   viewport.scale[1] =  height / y_scale;
+   viewport.scale[2] =  1.0;
+   viewport.scale[3] =  1.0;
+   viewport.translate[0] = width / 2.f;
+   viewport.translate[1] = height / 2.f;
+   viewport.translate[2] = 0.0;
+   viewport.translate[3] = 0.0;
+
+   cso_set_viewport(exa->cso, &viewport);
+}
+
+static void
+bind_viewport_state(struct exa_context *exa, PicturePtr pDstPicture)
+{
+   int width = pDstPicture->pDrawable->width;
+   int height = pDstPicture->pDrawable->height;
+
+   set_viewport(exa, width, height, Y0_TOP);
+}
+
+static void
+bind_blend_state(struct exa_context *exa, int op,
+                 PicturePtr pSrcPicture, PicturePtr pMaskPicture)
+{
+   boolean component_alpha = pSrcPicture->componentAlpha;
+   struct xorg_composite_blend blend_opt;
+   struct pipe_blend_state blend;
+
+   if (component_alpha) {
+      op = PictOpOver;
+   }
+   blend_opt = blend_for_op(op);
+
+   memset(&blend, 0, sizeof(struct pipe_blend_state));
+   blend.blend_enable = 1;
+   blend.colormask |= PIPE_MASK_R;
+   blend.colormask |= PIPE_MASK_G;
+   blend.colormask |= PIPE_MASK_B;
+   blend.colormask |= PIPE_MASK_A;
+
+   blend.rgb_src_factor   = blend_opt.rgb_src_factor;
+   blend.alpha_src_factor = blend_opt.alpha_src_factor;
+   blend.rgb_dst_factor   = blend_opt.rgb_dst_factor;
+   blend.alpha_dst_factor = blend_opt.alpha_dst_factor;
+
+   cso_set_blend(exa->cso, &blend);
+}
+
+static void
+bind_rasterizer_state(struct exa_context *exa)
+{
+   struct pipe_rasterizer_state raster;
+   memset(&raster, 0, sizeof(struct pipe_rasterizer_state));
+   raster.gl_rasterization_rules = 1;
+   cso_set_rasterizer(exa->cso, &raster);
+}
+
+static void
+bind_shaders(struct exa_context *exa, int op,
+             PicturePtr pSrcPicture, PicturePtr pMaskPicture)
+{
+   unsigned vs_traits = 0, fs_traits = 0;
+   struct xorg_shader shader;
+
+   if (pSrcPicture) {
+      if (pSrcPicture->pSourcePict) {
+         if (pSrcPicture->pSourcePict->type == SourcePictTypeSolidFill) {
+            fs_traits |= FS_SOLID_FILL;
+            vs_traits |= VS_SOLID_FILL;
+            pixel_to_float4(pSrcPicture->pFormat,
+                            pSrcPicture->pSourcePict->solidFill.color,
+                            exa->solid_color);
+         } else {
+            debug_assert("!gradients not supported");
+         }
+      } else {
+         fs_traits |= FS_COMPOSITE;
+         vs_traits |= VS_COMPOSITE;
+      }
+   }
+
+   if (pMaskPicture) {
+      vs_traits |= VS_MASK;
+      fs_traits |= FS_MASK;
+   }
+
+   shader = xorg_shaders_get(exa->shaders, vs_traits, fs_traits);
+   cso_set_vertex_shader_handle(exa->cso, shader.vs);
+   cso_set_fragment_shader_handle(exa->cso, shader.fs);
+}
+
+
+static void
+bind_samplers(struct exa_context *exa, int op,
+              PicturePtr pSrcPicture, PicturePtr pMaskPicture,
+              PicturePtr pDstPicture,
+              struct exa_pixmap_priv *pSrc,
+              struct exa_pixmap_priv *pMask,
+              struct exa_pixmap_priv *pDst)
+{
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_state src_sampler, mask_sampler;
+
+   exa->num_bound_samplers = 0;
+
+   memset(&src_sampler, 0, sizeof(struct pipe_sampler_state));
+   memset(&mask_sampler, 0, sizeof(struct pipe_sampler_state));
+
+   if (pSrcPicture && pSrc) {
+      unsigned src_wrap = render_repeat_to_gallium(
+         pSrcPicture->repeatType);
+      src_sampler.wrap_s = src_wrap;
+      src_sampler.wrap_t = src_wrap;
+      src_sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      src_sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      src_sampler.normalized_coords = 1;
+      samplers[0] = &src_sampler;
+      exa->bound_textures[0] = pSrc->tex;
+      ++exa->num_bound_samplers;
+   }
+
+   if (pMaskPicture && pMask) {
+      unsigned mask_wrap = render_repeat_to_gallium(
+         pMaskPicture->repeatType);
+      mask_sampler.wrap_s = mask_wrap;
+      mask_sampler.wrap_t = mask_wrap;
+      mask_sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      mask_sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      mask_sampler.normalized_coords = 1;
+      samplers[1] = &mask_sampler;
+      exa->bound_textures[1] = pMask->tex;
+      ++exa->num_bound_samplers;
+   }
+
+   cso_set_samplers(exa->cso, exa->num_bound_samplers,
+                    (const struct pipe_sampler_state **)samplers);
+   cso_set_sampler_textures(exa->cso, exa->num_bound_samplers,
+                            exa->bound_textures);
+}
+
+static void
+setup_vs_constant_buffer(struct exa_context *exa,
+                         int width, int height)
+{
+   const int param_bytes = 8 * sizeof(float);
+   float vs_consts[8] = {
+      2.f/width, 2.f/height, 1, 1,
+      -1, -1, 0, 0
+   };
+   struct pipe_constant_buffer *cbuf = &exa->vs_const_buffer;
+
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+   cbuf->buffer = pipe_buffer_create(exa->ctx->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      pipe_buffer_write(exa->ctx->screen, cbuf->buffer,
+                        0, param_bytes, vs_consts);
+   }
+   exa->ctx->set_constant_buffer(exa->ctx, PIPE_SHADER_VERTEX, 0, cbuf);
+}
+
+
+static void
+setup_fs_constant_buffer(struct exa_context *exa)
+{
+   const int param_bytes = 4 * sizeof(float);
+   float fs_consts[8] = {
+      0, 0, 0, 1,
+   };
+   struct pipe_constant_buffer *cbuf = &exa->fs_const_buffer;
+
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+   cbuf->buffer = pipe_buffer_create(exa->ctx->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      pipe_buffer_write(exa->ctx->screen, cbuf->buffer,
+                        0, param_bytes, fs_consts);
+   }
+   exa->ctx->set_constant_buffer(exa->ctx, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static void
+setup_constant_buffers(struct exa_context *exa, PicturePtr pDstPicture)
+{
+   int width = pDstPicture->pDrawable->width;
+   int height = pDstPicture->pDrawable->height;
+
+   setup_vs_constant_buffer(exa, width, height);
+   setup_fs_constant_buffer(exa);
+}
+
+boolean xorg_composite_bind_state(struct exa_context *exa,
+                                  int op,
+                                  PicturePtr pSrcPicture,
+                                  PicturePtr pMaskPicture,
+                                  PicturePtr pDstPicture,
+                                  struct exa_pixmap_priv *pSrc,
+                                  struct exa_pixmap_priv *pMask,
+                                  struct exa_pixmap_priv *pDst)
+{
+   bind_framebuffer_state(exa, pDstPicture, pDst);
+   bind_viewport_state(exa, pDstPicture);
+   bind_blend_state(exa, op, pSrcPicture, pMaskPicture);
+   bind_rasterizer_state(exa);
+   bind_shaders(exa, op, pSrcPicture, pMaskPicture);
+   bind_samplers(exa, op, pSrcPicture, pMaskPicture,
+                 pDstPicture, pSrc, pMask, pDst);
+
+   setup_constant_buffers(exa, pDstPicture);
+
+   return FALSE;
+}
+
+void xorg_composite(struct exa_context *exa,
+                    struct exa_pixmap_priv *dst,
+                    int srcX, int srcY, int maskX, int maskY,
+                    int dstX, int dstY, int width, int height)
+{
+   struct pipe_context *pipe = exa->ctx;
+   struct pipe_buffer *buf = 0;
+
+   if (exa->num_bound_samplers == 0 ) { /* solid fill */
+      buf = setup_vertex_data0(exa,
+                               srcX, srcY, maskX, maskY,
+                               dstX, dstY, width, height);
+   } else if (exa->num_bound_samplers == 1 ) /* src */
+      buf = setup_vertex_data1(exa,
+                               srcX, srcY, maskX, maskY,
+                               dstX, dstY, width, height);
+   else if (exa->num_bound_samplers == 2) /* src + mask */
+      buf = setup_vertex_data2(exa,
+                               srcX, srcY, maskX, maskY,
+                               dstX, dstY, width, height);
+   else if (exa->num_bound_samplers == 3) { /* src + mask + dst */
+      debug_assert(!"src/mask/dst not handled right now");
+#if 0
+      buf = setup_vertex_data2(exa,
+                               srcX, srcY, maskX, maskY,
+                               dstX, dstY, width, height);
+#endif
+   }
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              1 + exa->num_bound_samplers); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+}
+
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.h b/src/gallium/state_trackers/xorg/xorg_composite.h
new file mode 100644
index 0000000000..17dfcb199e
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_composite.h
@@ -0,0 +1,25 @@
+#ifndef XORG_COMPOSITE_H
+#define XORG_COMPOSITE_H
+
+#include "xorg_exa.h"
+
+boolean xorg_composite_accelerated(int op,
+                                   PicturePtr pSrcPicture,
+                                   PicturePtr pMaskPicture,
+                                   PicturePtr pDstPicture);
+
+boolean xorg_composite_bind_state(struct exa_context *exa,
+                                  int op,
+                                  PicturePtr pSrcPicture,
+                                  PicturePtr pMaskPicture,
+                                  PicturePtr pDstPicture,
+                                  struct exa_pixmap_priv *pSrc,
+                                  struct exa_pixmap_priv *pMask,
+                                  struct exa_pixmap_priv *pDst);
+
+void xorg_composite(struct exa_context *exa,
+                    struct exa_pixmap_priv *dst,
+                    int srcX, int srcY, int maskX, int maskY,
+                    int dstX, int dstY, int width, int height);
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c
index 7304113a65..fe08bde9ef 100644
--- a/src/gallium/state_trackers/xorg/xorg_crtc.c
+++ b/src/gallium/state_trackers/xorg/xorg_crtc.c
@@ -46,13 +46,14 @@
 #include <X11/extensions/dpms.h>
 
 #include "pipe/p_inlines.h"
+#include "util/u_rect.h"
 
 struct crtc_private
 {
     drmModeCrtcPtr drm_crtc;
 
     /* hwcursor */
-    struct pipe_buffer *cursor_buf;
+    struct pipe_texture *cursor_tex;
     unsigned cursor_handle;
 };
 
@@ -171,11 +172,10 @@ crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data)
 static void
 crtc_destroy(xf86CrtcPtr crtc)
 {
-    modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
 
-    if (crtcp->cursor_buf)
-	pipe_buffer_reference(&crtcp->cursor_buf, NULL);
+    if (crtcp->cursor_tex)
+	pipe_texture_reference(&crtcp->cursor_tex, NULL);
 
     drmModeFreeCrtc(crtcp->drm_crtc);
     xfree(crtcp);
@@ -187,24 +187,42 @@ crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image)
     unsigned char *ptr;
     modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
-
-    if (!crtcp->cursor_buf) {
-	crtcp->cursor_buf = pipe_buffer_create(ms->screen,
-					       0,
-					       PIPE_BUFFER_USAGE_CPU_WRITE |
-					       PIPE_BUFFER_USAGE_GPU_READ,
-					       64*64*4);
-	drm_api_hooks.handle_from_buffer(ms->screen,
-					 crtcp->cursor_buf,
-					 &crtcp->cursor_handle);
+    struct pipe_transfer *transfer;
+
+    if (!crtcp->cursor_tex) {
+	struct pipe_texture templat;
+	unsigned pitch;
+
+	memset(&templat, 0, sizeof(templat));
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY;
+	templat.target = PIPE_TEXTURE_2D;
+	templat.last_level = 0;
+	templat.depth[0] = 1;
+	templat.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	templat.width[0] = 64;
+	templat.height[0] = 64;
+	pf_get_block(templat.format, &templat.block);
+
+	crtcp->cursor_tex = ms->screen->texture_create(ms->screen,
+						       &templat);
+	ms->api->local_handle_from_texture(ms->api,
+					   ms->screen,
+					   crtcp->cursor_tex,
+					   &pitch,
+					   &crtcp->cursor_handle);
     }
 
-    ptr = pipe_buffer_map(ms->screen, crtcp->cursor_buf, PIPE_BUFFER_USAGE_CPU_WRITE);
-
-    if (ptr)
-	memcpy(ptr, image, 64 * 64 * 4);
-
-    pipe_buffer_unmap(ms->screen, crtcp->cursor_buf);
+    transfer = ms->screen->get_tex_transfer(ms->screen, crtcp->cursor_tex,
+					    0, 0, 0,
+					    PIPE_TRANSFER_WRITE,
+					    0, 0, 64, 64);
+    ptr = ms->screen->transfer_map(ms->screen, transfer);
+    util_copy_rect(ptr, &crtcp->cursor_tex->block,
+		   transfer->stride, 0, 0,
+		   64, 64, (void*)image, 64 * 4, 0, 0);
+    ms->screen->transfer_unmap(ms->screen, transfer);
+    ms->screen->tex_transfer_destroy(transfer);
 }
 
 static void
@@ -222,7 +240,7 @@ crtc_show_cursor(xf86CrtcPtr crtc)
     modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
 
-    if (crtcp->cursor_buf)
+    if (crtcp->cursor_tex)
 	drmModeSetCursor(ms->fd, crtcp->drm_crtc->crtc_id,
 			 crtcp->cursor_handle, 64, 64);
 }
@@ -260,13 +278,12 @@ static const xf86CrtcFuncsRec crtc_funcs = {
 };
 
 void
-cursor_destroy(xf86CrtcPtr crtc)
+crtc_cursor_destroy(xf86CrtcPtr crtc)
 {
-    modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
 
-    if (crtcp->cursor_buf) {
-	pipe_buffer_reference(&crtcp->cursor_buf, NULL);
+    if (crtcp->cursor_tex) {
+	pipe_texture_reference(&crtcp->cursor_tex, NULL);
     }
 }
 
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index 401bd39dac..6431a0fe25 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -33,97 +33,193 @@
 #include "xf86_OSproc.h"
 
 #include "xorg_tracker.h"
+#include "xorg_exa.h"
 
 #include "dri2.h"
 
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
 
+#include "util/u_rect.h"
+
 typedef struct {
     PixmapPtr pPixmap;
     struct pipe_texture *tex;
-    struct pipe_buffer *buf;
+    struct pipe_fence_handle *fence;
 } *BufferPrivatePtr;
 
-static DRI2BufferPtr
-driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
+static Bool
+driDoCreateBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int format)
 {
-    struct pipe_texture *depth, *tex;
-    struct pipe_buffer *buf;
+    struct pipe_texture *tex = NULL;
     ScreenPtr pScreen = pDraw->pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    BufferPrivatePtr privates;
-    DRI2BufferPtr buffers;
+    struct exa_pixmap_priv *exa_priv;
+    BufferPrivatePtr private = buffer->driverPrivate;
     PixmapPtr pPixmap;
     unsigned stride, handle;
-    int i;
-
-    buffers = xcalloc(count, sizeof *buffers);
-    if (!buffers)
-	goto fail_buffers;
 
-    privates = xcalloc(count, sizeof *privates);
-    if (!privates)
-	goto fail_privates;
+    if (pDraw->type == DRAWABLE_PIXMAP)
+	pPixmap = (PixmapPtr) pDraw;
+    else
+	pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw);
+    exa_priv = exaGetPixmapDriverPrivate(pPixmap);
 
-    depth = NULL;
-    for (i = 0; i < count; i++) {
-	pPixmap = NULL;
-	tex = NULL;
-	buf = NULL;
-	if (attachments[i] == DRI2BufferFrontLeft) {
-	    if (pDraw->type == DRAWABLE_PIXMAP)
-		pPixmap = (PixmapPtr) pDraw;
-	    else
-		pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw);
-	    pPixmap->refcnt++;
-	    tex = xorg_exa_get_texture(pPixmap);
-	} else if (attachments[i] == DRI2BufferStencil) {
-	    pipe_texture_reference(&tex, depth);
-	} else if (attachments[i] == DRI2BufferDepth) {
-	    struct pipe_texture template;
-
-	    memset(&template, 0, sizeof(template));
-	    template.target = PIPE_TEXTURE_2D;
-	    template.format = PIPE_FORMAT_S8Z24_UNORM;
-	    pf_get_block(template.format, &template.block);
-	    template.width[0] = pDraw->width;
-	    template.height[0] = pDraw->height;
-	    template.depth[0] = 1;
-	    template.last_level = 0;
-	    template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
-	    tex = ms->screen->texture_create(ms->screen, &template);
-	} else {
+    switch (buffer->attachment) {
+    default:
+	if (buffer->attachment != DRI2BufferFakeFrontLeft ||
+	    pDraw->type != DRAWABLE_PIXMAP) {
+	    private->pPixmap = (*pScreen->CreatePixmap)(pScreen, pDraw->width,
+							pDraw->height,
+							pDraw->depth,
+							0);
+	}
+	break;
+    case DRI2BufferFrontLeft:
+	break;
+    case DRI2BufferStencil:
+#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2
+    case DRI2BufferDepthStencil:
+	if (exa_priv->depth_stencil_tex &&
+	    !pf_is_depth_stencil(exa_priv->depth_stencil_tex->format))
+	    exa_priv->depth_stencil_tex = NULL;
+        /* Fall through */
+#endif
+    case DRI2BufferDepth:
+	if (exa_priv->depth_stencil_tex)
+	    pipe_texture_reference(&tex, exa_priv->depth_stencil_tex);
+        else {
 	    struct pipe_texture template;
 	    memset(&template, 0, sizeof(template));
 	    template.target = PIPE_TEXTURE_2D;
-	    template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	    if (buffer->attachment == DRI2BufferDepth)
+		template.format = ms->ds_depth_bits_last ?
+		    PIPE_FORMAT_X8Z24_UNORM : PIPE_FORMAT_Z24X8_UNORM;
+	    else
+		template.format = ms->ds_depth_bits_last ?
+		    PIPE_FORMAT_S8Z24_UNORM : PIPE_FORMAT_Z24S8_UNORM;
 	    pf_get_block(template.format, &template.block);
 	    template.width[0] = pDraw->width;
 	    template.height[0] = pDraw->height;
 	    template.depth[0] = 1;
 	    template.last_level = 0;
-	    template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
+	    template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
+		PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
 	    tex = ms->screen->texture_create(ms->screen, &template);
+	    pipe_texture_reference(&exa_priv->depth_stencil_tex, tex);
 	}
+	break;
+    }
+
+    if (!private->pPixmap) {
+	private->pPixmap = pPixmap;
+	pPixmap->refcnt++;
+    }
+
+    if (!tex) {
+	xorg_exa_set_shared_usage(private->pPixmap);
+	pScreen->ModifyPixmapHeader(private->pPixmap, 0, 0, 0, 0, 0, NULL);
+	tex = xorg_exa_get_texture(private->pPixmap);
+    }
+
+    if (!tex)
+	FatalError("NO TEXTURE IN DRI2\n");
+
+    ms->api->shared_handle_from_texture(ms->api, ms->screen, tex, &stride, &handle);
+
+    buffer->name = handle;
+    buffer->pitch = stride;
+    buffer->cpp = 4;
+    buffer->driverPrivate = private;
+    buffer->flags = 0; /* not tiled */
+    private->tex = tex;
+
+    return TRUE;
+}
+
+static void
+driDoDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer)
+{
+    ScreenPtr pScreen = pDraw->pScreen;
+    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
+    modesettingPtr ms = modesettingPTR(pScrn);
+    BufferPrivatePtr private = buffer->driverPrivate;
+    struct exa_pixmap_priv *exa_priv = exaGetPixmapDriverPrivate(private->pPixmap);
+
+    pipe_texture_reference(&private->tex, NULL);
+    ms->screen->fence_reference(ms->screen, &private->fence, NULL);
+    pipe_texture_reference(&exa_priv->depth_stencil_tex, NULL);
+    (*pScreen->DestroyPixmap)(private->pPixmap);
+}
+
+#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2
+
+static DRI2BufferPtr
+driCreateBuffer(DrawablePtr pDraw, unsigned int attachment, unsigned int format)
+{
+    DRI2BufferPtr buffer;
+    BufferPrivatePtr private;
+
+    buffer = xcalloc(1, sizeof *buffer);
+    if (!buffer)
+	return NULL;
+
+    private = xcalloc(1, sizeof *private);
+    if (!private) {
+	goto fail;
+    }
+
+    buffer->attachment = attachment;
+    buffer->driverPrivate = private;
+
+    if (driDoCreateBuffer(pDraw, buffer, format))
+	return buffer;
+
+    xfree(private);
+fail:
+    xfree(buffer);
+    return NULL;
+}
+
+static void
+driDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer)
+{
+    driDoDestroyBuffer(pDraw, buffer);
+
+    xfree(buffer->driverPrivate);
+    xfree(buffer);
+}
 
-	drm_api_hooks.buffer_from_texture(tex, &buf, &stride);
-	drm_api_hooks.global_handle_from_buffer(ms->screen, buf, &handle);
+#else /* DRI2INFOREC_VERSION <= 2 */
 
-	buffers[i].name = handle;
+static DRI2BufferPtr
+driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
+{
+    BufferPrivatePtr privates;
+    DRI2BufferPtr buffers;
+    int i;
+
+    buffers = xcalloc(count, sizeof *buffers);
+    if (!buffers)
+	goto fail_buffers;
+
+    privates = xcalloc(count, sizeof *privates);
+    if (!privates)
+	goto fail_privates;
+
+    for (i = 0; i < count; i++) {
 	buffers[i].attachment = attachments[i];
-	buffers[i].pitch = stride;
-	buffers[i].cpp = 4;
 	buffers[i].driverPrivate = &privates[i];
-	buffers[i].flags = 0; /* not tiled */
-	privates[i].pPixmap = pPixmap;
-	privates[i].buf = buf;
-	privates[i].tex = tex;
+
+	if (!driDoCreateBuffer(pDraw, &buffers[i], 0))
+	    goto fail;
     }
 
     return buffers;
 
+fail:
+    xfree(privates);
 fail_privates:
     xfree(buffers);
 fail_buffers:
@@ -133,20 +229,10 @@ fail_buffers:
 static void
 driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count)
 {
-    ScreenPtr pScreen = pDraw->pScreen;
-    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
-    modesettingPtr ms = modesettingPTR(pScrn);
-    BufferPrivatePtr private;
     int i;
 
     for (i = 0; i < count; i++) {
-	private = buffers[i].driverPrivate;
-
-	if (private->pPixmap)
-	    (*pScreen->DestroyPixmap)(private->pPixmap);
-
-	pipe_texture_reference(&private->tex, NULL);
-	pipe_buffer_reference(&private->buf, NULL);
+	driDoDestroyBuffer(pDraw, &buffers[i]);
     }
 
     if (buffers) {
@@ -155,6 +241,8 @@ driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count)
     }
 }
 
+#endif /* DRI2INFOREC_VERSION */
+
 static void
 driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
               DRI2BufferPtr pDestBuffer, DRI2BufferPtr pSrcBuffer)
@@ -164,19 +252,83 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
     modesettingPtr ms = modesettingPTR(pScrn);
     BufferPrivatePtr dst_priv = pDestBuffer->driverPrivate;
     BufferPrivatePtr src_priv = pSrcBuffer->driverPrivate;
+    PixmapPtr src_pixmap;
+    PixmapPtr dst_pixmap;
+    GCPtr gc;
+    RegionPtr copy_clip;
+
+    /*
+     * In driCreateBuffers we dewrap windows into the
+     * backing pixmaps in order to get to the texture.
+     * We need to use the real drawable in CopyArea
+     * so that cliprects and offsets are correct.
+     */
+    src_pixmap = src_priv->pPixmap;
+    dst_pixmap = dst_priv->pPixmap;
+    if (pSrcBuffer->attachment == DRI2BufferFrontLeft)
+	src_pixmap = (PixmapPtr)pDraw;
+    if (pDestBuffer->attachment == DRI2BufferFrontLeft)
+	dst_pixmap = (PixmapPtr)pDraw;
+
+    /*
+     * The clients implements glXWaitX with a copy front to fake and then
+     * waiting on the server to signal its completion of it. While
+     * glXWaitGL is a client side flush and a copy from fake to front.
+     * This is how it is done in the DRI2 protocol, how ever depending
+     * which type of drawables the server does things a bit differently
+     * then what the protocol says as the fake and front are the same.
+     *
+     * for pixmaps glXWaitX is a server flush.
+     * for pixmaps glXWaitGL is a client flush.
+     * for windows glXWaitX is a copy from front to fake then a server flush.
+     * for windows glXWaitGL is a client flush then a copy from fake to front.
+     *
+     * XXX in the windows case this code always flushes but that isn't a
+     * must in the glXWaitGL case but we don't know if this is a glXWaitGL
+     * or a glFlush/glFinish call.
+     */
+    if (dst_pixmap == src_pixmap) {
+	/* pixmap glXWaitX */
+	if (pSrcBuffer->attachment == DRI2BufferFrontLeft &&
+	    pDestBuffer->attachment == DRI2BufferFakeFrontLeft) {
+	    ms->ctx->flush(ms->ctx, PIPE_FLUSH_SWAPBUFFERS, NULL);
+	    return;
+	}
+	/* pixmap glXWaitGL */
+	if (pDestBuffer->attachment == DRI2BufferFrontLeft &&
+	    pSrcBuffer->attachment == DRI2BufferFakeFrontLeft) {
+	    return;
+	} else {
+	    xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
+		"copying between the same pixmap\n");
+	}
+    }
 
-    struct pipe_surface *dst_surf =
-	ms->screen->get_tex_surface(ms->screen, dst_priv->tex, 0, 0, 0,
-				    PIPE_BUFFER_USAGE_GPU_WRITE);
-    struct pipe_surface *src_surf =
-	ms->screen->get_tex_surface(ms->screen, src_priv->tex, 0, 0, 0,
-				    PIPE_BUFFER_USAGE_GPU_READ);
+    gc = GetScratchGC(pDraw->depth, pScreen);
+    copy_clip = REGION_CREATE(pScreen, NULL, 0);
+    REGION_COPY(pScreen, copy_clip, pRegion);
+    (*gc->funcs->ChangeClip) (gc, CT_REGION, copy_clip, 0);
+    ValidateGC(&dst_pixmap->drawable, gc);
 
-    ms->ctx->surface_copy(ms->ctx, dst_surf, 0, 0, src_surf,
-			  0, 0, pDraw->width, pDraw->height);
+    /* If this is a full buffer swap, throttle on the previous one */
+    if (dst_priv->fence && REGION_NUM_RECTS(pRegion) == 1) {
+	BoxPtr extents = REGION_EXTENTS(pScreen, pRegion);
 
-    pipe_surface_reference(&dst_surf, NULL);
-    pipe_surface_reference(&src_surf, NULL);
+	if (extents->x1 == 0 && extents->y1 == 0 &&
+	    extents->x2 == pDraw->width && extents->y2 == pDraw->height) {
+	    ms->screen->fence_finish(ms->screen, dst_priv->fence, 0);
+	    ms->screen->fence_reference(ms->screen, &dst_priv->fence, NULL);
+	}
+    }
+
+    (*gc->ops->CopyArea)(&src_pixmap->drawable, &dst_pixmap->drawable, gc,
+			 0, 0, pDraw->width, pDraw->height, 0, 0);
+
+    FreeScratchGC(gc);
+
+    ms->ctx->flush(ms->ctx, PIPE_FLUSH_SWAPBUFFERS,
+		   pDestBuffer->attachment == DRI2BufferFrontLeft ?
+		   &dst_priv->fence : NULL);
 }
 
 Bool
@@ -186,18 +338,34 @@ driScreenInit(ScreenPtr pScreen)
     modesettingPtr ms = modesettingPTR(pScrn);
     DRI2InfoRec dri2info;
 
-    dri2info.version = 1;
-    dri2info.fd = ms->fd;
-#if 0
-    dri2info.driverName = pScrn->name;
+#if defined(DRI2INFOREC_VERSION)
+    dri2info.version = DRI2INFOREC_VERSION;
 #else
-    dri2info.driverName = "i915"; /* FIXME */
+    dri2info.version = 1;
 #endif
+    dri2info.fd = ms->fd;
+
+    dri2info.driverName = pScrn->driverName;
     dri2info.deviceName = "/dev/dri/card0"; /* FIXME */
 
+#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2
+    dri2info.CreateBuffer = driCreateBuffer;
+    dri2info.DestroyBuffer = driDestroyBuffer;
+#else
     dri2info.CreateBuffers = driCreateBuffers;
     dri2info.DestroyBuffers = driDestroyBuffers;
+#endif
     dri2info.CopyRegion = driCopyRegion;
+    dri2info.Wait = NULL;
+
+    ms->d_depth_bits_last =
+	 ms->screen->is_format_supported(ms->screen, PIPE_FORMAT_X8Z24_UNORM,
+					 PIPE_TEXTURE_2D,
+					 PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
+    ms->ds_depth_bits_last =
+	 ms->screen->is_format_supported(ms->screen, PIPE_FORMAT_S8Z24_UNORM,
+					 PIPE_TEXTURE_2D,
+					 PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0);
 
     return DRI2ScreenInit(pScreen, &dri2info);
 }
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index 45e831f0c2..643b6b3b9e 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -33,10 +33,8 @@
 #include "xf86.h"
 #include "xf86_OSproc.h"
 #include "compiler.h"
-#include "xf86RAC.h"
 #include "xf86PciInfo.h"
 #include "xf86Pci.h"
-#include "xf86Resources.h"
 #include "mipointer.h"
 #include "micmap.h"
 #include <X11/extensions/randr.h>
@@ -54,6 +52,7 @@
 
 #include <pciaccess.h>
 
+#include "pipe/p_context.h"
 #include "xorg_tracker.h"
 #include "xorg_winsys.h"
 
@@ -84,41 +83,9 @@ static const OptionInfoRec Options[] = {
 };
 
 /*
- * Functions that might be needed
- */
-
-static const char *exaSymbols[] = {
-    "exaGetVersion",
-    "exaDriverInit",
-    "exaDriverFini",
-    "exaOffscreenAlloc",
-    "exaOffscreenFree",
-    "exaWaitSync",
-    NULL
-};
-
-static const char *fbSymbols[] = {
-    "fbPictureInit",
-    "fbScreenInit",
-    NULL
-};
-
-static const char *ddcSymbols[] = {
-    "xf86PrintEDID",
-    "xf86SetDDCproperties",
-    NULL
-};
-
-/*
  * Exported Xorg driver functions to winsys
  */
 
-void
-xorg_tracker_loader_ref_sym_lists()
-{
-    LoaderRefSymLists(exaSymbols, fbSymbols, ddcSymbols, NULL);
-}
-
 const OptionInfoRec *
 xorg_tracker_available_options(int chipid, int busid)
 {
@@ -179,8 +146,10 @@ CreateFrontBuffer(ScrnInfoPtr pScrn)
     modesettingPtr ms = modesettingPTR(pScrn);
     ScreenPtr pScreen = pScrn->pScreen;
     PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen);
+    unsigned handle, stride;
 
     ms->noEvict = TRUE;
+    xorg_exa_set_displayed_usage(rootPixmap);
     pScreen->ModifyPixmapHeader(rootPixmap,
 				pScrn->virtualX, pScrn->virtualY,
 				pScrn->depth, pScrn->bitsPerPixel,
@@ -188,13 +157,16 @@ CreateFrontBuffer(ScrnInfoPtr pScrn)
 				NULL);
     ms->noEvict = FALSE;
 
+    handle = xorg_exa_get_pixmap_handle(rootPixmap, &stride);
+
     drmModeAddFB(ms->fd,
 		 pScrn->virtualX,
 		 pScrn->virtualY,
 		 pScrn->depth,
 		 pScrn->bitsPerPixel,
-		 pScrn->displayWidth * pScrn->bitsPerPixel / 8,
-		 xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id);
+		 stride,
+		 handle,
+		 &ms->fb_id);
 
     pScrn->frameX0 = 0;
     pScrn->frameY0 = 0;
@@ -282,10 +254,6 @@ PreInit(ScrnInfoPtr pScrn, int flags)
     } else
 	ms->entityPrivate = NULL;
 
-    if (xf86RegisterResources(ms->pEnt->index, NULL, ResNone)) {
-	return FALSE;
-    }
-
     if (xf86IsEntityShared(pScrn->entityList[0])) {
 	if (xf86IsPrimInitDone(pScrn->entityList[0])) {
 	    /* do something */
@@ -300,12 +268,12 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 	    ms->PciInfo->dev, ms->PciInfo->func
 	);
 
+    ms->api = drm_api_create();
     ms->fd = drmOpen(NULL, BusID);
 
     if (ms->fd < 0)
 	return FALSE;
 
-    pScrn->racMemFlags = RAC_FB | RAC_COLORMAP;
     pScrn->monitor = pScrn->confScreen->monitor;
     pScrn->progClock = TRUE;
     pScrn->rgbBits = 8;
@@ -391,8 +359,6 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 	return FALSE;
     }
 
-    xf86LoaderReqSymLists(fbSymbols, NULL);
-
     xf86LoadSubModule(pScrn, "exa");
 
 #ifdef DRI2
@@ -418,6 +384,44 @@ RestoreHWState(ScrnInfoPtr pScrn)
     return TRUE;
 }
 
+static void xorgBlockHandler(int i, pointer blockData, pointer pTimeout,
+                             pointer pReadmask)
+{
+    ScreenPtr pScreen = screenInfo.screens[i];
+    modesettingPtr ms = modesettingPTR(xf86Screens[pScreen->myNum]);
+
+    pScreen->BlockHandler = ms->blockHandler;
+    pScreen->BlockHandler(i, blockData, pTimeout, pReadmask);
+    pScreen->BlockHandler = xorgBlockHandler;
+
+    ms->ctx->flush(ms->ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+    {
+	RegionPtr dirty = DamageRegion(ms->damage);
+	unsigned num_cliprects = REGION_NUM_RECTS(dirty);
+
+	if (num_cliprects) {
+	    drmModeClip *clip = alloca(num_cliprects * sizeof(drmModeClip));
+	    BoxPtr rect = REGION_RECTS(dirty);
+	    int i;
+
+	    for (i = 0; i < num_cliprects; i++, rect++) {
+		clip[i].x = rect->x1;
+		clip[i].y = rect->y1;
+		clip[i].width = rect->x2 - rect->x1;
+		clip[i].height = rect->y2 - rect->y1;
+	    }
+
+	    /* TODO query connector property to see if this is needed */
+	    drmModeDirtyFB(ms->fd, ms->fb_id, clip, num_cliprects);
+
+	    DamageEmpty(ms->damage);
+	}
+    }
+#endif
+}
+
 static Bool
 CreateScreenResources(ScreenPtr pScreen)
 {
@@ -425,6 +429,7 @@ CreateScreenResources(ScreenPtr pScreen)
     modesettingPtr ms = modesettingPTR(pScrn);
     PixmapPtr rootPixmap;
     Bool ret;
+    unsigned handle, stride;
 
     ms->noEvict = TRUE;
 
@@ -434,21 +439,41 @@ CreateScreenResources(ScreenPtr pScreen)
 
     rootPixmap = pScreen->GetScreenPixmap(pScreen);
 
+    xorg_exa_set_displayed_usage(rootPixmap);
+    xorg_exa_set_shared_usage(rootPixmap);
     if (!pScreen->ModifyPixmapHeader(rootPixmap, -1, -1, -1, -1, -1, NULL))
 	FatalError("Couldn't adjust screen pixmap\n");
 
     ms->noEvict = FALSE;
 
+    handle = xorg_exa_get_pixmap_handle(rootPixmap, &stride);
+
     drmModeAddFB(ms->fd,
 		 pScrn->virtualX,
 		 pScrn->virtualY,
 		 pScrn->depth,
 		 pScrn->bitsPerPixel,
-		 pScrn->displayWidth * pScrn->bitsPerPixel / 8,
-		 xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id);
+		 stride,
+		 handle,
+                 &ms->fb_id);
 
     AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
 
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+    ms->damage = DamageCreate(NULL, NULL, DamageReportNone, TRUE,
+                              pScreen, rootPixmap);
+
+    if (ms->damage) {
+       DamageRegister(&rootPixmap->drawable, ms->damage);
+
+       xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Damage tracking initialized\n");
+    } else {
+       xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
+                  "Failed to create screen damage record\n");
+       return FALSE;
+    }
+#endif
+
     return ret;
 }
 
@@ -476,7 +501,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
     }
 
     if (!ms->screen) {
-	ms->screen = drm_api_hooks.create_screen(ms->fd, NULL);
+	ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL);
 
 	if (!ms->screen) {
 	    FatalError("Could not init pipe_screen\n");
@@ -525,6 +550,8 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
 
     fbPictureInit(pScreen, NULL, 0);
 
+    ms->blockHandler = pScreen->BlockHandler;
+    pScreen->BlockHandler = xorgBlockHandler;
     ms->createScreenResources = pScreen->CreateScreenResources;
     pScreen->CreateScreenResources = CreateScreenResources;
 
@@ -593,10 +620,6 @@ FreeScreen(int scrnIndex, int flags)
     FreeRec(xf86Screens[scrnIndex]);
 }
 
-/* HACK */
-void
-cursor_destroy(xf86CrtcPtr crtc);
-
 static void
 LeaveVT(int scrnIndex, int flags)
 {
@@ -608,7 +631,7 @@ LeaveVT(int scrnIndex, int flags)
     for (o = 0; o < config->num_crtc; o++) {
 	xf86CrtcPtr crtc = config->crtc[o];
 
-	cursor_destroy(crtc);
+	crtc_cursor_destroy(crtc);
 
 	if (crtc->rotatedPixmap || crtc->rotatedData) {
 	    crtc->funcs->shadow_destroy(crtc, crtc->rotatedPixmap,
@@ -622,6 +645,10 @@ LeaveVT(int scrnIndex, int flags)
 
     RestoreHWState(pScrn);
 
+    if (drmDropMaster(ms->fd))
+	xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
+		   "drmDropMaster failed: %s\n", strerror(errno));
+
     pScrn->vtSema = FALSE;
 }
 
@@ -634,6 +661,17 @@ EnterVT(int scrnIndex, int flags)
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
     modesettingPtr ms = modesettingPTR(pScrn);
 
+    if (drmSetMaster(ms->fd)) {
+	if (errno == EINVAL) {
+	    xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
+		       "drmSetMaster failed: 2.6.29 or newer kernel required for "
+		       "multi-server DRI\n");
+	} else {
+	    xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
+		       "drmSetMaster failed: %s\n", strerror(errno));
+	}
+    }
+
     /*
      * Only save state once per server generation since that's what most
      * drivers do.  Could change this to save state at each VT enter.
@@ -673,11 +711,22 @@ CloseScreen(int scrnIndex, ScreenPtr pScreen)
     driCloseScreen(pScreen);
 #endif
 
+    pScreen->BlockHandler = ms->blockHandler;
     pScreen->CreateScreenResources = ms->createScreenResources;
 
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+    if (ms->damage) {
+	DamageUnregister(&pScreen->GetScreenPixmap(pScreen)->drawable, ms->damage);
+	DamageDestroy(ms->damage);
+	ms->damage = NULL;
+    }
+#endif
+
     if (ms->exa)
 	xorg_exa_close(pScrn);
 
+	ms->api->destroy(ms->api);
+	ms->api = NULL;
     drmClose(ms->fd);
     ms->fd = -1;
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index 7913174354..a17a71f23a 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -28,34 +28,24 @@
  *
  */
 
-#include "xorg-server.h"
-#include "xf86.h"
+#include "xorg_exa.h"
 #include "xorg_tracker.h"
+#include "xorg_composite.h"
+#include "xorg_exa_tgsi.h"
+
+#include <xorg-server.h>
+#include <xf86.h>
+#include <picturestr.h>
+#include <picture.h>
 
 #include "pipe/p_format.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
 
-#include "util/u_rect.h"
-
-struct exa_entity
-{
-    ExaDriverPtr pExa;
-    struct pipe_context *ctx;
-    struct pipe_screen *scrn;
-};
-
-struct PixmapPriv
-{
-    int flags;
-    struct pipe_texture *tex;
-    unsigned int color;
-    struct pipe_surface *src_surf; /* for copies */
+#include "cso_cache/cso_context.h"
 
-    struct pipe_transfer *map_transfer;
-    unsigned map_count;
-};
+#include "util/u_rect.h"
 
 /*
  * Helper functions
@@ -108,13 +98,75 @@ ExaMarkSync(ScreenPtr pScreen)
 }
 
 static Bool
+ExaDownloadFromScreen(PixmapPtr pPix, int x,  int y, int w,  int h, char *dst,
+		      int dst_pitch)
+{
+    ScreenPtr pScreen = pPix->drawable.pScreen;
+    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
+    modesettingPtr ms = modesettingPTR(pScrn);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPix);
+    struct pipe_transfer *transfer;
+
+    if (!priv || !priv->tex)
+	return FALSE;
+
+    if (exa->ctx->is_texture_referenced(exa->ctx, priv->tex, 0, 0) &
+	PIPE_REFERENCED_FOR_WRITE)
+	exa->ctx->flush(exa->ctx, 0, NULL);
+
+    transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
+					   PIPE_TRANSFER_READ, x, y, w, h);
+    if (!transfer)
+	return FALSE;
+
+    util_copy_rect((unsigned char*)dst, &priv->tex->block, dst_pitch, 0, 0,
+		   w, h, exa->scrn->transfer_map(exa->scrn, transfer),
+		   transfer->stride, 0, 0);
+
+    exa->scrn->transfer_unmap(exa->scrn, transfer);
+    exa->scrn->tex_transfer_destroy(transfer);
+
+    return TRUE;
+}
+
+static Bool
+ExaUploadToScreen(PixmapPtr pPix, int x, int y, int w, int h, char *src,
+		  int src_pitch)
+{
+    ScreenPtr pScreen = pPix->drawable.pScreen;
+    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
+    modesettingPtr ms = modesettingPTR(pScrn);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPix);
+    struct pipe_transfer *transfer;
+
+    if (!priv || !priv->tex)
+	return FALSE;
+
+    transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
+					   PIPE_TRANSFER_WRITE, x, y, w, h);
+    if (!transfer)
+	return FALSE;
+
+    util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer),
+		   &priv->tex->block, transfer->stride, 0, 0, w, h,
+		   (unsigned char*)src, src_pitch, 0, 0);
+
+    exa->scrn->transfer_unmap(exa->scrn, transfer);
+    exa->scrn->tex_transfer_destroy(transfer);
+
+    return TRUE;
+}
+
+static Bool
 ExaPrepareAccess(PixmapPtr pPix, int index)
 {
     ScreenPtr pScreen = pPix->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv;
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv;
 
     priv = exaGetPixmapDriverPrivate(pPix);
 
@@ -126,6 +178,10 @@ ExaPrepareAccess(PixmapPtr pPix, int index)
 
     if (priv->map_count++ == 0)
     {
+	if (exa->ctx->is_texture_referenced(exa->ctx, priv->tex, 0, 0) &
+	    PIPE_REFERENCED_FOR_WRITE)
+	    exa->ctx->flush(exa->ctx, 0, NULL);
+
 	priv->map_transfer =
 	    exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
 					PIPE_TRANSFER_READ_WRITE,
@@ -145,8 +201,8 @@ ExaFinishAccess(PixmapPtr pPix, int index)
     ScreenPtr pScreen = pPix->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv;
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv;
     priv = exaGetPixmapDriverPrivate(pPix);
 
     if (!priv)
@@ -160,6 +216,7 @@ ExaFinishAccess(PixmapPtr pPix, int index)
 	exa->scrn->transfer_unmap(exa->scrn, priv->map_transfer);
 	exa->scrn->tex_transfer_destroy(priv->map_transfer);
 	priv->map_transfer = NULL;
+	pPix->devPrivate.ptr = NULL;
     }
 }
 
@@ -168,8 +225,8 @@ ExaDone(PixmapPtr pPixmap)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct exa_entity *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
 
     if (!priv)
 	return;
@@ -190,11 +247,8 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct exa_entity *exa = ms->exa;
-
-    if (1)
-        return FALSE;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
 
     if (pPixmap->drawable.depth < 15)
 	return FALSE;
@@ -221,12 +275,9 @@ ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct pipe_surface *surf =
-	exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
-				   PIPE_BUFFER_USAGE_GPU_READ |
-				   PIPE_BUFFER_USAGE_GPU_WRITE);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct pipe_surface *surf = exa_gpu_surface(exa, priv);
 
     exa->ctx->surface_fill(exa->ctx, surf, x0, y0, x1 - x0, y1 - y0,
 			   priv->color);
@@ -240,12 +291,9 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
 {
     ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
-    struct PixmapPriv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap);
-
-    if (1)
-        return FALSE;
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
+    struct exa_pixmap_priv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap);
 
     if (alu != GXcopy)
 	return FALSE;
@@ -265,10 +313,7 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
     if (!exa->ctx || !exa->ctx->surface_copy)
 	return FALSE;
 
-    priv->src_surf =
-	exa->scrn->get_tex_surface(exa->scrn, src_priv->tex, 0, 0, 0,
-				   PIPE_BUFFER_USAGE_GPU_READ |
-				   PIPE_BUFFER_USAGE_GPU_WRITE);
+    priv->src_surf = exa_gpu_surface(exa, src_priv);
 
     return TRUE;
 }
@@ -279,12 +324,9 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY,
 {
     ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
-    struct pipe_surface *surf =
-	exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
-				   PIPE_BUFFER_USAGE_GPU_READ |
-				   PIPE_BUFFER_USAGE_GPU_WRITE);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
+    struct pipe_surface *surf = exa_gpu_surface(exa, priv);
 
     exa->ctx->surface_copy(exa->ctx, surf, dstX, dstY, priv->src_surf,
 			   srcX, srcY, width, height);
@@ -296,13 +338,28 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture,
 		    PicturePtr pMaskPicture, PicturePtr pDstPicture,
 		    PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst)
 {
-    return FALSE;
+   ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+
+   return xorg_composite_bind_state(exa, op, pSrcPicture, pMaskPicture,
+                                    pDstPicture,
+                                    exaGetPixmapDriverPrivate(pSrc),
+                                    exaGetPixmapDriverPrivate(pMask),
+                                    exaGetPixmapDriverPrivate(pDst));
 }
 
 static void
 ExaComposite(PixmapPtr pDst, int srcX, int srcY, int maskX, int maskY,
 	     int dstX, int dstY, int width, int height)
 {
+   ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+   struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDst);
+
+   xorg_composite(exa, priv, srcX, srcY, maskX, maskY,
+                  dstX, dstY, width, height);
 }
 
 static Bool
@@ -310,15 +367,18 @@ ExaCheckComposite(int op,
 		  PicturePtr pSrcPicture, PicturePtr pMaskPicture,
 		  PicturePtr pDstPicture)
 {
-    return FALSE;
+   return xorg_composite_accelerated(op,
+                                     pSrcPicture,
+                                     pMaskPicture,
+                                     pDstPicture);
 }
 
 static void *
 ExaCreatePixmap(ScreenPtr pScreen, int size, int align)
 {
-    struct PixmapPriv *priv;
+    struct exa_pixmap_priv *priv;
 
-    priv = xcalloc(1, sizeof(struct PixmapPriv));
+    priv = xcalloc(1, sizeof(struct exa_pixmap_priv));
     if (!priv)
 	return NULL;
 
@@ -328,7 +388,7 @@ ExaCreatePixmap(ScreenPtr pScreen, int size, int align)
 static void
 ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv)
 {
-    struct PixmapPriv *priv = (struct PixmapPriv *)dPriv;
+    struct exa_pixmap_priv *priv = (struct exa_pixmap_priv *)dPriv;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
 
@@ -344,7 +404,7 @@ ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv)
 static Bool
 ExaPixmapIsOffscreen(PixmapPtr pPixmap)
 {
-    struct PixmapPriv *priv;
+    struct exa_pixmap_priv *priv;
 
     priv = exaGetPixmapDriverPrivate(pPixmap);
 
@@ -357,14 +417,45 @@ ExaPixmapIsOffscreen(PixmapPtr pPixmap)
     return FALSE;
 }
 
+int
+xorg_exa_set_displayed_usage(PixmapPtr pPixmap)
+{
+    struct exa_pixmap_priv *priv;
+    priv = exaGetPixmapDriverPrivate(pPixmap);
+
+    if (!priv) {
+	FatalError("NO PIXMAP PRIVATE\n");
+	return 0;
+    }
+
+    priv->flags |= PIPE_TEXTURE_USAGE_PRIMARY;
+
+    return 0;
+}
+
+int
+xorg_exa_set_shared_usage(PixmapPtr pPixmap)
+{
+    struct exa_pixmap_priv *priv;
+    priv = exaGetPixmapDriverPrivate(pPixmap);
+
+    if (!priv) {
+	FatalError("NO PIXMAP PRIVATE\n");
+	return 0;
+    }
+
+    priv->flags |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+
+    return 0;
+}
+
 unsigned
-xorg_exa_get_pixmap_handle(PixmapPtr pPixmap)
+xorg_exa_get_pixmap_handle(PixmapPtr pPixmap, unsigned *stride_out)
 {
     ScreenPtr pScreen = pPixmap->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv;
-    struct pipe_buffer *buffer = NULL;
+    struct exa_pixmap_priv *priv;
     unsigned handle;
     unsigned stride;
 
@@ -380,9 +471,10 @@ xorg_exa_get_pixmap_handle(PixmapPtr pPixmap)
 	return 0;
     }
 
-    drm_api_hooks.buffer_from_texture(priv->tex, &buffer, &stride);
-    drm_api_hooks.handle_from_buffer(ms->screen, buffer, &handle);
-    pipe_buffer_reference(&buffer, NULL);
+    ms->api->local_handle_from_texture(ms->api, ms->screen, priv->tex, &stride, &handle);
+    if (stride_out)
+	*stride_out = stride;
+
     return handle;
 }
 
@@ -393,11 +485,11 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 {
     ScreenPtr pScreen = pPixmap->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
+    struct exa_context *exa = ms->exa;
 
-    if (!priv)
+    if (!priv || pPixData)
 	return FALSE;
 
     if (depth <= 0)
@@ -419,35 +511,69 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 			     bitsPerPixel, devKind, NULL);
 
     /* Deal with screen resize */
-    if (priv->tex && (priv->tex->width[0] != width || priv->tex->height[0] != height)) {
-	pipe_texture_reference(&priv->tex, NULL);
-    }
-
-    if (!priv->tex) {
-	struct pipe_texture template;
-
-	memset(&template, 0, sizeof(template));
-	template.target = PIPE_TEXTURE_2D;
-	exa_get_pipe_format(depth, &template.format, &bitsPerPixel);
-	pf_get_block(template.format, &template.block);
-	template.width[0] = width;
-	template.height[0] = height;
-	template.depth[0] = 1;
-	template.last_level = 0;
-	template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
-	priv->tex = exa->scrn->texture_create(exa->scrn, &template);
-    }
-
-    if (pPixData) {
-	struct pipe_transfer *transfer =
-	    exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
-					PIPE_TRANSFER_WRITE,
-					0, 0, width, height);
-        pipe_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer),
-                       &priv->tex->block, transfer->stride, 0, 0,
-                       width, height, pPixData, pPixmap->devKind, 0, 0);
-        exa->scrn->transfer_unmap(exa->scrn, transfer);
-        exa->scrn->tex_transfer_destroy(transfer);
+    if (!priv->tex ||
+        (priv->tex->width[0] != width ||
+         priv->tex->height[0] != height ||
+         priv->tex_flags != priv->flags)) {
+	struct pipe_texture *texture = NULL;
+
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+	if (priv->flags)
+#endif
+	{
+	    struct pipe_texture template;
+
+	    memset(&template, 0, sizeof(template));
+	    template.target = PIPE_TEXTURE_2D;
+	    exa_get_pipe_format(depth, &template.format, &bitsPerPixel);
+	    pf_get_block(template.format, &template.block);
+	    template.width[0] = width;
+	    template.height[0] = height;
+	    template.depth[0] = 1;
+	    template.last_level = 0;
+	    template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags;
+	    priv->tex_flags = priv->flags;
+	    texture = exa->scrn->texture_create(exa->scrn, &template);
+
+	    if (priv->tex) {
+		struct pipe_surface *dst_surf;
+
+		dst_surf = exa->scrn->get_tex_surface(exa->scrn, texture, 0, 0, 0,
+						      PIPE_BUFFER_USAGE_GPU_WRITE);
+		priv->src_surf = exa_gpu_surface(exa, priv);
+		exa->ctx->surface_copy(exa->ctx, dst_surf, 0, 0, priv->src_surf,
+				       0, 0, min(width, texture->width[0]),
+				       min(height, texture->height[0]));
+		exa->scrn->tex_surface_destroy(dst_surf);
+		exa->scrn->tex_surface_destroy(priv->src_surf);
+		priv->src_surf = NULL;
+	    } else if (pPixmap->devPrivate.ptr) {
+		struct pipe_transfer *transfer;
+
+		if (priv->map_count != 0)
+		     FatalError("doing ExaModifyPixmapHeader on mapped buffer\n");
+
+		transfer =
+		    exa->scrn->get_tex_transfer(exa->scrn, texture, 0, 0, 0,
+						PIPE_TRANSFER_WRITE,
+						0, 0, width, height);
+		util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer),
+			       &texture->block, transfer->stride, 0, 0,
+			       width, height, pPixmap->devPrivate.ptr,
+			       pPixmap->devKind, 0, 0);
+		exa->scrn->transfer_unmap(exa->scrn, transfer);
+		exa->scrn->tex_transfer_destroy(transfer);
+	    }
+	}
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+	else {
+	    xfree(pPixmap->devPrivate.ptr);
+	    pPixmap->devPrivate.ptr = xalloc(pPixmap->drawable.height *
+					     pPixmap->devKind);
+	}
+#endif
+
+	pipe_texture_reference(&priv->tex, texture);
     }
 
     return TRUE;
@@ -456,87 +582,119 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 struct pipe_texture *
 xorg_exa_get_texture(PixmapPtr pPixmap)
 {
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct pipe_texture *tex = NULL;
-    pipe_texture_reference(&tex, priv->tex);
-    return tex;
+   struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+   struct pipe_texture *tex = NULL;
+   pipe_texture_reference(&tex, priv->tex);
+   return tex;
 }
 
 void
 xorg_exa_close(ScrnInfoPtr pScrn)
 {
-    modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-
-    if (exa->ctx)
-	exa->ctx->destroy(exa->ctx);
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+   struct pipe_constant_buffer *vsbuf = &exa->vs_const_buffer;
+   struct pipe_constant_buffer *fsbuf = &exa->fs_const_buffer;
 
-    exaDriverFini(pScrn->pScreen);
-    xfree(exa);
-    ms->exa = NULL;
-}
+   if (exa->shaders) {
+      xorg_shaders_destroy(exa->shaders);
+   }
 
-void *
-xorg_exa_init(ScrnInfoPtr pScrn)
-{
-    modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa;
-    ExaDriverPtr pExa;
+   if (vsbuf && vsbuf->buffer)
+      pipe_buffer_reference(&vsbuf->buffer, NULL);
 
-    exa = xcalloc(1, sizeof(struct exa_entity));
-    if (!exa)
-	return NULL;
+   if (fsbuf && fsbuf->buffer)
+      pipe_buffer_reference(&fsbuf->buffer, NULL);
 
-    pExa = exaDriverAlloc();
-    if (!pExa) {
-	goto out_err;
-    }
+   if (exa->cso) {
+      cso_release_all(exa->cso);
+      cso_destroy_context(exa->cso);
+   }
 
-    memset(pExa, 0, sizeof(*pExa));
-    pExa->exa_major = 2;
-    pExa->exa_minor = 2;
-    pExa->memoryBase = 0;
-    pExa->memorySize = 0;
-    pExa->offScreenBase = 0;
-    pExa->pixmapOffsetAlign = 0;
-    pExa->pixmapPitchAlign = 1;
-    pExa->flags = EXA_OFFSCREEN_PIXMAPS | EXA_HANDLES_PIXMAPS;
-    pExa->maxX = 8191;		       /* FIXME */
-    pExa->maxY = 8191;		       /* FIXME */
-    pExa->WaitMarker = ExaWaitMarker;
-    pExa->MarkSync = ExaMarkSync;
-    pExa->PrepareSolid = ExaPrepareSolid;
-    pExa->Solid = ExaSolid;
-    pExa->DoneSolid = ExaDone;
-    pExa->PrepareCopy = ExaPrepareCopy;
-    pExa->Copy = ExaCopy;
-    pExa->DoneCopy = ExaDone;
-    pExa->CheckComposite = ExaCheckComposite;
-    pExa->PrepareComposite = ExaPrepareComposite;
-    pExa->Composite = ExaComposite;
-    pExa->DoneComposite = ExaDoneComposite;
-    pExa->PixmapIsOffscreen = ExaPixmapIsOffscreen;
-    pExa->PrepareAccess = ExaPrepareAccess;
-    pExa->FinishAccess = ExaFinishAccess;
-    pExa->CreatePixmap = ExaCreatePixmap;
-    pExa->DestroyPixmap = ExaDestroyPixmap;
-    pExa->ModifyPixmapHeader = ExaModifyPixmapHeader;
-
-    if (!exaDriverInit(pScrn->pScreen, pExa)) {
-	goto out_err;
-    }
+   if (exa->ctx)
+      exa->ctx->destroy(exa->ctx);
 
-    exa->scrn = ms->screen;
-    exa->ctx = drm_api_hooks.create_context(exa->scrn);
-    /* Share context with DRI */
-    ms->ctx = exa->ctx;
+   exaDriverFini(pScrn->pScreen);
+   xfree(exa);
+   ms->exa = NULL;
+}
 
-    return (void *)exa;
+void *
+xorg_exa_init(ScrnInfoPtr pScrn)
+{
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa;
+   ExaDriverPtr pExa;
+
+   exa = xcalloc(1, sizeof(struct exa_context));
+   if (!exa)
+      return NULL;
+
+   pExa = exaDriverAlloc();
+   if (!pExa) {
+      goto out_err;
+   }
+
+   memset(pExa, 0, sizeof(*pExa));
+
+   pExa->exa_major         = 2;
+   pExa->exa_minor         = 2;
+   pExa->memoryBase        = 0;
+   pExa->memorySize        = 0;
+   pExa->offScreenBase     = 0;
+   pExa->pixmapOffsetAlign = 0;
+   pExa->pixmapPitchAlign  = 1;
+   pExa->flags             = EXA_OFFSCREEN_PIXMAPS | EXA_HANDLES_PIXMAPS;
+   pExa->maxX              = 8191; /* FIXME */
+   pExa->maxY              = 8191; /* FIXME */
+
+   pExa->WaitMarker         = ExaWaitMarker;
+   pExa->MarkSync           = ExaMarkSync;
+   pExa->PrepareSolid       = ExaPrepareSolid;
+   pExa->Solid              = ExaSolid;
+   pExa->DoneSolid          = ExaDone;
+   pExa->PrepareCopy        = ExaPrepareCopy;
+   pExa->Copy               = ExaCopy;
+   pExa->DoneCopy           = ExaDone;
+   pExa->CheckComposite     = ExaCheckComposite;
+   pExa->PrepareComposite   = ExaPrepareComposite;
+   pExa->Composite          = ExaComposite;
+   pExa->DoneComposite      = ExaDoneComposite;
+   pExa->PixmapIsOffscreen  = ExaPixmapIsOffscreen;
+   pExa->DownloadFromScreen = ExaDownloadFromScreen;
+   pExa->UploadToScreen     = ExaUploadToScreen;
+   pExa->PrepareAccess      = ExaPrepareAccess;
+   pExa->FinishAccess       = ExaFinishAccess;
+   pExa->CreatePixmap       = ExaCreatePixmap;
+   pExa->DestroyPixmap      = ExaDestroyPixmap;
+   pExa->ModifyPixmapHeader = ExaModifyPixmapHeader;
+
+   if (!exaDriverInit(pScrn->pScreen, pExa)) {
+      goto out_err;
+   }
+
+   exa->scrn = ms->screen;
+   exa->ctx = ms->api->create_context(ms->api, exa->scrn);
+   /* Share context with DRI */
+   ms->ctx = exa->ctx;
+
+   exa->cso = cso_create_context(exa->ctx);
+   exa->shaders = xorg_shaders_create(exa);
+
+   return (void *)exa;
+
+out_err:
+   xorg_exa_close(pScrn);
+
+   return NULL;
+}
 
-  out_err:
-    xorg_exa_close(pScrn);
+struct pipe_surface *
+exa_gpu_surface(struct exa_context *exa, struct exa_pixmap_priv *priv)
+{
+   return exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_READ |
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
 
-    return NULL;
 }
 
-/* vim: set sw=4 ts=8 sts=4: */
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h
new file mode 100644
index 0000000000..5b515be139
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_exa.h
@@ -0,0 +1,50 @@
+#ifndef XORG_EXA_H
+#define XORG_EXA_H
+
+#include "xorg_tracker.h"
+
+#include <pipe/p_state.h>
+
+struct cso_context;
+struct xorg_shaders;
+
+/* src + mask + dst */
+#define MAX_EXA_SAMPLERS 3
+
+struct exa_context
+{
+   ExaDriverPtr pExa;
+   struct pipe_context *ctx;
+   struct pipe_screen *scrn;
+   struct cso_context *cso;
+   struct xorg_shaders *shaders;
+
+   struct pipe_constant_buffer vs_const_buffer;
+   struct pipe_constant_buffer fs_const_buffer;
+
+   struct pipe_texture *bound_textures[MAX_EXA_SAMPLERS];
+   int num_bound_samplers;
+
+   float solid_color[4];
+};
+
+
+struct exa_pixmap_priv
+{
+   int flags;
+   int tex_flags;
+
+   struct pipe_texture *tex;
+   struct pipe_texture *depth_stencil_tex;
+   unsigned int color;
+   struct pipe_surface *src_surf; /* for copies */
+
+   struct pipe_transfer *map_transfer;
+   unsigned map_count;
+};
+
+struct pipe_surface *
+exa_gpu_surface(struct exa_context *exa, struct exa_pixmap_priv *priv);
+
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
new file mode 100644
index 0000000000..cfee10c3b3
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
@@ -0,0 +1,489 @@
+#include "xorg_exa_tgsi.h"
+
+/*### stupidity defined in X11/extensions/XI.h */
+#undef Absolute
+
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+#include "util/u_simple_shaders.h"
+
+#include "tgsi/tgsi_ureg.h"
+
+#include "cso_cache/cso_context.h"
+#include "cso_cache/cso_hash.h"
+
+/* Vertex shader:
+ * IN[0]    = vertex pos
+ * IN[1]    = src tex coord | solid fill color
+ * IN[2]    = mask tex coord
+ * IN[3]    = dst tex coord
+ * CONST[0] = (2/dst_width, 2/dst_height, 1, 1)
+ * CONST[1] = (-1, -1, 0, 0)
+ *
+ * OUT[0]   = vertex pos
+ * OUT[1]   = src tex coord | solid fill color
+ * OUT[2]   = mask tex coord
+ * OUT[3]   = dst tex coord
+ */
+
+/* Fragment shader:
+ * SAMP[0]  = src
+ * SAMP[1]  = mask
+ * SAMP[2]  = dst
+ * IN[0]    = pos src | solid fill color
+ * IN[1]    = pos mask
+ * IN[2]    = pos dst
+ * CONST[0] = (0, 0, 0, 1)
+ *
+ * OUT[0] = color
+ */
+
+struct xorg_shaders {
+   struct exa_context *exa;
+
+   struct cso_hash *vs_hash;
+   struct cso_hash *fs_hash;
+};
+
+static const char over_op[] =
+   "SUB TEMP[3], CONST[0].wwww, TEMP[1].wwww\n"
+   "MUL TEMP[3], TEMP[0], TEMP[3]\n"
+   "ADD TEMP[0], TEMP[3], TEMP[0]\n";
+
+
+static INLINE void
+create_preamble(struct ureg_program *ureg)
+{
+}
+
+
+static INLINE void
+src_in_mask(struct ureg_program *ureg,
+            struct ureg_dst dst,
+            struct ureg_src src,
+            struct ureg_src mask)
+{
+   /* MUL dst, src, mask.wwww */
+   ureg_MUL(ureg, dst, src,
+            ureg_scalar(mask, TGSI_SWIZZLE_W));
+}
+
+static struct ureg_src
+vs_normalize_coords(struct ureg_program *ureg, struct ureg_src coords,
+                    struct ureg_src const0, struct ureg_src const1)
+{
+   struct ureg_dst tmp = ureg_DECL_temporary(ureg);
+   struct ureg_src ret;
+   ureg_MUL(ureg, tmp, coords, const0);
+   ureg_ADD(ureg, tmp, ureg_src(tmp), const1);
+   ret = ureg_src(tmp);
+   ureg_release_temporary(ureg, tmp);
+   return ret;
+}
+
+static void
+linear_gradient(struct ureg_program *ureg,
+                struct ureg_dst out,
+                struct ureg_src pos,
+                struct ureg_src sampler,
+                struct ureg_src coords,
+                struct ureg_src const0124,
+                struct ureg_src matrow0,
+                struct ureg_src matrow1,
+                struct ureg_src matrow2)
+{
+   struct ureg_dst temp0 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp1 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp2 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp3 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp4 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp5 = ureg_DECL_temporary(ureg);
+
+   ureg_MOV(ureg,
+            ureg_writemask(temp0, TGSI_WRITEMASK_XY), pos);
+   ureg_MOV(ureg,
+            ureg_writemask(temp0, TGSI_WRITEMASK_Z),
+            ureg_scalar(const0124, TGSI_SWIZZLE_Y));
+
+   ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0));
+   ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0));
+   ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0));
+   ureg_RCP(ureg, temp3, ureg_src(temp3));
+   ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3));
+   ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3));
+
+   ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_X),
+            ureg_src(temp1));
+   ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_Y),
+            ureg_src(temp2));
+
+   ureg_MUL(ureg, temp0,
+            ureg_scalar(coords, TGSI_SWIZZLE_Y),
+            ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_Y));
+   ureg_MAD(ureg, temp1,
+            ureg_scalar(coords, TGSI_SWIZZLE_X),
+            ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_X),
+            ureg_src(temp0));
+
+   ureg_MUL(ureg, temp2,
+            ureg_src(temp1),
+            ureg_scalar(coords, TGSI_SWIZZLE_Z));
+
+   ureg_TEX(ureg, out,
+            TGSI_TEXTURE_1D, ureg_src(temp2), sampler);
+
+   ureg_release_temporary(ureg, temp0);
+   ureg_release_temporary(ureg, temp1);
+   ureg_release_temporary(ureg, temp2);
+   ureg_release_temporary(ureg, temp3);
+   ureg_release_temporary(ureg, temp4);
+   ureg_release_temporary(ureg, temp5);
+}
+
+
+static void
+radial_gradient(struct ureg_program *ureg,
+                struct ureg_dst out,
+                struct ureg_src pos,
+                struct ureg_src sampler,
+                struct ureg_src coords,
+                struct ureg_src const0124,
+                struct ureg_src matrow0,
+                struct ureg_src matrow1,
+                struct ureg_src matrow2)
+{
+   struct ureg_dst temp0 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp1 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp2 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp3 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp4 = ureg_DECL_temporary(ureg);
+   struct ureg_dst temp5 = ureg_DECL_temporary(ureg);
+
+   ureg_MOV(ureg,
+            ureg_writemask(temp0, TGSI_WRITEMASK_XY),
+            pos);
+   ureg_MOV(ureg,
+            ureg_writemask(temp0, TGSI_WRITEMASK_Z),
+            ureg_scalar(const0124, TGSI_SWIZZLE_Y));
+
+   ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0));
+   ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0));
+   ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0));
+   ureg_RCP(ureg, temp3, ureg_src(temp3));
+   ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3));
+   ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3));
+
+   ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_X),
+            ureg_src(temp1));
+   ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_Y),
+            ureg_src(temp2));
+
+   ureg_MUL(ureg, temp0, ureg_scalar(coords, TGSI_SWIZZLE_Y),
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y));
+   ureg_MAD(ureg, temp1,
+            ureg_scalar(coords, TGSI_SWIZZLE_X),
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X),
+            ureg_src(temp0));
+   ureg_ADD(ureg, temp1,
+            ureg_src(temp1), ureg_src(temp1));
+   ureg_MUL(ureg, temp3,
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y),
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y));
+   ureg_MAD(ureg, temp4,
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X),
+            ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X),
+            ureg_src(temp3));
+   ureg_MOV(ureg, temp4, ureg_negate(ureg_src(temp4)));
+   ureg_MUL(ureg, temp2,
+            ureg_scalar(coords, TGSI_SWIZZLE_Z),
+            ureg_src(temp4));
+   ureg_MUL(ureg, temp0,
+            ureg_scalar(const0124, TGSI_SWIZZLE_W),
+            ureg_src(temp2));
+   ureg_MUL(ureg, temp3,
+            ureg_src(temp1), ureg_src(temp1));
+   ureg_SUB(ureg, temp2,
+            ureg_src(temp3), ureg_src(temp0));
+   ureg_RSQ(ureg, temp2, ureg_abs(ureg_src(temp2)));
+   ureg_RCP(ureg, temp2, ureg_src(temp2));
+   ureg_SUB(ureg, temp1,
+            ureg_src(temp2), ureg_src(temp1));
+   ureg_ADD(ureg, temp0,
+            ureg_scalar(coords, TGSI_SWIZZLE_Z),
+            ureg_scalar(coords, TGSI_SWIZZLE_Z));
+   ureg_RCP(ureg, temp0, ureg_src(temp0));
+   ureg_MUL(ureg, temp2,
+            ureg_src(temp1), ureg_src(temp0));
+   ureg_TEX(ureg, out, TGSI_TEXTURE_1D,
+            ureg_src(temp2), sampler);
+
+   ureg_release_temporary(ureg, temp0);
+   ureg_release_temporary(ureg, temp1);
+   ureg_release_temporary(ureg, temp2);
+   ureg_release_temporary(ureg, temp3);
+   ureg_release_temporary(ureg, temp4);
+   ureg_release_temporary(ureg, temp5);
+}
+
+static void *
+create_vs(struct pipe_context *pipe,
+          unsigned vs_traits)
+{
+   struct ureg_program *ureg;
+   struct ureg_src src;
+   struct ureg_dst dst;
+   struct ureg_src const0, const1;
+   boolean is_fill = vs_traits & VS_FILL;
+   boolean is_composite = vs_traits & VS_COMPOSITE;
+   boolean has_mask = vs_traits & VS_MASK;
+
+   ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
+   if (ureg == NULL)
+      return 0;
+
+   const0 = ureg_DECL_constant(ureg);
+   const1 = ureg_DECL_constant(ureg);
+
+   /* it has to be either a fill or a composite op */
+   debug_assert(is_fill ^ is_composite);
+
+   src = ureg_DECL_vs_input(ureg,
+                            TGSI_SEMANTIC_POSITION, 0);
+   dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
+   src = vs_normalize_coords(ureg, src,
+                             const0, const1);
+   ureg_MOV(ureg, dst, src);
+
+
+   if (is_composite) {
+      src = ureg_DECL_vs_input(ureg,
+                               TGSI_SEMANTIC_GENERIC, 1);
+      dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 1);
+      ureg_MOV(ureg, dst, src);
+   }
+   if (is_fill) {
+      src = ureg_DECL_vs_input(ureg,
+                               TGSI_SEMANTIC_COLOR, 1);
+      dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 1);
+      ureg_MOV(ureg, dst, src);
+   }
+
+   if (has_mask) {
+      src = ureg_DECL_vs_input(ureg,
+                               TGSI_SEMANTIC_GENERIC, 2);
+      dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 2);
+      ureg_MOV(ureg, dst, src);
+   }
+
+   ureg_END(ureg);
+
+   return ureg_create_shader_and_destroy(ureg, pipe);
+}
+
+static void *
+create_fs(struct pipe_context *pipe,
+          unsigned fs_traits)
+{
+   struct ureg_program *ureg;
+   struct ureg_src /*dst_sampler,*/ src_sampler, mask_sampler;
+   struct ureg_src /*dst_pos,*/ src_input, mask_pos;
+   struct ureg_dst src, mask;
+   struct ureg_dst out;
+   boolean has_mask = fs_traits & FS_MASK;
+   boolean is_fill = fs_traits & FS_FILL;
+   boolean is_composite = fs_traits & FS_COMPOSITE;
+   boolean is_solid   = fs_traits & FS_SOLID_FILL;
+   boolean is_lingrad = fs_traits & FS_LINGRAD_FILL;
+   boolean is_radgrad = fs_traits & FS_RADGRAD_FILL;
+
+   ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
+   if (ureg == NULL)
+      return 0;
+
+   /* it has to be either a fill or a composite op */
+   debug_assert(is_fill ^ is_composite);
+
+   out = ureg_DECL_output(ureg,
+                          TGSI_SEMANTIC_COLOR,
+                          0);
+
+   if (is_composite) {
+      src_sampler = ureg_DECL_sampler(ureg, 0);
+      src_input = ureg_DECL_fs_input(ureg,
+                                     TGSI_SEMANTIC_POSITION,
+                                     0,
+                                     TGSI_INTERPOLATE_PERSPECTIVE);
+   }
+   if (is_fill) {
+      if (is_solid)
+         src_input = ureg_DECL_fs_input(ureg,
+                                        TGSI_SEMANTIC_COLOR,
+                                        0,
+                                        TGSI_INTERPOLATE_PERSPECTIVE);
+      else
+         src_input = ureg_DECL_fs_input(ureg,
+                                        TGSI_SEMANTIC_POSITION,
+                                        0,
+                                        TGSI_INTERPOLATE_PERSPECTIVE);
+   }
+
+   if (has_mask) {
+      mask_sampler = ureg_DECL_sampler(ureg, 1);
+      mask_pos = ureg_DECL_fs_input(ureg,
+                                    TGSI_SEMANTIC_POSITION,
+                                    1,
+                                    TGSI_INTERPOLATE_PERSPECTIVE);
+   }
+
+#if 0  /* unused right now */
+   dst_sampler = ureg_DECL_sampler(ureg, 2);
+   dst_pos = ureg_DECL_fs_input(ureg,
+                                TGSI_SEMANTIC_POSITION,
+                                2,
+                                TGSI_INTERPOLATE_PERSPECTIVE);
+#endif
+
+   if (is_composite) {
+      if (has_mask)
+         src = ureg_DECL_temporary(ureg);
+      else
+         src = out;
+      ureg_TEX(ureg, src,
+               TGSI_TEXTURE_2D, src_input, src_sampler);
+   } else if (is_fill) {
+      if (is_solid) {
+         if (has_mask)
+            src = ureg_dst(src_input);
+         else
+            ureg_MOV(ureg, out, src_input);
+      } else if (is_lingrad || is_radgrad) {
+         struct ureg_src coords, const0124,
+            matrow0, matrow1, matrow2;
+
+         if (has_mask)
+            src = ureg_DECL_temporary(ureg);
+         else
+            src = out;
+
+         coords = ureg_DECL_constant(ureg);
+         const0124 = ureg_DECL_constant(ureg);
+         matrow0 = ureg_DECL_constant(ureg);
+         matrow1 = ureg_DECL_constant(ureg);
+         matrow2 = ureg_DECL_constant(ureg);
+
+         if (is_lingrad) {
+            linear_gradient(ureg, src,
+                            src_input, src_sampler,
+                            coords, const0124,
+                            matrow0, matrow1, matrow2);
+         } else if (is_radgrad) {
+            radial_gradient(ureg, src,
+                            src_input, src_sampler,
+                            coords, const0124,
+                            matrow0, matrow1, matrow2);
+         }
+      } else
+         debug_assert(!"Unknown fill type!");
+   }
+
+   if (has_mask) {
+      mask = ureg_DECL_temporary(ureg);
+      ureg_TEX(ureg, mask,
+               TGSI_TEXTURE_2D, mask_pos, mask_sampler);
+      /* src IN mask */
+      src_in_mask(ureg, out, ureg_src(src), ureg_src(mask));
+      ureg_release_temporary(ureg, mask);
+   }
+
+   ureg_END(ureg);
+
+   return ureg_create_shader_and_destroy(ureg, pipe);
+}
+
+struct xorg_shaders * xorg_shaders_create(struct exa_context *exa)
+{
+   struct xorg_shaders *sc = CALLOC_STRUCT(xorg_shaders);
+
+   sc->exa = exa;
+   sc->vs_hash = cso_hash_create();
+   sc->fs_hash = cso_hash_create();
+
+   return sc;
+}
+
+static void
+cache_destroy(struct cso_context *cso,
+              struct cso_hash *hash,
+              unsigned processor)
+{
+   struct cso_hash_iter iter = cso_hash_first_node(hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      void *shader = (void *)cso_hash_iter_data(iter);
+      if (processor == PIPE_SHADER_FRAGMENT) {
+         cso_delete_fragment_shader(cso, shader);
+      } else if (processor == PIPE_SHADER_VERTEX) {
+         cso_delete_vertex_shader(cso, shader);
+      }
+      iter = cso_hash_erase(hash, iter);
+   }
+   cso_hash_delete(hash);
+}
+
+void xorg_shaders_destroy(struct xorg_shaders *sc)
+{
+   cache_destroy(sc->exa->cso, sc->vs_hash,
+                 PIPE_SHADER_VERTEX);
+   cache_destroy(sc->exa->cso, sc->fs_hash,
+                 PIPE_SHADER_FRAGMENT);
+
+   free(sc);
+}
+
+static INLINE void *
+shader_from_cache(struct pipe_context *pipe,
+                  unsigned type,
+                  struct cso_hash *hash,
+                  unsigned key)
+{
+   void *shader = 0;
+
+   struct cso_hash_iter iter = cso_hash_find(hash, key);
+
+   if (cso_hash_iter_is_null(iter)) {
+      if (type == PIPE_SHADER_VERTEX)
+         shader = create_vs(pipe, key);
+      else
+         shader = create_fs(pipe, key);
+      cso_hash_insert(hash, key, shader);
+   } else
+      shader = (void *)cso_hash_iter_data(iter);
+
+   return shader;
+}
+
+struct xorg_shader xorg_shaders_get(struct xorg_shaders *sc,
+                                    unsigned vs_traits,
+                                    unsigned fs_traits)
+{
+   struct xorg_shader shader = {0};
+   void *vs, *fs;
+
+   vs = shader_from_cache(sc->exa->ctx, PIPE_SHADER_VERTEX,
+                          sc->vs_hash, vs_traits);
+   fs = shader_from_cache(sc->exa->ctx, PIPE_SHADER_FRAGMENT,
+                          sc->fs_hash, fs_traits);
+
+   debug_assert(vs && fs);
+   if (!vs || !fs)
+      return shader;
+
+   shader.vs = vs;
+   shader.fs = fs;
+
+   return shader;
+}
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h
new file mode 100644
index 0000000000..1535a0c8c3
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h
@@ -0,0 +1,43 @@
+#ifndef XORG_EXA_TGSI_H
+#define XORG_EXA_TGSI_H
+
+#include "xorg_exa.h"
+
+enum xorg_vs_traits {
+   VS_COMPOSITE        = 1 << 0,
+   VS_MASK             = 1 << 1,
+   VS_SOLID_FILL       = 1 << 2,
+   VS_LINGRAD_FILL     = 1 << 3,
+   VS_RADGRAD_FILL     = 1 << 4,
+   VS_FILL             = (VS_SOLID_FILL |
+                          VS_LINGRAD_FILL |
+                          VS_RADGRAD_FILL)
+   /*VS_TRANSFORM      = 1 << 5*/
+};
+
+enum xorg_fs_traits {
+   FS_COMPOSITE        = 1 << 0,
+   FS_MASK             = 1 << 1,
+   FS_SOLID_FILL       = 1 << 2,
+   FS_LINGRAD_FILL     = 1 << 3,
+   FS_RADGRAD_FILL     = 1 << 4,
+   FS_FILL             = (FS_SOLID_FILL |
+                          FS_LINGRAD_FILL |
+                          FS_RADGRAD_FILL)
+};
+
+struct xorg_shader {
+   void *fs;
+   void *vs;
+};
+
+struct xorg_shaders;
+
+struct xorg_shaders *xorg_shaders_create(struct exa_context *exa);
+void xorg_shaders_destroy(struct xorg_shaders *shaders);
+
+struct xorg_shader xorg_shaders_get(struct xorg_shaders *shaders,
+                                    unsigned vs_traits,
+                                    unsigned fs_traits);
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h
index 82c3890dfb..b1ab783a15 100644
--- a/src/gallium/state_trackers/xorg/xorg_tracker.h
+++ b/src/gallium/state_trackers/xorg/xorg_tracker.h
@@ -31,11 +31,20 @@
 #ifndef _XORG_TRACKER_H_
 #define _XORG_TRACKER_H_
 
+#include <stddef.h>
+#include <stdint.h>
 #include <errno.h>
 #include <drm.h>
 #include <xf86drm.h>
 #include <xf86drmMode.h>
-#include "exa.h"
+#include <xorg-server.h>
+#include <xf86.h>
+#include "xf86Crtc.h"
+#include <exa.h>
+
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+#include <damage.h>
+#endif
 
 #include "pipe/p_screen.h"
 #include "state_tracker/drm_api.h"
@@ -72,16 +81,23 @@ typedef struct _modesettingRec
 
     unsigned int SaveGeneration;
 
+    void (*blockHandler)(int, pointer, pointer, pointer);
     CreateScreenResourcesProcPtr createScreenResources;
 
     /* gallium */
+    struct drm_api *api;
     struct pipe_screen *screen;
     struct pipe_context *ctx;
+    boolean d_depth_bits_last;
+    boolean ds_depth_bits_last;
 
     /* exa */
     void *exa;
     Bool noEvict;
 
+#ifdef DRM_MODE_FEATURE_DIRTYFB
+    DamagePtr damage;
+#endif
 } modesettingRec, *modesettingPtr;
 
 #define modesettingPTR(p) ((modesettingPtr)((p)->driverPrivate))
@@ -94,7 +110,13 @@ struct pipe_texture *
 xorg_exa_get_texture(PixmapPtr pPixmap);
 
 unsigned
-xorg_exa_get_pixmap_handle(PixmapPtr pPixmap);
+xorg_exa_get_pixmap_handle(PixmapPtr pPixmap, unsigned *stride);
+
+int
+xorg_exa_set_displayed_usage(PixmapPtr pPixmap);
+
+int
+xorg_exa_set_shared_usage(PixmapPtr pPixmap);
 
 void *
 xorg_exa_init(ScrnInfoPtr pScrn);
@@ -119,6 +141,9 @@ driCloseScreen(ScreenPtr pScreen);
 void
 crtc_init(ScrnInfoPtr pScrn);
 
+void
+crtc_cursor_destroy(xf86CrtcPtr crtc);
+
 
 /***********************************************************************
  * xorg_output.c
diff --git a/src/gallium/state_trackers/xorg/xorg_winsys.h b/src/gallium/state_trackers/xorg/xorg_winsys.h
index d523080e90..47ee4b9ffd 100644
--- a/src/gallium/state_trackers/xorg/xorg_winsys.h
+++ b/src/gallium/state_trackers/xorg/xorg_winsys.h
@@ -37,7 +37,6 @@
 
 #include "xorg-server.h"
 #include "xf86.h"
-#include "xf86Resources.h"
 #include "pciaccess.h"
 
 #ifndef XSERVER_LIBPCIACCESS
@@ -46,6 +45,5 @@
 
 void xorg_tracker_set_functions(ScrnInfoPtr scrn);
 const OptionInfoRec * xorg_tracker_available_options(int chipid, int busid);
-void xorg_tracker_loader_ref_sym_lists(void);
 
 #endif
diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template
index 9f92cb4207..9635c3c50e 100644
--- a/src/gallium/winsys/drm/Makefile.template
+++ b/src/gallium/winsys/drm/Makefile.template
@@ -83,7 +83,9 @@ default: depend symlinks $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME)
 
 $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
 	$(MKLIB) -noprefix -o $@ \
-		$(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES)  $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS)
+		$(OBJECTS) $(PIPE_DRIVERS) \
+                -Wl,--start-group $(MESA_MODULES) -Wl,--end-group \
+                 $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS)
 
 $(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS)
 	$(MKLIB) -o $(LIBNAME_EGL) \
@@ -118,7 +120,7 @@ clean:
 
 install: $(LIBNAME)
 	$(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
 
 
 include depend
diff --git a/src/gallium/winsys/drm/intel/SConscript b/src/gallium/winsys/drm/intel/SConscript
new file mode 100644
index 0000000000..50d7b75ed6
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/SConscript
@@ -0,0 +1,7 @@
+Import('*')
+
+SConscript(['gem/SConscript',])
+
+if 'mesa' in env['statetrackers']:
+
+    SConscript(['dri/SConscript'])
diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile
index ac0891a646..5e212b62a4 100644
--- a/src/gallium/winsys/drm/intel/dri/Makefile
+++ b/src/gallium/winsys/drm/intel/dri/Makefile
@@ -6,7 +6,9 @@ LIBNAME = i915_dri.so
 PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/identity/libidentity.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
 
 
diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript
new file mode 100644
index 0000000000..6c00861f51
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/SConscript
@@ -0,0 +1,19 @@
+Import('*')
+
+env = drienv.Clone()
+
+env.ParseConfig('pkg-config --cflags --libs libdrm_intel')
+
+drivers = [
+    st_dri,
+    inteldrm,
+    softpipe,
+    i915simple,
+    trace,
+]
+
+env.SharedLibrary(
+    target ='i915_dri.so',
+    source = COMMON_GALLIUM_SOURCES,
+    LIBS = drivers + mesa + auxiliaries + env['LIBS'],
+)
diff --git a/src/gallium/winsys/drm/intel/egl/Makefile b/src/gallium/winsys/drm/intel/egl/Makefile
index c5217ad2d6..490baded66 100644
--- a/src/gallium/winsys/drm/intel/egl/Makefile
+++ b/src/gallium/winsys/drm/intel/egl/Makefile
@@ -8,6 +8,7 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/egl/libegldrm.a \
 	$(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
 
 DRIVER_SOURCES =
diff --git a/src/gallium/winsys/drm/intel/gem/Makefile b/src/gallium/winsys/drm/intel/gem/Makefile
index 7ab1a2a771..0d6d4e37db 100644
--- a/src/gallium/winsys/drm/intel/gem/Makefile
+++ b/src/gallium/winsys/drm/intel/gem/Makefile
@@ -4,10 +4,10 @@ include $(TOP)/configs/current
 LIBNAME = inteldrm
 
 C_SOURCES = \
-	intel_be_batchbuffer.c \
-	intel_be_context.c \
-	intel_be_device.c \
-	intel_be_api.c
+	intel_drm_batchbuffer.c \
+	intel_drm_buffer.c \
+	intel_drm_fence.c \
+	intel_drm_api.c
 
 LIBRARY_INCLUDES = $(shell pkg-config libdrm --cflags-only-I)
 
diff --git a/src/gallium/winsys/drm/intel/gem/SConscript b/src/gallium/winsys/drm/intel/gem/SConscript
new file mode 100644
index 0000000000..26717f391f
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/SConscript
@@ -0,0 +1,17 @@
+Import('*')
+
+env = drienv.Clone()
+
+inteldrm_sources = [
+    'intel_drm_api.c',
+    'intel_drm_batchbuffer.c',
+    'intel_drm_buffer.c',
+    'intel_drm_fence.c',
+]
+
+inteldrm = env.ConvenienceLibrary(
+    target ='inteldrm',
+    source = inteldrm_sources,
+)
+
+Export('inteldrm')
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c
deleted file mode 100644
index f4ef7c2d88..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_api.c
+++ /dev/null
@@ -1,15 +0,0 @@
-
-#include "intel_be_api.h"
-#include "i915simple/i915_winsys.h"
-
-struct drm_api drm_api_hooks =
-{
-	/* intel_be_context.c */
-	.create_context = intel_be_create_context,
-	/* intel_be_device.c */
-	.create_screen = intel_be_create_screen,
-	.buffer_from_texture = i915_get_texture_buffer,
-	.buffer_from_handle = intel_be_buffer_from_handle,
-	.handle_from_buffer = intel_be_handle_from_buffer,
-	.global_handle_from_buffer = intel_be_global_handle_from_buffer,
-};
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h
deleted file mode 100644
index 1c622f3b97..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h
+++ /dev/null
@@ -1,15 +0,0 @@
-
-#ifndef _INTEL_BE_API_H_
-#define _INTEL_BE_API_H_
-
-#include "pipe/p_compiler.h"
-
-#include "state_tracker/drm_api.h"
-
-#include "intel_be_device.h"
-
-struct pipe_screen *intel_be_create_screen(int drmFD,
-					   struct drm_create_screen_arg *arg);
-struct pipe_context *intel_be_create_context(struct pipe_screen *screen);
-
-#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c
deleted file mode 100644
index d5e63c3bae..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c
+++ /dev/null
@@ -1,139 +0,0 @@
-
-#include "i915simple/i915_debug.h"
-#include "intel_be_batchbuffer.h"
-#include "intel_be_context.h"
-#include "intel_be_device.h"
-#include "intel_be_fence.h"
-#include <errno.h>
-
-#include "util/u_memory.h"
-
-struct intel_be_batchbuffer *
-intel_be_batchbuffer_alloc(struct intel_be_context *intel)
-{
-	struct intel_be_batchbuffer *batch = CALLOC_STRUCT(intel_be_batchbuffer);
-
-
-	batch->base.buffer = NULL;
-	batch->base.winsys = &intel->base;
-	batch->base.map = NULL;
-	batch->base.ptr = NULL;
-	batch->base.size = 0;
-	batch->base.actual_size = intel->device->max_batch_size;
-	batch->base.relocs = 0;
-	batch->base.max_relocs = 500;/*INTEL_DEFAULT_RELOCS;*/
-
-	batch->base.map = malloc(batch->base.actual_size);
-	memset(batch->base.map, 0, batch->base.actual_size);
-
-	batch->base.ptr = batch->base.map;
-
-	intel_be_batchbuffer_reset(batch);
-
-	return batch;
-}
-
-void
-intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch)
-{
-	struct intel_be_context *intel = intel_be_context(batch->base.winsys);
-	struct intel_be_device *dev = intel->device;
-
-	if (batch->bo)
-		drm_intel_bo_unreference(batch->bo);
-
-	memset(batch->base.map, 0, batch->base.actual_size);
-	batch->base.ptr = batch->base.map;
-	batch->base.size = batch->base.actual_size - BATCH_RESERVED;
-
-	batch->base.relocs = 0;
-
-	batch->bo = drm_intel_bo_alloc(dev->pools.gem,
-	                               "gallium3d_batch_buffer",
-	                               batch->base.actual_size, 0);
-}
-
-int
-intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
-			   unsigned pre_add,
-			   drm_intel_bo *bo,
-			   uint32_t read_domains,
-			   uint32_t write_domain)
-{
-	unsigned offset;
-	int ret = 0;
-
-	assert(batch->base.relocs < batch->base.max_relocs);
-
-	offset = (unsigned)(batch->base.ptr - batch->base.map);
-
-	ret = drm_intel_bo_emit_reloc(batch->bo, offset,
-	                              bo, pre_add,
-	                              read_domains,
-	                              write_domain);
-
-	((uint32_t*)batch->base.ptr)[0] = bo->offset + pre_add;
-	batch->base.ptr += 4;
-
-	if (!ret)
-		batch->base.relocs++;
-
-	return ret;
-}
-
-void
-intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch,
-			   struct intel_be_fence **fence)
-{
-	struct i915_batchbuffer *i915 = &batch->base;
-	unsigned used = 0;
-	int ret = 0;
-
-	assert(i915_batchbuffer_space(i915) >= 0);
-
-	used = batch->base.ptr - batch->base.map;
-	assert((used & 3) == 0);
-
-	if (used & 4) {
-		i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); // MI_FLUSH | FLUSH_MAP_CACHE;
-		i915_batchbuffer_dword(i915, (0x0<<29)|(0x0<<23)); // MI_NOOP
-		i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END;
-	} else {
-		i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); //MI_FLUSH | FLUSH_MAP_CACHE;
-		i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END;
-	}
-
-	used = batch->base.ptr - batch->base.map;
-
-	drm_intel_bo_subdata(batch->bo, 0, used, batch->base.map);
-	ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0);
-
-	assert(ret == 0);
-
-	intel_be_batchbuffer_reset(batch);
-
-	if (fence) {
-		if (*fence)
-			intel_be_fence_reference(fence, NULL);
-
-		(*fence) = CALLOC_STRUCT(intel_be_fence);
-		pipe_reference_init(&(*fence)->reference, 1);
-		(*fence)->bo = NULL;
-	}
-}
-
-void
-intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch)
-{
-
-}
-
-void
-intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch)
-{
-	if (batch->bo)
-		drm_intel_bo_unreference(batch->bo);
-
-	free(batch->base.map);
-	free(batch);
-}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h
deleted file mode 100644
index 195bf8dee7..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h
+++ /dev/null
@@ -1,55 +0,0 @@
-
-#ifndef INTEL_BE_BATCHBUFFER_H
-#define INTEL_BE_BATCHBUFFER_H
-
-#include "i915simple/i915_batch.h"
-
-#include "drm.h"
-#include "intel_bufmgr.h"
-
-#define BATCH_RESERVED 16
-
-#define INTEL_DEFAULT_RELOCS 100
-#define INTEL_MAX_RELOCS 400
-
-#define INTEL_BATCH_NO_CLIPRECTS 0x1
-#define INTEL_BATCH_CLIPRECTS    0x2
-
-struct intel_be_context;
-struct intel_be_device;
-struct intel_be_fence;
-
-struct intel_be_batchbuffer
-{
-	struct i915_batchbuffer base;
-
-	struct intel_be_context *intel;
-	struct intel_be_device *device;
-
-	drm_intel_bo *bo;
-};
-
-struct intel_be_batchbuffer *
-intel_be_batchbuffer_alloc(struct intel_be_context *intel);
-
-void
-intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch);
-
-void
-intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch);
-
-void
-intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch,
-			   struct intel_be_fence **fence);
-
-void
-intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch);
-
-int
-intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
-			   unsigned pre_add,
-			   drm_intel_bo *bo,
-			   uint32_t read_domains,
-			   uint32_t write_doman);
-
-#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
deleted file mode 100644
index fe0b138fbe..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c
+++ /dev/null
@@ -1,120 +0,0 @@
-
-#include "pipe/p_screen.h"
-
-#include "softpipe/sp_winsys.h"
-
-#include "intel_be_device.h"
-#include "intel_be_context.h"
-#include "intel_be_batchbuffer.h"
-
-#include "i915_drm.h"
-
-#include "intel_be_api.h"
-
-static struct i915_batchbuffer *
-intel_be_batch_get(struct i915_winsys *sws)
-{
-	struct intel_be_context *intel = intel_be_context(sws);
-	return &intel->batch->base;
-}
-
-static void
-intel_be_batch_reloc(struct i915_winsys *sws,
-		     struct pipe_buffer *buf,
-		     unsigned access_flags,
-		     unsigned delta)
-{
-	struct intel_be_context *intel = intel_be_context(sws);
-	drm_intel_bo *bo = intel_bo(buf);
-	int ret;
-	uint32_t read = 0;
-	uint32_t write = 0;
-
-	if (access_flags & I915_BUFFER_ACCESS_WRITE) {
-		write = I915_GEM_DOMAIN_RENDER;
-		read = I915_GEM_DOMAIN_RENDER;
-	}
-
-	if (access_flags & I915_BUFFER_ACCESS_READ) {
-		read |= I915_GEM_DOMAIN_VERTEX;
-	}
-
-	ret = intel_be_offset_relocation(intel->batch,
-	                                 delta,
-	                                 bo,
-	                                 read,
-	                                 write);
-    assert(ret == 0);
-
-	/* TODO change return type */
-	/* return ret; */
-}
-
-static void
-intel_be_batch_flush(struct i915_winsys *sws,
-		     struct pipe_fence_handle **fence)
-{
-	struct intel_be_context *intel = intel_be_context(sws);
-	struct intel_be_fence **f = (struct intel_be_fence **)fence;
-
-	if (fence && *fence)
-		assert(0);
-
-	intel_be_batchbuffer_flush(intel->batch, f);
-}
-
-
-/*
- * Misc functions.
- */
-
-static void
-intel_be_destroy_context(struct i915_winsys *winsys)
-{
-	struct intel_be_context *intel = intel_be_context(winsys);
-
-	intel_be_batchbuffer_free(intel->batch);
-
-	free(intel);
-}
-
-boolean
-intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device)
-{
-	assert(intel);
-	assert(device);
-	intel->device = device;
-
-	intel->base.batch_get = intel_be_batch_get;
-	intel->base.batch_reloc = intel_be_batch_reloc;
-	intel->base.batch_flush = intel_be_batch_flush;
-
-	intel->base.destroy = intel_be_destroy_context;
-
-	intel->batch = intel_be_batchbuffer_alloc(intel);
-
-	return true;
-}
-
-struct pipe_context *
-intel_be_create_context(struct pipe_screen *screen)
-{
-	struct intel_be_context *intel;
-	struct pipe_context *pipe;
-	struct intel_be_device *device = intel_be_device(screen->winsys);
-
-	intel = (struct intel_be_context *)malloc(sizeof(*intel));
-	memset(intel, 0, sizeof(*intel));
-
-	intel_be_init_context(intel, device);
-
-	if (device->softpipe)
-		pipe = softpipe_create(screen);
-	else
-		pipe = i915_create_context(screen, &device->base, &intel->base);
-
-	if (pipe)
-		pipe->priv = intel;
-
-	return pipe;
-}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.h b/src/gallium/winsys/drm/intel/gem/intel_be_context.h
deleted file mode 100644
index 5a369669c0..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_context.h
+++ /dev/null
@@ -1,31 +0,0 @@
-
-#ifndef INTEL_BE_CONTEXT_H
-#define INTEL_BE_CONTEXT_H
-
-#include "i915simple/i915_winsys.h"
-
-struct intel_be_context
-{
-	/** Interface to i915simple driver */
-	struct i915_winsys base;
-
-	struct intel_be_device *device;
-	struct intel_be_batchbuffer *batch;
-};
-
-static INLINE struct intel_be_context *
-intel_be_context(struct i915_winsys *sws)
-{
-	return (struct intel_be_context *)sws;
-}
-
-/**
- * Intialize a allocated intel_be_context struct.
- *
- * Remember to set the hardware_* functions.
- */
-boolean
-intel_be_init_context(struct intel_be_context *intel,
-		      struct intel_be_device *device);
-
-#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
deleted file mode 100644
index 907ac86637..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c
+++ /dev/null
@@ -1,347 +0,0 @@
-
-#include "intel_be_device.h"
-
-#include "pipe/internal/p_winsys_screen.h"
-#include "pipe/p_defines.h"
-#include "pipe/p_state.h"
-#include "pipe/p_inlines.h"
-#include "util/u_memory.h"
-#include "util/u_debug.h"
-
-#include "intel_be_fence.h"
-
-#include "i915simple/i915_winsys.h"
-#include "softpipe/sp_winsys.h"
-
-#include "intel_be_api.h"
-#include <stdio.h>
-
-/*
- * Buffer
- */
-
-static void *
-intel_be_buffer_map(struct pipe_winsys *winsys,
-		    struct pipe_buffer *buf,
-		    unsigned flags)
-{
-	drm_intel_bo *bo = intel_bo(buf);
-	int write = 0;
-	int ret;
-
-        if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
-           /* Remove this when drm_intel_bo_map supports DONTBLOCK 
-            */
-           return NULL;
-        }
-
-	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
-		write = 1;
-
-	ret = drm_intel_bo_map(bo, write);
-
-	if (ret)
-		return NULL;
-
-	return bo->virtual;
-}
-
-static void
-intel_be_buffer_unmap(struct pipe_winsys *winsys,
-		      struct pipe_buffer *buf)
-{
-	drm_intel_bo_unmap(intel_bo(buf));
-}
-
-static void
-intel_be_buffer_destroy(struct pipe_buffer *buf)
-{
-	drm_intel_bo_unreference(intel_bo(buf));
-	free(buf);
-}
-
-static struct pipe_buffer *
-intel_be_buffer_create(struct pipe_winsys *winsys,
-		       unsigned alignment,
-		       unsigned usage,
-		       unsigned size)
-{
-	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
-	struct intel_be_device *dev = intel_be_device(winsys);
-	drm_intel_bufmgr *pool;
-	char *name;
-
-	if (!buffer)
-		return NULL;
-
-	pipe_reference_init(&buffer->base.reference, 1);
-	buffer->base.alignment = alignment;
-	buffer->base.usage = usage;
-	buffer->base.size = size;
-	buffer->flinked = FALSE;
-	buffer->flink = 0;
-
-	if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) {
-		/* Local buffer */
-		name = "gallium3d_local";
-		pool = dev->pools.gem;
-	} else if (usage & PIPE_BUFFER_USAGE_CUSTOM) {
-		/* For vertex buffers */
-		name = "gallium3d_internal_vertex";
-		pool = dev->pools.gem;
-	} else {
-		/* Regular buffers */
-		name = "gallium3d_regular";
-		pool = dev->pools.gem;
-	}
-
-	buffer->bo = drm_intel_bo_alloc(pool, name, size, alignment);
-
-	if (!buffer->bo)
-		goto err;
-
-	return &buffer->base;
-
-err:
-	free(buffer);
-	return NULL;
-}
-
-static struct pipe_buffer *
-intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes)
-{
-	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
-	struct intel_be_device *dev = intel_be_device(winsys);
-	int ret;
-
-	if (!buffer)
-		return NULL;
-
-	pipe_reference_init(&buffer->base.reference, 1);
-	buffer->base.alignment = 0;
-	buffer->base.usage = 0;
-	buffer->base.size = bytes;
-
-	buffer->bo = drm_intel_bo_alloc(dev->pools.gem,
-	                                "gallium3d_user_buffer",
-	                                bytes, 0);
-
-	if (!buffer->bo)
-		goto err;
-
-	ret = drm_intel_bo_subdata(buffer->bo,
-	                           0, bytes, ptr);
-
-	if (ret)
-		goto err;
-
-	return &buffer->base;
-
-err:
-	free(buffer);
-	return NULL;
-}
-
-struct pipe_buffer *
-intel_be_buffer_from_handle(struct pipe_screen *screen,
-                            const char* name, unsigned handle)
-{
-	struct intel_be_device *dev = intel_be_device(screen->winsys);
-	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
-
-	if (!buffer)
-		return NULL;
-
-	buffer->bo = drm_intel_bo_gem_create_from_name(dev->pools.gem, name, handle);
-
-	if (!buffer->bo)
-		goto err;
-
-	pipe_reference_init(&buffer->base.reference, 1);
-	buffer->base.screen = screen;
-	buffer->base.alignment = buffer->bo->align;
-	buffer->base.usage = PIPE_BUFFER_USAGE_GPU_READ |
-	                     PIPE_BUFFER_USAGE_GPU_WRITE |
-	                     PIPE_BUFFER_USAGE_CPU_READ |
-	                     PIPE_BUFFER_USAGE_CPU_WRITE;
-	buffer->base.size = buffer->bo->size;
-
-	return &buffer->base;
-
-err:
-	free(buffer);
-	return NULL;
-}
-
-boolean
-intel_be_handle_from_buffer(struct pipe_screen *screen,
-                            struct pipe_buffer *buffer,
-                            unsigned *handle)
-{
-	if (!buffer)
-		return FALSE;
-
-	*handle = intel_bo(buffer)->handle;
-	return TRUE;
-}
-
-boolean
-intel_be_global_handle_from_buffer(struct pipe_screen *screen,
-				   struct pipe_buffer *buffer,
-				   unsigned *handle)
-{
-	struct intel_be_buffer *buf = intel_be_buffer(buffer);
-
-	if (!buffer)
-		return FALSE;
-
-	if (!buf->flinked) {
-		if (drm_intel_bo_flink(intel_bo(buffer), &buf->flink))
-			return FALSE;
-		buf->flinked = TRUE;
-	}
-
-	*handle = buf->flink;
-	return TRUE;
-}
-/*
- * Fence
- */
-
-static void
-intel_be_fence_refunref(struct pipe_winsys *sws,
-			 struct pipe_fence_handle **ptr,
-			 struct pipe_fence_handle *fence)
-{
-	struct intel_be_fence **p = (struct intel_be_fence **)ptr;
-	struct intel_be_fence *f = (struct intel_be_fence *)fence;
-
-        intel_be_fence_reference(p, f);
-}
-
-static int
-intel_be_fence_signalled(struct pipe_winsys *sws,
-			 struct pipe_fence_handle *fence,
-			 unsigned flag)
-{
-	assert(0);
-
-	return 0;
-}
-
-static int
-intel_be_fence_finish(struct pipe_winsys *sws,
-		      struct pipe_fence_handle *fence,
-		      unsigned flag)
-{
-	struct intel_be_fence *f = (struct intel_be_fence *)fence;
-
-	/* fence already expired */
-	if (!f->bo)
-		return 0;
-
-	drm_intel_bo_wait_rendering(f->bo);
-	drm_intel_bo_unreference(f->bo);
-	f->bo = NULL;
-
-	return 0;
-}
-
-/*
- * Misc functions
- */
-
-static void
-intel_be_destroy_winsys(struct pipe_winsys *winsys)
-{
-	struct intel_be_device *dev = intel_be_device(winsys);
-
-	drm_intel_bufmgr_destroy(dev->pools.gem);
-
-	free(dev);
-}
-
-boolean
-intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id)
-{
-	dev->fd = fd;
-	dev->id = id;
-	dev->max_batch_size = 16 * 4096;
-	dev->max_vertex_size = 128 * 4096;
-
-	dev->base.buffer_create = intel_be_buffer_create;
-	dev->base.user_buffer_create = intel_be_user_buffer_create;
-	dev->base.buffer_map = intel_be_buffer_map;
-	dev->base.buffer_unmap = intel_be_buffer_unmap;
-	dev->base.buffer_destroy = intel_be_buffer_destroy;
-
-	/* Not used anymore */
-	dev->base.surface_buffer_create = NULL;
-
-	dev->base.fence_reference = intel_be_fence_refunref;
-	dev->base.fence_signalled = intel_be_fence_signalled;
-	dev->base.fence_finish = intel_be_fence_finish;
-
-	dev->base.destroy = intel_be_destroy_winsys;
-
-	dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size);
-
-	dev->softpipe = debug_get_bool_option("INTEL_SOFTPIPE", FALSE);
-
-	return true;
-}
-
-static void
-intel_be_get_device_id(unsigned int *device_id)
-{
-   char path[512];
-   FILE *file;
-
-   /*
-    * FIXME: Fix this up to use a drm ioctl or whatever.
-    */
-
-   snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device");
-   file = fopen(path, "r");
-   if (!file) {
-      return;
-   }
-
-   fgets(path, sizeof(path), file);
-   sscanf(path, "%x", device_id);
-   fclose(file);
-}
-
-struct pipe_screen *
-intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg)
-{
-	struct intel_be_device *dev;
-	struct pipe_screen *screen;
-	unsigned int deviceID;
-
-	if (arg != NULL) {
-		switch(arg->mode) {
-		case DRM_CREATE_NORMAL:
-			break;
-		default:
-			return NULL;
-		}
-	}
-
-	/* Allocate the private area */
-	dev = malloc(sizeof(*dev));
-	if (!dev)
-		return NULL;
-	memset(dev, 0, sizeof(*dev));
-
-	intel_be_get_device_id(&deviceID);
-	intel_be_init_device(dev, drmFD, deviceID);
-
-	if (dev->softpipe) {
-		screen = softpipe_create_screen(&dev->base);
-		drm_api_hooks.buffer_from_texture = softpipe_get_texture_buffer;
-	} else
-		screen = i915_create_screen(&dev->base, deviceID);
-
-	return screen;
-}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
deleted file mode 100644
index b32637ece2..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h
+++ /dev/null
@@ -1,94 +0,0 @@
-
-#ifndef INTEL_DRM_DEVICE_H
-#define INTEL_DRM_DEVICE_H
-
-#include "pipe/internal/p_winsys_screen.h"
-#include "pipe/p_context.h"
-
-#include "drm.h"
-#include "intel_bufmgr.h"
-
-/*
- * Device
- */
-
-struct intel_be_device
-{
-	struct pipe_winsys base;
-
-	boolean softpipe;
-
-	int fd; /**< Drm file discriptor */
-
-	unsigned id;
-
-	size_t max_batch_size;
-	size_t max_vertex_size;
-
-	struct {
-		drm_intel_bufmgr *gem;
-	} pools;
-};
-
-static INLINE struct intel_be_device *
-intel_be_device(struct pipe_winsys *winsys)
-{
-	return (struct intel_be_device *)winsys;
-}
-
-boolean
-intel_be_init_device(struct intel_be_device *device, int fd, unsigned id);
-
-/*
- * Buffer
- */
-
-struct intel_be_buffer {
-	struct pipe_buffer base;
-	drm_intel_bo *bo;
-	boolean flinked;
-	unsigned flink;
-};
-
-/**
- * Create a be buffer from a drm bo handle.
- *
- * Takes a reference.
- */
-struct pipe_buffer *
-intel_be_buffer_from_handle(struct pipe_screen *screen,
-                            const char* name, unsigned handle);
-
-/**
- * Gets a handle from a buffer.
- *
- * If buffer is destroyed handle may become invalid.
- */
-boolean
-intel_be_handle_from_buffer(struct pipe_screen *screen,
-                            struct pipe_buffer *buffer,
-                            unsigned *handle);
-
-/**
- * Gets the global handle from a buffer.
- *
- * If buffer is destroyed handle may become invalid.
- */
-boolean
-intel_be_global_handle_from_buffer(struct pipe_screen *screen,
-                                   struct pipe_buffer *buffer,
-                                   unsigned *handle);
-
-static INLINE struct intel_be_buffer *
-intel_be_buffer(struct pipe_buffer *buf)
-{
-	return (struct intel_be_buffer *)buf;
-}
-
-static INLINE drm_intel_bo *
-intel_bo(struct pipe_buffer *buf)
-{
-	return intel_be_buffer(buf)->bo;
-}
-
-#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h
deleted file mode 100644
index a8abb01a9e..0000000000
--- a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h
+++ /dev/null
@@ -1,34 +0,0 @@
-
-#ifndef INTEL_BE_FENCE_H
-#define INTEL_BE_FENCE_H
-
-#include "pipe/p_defines.h"
-
-#include "drm.h"
-#include "intel_bufmgr.h"
-
-/**
- * Because gem does not have fence's we have to create our own fences.
- *
- * They work by keeping the batchbuffer around and checking if that has
- * been idled. If bo is NULL fence has expired.
- */
-struct intel_be_fence
-{
-	struct pipe_reference reference;
-	drm_intel_bo *bo;
-};
-
-static INLINE void
-intel_be_fence_reference(struct intel_be_fence **ptr, struct intel_be_fence *f)
-{
-	struct intel_be_fence *old_fence = *ptr;
-
-        if (pipe_reference((struct pipe_reference**)ptr, &f->reference)) {
-		if (old_fence->bo)
-			drm_intel_bo_unreference(old_fence->bo);
-		free(old_fence);
-	}
-}
-
-#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
new file mode 100644
index 0000000000..4c5a1d2ea8
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
@@ -0,0 +1,202 @@
+
+#include "state_tracker/drm_api.h"
+
+#include "intel_drm_winsys.h"
+#include "util/u_memory.h"
+
+#include "i915simple/i915_context.h"
+#include "i915simple/i915_screen.h"
+
+
+/*
+ * Helper functions
+ */
+
+
+static void
+intel_drm_get_device_id(unsigned int *device_id)
+{
+   char path[512];
+   FILE *file;
+   void *shutup_gcc;
+
+   /*
+    * FIXME: Fix this up to use a drm ioctl or whatever.
+    */
+
+   snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device");
+   file = fopen(path, "r");
+   if (!file) {
+      return;
+   }
+
+   shutup_gcc = fgets(path, sizeof(path), file);
+   sscanf(path, "%x", device_id);
+   fclose(file);
+}
+
+static struct intel_buffer *
+intel_drm_buffer_from_handle(struct intel_drm_winsys *idws,
+                             const char* name, unsigned handle)
+{
+   struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer);
+
+   if (!buf)
+      return NULL;
+
+   buf->magic = 0xDEAD1337;
+   buf->bo = drm_intel_bo_gem_create_from_name(idws->pools.gem, name, handle);
+   buf->flinked = TRUE;
+   buf->flink = handle;
+
+   if (!buf->bo)
+      goto err;
+
+   return (struct intel_buffer *)buf;
+
+err:
+   FREE(buf);
+   return NULL;
+}
+
+
+/*
+ * Exported functions
+ */
+
+
+static struct pipe_texture *
+intel_drm_texture_from_shared_handle(struct drm_api *api,
+                                     struct pipe_screen *screen,
+                                     struct pipe_texture *templ,
+                                     const char* name,
+                                     unsigned pitch,
+                                     unsigned handle)
+{
+   struct intel_drm_winsys *idws = intel_drm_winsys(i915_screen(screen)->iws);
+   struct intel_buffer *buffer;
+
+   buffer = intel_drm_buffer_from_handle(idws, name, handle);
+   if (!buffer)
+      return NULL;
+
+   return i915_texture_blanket_intel(screen, templ, pitch, buffer);
+}
+
+static boolean
+intel_drm_shared_handle_from_texture(struct drm_api *api,
+                                     struct pipe_screen *screen,
+                                     struct pipe_texture *texture,
+                                     unsigned *pitch,
+                                     unsigned *handle)
+{
+   struct intel_drm_buffer *buf = NULL;
+   struct intel_buffer *buffer = NULL;
+   if (!i915_get_texture_buffer_intel(texture, &buffer, pitch))
+      return FALSE;
+
+   buf = intel_drm_buffer(buffer);
+   if (!buf->flinked) {
+      if (drm_intel_bo_flink(buf->bo, &buf->flink))
+         return FALSE;
+      buf->flinked = TRUE;
+   }
+
+   *handle = buf->flink;
+
+   return TRUE;
+}
+
+static boolean
+intel_drm_local_handle_from_texture(struct drm_api *api,
+                                    struct pipe_screen *screen,
+                                    struct pipe_texture *texture,
+                                    unsigned *pitch,
+                                    unsigned *handle)
+{
+   struct intel_buffer *buffer = NULL;
+   if (!i915_get_texture_buffer_intel(texture, &buffer, pitch))
+      return FALSE;
+
+   *handle = intel_drm_buffer(buffer)->bo->handle;
+
+   return TRUE;
+}
+
+static void
+intel_drm_winsys_destroy(struct intel_winsys *iws)
+{
+   struct intel_drm_winsys *idws = intel_drm_winsys(iws);
+
+   drm_intel_bufmgr_destroy(idws->pools.gem);
+
+   FREE(idws);
+}
+
+static struct pipe_screen *
+intel_drm_create_screen(struct drm_api *api, int drmFD,
+                        struct drm_create_screen_arg *arg)
+{
+   struct intel_drm_winsys *idws;
+   unsigned int deviceID;
+
+   if (arg != NULL) {
+      switch(arg->mode) {
+      case DRM_CREATE_NORMAL:
+         break;
+      default:
+         return NULL;
+      }
+   }
+
+   idws = CALLOC_STRUCT(intel_drm_winsys);
+   if (!idws)
+      return NULL;
+
+   intel_drm_get_device_id(&deviceID);
+
+   intel_drm_winsys_init_batchbuffer_functions(idws);
+   intel_drm_winsys_init_buffer_functions(idws);
+   intel_drm_winsys_init_fence_functions(idws);
+
+   idws->fd = drmFD;
+   idws->id = deviceID;
+   idws->max_batch_size = 16 * 4096;
+
+   idws->base.destroy = intel_drm_winsys_destroy;
+
+   idws->pools.gem = drm_intel_bufmgr_gem_init(idws->fd, idws->max_batch_size);
+
+   idws->softpipe = FALSE;
+   idws->dump_cmd = debug_get_bool_option("INTEL_DUMP_CMD", FALSE);
+
+   return i915_create_screen(&idws->base, deviceID);
+}
+
+static struct pipe_context *
+intel_drm_create_context(struct drm_api *api, struct pipe_screen *screen)
+{
+   return i915_create_context(screen);
+}
+
+static void
+destroy(struct drm_api *api)
+{
+
+}
+
+struct drm_api intel_drm_api =
+{
+   .create_context = intel_drm_create_context,
+   .create_screen = intel_drm_create_screen,
+   .texture_from_shared_handle = intel_drm_texture_from_shared_handle,
+   .shared_handle_from_texture = intel_drm_shared_handle_from_texture,
+   .local_handle_from_texture = intel_drm_local_handle_from_texture,
+   .destroy = destroy,
+};
+
+struct drm_api *
+drm_api_create()
+{
+   return &intel_drm_api;
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c
new file mode 100644
index 0000000000..5ca3ad9762
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c
@@ -0,0 +1,214 @@
+
+#include "intel_drm_winsys.h"
+#include "util/u_memory.h"
+
+#include "i915_drm.h"
+
+#define BATCH_RESERVED 16
+
+#define INTEL_DEFAULT_RELOCS 100
+#define INTEL_MAX_RELOCS 400
+
+#define INTEL_BATCH_NO_CLIPRECTS 0x1
+#define INTEL_BATCH_CLIPRECTS    0x2
+
+#undef INTEL_RUN_SYNC
+
+struct intel_drm_batchbuffer
+{
+   struct intel_batchbuffer base;
+
+   size_t actual_size;
+
+   drm_intel_bo *bo;
+};
+
+static INLINE struct intel_drm_batchbuffer *
+intel_drm_batchbuffer(struct intel_batchbuffer *batch)
+{
+   return (struct intel_drm_batchbuffer *)batch;
+}
+
+static void
+intel_drm_batchbuffer_reset(struct intel_drm_batchbuffer *batch)
+{
+   struct intel_drm_winsys *idws = intel_drm_winsys(batch->base.iws);
+
+   if (batch->bo)
+      drm_intel_bo_unreference(batch->bo);
+   batch->bo = drm_intel_bo_alloc(idws->pools.gem,
+                                  "gallium3d_batchbuffer",
+                                  batch->actual_size,
+                                  4096);
+   drm_intel_bo_map(batch->bo, TRUE);
+   batch->base.map = batch->bo->virtual;
+
+   memset(batch->base.map, 0, batch->actual_size);
+   batch->base.ptr = batch->base.map;
+   batch->base.size = batch->actual_size - BATCH_RESERVED;
+   batch->base.relocs = 0;
+}
+
+static struct intel_batchbuffer *
+intel_drm_batchbuffer_create(struct intel_winsys *iws)
+{
+   struct intel_drm_winsys *idws = intel_drm_winsys(iws);
+   struct intel_drm_batchbuffer *batch = CALLOC_STRUCT(intel_drm_batchbuffer);
+
+   batch->base.map = NULL;
+   batch->base.ptr = NULL;
+   batch->base.size = 0;
+
+   batch->base.relocs = 0;
+   batch->base.max_relocs = 300;/*INTEL_DEFAULT_RELOCS;*/
+
+   batch->base.iws = iws;
+
+   batch->actual_size = idws->max_batch_size;
+
+   intel_drm_batchbuffer_reset(batch);
+
+   return &batch->base;
+}
+
+static int
+intel_drm_batchbuffer_reloc(struct intel_batchbuffer *ibatch,
+                            struct intel_buffer *buffer,
+                            enum intel_buffer_usage usage,
+                            unsigned pre_add)
+{
+   struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch);
+   unsigned write_domain = 0;
+   unsigned read_domain = 0;
+   unsigned offset;
+   int ret = 0;
+
+   assert(batch->base.relocs < batch->base.max_relocs);
+
+   if (usage == INTEL_USAGE_SAMPLER) {
+      write_domain = 0;
+      read_domain = I915_GEM_DOMAIN_SAMPLER;
+
+   } else if (usage == INTEL_USAGE_RENDER) {
+      write_domain = I915_GEM_DOMAIN_RENDER;
+      read_domain = I915_GEM_DOMAIN_RENDER;
+
+   } else if (usage == INTEL_USAGE_2D_TARGET) {
+      write_domain = I915_GEM_DOMAIN_RENDER;
+      read_domain = I915_GEM_DOMAIN_RENDER;
+
+   } else if (usage == INTEL_USAGE_2D_SOURCE) {
+      write_domain = 0;
+      read_domain = I915_GEM_DOMAIN_RENDER;
+
+   } else if (usage == INTEL_USAGE_VERTEX) {
+      write_domain = 0;
+      read_domain = I915_GEM_DOMAIN_VERTEX;
+
+   } else {
+      assert(0);
+      return -1;
+   }
+
+   offset = (unsigned)(batch->base.ptr - batch->base.map);
+
+   ret = drm_intel_bo_emit_reloc(batch->bo, offset,
+                                 intel_bo(buffer), pre_add,
+                                 read_domain,
+                                 write_domain);
+
+   ((uint32_t*)batch->base.ptr)[0] = intel_bo(buffer)->offset + pre_add;
+   batch->base.ptr += 4;
+
+   if (!ret)
+      batch->base.relocs++;
+
+   return ret;
+}
+
+static void
+intel_drm_batchbuffer_flush(struct intel_batchbuffer *ibatch,
+                            struct pipe_fence_handle **fence)
+{
+   struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch);
+   unsigned used = 0;
+   int ret = 0;
+   int i;
+
+   assert(intel_batchbuffer_space(ibatch) >= 0);
+
+   used = batch->base.ptr - batch->base.map;
+   assert((used & 3) == 0);
+
+   if (used & 4) {
+      // MI_FLUSH | FLUSH_MAP_CACHE;
+      intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x4<<23)|(1<<0));
+      // MI_NOOP
+      intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x0<<23));
+      // MI_BATCH_BUFFER_END;
+      intel_batchbuffer_dword(ibatch, (0x0<<29)|(0xA<<23));
+   } else {
+      //MI_FLUSH | FLUSH_MAP_CACHE;
+      intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x4<<23)|(1<<0));
+      // MI_BATCH_BUFFER_END;
+      intel_batchbuffer_dword(ibatch, (0x0<<29)|(0xA<<23));
+   }
+
+   used = batch->base.ptr - batch->base.map;
+
+   drm_intel_bo_unmap(batch->bo);
+
+   /* Do the sending to HW */
+   ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0);
+   assert(ret == 0);
+
+   if (intel_drm_winsys(ibatch->iws)->dump_cmd) {
+      unsigned *ptr;
+      drm_intel_bo_map(batch->bo, FALSE);
+      ptr = (unsigned*)batch->bo->virtual;
+
+      debug_printf("%s:\n", __func__);
+      for (i = 0; i < used / 4; i++, ptr++) {
+         debug_printf("\t%08x:    %08x\n", i*4, *ptr);
+      }
+
+      drm_intel_bo_unmap(batch->bo);
+   } else {
+#ifdef INTEL_RUN_SYNC
+      drm_intel_bo_map(batch->bo, FALSE);
+      drm_intel_bo_unmap(batch->bo);
+#endif
+   }
+
+   if (fence) {
+      ibatch->iws->fence_reference(ibatch->iws, fence, NULL);
+
+#ifdef INTEL_RUN_SYNC
+      /* we run synced to GPU so just pass null */
+      (*fence) = intel_drm_fence_create(NULL);
+#else
+      (*fence) = intel_drm_fence_create(batch->bo);
+#endif
+   }
+
+   intel_drm_batchbuffer_reset(batch);
+}
+
+static void
+intel_drm_batchbuffer_destroy(struct intel_batchbuffer *ibatch)
+{
+   struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch);
+
+   if (batch->bo)
+      drm_intel_bo_unreference(batch->bo);
+
+   free(batch);
+}
+
+void intel_drm_winsys_init_batchbuffer_functions(struct intel_drm_winsys *idws)
+{
+   idws->base.batchbuffer_create = intel_drm_batchbuffer_create;
+   idws->base.batchbuffer_reloc = intel_drm_batchbuffer_reloc;
+   idws->base.batchbuffer_flush = intel_drm_batchbuffer_flush;
+   idws->base.batchbuffer_destroy = intel_drm_batchbuffer_destroy;
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c
new file mode 100644
index 0000000000..e017cd2e98
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c
@@ -0,0 +1,134 @@
+
+#include "intel_drm_winsys.h"
+#include "util/u_memory.h"
+
+#include "i915_drm.h"
+
+static struct intel_buffer *
+intel_drm_buffer_create(struct intel_winsys *iws,
+                        unsigned size, unsigned alignment,
+                        enum intel_buffer_type type)
+{
+   struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer);
+   struct intel_drm_winsys *idws = intel_drm_winsys(iws);
+   drm_intel_bufmgr *pool;
+   char *name;
+
+   if (!buf)
+      return NULL;
+
+   buf->magic = 0xDEAD1337;
+   buf->flinked = FALSE;
+   buf->flink = 0;
+   buf->map_gtt = FALSE;
+
+   if (type == INTEL_NEW_TEXTURE) {
+      name = "gallium3d_texture";
+      pool = idws->pools.gem;
+   } else if (type == INTEL_NEW_VERTEX) {
+      name = "gallium3d_vertex";
+      pool = idws->pools.gem;
+   } else if (type == INTEL_NEW_SCANOUT) {
+      name = "gallium3d_scanout";
+      pool = idws->pools.gem;
+      buf->map_gtt = TRUE;
+   } else {
+      assert(0);
+      name = "gallium3d_unknown";
+      pool = idws->pools.gem;
+   }
+
+   buf->bo = drm_intel_bo_alloc(pool, name, size, alignment);
+
+   if (!buf->bo)
+      goto err;
+
+   return (struct intel_buffer *)buf;
+
+err:
+   assert(0);
+   FREE(buf);
+   return NULL;
+}
+
+static int
+intel_drm_buffer_set_fence_reg(struct intel_winsys *iws,
+                               struct intel_buffer *buffer,
+                               unsigned stride,
+                               enum intel_buffer_tile tile)
+{
+   assert(I915_TILING_NONE == INTEL_TILE_NONE);
+   assert(I915_TILING_X == INTEL_TILE_X);
+   assert(I915_TILING_Y == INTEL_TILE_Y);
+
+   return drm_intel_bo_set_tiling(intel_bo(buffer), &tile, stride);
+}
+
+static void *
+intel_drm_buffer_map(struct intel_winsys *iws,
+                     struct intel_buffer *buffer,
+                     boolean write)
+{
+   struct intel_drm_buffer *buf = intel_drm_buffer(buffer);
+   drm_intel_bo *bo = intel_bo(buffer);
+   int ret = 0;
+
+   assert(bo);
+
+   if (buf->map_count)
+      goto out;
+
+   if (buf->map_gtt)
+      ret = drm_intel_gem_bo_map_gtt(bo);
+   else
+      ret = drm_intel_bo_map(bo, write);
+
+   buf->ptr = bo->virtual;
+
+   assert(ret == 0);
+out:
+   if (ret)
+      return NULL;
+
+   buf->map_count++;
+   return buf->ptr;
+}
+
+static void
+intel_drm_buffer_unmap(struct intel_winsys *iws,
+                       struct intel_buffer *buffer)
+{
+   struct intel_drm_buffer *buf = intel_drm_buffer(buffer);
+
+   if (--buf->map_count)
+      return;
+
+   if (buf->map_gtt)
+      drm_intel_gem_bo_unmap_gtt(intel_bo(buffer));
+   else
+      drm_intel_bo_unmap(intel_bo(buffer));
+}
+
+static void
+intel_drm_buffer_destroy(struct intel_winsys *iws,
+                         struct intel_buffer *buffer)
+{
+   drm_intel_bo_unreference(intel_bo(buffer));
+
+#ifdef DEBUG
+   intel_drm_buffer(buffer)->magic = 0;
+   intel_drm_buffer(buffer)->bo = NULL;
+#endif
+
+   FREE(buffer);
+}
+
+void
+intel_drm_winsys_init_buffer_functions(struct intel_drm_winsys *idws)
+{
+   idws->base.buffer_create = intel_drm_buffer_create;
+   idws->base.buffer_set_fence_reg = intel_drm_buffer_set_fence_reg;
+   idws->base.buffer_map = intel_drm_buffer_map;
+   idws->base.buffer_unmap = intel_drm_buffer_unmap;
+   idws->base.buffer_destroy = intel_drm_buffer_destroy;
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c
new file mode 100644
index 0000000000..e70bfe7b44
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c
@@ -0,0 +1,81 @@
+
+#include "intel_drm_winsys.h"
+#include "util/u_memory.h"
+#include "pipe/p_refcnt.h"
+
+/**
+ * Because gem does not have fence's we have to create our own fences.
+ *
+ * They work by keeping the batchbuffer around and checking if that has
+ * been idled. If bo is NULL fence has expired.
+ */
+struct intel_drm_fence
+{
+   struct pipe_reference reference;
+   drm_intel_bo *bo;
+};
+
+
+struct pipe_fence_handle *
+intel_drm_fence_create(drm_intel_bo *bo)
+{
+   struct intel_drm_fence *fence = CALLOC_STRUCT(intel_drm_fence);
+
+   pipe_reference_init(&fence->reference, 1);
+   /* bo is null if fence already expired */
+   if (bo) {
+      drm_intel_bo_reference(bo);
+      fence->bo = bo;
+   }
+
+   return (struct pipe_fence_handle *)fence;
+}
+
+static void
+intel_drm_fence_reference(struct intel_winsys *iws,
+                          struct pipe_fence_handle **ptr,
+                          struct pipe_fence_handle *fence)
+{
+   struct intel_drm_fence *old = (struct intel_drm_fence *)*ptr;
+   struct intel_drm_fence *f = (struct intel_drm_fence *)fence;
+
+   if (pipe_reference((struct pipe_reference**)ptr, &f->reference)) {
+      if (old->bo)
+         drm_intel_bo_unreference(old->bo);
+      FREE(old);
+   }
+}
+
+static int
+intel_drm_fence_signalled(struct intel_winsys *iws,
+                          struct pipe_fence_handle *fence)
+{
+   assert(0);
+
+   return 0;
+}
+
+static int
+intel_drm_fence_finish(struct intel_winsys *iws,
+                       struct pipe_fence_handle *fence)
+{
+   struct intel_drm_fence *f = (struct intel_drm_fence *)fence;
+
+   /* fence already expired */
+   if (!f->bo)
+      return 0;
+
+   drm_intel_bo_wait_rendering(f->bo);
+   drm_intel_bo_unreference(f->bo);
+   f->bo = NULL;
+
+   return 0;
+}
+
+void
+intel_drm_winsys_init_fence_functions(struct intel_drm_winsys *idws)
+{
+   idws->base.fence_reference = intel_drm_fence_reference;
+   idws->base.fence_signalled = intel_drm_fence_signalled;
+   idws->base.fence_finish = intel_drm_fence_finish;
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h b/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h
new file mode 100644
index 0000000000..415c45feea
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h
@@ -0,0 +1,78 @@
+
+#ifndef INTEL_DRM_WINSYS_H
+#define INTEL_DRM_WINSYS_H
+
+#include "i915simple/intel_batchbuffer.h"
+
+#include "drm.h"
+#include "intel_bufmgr.h"
+
+
+/*
+ * Winsys
+ */
+
+
+struct intel_drm_winsys
+{
+   struct intel_winsys base;
+
+   boolean softpipe;
+   boolean dump_cmd;
+
+   int fd; /**< Drm file discriptor */
+
+   unsigned id;
+
+   size_t max_batch_size;
+
+   struct {
+      drm_intel_bufmgr *gem;
+   } pools;
+};
+
+static INLINE struct intel_drm_winsys *
+intel_drm_winsys(struct intel_winsys *iws)
+{
+   return (struct intel_drm_winsys *)iws;
+}
+
+struct intel_drm_winsys * intel_drm_winsys_create(int fd, unsigned pci_id);
+struct pipe_fence_handle * intel_drm_fence_create(drm_intel_bo *bo);
+
+void intel_drm_winsys_init_batchbuffer_functions(struct intel_drm_winsys *idws);
+void intel_drm_winsys_init_buffer_functions(struct intel_drm_winsys *idws);
+void intel_drm_winsys_init_fence_functions(struct intel_drm_winsys *idws);
+
+
+/*
+ * Buffer
+ */
+
+
+struct intel_drm_buffer {
+   unsigned magic;
+
+   drm_intel_bo *bo;
+
+   void *ptr;
+   unsigned map_count;
+   boolean map_gtt;
+
+   boolean flinked;
+   unsigned flink;
+};
+
+static INLINE struct intel_drm_buffer *
+intel_drm_buffer(struct intel_buffer *buffer)
+{
+   return (struct intel_drm_buffer *)buffer;
+}
+
+static INLINE drm_intel_bo *
+intel_bo(struct intel_buffer *buffer)
+{
+   return intel_drm_buffer(buffer)->bo;
+}
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile
index b1b6b9362b..9e56853b02 100644
--- a/src/gallium/winsys/drm/intel/xorg/Makefile
+++ b/src/gallium/winsys/drm/intel/xorg/Makefile
@@ -1,36 +1,39 @@
 TARGET     = modesetting_drv.so
 CFILES     = $(wildcard ./*.c)
 OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../../../..
 TOP        = ../../../../../..
 
-include ${TOP}/configs/current
+include $(TOP)/configs/current
 
-CFLAGS = -DHAVE_CONFIG_H \
-         -g -Wall -Wimplicit-function-declaration -fPIC \
-         $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \
-	 -I../gem \
-         -I${GALLIUMDIR}/include \
-         -I${GALLIUMDIR}/drivers \
-         -I${GALLIUMDIR}/auxiliary \
-         -I${TOP}/src/mesa \
-         -I$(TOP)/include \
-         -I$(TOP)/src/egl/main
+INCLUDES = \
+	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
+	-I../gem \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include \
+	-I$(TOP)/src/egl/main
 
 LIBS = \
-	$(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \
-	$(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \
+	$(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a \
+	$(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
 	$(GALLIUM_AUXILIARIES)
 
+DRIVER_DEFINES = \
+	-DHAVE_CONFIG_H
+
+
 #############################################
 
 
 
 all default: $(TARGET)
 
-$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a $(LIBS)
+$(TARGET): $(OBJECTS) Makefile $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a $(LIBS)
 	$(TOP)/bin/mklib -noprefix -o $@ \
 	$(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel
 
@@ -39,6 +42,16 @@ clean:
 
 install:
 	$(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+
+
+##############################################
+
+
+.c.o:
+	$(CC) -c $(CFLAGS) $(INCLUDES) $(DRIVER_DEFINES) $< -o $@
+
+
+##############################################
 
 .PHONY	= all clean install
diff --git a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
index 46a7971f60..369dc356cf 100644
--- a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
+++ b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
@@ -47,8 +47,8 @@ static SymTabRec intel_xorg_chipsets[] = {
 };
 
 static PciChipsets intel_xorg_pci_devices[] = {
-    {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA},
-    {-1, -1, RES_UNDEFINED}
+    {PCI_MATCH_ANY, PCI_MATCH_ANY, NULL},
+    {-1, -1, NULL}
 };
 
 static XF86ModuleVersionInfo intel_xorg_version = {
@@ -105,12 +105,6 @@ intel_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin)
 	xf86AddDriver(&modesetting, module, HaveDriverFuncs);
 
 	/*
-	 * Tell the loader about symbols from other modules that this module
-	 * might refer to.
-	 */
-	xorg_tracker_loader_ref_sym_lists();
-
-	/*
 	 * The return value must be non-NULL on success even though there
 	 * is no TearDownProc.
 	 */
@@ -140,7 +134,7 @@ intel_xorg_pci_probe(DriverPtr driver,
 			       NULL, NULL, NULL, NULL, NULL);
     if (scrn != NULL) {
 	scrn->driverVersion = 1;
-	scrn->driverName = "modesetting";
+	scrn->driverName = "i915";
 	scrn->name = "modesetting";
 	scrn->Probe = NULL;
 
diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile
index f8c8135854..6c9cbef26d 100644
--- a/src/gallium/winsys/drm/nouveau/Makefile
+++ b/src/gallium/winsys/drm/nouveau/Makefile
@@ -2,7 +2,7 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
-SUBDIRS = drm dri dri2
+SUBDIRS = drm $(GALLIUM_STATE_TRACKERS_DIRS)
 
 default install clean:
 	@for dir in $(SUBDIRS) ; do \
diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile
index f7db6201fe..0937f68c34 100644
--- a/src/gallium/winsys/drm/nouveau/dri/Makefile
+++ b/src/gallium/winsys/drm/nouveau/dri/Makefile
@@ -3,32 +3,25 @@ include $(TOP)/configs/current
 
 LIBNAME = nouveau_dri.so
 
-MINIGLX_SOURCES =
-
 PIPE_DRIVERS = \
+	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
 	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
 	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
 	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
 	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
 	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
-	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-	
-DRIVER_SOURCES = \
-	nouveau_context.c \
-	nouveau_screen.c \
-	nouveau_swapbuffers.c \
-	nouveau_lock.c
+	$(TOP)/src/gallium/drivers/nv50/libnv50.a \
+	$(TOP)/src/gallium/drivers/nouveau/libnouveau.a
+
+DRIVER_SOURCES =
 
 C_SOURCES = \
 	$(COMMON_GALLIUM_SOURCES) \
 	$(DRIVER_SOURCES)
 
-ASM_SOURCES = 
+include ../../Makefile.template
 
-DRIVER_DEFINES = $(shell pkg-config libdrm_nouveau --cflags)
 DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
 
-include ../../Makefile.template
-
 symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
deleted file mode 100644
index deb6ffcff1..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
+++ /dev/null
@@ -1,118 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-#include <utils.h>
-
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/drm_api.h>
-#include <pipe/p_defines.h>
-#include <pipe/p_context.h>
-#include <pipe/p_screen.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-
-#include "nouveau_drmif.h"
-
-GLboolean
-nouveau_context_create(const __GLcontextModes *glVis,
-		       __DRIcontextPrivate *driContextPriv,
-		       void *sharedContextPrivate)
-{
-	__DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
-	struct nouveau_screen  *nv_screen = driScrnPriv->private;
-	struct nouveau_context *nv;
-	struct pipe_context *pipe;
-	struct st_context *st_share = NULL;
-
-	if (sharedContextPrivate)
-		st_share = ((struct nouveau_context *)sharedContextPrivate)->st;
-
-	nv = CALLOC_STRUCT(nouveau_context);
-	if (!nv)
-		return GL_FALSE;
-
-	{
-		struct nouveau_device_priv *nvdev =
-			nouveau_device(nv_screen->device);
-
-		nvdev->ctx  = driContextPriv->hHWContext;
-		nvdev->lock = (drmLock *)&driScrnPriv->pSAREA->lock;
-	}
-
-	pipe = drm_api_hooks.create_context(nv_screen->pscreen);
-	if (!pipe) {
-		FREE(nv);
-		return GL_FALSE;
-	}
-	pipe->priv = nv;
-
-	driContextPriv->driverPrivate = nv;
-	nv->dri_screen = driScrnPriv;
-
-	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
-			    nv->dri_screen->myNum, "nouveau");
-
-	nv->st = st_create_context(pipe, glVis, st_share);
-	return GL_TRUE;
-}
-
-void
-nouveau_context_destroy(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-
-	assert(nv);
-
-	st_finish(nv->st);
-	st_destroy_context(nv->st);
-
-	FREE(nv);
-}
-
-GLboolean
-nouveau_context_bind(__DRIcontextPrivate *driContextPriv,
-		     __DRIdrawablePrivate *driDrawPriv,
-		     __DRIdrawablePrivate *driReadPriv)
-{
-	struct nouveau_context *nv;
-	struct nouveau_framebuffer *draw, *read;
-
-	if (!driContextPriv) {
-		st_make_current(NULL, NULL, NULL);
-		return GL_TRUE;
-	}
-
-	nv = driContextPriv->driverPrivate;
-	draw = driDrawPriv->driverPrivate;
-	read = driReadPriv->driverPrivate;
-
-	st_make_current(nv->st, draw->stfb, read->stfb);
-
-	if ((nv->dri_drawable != driDrawPriv) ||
-	    (nv->last_stamp != driDrawPriv->lastStamp)) {
-		nv->dri_drawable = driDrawPriv;
-		st_resize_framebuffer(draw->stfb, driDrawPriv->w,
-				      driDrawPriv->h);
-		nv->last_stamp = driDrawPriv->lastStamp;
-	}
-
-	if (driDrawPriv != driReadPriv) {
-		st_resize_framebuffer(read->stfb, driReadPriv->w,
-				      driReadPriv->h);
-	}
-
-	return GL_TRUE;
-}
-
-GLboolean
-nouveau_context_unbind(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-	(void)nv;
-
-	st_flush(nv->st, 0, NULL);
-	return GL_TRUE;
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
deleted file mode 100644
index 2779b092e6..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
+++ /dev/null
@@ -1,53 +0,0 @@
-#ifndef __NOUVEAU_CONTEXT_DRI_H__
-#define __NOUVEAU_CONTEXT_DRI_H__
-
-#include <dri_util.h>
-#include <xmlconfig.h>
-
-#include "nouveau/nouveau_winsys.h"
-
-#define NOUVEAU_ERR(fmt, args...) debug_printf("%s: "fmt, __func__, ##args)
-
-struct nouveau_framebuffer {
-	struct st_framebuffer *stfb;
-};
-
-struct nouveau_context {
-	struct st_context *st;
-
-	/* DRI stuff */
-	__DRIscreenPrivate    *dri_screen;
-	__DRIdrawablePrivate  *dri_drawable;
-	unsigned int           last_stamp;
-	driOptionCache         dri_option_cache;
-	drm_context_t          drm_context;
-	drmLock                drm_lock;
-	int                    locked;
-};
-
-extern GLboolean nouveau_context_create(const __GLcontextModes *,
-					__DRIcontextPrivate *, void *);
-extern void nouveau_context_destroy(__DRIcontextPrivate *);
-extern GLboolean nouveau_context_bind(__DRIcontextPrivate *,
-				      __DRIdrawablePrivate *draw,
-				      __DRIdrawablePrivate *read);
-extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *);
-
-extern void nouveau_contended_lock(struct nouveau_context *nv);
-extern void LOCK_HARDWARE(struct nouveau_context *nv);
-extern void UNLOCK_HARDWARE(struct nouveau_context *nv);
-
-#ifdef DEBUG
-extern int __nouveau_debug;
-
-#define DEBUG_BO (1 << 0)
-
-#define DBG(flag, ...) do {                   \
-	if (__nouveau_debug & (DEBUG_##flag)) \
-		NOUVEAU_ERR(__VA_ARGS__);     \
-} while(0)
-#else
-#define DBG(flag, ...)
-#endif
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
deleted file mode 100644
index 4e9b76a909..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
+++ /dev/null
@@ -1,330 +0,0 @@
-#include <utils.h>
-#include <vblank.h>
-#include <xmlpool.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_cb_fbo.h>
-#include <state_tracker/drm_api.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-#include "nouveau_dri.h"
-
-#include "nouveau_drm.h"
-#include "nouveau_drmif.h"
-
-#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12
-#error nouveau_drm.h version does not match expected version
-#endif
-
-/* Extension stuff, enabling of extensions handled by Gallium's GL state
- * tracker.  But, we still need to define the entry points we want.
- */
-#define need_GL_ARB_fragment_program
-#define need_GL_ARB_multisample
-#define need_GL_ARB_occlusion_query
-#define need_GL_ARB_point_parameters
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_program
-#define need_GL_ARB_vertex_shader
-#define need_GL_ARB_vertex_buffer_object
-#define need_GL_EXT_compiled_vertex_array
-#define need_GL_EXT_fog_coord
-#define need_GL_EXT_secondary_color
-#define need_GL_EXT_framebuffer_object
-#define need_GL_VERSION_2_0
-#define need_GL_VERSION_2_1
-#include "extension_helper.h"
-
-const struct dri_extension card_extensions[] =
-{
-	{ "GL_ARB_multisample", GL_ARB_multisample_functions },
-	{ "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions },
-	{ "GL_ARB_point_parameters", GL_ARB_point_parameters_functions },
-	{ "GL_ARB_shader_objects", GL_ARB_shader_objects_functions },
-	{ "GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
-	{ "GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
-	{ "GL_ARB_texture_compression", GL_ARB_texture_compression_functions },
-	{ "GL_ARB_vertex_program", GL_ARB_vertex_program_functions },
-	{ "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions },
-	{ "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions },
-	{ "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions },
-	{ "GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
-	{ "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
-	{ "GL_EXT_secondary_color", GL_EXT_secondary_color_functions },
-	{ NULL, 0 }
-};
-
-PUBLIC const char __driConfigOptions[] =
-DRI_CONF_BEGIN
-DRI_CONF_END;
-static const GLuint __driNConfigOptions = 0;
-
-extern const struct dri_extension common_extensions[];
-extern const struct dri_extension nv40_extensions[];
-
-static GLboolean
-nouveau_create_buffer(__DRIscreenPrivate * driScrnPriv,
-		      __DRIdrawablePrivate * driDrawPriv,
-		      const __GLcontextModes *glVis, GLboolean pixmapBuffer)
-{
-	struct nouveau_framebuffer *nvfb;
-	enum pipe_format colour, depth, stencil;
-
-	if (pixmapBuffer)
-		return GL_FALSE;
-
-	nvfb = CALLOC_STRUCT(nouveau_framebuffer);
-	if (!nvfb)
-		return GL_FALSE;
-
-	if (glVis->redBits == 5)
-		colour = PIPE_FORMAT_R5G6B5_UNORM;
-	else
-		colour = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-	if (glVis->depthBits == 16)
-		depth = PIPE_FORMAT_Z16_UNORM;
-	else if (glVis->depthBits == 24)
-		depth = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		depth = PIPE_FORMAT_NONE;
-
-	if (glVis->stencilBits == 8)
-		stencil = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		stencil = PIPE_FORMAT_NONE;
-
-	nvfb->stfb = st_create_framebuffer(glVis, colour, depth, stencil,
-					   driDrawPriv->w, driDrawPriv->h,
-					   (void*)nvfb);
-	if (!nvfb->stfb) {
-		free(nvfb);
-		return  GL_FALSE;
-	}
-
-	driDrawPriv->driverPrivate = (void *)nvfb;
-	return GL_TRUE;
-}
-
-static void
-nouveau_destroy_buffer(__DRIdrawablePrivate * driDrawPriv)
-{
-	struct nouveau_framebuffer *nvfb;
-	
-	nvfb = (struct nouveau_framebuffer *)driDrawPriv->driverPrivate;
-	st_unreference_framebuffer(nvfb->stfb);
-	free(nvfb);
-}
-
-static __DRIconfig **
-nouveau_fill_in_modes(__DRIscreenPrivate *psp,
-		      unsigned pixel_bits, unsigned depth_bits,
-		      unsigned stencil_bits, GLboolean have_back_buffer)
-{
-	__DRIconfig **configs;
-	unsigned depth_buffer_factor;
-	unsigned back_buffer_factor;
-	GLenum fb_format;
-	GLenum fb_type;
-
-	static const GLenum back_buffer_modes[] = {
-		GLX_NONE, GLX_SWAP_UNDEFINED_OML,
-	};
-
-	uint8_t depth_bits_array[3];
-	uint8_t stencil_bits_array[3];
-	uint8_t msaa_samples_array[1];
-
-	depth_bits_array[0] = 0;
-	depth_bits_array[1] = depth_bits;
-	depth_bits_array[2] = depth_bits;
-
-	/* Just like with the accumulation buffer, always provide some modes
-	 * with a stencil buffer.  It will be a sw fallback, but some apps won't
-	 * care about that.
-	 */
-	stencil_bits_array[0] = 0;
-	stencil_bits_array[1] = 0;
-	if (depth_bits == 24)
-		stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
-	stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
-
-	msaa_samples_array[0] = 0;
-
-	depth_buffer_factor =
-		((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
-	back_buffer_factor = (have_back_buffer) ? 3 : 1;
-
-	if (pixel_bits == 16) {
-		fb_format = GL_RGB;
-		fb_type = GL_UNSIGNED_SHORT_5_6_5;
-	}
-	else {
-		fb_format = GL_BGRA;
-		fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
-	}
-
-	configs = driCreateConfigs(fb_format, fb_type,
-				   depth_bits_array, stencil_bits_array,
-				   depth_buffer_factor, back_buffer_modes,
-				   back_buffer_factor, msaa_samples_array, 1);
-	if (configs == NULL) {
-	 fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
-			 __func__, __LINE__);
-		return NULL;
-	}
-
-	return configs;
-}
-
-static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
-                        unsigned handle,
-                        enum pipe_format format,
-                        unsigned width,
-                        unsigned height,
-                        unsigned pitch)
-{
-   struct pipe_surface *surface = NULL;
-   struct pipe_texture *texture = NULL;
-   struct pipe_texture templat;
-   struct pipe_buffer *buf = NULL;
-
-   buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle);
-   if (!buf)
-      return NULL;
-
-   memset(&templat, 0, sizeof(templat));
-   templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
-                       NOUVEAU_TEXTURE_USAGE_LINEAR;
-   templat.target = PIPE_TEXTURE_2D;
-   templat.last_level = 0;
-   templat.depth[0] = 1;
-   templat.format = format;
-   templat.width[0] = width;
-   templat.height[0] = height;
-   pf_get_block(templat.format, &templat.block);
-
-   texture = screen->texture_blanket(screen,
-                                     &templat,
-                                     &pitch,
-                                     buf);
-
-   /* we don't need the buffer from this point on */
-   pipe_buffer_reference(&buf, NULL);
-
-   if (!texture)
-      return NULL;
-
-   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
-                                     PIPE_BUFFER_USAGE_GPU_READ |
-                                     PIPE_BUFFER_USAGE_GPU_WRITE);
-
-   /* we don't need the texture from this point on */
-   pipe_texture_reference(&texture, NULL);
-   return surface;
-}
-
-static const __DRIconfig **
-nouveau_screen_create(__DRIscreenPrivate *psp)
-{
-	struct nouveau_dri *nv_dri = psp->pDevPriv;
-	struct nouveau_screen *nv_screen;
-	static const __DRIversion ddx_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-	static const __DRIversion dri_expected = { 4, 0, 0 };
-	static const __DRIversion drm_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-
-	if (!driCheckDriDdxDrmVersions2("nouveau",
-					&psp->dri_version, &dri_expected,
-					&psp->ddx_version, &ddx_expected,
-					&psp->drm_version, &drm_expected)) {
-		return NULL;
-	}
-
-	if (drm_expected.patch != psp->drm_version.patch) {
-		fprintf(stderr, "Incompatible DRM patch level.\n"
-				"Expected: %d\n" "Current : %d\n",
-			drm_expected.patch, psp->drm_version.patch);
-		return NULL;
-	}
-
-	driInitExtensions(NULL, card_extensions, GL_FALSE);
-
-	if (psp->devPrivSize != sizeof(struct nouveau_dri)) {
-		NOUVEAU_ERR("DRI struct mismatch between DDX/DRI\n");
-		return NULL;
-	}
-
-	nv_screen = CALLOC_STRUCT(nouveau_screen);
-	if (!nv_screen)
-		return NULL;
-
-	nouveau_device_open_existing(&nv_screen->device, 0, psp->fd, 0);
-
-	nv_screen->pscreen = drm_api_hooks.create_screen(psp->fd, NULL);
-	if (!nv_screen->pscreen) {
-		FREE(nv_screen);
-		return NULL;
-	}
-	nv_screen->pscreen->flush_frontbuffer = nouveau_flush_frontbuffer;
-
-	{
-		enum pipe_format format;
-
-		if (nv_dri->bpp == 16)
-			format = PIPE_FORMAT_R5G6B5_UNORM;
-		else
-			format = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-		nv_screen->fb = dri_surface_from_handle(nv_screen->pscreen,
-							nv_dri->front_offset,
-							format,
-							nv_dri->width,
-							nv_dri->height,
-							nv_dri->front_pitch *
-							nv_dri->bpp / 8);
-	}
-						
-	driParseOptionInfo(&nv_screen->option_cache,
-			   __driConfigOptions, __driNConfigOptions);
-
-	nv_screen->driScrnPriv = psp;
-	psp->private = (void *)nv_screen;
-
-	return (const __DRIconfig **)
-		nouveau_fill_in_modes(psp, nv_dri->bpp,
-				      (nv_dri->bpp == 16) ? 16 : 24,
-				      (nv_dri->bpp == 16) ? 0 : 8, 1);
-}
-
-static void
-nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv)
-{
-	struct nouveau_screen *nv_screen = driScrnPriv->private;
-
-	driScrnPriv->private = NULL;
-	FREE(nv_screen);
-}
-
-const struct __DriverAPIRec
-driDriverAPI = {
-	.InitScreen	= nouveau_screen_create,
-	.DestroyScreen	= nouveau_screen_destroy,
-	.CreateContext	= nouveau_context_create,
-	.DestroyContext	= nouveau_context_destroy,
-	.CreateBuffer	= nouveau_create_buffer,
-	.DestroyBuffer	= nouveau_destroy_buffer,
-	.SwapBuffers	= nouveau_swap_buffers,
-	.MakeCurrent	= nouveau_context_bind,
-	.UnbindContext	= nouveau_context_unbind,
-	.CopySubBuffer	= nouveau_copy_sub_buffer,
-
-	.InitScreen2	= NULL, /* one day, I promise! */
-};
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
deleted file mode 100644
index ac078f3c63..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
+++ /dev/null
@@ -1,16 +0,0 @@
-#ifndef __NOUVEAU_SCREEN_DRI_H__
-#define __NOUVEAU_SCREEN_DRI_H__
-
-#include "xmlconfig.h"
-
-struct nouveau_screen {
-	__DRIscreenPrivate *driScrnPriv;
-	driOptionCache      option_cache;
-
-	struct nouveau_device *device;
-
-	struct pipe_screen *pscreen;
-	struct pipe_surface *fb;
-};
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
deleted file mode 100644
index 9c841a0b2d..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
+++ /dev/null
@@ -1,115 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/st_cb_fbo.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-
-#include "nouveau_pushbuf.h"
-
-void
-nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf,
-		    const drm_clip_rect_t *rect)
-{
-	struct nouveau_context *nv = dPriv->driContextPriv->driverPrivate;
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct pipe_context *pipe = nv->st->pipe;
-	drm_clip_rect_t *pbox;
-	int nbox, i;
-
-	LOCK_HARDWARE(nv);
-	if (!dPriv->numClipRects) {
-		UNLOCK_HARDWARE(nv);
-		return;
-	}
-	pbox = dPriv->pClipRects;
-	nbox = dPriv->numClipRects;
-
-	for (i = 0; i < nbox; i++, pbox++) {
-		int sx, sy, dx, dy, w, h;
-
-		sx = pbox->x1 - dPriv->x;
-		sy = pbox->y1 - dPriv->y;
-		dx = pbox->x1;
-		dy = pbox->y1;
-		w  = pbox->x2 - pbox->x1;
-		h  = pbox->y2 - pbox->y1;
-
-		pipe->surface_copy(pipe, nv_screen->fb, dx, dy, surf,
-				   sx, sy, w, h);
-	}
-
-	pipe->flush(pipe, 0, NULL);
-	UNLOCK_HARDWARE(nv);
-
-	if (nv->last_stamp != dPriv->lastStamp) {
-		struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-		st_resize_framebuffer(nvfb->stfb, dPriv->w, dPriv->h);
-		nv->last_stamp = dPriv->lastStamp;
-	}
-}
-
-void
-nouveau_copy_sub_buffer(__DRIdrawablePrivate *dPriv, int x, int y, int w, int h)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		drm_clip_rect_t rect;
-		rect.x1 = x;
-		rect.y1 = y;
-		rect.x2 = x + w;
-		rect.y2 = y + h;
-
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, &rect);
-	}
-}
-
-void
-nouveau_swap_buffers(__DRIdrawablePrivate *dPriv)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, NULL);
-	}
-}
-
-void
-nouveau_flush_frontbuffer(struct pipe_screen *pscreen, struct pipe_surface *ps,
-			  void *context_private)
-{
-	struct nouveau_context *nv = context_private;
-	__DRIdrawablePrivate *dPriv = nv->dri_drawable;
-
-	nouveau_copy_buffer(dPriv, ps, NULL);
-}
-
-void
-nouveau_contended_lock(struct nouveau_context *nv)
-{
-	struct nouveau_context *nv_sub = (struct nouveau_context*)nv;
-	__DRIdrawablePrivate *dPriv = nv_sub->dri_drawable;
-	__DRIscreenPrivate *sPriv = nv_sub->dri_screen;
-
-	/* If the window moved, may need to set a new cliprect now.
-	 *
-	 * NOTE: This releases and regains the hw lock, so all state
-	 * checking must be done *after* this call:
-	 */
-	if (dPriv)
-		DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
deleted file mode 100644
index 4ca9cc2283..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
+++ /dev/null
@@ -1,11 +0,0 @@
-#ifndef __NOUVEAU_SWAPBUFFERS_H__
-#define __NOUVEAU_SWAPBUFFERS_H__
-
-void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *,
-			 const drm_clip_rect_t *);
-void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, int x, int y, int w, int h);
-void nouveau_swap_buffers(__DRIdrawablePrivate *);
-void nouveau_flush_frontbuffer(struct pipe_screen *, struct pipe_surface *,
-			       void *context_private);
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri2/Makefile b/src/gallium/winsys/drm/nouveau/dri2/Makefile
deleted file mode 100644
index 377a80d518..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri2/Makefile
+++ /dev/null
@@ -1,26 +0,0 @@
-TOP = ../../../../../..
-include $(TOP)/configs/current
-
-LIBNAME = nouveau_dri2.so
-
-PIPE_DRIVERS = \
-	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
-	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
-	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
-	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
-	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-
-DRIVER_SOURCES =
-
-C_SOURCES = \
-	$(COMMON_GALLIUM_SOURCES) \
-	$(DRIVER_SOURCES)
-
-include ../../Makefile.template
-
-DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
-
-symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/drm/Makefile b/src/gallium/winsys/drm/nouveau/drm/Makefile
index 2da78d8690..54c3b26c75 100644
--- a/src/gallium/winsys/drm/nouveau/drm/Makefile
+++ b/src/gallium/winsys/drm/nouveau/drm/Makefile
@@ -3,9 +3,7 @@ include $(TOP)/configs/current
 
 LIBNAME = nouveaudrm
 
-C_SOURCES = nouveau_drm_api.c \
-	    nouveau_winsys_pipe.c \
-	    nouveau_winsys.c
+C_SOURCES = nouveau_drm_api.c
 
 LIBRARY_INCLUDES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-I)
 LIBRARY_DEFINES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-other)
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
index 1207c2d609..1207c2d609 100644
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index a558fda140..091cbbcfed 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -1,20 +1,70 @@
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
 #include "util/u_memory.h"
 
 #include "nouveau_drm_api.h"
-#include "nouveau_winsys_pipe.h"
 
 #include "nouveau_drmif.h"
 #include "nouveau_channel.h"
 #include "nouveau_bo.h"
 
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_screen.h"
+
+static struct pipe_surface *
+dri_surface_from_handle(struct drm_api *api, struct pipe_screen *pscreen,
+                        unsigned handle, enum pipe_format format,
+                        unsigned width, unsigned height, unsigned pitch)
+{
+	struct pipe_surface *ps = NULL;
+	struct pipe_texture *pt = NULL;
+	struct pipe_texture tmpl;
+
+	memset(&tmpl, 0, sizeof(tmpl));
+	tmpl.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
+			 NOUVEAU_TEXTURE_USAGE_LINEAR;
+	tmpl.target = PIPE_TEXTURE_2D;
+	tmpl.last_level = 0;
+	tmpl.depth[0] = 1;
+	tmpl.format = format;
+	tmpl.width[0] = width;
+	tmpl.height[0] = height;
+	pf_get_block(tmpl.format, &tmpl.block);
+
+	pt = api->texture_from_shared_handle(api, pscreen, &tmpl,
+					     "front buffer", pitch, handle);
+	if (!pt)
+		return NULL;
+
+	ps = pscreen->get_tex_surface(pscreen, pt, 0, 0, 0,
+				      PIPE_BUFFER_USAGE_GPU_READ |
+				      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+	/* we don't need the texture from this point on */
+	pipe_texture_reference(&pt, NULL);
+	return ps;
+}
+
+static struct pipe_surface *
+nouveau_dri1_front_surface(struct pipe_context *pipe)
+{
+	return nouveau_winsys_screen(pipe->screen)->front;
+}
+
+static struct dri1_api nouveau_dri1_api = {
+	nouveau_dri1_front_surface,
+};
+
 static struct pipe_screen *
-nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg)
+nouveau_drm_create_screen(struct drm_api *api, int fd,
+			  struct drm_create_screen_arg *arg)
 {
-	struct pipe_winsys *ws;
+	struct dri1_create_screen_arg *dri1 = (void *)arg;
 	struct nouveau_winsys *nvws;
+	struct pipe_winsys *ws;
 	struct nouveau_device *dev = NULL;
 	struct pipe_screen *(*init)(struct pipe_winsys *,
-				    struct nouveau_winsys *);
+				    struct nouveau_device *);
 	int ret;
 
 	ret = nouveau_device_open_existing(&dev, 0, fd, 0);
@@ -49,33 +99,53 @@ nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg)
 		return NULL;
 	}
 
-	ws = nouveau_pipe_winsys_new(dev);
-	if (!ws) {
+	nvws = CALLOC_STRUCT(nouveau_winsys);
+	if (!nvws) {
 		nouveau_device_close(&dev);
 		return NULL;
 	}
+	ws = &nvws->base;
 
-	nvws = nouveau_winsys_new(ws);
-	if (!nvws) {
+	nvws->pscreen = init(ws, dev);
+	if (!nvws->pscreen) {
 		ws->destroy(ws);
 		return NULL;
 	}
 
-	nouveau_pipe_winsys(ws)->pscreen = init(ws, nvws);
-	if (!nouveau_pipe_winsys(ws)->pscreen) {
-		ws->destroy(ws);
-		return NULL;
+	if (arg->mode == DRM_CREATE_DRI1) {
+		struct nouveau_dri *nvdri = dri1->ddx_info;
+		enum pipe_format format;
+
+		if (nvdri->bpp == 16)
+			format = PIPE_FORMAT_R5G6B5_UNORM;
+		else
+			format = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+		nvws->front = dri_surface_from_handle(api, nvws->pscreen,
+						      nvdri->front_offset,
+						      format, nvdri->width,
+						      nvdri->height,
+						      nvdri->front_pitch *
+						      (nvdri->bpp / 8));
+		if (!nvws->front) {
+			debug_printf("%s: error referencing front buffer\n",
+				     __func__);
+			ws->destroy(ws);
+			return NULL;
+		}
+
+		dri1->api = &nouveau_dri1_api;
 	}
 
-	return nouveau_pipe_winsys(ws)->pscreen;
+	return nvws->pscreen;
 }
 
 static struct pipe_context *
-nouveau_drm_create_context(struct pipe_screen *pscreen)
+nouveau_drm_create_context(struct drm_api *api, struct pipe_screen *pscreen)
 {
-	struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen);
+	struct nouveau_winsys *nvws = nouveau_winsys_screen(pscreen);
 	struct pipe_context *(*init)(struct pipe_screen *, unsigned);
-	unsigned chipset = nvpws->channel->device->chipset;
+	unsigned chipset = nouveau_screen(pscreen)->device->chipset;
 	int i;
 
 	switch (chipset & 0xf0) {
@@ -106,89 +176,88 @@ nouveau_drm_create_context(struct pipe_screen *pscreen)
 	}
 
 	/* Find a free slot for a pipe context, allocate a new one if needed */
-	for (i = 0; i < nvpws->nr_pctx; i++) {
-		if (nvpws->pctx[i] == NULL)
+	for (i = 0; i < nvws->nr_pctx; i++) {
+		if (nvws->pctx[i] == NULL)
 			break;
 	}
 
-	if (i == nvpws->nr_pctx) {
-		nvpws->nr_pctx++;
-		nvpws->pctx = realloc(nvpws->pctx,
-				      sizeof(*nvpws->pctx) * nvpws->nr_pctx);
+	if (i == nvws->nr_pctx) {
+		nvws->nr_pctx++;
+		nvws->pctx = realloc(nvws->pctx,
+				      sizeof(*nvws->pctx) * nvws->nr_pctx);
 	}
 
-	nvpws->pctx[i] = init(pscreen, i);
-	return nvpws->pctx[i];
-}
-
-static boolean
-nouveau_drm_pb_from_pt(struct pipe_texture *pt, struct pipe_buffer **ppb,
-		       unsigned *stride)
-{
-	return false;
+	nvws->pctx[i] = init(pscreen, i);
+	return nvws->pctx[i];
 }
 
-static struct pipe_buffer *
-nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name,
-			   unsigned handle)
+static struct pipe_texture *
+nouveau_drm_pt_from_name(struct drm_api *api, struct pipe_screen *pscreen,
+			 struct pipe_texture *templ, const char *name,
+			 unsigned stride, unsigned handle)
 {
-	struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvpb;
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct pipe_buffer *pb;
 	int ret;
 
-	nvpb = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvpb)
+	pb = CALLOC(1, sizeof(struct pipe_buffer) + sizeof(struct nouveau_bo*));
+	if (!pb)
 		return NULL;
 
-	ret = nouveau_bo_handle_ref(dev, handle, &nvpb->bo);
+	ret = nouveau_bo_handle_ref(dev, handle, (struct nouveau_bo**)(pb+1));
 	if (ret) {
 		debug_printf("%s: ref name 0x%08x failed with %d\n",
 			     __func__, handle, ret);
-		FREE(nvpb);
+		FREE(pb);
 		return NULL;
 	}
 
-	pipe_reference_init(&nvpb->base.reference, 1);
-	nvpb->base.screen = pscreen;
-	nvpb->base.alignment = 0;
-	nvpb->base.usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
-			   PIPE_BUFFER_USAGE_CPU_READ_WRITE;
-	nvpb->base.size = nvpb->bo->size;
-	return &nvpb->base;
+	pipe_reference_init(&pb->reference, 1);
+	pb->screen = pscreen;
+	pb->alignment = 0;
+	pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
+		    PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+	pb->size = nouveau_bo(pb)->size;
+	return pscreen->texture_blanket(pscreen, templ, &stride, pb);
 }
 
 static boolean
-nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb,
-			   unsigned *handle)
+nouveau_drm_name_from_pt(struct drm_api *api, struct pipe_screen *pscreen,
+			 struct pipe_texture *pt, unsigned *stride,
+			 unsigned *handle)
 {
-	struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb);
+	struct nouveau_miptree *mt = nouveau_miptree(pt);
 
-	if (!nvpb)
-		return FALSE;
+	if (!mt || !mt->bo)
+		return false;
 
-	*handle = nvpb->bo->handle;
-	return TRUE;
+	return nouveau_bo_handle_get(mt->bo, handle) == 0;
 }
 
 static boolean
-nouveau_drm_name_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb,
-			 unsigned *handle)
+nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen,
+			   struct pipe_texture *pt, unsigned *stride,
+			   unsigned *handle)
 {
-	struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb);
+	struct nouveau_miptree *mt = nouveau_miptree(pt);
 
-	if (!nvpb)
-		return FALSE;
+	if (!mt || !mt->bo)
+		return false;
 
-	return nouveau_bo_handle_get(nvpb->bo, handle) == 0;
+	*handle = mt->bo->handle;
+	return true;
 }
 
 struct drm_api drm_api_hooks = {
 	.create_screen = nouveau_drm_create_screen,
 	.create_context = nouveau_drm_create_context,
-	.buffer_from_texture = nouveau_drm_pb_from_pt,
-	.buffer_from_handle = nouveau_drm_pb_from_handle,
-	.handle_from_buffer = nouveau_drm_handle_from_pb,
-	.global_handle_from_buffer = nouveau_drm_name_from_pb,
+	.texture_from_shared_handle = nouveau_drm_pt_from_name,
+	.shared_handle_from_texture = nouveau_drm_name_from_pt,
+	.local_handle_from_texture = nouveau_drm_handle_from_pt,
 };
 
+struct drm_api *
+drm_api_create() {
+	return &drm_api_hooks;
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
index 2782c83c0e..e61e0e0957 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
@@ -1,5 +1,34 @@
 #ifndef __NOUVEAU_DRM_API_H__
 #define __NOUVEAU_DRM_API_H__
+
 #include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nouveau_dri.h"
+
+struct nouveau_winsys {
+	struct pipe_winsys base;
+
+	struct pipe_screen *pscreen;
+
+	unsigned nr_pctx;
+	struct pipe_context **pctx;
+
+	struct pipe_surface *front;
+};
+
+static INLINE struct nouveau_winsys *
+nouveau_winsys(struct pipe_winsys *ws)
+{
+	return (struct nouveau_winsys *)ws;
+}
+
+static INLINE struct nouveau_winsys *
+nouveau_winsys_screen(struct pipe_screen *pscreen)
+{
+	return nouveau_winsys(pscreen->winsys);
+}
 
 #endif
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c
deleted file mode 100644
index e3175fd775..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c
+++ /dev/null
@@ -1,94 +0,0 @@
-#include "util/u_memory.h"
-
-#include "nouveau_winsys_pipe.h"
-
-static int
-nouveau_pipe_notifier_alloc(struct nouveau_winsys *nvws, int count,
-			    struct nouveau_notifier **notify)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws);
-
-	return nouveau_notifier_alloc(nvpws->channel, nvpws->next_handle++,
-				      count, notify);
-}
-
-static int
-nouveau_pipe_grobj_alloc(struct nouveau_winsys *nvws, int grclass,
-			 struct nouveau_grobj **grobj)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws);
-	struct nouveau_channel *chan = nvpws->channel;
-	int ret;
-
-	ret = nouveau_grobj_alloc(chan, nvpws->next_handle++, grclass, grobj);
-	if (ret)
-		return ret;
-
-	BEGIN_RING(chan, *grobj, 0x0000, 1);
-	OUT_RING  (chan, (*grobj)->handle);
-	(*grobj)->bound = NOUVEAU_GROBJ_BOUND_EXPLICIT;
-	return 0;
-}
-
-static int
-nouveau_pipe_push_reloc(struct nouveau_winsys *nvws, void *ptr,
-			struct pipe_buffer *buf, uint32_t data,
-			uint32_t flags, uint32_t vor, uint32_t tor)
-{
-	struct nouveau_bo *bo = nouveau_pipe_buffer(buf)->bo;
-
-	return nouveau_pushbuf_emit_reloc(nvws->channel, ptr, bo,
-					  data, flags, vor, tor);
-}
-
-static int
-nouveau_pipe_push_flush(struct nouveau_winsys *nvws, unsigned size,
-			struct pipe_fence_handle **fence)
-{
-	if (fence)
-		*fence = NULL;
-
-	return nouveau_pushbuf_flush(nvws->channel, size);
-}
-
-static struct nouveau_bo *
-nouveau_pipe_get_bo(struct pipe_buffer *pb)
-{
-	return nouveau_pipe_buffer(pb)->bo;
-}
-
-struct nouveau_winsys *
-nouveau_winsys_new(struct pipe_winsys *ws)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_winsys *nvws;
-
-	nvws = CALLOC_STRUCT(nouveau_winsys);
-	if (!nvws)
-		return NULL;
-
-	nvws->ws		= ws;
-	nvws->channel		= nvpws->channel;
-
-	nvws->res_init		= nouveau_resource_init;
-	nvws->res_alloc		= nouveau_resource_alloc;
-	nvws->res_free		= nouveau_resource_free;
-
-	nvws->push_reloc        = nouveau_pipe_push_reloc;
-	nvws->push_flush	= nouveau_pipe_push_flush;
-
-	nvws->grobj_alloc	= nouveau_pipe_grobj_alloc;
-	nvws->grobj_free	= nouveau_grobj_free;
-
-	nvws->notifier_alloc	= nouveau_pipe_notifier_alloc;
-	nvws->notifier_free	= nouveau_notifier_free;
-	nvws->notifier_reset	= nouveau_notifier_reset;
-	nvws->notifier_status	= nouveau_notifier_status;
-	nvws->notifier_retval	= nouveau_notifier_return_val;
-	nvws->notifier_wait	= nouveau_notifier_wait_status;
-
-	nvws->get_bo		= nouveau_pipe_get_bo;
-
-	return nvws;
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c
deleted file mode 100644
index 9e03a9f5db..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c
+++ /dev/null
@@ -1,204 +0,0 @@
-#include "pipe/internal/p_winsys_screen.h"
-#include <pipe/p_defines.h>
-#include <pipe/p_inlines.h>
-#include <util/u_memory.h>
-
-#include "nouveau_winsys_pipe.h"
-
-#include "nouveau_drmif.h"
-#include "nouveau_bo.h"
-
-static const char *
-nouveau_get_name(struct pipe_winsys *pws)
-{
-	return "Nouveau/DRI";
-}
-
-static uint32_t
-nouveau_flags_from_usage(struct pipe_winsys *ws, unsigned usage)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct pipe_screen *pscreen = nvpws->pscreen;
-	uint32_t flags = NOUVEAU_BO_LOCAL;
-
-	if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER)
-		flags |= NOUVEAU_BO_GART;
-
-	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
-		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
-			flags |= NOUVEAU_BO_GART;
-		if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE))
-			flags |= NOUVEAU_BO_VRAM;
-
-		switch (nvpws->channel->device->chipset & 0xf0) {
-		case 0x50:
-		case 0x80:
-		case 0x90:
-			flags |= NOUVEAU_BO_TILED;
-			if (usage & NOUVEAU_BUFFER_USAGE_ZETA)
-				flags |= NOUVEAU_BO_ZTILE;
-			break;
-		default:
-			break;
-		}
-	}
-
-	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
-		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF))
-			flags |= NOUVEAU_BO_GART;
-	}
-
-	if (usage & PIPE_BUFFER_USAGE_INDEX) {
-		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF))
-			flags |= NOUVEAU_BO_GART;
-	}
-
-	return flags;
-}
-
-static struct pipe_buffer *
-nouveau_pipe_bo_create(struct pipe_winsys *ws, unsigned alignment,
-		       unsigned usage, unsigned size)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvbuf;
-	uint32_t flags;
-
-	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvbuf)
-		return NULL;
-	pipe_reference_init(&nvbuf->base.reference, 1);
-	nvbuf->base.alignment = alignment;
-	nvbuf->base.usage = usage;
-	nvbuf->base.size = size;
-
-	flags = nouveau_flags_from_usage(ws, usage);
-	if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) {
-		FREE(nvbuf);
-		return NULL;
-	}
-
-	return &nvbuf->base;
-}
-
-static struct pipe_buffer *
-nouveau_pipe_bo_user_create(struct pipe_winsys *ws, void *ptr, unsigned bytes)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvbuf;
-
-	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvbuf)
-		return NULL;
-	pipe_reference_init(&nvbuf->base.reference, 1);
-	nvbuf->base.size = bytes;
-
-	if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) {
-		FREE(nvbuf);
-		return NULL;
-	}
-
-	return &nvbuf->base;
-}
-
-static void
-nouveau_pipe_bo_del(struct pipe_buffer *buf)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-
-	nouveau_bo_ref(NULL, &nvbuf->bo);
-	FREE(nvbuf);
-}
-
-static void *
-nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
-		    unsigned flags)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-	uint32_t map_flags = 0;
-
-	if (flags & PIPE_BUFFER_USAGE_CPU_READ)
-		map_flags |= NOUVEAU_BO_RD;
-	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
-		map_flags |= NOUVEAU_BO_WR;
-
-	if (nouveau_bo_map(nvbuf->bo, map_flags))
-		return NULL;
-	return nvbuf->bo->map;
-}
-
-static void
-nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-
-	nouveau_bo_unmap(nvbuf->bo);
-}
-
-static void
-nouveau_pipe_fence_reference(struct pipe_winsys *ws,
-			     struct pipe_fence_handle **ptr,
-			     struct pipe_fence_handle *pfence)
-{
-	*ptr = pfence;
-}
-
-static int
-nouveau_pipe_fence_signalled(struct pipe_winsys *ws,
-			     struct pipe_fence_handle *pfence, unsigned flag)
-{
-	return 0;
-}
-
-static int
-nouveau_pipe_fence_finish(struct pipe_winsys *ws,
-			  struct pipe_fence_handle *pfence, unsigned flag)
-{
-	return 0;
-}
-
-static void
-nouveau_destroy(struct pipe_winsys *ws)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-
-	nouveau_device_close(&nvpws->channel->device);
-	FREE(nvpws);
-}
-
-struct pipe_winsys *
-nouveau_pipe_winsys_new(struct nouveau_device *dev)
-{
-	struct nouveau_pipe_winsys *nvpws;
-	int ret;
-
-	nvpws = CALLOC_STRUCT(nouveau_pipe_winsys);
-	if (!nvpws)
-		return NULL;
-
-	ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202,
-				    &nvpws->channel);
-	if (ret) {
-		debug_printf("%s: error opening GPU channel: %d\n",
-			     __func__, ret);
-		FREE(nvpws);
-		return NULL;
-	}
-	nvpws->next_handle = 0x77000000;
-
-	nvpws->base.buffer_create = nouveau_pipe_bo_create;
-	nvpws->base.buffer_destroy = nouveau_pipe_bo_del;
-	nvpws->base.user_buffer_create = nouveau_pipe_bo_user_create;
-	nvpws->base.buffer_map = nouveau_pipe_bo_map;
-	nvpws->base.buffer_unmap = nouveau_pipe_bo_unmap;
-
-	nvpws->base.fence_reference = nouveau_pipe_fence_reference;
-	nvpws->base.fence_signalled = nouveau_pipe_fence_signalled;
-	nvpws->base.fence_finish = nouveau_pipe_fence_finish;
-
-	nvpws->base.get_name = nouveau_get_name;
-	nvpws->base.destroy = nouveau_destroy;
-	return &nvpws->base;
-}
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
deleted file mode 100644
index 10e1e269e8..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
+++ /dev/null
@@ -1,52 +0,0 @@
-#ifndef NOUVEAU_PIPE_WINSYS_H
-#define NOUVEAU_PIPE_WINSYS_H
-
-#include "pipe/internal/p_winsys_screen.h"
-#include "pipe/p_context.h"
-
-#include "nouveau/nouveau_winsys.h"
-
-#include "nouveau_device.h"
-
-struct nouveau_pipe_buffer {
-	struct pipe_buffer base;
-	struct nouveau_bo *bo;
-};
-
-static INLINE struct nouveau_pipe_buffer *
-nouveau_pipe_buffer(struct pipe_buffer *buf)
-{
-	return (struct nouveau_pipe_buffer *)buf;
-}
-
-struct nouveau_pipe_winsys {
-	struct pipe_winsys base;
-
-	struct pipe_screen *pscreen;
-
-	struct nouveau_channel *channel;
-	uint32_t next_handle;
-
-	unsigned nr_pctx;
-	struct pipe_context **pctx;
-};
-
-static INLINE struct nouveau_pipe_winsys *
-nouveau_pipe_winsys(struct pipe_winsys *ws)
-{
-	return (struct nouveau_pipe_winsys *)ws;
-}
-
-static INLINE struct nouveau_pipe_winsys *
-nouveau_screen(struct pipe_screen *pscreen)
-{
-	return nouveau_pipe_winsys(pscreen->winsys);
-}
-
-struct pipe_winsys *
-nouveau_pipe_winsys_new(struct nouveau_device *);
-
-struct nouveau_winsys *
-nouveau_winsys_new(struct pipe_winsys *ws);
-
-#endif
diff --git a/src/gallium/winsys/drm/radeon/SConscript b/src/gallium/winsys/drm/radeon/SConscript
index 8f99055b2f..b2dfd504d4 100644
--- a/src/gallium/winsys/drm/radeon/SConscript
+++ b/src/gallium/winsys/drm/radeon/SConscript
@@ -4,4 +4,4 @@ SConscript(['core/SConscript',])
 
 if 'mesa' in env['statetrackers']:
 
-    SConscript(['dri2/SConscript'])
+    SConscript(['dri/SConscript'])
diff --git a/src/gallium/winsys/drm/radeon/core/SConscript b/src/gallium/winsys/drm/radeon/core/SConscript
index 578174e32b..2ad68e403f 100644
--- a/src/gallium/winsys/drm/radeon/core/SConscript
+++ b/src/gallium/winsys/drm/radeon/core/SConscript
@@ -11,7 +11,9 @@ radeon_sources = [
 
 env.Append(CPPPATH = '#/src/gallium/drivers/r300')
 
-env.ConvenienceLibrary(
+radeonwinsys = env.ConvenienceLibrary(
     target ='radeonwinsys',
     source = radeon_sources,
 )
+
+Export('radeonwinsys')
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index a15487352b..07551e7cd1 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -1,8 +1,8 @@
-/* 
+/*
  * Copyright © 2008 Jérôme Glisse
  *             2009 Corbin Simpson
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -10,14 +10,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -72,6 +72,7 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws,
             alignment, domain, 0);
     if (radeon_buffer->bo == NULL) {
         FREE(radeon_buffer);
+        return NULL;
     }
     return &radeon_buffer->base;
 }
@@ -93,6 +94,29 @@ static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws,
     return &radeon_buffer->base;
 }
 
+static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws,
+                                                        unsigned width,
+                                                        unsigned height,
+                                                        enum pipe_format format,
+                                                        unsigned usage,
+                                                        unsigned *stride)
+{
+    struct pipe_format_block block;
+    unsigned nblocksx, nblocksy, size;
+
+    pf_get_block(format, &block);
+
+    nblocksx = pf_get_nblocksx(&block, width);
+    nblocksy = pf_get_nblocksy(&block, height);
+
+    /* Radeons enjoy things in multiples of 32. */
+    /* XXX this can be 32 when POT */
+    *stride = (nblocksx * block.size + 63) & ~63;
+    size = *stride * nblocksy;
+
+    return radeon_buffer_create(ws, 64, usage, size);
+}
+
 static void radeon_buffer_del(struct pipe_buffer *buffer)
 {
     struct radeon_pipe_buffer *radeon_buffer =
@@ -110,16 +134,17 @@ static void *radeon_buffer_map(struct pipe_winsys *ws,
         (struct radeon_pipe_buffer*)buffer;
     int write = 0;
 
-    if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
-       /* XXX Remove this when radeon_bo_map supports DONTBLOCK */
-       return NULL;
+    if (!(flags & PIPE_BUFFER_USAGE_DONTBLOCK)) {
+        radeon_bo_wait(radeon_buffer->bo);
     }
     if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) {
         write = 1;
     }
 
-    if (radeon_bo_map(radeon_buffer->bo, write))
+    if (radeon_bo_map(radeon_buffer->bo, write)) {
         return NULL;
+    }
+
     return radeon_buffer->bo->ptr;
 }
 
@@ -175,15 +200,17 @@ struct radeon_winsys* radeon_pipe_winsys(int fd)
         return NULL;
     }
 
+    radeon_ws->priv->fd = fd;
     radeon_ws->priv->bom = radeon_bo_manager_gem_ctor(fd);
 
     radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer;
 
     radeon_ws->base.buffer_create = radeon_buffer_create;
-    radeon_ws->base.buffer_destroy = radeon_buffer_del;
     radeon_ws->base.user_buffer_create = radeon_buffer_user_create;
+    radeon_ws->base.surface_buffer_create = radeon_surface_buffer_create;
     radeon_ws->base.buffer_map = radeon_buffer_map;
     radeon_ws->base.buffer_unmap = radeon_buffer_unmap;
+    radeon_ws->base.buffer_destroy = radeon_buffer_del;
 
     radeon_ws->base.fence_reference = radeon_fence_reference;
     radeon_ws->base.fence_signalled = radeon_fence_signalled;
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
index ca8bbb3c11..f5153b06af 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
@@ -1,7 +1,7 @@
-/* 
+/*
  * Copyright © 2008 Jérôme Glisse
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -9,14 +9,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -48,19 +48,19 @@
 struct radeon_pipe_buffer {
     struct pipe_buffer  base;
     struct radeon_bo    *bo;
+    boolean flinked;
+    uint32_t flink;
 };
 
 #define RADEON_MAX_BOS 24
 
 struct radeon_winsys_priv {
+    /* DRM FD */
+    int fd;
+
     /* Radeon BO manager. */
     struct radeon_bo_manager* bom;
 
-    /* Radeon BO space checker. */
-    struct radeon_cs_space_check sc[RADEON_MAX_BOS];
-    /* Current BO count. */
-    unsigned bo_count;
-
     /* Radeon CS manager. */
     struct radeon_cs_manager* csm;
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index 428d3f65a1..47376a0f07 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -1,7 +1,7 @@
-/* 
+/*
  * Copyright © 2009 Corbin Simpson
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -9,14 +9,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -26,13 +26,15 @@
 /*
  * Authors:
  *      Corbin Simpson <MostAwesomeDude@gmail.com>
+ *      Joakim Sindholt <opensource@zhasha.com>
  */
 
 #include "radeon_drm.h"
 
 /* Create a pipe_screen. */
-struct pipe_screen* radeon_create_screen(int drmFB,
-					 struct drm_create_screen_arg *arg)
+struct pipe_screen* radeon_create_screen(struct drm_api* api,
+                                         int drmFB,
+                                         struct drm_create_screen_arg *arg)
 {
     struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB);
 
@@ -46,25 +48,30 @@ struct pipe_screen* radeon_create_screen(int drmFB,
 }
 
 /* Create a pipe_context. */
-struct pipe_context* radeon_create_context(struct pipe_screen* screen)
+struct pipe_context* radeon_create_context(struct drm_api* api,
+                                           struct pipe_screen* screen)
 {
     if (getenv("RADEON_SOFTPIPE")) {
         return radeon_create_softpipe(screen->winsys);
     } else {
-        return r300_create_context(screen, screen->winsys);
+        return r300_create_context(screen,
+                                   (struct r300_winsys*)screen->winsys);
     }
 }
 
-boolean radeon_buffer_from_texture(struct pipe_texture* texture,
+boolean radeon_buffer_from_texture(struct drm_api* api,
+                                   struct pipe_texture* texture,
                                    struct pipe_buffer** buffer,
                                    unsigned* stride)
 {
-    return FALSE;
+    /* XXX fix this */
+    return r300_get_texture_buffer(texture, buffer, stride);
 }
 
 /* Create a buffer from a handle. */
 /* XXX what's up with name? */
-struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
+struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
+                                              struct pipe_screen* screen,
                                               const char* name,
                                               unsigned handle)
 {
@@ -91,33 +98,89 @@ struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
     return &radeon_buffer->base;
 }
 
-boolean radeon_handle_from_buffer(struct pipe_screen* screen,
-                                  struct pipe_buffer* buffer,
-                                  unsigned* handle)
+struct pipe_texture*
+radeon_texture_from_shared_handle(struct drm_api *api,
+                                  struct pipe_screen *screen,
+                                  struct pipe_texture *templ,
+                                  const char *name,
+                                  unsigned stride,
+                                  unsigned handle)
+{
+    struct pipe_buffer *buffer;
+
+    buffer = radeon_buffer_from_handle(api, screen, name, handle);
+    if (!buffer) {
+        return NULL;
+    }
+
+    return screen->texture_blanket(screen, templ, &stride, buffer);
+}
+
+boolean radeon_shared_handle_from_texture(struct drm_api *api,
+                                          struct pipe_screen *screen,
+                                          struct pipe_texture *texture,
+                                          unsigned *stride,
+                                          unsigned *handle)
 {
-    struct radeon_pipe_buffer* radeon_buffer =
-        (struct radeon_pipe_buffer*)buffer;
-    *handle = radeon_buffer->bo->handle;
+    int retval, fd;
+    struct drm_gem_flink flink;
+    struct radeon_pipe_buffer* radeon_buffer;
+    struct pipe_buffer* buffer = &radeon_buffer->base;
+    if (!radeon_buffer_from_texture(api, texture, buffer, stride)) {
+        return FALSE;
+    }
+
+    if (!radeon_buffer->flinked) {
+        fd = ((struct radeon_winsys*)screen->winsys)->priv->fd;
+
+        flink.handle = radeon_buffer->bo->handle;
+
+        retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink);
+        if (retval) {
+            debug_printf("radeon: DRM_IOCTL_GEM_FLINK failed, error %d\n",
+                    retval);
+            return FALSE;
+        }
+
+        radeon_buffer->flink = flink.name;
+        radeon_buffer->flinked = TRUE;
+    }
+
+    *handle = radeon_buffer->flink;
     return TRUE;
 }
 
-boolean radeon_global_handle_from_buffer(struct pipe_screen* screen,
-                                         struct pipe_buffer* buffer,
-                                         unsigned* handle)
+boolean radeon_local_handle_from_texture(struct drm_api *api,
+                                         struct pipe_screen *screen,
+                                         struct pipe_texture *texture,
+                                         unsigned *stride,
+                                         unsigned *handle)
 {
-    /* XXX WTF is the difference here? global? */
-    struct radeon_pipe_buffer* radeon_buffer =
-        (struct radeon_pipe_buffer*)buffer;
-    *handle = radeon_buffer->bo->handle;
+    struct pipe_buffer *buffer;
+    if (!radeon_buffer_from_texture(api, texture, &buffer, stride)) {
+        return FALSE;
+    }
+
+    *handle = ((struct radeon_pipe_buffer*)buffer)->bo->handle;
+
+    pipe_buffer_reference(&buffer, NULL);
+
     return TRUE;
 }
 
 struct drm_api drm_api_hooks = {
     .create_screen = radeon_create_screen,
     .create_context = radeon_create_context,
-    /* XXX fix this */
-    .buffer_from_texture = r300_get_texture_buffer,
-    .buffer_from_handle = radeon_buffer_from_handle,
-    .handle_from_buffer = radeon_handle_from_buffer,
-    .global_handle_from_buffer = radeon_global_handle_from_buffer,
+    .texture_from_shared_handle = radeon_texture_from_shared_handle,
+    .shared_handle_from_texture = radeon_shared_handle_from_texture,
+    .local_handle_from_texture = radeon_local_handle_from_texture,
 };
+
+struct drm_api* drm_api_create()
+{
+#ifdef DEBUG
+    return trace_drm_create(&drm_api_hooks);
+#else
+    return &drm_api_hooks;
+#endif
+}
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
index 049f9984db..88a5c82b28 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
@@ -30,8 +30,13 @@
 #ifndef RADEON_DRM_H
 #define RADEON_DRM_H
 
+#include <sys/ioctl.h>
+
+#include "xf86drm.h"
+
 #include "pipe/p_screen.h"
 
+#include "trace/tr_drm.h"
 #include "util/u_memory.h"
 
 #include "state_tracker/drm_api.h"
@@ -40,25 +45,35 @@
 #include "radeon_r300.h"
 #include "radeon_winsys_softpipe.h"
 
-struct pipe_screen* radeon_create_screen(int drmFB,
+/* XXX */
+#include "r300_screen.h"
+
+struct pipe_screen* radeon_create_screen(struct drm_api* api,
+                                         int drmFB,
 					 struct drm_create_screen_arg *arg);
 
-struct pipe_context* radeon_create_context(struct pipe_screen* screen);
+struct pipe_context* radeon_create_context(struct drm_api* api,
+                                           struct pipe_screen* screen);
 
-boolean radeon_buffer_from_texture(struct pipe_texture* texture,
+boolean radeon_buffer_from_texture(struct drm_api* api,
+                                   struct pipe_texture* texture,
                                    struct pipe_buffer** buffer,
                                    unsigned* stride);
 
-struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
+struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
+                                              struct pipe_screen* screen,
                                               const char* name,
                                               unsigned handle);
 
-boolean radeon_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_handle_from_buffer(struct drm_api* api,
+                                  struct pipe_screen* screen,
                                   struct pipe_buffer* buffer,
                                   unsigned* handle);
 
-boolean radeon_global_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_global_handle_from_buffer(struct drm_api* api,
+                                         struct pipe_screen* screen,
                                          struct pipe_buffer* buffer,
                                          unsigned* handle);
 
+void radeon_destroy_drm_api(struct drm_api* api);
 #endif
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
index da233203d7..d723876221 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
@@ -22,54 +22,30 @@
 
 #include "radeon_r300.h"
 
-static void radeon_r300_add_buffer(struct r300_winsys* winsys,
-                                   struct pipe_buffer* pbuffer,
-                                   uint32_t rd,
-                                   uint32_t wd)
+static boolean radeon_r300_add_buffer(struct r300_winsys* winsys,
+                                      struct pipe_buffer* pbuffer,
+                                      uint32_t rd,
+                                      uint32_t wd)
 {
-    int i;
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
-    struct radeon_cs_space_check* sc = priv->sc;
     struct radeon_bo* bo = ((struct radeon_pipe_buffer*)pbuffer)->bo;
 
-    /* Check to see if this BO is already in line for validation;
-     * find a slot for it otherwise. */
-    for (i = 0; i < RADEON_MAX_BOS; i++) {
-        if (sc[i].bo == bo) {
-            return;
-        } else if (sc[i].bo == NULL) {
-            sc[i].bo = bo;
-            sc[i].read_domains = rd;
-            sc[i].write_domain = wd;
-            priv->bo_count = i + 1;
-            return;
-        }
-    }
-
-    assert(FALSE && "Oh God too many BOs!");
+    radeon_cs_space_add_persistent_bo(priv->cs, bo, rd, wd);
+    return TRUE;
 }
 
 static boolean radeon_r300_validate(struct r300_winsys* winsys)
 {
-    int retval;
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
-    struct radeon_cs_space_check* sc = priv->sc;
 
-    retval = radeon_cs_space_check(priv->cs, sc, priv->bo_count);
-
-    if (retval == RADEON_CS_SPACE_OP_TO_BIG) {
-        /* We might as well HCF, since this is not going to fit in the card,
-         * period. */
-	exit(1);
-    } else if (retval == RADEON_CS_SPACE_FLUSH) {
-        /* We must flush before more rendering can commence. */
-        return TRUE;
+    if (radeon_cs_space_check(priv->cs) < 0) {
+        return FALSE;
     }
 
     /* Things are fine, we can proceed as normal. */
-    return FALSE;
+    return TRUE;
 }
 
 static boolean radeon_r300_check_cs(struct r300_winsys* winsys, int size)
@@ -108,9 +84,15 @@ static void radeon_r300_write_cs_reloc(struct r300_winsys* winsys,
 {
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
+    int retval = 0;
 
-    radeon_cs_write_reloc(priv->cs,
+    retval = radeon_cs_write_reloc(priv->cs,
             ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags);
+
+    if (retval) {
+        debug_printf("radeon: Relocation of %p (%d, %d, %d) failed!\n",
+                pbuffer, rd, wd, flags);
+    }
 }
 
 static void radeon_r300_end_cs(struct r300_winsys* winsys,
@@ -128,57 +110,66 @@ static void radeon_r300_flush_cs(struct r300_winsys* winsys)
 {
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
-    int retval = 0;
+    int retval;
 
+    /* Emit the CS. */
     retval = radeon_cs_emit(priv->cs);
     if (retval) {
         debug_printf("radeon: Bad CS, dumping...\n");
         radeon_cs_print(priv->cs, stderr);
     }
+
+    /* Clean out BOs. */
+    radeon_cs_space_reset_bos(priv->cs);
+
+    /* Reset CS.
+     * Someday, when we care about performance, we should really find a way
+     * to rotate between two or three CS objects so that the GPU can be
+     * spinning through one CS while another one is being filled. */
     radeon_cs_erase(priv->cs);
 }
 
 /* Helper function to do the ioctls needed for setup and init. */
 static void do_ioctls(struct r300_winsys* winsys, int fd)
 {
-    struct drm_radeon_gem_info info;
-    drm_radeon_getparam_t gp;
-    int target;
+    struct drm_radeon_gem_info gem_info = {0};
+    struct drm_radeon_info info = {0};
+    int target = 0;
     int retval;
 
-    gp.value = &target;
+    info.value = &target;
 
     /* First, get the number of pixel pipes */
-    gp.param = RADEON_PARAM_NUM_GB_PIPES;
-    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    info.request = RADEON_INFO_NUM_GB_PIPES;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info));
     if (retval) {
-        fprintf(stderr, "%s: Failed to get GB pipe count, error number %d\n",
-                __FUNCTION__, retval);
+        fprintf(stderr, "%s: Failed to get GB pipe count, "
+                "error number %d\n", __FUNCTION__, retval);
         exit(1);
     }
     winsys->gb_pipes = target;
 
     /* Then, get PCI ID */
-    gp.param = RADEON_PARAM_DEVICE_ID;
-    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    info.request = RADEON_INFO_DEVICE_ID;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info));
     if (retval) {
-        fprintf(stderr, "%s: Failed to get PCI ID, error number %d\n",
-                __FUNCTION__, retval);
+        fprintf(stderr, "%s: Failed to get PCI ID, "
+                "error number %d\n", __FUNCTION__, retval);
         exit(1);
     }
     winsys->pci_id = target;
 
     /* Finally, retrieve MM info */
     retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO,
-            &info, sizeof(info));
+            &gem_info, sizeof(gem_info));
     if (retval) {
         fprintf(stderr, "%s: Failed to get MM info, error number %d\n",
                 __FUNCTION__, retval);
         exit(1);
     }
-    winsys->gart_size = info.gart_size;
+    winsys->gart_size = gem_info.gart_size;
     /* XXX */
-    winsys->vram_size = info.vram_visible;
+    winsys->vram_size = gem_info.vram_visible;
 }
 
 struct r300_winsys*
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
index 5c373cd084..775d7937fd 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
@@ -20,16 +20,23 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
+#ifndef RADEON_R300_H
+#define RADEON_R300_H
+
 /* XXX WTF is this! I shouldn't have to include those first three! FUCK! */
 #include <stdint.h>
 #include <stdlib.h>
 #include "drm.h"
 #include "radeon_drm.h"
-#include "radeon_cs.h"
+#include "radeon_cs_gem.h"
 
 #include "r300_winsys.h"
 
 #include "radeon_buffer.h"
 
+struct radeon_winsys;
+
 struct r300_winsys*
 radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys);
+
+#endif /* RADEON_R300_H */
diff --git a/src/gallium/winsys/drm/radeon/dri/Makefile b/src/gallium/winsys/drm/radeon/dri/Makefile
index c218ee9d01..a9889444de 100644
--- a/src/gallium/winsys/drm/radeon/dri/Makefile
+++ b/src/gallium/winsys/drm/radeon/dri/Makefile
@@ -10,6 +10,7 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/radeon/core/libradeonwinsys.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/r300/libr300.a
 
 C_SOURCES = \
diff --git a/src/gallium/winsys/drm/radeon/dri/SConscript b/src/gallium/winsys/drm/radeon/dri/SConscript
index f2cdee97d9..aea987a3ac 100644
--- a/src/gallium/winsys/drm/radeon/dri/SConscript
+++ b/src/gallium/winsys/drm/radeon/dri/SConscript
@@ -2,7 +2,10 @@ Import('*')
 
 env = drienv.Clone()
 
+env.ParseConfig('pkg-config --cflags --libs libdrm_radeon')
+
 drivers = [
+    trace,
     softpipe,
     r300
 ]
@@ -10,5 +13,5 @@ drivers = [
 env.SharedLibrary(
     target ='radeon_dri.so',
     source = COMMON_GALLIUM_SOURCES,
-    LIBS = drivers + mesa + auxiliaries + env['LIBS'],
+    LIBS = st_dri + radeonwinsys + mesa + drivers + auxiliaries + env['LIBS'],
 )
diff --git a/src/gallium/winsys/drm/radeon/egl/Makefile b/src/gallium/winsys/drm/radeon/egl/Makefile
index d989b3aa93..6a1448d1b9 100644
--- a/src/gallium/winsys/drm/radeon/egl/Makefile
+++ b/src/gallium/winsys/drm/radeon/egl/Makefile
@@ -8,6 +8,7 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/egl/libegldrm.a \
 	$(GALLIUMDIR)/winsys/drm/radeon/core/libradeonwinsys.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/r300/libr300.a
 
 DRIVER_SOURCES =
diff --git a/src/gallium/winsys/drm/radeon/xorg/Makefile b/src/gallium/winsys/drm/radeon/xorg/Makefile
index 6ffd4a3a54..0241625f69 100644
--- a/src/gallium/winsys/drm/radeon/xorg/Makefile
+++ b/src/gallium/winsys/drm/radeon/xorg/Makefile
@@ -37,6 +37,6 @@ clean:
 
 install:
 	$(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
 
 .PHONY	= all clean install
diff --git a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
index 36824251f0..837f2aa8fe 100644
--- a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
+++ b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
@@ -48,8 +48,8 @@ static SymTabRec radeon_xorg_chipsets[] = {
 };
 
 static PciChipsets radeon_xorg_pci_devices[] = {
-    {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA},
-    {-1, -1, RES_UNDEFINED}
+    {PCI_MATCH_ANY, PCI_MATCH_ANY, NULL},
+    {-1, -1, NULL}
 };
 
 static XF86ModuleVersionInfo radeon_xorg_version = {
@@ -106,12 +106,6 @@ radeon_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin)
 	xf86AddDriver(&modesetting, module, HaveDriverFuncs);
 
 	/*
-	 * Tell the loader about symbols from other modules that this module
-	 * might refer to.
-	 */
-	xorg_tracker_loader_ref_sym_lists();
-
-	/*
 	 * The return value must be non-NULL on success even though there
 	 * is no TearDownProc.
 	 */
@@ -141,7 +135,7 @@ radeon_xorg_pci_probe(DriverPtr driver,
 			       NULL, NULL, NULL, NULL, NULL);
     if (scrn != NULL) {
 	scrn->driverVersion = 1;
-	scrn->driverName = "modesetting";
+	scrn->driverName = "radeon";
 	scrn->name = "modesetting";
 	scrn->Probe = NULL;
 
diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile
index 8646ee3b52..3efb7ed4af 100644
--- a/src/gallium/winsys/egl_xlib/Makefile
+++ b/src/gallium/winsys/egl_xlib/Makefile
@@ -37,7 +37,7 @@ UNUSED_LIBS = \
 	$(TOP)/src/mesa/libmesagallium.a \
 
 
-LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1
+LOCAL_CFLAGS =
 
 
 .c.o:
@@ -74,7 +74,7 @@ depend: $(ALL_SOURCES)
 install: default
 	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
 	@if [ -e $(TOP)/$(LIB_DIR) ]; then \
-		$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \
+		$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \
 	fi
 
 
diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c
index b52f427e4a..d02f825047 100644
--- a/src/gallium/winsys/egl_xlib/egl_xlib.c
+++ b/src/gallium/winsys/egl_xlib/egl_xlib.c
@@ -33,6 +33,7 @@
 
 
 #include <dlfcn.h>
+#include <X11/Xlib.h>
 #include <X11/Xutil.h>
 
 #include "pipe/p_compiler.h"
@@ -61,6 +62,14 @@
 struct xlib_egl_driver
 {
    _EGLDriver Base;   /**< base class */
+   EGLint apis;
+};
+
+
+/** driver data of _EGLDisplay */
+struct xlib_egl_display
+{
+   Display *Dpy;
 
    struct pipe_winsys *winsys;
    struct pipe_screen *screen;
@@ -82,6 +91,7 @@ struct xlib_egl_surface
 {
    _EGLSurface Base;   /**< base class */
 
+   /* These are set for window surface */
    Display *Dpy;  /**< The X Display of the window */
    Window Win;    /**< The user-created window ID */
    GC Gc;
@@ -93,6 +103,12 @@ struct xlib_egl_surface
 };
 
 
+static void
+flush_frontbuffer(struct pipe_winsys *pws,
+                  struct pipe_surface *psurf,
+                  void *context_private);
+
+
 /** cast wrapper */
 static INLINE struct xlib_egl_driver *
 xlib_egl_driver(_EGLDriver *drv)
@@ -101,19 +117,24 @@ xlib_egl_driver(_EGLDriver *drv)
 }
 
 
-static struct xlib_egl_surface *
-lookup_surface(EGLSurface surf)
+static INLINE struct xlib_egl_display *
+xlib_egl_display(_EGLDisplay *dpy)
 {
-   _EGLSurface *surface = _eglLookupSurface(surf);
-   return (struct xlib_egl_surface *) surface;
+   return (struct xlib_egl_display *) dpy->DriverData;
 }
 
 
-static struct xlib_egl_context *
-lookup_context(EGLContext surf)
+static INLINE struct xlib_egl_surface *
+lookup_surface(_EGLSurface *surf)
 {
-   _EGLContext *context = _eglLookupContext(surf);
-   return (struct xlib_egl_context *) context;
+   return (struct xlib_egl_surface *) surf;
+}
+
+
+static INLINE struct xlib_egl_context *
+lookup_context(_EGLContext *ctx)
+{
+   return (struct xlib_egl_context *) ctx;
 }
 
 
@@ -132,19 +153,18 @@ bitcount(unsigned int n)
  * Create the EGLConfigs.  (one per X visual)
  */
 static void
-create_configs(_EGLDriver *drv, EGLDisplay dpy)
+create_configs(struct xlib_egl_display *xdpy, _EGLDisplay *disp)
 {
    static const EGLint all_apis = (EGL_OPENGL_ES_BIT |
                                    EGL_OPENGL_ES2_BIT |
                                    EGL_OPENVG_BIT |
                                    EGL_OPENGL_BIT);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
    XVisualInfo *visInfo, visTemplate;
    int num_visuals, i;
 
    /* get list of all X visuals, create an EGL config for each */
-   visTemplate.screen = DefaultScreen(disp->Xdpy);
-   visInfo = XGetVisualInfo(disp->Xdpy, VisualScreenMask,
+   visTemplate.screen = DefaultScreen(xdpy->Dpy);
+   visInfo = XGetVisualInfo(xdpy->Dpy, VisualScreenMask,
                             &visTemplate, &num_visuals);
    if (!visInfo) {
       printf("egl_xlib.c: couldn't get any X visuals\n");
@@ -180,10 +200,14 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy)
       SET_CONFIG_ATTRIB(config, EGL_NATIVE_RENDERABLE, EGL_FALSE);
       SET_CONFIG_ATTRIB(config, EGL_CONFORMANT, all_apis);
       SET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE, all_apis);
-      SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT);
+      SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT | EGL_PBUFFER_BIT);
+      SET_CONFIG_ATTRIB(config, EGL_BIND_TO_TEXTURE_RGBA, EGL_TRUE);
+      SET_CONFIG_ATTRIB(config, EGL_BIND_TO_TEXTURE_RGB, EGL_TRUE);
 
       _eglAddConfig(disp, config);
    }
+
+   XFree(visInfo);
 }
 
 
@@ -191,16 +215,47 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy)
  * Called via eglInitialize(), drv->API.Initialize().
  */
 static EGLBoolean
-xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
-                   EGLint *minor, EGLint *major)
+xlib_eglInitialize(_EGLDriver *drv, _EGLDisplay *dpy,
+                   EGLint *major, EGLint *minor)
 {
-   create_configs(drv, dpy);
+   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
+   struct xlib_egl_display *xdpy;
+
+   xdpy = CALLOC_STRUCT(xlib_egl_display);
+   if (!xdpy)
+      return _eglError(EGL_BAD_ALLOC, "eglInitialize");
+
+   xdpy->Dpy = (Display *) dpy->NativeDisplay;
+   if (!xdpy->Dpy) {
+      xdpy->Dpy = XOpenDisplay(NULL);
+      if (!xdpy->Dpy) {
+         free(xdpy);
+         return EGL_FALSE;
+      }
+   }
 
-   drv->Initialized = EGL_TRUE;
+   /* create winsys and pipe screen */
+   xdpy->winsys = create_sw_winsys();
+   if (!xdpy->winsys) {
+      free(xdpy);
+      return _eglError(EGL_BAD_ALLOC, "eglInitialize");
+   }
+   xdpy->winsys->flush_frontbuffer = flush_frontbuffer;
+   xdpy->screen = softpipe_create_screen(xdpy->winsys);
+   if (!xdpy->screen) {
+      free(xdpy->winsys);
+      free(xdpy);
+      return _eglError(EGL_BAD_ALLOC, "eglInitialize");
+   }
+
+   dpy->DriverData = (void *) xdpy;
+   dpy->ClientAPIsMask = xdrv->apis;
+
+   create_configs(xdpy, dpy);
 
    /* we're supporting EGL 1.4 */
-   *minor = 1;
-   *major = 4;
+   *major = 1;
+   *minor = 4;
 
    return EGL_TRUE;
 }
@@ -210,8 +265,20 @@ xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
  * Called via eglTerminate(), drv->API.Terminate().
  */
 static EGLBoolean
-xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+xlib_eglTerminate(_EGLDriver *drv, _EGLDisplay *dpy)
 {
+   struct xlib_egl_display *xdpy = xlib_egl_display(dpy);
+
+   _eglReleaseDisplayResources(drv, dpy);
+   _eglCleanupDisplay(dpy);
+
+   xdpy->screen->destroy(xdpy->screen);
+   free(xdpy->winsys);
+
+   if (!dpy->NativeDisplay)
+      XCloseDisplay(xdpy->Dpy);
+   free(xdpy);
+
    return EGL_TRUE;
 }
 
@@ -264,7 +331,13 @@ static void
 check_and_update_buffer_size(struct xlib_egl_surface *surface)
 {
    uint width, height;
-   get_drawable_size(surface->Dpy, surface->Win, &width, &height);
+   if (surface->Base.Type == EGL_PBUFFER_BIT) {
+      width = surface->Base.Width;
+      height = surface->Base.Height;
+   }
+   else {
+      get_drawable_size(surface->Dpy, surface->Win, &width, &height);
+   }
    st_resize_framebuffer(surface->Framebuffer, width, height);
    surface->Base.Width = width;
    surface->Base.Height = height;
@@ -281,6 +354,9 @@ display_surface(struct pipe_winsys *pws,
    XImage *ximage;
    void *data;
 
+   if (xsurf->Base.Type == EGL_PBUFFER_BIT)
+      return;
+
    ximage = XCreateImage(xsurf->Dpy,
                          xsurf->VisInfo.visual,
                          xsurf->VisInfo.depth,
@@ -330,24 +406,23 @@ flush_frontbuffer(struct pipe_winsys *pws,
 /**
  * Called via eglCreateContext(), drv->API.CreateContext().
  */
-static EGLContext
-xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
-                      EGLContext share_list, const EGLint *attrib_list)
+static _EGLContext *
+xlib_eglCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
+                      _EGLContext *share_list, const EGLint *attrib_list)
 {
-   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
-   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
+   struct xlib_egl_display *xdpy = xlib_egl_display(dpy);
    struct xlib_egl_context *ctx;
    struct st_context *share_ctx = NULL; /* XXX fix */
    __GLcontextModes visual;
 
    ctx = CALLOC_STRUCT(xlib_egl_context);
    if (!ctx)
-      return EGL_NO_CONTEXT;
+      return NULL;
 
    /* let EGL lib init the common stuff */
-   if (!_eglInitContext(drv, dpy, &ctx->Base, config, attrib_list)) {
+   if (!_eglInitContext(drv, &ctx->Base, conf, attrib_list)) {
       free(ctx);
-      return EGL_NO_CONTEXT;
+      return NULL;
    }
 
    /* API-dependent context creation */
@@ -359,7 +434,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       /* fall-through */
    case EGL_OPENGL_API:
       /* create a softpipe context */
-      ctx->pipe = softpipe_create(xdrv->screen);
+      ctx->pipe = softpipe_create(xdpy->screen);
       /* Now do xlib / state tracker inits here */
       _eglConfigToContextModesRec(conf, &visual);
       ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx);
@@ -367,43 +442,33 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    default:
       _eglError(EGL_BAD_MATCH, "eglCreateContext(unsupported API)");
       free(ctx);
-      return EGL_NO_CONTEXT;
+      return NULL;
    }
 
-   _eglSaveContext(&ctx->Base);
-
-   return _eglGetContextHandle(&ctx->Base);
+   return &ctx->Base;
 }
 
 
 static EGLBoolean
-xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
+xlib_eglDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx)
 {
    struct xlib_egl_context *context = lookup_context(ctx);
-   if (context) {
-      if (context->Base.IsBound) {
-         context->Base.DeletePending = EGL_TRUE;
-      }
-      else {
-         /* API-dependent clean-up */
-         switch (context->Base.ClientAPI) {
-         case EGL_OPENGL_ES_API:
-         case EGL_OPENVG_API:
-            /* fall-through */
-         case EGL_OPENGL_API:
-            st_destroy_context(context->Context);
-            break;
-         default:
-            assert(0);
-         }
-         free(context);
+
+   if (!_eglIsContextBound(&context->Base)) {
+      /* API-dependent clean-up */
+      switch (context->Base.ClientAPI) {
+      case EGL_OPENGL_ES_API:
+      case EGL_OPENVG_API:
+         /* fall-through */
+      case EGL_OPENGL_API:
+         st_destroy_context(context->Context);
+         break;
+      default:
+         assert(0);
       }
-      return EGL_TRUE;
-   }
-   else {
-      _eglError(EGL_BAD_CONTEXT, "eglDestroyContext");
-      return EGL_TRUE;
+      free(context);
    }
+   return EGL_TRUE;
 }
 
 
@@ -411,16 +476,25 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
  * Called via eglMakeCurrent(), drv->API.MakeCurrent().
  */
 static EGLBoolean
-xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy,
-                    EGLSurface draw, EGLSurface read, EGLContext ctx)
+xlib_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy,
+                    _EGLSurface *draw, _EGLSurface *read, _EGLContext *ctx)
 {
    struct xlib_egl_context *context = lookup_context(ctx);
    struct xlib_egl_surface *draw_surf = lookup_surface(draw);
    struct xlib_egl_surface *read_surf = lookup_surface(read);
+   struct st_context *oldcontext = NULL;
+   _EGLContext *oldctx;
+
+   oldctx = _eglGetCurrentContext();
+   if (oldctx && _eglIsContextLinked(oldctx))
+      oldcontext = st_get_current();
 
-   if (!_eglMakeCurrent(drv, dpy, draw, read, context))
+   if (!_eglMakeCurrent(drv, dpy, draw, read, ctx))
       return EGL_FALSE;
 
+   /* Flush before switching context.  Check client API? */
+   if (oldcontext)
+      st_flush(oldcontext, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, NULL);
    st_make_current((context ? context->Context : NULL),
                    (draw_surf ? draw_surf->Framebuffer : NULL),
                    (read_surf ? read_surf->Framebuffer : NULL));
@@ -474,39 +548,34 @@ choose_stencil_format(const __GLcontextModes *visual)
 /**
  * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
  */
-static EGLSurface
-xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+static _EGLSurface *
+xlib_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
                             NativeWindowType window, const EGLint *attrib_list)
 {
-   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
-   _EGLDisplay *disp = _eglLookupDisplay(dpy);
-   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
-
+   struct xlib_egl_display *xdpy = xlib_egl_display(disp);
    struct xlib_egl_surface *surf;
    __GLcontextModes visual;
    uint width, height;
 
    surf = CALLOC_STRUCT(xlib_egl_surface);
    if (!surf)
-      return EGL_NO_SURFACE;
+      return NULL;
 
    /* Let EGL lib init the common stuff */
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &surf->Base, EGL_WINDOW_BIT,
+                        conf, attrib_list)) {
       free(surf);
-      return EGL_NO_SURFACE;
+      return NULL;
    }
 
-   _eglSaveSurface(&surf->Base);
-
    /*
     * Now init the Xlib and gallium stuff
     */
    surf->Win = (Window) window;  /* The X window ID */
-   surf->Dpy = disp->Xdpy;  /* The X display */
+   surf->Dpy = xdpy->Dpy;  /* The X display */
    surf->Gc = XCreateGC(surf->Dpy, surf->Win, 0, NULL);
 
-   surf->winsys = xdrv->winsys;
+   surf->winsys = xdpy->winsys;
 
    _eglConfigToContextModesRec(conf, &visual);
    get_drawable_size(surf->Dpy, surf->Win, &width, &height);
@@ -525,35 +594,172 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    st_resize_framebuffer(surf->Framebuffer, width, height);
 
-   return _eglGetSurfaceHandle(&surf->Base);
+   return &surf->Base;
+}
+
+
+static _EGLSurface *
+xlib_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
+                             const EGLint *attrib_list)
+{
+   struct xlib_egl_display *xdpy = xlib_egl_display(disp);
+   struct xlib_egl_surface *surf;
+   __GLcontextModes visual;
+   uint width, height;
+   EGLBoolean bind_texture;
+
+   surf = CALLOC_STRUCT(xlib_egl_surface);
+   if (!surf) {
+      _eglError(EGL_BAD_ALLOC, "eglCreatePbufferSurface");
+      return NULL;
+   }
+
+   if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
+      free(surf);
+      return NULL;
+   }
+   if (surf->Base.Width < 0 || surf->Base.Height < 0) {
+      _eglError(EGL_BAD_PARAMETER, "eglCreatePbufferSurface");
+      free(surf);
+      return NULL;
+   }
+
+   bind_texture = (surf->Base.TextureFormat != EGL_NO_TEXTURE);
+   width = (uint) surf->Base.Width;
+   height = (uint) surf->Base.Height;
+   if ((surf->Base.TextureTarget == EGL_NO_TEXTURE && bind_texture) ||
+       (surf->Base.TextureTarget != EGL_NO_TEXTURE && !bind_texture)) {
+      _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface");
+      free(surf);
+      return NULL;
+   }
+   /* a framebuffer of zero width or height confuses st */
+   if (width == 0 || height == 0) {
+      _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface");
+      free(surf);
+      return NULL;
+   }
+   /* no mipmap generation */
+   if (surf->Base.MipmapTexture) {
+      _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface");
+      free(surf);
+      return NULL;
+   }
+
+   surf->winsys = xdpy->winsys;
+
+   _eglConfigToContextModesRec(conf, &visual);
+
+   /* Create GL statetracker framebuffer */
+   surf->Framebuffer = st_create_framebuffer(&visual,
+                                             choose_color_format(&visual),
+                                             choose_depth_format(&visual),
+                                             choose_stencil_format(&visual),
+                                             width, height,
+                                             (void *) surf);
+   st_resize_framebuffer(surf->Framebuffer, width, height);
+
+   return &surf->Base;
 }
 
 
 static EGLBoolean
-xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+xlib_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface)
 {
    struct xlib_egl_surface *surf = lookup_surface(surface);
-   if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->Base.IsBound) {
-         surf->Base.DeletePending = EGL_TRUE;
-      }
-      else {
+   if (!_eglIsSurfaceBound(&surf->Base)) {
+      if (surf->Base.Type != EGL_PBUFFER_BIT)
          XFreeGC(surf->Dpy, surf->Gc);
-         st_unreference_framebuffer(surf->Framebuffer);
-         free(surf);
-      }
-      return EGL_TRUE;
+      st_unreference_framebuffer(surf->Framebuffer);
+      free(surf);
    }
-   else {
-      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
-      return EGL_FALSE;
+   return EGL_TRUE;
+}
+
+
+static EGLBoolean
+xlib_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *dpy,
+                     _EGLSurface *surface, EGLint buffer)
+{
+   struct xlib_egl_surface *xsurf = lookup_surface(surface);
+   struct xlib_egl_context *xctx;
+   struct pipe_surface *psurf;
+   enum pipe_format format;
+   int target;
+
+   if (!xsurf || xsurf->Base.Type != EGL_PBUFFER_BIT)
+      return _eglError(EGL_BAD_SURFACE, "eglBindTexImage");
+   if (buffer != EGL_BACK_BUFFER)
+      return _eglError(EGL_BAD_PARAMETER, "eglBindTexImage");
+   if (xsurf->Base.BoundToTexture)
+      return _eglError(EGL_BAD_ACCESS, "eglBindTexImage");
+
+   /* this should be updated when choose_color_format is */
+   switch (xsurf->Base.TextureFormat) {
+   case EGL_TEXTURE_RGB:
+      format = PIPE_FORMAT_R8G8B8_UNORM;
+      break;
+   case EGL_TEXTURE_RGBA:
+      format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      break;
+   default:
+      return _eglError(EGL_BAD_MATCH, "eglBindTexImage");
+   }
+
+   switch (xsurf->Base.TextureTarget) {
+   case EGL_TEXTURE_2D:
+      target = ST_TEXTURE_2D;
+      break;
+   default:
+      return _eglError(EGL_BAD_MATCH, "eglBindTexImage");
+   }
+
+   /* flush properly */
+   if (eglGetCurrentSurface(EGL_DRAW) == surface) {
+      xctx = lookup_context(_eglGetCurrentContext());
+      st_flush(xctx->Context, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME,
+               NULL);
+   }
+   else if (_eglIsSurfaceBound(&xsurf->Base)) {
+      xctx = lookup_context(xsurf->Base.Binding);
+      if (xctx)
+         st_finish(xctx->Context);
    }
+
+   st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT,
+                              &psurf);
+   st_bind_texture_surface(psurf, target, xsurf->Base.MipmapLevel, format);
+   xsurf->Base.BoundToTexture = EGL_TRUE;
+
+   return EGL_TRUE;
+}
+
+
+static EGLBoolean
+xlib_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface,
+                        EGLint buffer)
+{
+   struct xlib_egl_surface *xsurf = lookup_surface(surface);
+   struct pipe_surface *psurf;
+
+   if (!xsurf || xsurf->Base.Type != EGL_PBUFFER_BIT ||
+       !xsurf->Base.BoundToTexture)
+      return _eglError(EGL_BAD_SURFACE, "eglReleaseTexImage");
+   if (buffer != EGL_BACK_BUFFER)
+      return _eglError(EGL_BAD_PARAMETER, "eglReleaseTexImage");
+
+   st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT,
+                              &psurf);
+   st_unbind_texture_surface(psurf, ST_TEXTURE_2D, xsurf->Base.MipmapLevel);
+   xsurf->Base.BoundToTexture = EGL_FALSE;
+
+   return EGL_TRUE;
 }
 
 
 static EGLBoolean
-xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+xlib_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw)
 {
    /* error checking step: */
    if (!_eglSwapBuffers(drv, dpy, draw))
@@ -588,7 +794,9 @@ find_supported_apis(void)
    EGLint mask = 0;
    void *handle;
 
-   handle = dlopen(NULL, 0);
+   handle = dlopen(NULL, RTLD_LAZY | RTLD_LOCAL);
+   if(!handle)
+      return mask;
 
    if (dlsym(handle, "st_api_OpenGL_ES1"))
       mask |= EGL_OPENGL_ES_BIT;
@@ -608,12 +816,20 @@ find_supported_apis(void)
 }
 
 
+static void
+xlib_Unload(_EGLDriver *drv)
+{
+   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
+   free(xdrv);
+}
+
+
 /**
  * This is the main entrypoint into the driver.
  * Called by libEGL to instantiate an _EGLDriver object.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *dpy, const char *args)
+_eglMain(const char *args)
 {
    struct xlib_egl_driver *xdrv;
 
@@ -623,10 +839,6 @@ _eglMain(_EGLDisplay *dpy, const char *args)
    if (!xdrv)
       return NULL;
 
-   if (!dpy->Xdpy) {
-      dpy->Xdpy = XOpenDisplay(NULL);
-   }
-
    _eglInitDriverFallbacks(&xdrv->Base);
    xdrv->Base.API.Initialize = xlib_eglInitialize;
    xdrv->Base.API.Terminate = xlib_eglTerminate;
@@ -634,27 +846,24 @@ _eglMain(_EGLDisplay *dpy, const char *args)
    xdrv->Base.API.CreateContext = xlib_eglCreateContext;
    xdrv->Base.API.DestroyContext = xlib_eglDestroyContext;
    xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface;
+   xdrv->Base.API.CreatePbufferSurface = xlib_eglCreatePbufferSurface;
    xdrv->Base.API.DestroySurface = xlib_eglDestroySurface;
+   xdrv->Base.API.BindTexImage = xlib_eglBindTexImage;
+   xdrv->Base.API.ReleaseTexImage = xlib_eglReleaseTexImage;
    xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent;
    xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers;
 
-   xdrv->Base.ClientAPIsMask = find_supported_apis();
-   if (xdrv->Base.ClientAPIsMask == 0x0) {
+   xdrv->apis = find_supported_apis();
+   if (xdrv->apis == 0x0) {
       /* the app isn't directly linked with any EGL-supprted APIs
        * (such as libGLESv2.so) so use an EGL utility to see what
        * APIs might be loaded dynamically on this system.
        */
-      xdrv->Base.ClientAPIsMask = _eglFindAPIs();
-   }      
+      xdrv->apis = _eglFindAPIs();
+   }
 
    xdrv->Base.Name = "Xlib/softpipe";
-
-   /* create one winsys and use it for all contexts/surfaces */
-   xdrv->winsys = create_sw_winsys();
-   xdrv->winsys->flush_frontbuffer = flush_frontbuffer;
-
-   xdrv->screen = softpipe_create_screen(xdrv->winsys);
+   xdrv->Base.Unload = xlib_Unload;
 
    return &xdrv->Base;
 }
-
diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c
index aa1bfa8e88..79ff2cc985 100644
--- a/src/gallium/winsys/egl_xlib/sw_winsys.c
+++ b/src/gallium/winsys/egl_xlib/sw_winsys.c
@@ -166,6 +166,7 @@ surface_buffer_create(struct pipe_winsys *winsys,
                       unsigned width, unsigned height,
                       enum pipe_format format, 
                       unsigned usage,
+                      unsigned tex_usage,
                       unsigned *stride)
 {
    const unsigned alignment = 64;
diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript
index aabab95f3a..86eb9ef55e 100644
--- a/src/gallium/winsys/gdi/SConscript
+++ b/src/gallium/winsys/gdi/SConscript
@@ -15,6 +15,7 @@ if env['platform'] == 'windows':
 		'gdi32',
 		'user32',
 		'kernel32',
+		'ws2_32',
 	])
 
 	sources = [
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
index 04309e67ee..3a1945d92c 100644
--- a/src/gallium/winsys/xlib/Makefile
+++ b/src/gallium/winsys/xlib/Makefile
@@ -34,6 +34,7 @@ XLIB_WINSYS_SOURCES = \
 	xlib_brw_aub.c \
 	xlib_brw_context.c \
 	xlib_brw_screen.c \
+	xlib_llvmpipe.c \
 	xlib_softpipe.c \
 	xlib_trace.c 
 
@@ -90,7 +91,7 @@ install: default
 	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
 	$(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL
 	@if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \
-		$(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \
+		$(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \
 	fi
 
 
diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript
index 0fb4b50f63..467d595d33 100644
--- a/src/gallium/winsys/xlib/SConscript
+++ b/src/gallium/winsys/xlib/SConscript
@@ -5,7 +5,7 @@ Import('*')
 
 if env['platform'] == 'linux' \
         and 'mesa' in env['statetrackers'] \
-        and ('softpipe' or 'i915simple' or 'trace') in env['drivers'] \
+        and set(('softpipe', 'llvmpipe', 'i915simple', 'trace')).intersection(env['drivers']) \
         and not env['dri']:
 
     env = env.Clone()
@@ -29,6 +29,14 @@ if env['platform'] == 'linux' \
         sources += ['xlib_softpipe.c']
         drivers += [softpipe]
 
+    if 'llvmpipe' in env['drivers']:
+        env.Tool('llvm')
+        if 'LLVM_VERSION' in env:
+            env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE')
+            env.Tool('udis86')
+            sources += ['xlib_llvmpipe.c']
+            drivers += [llvmpipe]
+
     if 'i965simple' in env['drivers']:
         env.Append(CPPDEFINES = 'GALLIUM_I965SIMPLE')
         sources += [
diff --git a/src/gallium/winsys/xlib/xlib.c b/src/gallium/winsys/xlib/xlib.c
index da72228215..4b71cf7ec3 100644
--- a/src/gallium/winsys/xlib/xlib.c
+++ b/src/gallium/winsys/xlib/xlib.c
@@ -45,6 +45,7 @@ enum mode {
    MODE_TRACE,
    MODE_BRW,
    MODE_CELL,
+   MODE_LLVMPIPE,
    MODE_SOFTPIPE
 };
 
@@ -62,7 +63,11 @@ static enum mode get_mode()
       return MODE_CELL;
 #endif
 
+#if defined(GALLIUM_LLVMPIPE)
+   return MODE_LLVMPIPE;
+#else
    return MODE_SOFTPIPE;
+#endif
 }
 
 static void _init( void ) __attribute__((constructor));
@@ -87,6 +92,11 @@ static void _init( void )
       xmesa_set_driver( &xlib_cell_driver );
 #endif
       break;
+   case MODE_LLVMPIPE:
+#if defined(GALLIUM_LLVMPIPE)
+      xmesa_set_driver( &xlib_llvmpipe_driver );
+#endif
+      break;
    case MODE_SOFTPIPE:
 #if defined(GALLIUM_SOFTPIPE)
       xmesa_set_driver( &xlib_softpipe_driver );
diff --git a/src/gallium/winsys/xlib/xlib.h b/src/gallium/winsys/xlib/xlib.h
index d602ab0b13..347d45f4d6 100644
--- a/src/gallium/winsys/xlib/xlib.h
+++ b/src/gallium/winsys/xlib/xlib.h
@@ -7,6 +7,7 @@
 
 extern struct xm_driver xlib_trace_driver;
 extern struct xm_driver xlib_softpipe_driver;
+extern struct xm_driver xlib_llvmpipe_driver;
 extern struct xm_driver xlib_cell_driver;
 extern struct xm_driver xlib_brw_driver;
 
diff --git a/src/gallium/winsys/xlib/xlib_brw_screen.c b/src/gallium/winsys/xlib/xlib_brw_screen.c
index a78efb10fa..ef545796f3 100644
--- a/src/gallium/winsys/xlib/xlib_brw_screen.c
+++ b/src/gallium/winsys/xlib/xlib_brw_screen.c
@@ -249,6 +249,7 @@ aub_i915_surface_buffer_create(struct pipe_winsys *winsys,
                                unsigned width, unsigned height,
                                enum pipe_format format,
                                unsigned usage,
+                               unsigned tex_usage,
                                unsigned *stride)
 {
    const unsigned alignment = 64;
diff --git a/src/gallium/winsys/xlib/xlib_llvmpipe.c b/src/gallium/winsys/xlib/xlib_llvmpipe.c
new file mode 100644
index 0000000000..3dd15e099b
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_llvmpipe.c
@@ -0,0 +1,464 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+
+#if defined(GALLIUM_LLVMPIPE)
+
+#include "xm_api.h"
+
+#undef ASSERT
+#undef Elements
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "llvmpipe/lp_winsys.h"
+#include "llvmpipe/lp_texture.h"
+
+#include "xlib.h"
+
+/**
+ * Subclass of pipe_buffer for Xlib winsys.
+ * Low-level OS/window system memory buffer
+ */
+struct xm_displaytarget
+{
+   enum pipe_format format;
+   struct pipe_format_block block;
+   unsigned width;
+   unsigned height;
+   unsigned stride;
+
+   void *data;
+   void *mapped;
+
+   XImage *tempImage;
+#ifdef USE_XSHM
+   int shm;
+   XShmSegmentInfo shminfo;
+#endif
+};
+
+
+/**
+ * Subclass of llvmpipe_winsys for Xlib winsys
+ */
+struct xmesa_llvmpipe_winsys
+{
+   struct llvmpipe_winsys base;
+/*   struct xmesa_visual *xm_visual; */
+};
+
+
+
+/** Cast wrapper */
+static INLINE struct xm_displaytarget *
+xm_displaytarget( struct llvmpipe_displaytarget *dt )
+{
+   return (struct xm_displaytarget *)dt;
+}
+
+
+/**
+ * X Shared Memory Image extension code
+ */
+
+#ifdef USE_XSHM
+
+static volatile int mesaXErrorFlag = 0;
+
+/**
+ * Catches potential Xlib errors.
+ */
+static int
+mesaHandleXError(Display *dpy, XErrorEvent *event)
+{
+   (void) dpy;
+   (void) event;
+   mesaXErrorFlag = 1;
+   return 0;
+}
+
+
+static char *alloc_shm(struct xm_displaytarget *buf, unsigned size)
+{
+   XShmSegmentInfo *const shminfo = & buf->shminfo;
+
+   shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777);
+   if (shminfo->shmid < 0) {
+      return NULL;
+   }
+
+   shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0);
+   if (shminfo->shmaddr == (char *) -1) {
+      shmctl(shminfo->shmid, IPC_RMID, 0);
+      return NULL;
+   }
+
+   shminfo->readOnly = False;
+   return shminfo->shmaddr;
+}
+
+
+/**
+ * Allocate a shared memory XImage back buffer for the given XMesaBuffer.
+ */
+static void
+alloc_shm_ximage(struct xm_displaytarget *xm_buffer,
+                 struct xmesa_buffer *xmb,
+                 unsigned width, unsigned height)
+{
+   /*
+    * We have to do a _lot_ of error checking here to be sure we can
+    * really use the XSHM extension.  It seems different servers trigger
+    * errors at different points if the extension won't work.  Therefore
+    * we have to be very careful...
+    */
+   int (*old_handler)(Display *, XErrorEvent *);
+
+   xm_buffer->tempImage = XShmCreateImage(xmb->xm_visual->display,
+                                  xmb->xm_visual->visinfo->visual,
+                                  xmb->xm_visual->visinfo->depth,
+                                  ZPixmap,
+                                  NULL,
+                                  &xm_buffer->shminfo,
+                                  width, height);
+   if (xm_buffer->tempImage == NULL) {
+      xm_buffer->shm = 0;
+      return;
+   }
+
+
+   mesaXErrorFlag = 0;
+   old_handler = XSetErrorHandler(mesaHandleXError);
+   /* This may trigger the X protocol error we're ready to catch: */
+   XShmAttach(xmb->xm_visual->display, &xm_buffer->shminfo);
+   XSync(xmb->xm_visual->display, False);
+
+   if (mesaXErrorFlag) {
+      /* we are on a remote display, this error is normal, don't print it */
+      XFlush(xmb->xm_visual->display);
+      mesaXErrorFlag = 0;
+      XDestroyImage(xm_buffer->tempImage);
+      xm_buffer->tempImage = NULL;
+      xm_buffer->shm = 0;
+      (void) XSetErrorHandler(old_handler);
+      return;
+   }
+
+   xm_buffer->shm = 1;
+}
+
+#endif /* USE_XSHM */
+
+static boolean
+xm_is_displaytarget_format_supported( struct llvmpipe_winsys *ws,
+                                      enum pipe_format format )
+{
+   /* TODO: check visuals or other sensible thing here */
+   return TRUE;
+}
+
+
+static void *
+xm_displaytarget_map(struct llvmpipe_winsys *ws,
+                     struct llvmpipe_displaytarget *dt,
+                     unsigned flags)
+{
+   struct xm_displaytarget *xm_dt = xm_displaytarget(dt);
+   xm_dt->mapped = xm_dt->data;
+   return xm_dt->mapped;
+}
+
+static void
+xm_displaytarget_unmap(struct llvmpipe_winsys *ws,
+                       struct llvmpipe_displaytarget *dt)
+{
+   struct xm_displaytarget *xm_dt = xm_displaytarget(dt);
+   xm_dt->mapped = NULL;
+}
+
+static void
+xm_displaytarget_destroy(struct llvmpipe_winsys *ws,
+                         struct llvmpipe_displaytarget *dt)
+{
+   struct xm_displaytarget *xm_dt = xm_displaytarget(dt);
+
+   if (xm_dt->data) {
+#ifdef USE_XSHM
+      if (xm_dt->shminfo.shmid >= 0) {
+         shmdt(xm_dt->shminfo.shmaddr);
+         shmctl(xm_dt->shminfo.shmid, IPC_RMID, 0);
+         
+         xm_dt->shminfo.shmid = -1;
+         xm_dt->shminfo.shmaddr = (char *) -1;
+      }
+      else
+#endif
+         FREE(xm_dt->data);
+   }
+
+   FREE(xm_dt);
+}
+
+
+/**
+ * Display/copy the image in the surface into the X window specified
+ * by the XMesaBuffer.
+ */
+static void
+xm_llvmpipe_display(struct xmesa_buffer *xm_buffer,
+                    struct llvmpipe_displaytarget *dt)
+{
+   XImage *ximage;
+   struct xm_displaytarget *xm_dt = xm_displaytarget(dt);
+   static boolean no_swap = 0;
+   static boolean firsttime = 1;
+
+   if (firsttime) {
+      no_swap = getenv("SP_NO_RAST") != NULL;
+      firsttime = 0;
+   }
+
+   if (no_swap)
+      return;
+
+#ifdef USE_XSHM
+   if (xm_dt->shm)
+   {
+      if (xm_dt->tempImage == NULL)
+      {
+         assert(xm_dt->block.width == 1);
+         assert(xm_dt->block.height == 1);
+         alloc_shm_ximage(xm_dt, xm_buffer,
+                          xm_dt->stride / xm_dt->block.size,
+                          xm_dt->height);
+      }
+
+      ximage = xm_dt->tempImage;
+      ximage->data = xm_dt->data;
+
+      /* _debug_printf("XSHM\n"); */
+      XShmPutImage(xm_buffer->xm_visual->display, xm_buffer->drawable, xm_buffer->gc,
+                   ximage, 0, 0, 0, 0, xm_dt->width, xm_dt->height, False);
+   }
+   else
+#endif
+   {
+      /* display image in Window */
+      ximage = xm_dt->tempImage;
+      ximage->data = xm_dt->data;
+
+      /* check that the XImage has been previously initialized */
+      assert(ximage->format);
+      assert(ximage->bitmap_unit);
+
+      /* update XImage's fields */
+      ximage->width = xm_dt->width;
+      ximage->height = xm_dt->height;
+      ximage->bytes_per_line = xm_dt->stride;
+
+      /* _debug_printf("XPUT\n"); */
+      XPutImage(xm_buffer->xm_visual->display, xm_buffer->drawable, xm_buffer->gc,
+                ximage, 0, 0, 0, 0, xm_dt->width, xm_dt->height);
+   }
+}
+
+/**
+ * Display/copy the image in the surface into the X window specified
+ * by the XMesaBuffer.
+ */
+static void
+xm_displaytarget_display(struct llvmpipe_winsys *ws,
+                         struct llvmpipe_displaytarget *dt,
+                         void *context_private)
+{
+   XMesaContext xmctx = (XMesaContext) context_private;
+   struct xmesa_buffer *xm_buffer = xmctx->xm_buffer;
+   xm_llvmpipe_display(xm_buffer, dt);
+}
+
+
+static struct llvmpipe_displaytarget *
+xm_displaytarget_create(struct llvmpipe_winsys *winsys,
+                        enum pipe_format format,
+                        unsigned width, unsigned height,
+                        unsigned alignment,
+                        unsigned *stride)
+{
+   struct xm_displaytarget *xm_dt = CALLOC_STRUCT(xm_displaytarget);
+   unsigned nblocksx, nblocksy, size;
+
+   xm_dt = CALLOC_STRUCT(xm_displaytarget);
+   if(!xm_dt)
+      goto no_xm_dt;
+
+   xm_dt->format = format;
+   xm_dt->width = width;
+   xm_dt->height = height;
+
+   pf_get_block(format, &xm_dt->block);
+   nblocksx = pf_get_nblocksx(&xm_dt->block, width);
+   nblocksy = pf_get_nblocksy(&xm_dt->block, height);
+   xm_dt->stride = align(nblocksx * xm_dt->block.size, alignment);
+   size = xm_dt->stride * nblocksy;
+
+#ifdef USE_XSHM
+   if (!debug_get_bool_option("XLIB_NO_SHM", FALSE))
+   {
+      xm_dt->shminfo.shmid = -1;
+      xm_dt->shminfo.shmaddr = (char *) -1;
+      xm_dt->shm = TRUE;
+         
+      xm_dt->data = alloc_shm(xm_dt, size);
+      if(!xm_dt->data)
+         goto no_data;
+   }
+#endif
+
+   if(!xm_dt->data) {
+      xm_dt->data = align_malloc(size, alignment);
+      if(!xm_dt->data)
+         goto no_data;
+   }
+
+   *stride = xm_dt->stride;
+   return (struct llvmpipe_displaytarget *)xm_dt;
+
+no_data:
+   FREE(xm_dt);
+no_xm_dt:
+   return NULL;
+}
+
+
+static void
+xm_destroy( struct llvmpipe_winsys *ws )
+{
+   FREE(ws);
+}
+
+
+static struct llvmpipe_winsys *
+xlib_create_llvmpipe_winsys( void )
+{
+   struct xmesa_llvmpipe_winsys *ws;
+
+   ws = CALLOC_STRUCT(xmesa_llvmpipe_winsys);
+   if (!ws)
+      return NULL;
+
+   ws->base.destroy = xm_destroy;
+
+   ws->base.is_displaytarget_format_supported = xm_is_displaytarget_format_supported;
+
+   ws->base.displaytarget_create = xm_displaytarget_create;
+   ws->base.displaytarget_map = xm_displaytarget_map;
+   ws->base.displaytarget_unmap = xm_displaytarget_unmap;
+   ws->base.displaytarget_destroy = xm_displaytarget_destroy;
+
+   ws->base.displaytarget_display = xm_displaytarget_display;
+
+   return &ws->base;
+}
+
+
+static struct pipe_screen *
+xlib_create_llvmpipe_screen( void )
+{
+   struct llvmpipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = xlib_create_llvmpipe_winsys();
+   if (winsys == NULL)
+      return NULL;
+
+   screen = llvmpipe_create_screen(winsys);
+   if (screen == NULL)
+      goto fail;
+
+   return screen;
+
+fail:
+   if (winsys)
+      winsys->destroy( winsys );
+
+   return NULL;
+}
+
+
+static struct pipe_context *
+xlib_create_llvmpipe_context( struct pipe_screen *screen,
+                              void *context_private )
+{
+   struct pipe_context *pipe;
+   
+   pipe = llvmpipe_create(screen);
+   if (pipe == NULL)
+      goto fail;
+
+   pipe->priv = context_private;
+   return pipe;
+
+fail:
+   /* Free stuff here */
+   return NULL;
+}
+
+
+static void
+xlib_llvmpipe_display_surface(struct xmesa_buffer *xm_buffer,
+                              struct pipe_surface *surf)
+{
+   struct llvmpipe_texture *texture = llvmpipe_texture(surf->texture);
+
+   assert(texture->dt);
+   if (texture->dt)
+      xm_llvmpipe_display(xm_buffer, texture->dt);
+}
+
+
+struct xm_driver xlib_llvmpipe_driver = 
+{
+   .create_pipe_screen = xlib_create_llvmpipe_screen,
+   .create_pipe_context = xlib_create_llvmpipe_context,
+   .display_surface = xlib_llvmpipe_display_surface
+};
+
+
+
+#endif /* GALLIUM_LLVMPIPE */
diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c
index 2a08b82908..67fea023a3 100644
--- a/src/gallium/winsys/xlib/xlib_softpipe.c
+++ b/src/gallium/winsys/xlib/xlib_softpipe.c
@@ -376,6 +376,7 @@ xm_surface_buffer_create(struct pipe_winsys *winsys,
                          unsigned width, unsigned height,
                          enum pipe_format format,
                          unsigned usage,
+                         unsigned tex_usage,
                          unsigned *stride)
 {
    const unsigned alignment = 64;
diff --git a/src/glu/Makefile b/src/glu/Makefile
index 5c26ead1bb..b268265976 100644
--- a/src/glu/Makefile
+++ b/src/glu/Makefile
@@ -30,7 +30,7 @@ glu.pc: glu.pc.in
 install: glu.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLU_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLU_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glu.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 clean:
diff --git a/src/glut/fbdev/Makefile b/src/glut/fbdev/Makefile
index 199d8c390a..c150ea88dc 100644
--- a/src/glut/fbdev/Makefile
+++ b/src/glut/fbdev/Makefile
@@ -72,7 +72,7 @@ install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_INC_DIR)/GL
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
 
 # Run 'make -f Makefile.solo dep' to update the dependencies if you change
 # what's included by any source file.
diff --git a/src/glut/glx/Makefile b/src/glut/glx/Makefile
index 1b072906c7..6889cd4b40 100644
--- a/src/glut/glx/Makefile
+++ b/src/glut/glx/Makefile
@@ -117,7 +117,7 @@ install: glut.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLUT_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLUT_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glut.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 
diff --git a/src/glut/glx/win32_winproc.c b/src/glut/glx/win32_winproc.c
index 4a9dc3781c..1b3a297828 100644
--- a/src/glut/glx/win32_winproc.c
+++ b/src/glut/glx/win32_winproc.c
@@ -548,8 +548,13 @@ __glutWindowProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam)
        window to be bigger than the screen, and smaller than 100x100
        (although it doesn't seem to help the y minimum). */
     minmax = (LPMINMAXINFO)lParam;
+#if 0
+    /* These two lines are disabled to fix incorrect handling of
+     * window maximization on Vista.  See bug 23182.
+     */
     minmax->ptMaxSize.x = __glutScreenWidth;
     minmax->ptMaxSize.y = __glutScreenHeight;
+#endif
     minmax->ptMinTrackSize.x = 0;
     minmax->ptMinTrackSize.y = 0;
     minmax->ptMaxTrackSize.x = __glutScreenWidth + 
diff --git a/src/glut/mini/Makefile b/src/glut/mini/Makefile
index 0e42436133..b82a758d2c 100644
--- a/src/glut/mini/Makefile
+++ b/src/glut/mini/Makefile
@@ -91,7 +91,7 @@ install: glut.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glut.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 
diff --git a/src/glw/Makefile b/src/glw/Makefile
index d88d773313..1fb3d3c320 100644
--- a/src/glw/Makefile
+++ b/src/glw/Makefile
@@ -43,7 +43,7 @@ install: glw.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 *.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLW_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLW_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glw.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 clean:
diff --git a/src/glx/x11/dri2.c b/src/glx/x11/dri2.c
index ebb2985924..e144ed3e1f 100644
--- a/src/glx/x11/dri2.c
+++ b/src/glx/x11/dri2.c
@@ -52,320 +52,328 @@
 static char dri2ExtensionName[] = DRI2_NAME;
 static XExtensionInfo *dri2Info;
 static XEXT_GENERATE_CLOSE_DISPLAY (DRI2CloseDisplay, dri2Info)
+
 static /* const */ XExtensionHooks dri2ExtensionHooks = {
-    NULL,				/* create_gc */
-    NULL,				/* copy_gc */
-    NULL,				/* flush_gc */
-    NULL,				/* free_gc */
-    NULL,				/* create_font */
-    NULL,				/* free_font */
-    DRI2CloseDisplay,			/* close_display */
-    NULL,				/* wire_to_event */
-    NULL,				/* event_to_wire */
-    NULL,				/* error */
-    NULL,				/* error_string */
+  NULL,                   /* create_gc */
+  NULL,                   /* copy_gc */
+  NULL,                   /* flush_gc */
+  NULL,                   /* free_gc */
+  NULL,                   /* create_font */
+  NULL,                   /* free_font */
+  DRI2CloseDisplay,       /* close_display */
+  NULL,                   /* wire_to_event */
+  NULL,                   /* event_to_wire */
+  NULL,                   /* error */
+  NULL,                   /* error_string */
 };
 
-static XEXT_GENERATE_FIND_DISPLAY (DRI2FindDisplay, dri2Info, 
-				   dri2ExtensionName, 
-				   &dri2ExtensionHooks, 
-				   0, NULL)
+static XEXT_GENERATE_FIND_DISPLAY (DRI2FindDisplay,
+                                   dri2Info,
+                                   dri2ExtensionName,
+                                   &dri2ExtensionHooks,
+                                   0, NULL)
 
-Bool DRI2QueryExtension(Display *dpy, int *eventBase, int *errorBase)
+Bool
+DRI2QueryExtension(Display * dpy, int *eventBase, int *errorBase)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
 
-    if (XextHasExtension(info)) {
-	*eventBase = info->codes->first_event;
-	*errorBase = info->codes->first_error;
-	return True;
-    }
+   if (XextHasExtension(info)) {
+      *eventBase = info->codes->first_event;
+      *errorBase = info->codes->first_error;
+      return True;
+   }
 
-    return False;
+   return False;
 }
 
-Bool DRI2QueryVersion(Display *dpy, int *major, int *minor)
+Bool
+DRI2QueryVersion(Display * dpy, int *major, int *minor)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay (dpy);
-    xDRI2QueryVersionReply rep;
-    xDRI2QueryVersionReq *req;
-
-    XextCheckExtension (dpy, info, dri2ExtensionName, False);
-
-    LockDisplay(dpy);
-    GetReq(DRI2QueryVersion, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2QueryVersion;
-    req->majorVersion = DRI2_MAJOR;
-    req->minorVersion = DRI2_MINOR;
-    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
-    *major = rep.majorVersion;
-    *minor = rep.minorVersion;
-    UnlockDisplay(dpy);
-    SyncHandle();
-
-    return True;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2QueryVersionReply rep;
+   xDRI2QueryVersionReq *req;
+
+   XextCheckExtension(dpy, info, dri2ExtensionName, False);
+
+   LockDisplay(dpy);
+   GetReq(DRI2QueryVersion, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2QueryVersion;
+   req->majorVersion = DRI2_MAJOR;
+   req->minorVersion = DRI2_MINOR;
+   if (!_XReply(dpy, (xReply *) & rep, 0, xFalse)) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
+   *major = rep.majorVersion;
+   *minor = rep.minorVersion;
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return True;
 }
 
-Bool DRI2Connect(Display *dpy, XID window,
-		 char **driverName, char **deviceName)
+Bool
+DRI2Connect(Display * dpy, XID window, char **driverName, char **deviceName)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2ConnectReply rep;
-    xDRI2ConnectReq *req;
-
-    XextCheckExtension (dpy, info, dri2ExtensionName, False);
-
-    LockDisplay(dpy);
-    GetReq(DRI2Connect, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2Connect;
-    req->window = window;
-    req->driverType = DRI2DriverDRI;
-    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
-
-    if (rep.driverNameLength == 0 && rep.deviceNameLength == 0) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
-
-    *driverName = Xmalloc(rep.driverNameLength + 1);
-    if (*driverName == NULL) {
-	_XEatData(dpy, 
-		  ((rep.driverNameLength + 3) & ~3) +
-		  ((rep.deviceNameLength + 3) & ~3));
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
-    _XReadPad(dpy, *driverName, rep.driverNameLength);
-    (*driverName)[rep.driverNameLength] = '\0';
-
-    *deviceName = Xmalloc(rep.deviceNameLength + 1);
-    if (*deviceName == NULL) {
-	Xfree(*driverName);
-	_XEatData(dpy, ((rep.deviceNameLength + 3) & ~3));
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
-    _XReadPad(dpy, *deviceName, rep.deviceNameLength);
-    (*deviceName)[rep.deviceNameLength] = '\0';
-
-    UnlockDisplay(dpy);
-    SyncHandle();
-
-    return True;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2ConnectReply rep;
+   xDRI2ConnectReq *req;
+
+   XextCheckExtension(dpy, info, dri2ExtensionName, False);
+
+   LockDisplay(dpy);
+   GetReq(DRI2Connect, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2Connect;
+   req->window = window;
+   req->driverType = DRI2DriverDRI;
+   if (!_XReply(dpy, (xReply *) & rep, 0, xFalse)) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
+
+   if (rep.driverNameLength == 0 && rep.deviceNameLength == 0) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
+
+   *driverName = Xmalloc(rep.driverNameLength + 1);
+   if (*driverName == NULL) {
+      _XEatData(dpy,
+                ((rep.driverNameLength + 3) & ~3) +
+                ((rep.deviceNameLength + 3) & ~3));
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
+   _XReadPad(dpy, *driverName, rep.driverNameLength);
+   (*driverName)[rep.driverNameLength] = '\0';
+
+   *deviceName = Xmalloc(rep.deviceNameLength + 1);
+   if (*deviceName == NULL) {
+      Xfree(*driverName);
+      _XEatData(dpy, ((rep.deviceNameLength + 3) & ~3));
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
+   _XReadPad(dpy, *deviceName, rep.deviceNameLength);
+   (*deviceName)[rep.deviceNameLength] = '\0';
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return True;
 }
 
-Bool DRI2Authenticate(Display *dpy, XID window, drm_magic_t magic)
+Bool
+DRI2Authenticate(Display * dpy, XID window, drm_magic_t magic)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2AuthenticateReq *req;
-    xDRI2AuthenticateReply rep;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2AuthenticateReq *req;
+   xDRI2AuthenticateReply rep;
 
-    XextCheckExtension (dpy, info, dri2ExtensionName, False);
+   XextCheckExtension(dpy, info, dri2ExtensionName, False);
 
-    LockDisplay(dpy);
-    GetReq(DRI2Authenticate, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2Authenticate;
-    req->window = window;
-    req->magic = magic;
+   LockDisplay(dpy);
+   GetReq(DRI2Authenticate, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2Authenticate;
+   req->window = window;
+   req->magic = magic;
 
-    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return False;
-    }
+   if (!_XReply(dpy, (xReply *) & rep, 0, xFalse)) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return False;
+   }
 
-    UnlockDisplay(dpy);
-    SyncHandle();
+   UnlockDisplay(dpy);
+   SyncHandle();
 
-    return rep.authenticated;
+   return rep.authenticated;
 }
 
-void DRI2CreateDrawable(Display *dpy, XID drawable)
+void
+DRI2CreateDrawable(Display * dpy, XID drawable)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2CreateDrawableReq *req;
-
-    XextSimpleCheckExtension (dpy, info, dri2ExtensionName);
-
-    LockDisplay(dpy);
-    GetReq(DRI2CreateDrawable, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2CreateDrawable;
-    req->drawable = drawable;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2CreateDrawableReq *req;
+
+   XextSimpleCheckExtension(dpy, info, dri2ExtensionName);
+
+   LockDisplay(dpy);
+   GetReq(DRI2CreateDrawable, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2CreateDrawable;
+   req->drawable = drawable;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
-void DRI2DestroyDrawable(Display *dpy, XID drawable)
+void
+DRI2DestroyDrawable(Display * dpy, XID drawable)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2DestroyDrawableReq *req;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2DestroyDrawableReq *req;
 
-    XextSimpleCheckExtension (dpy, info, dri2ExtensionName);
+   XextSimpleCheckExtension(dpy, info, dri2ExtensionName);
 
-    XSync(dpy, False);
+   XSync(dpy, False);
 
-    LockDisplay(dpy);
-    GetReq(DRI2DestroyDrawable, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2DestroyDrawable;
-    req->drawable = drawable;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   LockDisplay(dpy);
+   GetReq(DRI2DestroyDrawable, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2DestroyDrawable;
+   req->drawable = drawable;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
-DRI2Buffer *DRI2GetBuffers(Display *dpy, XID drawable,
-			   int *width, int *height,
-			   unsigned int *attachments, int count,
-			   int *outCount)
+DRI2Buffer *
+DRI2GetBuffers(Display * dpy, XID drawable,
+               int *width, int *height,
+               unsigned int *attachments, int count, int *outCount)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2GetBuffersReply rep;
-    xDRI2GetBuffersReq *req;
-    DRI2Buffer *buffers;
-    xDRI2Buffer repBuffer;
-    CARD32 *p;
-    int i;
-
-    XextCheckExtension (dpy, info, dri2ExtensionName, False);
-
-    LockDisplay(dpy);
-    GetReqExtra(DRI2GetBuffers, count * 4, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2GetBuffers;
-    req->drawable = drawable;
-    req->count = count;
-    p = (CARD32 *) &req[1];
-    for (i = 0; i < count; i++)
-	p[i] = attachments[i];
-
-    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return NULL;
-    }
-
-    *width = rep.width;
-    *height = rep.height;
-    *outCount = rep.count;
-
-    buffers = Xmalloc(rep.count * sizeof buffers[0]);
-    if (buffers == NULL) {
-	_XEatData(dpy, rep.count * sizeof repBuffer);
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return NULL;
-    }
-
-    for (i = 0; i < rep.count; i++) {
-	_XReadPad(dpy, (char *) &repBuffer, sizeof repBuffer);
-	buffers[i].attachment = repBuffer.attachment;
-	buffers[i].name = repBuffer.name;
-	buffers[i].pitch = repBuffer.pitch;
-	buffers[i].cpp = repBuffer.cpp;
-	buffers[i].flags = repBuffer.flags;
-    }
-
-    UnlockDisplay(dpy);
-    SyncHandle();
-
-    return buffers;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2GetBuffersReply rep;
+   xDRI2GetBuffersReq *req;
+   DRI2Buffer *buffers;
+   xDRI2Buffer repBuffer;
+   CARD32 *p;
+   int i;
+
+   XextCheckExtension(dpy, info, dri2ExtensionName, False);
+
+   LockDisplay(dpy);
+   GetReqExtra(DRI2GetBuffers, count * 4, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2GetBuffers;
+   req->drawable = drawable;
+   req->count = count;
+   p = (CARD32 *) & req[1];
+   for (i = 0; i < count; i++)
+      p[i] = attachments[i];
+
+   if (!_XReply(dpy, (xReply *) & rep, 0, xFalse)) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return NULL;
+   }
+
+   *width = rep.width;
+   *height = rep.height;
+   *outCount = rep.count;
+
+   buffers = Xmalloc(rep.count * sizeof buffers[0]);
+   if (buffers == NULL) {
+      _XEatData(dpy, rep.count * sizeof repBuffer);
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return NULL;
+   }
+
+   for (i = 0; i < rep.count; i++) {
+      _XReadPad(dpy, (char *) &repBuffer, sizeof repBuffer);
+      buffers[i].attachment = repBuffer.attachment;
+      buffers[i].name = repBuffer.name;
+      buffers[i].pitch = repBuffer.pitch;
+      buffers[i].cpp = repBuffer.cpp;
+      buffers[i].flags = repBuffer.flags;
+   }
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return buffers;
 }
 
 
-DRI2Buffer *DRI2GetBuffersWithFormat(Display *dpy, XID drawable,
-				     int *width, int *height,
-				     unsigned int *attachments, int count,
-				     int *outCount)
+DRI2Buffer *
+DRI2GetBuffersWithFormat(Display * dpy, XID drawable,
+                         int *width, int *height,
+                         unsigned int *attachments, int count, int *outCount)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2GetBuffersReply rep;
-    xDRI2GetBuffersReq *req;
-    DRI2Buffer *buffers;
-    xDRI2Buffer repBuffer;
-    CARD32 *p;
-    int i;
-
-    XextCheckExtension (dpy, info, dri2ExtensionName, False);
-
-    LockDisplay(dpy);
-    GetReqExtra(DRI2GetBuffers, count * (4 * 2), req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2GetBuffersWithFormat;
-    req->drawable = drawable;
-    req->count = count;
-    p = (CARD32 *) &req[1];
-    for (i = 0; i < (count * 2); i++)
-	p[i] = attachments[i];
-
-    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return NULL;
-    }
-
-    *width = rep.width;
-    *height = rep.height;
-    *outCount = rep.count;
-
-    buffers = Xmalloc(rep.count * sizeof buffers[0]);
-    if (buffers == NULL) {
-	_XEatData(dpy, rep.count * sizeof repBuffer);
-	UnlockDisplay(dpy);
-	SyncHandle();
-	return NULL;
-    }
-
-    for (i = 0; i < rep.count; i++) {
-	_XReadPad(dpy, (char *) &repBuffer, sizeof repBuffer);
-	buffers[i].attachment = repBuffer.attachment;
-	buffers[i].name = repBuffer.name;
-	buffers[i].pitch = repBuffer.pitch;
-	buffers[i].cpp = repBuffer.cpp;
-	buffers[i].flags = repBuffer.flags;
-    }
-
-    UnlockDisplay(dpy);
-    SyncHandle();
-
-    return buffers;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2GetBuffersReply rep;
+   xDRI2GetBuffersReq *req;
+   DRI2Buffer *buffers;
+   xDRI2Buffer repBuffer;
+   CARD32 *p;
+   int i;
+
+   XextCheckExtension(dpy, info, dri2ExtensionName, False);
+
+   LockDisplay(dpy);
+   GetReqExtra(DRI2GetBuffers, count * (4 * 2), req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2GetBuffersWithFormat;
+   req->drawable = drawable;
+   req->count = count;
+   p = (CARD32 *) & req[1];
+   for (i = 0; i < (count * 2); i++)
+      p[i] = attachments[i];
+
+   if (!_XReply(dpy, (xReply *) & rep, 0, xFalse)) {
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return NULL;
+   }
+
+   *width = rep.width;
+   *height = rep.height;
+   *outCount = rep.count;
+
+   buffers = Xmalloc(rep.count * sizeof buffers[0]);
+   if (buffers == NULL) {
+      _XEatData(dpy, rep.count * sizeof repBuffer);
+      UnlockDisplay(dpy);
+      SyncHandle();
+      return NULL;
+   }
+
+   for (i = 0; i < rep.count; i++) {
+      _XReadPad(dpy, (char *) &repBuffer, sizeof repBuffer);
+      buffers[i].attachment = repBuffer.attachment;
+      buffers[i].name = repBuffer.name;
+      buffers[i].pitch = repBuffer.pitch;
+      buffers[i].cpp = repBuffer.cpp;
+      buffers[i].flags = repBuffer.flags;
+   }
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return buffers;
 }
 
 
-void DRI2CopyRegion(Display *dpy, XID drawable, XserverRegion region,
-		    CARD32 dest, CARD32 src)
+void
+DRI2CopyRegion(Display * dpy, XID drawable, XserverRegion region,
+               CARD32 dest, CARD32 src)
 {
-    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-    xDRI2CopyRegionReq *req;
-    xDRI2CopyRegionReply rep;
+   XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+   xDRI2CopyRegionReq *req;
+   xDRI2CopyRegionReply rep;
 
-    XextSimpleCheckExtension (dpy, info, dri2ExtensionName);
+   XextSimpleCheckExtension(dpy, info, dri2ExtensionName);
 
-    LockDisplay(dpy);
-    GetReq(DRI2CopyRegion, req);
-    req->reqType = info->codes->major_opcode;
-    req->dri2ReqType = X_DRI2CopyRegion;
-    req->drawable = drawable;
-    req->region = region;
-    req->dest = dest;
-    req->src = src;
+   LockDisplay(dpy);
+   GetReq(DRI2CopyRegion, req);
+   req->reqType = info->codes->major_opcode;
+   req->dri2ReqType = X_DRI2CopyRegion;
+   req->drawable = drawable;
+   req->region = region;
+   req->dest = dest;
+   req->src = src;
 
-    _XReply(dpy, (xReply *)&rep, 0, xFalse);
+   _XReply(dpy, (xReply *) & rep, 0, xFalse);
 
-    UnlockDisplay(dpy);
-    SyncHandle();
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
diff --git a/src/glx/x11/dri2.h b/src/glx/x11/dri2.h
index b0e61f80d7..a6fe66e136 100644
--- a/src/glx/x11/dri2.h
+++ b/src/glx/x11/dri2.h
@@ -36,45 +36,53 @@
 #include <X11/extensions/Xfixes.h>
 #include <X11/extensions/dri2tokens.h>
 
-typedef struct {
-    unsigned int attachment;
-    unsigned int name;
-    unsigned int pitch;
-    unsigned int cpp;
-    unsigned int flags;
+typedef struct
+{
+   unsigned int attachment;
+   unsigned int name;
+   unsigned int pitch;
+   unsigned int cpp;
+   unsigned int flags;
 } DRI2Buffer;
 
 extern Bool
-DRI2QueryExtension(Display *display, int *eventBase, int *errorBase);
+DRI2QueryExtension(Display * display, int *eventBase, int *errorBase);
+
 extern Bool
-DRI2QueryVersion(Display *display, int *major, int *minor);
+DRI2QueryVersion(Display * display, int *major, int *minor);
+
 extern Bool
-DRI2Connect(Display *display, XID window,
-	    char **driverName, char **deviceName);
+DRI2Connect(Display * display, XID window,
+            char **driverName, char **deviceName);
+
 extern Bool
-DRI2Authenticate(Display *display, XID window, drm_magic_t magic);
+DRI2Authenticate(Display * display, XID window, drm_magic_t magic);
+
 extern void
-DRI2CreateDrawable(Display *display, XID drawable);
+DRI2CreateDrawable(Display * display, XID drawable);
+
 extern void
-DRI2DestroyDrawable(Display *display, XID handle);
-extern DRI2Buffer *
-DRI2GetBuffers(Display *dpy, XID drawable,
-	       int *width, int *height,
-	       unsigned int *attachments, int count,
-	       int *outCount);
+DRI2DestroyDrawable(Display * display, XID handle);
+
+extern DRI2Buffer*
+DRI2GetBuffers(Display * dpy, XID drawable,
+               int *width, int *height,
+               unsigned int *attachments, int count,
+               int *outCount);
 
 /**
  * \note
  * This function is only supported with DRI2 version 1.1 or later.
  */
-extern DRI2Buffer *
-DRI2GetBuffersWithFormat(Display *dpy, XID drawable,
-			 int *width, int *height,
-			 unsigned int *attachments, int count,
-			 int *outCount);
+extern DRI2Buffer*
+DRI2GetBuffersWithFormat(Display * dpy, XID drawable,
+                         int *width, int *height,
+                         unsigned int *attachments,
+                         int count, int *outCount);
 
 extern void
-DRI2CopyRegion(Display *dpy, XID drawable, XserverRegion region,
-	       CARD32 dest, CARD32 src);
+DRI2CopyRegion(Display * dpy, XID drawable,
+               XserverRegion region,
+               CARD32 dest, CARD32 src);
 
 #endif
diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c
index c5635a94ab..89efe3ab29 100644
--- a/src/glx/x11/dri2_glx.c
+++ b/src/glx/x11/dri2_glx.c
@@ -54,242 +54,260 @@ typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate;
 typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate;
 typedef struct __GLXDRIdrawablePrivateRec __GLXDRIdrawablePrivate;
 
-struct __GLXDRIdisplayPrivateRec {
-    __GLXDRIdisplay base;
+struct __GLXDRIdisplayPrivateRec
+{
+   __GLXDRIdisplay base;
 
-    /*
+   /*
     ** XFree86-DRI version information
     */
-    int driMajor;
-    int driMinor;
-    int driPatch;
+   int driMajor;
+   int driMinor;
+   int driPatch;
 };
 
-struct __GLXDRIcontextPrivateRec {
-    __GLXDRIcontext base;
-    __DRIcontext *driContext;
-    __GLXscreenConfigs *psc;
+struct __GLXDRIcontextPrivateRec
+{
+   __GLXDRIcontext base;
+   __DRIcontext *driContext;
+   __GLXscreenConfigs *psc;
 };
 
-struct __GLXDRIdrawablePrivateRec {
-    __GLXDRIdrawable base;
-    __DRIbuffer buffers[5];
-    int bufferCount;
-    int width, height;
-    int have_back;
-    int have_fake_front;
+struct __GLXDRIdrawablePrivateRec
+{
+   __GLXDRIdrawable base;
+   __DRIbuffer buffers[5];
+   int bufferCount;
+   int width, height;
+   int have_back;
+   int have_fake_front;
 };
 
-static void dri2DestroyContext(__GLXDRIcontext *context,
-			      __GLXscreenConfigs *psc, Display *dpy)
+static void dri2WaitX(__GLXDRIdrawable * pdraw);
+
+static void
+dri2DestroyContext(__GLXDRIcontext * context,
+                   __GLXscreenConfigs * psc, Display * dpy)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-    const __DRIcoreExtension *core = pcp->psc->core;
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
+   const __DRIcoreExtension *core = pcp->psc->core;
 
-    (*core->destroyContext)(pcp->driContext);
+   (*core->destroyContext) (pcp->driContext);
 
-    Xfree(pcp);
+   Xfree(pcp);
 }
 
-static Bool dri2BindContext(__GLXDRIcontext *context,
-			   __GLXDRIdrawable *draw, __GLXDRIdrawable *read)
+static Bool
+dri2BindContext(__GLXDRIcontext * context,
+                __GLXDRIdrawable * draw, __GLXDRIdrawable * read)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-    const __DRIcoreExtension *core = pcp->psc->core;
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
+   const __DRIcoreExtension *core = pcp->psc->core;
 
-    return (*core->bindContext)(pcp->driContext,
-				draw->driDrawable,
-				read->driDrawable);
+   return (*core->bindContext) (pcp->driContext,
+                                draw->driDrawable, read->driDrawable);
 }
 
-static void dri2UnbindContext(__GLXDRIcontext *context)
+static void
+dri2UnbindContext(__GLXDRIcontext * context)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-    const __DRIcoreExtension *core = pcp->psc->core;
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
+   const __DRIcoreExtension *core = pcp->psc->core;
 
-    (*core->unbindContext)(pcp->driContext);
+   (*core->unbindContext) (pcp->driContext);
 }
 
-static __GLXDRIcontext *dri2CreateContext(__GLXscreenConfigs *psc,
-					 const __GLcontextModes *mode,
-					 GLXContext gc,
-					 GLXContext shareList, int renderType)
+static __GLXDRIcontext *
+dri2CreateContext(__GLXscreenConfigs * psc,
+                  const __GLcontextModes * mode,
+                  GLXContext gc, GLXContext shareList, int renderType)
 {
-    __GLXDRIcontextPrivate *pcp, *pcp_shared;
-    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
-    __DRIcontext *shared = NULL;
-
-    if (shareList) {
-	pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext;
-	shared = pcp_shared->driContext;
-    }
-
-    pcp = Xmalloc(sizeof *pcp);
-    if (pcp == NULL)
-	return NULL;
-
-    pcp->psc = psc;
-    pcp->driContext = 
-	(*psc->dri2->createNewContext)(psc->__driScreen,
-				       config->driConfig, shared, pcp);
-    gc->__driContext = pcp->driContext;
-
-    if (pcp->driContext == NULL) {
-	Xfree(pcp);
-	return NULL;
-    }
-
-    pcp->base.destroyContext = dri2DestroyContext;
-    pcp->base.bindContext = dri2BindContext;
-    pcp->base.unbindContext = dri2UnbindContext;
-
-    return &pcp->base;
+   __GLXDRIcontextPrivate *pcp, *pcp_shared;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
+   __DRIcontext *shared = NULL;
+
+   if (shareList) {
+      pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext;
+      shared = pcp_shared->driContext;
+   }
+
+   pcp = Xmalloc(sizeof *pcp);
+   if (pcp == NULL)
+      return NULL;
+
+   pcp->psc = psc;
+   pcp->driContext =
+      (*psc->dri2->createNewContext) (psc->__driScreen,
+                                      config->driConfig, shared, pcp);
+   gc->__driContext = pcp->driContext;
+
+   if (pcp->driContext == NULL) {
+      Xfree(pcp);
+      return NULL;
+   }
+
+   pcp->base.destroyContext = dri2DestroyContext;
+   pcp->base.bindContext = dri2BindContext;
+   pcp->base.unbindContext = dri2UnbindContext;
+
+   return &pcp->base;
 }
 
-static void dri2DestroyDrawable(__GLXDRIdrawable *pdraw)
+static void
+dri2DestroyDrawable(__GLXDRIdrawable * pdraw)
 {
-    const __DRIcoreExtension *core = pdraw->psc->core;
+   const __DRIcoreExtension *core = pdraw->psc->core;
 
-    (*core->destroyDrawable)(pdraw->driDrawable);
-    DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->drawable);
-    Xfree(pdraw);
+   (*core->destroyDrawable) (pdraw->driDrawable);
+   DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->xDrawable);
+   Xfree(pdraw);
 }
 
-static __GLXDRIdrawable *dri2CreateDrawable(__GLXscreenConfigs *psc,
-					    XID xDrawable,
-					    GLXDrawable drawable,
-					    const __GLcontextModes *modes)
+static __GLXDRIdrawable *
+dri2CreateDrawable(__GLXscreenConfigs * psc,
+                   XID xDrawable,
+                   GLXDrawable drawable, const __GLcontextModes * modes)
 {
-    __GLXDRIdrawablePrivate *pdraw;
-    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
+   __GLXDRIdrawablePrivate *pdraw;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
 
-    pdraw = Xmalloc(sizeof(*pdraw));
-    if (!pdraw)
-	return NULL;
+   pdraw = Xmalloc(sizeof(*pdraw));
+   if (!pdraw)
+      return NULL;
 
-    pdraw->base.destroyDrawable = dri2DestroyDrawable;
-    pdraw->base.xDrawable = xDrawable;
-    pdraw->base.drawable = drawable;
-    pdraw->base.psc = psc;
-    pdraw->bufferCount = 0;
+   pdraw->base.destroyDrawable = dri2DestroyDrawable;
+   pdraw->base.xDrawable = xDrawable;
+   pdraw->base.drawable = drawable;
+   pdraw->base.psc = psc;
+   pdraw->bufferCount = 0;
 
-    DRI2CreateDrawable(psc->dpy, xDrawable);
+   DRI2CreateDrawable(psc->dpy, xDrawable);
 
-    /* Create a new drawable */
-    pdraw->base.driDrawable =
-	(*psc->dri2->createNewDrawable)(psc->__driScreen,
-					config->driConfig, pdraw);
+   /* Create a new drawable */
+   pdraw->base.driDrawable =
+      (*psc->dri2->createNewDrawable) (psc->__driScreen,
+                                       config->driConfig, pdraw);
 
-    if (!pdraw->base.driDrawable) {
-	DRI2DestroyDrawable(psc->dpy, drawable);
-	Xfree(pdraw);
-	return NULL;
-    }
+   if (!pdraw->base.driDrawable) {
+      DRI2DestroyDrawable(psc->dpy, xDrawable);
+      Xfree(pdraw);
+      return NULL;
+   }
 
-    return &pdraw->base;
+   return &pdraw->base;
 }
 
-static void dri2CopySubBuffer(__GLXDRIdrawable *pdraw,
-			      int x, int y, int width, int height)
+static void
+dri2CopySubBuffer(__GLXDRIdrawable * pdraw,
+                  int x, int y, int width, int height)
 {
-    __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
-    XRectangle xrect;
-    XserverRegion region;
+   __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
+   XRectangle xrect;
+   XserverRegion region;
 
-    /* Check we have the right attachments */
-    if (!priv->have_back)
-    	return;
+   /* Check we have the right attachments */
+   if (!priv->have_back)
+      return;
 
-    xrect.x = x;
-    xrect.y = priv->height - y - height;
-    xrect.width = width;
-    xrect.height = height;
+   xrect.x = x;
+   xrect.y = priv->height - y - height;
+   xrect.width = width;
+   xrect.height = height;
 
 #ifdef __DRI2_FLUSH
-    if (pdraw->psc->f)
-    	(*pdraw->psc->f->flush)(pdraw->driDrawable);
+   if (pdraw->psc->f)
+      (*pdraw->psc->f->flush) (pdraw->driDrawable);
 #endif
 
-    region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
-    /* should get a fence ID back from here at some point */
-    DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region,
-		   DRI2BufferFrontLeft, DRI2BufferBackLeft);
-    XFixesDestroyRegion(pdraw->psc->dpy, region);
+   region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
+   /* should get a fence ID back from here at some point */
+   DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region,
+                  DRI2BufferFrontLeft, DRI2BufferBackLeft);
+   XFixesDestroyRegion(pdraw->psc->dpy, region);
+
+   /* Refresh the fake front (if present) after we just damaged the real
+    * front.
+    */
+   dri2WaitX(pdraw);
 }
 
-static void dri2SwapBuffers(__GLXDRIdrawable *pdraw)
+static void
+dri2SwapBuffers(__GLXDRIdrawable * pdraw)
 {
-    __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
+   __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
 
-    dri2CopySubBuffer(pdraw, 0, 0, priv->width, priv->height);
+   dri2CopySubBuffer(pdraw, 0, 0, priv->width, priv->height);
 }
 
-static void dri2WaitX(__GLXDRIdrawable *pdraw)
+static void
+dri2WaitX(__GLXDRIdrawable * pdraw)
 {
-    __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
-    XRectangle xrect;
-    XserverRegion region;
+   __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
+   XRectangle xrect;
+   XserverRegion region;
 
-    /* Check we have the right attachments */
-    if (!priv->have_fake_front)
-    	return;
+   /* Check we have the right attachments */
+   if (!priv->have_fake_front)
+      return;
 
-    xrect.x = 0;
-    xrect.y = 0;
-    xrect.width = priv->width;
-    xrect.height = priv->height;
+   xrect.x = 0;
+   xrect.y = 0;
+   xrect.width = priv->width;
+   xrect.height = priv->height;
 
 #ifdef __DRI2_FLUSH
-    if (pdraw->psc->f)
-    	(*pdraw->psc->f->flush)(pdraw->driDrawable);
+   if (pdraw->psc->f)
+      (*pdraw->psc->f->flush) (pdraw->driDrawable);
 #endif
 
-    region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
-    DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region,
-		   DRI2BufferFakeFrontLeft, DRI2BufferFrontLeft);
-    XFixesDestroyRegion(pdraw->psc->dpy, region);
+   region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
+   DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region,
+                  DRI2BufferFakeFrontLeft, DRI2BufferFrontLeft);
+   XFixesDestroyRegion(pdraw->psc->dpy, region);
 }
 
-static void dri2WaitGL(__GLXDRIdrawable *pdraw)
+static void
+dri2WaitGL(__GLXDRIdrawable * pdraw)
 {
-    __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
-    XRectangle xrect;
-    XserverRegion region;
+   __GLXDRIdrawablePrivate *priv = (__GLXDRIdrawablePrivate *) pdraw;
+   XRectangle xrect;
+   XserverRegion region;
 
-    if (!priv->have_fake_front)
-    	return;
+   if (!priv->have_fake_front)
+      return;
 
-    xrect.x = 0;
-    xrect.y = 0;
-    xrect.width = priv->width;
-    xrect.height = priv->height;
+   xrect.x = 0;
+   xrect.y = 0;
+   xrect.width = priv->width;
+   xrect.height = priv->height;
 
 #ifdef __DRI2_FLUSH
-    if (pdraw->psc->f)
-    	(*pdraw->psc->f->flush)(pdraw->driDrawable);
+   if (pdraw->psc->f)
+      (*pdraw->psc->f->flush) (pdraw->driDrawable);
 #endif
 
-    region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
-    DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region,
-		   DRI2BufferFrontLeft, DRI2BufferFakeFrontLeft);
-    XFixesDestroyRegion(pdraw->psc->dpy, region);
+   region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1);
+   DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region,
+                  DRI2BufferFrontLeft, DRI2BufferFakeFrontLeft);
+   XFixesDestroyRegion(pdraw->psc->dpy, region);
 }
 
 
-static void dri2FlushFrontBuffer(__DRIdrawable *driDrawable,
-				 void *loaderPrivate)
+static void
+dri2FlushFrontBuffer(__DRIdrawable * driDrawable, void *loaderPrivate)
 {
-    (void) driDrawable;
-    dri2WaitGL((__GLXDRIdrawable *) loaderPrivate);
+   (void) driDrawable;
+   dri2WaitGL((__GLXDRIdrawable *) loaderPrivate);
 }
 
 
-static void dri2DestroyScreen(__GLXscreenConfigs *psc)
+static void
+dri2DestroyScreen(__GLXscreenConfigs * psc)
 {
-    /* Free the direct rendering per screen data */
-    (*psc->core->destroyScreen)(psc->__driScreen);
-    close(psc->fd);
-    psc->__driScreen = NULL;
+   /* Free the direct rendering per screen data */
+   (*psc->core->destroyScreen) (psc->__driScreen);
+   close(psc->fd);
+   psc->__driScreen = NULL;
 }
 
 /**
@@ -299,221 +317,222 @@ static void dri2DestroyScreen(__GLXscreenConfigs *psc)
  * \c DRI2GetBuffers or \c DRI2GetBuffersWithFormat.
  */
 static void
-process_buffers(__GLXDRIdrawablePrivate *pdraw, DRI2Buffer *buffers,
-		unsigned count)
+process_buffers(__GLXDRIdrawablePrivate * pdraw, DRI2Buffer * buffers,
+                unsigned count)
 {
-    int i;
-
-    pdraw->bufferCount = count;
-    pdraw->have_fake_front = 0;
-    pdraw->have_back = 0;
-
-    /* This assumes the DRI2 buffer attachment tokens matches the
-     * __DRIbuffer tokens. */
-    for (i = 0; i < count; i++) {
-	pdraw->buffers[i].attachment = buffers[i].attachment;
-	pdraw->buffers[i].name = buffers[i].name;
-	pdraw->buffers[i].pitch = buffers[i].pitch;
-	pdraw->buffers[i].cpp = buffers[i].cpp;
-	pdraw->buffers[i].flags = buffers[i].flags;
-	if (pdraw->buffers[i].attachment == __DRI_BUFFER_FAKE_FRONT_LEFT)
-	    pdraw->have_fake_front = 1;
-	if (pdraw->buffers[i].attachment == __DRI_BUFFER_BACK_LEFT)
-	    pdraw->have_back = 1;
-    }
+   int i;
+
+   pdraw->bufferCount = count;
+   pdraw->have_fake_front = 0;
+   pdraw->have_back = 0;
+
+   /* This assumes the DRI2 buffer attachment tokens matches the
+    * __DRIbuffer tokens. */
+   for (i = 0; i < count; i++) {
+      pdraw->buffers[i].attachment = buffers[i].attachment;
+      pdraw->buffers[i].name = buffers[i].name;
+      pdraw->buffers[i].pitch = buffers[i].pitch;
+      pdraw->buffers[i].cpp = buffers[i].cpp;
+      pdraw->buffers[i].flags = buffers[i].flags;
+      if (pdraw->buffers[i].attachment == __DRI_BUFFER_FAKE_FRONT_LEFT)
+         pdraw->have_fake_front = 1;
+      if (pdraw->buffers[i].attachment == __DRI_BUFFER_BACK_LEFT)
+         pdraw->have_back = 1;
+   }
 
 }
 
 static __DRIbuffer *
-dri2GetBuffers(__DRIdrawable *driDrawable,
-	       int *width, int *height,
-	       unsigned int *attachments, int count,
-	       int *out_count, void *loaderPrivate)
+dri2GetBuffers(__DRIdrawable * driDrawable,
+               int *width, int *height,
+               unsigned int *attachments, int count,
+               int *out_count, void *loaderPrivate)
 {
-    __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
-    DRI2Buffer *buffers;
+   __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
+   DRI2Buffer *buffers;
 
-    buffers = DRI2GetBuffers(pdraw->base.psc->dpy, pdraw->base.xDrawable,
-			     width, height, attachments, count, out_count);
-    if (buffers == NULL)
-       return NULL;
+   buffers = DRI2GetBuffers(pdraw->base.psc->dpy, pdraw->base.xDrawable,
+                            width, height, attachments, count, out_count);
+   if (buffers == NULL)
+      return NULL;
 
-    pdraw->width = *width;
-    pdraw->height = *height;
-    process_buffers(pdraw, buffers, *out_count);
+   pdraw->width = *width;
+   pdraw->height = *height;
+   process_buffers(pdraw, buffers, *out_count);
 
-    Xfree(buffers);
+   Xfree(buffers);
 
-    return pdraw->buffers;
+   return pdraw->buffers;
 }
 
 static __DRIbuffer *
-dri2GetBuffersWithFormat(__DRIdrawable *driDrawable,
-			 int *width, int *height,
-			 unsigned int *attachments, int count,
-			 int *out_count, void *loaderPrivate)
+dri2GetBuffersWithFormat(__DRIdrawable * driDrawable,
+                         int *width, int *height,
+                         unsigned int *attachments, int count,
+                         int *out_count, void *loaderPrivate)
 {
-    __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
-    DRI2Buffer *buffers;
+   __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
+   DRI2Buffer *buffers;
 
-    buffers = DRI2GetBuffersWithFormat(pdraw->base.psc->dpy,
-				       pdraw->base.xDrawable,
-				       width, height, attachments,
-				       count, out_count);
-    if (buffers == NULL)
-       return NULL;
+   buffers = DRI2GetBuffersWithFormat(pdraw->base.psc->dpy,
+                                      pdraw->base.xDrawable,
+                                      width, height, attachments,
+                                      count, out_count);
+   if (buffers == NULL)
+      return NULL;
 
-    pdraw->width = *width;
-    pdraw->height = *height;
-    process_buffers(pdraw, buffers, *out_count);
+   pdraw->width = *width;
+   pdraw->height = *height;
+   process_buffers(pdraw, buffers, *out_count);
 
-    Xfree(buffers);
+   Xfree(buffers);
 
-    return pdraw->buffers;
+   return pdraw->buffers;
 }
 
 static const __DRIdri2LoaderExtension dri2LoaderExtension = {
-    { __DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION },
-    dri2GetBuffers,
-    dri2FlushFrontBuffer,
-    dri2GetBuffersWithFormat,
+   {__DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION},
+   dri2GetBuffers,
+   dri2FlushFrontBuffer,
+   dri2GetBuffersWithFormat,
 };
 
 static const __DRIdri2LoaderExtension dri2LoaderExtension_old = {
-    { __DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION },
-    dri2GetBuffers,
-    dri2FlushFrontBuffer,
-    NULL,
+   {__DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION},
+   dri2GetBuffers,
+   dri2FlushFrontBuffer,
+   NULL,
 };
 
 static const __DRIextension *loader_extensions[] = {
-    &dri2LoaderExtension.base,
-    &systemTimeExtension.base,
-    NULL
+   &dri2LoaderExtension.base,
+   &systemTimeExtension.base,
+   NULL
 };
 
 static const __DRIextension *loader_extensions_old[] = {
-    &dri2LoaderExtension_old.base,
-    &systemTimeExtension.base,
-    NULL
+   &dri2LoaderExtension_old.base,
+   &systemTimeExtension.base,
+   NULL
 };
 
-static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen,
-					__GLXdisplayPrivate *priv)
+static __GLXDRIscreen *
+dri2CreateScreen(__GLXscreenConfigs * psc, int screen,
+                 __GLXdisplayPrivate * priv)
 {
-    const __DRIconfig **driver_configs;
-    const __DRIextension **extensions;
-    const __GLXDRIdisplayPrivate *const pdp = (__GLXDRIdisplayPrivate *)
+   const __DRIconfig **driver_configs;
+   const __DRIextension **extensions;
+   const __GLXDRIdisplayPrivate *const pdp = (__GLXDRIdisplayPrivate *)
       priv->dri2Display;
-    __GLXDRIscreen *psp;
-    char *driverName, *deviceName;
-    drm_magic_t magic;
-    int i;
-
-    psp = Xmalloc(sizeof *psp);
-    if (psp == NULL)
-	return NULL;
-
-    /* Initialize per screen dynamic client GLX extensions */
-    psc->ext_list_first_time = GL_TRUE;
-
-    if (!DRI2Connect(psc->dpy, RootWindow(psc->dpy, screen),
-		     &driverName, &deviceName))
-	return NULL;
-
-    psc->driver = driOpenDriver(driverName);
-    if (psc->driver == NULL) {
-	ErrorMessageF("driver pointer missing\n");
-	goto handle_error;
-    }
-
-    extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS);
-    if (extensions == NULL) {
-	ErrorMessageF("driver exports no extensions (%s)\n", dlerror());
-	goto handle_error;
-    }
-    
-    for (i = 0; extensions[i]; i++) {
-	if (strcmp(extensions[i]->name, __DRI_CORE) == 0)
-	    psc->core = (__DRIcoreExtension *) extensions[i];
-	if (strcmp(extensions[i]->name, __DRI_DRI2) == 0)
-	    psc->dri2 = (__DRIdri2Extension *) extensions[i];
-    }
-
-    if (psc->core == NULL || psc->dri2 == NULL) {
-	ErrorMessageF("core dri or dri2 extension not found\n");
-	goto handle_error;
-    }
-
-    psc->fd = open(deviceName, O_RDWR);
-    if (psc->fd < 0) {
-	ErrorMessageF("failed to open drm device: %s\n", strerror(errno));
-	return NULL;
-    }
-
-    if (drmGetMagic(psc->fd, &magic)) {
-	ErrorMessageF("failed to get magic\n");
-	return NULL;
-    }
-
-    if (!DRI2Authenticate(psc->dpy, RootWindow(psc->dpy, screen), magic)) {
-	ErrorMessageF("failed to authenticate magic %d\n", magic);
-	return NULL;
-    }
-
-    /* If the server does not support the protocol for
-     * DRI2GetBuffersWithFormat, don't supply that interface to the driver.
-     */
-    psc->__driScreen = 
-      psc->dri2->createNewScreen(screen, psc->fd,
-				 ((pdp->driMinor < 1)
-				  ? loader_extensions_old
-				  : loader_extensions),
-				 &driver_configs, psc);
-
-    if (psc->__driScreen == NULL) {
-	ErrorMessageF("failed to create dri screen\n");
-	return NULL;
-    }
-
-    driBindExtensions(psc, 1);
-
-    psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs);
-    psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs);
-
-    psc->driver_configs = driver_configs;
-
-    psp->destroyScreen = dri2DestroyScreen;
-    psp->createContext = dri2CreateContext;
-    psp->createDrawable = dri2CreateDrawable;
-    psp->swapBuffers = dri2SwapBuffers;
-    psp->waitGL = dri2WaitGL;
-    psp->waitX = dri2WaitX;
-
-    /* DRI2 suports SubBuffer through DRI2CopyRegion, so it's always
-     * available.*/
-    psp->copySubBuffer = dri2CopySubBuffer;
-    __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer");
-
-    Xfree(driverName);
-    Xfree(deviceName);
-
-    return psp;
+   __GLXDRIscreen *psp;
+   char *driverName, *deviceName;
+   drm_magic_t magic;
+   int i;
+
+   psp = Xmalloc(sizeof *psp);
+   if (psp == NULL)
+      return NULL;
+
+   /* Initialize per screen dynamic client GLX extensions */
+   psc->ext_list_first_time = GL_TRUE;
+
+   if (!DRI2Connect(psc->dpy, RootWindow(psc->dpy, screen),
+                    &driverName, &deviceName))
+      return NULL;
+
+   psc->driver = driOpenDriver(driverName);
+   if (psc->driver == NULL) {
+      ErrorMessageF("driver pointer missing\n");
+      goto handle_error;
+   }
+
+   extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS);
+   if (extensions == NULL) {
+      ErrorMessageF("driver exports no extensions (%s)\n", dlerror());
+      goto handle_error;
+   }
+
+   for (i = 0; extensions[i]; i++) {
+      if (strcmp(extensions[i]->name, __DRI_CORE) == 0)
+         psc->core = (__DRIcoreExtension *) extensions[i];
+      if (strcmp(extensions[i]->name, __DRI_DRI2) == 0)
+         psc->dri2 = (__DRIdri2Extension *) extensions[i];
+   }
+
+   if (psc->core == NULL || psc->dri2 == NULL) {
+      ErrorMessageF("core dri or dri2 extension not found\n");
+      goto handle_error;
+   }
+
+   psc->fd = open(deviceName, O_RDWR);
+   if (psc->fd < 0) {
+      ErrorMessageF("failed to open drm device: %s\n", strerror(errno));
+      return NULL;
+   }
+
+   if (drmGetMagic(psc->fd, &magic)) {
+      ErrorMessageF("failed to get magic\n");
+      return NULL;
+   }
+
+   if (!DRI2Authenticate(psc->dpy, RootWindow(psc->dpy, screen), magic)) {
+      ErrorMessageF("failed to authenticate magic %d\n", magic);
+      return NULL;
+   }
+
+   /* If the server does not support the protocol for
+    * DRI2GetBuffersWithFormat, don't supply that interface to the driver.
+    */
+   psc->__driScreen =
+      psc->dri2->createNewScreen(screen, psc->fd, ((pdp->driMinor < 1)
+                                                   ? loader_extensions_old
+                                                   : loader_extensions),
+                                 &driver_configs, psc);
+
+   if (psc->__driScreen == NULL) {
+      ErrorMessageF("failed to create dri screen\n");
+      return NULL;
+   }
+
+   driBindExtensions(psc, 1);
+
+   psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs);
+   psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs);
+
+   psc->driver_configs = driver_configs;
+
+   psp->destroyScreen = dri2DestroyScreen;
+   psp->createContext = dri2CreateContext;
+   psp->createDrawable = dri2CreateDrawable;
+   psp->swapBuffers = dri2SwapBuffers;
+   psp->waitGL = dri2WaitGL;
+   psp->waitX = dri2WaitX;
+
+   /* DRI2 suports SubBuffer through DRI2CopyRegion, so it's always
+    * available.*/
+   psp->copySubBuffer = dri2CopySubBuffer;
+   __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer");
+
+   Xfree(driverName);
+   Xfree(deviceName);
+
+   return psp;
 
  handle_error:
-    Xfree(driverName);
-    Xfree(deviceName);
+   Xfree(driverName);
+   Xfree(deviceName);
 
-    /* FIXME: clean up here */
+   /* FIXME: clean up here */
 
-    return NULL;
+   return NULL;
 }
 
 /* Called from __glXFreeDisplayPrivate.
  */
-static void dri2DestroyDisplay(__GLXDRIdisplay *dpy)
+static void
+dri2DestroyDisplay(__GLXDRIdisplay * dpy)
 {
-    Xfree(dpy);
+   Xfree(dpy);
 }
 
 /*
@@ -521,29 +540,30 @@ static void dri2DestroyDisplay(__GLXDRIdisplay *dpy)
  * This is called from __glXInitialize() when we are given a new
  * display pointer.
  */
-_X_HIDDEN __GLXDRIdisplay *dri2CreateDisplay(Display *dpy)
+_X_HIDDEN __GLXDRIdisplay *
+dri2CreateDisplay(Display * dpy)
 {
-    __GLXDRIdisplayPrivate *pdp;
-    int eventBase, errorBase;
+   __GLXDRIdisplayPrivate *pdp;
+   int eventBase, errorBase;
 
-    if (!DRI2QueryExtension(dpy, &eventBase, &errorBase))
-	return NULL;
+   if (!DRI2QueryExtension(dpy, &eventBase, &errorBase))
+      return NULL;
 
-    pdp = Xmalloc(sizeof *pdp);
-    if (pdp == NULL)
-	return NULL;
+   pdp = Xmalloc(sizeof *pdp);
+   if (pdp == NULL)
+      return NULL;
 
-    if (!DRI2QueryVersion(dpy, &pdp->driMajor, &pdp->driMinor)) {
-	Xfree(pdp);
-	return NULL;
-    }
+   if (!DRI2QueryVersion(dpy, &pdp->driMajor, &pdp->driMinor)) {
+      Xfree(pdp);
+      return NULL;
+   }
 
-    pdp->driPatch = 0;
+   pdp->driPatch = 0;
 
-    pdp->base.destroyDisplay = dri2DestroyDisplay;
-    pdp->base.createScreen = dri2CreateScreen;
+   pdp->base.destroyDisplay = dri2DestroyDisplay;
+   pdp->base.createScreen = dri2CreateScreen;
 
-    return &pdp->base;
+   return &pdp->base;
 }
 
 #endif /* GLX_DIRECT_RENDERING */
diff --git a/src/glx/x11/dri_common.c b/src/glx/x11/dri_common.c
index 6de4111113..9c825ad74a 100644
--- a/src/glx/x11/dri_common.c
+++ b/src/glx/x11/dri_common.c
@@ -49,33 +49,35 @@
 #define RTLD_GLOBAL 0
 #endif
 
-_X_HIDDEN void InfoMessageF(const char *f, ...)
+_X_HIDDEN void
+InfoMessageF(const char *f, ...)
 {
-    va_list args;
-    const char *env;
-
-    if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) {
-	fprintf(stderr, "libGL: ");
-	va_start(args, f);
-	vfprintf(stderr, f, args);
-	va_end(args);
-    }
+   va_list args;
+   const char *env;
+
+   if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) {
+      fprintf(stderr, "libGL: ");
+      va_start(args, f);
+      vfprintf(stderr, f, args);
+      va_end(args);
+   }
 }
 
 /**
  * Print error to stderr, unless LIBGL_DEBUG=="quiet".
  */
-_X_HIDDEN void ErrorMessageF(const char *f, ...)
+_X_HIDDEN void
+ErrorMessageF(const char *f, ...)
 {
-    va_list args;
-    const char *env;
-
-    if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) {
-	fprintf(stderr, "libGL error: ");
-	va_start(args, f);
-	vfprintf(stderr, f, args);
-	va_end(args);
-    }
+   va_list args;
+   const char *env;
+
+   if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) {
+      fprintf(stderr, "libGL error: ");
+      va_start(args, f);
+      vfprintf(stderr, f, args);
+      va_end(args);
+   }
 }
 
 #ifndef DEFAULT_DRIVER_DIR
@@ -95,7 +97,8 @@ _X_HIDDEN void ErrorMessageF(const char *f, ...)
  * \returns
  * A handle from \c dlopen, or \c NULL if driver file not found.
  */
-_X_HIDDEN void *driOpenDriver(const char *driverName)
+_X_HIDDEN void *
+driOpenDriver(const char *driverName)
 {
    void *glhandle, *handle;
    const char *libPaths, *p, *next;
@@ -110,40 +113,41 @@ _X_HIDDEN void *driOpenDriver(const char *driverName)
       /* don't allow setuid apps to use LIBGL_DRIVERS_PATH */
       libPaths = getenv("LIBGL_DRIVERS_PATH");
       if (!libPaths)
-         libPaths = getenv("LIBGL_DRIVERS_DIR"); /* deprecated */
+         libPaths = getenv("LIBGL_DRIVERS_DIR");        /* deprecated */
    }
    if (libPaths == NULL)
-       libPaths = DEFAULT_DRIVER_DIR;
+      libPaths = DEFAULT_DRIVER_DIR;
 
    handle = NULL;
    for (p = libPaths; *p; p = next) {
-       next = strchr(p, ':');
-       if (next == NULL) {
-	   len = strlen(p);
-	   next = p + len;
-       } else {
-	   len = next - p;
-	   next++;
-       }
+      next = strchr(p, ':');
+      if (next == NULL) {
+         len = strlen(p);
+         next = p + len;
+      }
+      else {
+         len = next - p;
+         next++;
+      }
 
 #ifdef GLX_USE_TLS
       snprintf(realDriverName, sizeof realDriverName,
-	       "%.*s/tls/%s_dri.so", len, p, driverName);
+               "%.*s/tls/%s_dri.so", len, p, driverName);
       InfoMessageF("OpenDriver: trying %s\n", realDriverName);
       handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL);
 #endif
 
-      if ( handle == NULL ) {
-	 snprintf(realDriverName, sizeof realDriverName,
-		  "%.*s/%s_dri.so", len, p, driverName);
-	 InfoMessageF("OpenDriver: trying %s\n", realDriverName);
-	 handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL);
+      if (handle == NULL) {
+         snprintf(realDriverName, sizeof realDriverName,
+                  "%.*s/%s_dri.so", len, p, driverName);
+         InfoMessageF("OpenDriver: trying %s\n", realDriverName);
+         handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL);
       }
 
-      if ( handle != NULL )
-	  break;
+      if (handle != NULL)
+         break;
       else
-	 ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror());
+         ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror());
    }
 
    if (!handle)
@@ -156,253 +160,258 @@ _X_HIDDEN void *driOpenDriver(const char *driverName)
 }
 
 _X_HIDDEN const __DRIsystemTimeExtension systemTimeExtension = {
-    { __DRI_SYSTEM_TIME, __DRI_SYSTEM_TIME_VERSION },
-    __glXGetUST,
-    __driGetMscRateOML
+   {__DRI_SYSTEM_TIME, __DRI_SYSTEM_TIME_VERSION},
+   __glXGetUST,
+   __driGetMscRateOML
 };
 
 #define __ATTRIB(attrib, field) \
     { attrib, offsetof(__GLcontextModes, field) }
 
-static const struct { unsigned int attrib, offset; } attribMap[] = {
-    __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE,			rgbBits),
-    __ATTRIB(__DRI_ATTRIB_LEVEL,			level),
-    __ATTRIB(__DRI_ATTRIB_RED_SIZE,			redBits),
-    __ATTRIB(__DRI_ATTRIB_GREEN_SIZE,			greenBits),
-    __ATTRIB(__DRI_ATTRIB_BLUE_SIZE,			blueBits),
-    __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE,			alphaBits),
-    __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE,			depthBits),
-    __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE,			stencilBits),
-    __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE,		accumRedBits),
-    __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE,		accumGreenBits),
-    __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE,		accumBlueBits),
-    __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE,		accumAlphaBits),
-    __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS,		sampleBuffers),
-    __ATTRIB(__DRI_ATTRIB_SAMPLES,			samples),
-    __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER,		doubleBufferMode),
-    __ATTRIB(__DRI_ATTRIB_STEREO,			stereoMode),
-    __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS,			numAuxBuffers),
+static const struct
+{
+   unsigned int attrib, offset;
+} attribMap[] = {
+   __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE, rgbBits),
+      __ATTRIB(__DRI_ATTRIB_LEVEL, level),
+      __ATTRIB(__DRI_ATTRIB_RED_SIZE, redBits),
+      __ATTRIB(__DRI_ATTRIB_GREEN_SIZE, greenBits),
+      __ATTRIB(__DRI_ATTRIB_BLUE_SIZE, blueBits),
+      __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE, alphaBits),
+      __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE, depthBits),
+      __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE, stencilBits),
+      __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE, accumRedBits),
+      __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE, accumGreenBits),
+      __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE, accumBlueBits),
+      __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE, accumAlphaBits),
+      __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS, sampleBuffers),
+      __ATTRIB(__DRI_ATTRIB_SAMPLES, samples),
+      __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER, doubleBufferMode),
+      __ATTRIB(__DRI_ATTRIB_STEREO, stereoMode),
+      __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS, numAuxBuffers),
 #if 0
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE,		transparentPixel),
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE,	transparentIndex),
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE,	transparentRed),
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE,	transparentGreen),
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE,	transparentBlue),
-    __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE,	transparentAlpha),
-    __ATTRIB(__DRI_ATTRIB_RED_MASK,			redMask),
-    __ATTRIB(__DRI_ATTRIB_GREEN_MASK,			greenMask),
-    __ATTRIB(__DRI_ATTRIB_BLUE_MASK,			blueMask),
-    __ATTRIB(__DRI_ATTRIB_ALPHA_MASK,			alphaMask),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE, transparentPixel),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE, transparentIndex),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE, transparentRed),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE, transparentGreen),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE, transparentBlue),
+      __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE, transparentAlpha),
+      __ATTRIB(__DRI_ATTRIB_RED_MASK, redMask),
+      __ATTRIB(__DRI_ATTRIB_GREEN_MASK, greenMask),
+      __ATTRIB(__DRI_ATTRIB_BLUE_MASK, blueMask),
+      __ATTRIB(__DRI_ATTRIB_ALPHA_MASK, alphaMask),
 #endif
-    __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH,		maxPbufferWidth),
-    __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT,		maxPbufferHeight),
-    __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS,		maxPbufferPixels),
-    __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH,	optimalPbufferWidth),
-    __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT,	optimalPbufferHeight),
+      __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH, maxPbufferWidth),
+      __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT, maxPbufferHeight),
+      __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS, maxPbufferPixels),
+      __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH, optimalPbufferWidth),
+      __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT, optimalPbufferHeight),
 #if 0
-    __ATTRIB(__DRI_ATTRIB_SWAP_METHOD,			swapMethod),
+      __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, swapMethod),
 #endif
-    __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB,		bindToTextureRgb),
-    __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA,		bindToTextureRgba),
-    __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE,	bindToMipmapTexture),
-    __ATTRIB(__DRI_ATTRIB_YINVERTED,			yInverted),
-};
+__ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb),
+      __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA, bindToTextureRgba),
+      __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE,
+                     bindToMipmapTexture),
+      __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted),};
 
 #define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
 
 static int
-scalarEqual(__GLcontextModes *mode, unsigned int attrib, unsigned int value)
+scalarEqual(__GLcontextModes * mode, unsigned int attrib, unsigned int value)
 {
-    unsigned int glxValue;
-    int i;
+   unsigned int glxValue;
+   int i;
 
-    for (i = 0; i < ARRAY_SIZE(attribMap); i++)
-	if (attribMap[i].attrib == attrib) {
-	    glxValue = *(unsigned int *) ((char *) mode + attribMap[i].offset);
-	    return glxValue == GLX_DONT_CARE || glxValue == value;
-	}
+   for (i = 0; i < ARRAY_SIZE(attribMap); i++)
+      if (attribMap[i].attrib == attrib) {
+         glxValue = *(unsigned int *) ((char *) mode + attribMap[i].offset);
+         return glxValue == GLX_DONT_CARE || glxValue == value;
+      }
 
-    return GL_TRUE; /* Is a non-existing attribute equal to value? */
+   return GL_TRUE;              /* Is a non-existing attribute equal to value? */
 }
 
 static int
-driConfigEqual(const __DRIcoreExtension *core,
-	       __GLcontextModes *modes, const __DRIconfig *driConfig)
+driConfigEqual(const __DRIcoreExtension * core,
+               __GLcontextModes * modes, const __DRIconfig * driConfig)
 {
-    unsigned int attrib, value, glxValue;
-    int i;
-
-    i = 0;
-    while (core->indexConfigAttrib(driConfig, i++, &attrib, &value)) {
-	switch (attrib) {
-	case __DRI_ATTRIB_RENDER_TYPE:
-	    glxValue = 0;
-	    if (value & __DRI_ATTRIB_RGBA_BIT) {
-		glxValue |= GLX_RGBA_BIT;
-	    } else if (value & __DRI_ATTRIB_COLOR_INDEX_BIT) {
-		glxValue |= GLX_COLOR_INDEX_BIT;
-	    }
-	    if (glxValue != modes->renderType)
-		return GL_FALSE;
-	    break;
-
-	case __DRI_ATTRIB_CONFIG_CAVEAT:
-	    if (value & __DRI_ATTRIB_NON_CONFORMANT_CONFIG)
-		glxValue = GLX_NON_CONFORMANT_CONFIG;
-	    else if (value & __DRI_ATTRIB_SLOW_BIT)
-		glxValue = GLX_SLOW_CONFIG;
-	    else
-		glxValue = GLX_NONE;
-	    if (glxValue != modes->visualRating)
-		return GL_FALSE;
-	    break;
-
-	case __DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS:
-	    glxValue = 0;
-	    if (value & __DRI_ATTRIB_TEXTURE_1D_BIT)
-		glxValue |= GLX_TEXTURE_1D_BIT_EXT;
-	    if (value & __DRI_ATTRIB_TEXTURE_2D_BIT)
-		glxValue |= GLX_TEXTURE_2D_BIT_EXT;
-	    if (value & __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT)
-		glxValue |= GLX_TEXTURE_RECTANGLE_BIT_EXT;
-	    if (modes->bindToTextureTargets != GLX_DONT_CARE &&
-		glxValue != modes->bindToTextureTargets)
-		return GL_FALSE;
-	    break;	
-
-	default:
-	    if (!scalarEqual(modes, attrib, value))
-		return GL_FALSE;
-	}
-    }
-
-    return GL_TRUE;
+   unsigned int attrib, value, glxValue;
+   int i;
+
+   i = 0;
+   while (core->indexConfigAttrib(driConfig, i++, &attrib, &value)) {
+      switch (attrib) {
+      case __DRI_ATTRIB_RENDER_TYPE:
+         glxValue = 0;
+         if (value & __DRI_ATTRIB_RGBA_BIT) {
+            glxValue |= GLX_RGBA_BIT;
+         }
+         else if (value & __DRI_ATTRIB_COLOR_INDEX_BIT) {
+            glxValue |= GLX_COLOR_INDEX_BIT;
+         }
+         if (glxValue != modes->renderType)
+            return GL_FALSE;
+         break;
+
+      case __DRI_ATTRIB_CONFIG_CAVEAT:
+         if (value & __DRI_ATTRIB_NON_CONFORMANT_CONFIG)
+            glxValue = GLX_NON_CONFORMANT_CONFIG;
+         else if (value & __DRI_ATTRIB_SLOW_BIT)
+            glxValue = GLX_SLOW_CONFIG;
+         else
+            glxValue = GLX_NONE;
+         if (glxValue != modes->visualRating)
+            return GL_FALSE;
+         break;
+
+      case __DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS:
+         glxValue = 0;
+         if (value & __DRI_ATTRIB_TEXTURE_1D_BIT)
+            glxValue |= GLX_TEXTURE_1D_BIT_EXT;
+         if (value & __DRI_ATTRIB_TEXTURE_2D_BIT)
+            glxValue |= GLX_TEXTURE_2D_BIT_EXT;
+         if (value & __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT)
+            glxValue |= GLX_TEXTURE_RECTANGLE_BIT_EXT;
+         if (modes->bindToTextureTargets != GLX_DONT_CARE &&
+             glxValue != modes->bindToTextureTargets)
+            return GL_FALSE;
+         break;
+
+      default:
+         if (!scalarEqual(modes, attrib, value))
+            return GL_FALSE;
+      }
+   }
+
+   return GL_TRUE;
 }
 
 static __GLcontextModes *
-createDriMode(const __DRIcoreExtension *core,
-	      __GLcontextModes *modes, const __DRIconfig **driConfigs)
+createDriMode(const __DRIcoreExtension * core,
+              __GLcontextModes * modes, const __DRIconfig ** driConfigs)
 {
-    __GLXDRIconfigPrivate *config;
-    int i;
+   __GLXDRIconfigPrivate *config;
+   int i;
 
-    for (i = 0; driConfigs[i]; i++) {
-	if (driConfigEqual(core, modes, driConfigs[i]))
-	    break;
-    }
+   for (i = 0; driConfigs[i]; i++) {
+      if (driConfigEqual(core, modes, driConfigs[i]))
+         break;
+   }
 
-    if (driConfigs[i] == NULL)
-	return NULL;
+   if (driConfigs[i] == NULL)
+      return NULL;
 
-    config = Xmalloc(sizeof *config);
-    if (config == NULL)
-	return NULL;
+   config = Xmalloc(sizeof *config);
+   if (config == NULL)
+      return NULL;
 
-    config->modes = *modes;
-    config->driConfig = driConfigs[i];
+   config->modes = *modes;
+   config->driConfig = driConfigs[i];
 
-    return &config->modes;
+   return &config->modes;
 }
 
 _X_HIDDEN __GLcontextModes *
-driConvertConfigs(const __DRIcoreExtension *core,
-		  __GLcontextModes *modes, const __DRIconfig **configs)
+driConvertConfigs(const __DRIcoreExtension * core,
+                  __GLcontextModes * modes, const __DRIconfig ** configs)
 {
-    __GLcontextModes head, *tail, *m;
-
-    tail = &head;
-    head.next = NULL;
-    for (m = modes; m; m = m->next) {
-	tail->next = createDriMode(core, m, configs);
-	if (tail->next == NULL) {
-	    /* no matching dri config for m */
-	    continue;
-	}
+   __GLcontextModes head, *tail, *m;
+
+   tail = &head;
+   head.next = NULL;
+   for (m = modes; m; m = m->next) {
+      tail->next = createDriMode(core, m, configs);
+      if (tail->next == NULL) {
+         /* no matching dri config for m */
+         continue;
+      }
 
 
-	tail = tail->next;
-    }
+      tail = tail->next;
+   }
 
-    _gl_context_modes_destroy(modes);
+   _gl_context_modes_destroy(modes);
 
-    return head.next;
+   return head.next;
 }
 
 _X_HIDDEN void
-driBindExtensions(__GLXscreenConfigs *psc, int dri2)
+driBindExtensions(__GLXscreenConfigs * psc, int dri2)
 {
-    const __DRIextension **extensions;
-    int i;
+   const __DRIextension **extensions;
+   int i;
 
-    extensions = psc->core->getExtensions(psc->__driScreen);
+   extensions = psc->core->getExtensions(psc->__driScreen);
 
-    for (i = 0; extensions[i]; i++) {
+   for (i = 0; extensions[i]; i++) {
 #ifdef __DRI_COPY_SUB_BUFFER
-	if (strcmp(extensions[i]->name, __DRI_COPY_SUB_BUFFER) == 0) {
-	    psc->driCopySubBuffer = (__DRIcopySubBufferExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer");
-	}
+      if (strcmp(extensions[i]->name, __DRI_COPY_SUB_BUFFER) == 0) {
+         psc->driCopySubBuffer =
+            (__DRIcopySubBufferExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer");
+      }
 #endif
 
 #ifdef __DRI_SWAP_CONTROL
-	/* No DRI2 support for swap_control at the moment, since SwapBuffers
-	 * is done by the X server */
-	if (strcmp(extensions[i]->name, __DRI_SWAP_CONTROL) == 0 && !dri2) {
-	    psc->swapControl = (__DRIswapControlExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_SGI_swap_control");
-	    __glXEnableDirectExtension(psc, "GLX_MESA_swap_control");
-	}
+      /* No DRI2 support for swap_control at the moment, since SwapBuffers
+       * is done by the X server */
+      if (strcmp(extensions[i]->name, __DRI_SWAP_CONTROL) == 0 && !dri2) {
+         psc->swapControl = (__DRIswapControlExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_SGI_swap_control");
+         __glXEnableDirectExtension(psc, "GLX_MESA_swap_control");
+      }
 #endif
 
 #ifdef __DRI_ALLOCATE
-	if (strcmp(extensions[i]->name, __DRI_ALLOCATE) == 0) {
-	    psc->allocate = (__DRIallocateExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_MESA_allocate_memory");
-	}
+      if (strcmp(extensions[i]->name, __DRI_ALLOCATE) == 0) {
+         psc->allocate = (__DRIallocateExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_MESA_allocate_memory");
+      }
 #endif
 
 #ifdef __DRI_FRAME_TRACKING
-	if (strcmp(extensions[i]->name, __DRI_FRAME_TRACKING) == 0) {
-	    psc->frameTracking = (__DRIframeTrackingExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_MESA_swap_frame_usage");
-	}
+      if (strcmp(extensions[i]->name, __DRI_FRAME_TRACKING) == 0) {
+         psc->frameTracking = (__DRIframeTrackingExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_MESA_swap_frame_usage");
+      }
 #endif
 
 #ifdef __DRI_MEDIA_STREAM_COUNTER
-	if (strcmp(extensions[i]->name, __DRI_MEDIA_STREAM_COUNTER) == 0) {
-	    psc->msc = (__DRImediaStreamCounterExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_SGI_video_sync");
-	}
+      if (strcmp(extensions[i]->name, __DRI_MEDIA_STREAM_COUNTER) == 0) {
+         psc->msc = (__DRImediaStreamCounterExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_SGI_video_sync");
+      }
 #endif
 
 #ifdef __DRI_SWAP_BUFFER_COUNTER
-	/* No driver supports this at this time and the extension is
-	 * not defined in dri_interface.h.  Will enable
-	 * GLX_OML_sync_control if implemented. */
+      /* No driver supports this at this time and the extension is
+       * not defined in dri_interface.h.  Will enable
+       * GLX_OML_sync_control if implemented. */
 #endif
 
 #ifdef __DRI_READ_DRAWABLE
-	if (strcmp(extensions[i]->name, __DRI_READ_DRAWABLE) == 0) {
-	    __glXEnableDirectExtension(psc, "GLX_SGI_make_current_read");
-	}
+      if (strcmp(extensions[i]->name, __DRI_READ_DRAWABLE) == 0) {
+         __glXEnableDirectExtension(psc, "GLX_SGI_make_current_read");
+      }
 #endif
 
 #ifdef __DRI_TEX_BUFFER
-	if ((strcmp(extensions[i]->name, __DRI_TEX_BUFFER) == 0) && dri2) {
-	    psc->texBuffer = (__DRItexBufferExtension *) extensions[i];
-	    __glXEnableDirectExtension(psc, "GLX_EXT_texture_from_pixmap");
-	}
+      if ((strcmp(extensions[i]->name, __DRI_TEX_BUFFER) == 0) && dri2) {
+         psc->texBuffer = (__DRItexBufferExtension *) extensions[i];
+         __glXEnableDirectExtension(psc, "GLX_EXT_texture_from_pixmap");
+      }
 #endif
 
 #ifdef __DRI2_FLUSH
-	if ((strcmp(extensions[i]->name, __DRI2_FLUSH) == 0) && dri2) {
-	    psc->f = (__DRI2flushExtension *) extensions[i];
-	    /* internal driver extension, no GL extension exposed */
-	}
+      if ((strcmp(extensions[i]->name, __DRI2_FLUSH) == 0) && dri2) {
+         psc->f = (__DRI2flushExtension *) extensions[i];
+         /* internal driver extension, no GL extension exposed */
+      }
 #endif
 
-	/* Ignore unknown extensions */
-    }
+      /* Ignore unknown extensions */
+   }
 }
 
 #endif /* GLX_DIRECT_RENDERING */
diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c
index d24471c436..ab24bd8ffe 100644
--- a/src/glx/x11/dri_glx.c
+++ b/src/glx/x11/dri_glx.c
@@ -51,22 +51,24 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate;
 typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate;
 
-struct __GLXDRIdisplayPrivateRec {
-    __GLXDRIdisplay base;
+struct __GLXDRIdisplayPrivateRec
+{
+   __GLXDRIdisplay base;
 
-    /*
+   /*
     ** XFree86-DRI version information
     */
-    int driMajor;
-    int driMinor;
-    int driPatch;
+   int driMajor;
+   int driMinor;
+   int driPatch;
 };
 
-struct __GLXDRIcontextPrivateRec {
-    __GLXDRIcontext base;
-    __DRIcontext *driContext;
-    XID hwContextID;
-    __GLXscreenConfigs *psc;
+struct __GLXDRIcontextPrivateRec
+{
+   __GLXDRIcontext base;
+   __DRIcontext *driContext;
+   XID hwContextID;
+   __GLXscreenConfigs *psc;
 };
 
 /*
@@ -74,47 +76,51 @@ struct __GLXDRIcontextPrivateRec {
  * the DRI driver for the screen. (I.e. "r128", "tdfx", etc).
  * Return True for success, False for failure.
  */
-static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName)
+static Bool
+driGetDriverName(Display * dpy, int scrNum, char **driverName)
 {
-    int directCapable;
-    Bool b;
-    int event, error;
-    int driverMajor, driverMinor, driverPatch;
-
-    *driverName = NULL;
-
-    if (XF86DRIQueryExtension(dpy, &event, &error)) { /* DRI1 */
-	if (!XF86DRIQueryDirectRenderingCapable(dpy, scrNum, &directCapable)) {
-	    ErrorMessageF("XF86DRIQueryDirectRenderingCapable failed\n");
-	    return False;
-	}
-	if (!directCapable) {
-	    ErrorMessageF("XF86DRIQueryDirectRenderingCapable returned false\n");
-	    return False;
-	}
-
-	b = XF86DRIGetClientDriverName(dpy, scrNum, &driverMajor, &driverMinor,
-		&driverPatch, driverName);
-	if (!b) {
-	    ErrorMessageF("Cannot determine driver name for screen %d\n", scrNum);
-	    return False;
-	}
-
-	InfoMessageF("XF86DRIGetClientDriverName: %d.%d.%d %s (screen %d)\n",
-		driverMajor, driverMinor, driverPatch, *driverName, scrNum);
-
-	return True;
-    } else if (DRI2QueryExtension(dpy, &event, &error)) { /* DRI2 */
-	char *dev;
-	Bool ret = DRI2Connect(dpy, RootWindow(dpy, scrNum), driverName, &dev);
-
-	if (ret)
-	    Xfree(dev);
-
-	return ret;
-    }
-
-    return False;
+   int directCapable;
+   Bool b;
+   int event, error;
+   int driverMajor, driverMinor, driverPatch;
+
+   *driverName = NULL;
+
+   if (XF86DRIQueryExtension(dpy, &event, &error)) {    /* DRI1 */
+      if (!XF86DRIQueryDirectRenderingCapable(dpy, scrNum, &directCapable)) {
+         ErrorMessageF("XF86DRIQueryDirectRenderingCapable failed\n");
+         return False;
+      }
+      if (!directCapable) {
+         ErrorMessageF("XF86DRIQueryDirectRenderingCapable returned false\n");
+         return False;
+      }
+
+      b = XF86DRIGetClientDriverName(dpy, scrNum, &driverMajor, &driverMinor,
+                                     &driverPatch, driverName);
+      if (!b) {
+         ErrorMessageF("Cannot determine driver name for screen %d\n",
+                       scrNum);
+         return False;
+      }
+
+      InfoMessageF("XF86DRIGetClientDriverName: %d.%d.%d %s (screen %d)\n",
+                   driverMajor, driverMinor, driverPatch, *driverName,
+                   scrNum);
+
+      return True;
+   }
+   else if (DRI2QueryExtension(dpy, &event, &error)) {  /* DRI2 */
+      char *dev;
+      Bool ret = DRI2Connect(dpy, RootWindow(dpy, scrNum), driverName, &dev);
+
+      if (ret)
+         Xfree(dev);
+
+      return ret;
+   }
+
+   return False;
 }
 
 /*
@@ -123,17 +129,19 @@ static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName)
  * The returned char pointer points to a static array that will be
  * overwritten by subsequent calls.
  */
-PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) {
+PUBLIC const char *
+glXGetScreenDriver(Display * dpy, int scrNum)
+{
    static char ret[32];
    char *driverName;
    if (driGetDriverName(dpy, scrNum, &driverName)) {
       int len;
       if (!driverName)
-	 return NULL;
-      len = strlen (driverName);
+         return NULL;
+      len = strlen(driverName);
       if (len >= 31)
-	 return NULL;
-      memcpy (ret, driverName, len+1);
+         return NULL;
+      memcpy(ret, driverName, len + 1);
       Xfree(driverName);
       return ret;
    }
@@ -151,121 +159,125 @@ PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) {
  *
  * Note: The driver remains opened after this function returns.
  */
-PUBLIC const char *glXGetDriverConfig (const char *driverName)
+PUBLIC const char *
+glXGetDriverConfig(const char *driverName)
 {
-   void *handle = driOpenDriver (driverName);
+   void *handle = driOpenDriver(driverName);
    if (handle)
-      return dlsym (handle, "__driConfigOptions");
+      return dlsym(handle, "__driConfigOptions");
    else
       return NULL;
 }
 
 #ifdef XDAMAGE_1_1_INTERFACE
 
-static GLboolean has_damage_post(Display *dpy)
+static GLboolean
+has_damage_post(Display * dpy)
 {
-    static GLboolean inited = GL_FALSE;
-    static GLboolean has_damage;
-
-    if (!inited) {
-	int major, minor;
-
-	if (XDamageQueryVersion(dpy, &major, &minor) &&
-	    major == 1 && minor >= 1)
-	{
-	    has_damage = GL_TRUE;
-	} else {
-	    has_damage = GL_FALSE;
-	}
-	inited = GL_TRUE;
-    }
-
-    return has_damage;
+   static GLboolean inited = GL_FALSE;
+   static GLboolean has_damage;
+
+   if (!inited) {
+      int major, minor;
+
+      if (XDamageQueryVersion(dpy, &major, &minor) &&
+          major == 1 && minor >= 1) {
+         has_damage = GL_TRUE;
+      }
+      else {
+         has_damage = GL_FALSE;
+      }
+      inited = GL_TRUE;
+   }
+
+   return has_damage;
 }
 
-static void __glXReportDamage(__DRIdrawable *driDraw,
-			      int x, int y,
-			      drm_clip_rect_t *rects, int num_rects,
-			      GLboolean front_buffer,
-			      void *loaderPrivate)
+static void
+__glXReportDamage(__DRIdrawable * driDraw,
+                  int x, int y,
+                  drm_clip_rect_t * rects, int num_rects,
+                  GLboolean front_buffer, void *loaderPrivate)
 {
-    XRectangle *xrects;
-    XserverRegion region;
-    int i;
-    int x_off, y_off;
-    __GLXDRIdrawable *glxDraw = loaderPrivate;
-    __GLXscreenConfigs *psc = glxDraw->psc;
-    Display *dpy = psc->dpy;
-    Drawable drawable;
-
-    if (!has_damage_post(dpy))
-	return;
-
-    if (front_buffer) {
-	x_off = x;
-	y_off = y;
-	drawable = RootWindow(dpy, psc->scr);
-    } else{
-	x_off = 0;
-	y_off = 0;
-	drawable = glxDraw->xDrawable;
-    }
-
-    xrects = malloc(sizeof(XRectangle) * num_rects);
-    if (xrects == NULL)
-	return;
-
-    for (i = 0; i < num_rects; i++) {
-	xrects[i].x = rects[i].x1 + x_off;
-	xrects[i].y = rects[i].y1 + y_off;
-	xrects[i].width = rects[i].x2 - rects[i].x1;
-	xrects[i].height = rects[i].y2 - rects[i].y1;
-    }
-    region = XFixesCreateRegion(dpy, xrects, num_rects);
-    free(xrects);
-    XDamageAdd(dpy, drawable, region);
-    XFixesDestroyRegion(dpy, region);
+   XRectangle *xrects;
+   XserverRegion region;
+   int i;
+   int x_off, y_off;
+   __GLXDRIdrawable *glxDraw = loaderPrivate;
+   __GLXscreenConfigs *psc = glxDraw->psc;
+   Display *dpy = psc->dpy;
+   Drawable drawable;
+
+   if (!has_damage_post(dpy))
+      return;
+
+   if (front_buffer) {
+      x_off = x;
+      y_off = y;
+      drawable = RootWindow(dpy, psc->scr);
+   }
+   else {
+      x_off = 0;
+      y_off = 0;
+      drawable = glxDraw->xDrawable;
+   }
+
+   xrects = malloc(sizeof(XRectangle) * num_rects);
+   if (xrects == NULL)
+      return;
+
+   for (i = 0; i < num_rects; i++) {
+      xrects[i].x = rects[i].x1 + x_off;
+      xrects[i].y = rects[i].y1 + y_off;
+      xrects[i].width = rects[i].x2 - rects[i].x1;
+      xrects[i].height = rects[i].y2 - rects[i].y1;
+   }
+   region = XFixesCreateRegion(dpy, xrects, num_rects);
+   free(xrects);
+   XDamageAdd(dpy, drawable, region);
+   XFixesDestroyRegion(dpy, region);
 }
 
 static const __DRIdamageExtension damageExtension = {
-    { __DRI_DAMAGE, __DRI_DAMAGE_VERSION },
-    __glXReportDamage,
+   {__DRI_DAMAGE, __DRI_DAMAGE_VERSION},
+   __glXReportDamage,
 };
 
 #endif
 
 static GLboolean
-__glXDRIGetDrawableInfo(__DRIdrawable *drawable,
-			unsigned int *index, unsigned int *stamp, 
-			int *X, int *Y, int *W, int *H,
-			int *numClipRects, drm_clip_rect_t ** pClipRects,
-			int *backX, int *backY,
-			int *numBackClipRects, drm_clip_rect_t **pBackClipRects,
-			void *loaderPrivate)
+__glXDRIGetDrawableInfo(__DRIdrawable * drawable,
+                        unsigned int *index, unsigned int *stamp,
+                        int *X, int *Y, int *W, int *H,
+                        int *numClipRects, drm_clip_rect_t ** pClipRects,
+                        int *backX, int *backY,
+                        int *numBackClipRects,
+                        drm_clip_rect_t ** pBackClipRects,
+                        void *loaderPrivate)
 {
-    __GLXDRIdrawable *glxDraw = loaderPrivate;
-    __GLXscreenConfigs *psc = glxDraw->psc;
-    Display *dpy = psc->dpy;
-
-    return XF86DRIGetDrawableInfo(dpy, psc->scr, glxDraw->drawable,
-				  index, stamp, X, Y, W, H,
-				  numClipRects, pClipRects,
-				  backX, backY,
-				  numBackClipRects, pBackClipRects);
+   __GLXDRIdrawable *glxDraw = loaderPrivate;
+   __GLXscreenConfigs *psc = glxDraw->psc;
+   Display *dpy = psc->dpy;
+
+   return XF86DRIGetDrawableInfo(dpy, psc->scr, glxDraw->drawable,
+                                 index, stamp, X, Y, W, H,
+                                 numClipRects, pClipRects,
+                                 backX, backY,
+                                 numBackClipRects, pBackClipRects);
 }
 
 static const __DRIgetDrawableInfoExtension getDrawableInfoExtension = {
-    { __DRI_GET_DRAWABLE_INFO, __DRI_GET_DRAWABLE_INFO_VERSION },
-    __glXDRIGetDrawableInfo
+   {__DRI_GET_DRAWABLE_INFO, __DRI_GET_DRAWABLE_INFO_VERSION},
+   __glXDRIGetDrawableInfo
 };
 
 static const __DRIextension *loader_extensions[] = {
-    &systemTimeExtension.base,
-    &getDrawableInfoExtension.base,
+   &systemTimeExtension.base,
+   &getDrawableInfoExtension.base,
 #ifdef XDAMAGE_1_1_INTERFACE
-    &damageExtension.base,
+   &damageExtension.base,
 #endif
-    NULL
+   NULL
 };
 
 #ifndef GLX_USE_APPLEGL
@@ -284,427 +296,429 @@ static const __DRIextension *loader_extensions[] = {
  *          the client-side driver on success, or \c NULL on failure.
  */
 static void *
-CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc,
-		    __GLXDRIdisplayPrivate * driDpy)
+CallCreateNewScreen(Display * dpy, int scrn, __GLXscreenConfigs * psc,
+                    __GLXDRIdisplayPrivate * driDpy)
 {
-    void *psp = NULL;
-    drm_handle_t hSAREA;
-    drmAddress pSAREA = MAP_FAILED;
-    char *BusID;
-    __DRIversion   ddx_version;
-    __DRIversion   dri_version;
-    __DRIversion   drm_version;
-    __DRIframebuffer  framebuffer;
-    int   fd = -1;
-    int   status;
-
-    drm_magic_t magic;
-    drmVersionPtr version;
-    int newlyopened;
-    char *driverName;
-    drm_handle_t  hFB;
-    int        junk;
-    const __DRIconfig **driver_configs;
-    __GLcontextModes *visual;
-
-    /* DRI protocol version. */
-    dri_version.major = driDpy->driMajor;
-    dri_version.minor = driDpy->driMinor;
-    dri_version.patch = driDpy->driPatch;
-
-    framebuffer.base = MAP_FAILED;
-    framebuffer.dev_priv = NULL;
-
-    if (!XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) {
-	ErrorMessageF("XF86DRIOpenConnection failed\n");
-	goto handle_error;
-    }
-
-    fd = drmOpenOnce(NULL, BusID, &newlyopened);
-
-    Xfree(BusID); /* No longer needed */
-
-    if (fd < 0) {
-	ErrorMessageF("drmOpenOnce failed (%s)\n", strerror(-fd));
-	goto handle_error;
-    }
-
-    if (drmGetMagic(fd, &magic)) {
-	ErrorMessageF("drmGetMagic failed\n");
-	goto handle_error;
-    }
-
-    version = drmGetVersion(fd);
-    if (version) {
-	drm_version.major = version->version_major;
-	drm_version.minor = version->version_minor;
-	drm_version.patch = version->version_patchlevel;
-	drmFreeVersion(version);
-    }
-    else {
-	drm_version.major = -1;
-	drm_version.minor = -1;
-	drm_version.patch = -1;
-    }
-
-    if (newlyopened && !XF86DRIAuthConnection(dpy, scrn, magic)) {
-	ErrorMessageF("XF86DRIAuthConnection failed\n");
-	goto handle_error;
-    }
-
-    /* Get device name (like "tdfx") and the ddx version numbers.
-     * We'll check the version in each DRI driver's "createNewScreen"
-     * function. */
-    if (!XF86DRIGetClientDriverName(dpy, scrn,
-				    &ddx_version.major,
-				    &ddx_version.minor,
-				    &ddx_version.patch,
-				    &driverName)) {
-	ErrorMessageF("XF86DRIGetClientDriverName failed\n");
-	goto handle_error;
-    }
-
-    Xfree(driverName); /* No longer needed. */
-
-    /*
-     * Get device-specific info.  pDevPriv will point to a struct
-     * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) that
-     * has information about the screen size, depth, pitch, ancilliary
-     * buffers, DRM mmap handles, etc.
-     */
-    if (!XF86DRIGetDeviceInfo(dpy, scrn, &hFB, &junk,
-			      &framebuffer.size, &framebuffer.stride,
-			      &framebuffer.dev_priv_size, &framebuffer.dev_priv)) {
-	ErrorMessageF("XF86DRIGetDeviceInfo failed");
-	goto handle_error;
-    }
-
-    framebuffer.width = DisplayWidth(dpy, scrn);
-    framebuffer.height = DisplayHeight(dpy, scrn);
-
-    /* Map the framebuffer region. */
-    status = drmMap(fd, hFB, framebuffer.size, 
-		    (drmAddressPtr)&framebuffer.base);
-    if (status != 0) {
-	ErrorMessageF("drmMap of framebuffer failed (%s)", strerror(-status));
-	goto handle_error;
-    }
-
-    /* Map the SAREA region.  Further mmap regions may be setup in
-     * each DRI driver's "createNewScreen" function.
-     */
-    status = drmMap(fd, hSAREA, SAREA_MAX, &pSAREA);
-    if (status != 0) {
-	ErrorMessageF("drmMap of SAREA failed (%s)", strerror(-status));
-	goto handle_error;
-    }
-
-    psp = (*psc->legacy->createNewScreen)(scrn,
-					  &ddx_version,
-					  &dri_version,
-					  &drm_version,
-					  &framebuffer,
-					  pSAREA,
-					  fd,
-					  loader_extensions,
-					  &driver_configs,
-					  psc);
-
-    if (psp == NULL) {
-	ErrorMessageF("Calling driver entry point failed");
-	goto handle_error;
-    }
-
-    psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs);
-    psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs);
-
-    psc->driver_configs = driver_configs;
-
-    /* Visuals with depth != screen depth are subject to automatic compositing
-     * in the X server, so DRI1 can't render to them properly. Mark them as
-     * non-conformant to prevent apps from picking them up accidentally.
-     */
-    for (visual = psc->visuals; visual; visual = visual->next) {
-	XVisualInfo template;
-	XVisualInfo *visuals;
-	int num_visuals;
-	long mask;
-
-	template.visualid = visual->visualID;
-	mask = VisualIDMask;
-	visuals = XGetVisualInfo(dpy, mask, &template, &num_visuals);
-
-	if (visuals) {
-	    if (num_visuals > 0 && visuals->depth != DefaultDepth(dpy, scrn))
-		visual->visualRating = GLX_NON_CONFORMANT_CONFIG;
-
-	    XFree(visuals);
-	}
-    }
-
-    return psp;
+   void *psp = NULL;
+   drm_handle_t hSAREA;
+   drmAddress pSAREA = MAP_FAILED;
+   char *BusID;
+   __DRIversion ddx_version;
+   __DRIversion dri_version;
+   __DRIversion drm_version;
+   __DRIframebuffer framebuffer;
+   int fd = -1;
+   int status;
+
+   drm_magic_t magic;
+   drmVersionPtr version;
+   int newlyopened;
+   char *driverName;
+   drm_handle_t hFB;
+   int junk;
+   const __DRIconfig **driver_configs;
+   __GLcontextModes *visual;
+
+   /* DRI protocol version. */
+   dri_version.major = driDpy->driMajor;
+   dri_version.minor = driDpy->driMinor;
+   dri_version.patch = driDpy->driPatch;
+
+   framebuffer.base = MAP_FAILED;
+   framebuffer.dev_priv = NULL;
+
+   if (!XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) {
+      ErrorMessageF("XF86DRIOpenConnection failed\n");
+      goto handle_error;
+   }
+
+   fd = drmOpenOnce(NULL, BusID, &newlyopened);
+
+   Xfree(BusID);                /* No longer needed */
+
+   if (fd < 0) {
+      ErrorMessageF("drmOpenOnce failed (%s)\n", strerror(-fd));
+      goto handle_error;
+   }
+
+   if (drmGetMagic(fd, &magic)) {
+      ErrorMessageF("drmGetMagic failed\n");
+      goto handle_error;
+   }
+
+   version = drmGetVersion(fd);
+   if (version) {
+      drm_version.major = version->version_major;
+      drm_version.minor = version->version_minor;
+      drm_version.patch = version->version_patchlevel;
+      drmFreeVersion(version);
+   }
+   else {
+      drm_version.major = -1;
+      drm_version.minor = -1;
+      drm_version.patch = -1;
+   }
+
+   if (newlyopened && !XF86DRIAuthConnection(dpy, scrn, magic)) {
+      ErrorMessageF("XF86DRIAuthConnection failed\n");
+      goto handle_error;
+   }
+
+   /* Get device name (like "tdfx") and the ddx version numbers.
+    * We'll check the version in each DRI driver's "createNewScreen"
+    * function. */
+   if (!XF86DRIGetClientDriverName(dpy, scrn,
+                                   &ddx_version.major,
+                                   &ddx_version.minor,
+                                   &ddx_version.patch, &driverName)) {
+      ErrorMessageF("XF86DRIGetClientDriverName failed\n");
+      goto handle_error;
+   }
+
+   Xfree(driverName);           /* No longer needed. */
+
+   /*
+    * Get device-specific info.  pDevPriv will point to a struct
+    * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) that
+    * has information about the screen size, depth, pitch, ancilliary
+    * buffers, DRM mmap handles, etc.
+    */
+   if (!XF86DRIGetDeviceInfo(dpy, scrn, &hFB, &junk,
+                             &framebuffer.size, &framebuffer.stride,
+                             &framebuffer.dev_priv_size,
+                             &framebuffer.dev_priv)) {
+      ErrorMessageF("XF86DRIGetDeviceInfo failed");
+      goto handle_error;
+   }
+
+   framebuffer.width = DisplayWidth(dpy, scrn);
+   framebuffer.height = DisplayHeight(dpy, scrn);
+
+   /* Map the framebuffer region. */
+   status = drmMap(fd, hFB, framebuffer.size,
+                   (drmAddressPtr) & framebuffer.base);
+   if (status != 0) {
+      ErrorMessageF("drmMap of framebuffer failed (%s)", strerror(-status));
+      goto handle_error;
+   }
+
+   /* Map the SAREA region.  Further mmap regions may be setup in
+    * each DRI driver's "createNewScreen" function.
+    */
+   status = drmMap(fd, hSAREA, SAREA_MAX, &pSAREA);
+   if (status != 0) {
+      ErrorMessageF("drmMap of SAREA failed (%s)", strerror(-status));
+      goto handle_error;
+   }
+
+   psp = (*psc->legacy->createNewScreen) (scrn,
+                                          &ddx_version,
+                                          &dri_version,
+                                          &drm_version,
+                                          &framebuffer,
+                                          pSAREA,
+                                          fd,
+                                          loader_extensions,
+                                          &driver_configs, psc);
+
+   if (psp == NULL) {
+      ErrorMessageF("Calling driver entry point failed");
+      goto handle_error;
+   }
+
+   psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs);
+   psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs);
+
+   psc->driver_configs = driver_configs;
+
+   /* Visuals with depth != screen depth are subject to automatic compositing
+    * in the X server, so DRI1 can't render to them properly. Mark them as
+    * non-conformant to prevent apps from picking them up accidentally.
+    */
+   for (visual = psc->visuals; visual; visual = visual->next) {
+      XVisualInfo template;
+      XVisualInfo *visuals;
+      int num_visuals;
+      long mask;
+
+      template.visualid = visual->visualID;
+      mask = VisualIDMask;
+      visuals = XGetVisualInfo(dpy, mask, &template, &num_visuals);
+
+      if (visuals) {
+         if (num_visuals > 0 && visuals->depth != DefaultDepth(dpy, scrn))
+            visual->visualRating = GLX_NON_CONFORMANT_CONFIG;
+
+         XFree(visuals);
+      }
+   }
+
+   return psp;
 
  handle_error:
-    if (pSAREA != MAP_FAILED)
-	drmUnmap(pSAREA, SAREA_MAX);
+   if (pSAREA != MAP_FAILED)
+      drmUnmap(pSAREA, SAREA_MAX);
 
-    if (framebuffer.base != MAP_FAILED)
-	drmUnmap((drmAddress)framebuffer.base, framebuffer.size);
+   if (framebuffer.base != MAP_FAILED)
+      drmUnmap((drmAddress) framebuffer.base, framebuffer.size);
 
-    if (framebuffer.dev_priv != NULL)
-	Xfree(framebuffer.dev_priv);
+   if (framebuffer.dev_priv != NULL)
+      Xfree(framebuffer.dev_priv);
 
-    if (fd >= 0)
-	drmCloseOnce(fd);
+   if (fd >= 0)
+      drmCloseOnce(fd);
 
-    XF86DRICloseConnection(dpy, scrn);
+   XF86DRICloseConnection(dpy, scrn);
 
-    ErrorMessageF("reverting to software direct rendering\n");
+   ErrorMessageF("reverting to software direct rendering\n");
 
-    return NULL;
+   return NULL;
 }
 
 #else /* !GLX_USE_APPLEGL */
 
 static void *
-CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc,
-		    __GLXDRIdisplayPrivate * driDpy)
+CallCreateNewScreen(Display * dpy, int scrn, __GLXscreenConfigs * psc,
+                    __GLXDRIdisplayPrivate * driDpy)
 {
-    return NULL;
+   return NULL;
 }
 
 #endif /* !GLX_USE_APPLEGL */
 
-static void driDestroyContext(__GLXDRIcontext *context,
-			      __GLXscreenConfigs *psc, Display *dpy)
+static void
+driDestroyContext(__GLXDRIcontext * context,
+                  __GLXscreenConfigs * psc, Display * dpy)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-			
-    (*psc->core->destroyContext)(pcp->driContext);
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
 
-    XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID);
-    Xfree(pcp);
+   (*psc->core->destroyContext) (pcp->driContext);
+
+   XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID);
+   Xfree(pcp);
 }
 
-static Bool driBindContext(__GLXDRIcontext *context,
-			   __GLXDRIdrawable *draw, __GLXDRIdrawable *read)
+static Bool
+driBindContext(__GLXDRIcontext * context,
+               __GLXDRIdrawable * draw, __GLXDRIdrawable * read)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-    const __DRIcoreExtension *core = pcp->psc->core;
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
+   const __DRIcoreExtension *core = pcp->psc->core;
 
-    return (*core->bindContext)(pcp->driContext,
-				draw->driDrawable,
-				read->driDrawable);
+   return (*core->bindContext) (pcp->driContext,
+                                draw->driDrawable, read->driDrawable);
 }
 
-static void driUnbindContext(__GLXDRIcontext *context)
+static void
+driUnbindContext(__GLXDRIcontext * context)
 {
-    __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
-    const __DRIcoreExtension *core = pcp->psc->core;
+   __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context;
+   const __DRIcoreExtension *core = pcp->psc->core;
 
-    (*core->unbindContext)(pcp->driContext);
+   (*core->unbindContext) (pcp->driContext);
 }
 
-static __GLXDRIcontext *driCreateContext(__GLXscreenConfigs *psc,
-					 const __GLcontextModes *mode,
-					 GLXContext gc,
-					 GLXContext shareList, int renderType)
+static __GLXDRIcontext *
+driCreateContext(__GLXscreenConfigs * psc,
+                 const __GLcontextModes * mode,
+                 GLXContext gc, GLXContext shareList, int renderType)
 {
-    __GLXDRIcontextPrivate *pcp, *pcp_shared;
-    drm_context_t hwContext;
-    __DRIcontext *shared = NULL;
-    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
-
-    if (!psc || !psc->driScreen)
-	return NULL;
-
-    if (shareList) {
-	pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext;
-	shared = pcp_shared->driContext;
-    }
-
-    pcp = Xmalloc(sizeof *pcp);
-    if (pcp == NULL)
-	return NULL;
-
-    pcp->psc = psc;
-    if (!XF86DRICreateContextWithConfig(psc->dpy, psc->scr,
-					mode->visualID,
-					&pcp->hwContextID, &hwContext)) {
-	Xfree(pcp);
-	return NULL;
-    }
-
-    pcp->driContext =
-        (*psc->legacy->createNewContext)(psc->__driScreen,
-					 config->driConfig,
-					 renderType,
-					 shared,
-					 hwContext,
-					 pcp);
-    if (pcp->driContext == NULL) {
-	XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID);
-	Xfree(pcp);
-	return NULL;
-    }
-
-    pcp->base.destroyContext = driDestroyContext;
-    pcp->base.bindContext = driBindContext;
-    pcp->base.unbindContext = driUnbindContext;
-
-    return &pcp->base;
+   __GLXDRIcontextPrivate *pcp, *pcp_shared;
+   drm_context_t hwContext;
+   __DRIcontext *shared = NULL;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
+
+   if (!psc || !psc->driScreen)
+      return NULL;
+
+   if (shareList) {
+      pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext;
+      shared = pcp_shared->driContext;
+   }
+
+   pcp = Xmalloc(sizeof *pcp);
+   if (pcp == NULL)
+      return NULL;
+
+   pcp->psc = psc;
+   if (!XF86DRICreateContextWithConfig(psc->dpy, psc->scr,
+                                       mode->visualID,
+                                       &pcp->hwContextID, &hwContext)) {
+      Xfree(pcp);
+      return NULL;
+   }
+
+   pcp->driContext =
+      (*psc->legacy->createNewContext) (psc->__driScreen,
+                                        config->driConfig,
+                                        renderType, shared, hwContext, pcp);
+   if (pcp->driContext == NULL) {
+      XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID);
+      Xfree(pcp);
+      return NULL;
+   }
+
+   pcp->base.destroyContext = driDestroyContext;
+   pcp->base.bindContext = driBindContext;
+   pcp->base.unbindContext = driUnbindContext;
+
+   return &pcp->base;
 }
 
-static void driDestroyDrawable(__GLXDRIdrawable *pdraw)
+static void
+driDestroyDrawable(__GLXDRIdrawable * pdraw)
 {
-    __GLXscreenConfigs *psc = pdraw->psc;
+   __GLXscreenConfigs *psc = pdraw->psc;
 
-    (*psc->core->destroyDrawable)(pdraw->driDrawable);
-    XF86DRIDestroyDrawable(psc->dpy, psc->scr, pdraw->drawable);
-    Xfree(pdraw);
+   (*psc->core->destroyDrawable) (pdraw->driDrawable);
+   XF86DRIDestroyDrawable(psc->dpy, psc->scr, pdraw->drawable);
+   Xfree(pdraw);
 }
 
-static __GLXDRIdrawable *driCreateDrawable(__GLXscreenConfigs *psc,
-					   XID xDrawable,
-					   GLXDrawable drawable,
-					   const __GLcontextModes *modes)
+static __GLXDRIdrawable *
+driCreateDrawable(__GLXscreenConfigs * psc,
+                  XID xDrawable,
+                  GLXDrawable drawable, const __GLcontextModes * modes)
 {
-    __GLXDRIdrawable *pdraw;
-    drm_drawable_t hwDrawable;
-    void *empty_attribute_list = NULL;
-    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
-
-    /* Old dri can't handle GLX 1.3+ drawable constructors. */
-    if (xDrawable != drawable)
-	return NULL;
-
-    pdraw = Xmalloc(sizeof(*pdraw));
-    if (!pdraw)
-	return NULL;
-
-    pdraw->drawable = drawable;
-    pdraw->psc = psc;
-
-    if (!XF86DRICreateDrawable(psc->dpy, psc->scr, drawable, &hwDrawable))
-	return NULL;
-
-    /* Create a new drawable */
-    pdraw->driDrawable =
-	(*psc->legacy->createNewDrawable)(psc->__driScreen,
-					  config->driConfig,
-					  hwDrawable,
-					  GLX_WINDOW_BIT,
-					  empty_attribute_list,
-					  pdraw);
-
-    if (!pdraw->driDrawable) {
-	XF86DRIDestroyDrawable(psc->dpy, psc->scr, drawable);
-	Xfree(pdraw);
-	return NULL;
-    }
-
-    pdraw->destroyDrawable = driDestroyDrawable;
-
-    return pdraw;
+   __GLXDRIdrawable *pdraw;
+   drm_drawable_t hwDrawable;
+   void *empty_attribute_list = NULL;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
+
+   /* Old dri can't handle GLX 1.3+ drawable constructors. */
+   if (xDrawable != drawable)
+      return NULL;
+
+   pdraw = Xmalloc(sizeof(*pdraw));
+   if (!pdraw)
+      return NULL;
+
+   pdraw->drawable = drawable;
+   pdraw->psc = psc;
+
+   if (!XF86DRICreateDrawable(psc->dpy, psc->scr, drawable, &hwDrawable))
+      return NULL;
+
+   /* Create a new drawable */
+   pdraw->driDrawable =
+      (*psc->legacy->createNewDrawable) (psc->__driScreen,
+                                         config->driConfig,
+                                         hwDrawable,
+                                         GLX_WINDOW_BIT,
+                                         empty_attribute_list, pdraw);
+
+   if (!pdraw->driDrawable) {
+      XF86DRIDestroyDrawable(psc->dpy, psc->scr, drawable);
+      Xfree(pdraw);
+      return NULL;
+   }
+
+   pdraw->destroyDrawable = driDestroyDrawable;
+
+   return pdraw;
 }
 
-static void driSwapBuffers(__GLXDRIdrawable *pdraw)
+static void
+driSwapBuffers(__GLXDRIdrawable * pdraw)
 {
-   (*pdraw->psc->core->swapBuffers)(pdraw->driDrawable);
+   (*pdraw->psc->core->swapBuffers) (pdraw->driDrawable);
 }
 
-static void driCopySubBuffer(__GLXDRIdrawable *pdraw,
-			     int x, int y, int width, int height)
+static void
+driCopySubBuffer(__GLXDRIdrawable * pdraw,
+                 int x, int y, int width, int height)
 {
-    (*pdraw->psc->driCopySubBuffer->copySubBuffer)(pdraw->driDrawable,
-						   x, y, width, height);
+   (*pdraw->psc->driCopySubBuffer->copySubBuffer) (pdraw->driDrawable,
+                                                   x, y, width, height);
 }
 
-static void driDestroyScreen(__GLXscreenConfigs *psc)
+static void
+driDestroyScreen(__GLXscreenConfigs * psc)
 {
-    /* Free the direct rendering per screen data */
-    if (psc->__driScreen)
-	(*psc->core->destroyScreen)(psc->__driScreen);
-    psc->__driScreen = NULL;
-    if (psc->driver)
-	dlclose(psc->driver);
+   /* Free the direct rendering per screen data */
+   if (psc->__driScreen)
+      (*psc->core->destroyScreen) (psc->__driScreen);
+   psc->__driScreen = NULL;
+   if (psc->driver)
+      dlclose(psc->driver);
 }
 
-static __GLXDRIscreen *driCreateScreen(__GLXscreenConfigs *psc, int screen,
-				       __GLXdisplayPrivate *priv)
+static __GLXDRIscreen *
+driCreateScreen(__GLXscreenConfigs * psc, int screen,
+                __GLXdisplayPrivate * priv)
 {
-    __GLXDRIdisplayPrivate *pdp;
-    __GLXDRIscreen *psp;
-    const __DRIextension **extensions;
-    char *driverName;
-    int i;
-
-    psp = Xcalloc(1, sizeof *psp);
-    if (psp == NULL)
-	return NULL;
-
-    /* Initialize per screen dynamic client GLX extensions */
-    psc->ext_list_first_time = GL_TRUE;
-
-    if (!driGetDriverName(priv->dpy, screen, &driverName)) {
-	Xfree(psp);
-	return NULL;
-    }
-
-    psc->driver = driOpenDriver(driverName);
-    Xfree(driverName);
-    if (psc->driver == NULL) {
-	Xfree(psp);
-	return NULL;
-    }
-
-    extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS);
-    if (extensions == NULL) {
- 	ErrorMessageF("driver exports no extensions (%s)\n", dlerror());
-	Xfree(psp);
-	return NULL;
-    }
-
-    for (i = 0; extensions[i]; i++) {
-	if (strcmp(extensions[i]->name, __DRI_CORE) == 0)
-	    psc->core = (__DRIcoreExtension *) extensions[i];
-	if (strcmp(extensions[i]->name, __DRI_LEGACY) == 0)
-	    psc->legacy = (__DRIlegacyExtension *) extensions[i];
-    }
-
-    if (psc->core == NULL || psc->legacy == NULL) {
-	Xfree(psp);
-	return NULL;
-    }
-  
-    pdp = (__GLXDRIdisplayPrivate *) priv->driDisplay;
-    psc->__driScreen =
- 	CallCreateNewScreen(psc->dpy, screen, psc, pdp);
-    if (psc->__driScreen == NULL) {
- 	dlclose(psc->driver);
- 	Xfree(psp);
- 	return NULL;
-    }
-
-    driBindExtensions(psc, 0);
-    if (psc->driCopySubBuffer)
-	psp->copySubBuffer = driCopySubBuffer;
-
-    psp->destroyScreen = driDestroyScreen;
-    psp->createContext = driCreateContext;
-    psp->createDrawable = driCreateDrawable;
-    psp->swapBuffers = driSwapBuffers;
-    psp->waitX = NULL;
-    psp->waitGL = NULL;
-
-    return psp;
+   __GLXDRIdisplayPrivate *pdp;
+   __GLXDRIscreen *psp;
+   const __DRIextension **extensions;
+   char *driverName;
+   int i;
+
+   psp = Xcalloc(1, sizeof *psp);
+   if (psp == NULL)
+      return NULL;
+
+   /* Initialize per screen dynamic client GLX extensions */
+   psc->ext_list_first_time = GL_TRUE;
+
+   if (!driGetDriverName(priv->dpy, screen, &driverName)) {
+      Xfree(psp);
+      return NULL;
+   }
+
+   psc->driver = driOpenDriver(driverName);
+   Xfree(driverName);
+   if (psc->driver == NULL) {
+      Xfree(psp);
+      return NULL;
+   }
+
+   extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS);
+   if (extensions == NULL) {
+      ErrorMessageF("driver exports no extensions (%s)\n", dlerror());
+      Xfree(psp);
+      return NULL;
+   }
+
+   for (i = 0; extensions[i]; i++) {
+      if (strcmp(extensions[i]->name, __DRI_CORE) == 0)
+         psc->core = (__DRIcoreExtension *) extensions[i];
+      if (strcmp(extensions[i]->name, __DRI_LEGACY) == 0)
+         psc->legacy = (__DRIlegacyExtension *) extensions[i];
+   }
+
+   if (psc->core == NULL || psc->legacy == NULL) {
+      Xfree(psp);
+      return NULL;
+   }
+
+   pdp = (__GLXDRIdisplayPrivate *) priv->driDisplay;
+   psc->__driScreen = CallCreateNewScreen(psc->dpy, screen, psc, pdp);
+   if (psc->__driScreen == NULL) {
+      dlclose(psc->driver);
+      Xfree(psp);
+      return NULL;
+   }
+
+   driBindExtensions(psc, 0);
+   if (psc->driCopySubBuffer)
+      psp->copySubBuffer = driCopySubBuffer;
+
+   psp->destroyScreen = driDestroyScreen;
+   psp->createContext = driCreateContext;
+   psp->createDrawable = driCreateDrawable;
+   psp->swapBuffers = driSwapBuffers;
+   psp->waitX = NULL;
+   psp->waitGL = NULL;
+
+   return psp;
 }
 
 /* Called from __glXFreeDisplayPrivate.
  */
-static void driDestroyDisplay(__GLXDRIdisplay *dpy)
+static void
+driDestroyDisplay(__GLXDRIdisplay * dpy)
 {
-    Xfree(dpy);
+   Xfree(dpy);
 }
 
 /*
@@ -712,33 +726,34 @@ static void driDestroyDisplay(__GLXDRIdisplay *dpy)
  * This is called from __glXInitialize() when we are given a new
  * display pointer.
  */
-_X_HIDDEN __GLXDRIdisplay *driCreateDisplay(Display *dpy)
+_X_HIDDEN __GLXDRIdisplay *
+driCreateDisplay(Display * dpy)
 {
-    __GLXDRIdisplayPrivate *pdpyp;
-    int eventBase, errorBase;
-    int major, minor, patch;
+   __GLXDRIdisplayPrivate *pdpyp;
+   int eventBase, errorBase;
+   int major, minor, patch;
 
-    if (!XF86DRIQueryExtension(dpy, &eventBase, &errorBase)) {
-	return NULL;
-    }
+   if (!XF86DRIQueryExtension(dpy, &eventBase, &errorBase)) {
+      return NULL;
+   }
 
-    if (!XF86DRIQueryVersion(dpy, &major, &minor, &patch)) {
-	return NULL;
-    }
+   if (!XF86DRIQueryVersion(dpy, &major, &minor, &patch)) {
+      return NULL;
+   }
 
-    pdpyp = Xmalloc(sizeof *pdpyp);
-    if (!pdpyp) {
-	return NULL;
-    }
+   pdpyp = Xmalloc(sizeof *pdpyp);
+   if (!pdpyp) {
+      return NULL;
+   }
 
-    pdpyp->driMajor = major;
-    pdpyp->driMinor = minor;
-    pdpyp->driPatch = patch;
+   pdpyp->driMajor = major;
+   pdpyp->driMinor = minor;
+   pdpyp->driPatch = patch;
 
-    pdpyp->base.destroyDisplay = driDestroyDisplay;
-    pdpyp->base.createScreen = driCreateScreen;
+   pdpyp->base.destroyDisplay = driDestroyDisplay;
+   pdpyp->base.createScreen = driCreateScreen;
 
-    return &pdpyp->base;
+   return &pdpyp->base;
 }
 
 #endif /* GLX_DIRECT_RENDERING */
diff --git a/src/glx/x11/glx_pbuffer.c b/src/glx/x11/glx_pbuffer.c
index a06331fd7f..88a17df6f8 100644
--- a/src/glx/x11/glx_pbuffer.c
+++ b/src/glx/x11/glx_pbuffer.c
@@ -25,7 +25,7 @@
 /**
  * \file glx_pbuffer.c
  * Implementation of pbuffer related functions.
- * 
+ *
  * \author Ian Romanick <idr@us.ibm.com>
  */
 
@@ -114,7 +114,7 @@ ChangeDrawableAttribute(Display * dpy, GLXDrawable drawable,
  * \note
  * This function dynamically determines whether to use the SGIX_pbuffer
  * version of the protocol or the GLX 1.3 version of the protocol.
- * 
+ *
  * \todo
  * This function needs to be modified to work with direct-rendering drivers.
  */
@@ -198,7 +198,7 @@ determineTextureFormat(const int *attribs, int numAttribs)
 
    for (i = 0; i < numAttribs; i++) {
       if (attribs[2 * i] == GLX_TEXTURE_FORMAT_EXT)
-	 return attribs[2 * i + 1];
+         return attribs[2 * i + 1];
    }
 
    return 0;
@@ -681,8 +681,7 @@ GLX_ALIAS_VOID(glXDestroyGLXPbufferSGIX,
 PUBLIC
 GLX_ALIAS_VOID(glXSelectEventSGIX,
                (Display * dpy, GLXDrawable drawable,
-                unsigned long mask), (dpy, drawable, mask),
-               glXSelectEvent)
+                unsigned long mask), (dpy, drawable, mask), glXSelectEvent)
 
 PUBLIC
 GLX_ALIAS_VOID(glXGetSelectedEventSGIX,
diff --git a/src/glx/x11/glx_query.c b/src/glx/x11/glx_query.c
index 2789b841b1..efad13d376 100644
--- a/src/glx/x11/glx_query.c
+++ b/src/glx/x11/glx_query.c
@@ -25,7 +25,7 @@
 /**
  * \file glx_query.c
  * Generic utility functions to query internal data from the server.
- * 
+ *
  * \author Ian Romanick <idr@us.ibm.com>
  */
 
@@ -43,13 +43,10 @@
  * Exchange a protocol request for glXQueryServerString.
  */
 char *
-__glXQueryServerString(Display* dpy,
-                       int opcode,
-                       CARD32 screen,
-                       CARD32 name)
+__glXQueryServerString(Display * dpy, int opcode, CARD32 screen, CARD32 name)
 {
    xcb_connection_t *c = XGetXCBConnection(dpy);
-   xcb_glx_query_server_string_reply_t* reply =
+   xcb_glx_query_server_string_reply_t *reply =
       xcb_glx_query_server_string_reply(c,
                                         xcb_glx_query_server_string(c,
                                                                     screen,
@@ -59,7 +56,7 @@ __glXQueryServerString(Display* dpy,
    /* The spec doesn't mention this, but the Xorg server replies with
     * a string already terminated with '\0'. */
    uint32_t len = xcb_glx_query_server_string_string_length(reply);
-   char* buf = Xmalloc(len);
+   char *buf = Xmalloc(len);
    memcpy(buf, xcb_glx_query_server_string_string(reply), len);
    free(reply);
 
@@ -70,23 +67,20 @@ __glXQueryServerString(Display* dpy,
  * Exchange a protocol request for glGetString.
  */
 char *
-__glXGetString(Display* dpy,
-               int opcode,
-               CARD32 contextTag,
-               CARD32 name)
+__glXGetString(Display * dpy, int opcode, CARD32 contextTag, CARD32 name)
 {
    xcb_connection_t *c = XGetXCBConnection(dpy);
-   xcb_glx_get_string_reply_t* reply =
-      xcb_glx_get_string_reply(c,
-                               xcb_glx_get_string(c,
-                                                  contextTag,
-                                                  name),
-                               NULL);
+   xcb_glx_get_string_reply_t *reply = xcb_glx_get_string_reply(c,
+                                                                xcb_glx_get_string
+                                                                (c,
+                                                                 contextTag,
+                                                                 name),
+                                                                NULL);
 
    /* The spec doesn't mention this, but the Xorg server replies with
     * a string already terminated with '\0'. */
    uint32_t len = xcb_glx_get_string_string_length(reply);
-   char* buf = Xmalloc(len);
+   char *buf = Xmalloc(len);
    memcpy(buf, xcb_glx_get_string_string(reply), len);
    free(reply);
 
@@ -97,7 +91,7 @@ __glXGetString(Display* dpy,
 
 /**
  * GLX protocol structure for the ficticious "GXLGenericGetString" request.
- * 
+ *
  * This is a non-existant protocol packet.  It just so happens that all of
  * the real protocol packets used to request a string from the server have
  * an identical binary layout.  The only difference between them is the
@@ -167,25 +161,17 @@ __glXGetStringFromServer(Display * dpy, int opcode, CARD32 glxCode,
 }
 
 char *
-__glXQueryServerString(Display* dpy,
-                       int opcode,
-                       CARD32 screen,
-                       CARD32 name)
+__glXQueryServerString(Display * dpy, int opcode, CARD32 screen, CARD32 name)
 {
    return __glXGetStringFromServer(dpy, opcode,
-                                   X_GLXQueryServerString,
-                                   screen, name);
+                                   X_GLXQueryServerString, screen, name);
 }
 
 char *
-__glXGetString(Display* dpy,
-               int opcode,
-               CARD32 contextTag,
-               CARD32 name)
+__glXGetString(Display * dpy, int opcode, CARD32 contextTag, CARD32 name)
 {
    return __glXGetStringFromServer(dpy, opcode, X_GLsop_GetString,
                                    contextTag, name);
 }
 
 #endif /* USE_XCB */
-
diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h
index 2778ad878e..00ee14fb05 100644
--- a/src/glx/x11/glxclient.h
+++ b/src/glx/x11/glxclient.h
@@ -55,9 +55,7 @@
 #include "GL/internal/glcore.h"
 #include "glapi/glapitable.h"
 #include "glxhash.h"
-#if defined( USE_XTHREADS )
-# include <X11/Xthreads.h>
-#elif defined( PTHREADS )
+#if defined( PTHREADS )
 # include <pthread.h>
 #endif
 
@@ -67,7 +65,7 @@
 /* If we build the library with gcc's -fvisibility=hidden flag, we'll
  * use the PUBLIC macro to mark functions that are to be exported.
  *
- * We also need to define a USED attribute, so the optimizer doesn't 
+ * We also need to define a USED attribute, so the optimizer doesn't
  * inline a static function that we later use in an alias. - ajax
  */
 #if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 303
@@ -80,8 +78,8 @@
 
 
 
-#define GLX_MAJOR_VERSION	1	/* current version numbers */
-#define GLX_MINOR_VERSION	4
+#define GLX_MAJOR_VERSION 1       /* current version numbers */
+#define GLX_MINOR_VERSION 4
 
 #define __GLX_MAX_TEXTURE_UNITS 32
 
@@ -95,7 +93,7 @@ typedef struct _glapi_table __GLapi;
 
 #ifdef GLX_DIRECT_RENDERING
 
-#define containerOf(ptr, type, member)			\
+#define containerOf(ptr, type, member)              \
     (type *)( (char *)ptr - offsetof(type,member) )
 
 #include <GL/internal/dri_interface.h>
@@ -112,75 +110,78 @@ typedef struct __GLXDRIcontextRec __GLXDRIcontext;
 
 #include "glxextensions.h"
 
-struct __GLXDRIdisplayRec {
+struct __GLXDRIdisplayRec
+{
     /**
      * Method to destroy the private DRI display data.
      */
-    void (*destroyDisplay)(__GLXDRIdisplay *display);
+   void (*destroyDisplay) (__GLXDRIdisplay * display);
 
-    __GLXDRIscreen *(*createScreen)(__GLXscreenConfigs *psc, int screen,
-				    __GLXdisplayPrivate *priv);
+   __GLXDRIscreen *(*createScreen) (__GLXscreenConfigs * psc, int screen,
+                                    __GLXdisplayPrivate * priv);
 };
 
-struct __GLXDRIscreenRec {
-
-    void (*destroyScreen)(__GLXscreenConfigs *psc);
-
-    __GLXDRIcontext *(*createContext)(__GLXscreenConfigs *psc,
-				      const __GLcontextModes *mode,
-				      GLXContext gc,
-				      GLXContext shareList, int renderType);
-	
-    __GLXDRIdrawable *(*createDrawable)(__GLXscreenConfigs *psc,
-					XID drawable,
-					GLXDrawable glxDrawable,
-					const __GLcontextModes *modes);
-
-    void (*swapBuffers)(__GLXDRIdrawable *pdraw);
-    void (*copySubBuffer)(__GLXDRIdrawable *pdraw,
-			  int x, int y, int width, int height);
-    void (*waitX)(__GLXDRIdrawable *pdraw);
-    void (*waitGL)(__GLXDRIdrawable *pdraw);
+struct __GLXDRIscreenRec
+{
+
+   void (*destroyScreen) (__GLXscreenConfigs * psc);
+
+   __GLXDRIcontext *(*createContext) (__GLXscreenConfigs * psc,
+                                      const __GLcontextModes * mode,
+                                      GLXContext gc,
+                                      GLXContext shareList, int renderType);
+
+   __GLXDRIdrawable *(*createDrawable) (__GLXscreenConfigs * psc,
+                                        XID drawable,
+                                        GLXDrawable glxDrawable,
+                                        const __GLcontextModes * modes);
+
+   void (*swapBuffers) (__GLXDRIdrawable * pdraw);
+   void (*copySubBuffer) (__GLXDRIdrawable * pdraw,
+                          int x, int y, int width, int height);
+   void (*waitX) (__GLXDRIdrawable * pdraw);
+   void (*waitGL) (__GLXDRIdrawable * pdraw);
 };
 
-struct __GLXDRIcontextRec {
-    void (*destroyContext)(__GLXDRIcontext *context, __GLXscreenConfigs *psc,
-			   Display *dpy);
-    Bool (*bindContext)(__GLXDRIcontext *context,
-			__GLXDRIdrawable *pdraw,
-			__GLXDRIdrawable *pread);
-    
-    void (*unbindContext)(__GLXDRIcontext *context);
+struct __GLXDRIcontextRec
+{
+   void (*destroyContext) (__GLXDRIcontext * context,
+                           __GLXscreenConfigs * psc, Display * dpy);
+     Bool(*bindContext) (__GLXDRIcontext * context, __GLXDRIdrawable * pdraw,
+                         __GLXDRIdrawable * pread);
+
+   void (*unbindContext) (__GLXDRIcontext * context);
 };
 
-struct __GLXDRIdrawableRec {
-    void (*destroyDrawable)(__GLXDRIdrawable *drawable);
+struct __GLXDRIdrawableRec
+{
+   void (*destroyDrawable) (__GLXDRIdrawable * drawable);
 
-    XID xDrawable;
-    XID drawable;
-    __GLXscreenConfigs *psc;
-    GLenum textureTarget;
-    __DRIdrawable *driDrawable;
-    GLenum textureFormat; /* EXT_texture_from_pixmap support */
+   XID xDrawable;
+   XID drawable;
+   __GLXscreenConfigs *psc;
+   GLenum textureTarget;
+   __DRIdrawable *driDrawable;
+   GLenum textureFormat;        /* EXT_texture_from_pixmap support */
 };
 
 /*
 ** Function to create and DRI display data and initialize the display
 ** dependent methods.
 */
-extern __GLXDRIdisplay *driswCreateDisplay(Display *dpy);
-extern __GLXDRIdisplay *driCreateDisplay(Display *dpy);
-extern __GLXDRIdisplay *dri2CreateDisplay(Display *dpy);
+extern __GLXDRIdisplay *driswCreateDisplay(Display * dpy);
+extern __GLXDRIdisplay *driCreateDisplay(Display * dpy);
+extern __GLXDRIdisplay *dri2CreateDisplay(Display * dpy);
 
-extern void DRI_glXUseXFont( Font font, int first, int count, int listbase );
+extern void DRI_glXUseXFont(Font font, int first, int count, int listbase);
 
 /*
 ** Functions to obtain driver configuration information from a direct
 ** rendering client application
 */
-extern const char *glXGetScreenDriver (Display *dpy, int scrNum);
+extern const char *glXGetScreenDriver(Display * dpy, int scrNum);
 
-extern const char *glXGetDriverConfig (const char *driverName);
+extern const char *glXGetDriverConfig(const char *driverName);
 
 #endif
 
@@ -188,53 +189,57 @@ extern const char *glXGetDriverConfig (const char *driverName);
 
 #define __GL_CLIENT_ATTRIB_STACK_DEPTH 16
 
-typedef struct __GLXpixelStoreModeRec {
-    GLboolean swapEndian;
-    GLboolean lsbFirst;
-    GLuint rowLength;
-    GLuint imageHeight;
-    GLuint imageDepth;
-    GLuint skipRows;
-    GLuint skipPixels;
-    GLuint skipImages;
-    GLuint alignment;
+typedef struct __GLXpixelStoreModeRec
+{
+   GLboolean swapEndian;
+   GLboolean lsbFirst;
+   GLuint rowLength;
+   GLuint imageHeight;
+   GLuint imageDepth;
+   GLuint skipRows;
+   GLuint skipPixels;
+   GLuint skipImages;
+   GLuint alignment;
 } __GLXpixelStoreMode;
 
 
-typedef struct __GLXattributeRec {
-    GLuint mask;
+typedef struct __GLXattributeRec
+{
+   GLuint mask;
 
     /**
      * Pixel storage state.  Most of the pixel store mode state is kept
      * here and used by the client code to manage the packing and
      * unpacking of data sent to/received from the server.
      */
-    __GLXpixelStoreMode storePack, storeUnpack;
+   __GLXpixelStoreMode storePack, storeUnpack;
 
     /**
      * Is EXT_vertex_array / GL 1.1 DrawArrays protocol specifically
      * disabled?
      */
-    GLboolean NoDrawArraysProtocol;
-    
+   GLboolean NoDrawArraysProtocol;
+
     /**
      * Vertex Array storage state.  The vertex array component
      * state is stored here and is used to manage the packing of
      * DrawArrays data sent to the server.
      */
-    struct array_state_vector * array_state;
+   struct array_state_vector *array_state;
 } __GLXattribute;
 
-typedef struct __GLXattributeMachineRec {
-	__GLXattribute *stack[__GL_CLIENT_ATTRIB_STACK_DEPTH];
-	__GLXattribute **stackPointer;
+typedef struct __GLXattributeMachineRec
+{
+   __GLXattribute *stack[__GL_CLIENT_ATTRIB_STACK_DEPTH];
+   __GLXattribute **stackPointer;
 } __GLXattributeMachine;
 
 /**
  * GLX state that needs to be kept on the client.  One of these records
  * exist for each context that has been made current by this client.
  */
-struct __GLXcontextRec {
+struct __GLXcontextRec
+{
     /**
      * \name Drawing command buffer.
      *
@@ -251,13 +256,13 @@ struct __GLXcontextRec {
      * These must be the first 6 fields since they are static initialized
      * in the dummy context in glxext.c
      */
-    /*@{*/
-    GLubyte *buf;
-    GLubyte *pc;
-    GLubyte *limit;
-    GLubyte *bufEnd;
-    GLint bufSize;
-    /*@}*/
+   /*@{ */
+   GLubyte *buf;
+   GLubyte *pc;
+   GLubyte *limit;
+   GLubyte *bufEnd;
+   GLint bufSize;
+   /*@} */
 
     /**
      * The XID of this rendering context.  When the context is created a
@@ -265,24 +270,24 @@ struct __GLXcontextRec {
      * destroyed but is still current to some thread. In this case the
      * context will be freed on next MakeCurrent.
      */
-    XID xid;
+   XID xid;
 
     /**
      * The XID of the \c shareList context.
      */
-    XID share_xid;
+   XID share_xid;
 
     /**
      * Screen number.
      */
-    GLint screen;
-    __GLXscreenConfigs *psc;
+   GLint screen;
+   __GLXscreenConfigs *psc;
 
     /**
      * \c GL_TRUE if the context was created with ImportContext, which
      * means the server-side context was created by another X client.
      */
-    GLboolean imported;
+   GLboolean imported;
 
     /**
      * The context tag returned by MakeCurrent when this context is made
@@ -292,7 +297,7 @@ struct __GLXcontextRec {
      * \c WaitX, \c WaitGL, \c UseXFont, and \c MakeCurrent (for the old
      * context)).
      */
-    GLXContextTag currentContextTag;
+   GLXContextTag currentContextTag;
 
     /**
      * \name Rendering mode
@@ -301,11 +306,11 @@ struct __GLXcontextRec {
      * When \c glRenderMode is called, the buffer associated with the
      * previous rendering mode (feedback or select) is filled.
      */
-    /*@{*/
-    GLenum renderMode;
-    GLfloat *feedbackBuf;
-    GLuint *selectBuf;
-    /*@}*/
+   /*@{ */
+   GLenum renderMode;
+   GLfloat *feedbackBuf;
+   GLuint *selectBuf;
+   /*@} */
 
     /**
      * This is \c GL_TRUE if the pixel unpack modes are such that an image
@@ -313,43 +318,43 @@ struct __GLXcontextRec {
      * still be true that the server will have to do some work.  This
      * just promises that a straight copy will fetch the correct bytes.
      */
-    GLboolean fastImageUnpack;
+   GLboolean fastImageUnpack;
 
     /**
      * Fill newImage with the unpacked form of \c oldImage getting it
      * ready for transport to the server.
      */
-    void (*fillImage)(__GLXcontext*, GLint, GLint, GLint, GLint, GLenum,
-		      GLenum, const GLvoid*, GLubyte*, GLubyte*);
+   void (*fillImage) (__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+                      GLenum, const GLvoid *, GLubyte *, GLubyte *);
 
     /**
      * Client side attribs.
      */
-    __GLXattributeMachine attributes;
+   __GLXattributeMachine attributes;
 
     /**
      * Client side error code.  This is set when client side gl API
      * routines need to set an error because of a bad enumerant or
      * running out of memory, etc.
      */
-    GLenum error;
+   GLenum error;
 
     /**
      * Whether this context does direct rendering.
      */
-    Bool isDirect;
+   Bool isDirect;
 
     /**
      * \c dpy of current display for this context.  Will be \c NULL if not
      * current to any display, or if this is the "dummy context".
      */
-    Display *currentDpy;
+   Display *currentDpy;
 
     /**
      * The current drawable for this context.  Will be None if this
      * context is not current to any drawable.  currentReadable is below.
      */
-    GLXDrawable currentDrawable;
+   GLXDrawable currentDrawable;
 
     /**
      * \name GL Constant Strings
@@ -358,38 +363,38 @@ struct __GLXcontextRec {
      * These pertain to GL attributes, not to be confused with
      * GLX versioning attributes.
      */
-    /*@{*/
-    GLubyte *vendor;
-    GLubyte *renderer;
-    GLubyte *version;
-    GLubyte *extensions;
-    /*@}*/
+   /*@{ */
+   GLubyte *vendor;
+   GLubyte *renderer;
+   GLubyte *version;
+   GLubyte *extensions;
+   /*@} */
 
     /**
      * Record the dpy this context was created on for later freeing
      */
-    Display *createDpy;
+   Display *createDpy;
 
     /**
      * Maximum small render command size.  This is the smaller of 64k and
      * the size of the above buffer.
      */
-    GLint maxSmallRenderCommandSize;
+   GLint maxSmallRenderCommandSize;
 
     /**
      * Major opcode for the extension.  Copied here so a lookup isn't
      * needed.
      */
-    GLint majorOpcode;
+   GLint majorOpcode;
 
     /**
      * Pointer to the mode used to create this context.
      */
-    const __GLcontextModes * mode;
+   const __GLcontextModes *mode;
 
 #ifdef GLX_DIRECT_RENDERING
-    __GLXDRIcontext *driContext;
-    __DRIcontext *__driContext;
+   __GLXDRIcontext *driContext;
+   __DRIcontext *__driContext;
 #endif
 
     /**
@@ -398,22 +403,22 @@ struct __GLXcontextRec {
      *
      * \since Internal API version 20030606.
      */
-    GLXDrawable currentReadable;
+   GLXDrawable currentReadable;
 
-   /** 
+   /**
     * Pointer to client-state data that is private to libGL.  This is only
     * used for indirect rendering contexts.
     *
     * No internal API version change was made for this change.  Client-side
     * drivers should NEVER use this data or even care that it exists.
     */
-   void * client_state_private;
+   void *client_state_private;
 
    /**
     * Stored value for \c glXQueryContext attribute \c GLX_RENDER_TYPE.
     */
    int renderType;
-    
+
    /**
     * \name Raw server GL version
     *
@@ -421,23 +426,23 @@ struct __GLXcontextRec {
     * returned by the server, and it may not reflect what is actually
     * supported (or reported) by the client-side library.
     */
-    /*@{*/
+   /*@{ */
    int server_major;        /**< Major version number. */
    int server_minor;        /**< Minor version number. */
-    /*@}*/
+   /*@} */
 
    /**
     * Thread ID we're currently current in. Zero if none.
     */
    unsigned long thread_id;
 
-    char gl_extension_bits[ __GL_EXT_BYTES ];
+   char gl_extension_bits[__GL_EXT_BYTES];
 };
 
-#define __glXSetError(gc,code) \
-    if (!(gc)->error) {	       \
-	(gc)->error = code;    \
-    }
+#define __glXSetError(gc,code)  \
+   if (!(gc)->error) {          \
+      (gc)->error = code;       \
+   }
 
 extern void __glFreeAttributeState(__GLXcontext *);
 
@@ -448,7 +453,7 @@ extern void __glFreeAttributeState(__GLXcontext *);
  * that will use the GLXRender GLX command.  In this case it is
  * \c glPolygonStipple.
  */
-#define __GLX_MAX_SMALL_RENDER_CMD_SIZE	156
+#define __GLX_MAX_SMALL_RENDER_CMD_SIZE 156
 
 /**
  * To keep the implementation fast, the code uses a "limit" pointer
@@ -460,75 +465,76 @@ extern void __glFreeAttributeState(__GLXcontext *);
  * efficacy of the buffer.  The "+32" is just to keep the code working
  * in case somebody counts wrong.
  */
-#define __GLX_BUFFER_LIMIT_SIZE	(__GLX_MAX_SMALL_RENDER_CMD_SIZE + 32)
+#define __GLX_BUFFER_LIMIT_SIZE (__GLX_MAX_SMALL_RENDER_CMD_SIZE + 32)
 
 /**
  * This implementation uses a smaller threshold for switching
  * to the RenderLarge protocol than the protcol requires so that
  * large copies don't occur.
  */
-#define __GLX_RENDER_CMD_SIZE_LIMIT	4096
+#define __GLX_RENDER_CMD_SIZE_LIMIT 4096
 
 /**
  * One of these records exists per screen of the display.  It contains
  * a pointer to the config data for that screen (if the screen supports GL).
  */
-struct __GLXscreenConfigsRec {
+struct __GLXscreenConfigsRec
+{
     /**
      * GLX extension string reported by the X-server.
      */
-    const char *serverGLXexts;
+   const char *serverGLXexts;
 
     /**
      * GLX extension string to be reported to applications.  This is the
      * set of extensions that the application can actually use.
      */
-    char *effectiveGLXexts;
+   char *effectiveGLXexts;
 
 #ifdef GLX_DIRECT_RENDERING
     /**
      * Per screen direct rendering interface functions and data.
      */
-    __DRIscreen *__driScreen;
-    const __DRIcoreExtension *core;
-    const __DRIlegacyExtension *legacy;
-    const __DRIswrastExtension *swrast;
-    const __DRIdri2Extension *dri2;
-    __glxHashTable *drawHash;
-    Display *dpy;
-    int scr, fd;
-    void *driver;
+   __DRIscreen *__driScreen;
+   const __DRIcoreExtension *core;
+   const __DRIlegacyExtension *legacy;
+   const __DRIswrastExtension *swrast;
+   const __DRIdri2Extension *dri2;
+   __glxHashTable *drawHash;
+   Display *dpy;
+   int scr, fd;
+   void *driver;
 
-    __GLXDRIscreen *driScreen;
+   __GLXDRIscreen *driScreen;
 
-    const __DRIconfig** driver_configs;
+   const __DRIconfig **driver_configs;
 
 #ifdef __DRI_COPY_SUB_BUFFER
-    const __DRIcopySubBufferExtension *driCopySubBuffer;
+   const __DRIcopySubBufferExtension *driCopySubBuffer;
 #endif
 
 #ifdef __DRI_SWAP_CONTROL
-    const __DRIswapControlExtension *swapControl;
+   const __DRIswapControlExtension *swapControl;
 #endif
 
 #ifdef __DRI_ALLOCATE
-    const __DRIallocateExtension *allocate;
+   const __DRIallocateExtension *allocate;
 #endif
 
 #ifdef __DRI_FRAME_TRACKING
-    const __DRIframeTrackingExtension *frameTracking;
+   const __DRIframeTrackingExtension *frameTracking;
 #endif
 
 #ifdef __DRI_MEDIA_STREAM_COUNTER
-    const __DRImediaStreamCounterExtension *msc;
+   const __DRImediaStreamCounterExtension *msc;
 #endif
 
 #ifdef __DRI_TEX_BUFFER
-    const __DRItexBufferExtension *texBuffer;
+   const __DRItexBufferExtension *texBuffer;
 #endif
 
 #ifdef __DRI2_FLUSH
-    const __DRI2flushExtension *f;
+   const __DRI2flushExtension *f;
 #endif
 
 #endif
@@ -536,7 +542,7 @@ struct __GLXscreenConfigsRec {
     /**
      * Linked list of glx visuals and  fbconfigs for this screen.
      */
-    __GLcontextModes *visuals, *configs;
+   __GLcontextModes *visuals, *configs;
 
     /**
      * Per-screen dynamic GLX extension tracking.  The \c direct_support
@@ -545,10 +551,10 @@ struct __GLXscreenConfigsRec {
      * this field.  The \c __GLXscreenConfigs structure is not used outside
      * libGL.
      */
-    /*@{*/
-    unsigned char direct_support[8];
-    GLboolean ext_list_first_time;
-    /*@}*/
+   /*@{ */
+   unsigned char direct_support[8];
+   GLboolean ext_list_first_time;
+   /*@} */
 
 };
 
@@ -556,67 +562,68 @@ struct __GLXscreenConfigsRec {
  * Per display private data.  One of these records exists for each display
  * that is using the OpenGL (GLX) extension.
  */
-struct __GLXdisplayPrivateRec {
+struct __GLXdisplayPrivateRec
+{
     /**
      * Back pointer to the display
      */
-    Display *dpy;
+   Display *dpy;
 
     /**
      * The \c majorOpcode is common to all connections to the same server.
      * It is also copied into the context structure.
      */
-    int majorOpcode;
+   int majorOpcode;
 
     /**
      * \name Server Version
      *
      * Major and minor version returned by the server during initialization.
      */
-    /*@{*/
-    int majorVersion, minorVersion;
-    /*@}*/
+   /*@{ */
+   int majorVersion, minorVersion;
+   /*@} */
 
     /**
      * \name Storage for the servers GLX vendor and versions strings.
-     * 
+     *
      * These are the same for all screens on this display. These fields will
      * be filled in on demand.
      */
-    /*@{*/
-    const char *serverGLXvendor;
-    const char *serverGLXversion;
-    /*@}*/
+   /*@{ */
+   const char *serverGLXvendor;
+   const char *serverGLXversion;
+   /*@} */
 
     /**
      * Configurations of visuals for all screens on this display.
      * Also, per screen data which now includes the server \c GLX_EXTENSION
      * string.
      */
-    __GLXscreenConfigs *screenConfigs;
+   __GLXscreenConfigs *screenConfigs;
 
 #ifdef GLX_DIRECT_RENDERING
     /**
      * Per display direct rendering interface functions and data.
      */
-    __GLXDRIdisplay *driswDisplay;
-    __GLXDRIdisplay *driDisplay;
-    __GLXDRIdisplay *dri2Display;
+   __GLXDRIdisplay *driswDisplay;
+   __GLXDRIdisplay *driDisplay;
+   __GLXDRIdisplay *dri2Display;
 #endif
 };
 
 
-extern GLubyte *__glXFlushRenderBuffer(__GLXcontext*, GLubyte*);
+extern GLubyte *__glXFlushRenderBuffer(__GLXcontext *, GLubyte *);
 
-extern void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, 
-				GLint totalRequests,
-				const GLvoid * data, GLint dataLen);
+extern void __glXSendLargeChunk(__GLXcontext * gc, GLint requestNumber,
+                                GLint totalRequests,
+                                const GLvoid * data, GLint dataLen);
 
 extern void __glXSendLargeCommand(__GLXcontext *, const GLvoid *, GLint,
-				  const GLvoid *, GLint);
+                                  const GLvoid *, GLint);
 
 /* Initialize the GLX extension for dpy */
-extern __GLXdisplayPrivate *__glXInitialize(Display*);
+extern __GLXdisplayPrivate *__glXInitialize(Display *);
 
 extern void __glXPreferEGL(int state);
 
@@ -625,16 +632,16 @@ extern void __glXPreferEGL(int state);
 extern int __glXDebug;
 
 /* This is per-thread storage in an MT environment */
-#if defined( USE_XTHREADS ) || defined( PTHREADS )
+#if defined( PTHREADS )
 
-extern void __glXSetCurrentContext(__GLXcontext *c);
+extern void __glXSetCurrentContext(__GLXcontext * c);
 
 # if defined( GLX_USE_TLS )
 
-extern __thread void * __glX_tls_Context
-    __attribute__((tls_model("initial-exec")));
+extern __thread void *__glX_tls_Context
+   __attribute__ ((tls_model("initial-exec")));
 
-#  define __glXGetCurrentContext()	__glX_tls_Context
+#  define __glXGetCurrentContext() __glX_tls_Context
 
 # else
 
@@ -645,25 +652,21 @@ extern __GLXcontext *__glXGetCurrentContext(void);
 #else
 
 extern __GLXcontext *__glXcurrentContext;
-#define __glXGetCurrentContext()	__glXcurrentContext
-#define __glXSetCurrentContext(gc)	__glXcurrentContext = gc
+#define __glXGetCurrentContext() __glXcurrentContext
+#define __glXSetCurrentContext(gc) __glXcurrentContext = gc
 
-#endif /* defined( USE_XTHREADS ) || defined( PTHREADS ) */
+#endif /* defined( PTHREADS ) */
 
 extern void __glXSetCurrentContextNull(void);
 
-extern void __glXFreeContext(__GLXcontext*);
+extern void __glXFreeContext(__GLXcontext *);
 
 
 /*
 ** Global lock for all threads in this address space using the GLX
 ** extension
 */
-#if defined( USE_XTHREADS )
-extern xmutex_rec __glXmutex;
-#define __glXLock()    xmutex_lock(&__glXmutex)
-#define __glXUnlock()  xmutex_unlock(&__glXmutex)
-#elif defined( PTHREADS )
+#if defined( PTHREADS )
 extern pthread_mutex_t __glXmutex;
 #define __glXLock()    pthread_mutex_lock(&__glXmutex)
 #define __glXUnlock()  pthread_mutex_unlock(&__glXmutex)
@@ -675,7 +678,7 @@ extern pthread_mutex_t __glXmutex;
 /*
 ** Setup for a command.  Initialize the extension for dpy if necessary.
 */
-extern CARD8 __glXSetupForCommand(Display *dpy);
+extern CARD8 __glXSetupForCommand(Display * dpy);
 
 /************************************************************************/
 
@@ -686,9 +689,11 @@ extern CARD8 __glXSetupForCommand(Display *dpy);
 extern const GLuint __glXDefaultPixelStore[9];
 
 /* Send an image to the server using RenderLarge. */
-extern void __glXSendLargeImage(__GLXcontext *gc, GLint compsize, GLint dim,
-    GLint width, GLint height, GLint depth, GLenum format, GLenum type,
-    const GLvoid *src, GLubyte *pc, GLubyte *modes);
+extern void __glXSendLargeImage(__GLXcontext * gc, GLint compsize, GLint dim,
+                                GLint width, GLint height, GLint depth,
+                                GLenum format, GLenum type,
+                                const GLvoid * src, GLubyte * pc,
+                                GLubyte * modes);
 
 /* Return the size, in bytes, of some pixel data */
 extern GLint __glImageSize(GLint, GLint, GLint, GLenum, GLenum, GLenum);
@@ -708,23 +713,23 @@ extern GLint __glBytesPerElement(GLenum type);
 ** updated to contain the modes needed by the server to decode the
 ** sent data.
 */
-extern void __glFillImage(__GLXcontext*, GLint, GLint, GLint, GLint, GLenum,
-			  GLenum, const GLvoid*, GLubyte*, GLubyte*);
+extern void __glFillImage(__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+                          GLenum, const GLvoid *, GLubyte *, GLubyte *);
 
 /* Copy map data with a stride into a packed buffer */
 extern void __glFillMap1f(GLint, GLint, GLint, const GLfloat *, GLubyte *);
 extern void __glFillMap1d(GLint, GLint, GLint, const GLdouble *, GLubyte *);
 extern void __glFillMap2f(GLint, GLint, GLint, GLint, GLint,
-			  const GLfloat *, GLfloat *);
+                          const GLfloat *, GLfloat *);
 extern void __glFillMap2d(GLint, GLint, GLint, GLint, GLint,
-			  const GLdouble *, GLdouble *);
+                          const GLdouble *, GLdouble *);
 
 /*
 ** Empty an image out of the reply buffer into the clients memory applying
 ** the pack modes to pack back into the clients requested format.
 */
-extern void __glEmptyImage(__GLXcontext*, GLint, GLint, GLint, GLint, GLenum,
-		           GLenum, const GLubyte *, GLvoid *);
+extern void __glEmptyImage(__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+                           GLenum, const GLubyte *, GLvoid *);
 
 
 /*
@@ -737,7 +742,7 @@ extern void __glXFreeVertexArrayState(__GLXcontext *);
 ** Inform the Server of the major and minor numbers and of the client
 ** libraries extension string.
 */
-extern void __glXClientInfo (  Display *dpy, int opcode );
+extern void __glXClientInfo(Display * dpy, int opcode);
 
 /************************************************************************/
 
@@ -745,20 +750,22 @@ extern void __glXClientInfo (  Display *dpy, int opcode );
 ** Declarations that should be in Xlib
 */
 #ifdef __GL_USE_OUR_PROTOTYPES
-extern void _XFlush(Display*);
-extern Status _XReply(Display*, xReply*, int, Bool);
-extern void _XRead(Display*, void*, long);
-extern void _XSend(Display*, const void*, long);
+extern void _XFlush(Display *);
+extern Status _XReply(Display *, xReply *, int, Bool);
+extern void _XRead(Display *, void *, long);
+extern void _XSend(Display *, const void *, long);
 #endif
 
 
-extern void __glXInitializeVisualConfigFromTags( __GLcontextModes *config,
-    int count, const INT32 *bp, Bool tagged_only, Bool fbconfig_style_tags );
+extern void __glXInitializeVisualConfigFromTags(__GLcontextModes * config,
+                                                int count, const INT32 * bp,
+                                                Bool tagged_only,
+                                                Bool fbconfig_style_tags);
 
-extern char * __glXQueryServerString(Display* dpy, int opcode,
-                                     CARD32 screen, CARD32 name);
-extern char * __glXGetString(Display* dpy, int opcode,
-                             CARD32 screen, CARD32 name);
+extern char *__glXQueryServerString(Display * dpy, int opcode,
+                                    CARD32 screen, CARD32 name);
+extern char *__glXGetString(Display * dpy, int opcode,
+                            CARD32 screen, CARD32 name);
 
 extern char *__glXstrdup(const char *str);
 
@@ -767,15 +774,16 @@ extern const char __glXGLClientVersion[];
 extern const char __glXGLClientExtensions[];
 
 /* Get the unadjusted system time */
-extern int __glXGetUST( int64_t * ust );
+extern int __glXGetUST(int64_t * ust);
 
 extern GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable,
-				    int32_t * numerator, int32_t * denominator);
+                                    int32_t * numerator,
+                                    int32_t * denominator);
 
 #ifdef GLX_DIRECT_RENDERING
 GLboolean
-__driGetMscRateOML(__DRIdrawable *draw,
-		   int32_t *numerator, int32_t *denominator, void *private);
+__driGetMscRateOML(__DRIdrawable * draw,
+                   int32_t * numerator, int32_t * denominator, void *private);
 #endif
 
 #endif /* !__GLX_client_h__ */
diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c
index 3242ac7712..ea55cc49ef 100644
--- a/src/glx/x11/glxcmds.c
+++ b/src/glx/x11/glxcmds.c
@@ -59,56 +59,59 @@ static const char __glXGLXClientVersion[] = "1.4";
 #ifdef GLX_DIRECT_RENDERING
 
 static Bool windowExistsFlag;
-static int windowExistsErrorHandler(Display *dpy, XErrorEvent *xerr)
+static int
+windowExistsErrorHandler(Display * dpy, XErrorEvent * xerr)
 {
-    if (xerr->error_code == BadWindow) {
-	windowExistsFlag = GL_FALSE;
-    }
-    return 0;
+   if (xerr->error_code == BadWindow) {
+      windowExistsFlag = GL_FALSE;
+   }
+   return 0;
 }
 
 /**
  * Find drawables in the local hash that have been destroyed on the
  * server.
- * 
+ *
  * \param dpy    Display to destroy drawables for
  * \param screen Screen number to destroy drawables for
  */
-static void GarbageCollectDRIDrawables(Display *dpy, __GLXscreenConfigs *sc)
-{
-    XID draw;
-    __GLXDRIdrawable *pdraw;
-    XWindowAttributes xwa;
-    int (*oldXErrorHandler)(Display *, XErrorEvent *);
-
-    /* Set no-op error handler so Xlib doesn't bail out if the windows
-     * has alreay been destroyed on the server. */
-    XSync(dpy, GL_FALSE);
-    oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler);
-
-    if (__glxHashFirst(sc->drawHash, &draw, (void *)&pdraw) == 1) {
-	do {
-	    windowExistsFlag = GL_TRUE;
-	    XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */
-	    if (!windowExistsFlag) {
-		/* Destroy the local drawable data, if the drawable no
-		   longer exists in the Xserver */
-		(*pdraw->destroyDrawable)(pdraw);
-                __glxHashDelete(sc->drawHash, draw);
-	    }
-	} while (__glxHashNext(sc->drawHash, &draw, (void *)&pdraw) == 1);
-    }
-
-    XSync(dpy, GL_FALSE);
-    XSetErrorHandler(oldXErrorHandler);
-}
-
-extern __GLXDRIdrawable *
-GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num);
+static void
+GarbageCollectDRIDrawables(Display * dpy, __GLXscreenConfigs * sc)
+{
+   XID draw;
+   __GLXDRIdrawable *pdraw;
+   XWindowAttributes xwa;
+   int (*oldXErrorHandler) (Display *, XErrorEvent *);
+
+   /* Set no-op error handler so Xlib doesn't bail out if the windows
+    * has alreay been destroyed on the server. */
+   XSync(dpy, GL_FALSE);
+   oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler);
+
+   if (__glxHashFirst(sc->drawHash, &draw, (void *) &pdraw) == 1) {
+      do {
+         windowExistsFlag = GL_TRUE;
+         XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */
+         if (!windowExistsFlag) {
+            /* Destroy the local drawable data, if the drawable no
+               longer exists in the Xserver */
+            (*pdraw->destroyDrawable) (pdraw);
+            __glxHashDelete(sc->drawHash, draw);
+         }
+      } while (__glxHashNext(sc->drawHash, &draw, (void *) &pdraw) == 1);
+   }
+
+   XSync(dpy, GL_FALSE);
+   XSetErrorHandler(oldXErrorHandler);
+}
+
+extern __GLXDRIdrawable *GetGLXDRIDrawable(Display * dpy,
+                                           GLXDrawable drawable,
+                                           int *const scrn_num);
 
 /**
  * Get the __DRIdrawable for the drawable associated with a GLXContext
- * 
+ *
  * \param dpy       The display associated with \c drawable.
  * \param drawable  GLXDrawable whose __DRIdrawable part is to be retrieved.
  * \param scrn_num  If non-NULL, the drawables screen is stored there
@@ -116,30 +119,30 @@ GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num);
  *           the drawable is not associated with a direct-rendering context.
  */
 _X_HIDDEN __GLXDRIdrawable *
-GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num)
+GetGLXDRIDrawable(Display * dpy, GLXDrawable drawable, int *const scrn_num)
 {
-    __GLXdisplayPrivate *priv = __glXInitialize(dpy);
-    __GLXDRIdrawable *pdraw;
-    const unsigned  screen_count = ScreenCount(dpy);
-    unsigned   i;
-    __GLXscreenConfigs *psc;
+   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   __GLXDRIdrawable *pdraw;
+   const unsigned screen_count = ScreenCount(dpy);
+   unsigned i;
+   __GLXscreenConfigs *psc;
 
-    if (priv == NULL)
-	return NULL;
-    
-    for (i = 0; i < screen_count; i++) {
-	psc = &priv->screenConfigs[i];
-	if (psc->drawHash == NULL)
-	    continue;
+   if (priv == NULL)
+      return NULL;
 
-	if (__glxHashLookup(psc->drawHash, drawable, (void *) &pdraw) == 0) {
-	    if (scrn_num != NULL)
-		*scrn_num = i;
-	    return pdraw;
-	}
-    }
+   for (i = 0; i < screen_count; i++) {
+      psc = &priv->screenConfigs[i];
+      if (psc->drawHash == NULL)
+         continue;
+
+      if (__glxHashLookup(psc->drawHash, drawable, (void *) &pdraw) == 0) {
+         if (scrn_num != NULL)
+            *scrn_num = i;
+         return pdraw;
+      }
+   }
 
-    return NULL;
+   return NULL;
 }
 
 #endif
@@ -147,57 +150,59 @@ GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num)
 
 /**
  * Get the GLX per-screen data structure associated with a GLX context.
- * 
+ *
  * \param dpy   Display for which the GLX per-screen information is to be
  *              retrieved.
  * \param scrn  Screen on \c dpy for which the GLX per-screen information is
  *              to be retrieved.
  * \returns A pointer to the GLX per-screen data if \c dpy and \c scrn
  *          specify a valid GLX screen, or NULL otherwise.
- * 
+ *
  * \todo Should this function validate that \c scrn is within the screen
  *       number range for \c dpy?
  */
 
 static __GLXscreenConfigs *
-GetGLXScreenConfigs(Display *dpy, int scrn)
+GetGLXScreenConfigs(Display * dpy, int scrn)
 {
-    __GLXdisplayPrivate * const priv = __glXInitialize(dpy);
+   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
 
-    return (priv && priv->screenConfigs != NULL) ? &priv->screenConfigs[scrn] : NULL;
+   return (priv
+           && priv->screenConfigs !=
+           NULL) ? &priv->screenConfigs[scrn] : NULL;
 }
 
 
 static int
-GetGLXPrivScreenConfig( Display *dpy, int scrn, __GLXdisplayPrivate ** ppriv,
-			__GLXscreenConfigs ** ppsc )
+GetGLXPrivScreenConfig(Display * dpy, int scrn, __GLXdisplayPrivate ** ppriv,
+                       __GLXscreenConfigs ** ppsc)
 {
-    /* Initialize the extension, if needed .  This has the added value
-     * of initializing/allocating the display private 
-     */
-    
-    if ( dpy == NULL ) {
-	return GLX_NO_EXTENSION;
-    }
+   /* Initialize the extension, if needed .  This has the added value
+    * of initializing/allocating the display private
+    */
 
-    *ppriv = __glXInitialize(dpy);
-    if ( *ppriv == NULL ) {
-	return GLX_NO_EXTENSION;
-    }
+   if (dpy == NULL) {
+      return GLX_NO_EXTENSION;
+   }
 
-    /* Check screen number to see if its valid */
-    if ((scrn < 0) || (scrn >= ScreenCount(dpy))) {
-	return GLX_BAD_SCREEN;
-    }
+   *ppriv = __glXInitialize(dpy);
+   if (*ppriv == NULL) {
+      return GLX_NO_EXTENSION;
+   }
 
-    /* Check to see if the GL is supported on this screen */
-    *ppsc = &((*ppriv)->screenConfigs[scrn]);
-    if ( (*ppsc)->configs == NULL ) {
-	/* No support for GL on this screen regardless of visual */
-	return GLX_BAD_VISUAL;
-    }
+   /* Check screen number to see if its valid */
+   if ((scrn < 0) || (scrn >= ScreenCount(dpy))) {
+      return GLX_BAD_SCREEN;
+   }
 
-    return Success;
+   /* Check to see if the GL is supported on this screen */
+   *ppsc = &((*ppriv)->screenConfigs[scrn]);
+   if ((*ppsc)->configs == NULL) {
+      /* No support for GL on this screen regardless of visual */
+      return GLX_BAD_VISUAL;
+   }
+
+   return Success;
 }
 
 
@@ -212,27 +217,26 @@ GetGLXPrivScreenConfig( Display *dpy, int scrn, __GLXdisplayPrivate ** ppriv,
  *          is returned.
  */
 static __GLcontextModes *
-ValidateGLXFBConfig( Display * dpy, GLXFBConfig config )
-{
-    __GLXdisplayPrivate * const priv = __glXInitialize(dpy);
-    const unsigned num_screens = ScreenCount(dpy);
-    unsigned   i;
-    const __GLcontextModes * modes;
-
-
-    if ( priv != NULL ) {
-	for ( i = 0 ; i < num_screens ; i++ ) {
-	    for ( modes = priv->screenConfigs[i].configs
-		  ; modes != NULL
-		  ; modes = modes->next ) {
-		if ( modes == (__GLcontextModes *) config ) {
-		    return (__GLcontextModes *) config;
-		}
-	    }
-	}
-    }
+ValidateGLXFBConfig(Display * dpy, GLXFBConfig config)
+{
+   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+   const unsigned num_screens = ScreenCount(dpy);
+   unsigned i;
+   const __GLcontextModes *modes;
+
+
+   if (priv != NULL) {
+      for (i = 0; i < num_screens; i++) {
+         for (modes = priv->screenConfigs[i].configs; modes != NULL;
+              modes = modes->next) {
+            if (modes == (__GLcontextModes *) config) {
+               return (__GLcontextModes *) config;
+            }
+         }
+      }
+   }
 
-    return NULL;
+   return NULL;
 }
 
 
@@ -241,471 +245,497 @@ ValidateGLXFBConfig( Display * dpy, GLXFBConfig config )
  * later in the function for direct-rendering contexts.  Direct-rendering
  * contexts don't need to track client state, so they don't need that memory
  * at all.
- * 
+ *
  * \todo Eliminate \c __glXInitVertexArrayState.  Replace it with a new
  * function called \c __glXAllocateClientState that allocates the memory and
  * does all the initialization (including the pixel pack / unpack).
  */
-static
-GLXContext AllocateGLXContext( Display *dpy )
-{
-     GLXContext gc;
-     int bufSize;
-     CARD8 opcode;
-    __GLXattribute *state;
-
-    if (!dpy)
-        return NULL;
-
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return NULL;
-    }
-
-    /* Allocate our context record */
-    gc = (GLXContext) Xmalloc(sizeof(struct __GLXcontextRec));
-    if (!gc) {
-	/* Out of memory */
-	return NULL;
-    }
-    memset(gc, 0, sizeof(struct __GLXcontextRec));
-
-    state = Xmalloc(sizeof(struct __GLXattributeRec));
-    if (state == NULL) {
-	/* Out of memory */
-	Xfree(gc);
-	return NULL;
-    }
-    gc->client_state_private = state;
-    memset(gc->client_state_private, 0, sizeof(struct __GLXattributeRec));
-    state->NoDrawArraysProtocol = (getenv("LIBGL_NO_DRAWARRAYS") != NULL);
-
-    /*
+static GLXContext
+AllocateGLXContext(Display * dpy)
+{
+   GLXContext gc;
+   int bufSize;
+   CARD8 opcode;
+   __GLXattribute *state;
+
+   if (!dpy)
+      return NULL;
+
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return NULL;
+   }
+
+   /* Allocate our context record */
+   gc = (GLXContext) Xmalloc(sizeof(struct __GLXcontextRec));
+   if (!gc) {
+      /* Out of memory */
+      return NULL;
+   }
+   memset(gc, 0, sizeof(struct __GLXcontextRec));
+
+   state = Xmalloc(sizeof(struct __GLXattributeRec));
+   if (state == NULL) {
+      /* Out of memory */
+      Xfree(gc);
+      return NULL;
+   }
+   gc->client_state_private = state;
+   memset(gc->client_state_private, 0, sizeof(struct __GLXattributeRec));
+   state->NoDrawArraysProtocol = (getenv("LIBGL_NO_DRAWARRAYS") != NULL);
+
+   /*
     ** Create a temporary buffer to hold GLX rendering commands.  The size
     ** of the buffer is selected so that the maximum number of GLX rendering
     ** commands can fit in a single X packet and still have room in the X
     ** packet for the GLXRenderReq header.
     */
 
-    bufSize = (XMaxRequestSize(dpy) * 4) - sz_xGLXRenderReq;
-    gc->buf = (GLubyte *) Xmalloc(bufSize);
-    if (!gc->buf) {
-	Xfree(gc->client_state_private);
-	Xfree(gc);
-	return NULL;
-    }
-    gc->bufSize = bufSize;
+   bufSize = (XMaxRequestSize(dpy) * 4) - sz_xGLXRenderReq;
+   gc->buf = (GLubyte *) Xmalloc(bufSize);
+   if (!gc->buf) {
+      Xfree(gc->client_state_private);
+      Xfree(gc);
+      return NULL;
+   }
+   gc->bufSize = bufSize;
 
-    /* Fill in the new context */
-    gc->renderMode = GL_RENDER;
+   /* Fill in the new context */
+   gc->renderMode = GL_RENDER;
 
-    state->storePack.alignment = 4;
-    state->storeUnpack.alignment = 4;
+   state->storePack.alignment = 4;
+   state->storeUnpack.alignment = 4;
 
-    gc->attributes.stackPointer = &gc->attributes.stack[0];
+   gc->attributes.stackPointer = &gc->attributes.stack[0];
 
-    /*
+   /*
     ** PERFORMANCE NOTE: A mode dependent fill image can speed things up.
     ** Other code uses the fastImageUnpack bit, but it is never set
     ** to GL_TRUE.
     */
-    gc->fastImageUnpack = GL_FALSE;
-    gc->fillImage = __glFillImage;
-    gc->pc = gc->buf;
-    gc->bufEnd = gc->buf + bufSize;
-    gc->isDirect = GL_FALSE;
-    if (__glXDebug) {
-	/*
-	** Set limit register so that there will be one command per packet
-	*/
-	gc->limit = gc->buf;
-    } else {
-	gc->limit = gc->buf + bufSize - __GLX_BUFFER_LIMIT_SIZE;
-    }
-    gc->createDpy = dpy;
-    gc->majorOpcode = opcode;
-
-    /*
+   gc->fastImageUnpack = GL_FALSE;
+   gc->fillImage = __glFillImage;
+   gc->pc = gc->buf;
+   gc->bufEnd = gc->buf + bufSize;
+   gc->isDirect = GL_FALSE;
+   if (__glXDebug) {
+      /*
+       ** Set limit register so that there will be one command per packet
+       */
+      gc->limit = gc->buf;
+   }
+   else {
+      gc->limit = gc->buf + bufSize - __GLX_BUFFER_LIMIT_SIZE;
+   }
+   gc->createDpy = dpy;
+   gc->majorOpcode = opcode;
+
+   /*
     ** Constrain the maximum drawing command size allowed to be
     ** transfered using the X_GLXRender protocol request.  First
     ** constrain by a software limit, then constrain by the protocl
     ** limit.
     */
-    if (bufSize > __GLX_RENDER_CMD_SIZE_LIMIT) {
-        bufSize = __GLX_RENDER_CMD_SIZE_LIMIT;
-    }
-    if (bufSize > __GLX_MAX_RENDER_CMD_SIZE) {
-        bufSize = __GLX_MAX_RENDER_CMD_SIZE;
-    }
-    gc->maxSmallRenderCommandSize = bufSize;
-    return gc;
+   if (bufSize > __GLX_RENDER_CMD_SIZE_LIMIT) {
+      bufSize = __GLX_RENDER_CMD_SIZE_LIMIT;
+   }
+   if (bufSize > __GLX_MAX_RENDER_CMD_SIZE) {
+      bufSize = __GLX_MAX_RENDER_CMD_SIZE;
+   }
+   gc->maxSmallRenderCommandSize = bufSize;
+   return gc;
 }
 
 
 /**
  * Create a new context.  Exactly one of \c vis and \c fbconfig should be
  * non-NULL.
- * 
+ *
  * \param use_glx_1_3  For FBConfigs, should GLX 1.3 protocol or
  *                     SGIX_fbconfig protocol be used?
  * \param renderType   For FBConfigs, what is the rendering type?
  */
 
 static GLXContext
-CreateContext(Display *dpy, XVisualInfo *vis,
-	      const __GLcontextModes * const fbconfig,
-	      GLXContext shareList,
-	      Bool allowDirect, GLXContextID contextID,
-	      Bool use_glx_1_3, int renderType)
+CreateContext(Display * dpy, XVisualInfo * vis,
+              const __GLcontextModes * const fbconfig,
+              GLXContext shareList,
+              Bool allowDirect, GLXContextID contextID,
+              Bool use_glx_1_3, int renderType)
 {
-    GLXContext gc;
+   GLXContext gc;
 #ifdef GLX_DIRECT_RENDERING
-    int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen;
-    __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
+   int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen;
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
 #endif
 
-    if ( dpy == NULL )
-       return NULL;
+   if (dpy == NULL)
+      return NULL;
 
-    gc = AllocateGLXContext(dpy);
-    if (!gc)
-	return NULL;
+   gc = AllocateGLXContext(dpy);
+   if (!gc)
+      return NULL;
 
-    if (None == contextID) {
-	if ( (vis == NULL) && (fbconfig == NULL) )
-	    return NULL;
+   if (None == contextID) {
+      if ((vis == NULL) && (fbconfig == NULL))
+         return NULL;
 
 #ifdef GLX_DIRECT_RENDERING
-	if (allowDirect && psc->driScreen) {
-	    const __GLcontextModes * mode;
-
-	    if (fbconfig == NULL) {
-		mode = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
-		if (mode == NULL) {
-		   xError error;
-
-		   error.errorCode = BadValue;
-		   error.resourceID = vis->visualid;
-		   error.sequenceNumber = dpy->request;
-		   error.type = X_Error;
-		   error.majorCode = gc->majorOpcode;
-		   error.minorCode = X_GLXCreateContext;
-		   _XError(dpy, &error);
-		   return None;
-		}
-	    }
-	    else {
-		mode = fbconfig;
-	    }
-
-	    gc->driContext = psc->driScreen->createContext(psc, mode, gc,
-							   shareList,
-							   renderType);
-	    if (gc->driContext != NULL) {
-		gc->screen = mode->screen;
-		gc->psc = psc;
-		gc->mode = mode;
-		gc->isDirect = GL_TRUE;
-	    }
-	}
+      if (allowDirect && psc->driScreen) {
+         const __GLcontextModes *mode;
+
+         if (fbconfig == NULL) {
+            mode = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
+            if (mode == NULL) {
+               xError error;
+
+               error.errorCode = BadValue;
+               error.resourceID = vis->visualid;
+               error.sequenceNumber = dpy->request;
+               error.type = X_Error;
+               error.majorCode = gc->majorOpcode;
+               error.minorCode = X_GLXCreateContext;
+               _XError(dpy, &error);
+               return None;
+            }
+         }
+         else {
+            mode = fbconfig;
+         }
+
+         gc->driContext = psc->driScreen->createContext(psc, mode, gc,
+                                                        shareList,
+                                                        renderType);
+         if (gc->driContext != NULL) {
+            gc->screen = mode->screen;
+            gc->psc = psc;
+            gc->mode = mode;
+            gc->isDirect = GL_TRUE;
+         }
+      }
 #endif
 
-	LockDisplay(dpy);
-	if ( fbconfig == NULL ) {
-	    xGLXCreateContextReq *req;
-
-	    /* Send the glXCreateContext request */
-	    GetReq(GLXCreateContext,req);
-	    req->reqType = gc->majorOpcode;
-	    req->glxCode = X_GLXCreateContext;
-	    req->context = gc->xid = XAllocID(dpy);
-	    req->visual = vis->visualid;
-	    req->screen = vis->screen;
-	    req->shareList = shareList ? shareList->xid : None;
+      LockDisplay(dpy);
+      if (fbconfig == NULL) {
+         xGLXCreateContextReq *req;
+
+         /* Send the glXCreateContext request */
+         GetReq(GLXCreateContext, req);
+         req->reqType = gc->majorOpcode;
+         req->glxCode = X_GLXCreateContext;
+         req->context = gc->xid = XAllocID(dpy);
+         req->visual = vis->visualid;
+         req->screen = vis->screen;
+         req->shareList = shareList ? shareList->xid : None;
 #ifdef GLX_DIRECT_RENDERING
-	    req->isDirect = gc->driContext != NULL;
+         req->isDirect = gc->driContext != NULL;
 #else
-	    req->isDirect = 0;
+         req->isDirect = 0;
 #endif
-	}
-	else if ( use_glx_1_3 ) {
-	    xGLXCreateNewContextReq *req;
-
-	    /* Send the glXCreateNewContext request */
-	    GetReq(GLXCreateNewContext,req);
-	    req->reqType = gc->majorOpcode;
-	    req->glxCode = X_GLXCreateNewContext;
-	    req->context = gc->xid = XAllocID(dpy);
-	    req->fbconfig = fbconfig->fbconfigID;
-	    req->screen = fbconfig->screen;
-	    req->renderType = renderType;
-	    req->shareList = shareList ? shareList->xid : None;
+      }
+      else if (use_glx_1_3) {
+         xGLXCreateNewContextReq *req;
+
+         /* Send the glXCreateNewContext request */
+         GetReq(GLXCreateNewContext, req);
+         req->reqType = gc->majorOpcode;
+         req->glxCode = X_GLXCreateNewContext;
+         req->context = gc->xid = XAllocID(dpy);
+         req->fbconfig = fbconfig->fbconfigID;
+         req->screen = fbconfig->screen;
+         req->renderType = renderType;
+         req->shareList = shareList ? shareList->xid : None;
 #ifdef GLX_DIRECT_RENDERING
-	    req->isDirect = gc->driContext != NULL;
+         req->isDirect = gc->driContext != NULL;
 #else
-	    req->isDirect = 0;
+         req->isDirect = 0;
 #endif
-	}
-	else {
-	    xGLXVendorPrivateWithReplyReq *vpreq;
-	    xGLXCreateContextWithConfigSGIXReq *req;
-
-	    /* Send the glXCreateNewContext request */
-	    GetReqExtra(GLXVendorPrivateWithReply,
-			sz_xGLXCreateContextWithConfigSGIXReq-sz_xGLXVendorPrivateWithReplyReq,vpreq);
-	    req = (xGLXCreateContextWithConfigSGIXReq *)vpreq;
-	    req->reqType = gc->majorOpcode;
-	    req->glxCode = X_GLXVendorPrivateWithReply;
-	    req->vendorCode = X_GLXvop_CreateContextWithConfigSGIX;
-	    req->context = gc->xid = XAllocID(dpy);
-	    req->fbconfig = fbconfig->fbconfigID;
-	    req->screen = fbconfig->screen;
-	    req->renderType = renderType;
-	    req->shareList = shareList ? shareList->xid : None;
+      }
+      else {
+         xGLXVendorPrivateWithReplyReq *vpreq;
+         xGLXCreateContextWithConfigSGIXReq *req;
+
+         /* Send the glXCreateNewContext request */
+         GetReqExtra(GLXVendorPrivateWithReply,
+                     sz_xGLXCreateContextWithConfigSGIXReq -
+                     sz_xGLXVendorPrivateWithReplyReq, vpreq);
+         req = (xGLXCreateContextWithConfigSGIXReq *) vpreq;
+         req->reqType = gc->majorOpcode;
+         req->glxCode = X_GLXVendorPrivateWithReply;
+         req->vendorCode = X_GLXvop_CreateContextWithConfigSGIX;
+         req->context = gc->xid = XAllocID(dpy);
+         req->fbconfig = fbconfig->fbconfigID;
+         req->screen = fbconfig->screen;
+         req->renderType = renderType;
+         req->shareList = shareList ? shareList->xid : None;
 #ifdef GLX_DIRECT_RENDERING
-	    req->isDirect = gc->driContext != NULL;
+         req->isDirect = gc->driContext != NULL;
 #else
-	    req->isDirect = 0;
+         req->isDirect = 0;
 #endif
-	}
+      }
 
-	UnlockDisplay(dpy);
-	SyncHandle();
-	gc->imported = GL_FALSE;
-    }
-    else {
-	gc->xid = contextID;
-	gc->imported = GL_TRUE;
-    }
+      UnlockDisplay(dpy);
+      SyncHandle();
+      gc->imported = GL_FALSE;
+   }
+   else {
+      gc->xid = contextID;
+      gc->imported = GL_TRUE;
+   }
 
-    return gc;
+   return gc;
 }
 
-PUBLIC GLXContext glXCreateContext(Display *dpy, XVisualInfo *vis,
-				   GLXContext shareList, Bool allowDirect)
+PUBLIC GLXContext
+glXCreateContext(Display * dpy, XVisualInfo * vis,
+                 GLXContext shareList, Bool allowDirect)
 {
    return CreateContext(dpy, vis, NULL, shareList, allowDirect, None,
-			False, 0);
+                        False, 0);
 }
 
-_X_HIDDEN void __glXFreeContext(__GLXcontext *gc)
+_X_HIDDEN void
+__glXFreeContext(__GLXcontext * gc)
 {
-    if (gc->vendor) XFree((char *) gc->vendor);
-    if (gc->renderer) XFree((char *) gc->renderer);
-    if (gc->version) XFree((char *) gc->version);
-    if (gc->extensions) XFree((char *) gc->extensions);
-    __glFreeAttributeState(gc);
-    XFree((char *) gc->buf);
-    Xfree((char *) gc->client_state_private);
-    XFree((char *) gc);
-    
+   if (gc->vendor)
+      XFree((char *) gc->vendor);
+   if (gc->renderer)
+      XFree((char *) gc->renderer);
+   if (gc->version)
+      XFree((char *) gc->version);
+   if (gc->extensions)
+      XFree((char *) gc->extensions);
+   __glFreeAttributeState(gc);
+   XFree((char *) gc->buf);
+   Xfree((char *) gc->client_state_private);
+   XFree((char *) gc);
+
 }
 
 /*
 ** Destroy the named context
 */
-static void 
-DestroyContext(Display *dpy, GLXContext gc)
+static void
+DestroyContext(Display * dpy, GLXContext gc)
 {
-    xGLXDestroyContextReq *req;
-    GLXContextID xid;
-    CARD8 opcode;
-    GLboolean imported;
+   xGLXDestroyContextReq *req;
+   GLXContextID xid;
+   CARD8 opcode;
+   GLboolean imported;
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode || !gc) {
-	return;
-    }
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode || !gc) {
+      return;
+   }
 
-    __glXLock();
-    xid = gc->xid;
-    imported = gc->imported;
-    gc->xid = None;
+   __glXLock();
+   xid = gc->xid;
+   imported = gc->imported;
+   gc->xid = None;
 
 #ifdef GLX_DIRECT_RENDERING
-    /* Destroy the direct rendering context */
-    if (gc->driContext) {
-	(*gc->driContext->destroyContext)(gc->driContext, gc->psc, dpy);
-	gc->driContext = NULL;
-	GarbageCollectDRIDrawables(dpy, gc->psc);
-    }
+   /* Destroy the direct rendering context */
+   if (gc->driContext) {
+      (*gc->driContext->destroyContext) (gc->driContext, gc->psc, dpy);
+      gc->driContext = NULL;
+      GarbageCollectDRIDrawables(dpy, gc->psc);
+   }
 #endif
 
-    __glXFreeVertexArrayState(gc);
+   __glXFreeVertexArrayState(gc);
 
-    if (gc->currentDpy) {
-	/* Have to free later cuz it's in use now */
-	__glXUnlock();
-    } else {
-	/* Destroy the handle if not current to anybody */
-	__glXUnlock();
-	__glXFreeContext(gc);
-    }
+   if (gc->currentDpy) {
+      /* Have to free later cuz it's in use now */
+      __glXUnlock();
+   }
+   else {
+      /* Destroy the handle if not current to anybody */
+      __glXUnlock();
+      __glXFreeContext(gc);
+   }
 
-    if (!imported) {
-	/* 
-	** This dpy also created the server side part of the context.
-	** Send the glXDestroyContext request.
-	*/
-	LockDisplay(dpy);
-	GetReq(GLXDestroyContext,req);
-	req->reqType = opcode;
-	req->glxCode = X_GLXDestroyContext;
-	req->context = xid;
-	UnlockDisplay(dpy);
-	SyncHandle();
-    }
+   if (!imported) {
+      /*
+       ** This dpy also created the server side part of the context.
+       ** Send the glXDestroyContext request.
+       */
+      LockDisplay(dpy);
+      GetReq(GLXDestroyContext, req);
+      req->reqType = opcode;
+      req->glxCode = X_GLXDestroyContext;
+      req->context = xid;
+      UnlockDisplay(dpy);
+      SyncHandle();
+   }
 }
 
-PUBLIC void glXDestroyContext(Display *dpy, GLXContext gc)
+PUBLIC void
+glXDestroyContext(Display * dpy, GLXContext gc)
 {
-    DestroyContext(dpy, gc);
+   DestroyContext(dpy, gc);
 }
 
 /*
 ** Return the major and minor version #s for the GLX extension
 */
-PUBLIC Bool glXQueryVersion(Display *dpy, int *major, int *minor)
+PUBLIC Bool
+glXQueryVersion(Display * dpy, int *major, int *minor)
 {
-    __GLXdisplayPrivate *priv;
+   __GLXdisplayPrivate *priv;
 
-    /* Init the extension.  This fetches the major and minor version. */
-    priv = __glXInitialize(dpy);
-    if (!priv) return GL_FALSE;
+   /* Init the extension.  This fetches the major and minor version. */
+   priv = __glXInitialize(dpy);
+   if (!priv)
+      return GL_FALSE;
 
-    if (major) *major = priv->majorVersion;
-    if (minor) *minor = priv->minorVersion;
-    return GL_TRUE;
+   if (major)
+      *major = priv->majorVersion;
+   if (minor)
+      *minor = priv->minorVersion;
+   return GL_TRUE;
 }
 
 /*
 ** Query the existance of the GLX extension
 */
-PUBLIC Bool glXQueryExtension(Display *dpy, int *errorBase, int *eventBase)
-{
-    int major_op, erb, evb;
-    Bool rv;
-
-    rv = XQueryExtension(dpy, GLX_EXTENSION_NAME, &major_op, &evb, &erb);
-    if (rv) {
-	if (errorBase) *errorBase = erb;
-	if (eventBase) *eventBase = evb;
-    }
-    return rv;
+PUBLIC Bool
+glXQueryExtension(Display * dpy, int *errorBase, int *eventBase)
+{
+   int major_op, erb, evb;
+   Bool rv;
+
+   rv = XQueryExtension(dpy, GLX_EXTENSION_NAME, &major_op, &evb, &erb);
+   if (rv) {
+      if (errorBase)
+         *errorBase = erb;
+      if (eventBase)
+         *eventBase = evb;
+   }
+   return rv;
 }
 
 /*
 ** Put a barrier in the token stream that forces the GL to finish its
 ** work before X can proceed.
 */
-PUBLIC void glXWaitGL(void)
+PUBLIC void
+glXWaitGL(void)
 {
-    xGLXWaitGLReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    Display *dpy = gc->currentDpy;
+   xGLXWaitGLReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   Display *dpy = gc->currentDpy;
 
-    if (!dpy) return;
+   if (!dpy)
+      return;
 
-    /* Flush any pending commands out */
-    __glXFlushRenderBuffer(gc, gc->pc);
+   /* Flush any pending commands out */
+   __glXFlushRenderBuffer(gc, gc->pc);
 
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext) {
-    	int screen;
-    	__GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, gc->currentDrawable, &screen);
-
-    	if ( pdraw != NULL ) {
-	    __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
-	    glFlush();
-	    if (psc->driScreen->waitGL != NULL)
-	    	(*psc->driScreen->waitGL)(pdraw);
-	}
-	return;
-    }
+   if (gc->driContext) {
+      int screen;
+      __GLXDRIdrawable *pdraw =
+         GetGLXDRIDrawable(dpy, gc->currentDrawable, &screen);
+
+      if (pdraw != NULL) {
+         __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
+         glFlush();
+         if (psc->driScreen->waitGL != NULL)
+            (*psc->driScreen->waitGL) (pdraw);
+      }
+      return;
+   }
 #endif
 
-    /* Send the glXWaitGL request */
-    LockDisplay(dpy);
-    GetReq(GLXWaitGL,req);
-    req->reqType = gc->majorOpcode;
-    req->glxCode = X_GLXWaitGL;
-    req->contextTag = gc->currentContextTag;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   /* Send the glXWaitGL request */
+   LockDisplay(dpy);
+   GetReq(GLXWaitGL, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXWaitGL;
+   req->contextTag = gc->currentContextTag;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
 /*
 ** Put a barrier in the token stream that forces X to finish its
 ** work before GL can proceed.
 */
-PUBLIC void glXWaitX(void)
+PUBLIC void
+glXWaitX(void)
 {
-    xGLXWaitXReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    Display *dpy = gc->currentDpy;
+   xGLXWaitXReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   Display *dpy = gc->currentDpy;
 
-    if (!dpy) return;
+   if (!dpy)
+      return;
 
-    /* Flush any pending commands out */
-    __glXFlushRenderBuffer(gc, gc->pc);
+   /* Flush any pending commands out */
+   __glXFlushRenderBuffer(gc, gc->pc);
 
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext) {
-    	int screen;
-    	__GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, gc->currentDrawable, &screen);
-
-    	if ( pdraw != NULL ) {
-	    __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
-	    if (psc->driScreen->waitX != NULL)
-	    	(*psc->driScreen->waitX)(pdraw);
-	} else
-	    XSync(dpy, False);
-	return;
-    }
+   if (gc->driContext) {
+      int screen;
+      __GLXDRIdrawable *pdraw =
+         GetGLXDRIDrawable(dpy, gc->currentDrawable, &screen);
+
+      if (pdraw != NULL) {
+         __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
+         if (psc->driScreen->waitX != NULL)
+            (*psc->driScreen->waitX) (pdraw);
+      }
+      else
+         XSync(dpy, False);
+      return;
+   }
 #endif
 
-    /*
+   /*
     ** Send the glXWaitX request.
     */
-    LockDisplay(dpy);
-    GetReq(GLXWaitX,req);
-    req->reqType = gc->majorOpcode;
-    req->glxCode = X_GLXWaitX;
-    req->contextTag = gc->currentContextTag;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   LockDisplay(dpy);
+   GetReq(GLXWaitX, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXWaitX;
+   req->contextTag = gc->currentContextTag;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
-PUBLIC void glXUseXFont(Font font, int first, int count, int listBase)
+PUBLIC void
+glXUseXFont(Font font, int first, int count, int listBase)
 {
-    xGLXUseXFontReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    Display *dpy = gc->currentDpy;
+   xGLXUseXFontReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   Display *dpy = gc->currentDpy;
 
-    if (!dpy) return;
+   if (!dpy)
+      return;
 
-    /* Flush any pending commands out */
-    (void) __glXFlushRenderBuffer(gc, gc->pc);
+   /* Flush any pending commands out */
+   (void) __glXFlushRenderBuffer(gc, gc->pc);
 
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext) {
+   if (gc->driContext) {
       DRI_glXUseXFont(font, first, count, listBase);
       return;
-    }
+   }
 #endif
 
-    /* Send the glXUseFont request */
-    LockDisplay(dpy);
-    GetReq(GLXUseXFont,req);
-    req->reqType = gc->majorOpcode;
-    req->glxCode = X_GLXUseXFont;
-    req->contextTag = gc->currentContextTag;
-    req->font = font;
-    req->first = first;
-    req->count = count;
-    req->listBase = listBase;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   /* Send the glXUseFont request */
+   LockDisplay(dpy);
+   GetReq(GLXUseXFont, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXUseXFont;
+   req->contextTag = gc->currentContextTag;
+   req->font = font;
+   req->first = first;
+   req->count = count;
+   req->listBase = listBase;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
 /************************************************************************/
@@ -714,46 +744,48 @@ PUBLIC void glXUseXFont(Font font, int first, int count, int listBase)
 ** Copy the source context to the destination context using the
 ** attribute "mask".
 */
-PUBLIC void glXCopyContext(Display *dpy, GLXContext source,
-			   GLXContext dest, unsigned long mask)
+PUBLIC void
+glXCopyContext(Display * dpy, GLXContext source,
+               GLXContext dest, unsigned long mask)
 {
-    xGLXCopyContextReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    GLXContextTag tag;
-    CARD8 opcode;
+   xGLXCopyContextReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   GLXContextTag tag;
+   CARD8 opcode;
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return;
-    }
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return;
+   }
 
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext) {
-	/* NOT_DONE: This does not work yet */
-    }
+   if (gc->driContext) {
+      /* NOT_DONE: This does not work yet */
+   }
 #endif
 
-    /*
+   /*
     ** If the source is the current context, send its tag so that the context
     ** can be flushed before the copy.
     */
-    if (source == gc && dpy == gc->currentDpy) {
-	tag = gc->currentContextTag;
-    } else {
-	tag = 0;
-    }
-
-    /* Send the glXCopyContext request */
-    LockDisplay(dpy);
-    GetReq(GLXCopyContext,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXCopyContext;
-    req->source = source ? source->xid : None;
-    req->dest = dest ? dest->xid : None;
-    req->mask = mask;
-    req->contextTag = tag;
-    UnlockDisplay(dpy);
-    SyncHandle();
+   if (source == gc && dpy == gc->currentDpy) {
+      tag = gc->currentContextTag;
+   }
+   else {
+      tag = 0;
+   }
+
+   /* Send the glXCopyContext request */
+   LockDisplay(dpy);
+   GetReq(GLXCopyContext, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXCopyContext;
+   req->source = source ? source->xid : None;
+   req->dest = dest ? dest->xid : None;
+   req->mask = mask;
+   req->contextTag = tag;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
 
@@ -765,42 +797,43 @@ PUBLIC void glXCopyContext(Display *dpy, GLXContext source,
  *
  * \returns \c GL_TRUE if the context is direct rendering or not.
  */
-static Bool __glXIsDirect(Display *dpy, GLXContextID contextID)
+static Bool
+__glXIsDirect(Display * dpy, GLXContextID contextID)
 {
 #if !defined(USE_XCB)
-    xGLXIsDirectReq *req;
-    xGLXIsDirectReply reply;
+   xGLXIsDirectReq *req;
+   xGLXIsDirectReply reply;
 #endif
-    CARD8 opcode;
+   CARD8 opcode;
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return GL_FALSE;
-    }
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return GL_FALSE;
+   }
 
 #ifdef USE_XCB
-   xcb_connection_t* c = XGetXCBConnection(dpy);
-   xcb_glx_is_direct_reply_t* reply =
-      xcb_glx_is_direct_reply(c,
-                              xcb_glx_is_direct(c, contextID),
-                              NULL);
+   xcb_connection_t *c = XGetXCBConnection(dpy);
+   xcb_glx_is_direct_reply_t *reply = xcb_glx_is_direct_reply(c,
+                                                              xcb_glx_is_direct
+                                                              (c, contextID),
+                                                              NULL);
 
    const Bool is_direct = reply->is_direct ? True : False;
    free(reply);
 
    return is_direct;
 #else
-    /* Send the glXIsDirect request */
-    LockDisplay(dpy);
-    GetReq(GLXIsDirect,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXIsDirect;
-    req->context = contextID;
-    _XReply(dpy, (xReply*) &reply, 0, False);
-    UnlockDisplay(dpy);
-    SyncHandle();
-
-    return reply.isDirect;
+   /* Send the glXIsDirect request */
+   LockDisplay(dpy);
+   GetReq(GLXIsDirect, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXIsDirect;
+   req->context = contextID;
+   _XReply(dpy, (xReply *) & reply, 0, False);
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return reply.isDirect;
 #endif /* USE_XCB */
 }
 
@@ -810,162 +843,168 @@ static Bool __glXIsDirect(Display *dpy, GLXContextID contextID)
  * \c GLX_DIRECT_RENDERING is not defined?  Do we really need to bother with
  * the GLX protocol here at all?
  */
-PUBLIC Bool glXIsDirect(Display *dpy, GLXContext gc)
+PUBLIC Bool
+glXIsDirect(Display * dpy, GLXContext gc)
 {
-    if (!gc) {
-	return GL_FALSE;
+   if (!gc) {
+      return GL_FALSE;
 #ifdef GLX_DIRECT_RENDERING
-    } else if (gc->driContext) {
-	return GL_TRUE;
+   }
+   else if (gc->driContext) {
+      return GL_TRUE;
 #endif
-    }
-    return __glXIsDirect(dpy, gc->xid);
-}
-
-PUBLIC GLXPixmap glXCreateGLXPixmap(Display *dpy, XVisualInfo *vis, 
-				    Pixmap pixmap)
-{
-    xGLXCreateGLXPixmapReq *req;
-    GLXPixmap xid;
-    CARD8 opcode;
-
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return None;
-    }
-
-    /* Send the glXCreateGLXPixmap request */
-    LockDisplay(dpy);
-    GetReq(GLXCreateGLXPixmap,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXCreateGLXPixmap;
-    req->screen = vis->screen;
-    req->visual = vis->visualid;
-    req->pixmap = pixmap;
-    req->glxpixmap = xid = XAllocID(dpy);
-    UnlockDisplay(dpy);
-    SyncHandle();
+   }
+   return __glXIsDirect(dpy, gc->xid);
+}
+
+PUBLIC GLXPixmap
+glXCreateGLXPixmap(Display * dpy, XVisualInfo * vis, Pixmap pixmap)
+{
+   xGLXCreateGLXPixmapReq *req;
+   GLXPixmap xid;
+   CARD8 opcode;
+
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return None;
+   }
+
+   /* Send the glXCreateGLXPixmap request */
+   LockDisplay(dpy);
+   GetReq(GLXCreateGLXPixmap, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXCreateGLXPixmap;
+   req->screen = vis->screen;
+   req->visual = vis->visualid;
+   req->pixmap = pixmap;
+   req->glxpixmap = xid = XAllocID(dpy);
+   UnlockDisplay(dpy);
+   SyncHandle();
 
 #ifdef GLX_DIRECT_RENDERING
-    do {
-	/* FIXME: Maybe delay __DRIdrawable creation until the drawable
-	 * is actually bound to a context... */
-
-	__GLXdisplayPrivate *const priv = __glXInitialize(dpy);
-	__GLXDRIdrawable *pdraw;
-	__GLXscreenConfigs *psc;
-	__GLcontextModes *modes;
-
-	psc = &priv->screenConfigs[vis->screen];
-	if (psc->driScreen == NULL)
-	    break;
-	modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
-	pdraw = psc->driScreen->createDrawable(psc, pixmap, req->glxpixmap, modes);
-	if (pdraw == NULL) {
-	    fprintf(stderr, "failed to create pixmap\n");
-	    break;
-	}
-
-	if (__glxHashInsert(psc->drawHash, req->glxpixmap, pdraw)) {
-	    (*pdraw->destroyDrawable) (pdraw);
-	    return None;           /* FIXME: Check what we're supposed to do here... */
-	}
-    } while (0);
+   do {
+      /* FIXME: Maybe delay __DRIdrawable creation until the drawable
+       * is actually bound to a context... */
+
+      __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+      __GLXDRIdrawable *pdraw;
+      __GLXscreenConfigs *psc;
+      __GLcontextModes *modes;
+
+      psc = &priv->screenConfigs[vis->screen];
+      if (psc->driScreen == NULL)
+         break;
+      modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
+      pdraw = psc->driScreen->createDrawable(psc, pixmap, req->glxpixmap, modes);
+      if (pdraw == NULL) {
+         fprintf(stderr, "failed to create pixmap\n");
+         break;
+      }
+
+      if (__glxHashInsert(psc->drawHash, req->glxpixmap, pdraw)) {
+         (*pdraw->destroyDrawable) (pdraw);
+         return None;           /* FIXME: Check what we're supposed to do here... */
+      }
+   } while (0);
 #endif
 
-    return xid;
+   return xid;
 }
 
 /*
 ** Destroy the named pixmap
 */
-PUBLIC void glXDestroyGLXPixmap(Display *dpy, GLXPixmap glxpixmap)
-{
-    xGLXDestroyGLXPixmapReq *req;
-    CARD8 opcode;
-
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return;
-    }
-    
-    /* Send the glXDestroyGLXPixmap request */
-    LockDisplay(dpy);
-    GetReq(GLXDestroyGLXPixmap,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXDestroyGLXPixmap;
-    req->glxpixmap = glxpixmap;
-    UnlockDisplay(dpy);
-    SyncHandle();
+PUBLIC void
+glXDestroyGLXPixmap(Display * dpy, GLXPixmap glxpixmap)
+{
+   xGLXDestroyGLXPixmapReq *req;
+   CARD8 opcode;
+
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return;
+   }
+
+   /* Send the glXDestroyGLXPixmap request */
+   LockDisplay(dpy);
+   GetReq(GLXDestroyGLXPixmap, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXDestroyGLXPixmap;
+   req->glxpixmap = glxpixmap;
+   UnlockDisplay(dpy);
+   SyncHandle();
 
 #ifdef GLX_DIRECT_RENDERING
-    {
-	int screen;
-	__GLXdisplayPrivate *const priv = __glXInitialize(dpy);
-	__GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, glxpixmap, &screen);
-	__GLXscreenConfigs *psc = &priv->screenConfigs[screen];
-
-	if (pdraw != NULL) {
-	    (*pdraw->destroyDrawable) (pdraw);
-	    __glxHashDelete(psc->drawHash, glxpixmap);
-	}
-    }
+   {
+      int screen;
+      __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+      __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, glxpixmap, &screen);
+      __GLXscreenConfigs *psc = &priv->screenConfigs[screen];
+
+      if (pdraw != NULL) {
+         (*pdraw->destroyDrawable) (pdraw);
+         __glxHashDelete(psc->drawHash, glxpixmap);
+      }
+   }
 #endif
 }
 
-PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable)
+PUBLIC void
+glXSwapBuffers(Display * dpy, GLXDrawable drawable)
 {
-    GLXContext gc;
-    GLXContextTag tag;
-    CARD8 opcode;
+   GLXContext gc;
+   GLXContextTag tag;
+   CARD8 opcode;
 #ifdef USE_XCB
-    xcb_connection_t *c;
+   xcb_connection_t *c;
 #else
-    xGLXSwapBuffersReq *req;
+   xGLXSwapBuffersReq *req;
 #endif
 
 #ifdef GLX_DIRECT_RENDERING
-    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL);
+   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL);
 
-    if (pdraw != NULL) {
-	glFlush();	    
-	(*pdraw->psc->driScreen->swapBuffers)(pdraw);
-	return;
-    }
+   if (pdraw != NULL) {
+      glFlush();
+      (*pdraw->psc->driScreen->swapBuffers) (pdraw);
+      return;
+   }
 #endif
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return;
-    }
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return;
+   }
 
-    /*
+   /*
     ** The calling thread may or may not have a current context.  If it
     ** does, send the context tag so the server can do a flush.
     */
-    gc = __glXGetCurrentContext();
-    if ((gc != NULL) && (dpy == gc->currentDpy) && 
- 	((drawable == gc->currentDrawable) || (drawable == gc->currentReadable)) ) {
-	tag = gc->currentContextTag;
-    } else {
-	tag = 0;
-    }
+   gc = __glXGetCurrentContext();
+   if ((gc != NULL) && (dpy == gc->currentDpy) &&
+       ((drawable == gc->currentDrawable)
+        || (drawable == gc->currentReadable))) {
+      tag = gc->currentContextTag;
+   }
+   else {
+      tag = 0;
+   }
 
 #ifdef USE_XCB
-    c = XGetXCBConnection(dpy);
-    xcb_glx_swap_buffers(c, tag, drawable);
-    xcb_flush(c);
+   c = XGetXCBConnection(dpy);
+   xcb_glx_swap_buffers(c, tag, drawable);
+   xcb_flush(c);
 #else
-    /* Send the glXSwapBuffers request */
-    LockDisplay(dpy);
-    GetReq(GLXSwapBuffers,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXSwapBuffers;
-    req->drawable = drawable;
-    req->contextTag = tag;
-    UnlockDisplay(dpy);
-    SyncHandle();
-    XFlush(dpy);
+   /* Send the glXSwapBuffers request */
+   LockDisplay(dpy);
+   GetReq(GLXSwapBuffers, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXSwapBuffers;
+   req->drawable = drawable;
+   req->contextTag = tag;
+   UnlockDisplay(dpy);
+   SyncHandle();
+   XFlush(dpy);
 #endif /* USE_XCB */
 }
 
@@ -974,94 +1013,96 @@ PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable)
 ** Return configuration information for the given display, screen and
 ** visual combination.
 */
-PUBLIC int glXGetConfig(Display *dpy, XVisualInfo *vis, int attribute,
-			int *value_return)
-{
-    __GLXdisplayPrivate *priv;
-    __GLXscreenConfigs *psc;
-    __GLcontextModes *modes;
-    int   status;
-
-    status = GetGLXPrivScreenConfig( dpy, vis->screen, & priv, & psc );
-    if ( status == Success ) {
-	modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
-
-	/* Lookup attribute after first finding a match on the visual */
-	if ( modes != NULL ) {
-	    return _gl_get_context_mode_data( modes, attribute, value_return );
-	}
-	
-	status = GLX_BAD_VISUAL;
-    }
-
-    /*
+PUBLIC int
+glXGetConfig(Display * dpy, XVisualInfo * vis, int attribute,
+             int *value_return)
+{
+   __GLXdisplayPrivate *priv;
+   __GLXscreenConfigs *psc;
+   __GLcontextModes *modes;
+   int status;
+
+   status = GetGLXPrivScreenConfig(dpy, vis->screen, &priv, &psc);
+   if (status == Success) {
+      modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
+
+      /* Lookup attribute after first finding a match on the visual */
+      if (modes != NULL) {
+         return _gl_get_context_mode_data(modes, attribute, value_return);
+      }
+
+      status = GLX_BAD_VISUAL;
+   }
+
+   /*
     ** If we can't find the config for this visual, this visual is not
     ** supported by the OpenGL implementation on the server.
     */
-    if ( (status == GLX_BAD_VISUAL) && (attribute == GLX_USE_GL) ) {
-	*value_return = GL_FALSE;
-	status = Success;
-    }
+   if ((status == GLX_BAD_VISUAL) && (attribute == GLX_USE_GL)) {
+      *value_return = GL_FALSE;
+      status = Success;
+   }
 
-    return status;
+   return status;
 }
 
 /************************************************************************/
 
 static void
-init_fbconfig_for_chooser( __GLcontextModes * config,
-			   GLboolean fbconfig_style_tags )
-{
-    memset( config, 0, sizeof( __GLcontextModes ) );
-    config->visualID = (XID) GLX_DONT_CARE;
-    config->visualType = GLX_DONT_CARE;
-
-    /* glXChooseFBConfig specifies different defaults for these two than
-     * glXChooseVisual.
-     */
-    if ( fbconfig_style_tags ) {
-	config->rgbMode = GL_TRUE;
-	config->doubleBufferMode = GLX_DONT_CARE;
-    }
-
-    config->visualRating = GLX_DONT_CARE;
-    config->transparentPixel = GLX_NONE;
-    config->transparentRed = GLX_DONT_CARE;
-    config->transparentGreen = GLX_DONT_CARE;
-    config->transparentBlue = GLX_DONT_CARE;
-    config->transparentAlpha = GLX_DONT_CARE;
-    config->transparentIndex = GLX_DONT_CARE;
-
-    config->drawableType = GLX_WINDOW_BIT;
-    config->renderType = (config->rgbMode) ? GLX_RGBA_BIT : GLX_COLOR_INDEX_BIT;
-    config->xRenderable = GLX_DONT_CARE;
-    config->fbconfigID = (GLXFBConfigID)(GLX_DONT_CARE);
-
-    config->swapMethod = GLX_DONT_CARE;
-}
-
-#define MATCH_DONT_CARE( param ) \
-	do { \
-	    if ( (a-> param != GLX_DONT_CARE) \
-		 && (a-> param != b-> param) ) { \
-		return False; \
-	    } \
-	} while ( 0 )
-
-#define MATCH_MINIMUM( param ) \
-	do { \
-	    if ( (a-> param != GLX_DONT_CARE) \
-		 && (a-> param > b-> param) ) { \
-		return False; \
-	    } \
-	} while ( 0 )
-
-#define MATCH_EXACT( param ) \
-	do { \
-	    if ( a-> param != b-> param) { \
-		return False; \
-	    } \
-	} while ( 0 )
+init_fbconfig_for_chooser(__GLcontextModes * config,
+                          GLboolean fbconfig_style_tags)
+{
+   memset(config, 0, sizeof(__GLcontextModes));
+   config->visualID = (XID) GLX_DONT_CARE;
+   config->visualType = GLX_DONT_CARE;
+
+   /* glXChooseFBConfig specifies different defaults for these two than
+    * glXChooseVisual.
+    */
+   if (fbconfig_style_tags) {
+      config->rgbMode = GL_TRUE;
+      config->doubleBufferMode = GLX_DONT_CARE;
+   }
+
+   config->visualRating = GLX_DONT_CARE;
+   config->transparentPixel = GLX_NONE;
+   config->transparentRed = GLX_DONT_CARE;
+   config->transparentGreen = GLX_DONT_CARE;
+   config->transparentBlue = GLX_DONT_CARE;
+   config->transparentAlpha = GLX_DONT_CARE;
+   config->transparentIndex = GLX_DONT_CARE;
+
+   config->drawableType = GLX_WINDOW_BIT;
+   config->renderType =
+      (config->rgbMode) ? GLX_RGBA_BIT : GLX_COLOR_INDEX_BIT;
+   config->xRenderable = GLX_DONT_CARE;
+   config->fbconfigID = (GLXFBConfigID) (GLX_DONT_CARE);
+
+   config->swapMethod = GLX_DONT_CARE;
+}
+
+#define MATCH_DONT_CARE( param )        \
+  do {                                  \
+    if ( (a-> param != GLX_DONT_CARE)   \
+         && (a-> param != b-> param) ) {        \
+      return False;                             \
+    }                                           \
+  } while ( 0 )
+
+#define MATCH_MINIMUM( param )                  \
+  do {                                          \
+    if ( (a-> param != GLX_DONT_CARE)           \
+         && (a-> param > b-> param) ) {         \
+      return False;                             \
+    }                                           \
+  } while ( 0 )
+
+#define MATCH_EXACT( param )                    \
+  do {                                          \
+    if ( a-> param != b-> param) {              \
+      return False;                             \
+    }                                           \
+  } while ( 0 )
 
 /**
  * Determine if two GLXFBConfigs are compatible.
@@ -1070,80 +1111,80 @@ init_fbconfig_for_chooser( __GLcontextModes * config,
  * \param b  Server specified config to test against \c a.
  */
 static Bool
-fbconfigs_compatible( const __GLcontextModes * const a,
-		      const __GLcontextModes * const b )
-{
-    MATCH_DONT_CARE( doubleBufferMode );
-    MATCH_DONT_CARE( visualType );
-    MATCH_DONT_CARE( visualRating );
-    MATCH_DONT_CARE( xRenderable );
-    MATCH_DONT_CARE( fbconfigID );
-    MATCH_DONT_CARE( swapMethod );
-
-    MATCH_MINIMUM( rgbBits );
-    MATCH_MINIMUM( numAuxBuffers );
-    MATCH_MINIMUM( redBits );
-    MATCH_MINIMUM( greenBits );
-    MATCH_MINIMUM( blueBits );
-    MATCH_MINIMUM( alphaBits );
-    MATCH_MINIMUM( depthBits );
-    MATCH_MINIMUM( stencilBits );
-    MATCH_MINIMUM( accumRedBits );
-    MATCH_MINIMUM( accumGreenBits );
-    MATCH_MINIMUM( accumBlueBits );
-    MATCH_MINIMUM( accumAlphaBits );
-    MATCH_MINIMUM( sampleBuffers );
-    MATCH_MINIMUM( maxPbufferWidth );
-    MATCH_MINIMUM( maxPbufferHeight );
-    MATCH_MINIMUM( maxPbufferPixels );
-    MATCH_MINIMUM( samples );
-
-    MATCH_DONT_CARE( stereoMode );
-    MATCH_EXACT( level );
-
-    if ( ((a->drawableType & b->drawableType) == 0)
-	 || ((a->renderType & b->renderType) == 0) ) {
-	return False;
-    }
-
-
-    /* There is a bug in a few of the XFree86 DDX drivers.  They contain
-     * visuals with a "transparent type" of 0 when they really mean GLX_NONE.
-     * Technically speaking, it is a bug in the DDX driver, but there is
-     * enough of an installed base to work around the problem here.  In any
-     * case, 0 is not a valid value of the transparent type, so we'll treat 0 
-     * from the app as GLX_DONT_CARE. We'll consider GLX_NONE from the app and
-     * 0 from the server to be a match to maintain backward compatibility with
-     * the (broken) drivers.
-     */
-
-    if ( a->transparentPixel != GLX_DONT_CARE
-         && a->transparentPixel != 0 ) {
-        if ( a->transparentPixel == GLX_NONE ) {
-            if ( b->transparentPixel != GLX_NONE && b->transparentPixel != 0 )
-                return False;
-        } else {
-            MATCH_EXACT( transparentPixel );
-        }
-
-	switch ( a->transparentPixel ) {
-	  case GLX_TRANSPARENT_RGB:
-	    MATCH_DONT_CARE( transparentRed );
-	    MATCH_DONT_CARE( transparentGreen );
-	    MATCH_DONT_CARE( transparentBlue );
-	    MATCH_DONT_CARE( transparentAlpha );
-	    break;
-
-	  case GLX_TRANSPARENT_INDEX:
-	    MATCH_DONT_CARE( transparentIndex );
-	    break;
-
-	  default:
-	    break;
-	}
-    }
-
-    return True;
+fbconfigs_compatible(const __GLcontextModes * const a,
+                     const __GLcontextModes * const b)
+{
+   MATCH_DONT_CARE(doubleBufferMode);
+   MATCH_DONT_CARE(visualType);
+   MATCH_DONT_CARE(visualRating);
+   MATCH_DONT_CARE(xRenderable);
+   MATCH_DONT_CARE(fbconfigID);
+   MATCH_DONT_CARE(swapMethod);
+
+   MATCH_MINIMUM(rgbBits);
+   MATCH_MINIMUM(numAuxBuffers);
+   MATCH_MINIMUM(redBits);
+   MATCH_MINIMUM(greenBits);
+   MATCH_MINIMUM(blueBits);
+   MATCH_MINIMUM(alphaBits);
+   MATCH_MINIMUM(depthBits);
+   MATCH_MINIMUM(stencilBits);
+   MATCH_MINIMUM(accumRedBits);
+   MATCH_MINIMUM(accumGreenBits);
+   MATCH_MINIMUM(accumBlueBits);
+   MATCH_MINIMUM(accumAlphaBits);
+   MATCH_MINIMUM(sampleBuffers);
+   MATCH_MINIMUM(maxPbufferWidth);
+   MATCH_MINIMUM(maxPbufferHeight);
+   MATCH_MINIMUM(maxPbufferPixels);
+   MATCH_MINIMUM(samples);
+
+   MATCH_DONT_CARE(stereoMode);
+   MATCH_EXACT(level);
+
+   if (((a->drawableType & b->drawableType) == 0)
+       || ((a->renderType & b->renderType) == 0)) {
+      return False;
+   }
+
+
+   /* There is a bug in a few of the XFree86 DDX drivers.  They contain
+    * visuals with a "transparent type" of 0 when they really mean GLX_NONE.
+    * Technically speaking, it is a bug in the DDX driver, but there is
+    * enough of an installed base to work around the problem here.  In any
+    * case, 0 is not a valid value of the transparent type, so we'll treat 0
+    * from the app as GLX_DONT_CARE. We'll consider GLX_NONE from the app and
+    * 0 from the server to be a match to maintain backward compatibility with
+    * the (broken) drivers.
+    */
+
+   if (a->transparentPixel != GLX_DONT_CARE && a->transparentPixel != 0) {
+      if (a->transparentPixel == GLX_NONE) {
+         if (b->transparentPixel != GLX_NONE && b->transparentPixel != 0)
+            return False;
+      }
+      else {
+         MATCH_EXACT(transparentPixel);
+      }
+
+      switch (a->transparentPixel) {
+      case GLX_TRANSPARENT_RGB:
+         MATCH_DONT_CARE(transparentRed);
+         MATCH_DONT_CARE(transparentGreen);
+         MATCH_DONT_CARE(transparentBlue);
+         MATCH_DONT_CARE(transparentAlpha);
+         break;
+
+      case GLX_TRANSPARENT_INDEX:
+         MATCH_DONT_CARE(transparentIndex);
+         break;
+
+      default:
+         break;
+      }
+   }
+
+   return True;
 }
 
 
@@ -1153,39 +1194,39 @@ fbconfigs_compatible( const __GLcontextModes * const a,
  * Well, that's really hard to do with the code as-is.  This behavior is
  * closer to correct, but still not technically right.
  */
-#define PREFER_LARGER_OR_ZERO(comp) \
-    do { \
-	if ( ((*a)-> comp) != ((*b)-> comp) ) { \
-	    if ( ((*a)-> comp) == 0 ) { \
-		return -1; \
-	    } \
-	    else if ( ((*b)-> comp) == 0 ) { \
-		return 1; \
-	    } \
-	    else { \
-		return ((*b)-> comp) - ((*a)-> comp) ; \
-	    } \
-	} \
-    } while( 0 )
-
-#define PREFER_LARGER(comp) \
-    do { \
-	if ( ((*a)-> comp) != ((*b)-> comp) ) { \
-	    return ((*b)-> comp) - ((*a)-> comp) ; \
-	} \
-    } while( 0 )
-
-#define PREFER_SMALLER(comp) \
-    do { \
-	if ( ((*a)-> comp) != ((*b)-> comp) ) { \
-	    return ((*a)-> comp) - ((*b)-> comp) ; \
-	} \
-    } while( 0 )
+#define PREFER_LARGER_OR_ZERO(comp)             \
+  do {                                          \
+    if ( ((*a)-> comp) != ((*b)-> comp) ) {     \
+      if ( ((*a)-> comp) == 0 ) {               \
+        return -1;                              \
+      }                                         \
+      else if ( ((*b)-> comp) == 0 ) {          \
+        return 1;                               \
+      }                                         \
+      else {                                    \
+        return ((*b)-> comp) - ((*a)-> comp) ;  \
+      }                                         \
+    }                                           \
+  } while( 0 )
+
+#define PREFER_LARGER(comp)                     \
+  do {                                          \
+    if ( ((*a)-> comp) != ((*b)-> comp) ) {     \
+      return ((*b)-> comp) - ((*a)-> comp) ;    \
+    }                                           \
+  } while( 0 )
+
+#define PREFER_SMALLER(comp)                    \
+  do {                                          \
+    if ( ((*a)-> comp) != ((*b)-> comp) ) {     \
+      return ((*a)-> comp) - ((*b)-> comp) ;    \
+    }                                           \
+  } while( 0 )
 
 /**
  * Compare two GLXFBConfigs.  This function is intended to be used as the
  * compare function passed in to qsort.
- * 
+ *
  * \returns If \c a is a "better" config, according to the specification of
  *          SGIX_fbconfig, a number less than zero is returned.  If \c b is
  *          better, then a number greater than zero is return.  If both are
@@ -1193,66 +1234,66 @@ fbconfigs_compatible( const __GLcontextModes * const a,
  * \sa qsort, glXChooseVisual, glXChooseFBConfig, glXChooseFBConfigSGIX
  */
 static int
-fbconfig_compare( const __GLcontextModes * const * const a,
-		  const __GLcontextModes * const * const b )
+fbconfig_compare(const __GLcontextModes * const *const a,
+                 const __GLcontextModes * const *const b)
 {
-    /* The order of these comparisons must NOT change.  It is defined by
-     * the GLX 1.3 spec and ARB_multisample.
-     */
+   /* The order of these comparisons must NOT change.  It is defined by
+    * the GLX 1.3 spec and ARB_multisample.
+    */
 
-    PREFER_SMALLER( visualSelectGroup );
+   PREFER_SMALLER(visualSelectGroup);
 
-    /* The sort order for the visualRating is GLX_NONE, GLX_SLOW, and
-     * GLX_NON_CONFORMANT_CONFIG.  It just so happens that this is the
-     * numerical sort order of the enums (0x8000, 0x8001, and 0x800D).
-     */
-    PREFER_SMALLER( visualRating );
+   /* The sort order for the visualRating is GLX_NONE, GLX_SLOW, and
+    * GLX_NON_CONFORMANT_CONFIG.  It just so happens that this is the
+    * numerical sort order of the enums (0x8000, 0x8001, and 0x800D).
+    */
+   PREFER_SMALLER(visualRating);
 
-    /* This isn't quite right.  It is supposed to compare the sum of the
-     * components the user specifically set minimums for.
-     */
-    PREFER_LARGER_OR_ZERO( redBits );
-    PREFER_LARGER_OR_ZERO( greenBits );
-    PREFER_LARGER_OR_ZERO( blueBits );
-    PREFER_LARGER_OR_ZERO( alphaBits );
+   /* This isn't quite right.  It is supposed to compare the sum of the
+    * components the user specifically set minimums for.
+    */
+   PREFER_LARGER_OR_ZERO(redBits);
+   PREFER_LARGER_OR_ZERO(greenBits);
+   PREFER_LARGER_OR_ZERO(blueBits);
+   PREFER_LARGER_OR_ZERO(alphaBits);
 
-    PREFER_SMALLER( rgbBits );
+   PREFER_SMALLER(rgbBits);
 
-    if ( ((*a)->doubleBufferMode != (*b)->doubleBufferMode) ) {
-	/* Prefer single-buffer.
-	 */
-	return ( !(*a)->doubleBufferMode ) ? -1 : 1;
-    }
+   if (((*a)->doubleBufferMode != (*b)->doubleBufferMode)) {
+      /* Prefer single-buffer.
+       */
+      return (!(*a)->doubleBufferMode) ? -1 : 1;
+   }
 
-    PREFER_SMALLER( numAuxBuffers );
+   PREFER_SMALLER(numAuxBuffers);
 
-    PREFER_LARGER_OR_ZERO( depthBits );
-    PREFER_SMALLER( stencilBits );
+   PREFER_LARGER_OR_ZERO(depthBits);
+   PREFER_SMALLER(stencilBits);
 
-    /* This isn't quite right.  It is supposed to compare the sum of the
-     * components the user specifically set minimums for.
-     */
-    PREFER_LARGER_OR_ZERO( accumRedBits );
-    PREFER_LARGER_OR_ZERO( accumGreenBits );
-    PREFER_LARGER_OR_ZERO( accumBlueBits );
-    PREFER_LARGER_OR_ZERO( accumAlphaBits );
+   /* This isn't quite right.  It is supposed to compare the sum of the
+    * components the user specifically set minimums for.
+    */
+   PREFER_LARGER_OR_ZERO(accumRedBits);
+   PREFER_LARGER_OR_ZERO(accumGreenBits);
+   PREFER_LARGER_OR_ZERO(accumBlueBits);
+   PREFER_LARGER_OR_ZERO(accumAlphaBits);
 
-    PREFER_SMALLER( visualType );
+   PREFER_SMALLER(visualType);
 
-    /* None of the multisample specs say where this comparison should happen,
-     * so I put it near the end.
-     */
-    PREFER_SMALLER( sampleBuffers );
-    PREFER_SMALLER( samples );
+   /* None of the multisample specs say where this comparison should happen,
+    * so I put it near the end.
+    */
+   PREFER_SMALLER(sampleBuffers);
+   PREFER_SMALLER(samples);
 
-    /* None of the pbuffer or fbconfig specs say that this comparison needs
-     * to happen at all, but it seems like it should.
-     */
-    PREFER_LARGER( maxPbufferWidth );
-    PREFER_LARGER( maxPbufferHeight );
-    PREFER_LARGER( maxPbufferPixels );
+   /* None of the pbuffer or fbconfig specs say that this comparison needs
+    * to happen at all, but it seems like it should.
+    */
+   PREFER_LARGER(maxPbufferWidth);
+   PREFER_LARGER(maxPbufferHeight);
+   PREFER_LARGER(maxPbufferPixels);
 
-    return 0;
+   return 0;
 }
 
 
@@ -1260,7 +1301,7 @@ fbconfig_compare( const __GLcontextModes * const * const a,
  * Selects and sorts a subset of the supplied configs based on the attributes.
  * This function forms to basis of \c glXChooseVisual, \c glXChooseFBConfig,
  * and \c glXChooseFBConfigSGIX.
- * 
+ *
  * \param configs   Array of pointers to possible configs.  The elements of
  *                  this array that do not meet the criteria will be set to
  *                  NULL.  The remaining elements will be sorted according to
@@ -1275,53 +1316,52 @@ fbconfig_compare( const __GLcontextModes * const * const a,
  *                             \c glXChooseVisual style or
  *                             \c glXChooseFBConfig style.
  * \returns The number of valid elements left in \c configs.
- * 
+ *
  * \sa glXChooseVisual, glXChooseFBConfig, glXChooseFBConfigSGIX
  */
 static int
-choose_visual( __GLcontextModes ** configs, int num_configs,
-	       const int *attribList, GLboolean fbconfig_style_tags )
-{
-    __GLcontextModes    test_config;
-    int   base;
-    int   i;
-
-    /* This is a fairly direct implementation of the selection method
-     * described by GLX_SGIX_fbconfig.  Start by culling out all the
-     * configs that are not compatible with the selected parameter
-     * list.
-     */
-
-    init_fbconfig_for_chooser( & test_config, fbconfig_style_tags );
-    __glXInitializeVisualConfigFromTags( & test_config, 512,
-					 (const INT32 *) attribList,
-					 GL_TRUE, fbconfig_style_tags );
-
-    base = 0;
-    for ( i = 0 ; i < num_configs ; i++ ) {
-	if ( fbconfigs_compatible( & test_config, configs[i] ) ) {
-	    configs[ base ] = configs[ i ];
-	    base++;
-	}
-    }
-
-    if ( base == 0 ) {
-	return 0;
-    }
- 
-    if ( base < num_configs ) {
-	(void) memset( & configs[ base ], 0, 
-		       sizeof( void * ) * (num_configs - base) );
-    }
-
-    /* After the incompatible configs are removed, the resulting
-     * list is sorted according to the rules set out in the various
-     * specifications.
-     */
-    
-    qsort( configs, base, sizeof( __GLcontextModes * ),
-	   (int (*)(const void*, const void*)) fbconfig_compare );
-    return base;
+choose_visual(__GLcontextModes ** configs, int num_configs,
+              const int *attribList, GLboolean fbconfig_style_tags)
+{
+   __GLcontextModes test_config;
+   int base;
+   int i;
+
+   /* This is a fairly direct implementation of the selection method
+    * described by GLX_SGIX_fbconfig.  Start by culling out all the
+    * configs that are not compatible with the selected parameter
+    * list.
+    */
+
+   init_fbconfig_for_chooser(&test_config, fbconfig_style_tags);
+   __glXInitializeVisualConfigFromTags(&test_config, 512,
+                                       (const INT32 *) attribList,
+                                       GL_TRUE, fbconfig_style_tags);
+
+   base = 0;
+   for (i = 0; i < num_configs; i++) {
+      if (fbconfigs_compatible(&test_config, configs[i])) {
+         configs[base] = configs[i];
+         base++;
+      }
+   }
+
+   if (base == 0) {
+      return 0;
+   }
+
+   if (base < num_configs) {
+      (void) memset(&configs[base], 0, sizeof(void *) * (num_configs - base));
+   }
+
+   /* After the incompatible configs are removed, the resulting
+    * list is sorted according to the rules set out in the various
+    * specifications.
+    */
+
+   qsort(configs, base, sizeof(__GLcontextModes *),
+         (int (*)(const void *, const void *)) fbconfig_compare);
+   return base;
 }
 
 
@@ -1331,33 +1371,34 @@ choose_visual( __GLcontextModes ** configs, int num_configs,
 ** Return the visual that best matches the template.  Return None if no
 ** visual matches the template.
 */
-PUBLIC XVisualInfo *glXChooseVisual(Display *dpy, int screen, int *attribList)
+PUBLIC XVisualInfo *
+glXChooseVisual(Display * dpy, int screen, int *attribList)
 {
-    XVisualInfo *visualList = NULL;
-    __GLXdisplayPrivate *priv;
-    __GLXscreenConfigs *psc;
-    __GLcontextModes  test_config;
-    __GLcontextModes *modes;
-    const __GLcontextModes *best_config = NULL;
+   XVisualInfo *visualList = NULL;
+   __GLXdisplayPrivate *priv;
+   __GLXscreenConfigs *psc;
+   __GLcontextModes test_config;
+   __GLcontextModes *modes;
+   const __GLcontextModes *best_config = NULL;
 
-    /*
+   /*
     ** Get a list of all visuals, return if list is empty
     */
-    if ( GetGLXPrivScreenConfig( dpy, screen, & priv, & psc ) != Success ) {
-	return None;
-    }
-   
+   if (GetGLXPrivScreenConfig(dpy, screen, &priv, &psc) != Success) {
+      return None;
+   }
+
 
-    /*
+   /*
     ** Build a template from the defaults and the attribute list
     ** Free visual list and return if an unexpected token is encountered
     */
-    init_fbconfig_for_chooser( & test_config, GL_FALSE );
-    __glXInitializeVisualConfigFromTags( & test_config, 512, 
-					 (const INT32 *) attribList,
-					 GL_TRUE, GL_FALSE );
+   init_fbconfig_for_chooser(&test_config, GL_FALSE);
+   __glXInitializeVisualConfigFromTags(&test_config, 512,
+                                       (const INT32 *) attribList,
+                                       GL_TRUE, GL_FALSE);
 
-    /*
+   /*
     ** Eliminate visuals that don't meet minimum requirements
     ** Compute a score for those that do
     ** Remember which visual, if any, got the highest score
@@ -1365,136 +1406,141 @@ PUBLIC XVisualInfo *glXChooseVisual(Display *dpy, int screen, int *attribList)
     ** Otherwise, create an XVisualInfo list with just the selected X visual
     ** and return this.
     */
-    for ( modes = psc->visuals ; modes != NULL ; modes = modes->next ) {
-	if ( fbconfigs_compatible( & test_config, modes )
-	     && ((best_config == NULL)
-		 || (fbconfig_compare( (const __GLcontextModes * const * const)&modes, &best_config ) < 0)) ) {
-	    XVisualInfo visualTemplate;
-	    XVisualInfo *newList;
-	    int  i;
-
-	    visualTemplate.screen = screen;
-	    visualTemplate.visualid = modes->visualID;
-	    newList = XGetVisualInfo( dpy, VisualScreenMask|VisualIDMask,
-				      &visualTemplate, &i );
-
-	    if (newList) {
-		Xfree(visualList);
-		visualList = newList;
-		best_config = modes;
-	    }
-	}
-    }
+   for (modes = psc->visuals; modes != NULL; modes = modes->next) {
+      if (fbconfigs_compatible(&test_config, modes)
+          && ((best_config == NULL)
+              ||
+              (fbconfig_compare
+               ((const __GLcontextModes * const *const) &modes,
+                &best_config) < 0))) {
+         XVisualInfo visualTemplate;
+         XVisualInfo *newList;
+         int i;
+
+         visualTemplate.screen = screen;
+         visualTemplate.visualid = modes->visualID;
+         newList = XGetVisualInfo(dpy, VisualScreenMask | VisualIDMask,
+                                  &visualTemplate, &i);
+
+         if (newList) {
+            Xfree(visualList);
+            visualList = newList;
+            best_config = modes;
+         }
+      }
+   }
 
-    return visualList;
+   return visualList;
 }
 
 
-PUBLIC const char *glXQueryExtensionsString( Display *dpy, int screen )
+PUBLIC const char *
+glXQueryExtensionsString(Display * dpy, int screen)
 {
-    __GLXscreenConfigs *psc;
-    __GLXdisplayPrivate *priv;
+   __GLXscreenConfigs *psc;
+   __GLXdisplayPrivate *priv;
 
-    if ( GetGLXPrivScreenConfig( dpy, screen, & priv, & psc ) != Success ) {
-	return NULL;
-    }
+   if (GetGLXPrivScreenConfig(dpy, screen, &priv, &psc) != Success) {
+      return NULL;
+   }
 
-    if (!psc->effectiveGLXexts) {
-        if (!psc->serverGLXexts) {
-           psc->serverGLXexts =
-              __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
-	}
+   if (!psc->effectiveGLXexts) {
+      if (!psc->serverGLXexts) {
+         psc->serverGLXexts =
+            __glXQueryServerString(dpy, priv->majorOpcode, screen,
+                                   GLX_EXTENSIONS);
+      }
 
-	__glXCalculateUsableExtensions(psc,
+      __glXCalculateUsableExtensions(psc,
 #ifdef GLX_DIRECT_RENDERING
-				       (psc->driScreen != NULL),
+                                     (psc->driScreen != NULL),
 #else
-				       GL_FALSE,
+                                     GL_FALSE,
 #endif
-				       priv->minorVersion);
-    }
+                                     priv->minorVersion);
+   }
 
-    return psc->effectiveGLXexts;
+   return psc->effectiveGLXexts;
 }
 
-PUBLIC const char *glXGetClientString( Display *dpy, int name )
+PUBLIC const char *
+glXGetClientString(Display * dpy, int name)
 {
-    switch(name) {
-	case GLX_VENDOR:
-	    return (__glXGLXClientVendorName);
-	case GLX_VERSION:
-	    return (__glXGLXClientVersion);
-	case GLX_EXTENSIONS:
-	    return (__glXGetClientExtensions());
-	default:
-	    return NULL;
-    }
+   switch (name) {
+   case GLX_VENDOR:
+      return (__glXGLXClientVendorName);
+   case GLX_VERSION:
+      return (__glXGLXClientVersion);
+   case GLX_EXTENSIONS:
+      return (__glXGetClientExtensions());
+   default:
+      return NULL;
+   }
 }
 
-PUBLIC const char *glXQueryServerString( Display *dpy, int screen, int name )
+PUBLIC const char *
+glXQueryServerString(Display * dpy, int screen, int name)
 {
-    __GLXscreenConfigs *psc;
-    __GLXdisplayPrivate *priv;
-    const char ** str;
+   __GLXscreenConfigs *psc;
+   __GLXdisplayPrivate *priv;
+   const char **str;
 
 
-    if ( GetGLXPrivScreenConfig( dpy, screen, & priv, & psc ) != Success ) {
-	return NULL;
-    }
+   if (GetGLXPrivScreenConfig(dpy, screen, &priv, &psc) != Success) {
+      return NULL;
+   }
+
+   switch (name) {
+   case GLX_VENDOR:
+      str = &priv->serverGLXvendor;
+      break;
+   case GLX_VERSION:
+      str = &priv->serverGLXversion;
+      break;
+   case GLX_EXTENSIONS:
+      str = &psc->serverGLXexts;
+      break;
+   default:
+      return NULL;
+   }
 
-    switch(name) {
-	case GLX_VENDOR:
-	    str = & priv->serverGLXvendor;
-	    break;
-	case GLX_VERSION:
-	    str = & priv->serverGLXversion;
-	    break;
-	case GLX_EXTENSIONS:
-	    str = & psc->serverGLXexts;
-	    break;
-	default:
-	    return NULL;
-    }
+   if (*str == NULL) {
+      *str = __glXQueryServerString(dpy, priv->majorOpcode, screen, name);
+   }
 
-    if ( *str == NULL ) {
-       *str = __glXQueryServerString(dpy, priv->majorOpcode, screen, name);
-    }
-    
-    return *str;
+   return *str;
 }
 
-void __glXClientInfo (  Display *dpy, int opcode  )
+void
+__glXClientInfo(Display * dpy, int opcode)
 {
-    char * ext_str = __glXGetClientGLExtensionString();
-    int size = strlen( ext_str ) + 1;
+   char *ext_str = __glXGetClientGLExtensionString();
+   int size = strlen(ext_str) + 1;
 
 #ifdef USE_XCB
    xcb_connection_t *c = XGetXCBConnection(dpy);
    xcb_glx_client_info(c,
-                       GLX_MAJOR_VERSION,
-                       GLX_MINOR_VERSION,
-                       size,
-                       (const uint8_t *)ext_str);
+                       GLX_MAJOR_VERSION, GLX_MINOR_VERSION, size, ext_str);
 #else
-    xGLXClientInfoReq *req;
-
-    /* Send the glXClientInfo request */
-    LockDisplay(dpy);
-    GetReq(GLXClientInfo,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXClientInfo;
-    req->major = GLX_MAJOR_VERSION;
-    req->minor = GLX_MINOR_VERSION;
-
-    req->length += (size + 3) >> 2;
-    req->numbytes = size;
-    Data(dpy, ext_str, size);
-
-    UnlockDisplay(dpy);
-    SyncHandle();
+   xGLXClientInfoReq *req;
+
+   /* Send the glXClientInfo request */
+   LockDisplay(dpy);
+   GetReq(GLXClientInfo, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXClientInfo;
+   req->major = GLX_MAJOR_VERSION;
+   req->minor = GLX_MINOR_VERSION;
+
+   req->length += (size + 3) >> 2;
+   req->numbytes = size;
+   Data(dpy, ext_str, size);
+
+   UnlockDisplay(dpy);
+   SyncHandle();
 #endif /* USE_XCB */
 
-    Xfree( ext_str );
+   Xfree(ext_str);
 }
 
 
@@ -1502,194 +1548,201 @@ void __glXClientInfo (  Display *dpy, int opcode  )
 ** EXT_import_context
 */
 
-PUBLIC Display *glXGetCurrentDisplay(void)
+PUBLIC Display *
+glXGetCurrentDisplay(void)
 {
-    GLXContext gc = __glXGetCurrentContext();
-    if (NULL == gc) return NULL;
-    return gc->currentDpy;
+   GLXContext gc = __glXGetCurrentContext();
+   if (NULL == gc)
+      return NULL;
+   return gc->currentDpy;
 }
 
-PUBLIC GLX_ALIAS(Display *, glXGetCurrentDisplayEXT, (void), (),
-	  glXGetCurrentDisplay)
+PUBLIC
+GLX_ALIAS(Display *, glXGetCurrentDisplayEXT, (void), (),
+          glXGetCurrentDisplay)
 
 /**
  * Used internally by libGL to send \c xGLXQueryContextinfoExtReq requests
  * to the X-server.
- * 
+ *
  * \param dpy  Display where \c ctx was created.
  * \param ctx  Context to query.
  * \returns  \c Success on success.  \c GLX_BAD_CONTEXT if \c ctx is invalid,
  *           or zero if the request failed due to internal problems (i.e.,
  *           unable to allocate temporary memory, etc.)
- * 
+ *
  * \note
  * This function dynamically determines whether to use the EXT_import_context
  * version of the protocol or the GLX 1.3 version of the protocol.
  */
-static int __glXQueryContextInfo(Display *dpy, GLXContext ctx)
-{
-    __GLXdisplayPrivate *priv = __glXInitialize(dpy);
-    xGLXQueryContextReply reply;
-    CARD8 opcode;
-    GLuint numValues;
-    int retval;
-
-    if (ctx == NULL) {
-	return GLX_BAD_CONTEXT;
-    }
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode) {
-	return 0;
-    }
-
-    /* Send the glXQueryContextInfoEXT request */
-    LockDisplay(dpy);
-
-    if ( (priv->majorVersion > 1) || (priv->minorVersion >= 3) ) {
-	xGLXQueryContextReq *req;
-
-	GetReq(GLXQueryContext, req);
-
-	req->reqType = opcode;
-	req->glxCode = X_GLXQueryContext;
-	req->context = (unsigned int)(ctx->xid);
-    }
-    else {
-	xGLXVendorPrivateReq *vpreq;
-	xGLXQueryContextInfoEXTReq *req;
-
-	GetReqExtra( GLXVendorPrivate,
-		     sz_xGLXQueryContextInfoEXTReq - sz_xGLXVendorPrivateReq,
-		     vpreq );
-	req = (xGLXQueryContextInfoEXTReq *)vpreq;
-	req->reqType = opcode;
-	req->glxCode = X_GLXVendorPrivateWithReply;
-	req->vendorCode = X_GLXvop_QueryContextInfoEXT;
-	req->context = (unsigned int)(ctx->xid);
-    }
-
-    _XReply(dpy, (xReply*) &reply, 0, False);
-
-    numValues = reply.n;
-    if (numValues == 0)
-	retval = Success;
-    else if (numValues > __GLX_MAX_CONTEXT_PROPS)
-	retval = 0;
-    else
-    {
-	int *propList, *pProp;
-	int nPropListBytes;
-	int i;
-
-	nPropListBytes = numValues << 3;
-	propList = (int *) Xmalloc(nPropListBytes);
-	if (NULL == propList) {
-	    retval = 0;
-	} else {
-	    _XRead(dpy, (char *)propList, nPropListBytes);
-	    pProp = propList;
-	    for (i=0; i < numValues; i++) {
-		switch (*pProp++) {
-		case GLX_SHARE_CONTEXT_EXT:
-		    ctx->share_xid = *pProp++;
-		    break;
-		case GLX_VISUAL_ID_EXT:
-		    ctx->mode =
-			_gl_context_modes_find_visual(ctx->psc->visuals, *pProp++);
-		    break;
-		case GLX_SCREEN:
-		    ctx->screen = *pProp++;
-		    break;
-		case GLX_FBCONFIG_ID:
-		    ctx->mode =
-			_gl_context_modes_find_fbconfig(ctx->psc->configs, *pProp++);
-		    break;
-		case GLX_RENDER_TYPE:
-		    ctx->renderType = *pProp++;
-		    break;
-		default:
-		    pProp++;
-		    continue;
-		}
-	    }
-	    Xfree((char *)propList);
-	    retval = Success;
-	}
-    }
-    UnlockDisplay(dpy);
-    SyncHandle();
-    return retval;
+     static int __glXQueryContextInfo(Display * dpy, GLXContext ctx)
+{
+   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   xGLXQueryContextReply reply;
+   CARD8 opcode;
+   GLuint numValues;
+   int retval;
+
+   if (ctx == NULL) {
+      return GLX_BAD_CONTEXT;
+   }
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode) {
+      return 0;
+   }
+
+   /* Send the glXQueryContextInfoEXT request */
+   LockDisplay(dpy);
+
+   if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) {
+      xGLXQueryContextReq *req;
+
+      GetReq(GLXQueryContext, req);
+
+      req->reqType = opcode;
+      req->glxCode = X_GLXQueryContext;
+      req->context = (unsigned int) (ctx->xid);
+   }
+   else {
+      xGLXVendorPrivateReq *vpreq;
+      xGLXQueryContextInfoEXTReq *req;
+
+      GetReqExtra(GLXVendorPrivate,
+                  sz_xGLXQueryContextInfoEXTReq - sz_xGLXVendorPrivateReq,
+                  vpreq);
+      req = (xGLXQueryContextInfoEXTReq *) vpreq;
+      req->reqType = opcode;
+      req->glxCode = X_GLXVendorPrivateWithReply;
+      req->vendorCode = X_GLXvop_QueryContextInfoEXT;
+      req->context = (unsigned int) (ctx->xid);
+   }
+
+   _XReply(dpy, (xReply *) & reply, 0, False);
+
+   numValues = reply.n;
+   if (numValues == 0)
+      retval = Success;
+   else if (numValues > __GLX_MAX_CONTEXT_PROPS)
+      retval = 0;
+   else {
+      int *propList, *pProp;
+      int nPropListBytes;
+      int i;
+
+      nPropListBytes = numValues << 3;
+      propList = (int *) Xmalloc(nPropListBytes);
+      if (NULL == propList) {
+         retval = 0;
+      }
+      else {
+         _XRead(dpy, (char *) propList, nPropListBytes);
+         pProp = propList;
+         for (i = 0; i < numValues; i++) {
+            switch (*pProp++) {
+            case GLX_SHARE_CONTEXT_EXT:
+               ctx->share_xid = *pProp++;
+               break;
+            case GLX_VISUAL_ID_EXT:
+               ctx->mode =
+                  _gl_context_modes_find_visual(ctx->psc->visuals, *pProp++);
+               break;
+            case GLX_SCREEN:
+               ctx->screen = *pProp++;
+               break;
+            case GLX_FBCONFIG_ID:
+               ctx->mode =
+                  _gl_context_modes_find_fbconfig(ctx->psc->configs,
+                                                  *pProp++);
+               break;
+            case GLX_RENDER_TYPE:
+               ctx->renderType = *pProp++;
+               break;
+            default:
+               pProp++;
+               continue;
+            }
+         }
+         Xfree((char *) propList);
+         retval = Success;
+      }
+   }
+   UnlockDisplay(dpy);
+   SyncHandle();
+   return retval;
 }
 
 PUBLIC int
-glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value)
+glXQueryContext(Display * dpy, GLXContext ctx, int attribute, int *value)
 {
-    int retVal;
+   int retVal;
 
-    /* get the information from the server if we don't have it already */
+   /* get the information from the server if we don't have it already */
 #ifdef GLX_DIRECT_RENDERING
-    if (!ctx->driContext && (ctx->mode == NULL)) {
+   if (!ctx->driContext && (ctx->mode == NULL)) {
 #else
-    if (ctx->mode == NULL) {
+   if (ctx->mode == NULL) {
 #endif
-	retVal = __glXQueryContextInfo(dpy, ctx);
-	if (Success != retVal) return retVal;
-    }
-    switch (attribute) {
-    case GLX_SHARE_CONTEXT_EXT:
-	*value = (int)(ctx->share_xid);
-	break;
-    case GLX_VISUAL_ID_EXT:
-	*value = ctx->mode ? ctx->mode->visualID : None;
-	break;
-    case GLX_SCREEN:
-	*value = (int)(ctx->screen);
-	break;
-    case GLX_FBCONFIG_ID:
-	*value = ctx->mode ? ctx->mode->fbconfigID : None;
-	break;
-    case GLX_RENDER_TYPE:
-	*value = (int)(ctx->renderType);
-	break;
-    default:
-	return GLX_BAD_ATTRIBUTE;
-    }
-    return Success;
+      retVal = __glXQueryContextInfo(dpy, ctx);
+      if (Success != retVal)
+         return retVal;
+   }
+   switch (attribute) {
+   case GLX_SHARE_CONTEXT_EXT:
+      *value = (int) (ctx->share_xid);
+      break;
+   case GLX_VISUAL_ID_EXT:
+      *value = ctx->mode ? ctx->mode->visualID : None;
+      break;
+   case GLX_SCREEN:
+      *value = (int) (ctx->screen);
+      break;
+   case GLX_FBCONFIG_ID:
+      *value = ctx->mode ? ctx->mode->fbconfigID : None;
+      break;
+   case GLX_RENDER_TYPE:
+      *value = (int) (ctx->renderType);
+      break;
+   default:
+      return GLX_BAD_ATTRIBUTE;
+   }
+   return Success;
 }
 
-PUBLIC GLX_ALIAS( int, glXQueryContextInfoEXT,
-	   (Display *dpy, GLXContext ctx, int attribute, int *value),
-	   (dpy, ctx, attribute, value),
-	   glXQueryContext )
+PUBLIC
+GLX_ALIAS(int, glXQueryContextInfoEXT,
+          (Display * dpy, GLXContext ctx, int attribute, int *value),
+          (dpy, ctx, attribute, value), glXQueryContext)
 
-PUBLIC GLXContextID glXGetContextIDEXT(const GLXContext ctx)
+     PUBLIC GLXContextID glXGetContextIDEXT(const GLXContext ctx)
 {
-    return ctx->xid;
+   return ctx->xid;
 }
 
-PUBLIC GLXContext glXImportContextEXT(Display *dpy, GLXContextID contextID)
+PUBLIC GLXContext
+glXImportContextEXT(Display * dpy, GLXContextID contextID)
 {
-    GLXContext ctx;
+   GLXContext ctx;
 
-    if (contextID == None) {
-	return NULL;
-    }
-    if (__glXIsDirect(dpy, contextID)) {
-	return NULL;
-    }
+   if (contextID == None) {
+      return NULL;
+   }
+   if (__glXIsDirect(dpy, contextID)) {
+      return NULL;
+   }
 
-    ctx = CreateContext(dpy, NULL, NULL, NULL, False, contextID, False, 0);
-    if (NULL != ctx) {
-	if (Success != __glXQueryContextInfo(dpy, ctx)) {
-	   return NULL;
-	}
-    }
-    return ctx;
+   ctx = CreateContext(dpy, NULL, NULL, NULL, False, contextID, False, 0);
+   if (NULL != ctx) {
+      if (Success != __glXQueryContextInfo(dpy, ctx)) {
+         return NULL;
+      }
+   }
+   return ctx;
 }
 
-PUBLIC void glXFreeContextEXT(Display *dpy, GLXContext ctx)
+PUBLIC void
+glXFreeContextEXT(Display * dpy, GLXContext ctx)
 {
-    DestroyContext(dpy, ctx);
+   DestroyContext(dpy, ctx);
 }
 
 
@@ -1698,147 +1751,149 @@ PUBLIC void glXFreeContextEXT(Display *dpy, GLXContext ctx)
  * GLX 1.3 functions - these are just stubs for now!
  */
 
-PUBLIC GLXFBConfig *glXChooseFBConfig(Display *dpy, int screen,
-				      const int *attribList, int *nitems)
+PUBLIC GLXFBConfig *
+glXChooseFBConfig(Display * dpy, int screen,
+                  const int *attribList, int *nitems)
 {
-    __GLcontextModes ** config_list;
-    int   list_size;
+   __GLcontextModes **config_list;
+   int list_size;
 
 
-    config_list = (__GLcontextModes **) 
-	glXGetFBConfigs( dpy, screen, & list_size );
+   config_list = (__GLcontextModes **)
+      glXGetFBConfigs(dpy, screen, &list_size);
 
-    if ( (config_list != NULL) && (list_size > 0) && (attribList != NULL) ) {
-	list_size = choose_visual( config_list, list_size, attribList,
-				   GL_TRUE );
-	if ( list_size == 0 ) {
-	    XFree( config_list );
-	    config_list = NULL;
-	}
-    }
+   if ((config_list != NULL) && (list_size > 0) && (attribList != NULL)) {
+      list_size = choose_visual(config_list, list_size, attribList, GL_TRUE);
+      if (list_size == 0) {
+         XFree(config_list);
+         config_list = NULL;
+      }
+   }
 
-    *nitems = list_size;
-    return (GLXFBConfig *) config_list;
+   *nitems = list_size;
+   return (GLXFBConfig *) config_list;
 }
 
 
-PUBLIC GLXContext glXCreateNewContext(Display *dpy, GLXFBConfig config,
-				      int renderType, GLXContext shareList,
-				      Bool allowDirect)
+PUBLIC GLXContext
+glXCreateNewContext(Display * dpy, GLXFBConfig config,
+                    int renderType, GLXContext shareList, Bool allowDirect)
 {
-    return CreateContext( dpy, NULL, (__GLcontextModes *) config, shareList,
-			  allowDirect, None, True, renderType );
+   return CreateContext(dpy, NULL, (__GLcontextModes *) config, shareList,
+                        allowDirect, None, True, renderType);
 }
 
 
-PUBLIC GLXDrawable glXGetCurrentReadDrawable(void)
+PUBLIC GLXDrawable
+glXGetCurrentReadDrawable(void)
 {
-    GLXContext gc = __glXGetCurrentContext();
-    return gc->currentReadable;
+   GLXContext gc = __glXGetCurrentContext();
+   return gc->currentReadable;
 }
 
 
-PUBLIC GLXFBConfig *glXGetFBConfigs(Display *dpy, int screen, int *nelements)
+PUBLIC GLXFBConfig *
+glXGetFBConfigs(Display * dpy, int screen, int *nelements)
 {
-    __GLXdisplayPrivate *priv = __glXInitialize(dpy);
-    __GLcontextModes ** config = NULL;
-    int   i;
+   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   __GLcontextModes **config = NULL;
+   int i;
 
-    *nelements = 0;
-    if ( priv
-         && (priv->screenConfigs != NULL)
-	 && (screen >= 0) && (screen <= ScreenCount(dpy))
-	 && (priv->screenConfigs[screen].configs != NULL)
-	 && (priv->screenConfigs[screen].configs->fbconfigID != GLX_DONT_CARE) ) {
-	unsigned num_configs = 0;
-	__GLcontextModes * modes;
+   *nelements = 0;
+   if (priv && (priv->screenConfigs != NULL)
+       && (screen >= 0) && (screen <= ScreenCount(dpy))
+       && (priv->screenConfigs[screen].configs != NULL)
+       && (priv->screenConfigs[screen].configs->fbconfigID != GLX_DONT_CARE)) {
+      unsigned num_configs = 0;
+      __GLcontextModes *modes;
 
 
-	for ( modes = priv->screenConfigs[screen].configs
-	      ; modes != NULL
-	      ; modes = modes->next ) {
-	    if ( modes->fbconfigID != GLX_DONT_CARE ) {
-		num_configs++;
-	    }
-	}
+      for (modes = priv->screenConfigs[screen].configs; modes != NULL;
+           modes = modes->next) {
+         if (modes->fbconfigID != GLX_DONT_CARE) {
+            num_configs++;
+         }
+      }
 
-	config = (__GLcontextModes **) Xmalloc( sizeof(__GLcontextModes *)
-						* num_configs );
-	if ( config != NULL ) {
-	    *nelements = num_configs;
-	    i = 0;
-	    for ( modes = priv->screenConfigs[screen].configs
-		  ; modes != NULL
-		  ; modes = modes->next ) {
-		if ( modes->fbconfigID != GLX_DONT_CARE ) {
-		    config[i] = modes;
-		    i++;
-		}
-	    }
-	}
-    }
-    return (GLXFBConfig *) config;
+      config = (__GLcontextModes **) Xmalloc(sizeof(__GLcontextModes *)
+                                             * num_configs);
+      if (config != NULL) {
+         *nelements = num_configs;
+         i = 0;
+         for (modes = priv->screenConfigs[screen].configs; modes != NULL;
+              modes = modes->next) {
+            if (modes->fbconfigID != GLX_DONT_CARE) {
+               config[i] = modes;
+               i++;
+            }
+         }
+      }
+   }
+   return (GLXFBConfig *) config;
 }
 
 
-PUBLIC int glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config,
-				int attribute, int *value)
+PUBLIC int
+glXGetFBConfigAttrib(Display * dpy, GLXFBConfig config,
+                     int attribute, int *value)
 {
-    __GLcontextModes * const modes = ValidateGLXFBConfig( dpy, config );
+   __GLcontextModes *const modes = ValidateGLXFBConfig(dpy, config);
 
-    return (modes != NULL)
-	? _gl_get_context_mode_data( modes, attribute, value )
-	: GLXBadFBConfig;
+   return (modes != NULL)
+      ? _gl_get_context_mode_data(modes, attribute, value)
+      : GLXBadFBConfig;
 }
 
 
-PUBLIC XVisualInfo *glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config)
+PUBLIC XVisualInfo *
+glXGetVisualFromFBConfig(Display * dpy, GLXFBConfig config)
 {
-    XVisualInfo visualTemplate;
-    __GLcontextModes * fbconfig = (__GLcontextModes *) config;
-    int  count;
+   XVisualInfo visualTemplate;
+   __GLcontextModes *fbconfig = (__GLcontextModes *) config;
+   int count;
 
-    /*
+   /*
     ** Get a list of all visuals, return if list is empty
     */
-    visualTemplate.visualid = fbconfig->visualID;
-    return XGetVisualInfo(dpy,VisualIDMask,&visualTemplate,&count);
+   visualTemplate.visualid = fbconfig->visualID;
+   return XGetVisualInfo(dpy, VisualIDMask, &visualTemplate, &count);
 }
 
 
 /*
 ** GLX_SGI_swap_control
 */
-static int __glXSwapIntervalSGI(int interval)
+static int
+__glXSwapIntervalSGI(int interval)
 {
    xGLXVendorPrivateReq *req;
    GLXContext gc = __glXGetCurrentContext();
-   Display * dpy;
-   CARD32 * interval_ptr;
+   Display *dpy;
+   CARD32 *interval_ptr;
    CARD8 opcode;
 
-   if ( gc == NULL ) {
+   if (gc == NULL) {
       return GLX_BAD_CONTEXT;
    }
-   
-   if ( interval <= 0 ) {
+
+   if (interval <= 0) {
       return GLX_BAD_VALUE;
    }
 
 #ifdef __DRI_SWAP_CONTROL
    if (gc->driContext) {
-       __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy,
-							     gc->screen );
-       __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(gc->currentDpy,
-						   gc->currentDrawable,
-						   NULL);
-       if (psc->swapControl != NULL && pdraw != NULL) {
-	   psc->swapControl->setSwapInterval(pdraw->driDrawable, interval);
-	   return 0;
-       }
-       else {
-	   return GLX_BAD_CONTEXT;
-       }
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(gc->currentDpy,
+                                                          gc->screen);
+      __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(gc->currentDpy,
+                                                  gc->currentDrawable,
+                                                  NULL);
+      if (psc->swapControl != NULL && pdraw != NULL) {
+         psc->swapControl->setSwapInterval(pdraw->driDrawable, interval);
+         return 0;
+      }
+      else {
+         return GLX_BAD_CONTEXT;
+      }
    }
 #endif
    dpy = gc->currentDpy;
@@ -1849,7 +1904,7 @@ static int __glXSwapIntervalSGI(int interval)
 
    /* Send the glXSwapIntervalSGI request */
    LockDisplay(dpy);
-   GetReqExtra(GLXVendorPrivate,sizeof(CARD32),req);
+   GetReqExtra(GLXVendorPrivate, sizeof(CARD32), req);
    req->reqType = opcode;
    req->glxCode = X_GLXVendorPrivate;
    req->vendorCode = X_GLXvop_SwapIntervalSGI;
@@ -1869,26 +1924,27 @@ static int __glXSwapIntervalSGI(int interval)
 /*
 ** GLX_MESA_swap_control
 */
-static int __glXSwapIntervalMESA(unsigned int interval)
+static int
+__glXSwapIntervalMESA(unsigned int interval)
 {
 #ifdef __DRI_SWAP_CONTROL
    GLXContext gc = __glXGetCurrentContext();
 
-   if ( interval < 0 ) {
+   if (interval < 0) {
       return GLX_BAD_VALUE;
    }
 
    if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy,
-							    gc->screen );
-      
-      if ( (psc != NULL) && (psc->driScreen != NULL) ) {
-	 __GLXDRIdrawable *pdraw = 
-	     GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
-	 if (psc->swapControl != NULL && pdraw != NULL) {
-	    psc->swapControl->setSwapInterval(pdraw->driDrawable, interval);
-	    return 0;
-	 }
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(gc->currentDpy,
+                                                          gc->screen);
+
+      if ((psc != NULL) && (psc->driScreen != NULL)) {
+         __GLXDRIdrawable *pdraw =
+            GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
+         if (psc->swapControl != NULL && pdraw != NULL) {
+            psc->swapControl->setSwapInterval(pdraw->driDrawable, interval);
+            return 0;
+         }
       }
    }
 #else
@@ -1897,23 +1953,24 @@ static int __glXSwapIntervalMESA(unsigned int interval)
 
    return GLX_BAD_CONTEXT;
 }
- 
 
-static int __glXGetSwapIntervalMESA(void)
+
+static int
+__glXGetSwapIntervalMESA(void)
 {
 #ifdef __DRI_SWAP_CONTROL
    GLXContext gc = __glXGetCurrentContext();
 
    if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy,
-							    gc->screen );
-      
-      if ( (psc != NULL) && (psc->driScreen != NULL) ) {
-	 __GLXDRIdrawable *pdraw = 
-	     GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
-	 if (psc->swapControl != NULL && pdraw != NULL) {
-	    return psc->swapControl->getSwapInterval(pdraw->driDrawable);
-	 }
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(gc->currentDpy,
+                                                          gc->screen);
+
+      if ((psc != NULL) && (psc->driScreen != NULL)) {
+         __GLXDRIdrawable *pdraw =
+            GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
+         if (psc->swapControl != NULL && pdraw != NULL) {
+            return psc->swapControl->getSwapInterval(pdraw->driDrawable);
+         }
       }
    }
 #endif
@@ -1926,16 +1983,17 @@ static int __glXGetSwapIntervalMESA(void)
 ** GLX_MESA_swap_frame_usage
 */
 
-static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable)
+static GLint
+__glXBeginFrameTrackingMESA(Display * dpy, GLXDrawable drawable)
 {
-   int   status = GLX_BAD_CONTEXT;
+   int status = GLX_BAD_CONTEXT;
 #ifdef __DRI_FRAME_TRACKING
-   int screen;
+   int screen = 0;
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
 
    if (pdraw != NULL && psc->frameTracking != NULL)
-       status = psc->frameTracking->frameTracking(pdraw->driDrawable, GL_TRUE);
+      status = psc->frameTracking->frameTracking(pdraw->driDrawable, GL_TRUE);
 #else
    (void) dpy;
    (void) drawable;
@@ -1943,18 +2001,19 @@ static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable)
    return status;
 }
 
-    
-static GLint __glXEndFrameTrackingMESA(Display *dpy, GLXDrawable drawable)
+
+static GLint
+__glXEndFrameTrackingMESA(Display * dpy, GLXDrawable drawable)
 {
-   int   status = GLX_BAD_CONTEXT;
+   int status = GLX_BAD_CONTEXT;
 #ifdef __DRI_FRAME_TRACKING
-   int screen;
-   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen);
+   int screen = 0;
+   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
    __GLXscreenConfigs *psc = GetGLXScreenConfigs(dpy, screen);
 
    if (pdraw != NULL && psc->frameTracking != NULL)
-       status = psc->frameTracking->frameTracking(pdraw->driDrawable,
-						  GL_FALSE);
+      status = psc->frameTracking->frameTracking(pdraw->driDrawable,
+                                                 GL_FALSE);
 #else
    (void) dpy;
    (void) drawable;
@@ -1963,24 +2022,24 @@ static GLint __glXEndFrameTrackingMESA(Display *dpy, GLXDrawable drawable)
 }
 
 
-static GLint __glXGetFrameUsageMESA(Display *dpy, GLXDrawable drawable,
-				    GLfloat *usage)
+static GLint
+__glXGetFrameUsageMESA(Display * dpy, GLXDrawable drawable, GLfloat * usage)
 {
-   int   status = GLX_BAD_CONTEXT;
+   int status = GLX_BAD_CONTEXT;
 #ifdef __DRI_FRAME_TRACKING
-   int screen;
-   __GLXDRIdrawable * const pdraw = GetGLXDRIDrawable(dpy, drawable, & screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
+   int screen = 0;
+   __GLXDRIdrawable *const pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
 
    if (pdraw != NULL && psc->frameTracking != NULL) {
-       int64_t sbc, missedFrames;
-       float   lastMissedUsage;
+      int64_t sbc, missedFrames;
+      float lastMissedUsage;
 
-       status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable,
-						       &sbc,
-						       &missedFrames,
-						       &lastMissedUsage,
-						       usage);
+      status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable,
+                                                      &sbc,
+                                                      &missedFrames,
+                                                      &lastMissedUsage,
+                                                      usage);
    }
 #else
    (void) dpy;
@@ -1991,22 +2050,24 @@ static GLint __glXGetFrameUsageMESA(Display *dpy, GLXDrawable drawable,
 }
 
 
-static GLint __glXQueryFrameTrackingMESA(Display *dpy, GLXDrawable drawable,
-					 int64_t *sbc, int64_t *missedFrames,
-					 GLfloat *lastMissedUsage)
+static GLint
+__glXQueryFrameTrackingMESA(Display * dpy, GLXDrawable drawable,
+                            int64_t * sbc, int64_t * missedFrames,
+                            GLfloat * lastMissedUsage)
 {
-   int   status = GLX_BAD_CONTEXT;
+   int status = GLX_BAD_CONTEXT;
 #ifdef __DRI_FRAME_TRACKING
-   int screen;
-   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
+   int screen = 0;
+   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
 
    if (pdraw != NULL && psc->frameTracking != NULL) {
-      float   usage;
+      float usage;
 
       status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable,
-						      sbc, missedFrames,
-						      lastMissedUsage, &usage);
+                                                      sbc, missedFrames,
+                                                      lastMissedUsage,
+                                                      &usage);
    }
 #else
    (void) dpy;
@@ -2022,7 +2083,8 @@ static GLint __glXQueryFrameTrackingMESA(Display *dpy, GLXDrawable drawable,
 /*
 ** GLX_SGI_video_sync
 */
-static int __glXGetVideoSyncSGI(unsigned int *count)
+static int
+__glXGetVideoSyncSGI(unsigned int *count)
 {
    /* FIXME: Looking at the GLX_SGI_video_sync spec in the extension registry,
     * FIXME: there should be a GLX encoding for this call.  I can find no
@@ -2033,49 +2095,50 @@ static int __glXGetVideoSyncSGI(unsigned int *count)
 
 
    if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy,
-							    gc->screen );
-      if ( psc->msc && psc->driScreen ) {
-          __GLXDRIdrawable *pdraw = 
-              GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
-	  int64_t temp; 
-	  int ret;
- 
-	  ret = (*psc->msc->getDrawableMSC)(psc->__driScreen,
-					    pdraw->driDrawable, &temp);
-	  *count = (unsigned) temp;
-
-	  return (ret == 0) ? 0 : GLX_BAD_CONTEXT;
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(gc->currentDpy,
+                                                          gc->screen);
+      if (psc->msc && psc->driScreen) {
+         __GLXDRIdrawable *pdraw =
+            GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
+         int64_t temp;
+         int ret;
+
+         ret = (*psc->msc->getDrawableMSC) (psc->__driScreen,
+                                            pdraw->driDrawable, &temp);
+         *count = (unsigned) temp;
+
+         return (ret == 0) ? 0 : GLX_BAD_CONTEXT;
       }
    }
 #else
-    (void) count;
+   (void) count;
 #endif
    return GLX_BAD_CONTEXT;
 }
 
-static int __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
+static int
+__glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
 {
 #ifdef __DRI_MEDIA_STREAM_COUNTER
    GLXContext gc = __glXGetCurrentContext();
 
-   if ( divisor <= 0 || remainder < 0 )
-     return GLX_BAD_VALUE;
+   if (divisor <= 0 || remainder < 0)
+      return GLX_BAD_VALUE;
 
    if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy,
-							    gc->screen );
-      if (psc->msc != NULL && psc->driScreen ) {
-	 __GLXDRIdrawable *pdraw = 
-	     GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
-	 int       ret;
-	 int64_t   msc;
-	 int64_t   sbc;
-
-	 ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, 0,
-				       divisor, remainder, &msc, &sbc);
-	 *count = (unsigned) msc;
-	 return (ret == 0) ? 0 : GLX_BAD_CONTEXT;
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(gc->currentDpy,
+                                                          gc->screen);
+      if (psc->msc != NULL && psc->driScreen) {
+         __GLXDRIdrawable *pdraw =
+            GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL);
+         int ret;
+         int64_t msc;
+         int64_t sbc;
+
+         ret = (*psc->msc->waitForMSC) (pdraw->driDrawable, 0,
+                                        divisor, remainder, &msc, &sbc);
+         *count = (unsigned) msc;
+         return (ret == 0) ? 0 : GLX_BAD_CONTEXT;
       }
    }
 #else
@@ -2091,109 +2154,112 @@ static int __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count
 ** GLX_functions table.
 */
 
-PUBLIC GLX_ALIAS(int, glXGetFBConfigAttribSGIX,
-	  (Display *dpy, GLXFBConfigSGIX config, int attribute, int *value),
-	  (dpy, config, attribute, value),
-	  glXGetFBConfigAttrib)
+PUBLIC
+GLX_ALIAS(int, glXGetFBConfigAttribSGIX,
+          (Display * dpy, GLXFBConfigSGIX config, int attribute, int *value),
+          (dpy, config, attribute, value), glXGetFBConfigAttrib)
 
-PUBLIC GLX_ALIAS(GLXFBConfigSGIX *, glXChooseFBConfigSGIX,
-	  (Display *dpy, int screen, int *attrib_list, int *nelements),
-	  (dpy, screen, attrib_list, nelements),
-	  glXChooseFBConfig)
+     PUBLIC GLX_ALIAS(GLXFBConfigSGIX *, glXChooseFBConfigSGIX,
+                      (Display * dpy, int screen, int *attrib_list,
+                       int *nelements), (dpy, screen, attrib_list, nelements),
+                      glXChooseFBConfig)
 
-PUBLIC GLX_ALIAS(XVisualInfo *, glXGetVisualFromFBConfigSGIX,
-	  (Display * dpy, GLXFBConfigSGIX config),
-	  (dpy, config),
-	  glXGetVisualFromFBConfig)
+     PUBLIC GLX_ALIAS(XVisualInfo *, glXGetVisualFromFBConfigSGIX,
+                      (Display * dpy, GLXFBConfigSGIX config),
+                      (dpy, config), glXGetVisualFromFBConfig)
 
-PUBLIC GLXPixmap glXCreateGLXPixmapWithConfigSGIX(Display *dpy,
-                            GLXFBConfigSGIX config, Pixmap pixmap)
+     PUBLIC GLXPixmap glXCreateGLXPixmapWithConfigSGIX(Display * dpy,
+                                                       GLXFBConfigSGIX config,
+                                                       Pixmap pixmap)
 {
-    xGLXVendorPrivateWithReplyReq *vpreq;
-    xGLXCreateGLXPixmapWithConfigSGIXReq *req;
-    GLXPixmap xid = None;
-    CARD8 opcode;
-    const __GLcontextModes * const fbconfig = (__GLcontextModes *) config;
-    __GLXscreenConfigs * psc;
+   xGLXVendorPrivateWithReplyReq *vpreq;
+   xGLXCreateGLXPixmapWithConfigSGIXReq *req;
+   GLXPixmap xid = None;
+   CARD8 opcode;
+   const __GLcontextModes *const fbconfig = (__GLcontextModes *) config;
+   __GLXscreenConfigs *psc;
 
 
-    if ( (dpy == NULL) || (config == NULL) ) {
-	return None;
-    }
+   if ((dpy == NULL) || (config == NULL)) {
+      return None;
+   }
 
-    psc = GetGLXScreenConfigs( dpy, fbconfig->screen );
-    if ( (psc != NULL) 
-	 && __glXExtensionBitIsEnabled( psc, SGIX_fbconfig_bit ) ) {
-	opcode = __glXSetupForCommand(dpy);
-	if (!opcode) {
-	    return None;
-	}
+   psc = GetGLXScreenConfigs(dpy, fbconfig->screen);
+   if ((psc != NULL)
+       && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)) {
+      opcode = __glXSetupForCommand(dpy);
+      if (!opcode) {
+         return None;
+      }
 
-	/* Send the glXCreateGLXPixmapWithConfigSGIX request */
-	LockDisplay(dpy);
-	GetReqExtra(GLXVendorPrivateWithReply,
-		    sz_xGLXCreateGLXPixmapWithConfigSGIXReq-sz_xGLXVendorPrivateWithReplyReq,vpreq);
-	req = (xGLXCreateGLXPixmapWithConfigSGIXReq *)vpreq;
-	req->reqType = opcode;
-	req->glxCode = X_GLXVendorPrivateWithReply;
-	req->vendorCode = X_GLXvop_CreateGLXPixmapWithConfigSGIX;
-	req->screen = fbconfig->screen;
-	req->fbconfig = fbconfig->fbconfigID;
-	req->pixmap = pixmap;
-	req->glxpixmap = xid = XAllocID(dpy);
-	UnlockDisplay(dpy);
-	SyncHandle();
-    }
+      /* Send the glXCreateGLXPixmapWithConfigSGIX request */
+      LockDisplay(dpy);
+      GetReqExtra(GLXVendorPrivateWithReply,
+                  sz_xGLXCreateGLXPixmapWithConfigSGIXReq -
+                  sz_xGLXVendorPrivateWithReplyReq, vpreq);
+      req = (xGLXCreateGLXPixmapWithConfigSGIXReq *) vpreq;
+      req->reqType = opcode;
+      req->glxCode = X_GLXVendorPrivateWithReply;
+      req->vendorCode = X_GLXvop_CreateGLXPixmapWithConfigSGIX;
+      req->screen = fbconfig->screen;
+      req->fbconfig = fbconfig->fbconfigID;
+      req->pixmap = pixmap;
+      req->glxpixmap = xid = XAllocID(dpy);
+      UnlockDisplay(dpy);
+      SyncHandle();
+   }
 
-    return xid;
+   return xid;
 }
 
-PUBLIC GLXContext glXCreateContextWithConfigSGIX(Display *dpy,
-                             GLXFBConfigSGIX config, int renderType,
-                             GLXContext shareList, Bool allowDirect)
+PUBLIC GLXContext
+glXCreateContextWithConfigSGIX(Display * dpy,
+                               GLXFBConfigSGIX config, int renderType,
+                               GLXContext shareList, Bool allowDirect)
 {
-    GLXContext gc = NULL;
-    const __GLcontextModes * const fbconfig = (__GLcontextModes *) config;
-    __GLXscreenConfigs * psc;
+   GLXContext gc = NULL;
+   const __GLcontextModes *const fbconfig = (__GLcontextModes *) config;
+   __GLXscreenConfigs *psc;
 
 
-    if ( (dpy == NULL) || (config == NULL) ) {
-	return None;
-    }
+   if ((dpy == NULL) || (config == NULL)) {
+      return None;
+   }
 
-    psc = GetGLXScreenConfigs( dpy, fbconfig->screen );
-    if ( (psc != NULL) 
-	 && __glXExtensionBitIsEnabled( psc, SGIX_fbconfig_bit ) ) {
-	gc = CreateContext( dpy, NULL, (__GLcontextModes *) config, shareList,
-			    allowDirect, None, False, renderType );
-    }
+   psc = GetGLXScreenConfigs(dpy, fbconfig->screen);
+   if ((psc != NULL)
+       && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)) {
+      gc = CreateContext(dpy, NULL, (__GLcontextModes *) config, shareList,
+                         allowDirect, None, False, renderType);
+   }
 
-    return gc;
+   return gc;
 }
 
 
-PUBLIC GLXFBConfigSGIX glXGetFBConfigFromVisualSGIX(Display *dpy,
-						    XVisualInfo *vis)
+PUBLIC GLXFBConfigSGIX
+glXGetFBConfigFromVisualSGIX(Display * dpy, XVisualInfo * vis)
 {
-    __GLXdisplayPrivate *priv;
-    __GLXscreenConfigs *psc;
+   __GLXdisplayPrivate *priv;
+   __GLXscreenConfigs *psc = NULL;
 
-    if ( (GetGLXPrivScreenConfig( dpy, vis->screen, & priv, & psc ) != Success)
-	 && __glXExtensionBitIsEnabled( psc, SGIX_fbconfig_bit )
-	 && (psc->configs->fbconfigID != GLX_DONT_CARE) ) {
-	return (GLXFBConfigSGIX) _gl_context_modes_find_visual( psc->configs,
-								vis->visualid );
-    }
+   if ((GetGLXPrivScreenConfig(dpy, vis->screen, &priv, &psc) != Success)
+       && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)
+       && (psc->configs->fbconfigID != GLX_DONT_CARE)) {
+      return (GLXFBConfigSGIX) _gl_context_modes_find_visual(psc->configs,
+                                                             vis->visualid);
+   }
 
-    return NULL;
+   return NULL;
 }
 
 
 /*
 ** GLX_SGIX_swap_group
 */
-static void __glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable,
-				   GLXDrawable member)
+static void
+__glXJoinSwapGroupSGIX(Display * dpy, GLXDrawable drawable,
+                       GLXDrawable member)
 {
    (void) dpy;
    (void) drawable;
@@ -2204,15 +2270,16 @@ static void __glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable,
 /*
 ** GLX_SGIX_swap_barrier
 */
-static void __glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable,
-				     int barrier)
+static void
+__glXBindSwapBarrierSGIX(Display * dpy, GLXDrawable drawable, int barrier)
 {
    (void) dpy;
    (void) drawable;
    (void) barrier;
 }
 
-static Bool __glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
+static Bool
+__glXQueryMaxSwapBarriersSGIX(Display * dpy, int screen, int *max)
 {
    (void) dpy;
    (void) screen;
@@ -2224,23 +2291,24 @@ static Bool __glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
 /*
 ** GLX_OML_sync_control
 */
-static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable,
-				  int64_t *ust, int64_t *msc, int64_t *sbc)
+static Bool
+__glXGetSyncValuesOML(Display * dpy, GLXDrawable drawable,
+                      int64_t * ust, int64_t * msc, int64_t * sbc)
 {
 #if defined(__DRI_SWAP_BUFFER_COUNTER) && defined(__DRI_MEDIA_STREAM_COUNTER)
-    __GLXdisplayPrivate * const priv = __glXInitialize(dpy);
-
-    if ( priv != NULL ) {
-	int   i;
-	__GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &i);
-	__GLXscreenConfigs * const psc = &priv->screenConfigs[i];
-
-	assert( (pdraw == NULL) || (i != -1) );
-	return ( (pdraw && psc->sbc && psc->msc)
-		 && ((*psc->msc->getMSC)(psc->driScreen, msc) == 0)
-		 && ((*psc->sbc->getSBC)(pdraw->driDrawable, sbc) == 0)
-		 && (__glXGetUST(ust) == 0) );
-    }
+   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+
+   if (priv != NULL) {
+      int i;
+      __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &i);
+      __GLXscreenConfigs *const psc = &priv->screenConfigs[i];
+
+      assert((pdraw == NULL) || (i != -1));
+      return ((pdraw && psc->sbc && psc->msc)
+              && ((*psc->msc->getMSC) (psc->driScreen, msc) == 0)
+              && ((*psc->sbc->getSBC) (pdraw->driDrawable, sbc) == 0)
+              && (__glXGetUST(ust) == 0));
+   }
 #else
    (void) dpy;
    (void) drawable;
@@ -2253,94 +2321,94 @@ static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable,
 
 #ifdef GLX_DIRECT_RENDERING
 _X_HIDDEN GLboolean
-__driGetMscRateOML(__DRIdrawable *draw,
-		   int32_t *numerator, int32_t *denominator, void *private)
+__driGetMscRateOML(__DRIdrawable * draw,
+                   int32_t * numerator, int32_t * denominator, void *private)
 {
 #ifdef XF86VIDMODE
-    __GLXscreenConfigs *psc;
-    XF86VidModeModeLine   mode_line;
-    int   dot_clock;
-    int   i;
-    __GLXDRIdrawable *glxDraw = private;
-
-    psc = glxDraw->psc;
-    if (XF86VidModeQueryVersion(psc->dpy, &i, &i) &&
-	XF86VidModeGetModeLine(psc->dpy, psc->scr, &dot_clock, &mode_line) ) {
-	unsigned   n = dot_clock * 1000;
-	unsigned   d = mode_line.vtotal * mode_line.htotal;
-	
+   __GLXscreenConfigs *psc;
+   XF86VidModeModeLine mode_line;
+   int dot_clock;
+   int i;
+   __GLXDRIdrawable *glxDraw = private;
+
+   psc = glxDraw->psc;
+   if (XF86VidModeQueryVersion(psc->dpy, &i, &i) &&
+       XF86VidModeGetModeLine(psc->dpy, psc->scr, &dot_clock, &mode_line)) {
+      unsigned n = dot_clock * 1000;
+      unsigned d = mode_line.vtotal * mode_line.htotal;
+
 # define V_INTERLACE 0x010
 # define V_DBLSCAN   0x020
 
-	if (mode_line.flags & V_INTERLACE)
-	    n *= 2;
-	else if (mode_line.flags & V_DBLSCAN)
-	    d *= 2;
-
-	/* The OML_sync_control spec requires that if the refresh rate is a
-	 * whole number, that the returned numerator be equal to the refresh
-	 * rate and the denominator be 1.
-	 */
-
-	if (n % d == 0) {
-	    n /= d;
-	    d = 1;
-	}
-	else {
-	    static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 };
-
-	    /* This is a poor man's way to reduce a fraction.  It's far from
-	     * perfect, but it will work well enough for this situation.
-	     */
-
-	    for (i = 0; f[i] != 0; i++) {
-		while (n % f[i] == 0 && d % f[i] == 0) {
-		    d /= f[i];
-		    n /= f[i];
-		}
-	    }
-	}
-
-	*numerator = n;
-	*denominator = d;
-
-	return True;
-    }
-    else
-	return False;
+      if (mode_line.flags & V_INTERLACE)
+         n *= 2;
+      else if (mode_line.flags & V_DBLSCAN)
+         d *= 2;
+
+      /* The OML_sync_control spec requires that if the refresh rate is a
+       * whole number, that the returned numerator be equal to the refresh
+       * rate and the denominator be 1.
+       */
+
+      if (n % d == 0) {
+         n /= d;
+         d = 1;
+      }
+      else {
+         static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 };
+
+         /* This is a poor man's way to reduce a fraction.  It's far from
+          * perfect, but it will work well enough for this situation.
+          */
+
+         for (i = 0; f[i] != 0; i++) {
+            while (n % f[i] == 0 && d % f[i] == 0) {
+               d /= f[i];
+               n /= f[i];
+            }
+         }
+      }
+
+      *numerator = n;
+      *denominator = d;
+
+      return True;
+   }
+   else
+      return False;
 #else
-    return False;
+   return False;
 #endif
 }
 #endif
 
 /**
  * Determine the refresh rate of the specified drawable and display.
- * 
+ *
  * \param dpy          Display whose refresh rate is to be determined.
  * \param drawable     Drawable whose refresh rate is to be determined.
  * \param numerator    Numerator of the refresh rate.
  * \param demoninator  Denominator of the refresh rate.
  * \return  If the refresh rate for the specified display and drawable could
  *          be calculated, True is returned.  Otherwise False is returned.
- * 
+ *
  * \note This function is implemented entirely client-side.  A lot of other
  *       functionality is required to export GLX_OML_sync_control, so on
  *       XFree86 this function can be called for direct-rendering contexts
  *       when GLX_OML_sync_control appears in the client extension string.
  */
 
-_X_HIDDEN GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable,
-				       int32_t * numerator,
-				       int32_t * denominator)
+_X_HIDDEN GLboolean
+__glXGetMscRateOML(Display * dpy, GLXDrawable drawable,
+                   int32_t * numerator, int32_t * denominator)
 {
 #if defined( GLX_DIRECT_RENDERING ) && defined( XF86VIDMODE )
-    __GLXDRIdrawable *draw = GetGLXDRIDrawable(dpy, drawable, NULL);
+   __GLXDRIdrawable *draw = GetGLXDRIDrawable(dpy, drawable, NULL);
 
-    if (draw == NULL)
-	return False;
+   if (draw == NULL)
+      return False;
 
-    return __driGetMscRateOML(draw->driDrawable, numerator, denominator, draw);
+   return __driGetMscRateOML(draw->driDrawable, numerator, denominator, draw);
 #else
    (void) dpy;
    (void) drawable;
@@ -2351,28 +2419,28 @@ _X_HIDDEN GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable,
 }
 
 
-static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable,
-				      int64_t target_msc, int64_t divisor,
-				      int64_t remainder)
+static int64_t
+__glXSwapBuffersMscOML(Display * dpy, GLXDrawable drawable,
+                       int64_t target_msc, int64_t divisor, int64_t remainder)
 {
 #ifdef __DRI_SWAP_BUFFER_COUNTER
    int screen;
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen );
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
 
    /* The OML_sync_control spec says these should "generate a GLX_BAD_VALUE
     * error", but it also says "It [glXSwapBuffersMscOML] will return a value
     * of -1 if the function failed because of errors detected in the input
     * parameters"
     */
-   if ( divisor < 0 || remainder < 0 || target_msc < 0 )
+   if (divisor < 0 || remainder < 0 || target_msc < 0)
       return -1;
-   if ( divisor > 0 && remainder >= divisor )
+   if (divisor > 0 && remainder >= divisor)
       return -1;
 
    if (pdraw != NULL && psc->counters != NULL)
-      return (*psc->sbc->swapBuffersMSC)(pdraw->driDrawable, target_msc,
-					 divisor, remainder);
+      return (*psc->sbc->swapBuffersMSC) (pdraw->driDrawable, target_msc,
+                                          divisor, remainder);
 
 #else
    (void) dpy;
@@ -2385,33 +2453,34 @@ static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable,
 }
 
 
-static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable,
-			       int64_t target_msc, int64_t divisor,
-			       int64_t remainder, int64_t *ust,
-			       int64_t *msc, int64_t *sbc)
+static Bool
+__glXWaitForMscOML(Display * dpy, GLXDrawable drawable,
+                   int64_t target_msc, int64_t divisor,
+                   int64_t remainder, int64_t * ust,
+                   int64_t * msc, int64_t * sbc)
 {
 #ifdef __DRI_MEDIA_STREAM_COUNTER
-   int screen;
+   int screen = 0;
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen );
-   int  ret;
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
+   int ret;
 
    /* The OML_sync_control spec says these should "generate a GLX_BAD_VALUE
     * error", but the return type in the spec is Bool.
     */
-   if ( divisor < 0 || remainder < 0 || target_msc < 0 )
+   if (divisor < 0 || remainder < 0 || target_msc < 0)
       return False;
-   if ( divisor > 0 && remainder >= divisor )
+   if (divisor > 0 && remainder >= divisor)
       return False;
 
    if (pdraw != NULL && psc->msc != NULL) {
-      ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, target_msc,
-				    divisor, remainder, msc, sbc);
+      ret = (*psc->msc->waitForMSC) (pdraw->driDrawable, target_msc,
+                                     divisor, remainder, msc, sbc);
 
       /* __glXGetUST returns zero on success and non-zero on failure.
        * This function returns True on success and False on failure.
        */
-      return ( (ret == 0) && (__glXGetUST( ust ) == 0) );
+      return ((ret == 0) && (__glXGetUST(ust) == 0));
    }
 #else
    (void) dpy;
@@ -2427,29 +2496,31 @@ static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable,
 }
 
 
-static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable,
-			       int64_t target_sbc, int64_t *ust,
-			       int64_t *msc, int64_t *sbc )
+static Bool
+__glXWaitForSbcOML(Display * dpy, GLXDrawable drawable,
+                   int64_t target_sbc, int64_t * ust,
+                   int64_t * msc, int64_t * sbc)
 {
 #ifdef __DRI_SWAP_BUFFER_COUNTER
    int screen;
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen );
-   int  ret;
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
+   int ret;
 
    /* The OML_sync_control spec says this should "generate a GLX_BAD_VALUE
     * error", but the return type in the spec is Bool.
     */
-   if ( target_sbc < 0 )
+   if (target_sbc < 0)
       return False;
 
    if (pdraw != NULL && psc->sbc != NULL) {
-      ret = (*psc->sbc->waitForSBC)(pdraw->driDrawable, target_sbc, msc, sbc);
+      ret =
+         (*psc->sbc->waitForSBC) (pdraw->driDrawable, target_sbc, msc, sbc);
 
       /* __glXGetUST returns zero on success and non-zero on failure.
        * This function returns True on success and False on failure.
        */
-      return( (ret == 0) && (__glXGetUST( ust ) == 0) );
+      return ((ret == 0) && (__glXGetUST(ust) == 0));
    }
 #else
    (void) dpy;
@@ -2468,16 +2539,17 @@ static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable,
  */
 /*@{*/
 
-PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn,
-				   size_t size, float readFreq,
-				   float writeFreq, float priority)
+PUBLIC void *
+glXAllocateMemoryMESA(Display * dpy, int scrn,
+                      size_t size, float readFreq,
+                      float writeFreq, float priority)
 {
 #ifdef __DRI_ALLOCATE
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn );
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, scrn);
 
    if (psc && psc->allocate)
-       return (*psc->allocate->allocateMemory)(psc->__driScreen, size,
-					       readFreq, writeFreq, priority);
+      return (*psc->allocate->allocateMemory) (psc->__driScreen, size,
+                                               readFreq, writeFreq, priority);
 
 #else
    (void) dpy;
@@ -2492,13 +2564,14 @@ PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn,
 }
 
 
-PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer)
+PUBLIC void
+glXFreeMemoryMESA(Display * dpy, int scrn, void *pointer)
 {
 #ifdef __DRI_ALLOCATE
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn );
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, scrn);
 
    if (psc && psc->allocate)
-	 (*psc->allocate->freeMemory)(psc->__driScreen, pointer);
+      (*psc->allocate->freeMemory) (psc->__driScreen, pointer);
 
 #else
    (void) dpy;
@@ -2508,14 +2581,14 @@ PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer)
 }
 
 
-PUBLIC GLuint glXGetMemoryOffsetMESA( Display *dpy, int scrn,
-				      const void *pointer )
+PUBLIC GLuint
+glXGetMemoryOffsetMESA(Display * dpy, int scrn, const void *pointer)
 {
 #ifdef __DRI_ALLOCATE
-   __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn );
+   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, scrn);
 
    if (psc && psc->allocate)
-       return (*psc->allocate->memoryOffset)(psc->__driScreen, pointer);
+      return (*psc->allocate->memoryOffset) (psc->__driScreen, pointer);
 
 #else
    (void) dpy;
@@ -2525,6 +2598,7 @@ PUBLIC GLuint glXGetMemoryOffsetMESA( Display *dpy, int scrn,
 
    return ~0L;
 }
+
 /*@}*/
 
 
@@ -2555,7 +2629,8 @@ PUBLIC GLuint glXGetMemoryOffsetMESA( Display *dpy, int scrn,
  *     glXDestroyPbuffer glXDestroyPixmap glXDestroyWindow
  *     glXDestroyGLXPbufferSGIX glXDestroyGLXVideoSourceSGIX
  */
-static Bool __glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
+static Bool
+__glXReleaseBuffersMESA(Display * dpy, GLXDrawable d)
 {
    (void) dpy;
    (void) d;
@@ -2563,8 +2638,9 @@ static Bool __glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
 }
 
 
-PUBLIC GLXPixmap glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visual,
-					 Pixmap pixmap, Colormap cmap )
+PUBLIC GLXPixmap
+glXCreateGLXPixmapMESA(Display * dpy, XVisualInfo * visual,
+                       Pixmap pixmap, Colormap cmap)
 {
    (void) dpy;
    (void) visual;
@@ -2572,6 +2648,7 @@ PUBLIC GLXPixmap glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visual,
    (void) cmap;
    return 0;
 }
+
 /*@}*/
 
 
@@ -2579,68 +2656,70 @@ PUBLIC GLXPixmap glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visual,
  * GLX_MESA_copy_sub_buffer
  */
 #define X_GLXvop_CopySubBufferMESA 5154 /* temporary */
-static void __glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable,
-				   int x, int y, int width, int height)
+static void
+__glXCopySubBufferMESA(Display * dpy, GLXDrawable drawable,
+                       int x, int y, int width, int height)
 {
-    xGLXVendorPrivateReq *req;
-    GLXContext gc;
-    GLXContextTag tag;
-    CARD32 *drawable_ptr;
-    INT32 *x_ptr, *y_ptr, *w_ptr, *h_ptr;
-    CARD8 opcode;
+   xGLXVendorPrivateReq *req;
+   GLXContext gc;
+   GLXContextTag tag;
+   CARD32 *drawable_ptr;
+   INT32 *x_ptr, *y_ptr, *w_ptr, *h_ptr;
+   CARD8 opcode;
 
 #ifdef __DRI_COPY_SUB_BUFFER
-    int screen;
-    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
-    if ( pdraw != NULL ) {
-	__GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen);
-	if (psc->driScreen->copySubBuffer != NULL) {
-	    glFlush();	    
-	    (*psc->driScreen->copySubBuffer)(pdraw, x, y, width, height);
-	}
-
-	return;
-    }
+   int screen;
+   __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen);
+   if (pdraw != NULL) {
+      __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
+      if (psc->driScreen->copySubBuffer != NULL) {
+         glFlush();
+         (*psc->driScreen->copySubBuffer) (pdraw, x, y, width, height);
+      }
+
+      return;
+   }
 #endif
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode)
-	return;
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode)
+      return;
 
-    /*
+   /*
     ** The calling thread may or may not have a current context.  If it
     ** does, send the context tag so the server can do a flush.
     */
-    gc = __glXGetCurrentContext();
-    if ((gc != NULL) && (dpy == gc->currentDpy) &&
-	((drawable == gc->currentDrawable) ||
-	 (drawable == gc->currentReadable)) ) {
-	tag = gc->currentContextTag;
-    } else {
-	tag = 0;
-    }
-
-    LockDisplay(dpy);
-    GetReqExtra(GLXVendorPrivate, sizeof(CARD32) + sizeof(INT32) * 4,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXVendorPrivate;
-    req->vendorCode = X_GLXvop_CopySubBufferMESA;
-    req->contextTag = tag;
-
-    drawable_ptr = (CARD32 *) (req + 1);
-    x_ptr = (INT32 *) (drawable_ptr + 1);
-    y_ptr = (INT32 *) (drawable_ptr + 2);
-    w_ptr = (INT32 *) (drawable_ptr + 3);
-    h_ptr = (INT32 *) (drawable_ptr + 4);
-
-    *drawable_ptr = drawable;
-    *x_ptr = x;
-    *y_ptr = y;
-    *w_ptr = width;
-    *h_ptr = height;
-
-    UnlockDisplay(dpy);
-    SyncHandle();
+   gc = __glXGetCurrentContext();
+   if ((gc != NULL) && (dpy == gc->currentDpy) &&
+       ((drawable == gc->currentDrawable) ||
+        (drawable == gc->currentReadable))) {
+      tag = gc->currentContextTag;
+   }
+   else {
+      tag = 0;
+   }
+
+   LockDisplay(dpy);
+   GetReqExtra(GLXVendorPrivate, sizeof(CARD32) + sizeof(INT32) * 4, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXVendorPrivate;
+   req->vendorCode = X_GLXvop_CopySubBufferMESA;
+   req->contextTag = tag;
+
+   drawable_ptr = (CARD32 *) (req + 1);
+   x_ptr = (INT32 *) (drawable_ptr + 1);
+   y_ptr = (INT32 *) (drawable_ptr + 2);
+   w_ptr = (INT32 *) (drawable_ptr + 3);
+   h_ptr = (INT32 *) (drawable_ptr + 4);
+
+   *drawable_ptr = drawable;
+   *x_ptr = x;
+   *y_ptr = y;
+   *w_ptr = width;
+   *h_ptr = height;
+
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
 
@@ -2648,128 +2727,127 @@ static void __glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable,
  * GLX_EXT_texture_from_pixmap
  */
 /*@{*/
-static void __glXBindTexImageEXT(Display *dpy,
-				 GLXDrawable drawable,
-				 int buffer,
-				 const int *attrib_list)
-{
-    xGLXVendorPrivateReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    CARD32 *drawable_ptr;
-    INT32 *buffer_ptr;
-    CARD32 *num_attrib_ptr;
-    CARD32 *attrib_ptr;
-    CARD8 opcode;
-    unsigned int i;
-
-    if (gc == NULL)
-	return;
-
-    i = 0;
-    if (attrib_list) {
- 	while (attrib_list[i * 2] != None)
- 	    i++;
-    }
- 
+static void
+__glXBindTexImageEXT(Display * dpy,
+                     GLXDrawable drawable, int buffer, const int *attrib_list)
+{
+   xGLXVendorPrivateReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   CARD32 *drawable_ptr;
+   INT32 *buffer_ptr;
+   CARD32 *num_attrib_ptr;
+   CARD32 *attrib_ptr;
+   CARD8 opcode;
+   unsigned int i;
+
+   if (gc == NULL)
+      return;
+
+   i = 0;
+   if (attrib_list) {
+      while (attrib_list[i * 2] != None)
+         i++;
+   }
+
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext) {
-	__GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL);
-
-	if (pdraw != NULL) {
-	    if (pdraw->psc->texBuffer->base.version >= 2 &&
-		pdraw->psc->texBuffer->setTexBuffer2 != NULL) {
-		(*pdraw->psc->texBuffer->setTexBuffer2)(gc->__driContext,
-							pdraw->textureTarget,
-							pdraw->textureFormat,
-							pdraw->driDrawable);
-	    } else {
-		(*pdraw->psc->texBuffer->setTexBuffer)(gc->__driContext,
-						       pdraw->textureTarget,
-						       pdraw->driDrawable);
-	    }
-	}
-	return;
-    }
+   if (gc->driContext) {
+      __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL);
+
+      if (pdraw != NULL) {
+         if (pdraw->psc->texBuffer->base.version >= 2 &&
+             pdraw->psc->texBuffer->setTexBuffer2 != NULL) {
+            (*pdraw->psc->texBuffer->setTexBuffer2) (gc->__driContext,
+                                                     pdraw->textureTarget,
+                                                     pdraw->textureFormat,
+                                                     pdraw->driDrawable);
+         }
+         else {
+            (*pdraw->psc->texBuffer->setTexBuffer) (gc->__driContext,
+                                                    pdraw->textureTarget,
+                                                    pdraw->driDrawable);
+         }
+      }
+      return;
+   }
 #endif
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode)
-	return;
-
-    LockDisplay(dpy);
-    GetReqExtra(GLXVendorPrivate, 12 + 8 * i,req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXVendorPrivate;
-    req->vendorCode = X_GLXvop_BindTexImageEXT;
-    req->contextTag = gc->currentContextTag;
-
-    drawable_ptr = (CARD32 *) (req + 1);
-    buffer_ptr = (INT32 *) (drawable_ptr + 1);
-    num_attrib_ptr = (CARD32 *) (buffer_ptr + 1);
-    attrib_ptr = (CARD32 *) (num_attrib_ptr + 1);
-
-    *drawable_ptr = drawable;
-    *buffer_ptr = buffer;
-    *num_attrib_ptr = (CARD32) i;
-
-    i = 0;
-    if (attrib_list) {
- 	while (attrib_list[i * 2] != None)
- 	{
- 	    *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 0];
- 	    *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 1];
- 	    i++;
- 	}
-    }
-
-    UnlockDisplay(dpy);
-    SyncHandle();
-}
-
-static void __glXReleaseTexImageEXT(Display *dpy,
-				    GLXDrawable drawable,
-				    int buffer)
-{
-    xGLXVendorPrivateReq *req;
-    GLXContext gc = __glXGetCurrentContext();
-    CARD32 *drawable_ptr;
-    INT32 *buffer_ptr;
-    CARD8 opcode;
-
-    if (gc == NULL)
-	return;
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode)
+      return;
+
+   LockDisplay(dpy);
+   GetReqExtra(GLXVendorPrivate, 12 + 8 * i, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXVendorPrivate;
+   req->vendorCode = X_GLXvop_BindTexImageEXT;
+   req->contextTag = gc->currentContextTag;
+
+   drawable_ptr = (CARD32 *) (req + 1);
+   buffer_ptr = (INT32 *) (drawable_ptr + 1);
+   num_attrib_ptr = (CARD32 *) (buffer_ptr + 1);
+   attrib_ptr = (CARD32 *) (num_attrib_ptr + 1);
+
+   *drawable_ptr = drawable;
+   *buffer_ptr = buffer;
+   *num_attrib_ptr = (CARD32) i;
+
+   i = 0;
+   if (attrib_list) {
+      while (attrib_list[i * 2] != None) {
+         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 0];
+         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 1];
+         i++;
+      }
+   }
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static void
+__glXReleaseTexImageEXT(Display * dpy, GLXDrawable drawable, int buffer)
+{
+   xGLXVendorPrivateReq *req;
+   GLXContext gc = __glXGetCurrentContext();
+   CARD32 *drawable_ptr;
+   INT32 *buffer_ptr;
+   CARD8 opcode;
+
+   if (gc == NULL)
+      return;
 
 #ifdef GLX_DIRECT_RENDERING
-    if (gc->driContext)
-	return;
+   if (gc->driContext)
+      return;
 #endif
 
-    opcode = __glXSetupForCommand(dpy);
-    if (!opcode)
-	return;
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode)
+      return;
 
-    LockDisplay(dpy);
-    GetReqExtra(GLXVendorPrivate, sizeof(CARD32)+sizeof(INT32),req);
-    req->reqType = opcode;
-    req->glxCode = X_GLXVendorPrivate;
-    req->vendorCode = X_GLXvop_ReleaseTexImageEXT;
-    req->contextTag = gc->currentContextTag;
+   LockDisplay(dpy);
+   GetReqExtra(GLXVendorPrivate, sizeof(CARD32) + sizeof(INT32), req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXVendorPrivate;
+   req->vendorCode = X_GLXvop_ReleaseTexImageEXT;
+   req->contextTag = gc->currentContextTag;
 
-    drawable_ptr = (CARD32 *) (req + 1);
-    buffer_ptr = (INT32 *) (drawable_ptr + 1);
+   drawable_ptr = (CARD32 *) (req + 1);
+   buffer_ptr = (INT32 *) (drawable_ptr + 1);
 
-    *drawable_ptr = drawable;
-    *buffer_ptr = buffer;
+   *drawable_ptr = drawable;
+   *buffer_ptr = buffer;
 
-    UnlockDisplay(dpy);
-    SyncHandle();
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
+
 /*@}*/
 
 /**
  * \c strdup is actually not a standard ANSI C or POSIX routine.
  * Irix will not define it if ANSI mode is in effect.
- * 
+ *
  * \sa strdup
  */
 _X_HIDDEN char *
@@ -2787,7 +2865,8 @@ __glXstrdup(const char *str)
 ** glXGetProcAddress support
 */
 
-struct name_address_pair {
+struct name_address_pair
+{
    const char *Name;
    GLvoid *Address;
 };
@@ -2797,139 +2876,139 @@ struct name_address_pair {
 
 static const struct name_address_pair GLX_functions[] = {
    /*** GLX_VERSION_1_0 ***/
-   GLX_FUNCTION( glXChooseVisual ),
-   GLX_FUNCTION( glXCopyContext ),
-   GLX_FUNCTION( glXCreateContext ),
-   GLX_FUNCTION( glXCreateGLXPixmap ),
-   GLX_FUNCTION( glXDestroyContext ),
-   GLX_FUNCTION( glXDestroyGLXPixmap ),
-   GLX_FUNCTION( glXGetConfig ),
-   GLX_FUNCTION( glXGetCurrentContext ),
-   GLX_FUNCTION( glXGetCurrentDrawable ),
-   GLX_FUNCTION( glXIsDirect ),
-   GLX_FUNCTION( glXMakeCurrent ),
-   GLX_FUNCTION( glXQueryExtension ),
-   GLX_FUNCTION( glXQueryVersion ),
-   GLX_FUNCTION( glXSwapBuffers ),
-   GLX_FUNCTION( glXUseXFont ),
-   GLX_FUNCTION( glXWaitGL ),
-   GLX_FUNCTION( glXWaitX ),
+   GLX_FUNCTION(glXChooseVisual),
+   GLX_FUNCTION(glXCopyContext),
+   GLX_FUNCTION(glXCreateContext),
+   GLX_FUNCTION(glXCreateGLXPixmap),
+   GLX_FUNCTION(glXDestroyContext),
+   GLX_FUNCTION(glXDestroyGLXPixmap),
+   GLX_FUNCTION(glXGetConfig),
+   GLX_FUNCTION(glXGetCurrentContext),
+   GLX_FUNCTION(glXGetCurrentDrawable),
+   GLX_FUNCTION(glXIsDirect),
+   GLX_FUNCTION(glXMakeCurrent),
+   GLX_FUNCTION(glXQueryExtension),
+   GLX_FUNCTION(glXQueryVersion),
+   GLX_FUNCTION(glXSwapBuffers),
+   GLX_FUNCTION(glXUseXFont),
+   GLX_FUNCTION(glXWaitGL),
+   GLX_FUNCTION(glXWaitX),
 
    /*** GLX_VERSION_1_1 ***/
-   GLX_FUNCTION( glXGetClientString ),
-   GLX_FUNCTION( glXQueryExtensionsString ),
-   GLX_FUNCTION( glXQueryServerString ),
+   GLX_FUNCTION(glXGetClientString),
+   GLX_FUNCTION(glXQueryExtensionsString),
+   GLX_FUNCTION(glXQueryServerString),
 
    /*** GLX_VERSION_1_2 ***/
-   GLX_FUNCTION( glXGetCurrentDisplay ),
+   GLX_FUNCTION(glXGetCurrentDisplay),
 
    /*** GLX_VERSION_1_3 ***/
-   GLX_FUNCTION( glXChooseFBConfig ),
-   GLX_FUNCTION( glXCreateNewContext ),
-   GLX_FUNCTION( glXCreatePbuffer ),
-   GLX_FUNCTION( glXCreatePixmap ),
-   GLX_FUNCTION( glXCreateWindow ),
-   GLX_FUNCTION( glXDestroyPbuffer ),
-   GLX_FUNCTION( glXDestroyPixmap ),
-   GLX_FUNCTION( glXDestroyWindow ),
-   GLX_FUNCTION( glXGetCurrentReadDrawable ),
-   GLX_FUNCTION( glXGetFBConfigAttrib ),
-   GLX_FUNCTION( glXGetFBConfigs ),
-   GLX_FUNCTION( glXGetSelectedEvent ),
-   GLX_FUNCTION( glXGetVisualFromFBConfig ),
-   GLX_FUNCTION( glXMakeContextCurrent ),
-   GLX_FUNCTION( glXQueryContext ),
-   GLX_FUNCTION( glXQueryDrawable ),
-   GLX_FUNCTION( glXSelectEvent ),
+   GLX_FUNCTION(glXChooseFBConfig),
+   GLX_FUNCTION(glXCreateNewContext),
+   GLX_FUNCTION(glXCreatePbuffer),
+   GLX_FUNCTION(glXCreatePixmap),
+   GLX_FUNCTION(glXCreateWindow),
+   GLX_FUNCTION(glXDestroyPbuffer),
+   GLX_FUNCTION(glXDestroyPixmap),
+   GLX_FUNCTION(glXDestroyWindow),
+   GLX_FUNCTION(glXGetCurrentReadDrawable),
+   GLX_FUNCTION(glXGetFBConfigAttrib),
+   GLX_FUNCTION(glXGetFBConfigs),
+   GLX_FUNCTION(glXGetSelectedEvent),
+   GLX_FUNCTION(glXGetVisualFromFBConfig),
+   GLX_FUNCTION(glXMakeContextCurrent),
+   GLX_FUNCTION(glXQueryContext),
+   GLX_FUNCTION(glXQueryDrawable),
+   GLX_FUNCTION(glXSelectEvent),
 
    /*** GLX_SGI_swap_control ***/
-   GLX_FUNCTION2( glXSwapIntervalSGI, __glXSwapIntervalSGI ),
+   GLX_FUNCTION2(glXSwapIntervalSGI, __glXSwapIntervalSGI),
 
    /*** GLX_SGI_video_sync ***/
-   GLX_FUNCTION2( glXGetVideoSyncSGI, __glXGetVideoSyncSGI ),
-   GLX_FUNCTION2( glXWaitVideoSyncSGI, __glXWaitVideoSyncSGI ),
+   GLX_FUNCTION2(glXGetVideoSyncSGI, __glXGetVideoSyncSGI),
+   GLX_FUNCTION2(glXWaitVideoSyncSGI, __glXWaitVideoSyncSGI),
 
    /*** GLX_SGI_make_current_read ***/
-   GLX_FUNCTION2( glXMakeCurrentReadSGI, glXMakeContextCurrent ),
-   GLX_FUNCTION2( glXGetCurrentReadDrawableSGI, glXGetCurrentReadDrawable ),
+   GLX_FUNCTION2(glXMakeCurrentReadSGI, glXMakeContextCurrent),
+   GLX_FUNCTION2(glXGetCurrentReadDrawableSGI, glXGetCurrentReadDrawable),
 
    /*** GLX_EXT_import_context ***/
-   GLX_FUNCTION( glXFreeContextEXT ),
-   GLX_FUNCTION( glXGetContextIDEXT ),
-   GLX_FUNCTION2( glXGetCurrentDisplayEXT, glXGetCurrentDisplay ),
-   GLX_FUNCTION( glXImportContextEXT ),
-   GLX_FUNCTION2( glXQueryContextInfoEXT, glXQueryContext ),
+   GLX_FUNCTION(glXFreeContextEXT),
+   GLX_FUNCTION(glXGetContextIDEXT),
+   GLX_FUNCTION2(glXGetCurrentDisplayEXT, glXGetCurrentDisplay),
+   GLX_FUNCTION(glXImportContextEXT),
+   GLX_FUNCTION2(glXQueryContextInfoEXT, glXQueryContext),
 
    /*** GLX_SGIX_fbconfig ***/
-   GLX_FUNCTION2( glXGetFBConfigAttribSGIX, glXGetFBConfigAttrib ),
-   GLX_FUNCTION2( glXChooseFBConfigSGIX, glXChooseFBConfig ),
-   GLX_FUNCTION( glXCreateGLXPixmapWithConfigSGIX ),
-   GLX_FUNCTION( glXCreateContextWithConfigSGIX ),
-   GLX_FUNCTION2( glXGetVisualFromFBConfigSGIX, glXGetVisualFromFBConfig ),
-   GLX_FUNCTION( glXGetFBConfigFromVisualSGIX ),
+   GLX_FUNCTION2(glXGetFBConfigAttribSGIX, glXGetFBConfigAttrib),
+   GLX_FUNCTION2(glXChooseFBConfigSGIX, glXChooseFBConfig),
+   GLX_FUNCTION(glXCreateGLXPixmapWithConfigSGIX),
+   GLX_FUNCTION(glXCreateContextWithConfigSGIX),
+   GLX_FUNCTION2(glXGetVisualFromFBConfigSGIX, glXGetVisualFromFBConfig),
+   GLX_FUNCTION(glXGetFBConfigFromVisualSGIX),
 
    /*** GLX_SGIX_pbuffer ***/
-   GLX_FUNCTION( glXCreateGLXPbufferSGIX ),
-   GLX_FUNCTION( glXDestroyGLXPbufferSGIX ),
-   GLX_FUNCTION( glXQueryGLXPbufferSGIX ),
-   GLX_FUNCTION( glXSelectEventSGIX ),
-   GLX_FUNCTION( glXGetSelectedEventSGIX ),
+   GLX_FUNCTION(glXCreateGLXPbufferSGIX),
+   GLX_FUNCTION(glXDestroyGLXPbufferSGIX),
+   GLX_FUNCTION(glXQueryGLXPbufferSGIX),
+   GLX_FUNCTION(glXSelectEventSGIX),
+   GLX_FUNCTION(glXGetSelectedEventSGIX),
 
    /*** GLX_SGIX_swap_group ***/
-   GLX_FUNCTION2( glXJoinSwapGroupSGIX, __glXJoinSwapGroupSGIX ),
+   GLX_FUNCTION2(glXJoinSwapGroupSGIX, __glXJoinSwapGroupSGIX),
 
    /*** GLX_SGIX_swap_barrier ***/
-   GLX_FUNCTION2( glXBindSwapBarrierSGIX, __glXBindSwapBarrierSGIX ),
-   GLX_FUNCTION2( glXQueryMaxSwapBarriersSGIX, __glXQueryMaxSwapBarriersSGIX ),
+   GLX_FUNCTION2(glXBindSwapBarrierSGIX, __glXBindSwapBarrierSGIX),
+   GLX_FUNCTION2(glXQueryMaxSwapBarriersSGIX, __glXQueryMaxSwapBarriersSGIX),
 
    /*** GLX_MESA_allocate_memory ***/
-   GLX_FUNCTION( glXAllocateMemoryMESA ),
-   GLX_FUNCTION( glXFreeMemoryMESA ),
-   GLX_FUNCTION( glXGetMemoryOffsetMESA ),
+   GLX_FUNCTION(glXAllocateMemoryMESA),
+   GLX_FUNCTION(glXFreeMemoryMESA),
+   GLX_FUNCTION(glXGetMemoryOffsetMESA),
 
    /*** GLX_MESA_copy_sub_buffer ***/
-   GLX_FUNCTION2( glXCopySubBufferMESA, __glXCopySubBufferMESA ),
+   GLX_FUNCTION2(glXCopySubBufferMESA, __glXCopySubBufferMESA),
 
    /*** GLX_MESA_pixmap_colormap ***/
-   GLX_FUNCTION( glXCreateGLXPixmapMESA ),
+   GLX_FUNCTION(glXCreateGLXPixmapMESA),
 
    /*** GLX_MESA_release_buffers ***/
-   GLX_FUNCTION2( glXReleaseBuffersMESA, __glXReleaseBuffersMESA ),
+   GLX_FUNCTION2(glXReleaseBuffersMESA, __glXReleaseBuffersMESA),
 
    /*** GLX_MESA_swap_control ***/
-   GLX_FUNCTION2( glXSwapIntervalMESA, __glXSwapIntervalMESA ),
-   GLX_FUNCTION2( glXGetSwapIntervalMESA, __glXGetSwapIntervalMESA ),
+   GLX_FUNCTION2(glXSwapIntervalMESA, __glXSwapIntervalMESA),
+   GLX_FUNCTION2(glXGetSwapIntervalMESA, __glXGetSwapIntervalMESA),
 
    /*** GLX_MESA_swap_frame_usage ***/
-   GLX_FUNCTION2( glXBeginFrameTrackingMESA, __glXBeginFrameTrackingMESA ),
-   GLX_FUNCTION2( glXEndFrameTrackingMESA, __glXEndFrameTrackingMESA ),
-   GLX_FUNCTION2( glXGetFrameUsageMESA, __glXGetFrameUsageMESA ),
-   GLX_FUNCTION2( glXQueryFrameTrackingMESA, __glXQueryFrameTrackingMESA ),
+   GLX_FUNCTION2(glXBeginFrameTrackingMESA, __glXBeginFrameTrackingMESA),
+   GLX_FUNCTION2(glXEndFrameTrackingMESA, __glXEndFrameTrackingMESA),
+   GLX_FUNCTION2(glXGetFrameUsageMESA, __glXGetFrameUsageMESA),
+   GLX_FUNCTION2(glXQueryFrameTrackingMESA, __glXQueryFrameTrackingMESA),
 
    /*** GLX_ARB_get_proc_address ***/
-   GLX_FUNCTION( glXGetProcAddressARB ),
+   GLX_FUNCTION(glXGetProcAddressARB),
 
    /*** GLX 1.4 ***/
-   GLX_FUNCTION2( glXGetProcAddress, glXGetProcAddressARB ),
+   GLX_FUNCTION2(glXGetProcAddress, glXGetProcAddressARB),
 
    /*** GLX_OML_sync_control ***/
-   GLX_FUNCTION2( glXWaitForSbcOML, __glXWaitForSbcOML ),
-   GLX_FUNCTION2( glXWaitForMscOML, __glXWaitForMscOML ),
-   GLX_FUNCTION2( glXSwapBuffersMscOML, __glXSwapBuffersMscOML ),
-   GLX_FUNCTION2( glXGetMscRateOML, __glXGetMscRateOML ),
-   GLX_FUNCTION2( glXGetSyncValuesOML, __glXGetSyncValuesOML ),
+   GLX_FUNCTION2(glXWaitForSbcOML, __glXWaitForSbcOML),
+   GLX_FUNCTION2(glXWaitForMscOML, __glXWaitForMscOML),
+   GLX_FUNCTION2(glXSwapBuffersMscOML, __glXSwapBuffersMscOML),
+   GLX_FUNCTION2(glXGetMscRateOML, __glXGetMscRateOML),
+   GLX_FUNCTION2(glXGetSyncValuesOML, __glXGetSyncValuesOML),
 
    /*** GLX_EXT_texture_from_pixmap ***/
-   GLX_FUNCTION2( glXBindTexImageEXT, __glXBindTexImageEXT ),
-   GLX_FUNCTION2( glXReleaseTexImageEXT, __glXReleaseTexImageEXT ),
+   GLX_FUNCTION2(glXBindTexImageEXT, __glXBindTexImageEXT),
+   GLX_FUNCTION2(glXReleaseTexImageEXT, __glXReleaseTexImageEXT),
 
 #ifdef GLX_DIRECT_RENDERING
    /*** DRI configuration ***/
-   GLX_FUNCTION( glXGetScreenDriver ),
-   GLX_FUNCTION( glXGetDriverConfig ),
+   GLX_FUNCTION(glXGetScreenDriver),
+   GLX_FUNCTION(glXGetDriverConfig),
 #endif
 
-   { NULL, NULL }   /* end of list */
+   {NULL, NULL}                 /* end of list */
 };
 
 
@@ -2941,7 +3020,7 @@ get_glx_proc_address(const char *funcName)
    /* try static functions */
    for (i = 0; GLX_functions[i].Name; i++) {
       if (strcmp(GLX_functions[i].Name, funcName) == 0)
-	 return GLX_functions[i].Address;
+         return GLX_functions[i].Address;
    }
 
    return NULL;
@@ -2957,9 +3036,9 @@ get_glx_proc_address(const char *funcName)
  *
  * \sa glXGetProcAddress
  */
-PUBLIC void (*glXGetProcAddressARB(const GLubyte *procName))( void )
+PUBLIC void (*glXGetProcAddressARB(const GLubyte * procName)) (void)
 {
-   typedef void (*gl_function)( void );
+   typedef void (*gl_function) (void);
    gl_function f;
 
 
@@ -2971,8 +3050,8 @@ PUBLIC void (*glXGetProcAddressARB(const GLubyte *procName))( void )
     */
 
    f = (gl_function) get_glx_proc_address((const char *) procName);
-   if ( (f == NULL) && (procName[0] == 'g') && (procName[1] == 'l')
-	&& (procName[2] != 'X') ) {
+   if ((f == NULL) && (procName[0] == 'g') && (procName[1] == 'l')
+       && (procName[2] != 'X')) {
       f = (gl_function) _glapi_get_proc_address((const char *) procName);
    }
 
@@ -2988,9 +3067,9 @@ PUBLIC void (*glXGetProcAddressARB(const GLubyte *procName))( void )
  *
  * \sa glXGetProcAddressARB
  */
-PUBLIC void (*glXGetProcAddress(const GLubyte *procName))( void )
+PUBLIC void (*glXGetProcAddress(const GLubyte * procName)) (void)
 #if defined(__GNUC__) && !defined(GLX_ALIAS_UNSUPPORTED)
-    __attribute__ ((alias ("glXGetProcAddressARB")));
+   __attribute__ ((alias("glXGetProcAddressARB")));
 #else
 {
    return glXGetProcAddressARB(procName);
@@ -3008,24 +3087,26 @@ PUBLIC void (*glXGetProcAddress(const GLubyte *procName))( void )
  *
  * \param ust Location to store the 64-bit UST
  * \returns Zero on success or a negative errno value on failure.
- * 
+ *
  * \sa glXGetProcAddress, PFNGLXGETUSTPROC
  *
  * \since Internal API version 20030317.
  */
-_X_HIDDEN int __glXGetUST( int64_t * ust )
-{
-    struct timeval  tv;
-    
-    if ( ust == NULL ) {
-	return -EFAULT;
-    }
-
-    if ( gettimeofday( & tv, NULL ) == 0 ) {
-	ust[0] = (tv.tv_sec * 1000000) + tv.tv_usec;
-	return 0;
-    } else {
-	return -errno;
-    }
+_X_HIDDEN int
+__glXGetUST(int64_t * ust)
+{
+   struct timeval tv;
+
+   if (ust == NULL) {
+      return -EFAULT;
+   }
+
+   if (gettimeofday(&tv, NULL) == 0) {
+      ust[0] = (tv.tv_sec * 1000000) + tv.tv_usec;
+      return 0;
+   }
+   else {
+      return -errno;
+   }
 }
 #endif /* GLX_DIRECT_RENDERING */
diff --git a/src/glx/x11/glxcurrent.c b/src/glx/x11/glxcurrent.c
index d44e0dd1fc..f1e3e161be 100644
--- a/src/glx/x11/glxcurrent.c
+++ b/src/glx/x11/glxcurrent.c
@@ -73,45 +73,7 @@ static __GLapi *IndirectAPI = NULL;
  * Current context management and locking
  */
 
-#if defined( USE_XTHREADS )
-
-/* thread safe */
-static GLboolean TSDinitialized = GL_FALSE;
-static xthread_key_t ContextTSD;
-
-_X_HIDDEN __GLXcontext *
-__glXGetCurrentContext(void)
-{
-   if (!TSDinitialized) {
-      xthread_key_create(&ContextTSD, NULL);
-      TSDinitialized = GL_TRUE;
-      return &dummyContext;
-   }
-   else {
-      void *p;
-      xthread_get_specific(ContextTSD, &p);
-      if (!p)
-         return &dummyContext;
-      else
-         return (__GLXcontext *) p;
-   }
-}
-
-_X_HIDDEN void
-__glXSetCurrentContext(__GLXcontext * c)
-{
-   if (!TSDinitialized) {
-      xthread_key_create(&ContextTSD, NULL);
-      TSDinitialized = GL_TRUE;
-   }
-   xthread_set_specific(ContextTSD, c);
-}
-
-
-/* Used by the __glXLock() and __glXUnlock() macros */
-_X_HIDDEN xmutex_rec __glXmutex;
-
-#elif defined( PTHREADS )
+#if defined( PTHREADS )
 
 _X_HIDDEN pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER;
 
@@ -119,7 +81,7 @@ _X_HIDDEN pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER;
 
 /**
  * Per-thread GLX context pointer.
- * 
+ *
  * \c __glXSetCurrentContext is written is such a way that this pointer can
  * \b never be \c NULL.  This is important!  Because of this
  * \c __glXGetCurrentContext can be implemented as trivial macro.
@@ -139,7 +101,7 @@ static pthread_once_t once_control = PTHREAD_ONCE_INIT;
 
 /**
  * Per-thread data key.
- * 
+ *
  * Once \c init_thread_data has been called, the per-thread data key will
  * take a value of \c NULL.  As each new thread is created the default
  * value, in that thread, will be \c NULL.
@@ -148,7 +110,7 @@ static pthread_key_t ContextTSD;
 
 /**
  * Initialize the per-thread data key.
- * 
+ *
  * This function is called \b exactly once per-process (not per-thread!) to
  * initialize the per-thread data key.  This is ideally done using the
  * \c pthread_once mechanism.
@@ -340,8 +302,8 @@ FetchDRIDrawable(Display * dpy, GLXDrawable glxDrawable, GLXContext gc)
 #endif /* GLX_DIRECT_RENDERING */
 
 static void
-__glXGenerateError(Display *dpy, GLXContext gc, XID resource,
-		   BYTE errorCode, CARD16 minorCode)
+__glXGenerateError(Display * dpy, GLXContext gc, XID resource,
+                   BYTE errorCode, CARD16 minorCode)
 {
    xError error;
 
@@ -356,7 +318,7 @@ __glXGenerateError(Display *dpy, GLXContext gc, XID resource,
 
 /**
  * Make a particular context current.
- * 
+ *
  * \note This is in this file so that it can access dummyContext.
  */
 static Bool
@@ -385,21 +347,19 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
 
    if (gc == NULL && (draw != None || read != None)) {
       __glXGenerateError(dpy, gc, (draw != None) ? draw : read,
-			 BadMatch, X_GLXMakeContextCurrent);
+                         BadMatch, X_GLXMakeContextCurrent);
       return False;
    }
    if (gc != NULL && (draw == None || read == None)) {
-      __glXGenerateError(dpy, gc, None,
-			 BadMatch, X_GLXMakeContextCurrent);
+      __glXGenerateError(dpy, gc, None, BadMatch, X_GLXMakeContextCurrent);
       return False;
    }
 
    _glapi_check_multithread();
 
-   if (gc != NULL && gc->thread_id != 0 &&
-       gc->thread_id != _glthread_GetID()) {
+   if (gc != NULL && gc->thread_id != 0 && gc->thread_id != _glthread_GetID()) {
       __glXGenerateError(dpy, gc, gc->xid,
-			 BadAccess, X_GLXMakeContextCurrent);
+                         BadAccess, X_GLXMakeContextCurrent);
       return False;
    }
 
@@ -411,7 +371,7 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
 
       if ((pdraw == NULL) || (pread == NULL)) {
          __glXGenerateError(dpy, gc, (pdraw == NULL) ? draw : read,
-			    GLXBadDrawable, X_GLXMakeContextCurrent);
+                            GLXBadDrawable, X_GLXMakeContextCurrent);
          return False;
       }
 
@@ -559,6 +519,5 @@ GLX_ALIAS(Bool, glXMakeCurrentReadSGI,
 
 PUBLIC
 GLX_ALIAS(Bool, glXMakeContextCurrent,
-	  (Display * dpy, GLXDrawable d, GLXDrawable r,
-	   GLXContext ctx), (dpy, d, r, ctx),
-	  MakeContextCurrent)
+          (Display * dpy, GLXDrawable d, GLXDrawable r,
+           GLXContext ctx), (dpy, d, r, ctx), MakeContextCurrent)
diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c
index 3078662c9d..e5553cbf76 100644
--- a/src/glx/x11/glxext.c
+++ b/src/glx/x11/glxext.c
@@ -102,19 +102,19 @@ static
 XEXT_GENERATE_ERROR_STRING(__glXErrorString, __glXExtensionName,
                            __GLX_NUMBER_ERRORS, error_list)
 
-     static /* const */ XExtensionHooks __glXExtensionHooks = {
-        NULL,                   /* create_gc */
-        NULL,                   /* copy_gc */
-        NULL,                   /* flush_gc */
-        NULL,                   /* free_gc */
-        NULL,                   /* create_font */
-        NULL,                   /* free_font */
-        __glXCloseDisplay,      /* close_display */
-        NULL,                   /* wire_to_event */
-        NULL,                   /* event_to_wire */
-        NULL,                   /* error */
-        __glXErrorString,       /* error_string */
-     };
+static /* const */ XExtensionHooks __glXExtensionHooks = {
+  NULL,                   /* create_gc */
+  NULL,                   /* copy_gc */
+  NULL,                   /* flush_gc */
+  NULL,                   /* free_gc */
+  NULL,                   /* create_font */
+  NULL,                   /* free_font */
+  __glXCloseDisplay,      /* close_display */
+  NULL,                   /* wire_to_event */
+  NULL,                   /* event_to_wire */
+  NULL,                   /* error */
+  __glXErrorString,       /* error_string */
+};
 
 static
 XEXT_GENERATE_FIND_DISPLAY(__glXFindDisplay, __glXExtensionInfo,
@@ -150,8 +150,8 @@ FreeScreenConfigs(__GLXdisplayPrivate * priv)
 
 #ifdef GLX_DIRECT_RENDERING
       if (psc->driver_configs) {
-         for(unsigned int i = 0; psc->driver_configs[i]; i++)
-            free((__DRIconfig*)psc->driver_configs[i]);
+         for (unsigned int i = 0; psc->driver_configs[i]; i++)
+            free((__DRIconfig *) psc->driver_configs[i]);
          free(psc->driver_configs);
          psc->driver_configs = NULL;
       }
@@ -217,15 +217,14 @@ QueryVersion(Display * dpy, int opcode, int *major, int *minor)
 {
 #ifdef USE_XCB
    xcb_connection_t *c = XGetXCBConnection(dpy);
-   xcb_glx_query_version_reply_t* reply =
-      xcb_glx_query_version_reply(c,
-                                  xcb_glx_query_version(c,
-                                                        GLX_MAJOR_VERSION,
-                                                        GLX_MINOR_VERSION),
-                                  NULL);
-
-   if(reply->major_version != GLX_MAJOR_VERSION)
-   {
+   xcb_glx_query_version_reply_t *reply = xcb_glx_query_version_reply(c,
+                                                                      xcb_glx_query_version
+                                                                      (c,
+                                                                       GLX_MAJOR_VERSION,
+                                                                       GLX_MINOR_VERSION),
+                                                                      NULL);
+
+   if (reply->major_version != GLX_MAJOR_VERSION) {
       free(reply);
       return GL_FALSE;
    }
@@ -543,7 +542,8 @@ getFBConfigs(Display * dpy, __GLXdisplayPrivate * priv, int screen)
    __GLXscreenConfigs *psc;
 
    psc = priv->screenConfigs + screen;
-   psc->serverGLXexts = __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
+   psc->serverGLXexts =
+      __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
 
    LockDisplay(dpy);
 
@@ -601,7 +601,8 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
    memset(psc, 0, screens * sizeof(__GLXscreenConfigs));
    priv->screenConfigs = psc;
 
-   priv->serverGLXversion = __glXQueryServerString(dpy, priv->majorOpcode, 0, GLX_VERSION);
+   priv->serverGLXversion =
+      __glXQueryServerString(dpy, priv->majorOpcode, 0, GLX_VERSION);
    if (priv->serverGLXversion == NULL) {
       FreeScreenConfigs(priv);
       return GL_FALSE;
@@ -652,17 +653,6 @@ __glXInitialize(Display * dpy)
    Bool glx_direct, glx_accel;
 #endif
 
-#if defined(USE_XTHREADS)
-   {
-      static int firstCall = 1;
-      if (firstCall) {
-         /* initialize the GLX mutexes */
-         xmutex_init(&__glXmutex);
-         firstCall = 0;
-      }
-   }
-#endif
-
    /* The one and only long long lock */
    __glXLock();
 
@@ -797,10 +787,10 @@ __glXSetupForCommand(Display * dpy)
 
 /**
  * Flush the drawing command transport buffer.
- * 
+ *
  * \param ctx  Context whose transport buffer is to be flushed.
  * \param pc   Pointer to first unused buffer location.
- * 
+ *
  * \todo
  * Modify this function to use \c ctx->pc instead of the explicit
  * \c pc parameter.
@@ -892,11 +882,11 @@ __glXSendLargeChunk(__GLXcontext * gc, GLint requestNumber,
 
 /**
  * Send a command that is too large for the GLXRender protocol request.
- * 
+ *
  * Send a large command, one that is too large for some reason to
  * send using the GLXRender protocol request.  One reason to send
  * a large command is to avoid copying the data.
- * 
+ *
  * \param ctx        GLX context
  * \param header     Header data.
  * \param headerLen  Size, in bytes, of the header data.  It is assumed that
diff --git a/src/glx/x11/glxextensions.c b/src/glx/x11/glxextensions.c
index 051433a006..473f46d1e5 100644
--- a/src/glx/x11/glxextensions.c
+++ b/src/glx/x11/glxextensions.c
@@ -315,10 +315,10 @@ set_glx_extension(const struct extension_info *ext,
 
 /**
  * Convert the server's extension string to a bit-field.
- * 
+ *
  * \param server_string   GLX extension string from the server.
  * \param server_support  Bit-field of supported extensions.
- * 
+ *
  * \note
  * This function is used to process both GLX and GL extension strings.  The
  * bit-fields used to track each of these have different sizes.  Therefore,
@@ -336,8 +336,7 @@ __glXProcessServerString(const struct extension_info *ext,
       /* Determine the length of the next extension name.
        */
       for (len = 0; (server_string[base + len] != SEPARATOR)
-           && (server_string[base + len] != NUL);
-           len++) {
+           && (server_string[base + len] != NUL); len++) {
          /* empty */
       }
 
@@ -351,8 +350,7 @@ __glXProcessServerString(const struct extension_info *ext,
        * over the previous string and any trialing white-space.
        */
       for (base += len; (server_string[base] == SEPARATOR)
-           && (server_string[base] != NUL);
-           base++) {
+           && (server_string[base] != NUL); base++) {
          /* empty */
       }
    }
@@ -549,7 +547,7 @@ __glXGetClientExtensions(void)
 /**
  * Calculate the list of application usable extensions.  The resulting
  * string is stored in \c psc->effectiveGLXexts.
- * 
+ *
  * \param psc                        Pointer to GLX per-screen record.
  * \param display_is_direct_capable  True if the display is capable of
  *                                   direct rendering.
@@ -600,7 +598,7 @@ __glXCalculateUsableExtensions(__GLXscreenConfigs * psc,
     * it and the "server" supports it.  In this case that means that either
     * the true server supports it or it is only for direct-rendering and
     * the direct rendering driver supports it.
-    * 
+    *
     * If the display is not capable of direct rendering, then the extension
     * is enabled if and only if the client-side library and the server
     * support it.
@@ -609,10 +607,10 @@ __glXCalculateUsableExtensions(__GLXscreenConfigs * psc,
    if (display_is_direct_capable) {
       for (i = 0; i < 8; i++) {
          usable[i] = (client_glx_support[i] & client_glx_only[i])
-            | (client_glx_support[i] & psc->
-               direct_support[i] & server_support[i])
-            | (client_glx_support[i] & psc->
-               direct_support[i] & direct_glx_only[i]);
+            | (client_glx_support[i] & psc->direct_support[i] &
+               server_support[i])
+            | (client_glx_support[i] & psc->direct_support[i] &
+               direct_glx_only[i]);
       }
    }
    else {
@@ -630,7 +628,7 @@ __glXCalculateUsableExtensions(__GLXscreenConfigs * psc,
 /**
  * Calculate the list of application usable extensions.  The resulting
  * string is stored in \c gc->extensions.
- * 
+ *
  * \param gc             Pointer to GLX context.
  * \param server_string  Extension string from the server.
  * \param major_version  GL major version from the server.
diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c
index aea117ec32..e0cafd43bc 100644
--- a/src/glx/x11/indirect.c
+++ b/src/glx/x11/indirect.c
@@ -5201,10 +5201,9 @@ glDeleteTexturesEXT(GLsizei n, const GLuint * textures)
 #ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_DeleteTextures(GET_DISPATCH(), (n, textures));
-    } else {
-#else
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -5270,13 +5269,12 @@ glGenTexturesEXT(GLsizei n, GLuint * textures)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GenTextures(GET_DISPATCH(), (n, textures));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4;
@@ -5336,13 +5334,12 @@ glIsTextureEXT(GLuint texture)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         return CALL_IsTexture(GET_DISPATCH(), (texture));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         GLboolean retval = (GLboolean) 0;
@@ -5653,13 +5650,12 @@ glGetColorTableEXT(GLenum target, GLenum format, GLenum type, GLvoid * table)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTable(GET_DISPATCH(), (target, format, type, table));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -5730,14 +5726,13 @@ glGetColorTableParameterfvEXT(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTableParameterfv(GET_DISPATCH(),
                                       (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -5804,14 +5799,13 @@ glGetColorTableParameterivEXT(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTableParameteriv(GET_DISPATCH(),
                                       (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6131,14 +6125,13 @@ gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionFilter(GET_DISPATCH(),
                                   (target, format, type, image));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6210,14 +6203,13 @@ gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionParameterfv(GET_DISPATCH(),
                                        (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6284,14 +6276,13 @@ gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionParameteriv(GET_DISPATCH(),
                                        (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6365,14 +6356,13 @@ gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogram(GET_DISPATCH(),
                           (target, reset, format, type, values));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6443,13 +6433,12 @@ gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogramParameterfv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6515,13 +6504,12 @@ gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogramParameteriv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6591,13 +6579,12 @@ gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmax(GET_DISPATCH(), (target, reset, format, type, values));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6666,13 +6653,12 @@ gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmaxParameterfv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6735,13 +6721,12 @@ gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmaxParameteriv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
diff --git a/src/glx/x11/indirect_size.c b/src/glx/x11/indirect_size.c
index 54c039dd6c..cdaf02ffe6 100644
--- a/src/glx/x11/indirect_size.c
+++ b/src/glx/x11/indirect_size.c
@@ -192,6 +192,10 @@ __glTexParameterfv_size(GLenum e)
     case GL_TEXTURE_MAX_ANISOTROPY_EXT:
     case GL_TEXTURE_LOD_BIAS:
 /*      case GL_TEXTURE_LOD_BIAS_EXT:*/
+    case GL_TEXTURE_STORAGE_HINT_APPLE:
+    case GL_STORAGE_PRIVATE_APPLE:
+    case GL_STORAGE_CACHED_APPLE:
+    case GL_STORAGE_SHARED_APPLE:
     case GL_DEPTH_TEXTURE_MODE:
 /*      case GL_DEPTH_TEXTURE_MODE_ARB:*/
     case GL_TEXTURE_COMPARE_MODE:
diff --git a/src/glx/x11/indirect_vertex_array.c b/src/glx/x11/indirect_vertex_array.c
index bb5232f618..ad9882528f 100644
--- a/src/glx/x11/indirect_vertex_array.c
+++ b/src/glx/x11/indirect_vertex_array.c
@@ -1,18 +1,18 @@
 /*
  * (C) Copyright IBM Corporation 2004, 2005
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
  * to deal in the Software without restriction, including without limitation
  * the rights to use, copy, modify, merge, publish, distribute, sub license,
  * and/or sell copies of the Software, and to permit persons to whom the
  * Software is furnished to do so, subject to the following conditions:
- * 
+ *
  * The above copyright notice and this permission notice (including the next
  * paragraph) shall be included in all copies or substantial portions of the
  * Software.
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
@@ -39,7 +39,7 @@
 /**
  * \file indirect_vertex_array.c
  * Implement GLX protocol for vertex arrays and vertex buffer objects.
- * 
+ *
  * The most important function in this fill is \c fill_array_info_cache.
  * The \c array_state_vector contains a cache of the ARRAY_INFO data sent
  * in the DrawArrays protocol.  Certain operations, such as enabling or
@@ -47,7 +47,7 @@
  * fills-in this data.  Additionally, it examines the enabled state and
  * other factors to determine what "version" of DrawArrays protocoal can be
  * used.
- * 
+ *
  * Current, only two versions of DrawArrays protocol are implemented.  The
  * first version is the "none" protocol.  This is the fallback when the
  * server does not support GL 1.1 / EXT_vertex_arrays.  It is implemented
@@ -90,11 +90,11 @@ static GLboolean validate_type(__GLXcontext * gc, GLenum type);
 
 
 /**
- * Table of sizes, in bytes, of a GL types.  All of the type enums are be in 
+ * Table of sizes, in bytes, of a GL types.  All of the type enums are be in
  * the range 0x1400 - 0x140F.  That includes types added by extensions (i.e.,
  * \c GL_HALF_FLOAT_NV).  This elements of this table correspond to the
  * type enums masked with 0x0f.
- * 
+ *
  * \notes
  * \c GL_HALF_FLOAT_NV is not included.  Neither are \c GL_2_BYTES,
  * \c GL_3_BYTES, or \c GL_4_BYTES.
@@ -131,15 +131,15 @@ __glXFreeVertexArrayState(__GLXcontext * gc)
 
 /**
  * Initialize vertex array state of a GLX context.
- * 
+ *
  * \param gc  GLX context whose vertex array state is to be initialized.
- * 
+ *
  * \warning
  * This function may only be called after __GLXcontext::gl_extension_bits,
  * __GLXcontext::server_minor, and __GLXcontext::server_major have been
  * initialized.  These values are used to determine what vertex arrays are
  * supported.
- * 
+ *
  * \bug
  * Return values from malloc are not properly tested.
  */
@@ -171,7 +171,7 @@ __glXInitVertexArrayState(__GLXcontext * gc)
     * are supported by the server are create.  For example, if the server
     * supports only 2 texture units, then only 2 texture coordinate arrays
     * are created.
-    * 
+    *
     * At the very least, GL_VERTEX_ARRAY, GL_NORMAL_ARRAY,
     * GL_COLOR_ARRAY, GL_INDEX_ARRAY, GL_TEXTURE_COORD_ARRAY, and
     * GL_EDGE_FLAG_ARRAY are supported.
@@ -532,7 +532,7 @@ emit_DrawArrays_none(GLenum mode, GLint first, GLsizei count)
 /**
  * Emit the header data for the GL 1.1 / EXT_vertex_arrays DrawArrays
  * protocol.
- * 
+ *
  * \param gc                    GLX context.
  * \param arrays                Array state.
  * \param elements_per_request  Location to store the number of elements that
@@ -543,7 +543,7 @@ emit_DrawArrays_none(GLenum mode, GLint first, GLsizei count)
  *                              will be zero.
  * \param mode                  Drawing mode.
  * \param count                 Number of vertices.
- * 
+ *
  * \returns
  * A pointer to the buffer for array data.
  */
@@ -594,7 +594,7 @@ emit_DrawArrays_header_old(__GLXcontext * gc,
       /* Calculate the number of data packets that will be required to send
        * the whole command.  To do this, the number of verticies that
        * will fit in a single buffer must be calculated.
-       * 
+       *
        * The important value here is elements_per_request.  This is the
        * number of complete array elements that will fit in a single
        * buffer.  There may be some wasted space at the end of the buffer,
@@ -869,7 +869,7 @@ emit_DrawElements_old(GLenum mode, GLsizei count, GLenum type,
 /**
  * Validate that the \c mode parameter to \c glDrawArrays, et. al. is valid.
  * If it is not valid, then an error code is set in the GLX context.
- * 
+ *
  * \returns
  * \c GL_TRUE if the argument is valid, \c GL_FALSE if is not.
  */
@@ -902,7 +902,7 @@ validate_mode(__GLXcontext * gc, GLenum mode)
  * A value less than zero is invalid and will result in \c GL_INVALID_VALUE
  * being set.  A value of zero will not result in an error being set, but
  * will result in \c GL_FALSE being returned.
- * 
+ *
  * \returns
  * \c GL_TRUE if the argument is valid, \c GL_FALSE if it is not.
  */
@@ -1084,21 +1084,21 @@ __indirect_glMultiDrawElementsEXT(GLenum mode, const GLsizei * count,
 
 
 #define COMMON_ARRAY_DATA_INIT(a, PTR, TYPE, STRIDE, COUNT, NORMALIZED, HDR_SIZE, OPCODE) \
-    do { \
-	(a)->data = PTR; \
-	(a)->data_type = TYPE; \
-	(a)->user_stride = STRIDE; \
-	(a)->count = COUNT; \
-	(a)->normalized = NORMALIZED; \
-	\
-	(a)->element_size = __glXTypeSize( TYPE ) * COUNT; \
-	(a)->true_stride = (STRIDE == 0) \
-	  ? (a)->element_size : STRIDE; \
-	\
-	(a)->header_size = HDR_SIZE; \
-	((uint16_t *) (a)->header)[0] = __GLX_PAD((a)->header_size + (a)->element_size); \
-	((uint16_t *) (a)->header)[1] = OPCODE; \
-    } while(0)
+  do {                                                                  \
+    (a)->data = PTR;                                                    \
+    (a)->data_type = TYPE;                                              \
+    (a)->user_stride = STRIDE;                                          \
+    (a)->count = COUNT;                                                 \
+    (a)->normalized = NORMALIZED;                                       \
+                                                                        \
+    (a)->element_size = __glXTypeSize( TYPE ) * COUNT;                  \
+    (a)->true_stride = (STRIDE == 0)                                    \
+      ? (a)->element_size : STRIDE;                                     \
+                                                                        \
+    (a)->header_size = HDR_SIZE;                                        \
+    ((uint16_t *) (a)->header)[0] = __GLX_PAD((a)->header_size + (a)->element_size); \
+    ((uint16_t *) (a)->header)[1] = OPCODE;                             \
+  } while(0)
 
 
 void
@@ -1363,36 +1363,36 @@ __indirect_glTexCoordPointer(GLint size, GLenum type, GLsizei stride,
 {
    static const uint16_t short_ops[5] = {
       0, X_GLrop_TexCoord1sv, X_GLrop_TexCoord2sv, X_GLrop_TexCoord3sv,
-         X_GLrop_TexCoord4sv
+      X_GLrop_TexCoord4sv
    };
    static const uint16_t int_ops[5] = {
       0, X_GLrop_TexCoord1iv, X_GLrop_TexCoord2iv, X_GLrop_TexCoord3iv,
-         X_GLrop_TexCoord4iv
+      X_GLrop_TexCoord4iv
    };
    static const uint16_t float_ops[5] = {
       0, X_GLrop_TexCoord1dv, X_GLrop_TexCoord2fv, X_GLrop_TexCoord3fv,
-         X_GLrop_TexCoord4fv
+      X_GLrop_TexCoord4fv
    };
    static const uint16_t double_ops[5] = {
       0, X_GLrop_TexCoord1dv, X_GLrop_TexCoord2dv, X_GLrop_TexCoord3dv,
-         X_GLrop_TexCoord4dv
+      X_GLrop_TexCoord4dv
    };
 
    static const uint16_t mshort_ops[5] = {
       0, X_GLrop_MultiTexCoord1svARB, X_GLrop_MultiTexCoord2svARB,
-         X_GLrop_MultiTexCoord3svARB, X_GLrop_MultiTexCoord4svARB
+      X_GLrop_MultiTexCoord3svARB, X_GLrop_MultiTexCoord4svARB
    };
    static const uint16_t mint_ops[5] = {
       0, X_GLrop_MultiTexCoord1ivARB, X_GLrop_MultiTexCoord2ivARB,
-         X_GLrop_MultiTexCoord3ivARB, X_GLrop_MultiTexCoord4ivARB
+      X_GLrop_MultiTexCoord3ivARB, X_GLrop_MultiTexCoord4ivARB
    };
    static const uint16_t mfloat_ops[5] = {
       0, X_GLrop_MultiTexCoord1dvARB, X_GLrop_MultiTexCoord2fvARB,
-         X_GLrop_MultiTexCoord3fvARB, X_GLrop_MultiTexCoord4fvARB
+      X_GLrop_MultiTexCoord3fvARB, X_GLrop_MultiTexCoord4fvARB
    };
    static const uint16_t mdouble_ops[5] = {
       0, X_GLrop_MultiTexCoord1dvARB, X_GLrop_MultiTexCoord2dvARB,
-         X_GLrop_MultiTexCoord3dvARB, X_GLrop_MultiTexCoord4dvARB
+      X_GLrop_MultiTexCoord3dvARB, X_GLrop_MultiTexCoord4dvARB
    };
 
    uint16_t opcode;
diff --git a/src/glx/x11/single2.c b/src/glx/x11/single2.c
index 0b2d5113cc..d128ba2053 100644
--- a/src/glx/x11/single2.c
+++ b/src/glx/x11/single2.c
@@ -147,7 +147,7 @@ __indirect_glGetError(void)
 
 /**
  * Get the selected attribute from the client state.
- * 
+ *
  * \returns
  * On success \c GL_TRUE is returned.  Otherwise, \c GL_FALSE is returned.
  */
@@ -679,7 +679,8 @@ __indirect_glGetString(GLenum name)
     */
 
    (void) __glXFlushRenderBuffer(gc, gc->pc);
-   s = (GLubyte *) __glXGetString(dpy, gc->majorOpcode, gc->currentContextTag, name);
+   s = (GLubyte *) __glXGetString(dpy, gc->majorOpcode, gc->currentContextTag,
+                                  name);
    if (!s) {
       /* Throw data on the floor */
       __glXSetError(gc, GL_OUT_OF_MEMORY);
@@ -753,7 +754,7 @@ __indirect_glGetString(GLenum name)
              * hardware accelerated.  For example, a TNT will show core
              * version 1.5, but most of the post-1.2 functionality is a
              * software fallback.
-             * 
+             *
              * I don't want to break applications that rely on this odd
              * behavior.  At the same time, the code is written and tested,
              * so I didn't want to throw it away.  Therefore, the code is here
@@ -766,7 +767,7 @@ __indirect_glGetString(GLenum name)
              * gc->server_minor are set.  This version may be higher than we
              * can completely support, but it may imply support for some
              * extensions that we can support.
-             * 
+             *
              * For example, at the time of this writing, the client-side
              * library only supports upto core GL version 1.2.  However, cubic
              * textures, multitexture, multisampling, and some other 1.3
@@ -880,50 +881,50 @@ GLboolean
 __indirect_glAreTexturesResident(GLsizei n, const GLuint * textures,
                                  GLboolean * residences)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
-    Display *const dpy = gc->currentDpy;
-    GLboolean retval = (GLboolean) 0;
-    const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
-    if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) {
+   __GLXcontext *const gc = __glXGetCurrentContext();
+   Display *const dpy = gc->currentDpy;
+   GLboolean retval = (GLboolean) 0;
+   const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
+   if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) {
 #ifdef USE_XCB
-        xcb_connection_t *c = XGetXCBConnection(dpy);
-        (void) __glXFlushRenderBuffer(gc, gc->pc);
-        xcb_glx_are_textures_resident_reply_t *reply =
-            xcb_glx_are_textures_resident_reply(c,
-                                                xcb_glx_are_textures_resident
-                                                (c, gc->currentContextTag, n,
-                                                 textures), NULL);
-        (void) memcpy(residences, xcb_glx_are_textures_resident_data(reply),
-                      xcb_glx_are_textures_resident_data_length(reply) *
-                      sizeof(GLboolean));
-        retval = reply->ret_val;
-        free(reply);
+      xcb_connection_t *c = XGetXCBConnection(dpy);
+      (void) __glXFlushRenderBuffer(gc, gc->pc);
+      xcb_glx_are_textures_resident_reply_t *reply =
+         xcb_glx_are_textures_resident_reply(c,
+                                             xcb_glx_are_textures_resident
+                                             (c, gc->currentContextTag, n,
+                                              textures), NULL);
+      (void) memcpy(residences, xcb_glx_are_textures_resident_data(reply),
+                    xcb_glx_are_textures_resident_data_length(reply) *
+                    sizeof(GLboolean));
+      retval = reply->ret_val;
+      free(reply);
 #else
-        GLubyte const *pc =
-            __glXSetupSingleRequest(gc, X_GLsop_AreTexturesResident, cmdlen);
-        (void) memcpy((void *) (pc + 0), (void *) (&n), 4);
-        (void) memcpy((void *) (pc + 4), (void *) (textures), (n * 4));
-        if (n & 3) {
-           /* n is not a multiple of four.
-            * When reply_is_always_array is TRUE, __glXReadReply() will
-            * put a multiple of four bytes into the dest buffer.  If the
-            * caller's buffer is not a multiple of four in size, we'll write
-            * out of bounds.  So use a temporary buffer that's a few bytes
-            * larger.
-            */
-           GLboolean *res4 = malloc((n + 3) & ~3);
-           retval = (GLboolean) __glXReadReply(dpy, 1, res4, GL_TRUE);
-           memcpy(residences, res4, n);
-           free(res4);
-        }
-        else {
-           retval = (GLboolean) __glXReadReply(dpy, 1, residences, GL_TRUE);
-        }
-        UnlockDisplay(dpy);
-        SyncHandle();
+      GLubyte const *pc =
+         __glXSetupSingleRequest(gc, X_GLsop_AreTexturesResident, cmdlen);
+      (void) memcpy((void *) (pc + 0), (void *) (&n), 4);
+      (void) memcpy((void *) (pc + 4), (void *) (textures), (n * 4));
+      if (n & 3) {
+         /* n is not a multiple of four.
+          * When reply_is_always_array is TRUE, __glXReadReply() will
+          * put a multiple of four bytes into the dest buffer.  If the
+          * caller's buffer is not a multiple of four in size, we'll write
+          * out of bounds.  So use a temporary buffer that's a few bytes
+          * larger.
+          */
+         GLboolean *res4 = malloc((n + 3) & ~3);
+         retval = (GLboolean) __glXReadReply(dpy, 1, res4, GL_TRUE);
+         memcpy(residences, res4, n);
+         free(res4);
+      }
+      else {
+         retval = (GLboolean) __glXReadReply(dpy, 1, residences, GL_TRUE);
+      }
+      UnlockDisplay(dpy);
+      SyncHandle();
 #endif /* USE_XCB */
-    }
-    return retval;
+   }
+   return retval;
 }
 
 
@@ -936,36 +937,37 @@ GLboolean
 glAreTexturesResidentEXT(GLsizei n, const GLuint * textures,
                          GLboolean * residences)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
-
-    if (gc->isDirect) {
-        return CALL_AreTexturesResident(GET_DISPATCH(),
-                                        (n, textures, residences));
-    } else {
-        __GLXcontext *const gc = __glXGetCurrentContext();
-        Display *const dpy = gc->currentDpy;
-        GLboolean retval = (GLboolean) 0;
-        const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
-        if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) {
-            GLubyte const *pc =
-                __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply,
-                                        X_GLvop_AreTexturesResidentEXT,
-                                        cmdlen);
-            (void) memcpy((void *) (pc + 0), (void *) (&n), 4);
-            (void) memcpy((void *) (pc + 4), (void *) (textures), (n * 4));
-            if (n & 3) {
-               /* see comments in __indirect_glAreTexturesResident() */
-               GLboolean *res4 = malloc((n + 3) & ~3);
-               retval = (GLboolean) __glXReadReply(dpy, 1, res4, GL_TRUE);
-               memcpy(residences, res4, n);
-               free(res4);
-            }
-            else {
-               retval = (GLboolean) __glXReadReply(dpy, 1, residences, GL_TRUE);
-            }
-            UnlockDisplay(dpy);
-            SyncHandle();
-        }
-        return retval;
-    }
+   __GLXcontext *const gc = __glXGetCurrentContext();
+
+   if (gc->isDirect) {
+      return CALL_AreTexturesResident(GET_DISPATCH(),
+                                      (n, textures, residences));
+   }
+   else {
+      __GLXcontext *const gc = __glXGetCurrentContext();
+      Display *const dpy = gc->currentDpy;
+      GLboolean retval = (GLboolean) 0;
+      const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
+      if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) {
+         GLubyte const *pc =
+            __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply,
+                                    X_GLvop_AreTexturesResidentEXT,
+                                    cmdlen);
+         (void) memcpy((void *) (pc + 0), (void *) (&n), 4);
+         (void) memcpy((void *) (pc + 4), (void *) (textures), (n * 4));
+         if (n & 3) {
+            /* see comments in __indirect_glAreTexturesResident() */
+            GLboolean *res4 = malloc((n + 3) & ~3);
+            retval = (GLboolean) __glXReadReply(dpy, 1, res4, GL_TRUE);
+            memcpy(residences, res4, n);
+            free(res4);
+         }
+         else {
+            retval = (GLboolean) __glXReadReply(dpy, 1, residences, GL_TRUE);
+         }
+         UnlockDisplay(dpy);
+         SyncHandle();
+      }
+      return retval;
+   }
 }
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index bb18bee8ea..8300b30144 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -129,14 +129,14 @@ install-headers:
 install-libgl: default gl.pc install-headers
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GL_LIB_GLOB) \
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GL_LIB_GLOB) \
 		$(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 gl.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 install-osmesa: default osmesa.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_GLOB) \
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_GLOB) \
 		$(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 osmesa.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
index 9ffc15eeb7..cad5676320 100644
--- a/src/mesa/SConscript
+++ b/src/mesa/SConscript
@@ -84,6 +84,7 @@ if env['platform'] != 'winddk':
 		'main/shared.c',
 		'main/state.c',
 		'main/stencil.c',
+		'main/syncobj.c',
 		'main/texcompress.c',
 		'main/texcompress_s3tc.c',
 		'main/texcompress_fxt1.c',
@@ -187,20 +188,26 @@ if env['platform'] != 'winddk':
 		'shader/arbprogram.c',
 		'shader/atifragshader.c',
 		'shader/grammar/grammar_mesa.c',
+		'shader/hash_table.c',
+		'shader/lex.yy.c',
 		'shader/nvfragparse.c',
 		'shader/nvprogram.c',
 		'shader/nvvertparse.c',
 		'shader/program.c',
+		'shader/program_parse.tab.c',
+		'shader/program_parse_extra.c',
 		'shader/prog_cache.c',
 		'shader/prog_execute.c',
 		'shader/prog_instruction.c',
 		'shader/prog_noise.c',
 		'shader/prog_optimize.c',
 		'shader/prog_parameter.c',
+		'shader/prog_parameter_layout.c',
 		'shader/prog_print.c',
 		'shader/prog_statevars.c',
 		'shader/prog_uniform.c',
 		'shader/programopt.c',
+		'shader/symbol_table.c',
 		'shader/shader_api.c',
 	]
 	
diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c
index 6a98c29a3d..a9f3c8e727 100644
--- a/src/mesa/drivers/common/driverfuncs.c
+++ b/src/mesa/drivers/common/driverfuncs.c
@@ -45,6 +45,9 @@
 #include "main/fbobject.h"
 #include "main/texrender.h"
 #endif
+#if FEATURE_ARB_sync
+#include "main/syncobj.h"
+#endif
 
 #include "shader/program.h"
 #include "shader/prog_execute.h"
@@ -200,17 +203,17 @@ _mesa_init_driver_functions(struct dd_function_table *driver)
    driver->GetDoublev = NULL;
    driver->GetFloatv = NULL;
    driver->GetIntegerv = NULL;
+   driver->GetInteger64v = NULL;
    driver->GetPointerv = NULL;
    
-#if FEATURE_ARB_vertex_buffer_object
-   driver->NewBufferObject = _mesa_new_buffer_object;
-   driver->DeleteBuffer = _mesa_delete_buffer_object;
-   driver->BindBuffer = NULL;
-   driver->BufferData = _mesa_buffer_data;
-   driver->BufferSubData = _mesa_buffer_subdata;
-   driver->GetBufferSubData = _mesa_buffer_get_subdata;
-   driver->MapBuffer = _mesa_buffer_map;
-   driver->UnmapBuffer = _mesa_buffer_unmap;
+   /* buffer objects */
+   _mesa_init_buffer_object_functions(driver);
+
+   /* query objects */
+   _mesa_init_query_object_functions(driver);
+
+#if FEATURE_ARB_sync
+   _mesa_init_sync_object_functions(driver);
 #endif
 
 #if FEATURE_EXT_framebuffer_object
@@ -225,14 +228,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver)
    driver->BlitFramebuffer = _swrast_BlitFramebuffer;
 #endif
 
-   /* query objects */
-   driver->NewQueryObject = _mesa_new_query_object;
-   driver->DeleteQuery = _mesa_delete_query;
-   driver->BeginQuery = _mesa_begin_query;
-   driver->EndQuery = _mesa_end_query;
-   driver->WaitQuery = _mesa_wait_query;
-   driver->CheckQuery = _mesa_check_query;
-
    /* APPLE_vertex_array_object */
    driver->NewArrayObject = _mesa_new_array_object;
    driver->DeleteArrayObject = _mesa_delete_array_object;
diff --git a/src/mesa/drivers/common/meta.c b/src/mesa/drivers/common/meta.c
new file mode 100644
index 0000000000..47090ba297
--- /dev/null
+++ b/src/mesa/drivers/common/meta.c
@@ -0,0 +1,1650 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.6
+ *
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * Meta operations.  Some GL operations can be expressed in terms of
+ * other GL operations.  For example, glBlitFramebuffer() can be done
+ * with texture mapping and glClear() can be done with polygon rendering.
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/imports.h"
+#include "main/arrayobj.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/depth.h"
+#include "main/enable.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/matrix.h"
+#include "main/polygon.h"
+#include "main/readpix.h"
+#include "main/scissor.h"
+#include "main/shaders.h"
+#include "main/stencil.h"
+#include "main/texobj.h"
+#include "main/texenv.h"
+#include "main/teximage.h"
+#include "main/texparam.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "main/viewport.h"
+#include "shader/program.h"
+#include "shader/arbprogram.h"
+#include "swrast/swrast.h"
+#include "drivers/common/meta.h"
+
+
+/**
+ * State which we may save/restore across meta ops.
+ * XXX this may be incomplete...
+ */
+struct save_state
+{
+   GLbitfield SavedState;  /**< bitmask of META_* flags */
+
+   /** META_ALPHA_TEST */
+   GLboolean AlphaEnabled;
+
+   /** META_BLEND */
+   GLboolean BlendEnabled;
+   GLboolean ColorLogicOpEnabled;
+
+   /** META_COLOR_MASK */
+   GLubyte ColorMask[4];
+
+   /** META_DEPTH_TEST */
+   struct gl_depthbuffer_attrib Depth;
+
+   /** META_FOG */
+   GLboolean Fog;
+
+   /** META_PIXEL_STORE */
+   struct gl_pixelstore_attrib Pack, Unpack;
+
+   /** META_RASTERIZATION */
+   GLenum FrontPolygonMode, BackPolygonMode;
+   GLboolean PolygonOffset;
+   GLboolean PolygonSmooth;
+   GLboolean PolygonStipple;
+   GLboolean PolygonCull;
+
+   /** META_SCISSOR */
+   struct gl_scissor_attrib Scissor;
+
+   /** META_SHADER */
+   GLboolean VertexProgramEnabled;
+   struct gl_vertex_program *VertexProgram;
+   GLboolean FragmentProgramEnabled;
+   struct gl_fragment_program *FragmentProgram;
+   GLuint Shader;
+
+   /** META_STENCIL_TEST */
+   struct gl_stencil_attrib Stencil;
+
+   /** META_TRANSFORM */
+   GLenum MatrixMode;
+   GLfloat ModelviewMatrix[16];
+   GLfloat ProjectionMatrix[16];
+   GLfloat TextureMatrix[16];
+   GLbitfield ClipPlanesEnabled;
+
+   /** META_TEXTURE */
+   GLuint ActiveUnit;
+   GLuint ClientActiveUnit;
+   /** for unit[0] only */
+   struct gl_texture_object *CurrentTexture[NUM_TEXTURE_TARGETS];
+   /** mask of TEXTURE_2D_BIT, etc */
+   GLbitfield TexEnabled[MAX_TEXTURE_UNITS];
+   GLbitfield TexGenEnabled[MAX_TEXTURE_UNITS];
+   GLuint EnvMode;  /* unit[0] only */
+
+   /** META_VERTEX */
+   struct gl_array_object *ArrayObj;
+   struct gl_buffer_object *ArrayBufferObj;
+
+   /** META_VIEWPORT */
+   GLint ViewportX, ViewportY, ViewportW, ViewportH;
+   GLclampd DepthNear, DepthFar;
+
+   /** Miscellaneous (always disabled) */
+   GLboolean Lighting;
+};
+
+
+/**
+ * State for glBlitFramebufer()
+ */
+struct blit_state
+{
+   GLuint ArrayObj;
+   GLuint VBO;
+   GLuint DepthFP;
+};
+
+
+/**
+ * State for glClear()
+ */
+struct clear_state
+{
+   GLuint ArrayObj;
+   GLuint VBO;
+};
+
+
+/**
+ * State for glCopyPixels()
+ */
+struct copypix_state
+{
+   GLuint ArrayObj;
+   GLuint VBO;
+};
+
+
+/**
+ * State for glDrawPixels()
+ */
+struct drawpix_state
+{
+   GLuint ArrayObj;
+   GLuint VBO;
+
+   GLuint StencilFP;  /**< Fragment program for drawing stencil images */
+   GLuint DepthFP;  /**< Fragment program for drawing depth images */
+};
+
+
+/**
+ * Temporary texture used for glBlitFramebuffer, glDrawPixels, etc.
+ * This is currently shared by all the meta ops.  But we could create a
+ * separate one for each of glDrawPixel, glBlitFramebuffer, glCopyPixels, etc.
+ */
+struct temp_texture
+{
+   GLuint TexObj;
+   GLenum Target;         /**< GL_TEXTURE_2D or GL_TEXTURE_RECTANGLE */
+   GLsizei MaxSize;       /**< Max possible texture size */
+   GLboolean NPOT;        /**< Non-power of two size OK? */
+   GLsizei Width, Height; /**< Current texture size */
+   GLenum IntFormat;
+   GLfloat Sright, Ttop;  /**< right, top texcoords */
+};
+
+
+/**
+ * All per-context meta state.
+ */
+struct gl_meta_state
+{
+   struct save_state Save;    /**< state saved during meta-ops */
+
+   struct temp_texture TempTex;
+
+   struct blit_state Blit;    /**< For _mesa_meta_blit_framebuffer() */
+   struct clear_state Clear;  /**< For _mesa_meta_clear() */
+   struct copypix_state CopyPix;  /**< For _mesa_meta_copy_pixels() */
+   struct drawpix_state DrawPix;  /**< For _mesa_meta_draw_pixels() */
+
+   /* other possible meta-ops:
+    * glBitmap()
+    * glGenerateMipmap()
+    */
+};
+
+
+/**
+ * Initialize meta-ops for a context.
+ * To be called once during context creation.
+ */
+void
+_mesa_meta_init(GLcontext *ctx)
+{
+   ASSERT(!ctx->Meta);
+
+   ctx->Meta = CALLOC_STRUCT(gl_meta_state);
+}
+
+
+/**
+ * Free context meta-op state.
+ * To be called once during context destruction.
+ */
+void
+_mesa_meta_free(GLcontext *ctx)
+{
+   struct gl_meta_state *meta = ctx->Meta;
+
+   if (_mesa_get_current_context()) {
+      /* if there's no current context, these textures, buffers, etc should
+       * still get freed by _mesa_free_context_data().
+       */
+
+      _mesa_DeleteTextures(1, &meta->TempTex.TexObj);
+
+      /* glBlitFramebuffer */
+      _mesa_DeleteBuffersARB(1, & meta->Blit.VBO);
+      _mesa_DeleteVertexArraysAPPLE(1, &meta->Blit.ArrayObj);
+      _mesa_DeletePrograms(1, &meta->Blit.DepthFP);
+
+      /* glClear */
+      _mesa_DeleteBuffersARB(1, & meta->Clear.VBO);
+      _mesa_DeleteVertexArraysAPPLE(1, &meta->Clear.ArrayObj);
+
+      /* glCopyPixels */
+      _mesa_DeleteBuffersARB(1, & meta->CopyPix.VBO);
+      _mesa_DeleteVertexArraysAPPLE(1, &meta->CopyPix.ArrayObj);
+
+      /* glDrawPixels */
+      _mesa_DeleteBuffersARB(1, & meta->DrawPix.VBO);
+      _mesa_DeleteVertexArraysAPPLE(1, &meta->DrawPix.ArrayObj);
+      _mesa_DeletePrograms(1, &meta->DrawPix.DepthFP);
+      _mesa_DeletePrograms(1, &meta->DrawPix.StencilFP);
+   }
+
+   _mesa_free(ctx->Meta);
+   ctx->Meta = NULL;
+}
+
+
+/**
+ * Enter meta state.  This is like a light-weight version of glPushAttrib
+ * but it also resets most GL state back to default values.
+ *
+ * \param state  bitmask of META_* flags indicating which attribute groups
+ *               to save and reset to their defaults
+ */
+static void
+_mesa_meta_begin(GLcontext *ctx, GLbitfield state)
+{
+   struct save_state *save = &ctx->Meta->Save;
+
+   save->SavedState = state;
+
+   if (state & META_ALPHA_TEST) {
+      save->AlphaEnabled = ctx->Color.AlphaEnabled;
+      if (ctx->Color.AlphaEnabled)
+         _mesa_Disable(GL_ALPHA_TEST);
+   }
+
+   if (state & META_BLEND) {
+      save->BlendEnabled = ctx->Color.BlendEnabled;
+      if (ctx->Color.BlendEnabled)
+         _mesa_Disable(GL_BLEND);
+      save->ColorLogicOpEnabled = ctx->Color.ColorLogicOpEnabled;
+      if (ctx->Color.ColorLogicOpEnabled)
+         _mesa_Disable(GL_COLOR_LOGIC_OP);
+   }
+
+   if (state & META_COLOR_MASK) {
+      COPY_4V(save->ColorMask, ctx->Color.ColorMask);
+      if (!ctx->Color.ColorMask[0] ||
+          !ctx->Color.ColorMask[1] ||
+          !ctx->Color.ColorMask[2] ||
+          !ctx->Color.ColorMask[3])
+         _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+   }
+
+   if (state & META_DEPTH_TEST) {
+      save->Depth = ctx->Depth; /* struct copy */
+      if (ctx->Depth.Test)
+         _mesa_Disable(GL_DEPTH_TEST);
+   }
+
+   if (state & META_FOG) {
+      save->Fog = ctx->Fog.Enabled;
+      if (ctx->Fog.Enabled)
+         _mesa_set_enable(ctx, GL_FOG, GL_FALSE);
+   }
+
+   if (state & META_PIXEL_STORE) {
+      save->Pack = ctx->Pack;
+      save->Unpack = ctx->Unpack;
+      ctx->Pack = ctx->DefaultPacking;
+      ctx->Unpack = ctx->DefaultPacking;
+   }
+
+   if (state & META_RASTERIZATION) {
+      save->FrontPolygonMode = ctx->Polygon.FrontMode;
+      save->BackPolygonMode = ctx->Polygon.BackMode;
+      save->PolygonOffset = ctx->Polygon.OffsetFill;
+      save->PolygonSmooth = ctx->Polygon.SmoothFlag;
+      save->PolygonStipple = ctx->Polygon.StippleFlag;
+      save->PolygonCull = ctx->Polygon.CullFlag;
+      _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+      _mesa_set_enable(ctx, GL_POLYGON_OFFSET_FILL, GL_FALSE);
+      _mesa_set_enable(ctx, GL_POLYGON_SMOOTH, GL_FALSE);
+      _mesa_set_enable(ctx, GL_POLYGON_STIPPLE, GL_FALSE);
+      _mesa_set_enable(ctx, GL_CULL_FACE, GL_FALSE);
+   }
+
+   if (state & META_SCISSOR) {
+      save->Scissor = ctx->Scissor; /* struct copy */
+   }
+
+   if (state & META_SHADER) {
+      if (ctx->Extensions.ARB_vertex_program) {
+         save->VertexProgramEnabled = ctx->VertexProgram.Enabled;
+         save->VertexProgram = ctx->VertexProgram.Current;
+         _mesa_set_enable(ctx, GL_VERTEX_PROGRAM_ARB, GL_FALSE);
+      }
+
+      if (ctx->Extensions.ARB_fragment_program) {
+         save->FragmentProgramEnabled = ctx->FragmentProgram.Enabled;
+         save->FragmentProgram = ctx->FragmentProgram.Current;
+         _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB, GL_FALSE);
+      }
+
+      if (ctx->Extensions.ARB_shader_objects) {
+         save->Shader = ctx->Shader.CurrentProgram ?
+            ctx->Shader.CurrentProgram->Name : 0;
+         _mesa_UseProgramObjectARB(0);
+      }
+   }
+
+   if (state & META_STENCIL_TEST) {
+      save->Stencil = ctx->Stencil; /* struct copy */
+      if (ctx->Stencil.Enabled)
+         _mesa_Disable(GL_STENCIL_TEST);
+      /* NOTE: other stencil state not reset */
+   }
+
+   if (state & META_TEXTURE) {
+      GLuint u, tgt;
+
+      save->ActiveUnit = ctx->Texture.CurrentUnit;
+      save->ClientActiveUnit = ctx->Array.ActiveTexture;
+      save->EnvMode = ctx->Texture.Unit[0].EnvMode;
+
+      /* Disable all texture units */
+      for (u = 0; u < ctx->Const.MaxTextureUnits; u++) {
+         save->TexEnabled[u] = ctx->Texture.Unit[u].Enabled;
+         save->TexGenEnabled[u] = ctx->Texture.Unit[u].TexGenEnabled;
+         if (ctx->Texture.Unit[u].Enabled ||
+             ctx->Texture.Unit[u].TexGenEnabled) {
+            _mesa_ActiveTextureARB(GL_TEXTURE0 + u);
+            _mesa_set_enable(ctx, GL_TEXTURE_1D, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_2D, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_3D, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_CUBE_MAP, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_RECTANGLE, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_GEN_S, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_GEN_T, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_GEN_R, GL_FALSE);
+            _mesa_set_enable(ctx, GL_TEXTURE_GEN_Q, GL_FALSE);
+         }
+      }
+
+      /* save current texture objects for unit[0] only */
+      for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) {
+         _mesa_reference_texobj(&save->CurrentTexture[tgt],
+                                ctx->Texture.Unit[0].CurrentTex[tgt]);
+      }
+
+      /* set defaults for unit[0] */
+      _mesa_ActiveTextureARB(GL_TEXTURE0);
+      _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+      _mesa_TexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
+   }
+
+   if (state & META_TRANSFORM) {
+      GLuint activeTexture = ctx->Texture.CurrentUnit;
+      _mesa_memcpy(save->ModelviewMatrix, ctx->ModelviewMatrixStack.Top->m,
+                   16 * sizeof(GLfloat));
+      _mesa_memcpy(save->ProjectionMatrix, ctx->ProjectionMatrixStack.Top->m,
+                   16 * sizeof(GLfloat));
+      _mesa_memcpy(save->TextureMatrix, ctx->TextureMatrixStack[0].Top->m,
+                   16 * sizeof(GLfloat));
+      save->MatrixMode = ctx->Transform.MatrixMode;
+      /* set 1:1 vertex:pixel coordinate transform */
+      _mesa_ActiveTextureARB(GL_TEXTURE0);
+      _mesa_MatrixMode(GL_TEXTURE);
+      _mesa_LoadIdentity();
+      _mesa_ActiveTextureARB(GL_TEXTURE0 + activeTexture);
+      _mesa_MatrixMode(GL_MODELVIEW);
+      _mesa_LoadIdentity();
+      _mesa_MatrixMode(GL_PROJECTION);
+      _mesa_LoadIdentity();
+      _mesa_Ortho(0.0F, ctx->DrawBuffer->Width,
+                  0.0F, ctx->DrawBuffer->Height,
+                  -1.0F, 1.0F);
+      save->ClipPlanesEnabled = ctx->Transform.ClipPlanesEnabled;
+      if (ctx->Transform.ClipPlanesEnabled) {
+         GLuint i;
+         for (i = 0; i < ctx->Const.MaxClipPlanes; i++) {
+            _mesa_set_enable(ctx, GL_CLIP_PLANE0 + i, GL_FALSE);
+         }
+      }
+   }
+
+   if (state & META_VERTEX) {
+      /* save vertex array object state */
+      _mesa_reference_array_object(ctx, &save->ArrayObj,
+                                   ctx->Array.ArrayObj);
+      _mesa_reference_buffer_object(ctx, &save->ArrayBufferObj,
+                                    ctx->Array.ArrayBufferObj);
+      /* set some default state? */
+   }
+
+   if (state & META_VIEWPORT) {
+      /* save viewport state */
+      save->ViewportX = ctx->Viewport.X;
+      save->ViewportY = ctx->Viewport.Y;
+      save->ViewportW = ctx->Viewport.Width;
+      save->ViewportH = ctx->Viewport.Height;
+      /* set viewport to match window size */
+      if (ctx->Viewport.X != 0 ||
+          ctx->Viewport.Y != 0 ||
+          ctx->Viewport.Width != ctx->DrawBuffer->Width ||
+          ctx->Viewport.Height != ctx->DrawBuffer->Height) {
+         _mesa_set_viewport(ctx, 0, 0,
+                            ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+      }
+      /* save depth range state */
+      save->DepthNear = ctx->Viewport.Near;
+      save->DepthFar = ctx->Viewport.Far;
+      /* set depth range to default */
+      _mesa_DepthRange(0.0, 1.0);
+   }
+
+   /* misc */
+   {
+      save->Lighting = ctx->Light.Enabled;
+      if (ctx->Light.Enabled)
+         _mesa_set_enable(ctx, GL_LIGHTING, GL_FALSE);
+   }
+}
+
+
+/**
+ * Leave meta state.  This is like a light-weight version of glPopAttrib().
+ */
+static void
+_mesa_meta_end(GLcontext *ctx)
+{
+   struct save_state *save = &ctx->Meta->Save;
+   const GLbitfield state = save->SavedState;
+
+   if (state & META_ALPHA_TEST) {
+      if (ctx->Color.AlphaEnabled != save->AlphaEnabled)
+         _mesa_set_enable(ctx, GL_ALPHA_TEST, save->AlphaEnabled);
+   }
+
+   if (state & META_BLEND) {
+      if (ctx->Color.BlendEnabled != save->BlendEnabled)
+         _mesa_set_enable(ctx, GL_BLEND, save->BlendEnabled);
+      if (ctx->Color.ColorLogicOpEnabled != save->ColorLogicOpEnabled)
+         _mesa_set_enable(ctx, GL_COLOR_LOGIC_OP, save->ColorLogicOpEnabled);
+   }
+
+   if (state & META_COLOR_MASK) {
+      if (!TEST_EQ_4V(ctx->Color.ColorMask, save->ColorMask))
+         _mesa_ColorMask(save->ColorMask[0], save->ColorMask[1],
+                         save->ColorMask[2], save->ColorMask[3]);
+   }
+
+   if (state & META_DEPTH_TEST) {
+      if (ctx->Depth.Test != save->Depth.Test)
+         _mesa_set_enable(ctx, GL_DEPTH_TEST, save->Depth.Test);
+      _mesa_DepthFunc(save->Depth.Func);
+      _mesa_DepthMask(save->Depth.Mask);
+   }
+
+   if (state & META_FOG) {
+      _mesa_set_enable(ctx, GL_FOG, save->Fog);
+   }
+
+   if (state & META_PIXEL_STORE) {
+      ctx->Pack = save->Pack;
+      ctx->Unpack = save->Unpack;
+   }
+
+   if (state & META_RASTERIZATION) {
+      _mesa_PolygonMode(GL_FRONT, save->FrontPolygonMode);
+      _mesa_PolygonMode(GL_BACK, save->BackPolygonMode);
+      _mesa_set_enable(ctx, GL_POLYGON_STIPPLE, save->PolygonStipple);
+      _mesa_set_enable(ctx, GL_POLYGON_OFFSET_FILL, save->PolygonOffset);
+      _mesa_set_enable(ctx, GL_POLYGON_SMOOTH, save->PolygonSmooth);
+      _mesa_set_enable(ctx, GL_CULL_FACE, save->PolygonCull);
+   }
+
+   if (state & META_SCISSOR) {
+      _mesa_set_enable(ctx, GL_SCISSOR_TEST, save->Scissor.Enabled);
+      _mesa_Scissor(save->Scissor.X, save->Scissor.Y,
+                    save->Scissor.Width, save->Scissor.Height);
+   }
+
+   if (state & META_SHADER) {
+      if (ctx->Extensions.ARB_vertex_program) {
+         _mesa_set_enable(ctx, GL_VERTEX_PROGRAM_ARB,
+                          save->VertexProgramEnabled);
+         _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, 
+                                  save->VertexProgram);
+      }
+
+      if (ctx->Extensions.ARB_fragment_program) {
+         _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB,
+                          save->FragmentProgramEnabled);
+         _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+                                  save->FragmentProgram);
+      }
+
+      if (ctx->Extensions.ARB_shader_objects) {
+         _mesa_UseProgramObjectARB(save->Shader);
+      }
+   }
+
+   if (state & META_STENCIL_TEST) {
+      const struct gl_stencil_attrib *stencil = &save->Stencil;
+
+      _mesa_set_enable(ctx, GL_STENCIL_TEST, stencil->Enabled);
+      _mesa_ClearStencil(stencil->Clear);
+      if (ctx->Extensions.EXT_stencil_two_side) {
+         _mesa_set_enable(ctx, GL_STENCIL_TEST_TWO_SIDE_EXT,
+                          stencil->TestTwoSide);
+         _mesa_ActiveStencilFaceEXT(stencil->ActiveFace
+                                    ? GL_BACK : GL_FRONT);
+      }
+      /* front state */
+      _mesa_StencilFuncSeparate(GL_FRONT,
+                                stencil->Function[0],
+                                stencil->Ref[0],
+                                stencil->ValueMask[0]);
+      _mesa_StencilMaskSeparate(GL_FRONT, stencil->WriteMask[0]);
+      _mesa_StencilOpSeparate(GL_FRONT, stencil->FailFunc[0],
+                              stencil->ZFailFunc[0],
+                              stencil->ZPassFunc[0]);
+      /* back state */
+      _mesa_StencilFuncSeparate(GL_BACK,
+                                stencil->Function[1],
+                                stencil->Ref[1],
+                                stencil->ValueMask[1]);
+      _mesa_StencilMaskSeparate(GL_BACK, stencil->WriteMask[1]);
+      _mesa_StencilOpSeparate(GL_BACK, stencil->FailFunc[1],
+                              stencil->ZFailFunc[1],
+                              stencil->ZPassFunc[1]);
+   }
+
+   if (state & META_TEXTURE) {
+      GLuint u, tgt;
+
+      ASSERT(ctx->Texture.CurrentUnit == 0);
+
+      /* restore texenv for unit[0] */
+      _mesa_TexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, save->EnvMode);
+
+      /* restore texture objects for unit[0] only */
+      for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) {
+         _mesa_reference_texobj(&ctx->Texture.Unit[0].CurrentTex[tgt],
+                                save->CurrentTexture[tgt]);
+      }
+
+      /* Re-enable textures, texgen */
+      for (u = 0; u < ctx->Const.MaxTextureUnits; u++) {
+         if (save->TexEnabled[u]) {
+            _mesa_ActiveTextureARB(GL_TEXTURE0 + u);
+
+            if (save->TexEnabled[u] & TEXTURE_1D_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_1D, GL_TRUE);
+            if (save->TexEnabled[u] & TEXTURE_2D_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_2D, GL_TRUE);
+            if (save->TexEnabled[u] & TEXTURE_3D_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_3D, GL_TRUE);
+            if (save->TexEnabled[u] & TEXTURE_CUBE_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_CUBE_MAP, GL_TRUE);
+            if (save->TexEnabled[u] & TEXTURE_RECT_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_RECTANGLE, GL_TRUE);
+         }
+
+         if (save->TexGenEnabled[u]) {
+            _mesa_ActiveTextureARB(GL_TEXTURE0 + u);
+
+            if (save->TexGenEnabled[u] & S_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_GEN_S, GL_TRUE);
+            if (save->TexGenEnabled[u] & T_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_GEN_T, GL_TRUE);
+            if (save->TexGenEnabled[u] & R_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_GEN_R, GL_TRUE);
+            if (save->TexGenEnabled[u] & Q_BIT)
+               _mesa_set_enable(ctx, GL_TEXTURE_GEN_Q, GL_TRUE);
+         }
+      }
+
+      /* restore current unit state */
+      _mesa_ActiveTextureARB(GL_TEXTURE0 + save->ActiveUnit);
+      _mesa_ClientActiveTextureARB(GL_TEXTURE0 + save->ClientActiveUnit);
+   }
+
+   if (state & META_TRANSFORM) {
+      GLuint activeTexture = ctx->Texture.CurrentUnit;
+      _mesa_ActiveTextureARB(GL_TEXTURE0);
+      _mesa_MatrixMode(GL_TEXTURE);
+      _mesa_LoadMatrixf(save->TextureMatrix);
+      _mesa_ActiveTextureARB(GL_TEXTURE0 + activeTexture);
+
+      _mesa_MatrixMode(GL_MODELVIEW);
+      _mesa_LoadMatrixf(save->ModelviewMatrix);
+
+      _mesa_MatrixMode(GL_PROJECTION);
+      _mesa_LoadMatrixf(save->ProjectionMatrix);
+
+      _mesa_MatrixMode(save->MatrixMode);
+
+      save->ClipPlanesEnabled = ctx->Transform.ClipPlanesEnabled;
+      if (save->ClipPlanesEnabled) {
+         GLuint i;
+         for (i = 0; i < ctx->Const.MaxClipPlanes; i++) {
+            if (save->ClipPlanesEnabled & (1 << i)) {
+               _mesa_set_enable(ctx, GL_CLIP_PLANE0 + i, GL_TRUE);
+            }
+         }
+      }
+   }
+
+   if (state & META_VERTEX) {
+      /* restore vertex buffer object */
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, save->ArrayBufferObj->Name);
+      _mesa_reference_buffer_object(ctx, &save->ArrayBufferObj, NULL);
+
+      /* restore vertex array object */
+      _mesa_BindVertexArray(save->ArrayObj->Name);
+      _mesa_reference_array_object(ctx, &save->ArrayObj, NULL);
+   }
+
+   if (state & META_VIEWPORT) {
+      if (save->ViewportX != ctx->Viewport.X ||
+          save->ViewportY != ctx->Viewport.Y ||
+          save->ViewportW != ctx->Viewport.Width ||
+          save->ViewportH != ctx->Viewport.Height) {
+         _mesa_set_viewport(ctx, save->ViewportX, save->ViewportY,
+                            save->ViewportW, save->ViewportH);
+      }
+      _mesa_DepthRange(save->DepthNear, save->DepthFar);
+   }
+
+   /* misc */
+   if (save->Lighting) {
+      _mesa_set_enable(ctx, GL_LIGHTING, GL_TRUE);
+   }
+   if (save->Fog) {
+      _mesa_set_enable(ctx, GL_FOG, GL_TRUE);
+   }
+}
+
+
+/**
+ * Return pointer to temp_texture info.  This does some one-time init
+ * if needed.
+ */
+static struct temp_texture *
+get_temp_texture(GLcontext *ctx)
+{
+   struct temp_texture *tex = &ctx->Meta->TempTex;
+
+   if (!tex->TexObj) {
+      /* do one-time init */
+
+      /* prefer texture rectangle */
+      if (ctx->Extensions.NV_texture_rectangle) {
+         tex->Target = GL_TEXTURE_RECTANGLE;
+         tex->MaxSize = ctx->Const.MaxTextureRectSize;
+         tex->NPOT = GL_TRUE;
+      }
+      else {
+         /* use 2D texture, NPOT if possible */
+         tex->Target = GL_TEXTURE_2D;
+         tex->MaxSize = 1 << (ctx->Const.MaxTextureLevels - 1);
+         tex->NPOT = ctx->Extensions.ARB_texture_non_power_of_two;
+      }
+      assert(tex->MaxSize > 0);
+
+      _mesa_GenTextures(1, &tex->TexObj);
+      _mesa_BindTexture(tex->Target, tex->TexObj);
+   }
+
+   return tex;
+}
+
+
+/**
+ * Compute the width/height of texture needed to draw an image of the
+ * given size.  Return a flag indicating whether the current texture
+ * can be re-used (glTexSubImage2D) or if a new texture needs to be
+ * allocated (glTexImage2D).
+ * Also, compute s/t texcoords for drawing.
+ *
+ * \return GL_TRUE if new texture is needed, GL_FALSE otherwise
+ */
+static GLboolean
+alloc_texture(struct temp_texture *tex,
+              GLsizei width, GLsizei height, GLenum intFormat)
+{
+   GLboolean newTex = GL_FALSE;
+
+   if (width > tex->Width ||
+       height > tex->Height ||
+       intFormat != tex->IntFormat) {
+      /* alloc new texture (larger or different format) */
+
+      if (tex->NPOT) {
+         /* use non-power of two size */
+         tex->Width = width;
+         tex->Height = height;
+      }
+      else {
+         /* find power of two size */
+         GLsizei w, h;
+         w = h = 16;
+         while (w < width)
+            w *= 2;
+         while (h < height)
+            h *= 2;
+         tex->Width = w;
+         tex->Height = h;
+      }
+
+      tex->IntFormat = intFormat;
+
+      newTex = GL_TRUE;
+   }
+
+   /* compute texcoords */
+   if (tex->Target == GL_TEXTURE_RECTANGLE) {
+      tex->Sright = (GLfloat) width;
+      tex->Ttop = (GLfloat) height;
+   }
+   else {
+      tex->Sright = (GLfloat) width / tex->Width;
+      tex->Ttop = (GLfloat) height / tex->Height;
+   }
+
+   return newTex;
+}
+
+
+/**
+ * Setup/load texture for glCopyPixels or glBlitFramebuffer.
+ */
+static void
+setup_copypix_texture(struct temp_texture *tex,
+                      GLboolean newTex,
+                      GLint srcX, GLint srcY,
+                      GLsizei width, GLsizei height, GLenum intFormat,
+                      GLenum filter)
+{
+   _mesa_BindTexture(tex->Target, tex->TexObj);
+   _mesa_TexParameteri(tex->Target, GL_TEXTURE_MIN_FILTER, filter);
+   _mesa_TexParameteri(tex->Target, GL_TEXTURE_MAG_FILTER, filter);
+   _mesa_TexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
+
+   /* copy framebuffer image to texture */
+   if (newTex) {
+      /* create new tex image */
+      if (tex->Width == width && tex->Height == height) {
+         /* create new tex with framebuffer data */
+         _mesa_CopyTexImage2D(tex->Target, 0, tex->IntFormat,
+                              srcX, srcY, width, height, 0);
+      }
+      else {
+         /* create empty texture */
+         _mesa_TexImage2D(tex->Target, 0, tex->IntFormat,
+                          tex->Width, tex->Height, 0,
+                          intFormat, GL_UNSIGNED_BYTE, NULL);
+         /* load image */
+         _mesa_CopyTexSubImage2D(tex->Target, 0,
+                                 0, 0, srcX, srcY, width, height);
+      }
+   }
+   else {
+      /* replace existing tex image */
+      _mesa_CopyTexSubImage2D(tex->Target, 0,
+                              0, 0, srcX, srcY, width, height);
+   }
+}
+
+
+/**
+ * Setup/load texture for glDrawPixels.
+ */
+static void
+setup_drawpix_texture(struct temp_texture *tex,
+                      GLboolean newTex,
+                      GLenum texIntFormat,
+                      GLsizei width, GLsizei height,
+                      GLenum format, GLenum type,
+                      const GLvoid *pixels)
+{
+   _mesa_BindTexture(tex->Target, tex->TexObj);
+   _mesa_TexParameteri(tex->Target, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+   _mesa_TexParameteri(tex->Target, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+   _mesa_TexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
+
+   /* copy pixel data to texture */
+   if (newTex) {
+      /* create new tex image */
+      if (tex->Width == width && tex->Height == height) {
+         /* create new tex and load image data */
+         _mesa_TexImage2D(tex->Target, 0, tex->IntFormat,
+                          tex->Width, tex->Height, 0, format, type, pixels);
+      }
+      else {
+         /* create empty texture */
+         _mesa_TexImage2D(tex->Target, 0, tex->IntFormat,
+                          tex->Width, tex->Height, 0, format, type, NULL);
+         /* load image */
+         _mesa_TexSubImage2D(tex->Target, 0,
+                             0, 0, width, height, format, type, pixels);
+      }
+   }
+   else {
+      /* replace existing tex image */
+      _mesa_TexSubImage2D(tex->Target, 0,
+                          0, 0, width, height, format, type, pixels);
+   }
+}
+
+
+
+/**
+ * One-time init for drawing depth pixels.
+ */
+static void
+init_blit_depth_pixels(GLcontext *ctx)
+{
+   static const char *program =
+      "!!ARBfp1.0\n"
+      "TEX result.depth, fragment.texcoord[0], texture[0], %s; \n"
+      "END \n";
+   char program2[200];
+   struct blit_state *blit = &ctx->Meta->Blit;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   const char *texTarget;
+
+   assert(blit->DepthFP == 0);
+
+   /* replace %s with "RECT" or "2D" */
+   assert(strlen(program) + 4 < sizeof(program2));
+   if (tex->Target == GL_TEXTURE_RECTANGLE)
+      texTarget = "RECT";
+   else
+      texTarget = "2D";
+   _mesa_snprintf(program2, sizeof(program2), program, texTarget);
+
+   _mesa_GenPrograms(1, &blit->DepthFP);
+   _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, blit->DepthFP);
+   _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
+                          strlen(program2), (const GLubyte *) program2);
+}
+
+
+/**
+ * Meta implementation of ctx->Driver.BlitFramebuffer() in terms
+ * of texture mapping and polygon rendering.
+ */
+void
+_mesa_meta_blit_framebuffer(GLcontext *ctx,
+                            GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                            GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                            GLbitfield mask, GLenum filter)
+{
+   struct blit_state *blit = &ctx->Meta->Blit;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   const GLsizei maxTexSize = tex->MaxSize;
+   const GLint srcX = MIN2(srcX0, srcX1);
+   const GLint srcY = MIN2(srcY0, srcY1);
+   const GLint srcW = abs(srcX1 - srcX0);
+   const GLint srcH = abs(srcY1 - srcY0);
+   const GLboolean srcFlipX = srcX1 < srcX0;
+   const GLboolean srcFlipY = srcY1 < srcY0;
+   GLfloat verts[4][4]; /* four verts of X,Y,S,T */
+   GLboolean newTex;
+
+   if (srcW > maxTexSize || srcH > maxTexSize) {
+      /* XXX avoid this fallback */
+      _swrast_BlitFramebuffer(ctx, srcX0, srcY0, srcX1, srcY1,
+                              dstX0, dstY0, dstX1, dstY1, mask, filter);
+      return;
+   }
+
+   if (srcFlipX) {
+      GLint tmp = dstX0;
+      dstX0 = dstX1;
+      dstX1 = tmp;
+   }
+
+   if (srcFlipY) {
+      GLint tmp = dstY0;
+      dstY0 = dstY1;
+      dstY1 = tmp;
+   }
+
+   /* only scissor effects blit so save/clear all other relevant state */
+   _mesa_meta_begin(ctx, ~META_SCISSOR);
+
+   if (blit->ArrayObj == 0) {
+      /* one-time setup */
+
+      /* create vertex array object */
+      _mesa_GenVertexArrays(1, &blit->ArrayObj);
+      _mesa_BindVertexArray(blit->ArrayObj);
+
+      /* create vertex array buffer */
+      _mesa_GenBuffersARB(1, &blit->VBO);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, blit->VBO);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts),
+                          NULL, GL_DYNAMIC_DRAW_ARB);
+
+      /* setup vertex arrays */
+      _mesa_VertexPointer(2, GL_FLOAT, sizeof(verts[0]),
+                          (void *) (0 * sizeof(GLfloat)));
+      _mesa_TexCoordPointer(2, GL_FLOAT, sizeof(verts[0]),
+                            (void *) (2 * sizeof(GLfloat)));
+      _mesa_EnableClientState(GL_VERTEX_ARRAY);
+      _mesa_EnableClientState(GL_TEXTURE_COORD_ARRAY);
+   }
+   else {
+      _mesa_BindVertexArray(blit->ArrayObj);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, blit->VBO);
+   }
+
+   newTex = alloc_texture(tex, srcW, srcH, GL_RGBA);
+
+   /* vertex positions/texcoords (after texture allocation!) */
+   {
+      verts[0][0] = (GLfloat) dstX0;
+      verts[0][1] = (GLfloat) dstY0;
+      verts[1][0] = (GLfloat) dstX1;
+      verts[1][1] = (GLfloat) dstY0;
+      verts[2][0] = (GLfloat) dstX1;
+      verts[2][1] = (GLfloat) dstY1;
+      verts[3][0] = (GLfloat) dstX0;
+      verts[3][1] = (GLfloat) dstY1;
+
+      verts[0][2] = 0.0F;
+      verts[0][3] = 0.0F;
+      verts[1][2] = tex->Sright;
+      verts[1][3] = 0.0F;
+      verts[2][2] = tex->Sright;
+      verts[2][3] = tex->Ttop;
+      verts[3][2] = 0.0F;
+      verts[3][3] = tex->Ttop;
+
+      /* upload new vertex data */
+      _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
+   }
+
+   _mesa_Enable(tex->Target);
+
+   if (mask & GL_COLOR_BUFFER_BIT) {
+      setup_copypix_texture(tex, newTex, srcX, srcY, srcW, srcH,
+                            GL_RGBA, filter);
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+      mask &= ~GL_COLOR_BUFFER_BIT;
+   }
+
+   if (mask & GL_DEPTH_BUFFER_BIT) {
+      GLuint *tmp = (GLuint *) _mesa_malloc(srcW * srcH * sizeof(GLuint));
+      if (tmp) {
+         if (!blit->DepthFP)
+            init_blit_depth_pixels(ctx);
+
+         /* maybe change tex format here */
+         newTex = alloc_texture(tex, srcW, srcH, GL_DEPTH_COMPONENT);
+
+         _mesa_ReadPixels(srcX, srcY, srcW, srcH,
+                          GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, tmp);
+
+         setup_drawpix_texture(tex, newTex, GL_DEPTH_COMPONENT, srcW, srcH,
+                               GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, tmp);
+
+         _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, blit->DepthFP);
+         _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB, GL_TRUE);
+         _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+         _mesa_set_enable(ctx, GL_DEPTH_TEST, GL_TRUE);
+         _mesa_DepthFunc(GL_ALWAYS);
+         _mesa_DepthMask(GL_TRUE);
+
+         _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+         mask &= ~GL_DEPTH_BUFFER_BIT;
+
+         _mesa_free(tmp);
+      }
+   }
+
+   if (mask & GL_STENCIL_BUFFER_BIT) {
+      /* XXX can't easily do stencil */
+   }
+
+   _mesa_Disable(tex->Target);
+
+   _mesa_meta_end(ctx);
+
+   if (mask) {
+      _swrast_BlitFramebuffer(ctx, srcX0, srcY0, srcX1, srcY1,
+                              dstX0, dstY0, dstX1, dstY1, mask, filter);
+   }
+}
+
+
+/**
+ * Meta implementation of ctx->Driver.Clear() in terms of polygon rendering.
+ */
+void
+_mesa_meta_clear(GLcontext *ctx, GLbitfield buffers)
+{
+   struct clear_state *clear = &ctx->Meta->Clear;
+   GLfloat verts[4][7]; /* four verts of X,Y,Z,R,G,B,A */
+   /* save all state but scissor, pixel pack/unpack */
+   GLbitfield metaSave = META_ALL - META_SCISSOR - META_PIXEL_STORE;
+
+   if (buffers & BUFFER_BITS_COLOR) {
+      /* if clearing color buffers, don't save/restore colormask */
+      metaSave -= META_COLOR_MASK;
+   }
+
+   _mesa_meta_begin(ctx, metaSave);
+
+   if (clear->ArrayObj == 0) {
+      /* one-time setup */
+
+      /* create vertex array object */
+      _mesa_GenVertexArrays(1, &clear->ArrayObj);
+      _mesa_BindVertexArray(clear->ArrayObj);
+
+      /* create vertex array buffer */
+      _mesa_GenBuffersARB(1, &clear->VBO);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, clear->VBO);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts),
+                          NULL, GL_DYNAMIC_DRAW_ARB);
+
+      /* setup vertex arrays */
+      _mesa_VertexPointer(3, GL_FLOAT, sizeof(verts[0]),
+                          (void *) (0 * sizeof(GLfloat)));
+      _mesa_ColorPointer(4, GL_FLOAT, sizeof(verts[0]),
+                         (void *) (3 * sizeof(GLfloat)));
+      _mesa_EnableClientState(GL_VERTEX_ARRAY);
+      _mesa_EnableClientState(GL_COLOR_ARRAY);
+   }
+   else {
+      _mesa_BindVertexArray(clear->ArrayObj);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, clear->VBO);
+   }
+
+   /* GL_COLOR_BUFFER_BIT */
+   if (buffers & BUFFER_BITS_COLOR) {
+      /* leave colormask, glDrawBuffer state as-is */
+   }
+   else {
+      ASSERT(metaSave & META_COLOR_MASK);
+      _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+   }
+
+   /* GL_DEPTH_BUFFER_BIT */
+   if (buffers & BUFFER_BIT_DEPTH) {
+      _mesa_set_enable(ctx, GL_DEPTH_TEST, GL_TRUE);
+      _mesa_DepthFunc(GL_ALWAYS);
+      _mesa_DepthMask(GL_TRUE);
+   }
+   else {
+      assert(!ctx->Depth.Test);
+   }
+
+   /* GL_STENCIL_BUFFER_BIT */
+   if (buffers & BUFFER_BIT_STENCIL) {
+      _mesa_set_enable(ctx, GL_STENCIL_TEST, GL_TRUE);
+      _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
+                              GL_REPLACE, GL_REPLACE, GL_REPLACE);
+      _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
+                                ctx->Stencil.Clear & 0x7fffffff,
+                                ctx->Stencil.WriteMask[0]);
+   }
+   else {
+      assert(!ctx->Stencil.Enabled);
+   }
+
+   /* vertex positions/colors */
+   {
+      const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin;
+      const GLfloat y0 = (GLfloat) ctx->DrawBuffer->_Ymin;
+      const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax;
+      const GLfloat y1 = (GLfloat) ctx->DrawBuffer->_Ymax;
+      const GLfloat z = 1.0 - 2.0 * ctx->Depth.Clear;
+      GLuint i;
+
+      verts[0][0] = x0;
+      verts[0][1] = y0;
+      verts[0][2] = z;
+      verts[1][0] = x1;
+      verts[1][1] = y0;
+      verts[1][2] = z;
+      verts[2][0] = x1;
+      verts[2][1] = y1;
+      verts[2][2] = z;
+      verts[3][0] = x0;
+      verts[3][1] = y1;
+      verts[3][2] = z;
+
+      /* vertex colors */
+      for (i = 0; i < 4; i++) {
+         COPY_4FV(&verts[i][3], ctx->Color.ClearColor);
+      }
+
+      /* upload new vertex data */
+      _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
+   }
+
+   /* draw quad */
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   _mesa_meta_end(ctx);
+}
+
+
+/**
+ * Meta implementation of ctx->Driver.CopyPixels() in terms
+ * of texture mapping and polygon rendering.
+ */
+void
+_mesa_meta_copy_pixels(GLcontext *ctx, GLint srcX, GLint srcY,
+                       GLsizei width, GLsizei height,
+                       GLint dstX, GLint dstY, GLenum type)
+{
+   struct copypix_state *copypix = &ctx->Meta->CopyPix;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   GLfloat verts[4][5]; /* four verts of X,Y,Z,S,T */
+   GLboolean newTex;
+   GLenum intFormat = GL_RGBA;
+
+   if (type != GL_COLOR ||
+       ctx->_ImageTransferState ||
+       ctx->Fog.Enabled ||
+       width > tex->MaxSize ||
+       height > tex->MaxSize) {
+      /* XXX avoid this fallback */
+      _swrast_CopyPixels(ctx, srcX, srcY, width, height, dstX, dstY, type);
+      return;
+   }
+
+   /* Most GL state applies to glCopyPixels, but a there's a few things
+    * we need to override:
+    */
+   _mesa_meta_begin(ctx, (META_RASTERIZATION |
+                          META_SHADER |
+                          META_TEXTURE |
+                          META_TRANSFORM |
+                          META_VERTEX |
+                          META_VIEWPORT));
+
+   if (copypix->ArrayObj == 0) {
+      /* one-time setup */
+
+      /* create vertex array object */
+      _mesa_GenVertexArrays(1, &copypix->ArrayObj);
+      _mesa_BindVertexArray(copypix->ArrayObj);
+
+      /* create vertex array buffer */
+      _mesa_GenBuffersARB(1, &copypix->VBO);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, copypix->VBO);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts),
+                          NULL, GL_DYNAMIC_DRAW_ARB);
+
+      /* setup vertex arrays */
+      _mesa_VertexPointer(3, GL_FLOAT, sizeof(verts[0]),
+                          (void *) (0 * sizeof(GLfloat)));
+      _mesa_TexCoordPointer(2, GL_FLOAT, sizeof(verts[0]),
+                            (void *) (3 * sizeof(GLfloat)));
+      _mesa_EnableClientState(GL_VERTEX_ARRAY);
+      _mesa_EnableClientState(GL_TEXTURE_COORD_ARRAY);
+   }
+   else {
+      _mesa_BindVertexArray(copypix->ArrayObj);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, copypix->VBO);
+   }
+
+   newTex = alloc_texture(tex, width, height, intFormat);
+
+   /* vertex positions, texcoords (after texture allocation!) */
+   {
+      const GLfloat dstX0 = (GLfloat) dstX;
+      const GLfloat dstY0 = (GLfloat) dstY;
+      const GLfloat dstX1 = dstX + width * ctx->Pixel.ZoomX;
+      const GLfloat dstY1 = dstY + height * ctx->Pixel.ZoomY;
+      const GLfloat z = ctx->Current.RasterPos[2];
+
+      verts[0][0] = dstX0;
+      verts[0][1] = dstY0;
+      verts[0][2] = z;
+      verts[0][3] = 0.0F;
+      verts[0][4] = 0.0F;
+      verts[1][0] = dstX1;
+      verts[1][1] = dstY0;
+      verts[1][2] = z;
+      verts[1][3] = tex->Sright;
+      verts[1][4] = 0.0F;
+      verts[2][0] = dstX1;
+      verts[2][1] = dstY1;
+      verts[2][2] = z;
+      verts[2][3] = tex->Sright;
+      verts[2][4] = tex->Ttop;
+      verts[3][0] = dstX0;
+      verts[3][1] = dstY1;
+      verts[3][2] = z;
+      verts[3][3] = 0.0F;
+      verts[3][4] = tex->Ttop;
+
+      /* upload new vertex data */
+      _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
+   }
+
+   /* Alloc/setup texture */
+   setup_copypix_texture(tex, newTex, srcX, srcY, width, height,
+                         GL_RGBA, GL_NEAREST);
+
+   _mesa_Enable(tex->Target);
+
+   /* draw textured quad */
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   _mesa_Disable(tex->Target);
+
+   _mesa_meta_end(ctx);
+}
+
+
+
+/**
+ * When the glDrawPixels() image size is greater than the max rectangle
+ * texture size we use this function to break the glDrawPixels() image
+ * into tiles which fit into the max texture size.
+ */
+static void
+tiled_draw_pixels(GLcontext *ctx,
+                  GLint tileSize,
+                  GLint x, GLint y, GLsizei width, GLsizei height,
+                  GLenum format, GLenum type,
+                  const struct gl_pixelstore_attrib *unpack,
+                  const GLvoid *pixels)
+{
+   struct gl_pixelstore_attrib tileUnpack = *unpack;
+   GLint i, j;
+
+   if (tileUnpack.RowLength == 0)
+      tileUnpack.RowLength = width;
+
+   for (i = 0; i < width; i += tileSize) {
+      const GLint tileWidth = MIN2(tileSize, width - i);
+      const GLint tileX = (GLint) (x + i * ctx->Pixel.ZoomX);
+
+      tileUnpack.SkipPixels = unpack->SkipPixels + i;
+
+      for (j = 0; j < height; j += tileSize) {
+         const GLint tileHeight = MIN2(tileSize, height - j);
+         const GLint tileY = (GLint) (y + j * ctx->Pixel.ZoomY);
+
+         tileUnpack.SkipRows = unpack->SkipRows + j;
+
+         _mesa_meta_draw_pixels(ctx, tileX, tileY,
+                                tileWidth, tileHeight,
+                                format, type, &tileUnpack, pixels);
+      }
+   }
+}
+
+
+/**
+ * One-time init for drawing stencil pixels.
+ */
+static void
+init_draw_stencil_pixels(GLcontext *ctx)
+{
+   /* This program is run eight times, once for each stencil bit.
+    * The stencil values to draw are found in an 8-bit alpha texture.
+    * We read the texture/stencil value and test if bit 'b' is set.
+    * If the bit is not set, use KIL to kill the fragment.
+    * Finally, we use the stencil test to update the stencil buffer.
+    *
+    * The basic algorithm for checking if a bit is set is:
+    *   if (is_odd(value / (1 << bit)))
+    *      result is one (or non-zero).
+    *   else
+    *      result is zero.
+    * The program parameter contains three values:
+    *   parm.x = 255 / (1 << bit)
+    *   parm.y = 0.5
+    *   parm.z = 0.0
+    */
+   static const char *program =
+      "!!ARBfp1.0\n"
+      "PARAM parm = program.local[0]; \n"
+      "TEMP t; \n"
+      "TEX t, fragment.texcoord[0], texture[0], %s; \n"   /* NOTE %s here! */
+      "# t = t * 255 / bit \n"
+      "MUL t.x, t.a, parm.x; \n"
+      "# t = (int) t \n"
+      "FRC t.y, t.x; \n"
+      "SUB t.x, t.x, t.y; \n"
+      "# t = t * 0.5 \n"
+      "MUL t.x, t.x, parm.y; \n"
+      "# t = fract(t.x) \n"
+      "FRC t.x, t.x; # if t.x != 0, then the bit is set \n"
+      "# t.x = (t.x == 0 ? 1 : 0) \n"
+      "SGE t.x, -t.x, parm.z; \n"
+      "KIL -t.x; \n"
+      "# for debug only \n"
+      "#MOV result.color, t.x; \n"
+      "END \n";
+   char program2[1000];
+   struct drawpix_state *drawpix = &ctx->Meta->DrawPix;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   const char *texTarget;
+
+   assert(drawpix->StencilFP == 0);
+
+   /* replace %s with "RECT" or "2D" */
+   assert(strlen(program) + 4 < sizeof(program2));
+   if (tex->Target == GL_TEXTURE_RECTANGLE)
+      texTarget = "RECT";
+   else
+      texTarget = "2D";
+   _mesa_snprintf(program2, sizeof(program2), program, texTarget);
+
+   _mesa_GenPrograms(1, &drawpix->StencilFP);
+   _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, drawpix->StencilFP);
+   _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
+                          strlen(program2), (const GLubyte *) program2);
+}
+
+
+/**
+ * One-time init for drawing depth pixels.
+ */
+static void
+init_draw_depth_pixels(GLcontext *ctx)
+{
+   static const char *program =
+      "!!ARBfp1.0\n"
+      "PARAM color = program.local[0]; \n"
+      "TEX result.depth, fragment.texcoord[0], texture[0], %s; \n"
+      "MOV result.color, color; \n"
+      "END \n";
+   char program2[200];
+   struct drawpix_state *drawpix = &ctx->Meta->DrawPix;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   const char *texTarget;
+
+   assert(drawpix->DepthFP == 0);
+
+   /* replace %s with "RECT" or "2D" */
+   assert(strlen(program) + 4 < sizeof(program2));
+   if (tex->Target == GL_TEXTURE_RECTANGLE)
+      texTarget = "RECT";
+   else
+      texTarget = "2D";
+   _mesa_snprintf(program2, sizeof(program2), program, texTarget);
+
+   _mesa_GenPrograms(1, &drawpix->DepthFP);
+   _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, drawpix->DepthFP);
+   _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
+                          strlen(program2), (const GLubyte *) program2);
+}
+
+
+/**
+ * Meta implementation of ctx->Driver.DrawPixels() in terms
+ * of texture mapping and polygon rendering.
+ */
+void
+_mesa_meta_draw_pixels(GLcontext *ctx,
+		       GLint x, GLint y, GLsizei width, GLsizei height,
+		       GLenum format, GLenum type,
+		       const struct gl_pixelstore_attrib *unpack,
+		       const GLvoid *pixels)
+{
+   struct drawpix_state *drawpix = &ctx->Meta->DrawPix;
+   struct temp_texture *tex = get_temp_texture(ctx);
+   const struct gl_pixelstore_attrib unpackSave = ctx->Unpack;
+   const GLuint origStencilMask = ctx->Stencil.WriteMask[0];
+   GLfloat verts[4][5]; /* four verts of X,Y,Z,S,T */
+   GLenum texIntFormat;
+   GLboolean fallback, newTex;
+   GLbitfield metaExtraSave = 0x0;
+
+   /*
+    * Determine if we can do the glDrawPixels with texture mapping.
+    */
+   fallback = GL_FALSE;
+   if (ctx->_ImageTransferState ||
+       ctx->Fog.Enabled) {
+      fallback = GL_TRUE;
+   }
+
+   if (_mesa_is_color_format(format)) {
+      /* use more compact format when possible */
+      /* XXX disable special case for GL_LUMINANCE for now to work around
+       * apparent i965 driver bug (see bug #23670).
+       */
+      if (/*format == GL_LUMINANCE ||*/ format == GL_LUMINANCE_ALPHA)
+         texIntFormat = format;
+      else
+         texIntFormat = GL_RGBA;
+   }
+   else if (_mesa_is_stencil_format(format)) {
+      if (ctx->Extensions.ARB_fragment_program &&
+          ctx->Pixel.IndexShift == 0 &&
+          ctx->Pixel.IndexOffset == 0 &&
+          type == GL_UNSIGNED_BYTE) {
+         /* We'll store stencil as alpha.  This only works for GLubyte
+          * image data because of how incoming values are mapped to alpha
+          * in [0,1].
+          */
+         texIntFormat = GL_ALPHA;
+         metaExtraSave = (META_COLOR_MASK |
+                          META_DEPTH_TEST |
+                          META_SHADER |
+                          META_STENCIL_TEST);
+      }
+      else {
+         fallback = GL_TRUE;
+      }
+   }
+   else if (_mesa_is_depth_format(format)) {
+      if (ctx->Extensions.ARB_depth_texture &&
+          ctx->Extensions.ARB_fragment_program) {
+         texIntFormat = GL_DEPTH_COMPONENT;
+         metaExtraSave = (META_SHADER);
+      }
+      else {
+         fallback = GL_TRUE;
+      }
+   }
+   else {
+      fallback = GL_TRUE;
+   }
+
+   if (fallback) {
+      _swrast_DrawPixels(ctx, x, y, width, height,
+                         format, type, unpack, pixels);
+      return;
+   }
+
+   /*
+    * Check image size against max texture size, draw as tiles if needed.
+    */
+   if (width > tex->MaxSize || height > tex->MaxSize) {
+      tiled_draw_pixels(ctx, tex->MaxSize, x, y, width, height,
+                        format, type, unpack, pixels);
+      return;
+   }
+
+   /* Most GL state applies to glDrawPixels (like blending, stencil, etc),
+    * but a there's a few things we need to override:
+    */
+   _mesa_meta_begin(ctx, (META_RASTERIZATION |
+                          META_SHADER |
+                          META_TEXTURE |
+                          META_TRANSFORM |
+                          META_VERTEX |
+                          META_VIEWPORT |
+                          metaExtraSave));
+
+   if (drawpix->ArrayObj == 0) {
+      /* one-time setup */
+
+      /* create vertex array object */
+      _mesa_GenVertexArrays(1, &drawpix->ArrayObj);
+      _mesa_BindVertexArray(drawpix->ArrayObj);
+
+      /* create vertex array buffer */
+      _mesa_GenBuffersARB(1, &drawpix->VBO);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, drawpix->VBO);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts),
+                          NULL, GL_DYNAMIC_DRAW_ARB);
+
+      /* setup vertex arrays */
+      _mesa_VertexPointer(3, GL_FLOAT, sizeof(verts[0]),
+                          (void *) (0 * sizeof(GLfloat)));
+      _mesa_TexCoordPointer(2, GL_FLOAT, sizeof(verts[0]),
+                            (void *) (3 * sizeof(GLfloat)));
+      _mesa_EnableClientState(GL_VERTEX_ARRAY);
+      _mesa_EnableClientState(GL_TEXTURE_COORD_ARRAY);
+   }
+   else {
+      _mesa_BindVertexArray(drawpix->ArrayObj);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, drawpix->VBO);
+   }
+
+   newTex = alloc_texture(tex, width, height, texIntFormat);
+
+   /* vertex positions, texcoords (after texture allocation!) */
+   {
+      const GLfloat x0 = (GLfloat) x;
+      const GLfloat y0 = (GLfloat) y;
+      const GLfloat x1 = x + width * ctx->Pixel.ZoomX;
+      const GLfloat y1 = y + height * ctx->Pixel.ZoomY;
+      const GLfloat z = ctx->Current.RasterPos[2];
+
+      verts[0][0] = x0;
+      verts[0][1] = y0;
+      verts[0][2] = z;
+      verts[0][3] = 0.0F;
+      verts[0][4] = 0.0F;
+      verts[1][0] = x1;
+      verts[1][1] = y0;
+      verts[1][2] = z;
+      verts[1][3] = tex->Sright;
+      verts[1][4] = 0.0F;
+      verts[2][0] = x1;
+      verts[2][1] = y1;
+      verts[2][2] = z;
+      verts[2][3] = tex->Sright;
+      verts[2][4] = tex->Ttop;
+      verts[3][0] = x0;
+      verts[3][1] = y1;
+      verts[3][2] = z;
+      verts[3][3] = 0.0F;
+      verts[3][4] = tex->Ttop;
+
+      /* upload new vertex data */
+      _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
+   }
+
+   /* set given unpack params */
+   ctx->Unpack = *unpack;
+
+   _mesa_Enable(tex->Target);
+
+   if (_mesa_is_stencil_format(format)) {
+      /* Drawing stencil */
+      GLint bit;
+
+      if (!drawpix->StencilFP)
+         init_draw_stencil_pixels(ctx);
+
+      setup_drawpix_texture(tex, newTex, texIntFormat, width, height,
+                            GL_ALPHA, type, pixels);
+
+      _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+      _mesa_set_enable(ctx, GL_STENCIL_TEST, GL_TRUE);
+
+      /* set all stencil bits to 0 */
+      _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
+      _mesa_StencilFunc(GL_ALWAYS, 0, 255);
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+  
+      /* set stencil bits to 1 where needed */
+      _mesa_StencilOp(GL_KEEP, GL_KEEP, GL_REPLACE);
+
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, drawpix->StencilFP);
+      _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB, GL_TRUE);
+
+      for (bit = 0; bit < ctx->Visual.stencilBits; bit++) {
+         const GLuint mask = 1 << bit;
+         if (mask & origStencilMask) {
+            _mesa_StencilFunc(GL_ALWAYS, mask, mask);
+            _mesa_StencilMask(mask);
+
+            _mesa_ProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 0,
+                                             255.0 / mask, 0.5, 0.0, 0.0);
+
+            _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+         }
+      }
+   }
+   else if (_mesa_is_depth_format(format)) {
+      /* Drawing depth */
+      if (!drawpix->DepthFP)
+         init_draw_depth_pixels(ctx);
+
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, drawpix->DepthFP);
+      _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB, GL_TRUE);
+
+      /* polygon color = current raster color */
+      _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
+                                        ctx->Current.RasterColor);
+
+      setup_drawpix_texture(tex, newTex, texIntFormat, width, height,
+                            format, type, pixels);
+
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+   }
+   else {
+      /* Drawing RGBA */
+      setup_drawpix_texture(tex, newTex, texIntFormat, width, height,
+                            format, type, pixels);
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+   }
+
+   _mesa_Disable(tex->Target);
+
+   /* restore unpack params */
+   ctx->Unpack = unpackSave;
+
+   _mesa_meta_end(ctx);
+}
diff --git a/src/mesa/drivers/common/meta.h b/src/mesa/drivers/common/meta.h
new file mode 100644
index 0000000000..b03b64c48a
--- /dev/null
+++ b/src/mesa/drivers/common/meta.h
@@ -0,0 +1,81 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.6
+ *
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef META_H
+#define META_H
+
+
+/**
+ * Flags passed to _mesa_meta_begin().
+ * XXX these flags may evolve...
+ */
+/*@{*/
+#define META_ALPHA_TEST      0x1
+#define META_BLEND           0x2  /**< includes logicop */
+#define META_COLOR_MASK      0x4
+#define META_DEPTH_TEST      0x8
+#define META_FOG            0x10
+#define META_RASTERIZATION  0x20
+#define META_SCISSOR        0x40
+#define META_SHADER         0x80
+#define META_STENCIL_TEST  0x100
+#define META_TRANSFORM     0x200 /**< modelview, projection */
+#define META_TEXTURE       0x400
+#define META_VERTEX        0x800
+#define META_VIEWPORT     0x1000
+#define META_PIXEL_STORE  0x2000
+#define META_ALL            ~0x0
+/*@}*/
+
+
+extern void
+_mesa_meta_init(GLcontext *ctx);
+
+extern void
+_mesa_meta_free(GLcontext *ctx);
+
+extern void
+_mesa_meta_blit_framebuffer(GLcontext *ctx,
+                            GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                            GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                            GLbitfield mask, GLenum filter);
+
+extern void
+_mesa_meta_clear(GLcontext *ctx, GLbitfield buffers);
+
+extern void
+_mesa_meta_copy_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
+                       GLsizei width, GLsizei height,
+                       GLint dstx, GLint dsty, GLenum type);
+
+extern void
+_mesa_meta_draw_pixels(GLcontext *ctx,
+		       GLint x, GLint y, GLsizei width, GLsizei height,
+		       GLenum format, GLenum type,
+		       const struct gl_pixelstore_attrib *unpack,
+		       const GLvoid *pixels);
+
+
+#endif /* META_H */
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index 5c01d233c1..18dbeba24a 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -11,7 +11,8 @@ COMMON_GALLIUM_SOURCES = \
 COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
         ../../common/driverfuncs.c \
         ../common/texmem.c \
-        ../common/drirenderbuffer.c
+        ../common/drirenderbuffer.c \
+	../common/dri_metaops.c
 
 ifeq ($(WINDOW_SYSTEM),dri)
 WINOBJ=
@@ -92,7 +93,7 @@ clean:
 
 install: $(LIBNAME)
 	$(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
 
 
 -include depend
diff --git a/src/mesa/drivers/dri/common/.gitignore b/src/mesa/drivers/dri/common/.gitignore
new file mode 100644
index 0000000000..1edeb79fd1
--- /dev/null
+++ b/src/mesa/drivers/dri/common/.gitignore
@@ -0,0 +1 @@
+*.os
diff --git a/src/mesa/drivers/dri/common/dri_metaops.c b/src/mesa/drivers/dri/common/dri_metaops.c
new file mode 100644
index 0000000000..c7bea07dc9
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.c
@@ -0,0 +1,298 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "main/arrayobj.h"
+#include "main/attrib.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/enable.h"
+#include "main/matrix.h"
+#include "main/macros.h"
+#include "main/polygon.h"
+#include "main/shaders.h"
+#include "main/stencil.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "main/viewport.h"
+#include "shader/arbprogram.h"
+#include "shader/program.h"
+#include "dri_metaops.h"
+
+void
+meta_set_passthrough_transform(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   meta->saved_vp_x = ctx->Viewport.X;
+   meta->saved_vp_y = ctx->Viewport.Y;
+   meta->saved_vp_width = ctx->Viewport.Width;
+   meta->saved_vp_height = ctx->Viewport.Height;
+   meta->saved_matrix_mode = ctx->Transform.MatrixMode;
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+   meta->internal_viewport_call = GL_FALSE;
+
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+}
+
+void
+meta_restore_transform(struct dri_metaops *meta)
+{
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PopMatrix();
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PopMatrix();
+
+   _mesa_MatrixMode(meta->saved_matrix_mode);
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(meta->saved_vp_x, meta->saved_vp_y,
+		  meta->saved_vp_width, meta->saved_vp_height);
+   meta->internal_viewport_call = GL_FALSE;
+}
+
+
+/**
+ * Set up a vertex program to pass through the position and first texcoord
+ * for pixel path.
+ */
+void
+meta_set_passthrough_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   static const char *vp =
+      "!!ARBvp1.0\n"
+      "TEMP vertexClip;\n"
+      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
+      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
+      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
+      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
+      "MOV result.position, vertexClip;\n"
+      "MOV result.texcoord[0], vertex.texcoord[0];\n"
+      "MOV result.color, vertex.color;\n"
+      "END\n";
+
+   assert(meta->saved_vp == NULL);
+
+   _mesa_reference_vertprog(ctx, &meta->saved_vp,
+			    ctx->VertexProgram.Current);
+   if (meta->passthrough_vp == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(vp), (const GLubyte *)vp);
+      _mesa_reference_vertprog(ctx, &meta->passthrough_vp,
+			       ctx->VertexProgram.Current);
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->passthrough_vp);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &meta->passthrough_vp->Base);
+
+   meta->saved_vp_enable = ctx->VertexProgram.Enabled;
+   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous vertex program after
+ * meta_set_passthrough_vertex_program()
+ */
+void
+meta_restore_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->saved_vp);
+   _mesa_reference_vertprog(ctx, &meta->saved_vp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &ctx->VertexProgram.Current->Base);
+
+   if (!meta->saved_vp_enable)
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
+ * program object.
+ */
+void
+meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string)
+{
+   GLcontext *ctx = meta->ctx;
+   assert(meta->saved_fp == NULL);
+
+   _mesa_reference_fragprog(ctx, &meta->saved_fp,
+			    ctx->FragmentProgram.Current);
+   if (*prog == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(prog_string), (const GLubyte *)prog_string);
+      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
+      /* Note that DeletePrograms unbinds the program on us */
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
+
+   meta->saved_fp_enable = ctx->FragmentProgram.Enabled;
+   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous fragment program after
+ * meta_set_fragment_program()
+ */
+void
+meta_restore_fragment_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+			    meta->saved_fp);
+   _mesa_reference_fragprog(ctx, &meta->saved_fp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
+			   &ctx->FragmentProgram.Current->Base);
+
+   if (!meta->saved_fp_enable)
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+static const float default_texcoords[4][2] = { { 0.0, 0.0 },
+					       { 1.0, 0.0 },
+					       { 1.0, 1.0 },
+					       { 0.0, 1.0 } };
+
+void
+meta_set_default_texrect(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   struct gl_client_array *old_texcoord_array;
+
+   meta->saved_active_texture = ctx->Texture.CurrentUnit;
+   if (meta->saved_array_vbo == NULL) {
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo,
+				    ctx->Array.ArrayBufferObj);
+   }
+
+   old_texcoord_array = &ctx->Array.ArrayObj->TexCoord[0];
+   meta->saved_texcoord_type = old_texcoord_array->Type;
+   meta->saved_texcoord_size = old_texcoord_array->Size;
+   meta->saved_texcoord_stride = old_texcoord_array->Stride;
+   meta->saved_texcoord_enable = old_texcoord_array->Enabled;
+   meta->saved_texcoord_ptr = old_texcoord_array->Ptr;
+   _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo,
+				 old_texcoord_array->BufferObj);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+
+   if (meta->texcoord_vbo == NULL) {
+      GLuint vbo_name;
+
+      _mesa_GenBuffersARB(1, &vbo_name);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, vbo_name);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(default_texcoords),
+			  default_texcoords, GL_STATIC_DRAW_ARB);
+      _mesa_reference_buffer_object(ctx, &meta->texcoord_vbo,
+				    ctx->Array.ArrayBufferObj);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->texcoord_vbo->Name);
+   }
+   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), NULL);
+
+   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+}
+
+void
+meta_restore_texcoords(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   /* Restore the old TexCoordPointer */
+   if (meta->saved_texcoord_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_texcoord_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+
+   _mesa_TexCoordPointer(meta->saved_texcoord_size,
+			 meta->saved_texcoord_type,
+			 meta->saved_texcoord_stride,
+			 meta->saved_texcoord_ptr);
+   if (!meta->saved_texcoord_enable)
+      _mesa_Disable(GL_TEXTURE_COORD_ARRAY);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0 +
+				meta->saved_active_texture);
+
+   if (meta->saved_array_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_array_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+}
+
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta)
+{
+   meta->ctx = ctx;
+}
+
+void meta_destroy_metaops(struct dri_metaops *meta)
+{
+
+}
diff --git a/src/mesa/drivers/dri/common/dri_metaops.h b/src/mesa/drivers/dri/common/dri_metaops.h
new file mode 100644
index 0000000000..2487145326
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.h
@@ -0,0 +1,81 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRI_METAOPS_H
+#define DRI_METAOPS_H
+
+
+struct dri_metaops {
+    GLcontext *ctx;
+    GLboolean internal_viewport_call;
+    struct gl_fragment_program *bitmap_fp;
+    struct gl_vertex_program *passthrough_vp;
+    struct gl_buffer_object *texcoord_vbo;
+    
+    struct gl_fragment_program *saved_fp;
+    GLboolean saved_fp_enable;
+    struct gl_vertex_program *saved_vp;
+    GLboolean saved_vp_enable;
+
+    struct gl_fragment_program *tex2d_fp;
+    
+    GLboolean saved_texcoord_enable;
+    struct gl_buffer_object *saved_array_vbo, *saved_texcoord_vbo;
+    GLenum saved_texcoord_type;
+    GLsizei saved_texcoord_size, saved_texcoord_stride;
+    const void *saved_texcoord_ptr;
+    int saved_active_texture;
+
+    GLint saved_vp_x, saved_vp_y;
+    GLsizei saved_vp_width, saved_vp_height;
+    GLenum saved_matrix_mode;
+};
+
+
+void meta_set_passthrough_transform(struct dri_metaops *meta);
+
+void meta_restore_transform(struct dri_metaops *meta);
+
+void meta_set_passthrough_vertex_program(struct dri_metaops *meta);
+
+void meta_restore_vertex_program(struct dri_metaops *meta);
+
+void meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string);
+
+void meta_restore_fragment_program(struct dri_metaops *meta);
+
+void meta_set_default_texrect(struct dri_metaops *meta);
+
+void meta_restore_texcoords(struct dri_metaops *meta);
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta);
+void meta_destroy_metaops(struct dri_metaops *meta);
+
+#endif
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index ecccc01000..e48e10d7c0 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -1,4 +1,3 @@
-/* $XFree86: xc/lib/GL/dri/dri_util.c,v 1.7 2003/04/28 17:01:25 dawes Exp $ */
 /**
  * \file dri_util.c
  * DRI utility functions.
@@ -37,6 +36,9 @@
 typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRIdrawable *drawable, int32_t *numerator, int32_t *denominator);
 #endif
 
+static void dri_get_drawable(__DRIdrawable *pdp);
+static void dri_put_drawable(__DRIdrawable *pdp);
+
 /**
  * This is just a token extension used to signal that the driver
  * supports setting a read drawable.
@@ -130,7 +132,7 @@ static int driUnbindContext(__DRIcontext *pcp)
 	return GL_FALSE;
     }
 
-    pdp->refcount--;
+    dri_put_drawable(pdp);
 
     if (prp != pdp) {
         if (prp->refcount == 0) {
@@ -138,7 +140,7 @@ static int driUnbindContext(__DRIcontext *pcp)
 	    return GL_FALSE;
 	}
 
-	prp->refcount--;
+    	dri_put_drawable(prp);
     }
 
 
@@ -174,10 +176,10 @@ static int driBindContext(__DRIcontext *pcp,
 	pcp->driReadablePriv = prp;
 	if (pdp) {
 	    pdp->driContextPriv = pcp;
-	    pdp->refcount++;
+    	    dri_get_drawable(pdp);
 	}
 	if ( prp && pdp != prp ) {
-	    prp->refcount++;
+    	    dri_get_drawable(prp);
 	}
     }
 
@@ -434,7 +436,7 @@ driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config,
 
     pdp->loaderPrivate = data;
     pdp->hHWDrawable = hwDrawable;
-    pdp->refcount = 0;
+    pdp->refcount = 1;
     pdp->pStamp = NULL;
     pdp->lastStamp = 0;
     pdp->index = 0;
@@ -487,12 +489,19 @@ dri2CreateNewDrawable(__DRIscreen *screen,
     return pdraw;
 }
 
-
-static void
-driDestroyDrawable(__DRIdrawable *pdp)
+static void dri_get_drawable(__DRIdrawable *pdp)
+{
+    pdp->refcount++;
+}
+	
+static void dri_put_drawable(__DRIdrawable *pdp)
 {
     __DRIscreenPrivate *psp;
 
+    pdp->refcount--;
+    if (pdp->refcount)
+	return;
+
     if (pdp) {
 	psp = pdp->driScreenPriv;
         (*psp->DriverAPI.DestroyBuffer)(pdp);
@@ -508,6 +517,12 @@ driDestroyDrawable(__DRIdrawable *pdp)
     }
 }
 
+static void
+driDestroyDrawable(__DRIdrawable *pdp)
+{
+    dri_put_drawable(pdp);
+}
+
 /*@}*/
 
 
diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h
index 8dcaaee307..40a030ce0d 100644
--- a/src/mesa/drivers/dri/common/extension_helper.h
+++ b/src/mesa/drivers/dri/common/extension_helper.h
@@ -33,7 +33,7 @@
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char UniformMatrix3fvARB_names[] = 
+static const char UniformMatrix3fvARB_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix3fv\0"
     "glUniformMatrix3fvARB\0"
@@ -41,7 +41,7 @@ static const char UniformMatrix3fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_multisample)
-static const char SampleCoverageARB_names[] = 
+static const char SampleCoverageARB_names[] =
     "fi\0" /* Parameter signature */
     "glSampleCoverage\0"
     "glSampleCoverageARB\0"
@@ -49,7 +49,7 @@ static const char SampleCoverageARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionFilter1D_names[] = 
+static const char ConvolutionFilter1D_names[] =
     "iiiiip\0" /* Parameter signature */
     "glConvolutionFilter1D\0"
     "glConvolutionFilter1DEXT\0"
@@ -57,7 +57,7 @@ static const char ConvolutionFilter1D_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char BeginQueryARB_names[] = 
+static const char BeginQueryARB_names[] =
     "ii\0" /* Parameter signature */
     "glBeginQuery\0"
     "glBeginQueryARB\0"
@@ -65,7 +65,7 @@ static const char BeginQueryARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_NV_point_sprite)
-static const char PointParameteriNV_names[] = 
+static const char PointParameteriNV_names[] =
     "ii\0" /* Parameter signature */
     "glPointParameteri\0"
     "glPointParameteriNV\0"
@@ -73,14 +73,14 @@ static const char PointParameteriNV_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char GetProgramiv_names[] = 
+static const char GetProgramiv_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramiv\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3sARB_names[] = 
+static const char MultiTexCoord3sARB_names[] =
     "iiii\0" /* Parameter signature */
     "glMultiTexCoord3s\0"
     "glMultiTexCoord3sARB\0"
@@ -88,7 +88,7 @@ static const char MultiTexCoord3sARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3iEXT_names[] = 
+static const char SecondaryColor3iEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3i\0"
     "glSecondaryColor3iEXT\0"
@@ -96,7 +96,7 @@ static const char SecondaryColor3iEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3fMESA_names[] = 
+static const char WindowPos3fMESA_names[] =
     "fff\0" /* Parameter signature */
     "glWindowPos3f\0"
     "glWindowPos3fARB\0"
@@ -105,14 +105,14 @@ static const char WindowPos3fMESA_names[] =
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char PixelTexGenParameterfvSGIS_names[] = 
+static const char PixelTexGenParameterfvSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glPixelTexGenParameterfvSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char ActiveTextureARB_names[] = 
+static const char ActiveTextureARB_names[] =
     "i\0" /* Parameter signature */
     "glActiveTexture\0"
     "glActiveTextureARB\0"
@@ -120,7 +120,7 @@ static const char ActiveTextureARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_blit)
-static const char BlitFramebufferEXT_names[] = 
+static const char BlitFramebufferEXT_names[] =
     "iiiiiiiiii\0" /* Parameter signature */
     "glBlitFramebuffer\0"
     "glBlitFramebufferEXT\0"
@@ -128,21 +128,21 @@ static const char BlitFramebufferEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4ubvNV_names[] = 
+static const char VertexAttrib4ubvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4ubvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char GetProgramNamedParameterdvNV_names[] = 
+static const char GetProgramNamedParameterdvNV_names[] =
     "iipp\0" /* Parameter signature */
     "glGetProgramNamedParameterdvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char Histogram_names[] = 
+static const char Histogram_names[] =
     "iiii\0" /* Parameter signature */
     "glHistogram\0"
     "glHistogramEXT\0"
@@ -150,14 +150,14 @@ static const char Histogram_names[] =
 #endif
 
 #if defined(need_GL_SGIS_texture4D)
-static const char TexImage4DSGIS_names[] = 
+static const char TexImage4DSGIS_names[] =
     "iiiiiiiiiip\0" /* Parameter signature */
     "glTexImage4DSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2dvMESA_names[] = 
+static const char WindowPos2dvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos2dv\0"
     "glWindowPos2dvARB\0"
@@ -166,14 +166,14 @@ static const char WindowPos2dvMESA_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor3fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiColor3fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glReplacementCodeuiColor3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_EXT_blend_equation_separate) || defined(need_GL_ATI_blend_equation_separate)
-static const char BlendEquationSeparateEXT_names[] = 
+static const char BlendEquationSeparateEXT_names[] =
     "ii\0" /* Parameter signature */
     "glBlendEquationSeparate\0"
     "glBlendEquationSeparateEXT\0"
@@ -182,14 +182,14 @@ static const char BlendEquationSeparateEXT_names[] =
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char ListParameterfSGIX_names[] = 
+static const char ListParameterfSGIX_names[] =
     "iif\0" /* Parameter signature */
     "glListParameterfSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3bEXT_names[] = 
+static const char SecondaryColor3bEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3b\0"
     "glSecondaryColor3bEXT\0"
@@ -197,21 +197,21 @@ static const char SecondaryColor3bEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord4fColor4fNormal3fVertex4fvSUN_names[] = 
+static const char TexCoord4fColor4fNormal3fVertex4fvSUN_names[] =
     "pppp\0" /* Parameter signature */
     "glTexCoord4fColor4fNormal3fVertex4fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4svNV_names[] = 
+static const char VertexAttrib4svNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4svNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char GetBufferSubDataARB_names[] = 
+static const char GetBufferSubDataARB_names[] =
     "iiip\0" /* Parameter signature */
     "glGetBufferSubData\0"
     "glGetBufferSubDataARB\0"
@@ -219,7 +219,7 @@ static const char GetBufferSubDataARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char BufferSubDataARB_names[] = 
+static const char BufferSubDataARB_names[] =
     "iiip\0" /* Parameter signature */
     "glBufferSubData\0"
     "glBufferSubDataARB\0"
@@ -227,21 +227,21 @@ static const char BufferSubDataARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor4ubVertex3fvSUN_names[] = 
+static const char TexCoord2fColor4ubVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glTexCoord2fColor4ubVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char AttachShader_names[] = 
+static const char AttachShader_names[] =
     "ii\0" /* Parameter signature */
     "glAttachShader\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2fARB_names[] = 
+static const char VertexAttrib2fARB_names[] =
     "iff\0" /* Parameter signature */
     "glVertexAttrib2f\0"
     "glVertexAttrib2fARB\0"
@@ -249,14 +249,14 @@ static const char VertexAttrib2fARB_names[] =
 #endif
 
 #if defined(need_GL_MESA_shader_debug)
-static const char GetDebugLogLengthMESA_names[] = 
+static const char GetDebugLogLengthMESA_names[] =
     "iii\0" /* Parameter signature */
     "glGetDebugLogLengthMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3fARB_names[] = 
+static const char VertexAttrib3fARB_names[] =
     "ifff\0" /* Parameter signature */
     "glVertexAttrib3f\0"
     "glVertexAttrib3fARB\0"
@@ -264,7 +264,7 @@ static const char VertexAttrib3fARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char GetQueryivARB_names[] = 
+static const char GetQueryivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetQueryiv\0"
     "glGetQueryivARB\0"
@@ -272,7 +272,7 @@ static const char GetQueryivARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_texture3D)
-static const char TexImage3D_names[] = 
+static const char TexImage3D_names[] =
     "iiiiiiiiip\0" /* Parameter signature */
     "glTexImage3D\0"
     "glTexImage3DEXT\0"
@@ -280,14 +280,14 @@ static const char TexImage3D_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiVertex3fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glReplacementCodeuiVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char GetQueryObjectivARB_names[] = 
+static const char GetQueryObjectivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetQueryObjectiv\0"
     "glGetQueryObjectivARB\0"
@@ -295,14 +295,14 @@ static const char GetQueryObjectivARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexSubImage2DARB_names[] = 
+static const char CompressedTexSubImage2DARB_names[] =
     "iiiiiiiip\0" /* Parameter signature */
     "glCompressedTexSubImage2D\0"
     "glCompressedTexSubImage2DARB\0"
@@ -310,14 +310,14 @@ static const char CompressedTexSubImage2DARB_names[] =
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerOutputNV_names[] = 
+static const char CombinerOutputNV_names[] =
     "iiiiiiiiii\0" /* Parameter signature */
     "glCombinerOutputNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform2fARB_names[] = 
+static const char Uniform2fARB_names[] =
     "iff\0" /* Parameter signature */
     "glUniform2f\0"
     "glUniform2fARB\0"
@@ -325,7 +325,7 @@ static const char Uniform2fARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1svARB_names[] = 
+static const char VertexAttrib1svARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1sv\0"
     "glVertexAttrib1svARB\0"
@@ -333,14 +333,14 @@ static const char VertexAttrib1svARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs1dvNV_names[] = 
+static const char VertexAttribs1dvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs1dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform2ivARB_names[] = 
+static const char Uniform2ivARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform2iv\0"
     "glUniform2ivARB\0"
@@ -348,28 +348,28 @@ static const char Uniform2ivARB_names[] =
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char GetImageTransformParameterfvHP_names[] = 
+static const char GetImageTransformParameterfvHP_names[] =
     "iip\0" /* Parameter signature */
     "glGetImageTransformParameterfvHP\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightubvARB_names[] = 
+static const char WeightubvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightubvARB\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1fvNV_names[] = 
+static const char VertexAttrib1fvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char CopyConvolutionFilter1D_names[] = 
+static const char CopyConvolutionFilter1D_names[] =
     "iiiii\0" /* Parameter signature */
     "glCopyConvolutionFilter1D\0"
     "glCopyConvolutionFilter1DEXT\0"
@@ -377,21 +377,28 @@ static const char CopyConvolutionFilter1D_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiNormal3fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiNormal3fVertex3fSUN_names[] =
     "iffffff\0" /* Parameter signature */
     "glReplacementCodeuiNormal3fVertex3fSUN\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char DeleteSync_names[] =
+    "i\0" /* Parameter signature */
+    "glDeleteSync\0"
+    "";
+#endif
+
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentMaterialfvSGIX_names[] = 
+static const char FragmentMaterialfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glFragmentMaterialfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_blend_color)
-static const char BlendColor_names[] = 
+static const char BlendColor_names[] =
     "ffff\0" /* Parameter signature */
     "glBlendColor\0"
     "glBlendColorEXT\0"
@@ -399,43 +406,44 @@ static const char BlendColor_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char UniformMatrix4fvARB_names[] = 
+static const char UniformMatrix4fvARB_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix4fv\0"
     "glUniformMatrix4fvARB\0"
     "";
 #endif
 
-#if defined(need_GL_APPLE_vertex_array_object)
-static const char DeleteVertexArraysAPPLE_names[] = 
+#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object)
+static const char DeleteVertexArraysAPPLE_names[] =
     "ip\0" /* Parameter signature */
+    "glDeleteVertexArrays\0"
     "glDeleteVertexArraysAPPLE\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_instruments)
-static const char ReadInstrumentsSGIX_names[] = 
+static const char ReadInstrumentsSGIX_names[] =
     "i\0" /* Parameter signature */
     "glReadInstrumentsSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix2x4fv_names[] = 
+static const char UniformMatrix2x4fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix2x4fv\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4ubVertex3fvSUN_names[] = 
+static const char Color4ubVertex3fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glColor4ubVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_texture_array)
-static const char FramebufferTextureLayerEXT_names[] = 
+static const char FramebufferTextureLayerEXT_names[] =
     "iiiii\0" /* Parameter signature */
     "glFramebufferTextureLayer\0"
     "glFramebufferTextureLayerEXT\0"
@@ -443,14 +451,14 @@ static const char FramebufferTextureLayerEXT_names[] =
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char GetListParameterfvSGIX_names[] = 
+static const char GetListParameterfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetListParameterfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NusvARB_names[] = 
+static const char VertexAttrib4NusvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Nusv\0"
     "glVertexAttrib4NusvARB\0"
@@ -458,35 +466,35 @@ static const char VertexAttrib4NusvARB_names[] =
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4svMESA_names[] = 
+static const char WindowPos4svMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos4svMESA\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char CreateProgramObjectARB_names[] = 
+static const char CreateProgramObjectARB_names[] =
     "\0" /* Parameter signature */
     "glCreateProgramObjectARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightModelivSGIX_names[] = 
+static const char FragmentLightModelivSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glFragmentLightModelivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix4x3fv_names[] = 
+static const char UniformMatrix4x3fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix4x3fv\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char PrioritizeTextures_names[] = 
+static const char PrioritizeTextures_names[] =
     "ipp\0" /* Parameter signature */
     "glPrioritizeTextures\0"
     "glPrioritizeTexturesEXT\0"
@@ -494,28 +502,28 @@ static const char PrioritizeTextures_names[] =
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char AsyncMarkerSGIX_names[] = 
+static const char AsyncMarkerSGIX_names[] =
     "i\0" /* Parameter signature */
     "glAsyncMarkerSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactorubSUN_names[] = 
+static const char GlobalAlphaFactorubSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactorubSUN\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_shader_debug)
-static const char ClearDebugLogMESA_names[] = 
+static const char ClearDebugLogMESA_names[] =
     "iii\0" /* Parameter signature */
     "glClearDebugLogMESA\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char ResetHistogram_names[] = 
+static const char ResetHistogram_names[] =
     "i\0" /* Parameter signature */
     "glResetHistogram\0"
     "glResetHistogramEXT\0"
@@ -523,14 +531,14 @@ static const char ResetHistogram_names[] =
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char GetProgramNamedParameterfvNV_names[] = 
+static const char GetProgramNamedParameterfvNV_names[] =
     "iipp\0" /* Parameter signature */
     "glGetProgramNamedParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_point_parameters) || defined(need_GL_EXT_point_parameters) || defined(need_GL_SGIS_point_parameters)
-static const char PointParameterfEXT_names[] = 
+static const char PointParameterfEXT_names[] =
     "if\0" /* Parameter signature */
     "glPointParameterf\0"
     "glPointParameterfARB\0"
@@ -540,35 +548,35 @@ static const char PointParameterfEXT_names[] =
 #endif
 
 #if defined(need_GL_SGIX_polynomial_ffd)
-static const char LoadIdentityDeformationMapSGIX_names[] = 
+static const char LoadIdentityDeformationMapSGIX_names[] =
     "i\0" /* Parameter signature */
     "glLoadIdentityDeformationMapSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char GenFencesNV_names[] = 
+static const char GenFencesNV_names[] =
     "ip\0" /* Parameter signature */
     "glGenFencesNV\0"
     "";
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char ImageTransformParameterfHP_names[] = 
+static const char ImageTransformParameterfHP_names[] =
     "iif\0" /* Parameter signature */
     "glImageTransformParameterfHP\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_matrix_palette)
-static const char MatrixIndexusvARB_names[] = 
+static const char MatrixIndexusvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMatrixIndexusvARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char DisableVertexAttribArrayARB_names[] = 
+static const char DisableVertexAttribArrayARB_names[] =
     "i\0" /* Parameter signature */
     "glDisableVertexAttribArray\0"
     "glDisableVertexAttribArrayARB\0"
@@ -576,21 +584,21 @@ static const char DisableVertexAttribArrayARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char StencilMaskSeparate_names[] = 
+static const char StencilMaskSeparate_names[] =
     "ii\0" /* Parameter signature */
     "glStencilMaskSeparate\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char ProgramLocalParameter4dARB_names[] = 
+static const char ProgramLocalParameter4dARB_names[] =
     "iidddd\0" /* Parameter signature */
     "glProgramLocalParameter4dARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexImage3DARB_names[] = 
+static const char CompressedTexImage3DARB_names[] =
     "iiiiiiiip\0" /* Parameter signature */
     "glCompressedTexImage3D\0"
     "glCompressedTexImage3DARB\0"
@@ -598,7 +606,7 @@ static const char CompressedTexImage3DARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char GetConvolutionParameteriv_names[] = 
+static const char GetConvolutionParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glGetConvolutionParameteriv\0"
     "glGetConvolutionParameterivEXT\0"
@@ -606,7 +614,7 @@ static const char GetConvolutionParameteriv_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1fARB_names[] = 
+static const char VertexAttrib1fARB_names[] =
     "if\0" /* Parameter signature */
     "glVertexAttrib1f\0"
     "glVertexAttrib1fARB\0"
@@ -614,14 +622,14 @@ static const char VertexAttrib1fARB_names[] =
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char TestFenceNV_names[] = 
+static const char TestFenceNV_names[] =
     "i\0" /* Parameter signature */
     "glTestFenceNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1fvARB_names[] = 
+static const char MultiTexCoord1fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord1fv\0"
     "glMultiTexCoord1fvARB\0"
@@ -629,56 +637,56 @@ static const char MultiTexCoord1fvARB_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char ColorFragmentOp2ATI_names[] = 
+static const char ColorFragmentOp2ATI_names[] =
     "iiiiiiiiii\0" /* Parameter signature */
     "glColorFragmentOp2ATI\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char SecondaryColorPointerListIBM_names[] = 
+static const char SecondaryColorPointerListIBM_names[] =
     "iiipi\0" /* Parameter signature */
     "glSecondaryColorPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char GetPixelTexGenParameterivSGIS_names[] = 
+static const char GetPixelTexGenParameterivSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glGetPixelTexGenParameterivSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4fNV_names[] = 
+static const char VertexAttrib4fNV_names[] =
     "iffff\0" /* Parameter signature */
     "glVertexAttrib4fNV\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeubSUN_names[] = 
+static const char ReplacementCodeubSUN_names[] =
     "i\0" /* Parameter signature */
     "glReplacementCodeubSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char FinishAsyncSGIX_names[] = 
+static const char FinishAsyncSGIX_names[] =
     "p\0" /* Parameter signature */
     "glFinishAsyncSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_shader_debug)
-static const char GetDebugLogMESA_names[] = 
+static const char GetDebugLogMESA_names[] =
     "iiiipp\0" /* Parameter signature */
     "glGetDebugLogMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
-static const char FogCoorddEXT_names[] = 
+static const char FogCoorddEXT_names[] =
     "d\0" /* Parameter signature */
     "glFogCoordd\0"
     "glFogCoorddEXT\0"
@@ -686,14 +694,14 @@ static const char FogCoorddEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4ubVertex3fSUN_names[] = 
+static const char Color4ubVertex3fSUN_names[] =
     "iiiifff\0" /* Parameter signature */
     "glColor4ubVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
-static const char FogCoordfEXT_names[] = 
+static const char FogCoordfEXT_names[] =
     "f\0" /* Parameter signature */
     "glFogCoordf\0"
     "glFogCoordfEXT\0"
@@ -701,35 +709,35 @@ static const char FogCoordfEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fVertex3fSUN_names[] = 
+static const char TexCoord2fVertex3fSUN_names[] =
     "fffff\0" /* Parameter signature */
     "glTexCoord2fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactoriSUN_names[] = 
+static const char GlobalAlphaFactoriSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactoriSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2dNV_names[] = 
+static const char VertexAttrib2dNV_names[] =
     "idd\0" /* Parameter signature */
     "glVertexAttrib2dNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char GetProgramInfoLog_names[] = 
+static const char GetProgramInfoLog_names[] =
     "iipp\0" /* Parameter signature */
     "glGetProgramInfoLog\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NbvARB_names[] = 
+static const char VertexAttrib4NbvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Nbv\0"
     "glVertexAttrib4NbvARB\0"
@@ -737,7 +745,7 @@ static const char VertexAttrib4NbvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader)
-static const char GetActiveAttribARB_names[] = 
+static const char GetActiveAttribARB_names[] =
     "iiipppp\0" /* Parameter signature */
     "glGetActiveAttrib\0"
     "glGetActiveAttribARB\0"
@@ -745,84 +753,91 @@ static const char GetActiveAttribARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4ubNV_names[] = 
+static const char VertexAttrib4ubNV_names[] =
     "iiiii\0" /* Parameter signature */
     "glVertexAttrib4ubNV\0"
     "";
 #endif
 
+#if defined(need_GL_APPLE_texture_range)
+static const char TextureRangeAPPLE_names[] =
+    "iip\0" /* Parameter signature */
+    "glTextureRangeAPPLE\0"
+    "";
+#endif
+
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor4fNormal3fVertex3fSUN_names[] = 
+static const char TexCoord2fColor4fNormal3fVertex3fSUN_names[] =
     "ffffffffffff\0" /* Parameter signature */
     "glTexCoord2fColor4fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerParameterfvNV_names[] = 
+static const char CombinerParameterfvNV_names[] =
     "ip\0" /* Parameter signature */
     "glCombinerParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs3dvNV_names[] = 
+static const char VertexAttribs3dvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs3dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs4fvNV_names[] = 
+static const char VertexAttribs4fvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs4fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_array_range)
-static const char VertexArrayRangeNV_names[] = 
+static const char VertexArrayRangeNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexArrayRangeNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightiSGIX_names[] = 
+static const char FragmentLightiSGIX_names[] =
     "iii\0" /* Parameter signature */
     "glFragmentLightiSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_polygon_offset)
-static const char PolygonOffsetEXT_names[] = 
+static const char PolygonOffsetEXT_names[] =
     "ff\0" /* Parameter signature */
     "glPolygonOffsetEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char PollAsyncSGIX_names[] = 
+static const char PollAsyncSGIX_names[] =
     "p\0" /* Parameter signature */
     "glPollAsyncSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char DeleteFragmentShaderATI_names[] = 
+static const char DeleteFragmentShaderATI_names[] =
     "i\0" /* Parameter signature */
     "glDeleteFragmentShaderATI\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fNormal3fVertex3fvSUN_names[] = 
+static const char TexCoord2fNormal3fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glTexCoord2fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix)
-static const char MultTransposeMatrixdARB_names[] = 
+static const char MultTransposeMatrixdARB_names[] =
     "p\0" /* Parameter signature */
     "glMultTransposeMatrixd\0"
     "glMultTransposeMatrixdARB\0"
@@ -830,7 +845,7 @@ static const char MultTransposeMatrixdARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2svMESA_names[] = 
+static const char WindowPos2svMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos2sv\0"
     "glWindowPos2svARB\0"
@@ -839,7 +854,7 @@ static const char WindowPos2svMESA_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexImage1DARB_names[] = 
+static const char CompressedTexImage1DARB_names[] =
     "iiiiiip\0" /* Parameter signature */
     "glCompressedTexImage1D\0"
     "glCompressedTexImage1DARB\0"
@@ -847,35 +862,35 @@ static const char CompressedTexImage1DARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2sNV_names[] = 
+static const char VertexAttrib2sNV_names[] =
     "iii\0" /* Parameter signature */
     "glVertexAttrib2sNV\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char NormalPointerListIBM_names[] = 
+static const char NormalPointerListIBM_names[] =
     "iipi\0" /* Parameter signature */
     "glNormalPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char IndexPointerEXT_names[] = 
+static const char IndexPointerEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glIndexPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char NormalPointerEXT_names[] = 
+static const char NormalPointerEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glNormalPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3dARB_names[] = 
+static const char MultiTexCoord3dARB_names[] =
     "iddd\0" /* Parameter signature */
     "glMultiTexCoord3d\0"
     "glMultiTexCoord3dARB\0"
@@ -883,7 +898,7 @@ static const char MultiTexCoord3dARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2iARB_names[] = 
+static const char MultiTexCoord2iARB_names[] =
     "iii\0" /* Parameter signature */
     "glMultiTexCoord2i\0"
     "glMultiTexCoord2iARB\0"
@@ -891,14 +906,14 @@ static const char MultiTexCoord2iARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fSUN_names[] =
     "iffffffff\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2svARB_names[] = 
+static const char MultiTexCoord2svARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord2sv\0"
     "glMultiTexCoord2svARB\0"
@@ -906,14 +921,14 @@ static const char MultiTexCoord2svARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeubvSUN_names[] = 
+static const char ReplacementCodeubvSUN_names[] =
     "p\0" /* Parameter signature */
     "glReplacementCodeubvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform3iARB_names[] = 
+static const char Uniform3iARB_names[] =
     "iiii\0" /* Parameter signature */
     "glUniform3i\0"
     "glUniform3iARB\0"
@@ -921,42 +936,49 @@ static const char Uniform3iARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char GetFragmentMaterialfvSGIX_names[] = 
+static const char GetFragmentMaterialfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetFragmentMaterialfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char GetShaderInfoLog_names[] = 
+static const char GetShaderInfoLog_names[] =
     "iipp\0" /* Parameter signature */
     "glGetShaderInfoLog\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightivARB_names[] = 
+static const char WeightivARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightivARB\0"
     "";
 #endif
 
+#if defined(need_GL_SGIX_instruments)
+static const char PollInstrumentsSGIX_names[] =
+    "p\0" /* Parameter signature */
+    "glPollInstrumentsSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactordSUN_names[] = 
+static const char GlobalAlphaFactordSUN_names[] =
     "d\0" /* Parameter signature */
     "glGlobalAlphaFactordSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs3fvNV_names[] = 
+static const char VertexAttribs3fvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs3fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char GenerateMipmapEXT_names[] = 
+static const char GenerateMipmapEXT_names[] =
     "i\0" /* Parameter signature */
     "glGenerateMipmap\0"
     "glGenerateMipmapEXT\0"
@@ -964,35 +986,35 @@ static const char GenerateMipmapEXT_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char SetFragmentShaderConstantATI_names[] = 
+static const char SetFragmentShaderConstantATI_names[] =
     "ip\0" /* Parameter signature */
     "glSetFragmentShaderConstantATI\0"
     "";
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char GetMapAttribParameterivNV_names[] = 
+static const char GetMapAttribParameterivNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetMapAttribParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char CreateShaderObjectARB_names[] = 
+static const char CreateShaderObjectARB_names[] =
     "i\0" /* Parameter signature */
     "glCreateShaderObjectARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_sharpen_texture)
-static const char GetSharpenTexFuncSGIS_names[] = 
+static const char GetSharpenTexFuncSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glGetSharpenTexFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char BufferDataARB_names[] = 
+static const char BufferDataARB_names[] =
     "iipi\0" /* Parameter signature */
     "glBufferData\0"
     "glBufferDataARB\0"
@@ -1000,42 +1022,42 @@ static const char BufferDataARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_array_range)
-static const char FlushVertexArrayRangeNV_names[] = 
+static const char FlushVertexArrayRangeNV_names[] =
     "\0" /* Parameter signature */
     "glFlushVertexArrayRangeNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char SampleMapATI_names[] = 
+static const char SampleMapATI_names[] =
     "iii\0" /* Parameter signature */
     "glSampleMapATI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char VertexPointerEXT_names[] = 
+static const char VertexPointerEXT_names[] =
     "iiiip\0" /* Parameter signature */
     "glVertexPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_texture_filter4)
-static const char GetTexFilterFuncSGIS_names[] = 
+static const char GetTexFilterFuncSGIS_names[] =
     "iip\0" /* Parameter signature */
     "glGetTexFilterFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetCombinerOutputParameterfvNV_names[] = 
+static const char GetCombinerOutputParameterfvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetCombinerOutputParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_subtexture)
-static const char TexSubImage1D_names[] = 
+static const char TexSubImage1D_names[] =
     "iiiiiip\0" /* Parameter signature */
     "glTexSubImage1D\0"
     "glTexSubImage1DEXT\0"
@@ -1043,36 +1065,43 @@ static const char TexSubImage1D_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1sARB_names[] = 
+static const char VertexAttrib1sARB_names[] =
     "ii\0" /* Parameter signature */
     "glVertexAttrib1s\0"
     "glVertexAttrib1sARB\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char FenceSync_names[] =
+    "ii\0" /* Parameter signature */
+    "glFenceSync\0"
+    "";
+#endif
+
 #if defined(need_GL_NV_register_combiners)
-static const char FinalCombinerInputNV_names[] = 
+static const char FinalCombinerInputNV_names[] =
     "iiii\0" /* Parameter signature */
     "glFinalCombinerInputNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_flush_raster)
-static const char FlushRasterSGIX_names[] = 
+static const char FlushRasterSGIX_names[] =
     "\0" /* Parameter signature */
     "glFlushRasterSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fVertex3fSUN_names[] =
     "ifffff\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform1fARB_names[] = 
+static const char Uniform1fARB_names[] =
     "if\0" /* Parameter signature */
     "glUniform1f\0"
     "glUniform1fARB\0"
@@ -1080,7 +1109,7 @@ static const char Uniform1fARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char AreTexturesResident_names[] = 
+static const char AreTexturesResident_names[] =
     "ipp\0" /* Parameter signature */
     "glAreTexturesResident\0"
     "glAreTexturesResidentEXT\0"
@@ -1088,7 +1117,7 @@ static const char AreTexturesResident_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ATI_separate_stencil)
-static const char StencilOpSeparate_names[] = 
+static const char StencilOpSeparate_names[] =
     "iiii\0" /* Parameter signature */
     "glStencilOpSeparate\0"
     "glStencilOpSeparateATI\0"
@@ -1096,7 +1125,7 @@ static const char StencilOpSeparate_names[] =
 #endif
 
 #if defined(need_GL_SGI_color_table)
-static const char ColorTableParameteriv_names[] = 
+static const char ColorTableParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glColorTableParameteriv\0"
     "glColorTableParameterivSGI\0"
@@ -1104,14 +1133,14 @@ static const char ColorTableParameteriv_names[] =
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char FogCoordPointerListIBM_names[] = 
+static const char FogCoordPointerListIBM_names[] =
     "iipi\0" /* Parameter signature */
     "glFogCoordPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3dMESA_names[] = 
+static const char WindowPos3dMESA_names[] =
     "ddd\0" /* Parameter signature */
     "glWindowPos3d\0"
     "glWindowPos3dARB\0"
@@ -1120,7 +1149,7 @@ static const char WindowPos3dMESA_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_point_parameters) || defined(need_GL_EXT_point_parameters) || defined(need_GL_SGIS_point_parameters)
-static const char PointParameterfvEXT_names[] = 
+static const char PointParameterfvEXT_names[] =
     "ip\0" /* Parameter signature */
     "glPointParameterfv\0"
     "glPointParameterfvARB\0"
@@ -1130,7 +1159,7 @@ static const char PointParameterfvEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2fvMESA_names[] = 
+static const char WindowPos2fvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos2fv\0"
     "glWindowPos2fvARB\0"
@@ -1139,7 +1168,7 @@ static const char WindowPos2fvMESA_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3bvEXT_names[] = 
+static const char SecondaryColor3bvEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3bv\0"
     "glSecondaryColor3bvEXT\0"
@@ -1147,35 +1176,35 @@ static const char SecondaryColor3bvEXT_names[] =
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char VertexPointerListIBM_names[] = 
+static const char VertexPointerListIBM_names[] =
     "iiipi\0" /* Parameter signature */
     "glVertexPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramLocalParameterfvARB_names[] = 
+static const char GetProgramLocalParameterfvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramLocalParameterfvARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentMaterialfSGIX_names[] = 
+static const char FragmentMaterialfSGIX_names[] =
     "iif\0" /* Parameter signature */
     "glFragmentMaterialfSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fNormal3fVertex3fSUN_names[] = 
+static const char TexCoord2fNormal3fVertex3fSUN_names[] =
     "ffffffff\0" /* Parameter signature */
     "glTexCoord2fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char RenderbufferStorageEXT_names[] = 
+static const char RenderbufferStorageEXT_names[] =
     "iiii\0" /* Parameter signature */
     "glRenderbufferStorage\0"
     "glRenderbufferStorageEXT\0"
@@ -1183,28 +1212,28 @@ static const char RenderbufferStorageEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char IsFenceNV_names[] = 
+static const char IsFenceNV_names[] =
     "i\0" /* Parameter signature */
     "glIsFenceNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char AttachObjectARB_names[] = 
+static const char AttachObjectARB_names[] =
     "ii\0" /* Parameter signature */
     "glAttachObjectARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char GetFragmentLightivSGIX_names[] = 
+static const char GetFragmentLightivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetFragmentLightivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char UniformMatrix2fvARB_names[] = 
+static const char UniformMatrix2fvARB_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix2fv\0"
     "glUniformMatrix2fvARB\0"
@@ -1212,7 +1241,7 @@ static const char UniformMatrix2fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2fARB_names[] = 
+static const char MultiTexCoord2fARB_names[] =
     "iff\0" /* Parameter signature */
     "glMultiTexCoord2f\0"
     "glMultiTexCoord2fARB\0"
@@ -1220,7 +1249,7 @@ static const char MultiTexCoord2fARB_names[] =
 #endif
 
 #if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture)
-static const char ColorTable_names[] = 
+static const char ColorTable_names[] =
     "iiiiip\0" /* Parameter signature */
     "glColorTable\0"
     "glColorTableSGI\0"
@@ -1229,14 +1258,14 @@ static const char ColorTable_names[] =
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char MapControlPointsNV_names[] = 
+static const char MapControlPointsNV_names[] =
     "iiiiiiiip\0" /* Parameter signature */
     "glMapControlPointsNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionFilter2D_names[] = 
+static const char ConvolutionFilter2D_names[] =
     "iiiiiip\0" /* Parameter signature */
     "glConvolutionFilter2D\0"
     "glConvolutionFilter2DEXT\0"
@@ -1244,14 +1273,14 @@ static const char ConvolutionFilter2D_names[] =
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char MapParameterfvNV_names[] = 
+static const char MapParameterfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glMapParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3dvARB_names[] = 
+static const char VertexAttrib3dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3dv\0"
     "glVertexAttrib3dvARB\0"
@@ -1259,14 +1288,14 @@ static const char VertexAttrib3dvARB_names[] =
 #endif
 
 #if defined(need_GL_PGI_misc_hints)
-static const char HintPGI_names[] = 
+static const char HintPGI_names[] =
     "ii\0" /* Parameter signature */
     "glHintPGI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionParameteriv_names[] = 
+static const char ConvolutionParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glConvolutionParameteriv\0"
     "glConvolutionParameterivEXT\0"
@@ -1274,28 +1303,28 @@ static const char ConvolutionParameteriv_names[] =
 #endif
 
 #if defined(need_GL_EXT_cull_vertex)
-static const char CullParameterdvEXT_names[] = 
+static const char CullParameterdvEXT_names[] =
     "ip\0" /* Parameter signature */
     "glCullParameterdvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char ProgramNamedParameter4fNV_names[] = 
+static const char ProgramNamedParameter4fNV_names[] =
     "iipffff\0" /* Parameter signature */
     "glProgramNamedParameter4fNV\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color3fVertex3fSUN_names[] = 
+static const char Color3fVertex3fSUN_names[] =
     "ffffff\0" /* Parameter signature */
     "glColor3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char ProgramEnvParameter4fvARB_names[] = 
+static const char ProgramEnvParameter4fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glProgramEnvParameter4fvARB\0"
     "glProgramParameter4fvNV\0"
@@ -1303,14 +1332,14 @@ static const char ProgramEnvParameter4fvARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightModeliSGIX_names[] = 
+static const char FragmentLightModeliSGIX_names[] =
     "ii\0" /* Parameter signature */
     "glFragmentLightModeliSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionParameterfv_names[] = 
+static const char ConvolutionParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glConvolutionParameterfv\0"
     "glConvolutionParameterfvEXT\0"
@@ -1318,35 +1347,42 @@ static const char ConvolutionParameterfv_names[] =
 #endif
 
 #if defined(need_GL_3DFX_tbuffer)
-static const char TbufferMask3DFX_names[] = 
+static const char TbufferMask3DFX_names[] =
     "i\0" /* Parameter signature */
     "glTbufferMask3DFX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char LoadProgramNV_names[] = 
+static const char LoadProgramNV_names[] =
     "iiip\0" /* Parameter signature */
     "glLoadProgramNV\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char WaitSync_names[] =
+    "iii\0" /* Parameter signature */
+    "glWaitSync\0"
+    "";
+#endif
+
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4fvNV_names[] = 
+static const char VertexAttrib4fvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char GetAttachedObjectsARB_names[] = 
+static const char GetAttachedObjectsARB_names[] =
     "iipp\0" /* Parameter signature */
     "glGetAttachedObjectsARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform3fvARB_names[] = 
+static const char Uniform3fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform3fv\0"
     "glUniform3fvARB\0"
@@ -1354,7 +1390,7 @@ static const char Uniform3fvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_draw_range_elements)
-static const char DrawRangeElements_names[] = 
+static const char DrawRangeElements_names[] =
     "iiiiip\0" /* Parameter signature */
     "glDrawRangeElements\0"
     "glDrawRangeElementsEXT\0"
@@ -1362,14 +1398,14 @@ static const char DrawRangeElements_names[] =
 #endif
 
 #if defined(need_GL_SGIX_sprite)
-static const char SpriteParameterfvSGIX_names[] = 
+static const char SpriteParameterfvSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glSpriteParameterfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char CheckFramebufferStatusEXT_names[] = 
+static const char CheckFramebufferStatusEXT_names[] =
     "i\0" /* Parameter signature */
     "glCheckFramebufferStatus\0"
     "glCheckFramebufferStatusEXT\0"
@@ -1377,21 +1413,21 @@ static const char CheckFramebufferStatusEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactoruiSUN_names[] = 
+static const char GlobalAlphaFactoruiSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactoruiSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char GetHandleARB_names[] = 
+static const char GetHandleARB_names[] =
     "i\0" /* Parameter signature */
     "glGetHandleARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char GetVertexAttribivARB_names[] = 
+static const char GetVertexAttribivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribiv\0"
     "glGetVertexAttribivARB\0"
@@ -1399,21 +1435,21 @@ static const char GetVertexAttribivARB_names[] =
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetCombinerInputParameterfvNV_names[] = 
+static const char GetCombinerInputParameterfvNV_names[] =
     "iiiip\0" /* Parameter signature */
     "glGetCombinerInputParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char CreateProgram_names[] = 
+static const char CreateProgram_names[] =
     "\0" /* Parameter signature */
     "glCreateProgram\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix)
-static const char LoadTransposeMatrixdARB_names[] = 
+static const char LoadTransposeMatrixdARB_names[] =
     "p\0" /* Parameter signature */
     "glLoadTransposeMatrixd\0"
     "glLoadTransposeMatrixdARB\0"
@@ -1421,7 +1457,7 @@ static const char LoadTransposeMatrixdARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetMinmax_names[] = 
+static const char GetMinmax_names[] =
     "iiiip\0" /* Parameter signature */
     "glGetMinmax\0"
     "glGetMinmaxEXT\0"
@@ -1429,14 +1465,14 @@ static const char GetMinmax_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char StencilFuncSeparate_names[] = 
+static const char StencilFuncSeparate_names[] =
     "iiii\0" /* Parameter signature */
     "glStencilFuncSeparate\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3sEXT_names[] = 
+static const char SecondaryColor3sEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3s\0"
     "glSecondaryColor3sEXT\0"
@@ -1444,28 +1480,28 @@ static const char SecondaryColor3sEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color3fVertex3fvSUN_names[] = 
+static const char Color3fVertex3fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glColor3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactorbSUN_names[] = 
+static const char GlobalAlphaFactorbSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactorbSUN\0"
     "";
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char ImageTransformParameterfvHP_names[] = 
+static const char ImageTransformParameterfvHP_names[] =
     "iip\0" /* Parameter signature */
     "glImageTransformParameterfvHP\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4ivARB_names[] = 
+static const char VertexAttrib4ivARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4iv\0"
     "glVertexAttrib4ivARB\0"
@@ -1473,28 +1509,28 @@ static const char VertexAttrib4ivARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3fNV_names[] = 
+static const char VertexAttrib3fNV_names[] =
     "ifff\0" /* Parameter signature */
     "glVertexAttrib3fNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs2dvNV_names[] = 
+static const char VertexAttribs2dvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs2dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_timer_query)
-static const char GetQueryObjectui64vEXT_names[] = 
+static const char GetQueryObjectui64vEXT_names[] =
     "iip\0" /* Parameter signature */
     "glGetQueryObjectui64vEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3fvARB_names[] = 
+static const char MultiTexCoord3fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord3fv\0"
     "glMultiTexCoord3fvARB\0"
@@ -1502,7 +1538,7 @@ static const char MultiTexCoord3fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3dEXT_names[] = 
+static const char SecondaryColor3dEXT_names[] =
     "ddd\0" /* Parameter signature */
     "glSecondaryColor3d\0"
     "glSecondaryColor3dEXT\0"
@@ -1510,42 +1546,42 @@ static const char SecondaryColor3dEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetProgramParameterfvNV_names[] = 
+static const char GetProgramParameterfvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetProgramParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char TangentPointerEXT_names[] = 
+static const char TangentPointerEXT_names[] =
     "iip\0" /* Parameter signature */
     "glTangentPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4fNormal3fVertex3fvSUN_names[] = 
+static const char Color4fNormal3fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glColor4fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_instruments)
-static const char GetInstrumentsSGIX_names[] = 
+static const char GetInstrumentsSGIX_names[] =
     "\0" /* Parameter signature */
     "glGetInstrumentsSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char EvalMapsNV_names[] = 
+static const char EvalMapsNV_names[] =
     "ii\0" /* Parameter signature */
     "glEvalMapsNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_subtexture)
-static const char TexSubImage2D_names[] = 
+static const char TexSubImage2D_names[] =
     "iiiiiiiip\0" /* Parameter signature */
     "glTexSubImage2D\0"
     "glTexSubImage2DEXT\0"
@@ -1553,21 +1589,28 @@ static const char TexSubImage2D_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightivSGIX_names[] = 
+static const char FragmentLightivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glFragmentLightivSGIX\0"
     "";
 #endif
 
+#if defined(need_GL_APPLE_texture_range)
+static const char GetTexParameterPointervAPPLE_names[] =
+    "iip\0" /* Parameter signature */
+    "glGetTexParameterPointervAPPLE\0"
+    "";
+#endif
+
 #if defined(need_GL_EXT_pixel_transform)
-static const char PixelTransformParameterfvEXT_names[] = 
+static const char PixelTransformParameterfvEXT_names[] =
     "iip\0" /* Parameter signature */
     "glPixelTransformParameterfvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4bvARB_names[] = 
+static const char VertexAttrib4bvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4bv\0"
     "glVertexAttrib4bvARB\0"
@@ -1575,14 +1618,14 @@ static const char VertexAttrib4bvARB_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char AlphaFragmentOp2ATI_names[] = 
+static const char AlphaFragmentOp2ATI_names[] =
     "iiiiiiiii\0" /* Parameter signature */
     "glAlphaFragmentOp2ATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4sARB_names[] = 
+static const char MultiTexCoord4sARB_names[] =
     "iiiii\0" /* Parameter signature */
     "glMultiTexCoord4s\0"
     "glMultiTexCoord4sARB\0"
@@ -1590,28 +1633,28 @@ static const char MultiTexCoord4sARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char GetFragmentMaterialivSGIX_names[] = 
+static const char GetFragmentMaterialivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetFragmentMaterialivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4dMESA_names[] = 
+static const char WindowPos4dMESA_names[] =
     "dddd\0" /* Parameter signature */
     "glWindowPos4dMESA\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightPointerARB_names[] = 
+static const char WeightPointerARB_names[] =
     "iiip\0" /* Parameter signature */
     "glWeightPointerARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2dMESA_names[] = 
+static const char WindowPos2dMESA_names[] =
     "dd\0" /* Parameter signature */
     "glWindowPos2d\0"
     "glWindowPos2dARB\0"
@@ -1620,7 +1663,7 @@ static const char WindowPos2dMESA_names[] =
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char FramebufferTexture3DEXT_names[] = 
+static const char FramebufferTexture3DEXT_names[] =
     "iiiiii\0" /* Parameter signature */
     "glFramebufferTexture3D\0"
     "glFramebufferTexture3DEXT\0"
@@ -1628,7 +1671,7 @@ static const char FramebufferTexture3DEXT_names[] =
 #endif
 
 #if defined(need_GL_EXT_blend_minmax)
-static const char BlendEquation_names[] = 
+static const char BlendEquation_names[] =
     "i\0" /* Parameter signature */
     "glBlendEquation\0"
     "glBlendEquationEXT\0"
@@ -1636,14 +1679,14 @@ static const char BlendEquation_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3dNV_names[] = 
+static const char VertexAttrib3dNV_names[] =
     "iddd\0" /* Parameter signature */
     "glVertexAttrib3dNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3dARB_names[] = 
+static const char VertexAttrib3dARB_names[] =
     "iddd\0" /* Parameter signature */
     "glVertexAttrib3d\0"
     "glVertexAttrib3dARB\0"
@@ -1651,14 +1694,14 @@ static const char VertexAttrib3dARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN_names[] =
     "ppppp\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4fARB_names[] = 
+static const char VertexAttrib4fARB_names[] =
     "iffff\0" /* Parameter signature */
     "glVertexAttrib4f\0"
     "glVertexAttrib4fARB\0"
@@ -1666,14 +1709,14 @@ static const char VertexAttrib4fARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_index_func)
-static const char IndexFuncEXT_names[] = 
+static const char IndexFuncEXT_names[] =
     "if\0" /* Parameter signature */
     "glIndexFuncEXT\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char FramebufferTexture2DEXT_names[] = 
+static const char FramebufferTexture2DEXT_names[] =
     "iiiii\0" /* Parameter signature */
     "glFramebufferTexture2D\0"
     "glFramebufferTexture2DEXT\0"
@@ -1681,7 +1724,7 @@ static const char FramebufferTexture2DEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2dvARB_names[] = 
+static const char MultiTexCoord2dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord2dv\0"
     "glMultiTexCoord2dvARB\0"
@@ -1689,21 +1732,21 @@ static const char MultiTexCoord2dvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_cull_vertex)
-static const char CullParameterfvEXT_names[] = 
+static const char CullParameterfvEXT_names[] =
     "ip\0" /* Parameter signature */
     "glCullParameterfvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char ProgramNamedParameter4fvNV_names[] = 
+static const char ProgramNamedParameter4fvNV_names[] =
     "iipp\0" /* Parameter signature */
     "glProgramNamedParameter4fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColorPointerEXT_names[] = 
+static const char SecondaryColorPointerEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glSecondaryColorPointer\0"
     "glSecondaryColorPointerEXT\0"
@@ -1711,7 +1754,7 @@ static const char SecondaryColorPointerEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4fvARB_names[] = 
+static const char VertexAttrib4fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4fv\0"
     "glVertexAttrib4fvARB\0"
@@ -1719,14 +1762,14 @@ static const char VertexAttrib4fvARB_names[] =
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char ColorPointerListIBM_names[] = 
+static const char ColorPointerListIBM_names[] =
     "iiipi\0" /* Parameter signature */
     "glColorPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char GetActiveUniformARB_names[] = 
+static const char GetActiveUniformARB_names[] =
     "iiipppp\0" /* Parameter signature */
     "glGetActiveUniform\0"
     "glGetActiveUniformARB\0"
@@ -1734,14 +1777,14 @@ static const char GetActiveUniformARB_names[] =
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char ImageTransformParameteriHP_names[] = 
+static const char ImageTransformParameteriHP_names[] =
     "iii\0" /* Parameter signature */
     "glImageTransformParameteriHP\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1svARB_names[] = 
+static const char MultiTexCoord1svARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord1sv\0"
     "glMultiTexCoord1svARB\0"
@@ -1749,7 +1792,7 @@ static const char MultiTexCoord1svARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char EndQueryARB_names[] = 
+static const char EndQueryARB_names[] =
     "i\0" /* Parameter signature */
     "glEndQuery\0"
     "glEndQueryARB\0"
@@ -1757,35 +1800,42 @@ static const char EndQueryARB_names[] =
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char DeleteFencesNV_names[] = 
+static const char DeleteFencesNV_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteFencesNV\0"
     "";
 #endif
 
+#if defined(need_GL_SGIX_polynomial_ffd)
+static const char DeformationMap3dSGIX_names[] =
+    "iddiiddiiddiip\0" /* Parameter signature */
+    "glDeformationMap3dSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_2_0)
-static const char IsShader_names[] = 
+static const char IsShader_names[] =
     "i\0" /* Parameter signature */
     "glIsShader\0"
     "";
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char GetImageTransformParameterivHP_names[] = 
+static const char GetImageTransformParameterivHP_names[] =
     "iip\0" /* Parameter signature */
     "glGetImageTransformParameterivHP\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4ivMESA_names[] = 
+static const char WindowPos4ivMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos4ivMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3svARB_names[] = 
+static const char MultiTexCoord3svARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord3sv\0"
     "glMultiTexCoord3svARB\0"
@@ -1793,7 +1843,7 @@ static const char MultiTexCoord3svARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4iARB_names[] = 
+static const char MultiTexCoord4iARB_names[] =
     "iiiii\0" /* Parameter signature */
     "glMultiTexCoord4i\0"
     "glMultiTexCoord4iARB\0"
@@ -1801,21 +1851,21 @@ static const char MultiTexCoord4iARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3ivEXT_names[] = 
+static const char Binormal3ivEXT_names[] =
     "p\0" /* Parameter signature */
     "glBinormal3ivEXT\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_resize_buffers)
-static const char ResizeBuffersMESA_names[] = 
+static const char ResizeBuffersMESA_names[] =
     "\0" /* Parameter signature */
     "glResizeBuffersMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char GetUniformivARB_names[] = 
+static const char GetUniformivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetUniformiv\0"
     "glGetUniformivARB\0"
@@ -1823,28 +1873,28 @@ static const char GetUniformivARB_names[] =
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char PixelTexGenParameteriSGIS_names[] = 
+static const char PixelTexGenParameteriSGIS_names[] =
     "ii\0" /* Parameter signature */
     "glPixelTexGenParameteriSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_INTEL_parallel_arrays)
-static const char VertexPointervINTEL_names[] = 
+static const char VertexPointervINTEL_names[] =
     "iip\0" /* Parameter signature */
     "glVertexPointervINTEL\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor4fNormal3fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiColor4fNormal3fVertex3fvSUN_names[] =
     "pppp\0" /* Parameter signature */
     "glReplacementCodeuiColor4fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3uiEXT_names[] = 
+static const char SecondaryColor3uiEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3ui\0"
     "glSecondaryColor3uiEXT\0"
@@ -1852,14 +1902,14 @@ static const char SecondaryColor3uiEXT_names[] =
 #endif
 
 #if defined(need_GL_SGIX_instruments)
-static const char StartInstrumentsSGIX_names[] = 
+static const char StartInstrumentsSGIX_names[] =
     "\0" /* Parameter signature */
     "glStartInstrumentsSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3usvEXT_names[] = 
+static const char SecondaryColor3usvEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3usv\0"
     "glSecondaryColor3usvEXT\0"
@@ -1867,49 +1917,49 @@ static const char SecondaryColor3usvEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2fvNV_names[] = 
+static const char VertexAttrib2fvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char ProgramLocalParameter4dvARB_names[] = 
+static const char ProgramLocalParameter4dvARB_names[] =
     "iip\0" /* Parameter signature */
     "glProgramLocalParameter4dvARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_matrix_palette)
-static const char MatrixIndexuivARB_names[] = 
+static const char MatrixIndexuivARB_names[] =
     "ip\0" /* Parameter signature */
     "glMatrixIndexuivARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object)
-static const char RenderbufferStorageMultisample_names[] = 
+static const char RenderbufferStorageMultisample_names[] =
     "iiiii\0" /* Parameter signature */
     "glRenderbufferStorageMultisample\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3sEXT_names[] = 
+static const char Tangent3sEXT_names[] =
     "iii\0" /* Parameter signature */
     "glTangent3sEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactorfSUN_names[] = 
+static const char GlobalAlphaFactorfSUN_names[] =
     "f\0" /* Parameter signature */
     "glGlobalAlphaFactorfSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3iARB_names[] = 
+static const char MultiTexCoord3iARB_names[] =
     "iiii\0" /* Parameter signature */
     "glMultiTexCoord3i\0"
     "glMultiTexCoord3iARB\0"
@@ -1917,35 +1967,35 @@ static const char MultiTexCoord3iARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char IsProgram_names[] = 
+static const char IsProgram_names[] =
     "i\0" /* Parameter signature */
     "glIsProgram\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char TexCoordPointerListIBM_names[] = 
+static const char TexCoordPointerListIBM_names[] =
     "iiipi\0" /* Parameter signature */
     "glTexCoordPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactorusSUN_names[] = 
+static const char GlobalAlphaFactorusSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactorusSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2dvNV_names[] = 
+static const char VertexAttrib2dvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char FramebufferRenderbufferEXT_names[] = 
+static const char FramebufferRenderbufferEXT_names[] =
     "iiii\0" /* Parameter signature */
     "glFramebufferRenderbuffer\0"
     "glFramebufferRenderbufferEXT\0"
@@ -1953,14 +2003,14 @@ static const char FramebufferRenderbufferEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1dvNV_names[] = 
+static const char VertexAttrib1dvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char GenTextures_names[] = 
+static const char GenTextures_names[] =
     "ip\0" /* Parameter signature */
     "glGenTextures\0"
     "glGenTexturesEXT\0"
@@ -1968,14 +2018,14 @@ static const char GenTextures_names[] =
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char SetFenceNV_names[] = 
+static const char SetFenceNV_names[] =
     "ii\0" /* Parameter signature */
     "glSetFenceNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char FramebufferTexture1DEXT_names[] = 
+static const char FramebufferTexture1DEXT_names[] =
     "iiiii\0" /* Parameter signature */
     "glFramebufferTexture1D\0"
     "glFramebufferTexture1DEXT\0"
@@ -1983,49 +2033,49 @@ static const char FramebufferTexture1DEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetCombinerOutputParameterivNV_names[] = 
+static const char GetCombinerOutputParameterivNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetCombinerOutputParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char PixelTexGenParameterivSGIS_names[] = 
+static const char PixelTexGenParameterivSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glPixelTexGenParameterivSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_texture_perturb_normal)
-static const char TextureNormalEXT_names[] = 
+static const char TextureNormalEXT_names[] =
     "i\0" /* Parameter signature */
     "glTextureNormalEXT\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char IndexPointerListIBM_names[] = 
+static const char IndexPointerListIBM_names[] =
     "iipi\0" /* Parameter signature */
     "glIndexPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightfvARB_names[] = 
+static const char WeightfvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightfvARB\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4fMESA_names[] = 
+static const char WindowPos4fMESA_names[] =
     "ffff\0" /* Parameter signature */
     "glWindowPos4fMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3dvMESA_names[] = 
+static const char WindowPos3dvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos3dv\0"
     "glWindowPos3dvARB\0"
@@ -2034,14 +2084,14 @@ static const char WindowPos3dvMESA_names[] =
 #endif
 
 #if defined(need_GL_EXT_timer_query)
-static const char GetQueryObjecti64vEXT_names[] = 
+static const char GetQueryObjecti64vEXT_names[] =
     "iip\0" /* Parameter signature */
     "glGetQueryObjecti64vEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1dARB_names[] = 
+static const char MultiTexCoord1dARB_names[] =
     "id\0" /* Parameter signature */
     "glMultiTexCoord1d\0"
     "glMultiTexCoord1dARB\0"
@@ -2049,7 +2099,7 @@ static const char MultiTexCoord1dARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_NV_point_sprite)
-static const char PointParameterivNV_names[] = 
+static const char PointParameterivNV_names[] =
     "ip\0" /* Parameter signature */
     "glPointParameteriv\0"
     "glPointParameterivNV\0"
@@ -2057,15 +2107,22 @@ static const char PointParameterivNV_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform2fvARB_names[] = 
+static const char Uniform2fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform2fv\0"
     "glUniform2fvARB\0"
     "";
 #endif
 
+#if defined(need_GL_APPLE_flush_buffer_range)
+static const char BufferParameteriAPPLE_names[] =
+    "iii\0" /* Parameter signature */
+    "glBufferParameteriAPPLE\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3dvARB_names[] = 
+static const char MultiTexCoord3dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord3dv\0"
     "glMultiTexCoord3dvARB\0"
@@ -2073,42 +2130,49 @@ static const char MultiTexCoord3dvARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN_names[] =
     "pppp\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char DeleteObjectARB_names[] = 
+static const char DeleteObjectARB_names[] =
     "i\0" /* Parameter signature */
     "glDeleteObjectARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_matrix_palette)
-static const char MatrixIndexPointerARB_names[] = 
+static const char MatrixIndexPointerARB_names[] =
     "iiip\0" /* Parameter signature */
     "glMatrixIndexPointerARB\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char ProgramNamedParameter4dvNV_names[] = 
+static const char ProgramNamedParameter4dvNV_names[] =
     "iipp\0" /* Parameter signature */
     "glProgramNamedParameter4dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3fvEXT_names[] = 
+static const char Tangent3fvEXT_names[] =
     "p\0" /* Parameter signature */
     "glTangent3fvEXT\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_vertex_array_object)
+static const char GenVertexArrays_names[] =
+    "ip\0" /* Parameter signature */
+    "glGenVertexArrays\0"
+    "";
+#endif
+
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char BindFramebufferEXT_names[] = 
+static const char BindFramebufferEXT_names[] =
     "ii\0" /* Parameter signature */
     "glBindFramebuffer\0"
     "glBindFramebufferEXT\0"
@@ -2116,14 +2180,14 @@ static const char BindFramebufferEXT_names[] =
 #endif
 
 #if defined(need_GL_SGIX_reference_plane)
-static const char ReferencePlaneSGIX_names[] = 
+static const char ReferencePlaneSGIX_names[] =
     "p\0" /* Parameter signature */
     "glReferencePlaneSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char ValidateProgramARB_names[] = 
+static const char ValidateProgramARB_names[] =
     "i\0" /* Parameter signature */
     "glValidateProgram\0"
     "glValidateProgramARB\0"
@@ -2131,21 +2195,21 @@ static const char ValidateProgramARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_compiled_vertex_array)
-static const char UnlockArraysEXT_names[] = 
+static const char UnlockArraysEXT_names[] =
     "\0" /* Parameter signature */
     "glUnlockArraysEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor3fVertex3fSUN_names[] = 
+static const char TexCoord2fColor3fVertex3fSUN_names[] =
     "ffffffff\0" /* Parameter signature */
     "glTexCoord2fColor3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3fvMESA_names[] = 
+static const char WindowPos3fvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos3fv\0"
     "glWindowPos3fvARB\0"
@@ -2154,14 +2218,14 @@ static const char WindowPos3fvMESA_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1svNV_names[] = 
+static const char VertexAttrib1svNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1svNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_copy_texture)
-static const char CopyTexSubImage3D_names[] = 
+static const char CopyTexSubImage3D_names[] =
     "iiiiiiiii\0" /* Parameter signature */
     "glCopyTexSubImage3D\0"
     "glCopyTexSubImage3DEXT\0"
@@ -2169,22 +2233,29 @@ static const char CopyTexSubImage3D_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2dARB_names[] = 
+static const char VertexAttrib2dARB_names[] =
     "idd\0" /* Parameter signature */
     "glVertexAttrib2d\0"
     "glVertexAttrib2dARB\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char GetInteger64v_names[] =
+    "ip\0" /* Parameter signature */
+    "glGetInteger64v\0"
+    "";
+#endif
+
 #if defined(need_GL_SGIS_texture_color_mask)
-static const char TextureColorMaskSGIS_names[] = 
+static const char TextureColorMaskSGIS_names[] =
     "iiii\0" /* Parameter signature */
     "glTextureColorMaskSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture)
-static const char GetColorTable_names[] = 
+static const char GetColorTable_names[] =
     "iiip\0" /* Parameter signature */
     "glGetColorTable\0"
     "glGetColorTableSGI\0"
@@ -2193,7 +2264,7 @@ static const char GetColorTable_names[] =
 #endif
 
 #if defined(need_GL_SGI_color_table)
-static const char CopyColorTable_names[] = 
+static const char CopyColorTable_names[] =
     "iiiii\0" /* Parameter signature */
     "glCopyColorTable\0"
     "glCopyColorTableSGI\0"
@@ -2201,7 +2272,7 @@ static const char CopyColorTable_names[] =
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetHistogramParameterfv_names[] = 
+static const char GetHistogramParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glGetHistogramParameterfv\0"
     "glGetHistogramParameterfvEXT\0"
@@ -2209,21 +2280,21 @@ static const char GetHistogramParameterfv_names[] =
 #endif
 
 #if defined(need_GL_INTEL_parallel_arrays)
-static const char ColorPointervINTEL_names[] = 
+static const char ColorPointervINTEL_names[] =
     "iip\0" /* Parameter signature */
     "glColorPointervINTEL\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char AlphaFragmentOp1ATI_names[] = 
+static const char AlphaFragmentOp1ATI_names[] =
     "iiiiii\0" /* Parameter signature */
     "glAlphaFragmentOp1ATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3ivARB_names[] = 
+static const char MultiTexCoord3ivARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord3iv\0"
     "glMultiTexCoord3ivARB\0"
@@ -2231,7 +2302,7 @@ static const char MultiTexCoord3ivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2sARB_names[] = 
+static const char MultiTexCoord2sARB_names[] =
     "iii\0" /* Parameter signature */
     "glMultiTexCoord2s\0"
     "glMultiTexCoord2sARB\0"
@@ -2239,7 +2310,7 @@ static const char MultiTexCoord2sARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1dvARB_names[] = 
+static const char VertexAttrib1dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1dv\0"
     "glVertexAttrib1dvARB\0"
@@ -2247,7 +2318,7 @@ static const char VertexAttrib1dvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char DeleteTextures_names[] = 
+static const char DeleteTextures_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteTextures\0"
     "glDeleteTexturesEXT\0"
@@ -2255,49 +2326,49 @@ static const char DeleteTextures_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char TexCoordPointerEXT_names[] = 
+static const char TexCoordPointerEXT_names[] =
     "iiiip\0" /* Parameter signature */
     "glTexCoordPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_texture4D)
-static const char TexSubImage4DSGIS_names[] = 
+static const char TexSubImage4DSGIS_names[] =
     "iiiiiiiiiiiip\0" /* Parameter signature */
     "glTexSubImage4DSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners2)
-static const char CombinerStageParameterfvNV_names[] = 
+static const char CombinerStageParameterfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glCombinerStageParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_instruments)
-static const char StopInstrumentsSGIX_names[] = 
+static const char StopInstrumentsSGIX_names[] =
     "i\0" /* Parameter signature */
     "glStopInstrumentsSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord4fColor4fNormal3fVertex4fSUN_names[] = 
+static const char TexCoord4fColor4fNormal3fVertex4fSUN_names[] =
     "fffffffffffffff\0" /* Parameter signature */
     "glTexCoord4fColor4fNormal3fVertex4fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_polynomial_ffd)
-static const char DeformSGIX_names[] = 
+static const char DeformSGIX_names[] =
     "i\0" /* Parameter signature */
     "glDeformSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char GetVertexAttribfvARB_names[] = 
+static const char GetVertexAttribfvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribfv\0"
     "glGetVertexAttribfvARB\0"
@@ -2305,7 +2376,7 @@ static const char GetVertexAttribfvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3ivEXT_names[] = 
+static const char SecondaryColor3ivEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3iv\0"
     "glSecondaryColor3ivEXT\0"
@@ -2313,49 +2384,49 @@ static const char SecondaryColor3ivEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix4x2fv_names[] = 
+static const char UniformMatrix4x2fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix4x2fv\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_detail_texture)
-static const char GetDetailTexFuncSGIS_names[] = 
+static const char GetDetailTexFuncSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glGetDetailTexFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners2)
-static const char GetCombinerStageParameterfvNV_names[] = 
+static const char GetCombinerStageParameterfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetCombinerStageParameterfvNV\0"
     "";
 #endif
 
-#if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3fEXT_names[] = 
-    "fff\0" /* Parameter signature */
-    "glBinormal3fEXT\0"
+#if defined(need_GL_ARB_vertex_array_object)
+static const char BindVertexArray_names[] =
+    "i\0" /* Parameter signature */
+    "glBindVertexArray\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4ubVertex2fvSUN_names[] = 
+static const char Color4ubVertex2fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glColor4ubVertex2fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_texture_filter4)
-static const char TexFilterFuncSGIS_names[] = 
+static const char TexFilterFuncSGIS_names[] =
     "iiip\0" /* Parameter signature */
     "glTexFilterFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_multisample) || defined(need_GL_EXT_multisample)
-static const char SampleMaskSGIS_names[] = 
+static const char SampleMaskSGIS_names[] =
     "fi\0" /* Parameter signature */
     "glSampleMaskSGIS\0"
     "glSampleMaskEXT\0"
@@ -2363,7 +2434,7 @@ static const char SampleMaskSGIS_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader)
-static const char GetAttribLocationARB_names[] = 
+static const char GetAttribLocationARB_names[] =
     "ip\0" /* Parameter signature */
     "glGetAttribLocation\0"
     "glGetAttribLocationARB\0"
@@ -2371,7 +2442,7 @@ static const char GetAttribLocationARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4ubvARB_names[] = 
+static const char VertexAttrib4ubvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4ubv\0"
     "glVertexAttrib4ubvARB\0"
@@ -2379,21 +2450,21 @@ static const char VertexAttrib4ubvARB_names[] =
 #endif
 
 #if defined(need_GL_SGIS_detail_texture)
-static const char DetailTexFuncSGIS_names[] = 
+static const char DetailTexFuncSGIS_names[] =
     "iip\0" /* Parameter signature */
     "glDetailTexFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Normal3fVertex3fSUN_names[] = 
+static const char Normal3fVertex3fSUN_names[] =
     "ffffff\0" /* Parameter signature */
     "glNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_copy_texture)
-static const char CopyTexImage2D_names[] = 
+static const char CopyTexImage2D_names[] =
     "iiiiiiii\0" /* Parameter signature */
     "glCopyTexImage2D\0"
     "glCopyTexImage2DEXT\0"
@@ -2401,7 +2472,7 @@ static const char CopyTexImage2D_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char GetBufferPointervARB_names[] = 
+static const char GetBufferPointervARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetBufferPointerv\0"
     "glGetBufferPointervARB\0"
@@ -2409,7 +2480,7 @@ static const char GetBufferPointervARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char ProgramEnvParameter4fARB_names[] = 
+static const char ProgramEnvParameter4fARB_names[] =
     "iiffff\0" /* Parameter signature */
     "glProgramEnvParameter4fARB\0"
     "glProgramParameter4fNV\0"
@@ -2417,7 +2488,7 @@ static const char ProgramEnvParameter4fARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform3ivARB_names[] = 
+static const char Uniform3ivARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform3iv\0"
     "glUniform3ivARB\0"
@@ -2425,21 +2496,21 @@ static const char Uniform3ivARB_names[] =
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char GetFenceivNV_names[] = 
+static const char GetFenceivNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetFenceivNV\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4dvMESA_names[] = 
+static const char WindowPos4dvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos4dvMESA\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_color_subtable)
-static const char ColorSubTable_names[] = 
+static const char ColorSubTable_names[] =
     "iiiiip\0" /* Parameter signature */
     "glColorSubTable\0"
     "glColorSubTableEXT\0"
@@ -2447,7 +2518,7 @@ static const char ColorSubTable_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4ivARB_names[] = 
+static const char MultiTexCoord4ivARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord4iv\0"
     "glMultiTexCoord4ivARB\0"
@@ -2455,21 +2526,21 @@ static const char MultiTexCoord4ivARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_gpu_program_parameters)
-static const char ProgramLocalParameters4fvEXT_names[] = 
+static const char ProgramLocalParameters4fvEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glProgramLocalParameters4fvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char GetMapAttribParameterfvNV_names[] = 
+static const char GetMapAttribParameterfvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetMapAttribParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4sARB_names[] = 
+static const char VertexAttrib4sARB_names[] =
     "iiiii\0" /* Parameter signature */
     "glVertexAttrib4s\0"
     "glVertexAttrib4sARB\0"
@@ -2477,7 +2548,7 @@ static const char VertexAttrib4sARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char GetQueryObjectuivARB_names[] = 
+static const char GetQueryObjectuivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetQueryObjectuiv\0"
     "glGetQueryObjectuivARB\0"
@@ -2485,14 +2556,14 @@ static const char GetQueryObjectuivARB_names[] =
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char MapParameterivNV_names[] = 
+static const char MapParameterivNV_names[] =
     "iip\0" /* Parameter signature */
     "glMapParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char GenRenderbuffersEXT_names[] = 
+static const char GenRenderbuffersEXT_names[] =
     "ip\0" /* Parameter signature */
     "glGenRenderbuffers\0"
     "glGenRenderbuffersEXT\0"
@@ -2500,7 +2571,7 @@ static const char GenRenderbuffersEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2dvARB_names[] = 
+static const char VertexAttrib2dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2dv\0"
     "glVertexAttrib2dvARB\0"
@@ -2508,28 +2579,28 @@ static const char VertexAttrib2dvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char EdgeFlagPointerEXT_names[] = 
+static const char EdgeFlagPointerEXT_names[] =
     "iip\0" /* Parameter signature */
     "glEdgeFlagPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs2svNV_names[] = 
+static const char VertexAttribs2svNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs2svNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightbvARB_names[] = 
+static const char WeightbvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightbvARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2fvARB_names[] = 
+static const char VertexAttrib2fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2fv\0"
     "glVertexAttrib2fvARB\0"
@@ -2537,7 +2608,7 @@ static const char VertexAttrib2fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char GetBufferParameterivARB_names[] = 
+static const char GetBufferParameterivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetBufferParameteriv\0"
     "glGetBufferParameterivARB\0"
@@ -2545,28 +2616,28 @@ static const char GetBufferParameterivARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char ListParameteriSGIX_names[] = 
+static const char ListParameteriSGIX_names[] =
     "iii\0" /* Parameter signature */
     "glListParameteriSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor4fNormal3fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiColor4fNormal3fVertex3fSUN_names[] =
     "iffffffffff\0" /* Parameter signature */
     "glReplacementCodeuiColor4fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_instruments)
-static const char InstrumentsBufferSGIX_names[] = 
+static const char InstrumentsBufferSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glInstrumentsBufferSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NivARB_names[] = 
+static const char VertexAttrib4NivARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Niv\0"
     "glVertexAttrib4NivARB\0"
@@ -2574,35 +2645,35 @@ static const char VertexAttrib4NivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char GetAttachedShaders_names[] = 
+static const char GetAttachedShaders_names[] =
     "iipp\0" /* Parameter signature */
     "glGetAttachedShaders\0"
     "";
 #endif
 
 #if defined(need_GL_APPLE_vertex_array_object)
-static const char GenVertexArraysAPPLE_names[] = 
+static const char GenVertexArraysAPPLE_names[] =
     "ip\0" /* Parameter signature */
     "glGenVertexArraysAPPLE\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_gpu_program_parameters)
-static const char ProgramEnvParameters4fvEXT_names[] = 
+static const char ProgramEnvParameters4fvEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glProgramEnvParameters4fvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor4fNormal3fVertex3fvSUN_names[] = 
+static const char TexCoord2fColor4fNormal3fVertex3fvSUN_names[] =
     "pppp\0" /* Parameter signature */
     "glTexCoord2fColor4fNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2iMESA_names[] = 
+static const char WindowPos2iMESA_names[] =
     "ii\0" /* Parameter signature */
     "glWindowPos2i\0"
     "glWindowPos2iARB\0"
@@ -2611,7 +2682,7 @@ static const char WindowPos2iMESA_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3fvEXT_names[] = 
+static const char SecondaryColor3fvEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3fv\0"
     "glSecondaryColor3fvEXT\0"
@@ -2619,7 +2690,7 @@ static const char SecondaryColor3fvEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexSubImage1DARB_names[] = 
+static const char CompressedTexSubImage1DARB_names[] =
     "iiiiiip\0" /* Parameter signature */
     "glCompressedTexSubImage1D\0"
     "glCompressedTexSubImage1DARB\0"
@@ -2627,28 +2698,28 @@ static const char CompressedTexSubImage1DARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetVertexAttribivNV_names[] = 
+static const char GetVertexAttribivNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribivNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramStringARB_names[] = 
+static const char GetProgramStringARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramStringARB\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_envmap_bumpmap)
-static const char TexBumpParameterfvATI_names[] = 
+static const char TexBumpParameterfvATI_names[] =
     "ip\0" /* Parameter signature */
     "glTexBumpParameterfvATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char CompileShaderARB_names[] = 
+static const char CompileShaderARB_names[] =
     "i\0" /* Parameter signature */
     "glCompileShader\0"
     "glCompileShaderARB\0"
@@ -2656,14 +2727,14 @@ static const char CompileShaderARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char DeleteShader_names[] = 
+static const char DeleteShader_names[] =
     "i\0" /* Parameter signature */
     "glDeleteShader\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform3fARB_names[] = 
+static const char Uniform3fARB_names[] =
     "ifff\0" /* Parameter signature */
     "glUniform3f\0"
     "glUniform3fARB\0"
@@ -2671,28 +2742,28 @@ static const char Uniform3fARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char ListParameterfvSGIX_names[] = 
+static const char ListParameterfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glListParameterfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3dvEXT_names[] = 
+static const char Tangent3dvEXT_names[] =
     "p\0" /* Parameter signature */
     "glTangent3dvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetVertexAttribfvNV_names[] = 
+static const char GetVertexAttribfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3sMESA_names[] = 
+static const char WindowPos3sMESA_names[] =
     "iii\0" /* Parameter signature */
     "glWindowPos3s\0"
     "glWindowPos3sARB\0"
@@ -2701,35 +2772,35 @@ static const char WindowPos3sMESA_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2svNV_names[] = 
+static const char VertexAttrib2svNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2svNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs1fvNV_names[] = 
+static const char VertexAttribs1fvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs1fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fVertex3fvSUN_names[] = 
+static const char TexCoord2fVertex3fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glTexCoord2fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4sMESA_names[] = 
+static const char WindowPos4sMESA_names[] =
     "iiii\0" /* Parameter signature */
     "glWindowPos4sMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NuivARB_names[] = 
+static const char VertexAttrib4NuivARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Nuiv\0"
     "glVertexAttrib4NuivARB\0"
@@ -2737,7 +2808,7 @@ static const char VertexAttrib4NuivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char ClientActiveTextureARB_names[] = 
+static const char ClientActiveTextureARB_names[] =
     "i\0" /* Parameter signature */
     "glClientActiveTexture\0"
     "glClientActiveTextureARB\0"
@@ -2745,51 +2816,58 @@ static const char ClientActiveTextureARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_pixel_texture)
-static const char PixelTexGenSGIX_names[] = 
+static const char PixelTexGenSGIX_names[] =
     "i\0" /* Parameter signature */
     "glPixelTexGenSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeusvSUN_names[] = 
+static const char ReplacementCodeusvSUN_names[] =
     "p\0" /* Parameter signature */
     "glReplacementCodeusvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform4fARB_names[] = 
+static const char Uniform4fARB_names[] =
     "iffff\0" /* Parameter signature */
     "glUniform4f\0"
     "glUniform4fARB\0"
     "";
 #endif
 
-#if defined(need_GL_IBM_multimode_draw_arrays)
-static const char MultiModeDrawArraysIBM_names[] = 
-    "pppii\0" /* Parameter signature */
-    "glMultiModeDrawArraysIBM\0"
+#if defined(need_GL_ARB_map_buffer_range)
+static const char FlushMappedBufferRange_names[] =
+    "iii\0" /* Parameter signature */
+    "glFlushMappedBufferRange\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char IsProgramNV_names[] = 
+static const char IsProgramNV_names[] =
     "i\0" /* Parameter signature */
     "glIsProgramARB\0"
     "glIsProgramNV\0"
     "";
 #endif
 
+#if defined(need_GL_APPLE_flush_buffer_range)
+static const char FlushMappedBufferRangeAPPLE_names[] =
+    "iii\0" /* Parameter signature */
+    "glFlushMappedBufferRangeAPPLE\0"
+    "";
+#endif
+
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodePointerSUN_names[] = 
+static const char ReplacementCodePointerSUN_names[] =
     "iip\0" /* Parameter signature */
     "glReplacementCodePointerSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char ProgramEnvParameter4dARB_names[] = 
+static const char ProgramEnvParameter4dARB_names[] =
     "iidddd\0" /* Parameter signature */
     "glProgramEnvParameter4dARB\0"
     "glProgramParameter4dNV\0"
@@ -2797,7 +2875,7 @@ static const char ProgramEnvParameter4dARB_names[] =
 #endif
 
 #if defined(need_GL_SGI_color_table)
-static const char ColorTableParameterfv_names[] = 
+static const char ColorTableParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glColorTableParameterfv\0"
     "glColorTableParameterfvSGI\0"
@@ -2805,21 +2883,21 @@ static const char ColorTableParameterfv_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightModelfSGIX_names[] = 
+static const char FragmentLightModelfSGIX_names[] =
     "if\0" /* Parameter signature */
     "glFragmentLightModelfSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3bvEXT_names[] = 
+static const char Binormal3bvEXT_names[] =
     "p\0" /* Parameter signature */
     "glBinormal3bvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char IsTexture_names[] = 
+static const char IsTexture_names[] =
     "i\0" /* Parameter signature */
     "glIsTexture\0"
     "glIsTextureEXT\0"
@@ -2827,14 +2905,14 @@ static const char IsTexture_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_weighting)
-static const char VertexWeightfvEXT_names[] = 
+static const char VertexWeightfvEXT_names[] =
     "p\0" /* Parameter signature */
     "glVertexWeightfvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1dARB_names[] = 
+static const char VertexAttrib1dARB_names[] =
     "id\0" /* Parameter signature */
     "glVertexAttrib1d\0"
     "glVertexAttrib1dARB\0"
@@ -2842,14 +2920,14 @@ static const char VertexAttrib1dARB_names[] =
 #endif
 
 #if defined(need_GL_HP_image_transform)
-static const char ImageTransformParameterivHP_names[] = 
+static const char ImageTransformParameterivHP_names[] =
     "iip\0" /* Parameter signature */
     "glImageTransformParameterivHP\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char DeleteQueriesARB_names[] = 
+static const char DeleteQueriesARB_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteQueries\0"
     "glDeleteQueriesARB\0"
@@ -2857,28 +2935,28 @@ static const char DeleteQueriesARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4ubVertex2fSUN_names[] = 
+static const char Color4ubVertex2fSUN_names[] =
     "iiiiff\0" /* Parameter signature */
     "glColor4ubVertex2fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentColorMaterialSGIX_names[] = 
+static const char FragmentColorMaterialSGIX_names[] =
     "ii\0" /* Parameter signature */
     "glFragmentColorMaterialSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_matrix_palette)
-static const char CurrentPaletteMatrixARB_names[] = 
+static const char CurrentPaletteMatrixARB_names[] =
     "i\0" /* Parameter signature */
     "glCurrentPaletteMatrixARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_multisample) || defined(need_GL_EXT_multisample)
-static const char SamplePatternSGIS_names[] = 
+static const char SamplePatternSGIS_names[] =
     "i\0" /* Parameter signature */
     "glSamplePatternSGIS\0"
     "glSamplePatternEXT\0"
@@ -2886,7 +2964,7 @@ static const char SamplePatternSGIS_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char IsQueryARB_names[] = 
+static const char IsQueryARB_names[] =
     "i\0" /* Parameter signature */
     "glIsQuery\0"
     "glIsQueryARB\0"
@@ -2894,14 +2972,14 @@ static const char IsQueryARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor4ubVertex3fSUN_names[] = 
+static const char ReplacementCodeuiColor4ubVertex3fSUN_names[] =
     "iiiiifff\0" /* Parameter signature */
     "glReplacementCodeuiColor4ubVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4usvARB_names[] = 
+static const char VertexAttrib4usvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4usv\0"
     "glVertexAttrib4usvARB\0"
@@ -2909,7 +2987,7 @@ static const char VertexAttrib4usvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char LinkProgramARB_names[] = 
+static const char LinkProgramARB_names[] =
     "i\0" /* Parameter signature */
     "glLinkProgram\0"
     "glLinkProgramARB\0"
@@ -2917,14 +2995,14 @@ static const char LinkProgramARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib2fNV_names[] = 
+static const char VertexAttrib2fNV_names[] =
     "iff\0" /* Parameter signature */
     "glVertexAttrib2fNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char ShaderSourceARB_names[] = 
+static const char ShaderSourceARB_names[] =
     "iipp\0" /* Parameter signature */
     "glShaderSource\0"
     "glShaderSourceARB\0"
@@ -2932,14 +3010,14 @@ static const char ShaderSourceARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentMaterialiSGIX_names[] = 
+static const char FragmentMaterialiSGIX_names[] =
     "iii\0" /* Parameter signature */
     "glFragmentMaterialiSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3svARB_names[] = 
+static const char VertexAttrib3svARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3sv\0"
     "glVertexAttrib3svARB\0"
@@ -2947,7 +3025,7 @@ static const char VertexAttrib3svARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexSubImage3DARB_names[] = 
+static const char CompressedTexSubImage3DARB_names[] =
     "iiiiiiiiiip\0" /* Parameter signature */
     "glCompressedTexSubImage3D\0"
     "glCompressedTexSubImage3DARB\0"
@@ -2955,7 +3033,7 @@ static const char CompressedTexSubImage3DARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2ivMESA_names[] = 
+static const char WindowPos2ivMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos2iv\0"
     "glWindowPos2ivARB\0"
@@ -2964,7 +3042,7 @@ static const char WindowPos2ivMESA_names[] =
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char IsFramebufferEXT_names[] = 
+static const char IsFramebufferEXT_names[] =
     "i\0" /* Parameter signature */
     "glIsFramebuffer\0"
     "glIsFramebufferEXT\0"
@@ -2972,7 +3050,7 @@ static const char IsFramebufferEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform4ivARB_names[] = 
+static const char Uniform4ivARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform4iv\0"
     "glUniform4ivARB\0"
@@ -2980,7 +3058,7 @@ static const char Uniform4ivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char GetVertexAttribdvARB_names[] = 
+static const char GetVertexAttribdvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribdv\0"
     "glGetVertexAttribdvARB\0"
@@ -2988,14 +3066,14 @@ static const char GetVertexAttribdvARB_names[] =
 #endif
 
 #if defined(need_GL_ATI_envmap_bumpmap)
-static const char TexBumpParameterivATI_names[] = 
+static const char TexBumpParameterivATI_names[] =
     "ip\0" /* Parameter signature */
     "glTexBumpParameterivATI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char GetSeparableFilter_names[] = 
+static const char GetSeparableFilter_names[] =
     "iiippp\0" /* Parameter signature */
     "glGetSeparableFilter\0"
     "glGetSeparableFilterEXT\0"
@@ -3003,49 +3081,49 @@ static const char GetSeparableFilter_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3dEXT_names[] = 
+static const char Binormal3dEXT_names[] =
     "ddd\0" /* Parameter signature */
     "glBinormal3dEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_sprite)
-static const char SpriteParameteriSGIX_names[] = 
+static const char SpriteParameteriSGIX_names[] =
     "ii\0" /* Parameter signature */
     "glSpriteParameteriSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char RequestResidentProgramsNV_names[] = 
+static const char RequestResidentProgramsNV_names[] =
     "ip\0" /* Parameter signature */
     "glRequestResidentProgramsNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_tag_sample_buffer)
-static const char TagSampleBufferSGIX_names[] = 
+static const char TagSampleBufferSGIX_names[] =
     "\0" /* Parameter signature */
     "glTagSampleBufferSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeusSUN_names[] = 
+static const char ReplacementCodeusSUN_names[] =
     "i\0" /* Parameter signature */
     "glReplacementCodeusSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char ListParameterivSGIX_names[] = 
+static const char ListParameterivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glListParameterivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_multi_draw_arrays)
-static const char MultiDrawElementsEXT_names[] = 
+static const char MultiDrawElementsEXT_names[] =
     "ipipi\0" /* Parameter signature */
     "glMultiDrawElements\0"
     "glMultiDrawElementsEXT\0"
@@ -3053,7 +3131,7 @@ static const char MultiDrawElementsEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform1ivARB_names[] = 
+static const char Uniform1ivARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform1iv\0"
     "glUniform1ivARB\0"
@@ -3061,7 +3139,7 @@ static const char Uniform1ivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2sMESA_names[] = 
+static const char WindowPos2sMESA_names[] =
     "ii\0" /* Parameter signature */
     "glWindowPos2s\0"
     "glWindowPos2sARB\0"
@@ -3070,14 +3148,14 @@ static const char WindowPos2sMESA_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightusvARB_names[] = 
+static const char WeightusvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightusvARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
-static const char FogCoordPointerEXT_names[] = 
+static const char FogCoordPointerEXT_names[] =
     "iip\0" /* Parameter signature */
     "glFogCoordPointer\0"
     "glFogCoordPointerEXT\0"
@@ -3085,14 +3163,14 @@ static const char FogCoordPointerEXT_names[] =
 #endif
 
 #if defined(need_GL_EXT_index_material)
-static const char IndexMaterialEXT_names[] = 
+static const char IndexMaterialEXT_names[] =
     "ii\0" /* Parameter signature */
     "glIndexMaterialEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3ubvEXT_names[] = 
+static const char SecondaryColor3ubvEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3ubv\0"
     "glSecondaryColor3ubvEXT\0"
@@ -3100,7 +3178,7 @@ static const char SecondaryColor3ubvEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4dvARB_names[] = 
+static const char VertexAttrib4dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4dv\0"
     "glVertexAttrib4dvARB\0"
@@ -3108,7 +3186,7 @@ static const char VertexAttrib4dvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader)
-static const char BindAttribLocationARB_names[] = 
+static const char BindAttribLocationARB_names[] =
     "iip\0" /* Parameter signature */
     "glBindAttribLocation\0"
     "glBindAttribLocationARB\0"
@@ -3116,7 +3194,7 @@ static const char BindAttribLocationARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2dARB_names[] = 
+static const char MultiTexCoord2dARB_names[] =
     "idd\0" /* Parameter signature */
     "glMultiTexCoord2d\0"
     "glMultiTexCoord2dARB\0"
@@ -3124,35 +3202,35 @@ static const char MultiTexCoord2dARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char ExecuteProgramNV_names[] = 
+static const char ExecuteProgramNV_names[] =
     "iip\0" /* Parameter signature */
     "glExecuteProgramNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char LightEnviSGIX_names[] = 
+static const char LightEnviSGIX_names[] =
     "ii\0" /* Parameter signature */
     "glLightEnviSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeuiSUN_names[] = 
+static const char ReplacementCodeuiSUN_names[] =
     "i\0" /* Parameter signature */
     "glReplacementCodeuiSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribPointerNV_names[] = 
+static const char VertexAttribPointerNV_names[] =
     "iiiip\0" /* Parameter signature */
     "glVertexAttribPointerNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char GetFramebufferAttachmentParameterivEXT_names[] = 
+static const char GetFramebufferAttachmentParameterivEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glGetFramebufferAttachmentParameteriv\0"
     "glGetFramebufferAttachmentParameterivEXT\0"
@@ -3160,14 +3238,14 @@ static const char GetFramebufferAttachmentParameterivEXT_names[] =
 #endif
 
 #if defined(need_GL_EXT_pixel_transform)
-static const char PixelTransformParameterfEXT_names[] = 
+static const char PixelTransformParameterfEXT_names[] =
     "iif\0" /* Parameter signature */
     "glPixelTransformParameterfEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4dvARB_names[] = 
+static const char MultiTexCoord4dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord4dv\0"
     "glMultiTexCoord4dvARB\0"
@@ -3175,21 +3253,21 @@ static const char MultiTexCoord4dvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_pixel_transform)
-static const char PixelTransformParameteriEXT_names[] = 
+static const char PixelTransformParameteriEXT_names[] =
     "iii\0" /* Parameter signature */
     "glPixelTransformParameteriEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor4ubVertex3fSUN_names[] = 
+static const char TexCoord2fColor4ubVertex3fSUN_names[] =
     "ffiiiifff\0" /* Parameter signature */
     "glTexCoord2fColor4ubVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform1iARB_names[] = 
+static const char Uniform1iARB_names[] =
     "ii\0" /* Parameter signature */
     "glUniform1i\0"
     "glUniform1iARB\0"
@@ -3197,7 +3275,7 @@ static const char Uniform1iARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttribPointerARB_names[] = 
+static const char VertexAttribPointerARB_names[] =
     "iiiiip\0" /* Parameter signature */
     "glVertexAttribPointer\0"
     "glVertexAttribPointerARB\0"
@@ -3205,14 +3283,14 @@ static const char VertexAttribPointerARB_names[] =
 #endif
 
 #if defined(need_GL_SGIS_sharpen_texture)
-static const char SharpenTexFuncSGIS_names[] = 
+static const char SharpenTexFuncSGIS_names[] =
     "iip\0" /* Parameter signature */
     "glSharpenTexFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4fvARB_names[] = 
+static const char MultiTexCoord4fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord4fv\0"
     "glMultiTexCoord4fvARB\0"
@@ -3220,56 +3298,56 @@ static const char MultiTexCoord4fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix2x3fv_names[] = 
+static const char UniformMatrix2x3fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix2x3fv\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char TrackMatrixNV_names[] = 
+static const char TrackMatrixNV_names[] =
     "iiii\0" /* Parameter signature */
     "glTrackMatrixNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerParameteriNV_names[] = 
+static const char CombinerParameteriNV_names[] =
     "ii\0" /* Parameter signature */
     "glCombinerParameteriNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char DeleteAsyncMarkersSGIX_names[] = 
+static const char DeleteAsyncMarkersSGIX_names[] =
     "ii\0" /* Parameter signature */
     "glDeleteAsyncMarkersSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char IsAsyncMarkerSGIX_names[] = 
+static const char IsAsyncMarkerSGIX_names[] =
     "i\0" /* Parameter signature */
     "glIsAsyncMarkerSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_framezoom)
-static const char FrameZoomSGIX_names[] = 
+static const char FrameZoomSGIX_names[] =
     "i\0" /* Parameter signature */
     "glFrameZoomSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Normal3fVertex3fvSUN_names[] = 
+static const char Normal3fVertex3fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NsvARB_names[] = 
+static const char VertexAttrib4NsvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Nsv\0"
     "glVertexAttrib4NsvARB\0"
@@ -3277,15 +3355,22 @@ static const char VertexAttrib4NsvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3fvARB_names[] = 
+static const char VertexAttrib3fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3fv\0"
     "glVertexAttrib3fvARB\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char GetSynciv_names[] =
+    "iiipp\0" /* Parameter signature */
+    "glGetSynciv\0"
+    "";
+#endif
+
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char DeleteFramebuffersEXT_names[] = 
+static const char DeleteFramebuffersEXT_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteFramebuffers\0"
     "glDeleteFramebuffersEXT\0"
@@ -3293,14 +3378,14 @@ static const char DeleteFramebuffersEXT_names[] =
 #endif
 
 #if defined(need_GL_SUN_global_alpha)
-static const char GlobalAlphaFactorsSUN_names[] = 
+static const char GlobalAlphaFactorsSUN_names[] =
     "i\0" /* Parameter signature */
     "glGlobalAlphaFactorsSUN\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_texture3D)
-static const char TexSubImage3D_names[] = 
+static const char TexSubImage3D_names[] =
     "iiiiiiiiiip\0" /* Parameter signature */
     "glTexSubImage3D\0"
     "glTexSubImage3DEXT\0"
@@ -3308,14 +3393,14 @@ static const char TexSubImage3D_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3fEXT_names[] = 
+static const char Tangent3fEXT_names[] =
     "fff\0" /* Parameter signature */
     "glTangent3fEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3uivEXT_names[] = 
+static const char SecondaryColor3uivEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3uiv\0"
     "glSecondaryColor3uivEXT\0"
@@ -3323,35 +3408,35 @@ static const char SecondaryColor3uivEXT_names[] =
 #endif
 
 #if defined(need_GL_ARB_matrix_palette)
-static const char MatrixIndexubvARB_names[] = 
+static const char MatrixIndexubvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMatrixIndexubvARB\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char Color4fNormal3fVertex3fSUN_names[] = 
+static const char Color4fNormal3fVertex3fSUN_names[] =
     "ffffffffff\0" /* Parameter signature */
     "glColor4fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char PixelTexGenParameterfSGIS_names[] = 
+static const char PixelTexGenParameterfSGIS_names[] =
     "if\0" /* Parameter signature */
     "glPixelTexGenParameterfSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char CreateShader_names[] = 
+static const char CreateShader_names[] =
     "i\0" /* Parameter signature */
     "glCreateShader\0"
     "";
 #endif
 
 #if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture)
-static const char GetColorTableParameterfv_names[] = 
+static const char GetColorTableParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glGetColorTableParameterfv\0"
     "glGetColorTableParameterfvSGI\0"
@@ -3360,14 +3445,14 @@ static const char GetColorTableParameterfv_names[] =
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightModelfvSGIX_names[] = 
+static const char FragmentLightModelfvSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glFragmentLightModelfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord3fARB_names[] = 
+static const char MultiTexCoord3fARB_names[] =
     "ifff\0" /* Parameter signature */
     "glMultiTexCoord3f\0"
     "glMultiTexCoord3fARB\0"
@@ -3375,14 +3460,14 @@ static const char MultiTexCoord3fARB_names[] =
 #endif
 
 #if defined(need_GL_SGIS_pixel_texture)
-static const char GetPixelTexGenParameterfvSGIS_names[] = 
+static const char GetPixelTexGenParameterfvSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glGetPixelTexGenParameterfvSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char GenFramebuffersEXT_names[] = 
+static const char GenFramebuffersEXT_names[] =
     "ip\0" /* Parameter signature */
     "glGenFramebuffers\0"
     "glGenFramebuffersEXT\0"
@@ -3390,35 +3475,36 @@ static const char GenFramebuffersEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetProgramParameterdvNV_names[] = 
+static const char GetProgramParameterdvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetProgramParameterdvNV\0"
     "";
 #endif
 
-#if defined(need_GL_SGIX_instruments)
-static const char PollInstrumentsSGIX_names[] = 
-    "p\0" /* Parameter signature */
-    "glPollInstrumentsSGIX\0"
+#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object)
+static const char IsVertexArrayAPPLE_names[] =
+    "i\0" /* Parameter signature */
+    "glIsVertexArray\0"
+    "glIsVertexArrayAPPLE\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightfvSGIX_names[] = 
+static const char FragmentLightfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glFragmentLightfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char DetachShader_names[] = 
+static const char DetachShader_names[] =
     "ii\0" /* Parameter signature */
     "glDetachShader\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NubARB_names[] = 
+static const char VertexAttrib4NubARB_names[] =
     "iiiii\0" /* Parameter signature */
     "glVertexAttrib4Nub\0"
     "glVertexAttrib4NubARB\0"
@@ -3426,28 +3512,28 @@ static const char VertexAttrib4NubARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramEnvParameterfvARB_names[] = 
+static const char GetProgramEnvParameterfvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramEnvParameterfvARB\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetTrackMatrixivNV_names[] = 
+static const char GetTrackMatrixivNV_names[] =
     "iiip\0" /* Parameter signature */
     "glGetTrackMatrixivNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3svNV_names[] = 
+static const char VertexAttrib3svNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3svNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform4fvARB_names[] = 
+static const char Uniform4fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform4fv\0"
     "glUniform4fvARB\0"
@@ -3455,7 +3541,7 @@ static const char Uniform4fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix)
-static const char MultTransposeMatrixfARB_names[] = 
+static const char MultTransposeMatrixfARB_names[] =
     "p\0" /* Parameter signature */
     "glMultTransposeMatrixf\0"
     "glMultTransposeMatrixfARB\0"
@@ -3463,14 +3549,14 @@ static const char MultTransposeMatrixfARB_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char ColorFragmentOp1ATI_names[] = 
+static const char ColorFragmentOp1ATI_names[] =
     "iiiiiii\0" /* Parameter signature */
     "glColorFragmentOp1ATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char GetUniformfvARB_names[] = 
+static const char GetUniformfvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetUniformfv\0"
     "glGetUniformfvARB\0"
@@ -3478,28 +3564,28 @@ static const char GetUniformfvARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN_names[] =
     "iffffffffffff\0" /* Parameter signature */
     "glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char DetachObjectARB_names[] = 
+static const char DetachObjectARB_names[] =
     "ii\0" /* Parameter signature */
     "glDetachObjectARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char VertexBlendARB_names[] = 
+static const char VertexBlendARB_names[] =
     "i\0" /* Parameter signature */
     "glVertexBlendARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3iMESA_names[] = 
+static const char WindowPos3iMESA_names[] =
     "iii\0" /* Parameter signature */
     "glWindowPos3i\0"
     "glWindowPos3iARB\0"
@@ -3508,7 +3594,7 @@ static const char WindowPos3iMESA_names[] =
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char SeparableFilter2D_names[] = 
+static const char SeparableFilter2D_names[] =
     "iiiiiipp\0" /* Parameter signature */
     "glSeparableFilter2D\0"
     "glSeparableFilter2DEXT\0"
@@ -3516,14 +3602,14 @@ static const char SeparableFilter2D_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor4ubVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiColor4ubVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glReplacementCodeuiColor4ubVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char CompressedTexImage2DARB_names[] = 
+static const char CompressedTexImage2DARB_names[] =
     "iiiiiiip\0" /* Parameter signature */
     "glCompressedTexImage2D\0"
     "glCompressedTexImage2DARB\0"
@@ -3531,7 +3617,7 @@ static const char CompressedTexImage2DARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char ArrayElement_names[] = 
+static const char ArrayElement_names[] =
     "i\0" /* Parameter signature */
     "glArrayElement\0"
     "glArrayElementEXT\0"
@@ -3539,35 +3625,35 @@ static const char ArrayElement_names[] =
 #endif
 
 #if defined(need_GL_EXT_depth_bounds_test)
-static const char DepthBoundsEXT_names[] = 
+static const char DepthBoundsEXT_names[] =
     "dd\0" /* Parameter signature */
     "glDepthBoundsEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char ProgramParameters4fvNV_names[] = 
+static const char ProgramParameters4fvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glProgramParameters4fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_polynomial_ffd)
-static const char DeformationMap3fSGIX_names[] = 
+static const char DeformationMap3fSGIX_names[] =
     "iffiiffiiffiip\0" /* Parameter signature */
     "glDeformationMap3fSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetProgramivNV_names[] = 
+static const char GetProgramivNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramivNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetMinmaxParameteriv_names[] = 
+static const char GetMinmaxParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glGetMinmaxParameteriv\0"
     "glGetMinmaxParameterivEXT\0"
@@ -3575,7 +3661,7 @@ static const char GetMinmaxParameteriv_names[] =
 #endif
 
 #if defined(need_GL_EXT_copy_texture)
-static const char CopyTexImage1D_names[] = 
+static const char CopyTexImage1D_names[] =
     "iiiiiii\0" /* Parameter signature */
     "glCopyTexImage1D\0"
     "glCopyTexImage1DEXT\0"
@@ -3583,42 +3669,42 @@ static const char CopyTexImage1D_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char AlphaFragmentOp3ATI_names[] = 
+static const char AlphaFragmentOp3ATI_names[] =
     "iiiiiiiiiiii\0" /* Parameter signature */
     "glAlphaFragmentOp3ATI\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetVertexAttribdvNV_names[] = 
+static const char GetVertexAttribdvNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribdvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3fvNV_names[] = 
+static const char VertexAttrib3fvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetFinalCombinerInputParameterivNV_names[] = 
+static const char GetFinalCombinerInputParameterivNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetFinalCombinerInputParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char GetMapParameterivNV_names[] = 
+static const char GetMapParameterivNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetMapParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform4iARB_names[] = 
+static const char Uniform4iARB_names[] =
     "iiiii\0" /* Parameter signature */
     "glUniform4i\0"
     "glUniform4iARB\0"
@@ -3626,7 +3712,7 @@ static const char Uniform4iARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionParameteri_names[] = 
+static const char ConvolutionParameteri_names[] =
     "iii\0" /* Parameter signature */
     "glConvolutionParameteri\0"
     "glConvolutionParameteriEXT\0"
@@ -3634,14 +3720,14 @@ static const char ConvolutionParameteri_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3sEXT_names[] = 
+static const char Binormal3sEXT_names[] =
     "iii\0" /* Parameter signature */
     "glBinormal3sEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char ConvolutionParameterf_names[] = 
+static const char ConvolutionParameterf_names[] =
     "iif\0" /* Parameter signature */
     "glConvolutionParameterf\0"
     "glConvolutionParameterfEXT\0"
@@ -3649,7 +3735,7 @@ static const char ConvolutionParameterf_names[] =
 #endif
 
 #if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture)
-static const char GetColorTableParameteriv_names[] = 
+static const char GetColorTableParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glGetColorTableParameteriv\0"
     "glGetColorTableParameterivSGI\0"
@@ -3658,7 +3744,7 @@ static const char GetColorTableParameteriv_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char ProgramEnvParameter4dvARB_names[] = 
+static const char ProgramEnvParameter4dvARB_names[] =
     "iip\0" /* Parameter signature */
     "glProgramEnvParameter4dvARB\0"
     "glProgramParameter4dvNV\0"
@@ -3666,14 +3752,14 @@ static const char ProgramEnvParameter4dvARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs2fvNV_names[] = 
+static const char VertexAttribs2fvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs2fvNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char UseProgramObjectARB_names[] = 
+static const char UseProgramObjectARB_names[] =
     "i\0" /* Parameter signature */
     "glUseProgram\0"
     "glUseProgramObjectARB\0"
@@ -3681,42 +3767,42 @@ static const char UseProgramObjectARB_names[] =
 #endif
 
 #if defined(need_GL_NV_evaluators)
-static const char GetMapParameterfvNV_names[] = 
+static const char GetMapParameterfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetMapParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char PassTexCoordATI_names[] = 
+static const char PassTexCoordATI_names[] =
     "iii\0" /* Parameter signature */
     "glPassTexCoordATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char DeleteProgram_names[] = 
+static const char DeleteProgram_names[] =
     "i\0" /* Parameter signature */
     "glDeleteProgram\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3ivEXT_names[] = 
+static const char Tangent3ivEXT_names[] =
     "p\0" /* Parameter signature */
     "glTangent3ivEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3dEXT_names[] = 
+static const char Tangent3dEXT_names[] =
     "ddd\0" /* Parameter signature */
     "glTangent3dEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3dvEXT_names[] = 
+static const char SecondaryColor3dvEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3dv\0"
     "glSecondaryColor3dvEXT\0"
@@ -3724,7 +3810,7 @@ static const char SecondaryColor3dvEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_multi_draw_arrays)
-static const char MultiDrawArraysEXT_names[] = 
+static const char MultiDrawArraysEXT_names[] =
     "ippi\0" /* Parameter signature */
     "glMultiDrawArrays\0"
     "glMultiDrawArraysEXT\0"
@@ -3732,7 +3818,7 @@ static const char MultiDrawArraysEXT_names[] =
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char BindRenderbufferEXT_names[] = 
+static const char BindRenderbufferEXT_names[] =
     "ii\0" /* Parameter signature */
     "glBindRenderbuffer\0"
     "glBindRenderbufferEXT\0"
@@ -3740,7 +3826,7 @@ static const char BindRenderbufferEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4dARB_names[] = 
+static const char MultiTexCoord4dARB_names[] =
     "idddd\0" /* Parameter signature */
     "glMultiTexCoord4d\0"
     "glMultiTexCoord4dARB\0"
@@ -3748,7 +3834,7 @@ static const char MultiTexCoord4dARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3usEXT_names[] = 
+static const char SecondaryColor3usEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3us\0"
     "glSecondaryColor3usEXT\0"
@@ -3756,14 +3842,14 @@ static const char SecondaryColor3usEXT_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char ProgramLocalParameter4fvARB_names[] = 
+static const char ProgramLocalParameter4fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glProgramLocalParameter4fvARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char DeleteProgramsNV_names[] = 
+static const char DeleteProgramsNV_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteProgramsARB\0"
     "glDeleteProgramsNV\0"
@@ -3771,7 +3857,7 @@ static const char DeleteProgramsNV_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1sARB_names[] = 
+static const char MultiTexCoord1sARB_names[] =
     "ii\0" /* Parameter signature */
     "glMultiTexCoord1s\0"
     "glMultiTexCoord1sARB\0"
@@ -3779,14 +3865,14 @@ static const char MultiTexCoord1sARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiColor3fVertex3fSUN_names[] = 
+static const char ReplacementCodeuiColor3fVertex3fSUN_names[] =
     "iffffff\0" /* Parameter signature */
     "glReplacementCodeuiColor3fVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char GetVertexAttribPointervNV_names[] = 
+static const char GetVertexAttribPointervNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetVertexAttribPointerv\0"
     "glGetVertexAttribPointervARB\0"
@@ -3795,7 +3881,7 @@ static const char GetVertexAttribPointervNV_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1dvARB_names[] = 
+static const char MultiTexCoord1dvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord1dv\0"
     "glMultiTexCoord1dvARB\0"
@@ -3803,7 +3889,7 @@ static const char MultiTexCoord1dvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform2iARB_names[] = 
+static const char Uniform2iARB_names[] =
     "iii\0" /* Parameter signature */
     "glUniform2i\0"
     "glUniform2iARB\0"
@@ -3811,21 +3897,21 @@ static const char Uniform2iARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char GetProgramStringNV_names[] = 
+static const char GetProgramStringNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramStringNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char ColorPointerEXT_names[] = 
+static const char ColorPointerEXT_names[] =
     "iiiip\0" /* Parameter signature */
     "glColorPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char MapBufferARB_names[] = 
+static const char MapBufferARB_names[] =
     "ii\0" /* Parameter signature */
     "glMapBuffer\0"
     "glMapBufferARB\0"
@@ -3833,35 +3919,35 @@ static const char MapBufferARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3svEXT_names[] = 
+static const char Binormal3svEXT_names[] =
     "p\0" /* Parameter signature */
     "glBinormal3svEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_light_texture)
-static const char ApplyTextureEXT_names[] = 
+static const char ApplyTextureEXT_names[] =
     "i\0" /* Parameter signature */
     "glApplyTextureEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_light_texture)
-static const char TextureMaterialEXT_names[] = 
+static const char TextureMaterialEXT_names[] =
     "ii\0" /* Parameter signature */
     "glTextureMaterialEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_light_texture)
-static const char TextureLightEXT_names[] = 
+static const char TextureLightEXT_names[] =
     "i\0" /* Parameter signature */
     "glTextureLightEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char ResetMinmax_names[] = 
+static const char ResetMinmax_names[] =
     "i\0" /* Parameter signature */
     "glResetMinmax\0"
     "glResetMinmaxEXT\0"
@@ -3869,21 +3955,21 @@ static const char ResetMinmax_names[] =
 #endif
 
 #if defined(need_GL_SGIX_sprite)
-static const char SpriteParameterfSGIX_names[] = 
+static const char SpriteParameterfSGIX_names[] =
     "if\0" /* Parameter signature */
     "glSpriteParameterfSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4sNV_names[] = 
+static const char VertexAttrib4sNV_names[] =
     "iiiii\0" /* Parameter signature */
     "glVertexAttrib4sNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char GetConvolutionParameterfv_names[] = 
+static const char GetConvolutionParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glGetConvolutionParameterfv\0"
     "glGetConvolutionParameterfvEXT\0"
@@ -3891,14 +3977,21 @@ static const char GetConvolutionParameterfv_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs4dvNV_names[] = 
+static const char VertexAttribs4dvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs4dvNV\0"
     "";
 #endif
 
+#if defined(need_GL_IBM_multimode_draw_arrays)
+static const char MultiModeDrawArraysIBM_names[] =
+    "pppii\0" /* Parameter signature */
+    "glMultiModeDrawArraysIBM\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4dARB_names[] = 
+static const char VertexAttrib4dARB_names[] =
     "idddd\0" /* Parameter signature */
     "glVertexAttrib4d\0"
     "glVertexAttrib4dARB\0"
@@ -3906,28 +3999,35 @@ static const char VertexAttrib4dARB_names[] =
 #endif
 
 #if defined(need_GL_ATI_envmap_bumpmap)
-static const char GetTexBumpParameterfvATI_names[] = 
+static const char GetTexBumpParameterfvATI_names[] =
     "ip\0" /* Parameter signature */
     "glGetTexBumpParameterfvATI\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fragment_program)
-static const char ProgramNamedParameter4dNV_names[] = 
+static const char ProgramNamedParameter4dNV_names[] =
     "iipdddd\0" /* Parameter signature */
     "glProgramNamedParameter4dNV\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_weighting)
-static const char VertexWeightfEXT_names[] = 
+static const char VertexWeightfEXT_names[] =
     "f\0" /* Parameter signature */
     "glVertexWeightfEXT\0"
     "";
 #endif
 
+#if defined(need_GL_EXT_coordinate_frame)
+static const char Binormal3fEXT_names[] =
+    "fff\0" /* Parameter signature */
+    "glBinormal3fEXT\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
-static const char FogCoordfvEXT_names[] = 
+static const char FogCoordfvEXT_names[] =
     "p\0" /* Parameter signature */
     "glFogCoordfv\0"
     "glFogCoordfvEXT\0"
@@ -3935,7 +4035,7 @@ static const char FogCoordfvEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1ivARB_names[] = 
+static const char MultiTexCoord1ivARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord1iv\0"
     "glMultiTexCoord1ivARB\0"
@@ -3943,7 +4043,7 @@ static const char MultiTexCoord1ivARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3ubEXT_names[] = 
+static const char SecondaryColor3ubEXT_names[] =
     "iii\0" /* Parameter signature */
     "glSecondaryColor3ub\0"
     "glSecondaryColor3ubEXT\0"
@@ -3951,7 +4051,7 @@ static const char SecondaryColor3ubEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2ivARB_names[] = 
+static const char MultiTexCoord2ivARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord2iv\0"
     "glMultiTexCoord2ivARB\0"
@@ -3959,14 +4059,14 @@ static const char MultiTexCoord2ivARB_names[] =
 #endif
 
 #if defined(need_GL_SGIS_fog_function)
-static const char FogFuncSGIS_names[] = 
+static const char FogFuncSGIS_names[] =
     "ip\0" /* Parameter signature */
     "glFogFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_copy_texture)
-static const char CopyTexSubImage2D_names[] = 
+static const char CopyTexSubImage2D_names[] =
     "iiiiiiii\0" /* Parameter signature */
     "glCopyTexSubImage2D\0"
     "glCopyTexSubImage2DEXT\0"
@@ -3974,35 +4074,35 @@ static const char CopyTexSubImage2D_names[] =
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char GetObjectParameterivARB_names[] = 
+static const char GetObjectParameterivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetObjectParameterivARB\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord4fVertex4fSUN_names[] = 
+static const char TexCoord4fVertex4fSUN_names[] =
     "ffffffff\0" /* Parameter signature */
     "glTexCoord4fVertex4fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_APPLE_vertex_array_object)
-static const char BindVertexArrayAPPLE_names[] = 
+static const char BindVertexArrayAPPLE_names[] =
     "i\0" /* Parameter signature */
     "glBindVertexArrayAPPLE\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramLocalParameterdvARB_names[] = 
+static const char GetProgramLocalParameterdvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramLocalParameterdvARB\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetHistogramParameteriv_names[] = 
+static const char GetHistogramParameteriv_names[] =
     "iip\0" /* Parameter signature */
     "glGetHistogramParameteriv\0"
     "glGetHistogramParameterivEXT\0"
@@ -4010,7 +4110,7 @@ static const char GetHistogramParameteriv_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1iARB_names[] = 
+static const char MultiTexCoord1iARB_names[] =
     "ii\0" /* Parameter signature */
     "glMultiTexCoord1i\0"
     "glMultiTexCoord1iARB\0"
@@ -4018,7 +4118,7 @@ static const char MultiTexCoord1iARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char GetConvolutionFilter_names[] = 
+static const char GetConvolutionFilter_names[] =
     "iiip\0" /* Parameter signature */
     "glGetConvolutionFilter\0"
     "glGetConvolutionFilterEXT\0"
@@ -4026,14 +4126,14 @@ static const char GetConvolutionFilter_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramivARB_names[] = 
+static const char GetProgramivARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramivARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_blend_func_separate) || defined(need_GL_INGR_blend_func_separate)
-static const char BlendFuncSeparateEXT_names[] = 
+static const char BlendFuncSeparateEXT_names[] =
     "iiii\0" /* Parameter signature */
     "glBlendFuncSeparate\0"
     "glBlendFuncSeparateEXT\0"
@@ -4041,50 +4141,50 @@ static const char BlendFuncSeparateEXT_names[] =
     "";
 #endif
 
-#if defined(need_GL_APPLE_vertex_array_object)
-static const char IsVertexArrayAPPLE_names[] = 
-    "i\0" /* Parameter signature */
-    "glIsVertexArrayAPPLE\0"
+#if defined(need_GL_ARB_map_buffer_range)
+static const char MapBufferRange_names[] =
+    "iiii\0" /* Parameter signature */
+    "glMapBufferRange\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char ProgramParameters4dvNV_names[] = 
+static const char ProgramParameters4dvNV_names[] =
     "iiip\0" /* Parameter signature */
     "glProgramParameters4dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord2fColor3fVertex3fvSUN_names[] = 
+static const char TexCoord2fColor3fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glTexCoord2fColor3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3dvEXT_names[] = 
+static const char Binormal3dvEXT_names[] =
     "p\0" /* Parameter signature */
     "glBinormal3dvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_fence)
-static const char FinishFenceNV_names[] = 
+static const char FinishFenceNV_names[] =
     "i\0" /* Parameter signature */
     "glFinishFenceNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIS_fog_function)
-static const char GetFogFuncSGIS_names[] = 
+static const char GetFogFuncSGIS_names[] =
     "p\0" /* Parameter signature */
     "glGetFogFuncSGIS\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char GetUniformLocationARB_names[] = 
+static const char GetUniformLocationARB_names[] =
     "ip\0" /* Parameter signature */
     "glGetUniformLocation\0"
     "glGetUniformLocationARB\0"
@@ -4092,7 +4192,7 @@ static const char GetUniformLocationARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3fEXT_names[] = 
+static const char SecondaryColor3fEXT_names[] =
     "fff\0" /* Parameter signature */
     "glSecondaryColor3f\0"
     "glSecondaryColor3fEXT\0"
@@ -4100,14 +4200,14 @@ static const char SecondaryColor3fEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerInputNV_names[] = 
+static const char CombinerInputNV_names[] =
     "iiiiii\0" /* Parameter signature */
     "glCombinerInputNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib3sARB_names[] = 
+static const char VertexAttrib3sARB_names[] =
     "iiii\0" /* Parameter signature */
     "glVertexAttrib3s\0"
     "glVertexAttrib3sARB\0"
@@ -4115,49 +4215,49 @@ static const char VertexAttrib3sARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiNormal3fVertex3fvSUN_names[] = 
+static const char ReplacementCodeuiNormal3fVertex3fvSUN_names[] =
     "ppp\0" /* Parameter signature */
     "glReplacementCodeuiNormal3fVertex3fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char ProgramStringARB_names[] = 
+static const char ProgramStringARB_names[] =
     "iiip\0" /* Parameter signature */
     "glProgramStringARB\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char TexCoord4fVertex4fvSUN_names[] = 
+static const char TexCoord4fVertex4fvSUN_names[] =
     "pp\0" /* Parameter signature */
     "glTexCoord4fVertex4fvSUN\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3sNV_names[] = 
+static const char VertexAttrib3sNV_names[] =
     "iiii\0" /* Parameter signature */
     "glVertexAttrib3sNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1fNV_names[] = 
+static const char VertexAttrib1fNV_names[] =
     "if\0" /* Parameter signature */
     "glVertexAttrib1fNV\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentLightfSGIX_names[] = 
+static const char FragmentLightfSGIX_names[] =
     "iif\0" /* Parameter signature */
     "glFragmentLightfSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression)
-static const char GetCompressedTexImageARB_names[] = 
+static const char GetCompressedTexImageARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetCompressedTexImage\0"
     "glGetCompressedTexImageARB\0"
@@ -4165,14 +4265,14 @@ static const char GetCompressedTexImageARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_weighting)
-static const char VertexWeightPointerEXT_names[] = 
+static const char VertexWeightPointerEXT_names[] =
     "iiip\0" /* Parameter signature */
     "glVertexWeightPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetHistogram_names[] = 
+static const char GetHistogram_names[] =
     "iiiip\0" /* Parameter signature */
     "glGetHistogram\0"
     "glGetHistogramEXT\0"
@@ -4180,21 +4280,21 @@ static const char GetHistogram_names[] =
 #endif
 
 #if defined(need_GL_EXT_stencil_two_side)
-static const char ActiveStencilFaceEXT_names[] = 
+static const char ActiveStencilFaceEXT_names[] =
     "i\0" /* Parameter signature */
     "glActiveStencilFaceEXT\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_separate_stencil)
-static const char StencilFuncSeparateATI_names[] = 
+static const char StencilFuncSeparateATI_names[] =
     "iiii\0" /* Parameter signature */
     "glStencilFuncSeparateATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char GetShaderSourceARB_names[] = 
+static const char GetShaderSourceARB_names[] =
     "iipp\0" /* Parameter signature */
     "glGetShaderSource\0"
     "glGetShaderSourceARB\0"
@@ -4202,28 +4302,28 @@ static const char GetShaderSourceARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_igloo_interface)
-static const char IglooInterfaceSGIX_names[] = 
+static const char IglooInterfaceSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glIglooInterfaceSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4dNV_names[] = 
+static const char VertexAttrib4dNV_names[] =
     "idddd\0" /* Parameter signature */
     "glVertexAttrib4dNV\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_multimode_draw_arrays)
-static const char MultiModeDrawElementsIBM_names[] = 
+static const char MultiModeDrawElementsIBM_names[] =
     "ppipii\0" /* Parameter signature */
     "glMultiModeDrawElementsIBM\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4svARB_names[] = 
+static const char MultiTexCoord4svARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord4sv\0"
     "glMultiTexCoord4svARB\0"
@@ -4231,7 +4331,7 @@ static const char MultiTexCoord4svARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query)
-static const char GenQueriesARB_names[] = 
+static const char GenQueriesARB_names[] =
     "ip\0" /* Parameter signature */
     "glGenQueries\0"
     "glGenQueriesARB\0"
@@ -4239,35 +4339,42 @@ static const char GenQueriesARB_names[] =
 #endif
 
 #if defined(need_GL_SUN_vertex)
-static const char ReplacementCodeuiVertex3fSUN_names[] = 
+static const char ReplacementCodeuiVertex3fSUN_names[] =
     "ifff\0" /* Parameter signature */
     "glReplacementCodeuiVertex3fSUN\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3iEXT_names[] = 
+static const char Tangent3iEXT_names[] =
     "iii\0" /* Parameter signature */
     "glTangent3iEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_mesh_array)
-static const char DrawMeshArraysSUN_names[] = 
+static const char DrawMeshArraysSUN_names[] =
     "iiii\0" /* Parameter signature */
     "glDrawMeshArraysSUN\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char IsSync_names[] =
+    "i\0" /* Parameter signature */
+    "glIsSync\0"
+    "";
+#endif
+
 #if defined(need_GL_NV_evaluators)
-static const char GetMapControlPointsNV_names[] = 
+static const char GetMapControlPointsNV_names[] =
     "iiiiiip\0" /* Parameter signature */
     "glGetMapControlPointsNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_draw_buffers) || defined(need_GL_ATI_draw_buffers)
-static const char DrawBuffersARB_names[] = 
+static const char DrawBuffersARB_names[] =
     "ip\0" /* Parameter signature */
     "glDrawBuffers\0"
     "glDrawBuffersARB\0"
@@ -4276,21 +4383,28 @@ static const char DrawBuffersARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char ProgramLocalParameter4fARB_names[] = 
+static const char ProgramLocalParameter4fARB_names[] =
     "iiffff\0" /* Parameter signature */
     "glProgramLocalParameter4fARB\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_sprite)
-static const char SpriteParameterivSGIX_names[] = 
+static const char SpriteParameterivSGIX_names[] =
     "ip\0" /* Parameter signature */
     "glSpriteParameterivSGIX\0"
     "";
 #endif
 
+#if defined(need_GL_EXT_provoking_vertex)
+static const char ProvokingVertexEXT_names[] =
+    "i\0" /* Parameter signature */
+    "glProvokingVertexEXT\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord1fARB_names[] = 
+static const char MultiTexCoord1fARB_names[] =
     "if\0" /* Parameter signature */
     "glMultiTexCoord1f\0"
     "glMultiTexCoord1fARB\0"
@@ -4298,21 +4412,21 @@ static const char MultiTexCoord1fARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs4ubvNV_names[] = 
+static const char VertexAttribs4ubvNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs4ubvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightsvARB_names[] = 
+static const char WeightsvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightsvARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects)
-static const char Uniform1fvARB_names[] = 
+static const char Uniform1fvARB_names[] =
     "iip\0" /* Parameter signature */
     "glUniform1fv\0"
     "glUniform1fvARB\0"
@@ -4320,7 +4434,7 @@ static const char Uniform1fvARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_copy_texture)
-static const char CopyTexSubImage1D_names[] = 
+static const char CopyTexSubImage1D_names[] =
     "iiiiii\0" /* Parameter signature */
     "glCopyTexSubImage1D\0"
     "glCopyTexSubImage1DEXT\0"
@@ -4328,7 +4442,7 @@ static const char CopyTexSubImage1D_names[] =
 #endif
 
 #if defined(need_GL_EXT_texture_object)
-static const char BindTexture_names[] = 
+static const char BindTexture_names[] =
     "ii\0" /* Parameter signature */
     "glBindTexture\0"
     "glBindTextureEXT\0"
@@ -4336,14 +4450,14 @@ static const char BindTexture_names[] =
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char BeginFragmentShaderATI_names[] = 
+static const char BeginFragmentShaderATI_names[] =
     "\0" /* Parameter signature */
     "glBeginFragmentShaderATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord4fARB_names[] = 
+static const char MultiTexCoord4fARB_names[] =
     "iffff\0" /* Parameter signature */
     "glMultiTexCoord4f\0"
     "glMultiTexCoord4fARB\0"
@@ -4351,21 +4465,21 @@ static const char MultiTexCoord4fARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs3svNV_names[] = 
+static const char VertexAttribs3svNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs3svNV\0"
     "";
 #endif
 
 #if defined(need_GL_SUN_triangle_list)
-static const char ReplacementCodeuivSUN_names[] = 
+static const char ReplacementCodeuivSUN_names[] =
     "p\0" /* Parameter signature */
     "glReplacementCodeuivSUN\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char EnableVertexAttribArrayARB_names[] = 
+static const char EnableVertexAttribArrayARB_names[] =
     "i\0" /* Parameter signature */
     "glEnableVertexAttribArray\0"
     "glEnableVertexAttribArrayARB\0"
@@ -4373,14 +4487,14 @@ static const char EnableVertexAttribArrayARB_names[] =
 #endif
 
 #if defined(need_GL_INTEL_parallel_arrays)
-static const char NormalPointervINTEL_names[] = 
+static const char NormalPointervINTEL_names[] =
     "ip\0" /* Parameter signature */
     "glNormalPointervINTEL\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_convolution)
-static const char CopyConvolutionFilter2D_names[] = 
+static const char CopyConvolutionFilter2D_names[] =
     "iiiiii\0" /* Parameter signature */
     "glCopyConvolutionFilter2D\0"
     "glCopyConvolutionFilter2DEXT\0"
@@ -4388,7 +4502,7 @@ static const char CopyConvolutionFilter2D_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3ivMESA_names[] = 
+static const char WindowPos3ivMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos3iv\0"
     "glWindowPos3ivARB\0"
@@ -4396,8 +4510,15 @@ static const char WindowPos3ivMESA_names[] =
     "";
 #endif
 
+#if defined(need_GL_ARB_copy_buffer)
+static const char CopyBufferSubData_names[] =
+    "iiiii\0" /* Parameter signature */
+    "glCopyBufferSubData\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char IsBufferARB_names[] = 
+static const char IsBufferARB_names[] =
     "i\0" /* Parameter signature */
     "glIsBuffer\0"
     "glIsBufferARB\0"
@@ -4405,14 +4526,14 @@ static const char IsBufferARB_names[] =
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4iMESA_names[] = 
+static const char WindowPos4iMESA_names[] =
     "iiii\0" /* Parameter signature */
     "glWindowPos4iMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4uivARB_names[] = 
+static const char VertexAttrib4uivARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4uiv\0"
     "glVertexAttrib4uivARB\0"
@@ -4420,35 +4541,35 @@ static const char VertexAttrib4uivARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3bvEXT_names[] = 
+static const char Tangent3bvEXT_names[] =
     "p\0" /* Parameter signature */
     "glTangent3bvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix3x4fv_names[] = 
+static const char UniformMatrix3x4fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix3x4fv\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3fvEXT_names[] = 
+static const char Binormal3fvEXT_names[] =
     "p\0" /* Parameter signature */
     "glBinormal3fvEXT\0"
     "";
 #endif
 
 #if defined(need_GL_INTEL_parallel_arrays)
-static const char TexCoordPointervINTEL_names[] = 
+static const char TexCoordPointervINTEL_names[] =
     "iip\0" /* Parameter signature */
     "glTexCoordPointervINTEL\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char DeleteBuffersARB_names[] = 
+static const char DeleteBuffersARB_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteBuffers\0"
     "glDeleteBuffersARB\0"
@@ -4456,21 +4577,21 @@ static const char DeleteBuffersARB_names[] =
 #endif
 
 #if defined(need_GL_MESA_window_pos)
-static const char WindowPos4fvMESA_names[] = 
+static const char WindowPos4fvMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos4fvMESA\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1sNV_names[] = 
+static const char VertexAttrib1sNV_names[] =
     "ii\0" /* Parameter signature */
     "glVertexAttrib1sNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color)
-static const char SecondaryColor3svEXT_names[] = 
+static const char SecondaryColor3svEXT_names[] =
     "p\0" /* Parameter signature */
     "glSecondaryColor3sv\0"
     "glSecondaryColor3svEXT\0"
@@ -4478,7 +4599,7 @@ static const char SecondaryColor3svEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix)
-static const char LoadTransposeMatrixfARB_names[] = 
+static const char LoadTransposeMatrixfARB_names[] =
     "p\0" /* Parameter signature */
     "glLoadTransposeMatrixf\0"
     "glLoadTransposeMatrixfARB\0"
@@ -4486,7 +4607,7 @@ static const char LoadTransposeMatrixfARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char GetPointerv_names[] = 
+static const char GetPointerv_names[] =
     "ip\0" /* Parameter signature */
     "glGetPointerv\0"
     "glGetPointervEXT\0"
@@ -4494,21 +4615,21 @@ static const char GetPointerv_names[] =
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3bEXT_names[] = 
+static const char Tangent3bEXT_names[] =
     "iii\0" /* Parameter signature */
     "glTangent3bEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerParameterfNV_names[] = 
+static const char CombinerParameterfNV_names[] =
     "if\0" /* Parameter signature */
     "glCombinerParameterfNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char BindProgramNV_names[] = 
+static const char BindProgramNV_names[] =
     "ii\0" /* Parameter signature */
     "glBindProgramARB\0"
     "glBindProgramNV\0"
@@ -4516,7 +4637,7 @@ static const char BindProgramNV_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4svARB_names[] = 
+static const char VertexAttrib4svARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4sv\0"
     "glVertexAttrib4svARB\0"
@@ -4524,28 +4645,35 @@ static const char VertexAttrib4svARB_names[] =
 #endif
 
 #if defined(need_GL_MESA_shader_debug)
-static const char CreateDebugObjectMESA_names[] = 
+static const char CreateDebugObjectMESA_names[] =
     "\0" /* Parameter signature */
     "glCreateDebugObjectMESA\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0)
-static const char GetShaderiv_names[] = 
+static const char GetShaderiv_names[] =
     "iip\0" /* Parameter signature */
     "glGetShaderiv\0"
     "";
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const char ClientWaitSync_names[] =
+    "iii\0" /* Parameter signature */
+    "glClientWaitSync\0"
+    "";
+#endif
+
 #if defined(need_GL_ATI_fragment_shader)
-static const char BindFragmentShaderATI_names[] = 
+static const char BindFragmentShaderATI_names[] =
     "i\0" /* Parameter signature */
     "glBindFragmentShaderATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char UnmapBufferARB_names[] = 
+static const char UnmapBufferARB_names[] =
     "i\0" /* Parameter signature */
     "glUnmapBuffer\0"
     "glUnmapBufferARB\0"
@@ -4553,22 +4681,15 @@ static const char UnmapBufferARB_names[] =
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char Minmax_names[] = 
+static const char Minmax_names[] =
     "iii\0" /* Parameter signature */
     "glMinmax\0"
     "glMinmaxEXT\0"
     "";
 #endif
 
-#if defined(need_GL_SGIX_polynomial_ffd)
-static const char DeformationMap3dSGIX_names[] = 
-    "iddiiddiiddiip\0" /* Parameter signature */
-    "glDeformationMap3dSGIX\0"
-    "";
-#endif
-
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
-static const char FogCoorddvEXT_names[] = 
+static const char FogCoorddvEXT_names[] =
     "p\0" /* Parameter signature */
     "glFogCoorddv\0"
     "glFogCoorddvEXT\0"
@@ -4576,35 +4697,35 @@ static const char FogCoorddvEXT_names[] =
 #endif
 
 #if defined(need_GL_SUNX_constant_data)
-static const char FinishTextureSUNX_names[] = 
+static const char FinishTextureSUNX_names[] =
     "\0" /* Parameter signature */
     "glFinishTextureSUNX\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char GetFragmentLightfvSGIX_names[] = 
+static const char GetFragmentLightfvSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetFragmentLightfvSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetFinalCombinerInputParameterfvNV_names[] = 
+static const char GetFinalCombinerInputParameterfvNV_names[] =
     "iip\0" /* Parameter signature */
     "glGetFinalCombinerInputParameterfvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char ColorFragmentOp3ATI_names[] = 
+static const char ColorFragmentOp3ATI_names[] =
     "iiiiiiiiiiiii\0" /* Parameter signature */
     "glColorFragmentOp3ATI\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2svARB_names[] = 
+static const char VertexAttrib2svARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib2sv\0"
     "glVertexAttrib2svARB\0"
@@ -4612,14 +4733,14 @@ static const char VertexAttrib2svARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char AreProgramsResidentNV_names[] = 
+static const char AreProgramsResidentNV_names[] =
     "ipp\0" /* Parameter signature */
     "glAreProgramsResidentNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos3svMESA_names[] = 
+static const char WindowPos3svMESA_names[] =
     "p\0" /* Parameter signature */
     "glWindowPos3sv\0"
     "glWindowPos3svARB\0"
@@ -4628,7 +4749,7 @@ static const char WindowPos3svMESA_names[] =
 #endif
 
 #if defined(need_GL_EXT_color_subtable)
-static const char CopyColorSubTable_names[] = 
+static const char CopyColorSubTable_names[] =
     "iiiii\0" /* Parameter signature */
     "glCopyColorSubTable\0"
     "glCopyColorSubTableEXT\0"
@@ -4636,14 +4757,14 @@ static const char CopyColorSubTable_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightdvARB_names[] = 
+static const char WeightdvARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightdvARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char DeleteRenderbuffersEXT_names[] = 
+static const char DeleteRenderbuffersEXT_names[] =
     "ip\0" /* Parameter signature */
     "glDeleteRenderbuffers\0"
     "glDeleteRenderbuffersEXT\0"
@@ -4651,7 +4772,7 @@ static const char DeleteRenderbuffersEXT_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib4NubvARB_names[] = 
+static const char VertexAttrib4NubvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4Nubv\0"
     "glVertexAttrib4NubvARB\0"
@@ -4659,63 +4780,63 @@ static const char VertexAttrib4NubvARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib3dvNV_names[] = 
+static const char VertexAttrib3dvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib3dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char GetObjectParameterfvARB_names[] = 
+static const char GetObjectParameterfvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetObjectParameterfvARB\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_vertex_program)
-static const char GetProgramEnvParameterdvARB_names[] = 
+static const char GetProgramEnvParameterdvARB_names[] =
     "iip\0" /* Parameter signature */
     "glGetProgramEnvParameterdvARB\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_compiled_vertex_array)
-static const char LockArraysEXT_names[] = 
+static const char LockArraysEXT_names[] =
     "ii\0" /* Parameter signature */
     "glLockArraysEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_pixel_transform)
-static const char PixelTransformParameterivEXT_names[] = 
+static const char PixelTransformParameterivEXT_names[] =
     "iip\0" /* Parameter signature */
     "glPixelTransformParameterivEXT\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char BinormalPointerEXT_names[] = 
+static const char BinormalPointerEXT_names[] =
     "iip\0" /* Parameter signature */
     "glBinormalPointerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib1dNV_names[] = 
+static const char VertexAttrib1dNV_names[] =
     "id\0" /* Parameter signature */
     "glVertexAttrib1dNV\0"
     "";
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char GetCombinerInputParameterivNV_names[] = 
+static const char GetCombinerInputParameterivNV_names[] =
     "iiiip\0" /* Parameter signature */
     "glGetCombinerInputParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_3)
-static const char MultiTexCoord2fvARB_names[] = 
+static const char MultiTexCoord2fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glMultiTexCoord2fv\0"
     "glMultiTexCoord2fvARB\0"
@@ -4723,7 +4844,7 @@ static const char MultiTexCoord2fvARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char GetRenderbufferParameterivEXT_names[] = 
+static const char GetRenderbufferParameterivEXT_names[] =
     "iip\0" /* Parameter signature */
     "glGetRenderbufferParameteriv\0"
     "glGetRenderbufferParameterivEXT\0"
@@ -4731,21 +4852,21 @@ static const char GetRenderbufferParameterivEXT_names[] =
 #endif
 
 #if defined(need_GL_NV_register_combiners)
-static const char CombinerParameterivNV_names[] = 
+static const char CombinerParameterivNV_names[] =
     "ip\0" /* Parameter signature */
     "glCombinerParameterivNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char GenFragmentShadersATI_names[] = 
+static const char GenFragmentShadersATI_names[] =
     "i\0" /* Parameter signature */
     "glGenFragmentShadersATI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_vertex_array)
-static const char DrawArrays_names[] = 
+static const char DrawArrays_names[] =
     "iii\0" /* Parameter signature */
     "glDrawArrays\0"
     "glDrawArraysEXT\0"
@@ -4753,14 +4874,14 @@ static const char DrawArrays_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_blend)
-static const char WeightuivARB_names[] = 
+static const char WeightuivARB_names[] =
     "ip\0" /* Parameter signature */
     "glWeightuivARB\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib2sARB_names[] = 
+static const char VertexAttrib2sARB_names[] =
     "iii\0" /* Parameter signature */
     "glVertexAttrib2s\0"
     "glVertexAttrib2sARB\0"
@@ -4768,28 +4889,28 @@ static const char VertexAttrib2sARB_names[] =
 #endif
 
 #if defined(need_GL_SGIX_async)
-static const char GenAsyncMarkersSGIX_names[] = 
+static const char GenAsyncMarkersSGIX_names[] =
     "i\0" /* Parameter signature */
     "glGenAsyncMarkersSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Tangent3svEXT_names[] = 
+static const char Tangent3svEXT_names[] =
     "p\0" /* Parameter signature */
     "glTangent3svEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char GetListParameterivSGIX_names[] = 
+static const char GetListParameterivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glGetListParameterivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char BindBufferARB_names[] = 
+static const char BindBufferARB_names[] =
     "ii\0" /* Parameter signature */
     "glBindBuffer\0"
     "glBindBufferARB\0"
@@ -4797,35 +4918,35 @@ static const char BindBufferARB_names[] =
 #endif
 
 #if defined(need_GL_ARB_shader_objects)
-static const char GetInfoLogARB_names[] = 
+static const char GetInfoLogARB_names[] =
     "iipp\0" /* Parameter signature */
     "glGetInfoLogARB\0"
     "";
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs4svNV_names[] = 
+static const char VertexAttribs4svNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs4svNV\0"
     "";
 #endif
 
 #if defined(need_GL_IBM_vertex_array_lists)
-static const char EdgeFlagPointerListIBM_names[] = 
+static const char EdgeFlagPointerListIBM_names[] =
     "ipi\0" /* Parameter signature */
     "glEdgeFlagPointerListIBM\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_2_1)
-static const char UniformMatrix3x2fv_names[] = 
+static const char UniformMatrix3x2fv_names[] =
     "iiip\0" /* Parameter signature */
     "glUniformMatrix3x2fv\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_histogram)
-static const char GetMinmaxParameterfv_names[] = 
+static const char GetMinmaxParameterfv_names[] =
     "iip\0" /* Parameter signature */
     "glGetMinmaxParameterfv\0"
     "glGetMinmaxParameterfvEXT\0"
@@ -4833,7 +4954,7 @@ static const char GetMinmaxParameterfv_names[] =
 #endif
 
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
-static const char VertexAttrib1fvARB_names[] = 
+static const char VertexAttrib1fvARB_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib1fv\0"
     "glVertexAttrib1fvARB\0"
@@ -4841,7 +4962,7 @@ static const char VertexAttrib1fvARB_names[] =
 #endif
 
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
-static const char GenBuffersARB_names[] = 
+static const char GenBuffersARB_names[] =
     "ip\0" /* Parameter signature */
     "glGenBuffers\0"
     "glGenBuffersARB\0"
@@ -4849,35 +4970,35 @@ static const char GenBuffersARB_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttribs1svNV_names[] = 
+static const char VertexAttribs1svNV_names[] =
     "iip\0" /* Parameter signature */
     "glVertexAttribs1svNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_envmap_bumpmap)
-static const char GetTexBumpParameterivATI_names[] = 
+static const char GetTexBumpParameterivATI_names[] =
     "ip\0" /* Parameter signature */
     "glGetTexBumpParameterivATI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3bEXT_names[] = 
+static const char Binormal3bEXT_names[] =
     "iii\0" /* Parameter signature */
     "glBinormal3bEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_fragment_lighting)
-static const char FragmentMaterialivSGIX_names[] = 
+static const char FragmentMaterialivSGIX_names[] =
     "iip\0" /* Parameter signature */
     "glFragmentMaterialivSGIX\0"
     "";
 #endif
 
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
-static const char IsRenderbufferEXT_names[] = 
+static const char IsRenderbufferEXT_names[] =
     "i\0" /* Parameter signature */
     "glIsRenderbuffer\0"
     "glIsRenderbufferEXT\0"
@@ -4885,7 +5006,7 @@ static const char IsRenderbufferEXT_names[] =
 #endif
 
 #if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
-static const char GenProgramsNV_names[] = 
+static const char GenProgramsNV_names[] =
     "ip\0" /* Parameter signature */
     "glGenProgramsARB\0"
     "glGenProgramsNV\0"
@@ -4893,28 +5014,28 @@ static const char GenProgramsNV_names[] =
 #endif
 
 #if defined(need_GL_NV_vertex_program)
-static const char VertexAttrib4dvNV_names[] = 
+static const char VertexAttrib4dvNV_names[] =
     "ip\0" /* Parameter signature */
     "glVertexAttrib4dvNV\0"
     "";
 #endif
 
 #if defined(need_GL_ATI_fragment_shader)
-static const char EndFragmentShaderATI_names[] = 
+static const char EndFragmentShaderATI_names[] =
     "\0" /* Parameter signature */
     "glEndFragmentShaderATI\0"
     "";
 #endif
 
 #if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3iEXT_names[] = 
+static const char Binormal3iEXT_names[] =
     "iii\0" /* Parameter signature */
     "glBinormal3iEXT\0"
     "";
 #endif
 
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos)
-static const char WindowPos2fMESA_names[] = 
+static const char WindowPos2fMESA_names[] =
     "ff\0" /* Parameter signature */
     "glWindowPos2f\0"
     "glWindowPos2fARB\0"
@@ -4929,12 +5050,35 @@ static const struct dri_extension_function GL_3DFX_tbuffer_functions[] = {
 };
 #endif
 
+#if defined(need_GL_APPLE_flush_buffer_range)
+static const struct dri_extension_function GL_APPLE_flush_buffer_range_functions[] = {
+    { BufferParameteriAPPLE_names, BufferParameteriAPPLE_remap_index, -1 },
+    { FlushMappedBufferRangeAPPLE_names, FlushMappedBufferRangeAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
+#if defined(need_GL_APPLE_texture_range)
+static const struct dri_extension_function GL_APPLE_texture_range_functions[] = {
+    { TextureRangeAPPLE_names, TextureRangeAPPLE_remap_index, -1 },
+    { GetTexParameterPointervAPPLE_names, GetTexParameterPointervAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_APPLE_vertex_array_object)
 static const struct dri_extension_function GL_APPLE_vertex_array_object_functions[] = {
     { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 },
     { GenVertexArraysAPPLE_names, GenVertexArraysAPPLE_remap_index, -1 },
-    { BindVertexArrayAPPLE_names, BindVertexArrayAPPLE_remap_index, -1 },
     { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 },
+    { BindVertexArrayAPPLE_names, BindVertexArrayAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
+#if defined(need_GL_ARB_copy_buffer)
+static const struct dri_extension_function GL_ARB_copy_buffer_functions[] = {
+    { CopyBufferSubData_names, CopyBufferSubData_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -4972,6 +5116,14 @@ static const struct dri_extension_function GL_ARB_framebuffer_object_functions[]
 };
 #endif
 
+#if defined(need_GL_ARB_map_buffer_range)
+static const struct dri_extension_function GL_ARB_map_buffer_range_functions[] = {
+    { FlushMappedBufferRange_names, FlushMappedBufferRange_remap_index, -1 },
+    { MapBufferRange_names, MapBufferRange_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_matrix_palette)
 static const struct dri_extension_function GL_ARB_matrix_palette_functions[] = {
     { MatrixIndexusvARB_names, MatrixIndexusvARB_remap_index, -1 },
@@ -5057,6 +5209,19 @@ static const struct dri_extension_function GL_ARB_shader_objects_functions[] = {
 };
 #endif
 
+#if defined(need_GL_ARB_sync)
+static const struct dri_extension_function GL_ARB_sync_functions[] = {
+    { DeleteSync_names, DeleteSync_remap_index, -1 },
+    { FenceSync_names, FenceSync_remap_index, -1 },
+    { WaitSync_names, WaitSync_remap_index, -1 },
+    { GetInteger64v_names, GetInteger64v_remap_index, -1 },
+    { GetSynciv_names, GetSynciv_remap_index, -1 },
+    { IsSync_names, IsSync_remap_index, -1 },
+    { ClientWaitSync_names, ClientWaitSync_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_texture_compression)
 static const struct dri_extension_function GL_ARB_texture_compression_functions[] = {
     { CompressedTexSubImage2DARB_names, CompressedTexSubImage2DARB_remap_index, -1 },
@@ -5080,6 +5245,16 @@ static const struct dri_extension_function GL_ARB_transpose_matrix_functions[] =
 };
 #endif
 
+#if defined(need_GL_ARB_vertex_array_object)
+static const struct dri_extension_function GL_ARB_vertex_array_object_functions[] = {
+    { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 },
+    { GenVertexArrays_names, GenVertexArrays_remap_index, -1 },
+    { BindVertexArray_names, BindVertexArray_remap_index, -1 },
+    { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_vertex_blend)
 static const struct dri_extension_function GL_ARB_vertex_blend_functions[] = {
     { WeightubvARB_names, WeightubvARB_remap_index, -1 },
@@ -5333,7 +5508,6 @@ static const struct dri_extension_function GL_EXT_coordinate_frame_functions[] =
     { Binormal3ivEXT_names, Binormal3ivEXT_remap_index, -1 },
     { Tangent3sEXT_names, Tangent3sEXT_remap_index, -1 },
     { Tangent3fvEXT_names, Tangent3fvEXT_remap_index, -1 },
-    { Binormal3fEXT_names, Binormal3fEXT_remap_index, -1 },
     { Tangent3dvEXT_names, Tangent3dvEXT_remap_index, -1 },
     { Binormal3bvEXT_names, Binormal3bvEXT_remap_index, -1 },
     { Binormal3dEXT_names, Binormal3dEXT_remap_index, -1 },
@@ -5342,6 +5516,7 @@ static const struct dri_extension_function GL_EXT_coordinate_frame_functions[] =
     { Tangent3ivEXT_names, Tangent3ivEXT_remap_index, -1 },
     { Tangent3dEXT_names, Tangent3dEXT_remap_index, -1 },
     { Binormal3svEXT_names, Binormal3svEXT_remap_index, -1 },
+    { Binormal3fEXT_names, Binormal3fEXT_remap_index, -1 },
     { Binormal3dvEXT_names, Binormal3dvEXT_remap_index, -1 },
     { Tangent3iEXT_names, Tangent3iEXT_remap_index, -1 },
     { Tangent3bvEXT_names, Tangent3bvEXT_remap_index, -1 },
@@ -5527,6 +5702,13 @@ static const struct dri_extension_function GL_EXT_polygon_offset_functions[] = {
 };
 #endif
 
+#if defined(need_GL_EXT_provoking_vertex)
+static const struct dri_extension_function GL_EXT_provoking_vertex_functions[] = {
+    { ProvokingVertexEXT_names, ProvokingVertexEXT_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_EXT_secondary_color)
 static const struct dri_extension_function GL_EXT_secondary_color_functions[] = {
     { SecondaryColor3iEXT_names, SecondaryColor3iEXT_remap_index, -1 },
@@ -6024,11 +6206,11 @@ static const struct dri_extension_function GL_SGIX_igloo_interface_functions[] =
 #if defined(need_GL_SGIX_instruments)
 static const struct dri_extension_function GL_SGIX_instruments_functions[] = {
     { ReadInstrumentsSGIX_names, ReadInstrumentsSGIX_remap_index, -1 },
+    { PollInstrumentsSGIX_names, PollInstrumentsSGIX_remap_index, -1 },
     { GetInstrumentsSGIX_names, GetInstrumentsSGIX_remap_index, -1 },
     { StartInstrumentsSGIX_names, StartInstrumentsSGIX_remap_index, -1 },
     { StopInstrumentsSGIX_names, StopInstrumentsSGIX_remap_index, -1 },
     { InstrumentsBufferSGIX_names, InstrumentsBufferSGIX_remap_index, -1 },
-    { PollInstrumentsSGIX_names, PollInstrumentsSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -6055,9 +6237,9 @@ static const struct dri_extension_function GL_SGIX_pixel_texture_functions[] = {
 #if defined(need_GL_SGIX_polynomial_ffd)
 static const struct dri_extension_function GL_SGIX_polynomial_ffd_functions[] = {
     { LoadIdentityDeformationMapSGIX_names, LoadIdentityDeformationMapSGIX_remap_index, -1 },
+    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { DeformSGIX_names, DeformSGIX_remap_index, -1 },
     { DeformationMap3fSGIX_names, DeformationMap3fSGIX_remap_index, -1 },
-    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
diff --git a/src/mesa/drivers/dri/common/xmlpool.h b/src/mesa/drivers/dri/common/xmlpool.h
index 7fbc6e800d..587517ea10 100644
--- a/src/mesa/drivers/dri/common/xmlpool.h
+++ b/src/mesa/drivers/dri/common/xmlpool.h
@@ -60,6 +60,10 @@
 #define DRI_CONF_OPT_BEGIN(name,type,def) \
 "<option name=\""#name"\" type=\""#type"\" default=\""#def"\">\n"
 
+/** \brief Begin an option definition with qouted default value */
+#define DRI_CONF_OPT_BEGIN_Q(name,type,def) \
+"<option name=\""#name"\" type=\""#type"\" default="#def">\n"
+
 /** \brief Begin an option definition with restrictions on valid values */
 #define DRI_CONF_OPT_BEGIN_V(name,type,def,valid) \
 "<option name=\""#name"\" type=\""#type"\" default=\""#def"\" valid=\""valid"\">\n"
diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c
index 35c268441c..4e41860d8c 100644
--- a/src/mesa/drivers/dri/fb/fb_egl.c
+++ b/src/mesa/drivers/dri/fb/fb_egl.c
@@ -472,8 +472,8 @@ fbCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext sh
    c->Base.DrawSurface = EGL_NO_SURFACE;
    c->Base.ReadSurface = EGL_NO_SURFACE;
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
+   /* link to display */
+   _eglLinkContext(&c->Base, disp);
    assert(c->Base.Handle);
 
    /* Init default driver functions then plug in our FBdev-specific functions
@@ -604,13 +604,9 @@ static EGLBoolean
 fbDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    fbSurface *fs = Lookup_fbSurface(surface);
-   _eglRemoveSurface(&fs->Base);
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkSurface(&fs->Base);
+   if (!_eglIsSurfaceBound(&fs->Base))
       free(fs);
-   }
    return EGL_TRUE;
 }
 
@@ -619,13 +615,9 @@ static EGLBoolean
 fbDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
 {
    fbContext *fc = Lookup_fbContext(context);
-   _eglRemoveContext(&fc->Base);
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkContext(&fc->Base);
+   if (!_eglIsContextBound(&fc->Base))
       free(fc);
-   }
    return EGL_TRUE;
 }
 
@@ -688,7 +680,7 @@ fbCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    surface->mesa_framebuffer = _mesa_create_framebuffer(&vis);
    if (!surface->mesa_framebuffer) {
       free(surface);
-      _eglRemoveSurface(&surface->Base);
+      _eglUnlinkSurface(&surface->Base);
       return EGL_NO_SURFACE;
    }
 
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index 9f4bd1699f..9d049dea8f 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -19,6 +19,7 @@ DRIVER_SOURCES = \
 	intel_batchbuffer.c \
 	intel_clear.c \
 	intel_extensions.c \
+	intel_generatemipmap.c \
 	intel_mipmap_tree.c \
 	intel_tex_layout.c \
 	intel_tex_image.c \
@@ -50,6 +51,7 @@ DRIVER_SOURCES = \
 	intel_screen.c \
 	intel_span.c \
 	intel_state.c \
+	intel_syncobj.c \
 	intel_tris.c \
 	intel_fbo.c
 
diff --git a/src/mesa/drivers/dri/i915/i830_context.h b/src/mesa/drivers/dri/i915/i830_context.h
index 1bdb32049d..f73cbbf88b 100644
--- a/src/mesa/drivers/dri/i915/i830_context.h
+++ b/src/mesa/drivers/dri/i915/i830_context.h
@@ -40,6 +40,7 @@
 #define I830_UPLOAD_BUFFERS          0x2
 #define I830_UPLOAD_STIPPLE          0x4
 #define I830_UPLOAD_INVARIENT        0x8
+#define I830_UPLOAD_RASTER_RULES     0x10
 #define I830_UPLOAD_TEX(i)           (0x10<<(i))
 #define I830_UPLOAD_TEXBLEND(i)      (0x100<<(i))
 #define I830_UPLOAD_TEX_ALL          (0x0f0)
@@ -99,6 +100,11 @@
 
 #define I830_TEXBLEND_SIZE	12      /* (4 args + op) * 2 + COLOR_FACTOR */
 
+enum {
+   I830_RASTER_RULES,
+   I830_RASTER_RULES_SIZE
+};
+
 struct i830_texture_object
 {
    struct intel_texture_object intel;
@@ -112,6 +118,7 @@ struct i830_hw_state
    GLuint Ctx[I830_CTX_SETUP_SIZE];
    GLuint Buffer[I830_DEST_SETUP_SIZE];
    GLuint Stipple[I830_STP_SETUP_SIZE];
+   GLuint RasterRules[I830_RASTER_RULES_SIZE];
    GLuint Tex[I830_TEX_UNITS][I830_TEX_SETUP_SIZE];
    GLuint TexBlend[I830_TEX_UNITS][I830_TEXBLEND_SIZE];
    GLuint TexBlendWordsUsed[I830_TEX_UNITS];
@@ -197,6 +204,7 @@ extern void i830InitStateFuncs(struct dd_function_table *functions);
 extern void i830EmitState(struct i830_context *i830);
 
 extern void i830InitState(struct i830_context *i830);
+extern void i830_update_provoking_vertex(GLcontext *ctx);
 
 /* i830_metaops.c
  */
diff --git a/src/mesa/drivers/dri/i915/i830_reg.h b/src/mesa/drivers/dri/i915/i830_reg.h
index d210c2d08e..ae1317029a 100644
--- a/src/mesa/drivers/dri/i915/i830_reg.h
+++ b/src/mesa/drivers/dri/i915/i830_reg.h
@@ -48,19 +48,6 @@
 #define AA_LINE_ENABLE			((1<<1) | 1)
 #define AA_LINE_DISABLE			(1<<1)
 
-#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
-/* Dword 1 */
-#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
-#define BUF_3D_ID_DEPTH 	(0x7<<24)
-#define BUF_3D_USE_FENCE	(1<<23)
-#define BUF_3D_TILED_SURFACE	(1<<22)
-#define BUF_3D_TILE_WALK_X	0
-#define BUF_3D_TILE_WALK_Y	(1<<21)
-#define BUF_3D_PITCH(x)         (((x)/4)<<2)
-/* Dword 2 */
-#define BUF_3D_ADDR(x)		((x) & ~0x3)
-
-
 #define _3DSTATE_COLOR_FACTOR_CMD	(CMD_3D | (0x1d<<24) | (0x1<<16))
 
 #define _3DSTATE_COLOR_FACTOR_N_CMD(stage)	(CMD_3D | (0x1d<<24) | \
@@ -433,8 +420,11 @@
 #define ENABLE_LINE_STRIP_PROVOKE_VRTX	(1<<8)
 #define ENABLE_TRI_FAN_PROVOKE_VRTX	(1<<5)
 #define ENABLE_TRI_STRIP_PROVOKE_VRTX	(1<<2)
+#define LINE_STRIP_PROVOKE_VRTX_MASK	(3<<6)
 #define LINE_STRIP_PROVOKE_VRTX(x)	((x)<<6)
+#define TRI_FAN_PROVOKE_VRTX_MASK	(3<<3)
 #define TRI_FAN_PROVOKE_VRTX(x) 	((x)<<3)
+#define TRI_STRIP_PROVOKE_VRTX_MASK	(3<<0)
 #define TRI_STRIP_PROVOKE_VRTX(x)	(x)
 
 /* _3DSTATE_SCISSOR_ENABLE, p200 */
diff --git a/src/mesa/drivers/dri/i915/i830_state.c b/src/mesa/drivers/dri/i915/i830_state.c
index 8ef6c9144f..645ebe3057 100644
--- a/src/mesa/drivers/dri/i915/i830_state.c
+++ b/src/mesa/drivers/dri/i915/i830_state.c
@@ -1047,6 +1047,16 @@ i830_init_packets(struct i830_context *i830)
                                          TEXBIND_SET1(TEXCOORDSRC_VTXSET_1) |
                                          TEXBIND_SET0(TEXCOORDSRC_VTXSET_0));
 
+   i830->state.RasterRules[I830_RASTER_RULES] = (_3DSTATE_RASTER_RULES_CMD |
+						 ENABLE_POINT_RASTER_RULE |
+						 OGL_POINT_RASTER_RULE |
+						 ENABLE_LINE_STRIP_PROVOKE_VRTX |
+						 ENABLE_TRI_FAN_PROVOKE_VRTX |
+						 ENABLE_TRI_STRIP_PROVOKE_VRTX |
+						 LINE_STRIP_PROVOKE_VRTX(1) |
+						 TRI_FAN_PROVOKE_VRTX(2) |
+						 TRI_STRIP_PROVOKE_VRTX(2));
+
 
    i830->state.Stipple[I830_STPREG_ST0] = _3DSTATE_STIPPLE;
 
@@ -1058,6 +1068,27 @@ i830_init_packets(struct i830_context *i830)
    i830->state.Buffer[I830_DESTREG_SR2] = 0;
 }
 
+void
+i830_update_provoking_vertex(GLcontext * ctx)
+{
+   struct i830_context *i830 = i830_context(ctx);
+
+   I830_STATECHANGE(i830, I830_UPLOAD_RASTER_RULES);
+   i830->state.RasterRules[I830_RASTER_RULES] &= ~(LINE_STRIP_PROVOKE_VRTX_MASK |
+						   TRI_FAN_PROVOKE_VRTX_MASK |
+						   TRI_STRIP_PROVOKE_VRTX_MASK);
+
+   /* _NEW_LIGHT */
+   if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION) {
+      i830->state.RasterRules[I830_RASTER_RULES] |= (LINE_STRIP_PROVOKE_VRTX(1) |
+						     TRI_FAN_PROVOKE_VRTX(2) |
+						     TRI_STRIP_PROVOKE_VRTX(2));
+   } else {
+      i830->state.RasterRules[I830_RASTER_RULES] |= (LINE_STRIP_PROVOKE_VRTX(0) |
+						     TRI_FAN_PROVOKE_VRTX(1) |
+						     TRI_STRIP_PROVOKE_VRTX(0));
+    }
+}
 
 void
 i830InitStateFuncs(struct dd_function_table *functions)
@@ -1101,6 +1132,7 @@ i830InitState(struct i830_context *i830)
    i830->current = &i830->state;
    i830->state.emitted = 0;
    i830->state.active = (I830_UPLOAD_INVARIENT |
+                         I830_UPLOAD_RASTER_RULES |
                          I830_UPLOAD_TEXBLEND(0) |
                          I830_UPLOAD_STIPPLE |
                          I830_UPLOAD_CTX | I830_UPLOAD_BUFFERS);
diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c
index 753c25b57e..6f998fa6f7 100644
--- a/src/mesa/drivers/dri/i915/i830_texstate.c
+++ b/src/mesa/drivers/dri/i915/i830_texstate.c
@@ -174,14 +174,16 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 |
                                (LOAD_TEXTURE_MAP0 << unit) | 4);
 
-/*    state[I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | */
-/* 			       t->intel.TextureOffset); */
-
-
    state[I830_TEXREG_TM0S1] =
       (((firstImage->Height - 1) << TM0S1_HEIGHT_SHIFT) |
        ((firstImage->Width - 1) << TM0S1_WIDTH_SHIFT) | format);
 
+   if (intelObj->mt->region->tiling != I915_TILING_NONE) {
+      state[I830_TEXREG_TM0S1] |= TM0S1_TILED_SURFACE;
+      if (intelObj->mt->region->tiling == I915_TILING_Y)
+	 state[I830_TEXREG_TM0S1] |= TM0S1_TILE_WALK;
+   }
+
    state[I830_TEXREG_TM0S2] =
       ((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK);
 
diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c
index 3bf02de61f..983f6724c9 100644
--- a/src/mesa/drivers/dri/i915/i830_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i830_vtbl.c
@@ -299,7 +299,7 @@ i830_emit_invarient_state(struct intel_context *intel)
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(30, IGNORE_CLIPRECTS);
+   BEGIN_BATCH(29, IGNORE_CLIPRECTS);
 
    OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD);
    OUT_BATCH(0);
@@ -351,15 +351,6 @@ i830_emit_invarient_state(struct intel_context *intel)
    OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM);
    OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(3));
 
-   OUT_BATCH(_3DSTATE_RASTER_RULES_CMD |
-             ENABLE_POINT_RASTER_RULE |
-             OGL_POINT_RASTER_RULE |
-             ENABLE_LINE_STRIP_PROVOKE_VRTX |
-             ENABLE_TRI_FAN_PROVOKE_VRTX |
-             ENABLE_TRI_STRIP_PROVOKE_VRTX |
-             LINE_STRIP_PROVOKE_VRTX(1) |
-             TRI_FAN_PROVOKE_VRTX(2) | TRI_STRIP_PROVOKE_VRTX(2));
-
    OUT_BATCH(_3DSTATE_VERTEX_TRANSFORM);
    OUT_BATCH(DISABLE_VIEWPORT_TRANSFORM | DISABLE_PERSPECTIVE_DIVIDE);
 
@@ -394,6 +385,9 @@ get_state_size(struct i830_hw_state *state)
    if (dirty & I830_UPLOAD_INVARIENT)
       sz += 40 * sizeof(int);
 
+   if (dirty & I830_UPLOAD_RASTER_RULES)
+      sz += sizeof(state->RasterRules);
+
    if (dirty & I830_UPLOAD_CTX)
       sz += sizeof(state->Ctx);
 
@@ -486,6 +480,11 @@ i830_emit_state(struct intel_context *intel)
       i830_emit_invarient_state(intel);
    }
 
+   if (dirty & I830_UPLOAD_RASTER_RULES) {
+      DBG("I830_UPLOAD_RASTER_RULES:\n");
+      emit(intel, state->RasterRules, sizeof(state->RasterRules));
+   }
+
    if (dirty & I830_UPLOAD_CTX) {
       DBG("I830_UPLOAD_CTX:\n");
       emit(intel, state->Ctx, sizeof(state->Ctx));
@@ -552,7 +551,7 @@ i830_emit_state(struct intel_context *intel)
          if (state->tex_buffer[i]) {
             OUT_RELOC(state->tex_buffer[i],
 		      I915_GEM_DOMAIN_SAMPLER, 0,
-                      state->tex_offset[i] | TM0S0_USE_FENCE);
+                      state->tex_offset[i]);
          }
 	 else if (state == &i830->meta) {
 	    assert(i == 0);
@@ -634,21 +633,11 @@ i830_state_draw_region(struct intel_context *intel,
    /*
     * Set stride/cpp values
     */
-   if (color_region) {
-      state->Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I830_DESTREG_CBUFADDR1] =
-         (BUF_3D_ID_COLOR_BACK |
-          BUF_3D_PITCH(color_region->pitch * color_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I830_DESTREG_CBUFADDR0],
+				color_region, BUF_3D_ID_COLOR_BACK);
 
-   if (depth_region) {
-      state->Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I830_DESTREG_DBUFADDR1] =
-         (BUF_3D_ID_DEPTH |
-          BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I830_DESTREG_DBUFADDR0],
+				depth_region, BUF_3D_ID_DEPTH);
 
    /*
     * Compute/set I830_DESTREG_DV1 value
@@ -718,26 +707,6 @@ i830_set_draw_region(struct intel_context *intel,
    i830_state_draw_region(intel, &i830->state, color_regions[0], depth_region);
 }
 
-#if 0
-static void
-i830_update_color_z_regions(intelContextPtr intel,
-                            const intelRegion * colorRegion,
-                            const intelRegion * depthRegion)
-{
-   i830ContextPtr i830 = I830_CONTEXT(intel);
-
-   i830->state.Buffer[I830_DESTREG_CBUFADDR1] =
-      (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) |
-       BUF_3D_USE_FENCE);
-   i830->state.Buffer[I830_DESTREG_CBUFADDR2] = colorRegion->offset;
-
-   i830->state.Buffer[I830_DESTREG_DBUFADDR1] =
-      (BUF_3D_ID_DEPTH | BUF_3D_PITCH(depthRegion->pitch) | BUF_3D_USE_FENCE);
-   i830->state.Buffer[I830_DESTREG_DBUFADDR2] = depthRegion->offset;
-}
-#endif
-
-
 /* This isn't really handled at the moment.
  */
 static void
@@ -768,9 +737,10 @@ i830_assert_not_dirty( struct intel_context *intel )
 }
 
 static void
-i830_note_unlock( struct intel_context *intel )
+i830_invalidate_state(struct intel_context *intel, GLuint new_state)
 {
-    /* nothing */
+   if (new_state & _NEW_LIGHT)
+      i830_update_provoking_vertex(&intel->ctx);
 }
 
 void
@@ -787,6 +757,6 @@ i830InitVtbl(struct i830_context *i830)
    i830->intel.vtbl.render_start = i830_render_start;
    i830->intel.vtbl.render_prevalidate = i830_render_prevalidate;
    i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty;
-   i830->intel.vtbl.note_unlock = i830_note_unlock; 
    i830->intel.vtbl.finish_batch = intel_finish_vb;
+   i830->intel.vtbl.invalidate_state = i830_invalidate_state;
 }
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index 1f9f363df9..3ab7d682ee 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -27,6 +27,7 @@
 
 #include "i915_context.h"
 #include "main/imports.h"
+#include "main/macros.h"
 #include "intel_tex.h"
 #include "intel_tris.h"
 #include "tnl/t_context.h"
@@ -73,8 +74,12 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
          p->params_uptodate = 0;
    }
 
-   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM))
+   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS))
       i915_update_fog(ctx);
+   if (new_state & (_NEW_STENCIL | _NEW_BUFFERS | _NEW_POLYGON))
+      i915_update_stencil(ctx);
+   if (new_state & (_NEW_LIGHT))
+       i915_update_provoking_vertex(ctx);
 }
 
 
@@ -162,6 +167,9 @@ i915CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.FragmentProgram.MaxNativeTexIndirections =
       I915_MAX_TEX_INDIRECT;
    ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* I don't think we have one */
+   ctx->Const.FragmentProgram.MaxEnvParams =
+      MIN2(ctx->Const.FragmentProgram.MaxNativeParameters,
+	   ctx->Const.FragmentProgram.MaxEnvParams);
 
    ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h
index 87bbf5f927..8de4a9d0d3 100644
--- a/src/mesa/drivers/dri/i915/i915_context.h
+++ b/src/mesa/drivers/dri/i915/i915_context.h
@@ -48,6 +48,7 @@
 #define I915_UPLOAD_FOG              0x20
 #define I915_UPLOAD_INVARIENT        0x40
 #define I915_UPLOAD_DEFAULTS         0x80
+#define I915_UPLOAD_RASTER_RULES     0x100
 #define I915_UPLOAD_TEX(i)           (0x00010000<<(i))
 #define I915_UPLOAD_TEX_ALL          (0x00ff0000)
 #define I915_UPLOAD_TEX_0_SHIFT      16
@@ -82,7 +83,9 @@
 #define I915_CTXREG_IAB   	 	6
 #define I915_CTXREG_BLENDCOLOR0		7
 #define I915_CTXREG_BLENDCOLOR1		8
-#define I915_CTX_SETUP_SIZE		9
+#define I915_CTXREG_BF_STENCIL_OPS	9
+#define I915_CTXREG_BF_STENCIL_MASKS	10
+#define I915_CTX_SETUP_SIZE		11
 
 #define I915_FOGREG_COLOR		0
 #define I915_FOGREG_MODE0		1
@@ -110,6 +113,10 @@
 #define I915_DEFREG_Z1    5
 #define I915_DEF_SETUP_SIZE    6
 
+enum {
+   I915_RASTER_RULES,
+   I915_RASTER_RULES_SETUP_SIZE,
+};
 
 #define I915_MAX_CONSTANT      32
 #define I915_CONSTANT_SIZE     (2+(4*I915_MAX_CONSTANT))
@@ -206,6 +213,7 @@ struct i915_hw_state
    GLuint Stipple[I915_STP_SETUP_SIZE];
    GLuint Fog[I915_FOG_SETUP_SIZE];
    GLuint Defaults[I915_DEF_SETUP_SIZE];
+   GLuint RasterRules[I915_RASTER_RULES_SETUP_SIZE];
    GLuint Tex[I915_TEX_UNITS][I915_TEX_SETUP_SIZE];
    GLuint Constant[I915_CONSTANT_SIZE];
    GLuint ConstantSize;
@@ -321,6 +329,8 @@ extern void i915_print_ureg(const char *msg, GLuint ureg);
 extern void i915InitStateFunctions(struct dd_function_table *functions);
 extern void i915InitState(struct i915_context *i915);
 extern void i915_update_fog(GLcontext * ctx);
+extern void i915_update_stencil(GLcontext * ctx);
+extern void i915_update_provoking_vertex(GLcontext *ctx);
 
 
 /*======================================================================
diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h
index 8891e11c6f..b5fa7fddb9 100644
--- a/src/mesa/drivers/dri/i915/i915_reg.h
+++ b/src/mesa/drivers/dri/i915/i915_reg.h
@@ -86,27 +86,15 @@
 #define BFM_ENABLE_STENCIL_WRITE_MASK     (1<<16)
 #define BFM_STENCIL_TEST_MASK_SHIFT       8
 #define BFM_STENCIL_TEST_MASK_MASK        (0xff<<8)
+#define BFM_STENCIL_TEST_MASK(x)	  (((x)&0xff) << 8)
 #define BFM_STENCIL_WRITE_MASK_SHIFT      0
 #define BFM_STENCIL_WRITE_MASK_MASK       (0xff<<0)
+#define BFM_STENCIL_WRITE_MASK(x)	  ((x)&0xff)
 
 
 
 /* 3DSTATE_BIN_CONTROL p141 */
 
-/* p143 */
-#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
-/* Dword 1 */
-#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
-#define BUF_3D_ID_DEPTH 	(0x7<<24)
-#define BUF_3D_USE_FENCE	(1<<23)
-#define BUF_3D_TILED_SURFACE	(1<<22)
-#define BUF_3D_TILE_WALK_X	0
-#define BUF_3D_TILE_WALK_Y	(1<<21)
-#define BUF_3D_PITCH(x)         (((x)/4)<<2)
-/* Dword 2 */
-#define BUF_3D_ADDR(x)		((x) & ~0x3)
-
-
 /* 3DSTATE_CHROMA_KEY */
 
 /* 3DSTATE_CLEAR_PARAMETERS, p150 */
@@ -155,6 +143,7 @@
 /* p161 */
 #define _3DSTATE_DST_BUF_VARS_CMD	(CMD_3D | (0x1d<<24) | (0x85<<16))
 /* Dword 1 */
+#define CLASSIC_EARLY_DEPTH		(1<<31)
 #define TEX_DEFAULT_COLOR_OGL           (0<<30)
 #define TEX_DEFAULT_COLOR_D3D           (1<<30)
 #define ZR_EARLY_DEPTH                  (1<<29)
@@ -308,7 +297,9 @@
 #define TEXKILL_4D                      (1<<9)
 #define ENABLE_LINE_STRIP_PROVOKE_VRTX	(1<<8)
 #define ENABLE_TRI_FAN_PROVOKE_VRTX	(1<<5)
+#define LINE_STRIP_PROVOKE_VRTX_MASK	(3 << 6)
 #define LINE_STRIP_PROVOKE_VRTX(x)	((x)<<6)
+#define TRI_FAN_PROVOKE_VRTX_MASK	(3 << 3)
 #define TRI_FAN_PROVOKE_VRTX(x) 	((x)<<3)
 
 /* _3DSTATE_SCISSOR_ENABLE, p256 */
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index 814fb59fd3..b60efea75b 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -48,73 +48,119 @@
 
 #define FILE_DEBUG_FLAG DEBUG_STATE
 
-static void
-i915StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref,
-                        GLuint mask)
+void
+i915_update_stencil(GLcontext * ctx)
 {
    struct i915_context *i915 = I915_CONTEXT(ctx);
-   int test = intel_translate_compare_func(func);
+   GLuint front_ref, front_writemask, front_mask;
+   GLenum front_func, front_fail, front_pass_z_fail, front_pass_z_pass;
+   GLuint back_ref, back_writemask, back_mask;
+   GLenum back_func, back_fail, back_pass_z_fail, back_pass_z_pass;
 
-   mask = mask & 0xff;
-
-   DBG("%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__,
-       _mesa_lookup_enum_by_nr(func), ref, mask);
+   I915_STATECHANGE(i915, I915_UPLOAD_CTX);
 
+   /* The 915 considers CW to be "front" for two-sided stencil, so choose
+    * appropriately.
+    */
+   /* _NEW_POLYGON | _NEW_STENCIL */
+   if (ctx->Polygon.FrontFace == GL_CW) {
+      front_ref = ctx->Stencil.Ref[0];
+      front_mask = ctx->Stencil.ValueMask[0];
+      front_writemask = ctx->Stencil.WriteMask[0];
+      front_func = ctx->Stencil.Function[0];
+      front_fail = ctx->Stencil.FailFunc[0];
+      front_pass_z_fail = ctx->Stencil.ZFailFunc[0];
+      front_pass_z_pass = ctx->Stencil.ZPassFunc[0];
+      back_ref = ctx->Stencil.Ref[ctx->Stencil._BackFace];
+      back_mask = ctx->Stencil.ValueMask[ctx->Stencil._BackFace];
+      back_writemask = ctx->Stencil.WriteMask[ctx->Stencil._BackFace];
+      back_func = ctx->Stencil.Function[ctx->Stencil._BackFace];
+      back_fail = ctx->Stencil.FailFunc[ctx->Stencil._BackFace];
+      back_pass_z_fail = ctx->Stencil.ZFailFunc[ctx->Stencil._BackFace];
+      back_pass_z_pass = ctx->Stencil.ZPassFunc[ctx->Stencil._BackFace];
+   } else {
+      front_ref = ctx->Stencil.Ref[ctx->Stencil._BackFace];
+      front_mask = ctx->Stencil.ValueMask[ctx->Stencil._BackFace];
+      front_writemask = ctx->Stencil.WriteMask[ctx->Stencil._BackFace];
+      front_func = ctx->Stencil.Function[ctx->Stencil._BackFace];
+      front_fail = ctx->Stencil.FailFunc[ctx->Stencil._BackFace];
+      front_pass_z_fail = ctx->Stencil.ZFailFunc[ctx->Stencil._BackFace];
+      front_pass_z_pass = ctx->Stencil.ZPassFunc[ctx->Stencil._BackFace];
+      back_ref = ctx->Stencil.Ref[0];
+      back_mask = ctx->Stencil.ValueMask[0];
+      back_writemask = ctx->Stencil.WriteMask[0];
+      back_func = ctx->Stencil.Function[0];
+      back_fail = ctx->Stencil.FailFunc[0];
+      back_pass_z_fail = ctx->Stencil.ZFailFunc[0];
+      back_pass_z_pass = ctx->Stencil.ZPassFunc[0];
+   }
 
-   I915_STATECHANGE(i915, I915_UPLOAD_CTX);
-   i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK;
+   /* Set front state. */
+   i915->state.Ctx[I915_CTXREG_STATE4] &= ~(MODE4_ENABLE_STENCIL_TEST_MASK |
+					    MODE4_ENABLE_STENCIL_WRITE_MASK);
    i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK |
-                                           STENCIL_TEST_MASK(mask));
+					   ENABLE_STENCIL_WRITE_MASK |
+					   STENCIL_TEST_MASK(front_mask) |
+					   STENCIL_WRITE_MASK(front_writemask));
 
    i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_REF_MASK |
-                                          S5_STENCIL_TEST_FUNC_MASK);
+					  S5_STENCIL_TEST_FUNC_MASK |
+					  S5_STENCIL_FAIL_MASK |
+					  S5_STENCIL_PASS_Z_FAIL_MASK |
+					  S5_STENCIL_PASS_Z_PASS_MASK);
+
+   i915->state.Ctx[I915_CTXREG_LIS5] |=
+      (front_ref << S5_STENCIL_REF_SHIFT) |
+      (intel_translate_compare_func(front_func) << S5_STENCIL_TEST_FUNC_SHIFT) |
+      (intel_translate_stencil_op(front_fail) << S5_STENCIL_FAIL_SHIFT) |
+      (intel_translate_stencil_op(front_pass_z_fail) <<
+       S5_STENCIL_PASS_Z_FAIL_SHIFT) |
+      (intel_translate_stencil_op(front_pass_z_pass) <<
+       S5_STENCIL_PASS_Z_PASS_SHIFT);
+
+   /* Set back state if different from front. */
+   if (ctx->Stencil._TestTwoSide) {
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_OPS] &=
+	 ~(BFO_STENCIL_REF_MASK |
+	   BFO_STENCIL_TEST_MASK |
+	   BFO_STENCIL_FAIL_MASK |
+	   BFO_STENCIL_PASS_Z_FAIL_MASK |
+	   BFO_STENCIL_PASS_Z_PASS_MASK);
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_OPS] |= BFO_STENCIL_TWO_SIDE |
+	 (back_ref << BFO_STENCIL_REF_SHIFT) |
+	 (intel_translate_compare_func(back_func) << BFO_STENCIL_TEST_SHIFT) |
+	 (intel_translate_stencil_op(back_fail) << BFO_STENCIL_FAIL_SHIFT) |
+	 (intel_translate_stencil_op(back_pass_z_fail) <<
+	  BFO_STENCIL_PASS_Z_FAIL_SHIFT) |
+	 (intel_translate_stencil_op(back_pass_z_pass) <<
+	  BFO_STENCIL_PASS_Z_PASS_SHIFT);
+
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_MASKS] &=
+	 ~(BFM_STENCIL_TEST_MASK_MASK |
+	   BFM_STENCIL_WRITE_MASK_MASK);
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_MASKS] |=
+	 BFM_STENCIL_TEST_MASK(back_mask) |
+	 BFM_STENCIL_WRITE_MASK(back_writemask);
+   } else {
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_OPS] &= ~BFO_STENCIL_TWO_SIDE;
+   }
+}
 
-   i915->state.Ctx[I915_CTXREG_LIS5] |= ((ref << S5_STENCIL_REF_SHIFT) |
-                                         (test <<
-                                          S5_STENCIL_TEST_FUNC_SHIFT));
+static void
+i915StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref,
+                        GLuint mask)
+{
 }
 
 static void
 i915StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask)
 {
-   struct i915_context *i915 = I915_CONTEXT(ctx);
-
-   DBG("%s : mask 0x%x\n", __FUNCTION__, mask);
-   
-   mask = mask & 0xff;
-
-   I915_STATECHANGE(i915, I915_UPLOAD_CTX);
-   i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK;
-   i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK |
-                                           STENCIL_WRITE_MASK(mask));
 }
 
-
 static void
 i915StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail, GLenum zfail,
                       GLenum zpass)
 {
-   struct i915_context *i915 = I915_CONTEXT(ctx);
-   int fop = intel_translate_stencil_op(fail);
-   int dfop = intel_translate_stencil_op(zfail);
-   int dpop = intel_translate_stencil_op(zpass);
-
-
-   DBG("%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__,
-       _mesa_lookup_enum_by_nr(fail),
-       _mesa_lookup_enum_by_nr(zfail), _mesa_lookup_enum_by_nr(zpass));
-
-   I915_STATECHANGE(i915, I915_UPLOAD_CTX);
-
-   i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK |
-                                          S5_STENCIL_PASS_Z_FAIL_MASK |
-                                          S5_STENCIL_PASS_Z_PASS_MASK);
-
-   i915->state.Ctx[I915_CTXREG_LIS5] |= ((fop << S5_STENCIL_FAIL_SHIFT) |
-                                         (dfop <<
-                                          S5_STENCIL_PASS_Z_FAIL_SHIFT) |
-                                         (dpop <<
-                                          S5_STENCIL_PASS_Z_PASS_SHIFT));
 }
 
 static void
@@ -945,6 +991,17 @@ i915_init_packets(struct i915_context *i915)
          _3DSTATE_CONST_BLEND_COLOR_CMD;
       i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = 0;
 
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_MASKS] =
+	 _3DSTATE_BACKFACE_STENCIL_MASKS |
+	 BFM_ENABLE_STENCIL_TEST_MASK |
+	 BFM_ENABLE_STENCIL_WRITE_MASK |
+	 (0xff << BFM_STENCIL_WRITE_MASK_SHIFT) |
+	 (0xff << BFM_STENCIL_TEST_MASK_SHIFT);
+      i915->state.Ctx[I915_CTXREG_BF_STENCIL_OPS] =
+	 _3DSTATE_BACKFACE_STENCIL_OPS |
+	 BFO_ENABLE_STENCIL_REF |
+	 BFO_ENABLE_STENCIL_FUNCS |
+	 BFO_ENABLE_STENCIL_TWO_SIDE;
    }
 
    {
@@ -976,6 +1033,13 @@ i915_init_packets(struct i915_context *i915)
       i915->state.Buffer[I915_DESTREG_SR2] = 0;
    }
 
+   i915->state.RasterRules[I915_RASTER_RULES] = _3DSTATE_RASTER_RULES_CMD |
+      ENABLE_POINT_RASTER_RULE |
+      OGL_POINT_RASTER_RULE |
+      ENABLE_LINE_STRIP_PROVOKE_VRTX |
+      ENABLE_TRI_FAN_PROVOKE_VRTX |
+      LINE_STRIP_PROVOKE_VRTX(1) |
+      TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D;
 
 #if 0
    {
@@ -996,7 +1060,33 @@ i915_init_packets(struct i915_context *i915)
    i915->state.active = (I915_UPLOAD_PROGRAM |
                          I915_UPLOAD_STIPPLE |
                          I915_UPLOAD_CTX |
-                         I915_UPLOAD_BUFFERS | I915_UPLOAD_INVARIENT);
+                         I915_UPLOAD_BUFFERS |
+			 I915_UPLOAD_INVARIENT |
+			 I915_UPLOAD_RASTER_RULES);
+}
+
+void
+i915_update_provoking_vertex(GLcontext * ctx)
+{
+   struct i915_context *i915 = I915_CONTEXT(ctx);
+
+   I915_STATECHANGE(i915, I915_UPLOAD_CTX);
+   i915->state.Ctx[I915_CTXREG_LIS6] &= ~(S6_TRISTRIP_PV_MASK);
+
+   I915_STATECHANGE(i915, I915_UPLOAD_RASTER_RULES);
+   i915->state.RasterRules[I915_RASTER_RULES] &= ~(LINE_STRIP_PROVOKE_VRTX_MASK |
+						   TRI_FAN_PROVOKE_VRTX_MASK);
+
+   /* _NEW_LIGHT */
+   if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION) {
+      i915->state.RasterRules[I915_RASTER_RULES] |= (LINE_STRIP_PROVOKE_VRTX(1) |
+						     TRI_FAN_PROVOKE_VRTX(2));
+      i915->state.Ctx[I915_CTXREG_LIS6] |= (2 << S6_TRISTRIP_PV_SHIFT);
+   } else {
+      i915->state.RasterRules[I915_RASTER_RULES] |= (LINE_STRIP_PROVOKE_VRTX(0) |
+						     TRI_FAN_PROVOKE_VRTX(1));
+      i915->state.Ctx[I915_CTXREG_LIS6] |= (0 << S6_TRISTRIP_PV_SHIFT);
+    }
 }
 
 void
diff --git a/src/mesa/drivers/dri/i915/i915_tex_layout.c b/src/mesa/drivers/dri/i915/i915_tex_layout.c
index 7cc1c096e4..d9588e5b56 100644
--- a/src/mesa/drivers/dri/i915/i915_tex_layout.c
+++ b/src/mesa/drivers/dri/i915/i915_tex_layout.c
@@ -55,6 +55,17 @@ static GLint step_offsets[6][2] = {
    [FACE_NEG_Z] = {-1, 1},
 };
 
+
+static GLint bottom_offsets[6] = {
+   [FACE_POS_X] = 16 + 0 * 8,
+   [FACE_POS_Y] = 16 + 1 * 8,
+   [FACE_POS_Z] = 16 + 2 * 8,
+   [FACE_NEG_X] = 16 + 3 * 8,
+   [FACE_NEG_Y] = 16 + 4 * 8,
+   [FACE_NEG_Z] = 16 + 5 * 8,
+};
+
+
 /**
  * Cube texture map layout for i830M-GM915.
  *
@@ -101,7 +112,8 @@ static GLint step_offsets[6][2] = {
  */
 static void
 i915_miptree_layout_cube(struct intel_context *intel,
-			 struct intel_mipmap_tree * mt)
+			 struct intel_mipmap_tree * mt,
+			 uint32_t tiling)
 {
    const GLuint dim = mt->width0;
    GLuint face;
@@ -111,7 +123,7 @@ i915_miptree_layout_cube(struct intel_context *intel,
    assert(lvlWidth == lvlHeight); /* cubemap images are square */
 
    /* double pitch for cube layouts */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, dim * 2);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
    mt->total_height = dim * 4;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -145,7 +157,8 @@ i915_miptree_layout_cube(struct intel_context *intel,
 
 static void
 i915_miptree_layout_3d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
@@ -154,7 +167,7 @@ i915_miptree_layout_3d(struct intel_context *intel,
    GLint level;
 
    /* Calculate the size of a single slice. */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
 
    /* XXX: hardware expects/requires 9 levels at minimum. */
    for (level = mt->first_level; level <= MAX2(8, mt->last_level); level++) {
@@ -189,14 +202,15 @@ i915_miptree_layout_3d(struct intel_context *intel,
 
 static void
 i915_miptree_layout_2d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
    GLuint img_height;
    GLint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
    mt->total_height = 0;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -217,19 +231,20 @@ i915_miptree_layout_2d(struct intel_context *intel,
 }
 
 GLboolean
-i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
+i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
+		    uint32_t tiling)
 {
    switch (mt->target) {
    case GL_TEXTURE_CUBE_MAP:
-      i915_miptree_layout_cube(intel, mt);
+      i915_miptree_layout_cube(intel, mt, tiling);
       break;
    case GL_TEXTURE_3D:
-      i915_miptree_layout_3d(intel, mt);
+      i915_miptree_layout_3d(intel, mt, tiling);
       break;
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_TEXTURE_RECTANGLE_ARB:
-      i915_miptree_layout_2d(intel, mt);
+      i915_miptree_layout_2d(intel, mt, tiling);
       break;
    default:
       _mesa_problem(NULL, "Unexpected tex target in i915_miptree_layout()");
@@ -297,7 +312,7 @@ i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
  * +---+   +---+   +---+   +---+   +---+   +---+
  *
  * The bottom row continues with the remaining 2x2 then the 1x1 mip contents
- * in order, with each of them aligned to a 4x4 block boundary.  Thus, for
+ * in order, with each of them aligned to a 8x8 block boundary.  Thus, for
  * 32x32 cube maps and smaller, the bottom row layout is going to dictate the
  * pitch of the tree.  For a tree with 4x4 images, the pitch is at least
  * 14 * 8 = 112 texels, for 2x2 it is at least 12 * 8 texels, and for 1x1
@@ -306,7 +321,8 @@ i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
 
 static void
 i945_miptree_layout_cube(struct intel_context *intel,
-			 struct intel_mipmap_tree * mt)
+			 struct intel_mipmap_tree * mt,
+			 uint32_t tiling)
 {
    const GLuint dim = mt->width0;
    GLuint face;
@@ -320,9 +336,9 @@ i945_miptree_layout_cube(struct intel_context *intel,
     * or the final row of 4x4, 2x2 and 1x1 faces below this.
     */
    if (dim > 32)
-      mt->pitch = intel_miptree_pitch_align (intel, mt, dim * 2);
+      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
    else
-      mt->pitch = intel_miptree_pitch_align (intel, mt, 14 * 8);
+      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, 14 * 8);
 
    if (dim >= 4)
       mt->total_height = dim * 4 + 4;
@@ -375,10 +391,11 @@ i945_miptree_layout_cube(struct intel_context *intel,
 	       x = (face - 4) * 8;
 	       break;
 	    }
+	    break;
 
 	 case 2:
 	    y = mt->total_height - 4;
-	    x = 16 + face * 8;
+	    x = bottom_offsets[face];
 	    break;
 
 	 case 1:
@@ -396,7 +413,8 @@ i945_miptree_layout_cube(struct intel_context *intel,
 
 static void
 i945_miptree_layout_3d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
@@ -405,7 +423,7 @@ i945_miptree_layout_3d(struct intel_context *intel,
    GLuint pack_y_pitch;
    GLuint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
    mt->total_height = 0;
 
    pack_y_pitch = MAX2(mt->height0, 2);
@@ -450,22 +468,23 @@ i945_miptree_layout_3d(struct intel_context *intel,
 }
 
 GLboolean
-i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
+i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
+		    uint32_t tiling)
 {
    switch (mt->target) {
    case GL_TEXTURE_CUBE_MAP:
       if (mt->compressed)
-	 i945_miptree_layout_cube(intel, mt);
+	 i945_miptree_layout_cube(intel, mt, tiling);
       else
-	 i915_miptree_layout_cube(intel, mt);
+	 i915_miptree_layout_cube(intel, mt, tiling);
       break;
    case GL_TEXTURE_3D:
-      i945_miptree_layout_3d(intel, mt);
+      i945_miptree_layout_3d(intel, mt, tiling);
       break;
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_TEXTURE_RECTANGLE_ARB:
-      i945_miptree_layout_2d(intel, mt);
+      i945_miptree_layout_2d(intel, mt, tiling);
       break;
    default:
       _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()");
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index a37dd7f4fb..32d4b30cf9 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -185,8 +185,13 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
    state[I915_TEXREG_MS3] =
       (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) |
-       ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format |
-       MS3_USE_FENCE_REGS);
+       ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format);
+
+   if (intelObj->mt->region->tiling != I915_TILING_NONE) {
+      state[I915_TEXREG_MS3] |= MS3_TILED_SURFACE;
+      if (intelObj->mt->region->tiling == I915_TILING_Y)
+	 state[I915_TEXREG_MS3] |= MS3_TILE_WALK;
+   }
 
    state[I915_TEXREG_MS4] =
      ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK |
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index 115004616f..9a723d3cd7 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -42,6 +42,7 @@
 #include "intel_regions.h"
 #include "intel_tris.h"
 #include "intel_fbo.h"
+#include "intel_chipset.h"
 
 #include "i915_reg.h"
 #include "i915_context.h"
@@ -175,7 +176,7 @@ i915_emit_invarient_state(struct intel_context *intel)
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(20, IGNORE_CLIPRECTS);
+   BEGIN_BATCH(17, IGNORE_CLIPRECTS);
 
    OUT_BATCH(_3DSTATE_AA_CMD |
              AA_LINE_ECAAR_WIDTH_ENABLE |
@@ -199,14 +200,6 @@ i915_emit_invarient_state(struct intel_context *intel)
              CSB_TCB(3, 3) |
              CSB_TCB(4, 4) | CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7));
 
-   OUT_BATCH(_3DSTATE_RASTER_RULES_CMD |
-             ENABLE_POINT_RASTER_RULE |
-             OGL_POINT_RASTER_RULE |
-             ENABLE_LINE_STRIP_PROVOKE_VRTX |
-             ENABLE_TRI_FAN_PROVOKE_VRTX |
-             LINE_STRIP_PROVOKE_VRTX(1) |
-             TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D);
-
    /* Need to initialize this to zero.
     */
    OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0));
@@ -224,11 +217,6 @@ i915_emit_invarient_state(struct intel_context *intel)
    OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0);       /* disable indirect state */
    OUT_BATCH(0);
 
-
-   /* Don't support twosided stencil yet */
-   OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0);
-   OUT_BATCH(0);
-
    ADVANCE_BATCH();
 }
 
@@ -262,6 +250,9 @@ get_state_size(struct i915_hw_state *state)
    if (dirty & I915_UPLOAD_INVARIENT)
       sz += 30 * 4;
 
+   if (dirty & I915_UPLOAD_RASTER_RULES)
+      sz += sizeof(state->RasterRules);
+
    if (dirty & I915_UPLOAD_CTX)
       sz += sizeof(state->Ctx);
 
@@ -370,6 +361,12 @@ i915_emit_state(struct intel_context *intel)
       i915_emit_invarient_state(intel);
    }
 
+   if (dirty & I915_UPLOAD_RASTER_RULES) {
+      if (INTEL_DEBUG & DEBUG_STATE)
+         fprintf(stderr, "I915_UPLOAD_RASTER_RULES:\n");
+      emit(intel, state->RasterRules, sizeof(state->RasterRules));
+   }
+
    if (dirty & I915_UPLOAD_CTX) {
       if (INTEL_DEBUG & DEBUG_STATE)
          fprintf(stderr, "I915_UPLOAD_CTX:\n");
@@ -529,6 +526,23 @@ i915_destroy_context(struct intel_context *intel)
    _tnl_free_vertices(&intel->ctx);
 }
 
+void
+i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
+			     uint32_t buffer_id)
+{
+   state[0] = _3DSTATE_BUF_INFO_CMD;
+   state[1] = buffer_id;
+
+   if (region != NULL) {
+      state[1] |= BUF_3D_PITCH(region->pitch * region->cpp);
+
+      if (region->tiling != I915_TILING_NONE) {
+	 state[1] |= BUF_3D_TILED_SURFACE;
+	 if (region->tiling == I915_TILING_Y)
+	    state[1] |= BUF_3D_TILE_WALK_Y;
+      }
+   }
+}
 
 /**
  * Set the drawing regions for the color and depth/stencil buffers.
@@ -562,21 +576,11 @@ i915_state_draw_region(struct intel_context *intel,
    /*
     * Set stride/cpp values
     */
-   if (color_region) {
-      state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I915_DESTREG_CBUFADDR1] =
-         (BUF_3D_ID_COLOR_BACK |
-          BUF_3D_PITCH(color_region->pitch * color_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I915_DESTREG_CBUFADDR0],
+				color_region, BUF_3D_ID_COLOR_BACK);
 
-   if (depth_region) {
-      state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I915_DESTREG_DBUFADDR1] =
-         (BUF_3D_ID_DEPTH |
-          BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I915_DESTREG_DBUFADDR0],
+				depth_region, BUF_3D_ID_DEPTH);
 
    /*
     * Compute/set I915_DESTREG_DV1 value
@@ -604,6 +608,14 @@ i915_state_draw_region(struct intel_context *intel,
       }
    }
 
+   /* This isn't quite safe, thus being hidden behind an option.  When changing
+    * the value of this bit, the pipeline needs to be MI_FLUSHed.  And it
+    * can only be set when a depth buffer is already defined.
+    */
+   if (IS_945(intel->intelScreen->deviceID) && intel->use_early_z &&
+       depth_region->tiling != I915_TILING_NONE)
+      value |= CLASSIC_EARLY_DEPTH;
+
    if (depth_region && depth_region->cpp == 4) {
       value |= DEPTH_FRMT_24_FIXED_8_OTHER;
    }
@@ -676,13 +688,6 @@ i915_assert_not_dirty( struct intel_context *intel )
    assert(!dirty);
 }
 
-static void
-i915_note_unlock( struct intel_context *intel )
-{
-    /* nothing */
-}
-
-
 void
 i915InitVtbl(struct i915_context *i915)
 {
@@ -697,6 +702,5 @@ i915InitVtbl(struct i915_context *i915)
    i915->intel.vtbl.update_texture_state = i915UpdateTextureState;
    i915->intel.vtbl.flush_cmd = i915_flush_cmd;
    i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty;
-   i915->intel.vtbl.note_unlock = i915_note_unlock; 
    i915->intel.vtbl.finish_batch = intel_finish_vb;
 }
diff --git a/src/mesa/drivers/dri/i915/intel_generatemipmap.c b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
new file mode 120000
index 0000000000..4c6b37ada0
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_syncobj.c b/src/mesa/drivers/dri/i915/intel_syncobj.c
new file mode 120000
index 0000000000..0b2e56ab24
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_syncobj.c
@@ -0,0 +1 @@
+../intel/intel_syncobj.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index 1d39278cbf..a905455342 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -1255,11 +1255,9 @@ intel_meta_draw_poly(struct intel_context *intel,
 {
    union fi *vb;
    GLint i;
-   GLboolean was_locked = intel->locked;
    unsigned int saved_vertex_size = intel->vertex_size;
 
-   if (!was_locked)
-       LOCK_HARDWARE(intel);
+   LOCK_HARDWARE(intel);
 
    intel->vertex_size = 6;
 
@@ -1283,8 +1281,7 @@ intel_meta_draw_poly(struct intel_context *intel,
 
    intel->vertex_size = saved_vertex_size;
 
-   if (!was_locked)
-       UNLOCK_HARDWARE(intel);
+   UNLOCK_HARDWARE(intel);
 }
 
 static void
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 81c2fdb0dc..6e9a9a29a3 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -14,6 +14,7 @@ DRIVER_SOURCES = \
 	intel_decode.c \
 	intel_extensions.c \
 	intel_fbo.c \
+	intel_generatemipmap.c \
 	intel_mipmap_tree.c \
 	intel_regions.c \
 	intel_screen.c \
@@ -25,6 +26,7 @@ DRIVER_SOURCES = \
 	intel_pixel_read.c \
 	intel_state.c \
 	intel_swapbuffers.c \
+	intel_syncobj.c \
 	intel_tex.c \
 	intel_tex_copy.c \
 	intel_tex_format.c \
@@ -42,6 +44,7 @@ DRIVER_SOURCES = \
 	brw_clip_util.c \
 	brw_context.c \
 	brw_curbe.c \
+	brw_disasm.c \
 	brw_draw.c \
 	brw_draw_upload.c \
 	brw_eu.c \
@@ -70,6 +73,7 @@ DRIVER_SOURCES = \
 	brw_vs_constval.c \
 	brw_vs_emit.c \
 	brw_vs_state.c \
+	brw_vs_surface_state.c \
 	brw_vtbl.c \
 	brw_wm.c \
 	brw_wm_debug.c \
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 8fc9f89cb7..20a927cf38 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -65,21 +65,31 @@ static void compile_clip_prog( struct brw_context *brw,
    c.func.single_program_flow = 1;
 
    c.key = *key;
-
+   c.need_ff_sync = BRW_IS_IGDNG(brw);
 
    /* Need to locate the two positions present in vertex + header.
     * These are currently hardcoded:
     */
    c.header_position_offset = ATTR_SIZE;
 
-   for (i = 0, delta = REG_SIZE; i < VERT_RESULT_MAX; i++)
+   if (BRW_IS_IGDNG(brw))
+       delta = 3 * REG_SIZE;
+   else
+       delta = REG_SIZE;
+
+   for (i = 0; i < VERT_RESULT_MAX; i++)
       if (c.key.attrs & (1<<i)) {
 	 c.offset[i] = delta;
 	 delta += ATTR_SIZE;
       }
 
    c.nr_attrs = brw_count_bits(c.key.attrs);
-   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+   
+   if (BRW_IS_IGDNG(brw))
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 3;  /* are vertices packed, or reg-aligned? */
+   else
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
    c.nr_bytes = c.nr_regs * REG_SIZE;
 
    c.prog_data.clip_mode = c.key.clip_mode; /* XXX */
@@ -148,7 +158,11 @@ static void upload_clip_prog(struct brw_context *brw)
    key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
    /* _NEW_TRANSFORM */
    key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
-   key.clip_mode = BRW_CLIPMODE_NORMAL;
+
+   if (BRW_IS_IGDNG(brw))
+       key.clip_mode = BRW_CLIPMODE_KERNEL_CLIP;
+   else
+       key.clip_mode = BRW_CLIPMODE_NORMAL;
 
    /* _NEW_POLYGON */
    if (key.primitive == GL_TRIANGLES) {
diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h
index e06747864b..957df441ab 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.h
+++ b/src/mesa/drivers/dri/i965/brw_clip.h
@@ -100,6 +100,8 @@ struct brw_clip_compile {
       
       struct brw_reg fixed_planes;
       struct brw_reg plane_equation;
+       
+      struct brw_reg ff_sync;
    } reg;
 
    /* 3 different ways of expressing vertex size:
@@ -117,6 +119,7 @@ struct brw_clip_compile {
 
    GLuint header_position_offset;
    GLuint offset[VERT_ATTRIB_MAX];
+   GLboolean need_ff_sync;
 };
 
 #define ATTR_SIZE  (4*4)
@@ -171,5 +174,6 @@ struct brw_reg get_tmp( struct brw_clip_compile *c );
 
 void brw_clip_project_position(struct brw_clip_compile *c,
              struct brw_reg pos );
-
+void brw_clip_ff_sync(struct brw_clip_compile *c);
+void brw_clip_init_ff_sync(struct brw_clip_compile *c);
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_clip_line.c b/src/mesa/drivers/dri/i965/brw_clip_line.c
index d830e49e50..048ca620fa 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_line.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_line.c
@@ -85,6 +85,10 @@ static void brw_clip_line_alloc_regs( struct brw_clip_compile *c )
       i++;
    }
 
+   if (c->need_ff_sync) {
+      c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
+      i++;
+   }
 
    c->first_tmp = i;
    c->last_tmp = i;
@@ -130,7 +134,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
    struct brw_instruction *plane_loop;
    struct brw_instruction *plane_active;
    struct brw_instruction *is_negative;
-   struct brw_instruction *is_neg2;
+   struct brw_instruction *is_neg2 = NULL;
    struct brw_instruction *not_culled;
    struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD);
 
@@ -148,7 +152,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
    brw_clip_init_clipmask(c);
 
    /* -ve rhw workaround */
-   if (!BRW_IS_G4X(p->brw)) {
+   if (BRW_IS_965(p->brw)) {
       brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
       brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
               brw_imm_ud(1<<20));
@@ -185,7 +189,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
               * Both can be negative on GM965/G965 due to RHW workaround
               * if so, this object should be rejected.
               */
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_LE, c->reg.dp0, brw_imm_f(0.0));
                  is_neg2 = brw_IF(p, BRW_EXECUTE_1);
                  {
@@ -210,7 +214,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
 
              /* If both are positive, do nothing */
              /* Only on GM965/G965 */
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0));
                  is_neg2 = brw_IF(p, BRW_EXECUTE_1);
              }
@@ -225,7 +229,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
                  brw_set_predicate_control(p, BRW_PREDICATE_NONE);
              }
 
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_ENDIF(p, is_neg2);
              }
          }
@@ -263,6 +267,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
 void brw_emit_line_clip( struct brw_clip_compile *c )
 {
    brw_clip_line_alloc_regs(c);
+   brw_clip_init_ff_sync(c);
 
    if (c->key.do_flat_shading)
       brw_clip_copy_colors(c, 0, 1);
diff --git a/src/mesa/drivers/dri/i965/brw_clip_point.c b/src/mesa/drivers/dri/i965/brw_clip_point.c
index d17b199b89..8458f61c5a 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_point.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_point.c
@@ -50,5 +50,7 @@ void brw_emit_point_clip( struct brw_clip_compile *c )
    /* Send an empty message to kill the thread:
     */
    brw_clip_tri_alloc_regs(c, 0);
+   brw_clip_init_ff_sync(c);
+
    brw_clip_kill_thread(c);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index 9b0d7eab7b..5762c9577c 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -95,7 +95,14 @@ clip_unit_create_from_key(struct brw_context *brw,
        * even number.
        */
       assert(key->nr_urb_entries % 2 == 0);
-      clip.thread4.max_threads = 2 - 1;
+      
+      /* Although up to 16 concurrent Clip threads are allowed on IGDNG, 
+       * only 2 threads can output VUEs at a time.
+       */
+      if (BRW_IS_IGDNG(brw))
+         clip.thread4.max_threads = 16 - 1;        
+      else
+         clip.thread4.max_threads = 2 - 1;
    } else {
       assert(key->nr_urb_entries >= 5);
       clip.thread4.max_threads = 1 - 1;
diff --git a/src/mesa/drivers/dri/i965/brw_clip_tri.c b/src/mesa/drivers/dri/i965/brw_clip_tri.c
index 7fd37bd05f..0efd77225e 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_tri.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_tri.c
@@ -77,6 +77,10 @@ void brw_clip_tri_alloc_regs( struct brw_clip_compile *c,
    if (c->nr_attrs & 1) {
       for (j = 0; j < 3; j++) {
 	 GLuint delta = c->nr_attrs*16 + 32;
+
+         if (BRW_IS_IGDNG(c->func.brw))
+             delta = c->nr_attrs * 16 + 32 * 3;
+
 	 brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0));
       }
    }
@@ -115,6 +119,11 @@ void brw_clip_tri_alloc_regs( struct brw_clip_compile *c,
       i++;
    }
 
+   if (c->need_ff_sync) {
+      c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
+      i++;
+   }
+
    c->first_tmp = i;
    c->last_tmp = i;
 
@@ -559,10 +568,11 @@ void brw_emit_tri_clip( struct brw_clip_compile *c )
    brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6);
    brw_clip_tri_init_vertices(c);
    brw_clip_init_clipmask(c);
+   brw_clip_init_ff_sync(c);
 
    /* if -ve rhw workaround bit is set, 
       do cliptest */
-   if (!BRW_IS_G4X(p->brw)) {
+   if (BRW_IS_965(p->brw)) {
       brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
       brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), 
               brw_imm_ud(1<<20));
@@ -579,11 +589,12 @@ void brw_emit_tri_clip( struct brw_clip_compile *c )
    if (c->key.do_flat_shading) 
       brw_clip_tri_flat_shade(c); 
       
-   if (c->key.clip_mode == BRW_CLIPMODE_NORMAL)
+   if ((c->key.clip_mode == BRW_CLIPMODE_NORMAL) ||
+       (c->key.clip_mode == BRW_CLIPMODE_KERNEL_CLIP))
       do_clip_tri(c);
    else 
       maybe_do_clip_tri(c);
-      
+
    brw_clip_tri_emit_polygon(c);
 
    /* Send an empty message to kill the thread:
diff --git a/src/mesa/drivers/dri/i965/brw_clip_unfilled.c b/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
index d7ca517927..ad1bfa435f 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
@@ -453,6 +453,7 @@ void brw_emit_unfilled_clip( struct brw_clip_compile *c )
 
    brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6);
    brw_clip_tri_init_vertices(c);
+   brw_clip_init_ff_sync(c);
 
    assert(c->offset[VERT_RESULT_EDGE]);
 
diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c
index 9d3b0be694..5a73abdfee 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_util.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_util.c
@@ -140,6 +140,10 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
 
    /* Just copy the vertex header:
     */
+   /*
+    * After CLIP stage, only first 256 bits of the VUE are read
+    * back on IGDNG, so needn't change it
+    */
    brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
       
    /* Iterate over each attribute (could be done in pairs?)
@@ -147,6 +151,9 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    for (i = 0; i < c->nr_attrs; i++) {
       GLuint delta = i*16 + 32;
 
+      if (BRW_IS_IGDNG(p->brw))
+          delta = i * 16 + 32 * 3;
+
       if (delta == c->offset[VERT_RESULT_EDGE]) {
 	 if (force_edgeflag) 
 	    brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1));
@@ -177,6 +184,10 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
 
    if (i & 1) {
       GLuint delta = i*16 + 32;
+
+      if (BRW_IS_IGDNG(p->brw))
+          delta = i * 16 + 32 * 3;
+
       brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
    }
 
@@ -202,6 +213,8 @@ void brw_clip_emit_vue(struct brw_clip_compile *c,
    struct brw_compile *p = &c->func;
    GLuint start = c->last_mrf;
 
+   brw_clip_ff_sync(c);
+
    assert(!(allocate && eot));
    
    /* Cycle through mrf regs - probably futile as we have to wait for
@@ -252,6 +265,7 @@ void brw_clip_kill_thread(struct brw_clip_compile *c)
 {
    struct brw_compile *p = &c->func;
 
+   brw_clip_ff_sync(c);
    /* Send an empty message to kill the thread and release any
     * allocated urb entry:
     */
@@ -343,3 +357,40 @@ void brw_clip_init_clipmask( struct brw_clip_compile *c )
    }
 }
 
+void brw_clip_ff_sync(struct brw_clip_compile *c)
+{
+    if (c->need_ff_sync) {
+        struct brw_compile *p = &c->func;
+        struct brw_instruction *need_ff_sync;
+
+        brw_set_conditionalmod(p, BRW_CONDITIONAL_Z);
+        brw_AND(p, brw_null_reg(), c->reg.ff_sync, brw_imm_ud(0x1));
+        need_ff_sync = brw_IF(p, BRW_EXECUTE_1);
+        {
+            brw_OR(p, c->reg.ff_sync, c->reg.ff_sync, brw_imm_ud(0x1));
+            brw_ff_sync(p, 
+                    c->reg.R0,
+                    0,
+                    c->reg.R0,
+                    1,	
+                    1,		/* used */
+                    1,  	/* msg length */
+                    1,		/* response length */
+                    0,		/* eot */
+                    1,		/* write compelete */
+                    0,		/* urb offset */
+                    BRW_URB_SWIZZLE_NONE);
+        }
+        brw_ENDIF(p, need_ff_sync);
+        brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    }
+}
+
+void brw_clip_init_ff_sync(struct brw_clip_compile *c)
+{
+    if (c->need_ff_sync) {
+	struct brw_compile *p = &c->func;
+        
+        brw_MOV(p, c->reg.ff_sync, brw_imm_ud(0));
+    }
+}
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index 4dbe551d83..3c5b848319 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -126,7 +126,32 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
    /* We want the GLSL compiler to emit code that uses condition codes */
    ctx->Shader.EmitCondCodes = GL_TRUE;
 
-/*    ctx->Const.MaxNativeVertexProgramTemps = 32; */
+   ctx->Const.VertexProgram.MaxNativeInstructions = (16 * 1024);
+   ctx->Const.VertexProgram.MaxAluInstructions = 0;
+   ctx->Const.VertexProgram.MaxTexInstructions = 0;
+   ctx->Const.VertexProgram.MaxTexIndirections = 0;
+   ctx->Const.VertexProgram.MaxNativeAluInstructions = 0;
+   ctx->Const.VertexProgram.MaxNativeTexInstructions = 0;
+   ctx->Const.VertexProgram.MaxNativeTexIndirections = 0;
+   ctx->Const.VertexProgram.MaxNativeAttribs = 16;
+   ctx->Const.VertexProgram.MaxNativeTemps = 256;
+   ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+   ctx->Const.VertexProgram.MaxNativeParameters = 1024;
+   ctx->Const.VertexProgram.MaxEnvParams =
+      MIN2(ctx->Const.VertexProgram.MaxNativeParameters,
+	   ctx->Const.VertexProgram.MaxEnvParams);
+
+   ctx->Const.FragmentProgram.MaxNativeInstructions = (16 * 1024);
+   ctx->Const.FragmentProgram.MaxNativeAluInstructions = (16 * 1024);
+   ctx->Const.FragmentProgram.MaxNativeTexInstructions = (16 * 1024);
+   ctx->Const.FragmentProgram.MaxNativeTexIndirections = (16 * 1024);
+   ctx->Const.FragmentProgram.MaxNativeAttribs = 12;
+   ctx->Const.FragmentProgram.MaxNativeTemps = 256;
+   ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+   ctx->Const.FragmentProgram.MaxNativeParameters = 1024;
+   ctx->Const.FragmentProgram.MaxEnvParams =
+      MIN2(ctx->Const.FragmentProgram.MaxNativeParameters,
+	   ctx->Const.FragmentProgram.MaxEnvParams);
 
    brw_init_state( brw );
 
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 5cf12fb353..a5209ac41b 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -129,8 +129,8 @@ struct brw_context;
 #define BRW_NEW_PRIMITIVE               0x40
 #define BRW_NEW_CONTEXT                 0x80
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
-#define BRW_NEW_INPUT_VARYING           0x200
 #define BRW_NEW_PSP                     0x800
+#define BRW_NEW_WM_SURFACES		0x1000
 #define BRW_NEW_FENCE                   0x2000
 #define BRW_NEW_INDICES			0x4000
 #define BRW_NEW_VERTICES		0x8000
@@ -143,6 +143,7 @@ struct brw_context;
 #define BRW_NEW_DEPTH_BUFFER		0x20000
 #define BRW_NEW_NR_WM_SURFACES		0x40000
 #define BRW_NEW_NR_VS_SURFACES		0x80000
+#define BRW_NEW_INDEX_BUFFER		0x100000
 
 struct brw_state_flags {
    /** State update flags signalled by mesa internals */
@@ -173,6 +174,9 @@ struct brw_fragment_program {
 
    dri_bo *const_buffer;    /** Program constant buffer/surface */
    GLboolean use_const_buffer;
+
+   /** for debugging, which texture units are referenced */
+   GLbitfield tex_units_used;
 };
 
 
@@ -403,7 +407,6 @@ struct brw_vertex_element {
 
 
 struct brw_vertex_info {
-   GLuint varying;  /* varying:1[VERT_ATTRIB_MAX] */
    GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 };
 
@@ -441,9 +444,13 @@ struct brw_query_object {
    unsigned int count;
 };
 
+
+/**
+ * brw_context is derived from intel_context.
+ */
 struct brw_context 
 {
-   struct intel_context intel;
+   struct intel_context intel;  /**< base class, must be first field */
    GLuint primitive;
 
    GLboolean emit_state_always;
@@ -452,8 +459,6 @@ struct brw_context
 
    struct {
       struct brw_state_flags dirty;
-      struct brw_tracked_state **atoms;
-      GLuint nr_atoms;
 
       GLuint nr_color_regions;
       struct intel_region *color_regions[MAX_DRAW_BUFFERS];
@@ -473,7 +478,8 @@ struct brw_context
       int validated_bo_count;
    } state;
 
-   struct brw_cache cache;
+   struct brw_cache cache;  /** non-surface items */
+   struct brw_cache surface_cache;  /* surface items */
    struct brw_cached_batch_item *cached_batch_items;
 
    struct {
@@ -505,8 +511,15 @@ struct brw_context
        */
       const struct _mesa_index_buffer *ib;
 
+      /* Updates to these fields are signaled by BRW_NEW_INDEX_BUFFER. */
       dri_bo *bo;
       unsigned int offset;
+      unsigned int size;
+      /* Offset to index buffer index to use in CMD_3D_PRIM so that we can
+       * avoid re-uploading the IB packet over and over if we're actually
+       * referencing the same index buffer.
+       */
+      unsigned int start_vertex_offset;
    } ib;
 
    /* Active vertex program: 
@@ -560,11 +573,6 @@ struct brw_context
       GLuint vs_size;
       GLuint total_size;
 
-      /* Dynamic tracker which changes to reflect the state referenced
-       * by active fp and vp program parameters:
-       */
-      struct brw_tracked_state tracked_state;
-
       dri_bo *curbe_bo;
       /** Offset within curbe_bo of space for current curbe entry */
       GLuint curbe_offset;
@@ -718,6 +726,8 @@ void brw_upload_urb_fence(struct brw_context *brw);
  */
 void brw_upload_cs_urb_state(struct brw_context *brw);
 
+/* brw_disasm.c */
+int brw_disasm (FILE *file, struct brw_instruction *inst);
 
 /*======================================================================
  * Inline conversion functions.  These are better-typed than the
diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c
index 9197fede2d..0b0e6931a0 100644
--- a/src/mesa/drivers/dri/i965/brw_curbe.c
+++ b/src/mesa/drivers/dri/i965/brw_curbe.c
@@ -36,6 +36,7 @@
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
@@ -188,13 +189,6 @@ static void prepare_constant_buffer(struct brw_context *brw)
    GLfloat *buf;
    GLuint i;
 
-   /* Update our own dependency flags.  This works because this
-    * function will also be called whenever fp or vp changes.
-    */
-   brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION);
-   brw->curbe.tracked_state.dirty.mesa |= vp->program.Base.Parameters->StateFlags;
-   brw->curbe.tracked_state.dirty.mesa |= fp->program.Base.Parameters->StateFlags;
-
    if (sz == 0) {
       if (brw->curbe.last_buf) {
 	 free(brw->curbe.last_buf);
@@ -254,6 +248,9 @@ static void prepare_constant_buffer(struct brw_context *brw)
       GLuint offset = brw->curbe.vs_start * 16;
       GLuint nr = brw->vs.prog_data->nr_params / 4;
 
+      /* Updates the ParamaterValues[i] pointers for all parameters of the
+       * basic type of PROGRAM_STATE_VAR.
+       */
       _mesa_load_state_parameters(ctx, vp->program.Base.Parameters); 
 
       /* XXX just use a memcpy here */
@@ -335,78 +332,11 @@ static void prepare_constant_buffer(struct brw_context *brw)
     */
 }
 
-
-/**
- * Copy Mesa program parameters into given constant buffer.
- */
-static void
-update_constant_buffer(struct brw_context *brw,
-                       const struct gl_program_parameter_list *params,
-                       dri_bo *const_buffer)
-{
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   /* copy Mesa program constants into the buffer */
-   if (const_buffer && size > 0) {
-
-      assert(const_buffer);
-      assert(const_buffer->size >= size);
-
-      if (intel->intelScreen->kernel_exec_fencing) {
-         drm_intel_gem_bo_map_gtt(const_buffer);
-         memcpy(const_buffer->virtual, params->ParameterValues, size);
-         drm_intel_gem_bo_unmap_gtt(const_buffer);
-      }
-      else {
-         dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
-      }
-
-      if (0) {
-         int i;
-         for (i = 0; i < params->NumParameters; i++) {
-            float *p = params->ParameterValues[i];
-            printf("%d: %f %f %f %f\n", i, p[0], p[1], p[2], p[3]);
-         }
-      }
-   }
-}
-
-
-/** Copy current vertex program's parameters into the constant buffer */
-static void
-update_vertex_constant_buffer(struct brw_context *brw)
-{
-   struct brw_vertex_program *vp =
-      (struct brw_vertex_program *) brw->vertex_program;
-   if (0) {
-      printf("update VS constants in buffer %p\n", vp->const_buffer);
-      printf("program %u\n", vp->program.Base.Id);
-   }
-   if (vp->use_const_buffer)
-      update_constant_buffer(brw, vp->program.Base.Parameters, vp->const_buffer);
-}
-
-
-/** Copy current fragment program's parameters into the constant buffer */
-static void
-update_fragment_constant_buffer(struct brw_context *brw)
-{
-   struct brw_fragment_program *fp =
-      (struct brw_fragment_program *) brw->fragment_program;
-   if (fp->use_const_buffer)
-      update_constant_buffer(brw, fp->program.Base.Parameters, fp->const_buffer);
-}
-
-
 static void emit_constant_buffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    GLuint sz = brw->curbe.total_size;
 
-   update_vertex_constant_buffer(brw);
-   update_fragment_constant_buffer(brw);
-
    BEGIN_BATCH(2, IGNORE_CLIPRECTS);
    if (sz == 0) {
       OUT_BATCH((CMD_CONST_BUFFER << 16) | (2 - 2));
@@ -428,7 +358,7 @@ static void emit_constant_buffer(struct brw_context *brw)
  */
 const struct brw_tracked_state brw_constant_buffer = {
    .dirty = {
-      .mesa = (_NEW_TRANSFORM|_NEW_PROJECTION),      /* plus fp and vp flags */
+      .mesa = _NEW_PROGRAM_CONSTANTS,
       .brw  = (BRW_NEW_FRAGMENT_PROGRAM |
 	       BRW_NEW_VERTEX_PROGRAM |
 	       BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 98fc909c2a..78d457ad2b 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -139,6 +139,7 @@
 #define BRW_CLIPMODE_CLIP_NON_REJECTED   2
 #define BRW_CLIPMODE_REJECT_ALL          3
 #define BRW_CLIPMODE_ACCEPT_ALL          4
+#define BRW_CLIPMODE_KERNEL_CLIP         5
 
 #define BRW_CLIP_NDCSPACE     0
 #define BRW_CLIP_SCREENSPACE  1
@@ -470,8 +471,9 @@
 #define BRW_CONDITIONAL_GE    4
 #define BRW_CONDITIONAL_L     5
 #define BRW_CONDITIONAL_LE    6
-#define BRW_CONDITIONAL_C     7
+#define BRW_CONDITIONAL_R     7
 #define BRW_CONDITIONAL_O     8
+#define BRW_CONDITIONAL_U     9
 
 #define BRW_DEBUG_NONE        0
 #define BRW_DEBUG_BREAKPOINT  1
@@ -511,6 +513,7 @@
 #define BRW_OPCODE_RSL        11
 #define BRW_OPCODE_ASR        12
 #define BRW_OPCODE_CMP        16
+#define BRW_OPCODE_CMPN       17
 #define BRW_OPCODE_JMPI       32
 #define BRW_OPCODE_IF         34
 #define BRW_OPCODE_IFF        35
@@ -670,6 +673,25 @@
 #define BRW_SAMPLER_MESSAGE_SIMD8_LD                  3
 #define BRW_SAMPLER_MESSAGE_SIMD16_LD                 3
 
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG            0
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_IGDNG          0
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG           0
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG       1
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_BIAS_IGDNG     1
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG      1
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_IGDNG        2
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD_IGDNG      2
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD_IGDNG       2
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG    3
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE_IGDNG  3
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG   3
+
+/* for IGDNG only */
+#define BRW_SAMPLER_SIMD_MODE_SIMD4X2                   0
+#define BRW_SAMPLER_SIMD_MODE_SIMD8                     1
+#define BRW_SAMPLER_SIMD_MODE_SIMD16                    2
+#define BRW_SAMPLER_SIMD_MODE_SIMD32_64                 3
+
 #define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW   0
 #define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH  1
 #define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS     2
@@ -819,8 +841,11 @@
 #include "intel_chipset.h"
 
 #define BRW_IS_G4X(brw)         (IS_G4X((brw)->intel.intelScreen->deviceID))
-#define CMD_PIPELINE_SELECT(brw)        (BRW_IS_G4X(brw) ? CMD_PIPELINE_SELECT_GM45 : CMD_PIPELINE_SELECT_965)
-#define CMD_VF_STATISTICS(brw)          (BRW_IS_G4X(brw) ? CMD_VF_STATISTICS_GM45 : CMD_VF_STATISTICS_965)
-#define URB_SIZES(brw)                  (BRW_IS_G4X(brw) ? 384 : 256)  /* 512 bit units */
+#define BRW_IS_IGDNG(brw)         (IS_IGDNG((brw)->intel.intelScreen->deviceID))
+#define BRW_IS_965(brw)         (!(BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)))
+#define CMD_PIPELINE_SELECT(brw)        ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_PIPELINE_SELECT_GM45 : CMD_PIPELINE_SELECT_965)
+#define CMD_VF_STATISTICS(brw)          ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_VF_STATISTICS_GM45 : CMD_VF_STATISTICS_965)
+#define URB_SIZES(brw)                  (BRW_IS_IGDNG(brw) ? 1024 : \
+                                         (BRW_IS_G4X(brw) ? 384 : 256))  /* 512 bit units */
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_disasm.c b/src/mesa/drivers/dri/i965/brw_disasm.c
new file mode 100644
index 0000000000..9fef230507
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_disasm.c
@@ -0,0 +1,903 @@
+/*
+ * Copyright © 2008 Keith Packard
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and its
+ * documentation for any purpose is hereby granted without fee, provided that
+ * the above copyright notice appear in all copies and that both that copyright
+ * notice and this permission notice appear in supporting documentation, and
+ * that the name of the copyright holders not be used in advertising or
+ * publicity pertaining to distribution of the software without specific,
+ * written prior permission.  The copyright holders make no representations
+ * about the suitability of this software for any purpose.  It is provided "as
+ * is" without express or implied warranty.
+ *
+ * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
+ * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO
+ * EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR
+ * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+ * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+ * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THIS SOFTWARE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <getopt.h>
+#include <unistd.h>
+#include <stdarg.h>
+
+#include "main/mtypes.h"
+
+#include "brw_context.h"
+#include "brw_defines.h"
+
+struct {
+    char    *name;
+    int	    nsrc;
+    int	    ndst;
+} opcode[128] = {
+    [BRW_OPCODE_MOV] = { .name = "mov", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_FRC] = { .name = "frc", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_RNDU] = { .name = "rndu", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_RNDD] = { .name = "rndd", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_RNDE] = { .name = "rnde", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_RNDZ] = { .name = "rndz", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_NOT] = { .name = "not", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_LZD] = { .name = "lzd", .nsrc = 1, .ndst = 1 },
+
+    [BRW_OPCODE_MUL] = { .name = "mul", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_MAC] = { .name = "mac", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_MACH] = { .name = "mach", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_LINE] = { .name = "line", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SAD2] = { .name = "sad2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SADA2] = { .name = "sada2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP4] = { .name = "dp4", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DPH] = { .name = "dph", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP3] = { .name = "dp3", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP2] = { .name = "dp2", .nsrc = 2, .ndst = 1 },
+
+    [BRW_OPCODE_AVG] = { .name = "avg", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_ADD] = { .name = "add", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SEL] = { .name = "sel", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_AND] = { .name = "and", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_OR] = { .name = "or", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_XOR] = { .name = "xor", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SHR] = { .name = "shr", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SHL] = { .name = "shl", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_ASR] = { .name = "asr", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_CMP] = { .name = "cmp", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_CMPN] = { .name = "cmpn", .nsrc = 2, .ndst = 1 },
+
+    [BRW_OPCODE_SEND] = { .name = "send", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_NOP] = { .name = "nop", .nsrc = 0, .ndst = 0 },
+    [BRW_OPCODE_JMPI] = { .name = "jmpi", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_IF] = { .name = "if", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_IFF] = { .name = "iff", .nsrc = 1, .ndst = 01 },
+    [BRW_OPCODE_WHILE] = { .name = "while", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_ELSE] = { .name = "else", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_BREAK] = { .name = "break", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_CONTINUE] = { .name = "cont", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_HALT] = { .name = "halt", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_MSAVE] = { .name = "msave", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_PUSH] = { .name = "push", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_MRESTORE] = { .name = "mrest", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_POP] = { .name = "pop", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_WAIT] = { .name = "wait", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_DO] = { .name = "do", .nsrc = 0, .ndst = 0 },
+    [BRW_OPCODE_ENDIF] = { .name = "endif", .nsrc = 2, .ndst = 0 },
+};
+
+char *conditional_modifier[16] = {
+    [BRW_CONDITIONAL_NONE] = "",
+    [BRW_CONDITIONAL_Z] = ".e",
+    [BRW_CONDITIONAL_NZ] = ".ne",
+    [BRW_CONDITIONAL_G] = ".g",
+    [BRW_CONDITIONAL_GE] = ".ge",
+    [BRW_CONDITIONAL_L] = ".l",
+    [BRW_CONDITIONAL_LE] = ".le",
+    [BRW_CONDITIONAL_R] = ".r",
+    [BRW_CONDITIONAL_O] = ".o",
+    [BRW_CONDITIONAL_U] = ".u",
+};
+
+char *negate[2] = {
+    [0] = "",
+    [1] = "-",
+};
+
+char *_abs[2] = {
+    [0] = "",
+    [1] = "(abs)",
+};
+
+char *vert_stride[16] = {
+    [0] = "0",
+    [1] = "1",
+    [2] = "2",
+    [3] = "4",
+    [4] = "8",
+    [5] = "16",
+    [6] = "32",
+    [15] = "VxH",
+};
+
+char *width[8] = {
+    [0] = "1",
+    [1] = "2",
+    [2] = "4",
+    [3] = "8",
+    [4] = "16",
+};
+
+char *horiz_stride[4] = {
+    [0] = "0",
+    [1] = "1",
+    [2] = "2",
+    [3] = "4"
+};
+
+char *chan_sel[4] = {
+    [0] = "x",
+    [1] = "y",
+    [2] = "z",
+    [3] = "w",
+};
+
+char *dest_condmod[16] = {
+};
+
+char *debug_ctrl[2] = {
+    [0] = "",
+    [1] = ".breakpoint"
+};
+
+char *saturate[2] = {
+    [0] = "",
+    [1] = ".sat"
+};
+
+char *exec_size[8] = {
+    [0] = "1",
+    [1] = "2",
+    [2] = "4",
+    [3] = "8",
+    [4] = "16",
+    [5] = "32"
+};
+
+char *pred_inv[2] = {
+    [0] = "+",
+    [1] = "-"
+};
+
+char *pred_ctrl_align16[16] = {
+    [1] = "",
+    [2] = ".x",
+    [3] = ".y",
+    [4] = ".z",
+    [5] = ".w",
+    [6] = ".any4h",
+    [7] = ".all4h",
+};
+
+char *pred_ctrl_align1[16] = {
+    [1] = "",
+    [2] = ".anyv",
+    [3] = ".allv",
+    [4] = ".any2h",
+    [5] = ".all2h",
+    [6] = ".any4h",
+    [7] = ".all4h",
+    [8] = ".any8h",
+    [9] = ".all8h",
+    [10] = ".any16h",
+    [11] = ".all16h",
+};
+
+char *thread_ctrl[4] = {
+    [0] = "",
+    [2] = "switch"
+};
+
+char *compr_ctrl[4] = {
+    [0] = "",
+    [1] = "sechalf",
+    [2] = "compr",
+};
+
+char *dep_ctrl[4] = {
+    [0] = "",
+    [1] = "NoDDClr",
+    [2] = "NoDDChk",
+    [3] = "NoDDClr,NoDDChk",
+};
+
+char *mask_ctrl[4] = {
+    [0] = "",
+    [1] = "nomask",
+};
+
+char *access_mode[2] = {
+    [0] = "align1",
+    [1] = "align16",
+};
+
+char *reg_encoding[8] = {
+    [0] = "UD",
+    [1] = "D",
+    [2] = "UW",
+    [3] = "W",
+    [4] = "UB",
+    [5] = "B",
+    [7] = "F"
+};
+
+char *imm_encoding[8] = {
+    [0] = "UD",
+    [1] = "D",
+    [2] = "UW",
+    [3] = "W",
+    [5] = "VF",
+    [5] = "V",
+    [7] = "F"
+};
+
+char *reg_file[4] = {
+    [0] = "A",
+    [1] = "g",
+    [2] = "m",
+    [3] = "imm",
+};
+
+char *writemask[16] = {
+    [0x0] = ".",
+    [0x1] = ".x",
+    [0x2] = ".y",
+    [0x3] = ".xy",
+    [0x4] = ".z",
+    [0x5] = ".xz",
+    [0x6] = ".yz",
+    [0x7] = ".xyz",
+    [0x8] = ".w",
+    [0x9] = ".xw",
+    [0xa] = ".yw",
+    [0xb] = ".xyw",
+    [0xc] = ".zw",
+    [0xd] = ".xzw",
+    [0xe] = ".yzw",
+    [0xf] = "",
+};
+
+char *end_of_thread[2] = {
+    [0] = "",
+    [1] = "EOT"
+};
+
+char *target_function[16] = {
+    [BRW_MESSAGE_TARGET_NULL] = "null",
+    [BRW_MESSAGE_TARGET_MATH] = "math",
+    [BRW_MESSAGE_TARGET_SAMPLER] = "sampler",
+    [BRW_MESSAGE_TARGET_GATEWAY] = "gateway",
+    [BRW_MESSAGE_TARGET_DATAPORT_READ] = "read",
+    [BRW_MESSAGE_TARGET_DATAPORT_WRITE] = "write",
+    [BRW_MESSAGE_TARGET_URB] = "urb",
+    [BRW_MESSAGE_TARGET_THREAD_SPAWNER] = "thread_spawner"
+};
+
+char *math_function[16] = {
+    [BRW_MATH_FUNCTION_INV] = "inv",
+    [BRW_MATH_FUNCTION_LOG] = "log",
+    [BRW_MATH_FUNCTION_EXP] = "exp",
+    [BRW_MATH_FUNCTION_SQRT] = "sqrt",
+    [BRW_MATH_FUNCTION_RSQ] = "rsq",
+    [BRW_MATH_FUNCTION_SIN] = "sin",
+    [BRW_MATH_FUNCTION_COS] = "cos",
+    [BRW_MATH_FUNCTION_SINCOS] = "sincos",
+    [BRW_MATH_FUNCTION_TAN] = "tan",
+    [BRW_MATH_FUNCTION_POW] = "pow",
+    [BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER] = "intdivmod",
+    [BRW_MATH_FUNCTION_INT_DIV_QUOTIENT] = "intmod",
+    [BRW_MATH_FUNCTION_INT_DIV_REMAINDER] = "intdiv",
+};
+
+char *math_saturate[2] = {
+    [0] = "",
+    [1] = "sat"
+};
+
+char *math_signed[2] = {
+    [0] = "",
+    [1] = "signed"
+};
+
+char *math_scalar[2] = {
+    [0] = "",
+    [1] = "scalar"
+};
+
+char *math_precision[2] = {
+    [0] = "",
+    [1] = "partial_precision"
+};
+
+char *urb_swizzle[4] = {
+    [BRW_URB_SWIZZLE_NONE] = "",
+    [BRW_URB_SWIZZLE_INTERLEAVE] = "interleave",
+    [BRW_URB_SWIZZLE_TRANSPOSE] = "transpose",
+};
+
+char *urb_allocate[2] = {
+    [0] = "",
+    [1] = "allocate"
+};
+
+char *urb_used[2] = {
+    [0] = "",
+    [1] = "used"
+};
+
+char *urb_complete[2] = {
+    [0] = "",
+    [1] = "complete"
+};
+
+char *sampler_target_format[4] = {
+    [0] = "F",
+    [2] = "UD",
+    [3] = "D"
+};
+
+
+static int column;
+
+static int string (FILE *file, char *string)
+{
+    fputs (string, file);
+    column += strlen (string);
+    return 0;
+}
+
+static int format (FILE *f, char *format, ...)
+{
+    char    buf[1024];
+    va_list	args;
+    va_start (args, format);
+
+    vsnprintf (buf, sizeof (buf) - 1, format, args);
+    string (f, buf);
+    return 0;
+}
+
+static int newline (FILE *f)
+{
+    putc ('\n', f);
+    column = 0;
+    return 0;
+}
+
+static int pad (FILE *f, int c)
+{
+    do
+	string (f, " ");
+    while (column < c);
+    return 0;
+}
+
+static int control (FILE *file, char *name, char *ctrl[], GLuint id, int *space)
+{
+    if (!ctrl[id]) {
+	fprintf (file, "*** invalid %s value %d ",
+		 name, id);
+	return 1;
+    }
+    if (ctrl[id][0])
+    {
+	if (space && *space)
+	    string (file, " ");
+	string (file, ctrl[id]);
+	if (space)
+	    *space = 1;
+    }
+    return 0;
+}
+
+static int print_opcode (FILE *file, int id)
+{
+    if (!opcode[id].name) {
+	format (file, "*** invalid opcode value %d ", id);
+	return 1;
+    }
+    string (file, opcode[id].name);
+    return 0;
+}
+
+static int reg (FILE *file, GLuint _reg_file, GLuint _reg_nr)
+{
+    int	err = 0;
+    if (_reg_file == BRW_ARCHITECTURE_REGISTER_FILE) {
+	switch (_reg_nr & 0xf0) {
+	case BRW_ARF_NULL:
+	    string (file, "null");
+	    return -1;
+	case BRW_ARF_ADDRESS:
+	    format (file, "a%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_ACCUMULATOR:
+	    format (file, "acc%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_MASK:
+	    format (file, "mask%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_MASK_STACK:
+	    format (file, "msd%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_STATE:
+	    format (file, "sr%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_CONTROL:
+	    format (file, "cr%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_NOTIFICATION_COUNT:
+	    format (file, "n%d", _reg_nr & 0x0f);
+	    break;
+	case BRW_ARF_IP:
+	    string (file, "ip");
+	    return -1;
+	    break;
+	default:
+	    format (file, "ARF%d", _reg_nr);
+	    break;
+	}
+    } else {
+	err  |= control (file, "src reg file", reg_file, _reg_file, NULL);
+	format (file, "%d", _reg_nr);
+    }
+    return err;
+}
+
+static int dest (FILE *file, struct brw_instruction *inst)
+{
+    int	err = 0;
+
+    if (inst->header.access_mode == BRW_ALIGN_1)
+    {
+	if (inst->bits1.da1.dest_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    err |= reg (file, inst->bits1.da1.dest_reg_file, inst->bits1.da1.dest_reg_nr);
+	    if (err == -1)
+		return 0;
+	    if (inst->bits1.da1.dest_subreg_nr)
+		format (file, ".%d", inst->bits1.da1.dest_subreg_nr);
+	    format (file, "<%d>", inst->bits1.da1.dest_horiz_stride);
+	    err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.da1.dest_reg_type, NULL);
+	}
+	else
+	{
+	    string (file, "g[a0");
+	    if (inst->bits1.ia1.dest_subreg_nr)
+		format (file, ".%d", inst->bits1.ia1.dest_subreg_nr);
+	    if (inst->bits1.ia1.dest_indirect_offset)
+		format (file, " %d", inst->bits1.ia1.dest_indirect_offset);
+	    string (file, "]");
+	    format (file, "<%d>", inst->bits1.ia1.dest_horiz_stride);
+	    err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.ia1.dest_reg_type, NULL);
+	}
+    }
+    else
+    {
+	if (inst->bits1.da16.dest_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    err |= reg (file, inst->bits1.da16.dest_reg_file, inst->bits1.da16.dest_reg_nr);
+	    if (err == -1)
+		return 0;
+	    if (inst->bits1.da16.dest_subreg_nr)
+		format (file, ".%d", inst->bits1.da16.dest_subreg_nr);
+	    string (file, "<1>");
+	    err |= control (file, "writemask", writemask, inst->bits1.da16.dest_writemask, NULL);
+	    err |= control (file, "dest reg encoding", reg_encoding, inst->bits1.da16.dest_reg_type, NULL);
+	}
+	else
+	{
+	    err = 1;
+	    string (file, "Indirect align16 address mode not supported");
+	}
+    }
+
+    return 0;
+}
+
+static int src_align1_region (FILE *file,
+			      GLuint _vert_stride, GLuint _width, GLuint _horiz_stride)
+{
+    int err = 0;
+    string (file, "<");
+    err |= control (file, "vert stride", vert_stride, _vert_stride, NULL);
+    string (file, ",");
+    err |= control (file, "width", width, _width, NULL);
+    string (file, ",");
+    err |= control (file, "horiz_stride", horiz_stride, _horiz_stride, NULL);
+    string (file, ">");
+    return err;
+}
+
+static int src_da1 (FILE *file, GLuint type, GLuint _reg_file,
+		    GLuint _vert_stride, GLuint _width, GLuint _horiz_stride,
+		    GLuint reg_num, GLuint sub_reg_num, GLuint __abs, GLuint _negate)
+{
+    int err = 0;
+    err |= control (file, "negate", negate, _negate, NULL);
+    err |= control (file, "abs", _abs, __abs, NULL);
+
+    err |= reg (file, _reg_file, reg_num);
+    if (err == -1)
+	return 0;
+    if (sub_reg_num)
+	format (file, ".%d", sub_reg_num);
+    src_align1_region (file, _vert_stride, _width, _horiz_stride);
+    err |= control (file, "src reg encoding", reg_encoding, type, NULL);
+    return err;
+}
+
+static int src_ia1 (FILE *file,
+		    GLuint type,
+		    GLuint _reg_file,
+		    GLint _addr_imm,
+		    GLuint _addr_subreg_nr,
+		    GLuint _negate,
+		    GLuint __abs,
+		    GLuint _addr_mode,
+		    GLuint _horiz_stride,
+		    GLuint _width,
+		    GLuint _vert_stride)
+{
+    int err = 0;
+    err |= control (file, "negate", negate, _negate, NULL);
+    err |= control (file, "abs", _abs, __abs, NULL);
+
+    string (file, "g[a0");
+    if (_addr_subreg_nr)
+	format (file, ".%d", _addr_subreg_nr);
+    if (_addr_imm)
+	format (file, " %d", _addr_imm);
+    string (file, "]");
+    src_align1_region (file, _vert_stride, _width, _horiz_stride);
+    err |= control (file, "src reg encoding", reg_encoding, type, NULL);
+    return err;
+}
+
+static int src_da16 (FILE *file,
+		     GLuint _reg_type,
+		     GLuint _reg_file,
+		     GLuint _vert_stride,
+		     GLuint _reg_nr,
+		     GLuint _subreg_nr,
+		     GLuint __abs,
+		     GLuint _negate,
+		     GLuint swz_x,
+		     GLuint swz_y,
+		     GLuint swz_z,
+		     GLuint swz_w)
+{
+    int err = 0;
+    err |= control (file, "negate", negate, _negate, NULL);
+    err |= control (file, "abs", _abs, __abs, NULL);
+
+    err |= reg (file, _reg_file, _reg_nr);
+    if (err == -1)
+	return 0;
+    if (_subreg_nr)
+	format (file, ".%d", _subreg_nr);
+    string (file, "<");
+    err |= control (file, "vert stride", vert_stride, _vert_stride, NULL);
+    string (file, ",1,1>");
+    err |= control (file, "src da16 reg type", reg_encoding, _reg_type, NULL);
+    /*
+     * Three kinds of swizzle display:
+     *  identity - nothing printed
+     *  1->all	 - print the single channel
+     *  1->1     - print the mapping
+     */
+    if (swz_x == BRW_CHANNEL_X &&
+	swz_y == BRW_CHANNEL_Y &&
+	swz_z == BRW_CHANNEL_Z &&
+	swz_w == BRW_CHANNEL_W)
+    {
+	;
+    }
+    else if (swz_x == swz_y && swz_x == swz_z && swz_x == swz_w)
+    {
+	string (file, ".");
+	err |= control (file, "channel select", chan_sel, swz_x, NULL);
+    }
+    else
+    {
+	string (file, ".");
+	err |= control (file, "channel select", chan_sel, swz_x, NULL);
+	err |= control (file, "channel select", chan_sel, swz_y, NULL);
+	err |= control (file, "channel select", chan_sel, swz_z, NULL);
+	err |= control (file, "channel select", chan_sel, swz_w, NULL);
+    }
+    return err;
+}
+
+
+static int imm (FILE *file, GLuint type, struct brw_instruction *inst) {
+    switch (type) {
+    case BRW_REGISTER_TYPE_UD:
+	format (file, "0x%08xUD", inst->bits3.ud);
+	break;
+    case BRW_REGISTER_TYPE_D:
+	format (file, "%dD", inst->bits3.d);
+	break;
+    case BRW_REGISTER_TYPE_UW:
+	format (file, "0x%04xUW", (uint16_t) inst->bits3.ud);
+	break;
+    case BRW_REGISTER_TYPE_W:
+	format (file, "%dW", (int16_t) inst->bits3.d);
+	break;
+    case BRW_REGISTER_TYPE_UB:
+	format (file, "0x%02xUB", (int8_t) inst->bits3.ud);
+	break;
+    case BRW_REGISTER_TYPE_VF:
+	format (file, "Vector Float");
+	break;
+    case BRW_REGISTER_TYPE_V:
+	format (file, "0x%08xV", inst->bits3.ud);
+	break;
+    case BRW_REGISTER_TYPE_F:
+	format (file, "%-gF", inst->bits3.f);
+    }
+    return 0;
+}
+
+static int src0 (FILE *file, struct brw_instruction *inst)
+{
+    if (inst->bits1.da1.src0_reg_file == BRW_IMMEDIATE_VALUE)
+	return imm (file, inst->bits1.da1.src0_reg_type,
+		    inst);
+    else if (inst->header.access_mode == BRW_ALIGN_1)
+    {
+	if (inst->bits2.da1.src0_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    return src_da1 (file,
+			    inst->bits1.da1.src0_reg_type,
+			    inst->bits1.da1.src0_reg_file,
+			    inst->bits2.da1.src0_vert_stride,
+			    inst->bits2.da1.src0_width,
+			    inst->bits2.da1.src0_horiz_stride,
+			    inst->bits2.da1.src0_reg_nr,
+			    inst->bits2.da1.src0_subreg_nr,
+			    inst->bits2.da1.src0_abs,
+			    inst->bits2.da1.src0_negate);
+	}
+	else
+	{
+	    return src_ia1 (file,
+			    inst->bits1.ia1.src0_reg_type,
+			    inst->bits1.ia1.src0_reg_file,
+			    inst->bits2.ia1.src0_indirect_offset,
+			    inst->bits2.ia1.src0_subreg_nr,
+			    inst->bits2.ia1.src0_negate,
+			    inst->bits2.ia1.src0_abs,
+			    inst->bits2.ia1.src0_address_mode,
+			    inst->bits2.ia1.src0_horiz_stride,
+			    inst->bits2.ia1.src0_width,
+			    inst->bits2.ia1.src0_vert_stride);
+	}
+    }
+    else
+    {
+	if (inst->bits2.da16.src0_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    return src_da16 (file,
+			     inst->bits1.da16.src0_reg_type,
+			     inst->bits1.da16.src0_reg_file,
+			     inst->bits2.da16.src0_vert_stride,
+			     inst->bits2.da16.src0_reg_nr,
+			     inst->bits2.da16.src0_subreg_nr,
+			     inst->bits2.da16.src0_abs,
+			     inst->bits2.da16.src0_negate,
+			     inst->bits2.da16.src0_swz_x,
+			     inst->bits2.da16.src0_swz_y,
+			     inst->bits2.da16.src0_swz_z,
+			     inst->bits2.da16.src0_swz_w);
+	}
+	else
+	{
+	    string (file, "Indirect align16 address mode not supported");
+	    return 1;
+	}
+    }
+}
+
+static int src1 (FILE *file, struct brw_instruction *inst)
+{
+    if (inst->bits1.da1.src1_reg_file == BRW_IMMEDIATE_VALUE)
+	return imm (file, inst->bits1.da1.src1_reg_type,
+		    inst);
+    else if (inst->header.access_mode == BRW_ALIGN_1)
+    {
+	if (inst->bits3.da1.src1_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    return src_da1 (file,
+			    inst->bits1.da1.src1_reg_type,
+			    inst->bits1.da1.src1_reg_file,
+			    inst->bits3.da1.src1_vert_stride,
+			    inst->bits3.da1.src1_width,
+			    inst->bits3.da1.src1_horiz_stride,
+			    inst->bits3.da1.src1_reg_nr,
+			    inst->bits3.da1.src1_subreg_nr,
+			    inst->bits3.da1.src1_abs,
+			    inst->bits3.da1.src1_negate);
+	}
+	else
+	{
+	    return src_ia1 (file,
+			    inst->bits1.ia1.src1_reg_type,
+			    inst->bits1.ia1.src1_reg_file,
+			    inst->bits3.ia1.src1_indirect_offset,
+			    inst->bits3.ia1.src1_subreg_nr,
+			    inst->bits3.ia1.src1_negate,
+			    inst->bits3.ia1.src1_abs,
+			    inst->bits3.ia1.src1_address_mode,
+			    inst->bits3.ia1.src1_horiz_stride,
+			    inst->bits3.ia1.src1_width,
+			    inst->bits3.ia1.src1_vert_stride);
+	}
+    }
+    else
+    {
+	if (inst->bits3.da16.src1_address_mode == BRW_ADDRESS_DIRECT)
+	{
+	    return src_da16 (file,
+			     inst->bits1.da16.src1_reg_type,
+			     inst->bits1.da16.src1_reg_file,
+			     inst->bits3.da16.src1_vert_stride,
+			     inst->bits3.da16.src1_reg_nr,
+			     inst->bits3.da16.src1_subreg_nr,
+			     inst->bits3.da16.src1_abs,
+			     inst->bits3.da16.src1_negate,
+			     inst->bits3.da16.src1_swz_x,
+			     inst->bits3.da16.src1_swz_y,
+			     inst->bits3.da16.src1_swz_z,
+			     inst->bits3.da16.src1_swz_w);
+	}
+	else
+	{
+	    string (file, "Indirect align16 address mode not supported");
+	    return 1;
+	}
+    }
+}
+
+int brw_disasm (FILE *file, struct brw_instruction *inst)
+{
+    int	err = 0;
+    int space = 0;
+
+    if (inst->header.predicate_control) {
+	string (file, "(");
+	err |= control (file, "predicate inverse", pred_inv, inst->header.predicate_inverse, NULL);
+	string (file, "f0");
+	if (inst->bits2.da1.flag_reg_nr)
+	    format (file, ".%d", inst->bits2.da1.flag_reg_nr);
+	if (inst->header.access_mode == BRW_ALIGN_1)
+	    err |= control (file, "predicate control align1", pred_ctrl_align1,
+			    inst->header.predicate_control, NULL);
+	else
+	    err |= control (file, "predicate control align16", pred_ctrl_align16,
+			    inst->header.predicate_control, NULL);
+	string (file, ") ");
+    }
+
+    err |= print_opcode (file, inst->header.opcode);
+    err |= control (file, "saturate", saturate, inst->header.saturate, NULL);
+    err |= control (file, "debug control", debug_ctrl, inst->header.debug_control, NULL);
+
+    if (inst->header.opcode != BRW_OPCODE_SEND)
+	err |= control (file, "conditional modifier", conditional_modifier,
+			inst->header.destreg__conditionalmod, NULL);
+
+    if (inst->header.opcode != BRW_OPCODE_NOP) {
+	string (file, "(");
+	err |= control (file, "execution size", exec_size, inst->header.execution_size, NULL);
+	string (file, ")");
+    }
+
+    if (inst->header.opcode == BRW_OPCODE_SEND)
+	format (file, " %d", inst->header.destreg__conditionalmod);
+
+    if (opcode[inst->header.opcode].ndst > 0) {
+	pad (file, 16);
+	err |= dest (file, inst);
+    }
+    if (opcode[inst->header.opcode].nsrc > 0) {
+	pad (file, 32);
+	err |= src0 (file, inst);
+    }
+    if (opcode[inst->header.opcode].nsrc > 1) {
+	pad (file, 48);
+	err |= src1 (file, inst);
+    }
+
+    if (inst->header.opcode == BRW_OPCODE_SEND) {
+	newline (file);
+	pad (file, 16);
+	space = 0;
+	err |= control (file, "target function", target_function,
+			inst->bits3.generic.msg_target, &space);
+	switch (inst->bits3.generic.msg_target) {
+	case BRW_MESSAGE_TARGET_MATH:
+	    err |= control (file, "math function", math_function,
+			    inst->bits3.math.function, &space);
+	    err |= control (file, "math saturate", math_saturate,
+			    inst->bits3.math.saturate, &space);
+	    err |= control (file, "math signed", math_signed,
+			    inst->bits3.math.int_type, &space);
+	    err |= control (file, "math scalar", math_scalar,
+			    inst->bits3.math.data_type, &space);
+	    err |= control (file, "math precision", math_precision,
+			    inst->bits3.math.precision, &space);
+	    break;
+	case BRW_MESSAGE_TARGET_SAMPLER:
+	    format (file, " (%d, %d, ",
+		    inst->bits3.sampler.binding_table_index,
+		    inst->bits3.sampler.sampler);
+	    err |= control (file, "sampler target format", sampler_target_format,
+			    inst->bits3.sampler.return_format, NULL);
+	    string (file, ")");
+	    break;
+	case BRW_MESSAGE_TARGET_DATAPORT_WRITE:
+	    format (file, " (%d, %d, %d, %d)",
+		    inst->bits3.dp_write.binding_table_index,
+		    (inst->bits3.dp_write.pixel_scoreboard_clear << 3) |
+		    inst->bits3.dp_write.msg_control,
+		    inst->bits3.dp_write.msg_type,
+		    inst->bits3.dp_write.send_commit_msg);
+	    break;
+	case BRW_MESSAGE_TARGET_URB:
+	    format (file, " %d", inst->bits3.urb.offset);
+	    space = 1;
+	    err |= control (file, "urb swizzle", urb_swizzle,
+			    inst->bits3.urb.swizzle_control, &space);
+	    err |= control (file, "urb allocate", urb_allocate,
+			    inst->bits3.urb.allocate, &space);
+	    err |= control (file, "urb used", urb_used,
+			    inst->bits3.urb.used, &space);
+	    err |= control (file, "urb complete", urb_complete,
+			    inst->bits3.urb.complete, &space);
+	    break;
+	case BRW_MESSAGE_TARGET_THREAD_SPAWNER:
+	    break;
+	default:
+	    format (file, "unsupported target %d", inst->bits3.generic.msg_target);
+	    break;
+	}
+	if (space)
+	    string (file, " ");
+	format (file, "mlen %d",
+		inst->bits3.generic.msg_length);
+	format (file, " rlen %d",
+		inst->bits3.generic.response_length);
+    }
+    pad (file, 64);
+    if (inst->header.opcode != BRW_OPCODE_NOP) {
+	string (file, "{");
+	space = 1;
+	err |= control(file, "access mode", access_mode, inst->header.access_mode, &space);
+	err |= control (file, "mask control", mask_ctrl, inst->header.mask_control, &space);
+	err |= control (file, "dependency control", dep_ctrl, inst->header.dependency_control, &space);
+	err |= control (file, "compression control", compr_ctrl, inst->header.compression_control, &space);
+	err |= control (file, "thread control", thread_ctrl, inst->header.thread_control, &space);
+	if (inst->header.opcode == BRW_OPCODE_SEND)
+	    err |= control (file, "end of thread", end_of_thread,
+			    inst->bits3.generic.end_of_thread, &space);
+	if (space)
+	    string (file, " ");
+	string (file, "}");
+    }
+    string (file, ";");
+    newline (file);
+    return err;
+}
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 54b0661db8..c53bd47bb5 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -141,6 +141,8 @@ static void brw_emit_prim(struct brw_context *brw,
 
    prim_packet.verts_per_instance = trim(prim->mode, prim->count);
    prim_packet.start_vert_location = prim->start;
+   if (prim->indexed)
+      prim_packet.start_vert_location += brw->ib.start_vertex_offset;
    prim_packet.instance_count = 1;
    prim_packet.start_instance_location = 0;
    prim_packet.base_vert_location = 0;
@@ -188,19 +190,13 @@ static void brw_merge_inputs( struct brw_context *brw,
       brw->vb.inputs[i].attrib = (gl_vert_attrib) i;
 
       if (arrays[i]->StrideB != 0)
-	 brw->vb.info.varying |= 1 << i;
-
 	 brw->vb.info.sizes[i/16] |= (brw->vb.inputs[i].glarray->Size - 1) <<
 	    ((i%16) * 2);
    }
 
-   /* Raise statechanges if input sizes and varying have changed: 
-    */
+   /* Raise statechanges if input sizes have changed. */
    if (memcmp(brw->vb.info.sizes, old.sizes, sizeof(old.sizes)) != 0)
       brw->state.dirty.brw |= BRW_NEW_INPUT_DIMENSIONS;
-
-   if (brw->vb.info.varying != old.varying)
-      brw->state.dirty.brw |= BRW_NEW_INPUT_VARYING;
 }
 
 /* XXX: could split the primitive list to fallback only on the
@@ -417,6 +413,8 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
  out:
    UNLOCK_HARDWARE(intel);
 
+   brw_state_cache_check_size(brw);
+
    if (warn)
       fprintf(stderr, "i965: Single primitive emit potentially exceeded "
 	      "available aperture space\n");
@@ -427,54 +425,31 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    return retval;
 }
 
-static GLboolean brw_need_rebase( GLcontext *ctx,
-				  const struct gl_client_array *arrays[],
-				  const struct _mesa_index_buffer *ib,
-				  GLuint min_index )
-{
-   if (min_index == 0) 
-      return GL_FALSE;
-
-   if (ib) {
-      if (!vbo_all_varyings_in_vbos(arrays))
-	 return GL_TRUE;
-      else
-	 return GL_FALSE;
-   }
-   else {
-      /* Hmm.  This isn't quite what I wanted.  BRW can actually
-       * handle the mixed case well enough that we shouldn't need to
-       * rebase.  However, it's probably not very common, nor hugely
-       * expensive to do it this way:
-       */
-      if (!vbo_all_varyings_in_vbos(arrays))
-	 return GL_TRUE;
-      else
-	 return GL_FALSE;
-   }
-}
-				  
-
 void brw_draw_prims( GLcontext *ctx,
 		     const struct gl_client_array *arrays[],
 		     const struct _mesa_prim *prim,
 		     GLuint nr_prims,
 		     const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
 		     GLuint min_index,
 		     GLuint max_index )
 {
    GLboolean retval;
 
-   /* Decide if we want to rebase.  If so we end up recursing once
-    * only into this function.
-    */
-   if (brw_need_rebase( ctx, arrays, ib, min_index )) {
-      vbo_rebase_prims( ctx, arrays, 
-			prim, nr_prims, 
-			ib, min_index, max_index, 
-			brw_draw_prims );
-      
-      return;
+   if (!vbo_all_varyings_in_vbos(arrays)) {
+      if (!index_bounds_valid)
+	 vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+
+      /* Decide if we want to rebase.  If so we end up recursing once
+       * only into this function.
+       */
+      if (min_index != 0) {
+	 vbo_rebase_prims(ctx, arrays,
+			  prim, nr_prims,
+			  ib, min_index, max_index,
+			  brw_draw_prims );
+	 return;
+      }
    }
 
    /* Make a first attempt at drawing:
diff --git a/src/mesa/drivers/dri/i965/brw_draw.h b/src/mesa/drivers/dri/i965/brw_draw.h
index 9aebbdb1b8..2a14db217f 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.h
+++ b/src/mesa/drivers/dri/i965/brw_draw.h
@@ -39,6 +39,7 @@ void brw_draw_prims( GLcontext *ctx,
 		     const struct _mesa_prim *prims,
 		     GLuint nr_prims,
 		     const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
 		     GLuint min_index,
 		     GLuint max_index );
 
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index fd9c3915c4..4aa17fa02d 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -383,7 +383,6 @@ static void brw_prepare_vertices(struct brw_context *brw)
       struct brw_vertex_element *input = brw->vb.enabled[i];
 
       input->element_size = get_size(input->glarray->Type) * input->glarray->Size;
-      input->count = input->glarray->StrideB ? max_index + 1 - min_index : 1;
 
       if (input->glarray->BufferObj->Name != 0) {
 	 struct intel_buffer_object *intel_buffer =
@@ -396,6 +395,7 @@ static void brw_prepare_vertices(struct brw_context *brw)
 	 dri_bo_reference(input->bo);
 	 input->offset = (unsigned long)input->glarray->Ptr;
 	 input->stride = input->glarray->StrideB;
+	 input->count = input->glarray->_MaxElement;
 
 	 /* This is a common place to reach if the user mistakenly supplies
 	  * a pointer in place of a VBO offset.  If we just let it go through,
@@ -411,6 +411,7 @@ static void brw_prepare_vertices(struct brw_context *brw)
 	  */
 	 assert(input->offset < input->bo->size);
       } else {
+	 input->count = input->glarray->StrideB ? max_index + 1 - min_index : 1;
 	 if (input->bo != NULL) {
 	    /* Already-uploaded vertex data is present from a previous
 	     * prepare_vertices, but we had to re-validate state due to
@@ -534,7 +535,19 @@ static void brw_emit_vertices(struct brw_context *brw)
       OUT_RELOC(input->bo,
 		I915_GEM_DOMAIN_VERTEX, 0,
 		input->offset);
-      OUT_BATCH(brw->vb.max_index);
+      if (BRW_IS_IGDNG(brw)) {
+          if (input->stride) {
+              OUT_RELOC(input->bo,
+                        I915_GEM_DOMAIN_VERTEX, 0,
+                        input->offset + input->stride * input->count);
+          } else {
+              assert(input->count == 1);
+              OUT_RELOC(input->bo,
+                        I915_GEM_DOMAIN_VERTEX, 0,
+                        input->offset + input->element_size);
+          }
+      } else
+          OUT_BATCH(input->stride ? input->count : 0);
       OUT_BATCH(0); /* Instance data step rate */
    }
    ADVANCE_BATCH();
@@ -564,11 +577,18 @@ static void brw_emit_vertices(struct brw_context *brw)
 		BRW_VE0_VALID |
 		(format << BRW_VE0_FORMAT_SHIFT) |
 		(0 << BRW_VE0_SRC_OFFSET_SHIFT));
-      OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
-		(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
-		(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
-		(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
-		((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
+
+      if (BRW_IS_IGDNG(brw))
+          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
+                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
+                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT));
+      else
+          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
+                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
+                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
+                    ((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
    }
    ADVANCE_BATCH();
 }
@@ -592,17 +612,20 @@ static void brw_prepare_indices(struct brw_context *brw)
    dri_bo *bo = NULL;
    struct gl_buffer_object *bufferobj;
    GLuint offset;
+   GLuint ib_type_size;
 
    if (index_buffer == NULL)
       return;
 
-   ib_size = get_size(index_buffer->type) * index_buffer->count;
+   ib_type_size = get_size(index_buffer->type);
+   ib_size = ib_type_size * index_buffer->count;
    bufferobj = index_buffer->obj;;
 
    /* Turn into a proper VBO:
     */
    if (!bufferobj->Name) {
-     
+      brw->ib.start_vertex_offset = 0;
+
       /* Get new bufferobj, offset:
        */
       get_space(brw, ib_size, &bo, &offset);
@@ -618,6 +641,7 @@ static void brw_prepare_indices(struct brw_context *brw)
       }
    } else {
       offset = (GLuint) (unsigned long) index_buffer->ptr;
+      brw->ib.start_vertex_offset = 0;
 
       /* If the index buffer isn't aligned to its element size, we have to
        * rebase it into a temporary.
@@ -638,39 +662,62 @@ static void brw_prepare_indices(struct brw_context *brw)
 	  bo = intel_bufferobj_buffer(intel, intel_buffer_object(bufferobj),
 				      INTEL_READ);
 	  dri_bo_reference(bo);
+
+	  /* Use CMD_3D_PRIM's start_vertex_offset to avoid re-uploading
+	   * the index buffer state when we're just moving the start index
+	   * of our drawing.
+	   */
+	  brw->ib.start_vertex_offset = offset / ib_type_size;
+	  offset = 0;
+	  ib_size = bo->size;
        }
    }
 
-   dri_bo_unreference(brw->ib.bo);
-   brw->ib.bo = bo;
-   brw->ib.offset = offset;
+   if (brw->ib.bo != bo ||
+       brw->ib.offset != offset ||
+       brw->ib.size != ib_size)
+   {
+      drm_intel_bo_unreference(brw->ib.bo);
+      brw->ib.bo = bo;
+      brw->ib.offset = offset;
+      brw->ib.size = ib_size;
+
+      brw->state.dirty.brw |= BRW_NEW_INDEX_BUFFER;
+   } else {
+      drm_intel_bo_unreference(bo);
+   }
 
    brw_add_validated_bo(brw, brw->ib.bo);
 }
 
-static void brw_emit_indices(struct brw_context *brw)
+const struct brw_tracked_state brw_indices = {
+   .dirty = {
+      .mesa = 0,
+      .brw = BRW_NEW_INDICES,
+      .cache = 0,
+   },
+   .prepare = brw_prepare_indices,
+};
+
+static void brw_emit_index_buffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    const struct _mesa_index_buffer *index_buffer = brw->ib.ib;
-   GLuint ib_size;
 
    if (index_buffer == NULL)
       return;
 
-   ib_size = get_size(index_buffer->type) * index_buffer->count - 1;
-
    /* Emit the indexbuffer packet:
     */
    {
       struct brw_indexbuffer ib;
 
       memset(&ib, 0, sizeof(ib));
-   
+
       ib.header.bits.opcode = CMD_INDEX_BUFFER;
       ib.header.bits.length = sizeof(ib)/4 - 2;
       ib.header.bits.index_format = get_index_type(index_buffer->type);
       ib.header.bits.cut_index_enable = 0;
-   
 
       BEGIN_BATCH(4, IGNORE_CLIPRECTS);
       OUT_BATCH( ib.header.dword );
@@ -679,18 +726,17 @@ static void brw_emit_indices(struct brw_context *brw)
 		brw->ib.offset);
       OUT_RELOC(brw->ib.bo,
 		I915_GEM_DOMAIN_VERTEX, 0,
-		brw->ib.offset + ib_size);
+		brw->ib.offset + brw->ib.size);
       OUT_BATCH( 0 );
       ADVANCE_BATCH();
    }
 }
 
-const struct brw_tracked_state brw_indices = {
+const struct brw_tracked_state brw_index_buffer = {
    .dirty = {
       .mesa = 0,
-      .brw = BRW_NEW_BATCH | BRW_NEW_INDICES,
+      .brw = BRW_NEW_BATCH | BRW_NEW_INDEX_BUFFER,
       .cache = 0,
    },
-   .prepare = brw_prepare_indices,
-   .emit = brw_emit_indices,
+   .emit = brw_emit_index_buffer,
 };
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 003332f78c..30603bdd0e 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -97,7 +97,7 @@ struct brw_glsl_call;
 
 
 #define BRW_EU_MAX_INSN_STACK 5
-#define BRW_EU_MAX_INSN 4000
+#define BRW_EU_MAX_INSN 10000
 
 struct brw_compile {
    struct brw_instruction store[BRW_EU_MAX_INSN];
@@ -171,9 +171,9 @@ static INLINE struct brw_reg brw_reg( GLuint file,
 {
    struct brw_reg reg;
    if (type == BRW_GENERAL_REGISTER_FILE)
-      assert(nr < 128);
+      assert(nr < BRW_MAX_GRF);
    else if (type == BRW_MESSAGE_REGISTER_FILE)
-      assert(nr < 9);
+      assert(nr < BRW_MAX_MRF);
    else if (type == BRW_ARCHITECTURE_REGISTER_FILE)
       assert(nr <= BRW_ARF_IP);
 
@@ -538,6 +538,7 @@ static INLINE struct brw_reg brw_mask_reg( GLuint subnr )
 
 static INLINE struct brw_reg brw_message_reg( GLuint nr )
 {
+   assert(nr < BRW_MAX_MRF);
    return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE,
 		       nr,
 		       0);
@@ -815,6 +816,19 @@ void brw_urb_WRITE(struct brw_compile *p,
 		   GLuint offset,
 		   GLuint swizzle);
 
+void brw_ff_sync(struct brw_compile *p,
+		   struct brw_reg dest,
+		   GLuint msg_reg_nr,
+		   struct brw_reg src0,
+		   GLboolean allocate,
+		   GLboolean used,
+		   GLuint msg_length,
+		   GLuint response_length,
+		   GLboolean eot,
+		   GLboolean writes_complete,
+		   GLuint offset,
+		   GLuint swizzle);
+
 void brw_fb_WRITE(struct brw_compile *p,
 		   struct brw_reg dest,
 		   GLuint msg_reg_nr,
@@ -834,7 +848,9 @@ void brw_SAMPLE(struct brw_compile *p,
 		GLuint msg_type,
 		GLuint response_length,
 		GLuint msg_length,
-		GLboolean eot);
+		GLboolean eot,
+		GLuint header_present,
+		GLuint simd_mode);
 
 void brw_math_16( struct brw_compile *p,
 		  struct brw_reg dest,
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 48243e1574..241cdc33f8 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -241,7 +241,8 @@ void brw_set_src1( struct brw_instruction *insn,
 
 
 
-static void brw_set_math_message( struct brw_instruction *insn,
+static void brw_set_math_message( struct brw_context *brw,
+				  struct brw_instruction *insn,
 				  GLuint msg_length,
 				  GLuint response_length,
 				  GLuint function,
@@ -252,18 +253,35 @@ static void brw_set_math_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.math.function = function;
-   insn->bits3.math.int_type = integer_type;
-   insn->bits3.math.precision = low_precision;
-   insn->bits3.math.saturate = saturate;
-   insn->bits3.math.data_type = dataType;
-   insn->bits3.math.response_length = response_length;
-   insn->bits3.math.msg_length = msg_length;
-   insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH;
-   insn->bits3.math.end_of_thread = 0;
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.math_igdng.function = function;
+       insn->bits3.math_igdng.int_type = integer_type;
+       insn->bits3.math_igdng.precision = low_precision;
+       insn->bits3.math_igdng.saturate = saturate;
+       insn->bits3.math_igdng.data_type = dataType;
+       insn->bits3.math_igdng.snapshot = 0;
+       insn->bits3.math_igdng.header_present = 0;
+       insn->bits3.math_igdng.response_length = response_length;
+       insn->bits3.math_igdng.msg_length = msg_length;
+       insn->bits3.math_igdng.end_of_thread = 0;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_MATH;
+       insn->bits2.send_igdng.end_of_thread = 0;
+   } else {
+       insn->bits3.math.function = function;
+       insn->bits3.math.int_type = integer_type;
+       insn->bits3.math.precision = low_precision;
+       insn->bits3.math.saturate = saturate;
+       insn->bits3.math.data_type = dataType;
+       insn->bits3.math.response_length = response_length;
+       insn->bits3.math.msg_length = msg_length;
+       insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH;
+       insn->bits3.math.end_of_thread = 0;
+   }
 }
 
-static void brw_set_urb_message( struct brw_instruction *insn,
+
+static void brw_set_ff_sync_message( struct brw_context *brw,
+				 struct brw_instruction *insn,
 				 GLboolean allocate,
 				 GLboolean used,
 				 GLuint msg_length,
@@ -273,21 +291,64 @@ static void brw_set_urb_message( struct brw_instruction *insn,
 				 GLuint offset,
 				 GLuint swizzle_control )
 {
-   brw_set_src1(insn, brw_imm_d(0));
+	brw_set_src1(insn, brw_imm_d(0));
+
+	insn->bits3.urb_igdng.opcode = 1;
+	insn->bits3.urb_igdng.offset = offset;
+	insn->bits3.urb_igdng.swizzle_control = swizzle_control;
+	insn->bits3.urb_igdng.allocate = allocate;
+	insn->bits3.urb_igdng.used = used;
+	insn->bits3.urb_igdng.complete = complete;
+	insn->bits3.urb_igdng.header_present = 1;
+	insn->bits3.urb_igdng.response_length = response_length;
+	insn->bits3.urb_igdng.msg_length = msg_length;
+	insn->bits3.urb_igdng.end_of_thread = end_of_thread;
+	insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB;
+	insn->bits2.send_igdng.end_of_thread = end_of_thread;
+}
 
-   insn->bits3.urb.opcode = 0;	/* ? */
-   insn->bits3.urb.offset = offset;
-   insn->bits3.urb.swizzle_control = swizzle_control;
-   insn->bits3.urb.allocate = allocate;
-   insn->bits3.urb.used = used;	/* ? */
-   insn->bits3.urb.complete = complete;
-   insn->bits3.urb.response_length = response_length;
-   insn->bits3.urb.msg_length = msg_length;
-   insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB;
-   insn->bits3.urb.end_of_thread = end_of_thread;
+static void brw_set_urb_message( struct brw_context *brw,
+				 struct brw_instruction *insn,
+				 GLboolean allocate,
+				 GLboolean used,
+				 GLuint msg_length,
+				 GLuint response_length,
+				 GLboolean end_of_thread,
+				 GLboolean complete,
+				 GLuint offset,
+				 GLuint swizzle_control )
+{
+    brw_set_src1(insn, brw_imm_d(0));
+
+    if (BRW_IS_IGDNG(brw)) {
+        insn->bits3.urb_igdng.opcode = 0;	/* ? */
+        insn->bits3.urb_igdng.offset = offset;
+        insn->bits3.urb_igdng.swizzle_control = swizzle_control;
+        insn->bits3.urb_igdng.allocate = allocate;
+        insn->bits3.urb_igdng.used = used;	/* ? */
+        insn->bits3.urb_igdng.complete = complete;
+        insn->bits3.urb_igdng.header_present = 1;
+        insn->bits3.urb_igdng.response_length = response_length;
+        insn->bits3.urb_igdng.msg_length = msg_length;
+        insn->bits3.urb_igdng.end_of_thread = end_of_thread;
+        insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB;
+        insn->bits2.send_igdng.end_of_thread = end_of_thread;
+    } else {
+        insn->bits3.urb.opcode = 0;	/* ? */
+        insn->bits3.urb.offset = offset;
+        insn->bits3.urb.swizzle_control = swizzle_control;
+        insn->bits3.urb.allocate = allocate;
+        insn->bits3.urb.used = used;	/* ? */
+        insn->bits3.urb.complete = complete;
+        insn->bits3.urb.response_length = response_length;
+        insn->bits3.urb.msg_length = msg_length;
+        insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB;
+        insn->bits3.urb.end_of_thread = end_of_thread;
+    }
 }
 
-static void brw_set_dp_write_message( struct brw_instruction *insn,
+static void brw_set_dp_write_message( struct brw_context *brw,
+				      struct brw_instruction *insn,
 				      GLuint binding_table_index,
 				      GLuint msg_control,
 				      GLuint msg_type,
@@ -298,18 +359,33 @@ static void brw_set_dp_write_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.dp_write.binding_table_index = binding_table_index;
-   insn->bits3.dp_write.msg_control = msg_control;
-   insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear;
-   insn->bits3.dp_write.msg_type = msg_type;
-   insn->bits3.dp_write.send_commit_msg = 0;
-   insn->bits3.dp_write.response_length = response_length;
-   insn->bits3.dp_write.msg_length = msg_length;
-   insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
-   insn->bits3.urb.end_of_thread = end_of_thread;
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.dp_write_igdng.binding_table_index = binding_table_index;
+       insn->bits3.dp_write_igdng.msg_control = msg_control;
+       insn->bits3.dp_write_igdng.pixel_scoreboard_clear = pixel_scoreboard_clear;
+       insn->bits3.dp_write_igdng.msg_type = msg_type;
+       insn->bits3.dp_write_igdng.send_commit_msg = 0;
+       insn->bits3.dp_write_igdng.header_present = 1;
+       insn->bits3.dp_write_igdng.response_length = response_length;
+       insn->bits3.dp_write_igdng.msg_length = msg_length;
+       insn->bits3.dp_write_igdng.end_of_thread = end_of_thread;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
+       insn->bits2.send_igdng.end_of_thread = end_of_thread;
+   } else {
+       insn->bits3.dp_write.binding_table_index = binding_table_index;
+       insn->bits3.dp_write.msg_control = msg_control;
+       insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear;
+       insn->bits3.dp_write.msg_type = msg_type;
+       insn->bits3.dp_write.send_commit_msg = 0;
+       insn->bits3.dp_write.response_length = response_length;
+       insn->bits3.dp_write.msg_length = msg_length;
+       insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
+       insn->bits3.dp_write.end_of_thread = end_of_thread;
+   }
 }
 
-static void brw_set_dp_read_message( struct brw_instruction *insn,
+static void brw_set_dp_read_message( struct brw_context *brw,
+				      struct brw_instruction *insn,
 				      GLuint binding_table_index,
 				      GLuint msg_control,
 				      GLuint msg_type,
@@ -320,15 +396,29 @@ static void brw_set_dp_read_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/
-   insn->bits3.dp_read.msg_control = msg_control;  /*8:11*/
-   insn->bits3.dp_read.msg_type = msg_type;  /*12:13*/
-   insn->bits3.dp_read.target_cache = target_cache;  /*14:15*/
-   insn->bits3.dp_read.response_length = response_length;  /*16:19*/
-   insn->bits3.dp_read.msg_length = msg_length;  /*20:23*/
-   insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; /*24:27*/
-   insn->bits3.dp_read.pad1 = 0;  /*28:30*/
-   insn->bits3.dp_read.end_of_thread = end_of_thread;  /*31*/
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.dp_read_igdng.binding_table_index = binding_table_index;
+       insn->bits3.dp_read_igdng.msg_control = msg_control;
+       insn->bits3.dp_read_igdng.msg_type = msg_type;
+       insn->bits3.dp_read_igdng.target_cache = target_cache;
+       insn->bits3.dp_read_igdng.header_present = 1;
+       insn->bits3.dp_read_igdng.response_length = response_length;
+       insn->bits3.dp_read_igdng.msg_length = msg_length;
+       insn->bits3.dp_read_igdng.pad1 = 0;
+       insn->bits3.dp_read_igdng.end_of_thread = end_of_thread;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_READ;
+       insn->bits2.send_igdng.end_of_thread = end_of_thread;
+   } else {
+       insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/
+       insn->bits3.dp_read.msg_control = msg_control;  /*8:11*/
+       insn->bits3.dp_read.msg_type = msg_type;  /*12:13*/
+       insn->bits3.dp_read.target_cache = target_cache;  /*14:15*/
+       insn->bits3.dp_read.response_length = response_length;  /*16:19*/
+       insn->bits3.dp_read.msg_length = msg_length;  /*20:23*/
+       insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; /*24:27*/
+       insn->bits3.dp_read.pad1 = 0;  /*28:30*/
+       insn->bits3.dp_read.end_of_thread = end_of_thread;  /*31*/
+   }
 }
 
 static void brw_set_sampler_message(struct brw_context *brw,
@@ -338,11 +428,25 @@ static void brw_set_sampler_message(struct brw_context *brw,
                                     GLuint msg_type,
                                     GLuint response_length,
                                     GLuint msg_length,
-                                    GLboolean eot)
+                                    GLboolean eot,
+                                    GLuint header_present,
+                                    GLuint simd_mode)
 {
+   assert(eot == 0);
    brw_set_src1(insn, brw_imm_d(0));
 
-   if (BRW_IS_G4X(brw)) {
+   if (BRW_IS_IGDNG(brw)) {
+      insn->bits3.sampler_igdng.binding_table_index = binding_table_index;
+      insn->bits3.sampler_igdng.sampler = sampler;
+      insn->bits3.sampler_igdng.msg_type = msg_type;
+      insn->bits3.sampler_igdng.simd_mode = simd_mode;
+      insn->bits3.sampler_igdng.header_present = header_present;
+      insn->bits3.sampler_igdng.response_length = response_length;
+      insn->bits3.sampler_igdng.msg_length = msg_length;
+      insn->bits3.sampler_igdng.end_of_thread = eot;
+      insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_SAMPLER;
+      insn->bits2.send_igdng.end_of_thread = eot;
+   } else if (BRW_IS_G4X(brw)) {
       insn->bits3.sampler_g4x.binding_table_index = binding_table_index;
       insn->bits3.sampler_g4x.sampler = sampler;
       insn->bits3.sampler_g4x.msg_type = msg_type;
@@ -484,6 +588,10 @@ struct brw_instruction *brw_JMPI(struct brw_compile *p,
 {
    struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1);
 
+   insn->header.execution_size = 1;
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+
    p->current->header.predicate_control = BRW_PREDICATE_NONE;
 
    return insn;
@@ -540,6 +648,10 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 				 struct brw_instruction *if_insn)
 {
    struct brw_instruction *insn;
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2;
 
    if (p->single_program_flow) {
       insn = next_insn(p, BRW_OPCODE_ADD);
@@ -566,8 +678,8 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
    } else {
       assert(if_insn->header.opcode == BRW_OPCODE_IF);
 
-      if_insn->bits3.if_else.jump_count = insn - if_insn;
-      if_insn->bits3.if_else.pop_count = 1;
+      if_insn->bits3.if_else.jump_count = br * (insn - if_insn);
+      if_insn->bits3.if_else.pop_count = 0;
       if_insn->bits3.if_else.pad0 = 0;
    }
 
@@ -577,6 +689,11 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 void brw_ENDIF(struct brw_compile *p, 
 	       struct brw_instruction *patch_insn)
 {
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2; 
+ 
    if (p->single_program_flow) {
       /* In single program flow mode, there's no need to execute an ENDIF,
        * since we don't need to do any stack operations, and if we're executing
@@ -608,11 +725,11 @@ void brw_ENDIF(struct brw_compile *p,
 	 /* Automagically turn it into an IFF:
 	  */
 	 patch_insn->header.opcode = BRW_OPCODE_IFF;
-	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
 	 patch_insn->bits3.if_else.pop_count = 0;
 	 patch_insn->bits3.if_else.pad0 = 0;
       } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
-	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
 	 patch_insn->bits3.if_else.pop_count = 1;
 	 patch_insn->bits3.if_else.pad0 = 0;
       } else {
@@ -686,6 +803,10 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p,
                                   struct brw_instruction *do_insn)
 {
    struct brw_instruction *insn;
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2;
 
    if (p->single_program_flow)
       insn = next_insn(p, BRW_OPCODE_ADD);
@@ -706,7 +827,7 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p,
       insn->header.execution_size = do_insn->header.execution_size;
 
       assert(do_insn->header.opcode == BRW_OPCODE_DO);
-      insn->bits3.if_else.jump_count = do_insn - insn + 1;
+      insn->bits3.if_else.jump_count = br * (do_insn - insn + 1);
       insn->bits3.if_else.pop_count = 0;
       insn->bits3.if_else.pad0 = 0;
    }
@@ -725,11 +846,15 @@ void brw_land_fwd_jump(struct brw_compile *p,
 		       struct brw_instruction *jmp_insn)
 {
    struct brw_instruction *landing = &p->store[p->nr_insn];
+   GLuint jmpi = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
 
    assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI);
    assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE);
 
-   jmp_insn->bits3.ud = (landing - jmp_insn) - 1; 
+   jmp_insn->bits3.ud = jmpi * ((landing - jmp_insn) - 1);
 }
 
 
@@ -794,7 +919,8 @@ void brw_math( struct brw_compile *p,
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw,
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -830,7 +956,8 @@ void brw_math_16( struct brw_compile *p,
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw,
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -846,7 +973,8 @@ void brw_math_16( struct brw_compile *p,
 
    brw_set_dest(insn, offset(dest,1));
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw, 
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -893,7 +1021,8 @@ void brw_dp_WRITE_16( struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, src);
 
-      brw_set_dp_write_message(insn,
+      brw_set_dp_write_message(p->brw,
+			       insn,
 			       255, /* binding table index (255=stateless) */
 			       BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */
 			       BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */
@@ -938,7 +1067,8 @@ void brw_dp_READ_16( struct brw_compile *p,
       brw_set_dest(insn, dest);	/* UW? */
       brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW));
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      255, /* binding table index (255=stateless) */
 			      3,  /* msg_control (3 means 4 Owords) */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -995,7 +1125,8 @@ void brw_dp_READ_4( struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, brw_null_reg());
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      bind_table_index,
 			      0,  /* msg_control (0 means 1 Oword) */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -1066,7 +1197,8 @@ void brw_dp_READ_4_vs(struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, brw_null_reg());
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      bind_table_index,
 			      oword,  /* 0 = lower Oword, 1 = upper Oword */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -1096,7 +1228,8 @@ void brw_fb_WRITE(struct brw_compile *p,
   
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src0);
-   brw_set_dp_write_message(insn,
+   brw_set_dp_write_message(p->brw,
+			    insn,
 			    binding_table_index,
 			    BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */
 			    BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */
@@ -1122,7 +1255,9 @@ void brw_SAMPLE(struct brw_compile *p,
 		GLuint msg_type,
 		GLuint response_length,
 		GLuint msg_length,
-		GLboolean eot)
+		GLboolean eot,
+		GLuint header_present,
+		GLuint simd_mode)
 {
    GLboolean need_stall = 0;
    
@@ -1197,7 +1332,9 @@ void brw_SAMPLE(struct brw_compile *p,
 			      msg_type,
 			      response_length, 
 			      msg_length,
-			      eot);
+			      eot,
+			      header_present,
+			      simd_mode);
    }
 
    if (need_stall) {
@@ -1232,7 +1369,7 @@ void brw_urb_WRITE(struct brw_compile *p,
 {
    struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
 
-   assert(msg_length < 16);
+   assert(msg_length < BRW_MAX_MRF);
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src0);
@@ -1240,7 +1377,8 @@ void brw_urb_WRITE(struct brw_compile *p,
 
    insn->header.destreg__conditionalmod = msg_reg_nr;
 
-   brw_set_urb_message(insn,
+   brw_set_urb_message(p->brw,
+		       insn,
 		       allocate,
 		       used,
 		       msg_length,
@@ -1251,3 +1389,37 @@ void brw_urb_WRITE(struct brw_compile *p,
 		       swizzle);
 }
 
+void brw_ff_sync(struct brw_compile *p,
+		   struct brw_reg dest,
+		   GLuint msg_reg_nr,
+		   struct brw_reg src0,
+		   GLboolean allocate,
+		   GLboolean used,
+		   GLuint msg_length,
+		   GLuint response_length,
+		   GLboolean eot,
+		   GLboolean writes_complete,
+		   GLuint offset,
+		   GLuint swizzle)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+   assert(msg_length < 16);
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->header.destreg__conditionalmod = msg_reg_nr;
+
+   brw_set_ff_sync_message(p->brw,
+		       insn,
+		       allocate,
+		       used,
+		       msg_length,
+		       response_length, 
+		       eot, 
+		       writes_complete, 
+		       offset,
+		       swizzle);
+}
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 299357409c..d27c6c24ca 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -37,6 +37,9 @@
 #include "tnl/tnl.h"
 #include "brw_context.h"
 #include "brw_fallback.h"
+#include "intel_chipset.h"
+#include "intel_fbo.h"
+#include "intel_regions.h"
 
 #include "glapi/glapi.h"
 
@@ -44,6 +47,7 @@
 
 static GLboolean do_check_fallback(struct brw_context *brw)
 {
+   struct intel_context *intel = &brw->intel;
    GLcontext *ctx = &brw->intel.ctx;
    GLuint i;
 
@@ -81,6 +85,33 @@ static GLboolean do_check_fallback(struct brw_context *brw)
       return GL_TRUE;
    }
 
+   /* _NEW_BUFFERS */
+   if (IS_965(intel->intelScreen->deviceID) &&
+       !IS_G4X(intel->intelScreen->deviceID)) {
+      for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
+	 struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+	 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+
+	 /* The original gen4 hardware couldn't set up WM surfaces pointing
+	  * at an offset within a tile, which can happen when rendering to
+	  * anything but the base level of a texture or the +X face/0 depth.
+	  * This was fixed with the 4 Series hardware.
+	  *
+	  * For these original chips, you would have to make the depth and
+	  * color destination surfaces include information on the texture
+	  * type, LOD, face, and various limits to use them as a destination.
+	  * I would have done this, but there's also a nasty requirement that
+	  * the depth and the color surfaces all be of the same LOD, which
+	  * may be a worse requirement than this alignment.  (Also, we may
+	  * want to just demote the texture to untiled, instead).
+	  */
+	 if (irb->region && irb->region->tiling != I915_TILING_NONE &&
+	     (irb->region->draw_offset & 4095)) {
+	    DBG("FALLBACK: non-tile-aligned destination for tiled FBO\n");
+	    return GL_TRUE;
+	 }
+      }
+   }
 
    return GL_FALSE;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c
index a8b74a0afe..48c2b9a41c 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.c
+++ b/src/mesa/drivers/dri/i965/brw_gs.c
@@ -54,12 +54,17 @@ static void compile_gs_prog( struct brw_context *brw,
    memset(&c, 0, sizeof(c));
    
    c.key = *key;
-
+   c.need_ff_sync = BRW_IS_IGDNG(brw);
    /* Need to locate the two positions present in vertex + header.
     * These are currently hardcoded:
     */
    c.nr_attrs = brw_count_bits(c.key.attrs);
-   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
+   if (BRW_IS_IGDNG(brw))
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 3;  /* are vertices packed, or reg-aligned? */
+   else
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
    c.nr_bytes = c.nr_regs * REG_SIZE;
 
    
diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h
index 18a4537c32..bbb991ea2e 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.h
+++ b/src/mesa/drivers/dri/i965/brw_gs.h
@@ -62,6 +62,7 @@ struct brw_gs_compile {
    GLuint nr_attrs;
    GLuint nr_regs;
    GLuint nr_bytes;
+   GLboolean need_ff_sync;
 };
 
 #define ATTR_SIZE  (4*4)
diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c
index 22e0d25c2e..a9b2aa2eac 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c
@@ -101,6 +101,23 @@ static void brw_gs_emit_vue(struct brw_gs_compile *c,
 		 BRW_URB_SWIZZLE_NONE);
 }
 
+static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim)
+{
+	struct brw_compile *p = &c->func;
+	brw_MOV(p, get_element_ud(c->reg.R0, 1), brw_imm_ud(num_prim));
+	brw_ff_sync(p, 
+				c->reg.R0,
+				0,
+				c->reg.R0,
+				1,	
+				1,		/* used */
+				1,  	/* msg length */
+				1,		/* response length */
+				0,		/* eot */
+				1,		/* write compelete */
+				0,		/* urb offset */
+				BRW_URB_SWIZZLE_NONE);
+}
 
 
 void brw_gs_quads( struct brw_gs_compile *c )
@@ -110,6 +127,8 @@ void brw_gs_quads( struct brw_gs_compile *c )
    /* Use polygons for correct edgeflag behaviour. Note that vertex 3
     * is the PV for quads, but vertex 0 for polygons:
     */
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);    
    brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2));
    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); 
@@ -120,6 +139,8 @@ void brw_gs_quad_strip( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 4);
    
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2));
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); 
@@ -129,6 +150,9 @@ void brw_gs_quad_strip( struct brw_gs_compile *c )
 void brw_gs_tris( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 3);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2));
    brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END));
@@ -137,6 +161,9 @@ void brw_gs_tris( struct brw_gs_compile *c )
 void brw_gs_lines( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 2);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END));
 }
@@ -144,6 +171,9 @@ void brw_gs_lines( struct brw_gs_compile *c )
 void brw_gs_points( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 1);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END));
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c
index 27023cf034..a761c03153 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_state.c
@@ -95,6 +95,9 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
 
    gs.thread4.max_threads = 0; /* Hardware requirement */
 
+   if (BRW_IS_IGDNG(brw))
+      gs.thread4.rendering_enable = 1;
+
    if (INTEL_DEBUG & DEBUG_STATS)
       gs.thread4.stats_enable = 1;
 
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 9bc5c35139..ea71857548 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -118,7 +118,10 @@ static void upload_binding_table_pointers(struct brw_context *brw)
 
    BEGIN_BATCH(6, IGNORE_CLIPRECTS);
    OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
-   OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   if (brw->vs.bind_bo != NULL)
+      OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   else
+      OUT_BATCH(0);
    OUT_BATCH(0); /* gs */
    OUT_BATCH(0); /* clip */
    OUT_BATCH(0); /* sf */
@@ -169,6 +172,7 @@ static void prepare_psp_urb_cbs(struct brw_context *brw)
    brw_add_validated_bo(brw, brw->vs.state_bo);
    brw_add_validated_bo(brw, brw->gs.state_bo);
    brw_add_validated_bo(brw, brw->clip.state_bo);
+   brw_add_validated_bo(brw, brw->sf.state_bo);
    brw_add_validated_bo(brw, brw->wm.state_bo);
    brw_add_validated_bo(brw, brw->cc.state_bo);
 }
@@ -208,7 +212,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    struct intel_region *region = brw->state.depth_region;
-   unsigned int len = BRW_IS_G4X(brw) ? 6 : 5;
+   unsigned int len = (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? 6 : 5;
 
    if (region == NULL) {
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
@@ -219,7 +223,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       OUT_BATCH(0);
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -241,6 +245,8 @@ static void emit_depthbuffer(struct brw_context *brw)
 	 return;
       }
 
+      assert(region->tiling != I915_TILING_X);
+
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
       OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
       OUT_BATCH(((region->pitch * region->cpp) - 1) |
@@ -256,7 +262,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 		((region->height - 1) << 19));
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -369,7 +375,7 @@ static void upload_aa_line_parameters(struct brw_context *brw)
 {
    struct brw_aa_line_parameters balp;
    
-   if (!BRW_IS_G4X(brw))
+   if (BRW_IS_965(brw))
       return;
 
    /* use legacy aa line coverage computation */
@@ -506,14 +512,27 @@ static void upload_state_base_address( struct brw_context *brw )
    /* Output the structure (brw_state_base_address) directly to the
     * batchbuffer, so we can emit relocations inline.
     */
-   BEGIN_BATCH(6, IGNORE_CLIPRECTS);
-   OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
-   OUT_BATCH(1); /* General state base address */
-   OUT_BATCH(1); /* Surface state base address */
-   OUT_BATCH(1); /* Indirect object base address */
-   OUT_BATCH(1); /* General state upper bound */
-   OUT_BATCH(1); /* Indirect object upper bound */
-   ADVANCE_BATCH();
+   if (BRW_IS_IGDNG(brw)) {
+       BEGIN_BATCH(8, IGNORE_CLIPRECTS);
+       OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2));
+       OUT_BATCH(1); /* General state base address */
+       OUT_BATCH(1); /* Surface state base address */
+       OUT_BATCH(1); /* Indirect object base address */
+       OUT_BATCH(1); /* Instruction base address */
+       OUT_BATCH(1); /* General state upper bound */
+       OUT_BATCH(1); /* Indirect object upper bound */
+       OUT_BATCH(1); /* Instruction access upper bound */
+       ADVANCE_BATCH();
+   } else {
+       BEGIN_BATCH(6, IGNORE_CLIPRECTS);
+       OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
+       OUT_BATCH(1); /* General state base address */
+       OUT_BATCH(1); /* Surface state base address */
+       OUT_BATCH(1); /* Indirect object base address */
+       OUT_BATCH(1); /* General state upper bound */
+       OUT_BATCH(1); /* Indirect object upper bound */
+       ADVANCE_BATCH();
+   }
 }
 
 const struct brw_tracked_state brw_state_base_address = {
diff --git a/src/mesa/drivers/dri/i965/brw_queryobj.c b/src/mesa/drivers/dri/i965/brw_queryobj.c
index cb9169e2ee..a195bc32b0 100644
--- a/src/mesa/drivers/dri/i965/brw_queryobj.c
+++ b/src/mesa/drivers/dri/i965/brw_queryobj.c
@@ -146,17 +146,12 @@ static void brw_wait_query(GLcontext *ctx, struct gl_query_object *q)
 
 static void brw_check_query(GLcontext *ctx, struct gl_query_object *q)
 {
-   /* XXX: Need to expose dri_bo_is_idle from bufmgr. */
-#if 0
    struct brw_query_object *query = (struct brw_query_object *)q;
 
-   if (dri_bo_is_idle(query->bo)) {
+   if (query->bo == NULL || !drm_intel_bo_busy(query->bo)) {
       brw_queryobj_get_results(query);
       query->Base.Ready = GL_TRUE;
    }
-#else
-   brw_wait_query(ctx, q);
-#endif
 }
 
 /** Called to set up the query BO and account for its aperture space */
diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c
index c3c85978f4..e1c2c7777b 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.c
+++ b/src/mesa/drivers/dri/i965/brw_sf.c
@@ -166,6 +166,9 @@ static void upload_sf_prog(struct brw_context *brw)
    key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
    key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide);
 
+   /* _NEW_HINT */
+   key.linear_color = (ctx->Hint.PerspectiveCorrection == GL_FASTEST);
+
    /* _NEW_POLYGON */
    if (key.do_twoside_color) {
       /* If we're rendering to a FBO, we have to invert the polygon
@@ -188,7 +191,7 @@ static void upload_sf_prog(struct brw_context *brw)
 
 const struct brw_tracked_state brw_sf_prog = {
    .dirty = {
-      .mesa  = (_NEW_LIGHT|_NEW_POLYGON|_NEW_POINT),
+      .mesa  = (_NEW_HINT | _NEW_LIGHT | _NEW_POLYGON | _NEW_POINT),
       .brw   = (BRW_NEW_REDUCED_PRIMITIVE),
       .cache = CACHE_NEW_VS_PROG
    },
diff --git a/src/mesa/drivers/dri/i965/brw_sf.h b/src/mesa/drivers/dri/i965/brw_sf.h
index 1c0fb70fe0..6426b6df9f 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.h
+++ b/src/mesa/drivers/dri/i965/brw_sf.h
@@ -51,7 +51,8 @@ struct brw_sf_prog_key {
    GLuint do_flat_shading:1;
    GLuint frontface_ccw:1;
    GLuint do_point_sprite:1;
-   GLuint pad:10;
+   GLuint linear_color:1;  /**< linear interp vs. perspective interp */
+   GLuint pad:25;
    GLenum SpriteOrigin;
 };
 
diff --git a/src/mesa/drivers/dri/i965/brw_sf_emit.c b/src/mesa/drivers/dri/i965/brw_sf_emit.c
index 862835f157..ca8f97f9f9 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_emit.c
@@ -151,6 +151,8 @@ static void do_flatshade_triangle( struct brw_sf_compile *c )
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
    GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
+   GLuint jmpi = 1;
+
    if (!nr)
       return;
 
@@ -159,18 +161,21 @@ static void do_flatshade_triangle( struct brw_sf_compile *c )
    if (c->key.primitive == SF_UNFILLED_TRIS)
       return;
 
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
+
    brw_push_insn_state(p);
    
-   brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr*2+1));
+   brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr*2+1)));
    brw_JMPI(p, ip, ip, c->pv);
 
    copy_colors(c, c->vert[1], c->vert[0]);
    copy_colors(c, c->vert[2], c->vert[0]);
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr*4+1));
+   brw_JMPI(p, ip, ip, brw_imm_d(jmpi*(nr*4+1)));
 
    copy_colors(c, c->vert[0], c->vert[1]);
    copy_colors(c, c->vert[2], c->vert[1]);
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr*2));
+   brw_JMPI(p, ip, ip, brw_imm_d(jmpi*nr*2));
 
    copy_colors(c, c->vert[0], c->vert[2]);
    copy_colors(c, c->vert[1], c->vert[2]);
@@ -184,7 +189,8 @@ static void do_flatshade_line( struct brw_sf_compile *c )
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
    GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
-   
+   GLuint jmpi = 1;
+
    if (!nr)
       return;
 
@@ -193,13 +199,16 @@ static void do_flatshade_line( struct brw_sf_compile *c )
    if (c->key.primitive == SF_UNFILLED_TRIS)
       return;
 
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
+
    brw_push_insn_state(p);
    
-   brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr+1));
+   brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr+1)));
    brw_JMPI(p, ip, ip, c->pv);
    copy_colors(c, c->vert[1], c->vert[0]);
 
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr));
+   brw_JMPI(p, ip, ip, brw_imm_ud(jmpi*nr));
    copy_colors(c, c->vert[0], c->vert[1]);
 
    brw_pop_insn_state(p);
@@ -218,7 +227,7 @@ static void alloc_regs( struct brw_sf_compile *c )
 
    /* Values computed by fixed function unit:
     */
-   c->pv  = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD);
+   c->pv  = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_D);
    c->det = brw_vec1_grf(1, 2);
    c->dx0 = brw_vec1_grf(1, 3);
    c->dx2 = brw_vec1_grf(1, 4);
@@ -295,9 +304,6 @@ static void invert_det( struct brw_sf_compile *c)
 
 }
 
-#define NON_PERPECTIVE_ATTRS  (FRAG_BIT_WPOS | \
-                               FRAG_BIT_COL0 | \
-			       FRAG_BIT_COL1)
 
 static GLboolean calculate_masks( struct brw_sf_compile *c,
 				  GLuint reg,
@@ -306,9 +312,16 @@ static GLboolean calculate_masks( struct brw_sf_compile *c,
 				  GLushort *pc_linear)
 {
    GLboolean is_last_attr = (reg == c->nr_setup_regs - 1);
-   GLuint persp_mask = c->key.attrs & ~NON_PERPECTIVE_ATTRS;
+   GLuint persp_mask;
    GLuint linear_mask;
 
+   if (c->key.do_flat_shading || c->key.linear_color)
+      persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS |
+                                    FRAG_BIT_COL0 |
+                                    FRAG_BIT_COL1);
+   else
+      persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS);
+
    if (c->key.do_flat_shading)
       linear_mask = c->key.attrs & ~(FRAG_BIT_COL0|FRAG_BIT_COL1);
    else
@@ -674,7 +687,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 					       (1<<_3DPRIM_POLYGON) |
 					       (1<<_3DPRIM_RECTLIST) |
 					       (1<<_3DPRIM_TRIFAN_NOSTIPPLE)));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
@@ -695,7 +708,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 					       (1<<_3DPRIM_LINESTRIP_CONT) |
 					       (1<<_3DPRIM_LINESTRIP_BF) |
 					       (1<<_3DPRIM_LINESTRIP_CONT_BF)));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
@@ -708,7 +721,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 
    brw_set_conditionalmod(p, BRW_CONDITIONAL_Z);
    brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index 38733c8c73..bc0f076073 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -113,7 +113,7 @@ struct brw_sf_unit_key {
 
    unsigned int nr_urb_entries, urb_size, sfsize;
 
-   GLenum front_face, cull_face;
+   GLenum front_face, cull_face, provoking_vertex;
    unsigned scissor:1;
    unsigned line_smooth:1;
    unsigned point_sprite:1;
@@ -153,6 +153,9 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
    key->point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
    key->point_attenuated = ctx->Point._Attenuated;
 
+   /* _NEW_LIGHT */
+   key->provoking_vertex = ctx->Light.ProvokingVertex;
+
    key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
 }
 
@@ -162,7 +165,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 {
    struct brw_sf_unit_state sf;
    dri_bo *bo;
-
+   int chipset_max_threads;
    memset(&sf, 0, sizeof(sf));
 
    sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
@@ -171,13 +174,26 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
 
    sf.thread3.dispatch_grf_start_reg = 3;
-   sf.thread3.urb_entry_read_offset = 1;
+
+   if (BRW_IS_IGDNG(brw))
+       sf.thread3.urb_entry_read_offset = 3;
+   else
+       sf.thread3.urb_entry_read_offset = 1;
+
    sf.thread3.urb_entry_read_length = key->urb_entry_read_length;
 
    sf.thread4.nr_urb_entries = key->nr_urb_entries;
    sf.thread4.urb_entry_allocation_size = key->sfsize - 1;
-   /* Each SF thread produces 1 PUE, and there can be up to 24 threads */
-   sf.thread4.max_threads = MIN2(24, key->nr_urb_entries) - 1;
+
+   /* Each SF thread produces 1 PUE, and there can be up to 24(Pre-IGDNG) or 
+    * 48(IGDNG) threads 
+    */
+   if (BRW_IS_IGDNG(brw))
+      chipset_max_threads = 48;
+   else
+      chipset_max_threads = 24;
+
+   sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1;
 
    if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
       sf.thread4.max_threads = 0;
@@ -271,9 +287,15 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 
    /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
     */
-   sf.sf7.trifan_pv = 2;
-   sf.sf7.linestrip_pv = 1;
-   sf.sf7.tristrip_pv = 2;
+   if (key->provoking_vertex == GL_LAST_VERTEX_CONVENTION) {
+      sf.sf7.trifan_pv = 2;
+      sf.sf7.linestrip_pv = 1;
+      sf.sf7.tristrip_pv = 2;
+   } else {
+      sf.sf7.trifan_pv = 1;
+      sf.sf7.linestrip_pv = 0;
+      sf.sf7.tristrip_pv = 0;
+   }
    sf.sf7.line_last_pixel_enable = 0;
 
    /* Set bias for OpenGL rasterization rules:
@@ -287,6 +309,9 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 			 &sf, sizeof(sf),
 			 NULL, NULL);
 
+   /* STATE_PREFETCH command description describes this state as being
+    * something loaded through the GPE (L2 ISC), so it's INSTRUCTION domain.
+    */
    /* Emit SF program relocation */
    dri_bo_emit_reloc(bo,
 		     I915_GEM_DOMAIN_INSTRUCTION, 0,
@@ -327,6 +352,7 @@ static void upload_sf_unit( struct brw_context *brw )
 const struct brw_tracked_state brw_sf_unit = {
    .dirty = {
       .mesa  = (_NEW_POLYGON | 
+		_NEW_LIGHT |
 		_NEW_LINE | 
 		_NEW_POINT | 
 		_NEW_SCISSOR |
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index 81b0a45998..78572356a3 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -72,11 +72,13 @@ const struct brw_tracked_state brw_sf_vp;
 const struct brw_tracked_state brw_state_base_address;
 const struct brw_tracked_state brw_urb_fence;
 const struct brw_tracked_state brw_vertex_state;
+const struct brw_tracked_state brw_vs_surfaces;
 const struct brw_tracked_state brw_vs_prog;
 const struct brw_tracked_state brw_vs_unit;
 const struct brw_tracked_state brw_wm_input_sizes;
 const struct brw_tracked_state brw_wm_prog;
 const struct brw_tracked_state brw_wm_samplers;
+const struct brw_tracked_state brw_wm_constant_surface;
 const struct brw_tracked_state brw_wm_surfaces;
 const struct brw_tracked_state brw_wm_unit;
 
@@ -90,6 +92,21 @@ const struct brw_tracked_state brw_clear_batch_cache;
 const struct brw_tracked_state brw_drawing_rect;
 const struct brw_tracked_state brw_indices;
 const struct brw_tracked_state brw_vertices;
+const struct brw_tracked_state brw_index_buffer;
+
+/**
+ * Use same key for WM and VS surfaces.
+ */
+struct brw_surface_key {
+   GLenum target, depthmode;
+   dri_bo *bo;
+   GLint format, internal_format;
+   GLint first_level, last_level;
+   GLint width, height, depth;
+   GLint pitch, cpp;
+   uint32_t tiling;
+   GLuint offset;
+};
 
 /***********************************************************************
  * brw_state.c
@@ -135,8 +152,8 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
 			  void *aux_return);
 void brw_state_cache_check_size( struct brw_context *brw );
 
-void brw_init_cache( struct brw_context *brw );
-void brw_destroy_cache( struct brw_context *brw );
+void brw_init_caches( struct brw_context *brw );
+void brw_destroy_caches( struct brw_context *brw );
 
 /***********************************************************************
  * brw_state_batch.c
@@ -150,4 +167,9 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 void brw_destroy_batch_cache( struct brw_context *brw );
 void brw_clear_batch_cache_flush( struct brw_context *brw );
 
+/* brw_wm_surface_state.c */
+dri_bo *
+brw_create_constant_surface( struct brw_context *brw,
+                             struct brw_surface_key *key );
+
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index d5b5166406..e40d7a0416 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -56,9 +56,9 @@
  * incorrect program is run for the other instance.
  */
 
+#include "main/imports.h"
 #include "brw_state.h"
 #include "intel_batchbuffer.h"
-#include "main/imports.h"
 
 /* XXX: Fixme - have to include these to get the sizes of the prog_key
  * structs:
@@ -69,8 +69,10 @@
 #include "brw_sf.h"
 #include "brw_gs.h"
 
-static GLuint hash_key( const void *key, GLuint key_size,
-			dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
+
+static GLuint
+hash_key(const void *key, GLuint key_size,
+         dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
 {
    GLuint *ikey = (GLuint *)key;
    GLuint hash = 0, i;
@@ -95,6 +97,7 @@ static GLuint hash_key( const void *key, GLuint key_size,
    return hash;
 }
 
+
 /**
  * Marks a new buffer as being chosen for the given cache id.
  */
@@ -111,6 +114,7 @@ update_cache_last(struct brw_cache *cache, enum brw_cache_id cache_id,
    cache->brw->state.dirty.cache |= 1 << cache_id;
 }
 
+
 static struct brw_cache_item *
 search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 	     GLuint hash, const void *key, GLuint key_size,
@@ -143,7 +147,8 @@ search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 }
 
 
-static void rehash( struct brw_cache *cache )
+static void
+rehash(struct brw_cache *cache)
 {
    struct brw_cache_item **items;
    struct brw_cache_item *c, *next;
@@ -164,15 +169,17 @@ static void rehash( struct brw_cache *cache )
    cache->size = size;
 }
 
+
 /**
  * Returns the buffer object matching cache_id and key, or NULL.
  */
-dri_bo *brw_search_cache( struct brw_cache *cache,
-			  enum brw_cache_id cache_id,
-			  const void *key,
-			  GLuint key_size,
-			  dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
-			  void *aux_return )
+dri_bo *
+brw_search_cache(struct brw_cache *cache,
+                 enum brw_cache_id cache_id,
+                 const void *key,
+                 GLuint key_size,
+                 dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
+                 void *aux_return)
 {
    struct brw_cache_item *item;
    GLuint hash = hash_key(key, key_size, reloc_bufs, nr_reloc_bufs);
@@ -192,6 +199,7 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
    return item->bo;
 }
 
+
 dri_bo *
 brw_upload_cache( struct brw_cache *cache,
 		  enum brw_cache_id cache_id,
@@ -265,7 +273,9 @@ brw_upload_cache( struct brw_cache *cache,
    return bo;
 }
 
-/* This doesn't really work with aux data.  Use search/upload instead
+
+/**
+ * This doesn't really work with aux data.  Use search/upload instead
  */
 dri_bo *
 brw_cache_data_sz(struct brw_cache *cache,
@@ -296,6 +306,7 @@ brw_cache_data_sz(struct brw_cache *cache,
    return bo;
 }
 
+
 /**
  * Wrapper around brw_cache_data_sz using the cache_id's canonical key size.
  *
@@ -319,21 +330,22 @@ enum pool_type {
    DW_GENERAL_STATE
 };
 
+
 static void
-brw_init_cache_id( struct brw_context *brw,
-		const char *name,
-		enum brw_cache_id id,
-		GLuint key_size,
-		GLuint aux_size)
+brw_init_cache_id(struct brw_cache *cache,
+                  const char *name,
+                  enum brw_cache_id id,
+                  GLuint key_size,
+                  GLuint aux_size)
 {
-   struct brw_cache *cache = &brw->cache;
-
    cache->name[id] = strdup(name);
    cache->key_size[id] = key_size;
    cache->aux_size[id] = aux_size;
 }
 
-void brw_init_cache( struct brw_context *brw )
+
+static void
+brw_init_non_surface_cache(struct brw_context *brw)
 {
    struct brw_cache *cache = &brw->cache;
 
@@ -342,114 +354,136 @@ void brw_init_cache( struct brw_context *brw )
    cache->size = 7;
    cache->n_items = 0;
    cache->items = (struct brw_cache_item **)
-      _mesa_calloc(cache->size * 
-		   sizeof(struct brw_cache_item));
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_VP",
 		     BRW_CC_VP,
 		     sizeof(struct brw_cc_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_UNIT",
 		     BRW_CC_UNIT,
 		     sizeof(struct brw_cc_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_PROG",
 		     BRW_WM_PROG,
 		     sizeof(struct brw_wm_prog_key),
 		     sizeof(struct brw_wm_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER_DEFAULT_COLOR",
 		     BRW_SAMPLER_DEFAULT_COLOR,
 		     sizeof(struct brw_sampler_default_color),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER",
 		     BRW_SAMPLER,
 		     0,		/* variable key/data size */
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_UNIT",
 		     BRW_WM_UNIT,
 		     sizeof(struct brw_wm_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_PROG",
 		     BRW_SF_PROG,
 		     sizeof(struct brw_sf_prog_key),
 		     sizeof(struct brw_sf_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_VP",
 		     BRW_SF_VP,
 		     sizeof(struct brw_sf_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_UNIT",
 		     BRW_SF_UNIT,
 		     sizeof(struct brw_sf_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_UNIT",
 		     BRW_VS_UNIT,
 		     sizeof(struct brw_vs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_PROG",
 		     BRW_VS_PROG,
 		     sizeof(struct brw_vs_prog_key),
 		     sizeof(struct brw_vs_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_UNIT",
 		     BRW_CLIP_UNIT,
 		     sizeof(struct brw_clip_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_PROG",
 		     BRW_CLIP_PROG,
 		     sizeof(struct brw_clip_prog_key),
 		     sizeof(struct brw_clip_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_UNIT",
 		     BRW_GS_UNIT,
 		     sizeof(struct brw_gs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_PROG",
 		     BRW_GS_PROG,
 		     sizeof(struct brw_gs_prog_key),
 		     sizeof(struct brw_gs_prog_data));
+}
+
+
+static void
+brw_init_surface_cache(struct brw_context *brw)
+{
+   struct brw_cache *cache = &brw->surface_cache;
+
+   cache->brw = brw;
 
-   brw_init_cache_id(brw,
+   cache->size = 7;
+   cache->n_items = 0;
+   cache->items = (struct brw_cache_item **)
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
+
+   brw_init_cache_id(cache,
 		     "SS_SURFACE",
 		     BRW_SS_SURFACE,
 		     sizeof(struct brw_surface_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SS_SURF_BIND",
 		     BRW_SS_SURF_BIND,
 		     0,
 		     0);
 }
 
+
+void
+brw_init_caches(struct brw_context *brw)
+{
+   brw_init_non_surface_cache(brw);
+   brw_init_surface_cache(brw);
+}
+
+
 static void
-brw_clear_cache( struct brw_context *brw )
+brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    struct brw_cache_item *c, *next;
    GLuint i;
@@ -457,8 +491,8 @@ brw_clear_cache( struct brw_context *brw )
    if (INTEL_DEBUG & DEBUG_STATE)
       _mesa_printf("%s\n", __FUNCTION__);
 
-   for (i = 0; i < brw->cache.size; i++) {
-      for (c = brw->cache.items[i]; c; c = next) {
+   for (i = 0; i < cache->size; i++) {
+      for (c = cache->items[i]; c; c = next) {
 	 int j;
 
 	 next = c->next;
@@ -468,10 +502,10 @@ brw_clear_cache( struct brw_context *brw )
 	 free((void *)c->key);
 	 free(c);
       }
-      brw->cache.items[i] = NULL;
+      cache->items[i] = NULL;
    }
 
-   brw->cache.n_items = 0;
+   cache->n_items = 0;
 
    if (brw->curbe.last_buf) {
       _mesa_free(brw->curbe.last_buf);
@@ -483,25 +517,46 @@ brw_clear_cache( struct brw_context *brw )
    brw->state.dirty.cache |= ~0;
 }
 
-void brw_state_cache_check_size( struct brw_context *brw )
+
+void
+brw_state_cache_check_size(struct brw_context *brw)
 {
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+
    /* un-tuned guess.  We've got around 20 state objects for a total of around
     * 32k, so 1000 of them is around 1.5MB.
     */
    if (brw->cache.n_items > 1000)
-      brw_clear_cache(brw);
+      brw_clear_cache(brw, &brw->cache);
+
+   if (brw->surface_cache.n_items > 1000)
+      brw_clear_cache(brw, &brw->surface_cache);
 }
 
-void brw_destroy_cache( struct brw_context *brw )
+
+static void
+brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   brw_clear_cache(brw);
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s\n", __FUNCTION__);
+
+   brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
-      dri_bo_unreference(brw->cache.last_bo[i]);
-      free(brw->cache.name[i]);
+      dri_bo_unreference(cache->last_bo[i]);
+      free(cache->name[i]);
    }
-   free(brw->cache.items);
-   brw->cache.items = NULL;
-   brw->cache.size = 0;
+   free(cache->items);
+   cache->items = NULL;
+   cache->size = 0;
+}
+
+
+void
+brw_destroy_caches(struct brw_context *brw)
+{
+   brw_destroy_cache(brw, &brw->cache);
+   brw_destroy_cache(brw, &brw->surface_cache);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_state_dump.c b/src/mesa/drivers/dri/i965/brw_state_dump.c
index a713262269..e94fa7d2b4 100644
--- a/src/mesa/drivers/dri/i965/brw_state_dump.c
+++ b/src/mesa/drivers/dri/i965/brw_state_dump.c
@@ -126,6 +126,8 @@ static void dump_wm_surface_state(struct brw_context *brw)
 		surf->ss3.pitch + 1, surf->ss3.tiled_surface ? "" : "not ");
       state_out(name, surf, surfoff, 4, "mip base %d\n",
 		surf->ss4.min_lod);
+      state_out(name, surf, surfoff, 5, "x,y offset: %d,%d\n",
+		surf->ss5.x_offset, surf->ss5.y_offset);
 
       dri_bo_unmap(surf_bo);
    }
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 5de1450e61..414620d0b3 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -59,11 +59,12 @@ const struct brw_tracked_state *atoms[] =
    &brw_curbe_offsets,
    &brw_recalculate_urb_fence,
 
-
    &brw_cc_vp,
    &brw_cc_unit,
 
-   &brw_wm_surfaces,		/* must do before samplers */
+   &brw_vs_surfaces,		/* must do before unit */
+   &brw_wm_constant_surface,	/* must do before wm surfaces/bind bo */
+   &brw_wm_surfaces,		/* must do before samplers and unit */
    &brw_wm_samplers,
 
    &brw_wm_unit,
@@ -88,54 +89,27 @@ const struct brw_tracked_state *atoms[] =
 
    &brw_line_stipple,
    &brw_aa_line_parameters,
-   /* Ordering of the commands below is documented as fixed.  
-    */
-#if 0
-   &brw_pipelined_state_pointers,
-   &brw_urb_fence,
-   &brw_constant_buffer_state,
-#else
+
    &brw_psp_urb_cbs,
-#endif
 
    &brw_drawing_rect,
    &brw_indices,
+   &brw_index_buffer,
    &brw_vertices,
 
-   NULL,			/* brw_constant_buffer */
+   &brw_constant_buffer
 };
 
 
 void brw_init_state( struct brw_context *brw )
 {
-   GLuint i;
-
-   brw_init_cache(brw);
-
-   brw->state.atoms = _mesa_malloc(sizeof(atoms));
-   brw->state.nr_atoms = sizeof(atoms)/sizeof(*atoms);
-   _mesa_memcpy(brw->state.atoms, atoms, sizeof(atoms));
-
-   /* Patch in a pointer to the dynamic state atom:
-    */
-   for (i = 0; i < brw->state.nr_atoms; i++)
-      if (brw->state.atoms[i] == NULL)
-	 brw->state.atoms[i] = &brw->curbe.tracked_state;
-
-   _mesa_memcpy(&brw->curbe.tracked_state, 
-		&brw_constant_buffer,
-		sizeof(brw_constant_buffer));
+   brw_init_caches(brw);
 }
 
 
 void brw_destroy_state( struct brw_context *brw )
 {
-   if (brw->state.atoms) {
-      _mesa_free(brw->state.atoms);
-      brw->state.atoms = NULL;
-   }
-
-   brw_destroy_cache(brw);
+   brw_destroy_caches(brw);
    brw_destroy_batch_cache(brw);
 }
 
@@ -218,6 +192,7 @@ static struct dirty_bit_map mesa_bits[] = {
    DEFINE_BIT(_NEW_MULTISAMPLE),
    DEFINE_BIT(_NEW_TRACK_MATRIX),
    DEFINE_BIT(_NEW_PROGRAM),
+   DEFINE_BIT(_NEW_PROGRAM_CONSTANTS),
    {0, 0, 0}
 };
 
@@ -231,10 +206,10 @@ static struct dirty_bit_map brw_bits[] = {
    DEFINE_BIT(BRW_NEW_PRIMITIVE),
    DEFINE_BIT(BRW_NEW_CONTEXT),
    DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS),
-   DEFINE_BIT(BRW_NEW_INPUT_VARYING),
    DEFINE_BIT(BRW_NEW_PSP),
    DEFINE_BIT(BRW_NEW_FENCE),
    DEFINE_BIT(BRW_NEW_INDICES),
+   DEFINE_BIT(BRW_NEW_INDEX_BUFFER),
    DEFINE_BIT(BRW_NEW_VERTICES),
    DEFINE_BIT(BRW_NEW_BATCH),
    DEFINE_BIT(BRW_NEW_DEPTH_BUFFER),
@@ -336,7 +311,7 @@ void brw_validate_state( struct brw_context *brw )
 
    /* do prepare stage for all atoms */
    for (i = 0; i < Elements(atoms); i++) {
-      const struct brw_tracked_state *atom = brw->state.atoms[i];
+      const struct brw_tracked_state *atom = atoms[i];
 
       if (brw->intel.Fallback)
          break;
@@ -347,6 +322,19 @@ void brw_validate_state( struct brw_context *brw )
         }
       }
    }
+
+   /* Make sure that the textures which are referenced by the current
+    * brw fragment program are actually present/valid.
+    * If this fails, we can experience GPU lock-ups.
+    */
+   {
+      const struct brw_fragment_program *fp;
+      fp = brw_fragment_program_const(brw->fragment_program);
+      if (fp) {
+         assert((fp->tex_units_used & ctx->Texture._EnabledUnits)
+                == fp->tex_units_used);
+      }
+   }
 }
 
 
@@ -367,8 +355,8 @@ void brw_upload_state(struct brw_context *brw)
       _mesa_memset(&examined, 0, sizeof(examined));
       prev = *state;
 
-      for (i = 0; i < brw->state.nr_atoms; i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+      for (i = 0; i < Elements(atoms); i++) {	 
+	 const struct brw_tracked_state *atom = atoms[i];
 	 struct brw_state_flags generated;
 
 	 assert(atom->dirty.mesa ||
@@ -397,7 +385,7 @@ void brw_upload_state(struct brw_context *brw)
    }
    else {
       for (i = 0; i < Elements(atoms); i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+	 const struct brw_tracked_state *atom = atoms[i];
 
 	 if (brw->intel.Fallback)
 	    break;
diff --git a/src/mesa/drivers/dri/i965/brw_structs.h b/src/mesa/drivers/dri/i965/brw_structs.h
index 6011a4aa0a..66d4127271 100644
--- a/src/mesa/drivers/dri/i965/brw_structs.h
+++ b/src/mesa/drivers/dri/i965/brw_structs.h
@@ -33,6 +33,14 @@
 #ifndef BRW_STRUCTS_H
 #define BRW_STRUCTS_H
 
+
+/** Number of general purpose registers (VS, WM, etc) */
+#define BRW_MAX_GRF 128
+
+/** Number of message register file registers */
+#define BRW_MAX_MRF 16
+
+
 /* Command packets:
  */
 struct header 
@@ -434,8 +442,8 @@ struct brw_urb_fence
    {
       GLuint sf_fence:10;  
       GLuint vf_fence:10;  
-      GLuint cs_fence:10;  
-      GLuint pad:2;
+      GLuint cs_fence:11;  
+      GLuint pad:1;
    } bits1;
 };
 
@@ -815,7 +823,9 @@ struct brw_gs_unit_state
 
    struct
    {
-      GLuint pad0:10;
+      GLuint pad0:8;
+      GLuint rendering_enable:1; /* for IGDNG */
+      GLuint pad4:1;
       GLuint stats_enable:1; 
       GLuint nr_urb_entries:7; 
       GLuint pad1:1;
@@ -923,6 +933,28 @@ struct brw_wm_unit_state
    
    GLfloat global_depth_offset_constant;  
    GLfloat global_depth_offset_scale;   
+   
+   /* for IGDNG only */
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_1:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_1:26;
+   } wm8;       
+
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_2:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_2:26;
+   } wm9;       
+
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_3:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_3:26;
+   } wm10;       
 };
 
 struct brw_sampler_default_color {
@@ -1075,7 +1107,7 @@ struct brw_surface_state
       GLuint y_offset:4;
       GLuint pad0:1;
       GLuint x_offset:7;
-   } ss5;   /* NEW in Integrated Graphics Device */
+   } ss5;   /* New in G4X */
 
 };
 
@@ -1196,7 +1228,9 @@ struct brw_instruction
 	 GLuint dest_reg_type:3;
 	 GLuint src0_reg_file:2;
 	 GLuint src0_reg_type:3;
-	 GLuint pad:6;
+	 GLuint src1_reg_file:2;        /* 0x00000c00 */
+	 GLuint src1_reg_type:3;        /* 0x00007000 */
+	 GLuint pad:1;
 	 GLint dest_indirect_offset:10;	/* offset against the deref'd address reg */
 	 GLuint dest_subreg_nr:3; /* subnr for the address reg a0.x */
 	 GLuint dest_horiz_stride:2;
@@ -1211,7 +1245,7 @@ struct brw_instruction
 	 GLuint src0_reg_type:3;
 	 GLuint src1_reg_file:2;
 	 GLuint src1_reg_type:3;
-	 GLuint pad0:1;
+	 GLuint pad:1;
 	 GLuint dest_writemask:4;
 	 GLuint dest_subreg_nr:1;
 	 GLuint dest_reg_nr:8;
@@ -1298,6 +1332,14 @@ struct brw_instruction
 	 GLuint pad1:6;
       } ia16;
 
+       struct 
+       {
+           GLuint pad:26;
+           GLuint end_of_thread:1;
+           GLuint pad1:1;
+           GLuint sfid:4;
+       } send_igdng;  /* for IGDNG only */
+
    } bits2;
 
    union
@@ -1308,7 +1350,7 @@ struct brw_instruction
 	 GLuint src1_reg_nr:8;
 	 GLuint src1_abs:1;
 	 GLuint src1_negate:1;
-	 GLuint pad:1;
+	 GLuint src1_address_mode:1;
 	 GLuint src1_horiz_stride:2;
 	 GLuint src1_width:3;
 	 GLuint src1_vert_stride:4;
@@ -1323,7 +1365,7 @@ struct brw_instruction
 	 GLuint src1_reg_nr:8;
 	 GLuint src1_abs:1;
 	 GLuint src1_negate:1;
-	 GLuint pad0:1;
+	 GLuint src1_address_mode:1;
 	 GLuint src1_swz_z:2;
 	 GLuint src1_swz_w:2;
 	 GLuint pad1:1;
@@ -1337,7 +1379,7 @@ struct brw_instruction
 	 GLuint src1_subreg_nr:3;
 	 GLuint src1_abs:1;
 	 GLuint src1_negate:1;
-	 GLuint pad0:1;
+	 GLuint src1_address_mode:1;
 	 GLuint src1_horiz_stride:2;
 	 GLuint src1_width:3;
 	 GLuint src1_vert_stride:4;
@@ -1385,6 +1427,21 @@ struct brw_instruction
       } math;
 
       struct {
+	 GLuint function:4;
+	 GLuint int_type:1;
+	 GLuint precision:1;
+	 GLuint saturate:1;
+	 GLuint data_type:1;
+	 GLuint snapshot:1;
+	 GLuint pad0:10;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } math_igdng;
+
+      struct {
 	 GLuint binding_table_index:8;
 	 GLuint sampler:4;
 	 GLuint return_format:2; 
@@ -1407,9 +1464,38 @@ struct brw_instruction
          GLuint end_of_thread:1;
       } sampler_g4x;
 
+      struct {
+	 GLuint binding_table_index:8;
+	 GLuint sampler:4;
+	 GLuint msg_type:4;
+	 GLuint simd_mode:2;
+	 GLuint pad0:1;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } sampler_igdng;
+
       struct brw_urb_immediate urb;
 
       struct {
+	 GLuint opcode:4;
+	 GLuint offset:6;
+	 GLuint swizzle_control:2; 
+	 GLuint pad:1;
+	 GLuint allocate:1;
+	 GLuint used:1;
+	 GLuint complete:1;
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } urb_igdng;
+
+      struct {
 	 GLuint binding_table_index:8;
 	 GLuint msg_control:4;  
 	 GLuint msg_type:2;  
@@ -1423,6 +1509,19 @@ struct brw_instruction
 
       struct {
 	 GLuint binding_table_index:8;
+	 GLuint msg_control:3;  
+	 GLuint msg_type:3;  
+	 GLuint target_cache:2;    
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } dp_read_igdng;
+
+      struct {
+	 GLuint binding_table_index:8;
 	 GLuint msg_control:3;
 	 GLuint pixel_scoreboard_clear:1;
 	 GLuint msg_type:3;    
@@ -1435,6 +1534,20 @@ struct brw_instruction
       } dp_write;
 
       struct {
+	 GLuint binding_table_index:8;
+	 GLuint msg_control:3;
+	 GLuint pixel_scoreboard_clear:1;
+	 GLuint msg_type:3;    
+	 GLuint send_commit_msg:1;
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } dp_write_igdng;
+
+      struct {
 	 GLuint pad:16;
 	 GLuint response_length:4;
 	 GLuint msg_length:4;
@@ -1443,8 +1556,18 @@ struct brw_instruction
 	 GLuint end_of_thread:1;
       } generic;
 
+      struct {
+	 GLuint pad:19;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } generic_igdng;
+
       GLint d;
       GLuint ud;
+      float f;
    } bits3;
 };
 
diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c
index 3ab27c26ef..5986cbffad 100644
--- a/src/mesa/drivers/dri/i965/brw_tex_layout.c
+++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c
@@ -28,7 +28,6 @@
   * Authors:
   *   Keith Whitwell <keith@tungstengraphics.com>
   */
-        
 
 /* Code to layout images in a mipmap tree for i965.
  */
@@ -37,17 +36,93 @@
 #include "intel_tex_layout.h"
 #include "intel_context.h"
 #include "main/macros.h"
+#include "intel_chipset.h"
 
 #define FILE_DEBUG_FLAG DEBUG_MIPTREE
 
-GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_tree *mt )
+GLboolean brw_miptree_layout(struct intel_context *intel,
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling)
 {
-   /* XXX: these vary depending on image format: 
-    */
-/*    GLint align_w = 4; */
+   /* XXX: these vary depending on image format: */
+   /* GLint align_w = 4; */
 
    switch (mt->target) {
-   case GL_TEXTURE_CUBE_MAP: 
+   case GL_TEXTURE_CUBE_MAP:
+      if (IS_IGDNG(intel->intelScreen->deviceID)) {
+          GLuint align_h = 2, align_w = 4;
+          GLuint level;
+          GLuint x = 0;
+          GLuint y = 0;
+          GLuint width = mt->width0;
+          GLuint height = mt->height0;
+          GLuint qpitch = 0;
+          GLuint y_pitch = 0;
+
+          mt->pitch = mt->width0;
+          intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
+          y_pitch = ALIGN(height, align_h);
+
+          if (mt->compressed) {
+              mt->pitch = ALIGN(mt->width0, align_w);
+          }
+
+          if (mt->first_level != mt->last_level) {
+              GLuint mip1_width;
+
+              if (mt->compressed) {
+                  mip1_width = ALIGN(minify(mt->width0), align_w)
+                      + ALIGN(minify(minify(mt->width0)), align_w);
+              } else {
+                  mip1_width = ALIGN(minify(mt->width0), align_w)
+                      + minify(minify(mt->width0));
+              }
+
+              if (mip1_width > mt->pitch) {
+                  mt->pitch = mip1_width;
+              }
+          }
+
+          mt->pitch = intel_miptree_pitch_align(intel, mt, tiling, mt->pitch);
+
+          if (mt->compressed) {
+              qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * mt->pitch * mt->cpp;
+              mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * 6;
+          } else {
+              qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * mt->pitch * mt->cpp;
+              mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * 6;
+          }
+
+          for (level = mt->first_level; level <= mt->last_level; level++) {
+              GLuint img_height;
+              GLuint nr_images = 6;
+              GLuint q = 0;
+
+              intel_miptree_set_level_info(mt, level, nr_images, x, y, width, 
+                                           height, 1);
+
+              for (q = 0; q < nr_images; q++)
+                  intel_miptree_set_image_offset_ex(mt, level, q, x, y, q * qpitch);
+
+              if (mt->compressed)
+                  img_height = MAX2(1, height/4);
+              else
+                  img_height = ALIGN(height, align_h);
+
+              if (level == mt->first_level + 1) {
+                  x += ALIGN(width, align_w);
+              }
+              else {
+                  y += img_height;
+              }
+
+              width  = minify(width);
+              height = minify(height);
+          }
+
+          break;
+      }
+
    case GL_TEXTURE_3D: {
       GLuint width  = mt->width0;
       GLuint height = mt->height0;
@@ -59,25 +134,25 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
       GLuint align_w = 4;
 
       mt->total_height = 0;
-      
+      intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
+
       if (mt->compressed) {
-          align_w = intel_compressed_alignment(mt->internal_format);
           mt->pitch = ALIGN(width, align_w);
           pack_y_pitch = (height + 3) / 4;
       } else {
-          mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
-          pack_y_pitch = ALIGN(mt->height0, align_h);
+	 mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+	 pack_y_pitch = ALIGN(mt->height0, align_h);
       }
 
-      pack_x_pitch = mt->pitch;
+      pack_x_pitch = width;
       pack_x_nr = 1;
 
-      for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
+      for (level = mt->first_level ; level <= mt->last_level ; level++) {
 	 GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6;
 	 GLint x = 0;
 	 GLint y = 0;
 	 GLint q, j;
-	    
+
 	 intel_miptree_set_level_info(mt, level, nr_images,
 				      0, mt->total_height,
 				      width, height, depth);
@@ -89,7 +164,7 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
 	    }
 
 	    x = 0;
-	    y += pack_y_pitch;	    
+	    y += pack_y_pitch;
 	 }
 
 
@@ -98,25 +173,25 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
 	 height = minify(height);
 	 depth  = minify(depth);
 
-    if (mt->compressed) {
-        pack_y_pitch = (height + 3) / 4;
-        
-        if (pack_x_pitch > ALIGN(width, align_w)) {
-            pack_x_pitch = ALIGN(width, align_w);
-            pack_x_nr <<= 1;
-        }
-    } else {
-        if (pack_x_pitch > 4) {
-            pack_x_pitch >>= 1;
-            pack_x_nr <<= 1;
-            assert(pack_x_pitch * pack_x_nr <= mt->pitch);
-        }
-
-        if (pack_y_pitch > 2) {
-            pack_y_pitch >>= 1;
-            pack_y_pitch = ALIGN(pack_y_pitch, align_h);
-        }
-    }
+	 if (mt->compressed) {
+	    pack_y_pitch = (height + 3) / 4;
+
+	    if (pack_x_pitch > ALIGN(width, align_w)) {
+	       pack_x_pitch = ALIGN(width, align_w);
+	       pack_x_nr <<= 1;
+	    }
+	 } else {
+	    if (pack_x_pitch > 4) {
+	       pack_x_pitch >>= 1;
+	       pack_x_nr <<= 1;
+	       assert(pack_x_pitch * pack_x_nr <= mt->pitch);
+	    }
+
+	    if (pack_y_pitch > 2) {
+	       pack_y_pitch >>= 1;
+	       pack_y_pitch = ALIGN(pack_y_pitch, align_h);
+	    }
+	 }
 
       }
       /* The 965's sampler lays cachelines out according to how accesses
@@ -133,15 +208,15 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
    }
 
    default:
-      i945_miptree_layout_2d(intel, mt);
+      i945_miptree_layout_2d(intel, mt, tiling);
       break;
    }
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, 
-		mt->pitch, 
+   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
+		mt->pitch,
 		mt->total_height,
 		mt->cpp,
 		mt->pitch * mt->total_height * mt->cpp );
-		
+
    return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c
index 7673dd36eb..8c6f4355a6 100644
--- a/src/mesa/drivers/dri/i965/brw_urb.c
+++ b/src/mesa/drivers/dri/i965/brw_urb.c
@@ -143,7 +143,29 @@ static void recalculate_urb_fence( struct brw_context *brw )
       brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries;
       brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;	
       brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries;	
-      
+
+      brw->urb.constrained = 0;
+
+      if (BRW_IS_IGDNG(brw)) {
+         brw->urb.nr_vs_entries = 128;
+         brw->urb.nr_sf_entries = 48;
+         if (check_urb_layout(brw)) {
+            goto done;
+         } else {
+            brw->urb.constrained = 1;
+            brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
+            brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;
+         }
+      } else if (BRW_IS_G4X(brw)) {
+	 brw->urb.nr_vs_entries = 64;
+	 if (check_urb_layout(brw)) {
+	    goto done;
+	 } else {
+	    brw->urb.constrained = 1;
+	    brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
+	 }
+      }
+
       if (!check_urb_layout(brw)) {
 	 brw->urb.nr_vs_entries = limits[VS].min_nr_entries;	
 	 brw->urb.nr_gs_entries = limits[GS].min_nr_entries;	
@@ -169,9 +191,8 @@ static void recalculate_urb_fence( struct brw_context *brw )
 	 if (INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
 	    _mesa_printf("URB CONSTRAINED\n");
       }
-      else 
-	 brw->urb.constrained = 0;
 
+done:
       if (INTEL_DEBUG & DEBUG_URB)
 	 _mesa_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
 		      brw->urb.vs_start,
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index 1e4f66091e..4a591365c9 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -58,6 +58,7 @@ struct brw_vs_compile {
 
    GLuint first_output;
    GLuint nr_outputs;
+   GLuint first_overflow_output; /**< VERT_ATTRIB_x */
 
    GLuint first_tmp;
    GLuint last_tmp;
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 72cd36a603..108e19cdbc 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -70,13 +70,18 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    GLuint i, reg = 0, mrf;
    int attributes_in_vue;
 
-#if 0
-   if (c->vp->program.Base.Parameters->NumParameters >= 6)
-      c->vp->use_const_buffer = 1;
+   /* Determine whether to use a real constant buffer or use a block
+    * of GRF registers for constants.  The later is faster but only
+    * works if everything fits in the GRF.
+    * XXX this heuristic/check may need some fine tuning...
+    */
+   if (c->vp->program.Base.Parameters->NumParameters +
+       c->vp->program.Base.NumTemporaries + 20 > BRW_MAX_GRF)
+      c->vp->use_const_buffer = GL_TRUE;
    else
-#endif
       c->vp->use_const_buffer = GL_FALSE;
-   /*printf("use_const_buffer = %d\n", c->use_const_buffer);*/
+
+   /*printf("use_const_buffer = %d\n", c->vp->use_const_buffer);*/
 
    /* r0 -- reserved as usual
     */
@@ -130,15 +135,21 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    if (c->nr_inputs == 0)
       reg++;
 
-   /* Allocate outputs: TODO: could organize the non-position outputs
-    * to go straight into message regs.
+   /* Allocate outputs.  The non-position outputs go straight into message regs.
     */
    c->nr_outputs = 0;
    c->first_output = reg;
-   mrf = 4;
+   c->first_overflow_output = 0;
+
+   if (BRW_IS_IGDNG(c->func.brw))
+       mrf = 8;
+   else
+       mrf = 4;
+
    for (i = 0; i < VERT_RESULT_MAX; i++) {
       if (c->prog_data.outputs_written & (1 << i)) {
 	 c->nr_outputs++;
+         assert(i < Elements(c->regs[PROGRAM_OUTPUT]));
 	 if (i == VERT_RESULT_HPOS) {
 	    c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
 	    reg++;
@@ -149,8 +160,17 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 	    mrf++;		/* just a placeholder?  XXX fix later stages & remove this */
 	 }
 	 else {
-	    c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
-	    mrf++;
+            if (mrf < 16) {
+               c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
+               mrf++;
+            }
+            else {
+               /* too many vertex results to fit in MRF, use GRF for overflow */
+               if (!c->first_overflow_output)
+                  c->first_overflow_output = i;
+               c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
+               reg++;
+            }
 	 }
       }
    }     
@@ -218,7 +238,10 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
     */
    attributes_in_vue = MAX2(c->nr_outputs, c->nr_inputs);
 
-   c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4;
+   if (BRW_IS_IGDNG(c->func.brw))
+       c->prog_data.urb_entry_size = (attributes_in_vue + 6 + 3) / 4;
+   else
+       c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4;
 
    c->prog_data.total_grf = reg;
 
@@ -727,10 +750,11 @@ get_constant(struct brw_vs_compile *c,
    struct brw_compile *p = &c->func;
    struct brw_reg const_reg;
    struct brw_reg const2_reg;
+   const GLboolean relAddr = src->RelAddr;
 
    assert(argIndex < 3);
 
-   if (c->current_const[argIndex].index != src->Index || src->RelAddr) {
+   if (c->current_const[argIndex].index != src->Index || relAddr) {
       struct brw_reg addrReg = c->regs[PROGRAM_ADDRESS][0];
 
       c->current_const[argIndex].index = src->Index;
@@ -743,13 +767,13 @@ get_constant(struct brw_vs_compile *c,
       brw_dp_READ_4_vs(p,
                        c->current_const[argIndex].reg,/* writeback dest */
                        0,                             /* oword */
-                       src->RelAddr,                  /* relative indexing? */
+                       relAddr,                       /* relative indexing? */
                        addrReg,                       /* address register */
                        16 * src->Index,               /* byte offset */
                        SURF_INDEX_VERT_CONST_BUFFER   /* binding table index */
                        );
 
-      if (src->RelAddr) {
+      if (relAddr) {
          /* second read */
          const2_reg = get_tmp(c);
 
@@ -760,7 +784,7 @@ get_constant(struct brw_vs_compile *c,
          brw_dp_READ_4_vs(p,
                           const2_reg,              /* writeback dest */
                           1,                       /* oword */
-                          src->RelAddr,            /* relative indexing? */
+                          relAddr,                 /* relative indexing? */
                           addrReg,                 /* address register */
                           16 * src->Index,         /* byte offset */
                           SURF_INDEX_VERT_CONST_BUFFER
@@ -770,7 +794,7 @@ get_constant(struct brw_vs_compile *c,
 
    const_reg = c->current_const[argIndex].reg;
 
-   if (src->RelAddr) {
+   if (relAddr) {
       /* merge the two Owords into the constant register */
       /* const_reg[7..4] = const2_reg[7..4] */
       brw_MOV(p,
@@ -887,6 +911,7 @@ get_src_reg( struct brw_vs_compile *c,
    case PROGRAM_STATE_VAR:
    case PROGRAM_CONSTANT:
    case PROGRAM_UNIFORM:
+   case PROGRAM_ENV_PARAM:
       if (c->vp->use_const_buffer) {
          return get_constant(c, inst, argIndex);
       }
@@ -906,7 +931,6 @@ get_src_reg( struct brw_vs_compile *c,
       return brw_null_reg();
 
    case PROGRAM_LOCAL_PARAM: 
-   case PROGRAM_ENV_PARAM: 
    case PROGRAM_WRITE_ONLY:
    default:
       assert(0);
@@ -1079,6 +1103,8 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    struct brw_reg m0 = brw_message_reg(0);
    struct brw_reg pos = c->regs[PROGRAM_OUTPUT][VERT_RESULT_HPOS];
    struct brw_reg ndc;
+   int eot;
+   GLuint len_vertext_header = 2;
 
    if (c->key.copy_edgeflag) {
       brw_MOV(p, 
@@ -1088,14 +1114,16 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 
    /* Build ndc coords */
    ndc = get_tmp(c);
+   /* ndc = 1.0 / pos.w */
    emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL);
+   /* ndc.xyz = pos * ndc */
    brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc);
 
    /* Update the header for point size, user clipping flags, and -ve rhw
     * workaround.
     */
    if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) ||
-       c->key.nr_userclip || !BRW_IS_G4X(p->brw))
+       c->key.nr_userclip || BRW_IS_965(p->brw))
    {
       struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
       GLuint i;
@@ -1126,7 +1154,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
        * Later, clipping will detect ucp[6] and ensure the primitive is
        * clipped against all fixed planes.
        */
-      if (!BRW_IS_G4X(p->brw)) {
+      if (BRW_IS_965(p->brw)) {
 	 brw_CMP(p,
 		 vec8(brw_null_reg()),
 		 BRW_CONDITIONAL_L,
@@ -1153,7 +1181,23 @@ static void emit_vertex_write( struct brw_vs_compile *c)
     */
    brw_set_access_mode(p, BRW_ALIGN_1);
    brw_MOV(p, offset(m0, 2), ndc);
-   brw_MOV(p, offset(m0, 3), pos);
+
+   if (BRW_IS_IGDNG(p->brw)) {
+       /* There are 20 DWs (D0-D19) in VUE vertex header on IGDNG */
+       brw_MOV(p, offset(m0, 3), pos); /* a portion of vertex header */
+       /* m4, m5 contain the distances from vertex to the user clip planeXXX. 
+        * Seems it is useless for us.
+        * m6 is used for aligning, so that the remainder of vertex element is 
+        * reg-aligned.
+        */
+       brw_MOV(p, offset(m0, 7), pos); /* the remainder of vertex element */
+       len_vertext_header = 6;
+   } else {
+       brw_MOV(p, offset(m0, 3), pos);
+       len_vertext_header = 2;
+   }
+
+   eot = (c->first_overflow_output == 0);
 
    brw_urb_WRITE(p, 
 		 brw_null_reg(), /* dest */
@@ -1161,12 +1205,43 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 		 c->r0,		/* src */
 		 0,		/* allocate */
 		 1,		/* used */
-		 c->nr_outputs + 3, /* msg len */
+		 MIN2(c->nr_outputs + 1 + len_vertext_header, (BRW_MAX_MRF-1)), /* msg len */
 		 0,		/* response len */
-		 1, 		/* eot */
+		 eot, 		/* eot */
 		 1, 		/* writes complete */
 		 0, 		/* urb destination offset */
 		 BRW_URB_SWIZZLE_INTERLEAVE);
+
+   if (c->first_overflow_output > 0) {
+      /* Not all of the vertex outputs/results fit into the MRF.
+       * Move the overflowed attributes from the GRF to the MRF and
+       * issue another brw_urb_WRITE().
+       */
+      /* XXX I'm not 100% sure about which MRF regs to use here.  Starting
+       * at mrf[4] atm...
+       */
+      GLuint i, mrf = 0;
+      for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) {
+         if (c->prog_data.outputs_written & (1 << i)) {
+            /* move from GRF to MRF */
+            brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]);
+            mrf++;
+         }
+      }
+
+      brw_urb_WRITE(p,
+                    brw_null_reg(), /* dest */
+                    4,              /* starting mrf reg nr */
+                    c->r0,          /* src */
+                    0,              /* allocate */
+                    1,              /* used */
+                    mrf+1,          /* msg len */
+                    0,              /* response len */
+                    1,              /* eot */
+                    1,              /* writes complete */
+                    BRW_MAX_MRF-1,  /* urb destination offset */
+                    BRW_URB_SWIZZLE_INTERLEAVE);
+   }
 }
 
 
@@ -1187,7 +1262,11 @@ post_vs_emit( struct brw_vs_compile *c,
 
    /* patch up the END code to jump past subroutines, etc */
    offset = last_inst - end_inst;
-   brw_set_src1(end_inst, brw_imm_d(offset * 16));
+   if (offset > 1) {
+      brw_set_src1(end_inst, brw_imm_d(offset * 16));
+   } else {
+      end_inst->header.opcode = BRW_OPCODE_NOP;
+   }
 }
 
 static uint32_t
@@ -1213,18 +1292,19 @@ get_predicate(uint32_t swizzle)
 void brw_vs_emit(struct brw_vs_compile *c )
 {
 #define MAX_IF_DEPTH 32
+#define MAX_LOOP_DEPTH 32
    struct brw_compile *p = &c->func;
    const GLuint nr_insns = c->vp->program.Base.NumInstructions;
-   GLuint insn, if_depth = 0;
+   GLuint insn, if_depth = 0, loop_depth = 0;
    GLuint end_offset = 0;
    struct brw_instruction *end_inst, *last_inst;
-   struct brw_instruction *if_inst[MAX_IF_DEPTH];
+   struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH];
    const struct brw_indirect stack_index = brw_indirect(0, 0);   
    GLuint index;
    GLuint file;
 
    if (INTEL_DEBUG & DEBUG_VS) {
-      _mesa_printf("vs-emit:\n");
+      _mesa_printf("vs-mesa:\n");
       _mesa_print_program(&c->vp->program.Base); 
       _mesa_printf("\n");
    }
@@ -1253,7 +1333,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    for (insn = 0; insn < nr_insns; insn++) {
 
-      struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
+      const struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
       struct brw_reg args[3], dst;
       GLuint i;
       
@@ -1266,7 +1346,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
        */
       if (inst->Opcode != OPCODE_SWZ)
 	  for (i = 0; i < 3; i++) {
-	      struct prog_src_register *src = &inst->SrcReg[i];
+	      const struct prog_src_register *src = &inst->SrcReg[i];
 	      index = src->Index;
 	      file = src->File;	
 	      if (file == PROGRAM_OUTPUT && c->output_regs[index].used_in_src)
@@ -1423,6 +1503,41 @@ void brw_vs_emit(struct brw_vs_compile *c )
          assert(if_depth > 0);
 	 brw_ENDIF(p, if_inst[--if_depth]);
 	 break;			
+#if 0
+      case OPCODE_BGNLOOP:
+         loop_inst[loop_depth++] = brw_DO(p, BRW_EXECUTE_8);
+         break;
+      case OPCODE_BRK:
+         brw_BREAK(p);
+         brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+         break;
+      case OPCODE_CONT:
+         brw_CONT(p);
+         brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+         break;
+      case OPCODE_ENDLOOP: 
+         {
+            struct brw_instruction *inst0, *inst1;
+            loop_depth--;
+            inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]);
+            /* patch all the BREAK/CONT instructions from last BEGINLOOP */
+            while (inst0 > loop_inst[loop_depth]) {
+               inst0--;
+               if (inst0->header.opcode == BRW_OPCODE_BREAK) {
+                  inst0->bits3.if_else.jump_count = inst1 - inst0 + 1;
+                  inst0->bits3.if_else.pop_count = 0;
+               }
+               else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+                  inst0->bits3.if_else.jump_count = inst1 - inst0;
+                  inst0->bits3.if_else.pop_count = 0;
+               }
+            }
+         }
+         break;
+#else
+         (void) loop_inst;
+         (void) loop_depth;
+#endif
       case OPCODE_BRA:
          brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
          brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
@@ -1517,4 +1632,13 @@ void brw_vs_emit(struct brw_vs_compile *c )
    emit_vertex_write(c);
 
    post_vs_emit(c, end_inst, last_inst);
+
+   if (INTEL_DEBUG & DEBUG_VS) {
+      int i;
+
+      _mesa_printf("vs-native:\n");
+      for (i = 0; i < p->nr_insn; i++)
+	 brw_disasm(stderr, &p->store[i]);
+      _mesa_printf("\n");
+   }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index 3d29538843..d790ab6555 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -97,7 +97,11 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
     * brw_urb_WRITE() results.
     */
    vs.thread1.single_program_flow = 0;
-   vs.thread1.binding_table_entry_count = key->nr_surfaces;
+
+   if (BRW_IS_IGDNG(brw))
+      vs.thread1.binding_table_entry_count = 0; /* hardware requirement */
+   else
+      vs.thread1.binding_table_entry_count = key->nr_surfaces;
 
    vs.thread3.urb_entry_read_length = key->urb_entry_read_length;
    vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
@@ -105,10 +109,16 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
    vs.thread3.urb_entry_read_offset = 0;
    vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
 
-   vs.thread4.nr_urb_entries = key->nr_urb_entries;
+   if (BRW_IS_IGDNG(brw))
+       vs.thread4.nr_urb_entries = key->nr_urb_entries >> 2;
+   else
+       vs.thread4.nr_urb_entries = key->nr_urb_entries;
+
    vs.thread4.urb_entry_allocation_size = key->urb_size - 1;
 
-   if (BRW_IS_G4X(brw))
+   if (BRW_IS_IGDNG(brw))
+      chipset_max_threads = 72;
+   else if (BRW_IS_G4X(brw))
       chipset_max_threads = 32;
    else
       chipset_max_threads = 16;
@@ -120,6 +130,8 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
 
    /* No samplers for ARB_vp programs:
     */
+   /* It has to be set to 0 for IGDNG
+    */
    vs.vs5.sampler_count = 0;
 
    if (INTEL_DEBUG & DEBUG_STATS)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
new file mode 100644
index 0000000000..89f47522a1
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
@@ -0,0 +1,226 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "main/mtypes.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "shader/prog_parameter.h"
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+/* Creates a new VS constant buffer reflecting the current VS program's
+ * constants, if needed by the VS program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_vs_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_VERTEX_PROGRAM */
+   if (!vp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
+
+/**
+ * Update the surface state for a VS constant buffer.
+ *
+ * Sets brw->vs.surf_bo[surf] and brw->vp->const_buffer.
+ */
+static void
+brw_update_vs_constant_surface( GLcontext *ctx,
+                                GLuint surf)
+{
+   struct brw_context *brw = brw_context(ctx);
+   struct brw_surface_key key;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+
+   assert(surf == 0);
+
+   /* If we're in this state update atom, we need to update VS constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(vp->const_buffer);
+   vp->const_buffer = brw_vs_update_constant_buffer(brw);
+
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (vp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+      brw->vs.surf_bo[surf] = NULL;
+      return;
+   }
+
+   memset(&key, 0, sizeof(key));
+
+   key.format = MESA_FORMAT_RGBA_FLOAT32;
+   key.internal_format = GL_RGBA;
+   key.bo = vp->const_buffer;
+   key.depthmode = GL_NONE;
+   key.pitch = params->NumParameters;
+   key.width = params->NumParameters;
+   key.height = 1;
+   key.depth = 1;
+   key.cpp = 16;
+
+   /*
+   printf("%s:\n", __FUNCTION__);
+   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
+          key.width, key.height, key.depth, key.cpp, key.pitch);
+   */
+
+   drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+   brw->vs.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
+   if (brw->vs.surf_bo[surf] == NULL) {
+      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
+   }
+}
+
+
+/**
+ * Constructs the binding table for the VS surface state.
+ */
+static dri_bo *
+brw_vs_get_binding_table(struct brw_context *brw)
+{
+   dri_bo *bind_bo;
+
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
+			      NULL, 0,
+			      brw->vs.surf_bo, BRW_VS_MAX_SURF,
+			      NULL);
+
+   if (bind_bo == NULL) {
+      GLuint data_size = BRW_VS_MAX_SURF * sizeof(GLuint);
+      uint32_t *data = malloc(data_size);
+      int i;
+
+      for (i = 0; i < BRW_VS_MAX_SURF; i++)
+         if (brw->vs.surf_bo[i])
+            data[i] = brw->vs.surf_bo[i]->offset;
+         else
+            data[i] = 0;
+
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
+				  NULL, 0,
+				  brw->vs.surf_bo, BRW_VS_MAX_SURF,
+				  data, data_size,
+				  NULL, NULL);
+
+      /* Emit binding table relocations to surface state */
+      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+	 if (brw->vs.surf_bo[i] != NULL) {
+	    /* The presumed offsets were set in the data values for
+	     * brw_upload_cache.
+	     */
+	    drm_intel_bo_emit_reloc(bind_bo, i * 4,
+				    brw->vs.surf_bo[i], 0,
+				    I915_GEM_DOMAIN_INSTRUCTION, 0);
+	 }
+      }
+
+      free(data);
+   }
+
+   return bind_bo;
+}
+
+/**
+ * Vertex shader surfaces (constant buffer).
+ *
+ * This consumes the state updates for the constant buffer needing
+ * to be updated, and produces BRW_NEW_NR_VS_SURFACES for the VS unit and
+ * CACHE_NEW_SURF_BIND for the binding table upload.
+ */
+static void prepare_vs_surfaces(struct brw_context *brw )
+{
+   GLcontext *ctx = &brw->intel.ctx;
+   int i;
+   int nr_surfaces = 0;
+
+   brw_update_vs_constant_surface(ctx, SURF_INDEX_VERT_CONST_BUFFER);
+
+   for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+      if (brw->vs.surf_bo[i] != NULL) {
+	 nr_surfaces = i + 1;
+      }
+   }
+
+   if (brw->vs.nr_surfaces != nr_surfaces) {
+      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
+      brw->vs.nr_surfaces = nr_surfaces;
+   }
+
+   /* Note that we don't end up updating the bind_bo if we don't have a
+    * surface to be pointing at.  This should be relatively harmless, as it
+    * just slightly increases our working set size.
+    */
+   if (brw->vs.nr_surfaces != 0) {
+      dri_bo_unreference(brw->vs.bind_bo);
+      brw->vs.bind_bo = brw_vs_get_binding_table(brw);
+   }
+}
+
+const struct brw_tracked_state brw_vs_surfaces = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_VERTEX_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_vs_surfaces,
+};
+
+
+
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index ba03afd6c1..ac11790151 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -177,14 +177,6 @@ static void brw_note_fence( struct intel_context *intel, GLuint fence )
    brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE;
 }
 
-
-static void brw_note_unlock( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-   brw_state_cache_check_size(brw);
-}
-
-
 /* called from intelWaitForIdle() and intelFlush()
  *
  * For now, just flush everything.  Could be smarter later.
@@ -194,7 +186,7 @@ static GLuint brw_flush_cmd( void )
    struct brw_mi_flush flush;
    flush.opcode = CMD_MI_FLUSH;
    flush.pad = 0;
-   flush.flags = BRW_FLUSH_READ_CACHE | BRW_FLUSH_STATE_CACHE;
+   flush.flags = BRW_FLUSH_STATE_CACHE;
    return *(GLuint *)&flush;
 }
 
@@ -215,7 +207,6 @@ void brwInitVtbl( struct brw_context *brw )
 
    brw->intel.vtbl.invalidate_state = brw_invalidate_state;
    brw->intel.vtbl.note_fence = brw_note_fence;
-   brw->intel.vtbl.note_unlock = brw_note_unlock;
    brw->intel.vtbl.new_batch = brw_new_batch;
    brw->intel.vtbl.finish_batch = brw_finish_batch;
    brw->intel.vtbl.destroy = brw_destroy_context;
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index d4e22d5939..2292de94c4 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -41,13 +41,13 @@ GLuint brw_wm_nr_args( GLuint opcode )
 {
    switch (opcode) {
    case WM_FRONTFACING:
-      return 0;
    case WM_PIXELXY:
+      return 0;
    case WM_CINTERP:
    case WM_WPOSXY:
+   case WM_DELTAXY:
       return 1;
    case WM_LINTERP:
-   case WM_DELTAXY:
    case WM_PIXELW:
       return 2;
    case WM_FB_WRITE:
@@ -171,9 +171,11 @@ static void do_wm_prog( struct brw_context *brw,
     * differently from "simple" shaders.
     */
    if (fp->isGLSL) {
+      c->dispatch_width = 8;
       brw_wm_glsl_emit(brw, c);
    }
    else {
+      c->dispatch_width = 16;
       brw_wm_non_glsl_emit(brw, c);
    }
 
@@ -274,6 +276,9 @@ static void brw_wm_populate_key( struct brw_context *brw,
    /* _NEW_LIGHT */
    key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT);
 
+   /* _NEW_HINT */
+   key->linear_color = (ctx->Hint.PerspectiveCorrection == GL_FASTEST);
+
    /* _NEW_TEXTURE */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
       const struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
@@ -353,6 +358,7 @@ const struct brw_tracked_state brw_wm_prog = {
    .dirty = {
       .mesa  = (_NEW_COLOR |
 		_NEW_DEPTH |
+                _NEW_HINT |
 		_NEW_STENCIL |
 		_NEW_POLYGON |
 		_NEW_LINE |
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 307b5e59dd..ae98b5492d 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -63,6 +63,7 @@ struct brw_wm_prog_key {
    GLuint computes_depth:1;	/* could be derived from program string */
    GLuint source_depth_to_render_target:1;
    GLuint flat_shade:1;
+   GLuint linear_color:1;  /**< linear interpolation vs perspective interp */
    GLuint runtime_check_aads_emit:1;
    
    GLbitfield proj_attrib_mask; /**< one bit per fragment program attribute */
@@ -241,19 +242,25 @@ struct brw_wm_compile {
    GLuint max_wm_grf;
    GLuint last_scratch;
 
+   GLuint cur_inst;  /**< index of current instruction */
+
+   GLboolean out_of_regs;  /**< ran out of GRF registers? */
+
    /** Mapping from Mesa registers to hardware registers */
    struct {
       GLboolean inited;
       struct brw_reg reg;
    } wm_regs[PROGRAM_PAYLOAD+1][256][4];
 
+   GLboolean used_grf[BRW_WM_MAX_GRF];
+   GLuint first_free_grf;
    struct brw_reg stack;
    struct brw_reg emit_mask_reg;
-   GLuint reg_index;  /**< Index of next free GRF register */
    GLuint tmp_regs[BRW_WM_MAX_GRF];
    GLuint tmp_index;
    GLuint tmp_max;
    GLuint subroutines[BRW_WM_MAX_SUBROUTINE];
+   GLuint dispatch_width;
 
    /** we may need up to 3 constants per instruction (if use_const_buffer) */
    struct {
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index a870e75e52..268f7965c0 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -65,8 +65,7 @@ static INLINE struct brw_reg sechalf( struct brw_reg reg )
 
 static void emit_pixel_xy(struct brw_compile *p,
 			  const struct brw_reg *dst,
-			  GLuint mask,
-			  const struct brw_reg *arg0)
+			  GLuint mask)
 {
    struct brw_reg r1 = brw_vec1_grf(1, 0);
    struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW);
@@ -98,8 +97,7 @@ static void emit_pixel_xy(struct brw_compile *p,
 static void emit_delta_xy(struct brw_compile *p,
 			  const struct brw_reg *dst,
 			  GLuint mask,
-			  const struct brw_reg *arg0,
-			  const struct brw_reg *arg1)
+			  const struct brw_reg *arg0)
 {
    struct brw_reg r1 = brw_vec1_grf(1, 0);
 
@@ -353,6 +351,19 @@ static void emit_mad( struct brw_compile *p,
    }
 }
 
+static void emit_trunc( struct brw_compile *p,
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0)
+{
+   GLuint i;
+
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 brw_RNDZ(p, dst[i], arg0[i]);
+      }
+   }
+}
 
 static void emit_lrp( struct brw_compile *p, 
 		      const struct brw_reg *dst,
@@ -532,16 +543,18 @@ static void emit_dp3( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
+   int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
    if (!(mask & WRITEMASK_XYZW))
       return; /* Do not emit dead code */
 
-   assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
+   assert(is_power_of_two(mask & WRITEMASK_XYZW));
 
    brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]);
    brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]);
 
    brw_set_saturate(p, (mask & SATURATE) ? 1 : 0);
-   brw_MAC(p, dst[0], arg0[2], arg1[2]);
+   brw_MAC(p, dst[dst_chan], arg0[2], arg1[2]);
    brw_set_saturate(p, 0);
 }
 
@@ -552,17 +565,19 @@ static void emit_dp4( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
+   int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
    if (!(mask & WRITEMASK_XYZW))
       return; /* Do not emit dead code */
 
-   assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
+   assert(is_power_of_two(mask & WRITEMASK_XYZW));
 
    brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]);
    brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]);
    brw_MAC(p, brw_null_reg(), arg0[2], arg1[2]);
 
    brw_set_saturate(p, (mask & SATURATE) ? 1 : 0);
-   brw_MAC(p, dst[0], arg0[3], arg1[3]);
+   brw_MAC(p, dst[dst_chan], arg0[3], arg1[3]);
    brw_set_saturate(p, 0);
 }
 
@@ -573,17 +588,19 @@ static void emit_dph( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
+   const int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
    if (!(mask & WRITEMASK_XYZW))
       return; /* Do not emit dead code */
 
-   assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
+   assert(is_power_of_two(mask & WRITEMASK_XYZW));
 
    brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]);
    brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]);
-   brw_MAC(p, dst[0], arg0[2], arg1[2]);
+   brw_MAC(p, dst[dst_chan], arg0[2], arg1[2]);
 
    brw_set_saturate(p, (mask & SATURATE) ? 1 : 0);
-   brw_ADD(p, dst[0], dst[0], arg1[3]);
+   brw_ADD(p, dst[dst_chan], dst[dst_chan], arg1[3]);
    brw_set_saturate(p, 0);
 }
 
@@ -619,18 +636,19 @@ static void emit_math1( struct brw_compile *p,
 			GLuint mask,
 			const struct brw_reg *arg0 )
 {
+   int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
    if (!(mask & WRITEMASK_XYZW))
       return; /* Do not emit dead code */
 
-   //assert((mask & WRITEMASK_XYZW) == WRITEMASK_X ||
-   //	  function == BRW_MATH_FUNCTION_SINCOS);
-   
+   assert(is_power_of_two(mask & WRITEMASK_XYZW));
+
    brw_MOV(p, brw_message_reg(2), arg0[0]);
 
    /* Send two messages to perform all 16 operations:
     */
    brw_math_16(p, 
-	       dst[0],
+	       dst[dst_chan],
 	       function,
 	       (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
 	       2,
@@ -646,10 +664,12 @@ static void emit_math2( struct brw_compile *p,
 			const struct brw_reg *arg0,
 			const struct brw_reg *arg1)
 {
+   int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
    if (!(mask & WRITEMASK_XYZW))
       return; /* Do not emit dead code */
 
-   assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
+   assert(is_power_of_two(mask & WRITEMASK_XYZW));
 
    brw_push_insn_state(p);
 
@@ -668,7 +688,7 @@ static void emit_math2( struct brw_compile *p,
     */
    brw_set_compression_control(p, BRW_COMPRESSION_NONE);
    brw_math(p, 
-	    dst[0],
+	    dst[dst_chan],
 	    function,
 	    (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
 	    2,
@@ -678,7 +698,7 @@ static void emit_math2( struct brw_compile *p,
 
    brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
    brw_math(p, 
-	    offset(dst[0],1),
+	    offset(dst[dst_chan],1),
 	    function,
 	    (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
 	    4,
@@ -701,6 +721,7 @@ static void emit_tex( struct brw_wm_compile *c,
    GLuint msgLength, responseLength;
    GLuint i, nr;
    GLuint emit;
+   GLuint msg_type;
 
    /* How many input regs are there?
     */
@@ -714,10 +735,14 @@ static void emit_tex( struct brw_wm_compile *c,
       emit = WRITEMASK_XY;
       nr = 2;
       break;
-   default:
+   case TEXTURE_3D_INDEX:
+   case TEXTURE_CUBE_INDEX:
       emit = WRITEMASK_XYZ;
       nr = 3;
       break;
+   default:
+      /* unexpected target */
+      abort();
    }
 
    if (inst->tex_shadow) {
@@ -738,6 +763,18 @@ static void emit_tex( struct brw_wm_compile *c,
 
    responseLength = 8;		/* always */
 
+   if (BRW_IS_IGDNG(p->brw)) {
+       if (inst->tex_shadow)
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG;
+       else
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG;
+   } else {
+       if (inst->tex_shadow)
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE;
+       else
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE;
+   }
+
    brw_SAMPLE(p, 
 	      retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW),
 	      1,
@@ -745,12 +782,12 @@ static void emit_tex( struct brw_wm_compile *c,
               SURF_INDEX_TEXTURE(inst->tex_unit),
 	      inst->tex_unit,	  /* sampler */
 	      inst->writemask,
-	      (inst->tex_shadow ? 
-	       BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE : 
-	       BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE),
+	      msg_type, 
 	      responseLength,
 	      msgLength,
-	      0);	
+	      0,	
+	      1,
+	      BRW_SAMPLER_SIMD_MODE_SIMD16);	
 }
 
 
@@ -762,7 +799,7 @@ static void emit_txb( struct brw_wm_compile *c,
 {
    struct brw_compile *p = &c->func;
    GLuint msgLength;
-
+   GLuint msg_type;
    /* Shadow ignored for txb.
     */
    switch (inst->tex_idx) {
@@ -777,16 +814,25 @@ static void emit_txb( struct brw_wm_compile *c,
       brw_MOV(p, brw_message_reg(4), arg[1]);
       brw_MOV(p, brw_message_reg(6), brw_imm_f(0));
       break;
-   default:
+   case TEXTURE_3D_INDEX:
+   case TEXTURE_CUBE_INDEX:
       brw_MOV(p, brw_message_reg(2), arg[0]);
       brw_MOV(p, brw_message_reg(4), arg[1]);
       brw_MOV(p, brw_message_reg(6), arg[2]);
       break;
+   default:
+      /* unexpected target */
+      abort();
    }
 
    brw_MOV(p, brw_message_reg(8), arg[3]);
    msgLength = 9;
 
+   if (BRW_IS_IGDNG(p->brw))
+       msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG;
+   else
+       msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS;
+
    brw_SAMPLE(p, 
 	      retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW),
 	      1,
@@ -794,10 +840,12 @@ static void emit_txb( struct brw_wm_compile *c,
               SURF_INDEX_TEXTURE(inst->tex_unit),
 	      inst->tex_unit,	  /* sampler */
 	      inst->writemask,
-	      BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,
+	      msg_type,
 	      8,		/* responseLength */
 	      msgLength,
-	      0);	
+	      0,	
+	      1,
+	      BRW_SAMPLER_SIMD_MODE_SIMD16);	
 }
 
 
@@ -1009,7 +1057,7 @@ static void emit_fb_write( struct brw_wm_compile *c,
 	      get_element_ud(brw_vec8_grf(1,0), 6), 
 	      brw_imm_ud(1<<26)); 
 
-      jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+      jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
       {
 	 emit_aa(c, arg1, 2);
 	 fire_fb_write(c, 0, nr, target, eot);
@@ -1161,11 +1209,11 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 /* Generated instructions for calculating triangle interpolants:
 	  */
       case WM_PIXELXY:
-	 emit_pixel_xy(p, dst, dst_flags, args[0]);
+	 emit_pixel_xy(p, dst, dst_flags);
 	 break;
 
       case WM_DELTAXY:
-	 emit_delta_xy(p, dst, dst_flags, args[0], args[1]);
+	 emit_delta_xy(p, dst, dst_flags, args[0]);
 	 break;
 
       case WM_WPOSXY:
@@ -1222,6 +1270,10 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 emit_dph(p, dst, dst_flags, args[0], args[1]);
 	 break;
 
+      case OPCODE_TRUNC:
+	 emit_trunc(p, dst, dst_flags, args[0]);
+	 break;
+
       case OPCODE_LRP:
 	 emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]);
 	 break;
@@ -1348,4 +1400,13 @@ void brw_wm_emit( struct brw_wm_compile *c )
 		      inst->dst[i]->hw_reg, 
 		      inst->dst[i]->spill_slot);
    }
+
+   if (INTEL_DEBUG & DEBUG_WM) {
+      int i;
+
+      _mesa_printf("wm-native:\n");
+      for (i = 0; i < p->nr_insn; i++)
+	 brw_disasm(stderr, &p->store[i]);
+      _mesa_printf("\n");
+   }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index 8cf3a1fd13..123fe841c3 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -42,6 +42,12 @@
 #include "shader/prog_statevars.h"
 
 
+/** An invalid texture target */
+#define TEX_TARGET_NONE NUM_TEXTURE_TARGETS
+
+/** An invalid texture unit */
+#define TEX_UNIT_NONE BRW_MAX_TEX_UNIT
+
 #define FIRST_INTERNAL_TEMP MAX_NV_FRAGMENT_PROGRAM_TEMPS
 
 #define X    0
@@ -199,6 +205,15 @@ static struct prog_instruction * emit_tex_op(struct brw_wm_compile *c,
 {
    struct prog_instruction *inst = get_fp_inst(c);
       
+   assert(tex_src_unit < BRW_MAX_TEX_UNIT ||
+          tex_src_unit == TEX_UNIT_NONE);
+   assert(tex_src_target < NUM_TEXTURE_TARGETS ||
+          tex_src_target == TEX_TARGET_NONE);
+
+   /* update mask of which texture units are referenced by this program */
+   if (tex_src_unit != TEX_UNIT_NONE)
+      c->fp->tex_units_used |= (1 << tex_src_unit);
+
    memset(inst, 0, sizeof(*inst));
 
    inst->Opcode = op;
@@ -223,12 +238,45 @@ static struct prog_instruction * emit_op(struct brw_wm_compile *c,
 				       struct prog_src_register src2 )
 {
    return emit_tex_op(c, op, dest, saturate,
-                      0, 0, 0,  /* tex unit, target, shadow */
+                      TEX_UNIT_NONE, TEX_TARGET_NONE, 0,  /* unit, tgt, shadow */
                       src0, src1, src2);
 }
-   
 
 
+/* Many Mesa opcodes produce the same value across all the result channels.
+ * We'd rather not have to support that splatting in the opcode implementations,
+ * and brw_wm_pass*.c wants to optimize them out by shuffling references around
+ * anyway.  We can easily get both by emitting the opcode to one channel, and
+ * then MOVing it to the others, which brw_wm_pass*.c already understands.
+ */
+static struct prog_instruction *emit_scalar_insn(struct brw_wm_compile *c,
+						 const struct prog_instruction *inst0)
+{
+   struct prog_instruction *inst;
+   unsigned int dst_chan;
+   unsigned int other_channel_mask;
+
+   if (inst0->DstReg.WriteMask == 0)
+      return NULL;
+
+   dst_chan = _mesa_ffs(inst0->DstReg.WriteMask) - 1;
+   inst = get_fp_inst(c);
+   *inst = *inst0;
+   inst->DstReg.WriteMask = 1 << dst_chan;
+
+   other_channel_mask = inst0->DstReg.WriteMask & ~(1 << dst_chan);
+   if (other_channel_mask != 0) {
+      inst = emit_op(c,
+		     OPCODE_MOV,
+		     dst_mask(inst0->DstReg, other_channel_mask),
+		     0,
+		     src_swizzle1(src_reg_from_dst(inst0->DstReg), dst_chan),
+		     src_undef(),
+		     src_undef());
+   }
+   return inst;
+}
+
 
 /***********************************************************************
  * Special instructions for interpolation and other tasks
@@ -354,24 +402,28 @@ static void emit_interp( struct brw_wm_compile *c,
 		 src_undef());
       }
       else {
-	 emit_op(c,
-		 WM_LINTERP,
-		 dst,
-		 0,
-		 interp,
-		 deltas,
-		 src_undef());
+         if (c->key.linear_color) {
+            emit_op(c,
+                    WM_LINTERP,
+                    dst,
+                    0,
+                    interp,
+                    deltas,
+                    src_undef());
+         }
+         else {
+            /* perspective-corrected color interpolation */
+            emit_op(c,
+                    WM_PINTERP,
+                    dst,
+                    0,
+                    interp,
+                    deltas,
+                    get_pixel_w(c));
+         }
       }
       break;
    case FRAG_ATTRIB_FOGC:
-      /* The FOGC input is really special.  When a program uses glFogFragCoord,
-       * the results returned are supposed to be (f,0,0,1).  But for Mesa GLSL,
-       * the glFrontFacing and glPointCoord values are also stashed in FOGC.
-       * So, write the interpolated fog value to X, then either 0, 1, or the
-       * stashed values to Y, Z, W.  Note that this means that
-       * glFogFragCoord.yzw can be wrong in those cases!
-       */
-
       /* Interpolate the fog coordinate */
       emit_op(c,
 	      WM_PINTERP,
@@ -381,26 +433,40 @@ static void emit_interp( struct brw_wm_compile *c,
 	      deltas,
 	      get_pixel_w(c));
 
-      /* Move the front facing value into FOGC.y if it's needed. */
-      if (c->fp->program.UsesFrontFacing) {
-	 emit_op(c,
-		 WM_FRONTFACING,
-		 dst_mask(dst, WRITEMASK_Y),
-		 0,
-		 src_undef(),
-		 src_undef(),
-		 src_undef());
-      } else {
-	 emit_op(c,
-		 OPCODE_MOV,
-		 dst_mask(dst, WRITEMASK_Y),
-		 0,
-		 src_swizzle1(interp, SWIZZLE_ZERO),
-		 src_undef(),
-		 src_undef());
-      }
+      emit_op(c,
+	      OPCODE_MOV,
+	      dst_mask(dst, WRITEMASK_YZW),
+	      0,
+	      src_swizzle(interp,
+			  SWIZZLE_ZERO,
+			  SWIZZLE_ZERO,
+			  SWIZZLE_ZERO,
+			  SWIZZLE_ONE),
+	      src_undef(),
+	      src_undef());
+      break;
+
+   case FRAG_ATTRIB_FACE:
+      /* XXX review/test this case */
+      emit_op(c,
+              WM_FRONTFACING,
+              dst_mask(dst, WRITEMASK_X),
+              0,
+              src_undef(),
+              src_undef(),
+              src_undef());
+      break;
+
+   case FRAG_ATTRIB_PNTC:
+      /* XXX review/test this case */
+      emit_op(c,
+	      WM_PINTERP,
+	      dst_mask(dst, WRITEMASK_XY),
+	      0,
+	      interp,
+	      deltas,
+	      get_pixel_w(c));
 
-      /* Should do the PointCoord thing here. */
       emit_op(c,
 	      OPCODE_MOV,
 	      dst_mask(dst, WRITEMASK_ZW),
@@ -413,6 +479,7 @@ static void emit_interp( struct brw_wm_compile *c,
 	      src_undef(),
 	      src_undef());
       break;
+
    default:
       emit_op(c,
 	      WM_PINTERP,
@@ -631,6 +698,8 @@ static void precalc_tex( struct brw_wm_compile *c,
    struct prog_dst_register tmpcoord;
    const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
 
+   assert(unit < BRW_MAX_TEX_UNIT);
+
    if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) {
        struct prog_instruction *out;
        struct prog_dst_register tmp0 = get_temp(c);
@@ -671,7 +740,7 @@ static void precalc_tex( struct brw_wm_compile *c,
 
        /* tmp0 = 1 / tmp1 */
        emit_op(c, OPCODE_RCP,
-               tmp0,
+               dst_mask(tmp0, WRITEMASK_X),
                0,
                tmp1src,
                src_undef(),
@@ -682,7 +751,7 @@ static void precalc_tex( struct brw_wm_compile *c,
                tmpcoord,
                0,
                src0,
-               tmp0src,
+               src_swizzle1(tmp0src, SWIZZLE_X),
                src_undef());
 
        release_temp(c, tmp0);
@@ -1033,6 +1102,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    c->delta_xy = src_undef();
    c->pixel_w = src_undef();
    c->nr_fp_insns = 0;
+   c->fp->tex_units_used = 0x0;
 
    /* Emit preamble instructions.  This is where special instructions such as
     * WM_CINTERP, WM_LINTERP, WM_PINTERP and WM_WPOSXY are emitted to
@@ -1100,6 +1170,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
       case OPCODE_TXB:
 	 out = emit_insn(c, inst);
 	 out->TexSrcUnit = fp->program.Base.SamplerUnits[inst->TexSrcUnit];
+         assert(out->TexSrcUnit < BRW_MAX_TEX_UNIT);
 	 break;
 
       case OPCODE_XPD: 
@@ -1126,9 +1197,11 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
 	 break;
       case OPCODE_PRINT:
 	 break;
-	 
       default:
-	 emit_insn(c, inst);
+	 if (brw_wm_is_scalar_result(inst->Opcode))
+	    emit_scalar_insn(c, inst);
+	 else
+	    emit_insn(c, inst);
 	 break;
       }
    }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index a907e1b788..7c210abbce 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -1,5 +1,7 @@
 #include "main/macros.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_optimize.h"
 #include "brw_context.h"
 #include "brw_eu.h"
 #include "brw_wm.h"
@@ -8,6 +10,9 @@ enum _subroutine {
     SUB_NOISE1, SUB_NOISE2, SUB_NOISE3, SUB_NOISE4
 };
 
+static struct brw_reg get_dst_reg(struct brw_wm_compile *c,
+                                  const struct prog_instruction *inst,
+                                  GLuint component);
 
 /**
  * Determine if the given fragment program uses GLSL features such
@@ -20,8 +25,8 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
     for (i = 0; i < fp->Base.NumInstructions; i++) {
 	const struct prog_instruction *inst = &fp->Base.Instructions[i];
 	switch (inst->Opcode) {
+	    case OPCODE_ARL:
 	    case OPCODE_IF:
-	    case OPCODE_TRUNC:
 	    case OPCODE_ENDIF:
 	    case OPCODE_CAL:
 	    case OPCODE_BRK:
@@ -42,6 +47,83 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 }
 
 
+
+static void
+reclaim_temps(struct brw_wm_compile *c);
+
+
+/** Mark GRF register as used. */
+static void
+prealloc_grf(struct brw_wm_compile *c, int r)
+{
+   c->used_grf[r] = GL_TRUE;
+}
+
+
+/** Mark given GRF register as not in use. */
+static void
+release_grf(struct brw_wm_compile *c, int r)
+{
+   /*assert(c->used_grf[r]);*/
+   c->used_grf[r] = GL_FALSE;
+   c->first_free_grf = MIN2(c->first_free_grf, r);
+}
+
+
+/** Return index of a free GRF, mark it as used. */
+static int
+alloc_grf(struct brw_wm_compile *c)
+{
+   GLuint r;
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   /* no free temps, try to reclaim some */
+   reclaim_temps(c);
+   c->first_free_grf = 0;
+
+   /* try alloc again */
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   for (r = 0; r < BRW_WM_MAX_GRF; r++) {
+      assert(c->used_grf[r]);
+   }
+
+   /* really, no free GRF regs found */
+   if (!c->out_of_regs) {
+      /* print warning once per compilation */
+      _mesa_warning(NULL, "i965: ran out of registers for fragment program");
+      c->out_of_regs = GL_TRUE;
+   }
+
+   return -1;
+}
+
+
+/** Return number of GRF registers used */
+static int
+num_grf_used(const struct brw_wm_compile *c)
+{
+   int r;
+   for (r = BRW_WM_MAX_GRF - 1; r >= 0; r--)
+      if (c->used_grf[r])
+         return r + 1;
+   return 0;
+}
+
+
+
 /**
  * Record the mapping of a Mesa register to a hardware register.
  */
@@ -52,27 +134,26 @@ static void set_reg(struct brw_wm_compile *c, int file, int index,
     c->wm_regs[file][index][component].inited = GL_TRUE;
 }
 
-/**
- * Examine instruction's write mask to find index of first component
- * enabled for writing.
- */
-static int get_scalar_dst_index(const struct prog_instruction *inst)
-{
-    int i;
-    for (i = 0; i < 4; i++)
-	if (inst->DstReg.WriteMask & (1<<i))
-	    break;
-    return i;
-}
-
 static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
 {
     struct brw_reg reg;
-    if(c->tmp_index == c->tmp_max)
-	c->tmp_regs[ c->tmp_max++ ] = c->reg_index++;
-    
+
+    /* if we need to allocate another temp, grow the tmp_regs[] array */
+    if (c->tmp_index == c->tmp_max) {
+       int r = alloc_grf(c);
+       if (r < 0) {
+          /*printf("Out of temps in %s\n", __FUNCTION__);*/
+          r = 50; /* XXX random register! */
+       }
+       c->tmp_regs[ c->tmp_max++ ] = r;
+    }
+
+    /* form the GRF register */
     reg = brw_vec8_grf(c->tmp_regs[ c->tmp_index++ ], 0);
+    /*printf("alloc_temp %d\n", reg.nr);*/
+    assert(reg.nr < BRW_WM_MAX_GRF);
     return reg;
+
 }
 
 /**
@@ -130,35 +211,29 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 	    return brw_null_reg();
     }
 
+    assert(index < 256);
+    assert(component < 4);
+
     /* see if we've already allocated a HW register for this Mesa register */
     if (c->wm_regs[file][index][component].inited) {
-	/* yes, re-use */
-	reg = c->wm_regs[file][index][component].reg;
+       /* yes, re-use */
+       reg = c->wm_regs[file][index][component].reg;
     }
     else {
 	/* no, allocate new register */
-	reg = brw_vec8_grf(c->reg_index, 0);
-    }
+       int grf = alloc_grf(c);
+       /*printf("alloc grf %d for reg %d:%d.%d\n", grf, file, index, component);*/
+       if (grf < 0) {
+          /* totally out of temps */
+          grf = 51; /* XXX random register! */
+       }
 
-    /* if this is a new register allocation, record it in the table */
-    if (!c->wm_regs[file][index][component].inited) {
-	set_reg(c, file, index, component, reg);
-	c->reg_index++;
-    }
+       reg = brw_vec8_grf(grf, 0);
+       /*printf("Alloc new grf %d for %d.%d\n", reg.nr, index, component);*/
 
-    if (c->reg_index >= BRW_WM_MAX_GRF - 12) {
-	/* ran out of temporary registers! */
-#if 1
-        /* This is a big hack for now.
-         * Return bad register index, just don't hang the GPU.
-         */
-        _mesa_fprintf(stderr, "out of regs %d\n", c->reg_index);
-        c->reg_index = BRW_WM_MAX_GRF - 13;
-#else
-	return brw_null_reg();
-#endif
+       set_reg(c, file, index, component, reg);
     }
- 
+
     if (neg & (1 << component)) {
 	reg = negate(reg);
     }
@@ -168,6 +243,46 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 }
 
 
+
+/**
+ * This is called if we run out of GRF registers.  Examine the live intervals
+ * of temp regs in the program and free those which won't be used again.
+ */
+static void
+reclaim_temps(struct brw_wm_compile *c)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS];
+   GLint intEnd[MAX_PROGRAM_TEMPS];
+   int index;
+
+   /*printf("Reclaim temps:\n");*/
+
+   _mesa_find_temp_intervals(c->prog_instructions, c->nr_fp_insns,
+                             intBegin, intEnd);
+
+   for (index = 0; index < MAX_PROGRAM_TEMPS; index++) {
+      if (intEnd[index] != -1 && intEnd[index] < c->cur_inst) {
+         /* program temp[i] can be freed */
+         int component;
+         /*printf("  temp[%d] is dead\n", index);*/
+         for (component = 0; component < 4; component++) {
+            if (c->wm_regs[PROGRAM_TEMPORARY][index][component].inited) {
+               int r = c->wm_regs[PROGRAM_TEMPORARY][index][component].reg.nr;
+               release_grf(c, r);
+               /*
+               printf("  Reclaim temp %d, reg %d at inst %d\n",
+                      index, r, c->cur_inst);
+               */
+               c->wm_regs[PROGRAM_TEMPORARY][index][component].inited = GL_FALSE;
+            }
+         }
+      }
+   }
+}
+
+
+
+
 /**
  * Preallocate registers.  This sets up the Mesa to hardware register
  * mapping for certain registers, such as constants (uniforms/state vars)
@@ -179,6 +294,10 @@ static void prealloc_reg(struct brw_wm_compile *c)
     struct brw_reg reg;
     int urb_read_length = 0;
     GLuint inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted;
+    GLuint reg_index = 0;
+
+    memset(c->used_grf, GL_FALSE, sizeof(c->used_grf));
+    c->first_free_grf = 0;
 
     for (i = 0; i < 4; i++) {
         if (i < c->key.nr_depth_regs) 
@@ -187,14 +306,20 @@ static void prealloc_reg(struct brw_wm_compile *c)
             reg = brw_vec8_grf(0, 0);
 	set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg);
     }
-    c->reg_index += 2 * c->key.nr_depth_regs;
+    reg_index += 2 * c->key.nr_depth_regs;
 
     /* constants */
     {
-        const int nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_temps = c->fp->program.Base.NumTemporaries;
 
         /* use a real constant buffer, or just use a section of the GRF? */
-        c->fp->use_const_buffer = GL_FALSE; /* (nr_params > 8);*/
+        /* XXX this heuristic may need adjustment... */
+        if ((nr_params + nr_temps) * 4 + reg_index > 80)
+           c->fp->use_const_buffer = GL_TRUE;
+        else
+           c->fp->use_const_buffer = GL_FALSE;
+        /*printf("WM use_const_buffer = %d\n", c->fp->use_const_buffer);*/
 
         if (c->fp->use_const_buffer) {
            /* We'll use a real constant buffer and fetch constants from
@@ -216,7 +341,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            for (i = 0; i < nr_params; i++) {
               /* loop over XYZW channels */
               for (j = 0; j < 4; j++, index++) {
-                 reg = brw_vec1_grf(c->reg_index + index / 8, index % 8);
+                 reg = brw_vec1_grf(reg_index + index / 8, index % 8);
                  /* Save pointer to parameter/constant value.
                   * Constants will be copied in prepare_constant_buffer()
                   */
@@ -226,7 +351,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            }
            /* number of constant regs used (each reg is float[8]) */
            c->nr_creg = 2 * ((4 * nr_params + 15) / 16);
-           c->reg_index += c->nr_creg;
+           reg_index += c->nr_creg;
         }
     }
 
@@ -242,23 +367,52 @@ static void prealloc_reg(struct brw_wm_compile *c)
 	  fp_input = -1;
 
        if (fp_input >= 0 && inputs & (1 << fp_input)) {
-	  urb_read_length = c->reg_index;
-	  reg = brw_vec8_grf(c->reg_index, 0);
+	  urb_read_length = reg_index;
+	  reg = brw_vec8_grf(reg_index, 0);
 	  for (j = 0; j < 4; j++)
 	     set_reg(c, PROGRAM_PAYLOAD, fp_input, j, reg);
        }
        if (c->key.vp_outputs_written & (1 << i)) {
-	  c->reg_index += 2;
+	  reg_index += 2;
        }
     }
 
     c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2;
     c->prog_data.urb_read_length = urb_read_length;
     c->prog_data.curb_read_length = c->nr_creg;
-    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index++;
-    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index += 2;
+    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index++;
+    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index += 2;
+
+    /* mark GRF regs [0..reg_index-1] as in-use */
+    for (i = 0; i < reg_index; i++)
+       prealloc_grf(c, i);
+
+    /* Don't use GRF 126, 127.  Using them seems to lead to GPU lock-ups */
+    prealloc_grf(c, 126);
+    prealloc_grf(c, 127);
+
+    for (i = 0; i < c->nr_fp_insns; i++) {
+	const struct prog_instruction *inst = &c->prog_instructions[i];
+	struct brw_reg dst[4];
+
+	switch (inst->Opcode) {
+	case OPCODE_TEX:
+	case OPCODE_TXB:
+	    /* Allocate the channels of texture results contiguously,
+	     * since they are written out that way by the sampler unit.
+	     */
+	    for (j = 0; j < 4; j++) {
+		dst[j] = get_dst_reg(c, inst, j);
+		if (j != 0)
+		    assert(dst[j].nr == dst[j - 1].nr + 1);
+	    }
+	    break;
+	default:
+	    break;
+	}
+    }
 
     /* An instruction may reference up to three constants.
      * They'll be found in these registers.
@@ -267,12 +421,12 @@ static void prealloc_reg(struct brw_wm_compile *c)
     if (c->fp->use_const_buffer) {
        for (i = 0; i < 3; i++) {
           c->current_const[i].index = -1;
-          c->current_const[i].reg = alloc_tmp(c);
+          c->current_const[i].reg = brw_vec8_grf(alloc_grf(c), 0);
        }
     }
 #if 0
     printf("USE CONST BUFFER? %d\n", c->fp->use_const_buffer);
-    printf("AFTER PRE_ALLOC, reg_index = %d\n", c->reg_index);
+    printf("AFTER PRE_ALLOC, reg_index = %d\n", reg_index);
 #endif
 }
 
@@ -376,6 +530,14 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c,
     const GLuint nr = 1;
     const GLuint component = GET_SWZ(src->Swizzle, channel);
 
+    /* Extended swizzle terms */
+    if (component == SWIZZLE_ZERO) {
+       return brw_imm_f(0.0F);
+    }
+    else if (component == SWIZZLE_ONE) {
+       return brw_imm_f(1.0F);
+    }
+
     if (c->fp->use_const_buffer &&
         (src->File == PROGRAM_STATE_VAR ||
          src->File == PROGRAM_CONSTANT ||
@@ -489,23 +651,6 @@ static void invoke_subroutine( struct brw_wm_compile *c,
     }
 }
 
-static void emit_abs( struct brw_wm_compile *c,
-                      const struct prog_instruction *inst)
-{
-    int i;
-    struct brw_compile *p = &c->func;
-    brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF);
-    for (i = 0; i < 4; i++) {
-	if (inst->DstReg.WriteMask & (1<<i)) {
-	    struct brw_reg src, dst;
-	    dst = get_dst_reg(c, inst, i);
-	    src = get_src_reg(c, inst, 0, i);
-	    brw_MOV(p, dst, brw_abs(src));
-	}
-    }
-    brw_set_saturate(p, 0);
-}
-
 static void emit_trunc( struct brw_wm_compile *c,
                         const struct prog_instruction *inst)
 {
@@ -673,27 +818,26 @@ static void emit_fb_write(struct brw_wm_compile *c,
     }
 
     if (c->key.dest_depth_reg) {
-        GLuint comp = c->key.dest_depth_reg / 2;
-        GLuint off = c->key.dest_depth_reg % 2;
+        const GLuint comp = c->key.dest_depth_reg / 2;
+        const GLuint off = c->key.dest_depth_reg % 2;
 
-        assert(comp == 1);
-        assert(off == 0);
-#if 0
-        /* XXX do we need this code?   comp always 1, off always 0, it seems */
         if (off != 0) {
+            /* XXX this code needs review/testing */
+            struct brw_reg arg1_0 = get_src_reg(c, inst, 1, comp);
+            struct brw_reg arg1_1 = get_src_reg(c, inst, 1, comp+1);
+
             brw_push_insn_state(p);
             brw_set_compression_control(p, BRW_COMPRESSION_NONE);
 
-            brw_MOV(p, brw_message_reg(nr), offset(arg1[comp],1));
+            brw_MOV(p, brw_message_reg(nr), offset(arg1_0, 1));
             /* 2nd half? */
-            brw_MOV(p, brw_message_reg(nr+1), arg1[comp+1]);
+            brw_MOV(p, brw_message_reg(nr+1), arg1_1);
             brw_pop_insn_state(p);
         }
         else
-#endif
         {
-           struct brw_reg src =  get_src_reg(c, inst, 1, 1);
-           brw_MOV(p, brw_message_reg(nr), src);
+            struct brw_reg src =  get_src_reg(c, inst, 1, 1);
+            brw_MOV(p, brw_message_reg(nr), src);
         }
         nr += 2;
    }
@@ -882,12 +1026,20 @@ static void emit_dp3(struct brw_wm_compile *c,
     struct brw_reg src0[3], src1[3], dst;
     int i;
     struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
+    if (!(mask & WRITEMASK_XYZW))
+	return;
+
+    assert(is_power_of_two(mask & WRITEMASK_XYZW));
+
     for (i = 0; i < 3; i++) {
 	src0[i] = get_src_reg(c, inst, 0, i);
 	src1[i] = get_src_reg_imm(c, inst, 1, i);
     }
 
-    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+    dst = get_dst_reg(c, inst, dst_chan);
     brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
     brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
     brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
@@ -901,11 +1053,19 @@ static void emit_dp4(struct brw_wm_compile *c,
     struct brw_reg src0[4], src1[4], dst;
     int i;
     struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
+    if (!(mask & WRITEMASK_XYZW))
+	return;
+
+    assert(is_power_of_two(mask & WRITEMASK_XYZW));
+
     for (i = 0; i < 4; i++) {
 	src0[i] = get_src_reg(c, inst, 0, i);
 	src1[i] = get_src_reg_imm(c, inst, 1, i);
     }
-    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+    dst = get_dst_reg(c, inst, dst_chan);
     brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
     brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
     brw_MAC(p, brw_null_reg(), src0[2], src1[2]);
@@ -920,11 +1080,19 @@ static void emit_dph(struct brw_wm_compile *c,
     struct brw_reg src0[4], src1[4], dst;
     int i;
     struct brw_compile *p = &c->func;
+    GLuint mask = inst->DstReg.WriteMask;
+    int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
+    if (!(mask & WRITEMASK_XYZW))
+	return;
+
+    assert(is_power_of_two(mask & WRITEMASK_XYZW));
+
     for (i = 0; i < 4; i++) {
 	src0[i] = get_src_reg(c, inst, 0, i);
 	src1[i] = get_src_reg_imm(c, inst, 1, i);
     }
-    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+    dst = get_dst_reg(c, inst, dst_chan);
     brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
     brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
     brw_MAC(p, dst, src0[2], src1[2]);
@@ -942,37 +1110,28 @@ static void emit_math1(struct brw_wm_compile *c,
                        const struct prog_instruction *inst, GLuint func)
 {
     struct brw_compile *p = &c->func;
-    struct brw_reg src0, dst, tmp;
-    const int mark = mark_tmps( c );
-    int i;
+    struct brw_reg src0, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
+    if (!(mask & WRITEMASK_XYZW))
+	return;
 
-    tmp = alloc_tmp(c);
+    assert(is_power_of_two(mask & WRITEMASK_XYZW));
 
     /* Get first component of source register */
+    dst = get_dst_reg(c, inst, dst_chan);
     src0 = get_src_reg(c, inst, 0, 0);
 
-    /* tmp = func(src0) */
     brw_MOV(p, brw_message_reg(2), src0);
     brw_math(p,
-             tmp,
+             dst,
              func,
              (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE,
              2,
              brw_null_reg(),
              BRW_MATH_DATA_VECTOR,
              BRW_MATH_PRECISION_FULL);
-
-    /*tmp.dw1.bits.swizzle = SWIZZLE_XXXX;*/
-
-    /* replicate tmp value across enabled dest channels */
-    for (i = 0; i < 4; i++) {
-       if (inst->DstReg.WriteMask & (1 << i)) {
-          dst = get_dst_reg(c, inst, i);
-          brw_MOV(p, dst, tmp);
-       }
-    }
-
-    release_tmps(c, mark);
 }
 
 static void emit_rcp(struct brw_wm_compile *c,
@@ -1043,24 +1202,6 @@ static void emit_arl(struct brw_wm_compile *c,
     brw_set_saturate(p, 0);
 }
 
-static void emit_sub(struct brw_wm_compile *c,
-                     const struct prog_instruction *inst)
-{
-    struct brw_compile *p = &c->func;
-    struct brw_reg src0, src1, dst;
-    GLuint mask = inst->DstReg.WriteMask;
-    int i;
-    brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
-    for (i = 0 ; i < 4; i++) {
-	if (mask & (1<<i)) {
-	    dst = get_dst_reg(c, inst, i);
-	    src0 = get_src_reg(c, inst, 0, i);
-	    src1 = get_src_reg_imm(c, inst, 1, i);
-	    brw_ADD(p, dst, src0, negate(src1));
-	}
-    }
-    brw_set_saturate(p, 0);
-}
 
 static void emit_mul(struct brw_wm_compile *c,
                      const struct prog_instruction *inst)
@@ -1172,7 +1313,15 @@ static void emit_pow(struct brw_wm_compile *c,
 {
     struct brw_compile *p = &c->func;
     struct brw_reg dst, src0, src1;
-    dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+    GLuint mask = inst->DstReg.WriteMask;
+    int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1;
+
+    if (!(mask & WRITEMASK_XYZW))
+	return;
+
+    assert(is_power_of_two(mask & WRITEMASK_XYZW));
+
+    dst = get_dst_reg(c, inst, dst_chan);
     src0 = get_src_reg_imm(c, inst, 0, 0);
     src1 = get_src_reg_imm(c, inst, 1, 0);
 
@@ -2474,8 +2623,12 @@ static void emit_txb(struct brw_wm_compile *c,
 {
     struct brw_compile *p = &c->func;
     struct brw_reg dst[4], src[4], payload_reg;
-    GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
+    /* Note: TexSrcUnit was already looked up through SamplerTextures[] */
+    const GLuint unit = inst->TexSrcUnit;
     GLuint i;
+    GLuint msg_type;
+
+    assert(unit < BRW_MAX_TEX_UNIT);
 
     payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
 
@@ -2496,14 +2649,26 @@ static void emit_txb(struct brw_wm_compile *c,
 	    brw_MOV(p, brw_message_reg(3), src[1]);
 	    brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
 	    break;
-	default:
+	case TEXTURE_3D_INDEX:
+	case TEXTURE_CUBE_INDEX:
 	    brw_MOV(p, brw_message_reg(2), src[0]);
 	    brw_MOV(p, brw_message_reg(3), src[1]);
 	    brw_MOV(p, brw_message_reg(4), src[2]);
 	    break;
+	default:
+            /* invalid target */
+            abort();
     }
     brw_MOV(p, brw_message_reg(5), src[3]);          /* bias */
     brw_MOV(p, brw_message_reg(6), brw_imm_f(0));    /* ref (unused?) */
+
+    if (BRW_IS_IGDNG(p->brw)) {
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG;
+    } else {
+        /* Does it work well on SIMD8? */
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS;
+    }
+
     brw_SAMPLE(p,
                retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),  /* dest */
                1,                                           /* msg_reg_nr */
@@ -2511,10 +2676,12 @@ static void emit_txb(struct brw_wm_compile *c,
                SURF_INDEX_TEXTURE(unit),
                unit,                                        /* sampler */
                inst->DstReg.WriteMask,                      /* writemask */
-               BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,      /* msg_type */
+               msg_type,                                    /* msg_type */
                4,                                           /* response_length */
                4,                                           /* msg_length */
-               0);                                          /* eot */
+               0,                                           /* eot */
+               1,
+               BRW_SAMPLER_SIMD_MODE_SIMD8);	
 }
 
 
@@ -2523,11 +2690,15 @@ static void emit_tex(struct brw_wm_compile *c,
 {
     struct brw_compile *p = &c->func;
     struct brw_reg dst[4], src[4], payload_reg;
-    GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
+    /* Note: TexSrcUnit was already looked up through SamplerTextures[] */
+    const GLuint unit = inst->TexSrcUnit;
     GLuint msg_len;
     GLuint i, nr;
     GLuint emit;
     GLboolean shadow = (c->key.shadowtex_mask & (1<<unit)) ? 1 : 0;
+    GLuint msg_type;
+
+    assert(unit < BRW_MAX_TEX_UNIT);
 
     payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
 
@@ -2546,10 +2717,14 @@ static void emit_tex(struct brw_wm_compile *c,
 	    emit = WRITEMASK_XY;
 	    nr = 2;
 	    break;
-	default:
+	case TEXTURE_3D_INDEX:
+	case TEXTURE_CUBE_INDEX:
 	    emit = WRITEMASK_XYZ;
 	    nr = 3;
 	    break;
+	default:
+           /* invalid target */
+           abort();
     }
     msg_len = 1;
 
@@ -2568,6 +2743,16 @@ static void emit_tex(struct brw_wm_compile *c,
        brw_MOV(p, brw_message_reg(6), src[2]);        /* ref value / R coord */
     }
 
+    if (BRW_IS_IGDNG(p->brw)) {
+        if (shadow)
+            msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG;
+        else
+            msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG;
+    } else {
+        /* Does it work for shadow on SIMD8 ? */
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE;
+    }
+    
     brw_SAMPLE(p,
                retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */
                1,                                          /* msg_reg_nr */
@@ -2575,10 +2760,12 @@ static void emit_tex(struct brw_wm_compile *c,
                SURF_INDEX_TEXTURE(unit),
                unit,                                       /* sampler */
                inst->DstReg.WriteMask,                     /* writemask */
-               BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE,           /* msg_type */
+               msg_type,                                   /* msg_type */
                4,                                          /* response_length */
                shadow ? 6 : 4,                             /* msg_length */
-               0);                                         /* eot */
+               0,                                          /* eot */
+               1,
+               BRW_SAMPLER_SIMD_MODE_SIMD8);	
 
     if (shadow)
 	brw_MOV(p, dst[3], brw_imm_f(1.0));
@@ -2595,15 +2782,15 @@ static void post_wm_emit( struct brw_wm_compile *c )
 
 static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 {
-#define MAX_IFSN 32
+#define MAX_IF_DEPTH 32
 #define MAX_LOOP_DEPTH 32
-    struct brw_instruction *if_inst[MAX_IFSN], *loop_inst[MAX_LOOP_DEPTH];
-    struct brw_instruction *inst0, *inst1;
-    int i, if_insn = 0, loop_insn = 0;
+    struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH];
+    GLuint i, if_depth = 0, loop_depth = 0;
     struct brw_compile *p = &c->func;
     struct brw_indirect stack_index = brw_indirect(0, 0);
 
-    c->reg_index = 0;
+    c->out_of_regs = GL_FALSE;
+
     prealloc_reg(c);
     brw_set_compression_control(p, BRW_COMPRESSION_NONE);
     brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
@@ -2611,6 +2798,8 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     for (i = 0; i < c->nr_fp_insns; i++) {
         const struct prog_instruction *inst = &c->prog_instructions[i];
 
+        c->cur_inst = i;
+
 #if 0
         _mesa_printf("Inst %d: ", i);
         _mesa_print_instruction(inst);
@@ -2653,18 +2842,12 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 	    case WM_FRONTFACING:
 		emit_frontfacing(c, inst);
 		break;
-	    case OPCODE_ABS:
-		emit_abs(c, inst);
-		break;
 	    case OPCODE_ADD:
 		emit_add(c, inst);
 		break;
 	    case OPCODE_ARL:
 		emit_arl(c, inst);
 		break;
-	    case OPCODE_SUB:
-		emit_sub(c, inst);
-		break;
 	    case OPCODE_FRC:
 		emit_frc(c, inst);
 		break;
@@ -2770,15 +2953,15 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		emit_kil(c);
 		break;
 	    case OPCODE_IF:
-		assert(if_insn < MAX_IFSN);
-		if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+		assert(if_depth < MAX_IF_DEPTH);
+		if_inst[if_depth++] = brw_IF(p, BRW_EXECUTE_8);
 		break;
 	    case OPCODE_ELSE:
-		if_inst[if_insn-1]  = brw_ELSE(p, if_inst[if_insn-1]);
+		if_inst[if_depth-1]  = brw_ELSE(p, if_inst[if_depth-1]);
 		break;
 	    case OPCODE_ENDIF:
-		assert(if_insn > 0);
-		brw_ENDIF(p, if_inst[--if_insn]);
+		assert(if_depth > 0);
+		brw_ENDIF(p, if_inst[--if_depth]);
 		break;
 	    case OPCODE_BGNSUB:
 		brw_save_label(p, inst->Comment, p->nr_insn);
@@ -2812,7 +2995,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		break;
 	    case OPCODE_BGNLOOP:
                 /* XXX may need to invalidate the current_constant regs */
-		loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8);
+		loop_inst[loop_depth++] = brw_DO(p, BRW_EXECUTE_8);
 		break;
 	    case OPCODE_BRK:
 		brw_BREAK(p);
@@ -2823,25 +3006,34 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		brw_set_predicate_control(p, BRW_PREDICATE_NONE);
 		break;
 	    case OPCODE_ENDLOOP: 
-		loop_insn--;
-		inst0 = inst1 = brw_WHILE(p, loop_inst[loop_insn]);
-		/* patch all the BREAK instructions from
-		   last BEGINLOOP */
-		while (inst0 > loop_inst[loop_insn]) {
-		    inst0--;
-		    if (inst0->header.opcode == BRW_OPCODE_BREAK) {
-			inst0->bits3.if_else.jump_count = inst1 - inst0 + 1;
+               {
+                  struct brw_instruction *inst0, *inst1;
+                  GLuint br = 1;
+
+                  if (BRW_IS_IGDNG(brw))
+                     br = 2;
+ 
+                  loop_depth--;
+                  inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]);
+                  /* patch all the BREAK/CONT instructions from last BGNLOOP */
+                  while (inst0 > loop_inst[loop_depth]) {
+                     inst0--;
+                     if (inst0->header.opcode == BRW_OPCODE_BREAK) {
+			inst0->bits3.if_else.jump_count = br * (inst1 - inst0 + 1);
 			inst0->bits3.if_else.pop_count = 0;
-		    } else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
-                        inst0->bits3.if_else.jump_count = inst1 - inst0;
+                     }
+                     else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+                        inst0->bits3.if_else.jump_count = br * (inst1 - inst0);
                         inst0->bits3.if_else.pop_count = 0;
-                    }
-		}
-		break;
+                     }
+                  }
+               }
+               break;
 	    default:
 		_mesa_printf("unsupported IR in fragment shader %d\n",
 			inst->Opcode);
 	}
+
 	if (inst->CondUpdate)
 	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
 	else
@@ -2849,13 +3041,14 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     }
     post_wm_emit(c);
 
-    if (c->reg_index >= BRW_WM_MAX_GRF) {
-        _mesa_problem(NULL, "Ran out of registers in brw_wm_emit_glsl()");
-        /* XXX we need to do some proper error recovery here */
+    if (INTEL_DEBUG & DEBUG_WM) {
+      _mesa_printf("wm-native:\n");
+      for (i = 0; i < p->nr_insn; i++)
+	 brw_disasm(stderr, &p->store[i]);
+      _mesa_printf("\n");
     }
 }
 
-
 /**
  * Do GPU code generation for shaders that use GLSL features such as
  * flow control.  Other shaders will be compiled with the 
@@ -2876,6 +3069,6 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
-    c->prog_data.total_grf = c->reg_index;
+    c->prog_data.total_grf = num_grf_used(c);
     c->prog_data.total_scratch = 0;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_iz.c b/src/mesa/drivers/dri/i965/brw_wm_iz.c
index 7e2b1c79de..5e399ac62a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_iz.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_iz.c
@@ -116,6 +116,10 @@ const struct {
  { C, 0, 1, 1, 1 } 
 };
 
+/**
+ * \param line_aa  AA_NEVER, AA_ALWAYS or AA_SOMETIMES
+ * \param lookup  bitmask of IZ_* flags
+ */
 void brw_wm_lookup_iz( GLuint line_aa,
 		       GLuint lookup,
 		       GLboolean ps_uses_depth,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
index 92142764f5..6279258339 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
@@ -257,34 +257,6 @@ static void pass0_set_dst( struct brw_wm_compile *c,
 }
 
 
-static void pass0_set_dst_scalar( struct brw_wm_compile *c,
-				  struct brw_wm_instruction *out,
-                                  const struct prog_instruction *inst,
-				  GLuint writemask )
-{
-   if (writemask) {
-      const struct prog_dst_register *dst = &inst->DstReg;
-      GLuint i;
-
-      /* Compute only the first (X) value:
-       */
-      out->writemask = WRITEMASK_X;
-      out->dst[0] = get_value(c);
-
-      /* Update our tracking register file for all the components in
-       * writemask:
-       */
-      for (i = 0; i < 4; i++) {
-	 if (writemask & (1<<i)) {
-	    pass0_set_fpreg_value(c, dst->File, dst->Index, i, out->dst[0]);
-	 }
-      }
-   }
-   else
-      out->writemask = 0;
-}
-
-
 static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c,
 						    struct prog_src_register src,
 						    GLuint i )
@@ -363,10 +335,7 @@ translate_insn(struct brw_wm_compile *c,
 
    /* Dst:
     */
-   if (brw_wm_is_scalar_result(out->opcode)) 
-      pass0_set_dst_scalar(c, out, inst, writemask);
-   else 
-      pass0_set_dst(c, out, inst, writemask);
+   pass0_set_dst(c, out, inst, writemask);
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index ab9aa2f10d..3436a24717 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -159,6 +159,7 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       case OPCODE_FRC:
       case OPCODE_MOV:
       case OPCODE_SWZ:
+      case OPCODE_TRUNC:
 	 read0 = writemask;
 	 break;
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index 3fc18ff1f3..dff466587a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -103,6 +103,10 @@ struct wm_sampler_key {
       GLenum minfilter, magfilter;
       GLenum comparemode, comparefunc;
       dri_bo *sdc_bo;
+
+      /** If target is cubemap, take context setting.
+       */
+      GLboolean seamless_cube_map;
    } sampler[BRW_MAX_TEX_UNIT];
 };
 
@@ -169,30 +173,33 @@ static void brw_update_sampler_state(struct wm_sampler_entry *key,
       }  
    }
 
-   if (key->tex_target == GL_TEXTURE_CUBE_MAP &&
-       (key->minfilter != GL_NEAREST || key->magfilter != GL_NEAREST)) {
-      /* If we're using anything but nearest sampling for a cube map, we
-       * need to set this wrap mode to avoid GPU lock-ups.
-       */
-      sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
-      sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
-      sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
-   }
-   else if (key->tex_target == GL_TEXTURE_1D) {
+   sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
+   sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
+   sampler->ss1.t_wrap_mode = translate_wrap_mode(key->wrap_t);
+
+   /* Cube-maps on 965 and later must use the same wrap mode for all 3
+    * coordinate dimensions.  Futher, only CUBE and CLAMP are valid.
+    */
+   if (key->tex_target == GL_TEXTURE_CUBE_MAP) {
+      if (key->seamless_cube_map &&
+	  (key->minfilter != GL_NEAREST || key->magfilter != GL_NEAREST)) {
+	 sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+	 sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+	 sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+      } else {
+	 sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
+	 sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
+	 sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
+      }
+   } else if (key->tex_target == GL_TEXTURE_1D) {
       /* There's a bug in 1D texture sampling - it actually pays
        * attention to the wrap_t value, though it should not.
        * Override the wrap_t value here to GL_REPEAT to keep
        * any nonexistent border pixels from floating in.
        */
-      sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
-      sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
       sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_WRAP;
    }
-   else {
-      sampler->ss1.r_wrap_mode = translate_wrap_mode(key->wrap_r);
-      sampler->ss1.s_wrap_mode = translate_wrap_mode(key->wrap_s);
-      sampler->ss1.t_wrap_mode = translate_wrap_mode(key->wrap_t);
-   }
+
 
    /* Set shadow function: 
     */
@@ -249,6 +256,9 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
 
          entry->tex_target = texObj->Target;
 
+	 entry->seamless_cube_map = (texObj->Target == GL_TEXTURE_CUBE_MAP)
+	    ? ctx->Texture.CubeMapSeamless : GL_FALSE;
+
 	 entry->wrap_r = texObj->WrapR;
 	 entry->wrap_s = texObj->WrapS;
 	 entry->wrap_t = texObj->WrapT;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 67b41173fb..39f8c6d522 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -71,7 +71,9 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
       key->max_threads = 1;
    else {
       /* WM maximum threads is number of EUs times number of threads per EU. */
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_IGDNG(brw))
+         key->max_threads = 12 * 6;
+      else if (BRW_IS_G4X(brw))
 	 key->max_threads = 10 * 5;
       else
 	 key->max_threads = 8 * 4;
@@ -141,7 +143,11 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
    wm.thread0.kernel_start_pointer = brw->wm.prog_bo->offset >> 6; /* reloc */
    wm.thread1.depth_coef_urb_read_offset = 1;
    wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
-   wm.thread1.binding_table_entry_count = key->nr_surfaces;
+
+   if (BRW_IS_IGDNG(brw))
+      wm.thread1.binding_table_entry_count = 0; /* hardware requirement */
+   else
+      wm.thread1.binding_table_entry_count = key->nr_surfaces;
 
    if (key->total_scratch != 0) {
       wm.thread2.scratch_space_base_pointer =
@@ -158,7 +164,11 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
    wm.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
    wm.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
 
-   wm.wm4.sampler_count = (key->sampler_count + 1) / 4;
+   if (BRW_IS_IGDNG(brw)) 
+      wm.wm4.sampler_count = 0; /* hardware requirement */
+   else
+      wm.wm4.sampler_count = (key->sampler_count + 1) / 4;
+
    if (brw->wm.sampler_bo != NULL) {
       /* reloc */
       wm.wm4.sampler_state_pointer = brw->wm.sampler_bo->offset >> 5;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 805df8a4af..3dcc592bde 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -176,22 +176,6 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format,
    }
 }
 
-
-/**
- * Use same key for WM and VS surfaces.
- */
-struct brw_surface_key {
-   GLenum target, depthmode;
-   dri_bo *bo;
-   GLint format, internal_format;
-   GLint first_level, last_level;
-   GLint width, height, depth;
-   GLint pitch, cpp;
-   uint32_t tiling;
-   GLuint offset;
-};
-
-
 static void
 brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling)
 {
@@ -268,7 +252,7 @@ brw_create_texture_surface( struct brw_context *brw,
       surf.ss0.cube_neg_z = 1;
    }
 
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -321,10 +305,11 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    key.tiling = intelObj->mt->region->tiling;
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                         &key, sizeof(key),
-                                         &key.bo, key.bo ? 1 : 0,
-                                         NULL);
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_texture_surface(brw, &key);
    }
@@ -336,7 +321,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
  * Create the constant buffer surface.  Vertex/fragment shader constants will be
  * read from this buffer with Data Port Read instructions/messages.
  */
-static dri_bo *
+dri_bo *
 brw_create_constant_surface( struct brw_context *brw,
                              struct brw_surface_key *key )
 {
@@ -362,7 +347,7 @@ brw_create_constant_surface( struct brw_context *brw,
    surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */
    brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */
  
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -380,39 +365,70 @@ brw_create_constant_surface( struct brw_context *brw,
    return bo;
 }
 
+/* Creates a new WM constant buffer reflecting the current fragment program's
+ * constants, if needed by the fragment program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_wm_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params = fp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   if (!fp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
 
 /**
  * Update the surface state for a WM constant buffer.
  * The constant buffer will be (re)allocated here if needed.
  */
-static dri_bo *
+static void
 brw_update_wm_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+                                GLuint surf)
 {
    struct brw_context *brw = brw_context(ctx);
    struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params =
+      fp->program.Base.Parameters;
 
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
+   /* If we're in this state update atom, we need to update WM constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer", size, 64);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+      brw->wm.surf_bo[surf] = NULL;
+      return;
    }
 
    memset(&key, 0, sizeof(key));
 
    key.format = MESA_FORMAT_RGBA_FLOAT32;
    key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
+   key.bo = fp->const_buffer;
    key.depthmode = GL_NONE;
    key.pitch = params->NumParameters;
    key.width = params->NumParameters;
@@ -427,77 +443,59 @@ brw_update_wm_constant_surface( GLcontext *ctx,
    */
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
                                             &key, sizeof(key),
                                             &key.bo, key.bo ? 1 : 0,
                                             NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_constant_surface(brw, &key);
    }
-
-   return const_buffer;
+   brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
 }
 
-
 /**
- * Update the surface state for a VS constant buffer.
- * The constant buffer will be (re)allocated here if needed.
+ * Updates surface / buffer for fragment shader constant buffer, if
+ * one is required.
+ *
+ * This consumes the state updates for the constant buffer, and produces
+ * BRW_NEW_WM_SURFACES to get picked up by brw_prepare_wm_surfaces for
+ * inclusion in the binding table.
  */
-static dri_bo *
-brw_update_vs_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+static void prepare_wm_constant_surface(struct brw_context *brw )
 {
-   struct brw_context *brw = brw_context(ctx);
-   struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   assert(surf == 0);
-
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
-
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer", size, 64);
-   }
-
-   memset(&key, 0, sizeof(key));
-
-   key.format = MESA_FORMAT_RGBA_FLOAT32;
-   key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
-   key.depthmode = GL_NONE;
-   key.pitch = params->NumParameters;
-   key.width = params->NumParameters;
-   key.height = 1;
-   key.depth = 1;
-   key.cpp = 16;
+   GLcontext *ctx = &brw->intel.ctx;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
 
-   /*
-   printf("%s:\n", __FUNCTION__);
-   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
-          key.width, key.height, key.depth, key.cpp, key.pitch);
-   */
+   drm_intel_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   dri_bo_unreference(brw->vs.surf_bo[surf]);
-   brw->vs.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                            &key, sizeof(key),
-                                            &key.bo, key.bo ? 1 : 0,
-                                            NULL);
-   if (brw->vs.surf_bo[surf] == NULL) {
-      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      if (brw->wm.surf_bo[surf] != NULL) {
+	 drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+	 brw->wm.surf_bo[surf] = NULL;
+	 brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
+      }
+      return;
    }
 
-   return const_buffer;
+   brw_update_wm_constant_surface(ctx, surf);
 }
 
+const struct brw_tracked_state brw_wm_constant_surface = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_FRAGMENT_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_wm_constant_surface,
+};
+
 
 /**
  * Sets up a surface state structure to point at the given region.
@@ -507,7 +505,7 @@ brw_update_vs_constant_surface( GLcontext *ctx,
 static void
 brw_update_renderbuffer_surface(struct brw_context *brw,
 				struct gl_renderbuffer *rb,
-				unsigned int unit, GLboolean cached)
+				unsigned int unit)
 {
    GLcontext *ctx = &brw->intel.ctx;
    dri_bo *region_bo = NULL;
@@ -567,12 +565,11 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 		      ctx->Color.BlendEnabled);
 
    dri_bo_unreference(brw->wm.surf_bo[unit]);
-   brw->wm.surf_bo[unit] = NULL;
-   if (cached) 
-       brw->wm.surf_bo[unit] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-	       &key, sizeof(key),
-	       &region_bo, 1,
-	       NULL);
+   brw->wm.surf_bo[unit] = brw_search_cache(&brw->surface_cache,
+					    BRW_SS_SURFACE,
+					    &key, sizeof(key),
+					    &region_bo, 1,
+					    NULL);
 
    if (brw->wm.surf_bo[unit] == NULL) {
       struct brw_surface_state surf;
@@ -581,7 +578,27 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 
       surf.ss0.surface_format = key.surface_format;
       surf.ss0.surface_type = key.surface_type;
-      surf.ss1.base_addr =  key.draw_offset;
+      if (key.tiling == I915_TILING_NONE) {
+	 surf.ss1.base_addr = key.draw_offset;
+      } else {
+	 uint32_t tile_offset = key.draw_offset % 4096;
+
+	 surf.ss1.base_addr = key.draw_offset - tile_offset;
+
+	 assert(BRW_IS_G4X(brw) || tile_offset == 0);
+	 if (BRW_IS_G4X(brw)) {
+	    if (key.tiling == I915_TILING_X) {
+	       /* Note that the low bits of these fields are missing, so
+		* there's the possibility of getting in trouble.
+		*/
+	       surf.ss5.x_offset = (tile_offset % 512) / key.cpp / 4;
+	       surf.ss5.y_offset = tile_offset / 512 / 2;
+	    } else {
+	       surf.ss5.x_offset = (tile_offset % 128) / key.cpp / 4;
+	       surf.ss5.y_offset = tile_offset / 128 / 2;
+	    }
+	 }
+      }
       if (region_bo != NULL)
 	 surf.ss1.base_addr += region_bo->offset; /* reloc */
 
@@ -598,7 +615,8 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       surf.ss0.writedisable_alpha = !key.color_mask[3];
 
       /* Key size will never match key size for textures, so we're safe. */
-      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->surface_cache,
+                                               BRW_SS_SURFACE,
                                                &key, sizeof(key),
 					       &region_bo, 1,
 					       &surf, sizeof(surf),
@@ -611,7 +629,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 	 drm_intel_bo_emit_reloc(brw->wm.surf_bo[unit],
 				 offsetof(struct brw_surface_state, ss1),
 				 region_bo,
-				 key.draw_offset,
+				 surf.ss1.base_addr - region_bo->offset,
 				 I915_GEM_DOMAIN_RENDER,
 				 I915_GEM_DOMAIN_RENDER);
       }
@@ -630,7 +648,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
 
    assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF);
 
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
 			      NULL, 0,
 			      brw->wm.surf_bo, brw->wm.nr_surfaces,
 			      NULL);
@@ -646,7 +664,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
          else
             data[i] = 0;
 
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
 				  NULL, 0,
 				  brw->wm.surf_bo, brw->wm.nr_surfaces,
 				  data, data_size,
@@ -682,27 +700,17 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
          brw_update_renderbuffer_surface(brw,
 					 ctx->DrawBuffer->_ColorDrawBuffers[i],
-					 i,
-					 GL_FALSE);
+					 i);
       }
    } else {
-      brw_update_renderbuffer_surface(brw, NULL, 0, GL_TRUE);
+      brw_update_renderbuffer_surface(brw, NULL, 0);
    }
 
    old_nr_surfaces = brw->wm.nr_surfaces;
    brw->wm.nr_surfaces = MAX_DRAW_BUFFERS;
 
-   /* Update surface / buffer for fragment shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
-      struct brw_fragment_program *fp =
-         (struct brw_fragment_program *) brw->fragment_program;
-      fp->const_buffer =
-         brw_update_wm_constant_surface(ctx, surf, fp->const_buffer,
-                                     fp->program.Base.Parameters);
-
-      brw->wm.nr_surfaces = surf + 1;
-   }
+   if (brw->wm.surf_bo[SURF_INDEX_FRAG_CONST_BUFFER] != NULL)
+       brw->wm.nr_surfaces = SURF_INDEX_FRAG_CONST_BUFFER + 1;
 
    /* Update surfaces for textures */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
@@ -735,100 +743,16 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES;
 }
 
-
-/**
- * Constructs the binding table for the VS surface state.
- */
-static dri_bo *
-brw_vs_get_binding_table(struct brw_context *brw)
-{
-   dri_bo *bind_bo;
-
-   assert(brw->vs.nr_surfaces <= BRW_VS_MAX_SURF);
-
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
-			      NULL, 0,
-			      brw->vs.surf_bo, brw->vs.nr_surfaces,
-			      NULL);
-
-   if (bind_bo == NULL) {
-      GLuint data_size = brw->vs.nr_surfaces * sizeof(GLuint);
-      uint32_t *data = malloc(data_size);
-      int i;
-
-      for (i = 0; i < brw->vs.nr_surfaces; i++)
-         if (brw->vs.surf_bo[i])
-            data[i] = brw->vs.surf_bo[i]->offset;
-         else
-            data[i] = 0;
-
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
-				  NULL, 0,
-				  brw->vs.surf_bo, brw->vs.nr_surfaces,
-				  data, data_size,
-				  NULL, NULL);
-
-      /* Emit binding table relocations to surface state */
-      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
-	 if (brw->vs.surf_bo[i] != NULL) {
-	    dri_bo_emit_reloc(bind_bo,
-			      I915_GEM_DOMAIN_INSTRUCTION, 0,
-			      0,
-			      i * sizeof(GLuint),
-			      brw->vs.surf_bo[i]);
-	 }
-      }
-
-      free(data);
-   }
-
-   return bind_bo;
-}
-
-
-/**
- * Vertex shader surfaces.  Just constant buffer for now.  Could add vertex 
- * shader textures in the future.
- */
-static void prepare_vs_surfaces(struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   /* Update surface / buffer for vertex shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_VERT_CONST_BUFFER;
-      struct brw_vertex_program *vp =
-         (struct brw_vertex_program *) brw->vertex_program;
-      vp->const_buffer =
-         brw_update_vs_constant_surface(ctx, surf, vp->const_buffer,
-                                        vp->program.Base.Parameters);
-
-      brw->vs.nr_surfaces = 1;
-   }
-
-   dri_bo_unreference(brw->vs.bind_bo);
-   brw->vs.bind_bo = brw_vs_get_binding_table(brw);
-
-   if (1)
-      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
-}
-
-
-static void
-prepare_surfaces(struct brw_context *brw)
-{
-   prepare_wm_surfaces(brw);
-   prepare_vs_surfaces(brw);
-}
-
-
 const struct brw_tracked_state brw_wm_surfaces = {
    .dirty = {
-      .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS | _NEW_PROGRAM,
-      .brw = BRW_NEW_CONTEXT,
+      .mesa = (_NEW_COLOR |
+               _NEW_TEXTURE |
+               _NEW_BUFFERS),
+      .brw = (BRW_NEW_CONTEXT |
+	      BRW_NEW_WM_SURFACES),
       .cache = 0
    },
-   .prepare = prepare_surfaces,
+   .prepare = prepare_wm_surfaces,
 };
 
 
diff --git a/src/mesa/drivers/dri/i965/intel_generatemipmap.c b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
new file mode 120000
index 0000000000..4c6b37ada0
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_syncobj.c b/src/mesa/drivers/dri/i965/intel_syncobj.c
new file mode 120000
index 0000000000..0b2e56ab24
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_syncobj.c
@@ -0,0 +1 @@
+../intel/intel_syncobj.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 82a92a9f45..6aa36d10b1 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -195,7 +195,11 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 {
    struct intel_context *intel = batch->intel;
    GLuint used = batch->ptr - batch->map;
-   GLboolean was_locked = intel->locked;
+
+   if (intel->first_post_swapbuffers_batch == NULL) {
+      intel->first_post_swapbuffers_batch = intel->batch->buf;
+      drm_intel_bo_reference(intel->first_post_swapbuffers_batch);
+   }
 
    if (intel->first_post_swapbuffers_batch == NULL) {
       intel->first_post_swapbuffers_batch = intel->batch->buf;
@@ -248,13 +252,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
    /* TODO: Just pass the relocation list and dma buffer up to the
     * kernel.
     */
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
+   LOCK_HARDWARE(intel);
    do_flush_locked(batch, used, GL_FALSE);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   UNLOCK_HARDWARE(intel);
 
    if (INTEL_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "waiting for idle\n");
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index 4919828131..0c5be4c798 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -108,6 +108,8 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
 	 CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       }
 
+      assert(src->tiling != I915_TILING_Y);
+      assert(dst->tiling != I915_TILING_Y);
 #ifndef I915
       if (src->tiling != I915_TILING_NONE) {
 	 CMD |= XY_SRC_TILED;
@@ -175,66 +177,6 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
    UNLOCK_HARDWARE(intel);
 }
 
-
-
-
-void
-intelEmitFillBlit(struct intel_context *intel,
-		  GLuint cpp,
-		  GLshort dst_pitch,
-		  dri_bo *dst_buffer,
-		  GLuint dst_offset,
-		  uint32_t dst_tiling,
-		  GLshort x, GLshort y,
-		  GLshort w, GLshort h,
-		  GLuint color)
-{
-   GLuint BR13, CMD;
-   BATCH_LOCALS;
-
-   dst_pitch *= cpp;
-
-   switch (cpp) {
-   case 1:
-      BR13 = (0xF0 << 16);
-      CMD = XY_COLOR_BLT_CMD;
-      break;
-   case 2:
-      BR13 = (0xF0 << 16) | BR13_565;
-      CMD = XY_COLOR_BLT_CMD;
-      break;
-   case 4:
-      BR13 = (0xF0 << 16) | BR13_8888;
-      CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
-      break;
-   default:
-      return;
-   }
-#ifndef I915
-   if (dst_tiling != I915_TILING_NONE) {
-      CMD |= XY_DST_TILED;
-      dst_pitch /= 4;
-   }
-#endif
-
-   DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
-       __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h);
-
-   assert(w > 0);
-   assert(h > 0);
-
-   BEGIN_BATCH(6, NO_LOOP_CLIPRECTS);
-   OUT_BATCH(CMD);
-   OUT_BATCH(BR13 | dst_pitch);
-   OUT_BATCH((y << 16) | x);
-   OUT_BATCH(((y + h) << 16) | (x + w));
-   OUT_RELOC(dst_buffer,
-	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-	     dst_offset);
-   OUT_BATCH(color);
-   ADVANCE_BATCH();
-}
-
 static GLuint translate_raster_op(GLenum logicop)
 {
    switch(logicop) {
@@ -261,7 +203,7 @@ static GLuint translate_raster_op(GLenum logicop)
 
 /* Copy BitBlt
  */
-void
+GLboolean
 intelEmitCopyBlit(struct intel_context *intel,
 		  GLuint cpp,
 		  GLshort src_pitch,
@@ -283,6 +225,19 @@ intelEmitCopyBlit(struct intel_context *intel,
    dri_bo *aper_array[3];
    BATCH_LOCALS;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+   if (src_tiling != I915_TILING_NONE) {
+      if (src_offset & 4095)
+	 return GL_FALSE;
+      if (src_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    /* do space/cliprects check before going any further */
    do {
        aper_array[0] = intel->batch->buf;
@@ -297,12 +252,7 @@ intelEmitCopyBlit(struct intel_context *intel,
    } while (pass < 2);
 
    if (pass >= 2) {
-       GLboolean locked = GL_FALSE;       
-       if (!intel->locked) {
-           LOCK_HARDWARE(intel);
-           locked = GL_TRUE;
-       }
-
+       LOCK_HARDWARE(intel);
        dri_bo_map(dst_buffer, GL_TRUE);
        dri_bo_map(src_buffer, GL_FALSE);
        _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
@@ -316,11 +266,9 @@ intelEmitCopyBlit(struct intel_context *intel,
        
        dri_bo_unmap(src_buffer);
        dri_bo_unmap(dst_buffer);
-       
-       if (locked)
-           UNLOCK_HARDWARE(intel);
+       UNLOCK_HARDWARE(intel);
 
-       return;
+       return GL_TRUE;
    }
 
    intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
@@ -347,7 +295,7 @@ intelEmitCopyBlit(struct intel_context *intel,
       CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       break;
    default:
-      return;
+      return GL_FALSE;
    }
 
 #ifndef I915
@@ -362,7 +310,7 @@ intelEmitCopyBlit(struct intel_context *intel,
 #endif
 
    if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
-      return;
+      return GL_TRUE;
    }
 
    assert(dst_x < dst_x2);
@@ -384,6 +332,8 @@ intelEmitCopyBlit(struct intel_context *intel,
    ADVANCE_BATCH();
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
 }
 
 
@@ -527,6 +477,8 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                   BR13 |= BR13_565;
                }
 
+	       assert(irb->region->tiling != I915_TILING_Y);
+
 #ifndef I915
 	       if (irb->region->tiling != I915_TILING_NONE) {
 		  CMD |= XY_DST_TILED;
@@ -596,7 +548,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
    UNLOCK_HARDWARE(intel);
 }
 
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
@@ -612,11 +564,19 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
    int dwords = ALIGN(src_size, 8) / 4;
    uint32_t opcode, br13, blit_cmd;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    assert( logic_op - GL_CLEAR >= 0 );
    assert( logic_op - GL_CLEAR < 0x10 );
+   assert(dst_pitch > 0);
 
    if (w < 0 || h < 0)
-      return;
+      return GL_TRUE;
 
    dst_pitch *= cpp;
 
@@ -673,4 +633,46 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 			   REFERENCES_CLIPRECTS );
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
+}
+
+/* We don't have a memmove-type blit like some other hardware, so we'll do a
+ * rectangular blit covering a large space, then emit 1-scanline blit at the
+ * end to cover the last if we need.
+ */
+void
+intel_emit_linear_blit(struct intel_context *intel,
+		       drm_intel_bo *dst_bo,
+		       unsigned int dst_offset,
+		       drm_intel_bo *src_bo,
+		       unsigned int src_offset,
+		       unsigned int size)
+{
+   GLuint pitch, height;
+
+   /* The pitch is a signed value. */
+   pitch = MIN2(size, (1 << 15) - 1);
+   height = size / pitch;
+   intelEmitCopyBlit(intel, 1,
+		     pitch, src_bo, src_offset, I915_TILING_NONE,
+		     pitch, dst_bo, dst_offset, I915_TILING_NONE,
+		     0, 0, /* src x/y */
+		     0, 0, /* dst x/y */
+		     pitch, height, /* w, h */
+		     GL_COPY);
+
+   src_offset += pitch * height;
+   dst_offset += pitch * height;
+   size -= pitch * height;
+   assert (size < (1 << 15));
+   if (size != 0) {
+      intelEmitCopyBlit(intel, 1,
+			size, src_bo, src_offset, I915_TILING_NONE,
+			size, dst_bo, dst_offset, I915_TILING_NONE,
+			0, 0, /* src x/y */
+			0, 0, /* dst x/y */
+			size, 1, /* w, h */
+			GL_COPY);
+   }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_blit.h b/src/mesa/drivers/dri/intel/intel_blit.h
index 52065b13ed..240cb7cd1b 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.h
+++ b/src/mesa/drivers/dri/intel/intel_blit.h
@@ -35,7 +35,8 @@ extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv,
 
 extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask);
 
-extern void intelEmitCopyBlit(struct intel_context *intel,
+GLboolean
+intelEmitCopyBlit(struct intel_context *intel,
                               GLuint cpp,
                               GLshort src_pitch,
                               dri_bo *src_buffer,
@@ -50,16 +51,7 @@ extern void intelEmitCopyBlit(struct intel_context *intel,
                               GLshort w, GLshort h,
 			      GLenum logicop );
 
-extern void intelEmitFillBlit(struct intel_context *intel,
-                              GLuint cpp,
-                              GLshort dst_pitch,
-                              dri_bo *dst_buffer,
-                              GLuint dst_offset,
-			      uint32_t dst_tiling,
-                              GLshort x, GLshort y,
-                              GLshort w, GLshort h, GLuint color);
-
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
@@ -71,5 +63,11 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLshort x, GLshort y,
 				  GLshort w, GLshort h,
 				  GLenum logic_op);
+void intel_emit_linear_blit(struct intel_context *intel,
+			    drm_intel_bo *dst_bo,
+			    unsigned int dst_offset,
+			    drm_intel_bo *src_bo,
+			    unsigned int src_offset,
+			    unsigned int size);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index 2e6b77824d..c55c5c426e 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -28,13 +28,18 @@
 
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "main/macros.h"
 #include "main/bufferobj.h"
 
 #include "intel_context.h"
+#include "intel_blit.h"
 #include "intel_buffer_objects.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
 
+static GLboolean
+intel_bufferobj_unmap(GLcontext * ctx,
+                      GLenum target, struct gl_buffer_object *obj);
 
 /** Allocates a new dri_bo to store the data for the buffer object. */
 static void
@@ -100,7 +105,13 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
 
    assert(intel_obj);
-   assert(!obj->Pointer); /* Mesa should have unmapped it */
+
+   /* Buffer objects are automatically unmapped when deleting according
+    * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy
+    * (though it does if you call glDeleteBuffers)
+    */
+   if (obj->Pointer)
+      intel_bufferobj_unmap(ctx, 0, obj);
 
    _mesa_free(intel_obj->sys_buffer);
    if (intel_obj->region) {
@@ -119,9 +130,10 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
  * Allocate space for and store data in a buffer object.  Any data that was
  * previously stored in the buffer object is lost.  If data is NULL,
  * memory will be allocated, but no copy will occur.
- * Called via glBufferDataARB().
+ * Called via ctx->Driver.BufferData().
+ * \return GL_TRUE for success, GL_FALSE if out of memory
  */
-static void
+static GLboolean
 intel_bufferobj_data(GLcontext * ctx,
                      GLenum target,
                      GLsizeiptrARB size,
@@ -156,15 +168,19 @@ intel_bufferobj_data(GLcontext * ctx,
 	 if (intel_obj->sys_buffer != NULL) {
 	    if (data != NULL)
 	       memcpy(intel_obj->sys_buffer, data, size);
-	    return;
+	    return GL_TRUE;
 	 }
       }
 #endif
       intel_bufferobj_alloc_buffer(intel, intel_obj);
+      if (!intel_obj->buffer)
+         return GL_FALSE;
 
       if (data != NULL)
 	 dri_bo_subdata(intel_obj->buffer, 0, size, data);
    }
+
+   return GL_TRUE;
 }
 
 
@@ -234,6 +250,11 @@ intel_bufferobj_map(GLcontext * ctx,
       return obj->Pointer;
    }
 
+   /* Flush any existing batchbuffer that might have written to this
+    * buffer.
+    */
+   intelFlush(ctx);
+
    if (intel_obj->region)
       intel_bufferobj_cow(intel, intel_obj);
 
@@ -251,32 +272,205 @@ intel_bufferobj_map(GLcontext * ctx,
    }
 
    obj->Pointer = intel_obj->buffer->virtual;
+   obj->Length = obj->Size;
+   obj->Offset = 0;
+
+   return obj->Pointer;
+}
+
+/**
+ * Called via glMapBufferRange().
+ *
+ * The goal of this extension is to allow apps to accumulate their rendering
+ * at the same time as they accumulate their buffer object.  Without it,
+ * you'd end up blocking on execution of rendering every time you mapped
+ * the buffer to put new data in.
+ *
+ * We support it in 3 ways: If unsynchronized, then don't bother
+ * flushing the batchbuffer before mapping the buffer, which can save blocking
+ * in many cases.  If we would still block, and they allow the whole buffer
+ * to be invalidated, then just allocate a new buffer to replace the old one.
+ * If not, and we'd block, and they allow the subrange of the buffer to be
+ * invalidated, then we can make a new little BO, let them write into that,
+ * and blit it into the real BO at unmap time.
+ */
+static void *
+intel_bufferobj_map_range(GLcontext * ctx,
+			  GLenum target, GLintptr offset, GLsizeiptr length,
+			  GLbitfield access, struct gl_buffer_object *obj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+
+   assert(intel_obj);
+
+   /* _mesa_MapBufferRange (GL entrypoint) sets these, but the vbo module also
+    * internally uses our functions directly.
+    */
+   obj->Offset = offset;
+   obj->Length = length;
+   obj->AccessFlags = access;
+
+   if (intel_obj->sys_buffer) {
+      obj->Pointer = intel_obj->sys_buffer + offset;
+      return obj->Pointer;
+   }
+
+   if (intel_obj->region)
+      intel_bufferobj_cow(intel, intel_obj);
+
+   /* If the mapping is synchronized with other GL operations, flush
+    * the batchbuffer so that GEM knows about the buffer access for later
+    * syncing.
+    */
+   if (!(access & GL_MAP_UNSYNCHRONIZED_BIT))
+      intelFlush(ctx);
+
+   if (intel_obj->buffer == NULL) {
+      obj->Pointer = NULL;
+      return NULL;
+   }
+
+   /* If the user doesn't care about existing buffer contents and mapping
+    * would cause us to block, then throw out the old buffer.
+    */
+   if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) &&
+       (access & GL_MAP_INVALIDATE_BUFFER_BIT) &&
+       drm_intel_bo_busy(intel_obj->buffer)) {
+      drm_intel_bo_unreference(intel_obj->buffer);
+      intel_obj->buffer = dri_bo_alloc(intel->bufmgr, "bufferobj",
+				       intel_obj->Base.Size, 64);
+   }
+
+   /* If the user is mapping a range of an active buffer object but
+    * doesn't require the current contents of that range, make a new
+    * BO, and we'll copy what they put in there out at unmap or
+    * FlushRange time.
+    */
+   if ((access & GL_MAP_INVALIDATE_RANGE_BIT) &&
+       drm_intel_bo_busy(intel_obj->buffer)) {
+      if (access & GL_MAP_FLUSH_EXPLICIT_BIT) {
+	 intel_obj->range_map_buffer = _mesa_malloc(length);
+	 obj->Pointer = intel_obj->range_map_buffer;
+      } else {
+	 intel_obj->range_map_bo = drm_intel_bo_alloc(intel->bufmgr,
+						      "range map",
+						      length, 64);
+	 if (!(access & GL_MAP_READ_BIT) &&
+	     intel->intelScreen->kernel_exec_fencing) {
+	    drm_intel_gem_bo_map_gtt(intel_obj->range_map_bo);
+	    intel_obj->mapped_gtt = GL_TRUE;
+	 } else {
+	    drm_intel_bo_map(intel_obj->range_map_bo,
+			     (access & GL_MAP_WRITE_BIT) != 0);
+	    intel_obj->mapped_gtt = GL_FALSE;
+	 }
+	 obj->Pointer = intel_obj->range_map_bo->virtual;
+      }
+      return obj->Pointer;
+   }
+
+   if (!(access & GL_MAP_READ_BIT) &&
+       intel->intelScreen->kernel_exec_fencing) {
+      drm_intel_gem_bo_map_gtt(intel_obj->buffer);
+      intel_obj->mapped_gtt = GL_TRUE;
+   } else {
+      drm_intel_bo_map(intel_obj->buffer, (access & GL_MAP_WRITE_BIT) != 0);
+      intel_obj->mapped_gtt = GL_FALSE;
+   }
+
+   obj->Pointer = intel_obj->buffer->virtual + offset;
    return obj->Pointer;
 }
 
+/* Ideally we'd use a BO to avoid taking up cache space for the temporary
+ * data, but FlushMappedBufferRange may be followed by further writes to
+ * the pointer, so we would have to re-map after emitting our blit, which
+ * would defeat the point.
+ */
+static void
+intel_bufferobj_flush_mapped_range(GLcontext *ctx, GLenum target,
+				   GLintptr offset, GLsizeiptr length,
+				   struct gl_buffer_object *obj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+   drm_intel_bo *temp_bo;
+
+   /* Unless we're in the range map using a temporary system buffer,
+    * there's no work to do.
+    */
+   if (intel_obj->range_map_buffer == NULL)
+      return;
+
+   temp_bo = drm_intel_bo_alloc(intel->bufmgr, "range map flush", length, 64);
+
+   drm_intel_bo_subdata(temp_bo, 0, length, intel_obj->range_map_buffer);
+
+   intel_emit_linear_blit(intel,
+			  intel_obj->buffer, obj->Offset + offset,
+			  temp_bo, 0,
+			  length);
+
+   drm_intel_bo_unreference(temp_bo);
+}
+
 
 /**
- * Called via glMapBufferARB().
+ * Called via glUnmapBuffer().
  */
 static GLboolean
 intel_bufferobj_unmap(GLcontext * ctx,
                       GLenum target, struct gl_buffer_object *obj)
 {
+   struct intel_context *intel = intel_context(ctx);
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
 
    assert(intel_obj);
+   assert(obj->Pointer);
    if (intel_obj->sys_buffer != NULL) {
-      assert(obj->Pointer);
-      obj->Pointer = NULL;
+      /* always keep the mapping around. */
+   } else if (intel_obj->range_map_buffer != NULL) {
+      /* Since we've emitted some blits to buffers that will (likely) be used
+       * in rendering operations in other cache domains in this batch, emit a
+       * flush.  Once again, we wish for a domain tracker in libdrm to cover
+       * usage inside of a batchbuffer.
+       */
+      intel_batchbuffer_emit_mi_flush(intel->batch);
+      free(intel_obj->range_map_buffer);
+      intel_obj->range_map_buffer = NULL;
+   } else if (intel_obj->range_map_bo != NULL) {
+      if (intel_obj->mapped_gtt) {
+	 drm_intel_gem_bo_unmap_gtt(intel_obj->range_map_bo);
+      } else {
+	 drm_intel_bo_unmap(intel_obj->range_map_bo);
+      }
+
+      intel_emit_linear_blit(intel,
+			     intel_obj->buffer, obj->Offset,
+			     intel_obj->range_map_bo, 0,
+			     obj->Length);
+
+      /* Since we've emitted some blits to buffers that will (likely) be used
+       * in rendering operations in other cache domains in this batch, emit a
+       * flush.  Once again, we wish for a domain tracker in libdrm to cover
+       * usage inside of a batchbuffer.
+       */
+      intel_batchbuffer_emit_mi_flush(intel->batch);
+
+      drm_intel_bo_unreference(intel_obj->range_map_bo);
+      intel_obj->range_map_bo = NULL;
    } else if (intel_obj->buffer != NULL) {
-      assert(obj->Pointer);
       if (intel_obj->mapped_gtt) {
 	 drm_intel_gem_bo_unmap_gtt(intel_obj->buffer);
       } else {
 	 drm_intel_bo_unmap(intel_obj->buffer);
       }
-      obj->Pointer = NULL;
    }
+   obj->Pointer = NULL;
+   obj->Offset = 0;
+   obj->Length = 0;
+
    return GL_TRUE;
 }
 
@@ -294,30 +488,94 @@ intel_bufferobj_buffer(struct intel_context *intel,
    }
 
    if (intel_obj->buffer == NULL) {
+      void *sys_buffer = intel_obj->sys_buffer;
+
+      /* only one of buffer and sys_buffer could be non-NULL */
       intel_bufferobj_alloc_buffer(intel, intel_obj);
+      intel_obj->sys_buffer = NULL;
+
       intel_bufferobj_subdata(&intel->ctx,
 			      GL_ARRAY_BUFFER_ARB,
 			      0,
 			      intel_obj->Base.Size,
-			      intel_obj->sys_buffer,
+			      sys_buffer,
 			      &intel_obj->Base);
-      _mesa_free(intel_obj->sys_buffer);
+      _mesa_free(sys_buffer);
       intel_obj->sys_buffer = NULL;
    }
 
    return intel_obj->buffer;
 }
 
+static void
+intel_bufferobj_copy_subdata(GLcontext *ctx,
+			     struct gl_buffer_object *src,
+			     struct gl_buffer_object *dst,
+			     GLintptr read_offset, GLintptr write_offset,
+			     GLsizeiptr size)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_src = intel_buffer_object(src);
+   struct intel_buffer_object *intel_dst = intel_buffer_object(dst);
+   drm_intel_bo *src_bo, *dst_bo;
+
+   if (size == 0)
+      return;
+
+   /* If we're in system memory, just map and memcpy. */
+   if (intel_src->sys_buffer || intel_dst->sys_buffer) {
+      /* The same buffer may be used, but note that regions copied may
+       * not overlap.
+       */
+      if (src == dst) {
+	 char *ptr = intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
+					 GL_READ_WRITE, dst);
+	 memcpy(ptr + write_offset, ptr + read_offset, size);
+	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
+      } else {
+	 const char *src_ptr;
+	 char *dst_ptr;
+
+	 src_ptr =  intel_bufferobj_map(ctx, GL_COPY_READ_BUFFER,
+					GL_READ_ONLY, src);
+	 dst_ptr =  intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
+					GL_WRITE_ONLY, dst);
+
+	 memcpy(dst_ptr + write_offset, src_ptr + read_offset, size);
+
+	 intel_bufferobj_unmap(ctx, GL_COPY_READ_BUFFER, src);
+	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
+      }
+   }
+
+   /* Otherwise, we have real BOs, so blit them. */
+
+   dst_bo = intel_bufferobj_buffer(intel, intel_dst, INTEL_WRITE_PART);
+   src_bo = intel_bufferobj_buffer(intel, intel_src, INTEL_READ);
+
+   intel_emit_linear_blit(intel,
+			  dst_bo, write_offset,
+			  src_bo, read_offset, size);
+
+   /* Since we've emitted some blits to buffers that will (likely) be used
+    * in rendering operations in other cache domains in this batch, emit a
+    * flush.  Once again, we wish for a domain tracker in libdrm to cover
+    * usage inside of a batchbuffer.
+    */
+   intel_batchbuffer_emit_mi_flush(intel->batch);
+}
+
 void
-intel_bufferobj_init(struct intel_context *intel)
+intelInitBufferObjectFuncs(struct dd_function_table *functions)
 {
-   GLcontext *ctx = &intel->ctx;
-
-   ctx->Driver.NewBufferObject = intel_bufferobj_alloc;
-   ctx->Driver.DeleteBuffer = intel_bufferobj_free;
-   ctx->Driver.BufferData = intel_bufferobj_data;
-   ctx->Driver.BufferSubData = intel_bufferobj_subdata;
-   ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata;
-   ctx->Driver.MapBuffer = intel_bufferobj_map;
-   ctx->Driver.UnmapBuffer = intel_bufferobj_unmap;
+   functions->NewBufferObject = intel_bufferobj_alloc;
+   functions->DeleteBuffer = intel_bufferobj_free;
+   functions->BufferData = intel_bufferobj_data;
+   functions->BufferSubData = intel_bufferobj_subdata;
+   functions->GetBufferSubData = intel_bufferobj_get_subdata;
+   functions->MapBuffer = intel_bufferobj_map;
+   functions->MapBufferRange = intel_bufferobj_map_range;
+   functions->FlushMappedBufferRange = intel_bufferobj_flush_mapped_range;
+   functions->UnmapBuffer = intel_bufferobj_unmap;
+   functions->CopyBufferSubData = intel_bufferobj_copy_subdata;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.h b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
index 0431015631..bf3e08a320 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.h
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
@@ -48,6 +48,12 @@ struct intel_buffer_object
    struct intel_region *region; /* Is there a zero-copy texture
                                    associated with this (pixel)
                                    buffer object? */
+
+   drm_intel_bo *range_map_bo;
+   void *range_map_buffer;
+   unsigned int range_map_offset;
+   GLsizei range_map_size;
+
    GLboolean mapped_gtt;
 };
 
@@ -60,7 +66,7 @@ dri_bo *intel_bufferobj_buffer(struct intel_context *intel,
 
 /* Hook the bufferobject implementation into mesa: 
  */
-void intel_bufferobj_init(struct intel_context *intel);
+void intelInitBufferObjectFuncs(struct dd_function_table *functions);
 
 
 
@@ -72,10 +78,7 @@ void intel_bufferobj_init(struct intel_context *intel);
 static INLINE struct intel_buffer_object *
 intel_buffer_object(struct gl_buffer_object *obj)
 {
-   if (obj->Name)
-      return (struct intel_buffer_object *) obj;
-   else
-      return NULL;
+   return (struct intel_buffer_object *) obj;
 }
 
 /* Helpers for zerocopy image uploads.  See also intel_regions.h:
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index 44e34330a6..e7357e78c5 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -157,7 +157,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
-   if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+   if (ctx->NewState & _NEW_BUFFERS) {
       /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
       _mesa_update_framebuffer(ctx);
       /* this updates the DrawBuffer's Width/Height if it's a FBO */
diff --git a/src/mesa/drivers/dri/intel/intel_chipset.h b/src/mesa/drivers/dri/intel/intel_chipset.h
index a528d996dc..3dc8653a73 100644
--- a/src/mesa/drivers/dri/intel/intel_chipset.h
+++ b/src/mesa/drivers/dri/intel/intel_chipset.h
@@ -68,13 +68,18 @@
 #define PCI_CHIP_G41_G                  0x2E32
 #define PCI_CHIP_B43_G                  0x2E42
 
+#define PCI_CHIP_ILD_G                  0x0042
+#define PCI_CHIP_ILM_G                  0x0046
+
 #define IS_MOBILE(devid)	(devid == PCI_CHIP_I855_GM || \
 				 devid == PCI_CHIP_I915_GM || \
 				 devid == PCI_CHIP_I945_GM || \
 				 devid == PCI_CHIP_I945_GME || \
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
-				 devid == PCI_CHIP_GM45_GM || IS_IGD(devid))
+				 devid == PCI_CHIP_GM45_GM || \
+				 IS_IGD(devid) || \
+				 devid == PCI_CHIP_ILM_G)
 
 #define IS_G45(devid)           (devid == PCI_CHIP_IGD_E_G || \
                                  devid == PCI_CHIP_Q45_G || \
@@ -84,6 +89,10 @@
 #define IS_GM45(devid)          (devid == PCI_CHIP_GM45_GM)
 #define IS_G4X(devid)		(IS_G45(devid) || IS_GM45(devid))
 
+#define IS_ILD(devid)           (devid == PCI_CHIP_ILD_G)
+#define IS_ILM(devid)           (devid == PCI_CHIP_ILM_G)
+#define IS_IGDNG(devid)           (IS_ILD(devid) || IS_ILM(devid))
+
 #define IS_915(devid)		(devid == PCI_CHIP_I915_G || \
 				 devid == PCI_CHIP_E7221_G || \
 				 devid == PCI_CHIP_I915_GM)
@@ -101,7 +110,8 @@
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
 				 devid == PCI_CHIP_I946_GZ || \
-				 IS_G4X(devid))
+				 IS_G4X(devid) || \
+				 IS_IGDNG(devid))
 
 #define IS_9XX(devid)		(IS_915(devid) || \
 				 IS_945(devid) || \
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
index 19f47632ac..bce23724b3 100644
--- a/src/mesa/drivers/dri/intel/intel_clear.c
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -27,25 +27,9 @@
  **************************************************************************/
 
 #include "main/glheader.h"
-#include "main/enums.h"
-#include "main/image.h"
 #include "main/mtypes.h"
-#include "main/arrayobj.h"
-#include "main/attrib.h"
-#include "main/blend.h"
-#include "main/bufferobj.h"
-#include "main/buffers.h"
-#include "main/depth.h"
-#include "main/enable.h"
-#include "main/macros.h"
-#include "main/matrix.h"
-#include "main/polygon.h"
-#include "main/texstate.h"
-#include "main/shaders.h"
-#include "main/stencil.h"
-#include "main/varray.h"
-#include "glapi/dispatch.h"
 #include "swrast/swrast.h"
+#include "drivers/common/meta.h"
 
 #include "intel_context.h"
 #include "intel_blit.h"
@@ -53,240 +37,10 @@
 #include "intel_clear.h"
 #include "intel_fbo.h"
 #include "intel_pixel.h"
+#include "intel_regions.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
-#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
-			      BUFFER_BIT_FRONT_LEFT |			\
-			      BUFFER_BIT_COLOR0 |			\
-			      BUFFER_BIT_COLOR1 |			\
-			      BUFFER_BIT_COLOR2 |			\
-			      BUFFER_BIT_COLOR3 |			\
-			      BUFFER_BIT_COLOR4 |			\
-			      BUFFER_BIT_COLOR5 |			\
-			      BUFFER_BIT_COLOR6 |			\
-			      BUFFER_BIT_COLOR7)
-
-
-/**
- * Per-context one-time init of things for intl_clear_tris().
- * Basically set up a private array object for vertex/color arrays.
- */
-static void
-init_clear(GLcontext *ctx)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_array_object *arraySave = NULL;
-   const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
-   const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
-
-   /* create new array object */
-   intel->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
-
-   /* save current array object, bind new one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   /* one-time setup of vertex arrays (pos, color) */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
-   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), intel->clear.color);
-   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), intel->clear.vertices);
-   _mesa_Enable(GL_COLOR_ARRAY);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-
-   /* restore original array object */
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-
-   /* restore original buffer objects */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
-}
-
-
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-void
-intel_clear_tris(GLcontext *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLuint saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   struct gl_array_object *arraySave = NULL;
-
-   if (!intel->clear.arrayObj)
-      init_clear(ctx);
-
-   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
-		    BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_CURRENT_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_POLYGON_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-
-   /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-
-   /* save current array object, bind our private one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   intel_meta_set_passthrough_transform(intel);
-
-   for (i = 0; i < 4; i++) {
-      COPY_4FV(intel->clear.color[i], ctx->Color.ClearColor);
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   intel->clear.vertices[0][0] = fb->_Xmin;
-   intel->clear.vertices[0][1] = fb->_Ymin;
-   intel->clear.vertices[0][2] = dst_z;
-   intel->clear.vertices[1][0] = fb->_Xmax;
-   intel->clear.vertices[1][1] = fb->_Ymin;
-   intel->clear.vertices[1][2] = dst_z;
-   intel->clear.vertices[2][0] = fb->_Xmax;
-   intel->clear.vertices[2][1] = fb->_Ymax;
-   intel->clear.vertices[2][2] = dst_z;
-   intel->clear.vertices[3][0] = fb->_Xmin;
-   intel->clear.vertices[3][1] = fb->_Ymax;
-   intel->clear.vertices[3][2] = dst_z;
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-      GLuint color_bit;
-
-      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
-      if (color_bit != 0)
-	 this_mask |= (1 << (color_bit - 1));
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else if (color_bit != 0)
-	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
-			  (color_bit - BUFFER_COLOR0 - 1));
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
-				 GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
-				   ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
-
-      mask &= ~this_mask;
-   }
-
-   intel_meta_restore_transform(intel);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopAttrib();
-
-   /* restore current array object */
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-}
-
 static const char *buffer_names[] = {
    [BUFFER_FRONT_LEFT] = "front",
    [BUFFER_BACK_LEFT] = "back",
@@ -340,7 +94,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
          = intel_get_rb_region(fb, BUFFER_STENCIL);
       if (stencilRegion) {
          /* have hw stencil */
-         if (IS_965(intel->intelScreen->deviceID) ||
+         if (stencilRegion->tiling == I915_TILING_Y ||
 	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
 	    /* We have to use the 3D engine if we're clearing a partial mask
 	     * of the stencil buffer, or if we're on a 965 which has a tiled
@@ -357,9 +111,10 @@ intelClear(GLcontext *ctx, GLbitfield mask)
 
    /* HW depth */
    if (mask & BUFFER_BIT_DEPTH) {
+      const struct intel_region *irb = intel_get_rb_region(fb, BUFFER_DEPTH);
+
       /* clear depth with whatever method is used for stencil (see above) */
-      if (IS_965(intel->intelScreen->deviceID) ||
-	  tri_mask & BUFFER_BIT_STENCIL)
+      if (irb->tiling == I915_TILING_Y || tri_mask & BUFFER_BIT_STENCIL)
          tri_mask |= BUFFER_BIT_DEPTH;
       else
          blit_mask |= BUFFER_BIT_DEPTH;
@@ -369,7 +124,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
     * buffer with it.
     */
    if (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
-      int color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
+      int color_bit = _mesa_ffs(mask & BUFFER_BITS_COLOR);
       if (color_bit != 0) {
 	 tri_mask |= blit_mask & (1 << (color_bit - 1));
 	 blit_mask &= ~(1 << (color_bit - 1));
@@ -379,14 +134,18 @@ intelClear(GLcontext *ctx, GLbitfield mask)
    /* SW fallback clearing */
    swrast_mask = mask & ~tri_mask & ~blit_mask;
 
-   for (i = 0; i < BUFFER_COUNT; i++) {
-      GLuint bufBit = 1 << i;
-      if ((blit_mask | tri_mask) & bufBit) {
+   {
+      /* look for non-Intel renderbuffers (clear them with swrast) */
+      GLbitfield blit_or_tri = blit_mask | tri_mask;
+      while (blit_or_tri) {
+         GLuint i = _mesa_ffs(blit_or_tri) - 1;
+         GLbitfield bufBit = 1 << i;
          if (!fb->Attachment[i].Renderbuffer->ClassID) {
             blit_mask &= ~bufBit;
             tri_mask &= ~bufBit;
             swrast_mask |= bufBit;
          }
+         blit_or_tri ^= bufBit;
       }
    }
 
@@ -411,7 +170,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
 	 }
 	 DBG("\n");
       }
-      intel_clear_tris(ctx, tri_mask);
+      _mesa_meta_clear(&intel->ctx, tri_mask);
    }
 
    if (swrast_mask) {
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index aecb317eb8..2364829a1e 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -38,6 +38,7 @@
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
 #include "drivers/common/driverfuncs.h"
+#include "drivers/common/meta.h"
 
 #include "i830_dri.h"
 
@@ -164,6 +165,12 @@ intelGetString(GLcontext * ctx, GLenum name)
       case PCI_CHIP_B43_G:
          chipset = "Intel(R) B43";
          break;
+      case PCI_CHIP_ILD_G:
+         chipset = "Intel(R) IGDNG_D";
+         break;
+      case PCI_CHIP_ILM_G:
+         chipset = "Intel(R) IGDNG_M";
+         break;
       default:
          chipset = "Unknown Intel Chipset";
          break;
@@ -401,7 +408,7 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     if (!driContext->driScreenPriv->dri2.enabled)
 	return;
 
-    if (!intel->internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+    if (!intel->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
        /* If we're rendering to the fake front buffer, make sure all the pending
 	* drawing has landed on the real front buffer.  Otherwise when we
 	* eventually get to DRI2GetBuffersWithFormat the stale real front
@@ -586,10 +593,15 @@ intelInitDriverFunctions(struct dd_function_table *functions)
    functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
 
    intelInitTextureFuncs(functions);
+   intelInitTextureImageFuncs(functions);
+   intelInitTextureSubImageFuncs(functions);
+   intelInitTextureCopyImageFuncs(functions);
    intelInitStateFuncs(functions);
    intelInitClearFuncs(functions);
    intelInitBufferFuncs(functions);
    intelInitPixelFuncs(functions);
+   intelInitBufferObjectFuncs(functions);
+   intel_init_syncobj_functions(functions);
 }
 
 
@@ -632,6 +644,10 @@ intelInitContext(struct intel_context *intel,
       intel->maxBatchSize = BATCH_SZ;
 
    intel->bufmgr = intelScreen->bufmgr;
+
+   if (0) /* for debug */
+      drm_intel_bufmgr_set_debug(intel->bufmgr, 1);
+
    intel->ttm = intelScreen->ttm;
    if (intel->ttm) {
       int bo_reuse_mode;
@@ -685,7 +701,15 @@ intelInitContext(struct intel_context *intel,
     */
    _mesa_init_point(ctx);
 
+   meta_init_metaops(ctx, &intel->meta);
    ctx->Const.MaxColorAttachments = 4;  /* XXX FBO: review this */
+   if (IS_965(intelScreen->deviceID)) {
+      if (MAX_WIDTH > 8192)
+	 ctx->Const.MaxRenderbufferSize = 8192;
+   } else {
+      if (MAX_WIDTH > 2048)
+	 ctx->Const.MaxRenderbufferSize = 2048;
+   }
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext(ctx);
@@ -697,6 +721,8 @@ intelInitContext(struct intel_context *intel,
    _swrast_allow_pixel_fog(ctx, GL_FALSE);
    _swrast_allow_vertex_fog(ctx, GL_TRUE);
 
+   _mesa_meta_init(ctx);
+
    intel->hw_stencil = mesaVis->stencilBits && mesaVis->depthBits == 24;
    intel->hw_stipple = 1;
 
@@ -743,7 +769,6 @@ intelInitContext(struct intel_context *intel,
 
    intel->batch = intel_batchbuffer_alloc(intel);
 
-   intel_bufferobj_init(intel);
    intel_fbo_init(intel);
 
    if (intel->ctx.Mesa_DXTn) {
@@ -753,6 +778,15 @@ intelInitContext(struct intel_context *intel,
    else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
    }
+   intel->use_texture_tiling = driQueryOptionb(&intel->optionCache,
+					       "texture_tiling");
+   if (intel->use_texture_tiling &&
+       !intel->intelScreen->kernel_exec_fencing) {
+      fprintf(stderr, "No kernel support for execution fencing, "
+	      "disabling texture tiling\n");
+      intel->use_texture_tiling = GL_FALSE;
+   }
+   intel->use_early_z = driQueryOptionb(&intel->optionCache, "early_z");
 
    intel->prim.primitive = ~0;
 
@@ -792,8 +826,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       INTEL_FIREVERTICES(intel);
 
-      if (intel->clear.arrayObj)
-         _mesa_delete_array_object(&intel->ctx, intel->clear.arrayObj);
+      _mesa_meta_free(&intel->ctx);
+
+      meta_destroy_metaops(&intel->meta);
 
       intel->vtbl.destroy(intel);
 
@@ -816,13 +851,64 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       intel->first_post_swapbuffers_batch = NULL;
 
       if (release_texture_heaps) {
-         /* This share group is about to go away, free our private
-          * texture object data.
+         /* Nothing is currently done here to free texture heaps;
+          * but we're not using the texture heap utilities, so I
+          * rather think we shouldn't.  I've taken a look, and can't
+          * find any private texture data hanging around anywhere, but
+          * I'm not yet certain there isn't any at all...
           */
-         if (INTEL_DEBUG & DEBUG_TEXTURE)
+         /* if (INTEL_DEBUG & DEBUG_TEXTURE)
             fprintf(stderr, "do something to free texture heaps\n");
+          */
       }
 
+      /* XXX In intelMakeCurrent() below, the context's static regions are 
+       * referenced inside the frame buffer; it's listed as a hack,
+       * with a comment of "XXX FBO temporary fix-ups!", but
+       * as long as it's there, we should release the regions here.
+       * The do/while loop around the block is used to allow the
+       * "continue" statements inside the block to exit the block,
+       * to avoid many layers of "if" constructs.
+       */
+      do {
+         __DRIdrawablePrivate * driDrawPriv = intel->driDrawable;
+         struct intel_framebuffer *intel_fb;
+         struct intel_renderbuffer *irbDepth, *irbStencil;
+         if (!driDrawPriv) {
+            /* We're already detached from the drawable; exit this block. */
+            continue;
+         }
+         intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate;
+         if (!intel_fb) {
+            /* The frame buffer is already gone; exit this block. */
+            continue;
+         }
+         irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+         irbStencil = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+         /* If the regions of the frame buffer still match the regions
+          * of the context, release them.  If they've changed somehow,
+          * leave them alone.
+          */
+         if (intel_fb->color_rb[0] && intel_fb->color_rb[0]->region == intel->front_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL);
+         }
+         if (intel_fb->color_rb[1] && intel_fb->color_rb[1]->region == intel->back_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL);
+         }
+
+         if (irbDepth && irbDepth->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbDepth, NULL);
+         }
+         /* Usually, the stencil buffer is the same as the depth buffer;
+          * but they're handled separately in MakeCurrent, so we'll
+          * handle them separately here.
+          */
+         if (irbStencil && irbStencil->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbStencil, NULL);
+         }
+      } while (0);
+
       intel_region_release(&intel->front_region);
       intel_region_release(&intel->back_region);
       intel_region_release(&intel->depth_region);
@@ -862,7 +948,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
           if (driDrawPriv != driReadPriv)
               intel_update_renderbuffers(driContextPriv, driReadPriv);
       } else {
-          /* XXX FBO temporary fix-ups! */
+          /* XXX FBO temporary fix-ups!  These are released in 
+           * intelDextroyContext(), above.  Changes here should be
+           * reflected there.
+           */
           /* if the renderbuffers don't have regions, init them from the context */
          struct intel_renderbuffer *irbDepth
             = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
@@ -947,7 +1036,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
    int me = intel->hHWContext;
 
    drmGetLock(intel->driFd, intel->hHWContext, flags);
-   intel->locked = 1;
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - got contended lock\n", __progname);
@@ -1004,9 +1092,12 @@ void LOCK_HARDWARE( struct intel_context *intel )
     struct intel_framebuffer *intel_fb = NULL;
     struct intel_renderbuffer *intel_rb = NULL;
 
-    _glthread_LOCK_MUTEX(lockMutex);
-    assert(!intel->locked);
-    intel->locked = 1;
+    intel->locked++;
+    if (intel->locked >= 2)
+       return;
+
+    if (!sPriv->dri2.enabled)
+       _glthread_LOCK_MUTEX(lockMutex);
 
     if (intel->driDrawable) {
        intel_fb = intel->driDrawable->driverPrivate;
@@ -1053,13 +1144,16 @@ void UNLOCK_HARDWARE( struct intel_context *intel )
 {
     __DRIscreen *sPriv = intel->driScreen;
 
-   intel->vtbl.note_unlock( intel );
-   intel->locked = 0;
+   intel->locked--;
+   if (intel->locked > 0)
+      return;
 
-   if (!sPriv->dri2.enabled)
-      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+   assert(intel->locked == 0);
 
-   _glthread_UNLOCK_MUTEX(lockMutex);
+   if (!sPriv->dri2.enabled) {
+      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+      _glthread_UNLOCK_MUTEX(lockMutex);
+   }
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - unlocked\n", __progname);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index e2b3943fb5..03e7cf39d6 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -33,6 +33,7 @@
 #include "main/mtypes.h"
 #include "main/mm.h"
 #include "texmem.h"
+#include "dri_metaops.h"
 #include "drm.h"
 #include "intel_bufmgr.h"
 
@@ -79,9 +80,19 @@ extern void intelFallback(struct intel_context *intel, GLuint bit,
 
 #define INTEL_MAX_FIXUP 64
 
+struct intel_sync_object {
+   struct gl_sync_object Base;
+
+   /** Batch associated with this sync object */
+   drm_intel_bo *bo;
+};
+
+/**
+ * intel_context is derived from Mesa's context class: GLcontext.
+ */
 struct intel_context
 {
-   GLcontext ctx;               /* the parent class */
+   GLcontext ctx;  /**< base class, must be first field */
 
    struct
    {
@@ -91,7 +102,6 @@ struct intel_context
       void (*new_batch) (struct intel_context * intel);
       void (*emit_invarient_state) (struct intel_context * intel);
       void (*note_fence) (struct intel_context *intel, GLuint fence);
-      void (*note_unlock) (struct intel_context *intel);
       void (*update_texture_state) (struct intel_context * intel);
 
       void (*render_start) (struct intel_context * intel);
@@ -158,19 +168,7 @@ struct intel_context
       void (*debug_batch)(struct intel_context *intel);
    } vtbl;
 
-   struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
+   struct dri_metaops meta;
 
    GLint refcount;
    GLuint Fallback;
@@ -182,7 +180,6 @@ struct intel_context
    struct intel_region *front_region;
    struct intel_region *back_region;
    struct intel_region *depth_region;
-   GLboolean internal_viewport_call;
 
    /**
     * This value indicates that the kernel memory manager is being used
@@ -216,13 +213,6 @@ struct intel_context
    GLuint ClearColor565;
    GLuint ClearColor8888;
 
-   /* info for intel_clear_tris() */
-   struct
-   {
-      struct gl_array_object *arrayObj;
-      GLfloat vertices[4][3];
-      GLfloat color[4][4];
-   } clear;
 
    /* Offsets of fields within the current vertex:
     */
@@ -304,6 +294,9 @@ struct intel_context
     */
    GLboolean is_front_buffer_reading;
 
+   GLboolean use_texture_tiling;
+   GLboolean use_early_z;
+
    drm_clip_rect_t fboRect;     /**< cliprect for FBO rendering */
 
    int perf_boxes;
@@ -325,7 +318,7 @@ struct intel_context
    __DRIdrawablePrivate *driReadDrawable;
    __DRIscreenPrivate *driScreen;
    intelScreenPrivate *intelScreen;
-   volatile struct drm_i915_sarea *sarea;
+   volatile drm_i915_sarea_t *sarea;
 
    GLuint lastStamp;
 
@@ -483,6 +476,8 @@ extern void intelFlush(GLcontext * ctx);
 
 extern void intelInitDriverFunctions(struct dd_function_table *functions);
 
+void intel_init_syncobj_functions(struct dd_function_table *functions);
+
 
 /* ================================================================
  * intel_state.c:
@@ -558,6 +553,9 @@ void intel_viewport(GLcontext * ctx, GLint x, GLint y,
 void intel_update_renderbuffers(__DRIcontext *context,
 				__DRIdrawable *drawable);
 
+void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
+				  uint32_t buffer_id);
+
 /*======================================================================
  * Inline conversion functions.  
  * These are better-typed than the macros used previously:
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c
index 9ec1b4ec2f..2e61c556d8 100644
--- a/src/mesa/drivers/dri/intel/intel_extensions.c
+++ b/src/mesa/drivers/dri/intel/intel_extensions.c
@@ -30,10 +30,14 @@
 #include "intel_extensions.h"
 
 
+#define need_GL_ARB_copy_buffer
 #define need_GL_ARB_framebuffer_object
+#define need_GL_ARB_map_buffer_range
 #define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_shader_objects
+#define need_GL_ARB_sync
+#define need_GL_ARB_vertex_array_object
 #define need_GL_ARB_vertex_program
 #define need_GL_ARB_vertex_shader
 #define need_GL_ARB_window_pos
@@ -45,9 +49,12 @@
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_framebuffer_blit
+#define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_point_parameters
+#define need_GL_EXT_provoking_vertex
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_stencil_two_side
+#define need_GL_APPLE_vertex_array_object
 #define need_GL_ATI_separate_stencil
 #define need_GL_ATI_envmap_bumpmap
 #define need_GL_NV_point_sprite
@@ -65,8 +72,13 @@
  * i965_dri.
  */
 static const struct dri_extension card_extensions[] = {
+   { "GL_ARB_copy_buffer",                GL_ARB_copy_buffer_functions },
+   { "GL_ARB_half_float_pixel",           NULL },
+   { "GL_ARB_map_buffer_range",           GL_ARB_map_buffer_range_functions },
    { "GL_ARB_multitexture",               NULL },
    { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
+   { "GL_ARB_point_sprite",               NULL },
+   { "GL_ARB_sync",                       GL_ARB_sync_functions },
    { "GL_ARB_texture_border_clamp",       NULL },
    { "GL_ARB_texture_cube_map",           NULL },
    { "GL_ARB_texture_env_add",            NULL },
@@ -75,6 +87,7 @@ static const struct dri_extension card_extensions[] = {
    { "GL_ARB_texture_env_dot3",           NULL },
    { "GL_ARB_texture_mirrored_repeat",    NULL },
    { "GL_ARB_texture_rectangle",          NULL },
+   { "GL_ARB_vertex_array_object",        GL_ARB_vertex_array_object_functions},
    { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
    { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
    { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
@@ -85,7 +98,9 @@ static const struct dri_extension card_extensions[] = {
    { "GL_EXT_blend_subtract",             NULL },
    { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
    { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+   { "GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions },
    { "GL_EXT_packed_depth_stencil",       NULL },
+   { "GL_EXT_provoking_vertex",           GL_EXT_provoking_vertex_functions },
    { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
    { "GL_EXT_stencil_wrap",               NULL },
    { "GL_EXT_texture_edge_clamp",         NULL },
@@ -95,6 +110,7 @@ static const struct dri_extension card_extensions[] = {
    { "GL_EXT_texture_lod_bias",           NULL },
    { "GL_3DFX_texture_compression_FXT1",  NULL },
    { "GL_APPLE_client_storage",           NULL },
+   { "GL_APPLE_vertex_array_object",      GL_APPLE_vertex_array_object_functions},
    { "GL_MESA_pack_invert",               NULL },
    { "GL_MESA_ycbcr_texture",             NULL },
    { "GL_NV_blend_square",                NULL },
@@ -112,8 +128,10 @@ static const struct dri_extension i915_extensions[] = {
    { "GL_ARB_fragment_program",           NULL },
    { "GL_ARB_shadow",                     NULL },
    { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
    { "GL_ATI_texture_env_combine3",       NULL },
    { "GL_EXT_shadow_funcs",               NULL },
+   { "GL_EXT_stencil_two_side",           GL_EXT_stencil_two_side_functions },
    { "GL_NV_texture_env_combine4",        NULL },
    { NULL,                                NULL }
 };
@@ -128,6 +146,7 @@ static const struct dri_extension brw_extensions[] = {
    { "GL_ARB_framebuffer_object",         GL_ARB_framebuffer_object_functions},
    { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
    { "GL_ARB_point_sprite", 		  NULL },
+   { "GL_ARB_seamless_cube_map",          NULL },
    { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
    { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
    { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index 30f58b1f44..804c034840 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -35,6 +35,7 @@
 #include "main/context.h"
 #include "main/texformat.h"
 #include "main/texrender.h"
+#include "drivers/common/meta.h"
 
 #include "intel_context.h"
 #include "intel_buffers.h"
@@ -217,7 +218,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width,
 	  height, pitch);
 
-      irb->region = intel_region_alloc(intel, cpp, width, height, pitch,
+      irb->region = intel_region_alloc(intel, I915_TILING_NONE,
+				       cpp, width, height, pitch,
 				       GL_TRUE);
       if (!irb->region)
          return GL_FALSE;       /* out of memory? */
@@ -575,9 +577,10 @@ intel_render_texture(GLcontext * ctx,
 
    ASSERT(newImage);
 
-   if (newImage->Border != 0) {
-      /* Fallback on drawing to a texture with a border, which won't have a
-       * miptree.
+   intel_image = intel_texture_image(newImage);
+   if (!intel_image->mt) {
+      /* Fallback on drawing to a texture that doesn't have a miptree
+       * (has a border, width/height 0, etc.)
        */
       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
       _mesa_render_texture(ctx, fb, att);
@@ -608,7 +611,6 @@ intel_render_texture(GLcontext * ctx,
        irb->Base.RefCount);
 
    /* point the renderbufer's region to the texture image region */
-   intel_image = intel_texture_image(newImage);
    if (irb->region != intel_image->mt->region) {
       if (irb->region)
 	 intel_region_release(&irb->region);
@@ -680,6 +682,11 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
       if (rb == NULL)
 	 continue;
 
+      if (irb == NULL) {
+	 fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+	 continue;
+      }
+
       switch (irb->texformat->MesaFormat) {
       case MESA_FORMAT_ARGB8888:
       case MESA_FORMAT_RGB565:
@@ -694,74 +701,6 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
 
 
 /**
- * Called from glBlitFramebuffer().
- * For now, we're doing an approximation with glCopyPixels().
- * XXX we need to bypass all the per-fragment operations, except scissor.
- */
-static void
-intel_blit_framebuffer(GLcontext *ctx,
-                       GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
-                       GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
-                       GLbitfield mask, GLenum filter)
-{
-   const GLfloat xZoomSave = ctx->Pixel.ZoomX;
-   const GLfloat yZoomSave = ctx->Pixel.ZoomY;
-   GLsizei width, height;
-   GLfloat xFlip = 1.0F, yFlip = 1.0F;
-
-   if (srcX1 < srcX0) {
-      GLint tmp = srcX1;
-      srcX1 = srcX0;
-      srcX0 = tmp;
-      xFlip = -1.0F;
-   }
-
-   if (srcY1 < srcY0) {
-      GLint tmp = srcY1;
-      srcY1 = srcY0;
-      srcY0 = tmp;
-      yFlip = -1.0F;
-   }
-
-   width = srcX1 - srcX0;
-   height = srcY1 - srcY0;
-
-   ctx->Pixel.ZoomX = xFlip * (dstX1 - dstX0) / (srcX1 - srcY0);
-   ctx->Pixel.ZoomY = yFlip * (dstY1 - dstY0) / (srcY1 - srcY0);
-
-   if (ctx->Pixel.ZoomX < 0.0F) {
-      dstX0 = MAX2(dstX0, dstX1);
-   }
-   else {
-      dstX0 = MIN2(dstX0, dstX1);
-   }
-
-   if (ctx->Pixel.ZoomY < 0.0F) {
-      dstY0 = MAX2(dstY0, dstY1);
-   }
-   else {
-      dstY0 = MIN2(dstY0, dstY1);
-   }
-
-   if (mask & GL_COLOR_BUFFER_BIT) {
-      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
-                             dstX0, dstY0, GL_COLOR);
-   }
-   if (mask & GL_DEPTH_BUFFER_BIT) {
-      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
-                             dstX0, dstY0, GL_DEPTH);
-   }
-   if (mask & GL_STENCIL_BUFFER_BIT) {
-      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
-                             dstX0, dstY0, GL_STENCIL);
-   }
-      
-   ctx->Pixel.ZoomX = xZoomSave;
-   ctx->Pixel.ZoomY = yZoomSave;
-}
-
-
-/**
  * Do one-time context initializations related to GL_EXT_framebuffer_object.
  * Hook in device driver functions.
  */
@@ -776,5 +715,5 @@ intel_fbo_init(struct intel_context *intel)
    intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture;
    intel->ctx.Driver.ResizeBuffers = intel_resize_buffers;
    intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer;
-   intel->ctx.Driver.BlitFramebuffer = intel_blit_framebuffer;
+   intel->ctx.Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
new file mode 100644
index 0000000000..fe986092db
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
@@ -0,0 +1,283 @@
+/*
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/bufferobj.h"
+#include "main/teximage.h"
+#include "main/texenv.h"
+#include "main/texobj.h"
+#include "main/texstate.h"
+#include "main/texparam.h"
+#include "main/varray.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/buffers.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/depth.h"
+#include "main/hash.h"
+#include "main/mipmap.h"
+#include "main/blend.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_pixel.h"
+#include "intel_tex.h"
+#include "intel_mipmap_tree.h"
+
+static const char *intel_fp_tex2d =
+      "!!ARBfp1.0\n"
+      "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n"
+      "END\n";
+
+static GLboolean
+intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
+			    int level, int width, int height)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLfloat vertices[4][2];
+   GLint status;
+
+   /* Set to source from the previous level */
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1);
+
+   /* Set to draw into the current level */
+   _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT,
+				 GL_COLOR_ATTACHMENT0_EXT,
+				 GL_TEXTURE_2D,
+				 tex_name,
+				 level);
+   /* Choose to render to the color attachment. */
+   _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
+
+   status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT);
+   if (status != GL_FRAMEBUFFER_COMPLETE_EXT)
+      return GL_FALSE;
+
+   meta_set_passthrough_transform(&intel->meta);
+
+   /* XXX: Doing it right would involve setting up the transformation to do
+    * 0-1 mapping or something, and not changing the vertex data.
+    */
+   vertices[0][0] = 0;
+   vertices[0][1] = 0;
+   vertices[1][0] = width;
+   vertices[1][1] = 0;
+   vertices[2][0] = width;
+   vertices[2][1] = height;
+   vertices[3][0] = 0;
+   vertices[3][1] = height;
+
+   _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+
+   return GL_TRUE;
+}
+
+static GLboolean
+intel_generate_mipmap_2d(GLcontext *ctx,
+			 GLenum target,
+			 struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLint old_active_texture;
+   int level, max_levels, start_level, end_level;
+   GLuint fb_name;
+   GLboolean success = GL_FALSE;
+   struct gl_framebuffer *saved_fbo = NULL;
+
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
+		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   old_active_texture = ctx->Texture.CurrentUnit;
+   _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer);
+
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_DEPTH_TEST);
+   _mesa_Disable(GL_STENCIL_TEST);
+   _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+   _mesa_DepthMask(GL_FALSE);
+
+   /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our
+    * minification.
+    */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
+   _mesa_Enable(GL_TEXTURE_2D);
+   _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+		       GL_LINEAR_MIPMAP_NEAREST);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+
+   /* Bind the new renderbuffer to the color attachment point. */
+   _mesa_GenFramebuffersEXT(1, &fb_name);
+   _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name);
+
+   meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp,
+			     intel_fp_tex2d);
+   meta_set_passthrough_vertex_program(&intel->meta);
+
+   max_levels = _mesa_max_texture_levels(ctx, texObj->Target);
+   start_level = texObj->BaseLevel;
+   end_level = texObj->MaxLevel;
+
+   /* Loop generating level+1 from level. */
+   for (level = start_level; level < end_level && level < max_levels - 1; level++) {
+      const struct gl_texture_image *srcImage;
+      int width, height;
+
+      srcImage = _mesa_select_tex_image(ctx, texObj, target, level);
+      if (srcImage->Border != 0)
+	 goto fail;
+
+      width = srcImage->Width / 2;
+      if (width < 1)
+	 width = 1;
+      height = srcImage->Height / 2;
+      if (height < 1)
+	 height = 1;
+
+      if (width == srcImage->Width &&
+	  height == srcImage->Height) {
+	 /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the
+	  * maximum texture level for the object, so break out when we've gone
+	  * over the edge.
+	  */
+	 break;
+      }
+
+      /* Make sure that there's space allocated for the target level.
+       * We could skip this if there's already space allocated and save some
+       * time.
+       */
+      _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat,
+		       width, height, 0,
+		       GL_RGBA, GL_UNSIGNED_INT, NULL);
+
+      if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1,
+				       width, height))
+	 goto fail;
+   }
+
+   success = GL_TRUE;
+
+fail:
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
+
+   _mesa_DeleteFramebuffersEXT(1, &fb_name);
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
+   if (saved_fbo)
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name);
+   _mesa_reference_framebuffer(&saved_fbo, NULL);
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+
+   return success;
+}
+
+
+/**
+ * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
+ * level).
+ *
+ * The texture object's miptree must be mapped.
+ *
+ * It would be really nice if this was just called by Mesa whenever mipmaps
+ * needed to be regenerated, rather than us having to remember to do so in
+ * each texture image modification path.
+ *
+ * This function should also include an accelerated path.
+ */
+void
+intel_generate_mipmap(GLcontext *ctx, GLenum target,
+                      struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_texture_object *intelObj = intel_texture_object(texObj);
+   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int face, i;
+
+   /* HW path */
+   if (target == GL_TEXTURE_2D &&
+       ctx->Extensions.EXT_framebuffer_object &&
+       ctx->Extensions.ARB_fragment_program &&
+       ctx->Extensions.ARB_vertex_program) {
+      GLboolean success;
+
+      /* We'll be accessing this texture using GL entrypoints, which should
+       * be resilient against other access to this texture.
+       */
+      _mesa_unlock_texture(ctx, texObj);
+      success = intel_generate_mipmap_2d(ctx, target, texObj);
+      _mesa_lock_texture(ctx, texObj);
+
+      if (success)
+	 return;
+   }
+
+   /* SW path */
+   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
+   _mesa_generate_mipmap(ctx, target, texObj);
+   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
+
+   /* Update the level information in our private data in the new images, since
+    * it didn't get set as part of a normal TexImage path.
+    */
+   for (face = 0; face < nr_faces; face++) {
+      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+         struct intel_texture_image *intelImage;
+
+	 intelImage = intel_texture_image(texObj->Image[face][i]);
+	 if (intelImage == NULL)
+	    break;
+
+	 intelImage->level = i;
+	 intelImage->face = face;
+	 /* Unreference the miptree to signal that the new Data is a bare
+	  * pointer from mesa.
+	  */
+	 intel_miptree_release(intel, &intelImage->mt);
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index 6e1e034e53..c985da5aa2 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -57,14 +57,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 			      GLuint last_level,
 			      GLuint width0,
 			      GLuint height0,
-			      GLuint depth0, GLuint cpp, GLuint compress_byte)
+			      GLuint depth0, GLuint cpp, GLuint compress_byte,
+			      uint32_t tiling)
 {
    GLboolean ok;
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
 
-   DBG("%s target %s format %s level %d..%d\n", __FUNCTION__,
+   DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
-       _mesa_lookup_enum_by_nr(internal_format), first_level, last_level);
+       _mesa_lookup_enum_by_nr(internal_format), 
+       first_level, last_level, mt);
 
    mt->target = target_to_target(target);
    mt->internal_format = internal_format;
@@ -80,15 +82,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 
 #ifdef I915
    if (IS_945(intel->intelScreen->deviceID))
-      ok = i945_miptree_layout(intel, mt);
+      ok = i945_miptree_layout(intel, mt, tiling);
    else
-      ok = i915_miptree_layout(intel, mt);
+      ok = i915_miptree_layout(intel, mt, tiling);
 #else
-   ok = brw_miptree_layout(intel, mt);
+   ok = brw_miptree_layout(intel, mt, tiling);
 #endif
 
    if (!ok) {
       free(mt);
+      DBG("%s not okay - returning NULL\n", __FUNCTION__);
       return NULL;
    }
 
@@ -98,6 +101,7 @@ intel_miptree_create_internal(struct intel_context *intel,
 struct intel_mipmap_tree *
 intel_miptree_create(struct intel_context *intel,
 		     GLenum target,
+		     GLenum base_format,
 		     GLenum internal_format,
 		     GLuint first_level,
 		     GLuint last_level,
@@ -107,10 +111,23 @@ intel_miptree_create(struct intel_context *intel,
 		     GLboolean expect_accelerated_upload)
 {
    struct intel_mipmap_tree *mt;
+   uint32_t tiling;
+
+   if (intel->use_texture_tiling && compress_byte == 0 &&
+       intel->intelScreen->kernel_exec_fencing) {
+      if (IS_965(intel->intelScreen->deviceID) &&
+	  (base_format == GL_DEPTH_COMPONENT ||
+	   base_format == GL_DEPTH_STENCIL_EXT))
+	 tiling = I915_TILING_Y;
+      else
+	 tiling = I915_TILING_X;
+   } else
+      tiling = I915_TILING_NONE;
 
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level, width0,
-				      height0, depth0, cpp, compress_byte);
+				      height0, depth0, cpp, compress_byte,
+				      tiling);
    /*
     * pitch == 0 || height == 0  indicates the null texture
     */
@@ -118,6 +135,7 @@ intel_miptree_create(struct intel_context *intel,
       return NULL;
 
    mt->region = intel_region_alloc(intel,
+				   tiling,
 				   mt->cpp,
 				   mt->pitch,
 				   mt->total_height,
@@ -147,7 +165,8 @@ intel_miptree_create_for_region(struct intel_context *intel,
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level,
 				      region->width, region->height, 1,
-				      region->cpp, compress_byte);
+				      region->cpp, compress_byte,
+				      I915_TILING_NONE);
    if (!mt)
       return mt;
 #if 0
@@ -185,6 +204,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch)
 {
 #ifdef I915
@@ -205,6 +225,11 @@ int intel_miptree_pitch_align (struct intel_context *intel,
 	 pitch_align = 4;
       }
 
+      if (tiling == I915_TILING_X)
+	 pitch_align = 512;
+      else if (tiling == I915_TILING_Y)
+	 pitch_align = 128;
+
       pitch = ALIGN(pitch * mt->cpp, pitch_align);
 
 #ifdef I915
@@ -328,23 +353,31 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
 }
 
 
-
 void
-intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
-			       GLuint level, GLuint img,
-			       GLuint x, GLuint y)
+intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                  GLuint level, GLuint img,
+                                  GLuint x, GLuint y, 
+                                  GLuint offset)
 {
    if (img == 0 && level == 0)
       assert(x == 0 && y == 0);
 
    assert(img < mt->level[level].nr_images);
 
-   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp;
+   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset;
 
    DBG("%s level %d img %d pos %d,%d image_offset %x\n",
        __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]);
 }
 
+void
+intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
+			       GLuint level, GLuint img,
+			       GLuint x, GLuint y)
+{
+    intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0);
+}
+
 
 /* Although we use the image_offset[] array to store relative offsets
  * to cube faces, Mesa doesn't know anything about this and expects
@@ -454,11 +487,11 @@ intel_miptree_image_data(struct intel_context *intel,
 			0, 0,                             /* source x, y */
 			dst->level[level].width, height); /* width, height */
 
-      src += src_image_pitch * dst->cpp;
+      src = (char *)src + src_image_pitch * dst->cpp;
    }
 }
 
-extern GLuint intel_compressed_alignment(GLenum);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
 /* Copy mipmap image between trees
  */
 void
@@ -475,20 +508,37 @@ intel_miptree_image_copy(struct intel_context *intel,
    const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level);
    const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level);
    GLuint i;
+   GLboolean success;
 
    if (dst->compressed) {
-       GLuint alignment = intel_compressed_alignment(dst->internal_format);
+       GLuint align_w, align_h;
+
+       intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h);
        height = (height + 3) / 4;
-       width = ((width + alignment - 1) & ~(alignment - 1));
+       width = ALIGN(width, align_w);
    }
 
    for (i = 0; i < depth; i++) {
-      intel_region_copy(intel,
-                        dst->region, dst_offset + dst_depth_offset[i],
-                        0,
-                        0,
-                        src->region, src_offset + src_depth_offset[i],
-                        0, 0, width, height);
+      success = intel_region_copy(intel,
+				  dst->region, dst_offset + dst_depth_offset[i],
+				  0, 0,
+				  src->region, src_offset + src_depth_offset[i],
+				  0, 0, width, height, GL_COPY);
+      if (!success) {
+	 GLubyte *src_ptr, *dst_ptr;
+
+	 src_ptr = intel_region_map(intel, src->region);
+	 dst_ptr = intel_region_map(intel, dst->region);
+
+	 _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i],
+			 dst->cpp,
+			 dst->pitch,
+			 0, 0, width, height,
+			 src_ptr + src_offset + src_depth_offset[i],
+			 src->pitch,
+			 0, 0);
+	 intel_region_unmap(intel, src->region);
+	 intel_region_unmap(intel, dst->region);
+      }
    }
-
 }
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
index 4060b9df78..c890b2a0d0 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
@@ -126,6 +126,7 @@ struct intel_mipmap_tree
 
 struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel,
                                                GLenum target,
+                                               GLenum base_format,
                                                GLenum internal_format,
                                                GLuint first_level,
                                                GLuint last_level,
@@ -148,6 +149,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch);
 
 void intel_miptree_reference(struct intel_mipmap_tree **dst,
@@ -194,6 +196,11 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
                                   GLuint x, GLuint y,
                                   GLuint w, GLuint h, GLuint d);
 
+void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                       GLuint level,
+                                       GLuint img, GLuint x, GLuint y,
+                                       GLuint offset);
+
 void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
                                     GLuint level,
                                     GLuint img, GLuint x, GLuint y);
@@ -218,10 +225,13 @@ void intel_miptree_image_copy(struct intel_context *intel,
 /* i915_mipmap_tree.c:
  */
 GLboolean i915_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean i945_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean brw_miptree_layout(struct intel_context *intel,
-			     struct intel_mipmap_tree *mt);
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index defb80f85d..a300141655 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -27,9 +27,12 @@
 
 #include "main/enums.h"
 #include "main/state.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/enable.h"
 #include "main/matrix.h"
+#include "main/texstate.h"
+#include "main/varray.h"
 #include "main/viewport.h"
 #include "swrast/swrast.h"
 #include "shader/arbprogram.h"
@@ -174,167 +177,6 @@ intel_check_blit_format(struct intel_region * region,
 }
 
 void
-intel_meta_set_passthrough_transform(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   intel->meta.saved_vp_x = ctx->Viewport.X;
-   intel->meta.saved_vp_y = ctx->Viewport.Y;
-   intel->meta.saved_vp_width = ctx->Viewport.Width;
-   intel->meta.saved_vp_height = ctx->Viewport.Height;
-   intel->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-   intel->internal_viewport_call = GL_FALSE;
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-void
-intel_meta_restore_transform(struct intel_context *intel)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(intel->meta.saved_matrix_mode);
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y,
-		  intel->meta.saved_vp_width, intel->meta.saved_vp_height);
-   intel->internal_viewport_call = GL_FALSE;
-}
-
-/**
- * Set up a vertex program to pass through the position and first texcoord
- * for pixel path.
- */
-void
-intel_meta_set_passthrough_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   static const char *vp =
-      "!!ARBvp1.0\n"
-      "TEMP vertexClip;\n"
-      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
-      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
-      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
-      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
-      "MOV result.position, vertexClip;\n"
-      "MOV result.texcoord[0], vertex.texcoord[0];\n"
-      "MOV result.color, vertex.color;\n"
-      "END\n";
-
-   assert(intel->meta.saved_vp == NULL);
-
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp,
-			    ctx->VertexProgram.Current);
-   if (intel->meta.passthrough_vp == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(vp), (const GLubyte *)vp);
-      _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp,
-			       ctx->VertexProgram.Current);
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.passthrough_vp);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &intel->meta.passthrough_vp->Base);
-
-   intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled;
-   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous vertex program after
- * intel_meta_set_passthrough_vertex_program()
- */
-void
-intel_meta_restore_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.saved_vp);
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &ctx->VertexProgram.Current->Base);
-
-   if (!intel->meta.saved_vp_enable)
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
- * program object.
- */
-void
-intel_meta_set_fragment_program(struct intel_context *intel,
-				struct gl_fragment_program **prog,
-				const char *prog_string)
-{
-   GLcontext *ctx = &intel->ctx;
-   assert(intel->meta.saved_fp == NULL);
-
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp,
-			    ctx->FragmentProgram.Current);
-   if (*prog == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(prog_string), (const GLubyte *)prog_string);
-      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
-      /* Note that DeletePrograms unbinds the program on us */
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
-
-   intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled;
-   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous fragment program after
- * intel_meta_set_fragment_program()
- */
-void
-intel_meta_restore_fragment_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
-			    intel->meta.saved_fp);
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
-			   &ctx->FragmentProgram.Current->Base);
-
-   if (!intel->meta.saved_fp_enable)
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-void
 intelInitPixelFuncs(struct dd_function_table *functions)
 {
    functions->Accum = _swrast_Accum;
@@ -346,12 +188,3 @@ intelInitPixelFuncs(struct dd_function_table *functions)
    functions->ReadPixels = intelReadPixels;
 }
 
-void
-intel_free_pixel_state(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
-   _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
-}
-
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index cb41fa182c..96a6dd17b2 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -31,16 +31,6 @@
 #include "main/mtypes.h"
 
 void intelInitPixelFuncs(struct dd_function_table *functions);
-void intel_meta_set_passthrough_transform(struct intel_context *intel);
-void intel_meta_restore_transform(struct intel_context *intel);
-void intel_meta_set_passthrough_vertex_program(struct intel_context *intel);
-void intel_meta_restore_vertex_program(struct intel_context *intel);
-void intel_meta_set_fragment_program(struct intel_context *intel,
-				     struct gl_fragment_program **prog,
-				     const char *prog_string);
-void intel_meta_restore_fragment_program(struct intel_context *intel);
-void intel_free_pixel_state(struct intel_context *intel);
-
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
 
@@ -76,6 +66,4 @@ void intelBitmap(GLcontext * ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLubyte * pixels);
 
-void intel_clear_tris(GLcontext *ctx, GLbitfield mask);
-
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index d137aef13d..b543a0bbc3 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -42,6 +42,7 @@
 #include "main/varray.h"
 #include "main/attrib.h"
 #include "main/enable.h"
+#include "main/viewport.h"
 #include "shader/arbprogram.h"
 #include "glapi/dispatch.h"
 #include "swrast/swrast.h"
@@ -92,19 +93,12 @@ static const GLubyte *map_pbo( GLcontext *ctx,
    return ADD_POINTERS(buf, bitmap);
 }
 
-static GLboolean test_bit( const GLubyte *src,
-			    GLuint bit )
+static GLboolean test_bit( const GLubyte *src, GLuint bit )
 {
    return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0;
 }
 
-static GLboolean test_msb_bit(const GLubyte *src, GLuint bit)
-{
-   return (src[bit/8] & (1<<(7 - (bit % 8)))) ? 1 : 0;
-}
-
-static void set_bit( GLubyte *dest,
-			  GLuint bit )
+static void set_bit( GLubyte *dest, GLuint bit )
 {
    dest[bit/8] |= 1 << (bit % 8);
 }
@@ -194,7 +188,7 @@ do_blit_bitmap( GLcontext *ctx,
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    GLfloat tmpColor[4];
    GLubyte ubcolor[4];
-   GLuint color8888, color565;
+   GLuint color;
    unsigned int num_cliprects;
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
@@ -232,8 +226,11 @@ do_blit_bitmap( GLcontext *ctx,
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[2], tmpColor[2]);
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]);
 
-   color8888 = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], ubcolor[2], ubcolor[3]);
-   color565 = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   if (dst->cpp == 2)
+      color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   else
+      color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1],
+				  ubcolor[2], ubcolor[3]);
 
    if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F))
       return GL_FALSE;
@@ -307,21 +304,21 @@ do_blit_bitmap( GLcontext *ctx,
 				   fb->Name == 0 ? GL_TRUE : GL_FALSE) == 0)
 		  continue;
 
-	       /* 
-		*/
-	       intelEmitImmediateColorExpandBlit( intel,
-						  dst->cpp,
-						  (GLubyte *)stipple, 
-						  sz,
-						  (dst->cpp == 2) ? color565 : color8888,
-						  dst->pitch,
-						  dst->buffer,
-						  0,
-						  dst->tiling,
-						  box_x + px,
-						  box_y + py,
-						  w, h,
-						  logic_op);
+	       if (!intelEmitImmediateColorExpandBlit(intel,
+						      dst->cpp,
+						      (GLubyte *)stipple,
+						      sz,
+						      color,
+						      dst->pitch,
+						      dst->buffer,
+						      0,
+						      dst->tiling,
+						      box_x + px,
+						      box_y + py,
+						      w, h,
+						      logic_op)) {
+		  return GL_FALSE;
+	       }
 	    } 
 	 } 
       }
@@ -360,11 +357,8 @@ intel_texture_bitmap(GLcontext * ctx,
       "END\n";
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLint old_active_texture;
-   GLubyte *unpacked_bitmap;
    GLubyte *a8_bitmap;
-   int x, y;
    GLfloat dst_z;
 
    /* We need a fragment program for the KIL effect */
@@ -424,22 +418,16 @@ intel_texture_bitmap(GLcontext * ctx,
       return GL_FALSE;
    }
 
-   /* Convert the A1 bitmap to an A8 format suitable for glTexImage */
    if (unpack->BufferObj->Name) {
       bitmap = map_pbo(ctx, width, height, unpack, bitmap);
       if (bitmap == NULL)
 	 return GL_TRUE;	/* even though this is an error, we're done */
    }
-   unpacked_bitmap = _mesa_unpack_bitmap(width, height, bitmap,
-					 unpack);
+
+   /* Convert the A1 bitmap to an A8 format suitable for glTexImage */
    a8_bitmap = _mesa_calloc(width * height);
-   for (y = 0; y < height; y++) {
-      for (x = 0; x < width; x++) {
-	 if (test_msb_bit(unpacked_bitmap, ALIGN(width, 8) * y + x))
-	    a8_bitmap[y * width + x] = 0xff;
-      }
-   }
-   _mesa_free(unpacked_bitmap);
+   _mesa_expand_bitmap(width, height, unpack, bitmap, a8_bitmap, width, 0xff);
+
    if (unpack->BufferObj->Name) {
       /* done with PBO so unmap it now */
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
@@ -473,15 +461,18 @@ intel_texture_bitmap(GLcontext * ctx,
 		    GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap);
    _mesa_free(a8_bitmap);
 
-   intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp);
+   meta_set_fragment_program(&intel->meta, &intel->meta.bitmap_fp, fp);
    _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
 				     ctx->Current.RasterColor);
-   intel_meta_set_passthrough_vertex_program(intel);
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_vertex_program(&intel->meta);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
 
+   /* RasterPos[2] already takes into account the DepthRange mapping. */
+   _mesa_DepthRange(0.0, 1.0);
+
    vertices[0][0] = dst_x;
    vertices[0][1] = dst_y;
    vertices[0][2] = dst_z;
@@ -499,25 +490,15 @@ intel_texture_bitmap(GLcontext * ctx,
    vertices[3][2] = dst_z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   meta_set_default_texrect(&intel->meta);
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
-   intel_meta_restore_fragment_program(intel);
-   intel_meta_restore_vertex_program(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
 
    _mesa_PopClientAttrib();
    _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index d50dd68092..07ca8f7ddb 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -26,18 +26,13 @@
  **************************************************************************/
 
 #include "main/glheader.h"
-#include "main/enums.h"
 #include "main/image.h"
 #include "main/state.h"
 #include "main/mtypes.h"
-#include "main/macros.h"
-#include "swrast/swrast.h"
+#include "drivers/common/meta.h"
 
-#include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_batchbuffer.h"
 #include "intel_buffers.h"
-#include "intel_blit.h"
 #include "intel_regions.h"
 #include "intel_pixel.h"
 
@@ -97,162 +92,6 @@ intel_check_copypixel_blit_fragment_ops(GLcontext * ctx)
 	    ctx->Color.BlendEnabled);
 }
 
-#ifdef I915
-/* Doesn't work for overlapping regions.  Could do a double copy or
- * just fallback.
- */
-static GLboolean
-do_texture_copypixels(GLcontext * ctx,
-                      GLint srcx, GLint srcy,
-                      GLsizei width, GLsizei height,
-                      GLint dstx, GLint dsty, GLenum type)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_region *dst = intel_drawbuf_region(intel);
-   struct intel_region *src = copypix_src_region(intel, type);
-   GLenum src_format;
-   GLenum src_type;
-
-   DBG("%s %d,%d %dx%d --> %d,%d\n", __FUNCTION__, 
-       srcx, srcy, width, height, dstx, dsty);
-
-   if (!src || !dst || type != GL_COLOR)
-      return GL_FALSE;
-
-   if (ctx->_ImageTransferState) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   /* Can't handle overlapping regions.  Don't have sufficient control
-    * over rasterization to pull it off in-place.  Punt on these for
-    * now.
-    * 
-    * XXX: do a copy to a temporary. 
-    */
-   if (src->buffer == dst->buffer) {
-      drm_clip_rect_t srcbox;
-      drm_clip_rect_t dstbox;
-      drm_clip_rect_t tmp;
-
-      srcbox.x1 = srcx;
-      srcbox.y1 = srcy;
-      srcbox.x2 = srcx + width;
-      srcbox.y2 = srcy + height;
-
-      if (ctx->Pixel.ZoomX > 0) {
-	 dstbox.x1 = dstx;
-	 dstbox.x2 = dstx + width * ctx->Pixel.ZoomX;
-      } else {
-	 dstbox.x1 = dstx + width * ctx->Pixel.ZoomX;
-	 dstbox.x2 = dstx;
-      }
-      if (ctx->Pixel.ZoomY > 0) {
-	 dstbox.y1 = dsty;
-	 dstbox.y2 = dsty + height * ctx->Pixel.ZoomY;
-      } else {
-	 dstbox.y1 = dsty + height * ctx->Pixel.ZoomY;
-	 dstbox.y2 = dsty;
-      }
-
-      DBG("src %d,%d %d,%d\n", srcbox.x1, srcbox.y1, srcbox.x2, srcbox.y2);
-      DBG("dst %d,%d %d,%d (%dx%d) (%f,%f)\n", dstbox.x1, dstbox.y1, dstbox.x2, dstbox.y2,
-	  width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
-
-      if (intel_intersect_cliprects(&tmp, &srcbox, &dstbox)) {
-         DBG("%s: regions overlap\n", __FUNCTION__);
-         return GL_FALSE;
-      }
-   }
-
-   intelFlush(&intel->ctx);
-
-   intel->vtbl.install_meta_state(intel);
-
-   /* Is this true?  Also will need to turn depth testing on according
-    * to state:
-    */
-   intel->vtbl.meta_no_stencil_write(intel);
-   intel->vtbl.meta_no_depth_write(intel);
-
-   /* Set the 3d engine to draw into the destination region:
-    */
-   intel->vtbl.meta_draw_region(intel, dst, intel->depth_region);
-
-   intel->vtbl.meta_import_pixel_state(intel);
-
-   if (src->cpp == 2) {
-      src_format = GL_RGB;
-      src_type = GL_UNSIGNED_SHORT_5_6_5;
-   }
-   else {
-      src_format = GL_BGRA;
-      src_type = GL_UNSIGNED_BYTE;
-   }
-
-   /* Set the frontbuffer up as a large rectangular texture.
-    */
-   if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0,
-                                         src->pitch,
-                                         src->height, src_format, src_type)) {
-      intel->vtbl.leave_meta_state(intel);
-      return GL_FALSE;
-   }
-
-
-   intel->vtbl.meta_texture_blend_replace(intel);
-
-   LOCK_HARDWARE(intel);
-
-   if (intel->driDrawable->numClipRects) {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-
-      srcy = dPriv->h - srcy - height;  /* convert from gl to hardware coords */
-
-      srcx += dPriv->x;
-      srcy += dPriv->y;
-
-      /* Clip against the source region.  This is the only source
-       * clipping we do.  XXX: Just set the texcord wrap mode to clamp
-       * or similar.
-       *
-       */
-      if (0) {
-         GLint orig_x = srcx;
-         GLint orig_y = srcy;
-
-         if (!_mesa_clip_to_region(0, 0, src->pitch, src->height,
-                                   &srcx, &srcy, &width, &height))
-            goto out;
-
-         dstx += srcx - orig_x;
-         dsty += (srcy - orig_y) * ctx->Pixel.ZoomY;
-      }
-
-      /* Just use the regular cliprect mechanism...  Does this need to
-       * even hold the lock???
-       */
-      intel->vtbl.meta_draw_quad(intel,
-				 dstx,
-				 dstx + width * ctx->Pixel.ZoomX,
-				 dPriv->h - (dsty + height * ctx->Pixel.ZoomY),
-				 dPriv->h - (dsty), 0, /* XXX: what z value? */
-				 0x00ff00ff,
-				 srcx, srcx + width, srcy, srcy + height);
-
-    out:
-      intel->vtbl.leave_meta_state(intel);
-      intel_batchbuffer_emit_mi_flush(intel->batch);
-   }
-   UNLOCK_HARDWARE(intel);
-
-   DBG("%s: success\n", __FUNCTION__);
-   return GL_TRUE;
-}
-#endif /* I915 */
-
 
 /**
  * CopyPixels with the blitter.  Don't support zooming, pixel transfer, etc.
@@ -272,6 +111,12 @@ do_blit_copypixels(GLcontext * ctx,
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
 
+   if (type == GL_DEPTH || type == GL_STENCIL) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
+      return GL_FALSE;
+   }
+
    /* Update draw buffer bounds */
    _mesa_update_state(ctx);
 
@@ -362,14 +207,16 @@ do_blit_copypixels(GLcontext * ctx,
 				   &clip_x, &clip_y, &clip_w, &clip_h))
             continue;
 
-         intelEmitCopyBlit(intel, dst->cpp,
-			   src->pitch, src->buffer, 0, src->tiling,
-			   dst->pitch, dst->buffer, 0, dst->tiling,
-			   clip_x + delta_x, clip_y + delta_y, /* srcx, srcy */
-			   clip_x, clip_y, /* dstx, dsty */
-			   clip_w, clip_h,
-			   ctx->Color.ColorLogicOpEnabled ?
-			   ctx->Color.LogicOp : GL_COPY);
+	 if (!intel_region_copy(intel,
+				dst, 0, clip_x, clip_y,
+				src, 0, clip_x + delta_x, clip_y + delta_y,
+				clip_w, clip_h,
+				ctx->Color.ColorLogicOpEnabled ?
+				ctx->Color.LogicOp : GL_COPY)) {
+	    DBG("%s: blit failure\n", __FUNCTION__);
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
       }
    }
 out:
@@ -392,12 +239,6 @@ intelCopyPixels(GLcontext * ctx,
    if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
       return;
 
-#ifdef I915
-   if (do_texture_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
-      return;
-#endif
-
-   DBG("fallback to _swrast_CopyPixels\n");
-
-   _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type);
+   /* this will use swrast if needed */
+   _mesa_meta_copy_pixels(ctx, srcx, srcy, width, height, destx, desty, type);
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index d80069dd58..7fbb89fd6a 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -29,8 +29,6 @@
 #include "main/enums.h"
 #include "main/image.h"
 #include "main/mtypes.h"
-#include "main/macros.h"
-#include "main/bufferobj.h"
 #include "main/teximage.h"
 #include "main/texenv.h"
 #include "main/texobj.h"
@@ -41,169 +39,22 @@
 #include "main/enable.h"
 #include "main/buffers.h"
 #include "main/fbobject.h"
-#include "main/renderbuffer.h"
 #include "main/depth.h"
 #include "main/hash.h"
 #include "main/blend.h"
-#include "glapi/dispatch.h"
 #include "swrast/swrast.h"
+#include "drivers/common/meta.h"
 
-#include "intel_screen.h"
 #include "intel_context.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
 #include "intel_regions.h"
 #include "intel_pixel.h"
-#include "intel_buffer_objects.h"
 #include "intel_fbo.h"
 
-static GLboolean
-intel_texture_drawpixels(GLcontext * ctx,
-			 GLint x, GLint y,
-			 GLsizei width, GLsizei height,
-			 GLenum format,
-			 GLenum type,
-			 const struct gl_pixelstore_attrib *unpack,
-			 const GLvoid *pixels)
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLuint texname;
-   GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
-   GLfloat z;
-   GLint old_active_texture;
-   GLenum internalFormat;
-
-   /* We're going to mess with texturing with no regard to existing texture
-    * state, so if there is some set up we have to bail.
-    */
-   if (ctx->Texture._EnabledUnits != 0) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: texturing enabled\n");
-      return GL_FALSE;
-   }
-
-   /* Can't do textured DrawPixels with a fragment program, unless we were
-    * to generate a new program that sampled our texture and put the results
-    * in the fragment color before the user's program started.
-    */
-   if (ctx->FragmentProgram.Enabled) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: fragment program enabled\n");
-      return GL_FALSE;
-   }
-
-   /* We don't have a way to generate fragments with stencil values which
-    * will set the resulting stencil value.
-    */
-   if (format == GL_STENCIL_INDEX || format == GL_DEPTH_STENCIL)
-      return GL_FALSE;
-
-   /* Check that we can load in a texture this big. */
-   if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
-       height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: bitmap too large (%dx%d)\n",
-		 width, height);
-      return GL_FALSE;
-   }
-
-   /* To do DEPTH_COMPONENT, we would need to change our setup to not draw to
-    * the color buffer, and sample the texture values into the fragment depth
-    * in a program.
-    */
-   if (format == GL_DEPTH_COMPONENT) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr,
-		 "glDrawPixels() fallback: format == GL_DEPTH_COMPONENT\n");
-      return GL_FALSE;
-   }
-
-   if (!ctx->Extensions.ARB_texture_non_power_of_two &&
-       (!is_power_of_two(width) || !is_power_of_two(height))) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr,
-		 "glDrawPixels() fallback: NPOT texture\n");
-      return GL_FALSE;
-   }
-
-   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
-		    GL_CURRENT_BIT);
-   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
-
-   /* XXX: pixel store stuff */
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-
-   old_active_texture = ctx->Texture.CurrentUnit;
-   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
-   _mesa_Enable(GL_TEXTURE_2D);
-   _mesa_GenTextures(1, &texname);
-   _mesa_BindTexture(GL_TEXTURE_2D, texname);
-   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
-   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
-   if (type == GL_ALPHA)
-      internalFormat = GL_ALPHA;
-   else
-      internalFormat = GL_RGBA;
-   _mesa_TexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format,
-		    type, pixels);
-
-   intel_meta_set_passthrough_transform(intel);
-
-   /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
-   z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
-
-   /* Create the vertex buffer based on the current raster pos.  The x and y
-    * we're handed are ctx->Current.RasterPos[0,1] rounded to integers.
-    * We also apply the depth.  However, the W component is already multiplied
-    * into ctx->Current.RasterPos[0,1,2] and we can ignore it at this point.
-    */
-   vertices[0][0] = x;
-   vertices[0][1] = y;
-   vertices[0][2] = z;
-   vertices[0][3] = 1.0;
-   vertices[1][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[1][1] = y;
-   vertices[1][2] = z;
-   vertices[1][3] = 1.0;
-   vertices[2][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[2][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[2][2] = z;
-   vertices[2][3] = 1.0;
-   vertices[3][0] = x;
-   vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[3][2] = z;
-   vertices[3][3] = 1.0;
-
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
-   _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
-   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
-
-   intel_meta_restore_transform(intel);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
-   _mesa_PopClientAttrib();
-   _mesa_PopAttrib();
-
-   _mesa_DeleteTextures(1, &texname);
-
-   return GL_TRUE;
-}
 
+/** XXX compare perf of this vs. _mesa_meta_draw_pixels(STENCIL) */
 static GLboolean
 intel_stencil_drawpixels(GLcontext * ctx,
 			 GLint x, GLint y,
@@ -216,14 +67,14 @@ intel_stencil_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname, rb_name, fb_name, old_fb_name;
    GLfloat vertices[4][2];
-   GLfloat texcoords[4][2];
    struct intel_renderbuffer *irb;
    struct intel_renderbuffer *depth_irb;
    struct gl_renderbuffer *rb;
    struct gl_pixelstore_attrib old_unpack;
    GLstencil *stencil_pixels;
-   int row;
+   int row, y1, y2;
    GLint old_active_texture;
+   GLboolean rendering_to_fbo = ctx->DrawBuffer->Name != 0;
 
    if (format != GL_STENCIL_INDEX)
       return GL_FALSE;
@@ -358,34 +209,35 @@ intel_stencil_drawpixels(GLcontext * ctx,
    ctx->Unpack = old_unpack;
    _mesa_free(stencil_pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
+   /* Since we're rendering to the framebuffer as if it was an FBO,
+    * if it's the window system we have to flip the coordinates.
+    */
+   if (rendering_to_fbo) {
+      y1 = y;
+      y2 = y + height * ctx->Pixel.ZoomY;
+   } else {
+      y1 = irb->Base.Height - (y + height * ctx->Pixel.ZoomY);
+      y2 = irb->Base.Height - y;
+   }
    vertices[0][0] = x;
-   vertices[0][1] = y;
+   vertices[0][1] = y1;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[1][1] = y;
+   vertices[1][1] = y1;
    vertices[2][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[2][1] = y + height * ctx->Pixel.ZoomY;
+   vertices[2][1] = y2;
    vertices[3][0] = x;
-   vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
+   vertices[3][1] = y2;
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name);
@@ -409,17 +261,25 @@ intelDrawPixels(GLcontext * ctx,
                 const struct gl_pixelstore_attrib *unpack,
                 const GLvoid * pixels)
 {
-   if (intel_texture_drawpixels(ctx, x, y, width, height, format, type,
-				unpack, pixels))
-      return;
-
+#if 0
+   /* XXX this function doesn't seem to work reliably even when all
+    * the pre-requisite conditions are met.
+    * Note that this function is never hit with conform.
+    * Fall back to swrast because even the _mesa_meta_draw_pixels() approach
+    * isn't working because of an apparent stencil bug.
+    */
    if (intel_stencil_drawpixels(ctx, x, y, width, height, format, type,
 				unpack, pixels))
       return;
+#else
+   (void) intel_stencil_drawpixels; /* silence warning */
+   if (format == GL_STENCIL_INDEX) {
+      _swrast_DrawPixels(ctx, x, y, width, height, format, type,
+                         unpack, pixels);
+      return;
+   }
+#endif
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
-
-   _swrast_DrawPixels(ctx, x, y, width, height, format, type,
-		      unpack, pixels);
+   _mesa_meta_draw_pixels(ctx, x, y, width, height, format, type,
+                          unpack, pixels);
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c
index 0370255614..8713463ace 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c
@@ -180,7 +180,7 @@ do_blit_readpixels(GLcontext * ctx,
    if (!src)
       return GL_FALSE;
 
-   if (dst) {
+   if (pack->BufferObj->Name) {
       /* XXX This validation should be done by core mesa:
        */
       if (!_mesa_validate_pbo_access(2, pack, width, height, 1,
@@ -261,16 +261,19 @@ do_blit_readpixels(GLcontext * ctx,
          if (!intel_intersect_cliprects(&rect, &src_rect, &box[i]))
             continue;
 
-         intelEmitCopyBlit(intel,
-                           src->cpp,
-                           src->pitch, src->buffer, 0, src->tiling,
-                           rowLength, dst_buffer, dst_offset, GL_FALSE,
-                           rect.x1,
-                           rect.y1,
-                           rect.x1 - src_rect.x1,
-                           rect.y2 - src_rect.y2,
-                           rect.x2 - rect.x1, rect.y2 - rect.y1,
-			   GL_COPY);
+         if (!intelEmitCopyBlit(intel,
+				src->cpp,
+				src->pitch, src->buffer, 0, src->tiling,
+				rowLength, dst_buffer, dst_offset, GL_FALSE,
+				rect.x1,
+				rect.y1,
+				rect.x1 - src_rect.x1,
+				rect.y2 - src_rect.y2,
+				rect.x2 - rect.x1, rect.y2 - rect.y1,
+				GL_COPY)) {
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
       }
    }
    UNLOCK_HARDWARE(intel);
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 57ac8f0cc1..d19f1bae34 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -189,6 +189,19 @@
 
 #define S7_DEPTH_OFFSET_CONST_MASK     ~0
 
+/* p143 */
+#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
+/* Dword 1 */
+#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
+#define BUF_3D_ID_DEPTH 	(0x7<<24)
+#define BUF_3D_USE_FENCE	(1<<23)
+#define BUF_3D_TILED_SURFACE	(1<<22)
+#define BUF_3D_TILE_WALK_X	0
+#define BUF_3D_TILE_WALK_Y	(1<<21)
+#define BUF_3D_PITCH(x)         (((x)/4)<<2)
+/* Dword 2 */
+#define BUF_3D_ADDR(x)		((x) & ~0x3)
+
 /* Primitive dispatch on 830-945 */
 #define _3DPRIMITIVE			(CMD_3D | (0x1f << 24))
 #define PRIM_INDIRECT            (1<<23)
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index b8d2dec4cb..a86c66a844 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -52,17 +52,77 @@
 
 #define FILE_DEBUG_FLAG DEBUG_REGION
 
+/* This should be set to the maximum backtrace size desired.
+ * Set it to 0 to disable backtrace debugging.
+ */
+#define DEBUG_BACKTRACE_SIZE 0
+
+#if DEBUG_BACKTRACE_SIZE == 0
+/* Use the standard debug output */
+#define _DBG(...) DBG(__VA_ARGS__)
+#else
+/* Use backtracing debug output */
+#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}
+
+/* Backtracing debug support */
+#include <execinfo.h>
+
+static void
+debug_backtrace(void)
+{
+   void *trace[DEBUG_BACKTRACE_SIZE];
+   char **strings = NULL;
+   int traceSize;
+   register int i;
+
+   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
+   strings = backtrace_symbols(trace, traceSize);
+   if (strings == NULL) {
+      DBG("no backtrace:");
+      return;
+   }
+
+   /* Spit out all the strings with a colon separator.  Ignore
+    * the first, since we don't really care about the call
+    * to debug_backtrace() itself.  Skip until the final "/" in
+    * the trace to avoid really long lines.
+    */
+   for (i = 1; i < traceSize; i++) {
+      char *p = strings[i], *slash = strings[i];
+      while (*p) {
+         if (*p++ == '/') {
+            slash = p;
+         }
+      }
+
+      DBG("%s:", slash);
+   }
+
+   /* Free up the memory, and we're done */
+   free(strings);
+}
+
+#endif
+
+
+
 /* XXX: Thread safety?
  */
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   intelFlush(&intel->ctx);
+
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)
          intel_region_cow(intel, region);
 
-      dri_bo_map(region->buffer, GL_TRUE);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_map_gtt(region->buffer);
+      else
+	 dri_bo_map(region->buffer, GL_TRUE);
       region->map = region->buffer->virtual;
    }
 
@@ -72,9 +132,13 @@ intel_region_map(struct intel_context *intel, struct intel_region *region)
 void
 intel_region_unmap(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!--region->map_refcount) {
-      dri_bo_unmap(region->buffer);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_unmap_gtt(region->buffer);
+      else
+	 dri_bo_unmap(region->buffer);
       region->map = NULL;
    }
 }
@@ -87,10 +151,10 @@ intel_region_alloc_internal(struct intel_context *intel,
 {
    struct intel_region *region;
 
-   DBG("%s\n", __FUNCTION__);
-
-   if (buffer == NULL)
+   if (buffer == NULL) {
+      _DBG("%s <-- NULL\n", __FUNCTION__);
       return NULL;
+   }
 
    region = calloc(sizeof(*region), 1);
    region->cpp = cpp;
@@ -104,15 +168,33 @@ intel_region_alloc_internal(struct intel_context *intel,
    region->tiling = I915_TILING_NONE;
    region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
+   _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
 }
 
 struct intel_region *
 intel_region_alloc(struct intel_context *intel,
+		   uint32_t tiling,
                    GLuint cpp, GLuint width, GLuint height, GLuint pitch,
 		   GLboolean expect_accelerated_upload)
 {
    dri_bo *buffer;
+   struct intel_region *region;
+
+   /* If we're tiled, our allocations are in 8 or 32-row blocks, so
+    * failure to align our height means that we won't allocate enough pages.
+    *
+    * If we're untiled, we still have to align to 2 rows high because the
+    * data port accesses 2x2 blocks even if the bottom row isn't to be
+    * rendered, so failure to align means we could walk off the end of the
+    * GTT and fault.
+    */
+   if (tiling == I915_TILING_X)
+      height = ALIGN(height, 8);
+   else if (tiling == I915_TILING_Y)
+      height = ALIGN(height, 32);
+   else
+      height = ALIGN(height, 2);
 
    /* If we're untiled, we have to align to 2 rows high because the
     * data port accesses 2x2 blocks even if the bottom row isn't to be
@@ -129,7 +211,16 @@ intel_region_alloc(struct intel_context *intel,
 				  pitch * cpp * height, 64);
    }
 
-   return intel_region_alloc_internal(intel, cpp, width, height, pitch, buffer);
+   region = intel_region_alloc_internal(intel, cpp, width, height,
+					pitch, buffer);
+
+   if (tiling != I915_TILING_NONE) {
+      assert(((pitch * cpp) & 127) == 0);
+      drm_intel_bo_set_tiling(buffer, &tiling, pitch * cpp);
+      drm_intel_bo_get_tiling(buffer, &region->tiling, &region->bit_6_swizzle);
+   }
+
+   return region;
 }
 
 struct intel_region *
@@ -165,7 +256,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
+      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -179,10 +270,12 @@ intel_region_release(struct intel_region **region_handle)
 {
    struct intel_region *region = *region_handle;
 
-   if (region == NULL)
+   if (region == NULL) {
+      _DBG("%s NULL\n", __FUNCTION__);
       return;
+   }
 
-   DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
+   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -256,9 +349,7 @@ intel_region_data(struct intel_context *intel,
                   const void *src, GLuint src_pitch,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height)
 {
-   GLboolean locked = GL_FALSE;
-
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -271,39 +362,33 @@ intel_region_data(struct intel_context *intel,
          intel_region_cow(intel, dst);
    }
 
-   if (!intel->locked) {
-      LOCK_HARDWARE(intel);
-      locked = GL_TRUE;
-   }
-
+   LOCK_HARDWARE(intel);
    _mesa_copy_rect(intel_region_map(intel, dst) + dst_offset,
                    dst->cpp,
                    dst->pitch,
                    dstx, dsty, width, height, src, src_pitch, srcx, srcy);
 
    intel_region_unmap(intel, dst);
-
-   if (locked)
-      UNLOCK_HARDWARE(intel);
-
+   UNLOCK_HARDWARE(intel);
 }
 
 /* Copy rectangular sub-regions. Need better logic about when to
  * push buffers into AGP - will currently do so whenever possible.
  */
-void
+GLboolean
 intel_region_copy(struct intel_context *intel,
                   struct intel_region *dst,
                   GLuint dst_offset,
                   GLuint dstx, GLuint dsty,
                   struct intel_region *src,
                   GLuint src_offset,
-                  GLuint srcx, GLuint srcy, GLuint width, GLuint height)
+                  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
-      return;
+      return GL_FALSE;
 
    if (dst->pbo) {
       if (dstx == 0 &&
@@ -315,41 +400,12 @@ intel_region_copy(struct intel_context *intel,
 
    assert(src->cpp == dst->cpp);
 
-   intelEmitCopyBlit(intel,
-                     dst->cpp,
-                     src->pitch, src->buffer, src_offset, src->tiling,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     srcx, srcy, dstx, dsty, width, height,
-		     GL_COPY);
-}
-
-/* Fill a rectangular sub-region.  Need better logic about when to
- * push buffers into AGP - will currently do so whenever possible.
- */
-void
-intel_region_fill(struct intel_context *intel,
-                  struct intel_region *dst,
-                  GLuint dst_offset,
-                  GLuint dstx, GLuint dsty,
-                  GLuint width, GLuint height, GLuint color)
-{
-   DBG("%s\n", __FUNCTION__);
-
-   if (intel == NULL)
-      return;   
-
-   if (dst->pbo) {
-      if (dstx == 0 &&
-          dsty == 0 && width == dst->pitch && height == dst->height)
-         intel_region_release_pbo(intel, dst);
-      else
-         intel_region_cow(intel, dst);
-   }
-
-   intelEmitFillBlit(intel,
-                     dst->cpp,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     dstx, dsty, width, height, color);
+   return intelEmitCopyBlit(intel,
+			    dst->cpp,
+			    src->pitch, src->buffer, src_offset, src->tiling,
+			    dst->pitch, dst->buffer, dst_offset, dst->tiling,
+			    srcx, srcy, dstx, dsty, width, height,
+			    logicop);
 }
 
 /* Attach to a pbo, discarding our data.  Effectively zero-copy upload
@@ -360,9 +416,13 @@ intel_region_attach_pbo(struct intel_context *intel,
                         struct intel_region *region,
                         struct intel_buffer_object *pbo)
 {
+   dri_bo *buffer;
+
    if (region->pbo == pbo)
       return;
 
+   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);
+
    /* If there is already a pbo attached, break the cow tie now.
     * Don't call intel_region_release_pbo() as that would
     * unnecessarily allocate a new buffer we would have to immediately
@@ -378,10 +438,13 @@ intel_region_attach_pbo(struct intel_context *intel,
       region->buffer = NULL;
    }
 
+   /* make sure pbo has a buffer of its own */
+   buffer = intel_bufferobj_buffer(intel, pbo, INTEL_WRITE_FULL);
+
    region->pbo = pbo;
    region->pbo->region = region;
-   dri_bo_reference(pbo->buffer);
-   region->buffer = pbo->buffer;
+   dri_bo_reference(buffer);
+   region->buffer = buffer;
 }
 
 
@@ -392,6 +455,7 @@ void
 intel_region_release_pbo(struct intel_context *intel,
                          struct intel_region *region)
 {
+   _DBG("%s %p\n", __FUNCTION__, region);
    assert(region->buffer == region->pbo->buffer);
    region->pbo->region = NULL;
    region->pbo = NULL;
@@ -410,34 +474,27 @@ void
 intel_region_cow(struct intel_context *intel, struct intel_region *region)
 {
    struct intel_buffer_object *pbo = region->pbo;
-   GLboolean was_locked = intel->locked;
-
-   if (intel == NULL)
-      return;
+   GLboolean ok;
 
    intel_region_release_pbo(intel, region);
 
    assert(region->cpp * region->pitch * region->height == pbo->Base.Size);
 
-   DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size);
+   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);
 
    /* Now blit from the texture buffer to the new buffer: 
     */
 
-   was_locked = intel->locked;
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
-   intelEmitCopyBlit(intel,
-		     region->cpp,
-		     region->pitch, region->buffer, 0, region->tiling,
-		     region->pitch, pbo->buffer, 0, region->tiling,
-		     0, 0, 0, 0,
-		     region->pitch, region->height,
-		     GL_COPY);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   LOCK_HARDWARE(intel);
+   ok = intelEmitCopyBlit(intel,
+                          region->cpp,
+                          region->pitch, pbo->buffer, 0, region->tiling,
+                          region->pitch, region->buffer, 0, region->tiling,
+                          0, 0, 0, 0,
+                          region->pitch, region->height,
+                          GL_COPY);
+   assert(ok);
+   UNLOCK_HARDWARE(intel);
 }
 
 dri_bo *
@@ -466,6 +523,10 @@ intel_recreate_static(struct intel_context *intel,
    if (region == NULL) {
       region = calloc(sizeof(*region), 1);
       region->refcount = 1;
+      _DBG("%s creating new region %p\n", __FUNCTION__, region);
+   }
+   else {
+      _DBG("%s %p\n", __FUNCTION__, region);
    }
 
    if (intel->ctx.Visual.rgbBits == 24)
diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h
index 45e2bf4e77..0d379bdc6e 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.h
+++ b/src/mesa/drivers/dri/intel/intel_regions.h
@@ -73,7 +73,8 @@ struct intel_region
  * copied by calling intel_reference_region().
  */
 struct intel_region *intel_region_alloc(struct intel_context *intel,
-                                        GLuint cpp, GLuint width,
+                                        uint32_t tiling,
+					GLuint cpp, GLuint width,
                                         GLuint height, GLuint pitch,
 					GLboolean expect_accelerated_upload);
 
@@ -109,21 +110,15 @@ void intel_region_data(struct intel_context *intel,
 
 /* Copy rectangular sub-regions
  */
-void intel_region_copy(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       struct intel_region *src,
-                       GLuint src_offset,
-                       GLuint srcx, GLuint srcy, GLuint width, GLuint height);
-
-/* Fill a rectangular sub-region
- */
-void intel_region_fill(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       GLuint width, GLuint height, GLuint color);
+GLboolean
+intel_region_copy(struct intel_context *intel,
+		  struct intel_region *dest,
+		  GLuint dest_offset,
+		  GLuint destx, GLuint desty,
+		  struct intel_region *src,
+		  GLuint src_offset,
+		  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop);
 
 /* Helpers for zerocopy uploads, particularly texture image uploads:
  */
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index f810850ef5..1b8c56e68d 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -49,6 +49,10 @@
 #include "i915_drm.h"
 #include "i830_dri.h"
 
+#define DRI_CONF_TEXTURE_TILING(def) \
+	DRI_CONF_OPT_BEGIN(texture_tiling, bool, def)		\
+		DRI_CONF_DESC(en, "Enable texture tiling")	\
+	DRI_CONF_OPT_END					\
 
 PUBLIC const char __driConfigOptions[] =
    DRI_CONF_BEGIN
@@ -64,6 +68,17 @@ PUBLIC const char __driConfigOptions[] =
 	    DRI_CONF_ENUM(1, "Enable reuse of all sizes of buffer objects")
 	 DRI_CONF_DESC_END
       DRI_CONF_OPT_END
+
+#ifdef I915
+     DRI_CONF_TEXTURE_TILING(false)
+#else
+     DRI_CONF_TEXTURE_TILING(true)
+#endif
+
+      DRI_CONF_OPT_BEGIN(early_z, bool, false)
+	 DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).")
+      DRI_CONF_OPT_END
+
    DRI_CONF_SECTION_END
    DRI_CONF_SECTION_QUALITY
       DRI_CONF_FORCE_S3TC_ENABLE(false)
@@ -76,7 +91,7 @@ PUBLIC const char __driConfigOptions[] =
    DRI_CONF_SECTION_END
 DRI_CONF_END;
 
-const GLuint __driNConfigOptions = 8;
+const GLuint __driNConfigOptions = 10;
 
 #ifdef USE_NEW_INTERFACE
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
@@ -587,7 +602,7 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen)
    GLboolean gem_supported;
    struct drm_i915_getparam gp;
    __DRIscreenPrivate *spriv = intelScreen->driScrnPriv;
-   int num_fences;
+   int num_fences = 0;
 
    intelScreen->no_hw = getenv("INTEL_NO_HW") != NULL;
 
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index 34b78ebc1a..8df4990880 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -501,7 +501,7 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map)
 
 
 /**
- * Prepare for softare rendering.  Map current read/draw framebuffers'
+ * Prepare for software rendering.  Map current read/draw framebuffers'
  * renderbuffes and all currently bound texture objects.
  *
  * Old note: Moved locking out to get reasonable span performance.
@@ -526,7 +526,7 @@ intelSpanRenderStart(GLcontext * ctx)
 }
 
 /**
- * Called when done softare rendering.  Unmap the buffers we mapped in
+ * Called when done software rendering.  Unmap the buffers we mapped in
  * the above function.
  */
 void
diff --git a/src/mesa/drivers/dri/intel/intel_syncobj.c b/src/mesa/drivers/dri/intel/intel_syncobj.c
new file mode 100644
index 0000000000..1286fe929b
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_syncobj.c
@@ -0,0 +1,132 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file intel_syncobj.c
+ *
+ * Support for ARB_sync
+ *
+ * ARB_sync is implemented by flushing the current batchbuffer and keeping a
+ * reference on it.  We can then check for completion or wait for compeltion
+ * using the normal buffer object mechanisms.  This does mean that if an
+ * application is using many sync objects, it will emit small batchbuffers
+ * which may end up being a significant overhead.  In other tests of removing
+ * gratuitous batchbuffer syncs in Mesa, it hasn't appeared to be a significant
+ * performance bottleneck, though.
+ */
+
+#include "main/simple_list.h"
+#include "main/imports.h"
+
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_reg.h"
+
+static struct gl_sync_object *
+intel_new_sync_object(GLcontext *ctx, GLuint id)
+{
+   struct intel_sync_object *sync;
+
+   sync = _mesa_calloc(sizeof(struct intel_sync_object));
+
+   return &sync->Base;
+}
+
+static void
+intel_delete_sync_object(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   drm_intel_bo_unreference(sync->bo);
+   _mesa_free(sync);
+}
+
+static void
+intel_fence_sync(GLcontext *ctx, struct gl_sync_object *s,
+	       GLenum condition, GLbitfield flags)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   assert(condition == GL_SYNC_GPU_COMMANDS_COMPLETE);
+   intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   sync->bo = intel->batch->buf;
+   drm_intel_bo_reference(sync->bo);
+
+   intelFlush(ctx);
+}
+
+/* We ignore the user-supplied timeout.  This is weaselly -- we're allowed to
+ * round to an implementation-dependent accuracy, and right now our
+ * implementation "rounds" to the wait-forever value.
+ *
+ * The fix would be a new kernel function to do the GTT transition with a
+ * timeout.
+ */
+static void intel_client_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo) {
+      drm_intel_bo_wait_rendering(sync->bo);
+      s->StatusFlag = 1;
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+   }
+}
+
+/* We have nothing to do for WaitSync.  Our GL command stream is sequential,
+ * so given that the sync object has already flushed the batchbuffer,
+ * any batchbuffers coming after this waitsync will naturally not occur until
+ * the previous one is done.
+ */
+static void intel_server_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+}
+
+static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo && drm_intel_bo_busy(sync->bo)) {
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+      s->StatusFlag = 1;
+   }
+}
+
+void intel_init_syncobj_functions(struct dd_function_table *functions)
+{
+   functions->NewSyncObject = intel_new_sync_object;
+   functions->DeleteSyncObject = intel_delete_sync_object;
+   functions->FenceSync = intel_fence_sync;
+   functions->CheckSync = intel_check_sync;
+   functions->ClientWaitSync = intel_client_wait_sync;
+   functions->ServerWaitSync = intel_server_wait_sync;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index ae0994b183..df63f29a42 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -158,81 +158,11 @@ timed_memcpy(void *dest, const void *src, size_t n)
 }
 #endif /* DO_DEBUG */
 
-/**
- * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
- * level).
- *
- * The texture object's miptree must be mapped.
- *
- * It would be really nice if this was just called by Mesa whenever mipmaps
- * needed to be regenerated, rather than us having to remember to do so in
- * each texture image modification path.
- *
- * This function should also include an accelerated path.
- */
-void
-intel_generate_mipmap(GLcontext *ctx, GLenum target,
-                      struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
-   int face, i;
-
-   _mesa_generate_mipmap(ctx, target, texObj);
-
-   /* Update the level information in our private data in the new images, since
-    * it didn't get set as part of a normal TexImage path.
-    */
-   for (face = 0; face < nr_faces; face++) {
-      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
-         struct intel_texture_image *intelImage;
-
-	 intelImage = intel_texture_image(texObj->Image[face][i]);
-	 if (intelImage == NULL)
-	    break;
-
-	 intelImage->level = i;
-	 intelImage->face = face;
-	 /* Unreference the miptree to signal that the new Data is a bare
-	  * pointer from mesa.
-	  */
-	 intel_miptree_release(intel, &intelImage->mt);
-      }
-   }
-}
-
-static void intelGenerateMipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-
-   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
-   intel_generate_mipmap(ctx, target, texObj);
-   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
-}
-
 void
 intelInitTextureFuncs(struct dd_function_table *functions)
 {
    functions->ChooseTextureFormat = intelChooseTextureFormat;
-   functions->TexImage1D = intelTexImage1D;
-   functions->TexImage2D = intelTexImage2D;
-   functions->TexImage3D = intelTexImage3D;
-   functions->TexSubImage1D = intelTexSubImage1D;
-   functions->TexSubImage2D = intelTexSubImage2D;
-   functions->TexSubImage3D = intelTexSubImage3D;
-   functions->CopyTexImage1D = intelCopyTexImage1D;
-   functions->CopyTexImage2D = intelCopyTexImage2D;
-   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
-   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
-   functions->GetTexImage = intelGetTexImage;
-   functions->GenerateMipmap = intelGenerateMipmap;
-
-   /* compressed texture functions */
-   functions->CompressedTexImage2D = intelCompressedTexImage2D;
-   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
-   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+   functions->GenerateMipmap = intel_generate_mipmap;
 
    functions->NewTextureObject = intelNewTextureObject;
    functions->NewTextureImage = intelNewTextureImage;
diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h
index f5372d82fb..471aa2a240 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.h
+++ b/src/mesa/drivers/dri/intel/intel_tex.h
@@ -35,116 +35,17 @@
 
 void intelInitTextureFuncs(struct dd_function_table *functions);
 
+void intelInitTextureImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureSubImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureCopyImageFuncs(struct dd_function_table *functions);
+
 const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx,
                                                          GLint internalFormat,
                                                          GLenum format,
                                                          GLenum type);
 
-
-void intelTexImage3D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint depth,
-                     GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage3D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset, GLint yoffset, GLint zoffset,
-                        GLsizei width, GLsizei height, GLsizei depth,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelTexImage2D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage2D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset, GLint yoffset,
-                        GLsizei width, GLsizei height,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelTexImage1D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage1D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset,
-                        GLsizei width,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLint border);
-
-void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLsizei height,
-                         GLint border);
-
-void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint x, GLint y, GLsizei width);
-
-void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint yoffset,
-                            GLint x, GLint y, GLsizei width, GLsizei height);
-
-void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
-                      GLenum format, GLenum type, GLvoid * pixels,
-                      struct gl_texture_object *texObj,
-                      struct gl_texture_image *texImage);
-
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage );
-
-void intelCompressedTexSubImage2D(GLcontext * ctx,
-				  GLenum target,
-				  GLint level,
-				  GLint xoffset, GLint yoffset,
-				  GLsizei width, GLsizei height,
-				  GLenum format, GLsizei imageSize,
-				  const GLvoid * pixels,
-				  struct gl_texture_object *texObj,
-				  struct gl_texture_image *texImage);
-
-void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
-				GLvoid *pixels,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage);
-
 void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		       unsigned long long offset, GLint depth, GLuint pitch);
 void intelSetTexBuffer(__DRIcontext *pDRICtx,
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 90bbb8c6bb..028b49c14d 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -73,11 +73,8 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat)
       return NULL;
    case GL_RGBA:
    case GL_RGBA8:
-      return intel_readbuf_region(intel);
    case GL_RGB:
-      if (intel->ctx.Visual.rgbBits == 16)
-         return intel_readbuf_region(intel);
-      return NULL;
+      return intel_readbuf_region(intel);
    default:
       return NULL;
    }
@@ -99,14 +96,24 @@ do_copy_texsubimage(struct intel_context *intel,
 
    if (!intelImage->mt || !src) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s fail %p %p\n",
-		 __FUNCTION__, intelImage->mt, src);
+	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
+		 __FUNCTION__, intelImage->mt, src, internalFormat);
+      return GL_FALSE;
+   }
+
+   if (intelImage->mt->cpp != src->cpp) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s fail %d vs %d cpp\n",
+		 __FUNCTION__, intelImage->mt->cpp, src->cpp);
       return GL_FALSE;
    }
 
    intelFlush(ctx);
    LOCK_HARDWARE(intel);
    {
+      drm_intel_bo *dst_bo = intel_region_buffer(intel,
+						 intelImage->mt->region,
+						 INTEL_WRITE_PART);
       GLuint image_offset = intel_miptree_image_offset(intelImage->mt,
                                                        intelImage->face,
                                                        intelImage->level);
@@ -118,8 +125,12 @@ do_copy_texsubimage(struct intel_context *intel,
       dstx += x - orig_x;
       dsty += y - orig_y;
 
-      /* image_offset may be non-page-aligned, but that's illegal for tiling. */
-      assert(intelImage->mt->region->tiling == I915_TILING_NONE);
+      /* Can't blit to tiled buffers with non-tile-aligned offset. */
+      if (intelImage->mt->region->tiling != I915_TILING_NONE &&
+	  (image_offset & 4095) != 0) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
 
       if (ctx->ReadBuffer->Name == 0) {
 	 /* reading from a window, adjust x, y */
@@ -140,35 +151,35 @@ do_copy_texsubimage(struct intel_context *intel,
 	 src_pitch = src->pitch;
       }
 
-      intelEmitCopyBlit(intel,
-			intelImage->mt->cpp,
-			src_pitch,
-			src->buffer,
-			0,
-			src->tiling,
-			intelImage->mt->pitch,
-			intelImage->mt->region->buffer,
-			image_offset,
-			intelImage->mt->region->tiling,
-			x, y, dstx, dsty, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_pitch,
+			     src->buffer,
+			     0,
+			     src->tiling,
+			     intelImage->mt->pitch,
+			     dst_bo,
+			     image_offset,
+			     intelImage->mt->region->tiling,
+			     x, y, dstx, dsty, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
 
    UNLOCK_HARDWARE(intel);
 
    /* GL_SGIS_generate_mipmap */
    if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+      intel_generate_mipmap(ctx, target, texObj);
    }
 
    return GL_TRUE;
 }
 
 
-
-
-
-void
+static void
 intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLint border)
@@ -214,7 +225,8 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                            width, border);
 }
 
-void
+
+static void
 intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLsizei height,
@@ -262,7 +274,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-void
+static void
 intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint x, GLint y, GLsizei width)
 {
@@ -287,8 +299,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-
-void
+static void
 intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint yoffset,
                        GLint x, GLint y, GLsizei width, GLsizei height)
@@ -301,7 +312,6 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
       _mesa_select_tex_image(ctx, texObj, target, level);
    GLenum internalFormat = texImage->InternalFormat;
 
-
    /* Need to check texture is compatible with source format. 
     */
 
@@ -316,3 +326,13 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                                  xoffset, yoffset, x, y, width, height);
    }
 }
+
+
+void
+intelInitTextureCopyImageFuncs(struct dd_function_table *functions)
+{
+   functions->CopyTexImage1D = intelCopyTexImage1D;
+   functions->CopyTexImage2D = intelCopyTexImage2D;
+   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
+   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 5e61e9e95e..c5f5220837 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -131,6 +131,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
       comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat);
    intelObj->mt = intel_miptree_create(intel,
                                        intelObj->base.Target,
+                                       intelImage->base._BaseFormat,
                                        intelImage->base.InternalFormat,
                                        firstLevel,
                                        lastLevel,
@@ -205,7 +206,7 @@ try_pbo_upload(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (unpack->BufferObj->Name == 0 ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
       DBG("%s: failure 1\n", __FUNCTION__);
@@ -235,12 +236,15 @@ try_pbo_upload(struct intel_context *intel,
 					       INTEL_WRITE_FULL);
 
 
-      intelEmitCopyBlit(intel,
-                        intelImage->mt->cpp,
-                        src_stride, src_buffer, src_offset, GL_FALSE,
-                        dst_stride, dst_buffer, dst_offset, GL_FALSE,
-                        0, 0, 0, 0, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_stride, src_buffer, src_offset, GL_FALSE,
+			     dst_stride, dst_buffer, dst_offset, GL_FALSE,
+			     0, 0, 0, 0, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
    UNLOCK_HARDWARE(intel);
 
@@ -248,7 +252,6 @@ try_pbo_upload(struct intel_context *intel,
 }
 
 
-
 static GLboolean
 try_pbo_zcopy(struct intel_context *intel,
               struct intel_texture_image *intelImage,
@@ -261,7 +264,7 @@ try_pbo_zcopy(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (unpack->BufferObj->Name == 0 ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
       DBG("%s: failure 1\n", __FUNCTION__);
@@ -293,10 +296,6 @@ try_pbo_zcopy(struct intel_context *intel,
 }
 
 
-
-
-
-
 static void
 intelTexImage(GLcontext * ctx,
               GLint dims,
@@ -307,7 +306,8 @@ intelTexImage(GLcontext * ctx,
               GLenum format, GLenum type, const void *pixels,
               const struct gl_pixelstore_attrib *unpack,
               struct gl_texture_object *texObj,
-              struct gl_texture_image *texImage, GLsizei imageSize, int compressed)
+              struct gl_texture_image *texImage, GLsizei imageSize,
+              GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_object *intelObj = intel_texture_object(texObj);
@@ -316,7 +316,6 @@ intelTexImage(GLcontext * ctx,
    GLint postConvHeight = height;
    GLint texelBytes, sizeInBytes;
    GLuint dstRowStride = 0, srcRowStride = texImage->RowStride;
-   GLboolean needs_map;
 
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
@@ -414,7 +413,9 @@ intelTexImage(GLcontext * ctx,
        * a miptree, so create one just for our level and store it in the image.
        * It'll get moved into the object miptree at validate time.
        */
-      intelImage->mt = intel_miptree_create(intel, target, internalFormat,
+      intelImage->mt = intel_miptree_create(intel, target,
+					    intelImage->base.TexFormat->BaseFormat,
+					    internalFormat,
 					    level, level,
 					    width, height, depth,
 					    intelImage->base.TexFormat->TexelBytes,
@@ -426,7 +427,7 @@ intelTexImage(GLcontext * ctx,
     */
    if (dims <= 2 &&
        intelImage->mt &&
-       intel_buffer_object(unpack->BufferObj) &&
+       unpack->BufferObj->Name != 0 &&
        check_pbo_format(internalFormat, format,
                         type, intelImage->base.TexFormat)) {
 
@@ -464,8 +465,6 @@ intelTexImage(GLcontext * ctx,
       DBG("pbo upload failed\n");
    }
 
-
-
    /* intelCopyTexImage calls this function with pixels == NULL, with
     * the expectation that the mipmap tree will be set up but nothing
     * more will be done.  This is where those calls return:
@@ -482,15 +481,8 @@ intelTexImage(GLcontext * ctx,
 
    LOCK_HARDWARE(intel);
 
-   /* Two cases where we need a mapping of the miptree: when the user supplied
-    * data is mapped as well (non-PBO, memcpy upload) or when we're going to do
-    * (software) mipmap generation.
-    */
-   needs_map = (pixels != NULL) || (level == texObj->BaseLevel &&
-				  texObj->GenerateMipmap);
-
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          texImage->Data = intel_miptree_image_map(intel,
                                                   intelImage->mt,
                                                   intelImage->face,
@@ -547,25 +539,26 @@ intelTexImage(GLcontext * ctx,
 						   format, type, pixels, unpack)) {
 	   _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
        }
-
-       /* GL_SGIS_generate_mipmap */
-       if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-	  intel_generate_mipmap(ctx, target, texObj);
-       }
    }
 
    _mesa_unmap_teximage_pbo(ctx, unpack);
 
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          intel_miptree_image_unmap(intel, intelImage->mt);
       texImage->Data = NULL;
    }
 
    UNLOCK_HARDWARE(intel);
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
 }
 
-void
+
+static void
 intelTexImage3D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -578,11 +571,11 @@ intelTexImage3D(GLcontext * ctx,
 {
    intelTexImage(ctx, 3, target, level,
                  internalFormat, width, height, depth, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
 
-void
+static void
 intelTexImage2D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -594,10 +587,11 @@ intelTexImage2D(GLcontext * ctx,
 {
    intelTexImage(ctx, 2, target, level,
                  internalFormat, width, height, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void
+
+static void
 intelTexImage1D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -609,21 +603,24 @@ intelTexImage1D(GLcontext * ctx,
 {
    intelTexImage(ctx, 1, target, level,
                  internalFormat, width, 1, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage )
+
+static void
+intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
+                           GLint internalFormat,
+                           GLint width, GLint height, GLint border,
+                           GLsizei imageSize, const GLvoid *data,
+                           struct gl_texture_object *texObj,
+                           struct gl_texture_image *texImage )
 {
    intelTexImage(ctx, 2, target, level,
 		 internalFormat, width, height, 1, border,
-		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE);
 }
 
+
 /**
  * Need to map texture image into memory before copying image data,
  * then unmap it.
@@ -632,7 +629,7 @@ static void
 intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
 		    GLenum format, GLenum type, GLvoid * pixels,
 		    struct gl_texture_object *texObj,
-		    struct gl_texture_image *texImage, int compressed)
+		    struct gl_texture_image *texImage, GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
@@ -686,28 +683,29 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
    }
 }
 
-void
+
+static void
 intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
                  GLenum format, GLenum type, GLvoid * pixels,
                  struct gl_texture_object *texObj,
                  struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, format, type, pixels,
-		       texObj, texImage, 0);
-
-
+		       texObj, texImage, GL_FALSE);
 }
 
-void
+
+static void
 intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
 			   GLvoid *pixels,
 			   struct gl_texture_object *texObj,
 			   struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, 0, 0, pixels,
-		       texObj, texImage, 1);
+		       texObj, texImage, GL_TRUE);
 }
 
+
 void
 intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		  unsigned long long offset, GLint depth, GLuint pitch)
@@ -816,3 +814,16 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
     */
    intelSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
 }
+
+
+void
+intelInitTextureImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexImage1D = intelTexImage1D;
+   functions->TexImage2D = intelTexImage2D;
+   functions->TexImage3D = intelTexImage3D;
+   functions->GetTexImage = intelGetTexImage;
+
+   functions->CompressedTexImage2D = intelCompressedTexImage2D;
+   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c
index e6f9a41779..7d69ea4484 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c
@@ -35,26 +35,39 @@
 #include "intel_context.h"
 #include "main/macros.h"
 
-GLuint intel_compressed_alignment(GLenum internalFormat)
+void intel_get_texture_alignment_unit(GLenum internalFormat, GLuint *w, GLuint *h)
 {
-    GLuint alignment = 4;
-
     switch (internalFormat) {
     case GL_COMPRESSED_RGB_FXT1_3DFX:
     case GL_COMPRESSED_RGBA_FXT1_3DFX:
-        alignment = 8;
+        *w = 8;
+        *h = 4;
+        break;
+
+    case GL_RGB_S3TC:
+    case GL_RGB4_S3TC:
+    case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+    case GL_RGBA_S3TC:
+    case GL_RGBA4_S3TC:
+    case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+        *w = 4;
+        *h = 4;
         break;
 
     default:
+        *w = 4;
+        *h = 2;
         break;
     }
-
-    return alignment;
 }
 
-void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt )
+void i945_miptree_layout_2d( struct intel_context *intel,
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling )
 {
-   GLint align_h = 2, align_w = 4;
+   GLuint align_h = 2, align_w = 4;
    GLuint level;
    GLuint x = 0;
    GLuint y = 0;
@@ -62,9 +75,9 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    GLuint height = mt->height0;
 
    mt->pitch = mt->width0;
+   intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
 
    if (mt->compressed) {
-       align_w = intel_compressed_alignment(mt->internal_format);
        mt->pitch = ALIGN(mt->width0, align_w);
    }
 
@@ -92,7 +105,7 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    /* Pitch must be a whole number of dwords, even though we
     * express it in texels.
     */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->pitch);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->pitch);
    mt->total_height = 0;
 
    for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h
index dbc90e6f9b..c9de9b5678 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.h
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h
@@ -38,5 +38,7 @@ static GLuint minify( GLuint d )
    return MAX2(1, d>>1);
 }
 
-extern void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt );
-extern GLuint intel_compressed_alignment(GLenum);
+extern void i945_miptree_layout_2d(struct intel_context *intel,
+				   struct intel_mipmap_tree *mt,
+				   uint32_t tiling);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index f86de56897..89037073f8 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -44,10 +44,12 @@ intelTexSubimage(GLcontext * ctx,
                  GLenum target, GLint level,
                  GLint xoffset, GLint yoffset, GLint zoffset,
                  GLint width, GLint height, GLint depth,
+                 GLsizei imageSize,
                  GLenum format, GLenum type, const void *pixels,
                  const struct gl_pixelstore_attrib *packing,
                  struct gl_texture_object *texObj,
-                 struct gl_texture_image *texImage)
+                 struct gl_texture_image *texImage,
+                 GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
@@ -59,9 +61,14 @@ intelTexSubimage(GLcontext * ctx,
 
    intelFlush(ctx);
 
-   pixels =
-      _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format,
-                                  type, pixels, packing, "glTexSubImage2D");
+   if (compressed)
+      pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize,
+                                                      pixels, packing,
+                                                      "glCompressedTexImage");
+   else
+      pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, depth,
+                                           format, type, pixels, packing,
+                                           "glTexSubImage");
    if (!pixels)
       return;
 
@@ -90,20 +97,28 @@ intelTexSubimage(GLcontext * ctx,
 
    assert(dstRowStride);
 
-   if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
-                                        texImage->TexFormat,
-                                        texImage->Data,
-                                        xoffset, yoffset, zoffset,
-                                        dstRowStride,
-                                        texImage->ImageOffsets,
-                                        width, height, depth,
-                                        format, type, pixels, packing)) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
+   if (compressed) {
+      if (intelImage->mt) {
+         struct intel_region *dst = intelImage->mt->region;
+         
+         _mesa_copy_rect(texImage->Data, dst->cpp, dst->pitch,
+                         xoffset, yoffset / 4,
+                         (width + 3)  & ~3, (height + 3) / 4,
+                         pixels, (width + 3) & ~3, 0, 0);
+      } else
+        memcpy(texImage->Data, pixels, imageSize);
    }
-
-   /* GL_SGIS_generate_mipmap */
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
+   else {
+      if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+                                           texImage->TexFormat,
+                                           texImage->Data,
+                                           xoffset, yoffset, zoffset,
+                                           dstRowStride,
+                                           texImage->ImageOffsets,
+                                           width, height, depth,
+                                           format, type, pixels, packing)) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
+      }
    }
 
    _mesa_unmap_teximage_pbo(ctx, packing);
@@ -114,13 +129,15 @@ intelTexSubimage(GLcontext * ctx,
    }
 
    UNLOCK_HARDWARE(intel);
-}
-
-
 
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
+}
 
 
-void
+static void
 intelTexSubImage3D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -132,18 +149,15 @@ intelTexSubImage3D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 3,
                     target, level,
                     xoffset, yoffset, zoffset,
-                    width, height, depth,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, height, depth, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
 
-
-void
+static void
 intelTexSubImage2D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -155,17 +169,15 @@ intelTexSubImage2D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 2,
                     target, level,
                     xoffset, yoffset, 0,
-                    width, height, 1,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, height, 1, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
 
-void
+static void
 intelTexSubImage1D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -180,12 +192,11 @@ intelTexSubImage1D(GLcontext * ctx,
    intelTexSubimage(ctx, 1,
                     target, level,
                     xoffset, 0, 0,
-                    width, 1, 1,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, 1, 1, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
-void
+static void
 intelCompressedTexSubImage2D(GLcontext * ctx,
 			     GLenum target,
 			     GLint level,
@@ -196,6 +207,20 @@ intelCompressedTexSubImage2D(GLcontext * ctx,
 			     struct gl_texture_object *texObj,
 			     struct gl_texture_image *texImage)
 {
-   fprintf(stderr, "stubbed CompressedTexSubImage2D: %dx%d@%dx%d\n",
-	   width, height, xoffset, yoffset);
+   intelTexSubimage(ctx, 2,
+                    target, level,
+                    xoffset, yoffset, 0,
+                    width, height, 1, imageSize,
+                    format, 0, pixels, &ctx->Unpack, texObj, texImage, GL_TRUE);
+}
+
+
+
+void
+intelInitTextureSubImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexSubImage1D = intelTexSubImage1D;
+   functions->TexSubImage2D = intelTexSubImage2D;
+   functions->TexSubImage3D = intelTexSubImage3D;
+   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c
index 05a375e1f3..a284d5475f 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c
@@ -199,6 +199,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
    if (!intelObj->mt) {
       intelObj->mt = intel_miptree_create(intel,
                                           intelObj->base.Target,
+                                          firstImage->base._BaseFormat,
                                           firstImage->base.InternalFormat,
                                           intelObj->firstLevel,
                                           intelObj->lastLevel,
@@ -241,7 +242,7 @@ intel_tex_map_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intelImage->base.Data =
 	    intel_miptree_image_map(intel,
 				    intelImage->mt,
@@ -268,7 +269,7 @@ intel_tex_unmap_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intel_miptree_image_unmap(intel, intelImage->mt);
 	 intelImage->base.Data = NULL;
       }
diff --git a/src/mesa/drivers/dri/r200/.gitignore b/src/mesa/drivers/dri/r200/.gitignore
deleted file mode 100644
index 3773d8ea73..0000000000
--- a/src/mesa/drivers/dri/r200/.gitignore
+++ /dev/null
@@ -1,3 +0,0 @@
-radeon_chipset.h
-radeon_screen.*
-server
diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile
index e9144ac75c..42635bf9d9 100644
--- a/src/mesa/drivers/dri/r200/Makefile
+++ b/src/mesa/drivers/dri/r200/Makefile
@@ -3,6 +3,8 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = r200_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c 
@@ -11,44 +13,52 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+RADEON_COMMON_SOURCES = \
+	radeon_bo_legacy.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_cs_legacy.c \
+	radeon_dma.c \
+	radeon_debug.c \
+	radeon_fbo.c \
+	radeon_lock.c \
+	radeon_mipmap_tree.c \
+	radeon_queryobj.c \
+	radeon_span.c \
+	radeon_texture.c
+
+
 DRIVER_SOURCES = r200_context.c \
 		 r200_ioctl.c \
-		 r200_lock.c \
 		 r200_state.c \
 		 r200_state_init.c \
 		 r200_cmdbuf.c \
 		 r200_pixel.c \
 		 r200_tex.c \
-		 r200_texmem.c \
 		 r200_texstate.c \
 		 r200_tcl.c \
 		 r200_swtcl.c \
-		 r200_span.c \
 		 r200_maos.c \
 		 r200_sanity.c \
 		 r200_fragshader.c \
 		 r200_vertprog.c \
 		 radeon_screen.c \
-		 $(EGL_SOURCES)
+		 $(EGL_SOURCES) \
+		 $(RADEON_COMMON_SOURCES) \
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
 X86_SOURCES = 
 
-DRIVER_DEFINES = -DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R200
-
-SYMLINKS = \
-	server/radeon_egl.c  \
-	server/radeon_dri.c  \
-	server/radeon_dri.h \
-	server/radeon.h \
-	server/radeon_macros.h \
-	server/radeon_reg.h
+DRIVER_DEFINES = -DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R200 \
+				 -Wall
 
-COMMON_SYMLINKS = \
-	radeon_chipset.h \
-	radeon_screen.c \
-	radeon_screen.h
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
 ##### TARGETS #####
 
@@ -57,15 +67,4 @@ include ../Makefile.template
 
 #INCLUDES += -I../radeon/server
 
-server:
-	mkdir -p server
-
-$(SYMLINKS): server
-	@[ -e $@ ] || ln -sf ../../radeon/$@ server/
-
-
-$(COMMON_SYMLINKS):
-	@[ -e $@ ] || ln -sf ../radeon/$@ ./
-
-symlinks: $(SYMLINKS) $(COMMON_SYMLINKS)
-
+symlinks:
diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
index e1633772a1..1d1bea6f5f 100644
--- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c
+++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
@@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "main/simple_list.h"
 
+#include "radeon_common.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -45,160 +46,71 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_sanity.h"
 #include "radeon_reg.h"
 
-static void print_state_atom( struct r200_state_atom *state )
-{
-   int i;
-
-   fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size);
-
-   if (0 & R200_DEBUG & DEBUG_VERBOSE) 
-      for (i = 0 ; i < state->cmd_size ; i++) 
-	 fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]);
-
-}
-
 /* The state atoms will be emitted in the order they appear in the atom list,
  * so this step is important.
  */
+#define insert_at_tail_if(atom_list, atom) \
+   do { \
+      struct radeon_state_atom* __atom = (atom); \
+      if (__atom->check) \
+	 insert_at_tail((atom_list), __atom); \
+   } while(0)
+
 void r200SetUpAtomList( r200ContextPtr rmesa )
 {
    int i, mtu;
 
-   mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   make_empty_list(&rmesa->hw.atomlist);
-   rmesa->hw.atomlist.name = "atom-list";
-
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ctx );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.set );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lin );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msk );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpt );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vtx );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vap );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vte );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msc );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cst );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.zbs );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcl );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msl );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcg );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.grd );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.fog );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tam );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tf );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.atf );
+   mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits;
+
+   make_empty_list(&rmesa->radeon.hw.atomlist);
+   rmesa->radeon.hw.atomlist.name = "atom-list";
+
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.ctx );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.set );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.lin );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.msk );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpt );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vtx );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vap );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vte );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.msc );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.cst );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.zbs );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcl );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.msl );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcg );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.grd );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.fog );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.tam );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.tf );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.atf );
    for (i = 0; i < mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tex[i] );
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i] );
    for (i = 0; i < mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cube[i] );
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i] );
    for (i = 0; i < 6; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pix[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[1] );
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.pix[i] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[0] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[1] );
    for (i = 0; i < 8; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lit[i] );
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i] );
    for (i = 0; i < 3 + mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mat[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.eye );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.glt );
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.eye );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.glt );
    for (i = 0; i < 2; ++i)
-      insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mtl[i] );
+      insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.mtl[i] );
    for (i = 0; i < 6; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ucp[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.spr );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ptp );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.prf );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pvs );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[1] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[1] );
-}
-
-static void r200SaveHwState( r200ContextPtr rmesa )
-{
-   struct r200_state_atom *atom;
-   char * dest = rmesa->backup_store.cmd_buf;
-
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->backup_store.cmd_used = 0;
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( atom->check( rmesa->glCtx, atom->idx ) ) {
-	 int size = atom->cmd_size * 4;
-	 memcpy( dest, atom->cmd, size);
-	 dest += size;
-	 rmesa->backup_store.cmd_used += size;
-	 if (R200_DEBUG & DEBUG_STATE)
-	    print_state_atom( atom );
-      }
-   }
-
-   assert( rmesa->backup_store.cmd_used <= R200_CMD_BUF_SZ );
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Returning to r200EmitState\n");
-}
-
-void r200EmitState( r200ContextPtr rmesa )
-{
-   char *dest;
-   int mtu;
-   struct r200_state_atom *atom;
-
-   if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->save_on_next_emit) {
-      r200SaveHwState(rmesa);
-      rmesa->save_on_next_emit = GL_FALSE;
-   }
-
-   if (!rmesa->hw.is_dirty && !rmesa->hw.all_dirty)
-      return;
-
-   mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   /* To avoid going across the entire set of states multiple times, just check
-    * for enough space for the case of emitting all state, and inline the
-    * r200AllocCmdBuf code here without all the checks.
-    */
-   r200EnsureCmdBufSpace( rmesa, rmesa->hw.max_state_size );
-
-   /* we need to calculate dest after EnsureCmdBufSpace
-      as we may flush the buffer - airlied */
-   dest = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-   if (R200_DEBUG & DEBUG_STATE) {
-      foreach( atom, &rmesa->hw.atomlist ) {
-	 if ( atom->dirty || rmesa->hw.all_dirty ) {
-	    if ( atom->check( rmesa->glCtx, atom->idx ) )
-	       print_state_atom( atom );
-	    else
-	       fprintf(stderr, "skip state %s\n", atom->name);
-	 }
-      }
-   }
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( rmesa->hw.all_dirty )
-	 atom->dirty = GL_TRUE;
-      if ( atom->dirty ) {
-	 if ( atom->check( rmesa->glCtx, atom->idx ) ) {
-	    int size = atom->cmd_size * 4;
-	    memcpy( dest, atom->cmd, size);
-	    dest += size;
-	    rmesa->store.cmd_used += size;
-	    atom->dirty = GL_FALSE;
-	 }
-      }
-   }
-
-   assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ );
-
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
+       insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.spr );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.ptp );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.prf );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.pvs );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[0] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[1] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[0] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[1] );
+   insert_at_tail_if( &rmesa->radeon.hw.atomlist, &rmesa->hw.sci );
 }
 
 /* Fire a section of the retained (indexed_verts) buffer as a regular
@@ -208,51 +120,81 @@ void r200EmitVbufPrim( r200ContextPtr rmesa,
                        GLuint primitive,
                        GLuint vertex_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
    assert(!(primitive & R200_VF_PRIM_WALK_IND));
    
-   r200EmitState( rmesa );
-   
-   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
-      fprintf(stderr, "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__,
-	      rmesa->store.cmd_used/4, primitive, vertex_nr);
+   radeonEmitState(&rmesa->radeon);
    
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VBUF_BUFSZ,
-						  __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_VBUF_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_LIST |
-	       R200_VF_COLOR_ORDER_RGBA |
-	       (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
+   radeon_print(RADEON_RENDER|RADEON_SWRENDER,RADEON_VERBOSE,
+           "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__,
+           rmesa->store.cmd_used/4, primitive, vertex_nr);
+ 
+   BEGIN_BATCH(3);
+   OUT_BATCH_PACKET3_CLIP(R200_CP_CMD_3D_DRAW_VBUF_2, 0);
+   OUT_BATCH(primitive | R200_VF_PRIM_WALK_LIST | R200_VF_COLOR_ORDER_RGBA |
+	     (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
+   END_BATCH();
 }
 
+static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type)
+{
+	BATCH_LOCALS(&rmesa->radeon);
+
+	if (vertex_count > 0) {
+		BEGIN_BATCH(8+2);
+		OUT_BATCH_PACKET3_CLIP(R200_CP_CMD_3D_DRAW_INDX_2, 0);
+		OUT_BATCH(R200_VF_PRIM_WALK_IND |
+			  R200_VF_COLOR_ORDER_RGBA | 
+			  ((vertex_count + 0) << 16) |
+			  type);
+		
+		if (!rmesa->radeon.radeonScreen->kernel_mm) {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset,
+					rmesa->radeon.tcl.elt_dma_bo,
+					rmesa->radeon.tcl.elt_dma_offset,
+					RADEON_GEM_DOMAIN_GTT, 0, 0);
+			OUT_BATCH((vertex_count + 1)/2);
+		} else {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
+			OUT_BATCH((vertex_count + 1)/2);
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.elt_dma_bo,
+					      RADEON_GEM_DOMAIN_GTT, 0, 0);
+		}
+		END_BATCH();
+	}
+}
 
-void r200FlushElts( r200ContextPtr rmesa )
+void r200FlushElts(GLcontext *ctx)
 {
-   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
-   int dwords;
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 12)) / 2;
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   int nr, elt_used = rmesa->tcl.elt_used;
+
+   radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x %d\n", __FUNCTION__, rmesa->tcl.hw_primitive, elt_used);
+
+   assert( rmesa->radeon.dma.flush == r200FlushElts );
+   rmesa->radeon.dma.flush = NULL;
 
-   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   nr = elt_used / 2;
 
-   assert( rmesa->dma.flush == r200FlushElts );
-   rmesa->dma.flush = NULL;
+   radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 
-   /* Cope with odd number of elts:
-    */
-   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
-   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+   r200FireEB(rmesa, nr, rmesa->tcl.hw_primitive);
 
-   cmd[1] |= (dwords - 3) << 16;
-   cmd[2] |= nr << R200_VF_VERTEX_NUMBER_SHIFT;
+   radeon_bo_unref(rmesa->radeon.tcl.elt_dma_bo);
+   rmesa->radeon.tcl.elt_dma_bo = NULL;
 
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-      r200Finish( rmesa->glCtx );
+   if (R200_ELT_BUF_SZ > elt_used)
+     radeonReturnDmaRegion(&rmesa->radeon, R200_ELT_BUF_SZ - elt_used);
+
+   if (radeon_is_debug_enabled(RADEON_SYNC, RADEON_CRITICAL)) {
+      radeon_print(RADEON_SYNC, RADEON_NORMAL, "%s: Syncing\n", __FUNCTION__);
+      radeonFinish( rmesa->radeon.glCtx );
    }
 }
 
@@ -261,168 +203,153 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 				    GLuint primitive,
 				    GLuint min_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
    GLushort *retval;
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
+   radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
 
    assert((primitive & R200_VF_PRIM_WALK_IND));
    
-   r200EmitState( rmesa );
-   
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, ELTS_BUFSZ(min_nr),
-						__FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_INDX_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_IND |
-	       R200_VF_COLOR_ORDER_RGBA);
-
-   
-   retval = (GLushort *)(cmd+3);
+   radeonEmitState(&rmesa->radeon);
 
-   if (R200_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x prim %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, primitive);
+   radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo,
+			&rmesa->radeon.tcl.elt_dma_offset, R200_ELT_BUF_SZ, 4);
+   rmesa->tcl.elt_used = min_nr * 2;
 
-   assert(!rmesa->dma.flush);
-   rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   rmesa->dma.flush = r200FlushElts;
-
-   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+   radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
+   retval = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
+   
+   assert(!rmesa->radeon.dma.flush);
+   rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+   rmesa->radeon.dma.flush = r200FlushElts;
 
    return retval;
 }
 
+void r200EmitMaxVtxIndex(r200ContextPtr rmesa, int count)
+{
+   BATCH_LOCALS(&rmesa->radeon);
 
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+	   BEGIN_BATCH_NO_AUTOSTATE(2);
+	   OUT_BATCH(CP_PACKET0(R200_SE_VF_MAX_VTX_INDX, 0));
+	   OUT_BATCH(count);
+	   END_BATCH();
+   }
+}
 
 void r200EmitVertexAOS( r200ContextPtr rmesa,
-			  GLuint vertex_size,
-			  GLuint offset )
+			GLuint vertex_size,
+ 			struct radeon_bo *bo,
+			GLuint offset )
 {
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
-   if (R200_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL))
-      fprintf(stderr, "%s:  vertex_size 0x%x offset 0x%x \n",
+   radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s:  vertex_size 0x%x offset 0x%x \n",
 	      __FUNCTION__, vertex_size, offset);
 
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VERT_AOS_BUFSZ,
-						  __FUNCTION__ );
 
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (2 << 16);
-   cmd[2].i = 1;
-   cmd[3].i = vertex_size | (vertex_size << 8);
-   cmd[4].i = offset;
+   BEGIN_BATCH(7);
+   OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, 2);
+   OUT_BATCH(1);
+   OUT_BATCH(vertex_size | (vertex_size << 8));
+   OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   END_BATCH();
 }
-		       
 
-void r200EmitAOS( r200ContextPtr rmesa,
-		    struct r200_dma_region **component,
-		    GLuint nr,
-		    GLuint offset )
+void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset)
 {
-   drm_radeon_cmd_header_t *cmd;
-   int sz = AOS_BUFSZ(nr);
+   BATCH_LOCALS(&rmesa->radeon);
+   uint32_t voffset;
+   int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
    int i;
-   int *tmp;
-
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s nr arrays: %d\n", __FUNCTION__, nr);
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sz, __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (((sz / sizeof(int)) - 3) << 16);
-   cmd[2].i = nr;
-   tmp = &cmd[0].i;
-   cmd += 3;
-
-   for (i = 0 ; i < nr ; i++) {
-      if (i & 1) {
-	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
-		      (component[i]->aos_size << 16));
-	 cmd[2].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-	 cmd += 3;
+   
+   radeon_print(RADEON_RENDER, RADEON_VERBOSE,
+           "%s: nr=%d, ofs=0x%08x\n",
+           __FUNCTION__, nr, offset);
+
+   BEGIN_BATCH(sz+2+ (nr*2));
+   OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, sz - 1);
+   OUT_BATCH(nr);
+
+    
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+			
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i+1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-      else {
-	 cmd[0].i = ((component[i]->aos_stride << 8) | 
-		     (component[i]->aos_size << 0));
-	 cmd[1].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
+      
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[nr - 1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+      }
+   } else {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+	 
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH(voffset);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH(voffset);
+      }
+      
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH(voffset);
+      }
+      for (i = 0; i + 1 < nr; i += 2) {
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+0].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+      }
+      if (nr & 1) {
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[nr-1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
    }
-
-   if (R200_DEBUG & DEBUG_VERTS) {
-      fprintf(stderr, "%s:\n", __FUNCTION__);
-      for (i = 0 ; i < sz ; i++)
-	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
-   }
-}
-
-void r200EmitBlit( r200ContextPtr rmesa,
-		   GLuint color_fmt,
-		   GLuint src_pitch,
-		   GLuint src_offset,
-		   GLuint dst_pitch,
-		   GLuint dst_offset,
-		   GLint srcx, GLint srcy,
-		   GLint dstx, GLint dsty,
-		   GLuint w, GLuint h )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-	      __FUNCTION__, 
-	      src_pitch, src_offset, srcx, srcy,
-	      dst_pitch, dst_offset, dstx, dsty,
-	      w, h);
-
-   assert( (src_pitch & 63) == 0 );
-   assert( (dst_pitch & 63) == 0 );
-   assert( (src_offset & 1023) == 0 );
-   assert( (dst_offset & 1023) == 0 );
-   assert( w < (1<<16) );
-   assert( h < (1<<16) );
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 8 * sizeof(int),
-						  __FUNCTION__ );
-
-
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_BITBLT_MULTI | (5 << 16);
-   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_BRUSH_NONE |
-	       (color_fmt << 8) |
-	       RADEON_GMC_SRC_DATATYPE_COLOR |
-	       RADEON_ROP3_S |
-	       RADEON_DP_SRC_SOURCE_MEMORY |
-	       RADEON_GMC_CLR_CMP_CNTL_DIS |
-	       RADEON_GMC_WR_MSK_DIS );
-
-   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
-   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
-   cmd[5].i = (srcx << 16) | srcy;
-   cmd[6].i = (dstx << 16) | dsty; /* dst */
-   cmd[7].i = (w << 16) | h;
-}
-
-
-void r200EmitWait( r200ContextPtr rmesa, GLuint flags )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 1 * sizeof(int),
-					   __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
-   cmd[0].wait.flags = flags;
+   END_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 058296eab2..3ddb5bf7d6 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -54,14 +54,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
-#include "r200_span.h"
 #include "r200_pixel.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
 #include "r200_tcl.h"
 #include "r200_maos.h"
 #include "r200_vertprog.h"
+#include "radeon_queryobj.h"
 
+#include "radeon_span.h"
+
+#define need_GL_ARB_occlusion_query
 #define need_GL_ARB_vertex_program
 #define need_GL_ATI_fragment_shader
 #define need_GL_EXT_blend_minmax
@@ -71,6 +74,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define need_GL_EXT_blend_func_separate
 #define need_GL_NV_vertex_program
 #define need_GL_ARB_point_parameters
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20060602"
@@ -78,9 +82,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h" /* for symbolic values of enum-type options */
-#ifndef R200_DEBUG
-int R200_DEBUG = (0);
-#endif
 
 /* Return various strings for glGetString().
  */
@@ -89,8 +90,8 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    static char buffer[128];
    unsigned   offset;
-   GLuint agp_mode = (rmesa->r200Screen->card_type == RADEON_CARD_PCI)? 0 :
-      rmesa->r200Screen->AGPMode;
+   GLuint agp_mode = (rmesa->radeon.radeonScreen->card_type == RADEON_CARD_PCI)? 0 :
+      rmesa->radeon.radeonScreen->AGPMode;
 
    switch ( name ) {
    case GL_VENDOR:
@@ -101,7 +102,7 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
 				     agp_mode );
 
       sprintf( & buffer[ offset ], " %sTCL",
-	       !(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)
+	       !(rmesa->radeon.TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)
 	       ? "" : "NO-" );
 
       return (GLubyte *)buffer;
@@ -117,6 +118,7 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
 const struct dri_extension card_extensions[] =
 {
     { "GL_ARB_multitexture",               NULL },
+    { "GL_ARB_occlusion_query",		   GL_ARB_occlusion_query_functions},
     { "GL_ARB_texture_border_clamp",       NULL },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
@@ -126,6 +128,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
     { "GL_EXT_blend_subtract",             NULL },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -167,6 +170,11 @@ const struct dri_extension point_extensions[] = {
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _r200_render_stage;
 extern const struct tnl_pipeline_stage _r200_tcl_stage;
 
@@ -213,26 +221,54 @@ static void r200InitDriverFuncs( struct dd_function_table *functions )
     functions->GetString		= r200GetString;
 }
 
-static const struct dri_debug_control debug_control[] =
+
+static void r200_get_lock(radeonContextPtr radeon)
 {
-    { "fall",  DEBUG_FALLBACKS },
-    { "tex",   DEBUG_TEXTURE },
-    { "ioctl", DEBUG_IOCTL },
-    { "prim",  DEBUG_PRIMS },
-    { "vert",  DEBUG_VERTS },
-    { "state", DEBUG_STATE },
-    { "code",  DEBUG_CODEGEN },
-    { "vfmt",  DEBUG_VFMT },
-    { "vtxf",  DEBUG_VFMT },
-    { "verb",  DEBUG_VERBOSE },
-    { "dri",   DEBUG_DRI },
-    { "dma",   DEBUG_DMA },
-    { "san",   DEBUG_SANITY },
-    { "sync",  DEBUG_SYNC },
-    { "pix",   DEBUG_PIXEL },
-    { "mem",   DEBUG_MEMORY },
-    { NULL,    0 }
-};
+   r200ContextPtr rmesa = (r200ContextPtr)radeon;
+   drm_radeon_sarea_t *sarea = radeon->sarea;
+
+   R200_STATECHANGE( rmesa, ctx );
+   if (rmesa->radeon.sarea->tiling_enabled) {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
+   }
+   else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE;
+
+   if ( sarea->ctx_owner != rmesa->radeon.dri.hwContext ) {
+      sarea->ctx_owner = rmesa->radeon.dri.hwContext;
+      if (!radeon->radeonScreen->kernel_mm)
+         radeon_bo_legacy_texture_age(radeon->radeonScreen->bom);
+   }
+
+}
+
+static void r200_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+}
+
+static void r200_emit_query_finish(radeonContextPtr radeon)
+{
+   BATCH_LOCALS(radeon);
+   struct radeon_query_object *query = radeon->query.current;
+
+   BEGIN_BATCH_NO_AUTOSTATE(4);
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZPASS_ADDR, 0));
+   OUT_BATCH_RELOC(0, query->bo, query->curr_offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+   END_BATCH();
+   query->curr_offset += sizeof(uint32_t);
+   assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
+   query->emitted_begin = GL_FALSE;
+}
+
+static void r200_init_vtbl(radeonContextPtr radeon)
+{
+   radeon->vtbl.get_lock = r200_get_lock;
+   radeon->vtbl.update_viewport_offset = r200UpdateViewportOffset;
+   radeon->vtbl.emit_cs_header = r200_vtbl_emit_cs_header;
+   radeon->vtbl.swtcl_flush = r200_swtcl_flush;
+   radeon->vtbl.fallback = r200Fallback;
+   radeon->vtbl.update_scissor = r200_vtbl_update_scissor;
+   radeon->vtbl.emit_query_finish = r200_emit_query_finish;
+}
 
 
 /* Create the device specific rendering context.
@@ -245,9 +281,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private);
    struct dd_function_table functions;
    r200ContextPtr rmesa;
-   GLcontext *ctx, *shareCtx;
+   GLcontext *ctx;
    int i;
-   int tcl_mode, fthrottle_mode;
+   int tcl_mode;
 
    assert(glVisual);
    assert(driContextPriv);
@@ -257,7 +293,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    rmesa = (r200ContextPtr) CALLOC( sizeof(*rmesa) );
    if ( !rmesa )
       return GL_FALSE;
-      
+
+   r200_init_vtbl(&rmesa->radeon);
    /* init exp fog table data */
    r200InitStaticFogData();
 
@@ -265,12 +302,13 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     * Do this here so that initialMaxAnisotropy is set before we create
     * the default textures.
     */
-   driParseConfigFiles (&rmesa->optionCache, &screen->optionCache,
+   driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
 			screen->driScreen->myNum, "r200");
-   rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache,
-                                                 "def_max_anisotropy");
+   rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
+							"def_max_anisotropy");
 
-   if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) {
+   if ( sPriv->drm_version.major == 1
+       && driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) {
       if ( sPriv->drm_version.minor < 13 )
 	 fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
 			  "disabling.\n", sPriv->drm_version.minor );
@@ -287,63 +325,20 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    _mesa_init_driver_functions(&functions);
    r200InitDriverFuncs(&functions);
    r200InitIoctlFuncs(&functions);
-   r200InitStateFuncs(&functions);
+   r200InitStateFuncs(&functions, screen->kernel_mm);
    r200InitTextureFuncs(&functions);
    r200InitShaderFuncs(&functions); 
+   radeonInitQueryObjFunctions(&functions);
 
-   /* Allocate and initialize the Mesa context */
-   if (sharedContextPrivate)
-      shareCtx = ((r200ContextPtr) sharedContextPrivate)->glCtx;
-   else
-      shareCtx = NULL;
-   rmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
-                                       &functions, (void *) rmesa);
-   if (!rmesa->glCtx) {
-      FREE(rmesa);
-      return GL_FALSE;
+   if (!radeonInitContext(&rmesa->radeon, &functions,
+			  glVisual, driContextPriv,
+			  sharedContextPrivate)) {
+     FREE(rmesa);
+     return GL_FALSE;
    }
-   driContextPriv->driverPrivate = rmesa;
-
-   /* Init r200 context data */
-   rmesa->dri.context = driContextPriv;
-   rmesa->dri.screen = sPriv;
-   rmesa->dri.drawable = NULL; /* Set by XMesaMakeCurrent */
-   rmesa->dri.hwContext = driContextPriv->hHWContext;
-   rmesa->dri.hwLock = &sPriv->pSAREA->lock;
-   rmesa->dri.fd = sPriv->fd;
-   rmesa->dri.drmMinor = sPriv->drm_version.minor;
-
-   rmesa->r200Screen = screen;
-   rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA +
-				       screen->sarea_priv_offset);
-
-
-   rmesa->dma.buf0_address = rmesa->r200Screen->buffers->list[0].address;
-
-   (void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) );
-   make_empty_list( & rmesa->swapped );
-
-   rmesa->nr_heaps = 1 /* screen->numTexHeaps */ ;
-   assert(rmesa->nr_heaps < RADEON_NR_TEX_HEAPS);
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa,
-	    screen->texSize[i],
-	    12,
-	    RADEON_NR_TEX_REGIONS,
-	    (drmTextureRegionPtr)rmesa->sarea->tex_list[i],
-	    & rmesa->sarea->tex_age[i],
-	    & rmesa->swapped,
-	    sizeof( r200TexObj ),
-	    (destroy_texture_object_t *) r200DestroyTexObj );
-   }
-   rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache,
-					   "texture_depth");
-   if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-      rmesa->texture_depth = ( screen->cpp == 4 ) ?
-	 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
 
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->hw.all_dirty = 1;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.hw.all_dirty = 1;
 
    /* Set the maximum texture size small enough that we can guarentee that
     * all texture units can bind a maximal texture and have all of them in
@@ -351,29 +346,20 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     * setting allow larger textures.
     */
 
-   ctx = rmesa->glCtx;
-   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache,
+   ctx = rmesa->radeon.glCtx;
+   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
 						 "texture_units");
    ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
    ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
 
-   i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures");
-
-   driCalculateMaxTextureLevels( rmesa->texture_heaps,
-				 rmesa->nr_heaps,
-				 & ctx->Const,
-				 4,
-				 11, /* max 2D texture size is 2048x2048 */
-#if ENABLE_HW_3D_TEXTURE
-				 8,  /* max 3D texture size is 256^3 */
-#else
-				 0,  /* 3D textures unsupported */
-#endif
-				 11, /* max cube texture size is 2048x2048 */
-				 11, /* max texture rectangle size is 2048x2048 */
-				 12,
-				 GL_FALSE,
-				 i );
+   i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures");
+
+   /* FIXME: When no memory manager is available we should set this 
+    * to some reasonable value based on texture memory pool size */
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
+   ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 
@@ -383,7 +369,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MinPointSizeAA = 1.0;
    ctx->Const.MaxPointSizeAA = 1.0;
    ctx->Const.PointSizeGranularity = 0.0625;
-   if (rmesa->r200Screen->drmSupportsPointSprites)
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites)
       ctx->Const.MaxPointSize = 2047.0;
    else
       ctx->Const.MaxPointSize = 1.0;
@@ -441,33 +427,39 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    _math_matrix_set_identity( &rmesa->tmpmat );
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
-   if (!(rmesa->r200Screen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) {
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
+   if (!(rmesa->radeon.radeonScreen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) {
      /* yuv textures don't work with some chips - R200 / rv280 okay so far
 	others get the bit ordering right but don't actually do YUV-RGB conversion */
       _mesa_enable_extension( ctx, "GL_MESA_ycbcr_texture" );
    }
-   if (rmesa->glCtx->Mesa_DXTn) {
+   if (rmesa->radeon.glCtx->Mesa_DXTn) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
       _mesa_enable_extension( ctx, "GL_S3_s3tc" );
    }
-   else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) {
+   else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
    }
 
-   if (rmesa->r200Screen->drmSupportsCubeMapsR200)
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200)
       _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
        driInitExtensions( ctx, blend_extensions, GL_FALSE );
    }
-   if(rmesa->r200Screen->drmSupportsVertexProgram)
+   if(rmesa->radeon.radeonScreen->drmSupportsVertexProgram)
       driInitSingleExtension( ctx, ARB_vp_extension );
-   if(driQueryOptionb(&rmesa->optionCache, "nv_vertex_program"))
+   if(driQueryOptionb(&rmesa->radeon.optionCache, "nv_vertex_program"))
       driInitSingleExtension( ctx, NV_vp_extension );
 
-   if ((ctx->Const.MaxTextureUnits == 6) && rmesa->r200Screen->drmSupportsFragShader)
+   if ((ctx->Const.MaxTextureUnits == 6) && rmesa->radeon.radeonScreen->drmSupportsFragShader)
       driInitSingleExtension( ctx, ATI_fs_extension );
-   if (rmesa->r200Screen->drmSupportsPointSprites)
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites)
       driInitExtensions( ctx, point_extensions, GL_FALSE );
+
+   if (!rmesa->radeon.radeonScreen->kernel_mm)
+      _mesa_disable_extension(ctx, "GL_ARB_occlusion_query");
 #if 0
    r200InitDriverFuncs( ctx );
    r200InitIoctlFuncs( ctx );
@@ -476,236 +468,43 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 #endif
    /* plug in a few more device driver functions */
    /* XXX these should really go right after _mesa_init_driver_functions() */
+   radeon_fbo_init(&rmesa->radeon);
+   radeonInitSpanFuncs( ctx );
    r200InitPixelFuncs( ctx );
-   r200InitSpanFuncs( ctx );
    r200InitTnlFuncs( ctx );
    r200InitState( rmesa );
    r200InitSwtcl( ctx );
 
-   fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode");
-   rmesa->iw.irq_seq = -1;
-   rmesa->irqsEmitted = 0;
-   rmesa->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
-		     rmesa->r200Screen->irq);
-
-   rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
-
-   if (!rmesa->do_irqs)
-      fprintf(stderr,
-	      "IRQ's not enabled, falling back to %s: %d %d\n",
-	      rmesa->do_usleeps ? "usleeps" : "busy waits",
-	      fthrottle_mode,
-	      rmesa->r200Screen->irq);
-
    rmesa->prefer_gart_client_texturing = 
       (getenv("R200_GART_CLIENT_TEXTURES") != 0);
 
-   (*sPriv->systemTime->getUST)( & rmesa->swap_ust );
-
-
-#if DO_DEBUG
-   R200_DEBUG  = driParseDebugString( getenv( "R200_DEBUG" ),
-				      debug_control );
-   R200_DEBUG |= driParseDebugString( getenv( "RADEON_DEBUG" ),
-				      debug_control );
-#endif
-
-   tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-   if (driQueryOptionb(&rmesa->optionCache, "no_rast")) {
+   tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
+   if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) {
       fprintf(stderr, "disabling 3D acceleration\n");
       FALLBACK(rmesa, R200_FALLBACK_DISABLE, 1);
    }
    else if (tcl_mode == DRI_CONF_TCL_SW || getenv("R200_NO_TCL") ||
-	    !(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) {
-      if (rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL) {
-	 rmesa->r200Screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	    !(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+	 rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
 	 fprintf(stderr, "Disabling HW TCL support\n");
       }
-      TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
+      TCL_FALLBACK(rmesa->radeon.glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
    return GL_TRUE;
 }
 
 
-/* Destroy the device specific context.
- */
-/* Destroy the Mesa and driver specific context data.
- */
 void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
 {
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
-   r200ContextPtr current = ctx ? R200_CONTEXT(ctx) : NULL;
-
-   /* check if we're deleting the currently bound context */
-   if (rmesa == current) {
-      R200_FIREVERTICES( rmesa );
-      _mesa_make_current(NULL, NULL, NULL);
-   }
-
-   /* Free r200 context resources */
-   assert(rmesa); /* should never be null */
-   if ( rmesa ) {
-      GLboolean   release_texture_heaps;
-
-
-      release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
-      _swsetup_DestroyContext( rmesa->glCtx );
-      _tnl_DestroyContext( rmesa->glCtx );
-      _vbo_DestroyContext( rmesa->glCtx );
-      _swrast_DestroyContext( rmesa->glCtx );
-
-      r200DestroySwtcl( rmesa->glCtx );
-      r200ReleaseArrays( rmesa->glCtx, ~0 );
-
-      if (rmesa->dma.current.buf) {
-	 r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-	 r200FlushCmdBuf( rmesa, __FUNCTION__ );
-      }
-
-      if (rmesa->state.scissor.pClipRects) {
-	 FREE(rmesa->state.scissor.pClipRects);
-	 rmesa->state.scissor.pClipRects = NULL;
-      }
-
-      if ( release_texture_heaps ) {
-         /* This share group is about to go away, free our private
-          * texture object data.
-          */
-         int i;
-
-         for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-	    driDestroyTextureHeap( rmesa->texture_heaps[ i ] );
-	    rmesa->texture_heaps[ i ] = NULL;
-         }
-
-	 assert( is_empty_list( & rmesa->swapped ) );
-      }
-
-      /* free the Mesa context */
-      rmesa->glCtx->DriverCtx = NULL;
-      _mesa_destroy_context( rmesa->glCtx );
-
-      /* free the option cache */
-      driDestroyOptionCache (&rmesa->optionCache);
-
-      FREE( rmesa );
-   }
-}
-
-
-
-
-void
-r200SwapBuffers( __DRIdrawablePrivate *dPriv )
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      r200ContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-         if ( rmesa->doPageFlip ) {
-            r200PageFlip( dPriv );
-         }
-         else {
-	     r200CopyBuffer( dPriv, NULL );
-         }
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-void
-r200CopySubBuffer( __DRIdrawablePrivate *dPriv,
-		   int x, int y, int w, int h )
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      r200ContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-	 drm_clip_rect_t rect;
-	 rect.x1 = x + dPriv->x;
-	 rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	 rect.x2 = rect.x1 + w;
-	 rect.y2 = rect.y1 + h;
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-	 r200CopyBuffer( dPriv, &rect );
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-/* Force the context `c' to be the current context and associate with it
- * buffer `b'.
- */
-GLboolean
-r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
-                   __DRIdrawablePrivate *driDrawPriv,
-                   __DRIdrawablePrivate *driReadPriv )
-{
-   if ( driContextPriv ) {
-      r200ContextPtr newCtx = 
-	 (r200ContextPtr) driContextPriv->driverPrivate;
-
-      if (R200_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)newCtx->glCtx);
-
-      newCtx->dri.readable = driReadPriv;
-
-      if ( newCtx->dri.drawable != driDrawPriv ||
-           newCtx->lastStamp != driDrawPriv->lastStamp ) {
-	 if (driDrawPriv->swap_interval == (unsigned)-1) {
-	    driDrawPriv->vblFlags = (newCtx->r200Screen->irq != 0)
-	       ? driGetDefaultVBlankFlags(&newCtx->optionCache)
-	       : VBLANK_FLAG_NO_IRQ;
-
-	    driDrawableInitVBlank( driDrawPriv );
-	 }
-
-	 newCtx->dri.drawable = driDrawPriv;
-
-	 r200SetCliprects(newCtx);
-	 r200UpdateViewportOffset( newCtx->glCtx );
-      }
-
-      _mesa_make_current( newCtx->glCtx,
-			  (GLframebuffer *) driDrawPriv->driverPrivate,
-			  (GLframebuffer *) driReadPriv->driverPrivate );
-
-      _mesa_update_state( newCtx->glCtx );
-      r200ValidateState( newCtx->glCtx );
-
-   } else {
-      if (R200_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-      _mesa_make_current( NULL, NULL, NULL );
-   }
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "End %s\n", __FUNCTION__);
-   return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean
-r200UnbindContext( __DRIcontextPrivate *driContextPriv )
-{
-   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->glCtx);
-
-   return GL_TRUE;
+	int i;
+	r200ContextPtr rmesa = (r200ContextPtr)driContextPriv->driverPrivate;
+	if (rmesa)
+	{
+		for ( i = 0 ; i < R200_MAX_TEXTURE_UNITS ; i++ ) {
+			_math_matrix_dtr( &rmesa->TexGenMatrix[i] );
+		}
+	}
+	radeonDestroyContext(driContextPriv);
 }
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index 14a1dda46a..246f98c6dc 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -53,51 +53,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #error This driver requires a newer libdrm to compile
 #endif
 
+#include "radeon_screen.h"
+#include "radeon_common.h"
+
+#include "radeon_lock.h"
+
 struct r200_context;
 typedef struct r200_context r200ContextRec;
 typedef struct r200_context *r200ContextPtr;
 
-/* This union is used to avoid warnings/miscompilation
-   with float to uint32_t casts due to strict-aliasing */
-typedef union { GLfloat f; uint32_t ui32; } float_ui32_type;
-
-#include "r200_lock.h"
-#include "radeon_screen.h"
 #include "main/mm.h"
 
-/* Flags for software fallback cases */
-/* See correponding strings in r200_swtcl.c */
-#define R200_FALLBACK_TEXTURE           0x01
-#define R200_FALLBACK_DRAW_BUFFER       0x02
-#define R200_FALLBACK_STENCIL           0x04
-#define R200_FALLBACK_RENDER_MODE       0x08
-#define R200_FALLBACK_DISABLE           0x10
-#define R200_FALLBACK_BORDER_MODE       0x20
-
-/* The blit width for texture uploads
- */
-#define BLIT_WIDTH_BYTES 1024
-
-/* Use the templated vertex format:
- */
-#define COLOR_IS_RGBA
-#define TAG(x) r200##x
-#include "tnl_dd/t_dd_vertex.h"
-#undef TAG
-
-typedef void (*r200_tri_func)( r200ContextPtr,
-				 r200Vertex *,
-				 r200Vertex *,
-				 r200Vertex * );
-
-typedef void (*r200_line_func)( r200ContextPtr,
-				  r200Vertex *,
-				  r200Vertex * );
-
-typedef void (*r200_point_func)( r200ContextPtr,
-				   r200Vertex * );
-
-
 struct r200_vertex_program {
         struct gl_vertex_program mesa_program; /* Must be first */
         int translated;
@@ -112,93 +78,11 @@ struct r200_vertex_program {
         int fogmode;
 };
 
-struct r200_colorbuffer_state {
-   GLuint clear;
-#if 000
-   GLint drawOffset, drawPitch;
-#endif
-   int roundEnable;
-};
-
-
-struct r200_depthbuffer_state {
-   GLuint clear;
-   GLfloat scale;
-};
-
-#if 000
-struct r200_pixel_state {
-   GLint readOffset, readPitch;
-};
-#endif
-
-struct r200_scissor_state {
-   drm_clip_rect_t rect;
-   GLboolean enabled;
-
-   GLuint numClipRects;			/* Cliprects active */
-   GLuint numAllocedClipRects;		/* Cliprects available */
-   drm_clip_rect_t *pClipRects;
-};
-
-struct r200_stencilbuffer_state {
-   GLboolean hwBuffer;
-   GLuint clear;			/* rb3d_stencilrefmask value */
-};
-
-struct r200_stipple_state {
-   GLuint mask[32];
-};
-
-
-
-#define TEX_0   0x1
-#define TEX_1   0x2
-#define TEX_2	0x4
-#define TEX_3	0x8
-#define TEX_4	0x10
-#define TEX_5	0x20
-#define TEX_ALL 0x3f
-
-typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr;
-
-/* Texture object in locally shared texture space.
- */
-struct r200_tex_obj {
-   driTextureObject   base;
-
-   GLuint bufAddr;			/* Offset to start of locally
-					   shared texture block */
-
-   GLuint dirty_state;		        /* Flags (1 per texunit) for
-					   whether or not this texobj
-					   has dirty hardware state
-					   (pp_*) that needs to be
-					   brought into the
-					   texunit. */
-
-   drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
-					/* Six, for the cube faces */
-   GLboolean image_override;		/* Image overridden by GLX_EXT_tfp */
-
-   GLuint pp_txfilter;		        /* hardware register values */
-   GLuint pp_txformat;
-   GLuint pp_txformat_x;
-   GLuint pp_txoffset;		        /* Image location in texmem.
-					   All cube faces follow. */
-   GLuint pp_txsize;		        /* npot only */
-   GLuint pp_txpitch;		        /* npot only */
-   GLuint pp_border_color;
-   GLuint pp_cubic_faces;	        /* cube face 1,2,3,4 log2 sizes */
-
-   GLboolean  border_fallback;
-
-   GLuint tile_bits;			/* hw texture tile bits used on this texture */
-};
+#define R200_TEX_ALL 0x3f
 
 
 struct r200_texture_env_state {
-   r200TexObjPtr texobj;
+   radeonTexObjPtr texobj;
    GLuint outputreg;
    GLuint unitneeded;
 };
@@ -210,19 +94,6 @@ struct r200_texture_state {
 };
 
 
-struct r200_state_atom {
-   struct r200_state_atom *next, *prev;
-   const char *name;		         /* for debug */
-   int cmd_size;		         /* size in bytes */
-   GLuint idx;
-   int *cmd;			         /* one or more cmd's */
-   int *lastcmd;			 /* one or more cmd's */
-   GLboolean dirty;
-   GLboolean (*check)( GLcontext *, int );    /* is this state active? */
-};
-   
-
-
 /* Trying to keep these relatively short as the variables are becoming
  * extravagently long.  Drop the driver name prefix off the front of
  * everything - I think we know which driver we're in by now, and keep the
@@ -596,182 +467,96 @@ struct r200_state_atom {
 #define PRF_STATE_SIZE    3
 
 
-struct r200_hw_state {
-   /* Head of the linked list of state atoms. */
-   struct r200_state_atom atomlist;
+#define SCI_CMD_0         0
+#define SCI_RE_AUX        1
+#define SCI_CMD_1         2
+#define SCI_XY_1          3
+#define SCI_CMD_2         4
+#define SCI_XY_2          5
+#define SCI_STATE_SIZE    6
+
+#define R200_QUERYOBJ_CMD_0  0
+#define R200_QUERYOBJ_DATA_0 1
+#define R200_QUERYOBJ_CMDSIZE  2
+
+#define STP_CMD_0 0
+#define STP_DATA_0 1
+#define STP_CMD_1 2
+#define STP_STATE_SIZE 35
 
+struct r200_hw_state {
    /* Hardware state, stored as cmdbuf commands:  
     *   -- Need to doublebuffer for
     *           - reviving state after loss of context
     *           - eliding noop statechange loops? (except line stipple count)
     */
-   struct r200_state_atom ctx;
-   struct r200_state_atom set;
-   struct r200_state_atom vte;
-   struct r200_state_atom lin;
-   struct r200_state_atom msk;
-   struct r200_state_atom vpt;
-   struct r200_state_atom vap;
-   struct r200_state_atom vtx;
-   struct r200_state_atom tcl;
-   struct r200_state_atom msl;
-   struct r200_state_atom tcg;
-   struct r200_state_atom msc;
-   struct r200_state_atom cst;
-   struct r200_state_atom tam;
-   struct r200_state_atom tf;
-   struct r200_state_atom tex[6];
-   struct r200_state_atom cube[6];
-   struct r200_state_atom zbs;
-   struct r200_state_atom mtl[2];
-   struct r200_state_atom mat[9];
-   struct r200_state_atom lit[8]; /* includes vec, scl commands */
-   struct r200_state_atom ucp[6];
-   struct r200_state_atom pix[6]; /* pixshader stages */
-   struct r200_state_atom eye; /* eye pos */
-   struct r200_state_atom grd; /* guard band clipping */
-   struct r200_state_atom fog;
-   struct r200_state_atom glt;
-   struct r200_state_atom prf;
-   struct r200_state_atom afs[2];
-   struct r200_state_atom pvs;
-   struct r200_state_atom vpi[2];
-   struct r200_state_atom vpp[2];
-   struct r200_state_atom atf;
-   struct r200_state_atom spr;
-   struct r200_state_atom ptp;
-
-   int max_state_size;	/* Number of bytes necessary for a full state emit. */
-   GLboolean is_dirty, all_dirty;
+   struct radeon_state_atom ctx;
+   struct radeon_state_atom set;
+   struct radeon_state_atom sci;
+   struct radeon_state_atom vte;
+   struct radeon_state_atom lin;
+   struct radeon_state_atom msk;
+   struct radeon_state_atom vpt;
+   struct radeon_state_atom vap;
+   struct radeon_state_atom vtx;
+   struct radeon_state_atom tcl;
+   struct radeon_state_atom msl;
+   struct radeon_state_atom tcg;
+   struct radeon_state_atom msc;
+   struct radeon_state_atom cst;
+   struct radeon_state_atom tam;
+   struct radeon_state_atom tf;
+   struct radeon_state_atom tex[6];
+   struct radeon_state_atom cube[6];
+   struct radeon_state_atom zbs;
+   struct radeon_state_atom mtl[2];
+   struct radeon_state_atom mat[9];
+   struct radeon_state_atom lit[8]; /* includes vec, scl commands */
+   struct radeon_state_atom ucp[6];
+   struct radeon_state_atom pix[6]; /* pixshader stages */
+   struct radeon_state_atom eye; /* eye pos */
+   struct radeon_state_atom grd; /* guard band clipping */
+   struct radeon_state_atom fog;
+   struct radeon_state_atom glt;
+   struct radeon_state_atom prf;
+   struct radeon_state_atom afs[2];
+   struct radeon_state_atom pvs;
+   struct radeon_state_atom vpi[2];
+   struct radeon_state_atom vpp[2];
+   struct radeon_state_atom atf;
+   struct radeon_state_atom spr;
+   struct radeon_state_atom ptp;
+   struct radeon_state_atom stp;
 };
 
 struct r200_state {
    /* Derived state for internal purposes:
     */
-   struct r200_colorbuffer_state color;
-   struct r200_depthbuffer_state depth;
-#if 00
-   struct r200_pixel_state pixel;
-#endif
-   struct r200_scissor_state scissor;
-   struct r200_stencilbuffer_state stencil;
-   struct r200_stipple_state stipple;
    struct r200_texture_state texture;
    GLuint envneeded;
 };
 
-/* Need refcounting on dma buffers:
- */
-struct r200_dma_buffer {
-   int refcount;		/* the number of retained regions in buf */
-   drmBufPtr buf;
-};
-
-#define GET_START(rvb) (rmesa->r200Screen->gart_buffer_offset +		\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-
-/* A retained region, eg vertices for indexed vertices.
- */
-struct r200_dma_region {
-   struct r200_dma_buffer *buf;
-   char *address;		/* == buf->address */
-   int start, end, ptr;		/* offsets from start of buf */
-   int aos_start;
-   int aos_stride;
-   int aos_size;
-};
-
-
-struct r200_dma {
-   /* Active dma region.  Allocations for vertices and retained
-    * regions come from here.  Also used for emitting random vertices,
-    * these may be flushed by calling flush_current();
-    */
-   struct r200_dma_region current;
-   
-   void (*flush)( r200ContextPtr );
-
-   char *buf0_address;		/* start of buf[0], for index calcs */
-   GLuint nr_released_bufs;	/* flush after so many buffers released */
-};
-
-struct r200_dri_mirror {
-   __DRIcontextPrivate	*context;	/* DRI context */
-   __DRIscreenPrivate	*screen;	/* DRI screen */
-   __DRIdrawablePrivate	*drawable;	/* DRI drawable bound to this ctx */
-   __DRIdrawablePrivate	*readable;	/* DRI readable bound to this ctx */
-
-   drm_context_t hwContext;
-   drm_hw_lock_t *hwLock;
-   int fd;
-   int drmMinor;
-};
-
-
 #define R200_CMD_BUF_SZ  (16*1024) 
 
-struct r200_store {
-   GLuint statenr;
-   GLuint primnr;
-   char cmd_buf[R200_CMD_BUF_SZ];
-   int cmd_used;   
-   int elts_start;
-};
-
-
+#define R200_ELT_BUF_SZ  (16*1024) 
 /* r200_tcl.c
  */
 struct r200_tcl_info {
    GLuint hw_primitive;
 
-/* hw can handle 12 components max */
-   struct r200_dma_region *aos_components[12];
-   GLuint nr_aos_components;
-
-   GLuint *Elts;
+   int elt_used;
 
-   struct r200_dma_region indexed_verts;
-   struct r200_dma_region vertex_data[15];
 };
 
 
 /* r200_swtcl.c
  */
 struct r200_swtcl_info {
-   GLuint RenderIndex;
-   
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
 
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
 
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
-
-   /* Fallback rasterization functions
-    */
-   r200_point_func draw_point;
-   r200_line_func draw_line;
-   r200_tri_func draw_tri;
-
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
+   radeon_point_func draw_point;
+   radeon_line_func draw_line;
+   radeon_tri_func draw_tri;
 
    /**
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
@@ -787,27 +572,10 @@ struct r200_swtcl_info {
     * Should Mesa project vertex data or will the hardware do it?
     */
    GLboolean needproj;
-
-   struct r200_dma_region indexed_verts;
-};
-
-
-struct r200_ioctl {
-   GLuint vertex_offset;
-   GLuint vertex_size;
 };
 
 
 
-#define R200_MAX_PRIMS 64
-
-
-
-struct r200_prim {
-   GLuint start;
-   GLuint end;
-   GLuint prim;
-};
 
    /* A maximum total of 29 elements per vertex:  3 floats for position, 3
     * floats for normal, 4 floats for color, 4 bytes for secondary color,
@@ -822,9 +590,8 @@ struct r200_prim {
 
 #define R200_MAX_VERTEX_SIZE ((3*6)+11)
 
-
 struct r200_context {
-   GLcontext *glCtx;			/* Mesa context */
+   struct radeon_context radeon;
 
    /* Driver and hardware state management
     */
@@ -832,56 +599,15 @@ struct r200_context {
    struct r200_state state;
    struct r200_vertex_program *curr_vp_hw;
 
-   /* Texture object bookkeeping
-    */
-   unsigned              nr_heaps;
-   driTexHeap          * texture_heaps[ RADEON_NR_TEX_HEAPS ];
-   driTextureObject      swapped;
-   int                   texture_depth;
-   float                 initialMaxAnisotropy;
-
-   /* Rasterization and vertex state:
-    */
-   GLuint TclFallback;
-   GLuint Fallback;
-   GLuint NewGLState;
-   DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
-
    /* Vertex buffers
     */
-   struct r200_ioctl ioctl;
-   struct r200_dma dma;
-   struct r200_store store;
-   /* A full state emit as of the first state emit in the main store, in case
-    * the context is lost.
-    */
-   struct r200_store backup_store;
-
-   /* Page flipping
-    */
-   GLuint doPageFlip;
-
-   /* Busy waiting
-    */
-   GLuint do_usleeps;
-   GLuint do_irqs;
-   GLuint irqsEmitted;
-   drm_radeon_irq_wait_t iw;
+   struct radeon_ioctl ioctl;
+   struct radeon_store store;
 
    /* Clientdata textures;
     */
    GLuint prefer_gart_client_texturing;
 
-   /* Drawable, cliprect and scissor information
-    */
-   GLuint numClipRects;			/* Cliprects for the draw buffer */
-   drm_clip_rect_t *pClipRects;
-   unsigned int lastStamp;
-   GLboolean lost_context;
-   GLboolean save_on_next_emit;
-   radeonScreenPtr r200Screen;	/* Screen private DRI data */
-   drm_radeon_sarea_t *sarea;		/* Private SAREA data */
-
    /* TCL stuff
     */
    GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS];
@@ -893,15 +619,6 @@ struct r200_context {
    GLuint TexGenCompSel;
    GLmatrix tmpmat;
 
-   /* buffer swap
-    */
-   int64_t swap_ust;
-   int64_t swap_missed_ust;
-
-   GLuint swap_count;
-   GLuint swap_missed_count;
-
-
    /* r200_tcl.c
     */
    struct r200_tcl_info tcl;
@@ -910,14 +627,6 @@ struct r200_context {
     */
    struct r200_swtcl_info swtcl;
 
-   /* Mirrors of some DRI state
-    */
-   struct r200_dri_mirror dri;
-
-   /* Configuration cache
-    */
-   driOptionCache optionCache;
-
    GLboolean using_hyperz;
    GLboolean texmicrotile;
 
@@ -927,28 +636,10 @@ struct r200_context {
 #define R200_CONTEXT(ctx)		((r200ContextPtr)(ctx->DriverCtx))
 
 
-static INLINE GLuint r200PackColor( GLuint cpp,
-					GLubyte r, GLubyte g,
-					GLubyte b, GLubyte a )
-{
-   switch ( cpp ) {
-   case 2:
-      return PACK_COLOR_565( r, g, b );
-   case 4:
-      return PACK_COLOR_8888( a, r, g, b );
-   default:
-      return 0;
-   }
-}
-
-
 extern void r200DestroyContext( __DRIcontextPrivate *driContextPriv );
 extern GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 				    __DRIcontextPrivate *driContextPriv,
 				    void *sharedContextPrivate);
-extern void r200SwapBuffers( __DRIdrawablePrivate *dPriv );
-extern void r200CopySubBuffer( __DRIdrawablePrivate * dPriv,
-			       int x, int y, int w, int h );
 extern GLboolean r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
 				  __DRIdrawablePrivate *driDrawPriv,
 				  __DRIdrawablePrivate *driReadPriv );
@@ -957,28 +648,9 @@ extern GLboolean r200UnbindContext( __DRIcontextPrivate *driContextPriv );
 /* ================================================================
  * Debugging:
  */
-#define DO_DEBUG		1
 
-#if DO_DEBUG
-extern int R200_DEBUG;
-#else
-#define R200_DEBUG		0
-#endif
+#define R200_DEBUG RADEON_DEBUG
+
 
-#define DEBUG_TEXTURE	0x001
-#define DEBUG_STATE	0x002
-#define DEBUG_IOCTL	0x004
-#define DEBUG_PRIMS	0x008
-#define DEBUG_VERTS	0x010
-#define DEBUG_FALLBACKS	0x020
-#define DEBUG_VFMT	0x040
-#define DEBUG_CODEGEN	0x080
-#define DEBUG_VERBOSE	0x100
-#define DEBUG_DRI       0x200
-#define DEBUG_DMA       0x400
-#define DEBUG_SANITY    0x800
-#define DEBUG_SYNC      0x1000
-#define DEBUG_PIXEL     0x2000
-#define DEBUG_MEMORY    0x4000
 
 #endif /* __R200_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_fragshader.c b/src/mesa/drivers/dri/r200/r200_fragshader.c
index d514b28219..85c1b7bdd1 100644
--- a/src/mesa/drivers/dri/r200/r200_fragshader.c
+++ b/src/mesa/drivers/dri/r200/r200_fragshader.c
@@ -522,7 +522,7 @@ static void r200UpdateFSConstants( GLcontext *ctx )
 	 CLAMPED_FLOAT_TO_UBYTE(con_byte[2], ctx->ATIFragmentShader.GlobalConstants[i][2]);
 	 CLAMPED_FLOAT_TO_UBYTE(con_byte[3], ctx->ATIFragmentShader.GlobalConstants[i][3]);
       }
-      rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = r200PackColor (
+      rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = radeonPackColor (
 	 4, con_byte[0], con_byte[1], con_byte[2], con_byte[3] );
    }
 }
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c
index 0741e57af7..b238adb972 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.c
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.c
@@ -31,7 +31,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Authors:
  *   Keith Whitwell <keith@tungstengraphics.com>
  */
- 
+
 #include <sched.h>
 #include <errno.h>
 
@@ -41,6 +41,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "swrast/swrast.h"
 
+
+
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -54,635 +58,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R200_TIMEOUT             512
 #define R200_IDLE_RETRY           16
 
-
-static void r200WaitForIdle( r200ContextPtr rmesa );
-
-
-/* At this point we were in FlushCmdBufLocked but we had lost our context, so
- * we need to unwire our current cmdbuf, hook the one with the saved state in
- * it, flush it, and then put the current one back.  This is so commands at the
- * start of a cmdbuf can rely on the state being kept from the previous one.
- */
-static void r200BackUpAndEmitLostStateLocked( r200ContextPtr rmesa )
-{
-   GLuint nr_released_bufs;
-   struct r200_store saved_store;
-
-   if (rmesa->backup_store.cmd_used == 0)
-      return;
-
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Emitting backup state on lost context\n");
-
-   rmesa->lost_context = GL_FALSE;
-
-   nr_released_bufs = rmesa->dma.nr_released_bufs;
-   saved_store = rmesa->store;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->store = rmesa->backup_store;
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-   rmesa->dma.nr_released_bufs = nr_released_bufs;
-   rmesa->store = saved_store;
-}
-
-int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller )
-{
-   int ret, i;
-   drm_radeon_cmd_buffer_t cmd;
-
-   if (rmesa->lost_context)
-      r200BackUpAndEmitLostStateLocked( rmesa );
-
-   if (R200_DEBUG & DEBUG_IOCTL) {
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-
-      if (0 & R200_DEBUG & DEBUG_VERBOSE) 
-	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
-	    fprintf(stderr, "%d: %x\n", i/4, 
-		    *(int *)(&rmesa->store.cmd_buf[i]));
-   }
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
-	      rmesa->dma.nr_released_bufs);
-
-
-   if (R200_DEBUG & DEBUG_SANITY) {
-      if (rmesa->state.scissor.enabled) 
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->state.scissor.numClipRects,
-				    rmesa->state.scissor.pClipRects);
-      else
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->numClipRects,
-				    rmesa->pClipRects);
-      if (ret) {
-	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
-	 goto out;
-      }
-   }
-
-
-   if (R200_DEBUG & DEBUG_MEMORY) {
-      if (! driValidateTextureHeaps( rmesa->texture_heaps, rmesa->nr_heaps,
-				     & rmesa->swapped ) ) {
-	 fprintf( stderr, "%s: texture memory is inconsistent - expect "
-		  "mangled textures\n", __FUNCTION__ );
-      }
-   }
-
-
-   cmd.bufsz = rmesa->store.cmd_used;
-   cmd.buf = rmesa->store.cmd_buf;
-
-   if (rmesa->state.scissor.enabled) {
-      cmd.nbox = rmesa->state.scissor.numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->state.scissor.pClipRects;
-   } else {
-      cmd.nbox = rmesa->numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->pClipRects;
-   }
-
-   ret = drmCommandWrite( rmesa->dri.fd,
-			  DRM_RADEON_CMDBUF,
-			  &cmd, sizeof(cmd) );
-
-   if (ret)
-      fprintf(stderr, "drmCommandWrite: %d\n", ret);
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
-      r200WaitForIdleLocked( rmesa );
-   }
-
-
- out:
-   rmesa->store.primnr = 0;
-   rmesa->store.statenr = 0;
-   rmesa->store.cmd_used = 0;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->save_on_next_emit = 1;
-
-   return ret;
-}
-
-
-/* Note: does not emit any commands to avoid recursion on
- * r200AllocCmdBuf.
- */
-void r200FlushCmdBuf( r200ContextPtr rmesa, const char *caller )
-{
-   int ret;
-
-   LOCK_HARDWARE( rmesa );
-
-   ret = r200FlushCmdBufLocked( rmesa, caller );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if (ret) {
-      fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret);
-      exit(ret);
-   }
-}
-
-
-/* =============================================================
- * Hardware vertex buffer handling
- */
-
-
-void r200RefillCurrentDmaRegion( r200ContextPtr rmesa )
-{
-   struct r200_dma_buffer *dmabuf;
-   int fd = rmesa->dri.fd;
-   int index = 0;
-   int size = 0;
-   drmDMAReq dma;
-   int ret;
-
-   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-      fprintf(stderr, "%s\n", __FUNCTION__);  
-
-   if (rmesa->dma.flush) {
-      rmesa->dma.flush( rmesa );
-   }
-
-   if (rmesa->dma.current.buf)
-      r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-
-   if (rmesa->dma.nr_released_bufs > 4)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-
-   dma.context = rmesa->dri.hwContext;
-   dma.send_count = 0;
-   dma.send_list = NULL;
-   dma.send_sizes = NULL;
-   dma.flags = 0;
-   dma.request_count = 1;
-   dma.request_size = RADEON_BUFFER_SIZE;
-   dma.request_list = &index;
-   dma.request_sizes = &size;
-   dma.granted_count = 0;
-
-   LOCK_HARDWARE(rmesa);	/* no need to validate */
-
-   while (1) {
-      ret = drmDMA( fd, &dma );
-      if (ret == 0)
-	 break;
-   
-      if (rmesa->dma.nr_released_bufs) {
-	 r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-      }
-
-      if (rmesa->do_usleeps) {
-	 UNLOCK_HARDWARE( rmesa );
-	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
-      }
-   }
-
-   UNLOCK_HARDWARE(rmesa);
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "Allocated buffer %d\n", index);
-
-   dmabuf = CALLOC_STRUCT( r200_dma_buffer );
-   dmabuf->buf = &rmesa->r200Screen->buffers->list[index];
-   dmabuf->refcount = 1;
-
-   rmesa->dma.current.buf = dmabuf;
-   rmesa->dma.current.address = dmabuf->buf->address;
-   rmesa->dma.current.end = dmabuf->buf->total;
-   rmesa->dma.current.start = 0;
-   rmesa->dma.current.ptr = 0;
-}
-
-void r200ReleaseDmaRegion( r200ContextPtr rmesa,
-			     struct r200_dma_region *region,
-			     const char *caller )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-   
-   if (!region->buf)
-      return;
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (--region->buf->refcount == 0) {
-      drm_radeon_cmd_header_t *cmd;
-
-      if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
-		 region->buf->buf->idx);  
-      
-      cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sizeof(*cmd), 
-						     __FUNCTION__ );
-      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
-      cmd->dma.buf_idx = region->buf->buf->idx;
-      FREE(region->buf);
-      rmesa->dma.nr_released_bufs++;
-   }
-
-   region->buf = NULL;
-   region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r200AllocDmaRegion( r200ContextPtr rmesa, 
-			   struct r200_dma_region *region,
-			   int bytes,
-			   int alignment )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (region->buf)
-      r200ReleaseDmaRegion( rmesa, region, __FUNCTION__ );
-
-   alignment--;
-   rmesa->dma.current.start = rmesa->dma.current.ptr = 
-      (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      r200RefillCurrentDmaRegion( rmesa );
-
-   region->start = rmesa->dma.current.start;
-   region->ptr = rmesa->dma.current.start;
-   region->end = rmesa->dma.current.start + bytes;
-   region->address = rmesa->dma.current.address;
-   region->buf = rmesa->dma.current.buf;
-   region->buf->refcount++;
-
-   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
-   rmesa->dma.current.start = 
-      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
-
-   assert( rmesa->dma.current.ptr <= rmesa->dma.current.end );
-}
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t r200GetLastFrame(r200ContextPtr rmesa)
-{
-   drm_radeon_getparam_t gp;
-   int ret;
-   uint32_t frame;
-
-   gp.param = RADEON_PARAM_LAST_FRAME;
-   gp.value = (int *)&frame;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM,
-			      &gp, sizeof(gp) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-
-   return frame;
-}
-
-static void r200EmitIrqLocked( r200ContextPtr rmesa )
-{
-   drm_radeon_irq_emit_t ie;
-   int ret;
-
-   ie.irq_seq = &rmesa->iw.irq_seq;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, 
-			      &ie, sizeof(ie) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void r200WaitIrq( r200ContextPtr rmesa )
-{
-   int ret;
-
-   do {
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT,
-			     &rmesa->iw, sizeof(rmesa->iw) );
-   } while (ret && (errno == EINTR || errno == EBUSY));
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void r200WaitForFrameCompletion( r200ContextPtr rmesa )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-
-   if (rmesa->do_irqs) {
-      if (r200GetLastFrame(rmesa) < sarea->last_frame) {
-	 if (!rmesa->irqsEmitted) {
-	    while (r200GetLastFrame (rmesa) < sarea->last_frame)
-	       ;
-	 }
-	 else {
-	    UNLOCK_HARDWARE( rmesa ); 
-	    r200WaitIrq( rmesa );	
-	    LOCK_HARDWARE( rmesa ); 
-	 }
-	 rmesa->irqsEmitted = 10;
-      }
-
-      if (rmesa->irqsEmitted) {
-	 r200EmitIrqLocked( rmesa );
-	 rmesa->irqsEmitted--;
-      }
-   } 
-   else {
-      while (r200GetLastFrame (rmesa) < sarea->last_frame) {
-	 UNLOCK_HARDWARE( rmesa ); 
-	 if (rmesa->do_usleeps) 
-	    DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa ); 
-      }
-   }
-}
-
-
-
-/* Copy the back color buffer to the front color buffer.
- */
-void r200CopyBuffer( __DRIdrawablePrivate *dPriv,
-		      const drm_clip_rect_t	 *rect)
-{
-   r200ContextPtr rmesa;
-   GLint nbox, i, ret;
-   GLboolean   missed_target;
-   int64_t ust;
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *)rmesa->glCtx );
-   }
-
-   R200_FIREVERTICES( rmesa );
-
-   LOCK_HARDWARE( rmesa );
-
-
-   /* Throttle the frame rate -- only allow one pending swap buffers
-    * request at a time.
-    */
-   r200WaitForFrameCompletion( rmesa );
-   if (!rect)
-   {
-       UNLOCK_HARDWARE( rmesa );
-       driWaitForVBlank( dPriv, & missed_target );
-       LOCK_HARDWARE( rmesa );
-   }
-
-   nbox = dPriv->numClipRects; /* must be in locked region */
-
-   for ( i = 0 ; i < nbox ; ) {
-      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      GLint n = 0;
-
-      for ( ; i < nr ; i++ ) {
-
-	  *b = box[i];
-
-	  if (rect)
-	  {
-	     if (rect->x1 > b->x1)
-		 b->x1 = rect->x1;
-	     if (rect->y1 > b->y1)
-		 b->y1 = rect->y1;
-	     if (rect->x2 < b->x2)
-		 b->x2 = rect->x2;
-	     if (rect->y2 < b->y2)
-		 b->y2 = rect->y2;
-
-	     if (b->x1 >= b->x2 || b->y1 >= b->y2)
-		 continue;
-	  }
-
-	  b++;
-	  n++;
-      }
-      rmesa->sarea->nbox = n;
-
-      if (!n)
-	 continue;
-
-      ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
-
-      if ( ret ) {
-	 fprintf( stderr, "DRM_R200_SWAP_BUFFERS: return = %d\n", ret );
-	 UNLOCK_HARDWARE( rmesa );
-	 exit( 1 );
-      }
-   }
-
-   UNLOCK_HARDWARE( rmesa );
-   if (!rect)
-   {
-       rmesa->hw.all_dirty = GL_TRUE;
-
-       rmesa->swap_count++;
-       (*psp->systemTime->getUST)( & ust );
-       if ( missed_target ) {
-	   rmesa->swap_missed_count++;
-	   rmesa->swap_missed_ust = ust - rmesa->swap_ust;
-       }
-
-       rmesa->swap_ust = ust;
-
-       sched_yield();
-   }
-}
-
-void r200PageFlip( __DRIdrawablePrivate *dPriv )
-{
-   r200ContextPtr rmesa;
-   GLint ret;
-   GLboolean   missed_target;
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-	      rmesa->sarea->pfCurrentPage);
-   }
-
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   if (!dPriv->numClipRects) {
-      UNLOCK_HARDWARE( rmesa );
-      usleep( 10000 );		/* throttle invisible client 10ms */
-      return;
-   }
-
-   /* Need to do this for the perf box placement:
-    */
-   {
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      b[0] = box[0];
-      rmesa->sarea->nbox = 1;
-   }
-
-   /* Throttle the frame rate -- only allow a few pending swap buffers
-    * request at a time.
-    */
-   r200WaitForFrameCompletion( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   driWaitForVBlank( dPriv, & missed_target );
-   if ( missed_target ) {
-      rmesa->swap_missed_count++;
-      (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust );
-   }
-   LOCK_HARDWARE( rmesa );
-
-   ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
-      exit( 1 );
-   }
-
-   rmesa->swap_count++;
-   (void) (*psp->systemTime->getUST)( & rmesa->swap_ust );
-
-#if 000
-   if ( rmesa->sarea->pfCurrentPage == 1 ) {
-	 rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-	 rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
-   } else {
-	 rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-	 rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
-   }
-
-   R200_STATECHANGE( rmesa, ctx );
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = rmesa->state.color.drawOffset
-					   + rmesa->r200Screen->fbLocation;
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH]  = rmesa->state.color.drawPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
-#else
-   /* Get ready for drawing next frame.  Update the renderbuffers'
-    * flippedOffset/Pitch fields so we draw into the right place.
-    */
-   driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                        rmesa->sarea->pfCurrentPage);
-
-
-   r200UpdateDrawBuffer(rmesa->glCtx);
-#endif
-}
-
-
-/* ================================================================
- * Buffer clear
- */
-static void r200Clear( GLcontext *ctx, GLbitfield mask )
+static void r200KernelClear(GLcontext *ctx, GLuint flags)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLuint flags = 0;
-   GLuint color_mask = 0;
-   GLint ret, i;
-   GLint cx, cy, cw, ch;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "r200Clear\n");
-   }
-
-   {
-      LOCK_HARDWARE( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      if ( dPriv->numClipRects == 0 ) 
-	 return;
-   }
-
-   r200Flush( ctx );
-
-   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
-      flags |= RADEON_FRONT;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_FRONT_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_BACK_LEFT ) {
-      flags |= RADEON_BACK;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_BACK_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_DEPTH ) {
-      flags |= RADEON_DEPTH;
-      mask &= ~BUFFER_BIT_DEPTH;
-   }
-
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->state.stencil.hwBuffer ) {
-      flags |= RADEON_STENCIL;
-      mask &= ~BUFFER_BIT_STENCIL;
-   }
-
-   if ( mask ) {
-      if (R200_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
-      _swrast_Clear( ctx, mask );
-   }
-
-   if ( !flags ) 
-      return;
-
-   if (rmesa->using_hyperz) {
-      flags |= RADEON_USE_COMP_ZBUF;
-/*      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200)
-	 flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
-	    ((rmesa->state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) {
-	  flags |= RADEON_CLEAR_FASTZ;
-      }
-   }
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLint cx, cy, cw, ch, ret;
+   GLuint i;
 
-   LOCK_HARDWARE( rmesa );
-
-   /* compute region after locking: */
-   cx = ctx->DrawBuffer->_Xmin;
-   cy = ctx->DrawBuffer->_Ymin;
-   cw = ctx->DrawBuffer->_Xmax - cx;
-   ch = ctx->DrawBuffer->_Ymax - cy;
-
-   /* Flip top to bottom */
-   cx += dPriv->x;
-   cy  = dPriv->y + dPriv->h - cy - ch;
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* Throttle the number of clear ioctls we do.
     */
@@ -693,7 +76,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 
       gp.param = RADEON_PARAM_LAST_CLEAR;
       gp.value = (int *)&clear;
-      ret = drmCommandWriteRead( rmesa->dri.fd,
+      ret = drmCommandWriteRead( rmesa->radeon.dri.fd,
 		      DRM_RADEON_GETPARAM, &gp, sizeof(gp) );
 
       if ( ret ) {
@@ -703,24 +86,34 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 
       /* Clear throttling needs more thought.
        */
-      if ( rmesa->sarea->last_clear - clear <= 25 ) {
+      if ( rmesa->radeon.sarea->last_clear - clear <= 25 ) {
 	 break;
       }
-      
-      if (rmesa->do_usleeps) {
-	 UNLOCK_HARDWARE( rmesa );
+
+      if (rmesa->radeon.do_usleeps) {
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
+	 LOCK_HARDWARE( &rmesa->radeon );
       }
    }
 
    /* Send current state to the hardware */
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
+
+
+  /* compute region after locking: */
+   cx = ctx->DrawBuffer->_Xmin;
+   cy = ctx->DrawBuffer->_Ymin;
+   cw = ctx->DrawBuffer->_Xmax - cx;
+   ch = ctx->DrawBuffer->_Ymax - cy;
 
+   /* Flip top to bottom */
+   cx += dPriv->x;
+   cy  = dPriv->y + dPriv->h - cy - ch;
    for ( i = 0 ; i < dPriv->numClipRects ; ) {
       GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects );
       drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
+      drm_clip_rect_t *b = rmesa->radeon.sarea->boxes;
       drm_radeon_clear_t clear;
       drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
       GLint n = 0;
@@ -755,17 +148,17 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 	 }
       }
 
-      rmesa->sarea->nbox = n;
+      rmesa->radeon.sarea->nbox = n;
 
       clear.flags       = flags;
-      clear.clear_color = rmesa->state.color.clear;
-      clear.clear_depth = rmesa->state.depth.clear;	/* needed for hyperz */
+      clear.clear_color = rmesa->radeon.state.color.clear;
+      clear.clear_depth = rmesa->radeon.state.depth.clear;	/* needed for hyperz */
       clear.color_mask  = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      clear.depth_mask  = rmesa->state.stencil.clear;
+      clear.depth_mask  = rmesa->radeon.state.stencil.clear;
       clear.depth_boxes = depth_boxes;
 
       n--;
-      b = rmesa->sarea->boxes;
+      b = rmesa->radeon.sarea->boxes;
       for ( ; n >= 0 ; n-- ) {
 	 depth_boxes[n].f[CLEAR_X1] = (float)b[n].x1;
 	 depth_boxes[n].f[CLEAR_Y1] = (float)b[n].y1;
@@ -774,84 +167,94 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 	 depth_boxes[n].f[CLEAR_DEPTH] = ctx->Depth.Clear;
       }
 
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR,
+      ret = drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_CLEAR,
 			     &clear, sizeof(clear));
 
 
       if ( ret ) {
-	 UNLOCK_HARDWARE( rmesa );
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret );
 	 exit( 1 );
       }
    }
-
-   UNLOCK_HARDWARE( rmesa );
-   rmesa->hw.all_dirty = GL_TRUE;
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
+/* ================================================================
+ * Buffer clear
+ */
+static void r200Clear( GLcontext *ctx, GLbitfield mask )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLuint flags = 0;
+   GLuint color_mask = 0;
+   GLuint orig_mask = mask;
 
+   if ( R200_DEBUG & RADEON_IOCTL ) {
+	   if (rmesa->radeon.sarea)
+	       fprintf( stderr, "r200Clear %x %d\n", mask, rmesa->radeon.sarea->pfCurrentPage);
+	   else
+	       fprintf( stderr, "r200Clear %x radeon->sarea is NULL\n", mask);
+   }
 
-void r200WaitForIdleLocked( r200ContextPtr rmesa )
-{
-    int ret;
-    int i = 0;
-    
-    do {
-       ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_CP_IDLE);
-       if (ret) 
-	  DO_USLEEP( 1 );
-    } while (ret && ++i < 100);
-    
-    if ( ret < 0 ) {
-       UNLOCK_HARDWARE( rmesa );
-       fprintf( stderr, "Error: R200 timed out... exiting\n" );
-       exit( -1 );
-    }
-}
+   {
+      LOCK_HARDWARE( &rmesa->radeon );
+      UNLOCK_HARDWARE( &rmesa->radeon );
+      if ( dPriv->numClipRects == 0 )
+	 return;
+   }
 
+   radeonFlush( ctx );
 
-static void r200WaitForIdle( r200ContextPtr rmesa )
-{
-   LOCK_HARDWARE(rmesa);
-   r200WaitForIdleLocked( rmesa );
-   UNLOCK_HARDWARE(rmesa);
-}
+   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
+      flags |= RADEON_FRONT;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_FRONT_LEFT;
+   }
 
+   if ( mask & BUFFER_BIT_BACK_LEFT ) {
+      flags |= RADEON_BACK;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
 
-void r200Flush( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+   if ( mask & BUFFER_BIT_DEPTH ) {
+      flags |= RADEON_DEPTH;
+      mask &= ~BUFFER_BIT_DEPTH;
+   }
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
+      flags |= RADEON_STENCIL;
+      mask &= ~BUFFER_BIT_STENCIL;
+   }
 
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+   if ( mask ) {
+      if (R200_DEBUG & RADEON_FALLBACKS)
+	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
+      _swrast_Clear( ctx, mask );
+   }
 
-   r200EmitState( rmesa );
-   
-   if (rmesa->store.cmd_used)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-}
+   if ( !flags )
+      return;
 
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void r200Finish( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200Flush( ctx );
+   if (rmesa->using_hyperz) {
+      flags |= RADEON_USE_COMP_ZBUF;
+/*      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200)
+	 flags |= RADEON_USE_HIERZ; */
+      if (!((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+	    ((rmesa->radeon.state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) {
+	  flags |= RADEON_CLEAR_FASTZ;
+      }
+   }
 
-   if (rmesa->do_irqs) {
-      LOCK_HARDWARE( rmesa );
-      r200EmitIrqLocked( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      r200WaitIrq( rmesa );
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+      radeonUserClear(ctx, orig_mask);
+   else {
+      r200KernelClear(ctx, flags);
+      rmesa->radeon.hw.all_dirty = GL_TRUE;
    }
-   else 
-      r200WaitForIdle( rmesa );
 }
 
-
 /* This version of AllocateMemoryMESA allocates only GART memory, and
  * only does so after the point at which the driver has been
  * initialized.
@@ -862,7 +265,7 @@ void r200Finish( GLcontext *ctx )
  * device fd.
  */
 void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
-			     GLfloat readfreq, GLfloat writefreq, 
+			     GLfloat readfreq, GLfloat writefreq,
 			     GLfloat priority)
 {
    GET_CURRENT_CONTEXT(ctx);
@@ -871,11 +274,11 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
    drm_radeon_mem_alloc_t alloc;
    int ret;
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, 
+   if (R200_DEBUG & RADEON_IOCTL)
+      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq,
 	      writefreq, priority);
 
-   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map)
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map)
       return NULL;
 
    if (getenv("R200_NO_ALLOC"))
@@ -886,17 +289,17 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
    alloc.size = size;
    alloc.region_offset = &region_offset;
 
-   ret = drmCommandWriteRead( rmesa->r200Screen->driScreen->fd,
+   ret = drmCommandWriteRead( rmesa->radeon.radeonScreen->driScreen->fd,
 			      DRM_RADEON_ALLOC,
 			      &alloc, sizeof(alloc));
-   
+
    if (ret) {
       fprintf(stderr, "%s: DRM_RADEON_ALLOC ret %d\n", __FUNCTION__, ret);
       return NULL;
    }
-   
+
    {
-      char *region_start = (char *)rmesa->r200Screen->gartTextures.map;
+      char *region_start = (char *)rmesa->radeon.radeonScreen->gartTextures.map;
       return (void *)(region_start + region_offset);
    }
 }
@@ -911,31 +314,31 @@ void r200FreeMemoryMESA(__DRIscreen *screen, GLvoid *pointer)
    drm_radeon_mem_free_t memfree;
    int ret;
 
-   if (R200_DEBUG & DEBUG_IOCTL)
+   if (R200_DEBUG & RADEON_IOCTL)
       fprintf(stderr, "%s %p\n", __FUNCTION__, pointer);
 
-   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map) {
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map) {
       fprintf(stderr, "%s: no context\n", __FUNCTION__);
       return;
    }
 
-   region_offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   region_offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
 
-   if (region_offset < 0 || 
-       region_offset > rmesa->r200Screen->gartTextures.size) {
+   if (region_offset < 0 ||
+       region_offset > rmesa->radeon.radeonScreen->gartTextures.size) {
       fprintf(stderr, "offset %d outside range 0..%d\n", region_offset,
-	      rmesa->r200Screen->gartTextures.size);
+	      rmesa->radeon.radeonScreen->gartTextures.size);
       return;
    }
 
    memfree.region = RADEON_MEM_REGION_GART;
    memfree.region_offset = region_offset;
-   
-   ret = drmCommandWrite( rmesa->r200Screen->driScreen->fd,
+
+   ret = drmCommandWrite( rmesa->radeon.radeonScreen->driScreen->fd,
 			  DRM_RADEON_FREE,
 			  &memfree, sizeof(memfree));
-   
-   if (ret) 
+
+   if (ret)
       fprintf(stderr, "%s: DRM_RADEON_FREE ret %d\n", __FUNCTION__, ret);
 }
 
@@ -956,32 +359,32 @@ GLuint r200GetMemoryOffsetMESA(__DRIscreen *screen, const GLvoid *pointer)
 
    card_offset = r200GartOffsetFromVirtual( rmesa, pointer );
 
-   return card_offset - rmesa->r200Screen->gart_base;
+   return card_offset - rmesa->radeon.radeonScreen->gart_base;
 }
 
 GLboolean r200IsGartMemory( r200ContextPtr rmesa, const GLvoid *pointer,
 			   GLint size )
 {
-   ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
    int valid = (size >= 0 &&
 		offset >= 0 &&
-		offset + size < rmesa->r200Screen->gartTextures.size);
+		offset + size < rmesa->radeon.radeonScreen->gartTextures.size);
 
-   if (R200_DEBUG & DEBUG_IOCTL)
+   if (R200_DEBUG & RADEON_IOCTL)
       fprintf(stderr, "r200IsGartMemory( %p ) : %d\n", pointer, valid );
-   
+
    return valid;
 }
 
 
 GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer )
 {
-   ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
 
-   if (offset < 0 || offset > rmesa->r200Screen->gartTextures.size)
+   if (offset < 0 || offset > rmesa->radeon.radeonScreen->gartTextures.size)
       return ~0;
    else
-      return rmesa->r200Screen->gart_texture_offset + offset;
+      return rmesa->radeon.radeonScreen->gart_texture_offset + offset;
 }
 
 
@@ -989,7 +392,7 @@ GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer )
 void r200InitIoctlFuncs( struct dd_function_table *functions )
 {
     functions->Clear = r200Clear;
-    functions->Finish = r200Finish;
-    functions->Flush = r200Flush;
+    functions->Finish = radeonFinish;
+    functions->Flush = radeonFlush;
 }
 
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.h b/src/mesa/drivers/dri/r200/r200_ioctl.h
index f7458e4a0e..8d51aefa04 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.h
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.h
@@ -37,65 +37,31 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/simple_list.h"
 #include "radeon_dri.h"
-#include "r200_lock.h"
+
+#include "radeon_bocs_wrapper.h"
 
 #include "xf86drm.h"
 #include "drm.h"
 #include "radeon_drm.h"
 
-extern void r200EmitState( r200ContextPtr rmesa );
+extern void r200EmitMaxVtxIndex(r200ContextPtr rmesa, int count);
 extern void r200EmitVertexAOS( r200ContextPtr rmesa,
-				 GLuint vertex_size,
-				 GLuint offset );
+			       GLuint vertex_size,
+			       struct radeon_bo *bo,
+			       GLuint offset );
 
 extern void r200EmitVbufPrim( r200ContextPtr rmesa,
 				GLuint primitive,
 				GLuint vertex_nr );
 
-extern void r200FlushElts( r200ContextPtr rmesa );
+extern void r200FlushElts(GLcontext *ctx);
 
 extern GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 					   GLuint primitive,
 					   GLuint min_nr );
 
-extern void r200EmitAOS( r200ContextPtr rmesa,
-			   struct r200_dma_region **regions,
-			   GLuint n,
-			   GLuint offset );
-
-extern void r200EmitBlit( r200ContextPtr rmesa,
-			  GLuint color_fmt,
-			  GLuint src_pitch,
-			  GLuint src_offset,
-			  GLuint dst_pitch,
-			  GLuint dst_offset,
-			  GLint srcx, GLint srcy,
-			  GLint dstx, GLint dsty,
-			  GLuint w, GLuint h );
-
-extern void r200EmitWait( r200ContextPtr rmesa, GLuint flags );
-
-extern void r200FlushCmdBuf( r200ContextPtr rmesa, const char * );
-extern int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller );
-
-extern void r200RefillCurrentDmaRegion( r200ContextPtr rmesa );
-
-extern void r200AllocDmaRegion( r200ContextPtr rmesa,
-				  struct r200_dma_region *region,
-				  int bytes, 
-				  int alignment );
-
-extern void r200ReleaseDmaRegion( r200ContextPtr rmesa,
-				    struct r200_dma_region *region,
-				    const char *caller );
-
-extern void r200CopyBuffer( __DRIdrawablePrivate *drawable,
-			    const drm_clip_rect_t      *rect);
-extern void r200PageFlip( __DRIdrawablePrivate *drawable );
-extern void r200Flush( GLcontext *ctx );
-extern void r200Finish( GLcontext *ctx );
-extern void r200WaitForIdleLocked( r200ContextPtr rmesa );
-extern void r200WaitForVBlank( r200ContextPtr rmesa );
+extern void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset);
+
 extern void r200InitIoctlFuncs( struct dd_function_table *functions );
 
 extern void *r200AllocateMemoryMESA( __DRIscreen *screen, GLsizei size, GLfloat readfreq,
@@ -119,8 +85,8 @@ void r200SetUpAtomList( r200ContextPtr rmesa );
  */
 #define R200_NEWPRIM( rmesa )			\
 do {						\
-   if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );	\
+   if ( rmesa->radeon.dma.flush )			\
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
 } while (0)
 
 /* Can accomodate several state changes and primitive changes without
@@ -130,22 +96,32 @@ do {						\
 do {								\
    R200_NEWPRIM( rmesa );					\
    rmesa->hw.ATOM.dirty = GL_TRUE;				\
-   rmesa->hw.is_dirty = GL_TRUE;				\
+   rmesa->radeon.hw.is_dirty = GL_TRUE;				\
 } while (0)
 
+#define R200_SET_STATE( rmesa, ATOM, index, newvalue ) 	\
+  do {	\
+    uint32_t __index = (index); \
+    uint32_t __dword = (newvalue); \
+    if (__dword != (rmesa)->hw.ATOM.cmd[__index]) { \
+      R200_STATECHANGE( (rmesa), ATOM ); \
+      (rmesa)->hw.ATOM.cmd[__index] = __dword; \
+    } \
+  } while(0)
+
 #define R200_DB_STATE( ATOM )			        \
    memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd,	\
 	   rmesa->hw.ATOM.cmd_size * 4)
 
 static INLINE int R200_DB_STATECHANGE( 
    r200ContextPtr rmesa,
-   struct r200_state_atom *atom )
+   struct radeon_state_atom *atom )
 {
    if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
-      int *tmp;
+      GLuint *tmp;
       R200_NEWPRIM( rmesa );
       atom->dirty = GL_TRUE;
-      rmesa->hw.is_dirty = GL_TRUE;
+      rmesa->radeon.hw.is_dirty = GL_TRUE;
       tmp = atom->cmd; 
       atom->cmd = atom->lastcmd;
       atom->lastcmd = tmp;
@@ -156,54 +132,47 @@ static INLINE int R200_DB_STATECHANGE(
 }
 
 
-/* Fire the buffered vertices no matter what.
- */
-#define R200_FIREVERTICES( rmesa )			\
-do {							\
-   if ( rmesa->store.cmd_used || rmesa->dma.flush ) {	\
-      r200Flush( rmesa->glCtx );			\
-   }							\
-} while (0)
-
 /* Command lengths.  Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ
  * are available, you will also be adding an rmesa->state.max_state_size because
  * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts.
  */
-#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int))
-#define VERT_AOS_BUFSZ	(5 * sizeof(int))
+#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2) + nr*2))
+#define VERT_AOS_BUFSZ	(5)
 #define ELTS_BUFSZ(nr)	(12 + nr * 2)
-#define VBUF_BUFSZ	(3 * sizeof(int))
-
-/* Ensure that a minimum amount of space is available in the command buffer.
- * This is used to ensure atomicity of state updates with the rendering requests
- * that rely on them.
- *
- * An alternative would be to implement a "soft lock" such that when the buffer
- * wraps at an inopportune time, we grab the lock, flush the current buffer,
- * and hang on to the lock until the critical section is finished and we flush
- * the buffer again and unlock.
- */
-static INLINE void r200EnsureCmdBufSpace( r200ContextPtr rmesa, int bytes )
-{
-   if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-   assert( bytes <= R200_CMD_BUF_SZ );
-}
+#define VBUF_BUFSZ	(3)
+#define SCISSOR_BUFSZ	(8)
+#define INDEX_BUFSZ	(8+2)
 
-/* Alloc space in the command buffer
- */
-static INLINE char *r200AllocCmdBuf( r200ContextPtr rmesa,
-					 int bytes, const char *where )
+static inline uint32_t cmdpacket3(int cmd_type)
 {
-   char * head;
+  drm_radeon_cmd_header_t cmd;
 
-   if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-      r200FlushCmdBuf( rmesa, where );
+  cmd.i = 0;
+  cmd.header.cmd_type = cmd_type;
+
+  return (uint32_t)cmd.i;
 
-   head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-   rmesa->store.cmd_used += bytes;
-   assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ );
-   return head;
 }
 
+#define OUT_BATCH_PACKET3(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3));				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP));	      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+
 #endif /* __R200_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_lock.c b/src/mesa/drivers/dri/r200/r200_lock.c
deleted file mode 100644
index 99661a4bfb..0000000000
--- a/src/mesa/drivers/dri/r200/r200_lock.c
+++ /dev/null
@@ -1,116 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
- 
-#include "r200_context.h"
-#include "r200_lock.h"
-#include "r200_tex.h"
-#include "r200_state.h"
-#include "r200_ioctl.h"
-
-#include "drirenderbuffer.h"
-
-
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-static void
-r200UpdatePageFlipping( r200ContextPtr rmesa )
-{
-   rmesa->doPageFlip = rmesa->sarea->pfState;
-   if (rmesa->glCtx->WinSysDrawBuffer) {
-      driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                           rmesa->sarea->pfCurrentPage);
-   }
-}
-
-
-
-/* Update the hardware state.  This is called if another main/context.has
- * grabbed the hardware lock, which includes the X server.  This
- * function also updates the driver's window state after the X server
- * moves, resizes or restacks a window -- the change will be reflected
- * in the drawable position and clip rects.  Since the X server grabs
- * the hardware lock when it changes the window state, this routine will
- * automatically be called after such a change.
- */
-void r200GetLock( r200ContextPtr rmesa, GLuint flags )
-{
-   __DRIdrawablePrivate *drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *readable = rmesa->dri.readable;
-   __DRIscreenPrivate *sPriv = rmesa->dri.screen;
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-   int i;
-
-   drmGetLock( rmesa->dri.fd, rmesa->dri.hwContext, flags );
-
-   /* The window might have moved, so we might need to get new clip
-    * rects.
-    *
-    * NOTE: This releases and regrabs the hw lock to allow the X server
-    * to respond to the DRI protocol request for new drawable info.
-    * Since the hardware state depends on having the latest drawable
-    * clip rects, all state checking must be done _after_ this call.
-    */
-   DRI_VALIDATE_DRAWABLE_INFO( sPriv, drawable );
-   if (drawable != readable) {
-      DRI_VALIDATE_DRAWABLE_INFO( sPriv, readable );
-   }
-
-   if ( rmesa->lastStamp != drawable->lastStamp ) {
-      r200UpdatePageFlipping( rmesa );
-      r200SetCliprects( rmesa );
-      r200UpdateViewportOffset( rmesa->glCtx );
-      driUpdateFramebufferSize(rmesa->glCtx, drawable);
-   }
-
-   R200_STATECHANGE( rmesa, ctx );
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
-   else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE;
-
-   if ( sarea->ctx_owner != rmesa->dri.hwContext ) {
-      sarea->ctx_owner = rmesa->dri.hwContext;
-   }
-
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      DRI_AGE_TEXTURES( rmesa->texture_heaps[ i ] );
-   }
-
-   rmesa->lost_context = GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_lock.h b/src/mesa/drivers/dri/r200/r200_lock.h
deleted file mode 100644
index 4ff98907fb..0000000000
--- a/src/mesa/drivers/dri/r200/r200_lock.h
+++ /dev/null
@@ -1,106 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_LOCK_H__
-#define __R200_LOCK_H__
-
-extern void r200GetLock( r200ContextPtr rmesa, GLuint flags );
-
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if ( prevLockFile ) {						\
-	 fprintf( stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__ );	\
-	 exit( 1 );							\
-      }									\
-   } while (0)
-
-#else
-
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
-
-#endif
-
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
-
-
-/* Lock the hardware and validate our state.
- */
-#define LOCK_HARDWARE( rmesa )					\
-   do {								\
-      char __ret = 0;						\
-      DEBUG_CHECK_LOCK();					\
-      DRM_CAS( rmesa->dri.hwLock, rmesa->dri.hwContext,		\
-	       (DRM_LOCK_HELD | rmesa->dri.hwContext), __ret );	\
-      if ( __ret )						\
-	 r200GetLock( rmesa, 0 );				\
-      DEBUG_LOCK();						\
-   } while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-   do {									\
-      DRM_UNLOCK( rmesa->dri.fd,					\
-		  rmesa->dri.hwLock,					\
-		  rmesa->dri.hwContext );				\
-      DEBUG_RESET();							\
-   } while (0)
-
-#endif /* __R200_LOCK_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_maos.h b/src/mesa/drivers/dri/r200/r200_maos.h
index d3ed06d402..16a70475e1 100644
--- a/src/mesa/drivers/dri/r200/r200_maos.h
+++ b/src/mesa/drivers/dri/r200/r200_maos.h
@@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 
 extern void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev );
-extern void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs );
 
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
index 8512b9af47..383a0c4b0d 100644
--- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c
+++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
@@ -50,110 +50,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_maos.h"
 #include "r200_tcl.h"
 
-
-#if 0
-/* Usage:
- *   - from r200_tcl_render
- *   - call r200EmitArrays to ensure uptodate arrays in dma
- *   - emit primitives (new type?) which reference the data
- *       -- need to use elts for lineloop, quads, quadstrip/flat
- *       -- other primitives are all well-formed (need tristrip-1,fake-poly)
- *
- */
-static void emit_ubyte_rgba3( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   r200_color_t *out = (r200_color_t *)(rvb->start + rvb->address);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p\n",
-	      __FUNCTION__, count, stride, (void *)out);
-
-   for (i = 0; i < count; i++) {
-      out->red   = *data;
-      out->green = *(data+1);
-      out->blue  = *(data+2);
-      out->alpha = 0xFF;
-      out++;
-      data += stride;
-   }
-}
-
-static void emit_ubyte_rgba4( GLcontext *ctx,
-			      struct r200_dma_region *rvb,
-			      char *data,
-			      int stride,
-			      int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4) {
-      for (i = 0; i < count; i++)
-	 ((int *)out)[i] = LE32_TO_CPU(((int *)data)[i]);
-   } else {
-      for (i = 0; i < count; i++) {
-	 *(int *)out++ = LE32_TO_CPU(*(int *)data);
-	 data += stride;
-      }
-   }
-}
-
-
-static void emit_ubyte_rgba( GLcontext *ctx,
-			     struct r200_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, 4 * count, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 3:
-      emit_ubyte_rgba3( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_ubyte_rgba4( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-}
-#endif
-
-
 #if defined(USE_X86_ASM)
 #define COPY_DWORDS( dst, src, nr )					\
 do {									\
@@ -174,204 +70,34 @@ do {						\
 } while (0)
 #endif
 
-
-static void emit_vecfog( GLcontext *ctx,
-			 struct r200_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos,
+			     GLvoid *data, int stride, int count)
 {
-   int i;
-   GLfloat *out;
-
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
-
-   /* Emit the data
-    */
-   out = (GLfloat *)(rvb->address + rvb->start);
-   for (i = 0; i < count; i++) {
-      out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
-      out++;
-      data += stride;
-   }
-
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	uint32_t *out;
+	int i;
+	int size = 1;
+
+	if (stride == 0) {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		count = 1;
+		aos->stride = 0;
+	} else {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		aos->stride = size;
+	}
+
+	aos->components = size;
+	aos->count = count;
+
+	out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
+	for (i = 0; i < count; i++) {
+	  out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
+	  out++;
+	  data += stride;
+	}
 }
 
-
-static void emit_vec4( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4)
-      COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out++;
-	 data += stride;
-      }
-}
-
-
-static void emit_vec8( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 8)
-      COPY_DWORDS( out, data, count*2 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out += 2;
-	 data += stride;
-      }
-}
-
-static void emit_vec12( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-	      __FUNCTION__, count, stride, (void *)out, (void *)data);
-
-   if (stride == 12)
-      COPY_DWORDS( out, data, count*3 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out += 3;
-	 data += stride;
-      }
-}
-
-static void emit_vec16( GLcontext *ctx,
-			struct r200_dma_region *rvb,
-			char *data,
-			int stride,
-			int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 16)
-      COPY_DWORDS( out, data, count*4 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out[3] = *(int *)(data+12);
-	 out += 4;
-	 data += stride;
-      }
-}
-
-
-static void emit_vector( GLcontext *ctx,
-			 struct r200_dma_region *rvb,
-			 char *data,
-			 int size,
-			 int stride,
-			 int count )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d size %d stride %d\n",
-	      __FUNCTION__, count, size, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, size * 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = size;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, size * count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = size;
-      rvb->aos_size = size;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 1:
-      emit_vec4( ctx, rvb, data, stride, count );
-      break;
-   case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
-      break;
-   case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_vec16( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-
-}
-
-
-
 /* Emit any changed arrays to new GART memory, re-emit a packet to
  * update the arrays.  
  */
@@ -379,12 +105,12 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 {
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
    struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
-   struct r200_dma_region **component = rmesa->tcl.aos_components;
    GLuint nr = 0;
    GLuint vfmt0 = 0, vfmt1 = 0;
    GLuint count = VB->Count;
    GLuint i, emitsize;
 
+   //   fprintf(stderr,"emit arrays\n");
    for ( i = 0; i < 15; i++ ) {
       GLubyte attrib = vimap_rev[i];
       if (attrib != 255) {
@@ -416,20 +142,20 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 	 case 3:
 	    /* special handling to fix up fog. Will get us into trouble with vbos...*/
 	    assert(attrib == VERT_ATTRIB_FOG);
-	    if (!rmesa->tcl.vertex_data[i].buf) {
+	    if (!rmesa->radeon.tcl.aos[i].bo) {
 	       if (ctx->VertexProgram._Enabled)
-		  emit_vector( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 1,
-			 VB->AttribPtr[attrib]->stride,
-			 count);
+		  rcommon_emit_vector( ctx,
+				       &(rmesa->radeon.tcl.aos[nr]),
+				       (char *)VB->AttribPtr[attrib]->data,
+				       1,
+				       VB->AttribPtr[attrib]->stride,
+				       count);
 	       else
-		  emit_vecfog( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 VB->AttribPtr[attrib]->stride,
-			 count);
+		 r200_emit_vecfog( ctx,
+				   &(rmesa->radeon.tcl.aos[nr]),
+				   (char *)VB->AttribPtr[attrib]->data,
+				   VB->AttribPtr[attrib]->stride,
+				   count);
 	    }
 	    vfmt0 |= R200_VTX_DISCRETE_FOG;
 	    goto after_emit;
@@ -473,17 +199,17 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 	 default:
 	    assert(0);
 	 }
-	 if (!rmesa->tcl.vertex_data[i].buf) {
-	    emit_vector( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 emitsize,
-			 VB->AttribPtr[attrib]->stride,
-			 count );
+	 if (!rmesa->radeon.tcl.aos[nr].bo) {
+	   rcommon_emit_vector( ctx,
+				&(rmesa->radeon.tcl.aos[nr]),
+				(char *)VB->AttribPtr[attrib]->data,
+				emitsize,
+				VB->AttribPtr[attrib]->stride,
+				count );
 	 }
 after_emit:
 	 assert(nr < 12);
-	 component[nr++] = &rmesa->tcl.vertex_data[i];
+	 nr++;
       }
    }
 
@@ -494,19 +220,6 @@ after_emit:
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
    }
 
-   rmesa->tcl.nr_aos_components = nr;
+   rmesa->radeon.tcl.aos_count = nr;
 }
 
-
-void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   /* only do it for changed inputs ? */
-   int i;
-   for (i = 0; i < 15; i++) {
-      if (newinputs & (1 << i))
-	 r200ReleaseDmaRegion( rmesa,
-	    &rmesa->tcl.vertex_data[i], __FUNCTION__ );
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_pixel.c b/src/mesa/drivers/dri/r200/r200_pixel.c
index 2797cbb3dc..95773871e0 100644
--- a/src/mesa/drivers/dri/r200/r200_pixel.c
+++ b/src/mesa/drivers/dri/r200/r200_pixel.c
@@ -51,29 +51,29 @@ check_color( const GLcontext *ctx, GLenum type, GLenum format,
 	     const void *pixels, GLint sz, GLint pitch )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
 
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    if (	(pitch & 63) ||
 	ctx->_ImageTransferState ||
 	packing->SwapBytes ||
 	packing->LsbFirst) {
-      if (R200_DEBUG & DEBUG_PIXEL)
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "%s: failed 1\n", __FUNCTION__);
       return GL_FALSE;
    }
 
-   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV && 
-	cpp == 4 && 
+   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV &&
+	cpp == 4 &&
 	format == GL_BGRA ) {
-      if (R200_DEBUG & DEBUG_PIXEL)
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "%s: passed 2\n", __FUNCTION__);
       return GL_TRUE;
    }
 
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s: failed\n", __FUNCTION__);
 
    return GL_FALSE;
@@ -83,7 +83,7 @@ static GLboolean
 check_color_per_fragment_ops( const GLcontext *ctx )
 {
    int result;
-   result = (!(     ctx->Color.AlphaEnabled || 
+   result = (!(     ctx->Color.AlphaEnabled ||
 		    ctx->Depth.Test ||
 		    ctx->Fog.Enabled ||
 		    ctx->Scissor.Enabled ||
@@ -96,12 +96,12 @@ check_color_per_fragment_ops( const GLcontext *ctx )
 		    ctx->Texture._EnabledUnits
            ) &&
 	   ctx->Current.RasterPosValid);
-   
+
    return result;
 }
 
 
-
+#if 0
 static GLboolean
 clip_pixelrect( const GLcontext *ctx,
 		const GLframebuffer *buffer,
@@ -137,11 +137,12 @@ clip_pixelrect( const GLcontext *ctx,
    if (*height <= 0)
       return GL_FALSE;
 
-   *size = ((*y + *height - 1) * rmesa->r200Screen->frontPitch +
-	    (*x + *width - 1) * rmesa->r200Screen->cpp);
+   *size = ((*y + *height - 1) * rmesa->radeon.radeonScreen->frontPitch +
+	    (*x + *width - 1) * rmesa->radeon.radeonScreen->cpp);
 
    return GL_TRUE;
 }
+#endif
 
 static GLboolean
 r200TryReadPixels( GLcontext *ctx,
@@ -150,29 +151,30 @@ r200TryReadPixels( GLcontext *ctx,
 		  const struct gl_pixelstore_attrib *pack,
 		  GLvoid *pixels )
 {
+   return GL_FALSE;
+#if 0
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLint pitch = pack->RowLength ? pack->RowLength : width;
    GLint blit_format;
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
    GLint size = width * height * cpp;
 
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    /* Only accelerate reading to GART buffers.
     */
-   if ( !r200IsGartMemory(rmesa, pixels, 
-			 pitch * height * rmesa->r200Screen->cpp ) ) {
-      if (R200_DEBUG & DEBUG_PIXEL)
+   if ( !r200IsGartMemory(rmesa, pixels,
+			 pitch * height * rmesa->radeon.radeonScreen->cpp ) ) {
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "%s: dest not GART\n", __FUNCTION__);
-      return GL_FALSE;
    }
 
    /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
     * blitter:
     */
    if (!pack->Invert) {
-      if (R200_DEBUG & DEBUG_PIXEL)
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "%s: MESA_PACK_INVERT not set\n", __FUNCTION__);
       return GL_FALSE;
    }
@@ -180,7 +182,7 @@ r200TryReadPixels( GLcontext *ctx,
    if (!check_color(ctx, type, format, pack, pixels, size, pitch))
       return GL_FALSE;
 
-   switch ( rmesa->r200Screen->cpp ) {
+   switch ( rmesa->radeon.radeonScreen->cpp ) {
    case 4:
       blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
       break;
@@ -197,40 +199,40 @@ r200TryReadPixels( GLcontext *ctx,
     * a full command buffer expects to be called unlocked.  As a
     * workaround, immediately flush the buffer on aquiring the lock.
     */
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    if (rmesa->store.cmd_used)
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    if (!clip_pixelrect(ctx, ctx->ReadBuffer, &x, &y, &width, &height,
 		       &size)) {
-      UNLOCK_HARDWARE( rmesa );
-      if (R200_DEBUG & DEBUG_PIXEL)
+      UNLOCK_HARDWARE( &rmesa->radeon );
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "%s totally clipped -- nothing to do\n",
 		 __FUNCTION__);
       return GL_TRUE;
    }
 
    {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+      __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
       driRenderbuffer *drb = (driRenderbuffer *) ctx->ReadBuffer->_ColorReadBuffer;
       int nbox = dPriv->numClipRects;
       int src_offset = drb->offset
-		     + rmesa->r200Screen->fbLocation;
+		     + rmesa->radeon.radeonScreen->fbLocation;
       int src_pitch = drb->pitch * drb->cpp;
       int dst_offset = r200GartOffsetFromVirtual( rmesa, pixels );
-      int dst_pitch = pitch * rmesa->r200Screen->cpp;
+      int dst_pitch = pitch * rmesa->radeon.radeonScreen->cpp;
       drm_clip_rect_t *box = dPriv->pClipRects;
       int i;
 
-      r200EmitWait( rmesa, RADEON_WAIT_3D ); 
+      r200EmitWait( rmesa, RADEON_WAIT_3D );
 
       y = dPriv->h - y - height;
       x += dPriv->x;
       y += dPriv->y;
 
 
-      if (R200_DEBUG & DEBUG_PIXEL)
+      if (R200_DEBUG & RADEON_PIXEL)
 	 fprintf(stderr, "readpixel blit src_pitch %d dst_pitch %d\n",
 		 src_pitch, dst_pitch);
 
@@ -240,7 +242,7 @@ r200TryReadPixels( GLcontext *ctx,
 	 GLint by = box[i].y1;
 	 GLint bw = box[i].x2 - bx;
 	 GLint bh = box[i].y2 - by;
-	 
+
 	 if (bx < x) bw -= x - bx, bx = x;
 	 if (by < y) bh -= y - by, by = y;
 	 if (bx + bw > x + width) bw = x + width - bx;
@@ -257,12 +259,12 @@ r200TryReadPixels( GLcontext *ctx,
 		       bw, bh );
       }
 
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
    }
-   UNLOCK_HARDWARE( rmesa );
-
-   r200Finish( ctx ); /* required by GL */
+   UNLOCK_HARDWARE( &rmesa->radeon );
 
+   radeonFinish( ctx ); /* required by GL */
+#endif
    return GL_TRUE;
 }
 
@@ -273,12 +275,12 @@ r200ReadPixels( GLcontext *ctx,
 		 const struct gl_pixelstore_attrib *pack,
 		 GLvoid *pixels )
 {
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-   if (!r200TryReadPixels( ctx, x, y, width, height, format, type, pack, 
+   if (!r200TryReadPixels( ctx, x, y, width, height, format, type, pack,
 			   pixels))
-      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, 
+      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack,
 			  pixels);
 }
 
@@ -291,8 +293,12 @@ static void do_draw_pix( GLcontext *ctx,
 			 const void *pixels,
 			 GLuint planemask)
 {
+   if (R200_DEBUG & RADEON_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+#if 0
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    drm_clip_rect_t *box = dPriv->pClipRects;
    struct gl_renderbuffer *rb = ctx->ReadBuffer->_ColorDrawBuffers[0];
    driRenderbuffer *drb = (driRenderbuffer *) rb;
@@ -301,12 +307,9 @@ static void do_draw_pix( GLcontext *ctx,
    int blit_format;
    int size;
    int src_offset = r200GartOffsetFromVirtual( rmesa, pixels );
-   int src_pitch = pitch * rmesa->r200Screen->cpp;
+   int src_pitch = pitch * rmesa->radeon.radeonScreen->cpp;
 
-   if (R200_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   switch ( rmesa->r200Screen->cpp ) {
+   switch ( rmesa->radeon.radeonScreen->cpp ) {
    case 2:
       blit_format = R200_CP_COLOR_FORMAT_RGB565;
       break;
@@ -318,17 +321,17 @@ static void do_draw_pix( GLcontext *ctx,
    }
 
 
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    if (rmesa->store.cmd_used)
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    y -= height;			/* cope with pixel zoom */
-   
+
    if (!clip_pixelrect(ctx, ctx->DrawBuffer,
 		       &x, &y, &width, &height,
 		       &size)) {
-      UNLOCK_HARDWARE( rmesa );
+      UNLOCK_HARDWARE( &rmesa->radeon );
       return;
    }
 
@@ -357,15 +360,16 @@ static void do_draw_pix( GLcontext *ctx,
 		    blit_format,
 		    src_pitch, src_offset,
 		    drb->pitch * drb->cpp,
-		    drb->offset + rmesa->r200Screen->fbLocation,
+		    drb->offset + rmesa->radeon.radeonScreen->fbLocation,
 		    bx - x, by - y,
 		    bx, by,
 		    bw, bh );
    }
 
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-   r200WaitForIdleLocked( rmesa ); /* required by GL */
-   UNLOCK_HARDWARE( rmesa );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
+   radeonWaitForIdleLocked( &rmesa->radeon ); /* required by GL */
+   UNLOCK_HARDWARE( &rmesa->radeon );
+#endif
 }
 
 
@@ -381,10 +385,10 @@ r200TryDrawPixels( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLint pitch = unpack->RowLength ? unpack->RowLength : width;
    GLuint planemask;
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
    GLint size = height * pitch * cpp;
 
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    /* check that we're drawing to exactly one color buffer */
@@ -395,7 +399,7 @@ r200TryDrawPixels( GLcontext *ctx,
    case GL_RGB:
    case GL_RGBA:
    case GL_BGRA:
-      planemask = r200PackColor(cpp,
+      planemask = radeonPackColor(cpp,
 				ctx->Color.ColorMask[RCOMP],
 				ctx->Color.ColorMask[GCOMP],
 				ctx->Color.ColorMask[BCOMP],
@@ -407,10 +411,10 @@ r200TryDrawPixels( GLcontext *ctx,
       if (planemask != ~0)
 	 return GL_FALSE;	/* fix me -- should be possible */
 
-      /* Can't do conversions on GART reads/draws. 
+      /* Can't do conversions on GART reads/draws.
        */
       if ( !r200IsGartMemory( rmesa, pixels, size ) ) {
-	 if (R200_DEBUG & DEBUG_PIXEL)
+	 if (R200_DEBUG & RADEON_PIXEL)
 	    fprintf(stderr, "%s: not GART memory\n", __FUNCTION__);
 	 return GL_FALSE;
       }
@@ -431,7 +435,7 @@ r200TryDrawPixels( GLcontext *ctx,
       return GL_FALSE;
    }
 
-   if ( r200IsGartMemory(rmesa, pixels, size) )
+   if (0)// r200IsGartMemory(rmesa, pixels, size) )
    {
       do_draw_pix( ctx, x, y, width, height, pitch, pixels, planemask );
       return GL_TRUE;
@@ -453,7 +457,7 @@ r200DrawPixels( GLcontext *ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLvoid *pixels )
 {
-   if (R200_DEBUG & DEBUG_PIXEL)
+   if (R200_DEBUG & RADEON_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    if (!r200TryDrawPixels( ctx, x, y, width, height, format, type,
@@ -471,7 +475,7 @@ r200Bitmap( GLcontext *ctx, GLint px, GLint py,
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   if (rmesa->Fallback)
+   if (rmesa->radeon.Fallback)
       _swrast_Bitmap( ctx, px, py, width, height, unpack, bitmap );
    else
       r200PointsBitmap( ctx, px, py, width, height, unpack, bitmap );
@@ -482,9 +486,9 @@ r200Bitmap( GLcontext *ctx, GLint px, GLint py,
 void r200InitPixelFuncs( GLcontext *ctx )
 {
    if (!getenv("R200_NO_BLITS")) {
-      ctx->Driver.ReadPixels = r200ReadPixels;  
-      ctx->Driver.DrawPixels = r200DrawPixels; 
-      if (getenv("R200_HW_BITMAP")) 
+      ctx->Driver.ReadPixels = r200ReadPixels;
+      ctx->Driver.DrawPixels = r200DrawPixels;
+      if (getenv("R200_HW_BITMAP"))
 	 ctx->Driver.Bitmap = r200Bitmap;
    }
 }
diff --git a/src/mesa/drivers/dri/r200/r200_reg.h b/src/mesa/drivers/dri/r200/r200_reg.h
index 5ce287f7a5..526a624b69 100644
--- a/src/mesa/drivers/dri/r200/r200_reg.h
+++ b/src/mesa/drivers/dri/r200/r200_reg.h
@@ -463,8 +463,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define     R200_VSC_UPDATE_USER_COLOR_1_ENABLE    0x00020000
 /* gap */
 #define R200_SE_TCL_VECTOR_INDX_REG                0x2200
+#       define RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT  16
+#       define RADEON_VEC_INDX_DWORD_COUNT_SHIFT     28
 #define R200_SE_TCL_VECTOR_DATA_REG                0x2204
 #define R200_SE_TCL_SCALAR_INDX_REG                0x2208
+#       define RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT  16
 #define R200_SE_TCL_SCALAR_DATA_REG                0x220c
 /* gap */
 #define R200_SE_TCL_MATRIX_SEL_0                   0x2230
@@ -949,6 +952,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define     R200_LOD_BIAS_MASK                        (0xfff80000)
 #define     R200_LOD_BIAS_SHIFT                       19
 #define R200_PP_TXSIZE_0                  0x2c0c /* NPOT only */
+#define R200_PP_TX_WIDTHMASK_SHIFT 0
+#define R200_PP_TX_HEIGHTMASK_SHIFT 16
+
 #define R200_PP_TXPITCH_0                 0x2c10 /* NPOT only */
 #define R200_PP_BORDER_COLOR_0            0x2c14
 #define R200_PP_CUBIC_FACES_0             0x2c18
diff --git a/src/mesa/drivers/dri/r200/r200_sanity.c b/src/mesa/drivers/dri/r200/r200_sanity.c
index 36530c224e..1241a926ba 100644
--- a/src/mesa/drivers/dri/r200/r200_sanity.c
+++ b/src/mesa/drivers/dri/r200/r200_sanity.c
@@ -48,11 +48,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define MORE_VERBOSE 1
 
 #if MORE_VERBOSE
-#define VERBOSE (R200_DEBUG & DEBUG_VERBOSE)
+#define VERBOSE (R200_DEBUG & RADEON_VERBOSE)
 #define NORMAL  (1)
 #else
 #define VERBOSE 0
-#define NORMAL  (R200_DEBUG & DEBUG_VERBOSE)
+#define NORMAL  (R200_DEBUG & RADEON_VERBOSE)
 #endif
 
 
diff --git a/src/mesa/drivers/dri/r200/r200_span.c b/src/mesa/drivers/dri/r200/r200_span.c
deleted file mode 100644
index 9783678028..0000000000
--- a/src/mesa/drivers/dri/r200/r200_span.c
+++ /dev/null
@@ -1,307 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/colormac.h"
-#include "swrast/swrast.h"
-
-#include "r200_context.h"
-#include "r200_ioctl.h"
-#include "r200_state.h"
-#include "r200_span.h"
-#include "r200_tex.h"
-
-#define DBG 0
-
-/*
- * Note that all information needed to access pixels in a renderbuffer
- * should be obtained through the gl_renderbuffer parameter, not per-context
- * information.
- */
-#define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
-
-#define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
-
-#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
-
-#define Y_FLIP(Y) (bottom - (Y))
-
-#define HW_LOCK() 
-
-#define HW_UNLOCK()							
-
-
-
-/* ================================================================
- * Color buffer
- */
-
-/* 16 bit, RGB565 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    r200##x##_RGB565
-#define TAG2(x,y) r200##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
-#include "spantmp2.h"
-
-/* 32 bit, ARGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    r200##x##_ARGB8888
-#define TAG2(x,y) r200##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
-#include "spantmp2.h"
-
-
-/* ================================================================
- * Depth buffer
- */
-
-/* The Radeon family has depth tiling on all the time, so we have to convert
- * the x,y coordinates into the memory bus address (mba) in the same
- * manner as the engine.  In each case, the linear block address (ba)
- * is calculated, and then wired with x and y to produce the final
- * memory address.
- * The chip will do address translation on its own if the surface registers
- * are set up correctly. It is not quite enough to get it working with hyperz too...
- */
-
-/* extract bit 'b' of x, result is zero or one */
-#define BIT(x,b) ((x & (1<<b))>>b)
-
-static GLuint
-r200_mba_z32( driRenderbuffer *drb, GLint x, GLint y )
-{
-   GLuint pitch = drb->pitch;
-   if (drb->depthHasSurface) {
-      return 4 * (x + y * pitch);
-   }
-   else {
-      GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x7FF) >> 5);
-      GLuint a = 
-         (BIT(x,0) << 2) |
-         (BIT(y,0) << 3) |
-         (BIT(x,1) << 4) |
-         (BIT(y,1) << 5) |
-         (BIT(x,3) << 6) |
-         (BIT(x,4) << 7) |
-         (BIT(x,2) << 8) |
-         (BIT(y,2) << 9) |
-         (BIT(y,3) << 10) |
-         (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-         ((b >> 1) << 12);
-      return a;
-   }
-}
-
-static GLuint
-r200_mba_z16( driRenderbuffer *drb, GLint x, GLint y )
-{
-   GLuint pitch = drb->pitch;
-   if (drb->depthHasSurface) {
-      return 2 * (x + y * pitch);
-   }
-   else {
-      GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x7FF) >> 6);
-      GLuint a = 
-         (BIT(x,0) << 1) |
-         (BIT(y,0) << 2) |
-         (BIT(x,1) << 3) |
-         (BIT(y,1) << 4) |
-         (BIT(x,2) << 5) |
-         (BIT(x,4) << 6) |
-         (BIT(x,5) << 7) |
-         (BIT(x,3) << 8) |
-         (BIT(y,2) << 9) |
-         (BIT(y,3) << 10) |
-         (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-         ((b >> 1) << 12);
-      return a;
-   }
-}
-
-
-/* 16-bit depth buffer functions
- */
-#define VALUE_TYPE GLushort
-
-#define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo )) = d;
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo ));
-
-#define TAG(x) r200##x##_z16
-#include "depthtmp.h"
-
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#define VALUE_TYPE GLuint
-
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + r200_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
-
-#define TAG(x) r200##x##_z24_s8
-#include "depthtmp.h"
-
-
-/* ================================================================
- * Stencil buffer
- */
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x00ffffff;							\
-   tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   d = tmp >> 24;							\
-} while (0)
-
-#define TAG(x) r200##x##_z24_s8
-#include "stenciltmp.h"
-
-
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
-
-static void r200SpanRenderStart( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-   r200WaitForIdleLocked( rmesa );
-
-   /* Read & rewrite the first pixel in the frame buffer.  This should
-    * be a noop, right?  In fact without this conform fails as reading
-    * from the framebuffer sometimes produces old results -- the
-    * on-card read cache gets mixed up and doesn't notice that the
-    * framebuffer has been updated.
-    *
-    * In the worst case this is buggy too as p might get the wrong
-    * value first time, so really need a hidden pixel somewhere for this.
-    */
-   {
-      int p;
-      driRenderbuffer *drb =
-	 (driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0];
-      volatile int *buf =
-	 (volatile int *)(rmesa->dri.screen->pFB + drb->offset);
-      p = *buf;
-      *buf = p;
-   }
-}
-
-static void r200SpanRenderFinish( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   _swrast_flush( ctx );
-   UNLOCK_HARDWARE( rmesa );
-}
-
-void r200InitSpanFuncs( GLcontext *ctx )
-{
-   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
-   swdd->SpanRenderStart          = r200SpanRenderStart;
-   swdd->SpanRenderFinish         = r200SpanRenderFinish; 
-}
-
-
-
-/**
- * Plug in the Get/Put routines for the given driRenderbuffer.
- */
-void
-radeonSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis)
-{
-   if (drb->Base.InternalFormat == GL_RGBA) {
-      if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
-         r200InitPointers_RGB565(&drb->Base);
-      }
-      else {
-         r200InitPointers_ARGB8888(&drb->Base);
-      }
-   }
-   else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-      r200InitDepthPointers_z16(&drb->Base);
-   }
-   else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-      r200InitDepthPointers_z24_s8(&drb->Base);
-   }
-   else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-      r200InitStencilPointers_z24_s8(&drb->Base);
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index 2fcc87c0f5..76852e315c 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -47,6 +47,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
@@ -77,7 +79,7 @@ static void r200AlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 
    switch ( func ) {
    case GL_NEVER:
-      pp_misc |= R200_ALPHA_TEST_FAIL; 
+      pp_misc |= R200_ALPHA_TEST_FAIL;
       break;
    case GL_LESS:
       pp_misc |= R200_ALPHA_TEST_LESS;
@@ -114,8 +116,8 @@ static void r200BlendColor( GLcontext *ctx, const GLfloat cf[4] )
    CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]);
    CLAMPED_FLOAT_TO_UBYTE(color[2], cf[2]);
    CLAMPED_FLOAT_TO_UBYTE(color[3], cf[3]);
-   if (rmesa->r200Screen->drmSupportsBlendColor)
-      rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = r200PackColor( 4, color[0], color[1], color[2], color[3] );
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+      rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = radeonPackColor( 4, color[0], color[1], color[2], color[3] );
 }
 
 /**
@@ -213,7 +215,7 @@ static void r200_set_blend_state( GLcontext * ctx )
 
    R200_STATECHANGE( rmesa, ctx );
 
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       if (ctx->Color.ColorLogicOpEnabled) {
          rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =  cntl | R200_ROP_ENABLE;
          rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = eqn | func;
@@ -278,7 +280,7 @@ static void r200_set_blend_state( GLcontext * ctx )
       return;
    }
 
-   if (!rmesa->r200Screen->drmSupportsBlendColor) {
+   if (!rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func;
       return;
    }
@@ -383,10 +385,10 @@ static void r200ClearDepth( GLcontext *ctx, GLclampd d )
 
    switch ( format ) {
    case R200_DEPTH_FORMAT_16BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x0000ffff;
+      rmesa->radeon.state.depth.clear = d * 0x0000ffff;
       break;
    case R200_DEPTH_FORMAT_24BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x00ffffff;
+      rmesa->radeon.state.depth.clear = d * 0x00ffffff;
       break;
    }
 }
@@ -477,10 +479,10 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 	 }
       }
       break;
-   case GL_FOG_COLOR: 
+   case GL_FOG_COLOR:
       R200_STATECHANGE( rmesa, ctx );
       UNCLAMPED_FLOAT_TO_RGB_CHAN( col, ctx->Fog.Color );
-      i = r200PackColor( 4, col[0], col[1], col[2], 0 );
+      i = radeonPackColor( 4, col[0], col[1], col[2], 0 );
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_COLOR_MASK;
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= i;
       break;
@@ -505,7 +507,7 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 
       if (out_0 != rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0]) {
 	 R200_STATECHANGE( rmesa, vtx );
-	 rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] = out_0;	 
+	 rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] = out_0;
       }
 
       break;
@@ -521,102 +523,6 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
    }
 }
 
-
-/* =============================================================
- * Scissoring
- */
-
-
-static GLboolean intersect_rect( drm_clip_rect_t *out,
-				 drm_clip_rect_t *a,
-				 drm_clip_rect_t *b )
-{
-   *out = *a;
-   if ( b->x1 > out->x1 ) out->x1 = b->x1;
-   if ( b->y1 > out->y1 ) out->y1 = b->y1;
-   if ( b->x2 < out->x2 ) out->x2 = b->x2;
-   if ( b->y2 < out->y2 ) out->y2 = b->y2;
-   if ( out->x1 >= out->x2 ) return GL_FALSE;
-   if ( out->y1 >= out->y2 ) return GL_FALSE;
-   return GL_TRUE;
-}
-
-
-void r200RecalcScissorRects( r200ContextPtr rmesa )
-{
-   drm_clip_rect_t *out;
-   int i;
-
-   /* Grow cliprect store?
-    */
-   if (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-      while (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-	 rmesa->state.scissor.numAllocedClipRects += 1;	/* zero case */
-	 rmesa->state.scissor.numAllocedClipRects *= 2;
-      }
-
-      if (rmesa->state.scissor.pClipRects)
-	 FREE(rmesa->state.scissor.pClipRects);
-
-      rmesa->state.scissor.pClipRects = 
-	 MALLOC( rmesa->state.scissor.numAllocedClipRects * 
-		 sizeof(drm_clip_rect_t) );
-
-      if ( rmesa->state.scissor.pClipRects == NULL ) {
-	 rmesa->state.scissor.numAllocedClipRects = 0;
-	 return;
-      }
-   }
-   
-   out = rmesa->state.scissor.pClipRects;
-   rmesa->state.scissor.numClipRects = 0;
-
-   for ( i = 0 ; i < rmesa->numClipRects ;  i++ ) {
-      if ( intersect_rect( out, 
-			   &rmesa->pClipRects[i], 
-			   &rmesa->state.scissor.rect ) ) {
-	 rmesa->state.scissor.numClipRects++;
-	 out++;
-      }
-   }
-}
-
-
-static void r200UpdateScissor( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( rmesa->dri.drawable ) {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-
-      int x = ctx->Scissor.X;
-      int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
-      int w = ctx->Scissor.X + ctx->Scissor.Width - 1;
-      int h = dPriv->h - ctx->Scissor.Y - 1;
-
-      rmesa->state.scissor.rect.x1 = x + dPriv->x;
-      rmesa->state.scissor.rect.y1 = y + dPriv->y;
-      rmesa->state.scissor.rect.x2 = w + dPriv->x + 1;
-      rmesa->state.scissor.rect.y2 = h + dPriv->y + 1;
-
-      r200RecalcScissorRects( rmesa );
-   }
-}
-
-
-static void r200Scissor( GLcontext *ctx,
-			   GLint x, GLint y, GLsizei w, GLsizei h )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( ctx->Scissor.Enabled ) {
-      R200_FIREVERTICES( rmesa );	/* don't pipeline cliprect changes */
-      r200UpdateScissor( ctx );
-   }
-
-}
-
-
 /* =============================================================
  * Culling
  */
@@ -668,6 +574,10 @@ static void r200FrontFace( GLcontext *ctx, GLenum mode )
    R200_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= R200_FFACE_CULL_CW;
@@ -790,7 +700,7 @@ static void r200LineStipple( GLcontext *ctx, GLint factor, GLushort pattern )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
    R200_STATECHANGE( rmesa, lin );
-   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] = 
+   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] =
       ((((GLuint)factor & 0xff) << 16) | ((GLuint)pattern));
 }
 
@@ -803,21 +713,27 @@ static void r200ColorMask( GLcontext *ctx,
 			   GLboolean b, GLboolean a )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint mask = r200PackColor( rmesa->r200Screen->cpp,
-				ctx->Color.ColorMask[RCOMP],
-				ctx->Color.ColorMask[GCOMP],
-				ctx->Color.ColorMask[BCOMP],
-				ctx->Color.ColorMask[ACOMP] );
-
+   GLuint mask;
+   struct radeon_renderbuffer *rrb;
    GLuint flag = rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] & ~R200_PLANE_MASK_ENABLE;
 
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+   mask = radeonPackColor( rrb->cpp,
+			   ctx->Color.ColorMask[RCOMP],
+			   ctx->Color.ColorMask[GCOMP],
+			   ctx->Color.ColorMask[BCOMP],
+			   ctx->Color.ColorMask[ACOMP] );
+
+
    if (!(r && g && b && a))
       flag |= R200_PLANE_MASK_ENABLE;
 
-   if ( rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] != flag ) { 
-      R200_STATECHANGE( rmesa, ctx ); 
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = flag; 
-   } 
+   if ( rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] != flag ) {
+      R200_STATECHANGE( rmesa, ctx );
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = flag;
+   }
 
    if ( rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK] != mask ) {
       R200_STATECHANGE( rmesa, msk );
@@ -834,7 +750,8 @@ static void r200PolygonOffset( GLcontext *ctx,
 			       GLfloat factor, GLfloat units )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->state.depth.scale };
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
 /*    factor *= 2; */
@@ -847,41 +764,16 @@ static void r200PolygonOffset( GLcontext *ctx,
    rmesa->hw.zbs.cmd[ZBS_SE_ZBIAS_CONSTANT] = constant.ui32;
 }
 
-static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint i;
-   drm_radeon_stipple_t stipple;
-
-   /* Must flip pattern upside down.
-    */
-   for ( i = 0 ; i < 32 ; i++ ) {
-      rmesa->state.stipple.mask[31 - i] = ((GLuint *) mask)[i];
-   }
-
-   /* TODO: push this into cmd mechanism
-    */
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* FIXME: Use window x,y offsets into stipple RAM.
-    */
-   stipple.mask = rmesa->state.stipple.mask;
-   drmCommandWrite( rmesa->dri.fd, DRM_RADEON_STIPPLE, 
-                    &stipple, sizeof(stipple) );
-   UNLOCK_HARDWARE( rmesa );
-}
-
 static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLboolean flag = (ctx->_TriangleCaps & DD_TRI_UNFILLED) != 0;
 
    /* Can't generally do unfilled via tcl, but some good special
-    * cases work. 
+    * cases work.
     */
    TCL_FALLBACK( ctx, R200_TCL_FALLBACK_UNFILLED, flag);
-   if (rmesa->TclFallback) {
+   if (rmesa->radeon.TclFallback) {
       r200ChooseRenderState( ctx );
       r200ChooseVertexState( ctx );
    }
@@ -920,34 +812,34 @@ static void r200UpdateSpecular( GLcontext *ctx )
 
    if (ctx->Light.Enabled &&
        ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
 	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT) |
-	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_0;
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_1;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHTING_ENABLE;
       p |=  R200_SPECULAR_ENABLE;
-      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= 
+      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &=
 	 ~R200_DIFFUSE_SPECULAR_COMBINE;
    }
    else if (ctx->Light.Enabled) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_0;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHTING_ENABLE;
    } else if (ctx->Fog.ColorSumEnabled ) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
 	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT) |
-	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       p |=  R200_SPECULAR_ENABLE;
    } else {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));
    }
 
    if (ctx->Fog.Enabled) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_1;
    }
 
@@ -958,7 +850,7 @@ static void r200UpdateSpecular( GLcontext *ctx )
 
    /* Update vertex/render formats
     */
-   if (rmesa->TclFallback) { 
+   if (rmesa->radeon.TclFallback) {
       r200ChooseRenderState( ctx );
       r200ChooseVertexState( ctx );
    }
@@ -970,7 +862,7 @@ static void r200UpdateSpecular( GLcontext *ctx )
  */
 
 
-/* Update on colormaterial, material emmissive/ambient, 
+/* Update on colormaterial, material emmissive/ambient,
  * lightmodel.globalambient
  */
 static void update_global_ambient( GLcontext *ctx )
@@ -984,23 +876,23 @@ static void update_global_ambient( GLcontext *ctx )
     */
    if ((rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_1] &
        ((3 << R200_FRONT_EMISSIVE_SOURCE_SHIFT) |
-	(3 << R200_FRONT_AMBIENT_SOURCE_SHIFT))) == 0) 
+	(3 << R200_FRONT_AMBIENT_SOURCE_SHIFT))) == 0)
    {
-      COPY_3V( &fcmd[GLT_RED], 
+      COPY_3V( &fcmd[GLT_RED],
 	       ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_EMISSION]);
       ACC_SCALE_3V( &fcmd[GLT_RED],
 		   ctx->Light.Model.Ambient,
 		   ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT]);
-   } 
+   }
    else
    {
       COPY_3V( &fcmd[GLT_RED], ctx->Light.Model.Ambient );
    }
-   
+
    R200_DB_STATECHANGE(rmesa, &rmesa->hw.glt);
 }
 
-/* Update on change to 
+/* Update on change to
  *    - light[p].colors
  *    - light[p].enabled
  */
@@ -1014,10 +906,10 @@ static void update_light_colors( GLcontext *ctx, GLuint p )
       r200ContextPtr rmesa = R200_CONTEXT(ctx);
       float *fcmd = (float *)R200_DB_STATE( lit[p] );
 
-      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );	 
+      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );
       COPY_4V( &fcmd[LIT_DIFFUSE_RED], l->Diffuse );
       COPY_4V( &fcmd[LIT_SPECULAR_RED], l->Specular );
-      
+
       R200_DB_STATECHANGE( rmesa, &rmesa->hw.lit[p] );
    }
 }
@@ -1037,7 +929,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 
    if (ctx->Light.ColorMaterialEnabled) {
       GLuint mask = ctx->Light.ColorMaterialBitmask;
-   
+
       if (mask & MAT_BIT_FRONT_EMISSION) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_EMISSIVE_SOURCE_SHIFT);
@@ -1053,7 +945,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       else
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_AMBIENT_SOURCE_SHIFT);
-	 
+
       if (mask & MAT_BIT_FRONT_DIFFUSE) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_DIFFUSE_SOURCE_SHIFT);
@@ -1061,7 +953,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       else
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_DIFFUSE_SOURCE_SHIFT);
-   
+
       if (mask & MAT_BIT_FRONT_SPECULAR) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_SPECULAR_SOURCE_SHIFT);
@@ -1070,7 +962,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_SPECULAR_SOURCE_SHIFT);
       }
-   
+
       if (mask & MAT_BIT_BACK_EMISSION) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_BACK_EMISSIVE_SOURCE_SHIFT);
@@ -1120,8 +1012,8 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       R200_STATECHANGE( rmesa, tcl );
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_1] = light_model_ctl1;
    }
-   
-   
+
+
 }
 
 void r200UpdateMaterial( GLcontext *ctx )
@@ -1131,12 +1023,12 @@ void r200UpdateMaterial( GLcontext *ctx )
    GLfloat *fcmd = (GLfloat *)R200_DB_STATE( mtl[0] );
    GLfloat *fcmd2 = (GLfloat *)R200_DB_STATE( mtl[1] );
    GLuint mask = ~0;
-   
+
    /* Might be possible and faster to update everything unconditionally? */
    if (ctx->Light.ColorMaterialEnabled)
       mask &= ~ctx->Light.ColorMaterialBitmask;
 
-   if (R200_DEBUG & DEBUG_STATE)
+   if (R200_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    if (mask & MAT_BIT_FRONT_EMISSION) {
@@ -1217,7 +1109,7 @@ void r200UpdateMaterial( GLcontext *ctx )
  *
  * which are calculated in light.c and are correct for the current
  * lighting space (model or eye), hence dependencies on _NEW_MODELVIEW
- * and _MESA_NEW_NEED_EYE_COORDS.  
+ * and _MESA_NEW_NEED_EYE_COORDS.
  */
 static void update_light( GLcontext *ctx )
 {
@@ -1234,8 +1126,8 @@ static void update_light( GLcontext *ctx )
 	 tmp &= ~R200_LIGHT_IN_MODELSPACE;
       else
 	 tmp |= R200_LIGHT_IN_MODELSPACE;
-      
-      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]) 
+
+      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0])
       {
 	 R200_STATECHANGE( rmesa, tcl );
 	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] = tmp;
@@ -1259,10 +1151,10 @@ static void update_light( GLcontext *ctx )
 	 if (ctx->Light.Light[p].Enabled) {
 	    struct gl_light *l = &ctx->Light.Light[p];
 	    GLfloat *fcmd = (GLfloat *)R200_DB_STATE( lit[p] );
-	    
+
 	    if (l->EyePosition[3] == 0.0) {
-	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm ); 
-	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm ); 
+	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm );
+	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm );
 	       fcmd[LIT_POSITION_W] = 0;
 	       fcmd[LIT_DIRECTION_W] = 0;
 	    } else {
@@ -1286,21 +1178,21 @@ static void r200Lightfv( GLcontext *ctx, GLenum light,
    GLint p = light - GL_LIGHT0;
    struct gl_light *l = &ctx->Light.Light[p];
    GLfloat *fcmd = (GLfloat *)rmesa->hw.lit[p].cmd;
-   
+
 
    switch (pname) {
-   case GL_AMBIENT:		
+   case GL_AMBIENT:
    case GL_DIFFUSE:
    case GL_SPECULAR:
       update_light_colors( ctx, p );
       break;
 
-   case GL_SPOT_DIRECTION: 
-      /* picked up in update_light */	
+   case GL_SPOT_DIRECTION:
+      /* picked up in update_light */
       break;
 
    case GL_POSITION: {
-      /* positions picked up in update_light, but can do flag here */	
+      /* positions picked up in update_light, but can do flag here */
       GLuint flag = (p&1)? R200_LIGHT_1_IS_LOCAL : R200_LIGHT_0_IS_LOCAL;
       GLuint idx = TCL_PER_LIGHT_CTL_0 + p/2;
 
@@ -1416,7 +1308,7 @@ static void r200LightModelfv( GLcontext *ctx, GLenum pname,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
    switch (pname) {
-      case GL_LIGHT_MODEL_AMBIENT: 
+      case GL_LIGHT_MODEL_AMBIENT:
 	 update_global_ambient( ctx );
 	 break;
 
@@ -1430,7 +1322,7 @@ static void r200LightModelfv( GLcontext *ctx, GLenum pname,
 	    rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHT_TWOSIDE;
 	 else
 	    rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= ~(R200_LIGHT_TWOSIDE);
-	 if (rmesa->TclFallback) {
+	 if (rmesa->radeon.TclFallback) {
 	    r200ChooseRenderState( ctx );
 	    r200ChooseVertexState( ctx );
 	 }
@@ -1675,7 +1567,7 @@ static void r200ClearStencil( GLcontext *ctx, GLint s )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   rmesa->state.stencil.clear = 
+   rmesa->radeon.state.stencil.clear =
       ((GLuint) (ctx->Stencil.Clear & 0xff) |
        (0xff << R200_STENCIL_MASK_SHIFT) |
        ((ctx->Stencil.WriteMask[0] & 0xff) << R200_STENCIL_WRITEMASK_SHIFT));
@@ -1700,19 +1592,29 @@ static void r200ClearStencil( GLcontext *ctx, GLint s )
 void r200UpdateWindow( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLfloat xoffset = (GLfloat)dPriv->x;
-   GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+   GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
-   R200_FIREVERTICES( rmesa );
    R200_STATECHANGE( rmesa, vpt );
 
    rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE]  = sx.ui32;
@@ -1723,6 +1625,30 @@ void r200UpdateWindow( GLcontext *ctx )
    rmesa->hw.vpt.cmd[VPT_SE_VPORT_ZOFFSET] = tz.ui32;
 }
 
+void r200_vtbl_update_scissor( GLcontext *ctx )
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   unsigned x1, y1, x2, y2;
+   struct radeon_renderbuffer *rrb;
+
+   R200_SET_STATE(r200, set, SET_RE_CNTL, R200_SCISSOR_ENABLE | r200->hw.set.cmd[SET_RE_CNTL]);
+
+   if (r200->radeon.state.scissor.enabled) {
+      x1 = r200->radeon.state.scissor.rect.x1;
+      y1 = r200->radeon.state.scissor.rect.y1;
+      x2 = r200->radeon.state.scissor.rect.x2;
+      y2 = r200->radeon.state.scissor.rect.y2;
+   } else {
+      rrb = radeon_get_colorbuffer(&r200->radeon);
+      x1 = 0;
+      y1 = 0;
+      x2 = rrb->base.Width - 1;
+      y2 = rrb->base.Height - 1;
+   }
+
+   R200_SET_STATE(r200, sci, SCI_XY_1, x1 | (y1 << 16));
+   R200_SET_STATE(r200, sci, SCI_XY_2, x2 | (y2 << 16));
+}
 
 
 static void r200Viewport( GLcontext *ctx, GLint x, GLint y,
@@ -1733,6 +1659,8 @@ static void r200Viewport( GLcontext *ctx, GLint x, GLint y,
     * values, or keep the originals hanging around.
     */
    r200UpdateWindow( ctx );
+
+   radeon_viewport(ctx, x, y, width, height);
 }
 
 static void r200DepthRange( GLcontext *ctx, GLclampd nearval,
@@ -1744,7 +1672,7 @@ static void r200DepthRange( GLcontext *ctx, GLclampd nearval,
 void r200UpdateViewportOffset( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    GLfloat xoffset = (GLfloat)dPriv->x;
    GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
@@ -1774,8 +1702,8 @@ void r200UpdateViewportOffset( GLcontext *ctx )
                 R200_STIPPLE_Y_OFFSET_MASK);
 
          /* add magic offsets, then invert */
-         stx = 31 - ((rmesa->dri.drawable->x - 1) & R200_STIPPLE_COORD_MASK);
-         sty = 31 - ((rmesa->dri.drawable->y + rmesa->dri.drawable->h - 1)
+         stx = 31 - ((dPriv->x - 1) & R200_STIPPLE_COORD_MASK);
+         sty = 31 - ((dPriv->y + dPriv->h - 1)
                      & R200_STIPPLE_COORD_MASK);
 
          m |= ((stx << R200_STIPPLE_X_OFFSET_SHIFT) |
@@ -1788,7 +1716,7 @@ void r200UpdateViewportOffset( GLcontext *ctx )
       }
    }
 
-   r200UpdateScissor( ctx );
+   radeonUpdateScissor( ctx );
 }
 
 
@@ -1801,11 +1729,16 @@ static void r200ClearColor( GLcontext *ctx, const GLfloat c[4] )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLubyte color[4];
+   struct radeon_renderbuffer *rrb;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
    CLAMPED_FLOAT_TO_UBYTE(color[0], c[0]);
    CLAMPED_FLOAT_TO_UBYTE(color[1], c[1]);
    CLAMPED_FLOAT_TO_UBYTE(color[2], c[2]);
    CLAMPED_FLOAT_TO_UBYTE(color[3], c[3]);
-   rmesa->state.color.clear = r200PackColor( rmesa->r200Screen->cpp,
+   rmesa->radeon.state.color.clear = radeonPackColor( rrb->cpp,
                                              color[0], color[1],
                                              color[2], color[3] );
 }
@@ -1848,96 +1781,6 @@ static void r200LogicOpCode( GLcontext *ctx, GLenum opcode )
    rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = r200_rop_tab[rop];
 }
 
-
-/*
- * Set up the cliprects for either front or back-buffer drawing.
- */
-void r200SetCliprects( r200ContextPtr rmesa )
-{
-   __DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *const readable = rmesa->dri.readable;
-   GLframebuffer *const draw_fb = (GLframebuffer*) drawable->driverPrivate;
-   GLframebuffer *const read_fb = (GLframebuffer*) readable->driverPrivate;
-
-   if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BIT_BACK_LEFT) {
-      /* Can't ignore 2d windows if we are page flipping.
-       */
-      if ( drawable->numBackClipRects == 0 || rmesa->doPageFlip ) {
-         rmesa->numClipRects = drawable->numClipRects;
-         rmesa->pClipRects = drawable->pClipRects;
-      }
-      else {
-         rmesa->numClipRects = drawable->numBackClipRects;
-         rmesa->pClipRects = drawable->pBackClipRects;
-      }
-   }
-   else {
-     /* front buffer (or none, or multiple buffers) */
-     rmesa->numClipRects = drawable->numClipRects;
-     rmesa->pClipRects = drawable->pClipRects;
-  }
-
-   if ((draw_fb->Width != drawable->w) || (draw_fb->Height != drawable->h)) {
-      _mesa_resize_framebuffer(rmesa->glCtx, draw_fb,
-			       drawable->w, drawable->h);
-      draw_fb->Initialized = GL_TRUE;
-   }
-
-   if (drawable != readable) {
-      if ((read_fb->Width != readable->w) ||
-	  (read_fb->Height != readable->h)) {
-	 _mesa_resize_framebuffer(rmesa->glCtx, read_fb,
-				  readable->w, readable->h);
-	 read_fb->Initialized = GL_TRUE;
-      }
-   }
-
-   if (rmesa->state.scissor.enabled)
-      r200RecalcScissorRects( rmesa );
-
-   rmesa->lastStamp = drawable->lastStamp;
-}
-
-
-static void r200DrawBuffer( GLcontext *ctx, GLenum mode )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s %s\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   R200_FIREVERTICES(rmesa);	/* don't pipeline cliprect changes */
-
-   if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) {
-      /* 0 (GL_NONE) buffers or multiple color drawing buffers */
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   switch ( ctx->DrawBuffer->_ColorDrawBufferIndexes[0] ) {
-   case BUFFER_FRONT_LEFT:
-   case BUFFER_BACK_LEFT:
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   default:
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   r200SetCliprects( rmesa );
-
-   /* We'll set the drawing engine's offset/pitch parameters later
-    * when we update other state.
-    */
-}
-
-
-static void r200ReadBuffer( GLcontext *ctx, GLenum mode )
-{
-   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
-}
-
 /* =============================================================
  * State enable/disable
  */
@@ -1947,7 +1790,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLuint p, flag;
 
-   if ( R200_DEBUG & DEBUG_STATE )
+   if ( R200_DEBUG & RADEON_STATE )
       fprintf( stderr, "%s( %s = %s )\n", __FUNCTION__,
 	       _mesa_lookup_enum_by_nr( cap ),
 	       state ? "GL_TRUE" : "GL_FALSE" );
@@ -1979,7 +1822,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_CLIP_PLANE2:
    case GL_CLIP_PLANE3:
    case GL_CLIP_PLANE4:
-   case GL_CLIP_PLANE5: 
+   case GL_CLIP_PLANE5:
       p = cap-GL_CLIP_PLANE0;
       R200_STATECHANGE( rmesa, tcl );
       if (state) {
@@ -2013,10 +1856,10 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
       R200_STATECHANGE(rmesa, ctx );
       if ( state ) {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->radeon.state.color.roundEnable;
       } else {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->radeon.state.color.roundEnable;
       }
       break;
 
@@ -2031,7 +1874,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
 	 rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_TCL_FOG_MASK;
       }
       r200UpdateSpecular( ctx ); /* for PK_SPEC */
-      if (rmesa->TclFallback) 
+      if (rmesa->radeon.TclFallback)
 	 r200ChooseVertexState( ctx );
       _mesa_allow_light_in_model( ctx, !state );
       break;
@@ -2046,13 +1889,13 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHT7:
       R200_STATECHANGE(rmesa, tcl);
       p = cap - GL_LIGHT0;
-      if (p&1) 
+      if (p&1)
 	 flag = (R200_LIGHT_1_ENABLE |
-		 R200_LIGHT_1_ENABLE_AMBIENT | 
+		 R200_LIGHT_1_ENABLE_AMBIENT |
 		 R200_LIGHT_1_ENABLE_SPECULAR);
       else
 	 flag = (R200_LIGHT_0_ENABLE |
-		 R200_LIGHT_0_ENABLE_AMBIENT | 
+		 R200_LIGHT_0_ENABLE_AMBIENT |
 		 R200_LIGHT_0_ENABLE_SPECULAR);
 
       if (state)
@@ -2060,7 +1903,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
       else
 	 rmesa->hw.tcl.cmd[p/2 + TCL_PER_LIGHT_CTL_0] &= ~flag;
 
-      /* 
+      /*
        */
       update_light_colors( ctx, p );
       break;
@@ -2068,7 +1911,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHTING:
       r200UpdateSpecular(ctx);
       /* for reflection map fixup - might set recheck_texgen for all units too */
-      rmesa->NewGLState |= _NEW_TEXTURE;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE;
       break;
 
    case GL_LINE_SMOOTH:
@@ -2181,21 +2024,30 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    }
 
    case GL_SCISSOR_TEST:
-      R200_FIREVERTICES( rmesa );
-      rmesa->state.scissor.enabled = state;
-      r200UpdateScissor( ctx );
+      radeon_firevertices(&rmesa->radeon);
+      rmesa->radeon.state.scissor.enabled = state;
+      radeonUpdateScissor( ctx );
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->state.stencil.hwBuffer ) {
-	 R200_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    R200_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
       }
       break;
 
@@ -2205,7 +2057,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_TEXTURE_GEN_T:
       /* Picked up in r200UpdateTextureState.
        */
-      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE; 
+      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE;
       break;
 
    case GL_COLOR_SUM_EXT:
@@ -2322,7 +2174,7 @@ void r200LightingSpaceChange( GLcontext *ctx )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLboolean tmp;
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s %d BEFORE %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]);
 
@@ -2338,7 +2190,7 @@ void r200LightingSpaceChange( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= ~R200_RESCALE_NORMALS;
    }
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s %d AFTER %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]);
 }
@@ -2381,7 +2233,7 @@ static void update_texturematrix( GLcontext *ctx )
    GLuint compsel = rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL];
    int unit;
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s before COMPSEL: %x\n", __FUNCTION__,
 	      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL]);
 
@@ -2389,7 +2241,7 @@ static void update_texturematrix( GLcontext *ctx )
    rmesa->TexMatCompSel = 0;
 
    for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (!ctx->Texture.Unit[unit]._ReallyEnabled) 
+      if (!ctx->Texture.Unit[unit]._ReallyEnabled)
 	 continue;
 
       if (ctx->TextureMatrixStack[unit].Top->type != MATRIX_IDENTITY) {
@@ -2399,21 +2251,21 @@ static void update_texturematrix( GLcontext *ctx )
 	 rmesa->TexMatCompSel |= R200_OUTPUT_TEX_0 << unit;
 
 	 if (rmesa->TexGenEnabled & (R200_TEXMAT_0_ENABLE << unit)) {
-	    /* Need to preconcatenate any active texgen 
+	    /* Need to preconcatenate any active texgen
 	     * obj/eyeplane matrices:
 	     */
 	    _math_matrix_mul_matrix( &rmesa->tmpmat,
-				     ctx->TextureMatrixStack[unit].Top, 
+				     ctx->TextureMatrixStack[unit].Top,
 				     &rmesa->TexGenMatrix[unit] );
 	    upload_matrix( rmesa, rmesa->tmpmat.m, R200_MTX_TEX0+unit );
-	 } 
+	 }
 	 else {
-	    upload_matrix( rmesa, ctx->TextureMatrixStack[unit].Top->m, 
+	    upload_matrix( rmesa, ctx->TextureMatrixStack[unit].Top->m,
 			   R200_MTX_TEX0+unit );
 	 }
       }
       else if (rmesa->TexGenEnabled & (R200_TEXMAT_0_ENABLE << unit)) {
-	 upload_matrix( rmesa, rmesa->TexGenMatrix[unit].m, 
+	 upload_matrix( rmesa, rmesa->TexGenMatrix[unit].m,
 			R200_MTX_TEX0+unit );
       }
    }
@@ -2432,69 +2284,84 @@ static void update_texturematrix( GLcontext *ctx )
    }
 }
 
-
-
-/**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void
-r200UpdateDrawBuffer(GLcontext *ctx)
+static GLboolean r200ValidateBuffers(GLcontext *ctx)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   driRenderbuffer *drb;
+   struct radeon_renderbuffer *rrb;
+   struct radeon_dma_bo *dma_bo;
+   int i, ret;
 
-   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-      /* draw to front */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
-   }
-   else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* draw to back */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
-   else {
-      /* drawing to multiple buffers, or none */
-      return;
+
+   /* depth buffer */
+   rrb = radeon_get_depthbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
-   assert(drb);
-   assert(drb->flippedPitch);
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+      radeonTexObj *t;
 
-   R200_STATECHANGE( rmesa, ctx );
+      if (!ctx->Texture.Unit[i]._ReallyEnabled)
+	 continue;
 
-   /* Note: we used the (possibly) page-flipped values */
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-     = ((drb->flippedOffset + rmesa->r200Screen->fbLocation)
-	& R200_COLOROFFSET_MASK);
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
+      t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+      if (t->image_override && t->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+      else if (t->mt->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
-}
-
 
+   dma_bo = first_elem(&rmesa->radeon.dma.reserved);
+   {
+       ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, dma_bo->bo, RADEON_GEM_DOMAIN_GTT, 0);
+       if (ret)
+	   return GL_FALSE;
+   }
+   return GL_TRUE;
+}
 
-void r200ValidateState( GLcontext *ctx )
+GLboolean r200ValidateState( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint new_state = rmesa->NewGLState;
+   GLuint new_state = rmesa->radeon.NewGLState;
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-     r200UpdateDrawBuffer(ctx);
+   if (new_state & _NEW_BUFFERS) {
+      _mesa_update_framebuffer(ctx);
+      /* this updates the DrawBuffer's Width/Height if it's a FBO */
+      _mesa_update_draw_buffer_bounds(ctx);
+
+      R200_STATECHANGE(rmesa, ctx);
    }
 
-   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM)) {
+   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)) {
       r200UpdateTextureState( ctx );
-      new_state |= rmesa->NewGLState; /* may add TEXTURE_MATRIX */
+      new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */
       r200UpdateLocalViewer( ctx );
    }
 
+   /* we need to do a space check here */
+   if (!r200ValidateBuffers(ctx))
+     return GL_FALSE;
+
 /* FIXME: don't really need most of these when vertex progs are enabled */
 
    /* Need an event driven matrix update?
     */
-   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION)) 
+   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       upload_matrix( rmesa, ctx->_ModelProjectMatrix.m, R200_MTX_MVP );
 
    /* Need these for lighting (shouldn't upload otherwise)
@@ -2518,11 +2385,12 @@ void r200ValidateState( GLcontext *ctx )
    /* emit all active clip planes if projection matrix changes.
     */
    if (new_state & (_NEW_PROJECTION)) {
-      if (ctx->Transform.ClipPlanesEnabled) 
+      if (ctx->Transform.ClipPlanesEnabled)
 	 r200UpdateClipPlanes( ctx );
    }
 
    if (new_state & (_NEW_PROGRAM|
+                    _NEW_PROGRAM_CONSTANTS |
    /* need to test for pretty much anything due to possible parameter bindings */
 	_NEW_MODELVIEW|_NEW_PROJECTION|_NEW_TRANSFORM|
 	_NEW_LIGHT|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX|
@@ -2533,7 +2401,8 @@ void r200ValidateState( GLcontext *ctx )
       else TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, 0);
    }
 
-   rmesa->NewGLState = 0;
+   rmesa->radeon.NewGLState = 0;
+   return GL_TRUE;
 }
 
 
@@ -2544,7 +2413,7 @@ static void r200InvalidateState( GLcontext *ctx, GLuint new_state )
    _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
-   R200_CONTEXT(ctx)->NewGLState |= new_state;
+   R200_CONTEXT(ctx)->radeon.NewGLState |= new_state;
 }
 
 /* A hack.  The r200 can actually cope just fine with materials
@@ -2573,12 +2442,13 @@ static void r200WrapRunPipeline( GLcontext *ctx )
    GLboolean has_material;
 
    if (0)
-      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->NewGLState);
+      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->radeon.NewGLState);
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!r200ValidateState( ctx ))
+	 FALLBACK(rmesa, RADEON_FALLBACK_TEXTURE, GL_TRUE);
 
    has_material = !ctx->VertexProgram._Enabled && ctx->Light.Enabled && check_material( ctx );
 
@@ -2587,7 +2457,7 @@ static void r200WrapRunPipeline( GLcontext *ctx )
    }
 
    /* Run the pipeline.
-    */ 
+    */
    _tnl_run_pipeline( ctx );
 
    if (has_material) {
@@ -2596,15 +2466,30 @@ static void r200WrapRunPipeline( GLcontext *ctx )
 }
 
 
+static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask )
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   GLint i;
+
+   radeon_firevertices(&r200->radeon);
+
+   R200_STATECHANGE(r200, stp);
+
+   /* Must flip pattern upside down.
+    */
+   for ( i = 31 ; i >= 0; i--) {
+     r200->hw.stp.cmd[3 + i] = ((GLuint *) mask)[i];
+   }
+}
 /* Initialize the driver's state functions.
  */
-void r200InitStateFuncs( struct dd_function_table *functions )
+void r200InitStateFuncs( struct dd_function_table *functions, GLboolean dri2 )
 {
    functions->UpdateState		= r200InvalidateState;
    functions->LightingSpaceChange	= r200LightingSpaceChange;
 
-   functions->DrawBuffer		= r200DrawBuffer;
-   functions->ReadBuffer		= r200ReadBuffer;
+   functions->DrawBuffer		= radeonDrawBuffer;
+   functions->ReadBuffer		= radeonReadBuffer;
 
    functions->AlphaFunc			= r200AlphaFunc;
    functions->BlendColor		= r200BlendColor;
@@ -2632,11 +2517,14 @@ void r200InitStateFuncs( struct dd_function_table *functions )
    functions->LogicOpcode		= r200LogicOpCode;
    functions->PolygonMode		= r200PolygonMode;
    functions->PolygonOffset		= r200PolygonOffset;
-   functions->PolygonStipple		= r200PolygonStipple;
+   if (dri2)
+      functions->PolygonStipple		= r200PolygonStipple;
+   else
+      functions->PolygonStipple		= radeonPolygonStipplePreKMS;
    functions->PointParameterfv		= r200PointParameter;
    functions->PointSize			= r200PointSize;
    functions->RenderMode		= r200RenderMode;
-   functions->Scissor			= r200Scissor;
+   functions->Scissor			= radeonScissor;
    functions->ShadeModel		= r200ShadeModel;
    functions->StencilFuncSeparate	= r200StencilFuncSeparate;
    functions->StencilMaskSeparate	= r200StencilMaskSeparate;
diff --git a/src/mesa/drivers/dri/r200/r200_state.h b/src/mesa/drivers/dri/r200/r200_state.h
index a917163a00..9c62f0a644 100644
--- a/src/mesa/drivers/dri/r200/r200_state.h
+++ b/src/mesa/drivers/dri/r200/r200_state.h
@@ -38,28 +38,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 
 extern void r200InitState( r200ContextPtr rmesa );
-extern void r200InitStateFuncs( struct dd_function_table *functions );
+extern void r200InitStateFuncs( struct dd_function_table *functions, GLboolean dri2 );
 extern void r200InitTnlFuncs( GLcontext *ctx );
 
 extern void r200UpdateMaterial( GLcontext *ctx );
 
-extern void r200SetCliprects( r200ContextPtr rmesa );
-extern void r200RecalcScissorRects( r200ContextPtr rmesa );
 extern void r200UpdateViewportOffset( GLcontext *ctx );
 extern void r200UpdateWindow( GLcontext *ctx );
 extern void r200UpdateDrawBuffer(GLcontext *ctx);
 
-extern void r200ValidateState( GLcontext *ctx );
-
-extern void r200PrintDirty( r200ContextPtr rmesa,
-			      const char *msg );
+extern GLboolean r200ValidateState( GLcontext *ctx );
 
+extern void r200_vtbl_update_scissor( GLcontext *ctx );
 
 extern void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode );
 #define FALLBACK( rmesa, bit, mode ) do {				\
    if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",		\
 		     __FUNCTION__, bit, mode );				\
-   r200Fallback( rmesa->glCtx, bit, mode );				\
+   r200Fallback( rmesa->radeon.glCtx, bit, mode );				\
 } while (0)
 
 extern void r200LightingSpaceChange( GLcontext *ctx );
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index 9e4677eda4..7697306d88 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -43,40 +43,141 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
 #include "r200_tcl.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
+#include "radeon_queryobj.h"
 
 #include "xmlpool.h"
 
+/* New (1.3) state mechanism.  3 commands (packet, scalar, vector) in
+ * 1.3 cmdbuffers allow all previous state to be updated as well as
+ * the tcl scalar and vector areas.
+ */
+static struct {
+	int start;
+	int len;
+	const char *name;
+} packet[RADEON_MAX_STATE_PACKETS] = {
+	{RADEON_PP_MISC, 7, "RADEON_PP_MISC"},
+	{RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"},
+	{RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"},
+	{RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"},
+	{RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"},
+	{RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"},
+	{RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"},
+	{RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"},
+	{RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"},
+	{RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"},
+	{RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"},
+	{RADEON_RE_MISC, 1, "RADEON_RE_MISC"},
+	{RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"},
+	{RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"},
+	{RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"},
+	{RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"},
+	{RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"},
+	{RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"},
+	{RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"},
+	{RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"},
+	{RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17,
+		    "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"},
+	{R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"},
+	{R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"},
+	{R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"},
+	{R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"},
+	{R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"},
+	{R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"},
+	{R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"},
+	{R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"},
+	{R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"},
+	{R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"},
+	{R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"},
+	{R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"},
+	{R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"},
+	{R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"},
+	{R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"},
+	{R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"},
+	{R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"},
+	{R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"},
+	{R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"},
+	{R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"},
+	{R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"},
+	{R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"},
+	{R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"},
+	{R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"},
+	{R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"},
+	{R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"},
+	{R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"},
+	{R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"},
+	{R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1,
+	 "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"},
+	{R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"},
+	{R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"},
+	{R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"},
+	{R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"},
+	{R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"},
+	{R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"},
+	{R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"},
+	{R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"},
+	{R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"},
+	{R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"},
+	{R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4,
+		    "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"},
+	{R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"},	/* 61 */
+	{R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */
+	{R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"},
+	{R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"},
+	{R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"},
+	{R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"},
+	{R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"},
+	{R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"},
+	{R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"},
+	{R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"},
+	{R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"},
+	{R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"},
+	{RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"},
+	{RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"},
+	{RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"},
+	{R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"},
+	{R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"},
+	{RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"},
+	{RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"},
+	{RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"},
+	{RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"},
+	{RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"},
+	{RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"},
+	{R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"},
+	{R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"},     /* 85 */
+	{R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"},
+	{R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"},
+	{R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"},
+	{R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"},
+	{R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"},
+	{R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"},
+	{R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"},
+	{R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"},
+	{R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"},
+};
+
 /* =============================================================
  * State initialization
  */
-
-void r200PrintDirty( r200ContextPtr rmesa, const char *msg )
+static int cmdpkt( r200ContextPtr rmesa, int id ) 
 {
-   struct r200_state_atom *l;
-
-   fprintf(stderr, msg);
-   fprintf(stderr, ": ");
+   drm_radeon_cmd_header_t h;
 
-   foreach(l, &rmesa->hw.atomlist) {
-      if (l->dirty || rmesa->hw.all_dirty)
-	 fprintf(stderr, "%s, ", l->name);
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     return CP_PACKET0(packet[id].start, packet[id].len - 1);
+   } else {
+     h.i = 0;
+     h.packet.cmd_type = RADEON_CMD_PACKET;
+     h.packet.packet_id = id;
    }
-
-   fprintf(stderr, "\n");
-}
-
-static int cmdpkt( int id ) 
-{
-   drm_radeon_cmd_header_t h;
-   h.i = 0;
-   h.packet.cmd_type = RADEON_CMD_PACKET;
-   h.packet.packet_id = id;
    return h.i;
 }
 
@@ -126,151 +227,607 @@ static int cmdscl2( int offset, int stride, int count )
    return h.i;
 }
 
-#define CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+/**
+ * Check functions are used to check if state is active.
+ * If it is active check function returns maximum emit size.
+ */
+#define CHECK( NM, FLAG, ADD )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \
 {							\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
    (void) rmesa;					\
-   return FLAG;						\
+   return (FLAG) ? atom->cmd_size + (ADD) : 0;			\
 }
 
-#define TCL_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
-{							\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && !ctx->VertexProgram._Enabled && (FLAG);	\
+#define TCL_CHECK( NM, FLAG, ADD )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
+   return (!rmesa->radeon.TclFallback && !ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size + (ADD) : 0; \
 }
 
-#define TCL_OR_VP_CHECK( NM, FLAG )			\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+#define TCL_OR_VP_CHECK( NM, FLAG, ADD )			\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \
 {							\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && (FLAG);		\
+   return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size + (ADD) : 0;	\
 }
 
-#define VP_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
-{							\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && ctx->VertexProgram._Enabled && (FLAG);		\
+#define VP_CHECK( NM, FLAG, ADD )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
+   (void) atom;								\
+   return (!rmesa->radeon.TclFallback && ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size + (ADD) : 0; \
+}
+
+CHECK( always, GL_TRUE, 0 )
+CHECK( always_add4, GL_TRUE, 4 )
+CHECK( never, GL_FALSE, 0 )
+CHECK( tex_any, ctx->Texture._EnabledUnits, 0 )
+CHECK( tf, (ctx->Texture._EnabledUnits && !ctx->ATIFragmentShader._Enabled), 0 );
+CHECK( pix_zero, !ctx->ATIFragmentShader._Enabled, 0 )
+   CHECK( texenv, (rmesa->state.envneeded & (1 << (atom->idx)) && !ctx->ATIFragmentShader._Enabled), 0 )
+CHECK( afs_pass1, (ctx->ATIFragmentShader._Enabled && (ctx->ATIFragmentShader.Current->NumPasses > 1)), 0 )
+CHECK( afs, ctx->ATIFragmentShader._Enabled, 0 )
+CHECK( tex_cube, rmesa->state.texture.unit[atom->idx].unitneeded & TEXTURE_CUBE_BIT, 3 + 3*5 - CUBE_STATE_SIZE )
+CHECK( tex_cube_cs, rmesa->state.texture.unit[atom->idx].unitneeded & TEXTURE_CUBE_BIT, 2 + 4*5 - CUBE_STATE_SIZE )
+TCL_CHECK( tcl_fog, ctx->Fog.Enabled, 0 )
+TCL_CHECK( tcl_fog_add4, ctx->Fog.Enabled, 4 )
+TCL_CHECK( tcl, GL_TRUE, 0 )
+TCL_CHECK( tcl_add8, GL_TRUE, 8 )
+TCL_CHECK( tcl_add4, GL_TRUE, 4 )
+TCL_CHECK( tcl_tex, rmesa->state.texture.unit[atom->idx].unitneeded, 0 )
+TCL_CHECK( tcl_lighting, ctx->Light.Enabled, 0 )
+TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[atom->idx].Enabled, 0 )
+TCL_CHECK( tcl_tex_add4, rmesa->state.texture.unit[atom->idx].unitneeded, 4 )
+TCL_CHECK( tcl_lighting_add4, ctx->Light.Enabled, 4 )
+TCL_CHECK( tcl_lighting_add6, ctx->Light.Enabled, 6 )
+TCL_CHECK( tcl_light_add8, ctx->Light.Enabled && ctx->Light.Light[atom->idx].Enabled, 8 )
+TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << (atom->idx))), 0 )
+TCL_OR_VP_CHECK( tcl_ucp_add4, (ctx->Transform.ClipPlanesEnabled & (1 << (atom->idx))), 4 )
+TCL_OR_VP_CHECK( tcl_or_vp, GL_TRUE, 0 )
+TCL_OR_VP_CHECK( tcl_or_vp_add2, GL_TRUE, 2 )
+VP_CHECK( tcl_vp, GL_TRUE, 0 )
+VP_CHECK( tcl_vp_add4, GL_TRUE, 4 )
+VP_CHECK( tcl_vp_size, ctx->VertexProgram.Current->Base.NumNativeInstructions > 64, 0 )
+VP_CHECK( tcl_vpp_size, ctx->VertexProgram.Current->Base.NumNativeParameters > 96, 0 )
+VP_CHECK( tcl_vp_size_add4, ctx->VertexProgram.Current->Base.NumNativeInstructions > 64, 4 )
+VP_CHECK( tcl_vpp_size_add4, ctx->VertexProgram.Current->Base.NumNativeParameters > 96, 4 )
+
+#define OUT_VEC(hdr, data) do {			\
+    drm_radeon_cmd_header_t h;					\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1));	\
+    OUT_BATCH_TABLE((data), h.vectors.count);				\
+  } while(0)
+
+#define OUT_VECLINEAR(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    uint32_t _start, _sz;						\
+    h.i = hdr;								\
+    _start = h.veclinear.addr_lo | (h.veclinear.addr_hi << 8);		\
+    _sz = h.veclinear.count * 4;					\
+    if (r200->radeon.radeonScreen->kernel_mm && _sz) { \
+    BEGIN_BATCH_NO_AUTOSTATE(dwords); \
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(_start | (1 << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT));	\
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, _sz - 1));	\
+    OUT_BATCH_TABLE((data), _sz);					\
+    END_BATCH(); \
+    } \
+  } while(0)
+
+#define OUT_SCL(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+
+#define OUT_SCL2(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset + 0x100) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+static int check_rrb(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb;
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo)
+      return 0;
+   return atom->cmd_size;
+}
+
+static void mtl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[MTL_CMD_0], (atom->cmd+1));
+   OUT_SCL2(atom->cmd[MTL_CMD_1], (atom->cmd + 18));
+   END_BATCH();
 }
 
+static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1);
+   OUT_VEC(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1);
+   END_BATCH();
+}
 
-CHECK( always, GL_TRUE )
-CHECK( never, GL_FALSE )
-CHECK( tex_any, ctx->Texture._EnabledUnits )
-CHECK( tf, (ctx->Texture._EnabledUnits && !ctx->ATIFragmentShader._Enabled) );
-CHECK( tex_pair, (rmesa->state.texture.unit[idx].unitneeded | rmesa->state.texture.unit[idx & ~1].unitneeded) )
-CHECK( tex, rmesa->state.texture.unit[idx].unitneeded )
-CHECK( pix_zero, !ctx->ATIFragmentShader._Enabled )
-CHECK( texenv, (rmesa->state.envneeded & (1 << idx) && !ctx->ATIFragmentShader._Enabled) )
-CHECK( afs_pass1, (ctx->ATIFragmentShader._Enabled && (ctx->ATIFragmentShader.Current->NumPasses > 1)) )
-CHECK( afs, ctx->ATIFragmentShader._Enabled )
-CHECK( tex_cube, rmesa->state.texture.unit[idx].unitneeded & TEXTURE_CUBE_BIT )
-TCL_CHECK( tcl_fog, ctx->Fog.Enabled )
-TCL_CHECK( tcl, GL_TRUE )
-TCL_CHECK( tcl_tex, rmesa->state.texture.unit[idx].unitneeded )
-TCL_CHECK( tcl_lighting, ctx->Light.Enabled )
-TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[idx].Enabled )
-TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << idx)) )
-TCL_OR_VP_CHECK( tcl_or_vp, GL_TRUE )
-VP_CHECK( tcl_vp, GL_TRUE )
-VP_CHECK( tcl_vp_size, ctx->VertexProgram.Current->Base.NumNativeInstructions > 64 )
-VP_CHECK( tcl_vpp_size, ctx->VertexProgram.Current->Base.NumNativeParameters > 96 )
+static void ptp_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[PTP_CMD_0], atom->cmd+1);
+   OUT_VEC(atom->cmd[PTP_CMD_1], atom->cmd+PTP_CMD_1+1);
+   END_BATCH();
+}
 
+static void veclinear_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
 
-/* Initialize the context's hardware state.
- */
-void r200InitState( r200ContextPtr rmesa )
+   OUT_VECLINEAR(atom->cmd[0], atom->cmd+1);
+}
+
+static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
 {
-   GLcontext *ctx = rmesa->glCtx;
-   GLuint color_fmt, depth_fmt, i;
-   GLint drawPitch, drawOffset;
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
 
-   switch ( rmesa->r200Screen->cpp ) {
-   case 2:
-      color_fmt = R200_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      color_fmt = R200_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" );
-      exit( -1 );
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_SCL(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
+
+
+static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
+
+static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   struct radeon_renderbuffer *rrb;
+   uint32_t cbpitch;
+   uint32_t zbpitch, depth_fmt;
+   uint32_t dwords = atom->check(ctx, atom);
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_BATCH_TABLE(atom->cmd, 5);
+
+   rrb = radeon_get_depthbuffer(&r200->radeon);
+   if (!rrb) {
+     OUT_BATCH(0);
+     OUT_BATCH(0);
+   } else {
+     zbpitch = (rrb->pitch / rrb->cpp);
+     if (r200->using_hyperz)
+       zbpitch |= RADEON_DEPTH_HYPERZ;
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+     OUT_BATCH(zbpitch);
+     if (rrb->cpp == 4) 
+       depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; 
+     else 
+       depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; 
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK; 
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt; 
+   }
+     
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(atom->cmd[CTX_CMD_1]);
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]);
+   } else {
+     atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); 
+     if (rrb->cpp == 4) 
+       atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; 
+     else 
+       atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; 
+ 
+     OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); 
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
    }
 
-   rmesa->state.color.clear = 0x00000000;
+   OUT_BATCH(atom->cmd[CTX_CMD_2]);
 
-   switch ( ctx->Visual.depthBits ) {
-   case 16:
-      rmesa->state.depth.clear = 0x0000ffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff;
-      depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z;
-      rmesa->state.stencil.clear = 0x00000000;
-      break;
-   case 24:
-      rmesa->state.depth.clear = 0x00ffffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff;
-      depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z;
-      rmesa->state.stencil.clear = 0xffff0000;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported depth %d... exiting\n",
-	       ctx->Visual.depthBits );
-      exit( -1 );
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]);
+   } else {
+     cbpitch = (rrb->pitch / rrb->cpp);
+     if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+     OUT_BATCH(cbpitch);
+   }
+
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM)
+     OUT_BATCH_TABLE((atom->cmd + 14), 4);
+
+   END_BATCH();
+}
+
+static int check_always_ctx( GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t dwords;
+
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo) {
+      return 0;
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
+   drb = radeon_get_depthbuffer(&r200->radeon);
 
-   rmesa->Fallback = 0;
+   dwords = 10;
+   if (drb)
+     dwords += 6;
+   if (rrb)
+     dwords += 8;
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM)
+     dwords += 4;
 
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      drawOffset = rmesa->r200Screen->backOffset;
-      drawPitch  = rmesa->r200Screen->backPitch;
-   } else {
-      drawOffset = rmesa->r200Screen->frontOffset;
-      drawPitch  = rmesa->r200Screen->frontPitch;
+
+   return dwords;
+}
+
+static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t cbpitch = 0;
+   uint32_t zbpitch = 0;
+   uint32_t dwords = atom->check(ctx, atom);
+   uint32_t depth_fmt;
+
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo) {
+      return;
    }
-#if 000
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-      rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+
+   atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+   if (rrb->cpp == 4)
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
+
+   cbpitch = (rrb->pitch / rrb->cpp);
+   if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+
+   drb = radeon_get_depthbuffer(&r200->radeon);
+   if (drb) {
+     zbpitch = (drb->pitch / drb->cpp);
+     if (drb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+   }
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   /* In the CS case we need to split this up */
+   OUT_BATCH(CP_PACKET0(packet[0].start, 3));
+   OUT_BATCH_TABLE((atom->cmd + 1), 4);
+
+   if (drb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0));
+     OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0));
+     OUT_BATCH(zbpitch);
+   }
+
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0));
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1));
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+   OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+
+
+   if (rrb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0));
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0));
+     OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+   }
+
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) {
+     OUT_BATCH_TABLE((atom->cmd + 14), 4);
+   }
+
+   END_BATCH();
+}
+
+static int get_tex_size(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   uint32_t dwords = atom->cmd_size + 2;
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   if (!(t && t->mt && !t->image_override))
+     dwords -= 2;
+
+   return dwords;
+}
+
+static int check_tex_pair(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   /** XOR is bit flip operation so use it for finding pair */
+   if (!(r200->state.texture.unit[atom->idx].unitneeded | r200->state.texture.unit[atom->idx ^ 1].unitneeded))
+     return 0;
+
+   return get_tex_size(ctx, atom);
+}
+
+static int check_tex(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   if (!(r200->state.texture.unit[atom->idx].unitneeded))
+     return 0;
+
+   return get_tex_size(ctx, atom);
+}
+
+
+static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   /* is this ok even with drm older than 1.18? */
+   OUT_BATCH_TABLE(atom->cmd, 10);
+
+   if (t && t->mt && !t->image_override) {
+     OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		  RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   } else if (!t) {
+     /* workaround for old CS mechanism */
+     OUT_BATCH(r200->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]);
    } else {
-      rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-      rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+     OUT_BATCH(t->override_offset);
    }
 
-   rmesa->state.pixel.readOffset = rmesa->state.color.drawOffset;
-   rmesa->state.pixel.readPitch  = rmesa->state.color.drawPitch;
-#endif
+   END_BATCH();
+}
+
+static int get_tex_mm_size(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   uint32_t dwords = atom->cmd_size + 2;
+   int hastexture = 1;
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   if (!t)
+	hastexture = 0;
+   else {
+	if (!t->mt && !t->bo)
+		hastexture = 0;
+   }
+
+   if (!hastexture)
+     dwords -= 4;
+   return dwords;
+}
+
+static int check_tex_pair_mm(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   /** XOR is bit flip operation so use it for finding pair */
+   if (!(r200->state.texture.unit[atom->idx].unitneeded | r200->state.texture.unit[atom->idx ^ 1].unitneeded))
+     return 0;
+
+   return get_tex_mm_size(ctx, atom);
+}
 
-   rmesa->hw.max_state_size = 0;
+static int check_tex_mm(GLcontext* ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   if (!(r200->state.texture.unit[atom->idx].unitneeded))
+     return 0;
+
+   return get_tex_mm_size(ctx, atom);
+}
+
+
+static void tex_emit_mm(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   if (!r200->state.texture.unit[i].unitneeded)
+        dwords -= 4;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH(CP_PACKET0(R200_PP_TXFILTER_0 + (32 * i), 7));
+   OUT_BATCH_TABLE((atom->cmd + 1), 8);
+
+   if (dwords > atom->cmd_size) {
+     OUT_BATCH(CP_PACKET0(R200_PP_TXOFFSET_0 + (24 * i), 0));
+     if (t->mt && !t->image_override) {
+        OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		  RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      } else {
+	if (t->bo)
+            OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+                            RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      }
+   }
+   END_BATCH();
+}
+
+
+static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx, j;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (!(t && !t->image_override))
+     dwords = 2;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   /* XXX that size won't really match with image_override... */
+   OUT_BATCH_TABLE(atom->cmd, 2);
+
+   if (t && !t->image_override) {
+     lvl = &t->mt->levels[0];
+     OUT_BATCH_TABLE((atom->cmd + 2), 1);
+     for (j = 1; j <= 5; j++) {
+       OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   }
+   END_BATCH();
+}
+
+static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx, j;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   if (!(t && !t->image_override))
+     dwords = 2;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_BATCH_TABLE(atom->cmd, 2);
+
+   if (t && !t->image_override) {
+     lvl = &t->mt->levels[0];
+     for (j = 1; j <= 5; j++) {
+       OUT_BATCH(CP_PACKET0(R200_PP_CUBIC_OFFSET_F1_0 + (24*i) + (4 * (j-1)), 0));
+       OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   }
+   END_BATCH();
+}
+
+/* Initialize the context's hardware state.
+ */
+void r200InitState( r200ContextPtr rmesa )
+{
+   GLcontext *ctx = rmesa->radeon.glCtx;
+   GLuint i;
+
+   rmesa->radeon.state.color.clear = 0x00000000;
+
+   switch ( ctx->Visual.depthBits ) {
+   case 16:
+      rmesa->radeon.state.depth.clear = 0x0000ffff;
+      rmesa->radeon.state.stencil.clear = 0x00000000;
+      break;
+   case 24:
+   default:
+      rmesa->radeon.state.depth.clear = 0x00ffffff;
+      rmesa->radeon.state.stencil.clear = 0xffff0000;
+      break;
+   }
+
+   rmesa->radeon.Fallback = 0;
+
+   rmesa->radeon.hw.max_state_size = 0;
 
 #define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX )				\
    do {								\
       rmesa->hw.ATOM.cmd_size = SZ;				\
-      rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
       rmesa->hw.ATOM.name = NM;					\
       rmesa->hw.ATOM.idx = IDX;					\
-      rmesa->hw.ATOM.check = check_##CHK;			\
+      if (check_##CHK != check_never) {				\
+         rmesa->hw.ATOM.check = check_##CHK;			\
+         rmesa->radeon.hw.max_state_size += SZ * sizeof(int);	\
+      } else {							\
+         rmesa->hw.ATOM.check = NULL;				\
+      }								\
       rmesa->hw.ATOM.dirty = GL_FALSE;				\
-      rmesa->hw.max_state_size += SZ * sizeof(int);		\
    } while (0)
 
 
    /* Allocate state buffers:
     */
-   if (rmesa->r200Screen->drmSupportsBlendColor)
-      ALLOC_STATE( ctx, always, CTX_STATE_SIZE_NEWDRM, "CTX/context", 0 );
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+      ALLOC_STATE( ctx, always_add4, CTX_STATE_SIZE_NEWDRM, "CTX/context", 0 );
+   else
+      ALLOC_STATE( ctx, always_add4, CTX_STATE_SIZE_OLDDRM, "CTX/context", 0 );
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+   {
+     rmesa->hw.ctx.emit = ctx_emit_cs;
+     rmesa->hw.ctx.check = check_always_ctx;
+   }
    else
-      ALLOC_STATE( ctx, always, CTX_STATE_SIZE_OLDDRM, "CTX/context", 0 );
+   {
+     rmesa->hw.ctx.emit = ctx_emit;
+   }
    ALLOC_STATE( set, always, SET_STATE_SIZE, "SET/setup", 0 );
    ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 );
    ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 );
@@ -282,52 +839,75 @@ void r200InitState( r200ContextPtr rmesa )
    ALLOC_STATE( cst, always, CST_STATE_SIZE, "CST/constant", 0 );
    ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 );
    ALLOC_STATE( tf, tf, TF_STATE_SIZE, "TF/tfactor", 0 );
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
-      /* make sure texture units 0/1 are emitted pair-wise for r200 t0 hang workaround */
-	 ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-0", 0 );
-	 ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-1", 1 );
-	 ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 );
+   {
+      int state_size = TEX_STATE_SIZE_NEWDRM;
+      if (!rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+         state_size = TEX_STATE_SIZE_OLDDRM;
       }
-      else {
-	 ALLOC_STATE( tex[0], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-0", 0 );
-	 ALLOC_STATE( tex[1], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-1", 1 );
-	 ALLOC_STATE( tam, never, TAM_STATE_SIZE, "TAM/tam", 0 );
+      if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+         if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
+            /* make sure texture units 0/1 are emitted pair-wise for r200 t0 hang workaround */
+            ALLOC_STATE( tex[0], tex_pair_mm, state_size, "TEX/tex-0", 0 );
+            ALLOC_STATE( tex[1], tex_pair_mm, state_size, "TEX/tex-1", 1 );
+            ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 );
+         }
+         else {
+            ALLOC_STATE( tex[0], tex_mm, state_size, "TEX/tex-0", 0 );
+            ALLOC_STATE( tex[1], tex_mm, state_size, "TEX/tex-1", 1 );
+            ALLOC_STATE( tam, never, TAM_STATE_SIZE, "TAM/tam", 0 );
+         }
+         ALLOC_STATE( tex[2], tex_mm, state_size, "TEX/tex-2", 2 );
+         ALLOC_STATE( tex[3], tex_mm, state_size, "TEX/tex-3", 3 );
+         ALLOC_STATE( tex[4], tex_mm, state_size, "TEX/tex-4", 4 );
+         ALLOC_STATE( tex[5], tex_mm, state_size, "TEX/tex-5", 5 );
+         if (!rmesa->radeon.radeonScreen->kernel_mm)
+         {
+            if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
+               rmesa->hw.tex[0].check = check_tex_pair;
+               rmesa->hw.tex[1].check = check_tex_pair;
+            } else {
+               rmesa->hw.tex[0].check = check_tex;
+               rmesa->hw.tex[1].check = check_tex;
+            }
+            rmesa->hw.tex[2].check = check_tex;
+            rmesa->hw.tex[3].check = check_tex;
+            rmesa->hw.tex[4].check = check_tex;
+            rmesa->hw.tex[5].check = check_tex;
+         }
+         if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+            ALLOC_STATE( atf, afs, ATF_STATE_SIZE, "ATF/tfactor", 0 );
+            ALLOC_STATE( afs[0], afs_pass1, AFS_STATE_SIZE, "AFS/afsinst-0", 0 );
+            ALLOC_STATE( afs[1], afs, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
+         } else {
+            ALLOC_STATE( atf, never, ATF_STATE_SIZE, "ATF/tfactor", 0 );
+            ALLOC_STATE( afs[0], never, AFS_STATE_SIZE, "AFS/afsinst-0", 0 );
+            ALLOC_STATE( afs[1], never, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
+         }
       }
-      ALLOC_STATE( tex[2], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-2", 2 );
-      ALLOC_STATE( tex[3], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-3", 3 );
-      ALLOC_STATE( tex[4], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-4", 4 );
-      ALLOC_STATE( tex[5], tex, TEX_STATE_SIZE_NEWDRM, "TEX/tex-5", 5 );
-      ALLOC_STATE( atf, afs, ATF_STATE_SIZE, "ATF/tfactor", 0 );
-      ALLOC_STATE( afs[0], afs_pass1, AFS_STATE_SIZE, "AFS/afsinst-0", 0 );
-      ALLOC_STATE( afs[1], afs, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
    }
-   else {
-      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
-	 ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-0", 0 );
-	 ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-1", 1 );
-	 ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 );
-      }
-      else {
-	 ALLOC_STATE( tex[0], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-0", 0 );
-	 ALLOC_STATE( tex[1], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-1", 1 );
-	 ALLOC_STATE( tam, never, TAM_STATE_SIZE, "TAM/tam", 0 );
-      }
-      ALLOC_STATE( tex[2], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-2", 2 );
-      ALLOC_STATE( tex[3], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-3", 3 );
-      ALLOC_STATE( tex[4], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-4", 4 );
-      ALLOC_STATE( tex[5], tex, TEX_STATE_SIZE_OLDDRM, "TEX/tex-5", 5 );
-      ALLOC_STATE( atf, never, ATF_STATE_SIZE, "TF/tfactor", 0 );
-      ALLOC_STATE( afs[0], never, AFS_STATE_SIZE, "AFS/afsinst-0", 0 );
-      ALLOC_STATE( afs[1], never, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
-   }
-   if (rmesa->r200Screen->drmSupportsCubeMapsR200) {
+   /* polygon stipple is done with irq for non-kms */
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+       ALLOC_STATE( stp, always, STP_STATE_SIZE, "STP/stp", 0 );
+   }
+
+   for (i = 0; i < 6; i++)
+      if (rmesa->radeon.radeonScreen->kernel_mm)
+          rmesa->hw.tex[i].emit = tex_emit_mm;
+      else
+          rmesa->hw.tex[i].emit = tex_emit;
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200) {
       ALLOC_STATE( cube[0], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-0", 0 );
       ALLOC_STATE( cube[1], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-1", 1 );
       ALLOC_STATE( cube[2], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-2", 2 );
       ALLOC_STATE( cube[3], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-3", 3 );
       ALLOC_STATE( cube[4], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-4", 4 );
       ALLOC_STATE( cube[5], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-5", 5 );
+      for (i = 0; i < 6; i++)
+          if (rmesa->radeon.radeonScreen->kernel_mm) {
+              rmesa->hw.cube[i].emit = cube_emit_cs;
+              rmesa->hw.cube[i].check = check_tex_cube_cs;
+          } else
+              rmesa->hw.cube[i].emit = cube_emit;
    }
    else {
       ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/tex-0", 0 );
@@ -337,12 +917,20 @@ void r200InitState( r200ContextPtr rmesa )
       ALLOC_STATE( cube[4], never, CUBE_STATE_SIZE, "CUBE/tex-4", 4 );
       ALLOC_STATE( cube[5], never, CUBE_STATE_SIZE, "CUBE/tex-5", 5 );
    }
-   if (rmesa->r200Screen->drmSupportsVertexProgram) {
+
+   if (rmesa->radeon.radeonScreen->drmSupportsVertexProgram) {
       ALLOC_STATE( pvs, tcl_vp, PVS_STATE_SIZE, "PVS/pvscntl", 0 );
-      ALLOC_STATE( vpi[0], tcl_vp, VPI_STATE_SIZE, "VP/vertexprog-0", 0 );
-      ALLOC_STATE( vpi[1], tcl_vp_size, VPI_STATE_SIZE, "VP/vertexprog-1", 1 );
-      ALLOC_STATE( vpp[0], tcl_vp, VPP_STATE_SIZE, "VPP/vertexparam-0", 0 );
-      ALLOC_STATE( vpp[1], tcl_vpp_size, VPP_STATE_SIZE, "VPP/vertexparam-1", 1 );
+      if (rmesa->radeon.radeonScreen->kernel_mm) {
+         ALLOC_STATE( vpi[0], tcl_vp_add4, VPI_STATE_SIZE, "VP/vertexprog-0", 0 );
+         ALLOC_STATE( vpi[1], tcl_vp_size_add4, VPI_STATE_SIZE, "VP/vertexprog-1", 1 );
+         ALLOC_STATE( vpp[0], tcl_vp_add4, VPP_STATE_SIZE, "VPP/vertexparam-0", 0 );
+         ALLOC_STATE( vpp[1], tcl_vpp_size_add4, VPP_STATE_SIZE, "VPP/vertexparam-1", 1 );
+      } else {
+         ALLOC_STATE( vpi[0], tcl_vp, VPI_STATE_SIZE, "VP/vertexprog-0", 0 );
+         ALLOC_STATE( vpi[1], tcl_vp_size, VPI_STATE_SIZE, "VP/vertexprog-1", 1 );
+         ALLOC_STATE( vpp[0], tcl_vp, VPP_STATE_SIZE, "VPP/vertexparam-0", 0 );
+         ALLOC_STATE( vpp[1], tcl_vpp_size, VPP_STATE_SIZE, "VPP/vertexparam-1", 1 );
+      }
    }
    else {
       ALLOC_STATE( pvs, never, PVS_STATE_SIZE, "PVS/pvscntl", 0 );
@@ -355,50 +943,87 @@ void r200InitState( r200ContextPtr rmesa )
    ALLOC_STATE( tcl, tcl_or_vp, TCL_STATE_SIZE, "TCL/tcl", 0 );
    ALLOC_STATE( msl, tcl, MSL_STATE_SIZE, "MSL/matrix-select", 0 );
    ALLOC_STATE( tcg, tcl, TCG_STATE_SIZE, "TCG/texcoordgen", 0 );
-   ALLOC_STATE( mtl[0], tcl_lighting, MTL_STATE_SIZE, "MTL0/material0", 0 );
-   ALLOC_STATE( mtl[1], tcl_lighting, MTL_STATE_SIZE, "MTL1/material1", 1 );
-   ALLOC_STATE( grd, tcl_or_vp, GRD_STATE_SIZE, "GRD/guard-band", 0 );
-   ALLOC_STATE( fog, tcl_fog, FOG_STATE_SIZE, "FOG/fog", 0 );
-   ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 0 );
-   ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 0 );
-   ALLOC_STATE( mat[R200_MTX_MV], tcl, MAT_STATE_SIZE, "MAT/modelview", 0 );
-   ALLOC_STATE( mat[R200_MTX_IMV], tcl, MAT_STATE_SIZE, "MAT/it-modelview", 0 );
-   ALLOC_STATE( mat[R200_MTX_MVP], tcl, MAT_STATE_SIZE, "MAT/modelproject", 0 );
-   ALLOC_STATE( mat[R200_MTX_TEX0], tcl_tex, MAT_STATE_SIZE, "MAT/texmat0", 0 );
-   ALLOC_STATE( mat[R200_MTX_TEX1], tcl_tex, MAT_STATE_SIZE, "MAT/texmat1", 1 );
-   ALLOC_STATE( mat[R200_MTX_TEX2], tcl_tex, MAT_STATE_SIZE, "MAT/texmat2", 2 );
-   ALLOC_STATE( mat[R200_MTX_TEX3], tcl_tex, MAT_STATE_SIZE, "MAT/texmat3", 3 );
-   ALLOC_STATE( mat[R200_MTX_TEX4], tcl_tex, MAT_STATE_SIZE, "MAT/texmat4", 4 );
-   ALLOC_STATE( mat[R200_MTX_TEX5], tcl_tex, MAT_STATE_SIZE, "MAT/texmat5", 5 );
-   ALLOC_STATE( ucp[0], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-0", 0 );
-   ALLOC_STATE( ucp[1], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
-   ALLOC_STATE( ucp[2], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-2", 2 );
-   ALLOC_STATE( ucp[3], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-3", 3 );
-   ALLOC_STATE( ucp[4], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-4", 4 );
-   ALLOC_STATE( ucp[5], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-5", 5 );
-   ALLOC_STATE( lit[0], tcl_light, LIT_STATE_SIZE, "LIT/light-0", 0 );
-   ALLOC_STATE( lit[1], tcl_light, LIT_STATE_SIZE, "LIT/light-1", 1 );
-   ALLOC_STATE( lit[2], tcl_light, LIT_STATE_SIZE, "LIT/light-2", 2 );
-   ALLOC_STATE( lit[3], tcl_light, LIT_STATE_SIZE, "LIT/light-3", 3 );
-   ALLOC_STATE( lit[4], tcl_light, LIT_STATE_SIZE, "LIT/light-4", 4 );
-   ALLOC_STATE( lit[5], tcl_light, LIT_STATE_SIZE, "LIT/light-5", 5 );
-   ALLOC_STATE( lit[6], tcl_light, LIT_STATE_SIZE, "LIT/light-6", 6 );
-   ALLOC_STATE( lit[7], tcl_light, LIT_STATE_SIZE, "LIT/light-7", 7 );
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      ALLOC_STATE( mtl[0], tcl_lighting_add6, MTL_STATE_SIZE, "MTL0/material0", 0 );
+      ALLOC_STATE( mtl[1], tcl_lighting_add6, MTL_STATE_SIZE, "MTL1/material1", 1 );
+      ALLOC_STATE( grd, tcl_or_vp_add2, GRD_STATE_SIZE, "GRD/guard-band", 0 );
+      ALLOC_STATE( fog, tcl_fog_add4, FOG_STATE_SIZE, "FOG/fog", 0 );
+      ALLOC_STATE( glt, tcl_lighting_add4, GLT_STATE_SIZE, "GLT/light-global", 0 );
+      ALLOC_STATE( eye, tcl_lighting_add4, EYE_STATE_SIZE, "EYE/eye-vector", 0 );
+      ALLOC_STATE( mat[R200_MTX_MV], tcl_add4, MAT_STATE_SIZE, "MAT/modelview", 0 );
+      ALLOC_STATE( mat[R200_MTX_IMV], tcl_add4, MAT_STATE_SIZE, "MAT/it-modelview", 0 );
+      ALLOC_STATE( mat[R200_MTX_MVP], tcl_add4, MAT_STATE_SIZE, "MAT/modelproject", 0 );
+      ALLOC_STATE( mat[R200_MTX_TEX0], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat0", 0 );
+      ALLOC_STATE( mat[R200_MTX_TEX1], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat1", 1 );
+      ALLOC_STATE( mat[R200_MTX_TEX2], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat2", 2 );
+      ALLOC_STATE( mat[R200_MTX_TEX3], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat3", 3 );
+      ALLOC_STATE( mat[R200_MTX_TEX4], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat4", 4 );
+      ALLOC_STATE( mat[R200_MTX_TEX5], tcl_tex_add4, MAT_STATE_SIZE, "MAT/texmat5", 5 );
+      ALLOC_STATE( ucp[0], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-0", 0 );
+      ALLOC_STATE( ucp[1], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
+      ALLOC_STATE( ucp[2], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-2", 2 );
+      ALLOC_STATE( ucp[3], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-3", 3 );
+      ALLOC_STATE( ucp[4], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-4", 4 );
+      ALLOC_STATE( ucp[5], tcl_ucp_add4, UCP_STATE_SIZE, "UCP/userclip-5", 5 );
+      ALLOC_STATE( lit[0], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-0", 0 );
+      ALLOC_STATE( lit[1], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-1", 1 );
+      ALLOC_STATE( lit[2], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-2", 2 );
+      ALLOC_STATE( lit[3], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-3", 3 );
+      ALLOC_STATE( lit[4], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-4", 4 );
+      ALLOC_STATE( lit[5], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-5", 5 );
+      ALLOC_STATE( lit[6], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-6", 6 );
+      ALLOC_STATE( lit[7], tcl_light_add8, LIT_STATE_SIZE, "LIT/light-7", 7 );
+      ALLOC_STATE( sci, rrb, SCI_STATE_SIZE, "SCI/scissor", 0 );
+   } else {
+      ALLOC_STATE( mtl[0], tcl_lighting, MTL_STATE_SIZE, "MTL0/material0", 0 );
+      ALLOC_STATE( mtl[1], tcl_lighting, MTL_STATE_SIZE, "MTL1/material1", 1 );
+      ALLOC_STATE( grd, tcl_or_vp, GRD_STATE_SIZE, "GRD/guard-band", 0 );
+      ALLOC_STATE( fog, tcl_fog, FOG_STATE_SIZE, "FOG/fog", 0 );
+      ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 0 );
+      ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 0 );
+      ALLOC_STATE( mat[R200_MTX_MV], tcl, MAT_STATE_SIZE, "MAT/modelview", 0 );
+      ALLOC_STATE( mat[R200_MTX_IMV], tcl, MAT_STATE_SIZE, "MAT/it-modelview", 0 );
+      ALLOC_STATE( mat[R200_MTX_MVP], tcl, MAT_STATE_SIZE, "MAT/modelproject", 0 );
+      ALLOC_STATE( mat[R200_MTX_TEX0], tcl_tex, MAT_STATE_SIZE, "MAT/texmat0", 0 );
+      ALLOC_STATE( mat[R200_MTX_TEX1], tcl_tex, MAT_STATE_SIZE, "MAT/texmat1", 1 );
+      ALLOC_STATE( mat[R200_MTX_TEX2], tcl_tex, MAT_STATE_SIZE, "MAT/texmat2", 2 );
+      ALLOC_STATE( mat[R200_MTX_TEX3], tcl_tex, MAT_STATE_SIZE, "MAT/texmat3", 3 );
+      ALLOC_STATE( mat[R200_MTX_TEX4], tcl_tex, MAT_STATE_SIZE, "MAT/texmat4", 4 );
+      ALLOC_STATE( mat[R200_MTX_TEX5], tcl_tex, MAT_STATE_SIZE, "MAT/texmat5", 5 );
+      ALLOC_STATE( ucp[0], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-0", 0 );
+      ALLOC_STATE( ucp[1], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
+      ALLOC_STATE( ucp[2], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-2", 2 );
+      ALLOC_STATE( ucp[3], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-3", 3 );
+      ALLOC_STATE( ucp[4], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-4", 4 );
+      ALLOC_STATE( ucp[5], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-5", 5 );
+      ALLOC_STATE( lit[0], tcl_light, LIT_STATE_SIZE, "LIT/light-0", 0 );
+      ALLOC_STATE( lit[1], tcl_light, LIT_STATE_SIZE, "LIT/light-1", 1 );
+      ALLOC_STATE( lit[2], tcl_light, LIT_STATE_SIZE, "LIT/light-2", 2 );
+      ALLOC_STATE( lit[3], tcl_light, LIT_STATE_SIZE, "LIT/light-3", 3 );
+      ALLOC_STATE( lit[4], tcl_light, LIT_STATE_SIZE, "LIT/light-4", 4 );
+      ALLOC_STATE( lit[5], tcl_light, LIT_STATE_SIZE, "LIT/light-5", 5 );
+      ALLOC_STATE( lit[6], tcl_light, LIT_STATE_SIZE, "LIT/light-6", 6 );
+      ALLOC_STATE( lit[7], tcl_light, LIT_STATE_SIZE, "LIT/light-7", 7 );
+      ALLOC_STATE( sci, never, SCI_STATE_SIZE, "SCI/scissor", 0 );
+   }
    ALLOC_STATE( pix[0], pix_zero, PIX_STATE_SIZE, "PIX/pixstage-0", 0 );
    ALLOC_STATE( pix[1], texenv, PIX_STATE_SIZE, "PIX/pixstage-1", 1 );
    ALLOC_STATE( pix[2], texenv, PIX_STATE_SIZE, "PIX/pixstage-2", 2 );
    ALLOC_STATE( pix[3], texenv, PIX_STATE_SIZE, "PIX/pixstage-3", 3 );
    ALLOC_STATE( pix[4], texenv, PIX_STATE_SIZE, "PIX/pixstage-4", 4 );
    ALLOC_STATE( pix[5], texenv, PIX_STATE_SIZE, "PIX/pixstage-5", 5 );
-   if (rmesa->r200Screen->drmSupportsTriPerf) {
+   if (rmesa->radeon.radeonScreen->drmSupportsTriPerf) {
       ALLOC_STATE( prf, always, PRF_STATE_SIZE, "PRF/performance-tri", 0 );
    }
    else {
       ALLOC_STATE( prf, never, PRF_STATE_SIZE, "PRF/performance-tri", 0 );
    }
-   if (rmesa->r200Screen->drmSupportsPointSprites) {
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites) {
       ALLOC_STATE( spr, always, SPR_STATE_SIZE, "SPR/pointsprite", 0 );
-      ALLOC_STATE( ptp, tcl, PTP_STATE_SIZE, "PTP/pointparams", 0 );
+      if (rmesa->radeon.radeonScreen->kernel_mm)
+         ALLOC_STATE( ptp, tcl_add8, PTP_STATE_SIZE, "PTP/pointparams", 0 );
+      else
+         ALLOC_STATE( ptp, tcl, PTP_STATE_SIZE, "PTP/pointparams", 0 );
    }
    else {
       ALLOC_STATE (spr, never, SPR_STATE_SIZE, "SPR/pointsprite", 0 );
@@ -409,87 +1034,125 @@ void r200InitState( r200ContextPtr rmesa )
 
    /* Fill in the packet headers:
     */
-   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC);
-   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL);
-   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH);
-   if (rmesa->r200Screen->drmSupportsBlendColor)
-      rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(R200_EMIT_RB3D_BLENDCOLOR);
-   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN);
-   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH);
-   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK);
-   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE);
-   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL);
-   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC);
-   rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(R200_EMIT_PP_CNTL_X);
-   rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(R200_EMIT_RB3D_DEPTHXY_OFFSET);
-   rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(R200_EMIT_RE_AUX_SCISSOR_CNTL);
-   rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(R200_EMIT_RE_SCISSOR_TL_0);
-   rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(R200_EMIT_SE_VAP_CNTL_STATUS);
-   rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(R200_EMIT_RE_POINTSIZE);
-   rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0);
-   rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(R200_EMIT_PP_TAM_DEBUG3);
-   rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(R200_EMIT_TFACTOR_0);
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(R200_EMIT_ATF_TFACTOR);
-      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_0);
-      rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0);
-      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_1);
-      rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1);
-      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_2);
-      rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2);
-      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_3);
-      rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3);
-      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_4);
-      rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4);
-      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_5);
-      rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5);
+   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC);
+   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL);
+   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH);
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+      rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(rmesa, R200_EMIT_RB3D_BLENDCOLOR);
+   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN);
+   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH);
+   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK);
+   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE);
+   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL);
+   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC);
+   rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CNTL_X);
+   rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(rmesa, R200_EMIT_RB3D_DEPTHXY_OFFSET);
+   rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(rmesa, R200_EMIT_RE_AUX_SCISSOR_CNTL);
+   rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(rmesa, R200_EMIT_RE_SCISSOR_TL_0);
+   rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(rmesa, R200_EMIT_SE_VAP_CNTL_STATUS);
+   rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(rmesa, R200_EMIT_RE_POINTSIZE);
+   rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(rmesa, R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0);
+   rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TAM_DEBUG3);
+   rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(rmesa, R200_EMIT_TFACTOR_0);
+   if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+      rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(rmesa, R200_EMIT_ATF_TFACTOR);
+      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_0);
+      rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0);
+      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_1);
+      rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1);
+      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_2);
+      rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2);
+      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_3);
+      rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3);
+      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_4);
+      rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4);
+      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_5);
+      rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5);
    } else {
-      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_0);
-      rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0);
-      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_1);
-      rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1);
-      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_2);
-      rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2);
-      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_3);
-      rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3);
-      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_4);
-      rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4);
-      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_5);
-      rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5);
-   }
-   rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_0);
-   rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_1);
-   rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(R200_EMIT_VAP_PVS_CNTL);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_0);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_0);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_1);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_1);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_2);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_2);
-   rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_3);
-   rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_3);
-   rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_4);
-   rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_4);
-   rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_5);
-   rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_5);
-   rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_0);
-   rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_1);
-   rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_2);
-   rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_3);
-   rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_4);
-   rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_5);
-   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
-   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(R200_EMIT_TCL_LIGHT_MODEL_CTL_0);
-   rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(R200_EMIT_TCL_UCP_VERT_BLEND_CTL);
-   rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(R200_EMIT_TEX_PROC_CTL_2);
-   rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(R200_EMIT_MATRIX_SELECT_0);
-   rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(R200_EMIT_VAP_CTL);
-   rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(R200_EMIT_VTX_FMT_0);
-   rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(R200_EMIT_OUTPUT_VTX_COMP_SEL);
-   rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(R200_EMIT_SE_VTX_STATE_CNTL);
-   rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(R200_EMIT_VTE_CNTL);
-   rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(R200_EMIT_PP_TRI_PERF_CNTL);
-   rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(R200_EMIT_TCL_POINT_SPRITE_CNTL);
+      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_0);
+      rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0);
+      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_1);
+      rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1);
+      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_2);
+      rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2);
+      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_3);
+      rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3);
+      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_4);
+      rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4);
+      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_5);
+      rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5);
+   }
+   rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_0);
+   rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_1);
+   rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_PVS_CNTL);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_1);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_2);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_2);
+   rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_3);
+   rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_3);
+   rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_4);
+   rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_4);
+   rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_5);
+   rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_5);
+   rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_0);
+   rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_1);
+   rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_2);
+   rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_3);
+   rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_4);
+   rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_5);
+   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR);
+   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_LIGHT_MODEL_CTL_0);
+   rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(rmesa, R200_EMIT_TCL_UCP_VERT_BLEND_CTL);
+   rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(rmesa, R200_EMIT_TEX_PROC_CTL_2);
+   rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(rmesa, R200_EMIT_MATRIX_SELECT_0);
+   rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_CTL);
+   rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTX_FMT_0);
+   rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(rmesa, R200_EMIT_OUTPUT_VTX_COMP_SEL);
+   rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(rmesa, R200_EMIT_SE_VTX_STATE_CNTL);
+   rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTE_CNTL);
+   rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TRI_PERF_CNTL);
+   rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_POINT_SPRITE_CNTL);
+
+   rmesa->hw.sci.cmd[SCI_CMD_0] = CP_PACKET0(R200_RE_AUX_SCISSOR_CNTL, 0);
+   rmesa->hw.sci.cmd[SCI_CMD_1] = CP_PACKET0(R200_RE_TOP_LEFT, 0);
+   rmesa->hw.sci.cmd[SCI_CMD_2] = CP_PACKET0(R200_RE_WIDTH_HEIGHT, 0);
+
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+
+	rmesa->hw.stp.cmd[STP_CMD_0] = CP_PACKET0(RADEON_RE_STIPPLE_ADDR, 0);
+	rmesa->hw.stp.cmd[STP_DATA_0] = 0;
+	rmesa->hw.stp.cmd[STP_CMD_1] = CP_PACKET0_ONE(RADEON_RE_STIPPLE_DATA, 31);
+
+        rmesa->hw.mtl[0].emit = mtl_emit;
+        rmesa->hw.mtl[1].emit = mtl_emit;
+
+        rmesa->hw.vpi[0].emit = veclinear_emit;
+        rmesa->hw.vpi[1].emit = veclinear_emit;
+        rmesa->hw.vpp[0].emit = veclinear_emit;
+        rmesa->hw.vpp[1].emit = veclinear_emit;
+
+        rmesa->hw.grd.emit = scl_emit;
+        rmesa->hw.fog.emit = vec_emit;
+        rmesa->hw.glt.emit = vec_emit;
+        rmesa->hw.eye.emit = vec_emit;
+
+	for (i = R200_MTX_MV; i <= R200_MTX_TEX5; i++)
+	  rmesa->hw.mat[i].emit = vec_emit;
+
+	for (i = 0; i < 8; i++)
+	  rmesa->hw.lit[i].emit = lit_emit;
+
+	for (i = 0; i < 6; i++)
+	  rmesa->hw.ucp[i].emit = vec_emit;
+
+	rmesa->hw.ptp.emit = ptp_emit;
+   }
+
+
+   
    rmesa->hw.mtl[0].cmd[MTL_CMD_0] = 
       cmdvec( R200_VS_MAT_0_EMISS, 1, 16 );
    rmesa->hw.mtl[0].cmd[MTL_CMD_1] = 
@@ -567,7 +1230,7 @@ void r200InitState( r200ContextPtr rmesa )
 				(R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
 				(R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT));
 
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = 0x00000000;
       rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = (R200_COMB_FCN_ADD_CLAMP |
 				(R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
@@ -578,18 +1241,17 @@ void r200InitState( r200ContextPtr rmesa )
    }
 
    rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] =
-      rmesa->r200Screen->depthOffset + rmesa->r200Screen->fbLocation;
+      rmesa->radeon.radeonScreen->depthOffset + rmesa->radeon.radeonScreen->fbLocation;
 
    rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = 
-      ((rmesa->r200Screen->depthPitch &
+      ((rmesa->radeon.radeonScreen->depthPitch &
 	R200_DEPTHPITCH_MASK) |
        R200_DEPTH_ENDIAN_NO_SWAP);
    
    if (rmesa->using_hyperz)
       rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] |= R200_DEPTH_HYPERZ;
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt |
-					       R200_Z_TEST_LESS |
+   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (R200_Z_TEST_LESS |
 					       R200_STENCIL_TEST_ALWAYS |
 					       R200_STENCIL_FAIL_KEEP |
 					       R200_STENCIL_ZPASS_KEEP |
@@ -599,15 +1261,14 @@ void r200InitState( r200ContextPtr rmesa )
    if (rmesa->using_hyperz) {
       rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= R200_Z_COMPRESSION_ENABLE |
 						  R200_Z_DECOMPRESSION_ENABLE;
-/*      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200)
+/*      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200)
 	 rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/
    }
 
    rmesa->hw.ctx.cmd[CTX_PP_CNTL] = (R200_ANTI_ALIAS_NONE 
  				     | R200_TEX_BLEND_0_ENABLE);
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = color_fmt;
-   switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) {
+   switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) {
    case DRI_CONF_DITHER_XERRORDIFFRESET:
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_INIT;
       break;
@@ -615,41 +1276,19 @@ void r200InitState( r200ContextPtr rmesa )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_SCALE_DITHER_ENABLE;
       break;
    }
-   if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) ==
+   if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) ==
 	DRI_CONF_ROUND_ROUND )
-      rmesa->state.color.roundEnable = R200_ROUND_ENABLE;
+      rmesa->radeon.state.color.roundEnable = R200_ROUND_ENABLE;
    else
-      rmesa->state.color.roundEnable = 0;
-   if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) ==
+      rmesa->radeon.state.color.roundEnable = 0;
+   if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) ==
 	DRI_CONF_COLOR_REDUCTION_DITHER )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_ENABLE;
    else
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable;
-
-#if 000
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((rmesa->state.color.drawOffset +
-					       rmesa->r200Screen->fbLocation)
-					      & R200_COLOROFFSET_MASK);
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((rmesa->state.color.drawPitch &
-					      R200_COLORPITCH_MASK) |
-					     R200_COLOR_ENDIAN_NO_SWAP);
-#else
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset +
-					       rmesa->r200Screen->fbLocation)
-					      & R200_COLOROFFSET_MASK);
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch &
-					      R200_COLORPITCH_MASK) |
-					     R200_COLOR_ENDIAN_NO_SWAP);
-#endif
-   /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable;
 
    rmesa->hw.prf.cmd[PRF_PP_TRI_PERF] = R200_TRI_CUTOFF_MASK - R200_TRI_CUTOFF_MASK * 
-			driQueryOptionf (&rmesa->optionCache,"texture_blend_quality");
+			driQueryOptionf (&rmesa->radeon.optionCache,"texture_blend_quality");
    rmesa->hw.prf.cmd[PRF_PP_PERF_CNTL] = 0;
 
    rmesa->hw.set.cmd[SET_SE_CNTL] = (R200_FFACE_CULL_CCW |
@@ -704,7 +1343,7 @@ void r200InitState( r200ContextPtr rmesa )
 						R200_VC_NO_SWAP;
 #endif
 
-   if (!(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) {
+   if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
       /* Bypass TCL */
       rmesa->hw.cst.cmd[CST_SE_VAP_CNTL_STATUS] |= (1<<8);
    }
@@ -743,28 +1382,28 @@ void r200InitState( r200ContextPtr rmesa )
       rmesa->hw.tex[i].cmd[TEX_PP_TXFORMAT_X] =
          (/* R200_TEXCOORD_PROJ | */
           0x100000);	/* Small default bias */
-      if (rmesa->r200Screen->drmSupportsFragShader) {
+      if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
 	 rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_NEWDRM] =
-	     rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	     rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 	 rmesa->hw.tex[i].cmd[TEX_PP_CUBIC_FACES] = 0;
 	 rmesa->hw.tex[i].cmd[TEX_PP_TXMULTI_CTL] = 0;
       }
       else {
 	  rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_OLDDRM] =
-	     rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	     rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
      }
 
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F1] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F2] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F3] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F4] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F5] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 
       rmesa->hw.pix[i].cmd[PIX_PP_TXCBLEND] =
          (R200_TXC_ARG_A_ZERO |
@@ -967,5 +1606,13 @@ void r200InitState( r200ContextPtr rmesa )
 
    r200LightingSpaceChange( ctx );
 
-   rmesa->hw.all_dirty = GL_TRUE;
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      radeon_init_query_stateobj(&rmesa->radeon, R200_QUERYOBJ_CMDSIZE);
+      rmesa->radeon.query.queryobj.cmd[R200_QUERYOBJ_CMD_0] = CP_PACKET0(RADEON_RB3D_ZPASS_DATA, 0);
+      rmesa->radeon.query.queryobj.cmd[R200_QUERYOBJ_DATA_0] = 0;
+   }
+
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
+
+   rcommonInitCmdBuf(&rmesa->radeon);
 }
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index b25f028244..240fb45078 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -39,6 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/image.h"
 #include "main/imports.h"
 #include "main/macros.h"
+#include "main/simple_list.h"
 
 #include "swrast/s_context.h"
 #include "swrast/s_fog.h"
@@ -55,27 +56,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 
 
-static void flush_last_swtcl_prim( r200ContextPtr rmesa  );
-
-
 /***********************************************************************
- *                         Initialization 
+ *                         Initialization
  ***********************************************************************/
 
 #define EMIT_ATTR( ATTR, STYLE, F0 )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
    fmt_0 |= F0;								\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static void r200SetVertexFormat( GLcontext *ctx )
@@ -100,7 +98,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    }
 
    assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-   rmesa->swtcl.vertex_attr_count = 0;
+   rmesa->radeon.swtcl.vertex_attr_count = 0;
 
    /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
     * build up a hardware vertex.
@@ -121,7 +119,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    }
 
    rmesa->swtcl.coloroffset = offset;
-#if MESA_LITTLE_ENDIAN 
+#if MESA_LITTLE_ENDIAN
    EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
 #else
    EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
@@ -132,7 +130,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) ||
        RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
 
-#if MESA_LITTLE_ENDIAN 
+#if MESA_LITTLE_ENDIAN
       if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
 	 rmesa->swtcl.specoffset = offset;
 	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_RGB, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
@@ -185,7 +183,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_SPEC_ALPHA;
    }
 
-   if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) ||
+   if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) ||
 	(rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0) ||
 	(rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
       R200_NEWPRIM(rmesa);
@@ -193,26 +191,42 @@ static void r200SetVertexFormat( GLcontext *ctx )
       rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0;
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1;
 
-      rmesa->swtcl.vertex_size =
+      rmesa->radeon.swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
-			      rmesa->swtcl.vertex_attrs, 
-			      rmesa->swtcl.vertex_attr_count,
+			      rmesa->radeon.swtcl.vertex_attrs,
+			      rmesa->radeon.swtcl.vertex_attr_count,
 			      NULL, 0 );
-      rmesa->swtcl.vertex_size /= 4;
-      RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
+      rmesa->radeon.swtcl.vertex_size /= 4;
+      RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
+   }
+}
+
+static void r200_predict_emit_size( r200ContextPtr rmesa )
+{
+   if (RADEON_DEBUG & RADEON_VERTS)
+      fprintf(stderr, "%s\n", __func__);
+   const int vertex_array_size = 7;
+   const int prim_size = 3;
+   if (!rmesa->radeon.swtcl.emit_prediction) {
+      const int state_size = radeonCountStateEmitSize(&rmesa->radeon);
+      if (rcommonEnsureCmdBufSpace(&rmesa->radeon,
+	       state_size +
+	       vertex_array_size + prim_size,
+	       __FUNCTION__))
+	 rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon);
+      else
+	 rmesa->radeon.swtcl.emit_prediction = state_size;
+      rmesa->radeon.swtcl.emit_prediction += vertex_array_size + prim_size
+	 + rmesa->radeon.cmdbuf.cs->cdw;
    }
 }
 
 
 static void r200RenderStart( GLcontext *ctx )
 {
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
    r200SetVertexFormat( ctx );
-
-   if (rmesa->dma.flush != 0 && 
-       rmesa->dma.flush != flush_last_swtcl_prim)
-      rmesa->dma.flush( rmesa );
+   if (RADEON_DEBUG & RADEON_VERTS)
+      fprintf(stderr, "%s\n", __func__);
 }
 
 
@@ -232,7 +246,7 @@ void r200ChooseVertexState( GLcontext *ctx )
     * rasterization fallback.  As this function will be called again when we
     * leave a rasterization fallback, we can just skip it for now.
     */
-   if (rmesa->Fallback != 0)
+   if (rmesa->radeon.Fallback != 0)
       return;
 
    vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL];
@@ -273,78 +287,32 @@ void r200ChooseVertexState( GLcontext *ctx )
    }
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( r200ContextPtr rmesa  )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->dma.flush = NULL;
-
-   if (rmesa->dma.current.buf) {
-      struct r200_dma_region *current = &rmesa->dma.current;
-      GLuint current_offset = (rmesa->r200Screen->gart_buffer_offset +
-			       current->buf->buf->idx * RADEON_BUFFER_SIZE + 
-			       current->start);
-
-      assert (!(rmesa->swtcl.hw_primitive & R200_VF_PRIM_WALK_IND));
-
-      assert (current->start + 
-	      rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	      current->ptr);
-
-      if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-	 r200EnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ +
-			        rmesa->hw.max_state_size + VBUF_BUFSZ );
-	 r200EmitVertexAOS( rmesa,
-			      rmesa->swtcl.vertex_size,
-			      current_offset);
-
-	 r200EmitVbufPrim( rmesa,
-			   rmesa->swtcl.hw_primitive,
-			   rmesa->swtcl.numverts);
-      }
-
-      rmesa->swtcl.numverts = 0;
-      current->start = current->ptr;
-   }
-}
-
-
-/* Alloc space in the current dma region.
- */
-static INLINE void *
-r200AllocDmaLowVerts( r200ContextPtr rmesa, int nverts, int vsize )
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 {
-   GLuint bytes = vsize * nverts;
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   if (RADEON_DEBUG & RADEON_VERTS)
+      fprintf(stderr, "%s\n", __func__);
 
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      r200RefillCurrentDmaRegion( rmesa );
 
-   if (!rmesa->dma.flush) {
-      rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-      rmesa->dma.flush = flush_last_swtcl_prim;
-   }
+   radeonEmitState(&rmesa->radeon);
+   r200EmitVertexAOS( rmesa,
+		      rmesa->radeon.swtcl.vertex_size,
+		      first_elem(&rmesa->radeon.dma.reserved)->bo,
+		      current_offset);
 
-   ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-   ASSERT( rmesa->dma.flush == flush_last_swtcl_prim );
-   ASSERT( rmesa->dma.current.start + 
-	   rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	   rmesa->dma.current.ptr );
 
+   r200EmitVbufPrim( rmesa,
+		     rmesa->radeon.swtcl.hw_primitive,
+		     rmesa->radeon.swtcl.numverts);
+   if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw )
+      WARN_ONCE("Rendering was %d commands larger than predicted size."
+	    " We might overflow  command buffer.\n",
+	    rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction );
 
-   {
-      GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->dma.current.ptr += bytes;
-      rmesa->swtcl.numverts += nverts;
-      return head;
-   }
+   rmesa->radeon.swtcl.emit_prediction = 0;
 
 }
 
-
 /**************************************************************************/
 
 
@@ -389,17 +357,27 @@ static void r200ResetLineStipple( GLcontext *ctx );
 #define HAVE_POLYGONS    1
 #define HAVE_ELTS        0
 
+static void* r200_alloc_verts( r200ContextPtr rmesa, GLuint n, GLuint size)
+{
+   void *rv;
+   do {
+      r200_predict_emit_size( rmesa );
+      rv = rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 );
+   } while(!rv);
+   return rv;
+}
+
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r200ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r200AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) r200_alloc_verts(rmesa, n, size)
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;
-#define VERT(x) (r200Vertex *)(r200verts + ((x) * vertsize * sizeof(int)))
-#define VERTEX r200Vertex 
-#define DO_DEBUG_VERTS (1 && (R200_DEBUG & DEBUG_VERTS))
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;
+#define VERT(x) (radeonVertex *)(r200verts + ((x) * vertsize * sizeof(int)))
+#define VERTEX radeonVertex
+#define DO_DEBUG_VERTS (1 && (R200_DEBUG & RADEON_VERTS))
 
 #undef TAG
 #define TAG(x) r200_##x
@@ -456,11 +434,11 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
 
 #define VERT_SET_RGBA( v, c )  					\
 do {								\
-   r200_color_t *color = (r200_color_t *)&((v)->ui[coloroffset]);	\
+   radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]);	\
    UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]);		\
    UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]);		\
    UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]);		\
@@ -472,7 +450,7 @@ do {								\
 #define VERT_SET_SPEC( v, c )					\
 do {								\
    if (specoffset) {						\
-      r200_color_t *spec = (r200_color_t *)&((v)->ui[specoffset]);	\
+      radeon_color_t *spec = (radeon_color_t *)&((v)->ui[specoffset]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->red, (c)[0]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->green, (c)[1]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->blue, (c)[2]);	\
@@ -481,8 +459,8 @@ do {								\
 #define VERT_COPY_SPEC( v0, v1 )			\
 do {							\
    if (specoffset) {					\
-      r200_color_t *spec0 = (r200_color_t *)&((v0)->ui[specoffset]);	\
-      r200_color_t *spec1 = (r200_color_t *)&((v1)->ui[specoffset]);	\
+      radeon_color_t *spec0 = (radeon_color_t *)&((v0)->ui[specoffset]);	\
+      radeon_color_t *spec1 = (radeon_color_t *)&((v1)->ui[specoffset]);	\
       spec0->red   = spec1->red;	\
       spec0->green = spec1->green;	\
       spec0->blue  = spec1->blue; 	\
@@ -503,7 +481,7 @@ do {							\
 
 #define LOCAL_VARS(n)							\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
-   GLuint color[n], spec[n];						\
+   GLuint color[n] = {0}, spec[n] = {0};						\
    GLuint coloroffset = rmesa->swtcl.coloroffset;	\
    GLuint specoffset = rmesa->swtcl.specoffset;			\
    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
@@ -513,7 +491,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, x) )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -569,8 +547,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -599,13 +577,13 @@ void r200ChooseRenderState( GLcontext *ctx )
    GLuint index = 0;
    GLuint flags = ctx->_TriangleCaps;
 
-   if (!rmesa->TclFallback || rmesa->Fallback) 
+   if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback)
       return;
 
    if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R200_TWOSIDE_BIT;
    if (flags & DD_TRI_UNFILLED)      index |= R200_UNFILLED_BIT;
 
-   if (index != rmesa->swtcl.RenderIndex) {
+   if (index != rmesa->radeon.swtcl.RenderIndex) {
       tnl->Driver.Render.Points = rast_tab[index].points;
       tnl->Driver.Render.Line = rast_tab[index].line;
       tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -622,7 +600,7 @@ void r200ChooseRenderState( GLcontext *ctx )
 	 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
       }
 
-      rmesa->swtcl.RenderIndex = index;
+      rmesa->radeon.swtcl.RenderIndex = index;
    }
 }
 
@@ -636,7 +614,7 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   if (rmesa->swtcl.hw_primitive != hwprim) {
+   if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       /* need to disable perspective-correct texturing for point sprites */
       if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) {
 	 if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) {
@@ -649,15 +627,15 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 	 rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE;
       }
       R200_NEWPRIM( rmesa );
-      rmesa->swtcl.hw_primitive = hwprim;
+      rmesa->radeon.swtcl.hw_primitive = hwprim;
    }
 }
 
 static void r200RenderPrimitive( GLcontext *ctx, GLenum prim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   rmesa->swtcl.render_primitive = prim;
-   if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) 
+   rmesa->radeon.swtcl.render_primitive = prim;
+   if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED))
       r200RasterPrimitive( ctx, reduced_hw_prim(ctx, prim) );
 }
 
@@ -701,23 +679,23 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->Fallback;
+   GLuint oldfallback = rmesa->radeon.Fallback;
 
    if (mode) {
-      rmesa->Fallback |= bit;
+      rmesa->radeon.Fallback |= bit;
       if (oldfallback == 0) {
-	 R200_FIREVERTICES( rmesa );
+	 radeon_firevertices(&rmesa->radeon);
 	 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE );
 	 _swsetup_Wakeup( ctx );
-	 rmesa->swtcl.RenderIndex = ~0;
-         if (R200_DEBUG & DEBUG_FALLBACKS) {
+	 rmesa->radeon.swtcl.RenderIndex = ~0;
+         if (R200_DEBUG & RADEON_FALLBACKS) {
             fprintf(stderr, "R200 begin rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
          }
       }
    }
    else {
-      rmesa->Fallback &= ~bit;
+      rmesa->radeon.Fallback &= ~bit;
       if (oldfallback == bit) {
 
 	 _swrast_flush( ctx );
@@ -731,18 +709,18 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 	 tnl->Driver.Render.ResetLineStipple = r200ResetLineStipple;
 	 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_FALSE );
-	 if (rmesa->TclFallback) {
-	    /* These are already done if rmesa->TclFallback goes to
+	 if (rmesa->radeon.TclFallback) {
+	    /* These are already done if rmesa->radeon.TclFallback goes to
 	     * zero above. But not if it doesn't (R200_NO_TCL for
 	     * example?)
 	     */
 	    _tnl_invalidate_vertex_state( ctx, ~0 );
 	    _tnl_invalidate_vertices( ctx, ~0 );
-	    RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
+	    RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset );
 	    r200ChooseVertexState( ctx );
 	    r200ChooseRenderState( ctx );
 	 }
-         if (R200_DEBUG & DEBUG_FALLBACKS) {
+         if (R200_DEBUG & RADEON_FALLBACKS) {
             fprintf(stderr, "R200 end rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
          }
@@ -755,7 +733,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 /**
  * Cope with depth operations by drawing individual pixels as points.
- * 
+ *
  * \todo
  * The way the vertex state is set in this routine is hokey.  It seems to
  * work, but it's very hackish.  This whole routine is pretty hackish.  If
@@ -770,14 +748,14 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 		  const GLubyte *bitmap )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   const GLfloat *rc = ctx->Current.RasterColor; 
+   const GLfloat *rc = ctx->Current.RasterColor;
    GLint row, col;
-   r200Vertex vert;
+   radeonVertex vert;
    GLuint orig_vte;
    GLuint h;
 
 
-   /* Turn off tcl.  
+   /* Turn off tcl.
     */
    TCL_FALLBACK( ctx, R200_TCL_FALLBACK_BITMAP, 1 );
 
@@ -794,7 +772,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
       vte |= R200_VTX_W0_FMT;
       vap &= ~R200_VAP_FORCE_W_TO_ONE;
 
-      rmesa->swtcl.vertex_size = 5;
+      rmesa->radeon.swtcl.vertex_size = 5;
 
       if ( (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0)
 	   || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
@@ -828,7 +806,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 					   R200_VPORT_Z_SCALE_ENA |
 					   R200_VPORT_X_OFFSET_ENA |
 					   R200_VPORT_Y_OFFSET_ENA |
-					   R200_VPORT_Z_OFFSET_ENA); 
+					   R200_VPORT_Z_OFFSET_ENA);
 
    /* Turn off other stuff:  Stipple?, texture?, blending?, etc.
     */
@@ -871,16 +849,16 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 
    /* Update window height
     */
-   LOCK_HARDWARE( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   h = rmesa->dri.drawable->h + rmesa->dri.drawable->y;
-   px += rmesa->dri.drawable->x;
+   LOCK_HARDWARE( &rmesa->radeon );
+   UNLOCK_HARDWARE( &rmesa->radeon );
+   h = radeon_get_drawable(&rmesa->radeon)->h + radeon_get_drawable(&rmesa->radeon)->y;
+   px += radeon_get_drawable(&rmesa->radeon)->x;
 
    /* Clipping handled by existing mechansims in r200_ioctl.c?
     */
    for (row=0; row<height; row++) {
-      const GLubyte *src = (const GLubyte *) 
-	 _mesa_image_address2d(unpack, bitmap, width, height, 
+      const GLubyte *src = (const GLubyte *)
+	 _mesa_image_address2d(unpack, bitmap, width, height,
                                GL_COLOR_INDEX, GL_BITMAP, row, 0 );
 
       if (unpack->LsbFirst) {
@@ -929,7 +907,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 
    /* Need to restore vertexformat?
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       r200ChooseVertexState( ctx );
 }
 
@@ -949,6 +927,7 @@ void r200InitSwtcl( GLcontext *ctx )
       init_rast_tab();
       firsttime = 0;
    }
+   rmesa->radeon.swtcl.emit_prediction = 0;
 
    tnl->Driver.Render.Start = r200RenderStart;
    tnl->Driver.Render.Finish = r200RenderFinish;
@@ -959,20 +938,12 @@ void r200InitSwtcl( GLcontext *ctx )
    tnl->Driver.Render.Interp = _tnl_interp;
 
    /* FIXME: what are these numbers? */
-   _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
+   _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
 		       36 * sizeof(GLfloat) );
-   
-   rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->swtcl.render_primitive = GL_TRIANGLES;
-   rmesa->swtcl.hw_primitive = 0;
-}
 
-
-void r200DestroySwtcl( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (rmesa->swtcl.indexed_verts.buf) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, __FUNCTION__ );
+   rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+   rmesa->radeon.swtcl.hw_primitive = 0;
 }
+
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h
index 8c29fd0c99..b0905879d7 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.h
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.h
@@ -39,7 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 
 extern void r200InitSwtcl( GLcontext *ctx );
-extern void r200DestroySwtcl( GLcontext *ctx );
 
 extern void r200ChooseRenderState( GLcontext *ctx );
 extern void r200ChooseVertexState( GLcontext *ctx );
@@ -52,15 +51,11 @@ extern void r200BuildVertices( GLcontext *ctx, GLuint start, GLuint count,
 extern void r200PrintSetupFlags(char *msg, GLuint flags );
 
 
-extern void r200_emit_indexed_verts( GLcontext *ctx,
-				       GLuint start,
-				       GLuint count );
-
 extern void r200_translate_vertex( GLcontext *ctx, 
-				     const r200Vertex *src, 
+				     const radeonVertex *src, 
 				     SWvertex *dst );
 
-extern void r200_print_vertex( GLcontext *ctx, const r200Vertex *v );
+extern void r200_print_vertex( GLcontext *ctx, const radeonVertex *v );
 
 extern void r200_import_float_colors( GLcontext *ctx );
 extern void r200_import_float_spec_colors( GLcontext *ctx );
@@ -70,5 +65,5 @@ extern void r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 			      const struct gl_pixelstore_attrib *unpack,
 			      const GLubyte *bitmap );
 
-
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 99aecfe1e9..c702910ef2 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -51,6 +51,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_swtcl.h"
 #include "r200_maos.h"
 
+#include "radeon_common_context.h"
+
 
 
 #define HAVE_POINTS      1
@@ -109,7 +111,7 @@ static GLboolean discrete_prim[0x10] = {
 #define ELT_INIT(prim, hw_prim) \
    r200TclPrimitive( ctx, prim, hw_prim | R200_VF_PRIM_WALK_IND )
 
-#define GET_MESA_ELTS() rmesa->tcl.Elts
+#define GET_MESA_ELTS() TNL_CONTEXT(ctx)->vb.Elts
 
 
 /* Don't really know how many elts will fit in what's left of cmdbuf,
@@ -123,7 +125,7 @@ static GLboolean discrete_prim[0x10] = {
 
 #define RESET_STIPPLE() do {			\
    R200_STATECHANGE( rmesa, lin );		\
-   r200EmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 #define AUTO_STIPPLE( mode )  do {		\
@@ -134,7 +136,7 @@ static GLboolean discrete_prim[0x10] = {
    else						\
       rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &=	\
 	 ~R200_LINE_PATTERN_AUTO_RESET;	\
-   r200EmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 
@@ -142,27 +144,24 @@ static GLboolean discrete_prim[0x10] = {
 
 static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) 
 {
-   if (rmesa->dma.flush == r200FlushElts &&
-       rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) {
+   if (rmesa->radeon.dma.flush == r200FlushElts &&
+       rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) {
 
-      GLushort *dest = (GLushort *)(rmesa->store.cmd_buf +
-				    rmesa->store.cmd_used);
+      GLushort *dest = (GLushort *)(rmesa->radeon.tcl.elt_dma_bo->ptr +
+				    rmesa->radeon.tcl.elt_dma_offset + rmesa->tcl.elt_used);
 
-      rmesa->store.cmd_used += nr*2;
+      rmesa->tcl.elt_used += nr*2;
 
       return dest;
    }
    else {
-      if (rmesa->dma.flush)
-	 rmesa->dma.flush( rmesa );
-
-      r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			     rmesa->hw.max_state_size + ELTS_BUFSZ(nr) );
+      if (rmesa->radeon.dma.flush)
+	 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
 
       r200EmitAOS( rmesa,
-		   rmesa->tcl.aos_components,
-		   rmesa->tcl.nr_aos_components, 0 );
+		   rmesa->radeon.tcl.aos_count, 0 );
 
+      r200EmitMaxVtxIndex(rmesa, rmesa->radeon.tcl.aos[0].count);
       return r200AllocEltsOpenEnded( rmesa, rmesa->tcl.hw_primitive, nr );
    }
 }
@@ -188,13 +187,11 @@ static void r200EmitPrim( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
    r200TclPrimitive( ctx, prim, hwprim );
    
-   r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			  rmesa->hw.max_state_size + VBUF_BUFSZ );
+   //   fprintf(stderr,"Emit prim %d\n", rmesa->radeon.tcl.aos_count);
 
    r200EmitAOS( rmesa,
-		  rmesa->tcl.aos_components,
-		  rmesa->tcl.nr_aos_components,
-		  start );
+		rmesa->radeon.tcl.aos_count,
+		start );
    
    /* Why couldn't this packet have taken an offset param?
     */
@@ -207,6 +204,7 @@ static void r200EmitPrim( GLcontext *ctx,
    r200EmitPrim( ctx, prim, hwprim, start, count );             \
    (void) rmesa; } while (0)
 
+#define MAX_CONVERSION_SIZE 40
 /* Try & join small primitives
  */
 #if 0
@@ -369,6 +367,66 @@ r200ComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord )
    }
 }
 
+/**
+ * Predict total emit size for next rendering operation so there is no flush in middle of rendering
+ * Prediction has to aim towards the best possible value that is worse than worst case scenario
+ */
+static GLuint r200EnsureEmitSize( GLcontext * ctx , GLubyte* vimap_rev )
+{
+  r200ContextPtr rmesa = R200_CONTEXT(ctx);
+  TNLcontext *tnl = TNL_CONTEXT(ctx);
+  struct vertex_buffer *VB = &tnl->vb;
+  GLuint space_required;
+  GLuint state_size;
+  GLuint nr_aos = 0;
+  int i;
+  /* predict number of aos to emit */
+  for (i = 0; i < 15; ++i)
+  {
+    if (vimap_rev[i] != 255)
+    {
+      ++nr_aos;
+    }
+  }
+
+  {
+    /* count the prediction for state size */
+    space_required = 0;
+    state_size = radeonCountStateEmitSize( &rmesa->radeon );
+    /* vtx may be changed in r200EmitArrays so account for it if not dirty */
+    if (!rmesa->hw.vtx.dirty)
+      state_size += rmesa->hw.vtx.check(rmesa->radeon.glCtx, &rmesa->hw.vtx);
+    /* predict size for elements */
+    for (i = 0; i < VB->PrimitiveCount; ++i)
+    {
+      if (!VB->Primitive[i].count)
+	continue;
+      /* If primitive.count is less than MAX_CONVERSION_SIZE
+         rendering code may decide convert to elts.
+	 In that case we have to make pessimistic prediction.
+	 and use larger of 2 paths. */
+      const GLuint elts = ELTS_BUFSZ(nr_aos);
+      const GLuint index = INDEX_BUFSZ;
+      const GLuint vbuf = VBUF_BUFSZ;
+      if ( (!VB->Elts && VB->Primitive[i].count >= MAX_CONVERSION_SIZE)
+	  || vbuf > index + elts)
+	space_required += vbuf;
+      else
+	space_required += index + elts;
+      space_required += AOS_BUFSZ(nr_aos);
+    }
+  }
+
+  radeon_print(RADEON_RENDER,RADEON_VERBOSE,
+      "%s space %u, aos %d\n",
+      __func__, space_required, AOS_BUFSZ(nr_aos) );
+  /* flush the buffer in case we need more than is left. */
+  if (rcommonEnsureCmdBufSpace(&rmesa->radeon, space_required + state_size, __FUNCTION__))
+    return space_required + radeonCountStateEmitSize( &rmesa->radeon );
+  else
+    return space_required + state_size;
+}
+
 
 /**********************************************************************/
 /*                          Render pipeline stage                     */
@@ -394,19 +452,19 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 
    /* TODO: separate this from the swtnl pipeline 
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       return GL_TRUE;	/* fallback to software t&l */
 
-   if (R200_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s\n", __FUNCTION__);
 
    if (VB->Count == 0)
       return GL_FALSE;
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!r200ValidateState( ctx ))
+         return GL_TRUE; /* fallback to sw t&l */
 
    if (!ctx->VertexProgram._Enabled) {
    /* NOTE: inputs != tnl->render_inputs - these are the untransformed
@@ -481,11 +539,11 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 
    /* Do the actual work:
     */
-   r200ReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
+   radeonReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
+   GLuint emit_end = r200EnsureEmitSize( ctx, vimap_rev )
+     + rmesa->radeon.cmdbuf.cs->cdw;
    r200EmitArrays( ctx, vimap_rev );
 
-   rmesa->tcl.Elts = VB->Elts;
-
    for (i = 0 ; i < VB->PrimitiveCount ; i++)
    {
       GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
@@ -495,11 +553,14 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
       if (!length)
 	 continue;
 
-      if (rmesa->tcl.Elts)
+      if (VB->Elts)
 	 r200EmitEltPrimitive( ctx, start, start+length, prim );
       else
 	 r200EmitPrimitive( ctx, start, start+length, prim );
    }
+   if ( emit_end < rmesa->radeon.cmdbuf.cs->cdw )
+     WARN_ONCE("Rendering was %d commands larger than predicted size."
+	 " We might overflow  command buffer.\n", rmesa->radeon.cmdbuf.cs->cdw - emit_end);
 
    return GL_FALSE;		/* finished the pipe */
 }
@@ -545,7 +606,7 @@ static void transition_to_swtnl( GLcontext *ctx )
    tnl->Driver.NotifyMaterialChange = 
       _mesa_validate_all_lighting_tables;
 
-   r200ReleaseArrays( ctx, ~0 );
+   radeonReleaseArrays( ctx, ~0 );
 
    /* Still using the D3D based hardware-rasterizer from the radeon;
     * need to put the card into D3D mode to make it work:
@@ -565,15 +626,11 @@ static void transition_to_hwtnl( GLcontext *ctx )
 
    tnl->Driver.NotifyMaterialChange = r200UpdateMaterial;
 
-   if ( rmesa->dma.flush )			
-      rmesa->dma.flush( rmesa );	
+   if ( rmesa->radeon.dma.flush )			
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	
 
-   rmesa->dma.flush = NULL;
+   rmesa->radeon.dma.flush = NULL;
    
-   if (rmesa->swtcl.indexed_verts.buf) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
-
    R200_STATECHANGE( rmesa, vap );
    rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE;
    rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_FORCE_W_TO_ONE;
@@ -594,7 +651,7 @@ static void transition_to_hwtnl( GLcontext *ctx )
    rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~(R200_VTX_XY_FMT|R200_VTX_Z_FMT);
    rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] |= R200_VTX_W0_FMT;
 
-   if (R200_DEBUG & DEBUG_FALLBACKS) 
+   if (R200_DEBUG & RADEON_FALLBACKS)
       fprintf(stderr, "R200 end tcl fallback\n");
 }
 
@@ -631,21 +688,21 @@ static char *getFallbackString(GLuint bit)
 void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->TclFallback;
+   GLuint oldfallback = rmesa->radeon.TclFallback;
 
    if (mode) {
-      rmesa->TclFallback |= bit;
+      rmesa->radeon.TclFallback |= bit;
       if (oldfallback == 0) {
-	 if (R200_DEBUG & DEBUG_FALLBACKS) 
+	 if (R200_DEBUG & RADEON_FALLBACKS)
 	    fprintf(stderr, "R200 begin tcl fallback %s\n",
 		    getFallbackString( bit ));
 	 transition_to_swtnl( ctx );
       }
    }
    else {
-      rmesa->TclFallback &= ~bit;
+      rmesa->radeon.TclFallback &= ~bit;
       if (oldfallback == bit) {
-	 if (R200_DEBUG & DEBUG_FALLBACKS) 
+	 if (R200_DEBUG & RADEON_FALLBACKS)
 	    fprintf(stderr, "R200 end tcl fallback %s\n",
 		    getFallbackString( bit ));
 	 transition_to_hwtnl( ctx );
diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c
index 259f35a34c..36d9e37d87 100644
--- a/src/mesa/drivers/dri/r200/r200_tex.c
+++ b/src/mesa/drivers/dri/r200/r200_tex.c
@@ -43,8 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/teximage.h"
 #include "main/texobj.h"
 
-#include "texmem.h"
-
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -63,10 +62,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \param twrap Wrap mode for the \a t texture coordinate
  */
 
-static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap )
+static void r200SetTexWrap( radeonTexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap )
 {
    GLboolean  is_clamp = GL_FALSE;
    GLboolean  is_clamp_to_border = GL_FALSE;
+   struct gl_texture_object *tObj = &t->base;
 
    t->pp_txfilter &= ~(R200_CLAMP_S_MASK | R200_CLAMP_T_MASK | R200_BORDER_MODE_D3D);
 
@@ -103,7 +103,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum
       _mesa_problem(NULL, "bad S wrap mode in %s", __FUNCTION__);
    }
 
-   if (t->base.tObj->Target != GL_TEXTURE_1D) {
+   if (tObj->Target != GL_TEXTURE_1D) {
       switch ( twrap ) {
       case GL_REPEAT:
          t->pp_txfilter |= R200_CLAMP_T_WRAP;
@@ -180,7 +180,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum
    t->border_fallback = (is_clamp && is_clamp_to_border);
 }
 
-static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max )
+static void r200SetTexMaxAnisotropy( radeonTexObjPtr t, GLfloat max )
 {
    t->pp_txfilter &= ~R200_MAX_ANISO_MASK;
 
@@ -205,10 +205,13 @@ static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max )
  * \param magf Texture magnification mode
  */
 
-static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf )
+static void r200SetTexFilter( radeonTexObjPtr t, GLenum minf, GLenum magf )
 {
    GLuint anisotropy = (t->pp_txfilter & R200_MAX_ANISO_MASK);
 
+   /* Force revalidation to account for switches from/to mipmapping. */
+   t->validated = GL_FALSE;
+
    t->pp_txfilter &= ~(R200_MIN_FILTER_MASK | R200_MAG_FILTER_MASK);
    t->pp_txformat_x &= ~R200_VOLUME_FILTER_MASK;
 
@@ -267,700 +270,15 @@ static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf )
    }
 }
 
-static void r200SetTexBorderColor( r200TexObjPtr t, const GLfloat color[4] )
+static void r200SetTexBorderColor( radeonTexObjPtr t, const GLfloat color[4] )
 {
    GLubyte c[4];
    CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
    CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]);
    CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]);
    CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]);
-   t->pp_border_color = r200PackColor( 4, c[0], c[1], c[2], c[3] );
-}
-
-
-/**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static r200TexObjPtr r200AllocTexObj( struct gl_texture_object *texObj )
-{
-   r200TexObjPtr t;
-
-   t = CALLOC_STRUCT( r200_tex_obj );
-   texObj->DriverData = t;
-   if ( t != NULL ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE ) {
-	 fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)texObj, 
-		  (void *)t );
-      }
-
-      /* Initialize non-image-dependent parts of the state:
-       */
-      t->base.tObj = texObj;
-      t->border_fallback = GL_FALSE;
-
-      make_empty_list( & t->base );
-
-      r200SetTexWrap( t, texObj->WrapS, texObj->WrapT, texObj->WrapR );
-      r200SetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      r200SetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
-      r200SetTexBorderColor( t, texObj->BorderColor );
-   }
-
-   return t;
-}
-
-/* try to find a format which will only need a memcopy */
-static const struct gl_texture_format *
-r200Choose8888TexFormat( GLenum srcFormat, GLenum srcType )
-{
-   const GLuint ui = 1;
-   const GLubyte littleEndian = *((const GLubyte *) &ui);
-
-   if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-       (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
-      return &_mesa_texformat_rgba8888;
-   }
-   else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-       (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
-      return &_mesa_texformat_rgba8888_rev;
-   }
-   else return _dri_texformat_argb8888;
-}
-
-static const struct gl_texture_format *
-r200ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
-                           GLenum format, GLenum type )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   const GLboolean do32bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32 );
-   const GLboolean force16bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16 );
-   (void) format;
-
-   switch ( internalFormat ) {
-   case 4:
-   case GL_RGBA:
-   case GL_COMPRESSED_RGBA:
-      switch ( type ) {
-      case GL_UNSIGNED_INT_10_10_10_2:
-      case GL_UNSIGNED_INT_2_10_10_10_REV:
-	 return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      default:
-         return do32bpt ?
-	    r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444;
-      }
-
-   case 3:
-   case GL_RGB:
-   case GL_COMPRESSED_RGB:
-      switch ( type ) {
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_5_6_5:
-      case GL_UNSIGNED_SHORT_5_6_5_REV:
-	 return _dri_texformat_rgb565;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-      }
-
-   case GL_RGBA8:
-   case GL_RGB10_A2:
-   case GL_RGBA12:
-   case GL_RGBA16:
-      return !force16bpt ?
-	  r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444;
-
-   case GL_RGBA4:
-   case GL_RGBA2:
-      return _dri_texformat_argb4444;
-
-   case GL_RGB5_A1:
-      return _dri_texformat_argb1555;
-
-   case GL_RGB8:
-   case GL_RGB10:
-   case GL_RGB12:
-   case GL_RGB16:
-      return !force16bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-
-   case GL_RGB5:
-   case GL_RGB4:
-   case GL_R3_G3_B2:
-      return _dri_texformat_rgb565;
-
-   case GL_ALPHA:
-   case GL_ALPHA4:
-   case GL_ALPHA8:
-   case GL_ALPHA12:
-   case GL_ALPHA16:
-   case GL_COMPRESSED_ALPHA:
-   /* can't use a8 format since interpreting hw I8 as a8 would result
-      in wrong rgb values (same as alpha value instead of 0). */
-      return _dri_texformat_al88;
-
-   case 1:
-   case GL_LUMINANCE:
-   case GL_LUMINANCE4:
-   case GL_LUMINANCE8:
-   case GL_LUMINANCE12:
-   case GL_LUMINANCE16:
-   case GL_COMPRESSED_LUMINANCE:
-      return _dri_texformat_l8;
-
-   case 2:
-   case GL_LUMINANCE_ALPHA:
-   case GL_LUMINANCE4_ALPHA4:
-   case GL_LUMINANCE6_ALPHA2:
-   case GL_LUMINANCE8_ALPHA8:
-   case GL_LUMINANCE12_ALPHA4:
-   case GL_LUMINANCE12_ALPHA12:
-   case GL_LUMINANCE16_ALPHA16:
-   case GL_COMPRESSED_LUMINANCE_ALPHA:
-      return _dri_texformat_al88;
-
-   case GL_INTENSITY:
-   case GL_INTENSITY4:
-   case GL_INTENSITY8:
-   case GL_INTENSITY12:
-   case GL_INTENSITY16:
-   case GL_COMPRESSED_INTENSITY:
-       return _dri_texformat_i8;
-
-   case GL_YCBCR_MESA:
-      if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-          type == GL_UNSIGNED_BYTE)
-         return &_mesa_texformat_ycbcr;
-      else
-         return &_mesa_texformat_ycbcr_rev;
-
-   case GL_RGB_S3TC:
-   case GL_RGB4_S3TC:
-   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgb_dxt1;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgba_dxt1;
-
-   case GL_RGBA_S3TC:
-   case GL_RGBA4_S3TC:
-   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-      return &_mesa_texformat_rgba_dxt3;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-      return &_mesa_texformat_rgba_dxt5;
-
-   default:
-      _mesa_problem(ctx,
-         "unexpected internalFormat 0x%x in r200ChooseTextureFormat",
-         (int) internalFormat);
-      return NULL;
-   }
-
-   return NULL; /* never get here */
-}
-
-
-static GLboolean
-r200ValidateClientStorage( GLcontext *ctx, GLenum target,
-			   GLint internalFormat,
-			   GLint srcWidth, GLint srcHeight, 
-                           GLenum format, GLenum type,  const void *pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( R200_DEBUG & DEBUG_TEXTURE )
-      fprintf(stderr, "intformat %s format %s type %s\n",
-	      _mesa_lookup_enum_by_nr( internalFormat ),
-	      _mesa_lookup_enum_by_nr( format ),
-	      _mesa_lookup_enum_by_nr( type ));
-
-   if (!ctx->Unpack.ClientStorage)
-      return 0;
-
-   if (ctx->_ImageTransferState ||
-       texImage->IsCompressed ||
-       texObj->GenerateMipmap)
-      return 0;
-
-
-   /* This list is incomplete, may be different on ppc???
-    */
-   switch ( internalFormat ) {
-   case GL_RGBA:
-      if ( format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV ) {
-	 texImage->TexFormat = _dri_texformat_argb8888;
-      }
-      else
-	 return 0;
-      break;
-
-   case GL_RGB:
-      if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) {
-	 texImage->TexFormat = _dri_texformat_rgb565;
-      }
-      else
-	 return 0;
-      break;
-
-   case GL_YCBCR_MESA:
-      if ( format == GL_YCBCR_MESA && 
-	   type == GL_UNSIGNED_SHORT_8_8_REV_APPLE ) {
-	 texImage->TexFormat = &_mesa_texformat_ycbcr_rev;
-      }
-      else if ( format == GL_YCBCR_MESA && 
-		(type == GL_UNSIGNED_SHORT_8_8_APPLE || 
-		 type == GL_UNSIGNED_BYTE)) {
-	 texImage->TexFormat = &_mesa_texformat_ycbcr;
-      }
-      else
-	 return 0;
-      break;
-
-   default:
-      return 0;
-   }
-
-   /* Could deal with these packing issues, but currently don't:
-    */
-   if (packing->SkipPixels || 
-       packing->SkipRows || 
-       packing->SwapBytes ||
-       packing->LsbFirst) {
-      return 0;
-   }
-
-   {      
-      GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth,
-						  format, type);
-
-      
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf(stderr, "%s: srcRowStride %d/%x\n", 
-		 __FUNCTION__, srcRowStride, srcRowStride);
-
-      /* Could check this later in upload, pitch restrictions could be
-       * relaxed, but would need to store the image pitch somewhere,
-       * as packing details might change before image is uploaded:
-       */
-      if (!r200IsGartMemory( rmesa, pixels, srcHeight * srcRowStride ) ||
-	  (srcRowStride & 63))
-	 return 0;
-
-
-      /* Have validated that _mesa_transfer_teximage would be a straight
-       * memcpy at this point.  NOTE: future calls to TexSubImage will
-       * overwrite the client data.  This is explicitly mentioned in the
-       * extension spec.
-       */
-      texImage->Data = (void *)pixels;
-      texImage->IsClientData = GL_TRUE;
-      texImage->RowStride = srcRowStride / texImage->TexFormat->TexelBytes;
-
-      return 1;
-   }
-}
-
-
-static void r200TexImage1D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage1d(ctx, target, level, internalFormat,
-                          width, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void r200TexSubImage1D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset,
-                                 GLsizei width,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-			     format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void r200TexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat, 
-				  width, height, 
-				  format, type, pixels, 
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
-   }
-   else {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_teximage2d(ctx, target, level, internalFormat,
-			     width, height, border, format, type, pixels,
-			     &ctx->Unpack, texObj, texImage);
-      
-      t->dirty_images[face] |= (1 << level);
-   }
-}
-
-
-static void r200TexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-			     height, format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void r200CompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLsizei imageSize, const GLvoid *data,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-/* can't call this, different parameters. Would never evaluate to true anyway currently
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat,
-				  width, height,
-				  format, type, pixels,
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__);
-   }
-   else */{
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__);
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width,
-                                 height, border, imageSize, data, texObj, texImage);
-
-      t->dirty_images[face] |= (1 << level);
-   }
-}
-
-
-static void r200CompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format,
-                                 GLsizei imageSize, const GLvoid *data,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-                            height, format, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-#if ENABLE_HW_3D_TEXTURE
-static void r200TexImage3D( GLcontext *ctx, GLenum target, GLint level,
-                            GLint internalFormat,
-                            GLint width, GLint height, GLint depth,
-                            GLint border,
-                            GLenum format, GLenum type, const GLvoid *pixels,
-                            const struct gl_pixelstore_attrib *packing,
-                            struct gl_texture_object *texObj,
-                            struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-
-#if 0
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat, 
-				  width, height, 
-				  format, type, pixels, 
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
-   }
-   else
-#endif
-   {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_teximage3d(ctx, target, level, internalFormat,
-			     width, height, depth, border,
-                             format, type, pixels,
-			     &ctx->Unpack, texObj, texImage);
-      
-      t->dirty_images[0] |= (1 << level);
-   }
+   t->pp_border_color = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
 }
-#endif
-
-
-#if ENABLE_HW_3D_TEXTURE
-static void
-r200TexSubImage3D( GLcontext *ctx, GLenum target, GLint level,
-                   GLint xoffset, GLint yoffset, GLint zoffset,
-                   GLsizei width, GLsizei height, GLsizei depth,
-                   GLenum format, GLenum type,
-                   const GLvoid *pixels,
-                   const struct gl_pixelstore_attrib *packing,
-                   struct gl_texture_object *texObj,
-                   struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-/*     fprintf(stderr, "%s\n", __FUNCTION__); */
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D");
-         return;
-      }
-      texObj->DriverData = t;
-   }
-
-   _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
-                             width, height, depth,
-                             format, type, pixels, packing, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-#endif
-
-
 
 static void r200TexEnv( GLcontext *ctx, GLenum target,
 			  GLenum pname, const GLfloat *param )
@@ -969,7 +287,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target,
    GLuint unit = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 
-   if ( R200_DEBUG & DEBUG_STATE ) {
+   if ( R200_DEBUG & RADEON_STATE ) {
       fprintf( stderr, "%s( %s )\n",
 	       __FUNCTION__, _mesa_lookup_enum_by_nr( pname ) );
    }
@@ -983,7 +301,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target,
       GLubyte c[4];
       GLuint envColor;
       UNCLAMPED_FLOAT_TO_RGBA_CHAN( c, texUnit->EnvColor );
-      envColor = r200PackColor( 4, c[0], c[1], c[2], c[3] );
+      envColor = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
       if ( rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] != envColor ) {
 	 R200_STATECHANGE( rmesa, tf );
 	 rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] = envColor;
@@ -1002,7 +320,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target,
        * NOTE: Add a small bias to the bias for conform mipsel.c test.
        */
       bias = *param + .01;
-      min = driQueryOptionb (&rmesa->optionCache, "no_neg_lod_bias") ?
+      min = driQueryOptionb (&rmesa->radeon.optionCache, "no_neg_lod_bias") ?
 	  0.0 : -16.0;
       bias = CLAMP( bias, min, 16.0 );
       b = (int)(bias * fixed_one) & R200_LOD_BIAS_MASK;
@@ -1039,9 +357,9 @@ static void r200TexParameter( GLcontext *ctx, GLenum target,
 				struct gl_texture_object *texObj,
 				GLenum pname, const GLfloat *params )
 {
-   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+   radeonTexObj* t = radeon_tex_obj(texObj);
 
-   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
+   if ( R200_DEBUG & (RADEON_STATE|RADEON_TEXTURE) ) {
       fprintf( stderr, "%s( %s )\n", __FUNCTION__,
 	       _mesa_lookup_enum_by_nr( pname ) );
    }
@@ -1073,59 +391,46 @@ static void r200TexParameter( GLcontext *ctx, GLenum target,
        * we just have to rely on loading the right subset of mipmap levels
        * to simulate a clamped LOD.
        */
-      driSwapOutTextureObject( (driTextureObject *) t );
+      if (t->mt) {
+         radeon_miptree_unreference(t->mt);
+	 t->mt = 0;
+	 t->validated = GL_FALSE;
+      }
       break;
 
    default:
       return;
    }
-
-   /* Mark this texobj as dirty (one bit per tex unit)
-    */
-   t->dirty_state = TEX_ALL;
 }
 
 
-
-static void r200BindTexture( GLcontext *ctx, GLenum target,
-			       struct gl_texture_object *texObj )
-{
-   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, (void *)texObj,
-	       ctx->Texture.CurrentUnit );
-   }
-
-   if ( (target == GL_TEXTURE_1D)
-	|| (target == GL_TEXTURE_2D) 
-#if ENABLE_HW_3D_TEXTURE
-	|| (target == GL_TEXTURE_3D)
-#endif
-	|| (target == GL_TEXTURE_CUBE_MAP)
-	|| (target == GL_TEXTURE_RECTANGLE_NV) ) {
-      assert( texObj->DriverData != NULL );
-   }
-}
-
-
-static void r200DeleteTexture( GLcontext *ctx,
-				 struct gl_texture_object *texObj )
+static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj,
-	       _mesa_lookup_enum_by_nr( texObj->Target ) );
+   radeonTexObj* t = radeon_tex_obj(texObj);
+
+   if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) {
+      fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+	      (void *)texObj,
+	      _mesa_lookup_enum_by_nr(texObj->Target));
+   }
+   
+   if (rmesa) {
+      int i;
+      radeon_firevertices(&rmesa->radeon);
+      for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) {
+	 if ( t == rmesa->state.texture.unit[i].texobj ) {
+	    rmesa->state.texture.unit[i].texobj = NULL;
+	    rmesa->hw.tex[i].dirty = GL_FALSE;
+	    rmesa->hw.cube[i].dirty = GL_FALSE;
+	 }
+      }      
    }
-
-   if ( t != NULL ) {
-      if ( rmesa ) {
-         R200_FIREVERTICES( rmesa );
-      }
-
-      driDestroyTextureObject( t );
+   
+   if (t->mt) {
+      radeon_miptree_unreference(t->mt);
+      t->mt = 0;
    }
-   /* Free mipmap images and the texture object itself */
    _mesa_delete_texture_object(ctx, texObj);
 }
 
@@ -1155,46 +460,59 @@ static void r200TexGen( GLcontext *ctx,
  * Called via ctx->Driver.NewTextureObject.
  * Note: this function will be called during context creation to
  * allocate the default texture objects.
- * Note: we could use containment here to 'derive' the driver-specific
- * texture object from the core mesa gl_texture_object.  Not done at this time.
  * Fixup MaxAnisotropy according to user preference.
  */
-static struct gl_texture_object *
-r200NewTextureObject( GLcontext *ctx, GLuint name, GLenum target )
+static struct gl_texture_object *r200NewTextureObject(GLcontext * ctx,
+						      GLuint name,
+						      GLenum target)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_object *obj;
-   obj = _mesa_new_texture_object(ctx, name, target);
-   if (!obj)
-      return NULL;
-   obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
-   r200AllocTexObj( obj );
-   return obj;
+   radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+
+   if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) {
+     fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+	     t, _mesa_lookup_enum_by_nr(target));
+   }
+
+   _mesa_initialize_texture_object(&t->base, name, target);
+   t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+   /* Initialize hardware state */
+   r200SetTexWrap( t, t->base.WrapS, t->base.WrapT, t->base.WrapR );
+   r200SetTexMaxAnisotropy( t, t->base.MaxAnisotropy );
+   r200SetTexFilter(t, t->base.MinFilter, t->base.MagFilter);
+   r200SetTexBorderColor(t, t->base.BorderColor);
+
+   return &t->base;
 }
 
 
+
 void r200InitTextureFuncs( struct dd_function_table *functions )
 {
    /* Note: we only plug in the functions we implement in the driver
     * since _mesa_init_driver_functions() was already called.
     */
-   functions->ChooseTextureFormat	= r200ChooseTextureFormat;
-   functions->TexImage1D		= r200TexImage1D;
-   functions->TexImage2D		= r200TexImage2D;
+   functions->ChooseTextureFormat	= radeonChooseTextureFormat_mesa;
+   functions->TexImage1D		= radeonTexImage1D;
+   functions->TexImage2D		= radeonTexImage2D;
 #if ENABLE_HW_3D_TEXTURE
-   functions->TexImage3D		= r200TexImage3D;
+   functions->TexImage3D		= radeonTexImage3D;
 #else
    functions->TexImage3D		= _mesa_store_teximage3d;
 #endif
-   functions->TexSubImage1D		= r200TexSubImage1D;
-   functions->TexSubImage2D		= r200TexSubImage2D;
+   functions->TexSubImage1D		= radeonTexSubImage1D;
+   functions->TexSubImage2D		= radeonTexSubImage2D;
 #if ENABLE_HW_3D_TEXTURE
-   functions->TexSubImage3D		= r200TexSubImage3D;
+   functions->TexSubImage3D		= radeonTexSubImage3D;
 #else
    functions->TexSubImage3D		= _mesa_store_texsubimage3d;
 #endif
+   functions->GetTexImage               = radeonGetTexImage;
+   functions->GetCompressedTexImage     = radeonGetCompressedTexImage;
    functions->NewTextureObject		= r200NewTextureObject;
-   functions->BindTexture		= r200BindTexture;
+   //   functions->BindTexture		= r200BindTexture;
    functions->DeleteTexture		= r200DeleteTexture;
    functions->IsTextureResident		= driIsTextureResident;
 
@@ -1202,22 +520,16 @@ void r200InitTextureFuncs( struct dd_function_table *functions )
    functions->TexParameter		= r200TexParameter;
    functions->TexGen			= r200TexGen;
 
-   functions->CompressedTexImage2D	= r200CompressedTexImage2D;
-   functions->CompressedTexSubImage2D	= r200CompressedTexSubImage2D;
+   functions->CompressedTexImage2D	= radeonCompressedTexImage2D;
+   functions->CompressedTexSubImage2D	= radeonCompressedTexSubImage2D;
 
-   driInitTextureFormats();
+   functions->GenerateMipmap = radeonGenerateMipmap;
 
-#if 000
-   /* moved or obsolete code */
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   driInitTextureObjects( ctx, & rmesa->swapped,
-			  DRI_TEXMGR_DO_TEXTURE_1D
-			  | DRI_TEXMGR_DO_TEXTURE_2D );
+   functions->NewTextureImage = radeonNewTextureImage;
+   functions->FreeTexImageData = radeonFreeTexImageData;
+   functions->MapTexture = radeonMapTexture;
+   functions->UnmapTexture = radeonUnmapTexture;
+
+   driInitTextureFormats();
 
-   /* Hack: r200NewTextureObject is not yet installed when the
-    * default textures are created. Therefore set MaxAnisotropy of the
-    * default 2D texture now. */
-   ctx->Shared->Default2D->MaxAnisotropy = driQueryOptionf (&rmesa->optionCache,
-							    "def_max_anisotropy");
-#endif
 }
diff --git a/src/mesa/drivers/dri/r200/r200_tex.h b/src/mesa/drivers/dri/r200/r200_tex.h
index 10ff8e8a66..e122de6e5e 100644
--- a/src/mesa/drivers/dri/r200/r200_tex.h
+++ b/src/mesa/drivers/dri/r200/r200_tex.h
@@ -35,15 +35,18 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __R200_TEX_H__
 #define __R200_TEX_H__
 
+extern void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv);
+extern void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			      __DRIdrawable *dPriv);
 extern void r200SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 			     unsigned long long offset, GLint depth,
 			     GLuint pitch);
 
 extern void r200UpdateTextureState( GLcontext *ctx );
 
-extern int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face );
+extern int r200UploadTexImages( r200ContextPtr rmesa, radeonTexObjPtr t, GLuint face );
 
-extern void r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t );
+extern void r200DestroyTexObj( r200ContextPtr rmesa, radeonTexObjPtr t );
 
 extern void r200InitTextureFuncs( struct dd_function_table *functions );
 
diff --git a/src/mesa/drivers/dri/r200/r200_texmem.c b/src/mesa/drivers/dri/r200/r200_texmem.c
deleted file mode 100644
index 3b81ac0c80..0000000000
--- a/src/mesa/drivers/dri/r200/r200_texmem.c
+++ /dev/null
@@ -1,530 +0,0 @@
-/**************************************************************************
-
-Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.  
-The Weather Channel, Inc. funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86
-license. This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *
- */
- 
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "r200_context.h"
-#include "r200_ioctl.h"
-#include "r200_tex.h"
-#include "radeon_reg.h"
-
-#include <unistd.h>  /* for usleep() */
-
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void
-r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t )
-{
-   if ( R200_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, 
-	       (void *)t, (void *)t->base.tObj );
-   }
-
-   if ( rmesa != NULL ) {
-      unsigned   i;
-
-
-      for ( i = 0 ; i < rmesa->glCtx->Const.MaxTextureUnits ; i++ ) {
-	 if ( t == rmesa->state.texture.unit[i].texobj ) {
-	    rmesa->state.texture.unit[i].texobj = NULL;
-	    rmesa->hw.tex[i].dirty = GL_FALSE;
-	    rmesa->hw.cube[i].dirty = GL_FALSE;
-	 }
-      }
-   }
-}
-
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-
-static void r200UploadGARTClientSubImage( r200ContextPtr rmesa,
-					  r200TexObjPtr t, 
-					  struct gl_texture_image *texImage,
-					  GLint hwlevel,
-					  GLint x, GLint y, 
-					  GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   GLuint srcPitch, dstPitch;
-   int blit_format;
-   int srcOffset;
-
-   /*
-    * XXX it appears that we always upload the full image, not a subimage.
-    * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
-    * changed, the src pitch will have to change.
-    */
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = R200_CP_COLOR_FORMAT_CI8;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   case 2:
-      blit_format = R200_CP_COLOR_FORMAT_RGB565;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   case 4:
-      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   default:
-      return;
-   }
-
-   t->image[0][hwlevel].data = texImage->Data;
-   srcOffset = r200GartOffsetFromVirtual( rmesa, texImage->Data );
-
-   assert( srcOffset != ~0 );
-
-   /* Don't currently need to cope with small pitches?
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-
-   r200EmitWait( rmesa, RADEON_WAIT_3D );
-
-   r200EmitBlit( rmesa, blit_format, 
-		 srcPitch,  
-		 srcOffset,   
-		 dstPitch,
-		 t->bufAddr,
-		 x, 
-		 y, 
-		 t->image[0][hwlevel].x + x,
-		 t->image[0][hwlevel].y + y, 
-		 width,
-		 height );
-
-   r200EmitWait( rmesa, RADEON_WAIT_2D );
-}
-
-static void r200UploadRectSubImage( r200ContextPtr rmesa,
-				    r200TexObjPtr t, 
-				    struct gl_texture_image *texImage,
-				    GLint x, GLint y, 
-				    GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   int blit_format, dstPitch, done;
-
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = R200_CP_COLOR_FORMAT_CI8;
-      break;
-   case 2:
-      blit_format = R200_CP_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      return;
-   }
-
-   t->image[0][0].data = texImage->Data;
-
-   /* Currently don't need to cope with small pitches.
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-   dstPitch = t->pp_txpitch + 32;
-
-   if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
-      /* In this case, could also use GART texturing.  This is
-       * currently disabled, but has been tested & works.
-       */
-      if ( !t->image_override )
-         t->pp_txoffset = r200GartOffsetFromVirtual( rmesa, texImage->Data );
-      t->pp_txpitch = texImage->RowStride * texFormat->TexelBytes - 32;
-
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, 
-		 "Using GART texturing for rectangular client texture\n");
-
-      /* Release FB memory allocated for this image:
-       */
-      /* FIXME This may not be correct as driSwapOutTextureObject sets
-       * FIXME dirty_images.  It may be fine, though.
-       */
-      if ( t->base.memBlock ) {
-	 driSwapOutTextureObject( (driTextureObject *) t );
-      }
-   }
-   else if (texImage->IsClientData) {
-      /* Data already in GART memory, with usable pitch.
-       */
-      GLuint srcPitch;
-      srcPitch = texImage->RowStride * texFormat->TexelBytes;
-      r200EmitBlit( rmesa, 
-		    blit_format, 
-		    srcPitch,
-		    r200GartOffsetFromVirtual( rmesa, texImage->Data ),   
-		    dstPitch, t->bufAddr,
-		    0, 0, 
-		    0, 0, 
-		    width, height );
-   }
-   else {
-      /* Data not in GART memory, or bad pitch.
-       */
-      for (done = 0; done < height ; ) {
-	 struct r200_dma_region region;
-	 int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch );
-	 int src_pitch;
-	 char *tex;
-
-         src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-	 tex = (char *)texImage->Data + done * src_pitch;
-
-	 memset(&region, 0, sizeof(region));
-	 r200AllocDmaRegion( rmesa, &region, lines * dstPitch, 1024 );
-
-	 /* Copy texdata to dma:
-	  */
-	 if (0)
-	    fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n",
-		    __FUNCTION__, src_pitch, dstPitch);
-
-	 if (src_pitch == dstPitch) {
-	    memcpy( region.address + region.start, tex, lines * src_pitch );
-	 } 
-	 else {
-	    char *buf = region.address + region.start;
-	    int i;
-	    for (i = 0 ; i < lines ; i++) {
-	       memcpy( buf, tex, src_pitch );
-	       buf += dstPitch;
-	       tex += src_pitch;
-	    }
-	 }
-
-	 r200EmitWait( rmesa, RADEON_WAIT_3D );
-
-	 /* Blit to framebuffer
-	  */
-	 r200EmitBlit( rmesa,
-		       blit_format,
-		       dstPitch, GET_START( &region ),
-		       dstPitch | (t->tile_bits >> 16),
-		       t->bufAddr,
-		       0, 0,
-		       0, done,
-		       width, lines );
-	 
-	 r200EmitWait( rmesa, RADEON_WAIT_2D );
-
-	 r200ReleaseDmaRegion( rmesa, &region, __FUNCTION__ );
-	 done += lines;
-      }
-   }
-}
-
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void uploadSubImage( r200ContextPtr rmesa, r200TexObjPtr t, 
-			    GLint hwlevel,
-			    GLint x, GLint y, GLint width, GLint height,
-			    GLuint face )
-{
-   struct gl_texture_image *texImage = NULL;
-   GLuint offset;
-   GLint imageWidth, imageHeight;
-   GLint ret;
-   drm_radeon_texture_t tex;
-   drm_radeon_tex_image_t tmp;
-   const int level = hwlevel + t->base.firstLevel;
-
-   if ( R200_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n", 
-	       __FUNCTION__, (void *)t, (void *)t->base.tObj,
-	       level, width, height, face );
-   }
-
-   ASSERT(face < 6);
-
-   /* Ensure we have a valid texture to upload */
-   if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) {
-      _mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-      return;
-   }
-
-   texImage = t->base.tObj->Image[face][level];
-
-   if ( !texImage ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level );
-      return;
-   }
-   if ( !texImage->Data ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ );
-      return;
-   }
-
-
-   if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-      assert(level == 0);
-      assert(hwlevel == 0);
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__);
-      r200UploadRectSubImage( rmesa, t, texImage, x, y, width, height );
-      return;
-   }
-   else if (texImage->IsClientData) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is in GART client storage\n",
-		  __FUNCTION__);
-      r200UploadGARTClientSubImage( rmesa, t, texImage, hwlevel,
-				   x, y, width, height );
-      return;
-   }
-   else if ( R200_DEBUG & DEBUG_TEXTURE )
-      fprintf( stderr, "%s: image data is in normal memory\n",
-	       __FUNCTION__);
-      
-
-   imageWidth = texImage->Width;
-   imageHeight = texImage->Height;
-
-   offset = t->bufAddr + t->base.totalSize / 6 * face;
-
-   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      GLint imageX = 0;
-      GLint imageY = 0;
-      GLint blitX = t->image[face][hwlevel].x;
-      GLint blitY = t->image[face][hwlevel].y;
-      GLint blitWidth = t->image[face][hwlevel].width;
-      GLint blitHeight = t->image[face][hwlevel].height;
-      fprintf( stderr, "   upload image: %d,%d at %d,%d\n",
-	       imageWidth, imageHeight, imageX, imageY );
-      fprintf( stderr, "   upload  blit: %d,%d at %d,%d\n",
-	       blitWidth, blitHeight, blitX, blitY );
-      fprintf( stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-	       (GLuint)offset, hwlevel, level );
-   }
-
-   t->image[face][hwlevel].data = texImage->Data;
-
-   /* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-    * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-    * We used to use 1, 2 and 4-byte texels and used to use the texture
-    * width to dictate the blit width - but that won't work for compressed
-    * textures. (Brian)
-    * NOTE: can't do that with texture tiling. (sroland)
-    */
-   tex.offset = offset;
-   tex.image = &tmp;
-   /* copy (x,y,width,height,data) */
-   memcpy( &tmp, &t->image[face][hwlevel], sizeof(tmp) );
-   
-   if (texImage->TexFormat->TexelBytes) {
-      /* use multi-byte upload scheme */
-      tex.height = imageHeight;
-      tex.width = imageWidth;
-      tex.format = t->pp_txformat & R200_TXFORMAT_FORMAT_MASK;
-      if (tex.format == R200_TXFORMAT_ABGR8888) {
-	 /* drm will refuse abgr8888 textures. */
-	 tex.format = R200_TXFORMAT_ARGB8888;
-      }
-      tex.pitch = MAX2((texImage->Width * texImage->TexFormat->TexelBytes) / 64, 1);
-      tex.offset += tmp.x & ~1023;
-      tmp.x = tmp.x % 1024;
-      if (t->tile_bits & R200_TXO_MICRO_TILE) {
-	 /* need something like "tiled coordinates" ? */
-	 tmp.y = tmp.x / (tex.pitch * 128) * 2;
-	 tmp.x = tmp.x % (tex.pitch * 128) / 2 / texImage->TexFormat->TexelBytes;
-	 tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-      }
-      else {
-	 tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-      }
-      if ((t->tile_bits & R200_TXO_MACRO_TILE) &&
-	 (texImage->Width * texImage->TexFormat->TexelBytes >= 256) &&
-	 ((!(t->tile_bits & R200_TXO_MICRO_TILE) && (texImage->Height >= 8)) ||
-	    (texImage->Height >= 16))) {
-	 /* weird: R200 disables macro tiling if mip width is smaller than 256 bytes,
-	    OR if height is smaller than 8 automatically, but if micro tiling is active
-	    the limit is height 16 instead ? */
-	 tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-      }
-   }
-   else {
-      /* In case of for instance 8x8 texture (2x2 dxt blocks), padding after the first two blocks is
-         needed (only with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-      /* set tex.height to 1/4 since 1 "macropixel" (dxt-block) has 4 real pixels. Needed
-         so the kernel module reads the right amount of data. */
-      tex.format = R200_TXFORMAT_I8; /* any 1-byte texel format */
-      tex.pitch = (BLIT_WIDTH_BYTES / 64);
-      tex.height = (imageHeight + 3) / 4;
-      tex.width = (imageWidth + 3) / 4;
-      switch (t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) {
-      case R200_TXFORMAT_DXT1:
-           tex.width *= 8;
-           break;
-      case R200_TXFORMAT_DXT23:
-      case R200_TXFORMAT_DXT45:
-           tex.width *= 16;
-           break;
-      default:
-          fprintf(stderr, "unknown compressed tex format in uploadSubImage\n");
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
-   do {
-      ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE,
-                                 &tex, sizeof(drm_radeon_texture_t) );
-      if (ret) {
-	 if (R200_DEBUG & DEBUG_IOCTL)
-	    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
-	 usleep(1);
-      }
-   } while ( ret == -EAGAIN );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret );
-      fprintf( stderr, "   offset=0x%08x\n",
-	       offset );
-      fprintf( stderr, "   image width=%d height=%d\n",
-	       imageWidth, imageHeight );
-      fprintf( stderr, "    blit width=%d height=%d data=%p\n",
-	       t->image[face][hwlevel].width, t->image[face][hwlevel].height,
-	       t->image[face][hwlevel].data );
-      exit( 1 );
-   }
-}
-
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- * 
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face )
-{
-   const int numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-	       (void *)rmesa->glCtx, (void *)t->base.tObj, t->base.totalSize,
-	       t->base.firstLevel, t->base.lastLevel );
-   }
-
-   if ( !t || t->base.totalSize == 0 || t->image_override )
-      return 0;
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      r200Finish( rmesa->glCtx );
-   }
-
-   LOCK_HARDWARE( rmesa );
-
-   if ( t->base.memBlock == NULL ) {
-      int heap;
-
-      heap = driAllocateTexture( rmesa->texture_heaps, rmesa->nr_heaps,
-				 (driTextureObject *) t );
-      if ( heap == -1 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 return -1;
-      }
-
-      /* Set the base offset of the texture image */
-      t->bufAddr = rmesa->r200Screen->texOffset[heap] 
-	   + t->base.memBlock->ofs;
-      t->pp_txoffset = t->bufAddr;
-       
-      if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-	 /* hope it's safe to add that here... */
-	 t->pp_txoffset |= t->tile_bits;
-      }
-
-      /* Mark this texobj as dirty on all units:
-       */
-      t->dirty_state = TEX_ALL;
-   }
-
-   /* Let the world know we've used this memory recently.
-    */
-   driUpdateTextureLRU( (driTextureObject *) t );
-   UNLOCK_HARDWARE( rmesa );
-
-   /* Upload any images that are new */
-   if (t->base.dirty_images[face]) {
-      int i;
-      for ( i = 0 ; i < numLevels ; i++ ) {
-         if ( (t->base.dirty_images[face] & (1 << (i+t->base.firstLevel))) != 0 ) {
-            uploadSubImage( rmesa, t, i, 0, 0, t->image[face][i].width,
-			    t->image[face][i].height, face );
-         }
-      }
-      t->base.dirty_images[face] = 0;
-   }
-
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      r200Finish( rmesa->glCtx );
-   }
-
-   return 0;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c
index 0ad5651cd4..c94834752e 100644
--- a/src/mesa/drivers/dri/r200/r200_texstate.c
+++ b/src/mesa/drivers/dri/r200/r200_texstate.c
@@ -37,9 +37,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/macros.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "main/texobj.h"
 #include "main/enums.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -139,257 +142,6 @@ static const struct tx_table tx_table_le[] =
 #undef _ALPHA
 #undef _INVALID
 
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c pp_txfilter, \c pp_txformat, etc. will be set here
- * too.
- * 
- * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
- */
-static void r200SetTexImages( r200ContextPtr rmesa,
-			      struct gl_texture_object *tObj )
-{
-   r200TexObjPtr t = (r200TexObjPtr)tObj->DriverData;
-   const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel];
-   GLint curOffset, blitWidth;
-   GLint i, texelBytes;
-   GLint numLevels;
-   GLint log2Width, log2Height, log2Depth;
-
-   /* Set the hardware texture format
-    */
-   if ( !t->image_override ) {
-      if ( VALID_FORMAT( baseImage->TexFormat->MesaFormat ) ) {
-	 const struct tx_table *table = _mesa_little_endian() ? tx_table_le :
-								tx_table_be;
-
-         t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK |
-                             R200_TXFORMAT_ALPHA_IN_MAP);
-         t->pp_txfilter &= ~R200_YUV_TO_RGB;
-
-	 t->pp_txformat |= table[ baseImage->TexFormat->MesaFormat ].format;
-	 t->pp_txfilter |= table[ baseImage->TexFormat->MesaFormat ].filter;
-      }
-      else {
-         _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__);
-         return;
-      }
-   }
-
-   texelBytes = baseImage->TexFormat->TexelBytes;
-
-   /* Compute which mipmap levels we really want to send to the hardware.
-    */
-
-   driCalculateTextureFirstLastLevel( (driTextureObject *) t );
-   log2Width  = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-   log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-   log2Depth  = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   assert(numLevels <= RADEON_MAX_TEXTURE_LEVELS);
-
-   /* Calculate mipmap offsets and dimensions for blitting (uploading)
-    * The idea is that we lay out the mipmap levels within a block of
-    * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-    */
-   curOffset = 0;
-   blitWidth = BLIT_WIDTH_BYTES;
-   t->tile_bits = 0;
-
-   /* figure out if this texture is suitable for tiling. */
-   if (texelBytes) {
-      if (rmesa->texmicrotile  && (tObj->Target != GL_TEXTURE_RECTANGLE_NV) &&
-      /* texrect might be able to use micro tiling too in theory? */
-	 (baseImage->Height > 1)) {
-	 /* allow 32 (bytes) x 1 mip (which will use two times the space
-	 the non-tiled version would use) max if base texture is large enough */
-	 if ((numLevels == 1) ||
-	   (((baseImage->Width * texelBytes / baseImage->Height) <= 32) &&
-	       (baseImage->Width * texelBytes > 64)) ||
-	    ((baseImage->Width * texelBytes / baseImage->Height) <= 16)) {
-	    t->tile_bits |= R200_TXO_MICRO_TILE;
-	 }
-      }
-      if (tObj->Target != GL_TEXTURE_RECTANGLE_NV) {
-	 /* we can set macro tiling even for small textures, they will be untiled anyway */
-	 t->tile_bits |= R200_TXO_MACRO_TILE;
-      }
-   }
-
-   for (i = 0; i < numLevels; i++) {
-      const struct gl_texture_image *texImage;
-      GLuint size;
-
-      texImage = tObj->Image[0][i + t->base.firstLevel];
-      if ( !texImage )
-	 break;
-
-      /* find image size in bytes */
-      if (texImage->IsCompressed) {
-      /* need to calculate the size AFTER padding even though the texture is
-         submitted without padding.
-         Only handle pot textures currently - don't know if npot is even possible,
-         size calculation would certainly need (trivial) adjustments.
-         Align (and later pad) to 32byte, not sure what that 64byte blit width is
-         good for? */
-         if ((t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) == R200_TXFORMAT_DXT1) {
-            /* RGB_DXT1/RGBA_DXT1, 8 bytes per block */
-            if ((texImage->Width + 3) < 8) /* width one block */
-               size = texImage->CompressedSize * 4;
-            else if ((texImage->Width + 3) < 16)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-         }
-         else /* DXT3/5, 16 bytes per block */
-            if ((texImage->Width + 3) < 8)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-      }
-      else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-	 size = ((texImage->Width * texelBytes + 63) & ~63) * texImage->Height;
-      }
-      else if (t->tile_bits & R200_TXO_MICRO_TILE) {
-	 /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-	    though the actual offset may be different (if texture is less than
-	    32 bytes width) to the untiled case */
-	 int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-	 size = (w * ((texImage->Height + 1) / 2)) * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      else {
-	 int w = (texImage->Width * texelBytes + 31) & ~31;
-	 size = w * texImage->Height * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      assert(size > 0);
-
-      /* Align to 32-byte offset.  It is faster to do this unconditionally
-       * (no branch penalty).
-       */
-
-      curOffset = (curOffset + 0x1f) & ~0x1f;
-
-      if (texelBytes) {
-	 t->image[0][i].x = curOffset; /* fix x and y coords up later together with offset */
-	 t->image[0][i].y = 0;
-	 t->image[0][i].width = MIN2(size / texelBytes, blitWidth);
-	 t->image[0][i].height = (size / texelBytes) / t->image[0][i].width;
-      }
-      else {
-         t->image[0][i].x = curOffset % BLIT_WIDTH_BYTES;
-         t->image[0][i].y = curOffset / BLIT_WIDTH_BYTES;
-         t->image[0][i].width  = MIN2(size, BLIT_WIDTH_BYTES);
-         t->image[0][i].height = size / t->image[0][i].width;     
-      }
-
-#if 0
-      /* for debugging only and only  applicable to non-rectangle targets */
-      assert(size % t->image[0][i].width == 0);
-      assert(t->image[0][i].x == 0
-             || (size < BLIT_WIDTH_BYTES && t->image[0][i].height == 1));
-#endif
-
-      if (0)
-         fprintf(stderr,
-                 "level %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-                 i, texImage->Width, texImage->Height,
-                 t->image[0][i].x, t->image[0][i].y,
-                 t->image[0][i].width, t->image[0][i].height, size, curOffset);
-
-      curOffset += size;
-
-   }
-
-   /* Align the total size of texture memory block.
-    */
-   t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-   /* Setup remaining cube face blits, if needed */
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      const GLuint faceSize = t->base.totalSize;
-      GLuint face;
-      /* reuse face 0 x/y/width/height - just update the offset when uploading */
-      for (face = 1; face < 6; face++) {
-         for (i = 0; i < numLevels; i++) {
-            t->image[face][i].x =  t->image[0][i].x;
-            t->image[face][i].y =  t->image[0][i].y;
-            t->image[face][i].width  = t->image[0][i].width;
-            t->image[face][i].height = t->image[0][i].height;
-         }
-      }
-      t->base.totalSize = 6 * faceSize; /* total texmem needed */
-   }
-
-
-   /* Hardware state:
-    */
-   t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK;
-   t->pp_txfilter |= (numLevels - 1) << R200_MAX_MIP_LEVEL_SHIFT;
-
-   t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK |
-		       R200_TXFORMAT_HEIGHT_MASK |
-                       R200_TXFORMAT_CUBIC_MAP_ENABLE |
-                       R200_TXFORMAT_F5_WIDTH_MASK |
-                       R200_TXFORMAT_F5_HEIGHT_MASK);
-   t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) |
-		      (log2Height << R200_TXFORMAT_HEIGHT_SHIFT));
-
-   t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK);
-   if (tObj->Target == GL_TEXTURE_3D) {
-      t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT);
-      t->pp_txformat_x |= R200_TEXCOORD_VOLUME;
-   }
-   else if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      ASSERT(log2Width == log2Height);
-      t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) |
-                         (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) |
-/* don't think we need this bit, if it exists at all - fglrx does not set it */
-                         (R200_TXFORMAT_CUBIC_MAP_ENABLE));
-      t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV;
-      t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_1_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_2_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_2_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_3_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_3_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_4_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_4_SHIFT));
-   }
-   else {
-      /* If we don't in fact send enough texture coordinates, q will be 1,
-       * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
-       */
-      t->pp_txformat_x |= R200_TEXCOORD_PROJ;
-   }
-
-   t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) |
-                   ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16));
-
-   /* Only need to round to nearest 32 for textures, but the blitter
-    * requires 64-byte aligned pitches, and we may/may not need the
-    * blitter.   NPOT only!
-    */
-   if ( !t->image_override ) {
-      if (baseImage->IsCompressed)
-         t->pp_txpitch = (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-      else
-         t->pp_txpitch = ((tObj->Image[0][t->base.firstLevel]->Width * texelBytes) + 63) & ~(63);
-      t->pp_txpitch -= 32;
-   }
-
-   t->dirty_state = TEX_ALL;
-
-   /* FYI: r200UploadTexImages( rmesa, t ) used to be called here */
-}
-
-
-
 /* ================================================================
  * Texture combine functions
  */
@@ -569,7 +321,7 @@ static GLboolean r200UpdateTextureEnv( GLcontext *ctx, int unit, int slot, GLuin
    assert( (texUnit->_ReallyEnabled == 0)
 	   || (texUnit->_Current != NULL) );
 
-   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+   if ( R200_DEBUG & RADEON_TEXTURE ) {
       fprintf( stderr, "%s( %p, %d )\n", __FUNCTION__, (void *)ctx, unit );
    }
 
@@ -981,20 +733,19 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 {
 	r200ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
-	    _mesa_lookup_texture(rmesa->glCtx, texname);
-	r200TexObjPtr t;
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 
 	if (!tObj)
 		return;
 
-	t = (r200TexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
 
-	t->pp_txoffset = offset;
+	t->bo = NULL;
+	t->override_offset = offset;
 	t->pp_txpitch = pitch - 32;
 
 	switch (depth) {
@@ -1014,6 +765,125 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 	}
 }
 
+void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+		       __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r200ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = tx_table_le[MESA_FORMAT_RGB888].format;
+		else
+			t->pp_txformat = tx_table_le[MESA_FORMAT_ARGB8888].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_ARGB8888].filter;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = tx_table_le[MESA_FORMAT_RGB888].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB888].filter;
+		break;
+	case 2:
+		t->pp_txformat = tx_table_le[MESA_FORMAT_RGB565].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB565].filter;
+		break;
+	}
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txformat |= R200_TXFORMAT_NON_POWER2;
+	t->pp_txpitch = pitch_val;
+        t->pp_txpitch -= 32;
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+
+void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        r200SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
+
 #define REF_COLOR 1
 #define REF_ALPHA 2
 
@@ -1207,12 +1077,43 @@ static GLboolean r200UpdateAllTexEnv( GLcontext *ctx )
                                 R200_VOLUME_FILTER_MASK)
 
 
+static void disable_tex_obj_state( r200ContextPtr rmesa, 
+				   int unit )
+{
+   
+   R200_STATECHANGE( rmesa, vtx );
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
+
+   R200_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~(R200_TEX_0_ENABLE << unit);
+   if (rmesa->radeon.TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) {
+      TCL_FALLBACK( rmesa->radeon.glCtx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
+   }
+
+   /* Actually want to keep all units less than max active texture
+    * enabled, right?  Fix this for >2 texunits.
+    */
+
+   {
+      GLuint tmp = rmesa->TexGenEnabled;
+
+      rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenNeedNormals[unit] = GL_FALSE;
+      rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit);
+
+      if (tmp != rmesa->TexGenEnabled) {
+	 rmesa->recheck_texgen[unit] = GL_TRUE;
+	 rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
+      }
+   }
+}
 static void import_tex_obj_state( r200ContextPtr rmesa,
 				  int unit,
-				  r200TexObjPtr texobj )
+				  radeonTexObjPtr texobj )
 {
 /* do not use RADEON_DB_STATE to avoid stale texture caches */
-   int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
+   GLuint *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
 
    R200_STATECHANGE( rmesa, tex[unit] );
 
@@ -1225,36 +1126,21 @@ static void import_tex_obj_state( r200ContextPtr rmesa,
    cmd[TEX_PP_TXSIZE] = texobj->pp_txsize; /* NPOT only! */
    cmd[TEX_PP_TXPITCH] = texobj->pp_txpitch; /* NPOT only! */
    cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color;
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      cmd[TEX_PP_TXOFFSET_NEWDRM] = texobj->pp_txoffset;
-   }
-   else {
-      cmd[TEX_PP_TXOFFSET_OLDDRM] = texobj->pp_txoffset;
-   }
 
-   if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
-      GLuint bytesPerFace = texobj->base.totalSize / 6;
-      ASSERT(texobj->base.totalSize % 6 == 0);
+   if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) {
+      GLuint *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
 
       R200_STATECHANGE( rmesa, cube[unit] );
       cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
-      if (rmesa->r200Screen->drmSupportsFragShader) {
+      if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
 	 /* that value is submitted twice. could change cube atom
 	    to not include that command when new drm is used */
 	 cmd[TEX_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
       }
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F1] = texobj->pp_txoffset + 1 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F2] = texobj->pp_txoffset + 2 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F3] = texobj->pp_txoffset + 3 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F4] = texobj->pp_txoffset + 4 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F5] = texobj->pp_txoffset + 5 * bytesPerFace;
    }
 
-   texobj->dirty_state &= ~(1<<unit);
 }
 
-
 static void set_texgen_matrix( r200ContextPtr rmesa, 
 			       GLuint unit,
 			       const GLfloat *s_plane,
@@ -1377,7 +1263,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    } else {
       tgcm |= R200_TEXGEN_COMP_T << (unit * 4);
    }
-
    if (texUnit->TexGenEnabled & R_BIT) {
       if (texUnit->GenR.Mode != mode)
 	 mixed_fallback = GL_TRUE;
@@ -1393,7 +1278,7 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    }
 
    if (mixed_fallback) {
-      if (R200_DEBUG & DEBUG_FALLBACKS)
+      if (R200_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback mixed texgen, 0x%x (0x%x 0x%x 0x%x 0x%x)\n",
 		 texUnit->TexGenEnabled, texUnit->GenS.Mode, texUnit->GenT.Mode,
 		 texUnit->GenR.Mode, texUnit->GenQ.Mode);
@@ -1419,7 +1304,7 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
                                                   texUnit->GenR.ObjectPlane,
                                                   texUnit->GenQ.ObjectPlane );
       if (needtgenable & (S_BIT | T_BIT)) {
-	 if (R200_DEBUG & DEBUG_FALLBACKS)
+	 if (R200_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback mixed texgen / obj plane, 0x%x\n",
 		 texUnit->TexGenEnabled);
 	 return GL_FALSE;
@@ -1447,7 +1332,7 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
                                                   texUnit->GenR.EyePlane,
                                                   texUnit->GenQ.EyePlane );
       if (needtgenable & (S_BIT | T_BIT)) {
-	 if (R200_DEBUG & DEBUG_FALLBACKS)
+	 if (R200_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback mixed texgen / eye plane, 0x%x\n",
 		 texUnit->TexGenEnabled);
 	 return GL_FALSE;
@@ -1497,7 +1382,7 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    default:
       /* Unsupported mode, fallback:
        */
-      if (R200_DEBUG & DEBUG_FALLBACKS)
+      if (R200_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback unsupported texgen, %d\n",
 		 texUnit->GenS.Mode);
       return GL_FALSE;
@@ -1517,52 +1402,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    return GL_TRUE;
 }
 
-
-static void disable_tex( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit)) {
-      /* Texture unit disabled */
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= ~(1UL << unit);
-	 rmesa->state.texture.unit[unit].texobj = NULL;
-      }
-
-      R200_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~(R200_TEX_0_ENABLE << unit);
-	 
-      R200_STATECHANGE( rmesa, vtx );
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
-	 
-      if (rmesa->TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) {
-	 TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
-      }
-
-      /* Actually want to keep all units less than max active texture
-       * enabled, right?  Fix this for >2 texunits.
-       */
-
-      {
-	 GLuint tmp = rmesa->TexGenEnabled;
-
-	 rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenNeedNormals[unit] = GL_FALSE;
-	 rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit);
-
-	 if (tmp != rmesa->TexGenEnabled) {
-	    rmesa->recheck_texgen[unit] = GL_TRUE;
-	    rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-	 }
-      }
-   }
-}
-
 void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
@@ -1579,237 +1418,169 @@ void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d )
    }
 }
 
-static GLboolean enable_tex_2d( GLcontext *ctx, int unit )
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t)
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
+   int firstlevel = t->mt ? t->mt->firstLevel : 0;
+   const struct gl_texture_image *firstImage = t->base.Image[0][firstlevel];
+   GLint log2Width, log2Height, log2Depth, texelBytes;
+   
+   if ( t->bo ) {
+       return;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
+   log2Width  = firstImage->WidthLog2;
+   log2Height = firstImage->HeightLog2;
+   log2Depth  = firstImage->DepthLog2;
+   texelBytes = firstImage->TexFormat->TexelBytes;
 
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock && !t->image_override ) 
-	 return GL_FALSE;
-   }
 
-   set_re_cntl_d3d( ctx, unit, GL_FALSE );
-
-   return GL_TRUE;
-}
-
-#if ENABLE_HW_3D_TEXTURE
-static GLboolean enable_tex_3d( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Need to load the 3d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
+   if (!t->image_override) {
+      if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+	 const struct tx_table *table = _mesa_little_endian() ? tx_table_le :
+	    tx_table_be;
+	 
+	 t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK |
+			     R200_TXFORMAT_ALPHA_IN_MAP);
+	 t->pp_txfilter &= ~R200_YUV_TO_RGB;
+	 
+	 t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format;
+	 t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter;
+      } else {
+	 _mesa_problem(NULL, "unexpected texture format in %s",
+		       __FUNCTION__);
+	 return;
+      }
    }
+   
+   t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK;
+   t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << R200_MAX_MIP_LEVEL_SHIFT;
+	
+   t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK |
+		       R200_TXFORMAT_HEIGHT_MASK |
+		       R200_TXFORMAT_CUBIC_MAP_ENABLE |
+		       R200_TXFORMAT_F5_WIDTH_MASK |
+		       R200_TXFORMAT_F5_HEIGHT_MASK);
+   t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) |
+		      (log2Height << R200_TXFORMAT_HEIGHT_SHIFT));
+   
+   t->tile_bits = 0;
+   
+   t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK);
+   if (t->base.Target == GL_TEXTURE_3D) {
+      t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT);
+      t->pp_txformat_x |= R200_TEXCOORD_VOLUME;
 
-   ASSERT(tObj->Target == GL_TEXTURE_3D);
-
-   /* R100 & R200 do not support mipmaps for 3D textures.
-    */
-   if ( (tObj->MinFilter != GL_NEAREST) && (tObj->MinFilter != GL_LINEAR) ) {
-      return GL_FALSE;
    }
-
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock ) 
-	 return GL_FALSE;
+   else if (t->base.Target == GL_TEXTURE_CUBE_MAP) {
+      ASSERT(log2Width == log2Height);
+      t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) |
+			 (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) |
+			 /* don't think we need this bit, if it exists at all - fglrx does not set it */
+			 (R200_TXFORMAT_CUBIC_MAP_ENABLE));
+      t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV;
+      t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_1_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_2_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_2_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_3_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_3_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_4_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_4_SHIFT));
    }
-
-   set_re_cntl_d3d( ctx, unit, GL_TRUE );
-
-   return GL_TRUE;
-}
-#endif
-
-static GLboolean enable_tex_cube( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-   GLuint face;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      for (face = 0; face < 6; face++)
-         t->base.dirty_images[face] = ~0;
+   else {
+      /* If we don't in fact send enough texture coordinates, q will be 1,
+       * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
+       */
+      t->pp_txformat_x |= R200_TEXCOORD_PROJ;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
+   t->pp_txsize = (((firstImage->Width - 1) << R200_PP_TX_WIDTHMASK_SHIFT)
+		   | ((firstImage->Height - 1) << R200_PP_TX_HEIGHTMASK_SHIFT));
 
-   if ( t->base.dirty_images[0] || t->base.dirty_images[1] ||
-        t->base.dirty_images[2] || t->base.dirty_images[3] ||
-        t->base.dirty_images[4] || t->base.dirty_images[5] ) {
-      /* flush */
-      R200_FIREVERTICES( rmesa );
-      /* layout memory space, once for all faces */
-      r200SetTexImages( rmesa, tObj );
-   }
-
-   /* upload (per face) */
-   for (face = 0; face < 6; face++) {
-      if (t->base.dirty_images[face]) {
-         r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, face );
-      }
-   }
-      
-   if ( !t->base.memBlock ) {
-      /* texmem alloc failed, use s/w fallback */
-      return GL_FALSE;
+   if ( !t->image_override ) {
+      if (firstImage->IsCompressed)
+         t->pp_txpitch = (firstImage->Width + 63) & ~(63);
+      else
+         t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63);
+      t->pp_txpitch -= 32;
    }
 
-   set_re_cntl_d3d( ctx, unit, GL_TRUE );
-
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_rect( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   if (!(t->pp_txformat & R200_TXFORMAT_NON_POWER2)) {
+   if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
       t->pp_txformat |= R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
-   }
-
-   ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock &&
-           !t->image_override &&
-           !rmesa->prefer_gart_client_texturing ) 
-	 return GL_FALSE;
    }
 
-   set_re_cntl_d3d( ctx, unit, GL_FALSE );
-
-   return GL_TRUE;
 }
 
-
-static GLboolean update_tex_common( GLcontext *ctx, int unit )
+static GLboolean r200_validate_texture(GLcontext *ctx, struct gl_texture_object *texObj, int unit)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Fallback if there's a texture border */
-   if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 )
-       return GL_FALSE;
-
-   /* Update state if this is a different texture object to last
-    * time.
-    */
-   if ( rmesa->state.texture.unit[unit].texobj != t ) {
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
+   radeonTexObj *t = radeon_tex_obj(texObj);
 
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= 
-	     ~(1UL << unit);
-      }
-
-      rmesa->state.texture.unit[unit].texobj = t;
-      t->base.bound |= (1UL << unit);
-      t->dirty_state |= 1<<unit;
-      driUpdateTextureLRU( (driTextureObject *) t ); /* XXX: should be locked! */
-   }
-
-
-   /* Newly enabled?
-    */
-   if ( 1|| !(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit))) {
-      R200_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit;
-
-      R200_STATECHANGE( rmesa, vtx );
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3);
+   if (!radeon_validate_texture_miptree(ctx, texObj))
+      return GL_FALSE;
 
-      rmesa->recheck_texgen[unit] = GL_TRUE;
-   }
+   r200_validate_texgen(ctx, unit);
+   /* Configure the hardware registers (more precisely, the cached version
+    * of the hardware registers). */
+   setup_hardware_state(rmesa, t);
+
+   if (texObj->Target == GL_TEXTURE_RECTANGLE_NV ||
+       texObj->Target == GL_TEXTURE_2D ||
+       texObj->Target == GL_TEXTURE_1D)
+      set_re_cntl_d3d( ctx, unit, GL_FALSE );
+   else
+      set_re_cntl_d3d( ctx, unit, GL_TRUE );
+   R200_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit;
+   
+   R200_STATECHANGE( rmesa, vtx );
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3);
 
-   if (t->dirty_state & (1<<unit)) {
-      import_tex_obj_state( rmesa, unit, t );
-   }
+   rmesa->recheck_texgen[unit] = GL_TRUE;
+   import_tex_obj_state( rmesa, unit, t );
 
    if (rmesa->recheck_texgen[unit]) {
       GLboolean fallback = !r200_validate_texgen( ctx, unit );
       TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), fallback);
       rmesa->recheck_texgen[unit] = 0;
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
-   FALLBACK( rmesa, R200_FALLBACK_BORDER_MODE, t->border_fallback );
-   return !t->border_fallback;
-}
+   t->validated = GL_TRUE;
 
+   FALLBACK( rmesa, RADEON_FALLBACK_BORDER_MODE, t->border_fallback );
 
+   return !t->border_fallback;
+}
 
-static GLboolean r200UpdateTextureUnit( GLcontext *ctx, int unit )
+static GLboolean r200UpdateTextureUnit(GLcontext *ctx, int unit)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLuint unitneeded = rmesa->state.texture.unit[unit].unitneeded;
 
-   if ( unitneeded & (TEXTURE_RECT_BIT) ) {
-      return (enable_tex_rect( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( unitneeded & (TEXTURE_1D_BIT | TEXTURE_2D_BIT) ) {
-      return (enable_tex_2d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-#if ENABLE_HW_3D_TEXTURE
-   else if ( unitneeded & (TEXTURE_3D_BIT) ) {
-      return (enable_tex_3d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-#endif
-   else if ( unitneeded & (TEXTURE_CUBE_BIT) ) {
-      return (enable_tex_cube( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( unitneeded ) {
-      return GL_FALSE;
-   }
-   else {
-      disable_tex( ctx, unit );
-      return GL_TRUE;
+   if (!unitneeded) {
+      /* disable the unit */
+     disable_tex_obj_state(rmesa, unit);
+     return GL_TRUE;
    }
+
+   if (!r200_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) {
+    _mesa_warning(ctx,
+		  "failed to validate texture for unit %d.\n",
+		  unit);
+    rmesa->state.texture.unit[unit].texobj = NULL;
+    return GL_FALSE;
+  }
+
+   rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+  return GL_TRUE;
 }
 
 
@@ -1850,11 +1621,11 @@ void r200UpdateTextureState( GLcontext *ctx )
 
    FALLBACK( rmesa, R200_FALLBACK_TEXTURE, !ok );
 
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       r200ChooseVertexState( ctx );
 
 
-   if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
+   if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
 
       /*
        * T0 hang workaround -------------
@@ -1867,7 +1638,7 @@ void r200UpdateTextureState( GLcontext *ctx )
 	 R200_STATECHANGE(rmesa, tex[1]);
 	 rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_1_ENABLE;
 	 if (!(rmesa->hw.cst.cmd[CST_PP_CNTL_X] & R200_PPX_TEX_1_ENABLE))
-	    rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
+	   rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
 	 rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] |= R200_TXFORMAT_LOOKUP_DISABLE;
       }
       else if (!ctx->ATIFragmentShader._Enabled) {
diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c
index 4ce93b5145..11405d7cae 100644
--- a/src/mesa/drivers/dri/r200/r200_vertprog.c
+++ b/src/mesa/drivers/dri/r200/r200_vertprog.c
@@ -423,7 +423,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
       ~(VERT_BIT_POS | VERT_BIT_NORMAL | VERT_BIT_COLOR0 | VERT_BIT_COLOR1 |
       VERT_BIT_FOG | VERT_BIT_TEX0 | VERT_BIT_TEX1 | VERT_BIT_TEX2 |
       VERT_BIT_TEX3 | VERT_BIT_TEX4 | VERT_BIT_TEX5)) != 0) {
-      if (R200_DEBUG & DEBUG_FALLBACKS) {
+      if (R200_DEBUG & RADEON_FALLBACKS) {
 	 fprintf(stderr, "can't handle vert prog inputs 0x%x\n",
 	    mesa_vp->Base.InputsRead);
       }
@@ -436,7 +436,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
       (1 << VERT_RESULT_FOGC) | (1 << VERT_RESULT_TEX0) | (1 << VERT_RESULT_TEX1) |
       (1 << VERT_RESULT_TEX2) | (1 << VERT_RESULT_TEX3) | (1 << VERT_RESULT_TEX4) |
       (1 << VERT_RESULT_TEX5) | (1 << VERT_RESULT_PSIZ))) != 0) {
-      if (R200_DEBUG & DEBUG_FALLBACKS) {
+      if (R200_DEBUG & RADEON_FALLBACKS) {
 	 fprintf(stderr, "can't handle vert prog outputs 0x%x\n",
 	    mesa_vp->Base.OutputsWritten);
       }
@@ -551,7 +551,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
       if (mesa_vp->Base.InputsRead & (1 << i)) {
 	 array_count++;
 	 if (array_count > 12) {
-	    if (R200_DEBUG & DEBUG_FALLBACKS) {
+	    if (R200_DEBUG & RADEON_FALLBACKS) {
 	       fprintf(stderr, "more than 12 attribs used in vert prog\n");
 	    }
 	    return GL_FALSE;
@@ -571,13 +571,13 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
    }
 
    if (!(mesa_vp->Base.OutputsWritten & (1 << VERT_RESULT_HPOS))) {
-      if (R200_DEBUG & DEBUG_FALLBACKS) {
+      if (R200_DEBUG & RADEON_FALLBACKS) {
 	 fprintf(stderr, "can't handle vert prog without position output\n");
       }
       return GL_FALSE;
    }
    if (free_inputs & 1) {
-      if (R200_DEBUG & DEBUG_FALLBACKS) {
+      if (R200_DEBUG & RADEON_FALLBACKS) {
 	 fprintf(stderr, "can't handle vert prog without position input\n");
       }
       return GL_FALSE;
@@ -1070,7 +1070,7 @@ else {
 	    mesa_vp->Base.NumTemporaries + u_temp_used;
       }
       if ((mesa_vp->Base.NumTemporaries + u_temp_used) > R200_VSF_MAX_TEMPS) {
-	 if (R200_DEBUG & DEBUG_FALLBACKS) {
+	 if (R200_DEBUG & RADEON_FALLBACKS) {
 	    fprintf(stderr, "Ran out of temps, num temps %d, us %d\n", mesa_vp->Base.NumTemporaries, u_temp_used);
 	 }
 	 return GL_FALSE;
@@ -1078,7 +1078,7 @@ else {
       u_temp_i = R200_VSF_MAX_TEMPS - 1;
       if(o_inst - vp->instr >= R200_VSF_MAX_INST) {
 	 mesa_vp->Base.NumNativeInstructions = 129;
-	 if (R200_DEBUG & DEBUG_FALLBACKS) {
+	 if (R200_DEBUG & RADEON_FALLBACKS) {
 	    fprintf(stderr, "more than 128 native instructions\n");
 	 }
 	 return GL_FALSE;
@@ -1110,9 +1110,9 @@ void r200SetupVertexProg( GLcontext *ctx ) {
    }
    /* could optimize setting up vertex progs away for non-tcl hw */
    fallback = !(vp->native && r200VertexProgUpdateParams(ctx, vp) &&
-      rmesa->r200Screen->drmSupportsVertexProgram);
+      rmesa->radeon.radeonScreen->drmSupportsVertexProgram);
    TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, fallback);
-   if (rmesa->TclFallback) return;
+   if (rmesa->radeon.TclFallback) return;
 
    R200_STATECHANGE( rmesa, vap );
    /* FIXME: fglrx sets R200_VAP_SINGLE_BUF_STATE_ENABLE too. Do we need it?
diff --git a/src/mesa/drivers/dri/r200/radeon_bo_legacy.c b/src/mesa/drivers/dri/r200/radeon_bo_legacy.c
new file mode 120000
index 0000000000..79ad050e6b
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_bo_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_bo_legacy.h b/src/mesa/drivers/dri/r200/radeon_bo_legacy.h
new file mode 120000
index 0000000000..83b0f7ffab
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_bo_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/r200/radeon_bocs_wrapper.h
new file mode 120000
index 0000000000..ca894b2443
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_bocs_wrapper.h
@@ -0,0 +1 @@
+../radeon/radeon_bocs_wrapper.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_chipset.h b/src/mesa/drivers/dri/r200/radeon_chipset.h
new file mode 120000
index 0000000000..eba99001ff
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_chipset.h
@@ -0,0 +1 @@
+../radeon/radeon_chipset.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_cmdbuf.h b/src/mesa/drivers/dri/r200/radeon_cmdbuf.h
new file mode 120000
index 0000000000..a799e1dc6d
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_cmdbuf.h
@@ -0,0 +1 @@
+../radeon/radeon_cmdbuf.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_common.c b/src/mesa/drivers/dri/r200/radeon_common.c
new file mode 120000
index 0000000000..67b19ba940
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_common.c
@@ -0,0 +1 @@
+../radeon/radeon_common.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_common.h b/src/mesa/drivers/dri/r200/radeon_common.h
new file mode 120000
index 0000000000..5bcb696a9f
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_common.h
@@ -0,0 +1 @@
+../radeon/radeon_common.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_common_context.c b/src/mesa/drivers/dri/r200/radeon_common_context.c
new file mode 120000
index 0000000000..86800f3819
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_common_context.c
@@ -0,0 +1 @@
+../radeon/radeon_common_context.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_common_context.h b/src/mesa/drivers/dri/r200/radeon_common_context.h
new file mode 120000
index 0000000000..4d66312550
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_common_context.h
@@ -0,0 +1 @@
+../radeon/radeon_common_context.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_cs_legacy.c b/src/mesa/drivers/dri/r200/radeon_cs_legacy.c
new file mode 120000
index 0000000000..006720f8a4
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_cs_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_cs_legacy.h b/src/mesa/drivers/dri/r200/radeon_cs_legacy.h
new file mode 120000
index 0000000000..a5f95e0a3d
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_cs_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_cs_space_drm.c b/src/mesa/drivers/dri/r200/radeon_cs_space_drm.c
new file mode 120000
index 0000000000..c248ea7d1a
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_cs_space_drm.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_space_drm.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_debug.c b/src/mesa/drivers/dri/r200/radeon_debug.c
new file mode 120000
index 0000000000..c98c2e074c
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_debug.c
@@ -0,0 +1 @@
+../radeon/radeon_debug.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_debug.h b/src/mesa/drivers/dri/r200/radeon_debug.h
new file mode 120000
index 0000000000..bd8aa28e89
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_debug.h
@@ -0,0 +1 @@
+../radeon/radeon_debug.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_dma.c b/src/mesa/drivers/dri/r200/radeon_dma.c
new file mode 120000
index 0000000000..43be000625
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_dma.c
@@ -0,0 +1 @@
+../radeon/radeon_dma.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_dma.h b/src/mesa/drivers/dri/r200/radeon_dma.h
new file mode 120000
index 0000000000..82e50634e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_dma.h
@@ -0,0 +1 @@
+../radeon/radeon_dma.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_fbo.c b/src/mesa/drivers/dri/r200/radeon_fbo.c
new file mode 120000
index 0000000000..0d738d8d78
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_fbo.c
@@ -0,0 +1 @@
+../radeon/radeon_fbo.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_lock.c b/src/mesa/drivers/dri/r200/radeon_lock.c
new file mode 120000
index 0000000000..af4108a8e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_lock.c
@@ -0,0 +1 @@
+../radeon/radeon_lock.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_lock.h b/src/mesa/drivers/dri/r200/radeon_lock.h
new file mode 120000
index 0000000000..64bdf94ee7
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_lock.h
@@ -0,0 +1 @@
+../radeon/radeon_lock.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_mipmap_tree.c b/src/mesa/drivers/dri/r200/radeon_mipmap_tree.c
new file mode 120000
index 0000000000..31c0cfbe94
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_mipmap_tree.c
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_mipmap_tree.h b/src/mesa/drivers/dri/r200/radeon_mipmap_tree.h
new file mode 120000
index 0000000000..254d50cf8c
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_mipmap_tree.h
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_queryobj.c b/src/mesa/drivers/dri/r200/radeon_queryobj.c
new file mode 120000
index 0000000000..1d6ebc1c48
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_queryobj.c
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_queryobj.h b/src/mesa/drivers/dri/r200/radeon_queryobj.h
new file mode 120000
index 0000000000..8f6f842b0a
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_queryobj.h
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_screen.c b/src/mesa/drivers/dri/r200/radeon_screen.c
new file mode 120000
index 0000000000..86161118dd
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_screen.c
@@ -0,0 +1 @@
+../radeon/radeon_screen.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_screen.h b/src/mesa/drivers/dri/r200/radeon_screen.h
new file mode 120000
index 0000000000..23bb6bd459
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_screen.h
@@ -0,0 +1 @@
+../radeon/radeon_screen.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_span.c b/src/mesa/drivers/dri/r200/radeon_span.c
new file mode 120000
index 0000000000..232868c4c9
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_span.c
@@ -0,0 +1 @@
+../radeon/radeon_span.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_span.h b/src/mesa/drivers/dri/r200/radeon_span.h
new file mode 120000
index 0000000000..f9d634508c
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_span.h
@@ -0,0 +1 @@
+../radeon/radeon_span.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_texture.c b/src/mesa/drivers/dri/r200/radeon_texture.c
new file mode 120000
index 0000000000..a822710915
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_texture.c
@@ -0,0 +1 @@
+../radeon/radeon_texture.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/radeon_texture.h b/src/mesa/drivers/dri/r200/radeon_texture.h
new file mode 120000
index 0000000000..17fac3d5ea
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/radeon_texture.h
@@ -0,0 +1 @@
+../radeon/radeon_texture.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon.h b/src/mesa/drivers/dri/r200/server/radeon.h
new file mode 120000
index 0000000000..81274a54f1
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon.h
@@ -0,0 +1 @@
+../../radeon/server/radeon.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon_dri.c b/src/mesa/drivers/dri/r200/server/radeon_dri.c
new file mode 120000
index 0000000000..d05847d650
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon_dri.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon_dri.h b/src/mesa/drivers/dri/r200/server/radeon_dri.h
new file mode 120000
index 0000000000..27c591d3c9
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon_dri.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon_egl.c b/src/mesa/drivers/dri/r200/server/radeon_egl.c
new file mode 120000
index 0000000000..d7735a7643
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon_egl.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_egl.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon_macros.h b/src/mesa/drivers/dri/r200/server/radeon_macros.h
new file mode 120000
index 0000000000..c56cd735b8
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon_macros.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_macros.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r200/server/radeon_reg.h b/src/mesa/drivers/dri/r200/server/radeon_reg.h
new file mode 120000
index 0000000000..e2349dcb68
--- /dev/null
+++ b/src/mesa/drivers/dri/r200/server/radeon_reg.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_reg.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/.gitignore b/src/mesa/drivers/dri/r300/.gitignore
deleted file mode 100644
index 3689a6a78e..0000000000
--- a/src/mesa/drivers/dri/r300/.gitignore
+++ /dev/null
@@ -1,4 +0,0 @@
-radeon_chipset.h
-radeon_screen.[ch]
-radeon_span.h
-server
diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 6ca934204f..5d8d6f6658 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -3,6 +3,8 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = r300_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c
@@ -11,6 +13,10 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
 COMMON_SOURCES = \
 	../../common/driverfuncs.c \
 	../common/mm.c \
@@ -20,67 +26,59 @@ COMMON_SOURCES = \
 	../common/xmlconfig.c \
 	../common/dri_util.c
 
+RADEON_COMMON_SOURCES = \
+	radeon_bo_legacy.c \
+	radeon_buffer_objects.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_cs_legacy.c \
+	radeon_dma.c \
+	radeon_debug.c \
+	radeon_fbo.c \
+	radeon_lock.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_queryobj.c \
+	radeon_texture.c
+
 DRIVER_SOURCES = \
 		 radeon_screen.c \
-		 radeon_context.c \
-		 radeon_ioctl.c \
-		 radeon_lock.c \
-		 radeon_span.c \
-		 radeon_state.c \
-		 r300_mem.c \
 		 r300_context.c \
+		 r300_draw.c \
 		 r300_ioctl.c \
 		 r300_cmdbuf.c \
 		 r300_state.c \
 		 r300_render.c \
-		 r300_texmem.c \
 		 r300_tex.c \
 		 r300_texstate.c \
-		 radeon_program.c \
-		 radeon_program_alu.c \
-		 radeon_program_pair.c \
-		 radeon_nqssadce.c \
 		 r300_vertprog.c \
-		 r300_fragprog.c \
-		 r300_fragprog_swizzle.c \
-		 r300_fragprog_emit.c \
-		 r500_fragprog.c \
-		 r500_fragprog_emit.c \
+		 r300_fragprog_common.c \
 		 r300_shader.c \
 		 r300_emit.c \
 		 r300_swtcl.c \
-		 $(EGL_SOURCES)
+		 $(RADEON_COMMON_SOURCES) \
+		 $(EGL_SOURCES) \
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
 DRIVER_DEFINES = -DCOMPILE_R300 -DR200_MERGED=0 \
-	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300
-
-SYMLINKS = \
-	server/radeon_dri.c \
-	server/radeon_dri.h \
-	server/radeon.h \
-	server/radeon_macros.h \
-	server/radeon_reg.h \
-	server/radeon_egl.c
-
-COMMON_SYMLINKS = \
-	radeon_chipset.h \
-	radeon_screen.c \
-	radeon_screen.h \
-	radeon_span.h
+	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300 \
+#	-DRADEON_BO_TRACK \
+	-Wall
+
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
+
+PIPE_DRIVERS =  compiler/libr300compiler.a
 
 ##### TARGETS #####
 
 include ../Makefile.template
 
-server:
-	mkdir -p server
-
-$(SYMLINKS): server
-	@[ -e $@ ] || ln -sf ../../radeon/$@ server/
+symlinks:
 
-$(COMMON_SYMLINKS):
-	@[ -e $@ ] || ln -sf ../radeon/$@ ./
+# Mark the archive phony so that we always check for recompilation
+.PHONY : compiler/libr300compiler.a
 
-symlinks: $(SYMLINKS) $(COMMON_SYMLINKS)
+compiler/libr300compiler.a:
+	cd compiler && $(MAKE)
diff --git a/src/mesa/drivers/dri/r300/compiler/Makefile b/src/mesa/drivers/dri/r300/compiler/Makefile
new file mode 100644
index 0000000000..d973844192
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/Makefile
@@ -0,0 +1,75 @@
+# src/mesa/drivers/dri/r300/compiler/Makefile
+
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = r300compiler
+
+C_SOURCES = \
+		radeon_code.c \
+		radeon_compiler.c \
+		radeon_nqssadce.c \
+		radeon_program.c \
+		radeon_program_alu.c \
+		radeon_program_pair.c \
+		r3xx_fragprog.c \
+		r300_fragprog.c \
+		r300_fragprog_swizzle.c \
+		r300_fragprog_emit.c \
+		r500_fragprog.c \
+		r500_fragprog_emit.c \
+		r3xx_vertprog.c \
+		r3xx_vertprog_dump.c \
+		\
+		memory_pool.c
+
+
+### Basic defines ###
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	$(CPP_SOURCES:.cpp=.o) \
+	$(ASM_SOURCES:.S=.o)
+
+INCLUDES = \
+	-I. \
+	-I$(TOP)/include \
+	-I$(TOP)/src/mesa \
+
+
+##### TARGETS #####
+
+default: depend lib$(LIBNAME).a
+
+lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/configs/current
+	$(MKLIB) -o $(LIBNAME) -static $(OBJECTS)
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) 2> /dev/null
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+# Remove .o and backup files
+clean:
+	rm -f $(OBJECTS) lib$(LIBNAME).a depend depend.bak
+
+# Dummy target
+install:
+	@echo -n ""
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(LIBRARY_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(CXXFLAGS) $(LIBRARY_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(LIBRARY_DEFINES)  $< -o $@
+
+
+sinclude depend
diff --git a/src/mesa/drivers/dri/r300/compiler/memory_pool.c b/src/mesa/drivers/dri/r300/compiler/memory_pool.c
new file mode 100644
index 0000000000..37aa2b6579
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/memory_pool.c
@@ -0,0 +1,95 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "memory_pool.h"
+
+#include <assert.h>
+#include <stdlib.h>
+#include <string.h>
+
+
+#define POOL_LARGE_ALLOC 4096
+#define POOL_ALIGN 4
+
+
+struct memory_block {
+	struct memory_block * next;
+};
+
+void memory_pool_init(struct memory_pool * pool)
+{
+	memset(pool, 0, sizeof(struct memory_pool));
+}
+
+
+void memory_pool_destroy(struct memory_pool * pool)
+{
+	while(pool->blocks) {
+		struct memory_block * block = pool->blocks;
+		pool->blocks = block->next;
+		free(block);
+	}
+}
+
+static void refill_pool(struct memory_pool * pool)
+{
+	unsigned int blocksize = pool->total_allocated;
+	struct memory_block * newblock;
+
+	if (!blocksize)
+		blocksize = 2*POOL_LARGE_ALLOC;
+
+	newblock = (struct memory_block*)malloc(blocksize);
+	newblock->next = pool->blocks;
+	pool->blocks = newblock;
+
+	pool->head = (unsigned char*)(newblock + 1);
+	pool->end = ((unsigned char*)newblock) + blocksize;
+	pool->total_allocated += blocksize;
+}
+
+
+void * memory_pool_malloc(struct memory_pool * pool, unsigned int bytes)
+{
+	if (bytes < POOL_LARGE_ALLOC) {
+		if (pool->head + bytes > pool->end)
+			refill_pool(pool);
+
+		assert(pool->head + bytes <= pool->end);
+
+		void * ptr = pool->head;
+
+		pool->head += bytes;
+		pool->head = (unsigned char*)(((unsigned long)pool->head + POOL_ALIGN - 1) & ~(POOL_ALIGN - 1));
+
+		return ptr;
+	} else {
+		struct memory_block * block = (struct memory_block*)malloc(bytes + sizeof(struct memory_block));
+
+		block->next = pool->blocks;
+		pool->blocks = block;
+
+		return (block + 1);
+	}
+}
+
+
diff --git a/src/mesa/drivers/dri/r300/compiler/memory_pool.h b/src/mesa/drivers/dri/r300/compiler/memory_pool.h
new file mode 100644
index 0000000000..ce23c319ad
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/memory_pool.h
@@ -0,0 +1,49 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef MEMORY_POOL_H
+#define MEMORY_POOL_H
+
+struct memory_block;
+
+/**
+ * Provides a pool of memory that can quickly be allocated from, at the
+ * cost of being unable to explicitly free one of the allocated blocks.
+ * Instead, the entire pool can be freed at once.
+ *
+ * The idea is to allow one to quickly allocate a flexible amount of
+ * memory during operations like shader compilation while avoiding
+ * reference counting headaches.
+ */
+struct memory_pool {
+	unsigned char * head;
+	unsigned char * end;
+	unsigned int total_allocated;
+	struct memory_block * blocks;
+};
+
+
+void memory_pool_init(struct memory_pool * pool);
+void memory_pool_destroy(struct memory_pool * pool);
+void * memory_pool_malloc(struct memory_pool * pool, unsigned int bytes);
+
+#endif /* MEMORY_POOL_H */
diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c
new file mode 100644
index 0000000000..6c9fba4914
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c
@@ -0,0 +1,416 @@
+/*
+ * Copyright (C) 2005 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "r300_fragprog.h"
+
+#include "shader/prog_parameter.h"
+
+#include "../r300_reg.h"
+
+static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int tmu)
+{
+	struct prog_src_register reg = { 0, };
+
+	reg.File = PROGRAM_STATE_VAR;
+	reg.Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_SHADOW_AMBIENT, tmu);
+	reg.Swizzle = SWIZZLE_WWWW;
+	return reg;
+}
+
+/**
+ * Transform TEX, TXP, TXB, and KIL instructions in the following way:
+ *  - premultiply texture coordinates for RECT
+ *  - extract operand swizzles
+ *  - introduce a temporary register when write masks are needed
+ */
+GLboolean r300_transform_TEX(
+	struct radeon_compiler * c,
+	struct rc_instruction* inst,
+	void* data)
+{
+	struct r300_fragment_program_compiler *compiler =
+		(struct r300_fragment_program_compiler*)data;
+
+	if (inst->I.Opcode != OPCODE_TEX &&
+	    inst->I.Opcode != OPCODE_TXB &&
+	    inst->I.Opcode != OPCODE_TXP &&
+	    inst->I.Opcode != OPCODE_KIL)
+		return GL_FALSE;
+
+	/* ARB_shadow & EXT_shadow_funcs */
+	if (inst->I.Opcode != OPCODE_KIL &&
+	    c->Program.ShadowSamplers & (1 << inst->I.TexSrcUnit)) {
+		GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func;
+
+		if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) {
+			inst->I.Opcode = OPCODE_MOV;
+
+			if (comparefunc == GL_ALWAYS) {
+				inst->I.SrcReg[0].File = PROGRAM_BUILTIN;
+				inst->I.SrcReg[0].Swizzle = SWIZZLE_1111;
+			} else {
+				inst->I.SrcReg[0] = shadow_ambient(c, inst->I.TexSrcUnit);
+			}
+
+			return GL_TRUE;
+		} else {
+			GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func;
+			GLuint depthmode = compiler->state.unit[inst->I.TexSrcUnit].depth_texture_mode;
+			struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, inst);
+			struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_rcp);
+			struct rc_instruction * inst_cmp = rc_insert_new_instruction(c, inst_mad);
+			int pass, fail;
+
+			inst_rcp->I.Opcode = OPCODE_RCP;
+			inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_rcp->I.DstReg.Index = rc_find_free_temporary(c);
+			inst_rcp->I.DstReg.WriteMask = WRITEMASK_W;
+			inst_rcp->I.SrcReg[0] = inst->I.SrcReg[0];
+			inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW;
+
+			inst_cmp->I.DstReg = inst->I.DstReg;
+			inst->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst->I.DstReg.Index = rc_find_free_temporary(c);
+			inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
+
+			inst_mad->I.Opcode = OPCODE_MAD;
+			inst_mad->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_mad->I.DstReg.Index = rc_find_free_temporary(c);
+			inst_mad->I.SrcReg[0] = inst->I.SrcReg[0];
+			inst_mad->I.SrcReg[0].Swizzle = SWIZZLE_ZZZZ;
+			inst_mad->I.SrcReg[1].File = PROGRAM_TEMPORARY;
+			inst_mad->I.SrcReg[1].Index = inst_rcp->I.DstReg.Index;
+			inst_mad->I.SrcReg[1].Swizzle = SWIZZLE_WWWW;
+			inst_mad->I.SrcReg[2].File = PROGRAM_TEMPORARY;
+			inst_mad->I.SrcReg[2].Index = inst->I.DstReg.Index;
+			if (depthmode == 0) /* GL_LUMINANCE */
+				inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z);
+			else if (depthmode == 2) /* GL_ALPHA */
+				inst_mad->I.SrcReg[2].Swizzle = SWIZZLE_WWWW;
+
+			/* Recall that SrcReg[0] is tex, SrcReg[2] is r and:
+			 *   r  < tex  <=>      -tex+r < 0
+			 *   r >= tex  <=> not (-tex+r < 0 */
+			if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
+				inst_mad->I.SrcReg[2].Negate = inst_mad->I.SrcReg[2].Negate ^ NEGATE_XYZW;
+			else
+				inst_mad->I.SrcReg[0].Negate = inst_mad->I.SrcReg[0].Negate ^ NEGATE_XYZW;
+
+			inst_cmp->I.Opcode = OPCODE_CMP;
+			/* DstReg has been filled out above */
+			inst_cmp->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+			inst_cmp->I.SrcReg[0].Index = inst_mad->I.DstReg.Index;
+
+			if (comparefunc == GL_LESS || comparefunc == GL_GREATER) {
+				pass = 1;
+				fail = 2;
+			} else {
+				pass = 2;
+				fail = 1;
+			}
+
+			inst_cmp->I.SrcReg[pass].File = PROGRAM_BUILTIN;
+			inst_cmp->I.SrcReg[pass].Swizzle = SWIZZLE_1111;
+			inst_cmp->I.SrcReg[fail] = shadow_ambient(c, inst->I.TexSrcUnit);
+		}
+	}
+
+	/* Hardware uses [0..1]x[0..1] range for rectangle textures
+	 * instead of [0..Width]x[0..Height].
+	 * Add a scaling instruction.
+	 */
+	if (inst->I.Opcode != OPCODE_KIL && inst->I.TexSrcTarget == TEXTURE_RECT_INDEX) {
+		struct rc_instruction * inst_mul = rc_insert_new_instruction(c, inst->Prev);
+
+		inst_mul->I.Opcode = OPCODE_MUL;
+		inst_mul->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst_mul->I.DstReg.Index = rc_find_free_temporary(c);
+		inst_mul->I.SrcReg[0] = inst->I.SrcReg[0];
+		inst_mul->I.SrcReg[1].File = PROGRAM_STATE_VAR;
+		inst_mul->I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_TEXRECT_FACTOR, inst->I.TexSrcUnit);
+
+		reset_srcreg(&inst->I.SrcReg[0]);
+		inst->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+		inst->I.SrcReg[0].Index = inst_mul->I.DstReg.Index;
+	}
+
+	/* Cannot write texture to output registers or with masks */
+	if (inst->I.Opcode != OPCODE_KIL &&
+	    (inst->I.DstReg.File != PROGRAM_TEMPORARY || inst->I.DstReg.WriteMask != WRITEMASK_XYZW)) {
+		struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst);
+
+		inst_mov->I.Opcode = OPCODE_MOV;
+		inst_mov->I.DstReg = inst->I.DstReg;
+		inst_mov->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+		inst_mov->I.SrcReg[0].Index = rc_find_free_temporary(c);
+
+		inst->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index;
+		inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
+	}
+
+
+	/* Cannot read texture coordinate from constants file */
+	if (inst->I.SrcReg[0].File != PROGRAM_TEMPORARY && inst->I.SrcReg[0].File != PROGRAM_INPUT) {
+		struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev);
+
+		inst_mov->I.Opcode = OPCODE_MOV;
+		inst_mov->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst_mov->I.DstReg.Index = rc_find_free_temporary(c);
+		inst_mov->I.SrcReg[0] = inst->I.SrcReg[0];
+
+		reset_srcreg(&inst->I.SrcReg[0]);
+		inst->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+		inst->I.SrcReg[0].Index = inst_mov->I.DstReg.Index;
+	}
+
+	return GL_TRUE;
+}
+
+/* just some random things... */
+void r300FragmentProgramDump(struct rX00_fragment_program_code *c)
+{
+	struct r300_fragment_program_code *code = &c->code.r300;
+	int n, i, j;
+	static int pc = 0;
+
+	fprintf(stderr, "pc=%d*************************************\n", pc++);
+
+	fprintf(stderr, "Hardware program\n");
+	fprintf(stderr, "----------------\n");
+
+	for (n = 0; n <= (code->config & 3); n++) {
+		uint32_t code_addr = code->code_addr[3 - (code->config & 3) + n];
+		int alu_offset = (code_addr & R300_ALU_START_MASK) >> R300_ALU_START_SHIFT;
+		int alu_end = (code_addr & R300_ALU_SIZE_MASK) >> R300_ALU_SIZE_SHIFT;
+		int tex_offset = (code_addr & R300_TEX_START_MASK) >> R300_TEX_START_SHIFT;
+		int tex_end = (code_addr & R300_TEX_SIZE_MASK) >> R300_TEX_SIZE_SHIFT;
+
+		fprintf(stderr, "NODE %d: alu_offset: %d, tex_offset: %d, "
+			"alu_end: %d, tex_end: %d  (code_addr: %08x)\n", n,
+			alu_offset, tex_offset, alu_end, tex_end, code_addr);
+
+		if (n > 0 || (code->config & R300_PFS_CNTL_FIRST_NODE_HAS_TEX)) {
+			fprintf(stderr, "  TEX:\n");
+			for (i = tex_offset;
+			     i <= tex_offset + tex_end;
+			     ++i) {
+				const char *instr;
+
+				switch ((code->tex.
+					 inst[i] >> R300_TEX_INST_SHIFT) &
+					15) {
+				case R300_TEX_OP_LD:
+					instr = "TEX";
+					break;
+				case R300_TEX_OP_KIL:
+					instr = "KIL";
+					break;
+				case R300_TEX_OP_TXP:
+					instr = "TXP";
+					break;
+				case R300_TEX_OP_TXB:
+					instr = "TXB";
+					break;
+				default:
+					instr = "UNKNOWN";
+				}
+
+				fprintf(stderr,
+					"    %s t%i, %c%i, texture[%i]   (%08x)\n",
+					instr,
+					(code->tex.
+					 inst[i] >> R300_DST_ADDR_SHIFT) & 31,
+					't',
+					(code->tex.
+					 inst[i] >> R300_SRC_ADDR_SHIFT) & 31,
+					(code->tex.
+					 inst[i] & R300_TEX_ID_MASK) >>
+					R300_TEX_ID_SHIFT,
+					code->tex.inst[i]);
+			}
+		}
+
+		for (i = alu_offset;
+		     i <= alu_offset + alu_end; ++i) {
+			char srcc[3][10], dstc[20];
+			char srca[3][10], dsta[20];
+			char argc[3][20];
+			char arga[3][20];
+			char flags[5], tmp[10];
+
+			for (j = 0; j < 3; ++j) {
+				int regc = code->alu.inst[i].rgb_addr >> (j * 6);
+				int rega = code->alu.inst[i].alpha_addr >> (j * 6);
+
+				sprintf(srcc[j], "%c%i",
+					(regc & 32) ? 'c' : 't', regc & 31);
+				sprintf(srca[j], "%c%i",
+					(rega & 32) ? 'c' : 't', rega & 31);
+			}
+
+			dstc[0] = 0;
+			sprintf(flags, "%s%s%s",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_REG_X) ? "x" : "",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_REG_Y) ? "y" : "",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_REG_Z) ? "z" : "");
+			if (flags[0] != 0) {
+				sprintf(dstc, "t%i.%s ",
+					(code->alu.inst[i].
+					 rgb_addr >> R300_ALU_DSTC_SHIFT) & 31,
+					flags);
+			}
+			sprintf(flags, "%s%s%s",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_OUTPUT_X) ? "x" : "",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_OUTPUT_Y) ? "y" : "",
+				(code->alu.inst[i].
+				 rgb_addr & R300_ALU_DSTC_OUTPUT_Z) ? "z" : "");
+			if (flags[0] != 0) {
+				sprintf(tmp, "o%i.%s",
+					(code->alu.inst[i].
+					 rgb_addr >> R300_ALU_DSTC_SHIFT) & 31,
+					flags);
+				strcat(dstc, tmp);
+			}
+
+			dsta[0] = 0;
+			if (code->alu.inst[i].alpha_addr & R300_ALU_DSTA_REG) {
+				sprintf(dsta, "t%i.w ",
+					(code->alu.inst[i].
+					 alpha_addr >> R300_ALU_DSTA_SHIFT) & 31);
+			}
+			if (code->alu.inst[i].alpha_addr & R300_ALU_DSTA_OUTPUT) {
+				sprintf(tmp, "o%i.w ",
+					(code->alu.inst[i].
+					 alpha_addr >> R300_ALU_DSTA_SHIFT) & 31);
+				strcat(dsta, tmp);
+			}
+			if (code->alu.inst[i].alpha_addr & R300_ALU_DSTA_DEPTH) {
+				strcat(dsta, "Z");
+			}
+
+			fprintf(stderr,
+				"%3i: xyz: %3s %3s %3s -> %-20s (%08x)\n"
+				"       w: %3s %3s %3s -> %-20s (%08x)\n", i,
+				srcc[0], srcc[1], srcc[2], dstc,
+				code->alu.inst[i].rgb_addr, srca[0], srca[1],
+				srca[2], dsta, code->alu.inst[i].alpha_addr);
+
+			for (j = 0; j < 3; ++j) {
+				int regc = code->alu.inst[i].rgb_inst >> (j * 7);
+				int rega = code->alu.inst[i].alpha_inst >> (j * 7);
+				int d;
+				char buf[20];
+
+				d = regc & 31;
+				if (d < 12) {
+					switch (d % 4) {
+					case R300_ALU_ARGC_SRC0C_XYZ:
+						sprintf(buf, "%s.xyz",
+							srcc[d / 4]);
+						break;
+					case R300_ALU_ARGC_SRC0C_XXX:
+						sprintf(buf, "%s.xxx",
+							srcc[d / 4]);
+						break;
+					case R300_ALU_ARGC_SRC0C_YYY:
+						sprintf(buf, "%s.yyy",
+							srcc[d / 4]);
+						break;
+					case R300_ALU_ARGC_SRC0C_ZZZ:
+						sprintf(buf, "%s.zzz",
+							srcc[d / 4]);
+						break;
+					}
+				} else if (d < 15) {
+					sprintf(buf, "%s.www", srca[d - 12]);
+				} else if (d == 20) {
+					sprintf(buf, "0.0");
+				} else if (d == 21) {
+					sprintf(buf, "1.0");
+				} else if (d == 22) {
+					sprintf(buf, "0.5");
+				} else if (d >= 23 && d < 32) {
+					d -= 23;
+					switch (d / 3) {
+					case 0:
+						sprintf(buf, "%s.yzx",
+							srcc[d % 3]);
+						break;
+					case 1:
+						sprintf(buf, "%s.zxy",
+							srcc[d % 3]);
+						break;
+					case 2:
+						sprintf(buf, "%s.Wzy",
+							srcc[d % 3]);
+						break;
+					}
+				} else {
+					sprintf(buf, "%i", d);
+				}
+
+				sprintf(argc[j], "%s%s%s%s",
+					(regc & 32) ? "-" : "",
+					(regc & 64) ? "|" : "",
+					buf, (regc & 64) ? "|" : "");
+
+				d = rega & 31;
+				if (d < 9) {
+					sprintf(buf, "%s.%c", srcc[d / 3],
+						'x' + (char)(d % 3));
+				} else if (d < 12) {
+					sprintf(buf, "%s.w", srca[d - 9]);
+				} else if (d == 16) {
+					sprintf(buf, "0.0");
+				} else if (d == 17) {
+					sprintf(buf, "1.0");
+				} else if (d == 18) {
+					sprintf(buf, "0.5");
+				} else {
+					sprintf(buf, "%i", d);
+				}
+
+				sprintf(arga[j], "%s%s%s%s",
+					(rega & 32) ? "-" : "",
+					(rega & 64) ? "|" : "",
+					buf, (rega & 64) ? "|" : "");
+			}
+
+			fprintf(stderr, "     xyz: %8s %8s %8s    op: %08x\n"
+				"       w: %8s %8s %8s    op: %08x\n",
+				argc[0], argc[1], argc[2],
+				code->alu.inst[i].rgb_inst, arga[0], arga[1],
+				arga[2], code->alu.inst[i].alpha_inst);
+		}
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h
new file mode 100644
index 0000000000..0ac46dbd9c
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h
@@ -0,0 +1,49 @@
+/*
+ * Copyright (C) 2005 Ben Skeggs.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/*
+ * Authors:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ *   Jerome Glisse <j.glisse@gmail.com>
+ */
+#ifndef __R300_FRAGPROG_H_
+#define __R300_FRAGPROG_H_
+
+#include "shader/program.h"
+#include "shader/prog_instruction.h"
+
+#include "radeon_compiler.h"
+#include "radeon_program.h"
+
+
+extern void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
+
+extern void r300FragmentProgramDump(struct rX00_fragment_program_code *c);
+
+extern GLboolean r300_transform_TEX(struct radeon_compiler * c, struct rc_instruction* inst, void* data);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c
index 9f0b7e3534..305dc074ee 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c
@@ -40,57 +40,43 @@
 
 #include "r300_fragprog.h"
 
+#include "../r300_reg.h"
+
 #include "radeon_program_pair.h"
 #include "r300_fragprog_swizzle.h"
-#include "r300_reg.h"
 
 
+struct r300_emit_state {
+	struct r300_fragment_program_compiler * compiler;
+
+	unsigned current_node : 2;
+	unsigned node_first_tex : 8;
+	unsigned node_first_alu : 8;
+	uint32_t node_flags;
+};
+
 #define PROG_CODE \
-	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
-	struct r300_fragment_program_code *code = c->code
+	struct r300_emit_state * emit = (struct r300_emit_state*)data; \
+	struct r300_fragment_program_compiler *c = emit->compiler; \
+	struct r300_fragment_program_code *code = &c->code->code.r300
 
 #define error(fmt, args...) do {			\
-		fprintf(stderr, "%s::%s(): " fmt "\n",	\
+		rc_error(&c->Base, "%s::%s(): " fmt "\n",	\
 			__FILE__, __FUNCTION__, ##args);	\
 	} while(0)
 
 
-static GLboolean emit_const(void* data, GLuint file, GLuint index, GLuint *hwindex)
-{
-	PROG_CODE;
-
-	for (*hwindex = 0; *hwindex < code->const_nr; ++*hwindex) {
-		if (code->constant[*hwindex].File == file &&
-		    code->constant[*hwindex].Index == index)
-			break;
-	}
-
-	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
-			error("Out of hw constants!\n");
-			return GL_FALSE;
-		}
-
-		code->const_nr++;
-		code->constant[*hwindex].File = file;
-		code->constant[*hwindex].Index = index;
-	}
-
-	return GL_TRUE;
-}
-
-
 /**
  * Mark a temporary register as used.
  */
 static void use_temporary(struct r300_fragment_program_code *code, GLuint index)
 {
-	if (index > code->max_temp_idx)
-		code->max_temp_idx = index;
+	if (index > code->pixsize)
+		code->pixsize = index;
 }
 
 
-static GLuint translate_rgb_opcode(GLuint opcode)
+static GLuint translate_rgb_opcode(struct r300_fragment_program_compiler * c, GLuint opcode)
 {
 	switch(opcode) {
 	case OPCODE_CMP: return R300_ALU_OUTC_CMP;
@@ -109,7 +95,7 @@ static GLuint translate_rgb_opcode(GLuint opcode)
 	}
 }
 
-static GLuint translate_alpha_opcode(GLuint opcode)
+static GLuint translate_alpha_opcode(struct r300_fragment_program_compiler * c, GLuint opcode)
 {
 	switch(opcode) {
 	case OPCODE_CMP: return R300_ALU_OUTA_CMP;
@@ -138,70 +124,69 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->alu.length >= PFS_MAX_ALU_INST) {
+	if (code->alu.length >= R300_PFS_MAX_ALU_INST) {
 		error("Too many ALU instructions");
 		return GL_FALSE;
 	}
 
 	int ip = code->alu.length++;
 	int j;
-	code->node[code->cur_node].alu_end++;
 
-	code->alu.inst[ip].inst0 = translate_rgb_opcode(inst->RGB.Opcode);
-	code->alu.inst[ip].inst2 = translate_alpha_opcode(inst->Alpha.Opcode);
+	code->alu.inst[ip].rgb_inst = translate_rgb_opcode(c, inst->RGB.Opcode);
+	code->alu.inst[ip].alpha_inst = translate_alpha_opcode(c, inst->Alpha.Opcode);
 
 	for(j = 0; j < 3; ++j) {
 		GLuint src = inst->RGB.Src[j].Index | (inst->RGB.Src[j].Constant << 5);
 		if (!inst->RGB.Src[j].Constant)
 			use_temporary(code, inst->RGB.Src[j].Index);
-		code->alu.inst[ip].inst1 |= src << (6*j);
+		code->alu.inst[ip].rgb_addr |= src << (6*j);
 
 		src = inst->Alpha.Src[j].Index | (inst->Alpha.Src[j].Constant << 5);
 		if (!inst->Alpha.Src[j].Constant)
 			use_temporary(code, inst->Alpha.Src[j].Index);
-		code->alu.inst[ip].inst3 |= src << (6*j);
+		code->alu.inst[ip].alpha_addr |= src << (6*j);
 
 		GLuint arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle);
 		arg |= inst->RGB.Arg[j].Abs << 6;
 		arg |= inst->RGB.Arg[j].Negate << 5;
-		code->alu.inst[ip].inst0 |= arg << (7*j);
+		code->alu.inst[ip].rgb_inst |= arg << (7*j);
 
 		arg = r300FPTranslateAlphaSwizzle(inst->Alpha.Arg[j].Source, inst->Alpha.Arg[j].Swizzle);
 		arg |= inst->Alpha.Arg[j].Abs << 6;
 		arg |= inst->Alpha.Arg[j].Negate << 5;
-		code->alu.inst[ip].inst2 |= arg << (7*j);
+		code->alu.inst[ip].alpha_inst |= arg << (7*j);
 	}
 
 	if (inst->RGB.Saturate)
-		code->alu.inst[ip].inst0 |= R300_ALU_OUTC_CLAMP;
+		code->alu.inst[ip].rgb_inst |= R300_ALU_OUTC_CLAMP;
 	if (inst->Alpha.Saturate)
-		code->alu.inst[ip].inst2 |= R300_ALU_OUTA_CLAMP;
+		code->alu.inst[ip].alpha_inst |= R300_ALU_OUTA_CLAMP;
 
 	if (inst->RGB.WriteMask) {
 		use_temporary(code, inst->RGB.DestIndex);
-		code->alu.inst[ip].inst1 |=
+		code->alu.inst[ip].rgb_addr |=
 			(inst->RGB.DestIndex << R300_ALU_DSTC_SHIFT) |
 			(inst->RGB.WriteMask << R300_ALU_DSTC_REG_MASK_SHIFT);
 	}
 	if (inst->RGB.OutputWriteMask) {
-		code->alu.inst[ip].inst1 |= (inst->RGB.OutputWriteMask << R300_ALU_DSTC_OUTPUT_MASK_SHIFT);
-		code->node[code->cur_node].flags |= R300_RGBA_OUT;
+		code->alu.inst[ip].rgb_addr |= (inst->RGB.OutputWriteMask << R300_ALU_DSTC_OUTPUT_MASK_SHIFT);
+		emit->node_flags |= R300_RGBA_OUT;
 	}
 
 	if (inst->Alpha.WriteMask) {
 		use_temporary(code, inst->Alpha.DestIndex);
-		code->alu.inst[ip].inst3 |=
+		code->alu.inst[ip].alpha_addr |=
 			(inst->Alpha.DestIndex << R300_ALU_DSTA_SHIFT) |
 			R300_ALU_DSTA_REG;
 	}
 	if (inst->Alpha.OutputWriteMask) {
-		code->alu.inst[ip].inst3 |= R300_ALU_DSTA_OUTPUT;
-		code->node[code->cur_node].flags |= R300_RGBA_OUT;
+		code->alu.inst[ip].alpha_addr |= R300_ALU_DSTA_OUTPUT;
+		emit->node_flags |= R300_RGBA_OUT;
 	}
 	if (inst->Alpha.DepthWriteMask) {
-		code->alu.inst[ip].inst3 |= R300_ALU_DSTA_DEPTH;
-		code->node[code->cur_node].flags |= R300_W_OUT;
-		c->fp->WritesDepth = GL_TRUE;
+		code->alu.inst[ip].alpha_addr |= R300_ALU_DSTA_DEPTH;
+		emit->node_flags |= R300_W_OUT;
+		c->code->writes_depth = GL_TRUE;
 	}
 
 	return GL_TRUE;
@@ -211,31 +196,50 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 /**
  * Finish the current node without advancing to the next one.
  */
-static GLboolean finish_node(struct r300_fragment_program_compiler *c)
+static GLboolean finish_node(struct r300_emit_state * emit)
 {
-	struct r300_fragment_program_code *code = c->code;
-	struct r300_fragment_program_node *node = &code->node[code->cur_node];
+	struct r300_fragment_program_compiler * c = emit->compiler;
+	struct r300_fragment_program_code *code = &emit->compiler->code->code.r300;
 
-	if (node->alu_end < 0) {
+	if (code->alu.length == emit->node_first_alu) {
 		/* Generate a single NOP for this node */
 		struct radeon_pair_instruction inst;
 		_mesa_bzero(&inst, sizeof(inst));
-		if (!emit_alu(c, &inst))
+		if (!emit_alu(emit, &inst))
 			return GL_FALSE;
 	}
 
-	if (node->tex_end < 0) {
-		if (code->cur_node == 0) {
-			node->tex_end = 0;
-		} else {
-			error("Node %i has no TEX instructions", code->cur_node);
+	unsigned alu_offset = emit->node_first_alu;
+	unsigned alu_end = code->alu.length - alu_offset - 1;
+	unsigned tex_offset = emit->node_first_tex;
+	unsigned tex_end = code->tex.length - tex_offset - 1;
+
+	if (code->tex.length == emit->node_first_tex) {
+		if (emit->current_node > 0) {
+			error("Node %i has no TEX instructions", emit->current_node);
 			return GL_FALSE;
 		}
+
+		tex_end = 0;
 	} else {
-		if (code->cur_node == 0)
-			code->first_node_has_tex = 1;
+		if (emit->current_node == 0)
+			code->config |= R300_PFS_CNTL_FIRST_NODE_HAS_TEX;
 	}
 
+	/* Write the config register.
+	 * Note: The order in which the words for each node are written
+	 * is not correct here and needs to be fixed up once we're entirely
+	 * done
+	 *
+	 * Also note that the register specification from AMD is slightly
+	 * incorrect in its description of this register. */
+	code->code_addr[emit->current_node] =
+			(alu_offset << R300_ALU_START_SHIFT) |
+			(alu_end << R300_ALU_SIZE_SHIFT) |
+			(tex_offset << R300_TEX_START_SHIFT) |
+			(tex_end << R300_TEX_SIZE_SHIFT) |
+			emit->node_flags;
+
 	return GL_TRUE;
 }
 
@@ -248,64 +252,61 @@ static GLboolean begin_tex(void* data)
 {
 	PROG_CODE;
 
-	if (code->cur_node == 0) {
-		if (code->node[0].alu_end < 0 &&
-		    code->node[0].tex_end < 0)
-			return GL_TRUE;
+	if (code->alu.length == emit->node_first_alu &&
+	    code->tex.length == emit->node_first_tex) {
+		return GL_TRUE;
 	}
 
-	if (code->cur_node == 3) {
+	if (emit->current_node == 3) {
 		error("Too many texture indirections");
 		return GL_FALSE;
 	}
 
-	if (!finish_node(c))
+	if (!finish_node(emit))
 		return GL_FALSE;
 
-	struct r300_fragment_program_node *node = &code->node[++code->cur_node];
-	node->alu_offset = code->alu.length;
-	node->alu_end = -1;
-	node->tex_offset = code->tex.length;
-	node->tex_end = -1;
+	emit->current_node++;
+	emit->node_first_tex = code->tex.length;
+	emit->node_first_alu = code->alu.length;
+	emit->node_flags = 0;
 	return GL_TRUE;
 }
 
 
-static GLboolean emit_tex(void* data, struct prog_instruction* inst)
+static GLboolean emit_tex(void* data, struct radeon_pair_texture_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->tex.length >= PFS_MAX_TEX_INST) {
+	if (code->tex.length >= R300_PFS_MAX_TEX_INST) {
 		error("Too many TEX instructions");
 		return GL_FALSE;
 	}
 
 	GLuint unit = inst->TexSrcUnit;
-	GLuint dest = inst->DstReg.Index;
+	GLuint dest = inst->DestIndex;
 	GLuint opcode;
 
 	switch(inst->Opcode) {
-	case OPCODE_KIL: opcode = R300_TEX_OP_KIL; break;
-	case OPCODE_TEX: opcode = R300_TEX_OP_LD; break;
-	case OPCODE_TXB: opcode = R300_TEX_OP_TXB; break;
-	case OPCODE_TXP: opcode = R300_TEX_OP_TXP; break;
+	case RADEON_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break;
+	case RADEON_OPCODE_TEX: opcode = R300_TEX_OP_LD; break;
+	case RADEON_OPCODE_TXB: opcode = R300_TEX_OP_TXB; break;
+	case RADEON_OPCODE_TXP: opcode = R300_TEX_OP_TXP; break;
 	default:
 		error("Unknown texture opcode %i", inst->Opcode);
 		return GL_FALSE;
 	}
 
-	if (inst->Opcode == OPCODE_KIL) {
+	if (inst->Opcode == RADEON_OPCODE_KIL) {
 		unit = 0;
 		dest = 0;
 	} else {
 		use_temporary(code, dest);
 	}
 
-	use_temporary(code, inst->SrcReg[0].Index);
+	use_temporary(code, inst->SrcIndex);
 
-	code->node[code->cur_node].tex_end++;
 	code->tex.inst[code->tex.length++] =
-		(inst->SrcReg[0].Index << R300_SRC_ADDR_SHIFT) |
+		(inst->SrcIndex << R300_SRC_ADDR_SHIFT) |
 		(dest << R300_DST_ADDR_SHIFT) |
 		(unit << R300_TEX_ID_SHIFT) |
 		(opcode << R300_TEX_INST_SHIFT);
@@ -314,31 +315,46 @@ static GLboolean emit_tex(void* data, struct prog_instruction* inst)
 
 
 static const struct radeon_pair_handler pair_handler = {
-	.EmitConst = &emit_const,
 	.EmitPaired = &emit_alu,
 	.EmitTex = &emit_tex,
 	.BeginTexBlock = &begin_tex,
-	.MaxHwTemps = PFS_NUM_TEMP_REGS
+	.MaxHwTemps = R300_PFS_NUM_TEMP_REGS
 };
 
 /**
  * Final compilation step: Turn the intermediate radeon_program into
  * machine-readable instructions.
  */
-GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler)
+void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r300_fragment_program_code *code = compiler->code;
-
-	_mesa_bzero(code, sizeof(struct r300_fragment_program_code));
-	code->node[0].alu_end = -1;
-	code->node[0].tex_end = -1;
+	struct r300_emit_state emit;
+	struct r300_fragment_program_code *code = &compiler->code->code.r300;
 
-	if (!radeonPairProgram(compiler->r300->radeon.glCtx, compiler->program, &pair_handler, compiler))
-		return GL_FALSE;
+	memset(&emit, 0, sizeof(emit));
+	emit.compiler = compiler;
 
-	if (!finish_node(compiler))
-		return GL_FALSE;
+	_mesa_bzero(code, sizeof(struct r300_fragment_program_code));
 
-	return GL_TRUE;
+	radeonPairProgram(compiler, &pair_handler, &emit);
+	if (compiler->Base.Error)
+		return;
+
+	/* Finish the program */
+	finish_node(&emit);
+
+	code->config |= emit.current_node; /* FIRST_NODE_HAS_TEX set by finish_node */
+	code->code_offset =
+		(0 << R300_PFS_CNTL_ALU_OFFSET_SHIFT) |
+		((code->alu.length-1) << R300_PFS_CNTL_ALU_END_SHIFT) |
+		(0 << R300_PFS_CNTL_TEX_OFFSET_SHIFT) |
+		((code->tex.length ? code->tex.length-1 : 0) << R300_PFS_CNTL_TEX_END_SHIFT);
+
+	if (emit.current_node < 3) {
+		int shift = 3 - emit.current_node;
+		int i;
+		for(i = 0; i <= emit.current_node; ++i)
+			code->code_addr[shift + i] = code->code_addr[i];
+		for(i = 0; i < shift; ++i)
+			code->code_addr[i] = 0;
+	}
 }
-
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c
index fc9d855bce..1b14cc3888 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c
+++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c
@@ -33,8 +33,9 @@
 
 #include "r300_fragprog_swizzle.h"
 
-#include "r300_reg.h"
+#include "../r300_reg.h"
 #include "radeon_nqssadce.h"
+#include "radeon_compiler.h"
 
 #define MAKE_SWZ3(x, y, z) (MAKE_SWIZZLE4(SWIZZLE_##x, SWIZZLE_##y, SWIZZLE_##z, SWIZZLE_ZERO))
 
@@ -174,18 +175,15 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst,
 			}
 		}
 
-		struct prog_instruction *inst;
-
-		_mesa_insert_instructions(s->Program, s->IP, 1);
-		inst = s->Program->Instructions + s->IP++;
-		inst->Opcode = OPCODE_MOV;
-		inst->DstReg = dst;
-		inst->DstReg.WriteMask &= (best_matchmask | WRITEMASK_W);
-		inst->SrcReg[0] = src;
-		inst->SrcReg[0].Negate = (best_matchmask & src.Negate) ? NEGATE_XYZW : NEGATE_NONE;
+		struct rc_instruction *inst = rc_insert_new_instruction(s->Compiler, s->IP->Prev);
+		inst->I.Opcode = OPCODE_MOV;
+		inst->I.DstReg = dst;
+		inst->I.DstReg.WriteMask &= (best_matchmask | WRITEMASK_W);
+		inst->I.SrcReg[0] = src;
+		inst->I.SrcReg[0].Negate = (best_matchmask & src.Negate) ? NEGATE_XYZW : NEGATE_NONE;
 		/* Note: We rely on NqSSA/DCE to set unused swizzle components to NIL */
 
-		dst.WriteMask &= ~inst->DstReg.WriteMask;
+		dst.WriteMask &= ~inst->I.DstReg.WriteMask;
 	}
 }
 
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.h b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h
index 231bf4eef5..231bf4eef5 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.h
+++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c
new file mode 100644
index 0000000000..76c3a7ecfd
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c
@@ -0,0 +1,149 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "radeon_compiler.h"
+
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_statevars.h"
+
+#include "radeon_nqssadce.h"
+#include "radeon_program_alu.h"
+#include "r300_fragprog.h"
+#include "r300_fragprog_swizzle.h"
+#include "r500_fragprog.h"
+
+
+static void nqssadce_init(struct nqssadce_state* s)
+{
+	struct r300_fragment_program_compiler * c = s->UserData;
+	s->Outputs[c->OutputColor].Sourced = WRITEMASK_XYZW;
+	s->Outputs[c->OutputDepth].Sourced = WRITEMASK_W;
+}
+
+static void rewrite_depth_out(struct r300_fragment_program_compiler * c)
+{
+	struct rc_instruction *rci;
+
+	for (rci = c->Base.Program.Instructions.Next; rci != &c->Base.Program.Instructions; rci = rci->Next) {
+		struct prog_instruction * inst = &rci->I;
+
+		if (inst->DstReg.File != PROGRAM_OUTPUT || inst->DstReg.Index != c->OutputDepth)
+			continue;
+
+		if (inst->DstReg.WriteMask & WRITEMASK_Z) {
+			inst->DstReg.WriteMask = WRITEMASK_W;
+		} else {
+			inst->DstReg.WriteMask = 0;
+			continue;
+		}
+
+		switch (inst->Opcode) {
+			case OPCODE_FRC:
+			case OPCODE_MOV:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				break;
+			case OPCODE_ADD:
+			case OPCODE_MAX:
+			case OPCODE_MIN:
+			case OPCODE_MUL:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				break;
+			case OPCODE_CMP:
+			case OPCODE_MAD:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
+				break;
+			default:
+				// Scalar instructions needn't be reswizzled
+				break;
+		}
+	}
+}
+
+void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c)
+{
+	rewrite_depth_out(c);
+
+	if (c->is_r500) {
+		struct radeon_program_transformation transformations[] = {
+			{ &r500_transform_TEX, c },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformDeriv, 0 },
+			{ &radeonTransformTrigScale, 0 }
+		};
+		radeonLocalTransform(&c->Base, 4, transformations);
+	} else {
+		struct radeon_program_transformation transformations[] = {
+			{ &r300_transform_TEX, c },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformTrigSimple, 0 }
+		};
+		radeonLocalTransform(&c->Base, 3, transformations);
+	}
+
+	if (c->Base.Debug) {
+		_mesa_printf("Fragment Program: After native rewrite:\n");
+		rc_print_program(&c->Base.Program);
+		fflush(stderr);
+	}
+
+	if (c->is_r500) {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r500FPIsNativeSwizzle,
+			.BuildSwizzle = &r500FPBuildSwizzle
+		};
+		radeonNqssaDce(&c->Base, &nqssadce, c);
+	} else {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
+			.BuildSwizzle = &r300FPBuildSwizzle
+		};
+		radeonNqssaDce(&c->Base, &nqssadce, c);
+	}
+
+	if (c->Base.Debug) {
+		_mesa_printf("Compiler: after NqSSA-DCE:\n");
+		rc_print_program(&c->Base.Program);
+		fflush(stderr);
+	}
+
+	if (c->is_r500) {
+		r500BuildFragmentProgramHwCode(c);
+	} else {
+		r300BuildFragmentProgramHwCode(c);
+	}
+
+	rc_constants_copy(&c->code->constants, &c->Base.Program.Constants);
+
+	if (c->Base.Debug) {
+		if (c->is_r500) {
+			r500FragmentProgramDump(c->code);
+		} else {
+			r300FragmentProgramDump(c->code);
+		}
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c
new file mode 100644
index 0000000000..93a516105e
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c
@@ -0,0 +1,655 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "radeon_compiler.h"
+
+#include "../r300_reg.h"
+
+#include "radeon_nqssadce.h"
+#include "radeon_program.h"
+#include "radeon_program_alu.h"
+
+#include "shader/prog_print.h"
+
+
+/*
+ * Take an already-setup and valid source then swizzle it appropriately to
+ * obtain a constant ZERO or ONE source.
+ */
+#define __CONST(x, y)	\
+	(PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[x]),	\
+			   t_swizzle(y),	\
+			   t_swizzle(y),	\
+			   t_swizzle(y),	\
+			   t_swizzle(y),	\
+			   t_src_class(vpi->SrcReg[x].File), \
+			   NEGATE_NONE) | (vpi->SrcReg[x].RelAddr << 4))
+
+
+static unsigned long t_dst_mask(GLuint mask)
+{
+	/* WRITEMASK_* is equivalent to VSF_FLAG_* */
+	return mask & WRITEMASK_XYZW;
+}
+
+static unsigned long t_dst_class(gl_register_file file)
+{
+
+	switch (file) {
+	case PROGRAM_TEMPORARY:
+		return PVS_DST_REG_TEMPORARY;
+	case PROGRAM_OUTPUT:
+		return PVS_DST_REG_OUT;
+	case PROGRAM_ADDRESS:
+		return PVS_DST_REG_A0;
+		/*
+		   case PROGRAM_INPUT:
+		   case PROGRAM_LOCAL_PARAM:
+		   case PROGRAM_ENV_PARAM:
+		   case PROGRAM_NAMED_PARAM:
+		   case PROGRAM_STATE_VAR:
+		   case PROGRAM_WRITE_ONLY:
+		   case PROGRAM_ADDRESS:
+		 */
+	default:
+		fprintf(stderr, "problem in %s", __FUNCTION__);
+		_mesa_exit(-1);
+		return -1;
+	}
+}
+
+static unsigned long t_dst_index(struct r300_vertex_program_code *vp,
+				 struct prog_dst_register *dst)
+{
+	if (dst->File == PROGRAM_OUTPUT)
+		return vp->outputs[dst->Index];
+
+	return dst->Index;
+}
+
+static unsigned long t_src_class(gl_register_file file)
+{
+	switch (file) {
+	case PROGRAM_TEMPORARY:
+		return PVS_SRC_REG_TEMPORARY;
+	case PROGRAM_INPUT:
+		return PVS_SRC_REG_INPUT;
+	case PROGRAM_LOCAL_PARAM:
+	case PROGRAM_ENV_PARAM:
+	case PROGRAM_NAMED_PARAM:
+	case PROGRAM_CONSTANT:
+	case PROGRAM_STATE_VAR:
+		return PVS_SRC_REG_CONSTANT;
+		/*
+		   case PROGRAM_OUTPUT:
+		   case PROGRAM_WRITE_ONLY:
+		   case PROGRAM_ADDRESS:
+		 */
+	default:
+		fprintf(stderr, "problem in %s", __FUNCTION__);
+		_mesa_exit(-1);
+		return -1;
+	}
+}
+
+static GLboolean t_src_conflict(struct prog_src_register a, struct prog_src_register b)
+{
+	unsigned long aclass = t_src_class(a.File);
+	unsigned long bclass = t_src_class(b.File);
+
+	if (aclass != bclass)
+		return GL_FALSE;
+	if (aclass == PVS_SRC_REG_TEMPORARY)
+		return GL_FALSE;
+
+	if (a.RelAddr || b.RelAddr)
+		return GL_TRUE;
+	if (a.Index != b.Index)
+		return GL_TRUE;
+
+	return GL_FALSE;
+}
+
+static INLINE unsigned long t_swizzle(GLubyte swizzle)
+{
+	/* this is in fact a NOP as the Mesa SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */
+	return swizzle;
+}
+
+static unsigned long t_src_index(struct r300_vertex_program_code *vp,
+				 struct prog_src_register *src)
+{
+	if (src->File == PROGRAM_INPUT) {
+		assert(vp->inputs[src->Index] != -1);
+		return vp->inputs[src->Index];
+	} else {
+		if (src->Index < 0) {
+			fprintf(stderr,
+				"negative offsets for indirect addressing do not work.\n");
+			return 0;
+		}
+		return src->Index;
+	}
+}
+
+/* these two functions should probably be merged... */
+
+static unsigned long t_src(struct r300_vertex_program_code *vp,
+			   struct prog_src_register *src)
+{
+	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
+	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
+	 */
+	return PVS_SRC_OPERAND(t_src_index(vp, src),
+			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 1)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 2)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 3)),
+			       t_src_class(src->File),
+			       src->Negate) | (src->RelAddr << 4);
+}
+
+static unsigned long t_src_scalar(struct r300_vertex_program_code *vp,
+				  struct prog_src_register *src)
+{
+	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
+	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
+	 */
+	return PVS_SRC_OPERAND(t_src_index(vp, src),
+			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
+			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
+			       t_src_class(src->File),
+			       src->Negate ? NEGATE_XYZW : NEGATE_NONE) |
+	    (src->RelAddr << 4);
+}
+
+static GLboolean valid_dst(struct r300_vertex_program_code *vp,
+			   struct prog_dst_register *dst)
+{
+	if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) {
+		return GL_FALSE;
+	} else if (dst->File == PROGRAM_ADDRESS) {
+		assert(dst->Index == 0);
+	}
+
+	return GL_TRUE;
+}
+
+static void ei_vector1(struct r300_vertex_program_code *vp,
+				GLuint hw_opcode,
+				struct prog_instruction *vpi,
+				GLuint * inst)
+{
+	inst[0] = PVS_OP_DST_OPERAND(hw_opcode,
+				     GL_FALSE,
+				     GL_FALSE,
+				     t_dst_index(vp, &vpi->DstReg),
+				     t_dst_mask(vpi->DstReg.WriteMask),
+				     t_dst_class(vpi->DstReg.File));
+	inst[1] = t_src(vp, &vpi->SrcReg[0]);
+	inst[2] = __CONST(0, SWIZZLE_ZERO);
+	inst[3] = __CONST(0, SWIZZLE_ZERO);
+}
+
+static void ei_vector2(struct r300_vertex_program_code *vp,
+				GLuint hw_opcode,
+				struct prog_instruction *vpi,
+				GLuint * inst)
+{
+	inst[0] = PVS_OP_DST_OPERAND(hw_opcode,
+				     GL_FALSE,
+				     GL_FALSE,
+				     t_dst_index(vp, &vpi->DstReg),
+				     t_dst_mask(vpi->DstReg.WriteMask),
+				     t_dst_class(vpi->DstReg.File));
+	inst[1] = t_src(vp, &vpi->SrcReg[0]);
+	inst[2] = t_src(vp, &vpi->SrcReg[1]);
+	inst[3] = __CONST(1, SWIZZLE_ZERO);
+}
+
+static void ei_math1(struct r300_vertex_program_code *vp,
+				GLuint hw_opcode,
+				struct prog_instruction *vpi,
+				GLuint * inst)
+{
+	inst[0] = PVS_OP_DST_OPERAND(hw_opcode,
+				     GL_TRUE,
+				     GL_FALSE,
+				     t_dst_index(vp, &vpi->DstReg),
+				     t_dst_mask(vpi->DstReg.WriteMask),
+				     t_dst_class(vpi->DstReg.File));
+	inst[1] = t_src_scalar(vp, &vpi->SrcReg[0]);
+	inst[2] = __CONST(0, SWIZZLE_ZERO);
+	inst[3] = __CONST(0, SWIZZLE_ZERO);
+}
+
+static void ei_lit(struct r300_vertex_program_code *vp,
+				      struct prog_instruction *vpi,
+				      GLuint * inst)
+{
+	//LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W}
+
+	inst[0] = PVS_OP_DST_OPERAND(ME_LIGHT_COEFF_DX,
+				     GL_TRUE,
+				     GL_FALSE,
+				     t_dst_index(vp, &vpi->DstReg),
+				     t_dst_mask(vpi->DstReg.WriteMask),
+				     t_dst_class(vpi->DstReg.File));
+	/* NOTE: Users swizzling might not work. */
+	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[0]), t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 0)),	// X
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 3)),	// W
+				  PVS_SRC_SELECT_FORCE_0,	// Z
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)),	// Y
+				  t_src_class(vpi->SrcReg[0].File),
+				  vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) |
+	    (vpi->SrcReg[0].RelAddr << 4);
+	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[0]), t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)),	// Y
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 3)),	// W
+				  PVS_SRC_SELECT_FORCE_0,	// Z
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 0)),	// X
+				  t_src_class(vpi->SrcReg[0].File),
+				  vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) |
+	    (vpi->SrcReg[0].RelAddr << 4);
+	inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[0]), t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)),	// Y
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 0)),	// X
+				  PVS_SRC_SELECT_FORCE_0,	// Z
+				  t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 3)),	// W
+				  t_src_class(vpi->SrcReg[0].File),
+				  vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) |
+	    (vpi->SrcReg[0].RelAddr << 4);
+}
+
+static void ei_mad(struct r300_vertex_program_code *vp,
+				      struct prog_instruction *vpi,
+				      GLuint * inst)
+{
+	/* Remarks about hardware limitations of MAD
+	 * (please preserve this comment, as this information is _NOT_
+	 * in the documentation provided by AMD).
+	 *
+	 * As described in the documentation, MAD with three unique temporary
+	 * source registers requires the use of the macro version.
+	 *
+	 * However (and this is not mentioned in the documentation), apparently
+	 * the macro version is _NOT_ a full superset of the normal version.
+	 * In particular, the macro version does not always work when relative
+	 * addressing is used in the source operands.
+	 *
+	 * This limitation caused incorrect rendering in Sauerbraten's OpenGL
+	 * assembly shader path when using medium quality animations
+	 * (i.e. animations with matrix blending instead of quaternion blending).
+	 *
+	 * Unfortunately, I (nha) have been unable to extract a Piglit regression
+	 * test for this issue - for some reason, it is possible to have vertex
+	 * programs whose prefix is *exactly* the same as the prefix of the
+	 * offending program in Sauerbraten up to the offending instruction
+	 * without causing any trouble.
+	 *
+	 * Bottom line: Only use the macro version only when really necessary;
+	 * according to AMD docs, this should improve performance by one clock
+	 * as a nice side bonus.
+	 */
+	if (vpi->SrcReg[0].File == PROGRAM_TEMPORARY &&
+	    vpi->SrcReg[1].File == PROGRAM_TEMPORARY &&
+	    vpi->SrcReg[2].File == PROGRAM_TEMPORARY &&
+	    vpi->SrcReg[0].Index != vpi->SrcReg[1].Index &&
+	    vpi->SrcReg[0].Index != vpi->SrcReg[2].Index &&
+	    vpi->SrcReg[1].Index != vpi->SrcReg[2].Index) {
+		inst[0] = PVS_OP_DST_OPERAND(PVS_MACRO_OP_2CLK_MADD,
+				GL_FALSE,
+				GL_TRUE,
+				t_dst_index(vp, &vpi->DstReg),
+				t_dst_mask(vpi->DstReg.WriteMask),
+				t_dst_class(vpi->DstReg.File));
+	} else {
+		inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD,
+				GL_FALSE,
+				GL_FALSE,
+				t_dst_index(vp, &vpi->DstReg),
+				t_dst_mask(vpi->DstReg.WriteMask),
+				t_dst_class(vpi->DstReg.File));
+	}
+	inst[1] = t_src(vp, &vpi->SrcReg[0]);
+	inst[2] = t_src(vp, &vpi->SrcReg[1]);
+	inst[3] = t_src(vp, &vpi->SrcReg[2]);
+}
+
+static void ei_pow(struct r300_vertex_program_code *vp,
+				      struct prog_instruction *vpi,
+				      GLuint * inst)
+{
+	inst[0] = PVS_OP_DST_OPERAND(ME_POWER_FUNC_FF,
+				     GL_TRUE,
+				     GL_FALSE,
+				     t_dst_index(vp, &vpi->DstReg),
+				     t_dst_mask(vpi->DstReg.WriteMask),
+				     t_dst_class(vpi->DstReg.File));
+	inst[1] = t_src_scalar(vp, &vpi->SrcReg[0]);
+	inst[2] = __CONST(0, SWIZZLE_ZERO);
+	inst[3] = t_src_scalar(vp, &vpi->SrcReg[1]);
+}
+
+
+static void translate_vertex_program(struct r300_vertex_program_compiler * compiler)
+{
+	struct rc_instruction *rci;
+
+	compiler->code->pos_end = 0;	/* Not supported yet */
+	compiler->code->length = 0;
+
+	compiler->SetHwInputOutput(compiler);
+
+	for(rci = compiler->Base.Program.Instructions.Next; rci != &compiler->Base.Program.Instructions; rci = rci->Next) {
+		struct prog_instruction *vpi = &rci->I;
+		GLuint *inst = compiler->code->body.d + compiler->code->length;
+
+		/* Skip instructions writing to non-existing destination */
+		if (!valid_dst(compiler->code, &vpi->DstReg))
+			continue;
+
+		if (compiler->code->length >= VSF_MAX_FRAGMENT_LENGTH) {
+			rc_error(&compiler->Base, "Vertex program has too many instructions\n");
+			return;
+		}
+
+		switch (vpi->Opcode) {
+		case OPCODE_ADD: ei_vector2(compiler->code, VE_ADD, vpi, inst); break;
+		case OPCODE_ARL: ei_vector1(compiler->code, VE_FLT2FIX_DX, vpi, inst); break;
+		case OPCODE_DP4: ei_vector2(compiler->code, VE_DOT_PRODUCT, vpi, inst); break;
+		case OPCODE_DST: ei_vector2(compiler->code, VE_DISTANCE_VECTOR, vpi, inst); break;
+		case OPCODE_EX2: ei_math1(compiler->code, ME_EXP_BASE2_FULL_DX, vpi, inst); break;
+		case OPCODE_EXP: ei_math1(compiler->code, ME_EXP_BASE2_DX, vpi, inst); break;
+		case OPCODE_FRC: ei_vector1(compiler->code, VE_FRACTION, vpi, inst); break;
+		case OPCODE_LG2: ei_math1(compiler->code, ME_LOG_BASE2_FULL_DX, vpi, inst); break;
+		case OPCODE_LIT: ei_lit(compiler->code, vpi, inst); break;
+		case OPCODE_LOG: ei_math1(compiler->code, ME_LOG_BASE2_DX, vpi, inst); break;
+		case OPCODE_MAD: ei_mad(compiler->code, vpi, inst); break;
+		case OPCODE_MAX: ei_vector2(compiler->code, VE_MAXIMUM, vpi, inst); break;
+		case OPCODE_MIN: ei_vector2(compiler->code, VE_MINIMUM, vpi, inst); break;
+		case OPCODE_MOV: ei_vector1(compiler->code, VE_ADD, vpi, inst); break;
+		case OPCODE_MUL: ei_vector2(compiler->code, VE_MULTIPLY, vpi, inst); break;
+		case OPCODE_POW: ei_pow(compiler->code, vpi, inst); break;
+		case OPCODE_RCP: ei_math1(compiler->code, ME_RECIP_DX, vpi, inst); break;
+		case OPCODE_RSQ: ei_math1(compiler->code, ME_RECIP_SQRT_DX, vpi, inst); break;
+		case OPCODE_SGE: ei_vector2(compiler->code, VE_SET_GREATER_THAN_EQUAL, vpi, inst); break;
+		case OPCODE_SLT: ei_vector2(compiler->code, VE_SET_LESS_THAN, vpi, inst); break;
+		default:
+			rc_error(&compiler->Base, "Unknown opcode %i\n", vpi->Opcode);
+			return;
+		}
+
+		compiler->code->length += 4;
+
+		if (compiler->Base.Error)
+			return;
+	}
+}
+
+struct temporary_allocation {
+	GLuint Allocated:1;
+	GLuint HwTemp:15;
+	struct rc_instruction * LastRead;
+};
+
+static void allocate_temporary_registers(struct r300_vertex_program_compiler * compiler)
+{
+	struct rc_instruction *inst;
+	GLuint num_orig_temps = 0;
+	GLboolean hwtemps[VSF_MAX_FRAGMENT_TEMPS];
+	struct temporary_allocation * ta;
+	GLuint i, j;
+
+	compiler->code->num_temporaries = 0;
+	memset(hwtemps, 0, sizeof(hwtemps));
+
+	/* Pass 1: Count original temporaries and allocate structures */
+	for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) {
+		GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode);
+		GLuint numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode);
+
+		for (i = 0; i < numsrcs; ++i) {
+			if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY) {
+				if (inst->I.SrcReg[i].Index >= num_orig_temps)
+					num_orig_temps = inst->I.SrcReg[i].Index + 1;
+			}
+		}
+
+		if (numdsts) {
+			if (inst->I.DstReg.File == PROGRAM_TEMPORARY) {
+				if (inst->I.DstReg.Index >= num_orig_temps)
+					num_orig_temps = inst->I.DstReg.Index + 1;
+			}
+		}
+	}
+
+	ta = (struct temporary_allocation*)memory_pool_malloc(&compiler->Base.Pool,
+			sizeof(struct temporary_allocation) * num_orig_temps);
+	memset(ta, 0, sizeof(struct temporary_allocation) * num_orig_temps);
+
+	/* Pass 2: Determine original temporary lifetimes */
+	for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) {
+		GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode);
+
+		for (i = 0; i < numsrcs; ++i) {
+			if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY)
+				ta[inst->I.SrcReg[i].Index].LastRead = inst;
+		}
+	}
+
+	/* Pass 3: Register allocation */
+	for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) {
+		GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode);
+		GLuint numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode);
+
+		for (i = 0; i < numsrcs; ++i) {
+			if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY) {
+				GLuint orig = inst->I.SrcReg[i].Index;
+				inst->I.SrcReg[i].Index = ta[orig].HwTemp;
+
+				if (ta[orig].Allocated && inst == ta[orig].LastRead)
+					hwtemps[ta[orig].HwTemp] = GL_FALSE;
+			}
+		}
+
+		if (numdsts) {
+			if (inst->I.DstReg.File == PROGRAM_TEMPORARY) {
+				GLuint orig = inst->I.DstReg.Index;
+
+				if (!ta[orig].Allocated) {
+					for(j = 0; j < VSF_MAX_FRAGMENT_TEMPS; ++j) {
+						if (!hwtemps[j])
+							break;
+					}
+					if (j >= VSF_MAX_FRAGMENT_TEMPS) {
+						fprintf(stderr, "Out of hw temporaries\n");
+					} else {
+						ta[orig].Allocated = GL_TRUE;
+						ta[orig].HwTemp = j;
+						hwtemps[j] = GL_TRUE;
+
+						if (j >= compiler->code->num_temporaries)
+							compiler->code->num_temporaries = j + 1;
+					}
+				}
+
+				inst->I.DstReg.Index = ta[orig].HwTemp;
+			}
+		}
+	}
+}
+
+
+/**
+ * Vertex engine cannot read two inputs or two constants at the same time.
+ * Introduce intermediate MOVs to temporary registers to account for this.
+ */
+static GLboolean transform_source_conflicts(
+	struct radeon_compiler *c,
+	struct rc_instruction* inst,
+	void* unused)
+{
+	GLuint num_operands = _mesa_num_inst_src_regs(inst->I.Opcode);
+
+	if (num_operands == 3) {
+		if (t_src_conflict(inst->I.SrcReg[1], inst->I.SrcReg[2])
+		    || t_src_conflict(inst->I.SrcReg[0], inst->I.SrcReg[2])) {
+			int tmpreg = rc_find_free_temporary(c);
+			struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev);
+			inst_mov->I.Opcode = OPCODE_MOV;
+			inst_mov->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_mov->I.DstReg.Index = tmpreg;
+			inst_mov->I.SrcReg[0] = inst->I.SrcReg[2];
+
+			reset_srcreg(&inst->I.SrcReg[2]);
+			inst->I.SrcReg[2].File = PROGRAM_TEMPORARY;
+			inst->I.SrcReg[2].Index = tmpreg;
+		}
+	}
+
+	if (num_operands >= 2) {
+		if (t_src_conflict(inst->I.SrcReg[1], inst->I.SrcReg[0])) {
+			int tmpreg = rc_find_free_temporary(c);
+			struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev);
+			inst_mov->I.Opcode = OPCODE_MOV;
+			inst_mov->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_mov->I.DstReg.Index = tmpreg;
+			inst_mov->I.SrcReg[0] = inst->I.SrcReg[1];
+
+			reset_srcreg(&inst->I.SrcReg[1]);
+			inst->I.SrcReg[1].File = PROGRAM_TEMPORARY;
+			inst->I.SrcReg[1].Index = tmpreg;
+		}
+	}
+
+	return GL_TRUE;
+}
+
+static void addArtificialOutputs(struct r300_vertex_program_compiler * compiler)
+{
+	int i;
+
+	for(i = 0; i < 32; ++i) {
+		if ((compiler->RequiredOutputs & (1 << i)) &&
+		    !(compiler->Base.Program.OutputsWritten & (1 << i))) {
+			struct rc_instruction * inst = rc_insert_new_instruction(&compiler->Base, compiler->Base.Program.Instructions.Prev);
+			inst->I.Opcode = OPCODE_MOV;
+
+			inst->I.DstReg.File = PROGRAM_OUTPUT;
+			inst->I.DstReg.Index = i;
+			inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
+
+			inst->I.SrcReg[0].File = PROGRAM_CONSTANT;
+			inst->I.SrcReg[0].Index = 0;
+			inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+			compiler->Base.Program.OutputsWritten |= 1 << i;
+		}
+	}
+}
+
+static void nqssadceInit(struct nqssadce_state* s)
+{
+	struct r300_vertex_program_compiler * compiler = s->UserData;
+	int i;
+
+	for(i = 0; i < VERT_RESULT_MAX; ++i) {
+		if (compiler->RequiredOutputs & (1 << i))
+			s->Outputs[i].Sourced = WRITEMASK_XYZW;
+	}
+}
+
+static GLboolean swizzleIsNative(GLuint opcode, struct prog_src_register reg)
+{
+	(void) opcode;
+	(void) reg;
+
+	return GL_TRUE;
+}
+
+
+
+void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler)
+{
+	addArtificialOutputs(compiler);
+
+	{
+		struct radeon_program_transformation transformations[] = {
+			{ &r300_transform_vertex_alu, 0 },
+		};
+		radeonLocalTransform(&compiler->Base, 1, transformations);
+	}
+
+	if (compiler->Base.Debug) {
+		fprintf(stderr, "Vertex program after native rewrite:\n");
+		rc_print_program(&compiler->Base.Program);
+		fflush(stderr);
+	}
+
+	{
+		/* Note: This pass has to be done seperately from ALU rewrite,
+		 * otherwise non-native ALU instructions with source conflits
+		 * will not be treated properly.
+		 */
+		struct radeon_program_transformation transformations[] = {
+			{ &transform_source_conflicts, 0 },
+		};
+		radeonLocalTransform(&compiler->Base, 1, transformations);
+	}
+
+	if (compiler->Base.Debug) {
+		fprintf(stderr, "Vertex program after source conflict resolve:\n");
+		rc_print_program(&compiler->Base.Program);
+		fflush(stderr);
+	}
+
+	{
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadceInit,
+			.IsNativeSwizzle = &swizzleIsNative,
+			.BuildSwizzle = NULL
+		};
+		radeonNqssaDce(&compiler->Base, &nqssadce, compiler);
+
+		/* We need this step for reusing temporary registers */
+		allocate_temporary_registers(compiler);
+
+		if (compiler->Base.Debug) {
+			fprintf(stderr, "Vertex program after NQSSADCE:\n");
+			rc_print_program(&compiler->Base.Program);
+			fflush(stderr);
+		}
+	}
+
+	translate_vertex_program(compiler);
+
+	rc_constants_copy(&compiler->code->constants, &compiler->Base.Program.Constants);
+
+	compiler->code->InputsRead = compiler->Base.Program.InputsRead;
+	compiler->code->OutputsWritten = compiler->Base.Program.OutputsWritten;
+
+	if (compiler->Base.Debug) {
+		fprintf(stderr, "Final vertex program code:\n");
+		r300_vertex_program_dump(compiler->code);
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c
new file mode 100644
index 0000000000..980ef3eaea
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "radeon_code.h"
+
+#include <stdio.h>
+
+static char* r300_vs_ve_ops[] = {
+	/* R300 vector ops */
+	"                 VE_NO_OP",
+	"           VE_DOT_PRODUCT",
+	"              VE_MULTIPLY",
+	"                   VE_ADD",
+	"          VE_MULTIPLY_ADD",
+	"       VE_DISTANCE_FACTOR",
+	"              VE_FRACTION",
+	"               VE_MAXIMUM",
+	"               VE_MINIMUM",
+	"VE_SET_GREATER_THAN_EQUAL",
+	"         VE_SET_LESS_THAN",
+	"        VE_MULTIPLYX2_ADD",
+	"        VE_MULTIPLY_CLAMP",
+	"            VE_FLT2FIX_DX",
+	"        VE_FLT2FIX_DX_RND",
+	/* R500 vector ops */
+	"      VE_PRED_SET_EQ_PUSH",
+	"      VE_PRED_SET_GT_PUSH",
+	"     VE_PRED_SET_GTE_PUSH",
+	"     VE_PRED_SET_NEQ_PUSH",
+	"         VE_COND_WRITE_EQ",
+	"         VE_COND_WRITE_GT",
+	"        VE_COND_WRITE_GTE",
+	"        VE_COND_WRITE_NEQ",
+	"      VE_SET_GREATER_THAN",
+	"             VE_SET_EQUAL",
+	"         VE_SET_NOT_EQUAL",
+	"               (reserved)",
+	"               (reserved)",
+	"               (reserved)",
+};
+
+static char* r300_vs_me_ops[] = {
+	/* R300 math ops */
+	"                 ME_NO_OP",
+	"          ME_EXP_BASE2_DX",
+	"          ME_LOG_BASE2_DX",
+	"          ME_EXP_BASEE_FF",
+	"        ME_LIGHT_COEFF_DX",
+	"         ME_POWER_FUNC_FF",
+	"              ME_RECIP_DX",
+	"              ME_RECIP_FF",
+	"         ME_RECIP_SQRT_DX",
+	"         ME_RECIP_SQRT_FF",
+	"              ME_MULTIPLY",
+	"     ME_EXP_BASE2_FULL_DX",
+	"     ME_LOG_BASE2_FULL_DX",
+	" ME_POWER_FUNC_FF_CLAMP_B",
+	"ME_POWER_FUNC_FF_CLAMP_B1",
+	"ME_POWER_FUNC_FF_CLAMP_01",
+	"                   ME_SIN",
+	"                   ME_COS",
+	/* R500 math ops */
+	"        ME_LOG_BASE2_IEEE",
+	"            ME_RECIP_IEEE",
+	"       ME_RECIP_SQRT_IEEE",
+	"           ME_PRED_SET_EQ",
+	"           ME_PRED_SET_GT",
+	"          ME_PRED_SET_GTE",
+	"          ME_PRED_SET_NEQ",
+	"          ME_PRED_SET_CLR",
+	"          ME_PRED_SET_INV",
+	"          ME_PRED_SET_POP",
+	"      ME_PRED_SET_RESTORE",
+	"               (reserved)",
+	"               (reserved)",
+	"               (reserved)",
+};
+
+/* XXX refactor to avoid clashing symbols */
+static char* r300_vs_src_debug[] = {
+	"t",
+	"i",
+	"c",
+	"a",
+};
+
+static char* r300_vs_dst_debug[] = {
+	"t",
+	"a0",
+	"o",
+	"ox",
+	"a",
+	"i",
+	"u",
+	"u",
+};
+
+static char* r300_vs_swiz_debug[] = {
+	"X",
+	"Y",
+	"Z",
+	"W",
+	"0",
+	"1",
+	"U",
+	"U",
+};
+
+
+static void r300_vs_op_dump(uint32_t op)
+{
+	fprintf(stderr, " dst: %d%s op: ",
+			(op >> 13) & 0x7f, r300_vs_dst_debug[(op >> 8) & 0x7]);
+	if (op & 0x80) {
+		if (op & 0x1) {
+			fprintf(stderr, "PVS_MACRO_OP_2CLK_M2X_ADD\n");
+		} else {
+			fprintf(stderr, "   PVS_MACRO_OP_2CLK_MADD\n");
+		}
+	} else if (op & 0x40) {
+		fprintf(stderr, "%s\n", r300_vs_me_ops[op & 0x1f]);
+	} else {
+		fprintf(stderr, "%s\n", r300_vs_ve_ops[op & 0x1f]);
+	}
+}
+
+static void r300_vs_src_dump(uint32_t src)
+{
+	fprintf(stderr, " reg: %d%s swiz: %s%s/%s%s/%s%s/%s%s\n",
+			(src >> 5) & 0x7f, r300_vs_src_debug[src & 0x3],
+			src & (1 << 25) ? "-" : " ",
+			r300_vs_swiz_debug[(src >> 13) & 0x7],
+			src & (1 << 26) ? "-" : " ",
+			r300_vs_swiz_debug[(src >> 16) & 0x7],
+			src & (1 << 27) ? "-" : " ",
+			r300_vs_swiz_debug[(src >> 19) & 0x7],
+			src & (1 << 28) ? "-" : " ",
+			r300_vs_swiz_debug[(src >> 22) & 0x7]);
+}
+
+void r300_vertex_program_dump(struct r300_vertex_program_code * vs)
+{
+	unsigned instrcount = vs->length / 4;
+	unsigned i;
+
+	for(i = 0; i < instrcount; i++) {
+		unsigned offset = i*4;
+		unsigned src;
+
+		fprintf(stderr, "%d: op: 0x%08x", i, vs->body.d[offset]);
+		r300_vs_op_dump(vs->body.d[offset]);
+
+		for(src = 0; src < 3; ++src) {
+			fprintf(stderr, " src%i: 0x%08x", src, vs->body.d[offset+1+src]);
+			r300_vs_src_dump(vs->body.d[offset+1+src]);
+		}
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c
new file mode 100644
index 0000000000..7e2faed690
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c
@@ -0,0 +1,449 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "r500_fragprog.h"
+
+#include "../r300_reg.h"
+
+static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int tmu)
+{
+	struct prog_src_register reg = { 0, };
+
+	reg.File = PROGRAM_STATE_VAR;
+	reg.Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_SHADOW_AMBIENT, tmu);
+	reg.Swizzle = SWIZZLE_WWWW;
+	return reg;
+}
+
+/**
+ * Transform TEX, TXP, TXB, and KIL instructions in the following way:
+ *  - implement texture compare (shadow extensions)
+ *  - extract non-native source / destination operands
+ */
+GLboolean r500_transform_TEX(
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
+	void* data)
+{
+	struct r300_fragment_program_compiler *compiler =
+		(struct r300_fragment_program_compiler*)data;
+
+	if (inst->I.Opcode != OPCODE_TEX &&
+	    inst->I.Opcode != OPCODE_TXB &&
+	    inst->I.Opcode != OPCODE_TXP &&
+	    inst->I.Opcode != OPCODE_KIL)
+		return GL_FALSE;
+
+	/* ARB_shadow & EXT_shadow_funcs */
+	if (inst->I.Opcode != OPCODE_KIL &&
+	    c->Program.ShadowSamplers & (1 << inst->I.TexSrcUnit)) {
+		GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func;
+
+		if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) {
+			inst->I.Opcode = OPCODE_MOV;
+
+			if (comparefunc == GL_ALWAYS) {
+				inst->I.SrcReg[0].File = PROGRAM_BUILTIN;
+				inst->I.SrcReg[0].Swizzle = SWIZZLE_1111;
+			} else {
+				inst->I.SrcReg[0] = shadow_ambient(c, inst->I.TexSrcUnit);
+			}
+
+			return GL_TRUE;
+		} else {
+			GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func;
+			GLuint depthmode = compiler->state.unit[inst->I.TexSrcUnit].depth_texture_mode;
+			struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, inst);
+			struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_rcp);
+			struct rc_instruction * inst_cmp = rc_insert_new_instruction(c, inst_mad);
+			int pass, fail;
+
+			inst_rcp->I.Opcode = OPCODE_RCP;
+			inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_rcp->I.DstReg.Index = rc_find_free_temporary(c);
+			inst_rcp->I.DstReg.WriteMask = WRITEMASK_W;
+			inst_rcp->I.SrcReg[0] = inst->I.SrcReg[0];
+			inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW;
+
+			inst_cmp->I.DstReg = inst->I.DstReg;
+			inst->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst->I.DstReg.Index = rc_find_free_temporary(c);
+			inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
+
+			inst_mad->I.Opcode = OPCODE_MAD;
+			inst_mad->I.DstReg.File = PROGRAM_TEMPORARY;
+			inst_mad->I.DstReg.Index = rc_find_free_temporary(c);
+			inst_mad->I.SrcReg[0] = inst->I.SrcReg[0];
+			inst_mad->I.SrcReg[0].Swizzle = SWIZZLE_ZZZZ;
+			inst_mad->I.SrcReg[1].File = PROGRAM_TEMPORARY;
+			inst_mad->I.SrcReg[1].Index = inst_rcp->I.DstReg.Index;
+			inst_mad->I.SrcReg[1].Swizzle = SWIZZLE_WWWW;
+			inst_mad->I.SrcReg[2].File = PROGRAM_TEMPORARY;
+			inst_mad->I.SrcReg[2].Index = inst->I.DstReg.Index;
+			if (depthmode == 0) /* GL_LUMINANCE */
+				inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z);
+			else if (depthmode == 2) /* GL_ALPHA */
+				inst_mad->I.SrcReg[2].Swizzle = SWIZZLE_WWWW;
+
+			/* Recall that SrcReg[0] is tex, SrcReg[2] is r and:
+			 *   r  < tex  <=>      -tex+r < 0
+			 *   r >= tex  <=> not (-tex+r < 0 */
+			if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
+				inst_mad->I.SrcReg[2].Negate = inst_mad->I.SrcReg[2].Negate ^ NEGATE_XYZW;
+			else
+				inst_mad->I.SrcReg[0].Negate = inst_mad->I.SrcReg[0].Negate ^ NEGATE_XYZW;
+
+			inst_cmp->I.Opcode = OPCODE_CMP;
+			/* DstReg has been filled out above */
+			inst_cmp->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+			inst_cmp->I.SrcReg[0].Index = inst_mad->I.DstReg.Index;
+
+			if (comparefunc == GL_LESS || comparefunc == GL_GREATER) {
+				pass = 1;
+				fail = 2;
+			} else {
+				pass = 2;
+				fail = 1;
+			}
+
+			inst_cmp->I.SrcReg[pass].File = PROGRAM_BUILTIN;
+			inst_cmp->I.SrcReg[pass].Swizzle = SWIZZLE_1111;
+			inst_cmp->I.SrcReg[fail] = shadow_ambient(c, inst->I.TexSrcUnit);
+		}
+	}
+
+	/* Cannot write texture to output registers */
+	if (inst->I.Opcode != OPCODE_KIL && inst->I.DstReg.File != PROGRAM_TEMPORARY) {
+		struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst);
+
+		inst_mov->I.Opcode = OPCODE_MOV;
+		inst_mov->I.DstReg = inst->I.DstReg;
+		inst_mov->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+		inst_mov->I.SrcReg[0].Index = rc_find_free_temporary(c);
+
+		inst->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index;
+		inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
+	}
+
+	/* Cannot read texture coordinate from constants file */
+	if (inst->I.SrcReg[0].File != PROGRAM_TEMPORARY && inst->I.SrcReg[0].File != PROGRAM_INPUT) {
+		struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev);
+
+		inst_mov->I.Opcode = OPCODE_MOV;
+		inst_mov->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst_mov->I.DstReg.Index = rc_find_free_temporary(c);
+		inst_mov->I.SrcReg[0] = inst->I.SrcReg[0];
+
+		reset_srcreg(&inst->I.SrcReg[0]);
+		inst->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+		inst->I.SrcReg[0].Index = inst_mov->I.DstReg.Index;
+	}
+
+	return GL_TRUE;
+}
+
+GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
+{
+	GLuint relevant;
+	int i;
+
+	if (opcode == OPCODE_TEX ||
+	    opcode == OPCODE_TXB ||
+	    opcode == OPCODE_TXP ||
+	    opcode == OPCODE_KIL) {
+		if (reg.Abs)
+			return GL_FALSE;
+
+		if (opcode == OPCODE_KIL && (reg.Swizzle != SWIZZLE_NOOP || reg.Negate != NEGATE_NONE))
+			return GL_FALSE;
+
+		if (reg.Negate)
+			reg.Negate ^= NEGATE_XYZW;
+
+		for(i = 0; i < 4; ++i) {
+			GLuint swz = GET_SWZ(reg.Swizzle, i);
+			if (swz == SWIZZLE_NIL) {
+				reg.Negate &= ~(1 << i);
+				continue;
+			}
+			if (swz >= 4)
+				return GL_FALSE;
+		}
+
+		if (reg.Negate)
+			return GL_FALSE;
+
+		return GL_TRUE;
+	} else if (opcode == OPCODE_DDX || opcode == OPCODE_DDY) {
+		/* DDX/MDH and DDY/MDV explicitly ignore incoming swizzles;
+		 * if it doesn't fit perfectly into a .xyzw case... */
+		if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs && !reg.Negate)
+			return GL_TRUE;
+
+		return GL_FALSE;
+	} else {
+		/* ALU instructions support almost everything */
+		if (reg.Abs)
+			return GL_TRUE;
+
+		relevant = 0;
+		for(i = 0; i < 3; ++i) {
+			GLuint swz = GET_SWZ(reg.Swizzle, i);
+			if (swz != SWIZZLE_NIL && swz != SWIZZLE_ZERO)
+				relevant |= 1 << i;
+		}
+		if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant))
+			return GL_FALSE;
+
+		return GL_TRUE;
+	}
+}
+
+/**
+ * Implement a MOV with a potentially non-native swizzle.
+ *
+ * The only thing we *cannot* do in an ALU instruction is per-component
+ * negation. Therefore, we split the MOV into two instructions when necessary.
+ */
+void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src)
+{
+	GLuint negatebase[2] = { 0, 0 };
+	int i;
+
+	for(i = 0; i < 4; ++i) {
+		GLuint swz = GET_SWZ(src.Swizzle, i);
+		if (swz == SWIZZLE_NIL)
+			continue;
+		negatebase[GET_BIT(src.Negate, i)] |= 1 << i;
+	}
+
+	for(i = 0; i <= 1; ++i) {
+		if (!negatebase[i])
+			continue;
+
+		struct rc_instruction *inst = rc_insert_new_instruction(s->Compiler, s->IP->Prev);
+		inst->I.Opcode = OPCODE_MOV;
+		inst->I.DstReg = dst;
+		inst->I.DstReg.WriteMask = negatebase[i];
+		inst->I.SrcReg[0] = src;
+		inst->I.SrcReg[0].Negate = (i == 0) ? NEGATE_NONE : NEGATE_XYZW;
+	}
+}
+
+
+static char *toswiz(int swiz_val) {
+  switch(swiz_val) {
+  case 0: return "R";
+  case 1: return "G";
+  case 2: return "B";
+  case 3: return "A";
+  case 4: return "0";
+  case 5: return "1/2";
+  case 6: return "1";
+  case 7: return "U";
+  }
+  return NULL;
+}
+
+static char *toop(int op_val)
+{
+  char *str = NULL;
+  switch (op_val) {
+  case 0: str = "MAD"; break;
+  case 1: str = "DP3"; break;
+  case 2: str = "DP4"; break;
+  case 3: str = "D2A"; break;
+  case 4: str = "MIN"; break;
+  case 5: str = "MAX"; break;
+  case 6: str = "Reserved"; break;
+  case 7: str = "CND"; break;
+  case 8: str = "CMP"; break;
+  case 9: str = "FRC"; break;
+  case 10: str = "SOP"; break;
+  case 11: str = "MDH"; break;
+  case 12: str = "MDV"; break;
+  }
+  return str;
+}
+
+static char *to_alpha_op(int op_val)
+{
+  char *str = NULL;
+  switch (op_val) {
+  case 0: str = "MAD"; break;
+  case 1: str = "DP"; break;
+  case 2: str = "MIN"; break;
+  case 3: str = "MAX"; break;
+  case 4: str = "Reserved"; break;
+  case 5: str = "CND"; break;
+  case 6: str = "CMP"; break;
+  case 7: str = "FRC"; break;
+  case 8: str = "EX2"; break;
+  case 9: str = "LN2"; break;
+  case 10: str = "RCP"; break;
+  case 11: str = "RSQ"; break;
+  case 12: str = "SIN"; break;
+  case 13: str = "COS"; break;
+  case 14: str = "MDH"; break;
+  case 15: str = "MDV"; break;
+  }
+  return str;
+}
+
+static char *to_mask(int val)
+{
+  char *str = NULL;
+  switch(val) {
+  case 0: str = "NONE"; break;
+  case 1: str = "R"; break;
+  case 2: str = "G"; break;
+  case 3: str = "RG"; break;
+  case 4: str = "B"; break;
+  case 5: str = "RB"; break;
+  case 6: str = "GB"; break;
+  case 7: str = "RGB"; break;
+  case 8: str = "A"; break;
+  case 9: str = "AR"; break;
+  case 10: str = "AG"; break;
+  case 11: str = "ARG"; break;
+  case 12: str = "AB"; break;
+  case 13: str = "ARB"; break;
+  case 14: str = "AGB"; break;
+  case 15: str = "ARGB"; break;
+  }
+  return str;
+}
+
+static char *to_texop(int val)
+{
+  switch(val) {
+  case 0: return "NOP";
+  case 1: return "LD";
+  case 2: return "TEXKILL";
+  case 3: return "PROJ";
+  case 4: return "LODBIAS";
+  case 5: return "LOD";
+  case 6: return "DXDY";
+  }
+  return NULL;
+}
+
+void r500FragmentProgramDump(struct rX00_fragment_program_code *c)
+{
+  struct r500_fragment_program_code *code = &c->code.r500;
+  fprintf(stderr, "R500 Fragment Program:\n--------\n");
+
+  int n;
+  uint32_t inst;
+  uint32_t inst0;
+  char *str = NULL;
+
+  for (n = 0; n < code->inst_end+1; n++) {
+    inst0 = inst = code->inst[n].inst0;
+    fprintf(stderr,"%d\t0:CMN_INST   0x%08x:", n, inst);
+    switch(inst & 0x3) {
+    case R500_INST_TYPE_ALU: str = "ALU"; break;
+    case R500_INST_TYPE_OUT: str = "OUT"; break;
+    case R500_INST_TYPE_FC: str = "FC"; break;
+    case R500_INST_TYPE_TEX: str = "TEX"; break;
+    };
+    fprintf(stderr,"%s %s %s %s %s ", str,
+	    inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "",
+	    inst & R500_INST_LAST ? "LAST" : "",
+	    inst & R500_INST_NOP ? "NOP" : "",
+	    inst & R500_INST_ALU_WAIT ? "ALU WAIT" : "");
+    fprintf(stderr,"wmask: %s omask: %s\n", to_mask((inst >> 11) & 0xf),
+	    to_mask((inst >> 15) & 0xf));
+
+    switch(inst0 & 0x3) {
+    case 0:
+    case 1:
+      fprintf(stderr,"\t1:RGB_ADDR   0x%08x:", code->inst[n].inst1);
+      inst = code->inst[n].inst1;
+
+      fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
+	      inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
+	      (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
+	      (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
+	      (inst >> 30));
+
+      fprintf(stderr,"\t2:ALPHA_ADDR 0x%08x:", code->inst[n].inst2);
+      inst = code->inst[n].inst2;
+      fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
+	      inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
+	      (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
+	      (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
+	      (inst >> 30));
+      fprintf(stderr,"\t3 RGB_INST:  0x%08x:", code->inst[n].inst3);
+      inst = code->inst[n].inst3;
+      fprintf(stderr,"rgb_A_src:%d %s/%s/%s %d rgb_B_src:%d %s/%s/%s %d\n",
+	      (inst) & 0x3, toswiz((inst >> 2) & 0x7), toswiz((inst >> 5) & 0x7), toswiz((inst >> 8) & 0x7),
+	      (inst >> 11) & 0x3,
+	      (inst >> 13) & 0x3, toswiz((inst >> 15) & 0x7), toswiz((inst >> 18) & 0x7), toswiz((inst >> 21) & 0x7),
+	      (inst >> 24) & 0x3);
+
+
+      fprintf(stderr,"\t4 ALPHA_INST:0x%08x:", code->inst[n].inst4);
+      inst = code->inst[n].inst4;
+      fprintf(stderr,"%s dest:%d%s alp_A_src:%d %s %d alp_B_src:%d %s %d w:%d\n", to_alpha_op(inst & 0xf),
+	      (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
+	      (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), (inst >> 17) & 0x3,
+	      (inst >> 19) & 0x3, toswiz((inst >> 21) & 0x7), (inst >> 24) & 0x3,
+	      (inst >> 31) & 0x1);
+
+      fprintf(stderr,"\t5 RGBA_INST: 0x%08x:", code->inst[n].inst5);
+      inst = code->inst[n].inst5;
+      fprintf(stderr,"%s dest:%d%s rgb_C_src:%d %s/%s/%s %d alp_C_src:%d %s %d\n", toop(inst & 0xf),
+	      (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
+	      (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), toswiz((inst >> 17) & 0x7), toswiz((inst >> 20) & 0x7),
+	      (inst >> 23) & 0x3,
+	      (inst >> 25) & 0x3, toswiz((inst >> 27) & 0x7), (inst >> 30) & 0x3);
+      break;
+    case 2:
+      break;
+    case 3:
+      inst = code->inst[n].inst1;
+      fprintf(stderr,"\t1:TEX_INST:  0x%08x: id: %d op:%s, %s, %s %s\n", inst, (inst >> 16) & 0xf,
+	      to_texop((inst >> 22) & 0x7), (inst & (1<<25)) ? "ACQ" : "",
+	      (inst & (1<<26)) ? "IGNUNC" : "", (inst & (1<<27)) ? "UNSCALED" : "SCALED");
+      inst = code->inst[n].inst2;
+      fprintf(stderr,"\t2:TEX_ADDR:  0x%08x: src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n", inst,
+	      inst & 127, inst & (1<<7) ? "(rel)" : "",
+	      toswiz((inst >> 8) & 0x3), toswiz((inst >> 10) & 0x3),
+	      toswiz((inst >> 12) & 0x3), toswiz((inst >> 14) & 0x3),
+	      (inst >> 16) & 127, inst & (1<<23) ? "(rel)" : "",
+	      toswiz((inst >> 24) & 0x3), toswiz((inst >> 26) & 0x3),
+	      toswiz((inst >> 28) & 0x3), toswiz((inst >> 30) & 0x3));
+
+      fprintf(stderr,"\t3:TEX_DXDY:  0x%08x\n", code->inst[n].inst3);
+      break;
+    }
+    fprintf(stderr,"\n");
+  }
+
+}
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h
index 1e45538f80..9091f65cd2 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog.h
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h
@@ -33,30 +33,23 @@
 #ifndef __R500_FRAGPROG_H_
 #define __R500_FRAGPROG_H_
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
 #include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-#include "shader/program.h"
 #include "shader/prog_instruction.h"
 
-#include "r300_context.h"
-#include "r300_state.h"
-#include "radeon_program.h"
+#include "radeon_compiler.h"
+#include "radeon_nqssadce.h"
 
-struct r500_fragment_program;
+extern void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
 
-extern void r500TranslateFragmentShader(r300ContextPtr r300,
-					struct r500_fragment_program *fp);
+extern void r500FragmentProgramDump(struct rX00_fragment_program_code *c);
 
-struct r500_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r500_fragment_program *fp;
-	struct r500_fragment_program_code *code;
-	struct gl_program *program;
-};
+extern GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg);
 
-extern GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler);
+extern void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src);
+
+extern GLboolean r500_transform_TEX(
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
+	void* data);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
index 4631235f0d..d694725c9b 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
@@ -45,47 +45,22 @@
 
 #include "r500_fragprog.h"
 
+#include "../r300_reg.h"
+
 #include "radeon_program_pair.h"
 
 
 #define PROG_CODE \
-	struct r500_fragment_program_compiler *c = (struct r500_fragment_program_compiler*)data; \
-	struct r500_fragment_program_code *code = c->code
+	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
+	struct r500_fragment_program_code *code = &c->code->code.r500
 
 #define error(fmt, args...) do {			\
-		fprintf(stderr, "%s::%s(): " fmt "\n",	\
+		rc_error(&c->Base, "%s::%s(): " fmt "\n",	\
 			__FILE__, __FUNCTION__, ##args);	\
 	} while(0)
 
 
-/**
- * Callback to register hardware constants.
- */
-static GLboolean emit_const(void *data, GLuint file, GLuint idx, GLuint *hwindex)
-{
-	PROG_CODE;
-
-	for (*hwindex = 0; *hwindex < code->const_nr; ++*hwindex) {
-		if (code->constant[*hwindex].File == file &&
-		    code->constant[*hwindex].Index == idx)
-			break;
-	}
-
-	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
-			error("Out of hw constants!\n");
-			return GL_FALSE;
-		}
-
-		code->const_nr++;
-		code->constant[*hwindex].File = file;
-		code->constant[*hwindex].Index = idx;
-	}
-
-	return GL_TRUE;
-}
-
-static GLuint translate_rgb_op(GLuint opcode)
+static GLuint translate_rgb_op(struct r300_fragment_program_compiler *c, GLuint opcode)
 {
 	switch(opcode) {
 	case OPCODE_CMP: return R500_ALU_RGBA_OP_CMP;
@@ -106,7 +81,7 @@ static GLuint translate_rgb_op(GLuint opcode)
 	}
 }
 
-static GLuint translate_alpha_op(GLuint opcode)
+static GLuint translate_alpha_op(struct r300_fragment_program_compiler *c, GLuint opcode)
 {
 	switch(opcode) {
 	case OPCODE_CMP: return R500_ALPHA_OP_CMP;
@@ -189,8 +164,8 @@ static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst)
 
 	int ip = ++code->inst_end;
 
-	code->inst[ip].inst5 = translate_rgb_op(inst->RGB.Opcode);
-	code->inst[ip].inst4 = translate_alpha_op(inst->Alpha.Opcode);
+	code->inst[ip].inst5 = translate_rgb_op(c, inst->RGB.Opcode);
+	code->inst[ip].inst4 = translate_alpha_op(c, inst->Alpha.Opcode);
 
 	if (inst->RGB.OutputWriteMask || inst->Alpha.OutputWriteMask || inst->Alpha.DepthWriteMask)
 		code->inst[ip].inst0 = R500_INST_TYPE_OUT;
@@ -202,7 +177,7 @@ static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst)
 	code->inst[ip].inst0 |= (inst->RGB.OutputWriteMask << 15) | (inst->Alpha.OutputWriteMask << 18);
 	if (inst->Alpha.DepthWriteMask) {
 		code->inst[ip].inst4 |= R500_ALPHA_W_OMASK;
-		c->fp->writes_depth = GL_TRUE;
+		c->code->writes_depth = GL_TRUE;
 	}
 
 	code->inst[ip].inst4 |= R500_ALPHA_ADDRD(inst->Alpha.DestIndex);
@@ -234,19 +209,19 @@ static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst)
 	return GL_TRUE;
 }
 
-static GLuint translate_strq_swizzle(struct prog_src_register src)
+static GLuint translate_strq_swizzle(GLuint swizzle)
 {
 	GLuint swiz = 0;
 	int i;
 	for (i = 0; i < 4; i++)
-		swiz |= (GET_SWZ(src.Swizzle, i) & 0x3) << i*2;
+		swiz |= (GET_SWZ(swizzle, i) & 0x3) << i*2;
 	return swiz;
 }
 
 /**
  * Emit a single TEX instruction
  */
-static GLboolean emit_tex(void *data, struct prog_instruction *inst)
+static GLboolean emit_tex(void *data, struct radeon_pair_texture_instruction *inst)
 {
 	PROG_CODE;
 
@@ -258,7 +233,7 @@ static GLboolean emit_tex(void *data, struct prog_instruction *inst)
 	int ip = ++code->inst_end;
 
 	code->inst[ip].inst0 = R500_INST_TYPE_TEX
-		| (inst->DstReg.WriteMask << 11)
+		| (inst->WriteMask << 11)
 		| R500_INST_TEX_SEM_WAIT;
 	code->inst[ip].inst1 = R500_TEX_ID(inst->TexSrcUnit)
 		| R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED;
@@ -267,25 +242,25 @@ static GLboolean emit_tex(void *data, struct prog_instruction *inst)
 	        code->inst[ip].inst1 |= R500_TEX_UNSCALED;
 
 	switch (inst->Opcode) {
-	case OPCODE_KIL:
+	case RADEON_OPCODE_KIL:
 		code->inst[ip].inst1 |= R500_TEX_INST_TEXKILL;
 		break;
-	case OPCODE_TEX:
+	case RADEON_OPCODE_TEX:
 		code->inst[ip].inst1 |= R500_TEX_INST_LD;
 		break;
-	case OPCODE_TXB:
+	case RADEON_OPCODE_TXB:
 		code->inst[ip].inst1 |= R500_TEX_INST_LODBIAS;
 		break;
-	case OPCODE_TXP:
+	case RADEON_OPCODE_TXP:
 		code->inst[ip].inst1 |= R500_TEX_INST_PROJ;
 		break;
 	default:
 		error("emit_tex can't handle opcode %x\n", inst->Opcode);
 	}
 
-	code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcReg[0].Index)
-		| (translate_strq_swizzle(inst->SrcReg[0]) << 8)
-		| R500_TEX_DST_ADDR(inst->DstReg.Index)
+	code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcIndex)
+		| (translate_strq_swizzle(inst->SrcSwizzle) << 8)
+		| R500_TEX_DST_ADDR(inst->DestIndex)
 		| R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G
 		| R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A;
 
@@ -293,35 +268,32 @@ static GLboolean emit_tex(void *data, struct prog_instruction *inst)
 }
 
 static const struct radeon_pair_handler pair_handler = {
-	.EmitConst = emit_const,
 	.EmitPaired = emit_paired,
 	.EmitTex = emit_tex,
 	.MaxHwTemps = 128
 };
 
-GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler)
+void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r500_fragment_program_code *code = compiler->code;
+	struct r500_fragment_program_code *code = &compiler->code->code.r500;
 
 	_mesa_bzero(code, sizeof(*code));
 	code->max_temp_idx = 1;
-	code->inst_offset = 0;
 	code->inst_end = -1;
 
-	if (!radeonPairProgram(compiler->r300->radeon.glCtx, compiler->program, &pair_handler, compiler))
-		return GL_FALSE;
+	radeonPairProgram(compiler, &pair_handler, compiler);
+	if (compiler->Base.Error)
+		return;
 
 	if ((code->inst[code->inst_end].inst0 & R500_INST_TYPE_MASK) != R500_INST_TYPE_OUT) {
 		/* This may happen when dead-code elimination is disabled or
 		 * when most of the fragment program logic is leading to a KIL */
 		if (code->inst_end >= 511) {
-			error("Introducing fake OUT: Too many instructions");
-			return GL_FALSE;
+			rc_error(&compiler->Base, "Introducing fake OUT: Too many instructions");
+			return;
 		}
 
 		int ip = ++code->inst_end;
 		code->inst[ip].inst0 = R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT;
 	}
-
-	return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.c b/src/mesa/drivers/dri/r300/compiler/radeon_code.c
new file mode 100644
index 0000000000..c7923004df
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.c
@@ -0,0 +1,170 @@
+/*
+ * Copyright (C) 2009 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "main/mtypes.h"
+#include "shader/prog_instruction.h"
+
+#include "radeon_code.h"
+
+void rc_constants_init(struct rc_constant_list * c)
+{
+	memset(c, 0, sizeof(*c));
+}
+
+/**
+ * Copy a constants structure, assuming that the destination structure
+ * is not initialized.
+ */
+void rc_constants_copy(struct rc_constant_list * dst, struct rc_constant_list * src)
+{
+	dst->Constants = malloc(sizeof(struct rc_constant) * src->Count);
+	memcpy(dst->Constants, src->Constants, sizeof(struct rc_constant) * src->Count);
+	dst->Count = src->Count;
+	dst->_Reserved = src->Count;
+}
+
+void rc_constants_destroy(struct rc_constant_list * c)
+{
+	free(c->Constants);
+	memset(c, 0, sizeof(*c));
+}
+
+unsigned rc_constants_add(struct rc_constant_list * c, struct rc_constant * constant)
+{
+	unsigned index = c->Count;
+
+	if (c->Count >= c->_Reserved) {
+		struct rc_constant * newlist;
+
+		c->_Reserved = c->_Reserved * 2;
+		if (!c->_Reserved)
+			c->_Reserved = 16;
+
+		newlist = malloc(sizeof(struct rc_constant) * c->_Reserved);
+		memcpy(newlist, c->Constants, sizeof(struct rc_constant) * c->Count);
+
+		free(c->Constants);
+		c->Constants = newlist;
+	}
+
+	c->Constants[index] = *constant;
+	c->Count++;
+
+	return index;
+}
+
+
+/**
+ * Add a state vector to the constant list, while trying to avoid duplicates.
+ */
+unsigned rc_constants_add_state(struct rc_constant_list * c, unsigned state0, unsigned state1)
+{
+	unsigned index;
+	struct rc_constant constant;
+
+	for(index = 0; index < c->Count; ++index) {
+		if (c->Constants[index].Type == RC_CONSTANT_STATE) {
+			if (c->Constants[index].u.State[0] == state0 &&
+			    c->Constants[index].u.State[1] == state1)
+				return index;
+		}
+	}
+
+	memset(&constant, 0, sizeof(constant));
+	constant.Type = RC_CONSTANT_STATE;
+	constant.Size = 4;
+	constant.u.State[0] = state0;
+	constant.u.State[1] = state1;
+
+	return rc_constants_add(c, &constant);
+}
+
+
+/**
+ * Add an immediate vector to the constant list, while trying to avoid
+ * duplicates.
+ */
+unsigned rc_constants_add_immediate_vec4(struct rc_constant_list * c, const float * data)
+{
+	unsigned index;
+	struct rc_constant constant;
+
+	for(index = 0; index < c->Count; ++index) {
+		if (c->Constants[index].Type == RC_CONSTANT_IMMEDIATE) {
+			if (!memcmp(c->Constants[index].u.Immediate, data, sizeof(float)*4))
+				return index;
+		}
+	}
+
+	memset(&constant, 0, sizeof(constant));
+	constant.Type = RC_CONSTANT_IMMEDIATE;
+	constant.Size = 4;
+	memcpy(constant.u.Immediate, data, sizeof(float) * 4);
+
+	return rc_constants_add(c, &constant);
+}
+
+
+/**
+ * Add an immediate scalar to the constant list, while trying to avoid
+ * duplicates.
+ */
+unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float data, unsigned * swizzle)
+{
+	unsigned index;
+	int free_index = -1;
+	struct rc_constant constant;
+
+	for(index = 0; index < c->Count; ++index) {
+		if (c->Constants[index].Type == RC_CONSTANT_IMMEDIATE) {
+			for(unsigned comp = 0; comp < c->Constants[index].Size; ++comp) {
+				if (c->Constants[index].u.Immediate[comp] == data) {
+					*swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp);
+					return index;
+				}
+			}
+
+			if (c->Constants[index].Size < 4)
+				free_index = index;
+		}
+	}
+
+	if (free_index >= 0) {
+		unsigned comp = c->Constants[free_index].Size++;
+		c->Constants[free_index].u.Immediate[comp] = data;
+		*swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp);
+		return free_index;
+	}
+
+	memset(&constant, 0, sizeof(constant));
+	constant.Type = RC_CONSTANT_IMMEDIATE;
+	constant.Size = 1;
+	constant.u.Immediate[0] = data;
+	*swizzle = SWIZZLE_XXXX;
+
+	return rc_constants_add(c, &constant);
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.h b/src/mesa/drivers/dri/r300/compiler/radeon_code.h
new file mode 100644
index 0000000000..3e88554ba1
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.h
@@ -0,0 +1,207 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef RADEON_CODE_H
+#define RADEON_CODE_H
+
+#include <stdint.h>
+
+#define R300_PFS_MAX_ALU_INST     64
+#define R300_PFS_MAX_TEX_INST     32
+#define R300_PFS_MAX_TEX_INDIRECT 4
+#define R300_PFS_NUM_TEMP_REGS    32
+#define R300_PFS_NUM_CONST_REGS   32
+
+#define R500_PFS_MAX_INST         512
+#define R500_PFS_NUM_TEMP_REGS    128
+#define R500_PFS_NUM_CONST_REGS   256
+
+
+#define STATE_R300_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
+
+enum {
+	/**
+	 * External constants are constants whose meaning is unknown to this
+	 * compiler. For example, a Mesa gl_program's constants are turned
+	 * into external constants.
+	 */
+	RC_CONSTANT_EXTERNAL = 0,
+
+	RC_CONSTANT_IMMEDIATE,
+
+	/**
+	 * Constant referring to state that is known by this compiler,
+	 * see RC_STATE_xxx, i.e. *not* arbitrary Mesa (or other) state.
+	 */
+	RC_CONSTANT_STATE
+};
+
+enum {
+	RC_STATE_SHADOW_AMBIENT = 0,
+
+	RC_STATE_R300_WINDOW_DIMENSION,
+	RC_STATE_R300_TEXRECT_FACTOR
+};
+
+struct rc_constant {
+	unsigned Type:2; /**< RC_CONSTANT_xxx */
+	unsigned Size:3;
+
+	union {
+		unsigned External;
+		float Immediate[4];
+		unsigned State[2];
+	} u;
+};
+
+struct rc_constant_list {
+	struct rc_constant * Constants;
+	unsigned Count;
+
+	unsigned _Reserved;
+};
+
+void rc_constants_init(struct rc_constant_list * c);
+void rc_constants_copy(struct rc_constant_list * dst, struct rc_constant_list * src);
+void rc_constants_destroy(struct rc_constant_list * c);
+unsigned rc_constants_add(struct rc_constant_list * c, struct rc_constant * constant);
+unsigned rc_constants_add_state(struct rc_constant_list * c, unsigned state1, unsigned state2);
+unsigned rc_constants_add_immediate_vec4(struct rc_constant_list * c, const float * data);
+unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float data, unsigned * swizzle);
+
+/**
+ * Stores state that influences the compilation of a fragment program.
+ */
+struct r300_fragment_program_external_state {
+	struct {
+		/**
+		 * If the sampler is used as a shadow sampler,
+		 * this field is:
+		 *  0 - GL_LUMINANCE
+		 *  1 - GL_INTENSITY
+		 *  2 - GL_ALPHA
+		 * depending on the depth texture mode.
+		 */
+		unsigned depth_texture_mode : 2;
+
+		/**
+		 * If the sampler is used as a shadow sampler,
+		 * this field is (texture_compare_func - GL_NEVER).
+		 * [e.g. if compare function is GL_LEQUAL, this field is 3]
+		 *
+		 * Otherwise, this field is 0.
+		 */
+		unsigned texture_compare_func : 3;
+	} unit[16];
+};
+
+
+
+struct r300_fragment_program_node {
+	int tex_offset; /**< first tex instruction */
+	int tex_end; /**< last tex instruction, relative to tex_offset */
+	int alu_offset; /**< first ALU instruction */
+	int alu_end; /**< last ALU instruction, relative to alu_offset */
+	int flags;
+};
+
+/**
+ * Stores an R300 fragment program in its compiled-to-hardware form.
+ */
+struct r300_fragment_program_code {
+	struct {
+		int length; /**< total # of texture instructions used */
+		uint32_t inst[R300_PFS_MAX_TEX_INST];
+	} tex;
+
+	struct {
+		int length; /**< total # of ALU instructions used */
+		struct {
+			uint32_t rgb_inst;
+			uint32_t rgb_addr;
+			uint32_t alpha_inst;
+			uint32_t alpha_addr;
+		} inst[R300_PFS_MAX_ALU_INST];
+	} alu;
+
+	uint32_t config; /* US_CONFIG */
+	uint32_t pixsize; /* US_PIXSIZE */
+	uint32_t code_offset; /* US_CODE_OFFSET */
+	uint32_t code_addr[4]; /* US_CODE_ADDR */
+};
+
+
+struct r500_fragment_program_code {
+	struct {
+		uint32_t inst0;
+		uint32_t inst1;
+		uint32_t inst2;
+		uint32_t inst3;
+		uint32_t inst4;
+		uint32_t inst5;
+	} inst[R500_PFS_MAX_INST];
+
+	int inst_end; /* Number of instructions - 1; also, last instruction to be executed */
+
+	int max_temp_idx;
+};
+
+struct rX00_fragment_program_code {
+	union {
+		struct r300_fragment_program_code r300;
+		struct r500_fragment_program_code r500;
+	} code;
+
+	unsigned writes_depth:1;
+
+	struct rc_constant_list constants;
+};
+
+
+#define VSF_MAX_FRAGMENT_LENGTH (255*4)
+#define VSF_MAX_FRAGMENT_TEMPS (14)
+
+#define VSF_MAX_INPUTS 32
+#define VSF_MAX_OUTPUTS 32
+
+struct r300_vertex_program_code {
+	int length;
+	union {
+		uint32_t d[VSF_MAX_FRAGMENT_LENGTH];
+		float f[VSF_MAX_FRAGMENT_LENGTH];
+	} body;
+
+	int pos_end;
+	int num_temporaries;	/* Number of temp vars used by program */
+	int inputs[VSF_MAX_INPUTS];
+	int outputs[VSF_MAX_OUTPUTS];
+
+	struct rc_constant_list constants;
+
+	uint32_t InputsRead;
+	uint32_t OutputsWritten;
+};
+
+void r300_vertex_program_dump(struct r300_vertex_program_code * vs);
+
+#endif /* RADEON_CODE_H */
+
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c
new file mode 100644
index 0000000000..da950d5289
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c
@@ -0,0 +1,262 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "radeon_compiler.h"
+
+#include <stdarg.h>
+
+#include "radeon_program.h"
+
+
+void rc_init(struct radeon_compiler * c)
+{
+	memset(c, 0, sizeof(*c));
+
+	memory_pool_init(&c->Pool);
+	c->Program.Instructions.Prev = &c->Program.Instructions;
+	c->Program.Instructions.Next = &c->Program.Instructions;
+	c->Program.Instructions.I.Opcode = OPCODE_END;
+}
+
+void rc_destroy(struct radeon_compiler * c)
+{
+	rc_constants_destroy(&c->Program.Constants);
+	memory_pool_destroy(&c->Pool);
+	free(c->ErrorMsg);
+}
+
+void rc_debug(struct radeon_compiler * c, const char * fmt, ...)
+{
+	va_list ap;
+
+	if (!c->Debug)
+		return;
+
+	va_start(ap, fmt);
+	vfprintf(stderr, fmt, ap);
+	va_end(ap);
+}
+
+void rc_error(struct radeon_compiler * c, const char * fmt, ...)
+{
+	va_list ap;
+
+	c->Error = GL_TRUE;
+
+	if (!c->ErrorMsg) {
+		/* Only remember the first error */
+		char buf[1024];
+		int written;
+
+		va_start(ap, fmt);
+		written = vsnprintf(buf, sizeof(buf), fmt, ap);
+		va_end(ap);
+
+		if (written < sizeof(buf)) {
+			c->ErrorMsg = strdup(buf);
+		} else {
+			c->ErrorMsg = malloc(written + 1);
+
+			va_start(ap, fmt);
+			vsnprintf(c->ErrorMsg, written + 1, fmt, ap);
+			va_end(ap);
+		}
+	}
+
+	if (c->Debug) {
+		fprintf(stderr, "r300compiler error: ");
+
+		va_start(ap, fmt);
+		vfprintf(stderr, fmt, ap);
+		va_end(ap);
+	}
+}
+
+/**
+ * Rewrite the program such that everything that source the given input
+ * register will source new_input instead.
+ */
+void rc_move_input(struct radeon_compiler * c, unsigned input, struct prog_src_register new_input)
+{
+	struct rc_instruction * inst;
+
+	c->Program.InputsRead &= ~(1 << input);
+
+	for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) {
+		const unsigned numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode);
+		unsigned i;
+
+		for(i = 0; i < numsrcs; ++i) {
+			if (inst->I.SrcReg[i].File == PROGRAM_INPUT && inst->I.SrcReg[i].Index == input) {
+				inst->I.SrcReg[i].File = new_input.File;
+				inst->I.SrcReg[i].Index = new_input.Index;
+				inst->I.SrcReg[i].Swizzle = combine_swizzles(new_input.Swizzle, inst->I.SrcReg[i].Swizzle);
+				if (!inst->I.SrcReg[i].Abs) {
+					inst->I.SrcReg[i].Negate ^= new_input.Negate;
+					inst->I.SrcReg[i].Abs = new_input.Abs;
+				}
+
+				c->Program.InputsRead |= 1 << new_input.Index;
+			}
+		}
+	}
+}
+
+
+/**
+ * Rewrite the program such that everything that writes into the given
+ * output register will instead write to new_output. The new_output
+ * writemask is honoured.
+ */
+void rc_move_output(struct radeon_compiler * c, unsigned output, unsigned new_output, unsigned writemask)
+{
+	struct rc_instruction * inst;
+
+	c->Program.OutputsWritten &= ~(1 << output);
+
+	for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) {
+		const unsigned numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode);
+
+		if (numdsts) {
+			if (inst->I.DstReg.File == PROGRAM_OUTPUT && inst->I.DstReg.Index == output) {
+				inst->I.DstReg.Index = new_output;
+				inst->I.DstReg.WriteMask &= writemask;
+
+				c->Program.OutputsWritten |= 1 << new_output;
+			}
+		}
+	}
+}
+
+
+/**
+ * Rewrite the program such that a given output is duplicated.
+ */
+void rc_copy_output(struct radeon_compiler * c, unsigned output, unsigned dup_output)
+{
+	unsigned tempreg = rc_find_free_temporary(c);
+	struct rc_instruction * inst;
+
+	for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) {
+		const unsigned numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode);
+
+		if (numdsts) {
+			if (inst->I.DstReg.File == PROGRAM_OUTPUT && inst->I.DstReg.Index == output) {
+				inst->I.DstReg.File = PROGRAM_TEMPORARY;
+				inst->I.DstReg.Index = tempreg;
+			}
+		}
+	}
+
+	inst = rc_insert_new_instruction(c, c->Program.Instructions.Prev);
+	inst->I.Opcode = OPCODE_MOV;
+	inst->I.DstReg.File = PROGRAM_OUTPUT;
+	inst->I.DstReg.Index = output;
+
+	inst->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+	inst->I.SrcReg[0].Index = tempreg;
+	inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	inst = rc_insert_new_instruction(c, c->Program.Instructions.Prev);
+	inst->I.Opcode = OPCODE_MOV;
+	inst->I.DstReg.File = PROGRAM_OUTPUT;
+	inst->I.DstReg.Index = dup_output;
+
+	inst->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+	inst->I.SrcReg[0].Index = tempreg;
+	inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	c->Program.OutputsWritten |= 1 << dup_output;
+}
+
+
+/**
+ * Introduce standard code fragment to deal with fragment.position.
+ */
+void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsigned new_input)
+{
+	unsigned tempregi = rc_find_free_temporary(c);
+
+	c->Program.InputsRead &= ~(1 << wpos);
+	c->Program.InputsRead |= 1 << new_input;
+
+	/* perspective divide */
+	struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, &c->Program.Instructions);
+	inst_rcp->I.Opcode = OPCODE_RCP;
+
+	inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY;
+	inst_rcp->I.DstReg.Index = tempregi;
+	inst_rcp->I.DstReg.WriteMask = WRITEMASK_W;
+
+	inst_rcp->I.SrcReg[0].File = PROGRAM_INPUT;
+	inst_rcp->I.SrcReg[0].Index = new_input;
+	inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW;
+
+	struct rc_instruction * inst_mul = rc_insert_new_instruction(c, inst_rcp);
+	inst_mul->I.Opcode = OPCODE_MUL;
+
+	inst_mul->I.DstReg.File = PROGRAM_TEMPORARY;
+	inst_mul->I.DstReg.Index = tempregi;
+	inst_mul->I.DstReg.WriteMask = WRITEMASK_XYZ;
+
+	inst_mul->I.SrcReg[0].File = PROGRAM_INPUT;
+	inst_mul->I.SrcReg[0].Index = new_input;
+
+	inst_mul->I.SrcReg[1].File = PROGRAM_TEMPORARY;
+	inst_mul->I.SrcReg[1].Index = tempregi;
+	inst_mul->I.SrcReg[1].Swizzle = SWIZZLE_WWWW;
+
+	/* viewport transformation */
+	struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_mul);
+	inst_mad->I.Opcode = OPCODE_MAD;
+
+	inst_mad->I.DstReg.File = PROGRAM_TEMPORARY;
+	inst_mad->I.DstReg.Index = tempregi;
+	inst_mad->I.DstReg.WriteMask = WRITEMASK_XYZ;
+
+	inst_mad->I.SrcReg[0].File = PROGRAM_TEMPORARY;
+	inst_mad->I.SrcReg[0].Index = tempregi;
+	inst_mad->I.SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	inst_mad->I.SrcReg[1].File = PROGRAM_STATE_VAR;
+	inst_mad->I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_WINDOW_DIMENSION, 0);
+	inst_mad->I.SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	inst_mad->I.SrcReg[2].File = PROGRAM_STATE_VAR;
+	inst_mad->I.SrcReg[2].Index = inst_mad->I.SrcReg[1].Index;
+	inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	struct rc_instruction * inst;
+	for (inst = inst_mad->Next; inst != &c->Program.Instructions; inst = inst->Next) {
+		const unsigned numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode);
+		unsigned i;
+
+		for(i = 0; i < numsrcs; i++) {
+			if (inst->I.SrcReg[i].File == PROGRAM_INPUT &&
+			    inst->I.SrcReg[i].Index == wpos) {
+				inst->I.SrcReg[i].File = PROGRAM_TEMPORARY;
+				inst->I.SrcReg[i].Index = tempregi;
+			}
+		}
+	}
+}
+
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
new file mode 100644
index 0000000000..e63ab8840a
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
@@ -0,0 +1,108 @@
+/*
+ * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef RADEON_COMPILER_H
+#define RADEON_COMPILER_H
+
+#include "main/mtypes.h"
+#include "shader/prog_instruction.h"
+
+#include "memory_pool.h"
+#include "radeon_code.h"
+
+
+struct rc_instruction {
+	struct rc_instruction * Prev;
+	struct rc_instruction * Next;
+	struct prog_instruction I;
+};
+
+struct rc_program {
+	/**
+	 * Instructions.Next points to the first instruction,
+	 * Instructions.Prev points to the last instruction.
+	 */
+	struct rc_instruction Instructions;
+
+	/* Long term, we should probably remove InputsRead & OutputsWritten,
+	 * since updating dependent state can be fragile, and they aren't
+	 * actually used very often. */
+	uint32_t InputsRead;
+	uint32_t OutputsWritten;
+	uint32_t ShadowSamplers; /**< Texture units used for shadow sampling. */
+
+	struct rc_constant_list Constants;
+};
+
+struct radeon_compiler {
+	struct memory_pool Pool;
+	struct rc_program Program;
+	unsigned Debug:1;
+	unsigned Error:1;
+	char * ErrorMsg;
+};
+
+void rc_init(struct radeon_compiler * c);
+void rc_destroy(struct radeon_compiler * c);
+
+void rc_debug(struct radeon_compiler * c, const char * fmt, ...);
+void rc_error(struct radeon_compiler * c, const char * fmt, ...);
+
+void rc_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program);
+
+void rc_calculate_inputs_outputs(struct radeon_compiler * c);
+
+void rc_move_input(struct radeon_compiler * c, unsigned input, struct prog_src_register new_input);
+void rc_move_output(struct radeon_compiler * c, unsigned output, unsigned new_output, unsigned writemask);
+void rc_copy_output(struct radeon_compiler * c, unsigned output, unsigned dup_output);
+void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsigned new_input);
+
+struct r300_fragment_program_compiler {
+	struct radeon_compiler Base;
+	struct rX00_fragment_program_code *code;
+	struct r300_fragment_program_external_state state;
+	unsigned is_r500;
+	unsigned OutputDepth;
+	unsigned OutputColor;
+
+	void * UserData;
+	void (*AllocateHwInputs)(
+		struct r300_fragment_program_compiler * c,
+		void (*allocate)(void * data, unsigned input, unsigned hwreg),
+		void * mydata);
+};
+
+void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c);
+
+
+struct r300_vertex_program_compiler {
+	struct radeon_compiler Base;
+	struct r300_vertex_program_code *code;
+	GLbitfield RequiredOutputs;
+
+	void * UserData;
+	void (*SetHwInputOutput)(struct r300_vertex_program_compiler * c);
+};
+
+void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* c);
+
+#endif /* RADEON_COMPILER_H */
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.c
index 4a2e1cba40..aaaa50ad1f 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.c
@@ -36,6 +36,8 @@
 
 #include "radeon_nqssadce.h"
 
+#include "radeon_compiler.h"
+
 
 /**
  * Return the @ref register_state for the given register (or 0 for untracked
@@ -46,6 +48,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
 	switch(file) {
 	case PROGRAM_TEMPORARY: return &s->Temps[index];
 	case PROGRAM_OUTPUT: return &s->Outputs[index];
+	case PROGRAM_ADDRESS: return &s->Address;
 	default: return 0;
 	}
 }
@@ -56,7 +59,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
  *
  * @note Works correctly only for X, Y, Z, W swizzles, not for constant swizzles.
  */
-static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
 {
 	struct prog_src_register tmp = srcreg;
 	int i;
@@ -75,9 +78,10 @@ static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_reg
 }
 
 
-static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
-	struct prog_instruction *inst, GLint src, GLuint sourced)
+static void track_used_srcreg(struct nqssadce_state* s,
+	GLint src, GLuint sourced)
 {
+	struct prog_instruction * inst = &s->IP->I;
 	int i;
 	GLuint deswz_source = 0;
 
@@ -94,12 +98,11 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 	if (!s->Descr->IsNativeSwizzle(inst->Opcode, inst->SrcReg[src])) {
 		struct prog_dst_register dstreg = inst->DstReg;
 		dstreg.File = PROGRAM_TEMPORARY;
-		dstreg.Index = _mesa_find_free_register(s->Program, PROGRAM_TEMPORARY);
+		dstreg.Index = rc_find_free_temporary(s->Compiler);
 		dstreg.WriteMask = sourced;
 
 		s->Descr->BuildSwizzle(s, dstreg, inst->SrcReg[src]);
 
-		inst = s->Program->Instructions + s->IP;
 		inst->SrcReg[src].File = PROGRAM_TEMPORARY;
 		inst->SrcReg[src].Index = dstreg.Index;
 		inst->SrcReg[src].Swizzle = 0;
@@ -114,67 +117,38 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 		deswz_source = sourced;
 	}
 
-	struct register_state *regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
-	if (regstate)
-		regstate->Sourced |= deswz_source & 0xf;
-
-	return inst;
-}
+	struct register_state *regstate;
 
-
-static void rewrite_depth_out(struct prog_instruction *inst)
-{
-	if (inst->DstReg.WriteMask & WRITEMASK_Z) {
-		inst->DstReg.WriteMask = WRITEMASK_W;
+	if (inst->SrcReg[src].RelAddr) {
+		regstate = get_reg_state(s, PROGRAM_ADDRESS, 0);
+		if (regstate)
+			regstate->Sourced |= WRITEMASK_X;
 	} else {
-		inst->DstReg.WriteMask = 0;
-		return;
-	}
-
-	switch (inst->Opcode) {
-	case OPCODE_FRC:
-	case OPCODE_MOV:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		break;
-	case OPCODE_ADD:
-	case OPCODE_MAX:
-	case OPCODE_MIN:
-	case OPCODE_MUL:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		break;
-	case OPCODE_CMP:
-	case OPCODE_MAD:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
-		break;
-	default:
-		// Scalar instructions needn't be reswizzled
-		break;
+		regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
+		if (regstate)
+			regstate->Sourced |= deswz_source & 0xf;
 	}
 }
 
-static void unalias_srcregs(struct prog_instruction *inst, GLuint oldindex, GLuint newindex)
+static void unalias_srcregs(struct rc_instruction *inst, GLuint oldindex, GLuint newindex)
 {
-	int nsrc = _mesa_num_inst_src_regs(inst->Opcode);
+	int nsrc = _mesa_num_inst_src_regs(inst->I.Opcode);
 	int i;
 	for(i = 0; i < nsrc; ++i)
-		if (inst->SrcReg[i].File == PROGRAM_TEMPORARY && inst->SrcReg[i].Index == oldindex)
-			inst->SrcReg[i].Index = newindex;
+		if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY && inst->I.SrcReg[i].Index == oldindex)
+			inst->I.SrcReg[i].Index = newindex;
 }
 
 static void unalias_temporary(struct nqssadce_state* s, GLuint oldindex)
 {
-	GLuint newindex = _mesa_find_free_register(s->Program, PROGRAM_TEMPORARY);
-	int ip;
-	for(ip = 0; ip < s->IP; ++ip) {
-		struct prog_instruction* inst = s->Program->Instructions + ip;
-		if (inst->DstReg.File == PROGRAM_TEMPORARY && inst->DstReg.Index == oldindex)
-			inst->DstReg.Index = newindex;
+	GLuint newindex = rc_find_free_temporary(s->Compiler);
+	struct rc_instruction * inst;
+	for(inst = s->Compiler->Program.Instructions.Next; inst != s->IP; inst = inst->Next) {
+		if (inst->I.DstReg.File == PROGRAM_TEMPORARY && inst->I.DstReg.Index == oldindex)
+			inst->I.DstReg.Index = newindex;
 		unalias_srcregs(inst, oldindex, newindex);
 	}
-	unalias_srcregs(s->Program->Instructions + s->IP, oldindex, newindex);
+	unalias_srcregs(s->IP, oldindex, newindex);
 }
 
 
@@ -183,20 +157,16 @@ static void unalias_temporary(struct nqssadce_state* s, GLuint oldindex)
  */
 static void process_instruction(struct nqssadce_state* s)
 {
-	struct prog_instruction *inst = s->Program->Instructions + s->IP;
+	struct prog_instruction *inst = &s->IP->I;
+	GLuint WriteMask;
 
 	if (inst->Opcode == OPCODE_END)
 		return;
 
 	if (inst->Opcode != OPCODE_KIL) {
-		if (s->Descr->RewriteDepthOut) {
-			if (inst->DstReg.File == PROGRAM_OUTPUT && inst->DstReg.Index == FRAG_RESULT_DEPTH)
-				rewrite_depth_out(inst);
-		}
-
 		struct register_state *regstate = get_reg_state(s, inst->DstReg.File, inst->DstReg.Index);
 		if (!regstate) {
-			_mesa_problem(s->Ctx, "NqssaDce: bad destination register (%i[%i])\n",
+			rc_error(s->Compiler, "NqssaDce: bad destination register (%i[%i])\n",
 				inst->DstReg.File, inst->DstReg.Index);
 			return;
 		}
@@ -205,7 +175,9 @@ static void process_instruction(struct nqssadce_state* s)
 		regstate->Sourced &= ~inst->DstReg.WriteMask;
 
 		if (inst->DstReg.WriteMask == 0) {
-			_mesa_delete_instructions(s->Program, s->IP, 1);
+			struct rc_instruction * inst_remove = s->IP;
+			s->IP = s->IP->Prev;
+			rc_remove_instruction(inst_remove);
 			return;
 		}
 
@@ -213,28 +185,30 @@ static void process_instruction(struct nqssadce_state* s)
 			unalias_temporary(s, inst->DstReg.Index);
 	}
 
-	/* Attention: Due to swizzle emulation code, the following
-	 * might change the instruction stream under us, so we have
-	 * to be careful with the inst pointer. */
+	WriteMask = inst->DstReg.WriteMask;
+
 	switch (inst->Opcode) {
+	case OPCODE_ARL:
 	case OPCODE_DDX:
 	case OPCODE_DDY:
 	case OPCODE_FRC:
 	case OPCODE_MOV:
-		inst = track_used_srcreg(s, inst, 0, inst->DstReg.WriteMask);
+		track_used_srcreg(s, 0, WriteMask);
 		break;
 	case OPCODE_ADD:
 	case OPCODE_MAX:
 	case OPCODE_MIN:
 	case OPCODE_MUL:
-		inst = track_used_srcreg(s, inst, 0, inst->DstReg.WriteMask);
-		inst = track_used_srcreg(s, inst, 1, inst->DstReg.WriteMask);
+	case OPCODE_SGE:
+	case OPCODE_SLT:
+		track_used_srcreg(s, 0, WriteMask);
+		track_used_srcreg(s, 1, WriteMask);
 		break;
 	case OPCODE_CMP:
 	case OPCODE_MAD:
-		inst = track_used_srcreg(s, inst, 0, inst->DstReg.WriteMask);
-		inst = track_used_srcreg(s, inst, 1, inst->DstReg.WriteMask);
-		inst = track_used_srcreg(s, inst, 2, inst->DstReg.WriteMask);
+		track_used_srcreg(s, 0, WriteMask);
+		track_used_srcreg(s, 1, WriteMask);
+		track_used_srcreg(s, 2, WriteMask);
 		break;
 	case OPCODE_COS:
 	case OPCODE_EX2:
@@ -242,42 +216,79 @@ static void process_instruction(struct nqssadce_state* s)
 	case OPCODE_RCP:
 	case OPCODE_RSQ:
 	case OPCODE_SIN:
-		inst = track_used_srcreg(s, inst, 0, 0x1);
+		track_used_srcreg(s, 0, 0x1);
 		break;
 	case OPCODE_DP3:
-		inst = track_used_srcreg(s, inst, 0, 0x7);
-		inst = track_used_srcreg(s, inst, 1, 0x7);
+		track_used_srcreg(s, 0, 0x7);
+		track_used_srcreg(s, 1, 0x7);
 		break;
 	case OPCODE_DP4:
-		inst = track_used_srcreg(s, inst, 0, 0xf);
-		inst = track_used_srcreg(s, inst, 1, 0xf);
+		track_used_srcreg(s, 0, 0xf);
+		track_used_srcreg(s, 1, 0xf);
 		break;
 	case OPCODE_KIL:
 	case OPCODE_TEX:
 	case OPCODE_TXB:
 	case OPCODE_TXP:
-		inst = track_used_srcreg(s, inst, 0, 0xf);
+		track_used_srcreg(s, 0, 0xf);
+		break;
+	case OPCODE_DST:
+		track_used_srcreg(s, 0, 0x6);
+		track_used_srcreg(s, 1, 0xa);
+		break;
+	case OPCODE_EXP:
+	case OPCODE_LOG:
+	case OPCODE_POW:
+		track_used_srcreg(s, 0, 0x3);
+		break;
+	case OPCODE_LIT:
+		track_used_srcreg(s, 0, 0xb);
 		break;
 	default:
-		_mesa_problem(s->Ctx, "NqssaDce: Unknown opcode %d\n", inst->Opcode);
+		rc_error(s->Compiler, "NqssaDce: Unknown opcode %d\n", inst->Opcode);
 		return;
 	}
+
+	s->IP = s->IP->Prev;
 }
 
+void rc_calculate_inputs_outputs(struct radeon_compiler * c)
+{
+	struct rc_instruction *inst;
+
+	c->Program.InputsRead = 0;
+	c->Program.OutputsWritten = 0;
+
+	for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next)
+	{
+		int i;
+		int num_src_regs = _mesa_num_inst_src_regs(inst->I.Opcode);
+
+		for (i = 0; i < num_src_regs; ++i) {
+			if (inst->I.SrcReg[i].File == PROGRAM_INPUT)
+				c->Program.InputsRead |= 1 << inst->I.SrcReg[i].Index;
+		}
+
+		if (_mesa_num_inst_dst_regs(inst->I.Opcode)) {
+			if (inst->I.DstReg.File == PROGRAM_OUTPUT)
+				c->Program.OutputsWritten |= 1 << inst->I.DstReg.Index;
+		}
+	}
+}
 
-void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr)
+void radeonNqssaDce(struct radeon_compiler * c, struct radeon_nqssadce_descr* descr, void * data)
 {
 	struct nqssadce_state s;
 
 	_mesa_bzero(&s, sizeof(s));
-	s.Ctx = ctx;
-	s.Program = p;
+	s.Compiler = c;
 	s.Descr = descr;
+	s.UserData = data;
 	s.Descr->Init(&s);
-	s.IP = p->NumInstructions;
+	s.IP = c->Program.Instructions.Prev;
 
-	while(s.IP > 0) {
-		s.IP--;
+	while(s.IP != &c->Program.Instructions && !c->Error)
 		process_instruction(&s);
-	}
+
+	rc_calculate_inputs_outputs(c);
 }
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.h b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.h
index a4f94abcb6..b3fc77a35a 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.h
@@ -30,7 +30,6 @@
 
 #include "radeon_program.h"
 
-
 struct register_state {
 	/**
 	 * Bitmask indicating which components of the register are sourced
@@ -44,20 +43,22 @@ struct register_state {
  * read from, etc.
  */
 struct nqssadce_state {
-	GLcontext *Ctx;
-	struct gl_program *Program;
+	struct radeon_compiler *Compiler;
 	struct radeon_nqssadce_descr *Descr;
 
 	/**
 	 * All instructions after this instruction pointer have been dealt with.
 	 */
-	int IP;
+	struct rc_instruction * IP;
 
 	/**
 	 * Which registers are read by subsequent instructions?
 	 */
 	struct register_state Temps[MAX_PROGRAM_TEMPS];
 	struct register_state Outputs[VERT_RESULT_MAX];
+	struct register_state Address;
+
+	void * UserData;
 };
 
 
@@ -82,15 +83,9 @@ struct radeon_nqssadce_descr {
 	 * The transformation will work recursively on the emitted instruction(s).
 	 */
 	void (*BuildSwizzle)(struct nqssadce_state*, struct prog_dst_register dst, struct prog_src_register src);
-
-	/**
-	 * Rewrite instructions that write to DEPR.z to write to DEPR.w
-	 * instead (rewriting is done *before* the WriteMask test).
-	 */
-	GLboolean RewriteDepthOut;
-	void *Data;
 };
 
-void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr);
+void radeonNqssaDce(struct radeon_compiler * c, struct radeon_nqssadce_descr* descr, void * data);
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg);
 
 #endif /* __RADEON_PROGRAM_NQSSADCE_H_ */
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.c b/src/mesa/drivers/dri/r300/compiler/radeon_program.c
new file mode 100644
index 0000000000..bbbf0dd776
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.c
@@ -0,0 +1,187 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "radeon_program.h"
+
+#include "radeon_compiler.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+
+/**
+ * Transform the given clause in the following way:
+ *  1. Replace it with an empty clause
+ *  2. For every instruction in the original clause, try the given
+ *     transformations in order.
+ *  3. If one of the transformations returns GL_TRUE, assume that it
+ *     has emitted the appropriate instruction(s) into the new clause;
+ *     otherwise, copy the instruction verbatim.
+ *
+ * \note The transformation is currently not recursive; in other words,
+ * instructions emitted by transformations are not transformed.
+ *
+ * \note The transform is called 'local' because it can only look at
+ * one instruction at a time.
+ */
+void radeonLocalTransform(
+	struct radeon_compiler * c,
+	int num_transformations,
+	struct radeon_program_transformation* transformations)
+{
+	struct rc_instruction * inst = c->Program.Instructions.Next;
+
+	while(inst != &c->Program.Instructions) {
+		struct rc_instruction * current = inst;
+		int i;
+
+		inst = inst->Next;
+
+		for(i = 0; i < num_transformations; ++i) {
+			struct radeon_program_transformation* t = transformations + i;
+
+			if (t->function(c, current, t->userData))
+				break;
+		}
+	}
+}
+
+
+GLint rc_find_free_temporary(struct radeon_compiler * c)
+{
+	GLboolean used[MAX_PROGRAM_TEMPS];
+	GLuint i;
+
+	memset(used, 0, sizeof(used));
+
+	for (struct rc_instruction * rcinst = c->Program.Instructions.Next; rcinst != &c->Program.Instructions; rcinst = rcinst->Next) {
+		const struct prog_instruction *inst = &rcinst->I;
+		const GLuint nsrc = _mesa_num_inst_src_regs(inst->Opcode);
+		const GLuint ndst = _mesa_num_inst_dst_regs(inst->Opcode);
+		GLuint k;
+
+		for (k = 0; k < nsrc; k++) {
+			if (inst->SrcReg[k].File == PROGRAM_TEMPORARY)
+				used[inst->SrcReg[k].Index] = GL_TRUE;
+		}
+
+		if (ndst) {
+			if (inst->DstReg.File == PROGRAM_TEMPORARY)
+				used[inst->DstReg.Index] = GL_TRUE;
+		}
+	}
+
+	for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
+		if (!used[i])
+			return i;
+	}
+
+	return -1;
+}
+
+
+struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c)
+{
+	struct rc_instruction * inst = memory_pool_malloc(&c->Pool, sizeof(struct rc_instruction));
+
+	inst->Prev = 0;
+	inst->Next = 0;
+
+	_mesa_init_instructions(&inst->I, 1);
+
+	return inst;
+}
+
+
+struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after)
+{
+	struct rc_instruction * inst = rc_alloc_instruction(c);
+
+	inst->Prev = after;
+	inst->Next = after->Next;
+
+	inst->Prev->Next = inst;
+	inst->Next->Prev = inst;
+
+	return inst;
+}
+
+void rc_remove_instruction(struct rc_instruction * inst)
+{
+	inst->Prev->Next = inst->Next;
+	inst->Next->Prev = inst->Prev;
+}
+
+
+void rc_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program)
+{
+	struct prog_instruction *source;
+	unsigned int i;
+
+	for(source = program->Instructions; source->Opcode != OPCODE_END; ++source) {
+		struct rc_instruction * dest = rc_insert_new_instruction(c, c->Program.Instructions.Prev);
+		dest->I = *source;
+	}
+
+	c->Program.ShadowSamplers = program->ShadowSamplers;
+	c->Program.InputsRead = program->InputsRead;
+	c->Program.OutputsWritten = program->OutputsWritten;
+
+	for(i = 0; i < program->Parameters->NumParameters; ++i) {
+		struct rc_constant constant;
+
+		constant.Type = RC_CONSTANT_EXTERNAL;
+		constant.Size = 4;
+		constant.u.External = i;
+
+		rc_constants_add(&c->Program.Constants, &constant);
+	}
+}
+
+
+/**
+ * Print program to stderr, default options.
+ */
+void rc_print_program(const struct rc_program *prog)
+{
+	GLuint indent = 0;
+	GLuint linenum = 1;
+	struct rc_instruction *inst;
+
+	fprintf(stderr, "# Radeon Compiler Program\n");
+
+	for(inst = prog->Instructions.Next; inst != &prog->Instructions; inst = inst->Next) {
+		fprintf(stderr, "%3d: ", linenum);
+
+		/* Massive hack: We rely on the fact that the printers do not actually
+		 * use the gl_program argument (last argument) in debug mode */
+		indent = _mesa_fprint_instruction_opt(
+				stderr, &inst->I,
+				indent, PROG_PRINT_DEBUG, 0);
+
+		linenum++;
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/radeon_program.h b/src/mesa/drivers/dri/r300/compiler/radeon_program.h
index b411f69bc8..561958608c 100644
--- a/src/mesa/drivers/dri/r300/radeon_program.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.h
@@ -34,12 +34,9 @@
 #include "shader/program.h"
 #include "shader/prog_instruction.h"
 
-
-enum {
-	CLAUSE_MIXED = 0,
-	CLAUSE_ALU,
-	CLAUSE_TEX
-};
+struct radeon_compiler;
+struct rc_instruction;
+struct rc_program;
 
 enum {
 	PROGRAM_BUILTIN = PROGRAM_FILE_MAX /**< not a real register, but a special swizzle constant */
@@ -52,18 +49,43 @@ enum {
 #define SWIZZLE_0000 MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO)
 #define SWIZZLE_1111 MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE)
 
-/**
- * Transformation context that is passed to local transformations.
- *
- * Care must be taken with some operations during transformation,
- * e.g. finding new temporary registers must use @ref radeonFindFreeTemporary
- */
-struct radeon_transform_context {
-	GLcontext *Ctx;
-	struct gl_program *Program;
-	struct prog_instruction *OldInstructions;
-	GLuint OldNumInstructions;
-};
+static inline GLuint get_swz(GLuint swz, GLuint idx)
+{
+	if (idx & 0x4)
+		return idx;
+	return GET_SWZ(swz, idx);
+}
+
+static inline GLuint combine_swizzles4(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, swz_x);
+	ret |= get_swz(src, swz_y) << 3;
+	ret |= get_swz(src, swz_z) << 6;
+	ret |= get_swz(src, swz_w) << 9;
+
+	return ret;
+}
+
+static inline GLuint combine_swizzles(GLuint src, GLuint swz)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_X));
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Y)) << 3;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Z)) << 6;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_W)) << 9;
+
+	return ret;
+}
+
+static INLINE void reset_srcreg(struct prog_src_register* reg)
+{
+	_mesa_bzero(reg, sizeof(*reg));
+	reg->Swizzle = SWIZZLE_NOOP;
+}
+
 
 /**
  * A transformation that can be passed to \ref radeonLocalTransform.
@@ -77,23 +99,23 @@ struct radeon_transform_context {
  */
 struct radeon_program_transformation {
 	GLboolean (*function)(
-		struct radeon_transform_context*,
-		struct prog_instruction*,
+		struct radeon_compiler*,
+		struct rc_instruction*,
 		void*);
 	void *userData;
 };
 
 void radeonLocalTransform(
-	GLcontext* ctx,
-	struct gl_program *program,
+	struct radeon_compiler *c,
 	int num_transformations,
 	struct radeon_program_transformation* transformations);
 
-/**
- * Find a usable free temporary register during program transformation
- */
-GLint radeonFindFreeTemporary(struct radeon_transform_context *ctx);
+GLint rc_find_free_temporary(struct radeon_compiler * c);
+
+struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c);
+struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after);
+void rc_remove_instruction(struct rc_instruction * inst);
 
-struct prog_instruction *radeonAppendInstructions(struct gl_program *program, int count);
+void rc_print_program(const struct rc_program *prog);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
index 8283723bad..8071899eaa 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_alu.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
@@ -35,49 +35,52 @@
 
 #include "radeon_program_alu.h"
 
-#include "shader/prog_parameter.h"
+#include "radeon_compiler.h"
 
 
-static struct prog_instruction *emit1(struct gl_program* p,
+static struct rc_instruction *emit1(
+	struct radeon_compiler * c, struct rc_instruction * after,
 	gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg,
 	struct prog_src_register SrcReg)
 {
-	struct prog_instruction *fpi = radeonAppendInstructions(p, 1);
+	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
 
-	fpi->Opcode = Opcode;
-	fpi->SaturateMode = Saturate;
-	fpi->DstReg = DstReg;
-	fpi->SrcReg[0] = SrcReg;
+	fpi->I.Opcode = Opcode;
+	fpi->I.SaturateMode = Saturate;
+	fpi->I.DstReg = DstReg;
+	fpi->I.SrcReg[0] = SrcReg;
 	return fpi;
 }
 
-static struct prog_instruction *emit2(struct gl_program* p,
+static struct rc_instruction *emit2(
+	struct radeon_compiler * c, struct rc_instruction * after,
 	gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg,
 	struct prog_src_register SrcReg0, struct prog_src_register SrcReg1)
 {
-	struct prog_instruction *fpi = radeonAppendInstructions(p, 1);
+	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
 
-	fpi->Opcode = Opcode;
-	fpi->SaturateMode = Saturate;
-	fpi->DstReg = DstReg;
-	fpi->SrcReg[0] = SrcReg0;
-	fpi->SrcReg[1] = SrcReg1;
+	fpi->I.Opcode = Opcode;
+	fpi->I.SaturateMode = Saturate;
+	fpi->I.DstReg = DstReg;
+	fpi->I.SrcReg[0] = SrcReg0;
+	fpi->I.SrcReg[1] = SrcReg1;
 	return fpi;
 }
 
-static struct prog_instruction *emit3(struct gl_program* p,
+static struct rc_instruction *emit3(
+	struct radeon_compiler * c, struct rc_instruction * after,
 	gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg,
 	struct prog_src_register SrcReg0, struct prog_src_register SrcReg1,
 	struct prog_src_register SrcReg2)
 {
-	struct prog_instruction *fpi = radeonAppendInstructions(p, 1);
-
-	fpi->Opcode = Opcode;
-	fpi->SaturateMode = Saturate;
-	fpi->DstReg = DstReg;
-	fpi->SrcReg[0] = SrcReg0;
-	fpi->SrcReg[1] = SrcReg1;
-	fpi->SrcReg[2] = SrcReg2;
+	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
+
+	fpi->I.Opcode = Opcode;
+	fpi->I.SaturateMode = Saturate;
+	fpi->I.DstReg = DstReg;
+	fpi->I.SrcReg[0] = SrcReg0;
+	fpi->I.SrcReg[1] = SrcReg1;
+	fpi->I.SrcReg[2] = SrcReg2;
 	return fpi;
 }
 
@@ -88,6 +91,7 @@ static struct prog_dst_register dstreg(int file, int index)
 	dst.Index = index;
 	dst.WriteMask = WRITEMASK_XYZW;
 	dst.CondMask = COND_TR;
+	dst.RelAddr = 0;
 	dst.CondSwizzle = SWIZZLE_NOOP;
 	dst.CondSrc = 0;
 	dst.pad = 0;
@@ -96,10 +100,11 @@ static struct prog_dst_register dstreg(int file, int index)
 
 static struct prog_dst_register dstregtmpmask(int index, int mask)
 {
-	struct prog_dst_register dst;
+	struct prog_dst_register dst = {0};
 	dst.File = PROGRAM_TEMPORARY;
 	dst.Index = index;
 	dst.WriteMask = mask;
+	dst.RelAddr = 0;
 	dst.CondMask = COND_TR;
 	dst.CondSwizzle = SWIZZLE_NOOP;
 	dst.CondSrc = 0;
@@ -171,44 +176,63 @@ static struct prog_src_register scalar(struct prog_src_register reg)
 	return swizzle(reg, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X);
 }
 
-static void transform_ABS(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_ABS(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	struct prog_src_register src = inst->SrcReg[0];
+	struct prog_src_register src = inst->I.SrcReg[0];
 	src.Abs = 1;
 	src.Negate = NEGATE_NONE;
-	emit1(t->Program, OPCODE_MOV, inst->SaturateMode, inst->DstReg, src);
+	emit1(c, inst->Prev, OPCODE_MOV, inst->I.SaturateMode, inst->I.DstReg, src);
+	rc_remove_instruction(inst);
 }
 
-static void transform_DPH(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_DP3(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	struct prog_src_register src0 = inst->SrcReg[0];
+	struct prog_src_register src0 = inst->I.SrcReg[0];
+	struct prog_src_register src1 = inst->I.SrcReg[1];
+	src0.Negate &= ~NEGATE_W;
+	src0.Swizzle &= ~(7 << (3 * 3));
+	src0.Swizzle |= SWIZZLE_ZERO << (3 * 3);
+	src1.Negate &= ~NEGATE_W;
+	src1.Swizzle &= ~(7 << (3 * 3));
+	src1.Swizzle |= SWIZZLE_ZERO << (3 * 3);
+	emit2(c, inst->Prev, OPCODE_DP4, inst->I.SaturateMode, inst->I.DstReg, src0, src1);
+	rc_remove_instruction(inst);
+}
+
+static void transform_DPH(struct radeon_compiler* c,
+	struct rc_instruction* inst)
+{
+	struct prog_src_register src0 = inst->I.SrcReg[0];
 	src0.Negate &= ~NEGATE_W;
 	src0.Swizzle &= ~(7 << (3 * 3));
 	src0.Swizzle |= SWIZZLE_ONE << (3 * 3);
-	emit2(t->Program, OPCODE_DP4, inst->SaturateMode, inst->DstReg, src0, inst->SrcReg[1]);
+	emit2(c, inst->Prev, OPCODE_DP4, inst->I.SaturateMode, inst->I.DstReg, src0, inst->I.SrcReg[1]);
+	rc_remove_instruction(inst);
 }
 
 /**
  * [1, src0.y*src1.y, src0.z, src1.w]
  * So basically MUL with lotsa swizzling.
  */
-static void transform_DST(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_DST(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	emit2(t->Program, OPCODE_MUL, inst->SaturateMode, inst->DstReg,
-		swizzle(inst->SrcReg[0], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE),
-		swizzle(inst->SrcReg[1], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_ONE, SWIZZLE_W));
+	emit2(c, inst->Prev, OPCODE_MUL, inst->I.SaturateMode, inst->I.DstReg,
+		swizzle(inst->I.SrcReg[0], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE),
+		swizzle(inst->I.SrcReg[1], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_ONE, SWIZZLE_W));
+	rc_remove_instruction(inst);
 }
 
-static void transform_FLR(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_FLR(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
-	emit1(t->Program, OPCODE_FRC, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->SrcReg[0]);
-	emit2(t->Program, OPCODE_ADD, inst->SaturateMode, inst->DstReg,
-		inst->SrcReg[0], negate(srcreg(PROGRAM_TEMPORARY, tempreg)));
+	int tempreg = rc_find_free_temporary(c);
+	emit1(c, inst->Prev, OPCODE_FRC, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0]);
+	emit2(c, inst->Prev, OPCODE_ADD, inst->I.SaturateMode, inst->I.DstReg,
+		inst->I.SrcReg[0], negate(srcreg(PROGRAM_TEMPORARY, tempreg)));
+	rc_remove_instruction(inst);
 }
 
 /**
@@ -229,152 +253,159 @@ static void transform_FLR(struct radeon_transform_context* t,
  * 5 slots, if the subsequent optimization passes are clever enough
  * to pair instructions correctly.
  */
-static void transform_LIT(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_LIT(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	static const GLfloat LitConst[4] = { -127.999999 };
-
 	GLuint constant;
 	GLuint constant_swizzle;
 	GLuint temp;
-	int needTemporary = 0;
 	struct prog_src_register srctemp;
 
-	constant = _mesa_add_unnamed_constant(t->Program->Parameters, LitConst, 1, &constant_swizzle);
+	constant = rc_constants_add_immediate_scalar(&c->Program.Constants, -127.999999, &constant_swizzle);
 
-	if (inst->DstReg.WriteMask != WRITEMASK_XYZW) {
-		needTemporary = 1;
-	} else if (inst->DstReg.File != PROGRAM_TEMPORARY) {
-		// LIT is typically followed by DP3/DP4, so there's no point
-		// in creating special code for this case
-		needTemporary = 1;
-	}
+	if (inst->I.DstReg.WriteMask != WRITEMASK_XYZW || inst->I.DstReg.File != PROGRAM_TEMPORARY) {
+		struct rc_instruction * inst_mov;
 
-	if (needTemporary) {
-		temp = radeonFindFreeTemporary(t);
-	} else {
-		temp = inst->DstReg.Index;
+		inst_mov = emit1(c, inst,
+			OPCODE_MOV, 0, inst->I.DstReg,
+			srcreg(PROGRAM_TEMPORARY, rc_find_free_temporary(c)));
+
+		inst->I.DstReg.File = PROGRAM_TEMPORARY;
+		inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index;
+		inst->I.DstReg.WriteMask = WRITEMASK_XYZW;
 	}
+
+	temp = inst->I.DstReg.Index;
 	srctemp = srcreg(PROGRAM_TEMPORARY, temp);
 
 	// tmp.x = max(0.0, Src.x);
 	// tmp.y = max(0.0, Src.y);
 	// tmp.w = clamp(Src.z, -128+eps, 128-eps);
-	emit2(t->Program, OPCODE_MAX, 0,
+	emit2(c, inst->Prev, OPCODE_MAX, 0,
 		dstregtmpmask(temp, WRITEMASK_XYW),
-		inst->SrcReg[0],
+		inst->I.SrcReg[0],
 		swizzle(srcreg(PROGRAM_CONSTANT, constant),
 			SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, constant_swizzle&3));
-	emit2(t->Program, OPCODE_MIN, 0,
+	emit2(c, inst->Prev, OPCODE_MIN, 0,
 		dstregtmpmask(temp, WRITEMASK_Z),
 		swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 		negate(srcregswz(PROGRAM_CONSTANT, constant, constant_swizzle)));
 
 	// tmp.w = Pow(tmp.y, tmp.w)
-	emit1(t->Program, OPCODE_LG2, 0,
+	emit1(c, inst->Prev, OPCODE_LG2, 0,
 		dstregtmpmask(temp, WRITEMASK_W),
 		swizzle(srctemp, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y));
-	emit2(t->Program, OPCODE_MUL, 0,
+	emit2(c, inst->Prev, OPCODE_MUL, 0,
 		dstregtmpmask(temp, WRITEMASK_W),
 		swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 		swizzle(srctemp, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z));
-	emit1(t->Program, OPCODE_EX2, 0,
+	emit1(c, inst->Prev, OPCODE_EX2, 0,
 		dstregtmpmask(temp, WRITEMASK_W),
 		swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W));
 
 	// tmp.z = (tmp.x > 0) ? tmp.w : 0.0
-	emit3(t->Program, OPCODE_CMP, inst->SaturateMode,
+	emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode,
 		dstregtmpmask(temp, WRITEMASK_Z),
 		negate(swizzle(srctemp, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)),
 		swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 		builtin_zero);
 
 	// tmp.x, tmp.y, tmp.w = 1.0, tmp.x, 1.0
-	emit1(t->Program, OPCODE_MOV, inst->SaturateMode,
+	emit1(c, inst->Prev, OPCODE_MOV, inst->I.SaturateMode,
 		dstregtmpmask(temp, WRITEMASK_XYW),
 		swizzle(srctemp, SWIZZLE_ONE, SWIZZLE_X, SWIZZLE_ONE, SWIZZLE_ONE));
 
-	if (needTemporary)
-		emit1(t->Program, OPCODE_MOV, 0, inst->DstReg, srctemp);
+	rc_remove_instruction(inst);
 }
 
-static void transform_LRP(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_LRP(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
+	int tempreg = rc_find_free_temporary(c);
 
-	emit2(t->Program, OPCODE_ADD, 0,
+	emit2(c, inst->Prev, OPCODE_ADD, 0,
 		dstreg(PROGRAM_TEMPORARY, tempreg),
-		inst->SrcReg[1], negate(inst->SrcReg[2]));
-	emit3(t->Program, OPCODE_MAD, inst->SaturateMode,
-		inst->DstReg,
-		inst->SrcReg[0], srcreg(PROGRAM_TEMPORARY, tempreg), inst->SrcReg[2]);
+		inst->I.SrcReg[1], negate(inst->I.SrcReg[2]));
+	emit3(c, inst->Prev, OPCODE_MAD, inst->I.SaturateMode,
+		inst->I.DstReg,
+		inst->I.SrcReg[0], srcreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[2]);
+
+	rc_remove_instruction(inst);
 }
 
-static void transform_POW(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_POW(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
+	int tempreg = rc_find_free_temporary(c);
 	struct prog_dst_register tempdst = dstreg(PROGRAM_TEMPORARY, tempreg);
 	struct prog_src_register tempsrc = srcreg(PROGRAM_TEMPORARY, tempreg);
 	tempdst.WriteMask = WRITEMASK_W;
 	tempsrc.Swizzle = SWIZZLE_WWWW;
 
-	emit1(t->Program, OPCODE_LG2, 0, tempdst, scalar(inst->SrcReg[0]));
-	emit2(t->Program, OPCODE_MUL, 0, tempdst, tempsrc, scalar(inst->SrcReg[1]));
-	emit1(t->Program, OPCODE_EX2, inst->SaturateMode, inst->DstReg, tempsrc);
+	emit1(c, inst->Prev, OPCODE_LG2, 0, tempdst, scalar(inst->I.SrcReg[0]));
+	emit2(c, inst->Prev, OPCODE_MUL, 0, tempdst, tempsrc, scalar(inst->I.SrcReg[1]));
+	emit1(c, inst->Prev, OPCODE_EX2, inst->I.SaturateMode, inst->I.DstReg, tempsrc);
+
+	rc_remove_instruction(inst);
 }
 
-static void transform_RSQ(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_RSQ(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	emit1(t->Program, OPCODE_RSQ, inst->SaturateMode, inst->DstReg, absolute(inst->SrcReg[0]));
+	inst->I.SrcReg[0] = absolute(inst->I.SrcReg[0]);
 }
 
-static void transform_SGE(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_SGE(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
+	int tempreg = rc_find_free_temporary(c);
 
-	emit2(t->Program, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->SrcReg[0], negate(inst->SrcReg[1]));
-	emit3(t->Program, OPCODE_CMP, inst->SaturateMode, inst->DstReg,
+	emit2(c, inst->Prev, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0], negate(inst->I.SrcReg[1]));
+	emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode, inst->I.DstReg,
 		srcreg(PROGRAM_TEMPORARY, tempreg), builtin_zero, builtin_one);
+
+	rc_remove_instruction(inst);
 }
 
-static void transform_SLT(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_SLT(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
+	int tempreg = rc_find_free_temporary(c);
 
-	emit2(t->Program, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->SrcReg[0], negate(inst->SrcReg[1]));
-	emit3(t->Program, OPCODE_CMP, inst->SaturateMode, inst->DstReg,
+	emit2(c, inst->Prev, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0], negate(inst->I.SrcReg[1]));
+	emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode, inst->I.DstReg,
 		srcreg(PROGRAM_TEMPORARY, tempreg), builtin_one, builtin_zero);
+
+	rc_remove_instruction(inst);
 }
 
-static void transform_SUB(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_SUB(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	emit2(t->Program, OPCODE_ADD, inst->SaturateMode, inst->DstReg, inst->SrcReg[0], negate(inst->SrcReg[1]));
+	inst->I.Opcode = OPCODE_ADD;
+	inst->I.SrcReg[1] = negate(inst->I.SrcReg[1]);
 }
 
-static void transform_SWZ(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_SWZ(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	emit1(t->Program, OPCODE_MOV, inst->SaturateMode, inst->DstReg, inst->SrcReg[0]);
+	inst->I.Opcode = OPCODE_MOV;
 }
 
-static void transform_XPD(struct radeon_transform_context* t,
-	struct prog_instruction* inst)
+static void transform_XPD(struct radeon_compiler* c,
+	struct rc_instruction* inst)
 {
-	int tempreg = radeonFindFreeTemporary(t);
-
-	emit2(t->Program, OPCODE_MUL, 0, dstreg(PROGRAM_TEMPORARY, tempreg),
-		swizzle(inst->SrcReg[0], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W),
-		swizzle(inst->SrcReg[1], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W));
-	emit3(t->Program, OPCODE_MAD, inst->SaturateMode, inst->DstReg,
-		swizzle(inst->SrcReg[0], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W),
-		swizzle(inst->SrcReg[1], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W),
+	int tempreg = rc_find_free_temporary(c);
+
+	emit2(c, inst->Prev, OPCODE_MUL, 0, dstreg(PROGRAM_TEMPORARY, tempreg),
+		swizzle(inst->I.SrcReg[0], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W),
+		swizzle(inst->I.SrcReg[1], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W));
+	emit3(c, inst->Prev, OPCODE_MAD, inst->I.SaturateMode, inst->I.DstReg,
+		swizzle(inst->I.SrcReg[0], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W),
+		swizzle(inst->I.SrcReg[1], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W),
 		negate(srcreg(PROGRAM_TEMPORARY, tempreg)));
+
+	rc_remove_instruction(inst);
 }
 
 
@@ -392,31 +423,64 @@ static void transform_XPD(struct radeon_transform_context* t,
  *
  * @note should be applicable to R300 and R500 fragment programs.
  */
-GLboolean radeonTransformALU(struct radeon_transform_context* t,
-	struct prog_instruction* inst,
+GLboolean radeonTransformALU(
+	struct radeon_compiler * c,
+	struct rc_instruction* inst,
 	void* unused)
 {
-	switch(inst->Opcode) {
-	case OPCODE_ABS: transform_ABS(t, inst); return GL_TRUE;
-	case OPCODE_DPH: transform_DPH(t, inst); return GL_TRUE;
-	case OPCODE_DST: transform_DST(t, inst); return GL_TRUE;
-	case OPCODE_FLR: transform_FLR(t, inst); return GL_TRUE;
-	case OPCODE_LIT: transform_LIT(t, inst); return GL_TRUE;
-	case OPCODE_LRP: transform_LRP(t, inst); return GL_TRUE;
-	case OPCODE_POW: transform_POW(t, inst); return GL_TRUE;
-	case OPCODE_RSQ: transform_RSQ(t, inst); return GL_TRUE;
-	case OPCODE_SGE: transform_SGE(t, inst); return GL_TRUE;
-	case OPCODE_SLT: transform_SLT(t, inst); return GL_TRUE;
-	case OPCODE_SUB: transform_SUB(t, inst); return GL_TRUE;
-	case OPCODE_SWZ: transform_SWZ(t, inst); return GL_TRUE;
-	case OPCODE_XPD: transform_XPD(t, inst); return GL_TRUE;
+	switch(inst->I.Opcode) {
+	case OPCODE_ABS: transform_ABS(c, inst); return GL_TRUE;
+	case OPCODE_DPH: transform_DPH(c, inst); return GL_TRUE;
+	case OPCODE_DST: transform_DST(c, inst); return GL_TRUE;
+	case OPCODE_FLR: transform_FLR(c, inst); return GL_TRUE;
+	case OPCODE_LIT: transform_LIT(c, inst); return GL_TRUE;
+	case OPCODE_LRP: transform_LRP(c, inst); return GL_TRUE;
+	case OPCODE_POW: transform_POW(c, inst); return GL_TRUE;
+	case OPCODE_RSQ: transform_RSQ(c, inst); return GL_TRUE;
+	case OPCODE_SGE: transform_SGE(c, inst); return GL_TRUE;
+	case OPCODE_SLT: transform_SLT(c, inst); return GL_TRUE;
+	case OPCODE_SUB: transform_SUB(c, inst); return GL_TRUE;
+	case OPCODE_SWZ: transform_SWZ(c, inst); return GL_TRUE;
+	case OPCODE_XPD: transform_XPD(c, inst); return GL_TRUE;
 	default:
 		return GL_FALSE;
 	}
 }
 
 
-static void sincos_constants(struct radeon_transform_context* t, GLuint *constants)
+static void transform_r300_vertex_ABS(struct radeon_compiler* c,
+	struct rc_instruction* inst)
+{
+	/* Note: r500 can take absolute values, but r300 cannot. */
+	inst->I.Opcode = OPCODE_MAX;
+	inst->I.SrcReg[1] = inst->I.SrcReg[0];
+	inst->I.SrcReg[1].Negate ^= NEGATE_XYZW;
+}
+
+/**
+ * For use with radeonLocalTransform, this transforms non-native ALU
+ * instructions of the r300 up to r500 vertex engine.
+ */
+GLboolean r300_transform_vertex_alu(
+	struct radeon_compiler * c,
+	struct rc_instruction* inst,
+	void* unused)
+{
+	switch(inst->I.Opcode) {
+	case OPCODE_ABS: transform_r300_vertex_ABS(c, inst); return GL_TRUE;
+	case OPCODE_DP3: transform_DP3(c, inst); return GL_TRUE;
+	case OPCODE_DPH: transform_DPH(c, inst); return GL_TRUE;
+	case OPCODE_FLR: transform_FLR(c, inst); return GL_TRUE;
+	case OPCODE_LRP: transform_LRP(c, inst); return GL_TRUE;
+	case OPCODE_SUB: transform_SUB(c, inst); return GL_TRUE;
+	case OPCODE_SWZ: transform_SWZ(c, inst); return GL_TRUE;
+	case OPCODE_XPD: transform_XPD(c, inst); return GL_TRUE;
+	default:
+		return GL_FALSE;
+	}
+}
+
+static void sincos_constants(struct radeon_compiler* c, GLuint *constants)
 {
 	static const GLfloat SinCosConsts[2][4] = {
 		{
@@ -434,11 +498,8 @@ static void sincos_constants(struct radeon_transform_context* t, GLuint *constan
 	};
 	int i;
 
-	for(i = 0; i < 2; ++i) {
-		GLuint swz;
-		constants[i] = _mesa_add_unnamed_constant(t->Program->Parameters, SinCosConsts[i], 4, &swz);
-		ASSERT(swz == SWIZZLE_NOOP);
-	}
+	for(i = 0; i < 2; ++i)
+		constants[i] = rc_constants_add_immediate_vec4(&c->Program.Constants, SinCosConsts[i]);
 }
 
 /**
@@ -449,23 +510,24 @@ static void sincos_constants(struct radeon_transform_context* t, GLuint *constan
  * MAD tmp.y, tmp.x, |tmp.x|, -tmp.x
  * MAD dest, tmp.y, weight, tmp.x
  */
-static void sin_approx(struct radeon_transform_context* t,
+static void sin_approx(
+	struct radeon_compiler* c, struct rc_instruction * after,
 	struct prog_dst_register dst, struct prog_src_register src, const GLuint* constants)
 {
-	GLuint tempreg = radeonFindFreeTemporary(t);
+	GLuint tempreg = rc_find_free_temporary(c);
 
-	emit2(t->Program, OPCODE_MUL, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
+	emit2(c, after->Prev, OPCODE_MUL, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
 		swizzle(src, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 		srcreg(PROGRAM_CONSTANT, constants[0]));
-	emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_X),
+	emit3(c, after->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_X),
 		swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y),
 		absolute(swizzle(src, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)),
 		swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X));
-	emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_Y),
+	emit3(c, after->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_Y),
 		swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 		absolute(swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)),
 		negate(swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)));
-	emit3(t->Program, OPCODE_MAD, 0, dst,
+	emit3(c, after->Prev, OPCODE_MAD, 0, dst,
 		swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y),
 		swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 		swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X));
@@ -476,78 +538,80 @@ static void sin_approx(struct radeon_transform_context* t,
  * using only the basic instructions
  *  MOV, ADD, MUL, MAD, FRC
  */
-GLboolean radeonTransformTrigSimple(struct radeon_transform_context* t,
-	struct prog_instruction* inst,
+GLboolean radeonTransformTrigSimple(struct radeon_compiler* c,
+	struct rc_instruction* inst,
 	void* unused)
 {
-	if (inst->Opcode != OPCODE_COS &&
-	    inst->Opcode != OPCODE_SIN &&
-	    inst->Opcode != OPCODE_SCS)
+	if (inst->I.Opcode != OPCODE_COS &&
+	    inst->I.Opcode != OPCODE_SIN &&
+	    inst->I.Opcode != OPCODE_SCS)
 		return GL_FALSE;
 
 	GLuint constants[2];
-	GLuint tempreg = radeonFindFreeTemporary(t);
+	GLuint tempreg = rc_find_free_temporary(c);
 
-	sincos_constants(t, constants);
+	sincos_constants(c, constants);
 
-	if (inst->Opcode == OPCODE_COS) {
+	if (inst->I.Opcode == OPCODE_COS) {
 		// MAD tmp.x, src, 1/(2*PI), 0.75
 		// FRC tmp.x, tmp.x
 		// MAD tmp.z, tmp.x, 2*PI, -PI
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
-			swizzle(inst->SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+			swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X));
-		emit1(t->Program, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+		emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W),
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W));
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z)));
 
-		sin_approx(t, inst->DstReg,
+		sin_approx(c, inst->Prev, inst->I.DstReg,
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			constants);
-	} else if (inst->Opcode == OPCODE_SIN) {
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
-			swizzle(inst->SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
+	} else if (inst->I.Opcode == OPCODE_SIN) {
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+			swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y));
-		emit1(t->Program, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+		emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W),
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W));
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W),
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z)));
 
-		sin_approx(t, inst->DstReg,
+		sin_approx(c, inst->Prev, inst->I.DstReg,
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			constants);
 	} else {
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
-			swizzle(inst->SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
+			swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W));
-		emit1(t->Program, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
+		emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
 			srcreg(PROGRAM_TEMPORARY, tempreg));
-		emit3(t->Program, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
+		emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY),
 			srcreg(PROGRAM_TEMPORARY, tempreg),
 			swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W),
 			negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z)));
 
-		struct prog_dst_register dst = inst->DstReg;
+		struct prog_dst_register dst = inst->I.DstReg;
 
-		dst.WriteMask = inst->DstReg.WriteMask & WRITEMASK_X;
-		sin_approx(t, dst,
+		dst.WriteMask = inst->I.DstReg.WriteMask & WRITEMASK_X;
+		sin_approx(c, inst->Prev, dst,
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 			constants);
 
-		dst.WriteMask = inst->DstReg.WriteMask & WRITEMASK_Y;
-		sin_approx(t, dst,
+		dst.WriteMask = inst->I.DstReg.WriteMask & WRITEMASK_Y;
+		sin_approx(c, inst->Prev, dst,
 			swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y),
 			constants);
 	}
 
+	rc_remove_instruction(inst);
+
 	return GL_TRUE;
 }
 
@@ -560,50 +624,52 @@ GLboolean radeonTransformTrigSimple(struct radeon_transform_context* t,
  *
  * @warning This transformation implicitly changes the semantics of SIN and COS!
  */
-GLboolean radeonTransformTrigScale(struct radeon_transform_context* t,
-	struct prog_instruction* inst,
+GLboolean radeonTransformTrigScale(struct radeon_compiler* c,
+	struct rc_instruction* inst,
 	void* unused)
 {
-	if (inst->Opcode != OPCODE_COS &&
-	    inst->Opcode != OPCODE_SIN &&
-	    inst->Opcode != OPCODE_SCS)
+	if (inst->I.Opcode != OPCODE_COS &&
+	    inst->I.Opcode != OPCODE_SIN &&
+	    inst->I.Opcode != OPCODE_SCS)
 		return GL_FALSE;
 
-	static const GLfloat RCP_2PI[] = { 0.15915494309189535 };
+	static const GLfloat RCP_2PI = 0.15915494309189535;
 	GLuint temp;
 	GLuint constant;
 	GLuint constant_swizzle;
 
-	temp = radeonFindFreeTemporary(t);
-	constant = _mesa_add_unnamed_constant(t->Program->Parameters, RCP_2PI, 1, &constant_swizzle);
+	temp = rc_find_free_temporary(c);
+	constant = rc_constants_add_immediate_scalar(&c->Program.Constants, RCP_2PI, &constant_swizzle);
 
-	emit2(t->Program, OPCODE_MUL, 0, dstregtmpmask(temp, WRITEMASK_W),
-		swizzle(inst->SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
+	emit2(c, inst->Prev, OPCODE_MUL, 0, dstregtmpmask(temp, WRITEMASK_W),
+		swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
 		srcregswz(PROGRAM_CONSTANT, constant, constant_swizzle));
-	emit1(t->Program, OPCODE_FRC, 0, dstregtmpmask(temp, WRITEMASK_W),
+	emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(temp, WRITEMASK_W),
 		srcreg(PROGRAM_TEMPORARY, temp));
 
-	if (inst->Opcode == OPCODE_COS) {
-		emit1(t->Program, OPCODE_COS, inst->SaturateMode, inst->DstReg,
+	if (inst->I.Opcode == OPCODE_COS) {
+		emit1(c, inst->Prev, OPCODE_COS, inst->I.SaturateMode, inst->I.DstReg,
 			srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW));
-	} else if (inst->Opcode == OPCODE_SIN) {
-		emit1(t->Program, OPCODE_SIN, inst->SaturateMode,
-			inst->DstReg, srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW));
-	} else if (inst->Opcode == OPCODE_SCS) {
-		struct prog_dst_register moddst = inst->DstReg;
+	} else if (inst->I.Opcode == OPCODE_SIN) {
+		emit1(c, inst->Prev, OPCODE_SIN, inst->I.SaturateMode,
+			inst->I.DstReg, srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW));
+	} else if (inst->I.Opcode == OPCODE_SCS) {
+		struct prog_dst_register moddst = inst->I.DstReg;
 
-		if (inst->DstReg.WriteMask & WRITEMASK_X) {
+		if (inst->I.DstReg.WriteMask & WRITEMASK_X) {
 			moddst.WriteMask = WRITEMASK_X;
-			emit1(t->Program, OPCODE_COS, inst->SaturateMode, moddst,
+			emit1(c, inst->Prev, OPCODE_COS, inst->I.SaturateMode, moddst,
 				srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW));
 		}
-		if (inst->DstReg.WriteMask & WRITEMASK_Y) {
+		if (inst->I.DstReg.WriteMask & WRITEMASK_Y) {
 			moddst.WriteMask = WRITEMASK_Y;
-			emit1(t->Program, OPCODE_SIN, inst->SaturateMode, moddst,
+			emit1(c, inst->Prev, OPCODE_SIN, inst->I.SaturateMode, moddst,
 				srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW));
 		}
 	}
 
+	rc_remove_instruction(inst);
+
 	return GL_TRUE;
 }
 
@@ -615,21 +681,15 @@ GLboolean radeonTransformTrigScale(struct radeon_transform_context* t,
  * @warning This explicitly changes the form of DDX and DDY!
  */
 
-GLboolean radeonTransformDeriv(struct radeon_transform_context* t,
-	struct prog_instruction* inst,
+GLboolean radeonTransformDeriv(struct radeon_compiler* c,
+	struct rc_instruction* inst,
 	void* unused)
 {
-	if (inst->Opcode != OPCODE_DDX && inst->Opcode != OPCODE_DDY)
+	if (inst->I.Opcode != OPCODE_DDX && inst->I.Opcode != OPCODE_DDY)
 		return GL_FALSE;
 
-	struct prog_src_register B = inst->SrcReg[1];
-
-	B.Swizzle = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE,
-						SWIZZLE_ONE, SWIZZLE_ONE);
-	B.Negate = NEGATE_XYZW;
-
-	emit2(t->Program, inst->Opcode, inst->SaturateMode, inst->DstReg,
-		inst->SrcReg[0], B);
+	inst->I.SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE);
+	inst->I.SrcReg[1].Negate = NEGATE_XYZW;
 
 	return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/r300/radeon_program_alu.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h
index b45958115c..147efec6fc 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_alu.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h
@@ -31,23 +31,28 @@
 #include "radeon_program.h"
 
 GLboolean radeonTransformALU(
-	struct radeon_transform_context *t,
-	struct prog_instruction*,
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
+	void*);
+
+GLboolean r300_transform_vertex_alu(
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
 	void*);
 
 GLboolean radeonTransformTrigSimple(
-	struct radeon_transform_context *t,
-	struct prog_instruction*,
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
 	void*);
 
 GLboolean radeonTransformTrigScale(
-	struct radeon_transform_context *t,
-	struct prog_instruction*,
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
 	void*);
 
 GLboolean radeonTransformDeriv(
-	struct radeon_transform_context *t,
-	struct prog_instruction*,
+	struct radeon_compiler * c,
+	struct rc_instruction * inst,
 	void*);
 
 #endif /* __RADEON_PROGRAM_ALU_H_ */
diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c
index 2e21f7bf66..4c26db5d24 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c
@@ -35,19 +35,20 @@
 
 #include "radeon_program_pair.h"
 
-#include "radeon_context.h"
-
+#include "memory_pool.h"
+#include "radeon_compiler.h"
 #include "shader/prog_print.h"
 
 #define error(fmt, args...) do { \
-	_mesa_problem(s->Ctx, "%s::%s(): " fmt "\n",	\
+	rc_error(&s->Compiler->Base, "%s::%s(): " fmt "\n",	\
 		__FILE__, __FUNCTION__, ##args);	\
-	s->Error = GL_TRUE;				\
 } while(0)
 
 struct pair_state_instruction {
+	struct prog_instruction Instruction;
+	GLuint IP; /**< Position of this instruction in original program */
+
 	GLuint IsTex:1; /**< Is a texture instruction */
-	GLuint IsOutput:1; /**< Is output instruction */
 	GLuint NeedRGB:1; /**< Needs the RGB ALU */
 	GLuint NeedAlpha:1; /**< Needs the Alpha ALU */
 	GLuint IsTranscendent:1; /**< Is a special transcendent instruction */
@@ -74,7 +75,7 @@ struct pair_state_instruction {
  * Used to keep track of which instructions read a value.
  */
 struct reg_value_reader {
-	GLuint IP; /**< IP of the instruction that performs this access */
+	struct pair_state_instruction *Reader;
 	struct reg_value_reader *Next;
 };
 
@@ -83,7 +84,7 @@ struct reg_value_reader {
  * PROGRAM_TEMPORARY.
  */
 struct reg_value {
-	GLuint IP; /**< IP of the instruction that writes this value */
+	struct pair_state_instruction *Writer;
 	struct reg_value *Next; /**< Pointer to the next value to be written to the same PROGRAM_TEMPORARY component */
 
 	/**
@@ -117,14 +118,10 @@ struct pair_register_translation {
 };
 
 struct pair_state {
-	GLcontext *Ctx;
-	struct gl_program *Program;
+	struct r300_fragment_program_compiler * Compiler;
 	const struct radeon_pair_handler *Handler;
-	GLboolean Error;
-	GLboolean Debug;
 	GLboolean Verbose;
 	void *UserData;
-	GLubyte NumKillInsts;
 
 	/**
 	 * Translate Mesa registers to hardware registers
@@ -132,11 +129,6 @@ struct pair_state {
 	struct pair_register_translation Inputs[FRAG_ATTRIB_MAX];
 	struct pair_register_translation Temps[MAX_PROGRAM_TEMPS];
 
-	/**
-	 * Derived information about program instructions.
-	 */
-	struct pair_state_instruction *Instructions;
-
 	struct {
 		GLuint RefCount; /**< # of times this occurs in an unscheduled SrcReg or DstReg */
 	} HwTemps[128];
@@ -149,19 +141,6 @@ struct pair_state {
 	struct pair_state_instruction *ReadyRGB;
 	struct pair_state_instruction *ReadyAlpha;
 	struct pair_state_instruction *ReadyTEX;
-
-	/**
-	 * Linked list of deferred instructions
-	 */
-	struct pair_state_instruction *DeferredInsts;
-
-	/**
-	 * Pool of @ref reg_value structures for fast allocation.
-	 */
-	struct reg_value *ValuePool;
-	GLuint ValuePoolUsed;
-	struct reg_value_reader *ReaderPool;
-	GLuint ReaderPoolUsed;
 };
 
 
@@ -190,7 +169,7 @@ static GLuint get_hw_reg(struct pair_state *s, GLuint file, GLuint index)
 
 	struct pair_register_translation *t = get_register(s, file, index);
 	if (!t) {
-		_mesa_problem(s->Ctx, "get_hw_reg: %i[%i]\n", file, index);
+		error("get_hw_reg: %i[%i]\n", file, index);
 		return 0;
 	}
 
@@ -228,19 +207,15 @@ static void add_pairinst_to_list(struct pair_state_instruction **list, struct pa
 }
 
 /**
- * The instruction at the given IP has become ready. Link it into the ready
+ * The given instruction has become ready. Link it into the ready
  * instructions.
  */
-static void instruction_ready(struct pair_state *s, int ip)
+static void instruction_ready(struct pair_state *s, struct pair_state_instruction *pairinst)
 {
-	struct pair_state_instruction *pairinst = s->Instructions + ip;
-
 	if (s->Verbose)
-		_mesa_printf("instruction_ready(%i)\n", ip);
+		_mesa_printf("instruction_ready(%i)\n", pairinst->IP);
 
-	if (s->NumKillInsts > 0 && pairinst->IsOutput)
-		add_pairinst_to_list(&s->DeferredInsts, pairinst);
-	else if (pairinst->IsTex)
+	if (pairinst->IsTex)
 		add_pairinst_to_list(&s->ReadyTEX, pairinst);
 	else if (!pairinst->NeedAlpha)
 		add_pairinst_to_list(&s->ReadyRGB, pairinst);
@@ -305,12 +280,12 @@ static void final_rewrite(struct pair_state *s, struct prog_instruction *inst)
  * Classify an instruction according to which ALUs etc. it needs
  */
 static void classify_instruction(struct pair_state *s,
-	struct prog_instruction *inst, struct pair_state_instruction *pairinst)
+	struct pair_state_instruction *psi)
 {
-	pairinst->NeedRGB = (inst->DstReg.WriteMask & WRITEMASK_XYZ) ? 1 : 0;
-	pairinst->NeedAlpha = (inst->DstReg.WriteMask & WRITEMASK_W) ? 1 : 0;
+	psi->NeedRGB = (psi->Instruction.DstReg.WriteMask & WRITEMASK_XYZ) ? 1 : 0;
+	psi->NeedAlpha = (psi->Instruction.DstReg.WriteMask & WRITEMASK_W) ? 1 : 0;
 
-	switch(inst->Opcode) {
+	switch(psi->Instruction.Opcode) {
 	case OPCODE_ADD:
 	case OPCODE_CMP:
 	case OPCODE_DDX:
@@ -328,28 +303,26 @@ static void classify_instruction(struct pair_state *s,
 	case OPCODE_RCP:
 	case OPCODE_RSQ:
 	case OPCODE_SIN:
-		pairinst->IsTranscendent = 1;
-		pairinst->NeedAlpha = 1;
+		psi->IsTranscendent = 1;
+		psi->NeedAlpha = 1;
 		break;
 	case OPCODE_DP4:
-		pairinst->NeedAlpha = 1;
+		psi->NeedAlpha = 1;
 		/* fall through */
 	case OPCODE_DP3:
-		pairinst->NeedRGB = 1;
+		psi->NeedRGB = 1;
 		break;
 	case OPCODE_KIL:
 	case OPCODE_TEX:
 	case OPCODE_TXB:
 	case OPCODE_TXP:
 	case OPCODE_END:
-		pairinst->IsTex = 1;
+		psi->IsTex = 1;
 		break;
 	default:
-		error("Unknown opcode %d\n", inst->Opcode);
+		error("Unknown opcode %d\n", psi->Instruction.Opcode);
 		break;
 	}
-
-	pairinst->IsOutput = (inst->DstReg.File == PROGRAM_OUTPUT);
 }
 
 
@@ -359,30 +332,34 @@ static void classify_instruction(struct pair_state *s,
  */
 static void scan_instructions(struct pair_state *s)
 {
-	struct prog_instruction *inst;
-	struct pair_state_instruction *pairinst;
+	struct rc_instruction *source;
 	GLuint ip;
 
-	for(inst = s->Program->Instructions, pairinst = s->Instructions, ip = 0;
-	    inst->Opcode != OPCODE_END;
-	    ++inst, ++pairinst, ++ip) {
-		final_rewrite(s, inst);
-		classify_instruction(s, inst, pairinst);
+	for(source = s->Compiler->Base.Program.Instructions.Next, ip = 0;
+	    source != &s->Compiler->Base.Program.Instructions;
+	    source = source->Next, ++ip) {
+		struct pair_state_instruction *pairinst = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*pairinst));
+		memset(pairinst, 0, sizeof(struct pair_state_instruction));
 
-		int nsrc = _mesa_num_inst_src_regs(inst->Opcode);
+		pairinst->Instruction = source->I;
+		pairinst->IP = ip;
+		final_rewrite(s, &pairinst->Instruction);
+		classify_instruction(s, pairinst);
+
+		int nsrc = _mesa_num_inst_src_regs(pairinst->Instruction.Opcode);
 		int j;
 		for(j = 0; j < nsrc; j++) {
 			struct pair_register_translation *t =
-				get_register(s, inst->SrcReg[j].File, inst->SrcReg[j].Index);
+				get_register(s, pairinst->Instruction.SrcReg[j].File, pairinst->Instruction.SrcReg[j].Index);
 			if (!t)
 				continue;
 
 			t->RefCount++;
 
-			if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) {
+			if (pairinst->Instruction.SrcReg[j].File == PROGRAM_TEMPORARY) {
 				int i;
 				for(i = 0; i < 4; ++i) {
-					GLuint swz = GET_SWZ(inst->SrcReg[j].Swizzle, i);
+					GLuint swz = GET_SWZ(pairinst->Instruction.SrcReg[j].Swizzle, i);
 					if (swz >= 4)
 						continue; /* constant or NIL swizzle */
 					if (!t->Value[swz])
@@ -392,36 +369,37 @@ static void scan_instructions(struct pair_state *s)
 					 * also rewrites the value. The code below adds
 					 * a dependency for the DstReg, which is a superset
 					 * of the SrcReg dependency. */
-					if (inst->DstReg.File == PROGRAM_TEMPORARY &&
-					    inst->DstReg.Index == inst->SrcReg[j].Index &&
-					    GET_BIT(inst->DstReg.WriteMask, swz))
+					if (pairinst->Instruction.DstReg.File == PROGRAM_TEMPORARY &&
+					    pairinst->Instruction.DstReg.Index == pairinst->Instruction.SrcReg[j].Index &&
+					    GET_BIT(pairinst->Instruction.DstReg.WriteMask, swz))
 						continue;
 
-					struct reg_value_reader* r = &s->ReaderPool[s->ReaderPoolUsed++];
+					struct reg_value_reader* r = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*r));
 					pairinst->NumDependencies++;
 					t->Value[swz]->NumReaders++;
-					r->IP = ip;
+					r->Reader = pairinst;
 					r->Next = t->Value[swz]->Readers;
 					t->Value[swz]->Readers = r;
 				}
 			}
 		}
 
-		int ndst = _mesa_num_inst_dst_regs(inst->Opcode);
+		int ndst = _mesa_num_inst_dst_regs(pairinst->Instruction.Opcode);
 		if (ndst) {
 			struct pair_register_translation *t =
-				get_register(s, inst->DstReg.File, inst->DstReg.Index);
+				get_register(s, pairinst->Instruction.DstReg.File, pairinst->Instruction.DstReg.Index);
 			if (t) {
 				t->RefCount++;
 
-				if (inst->DstReg.File == PROGRAM_TEMPORARY) {
+				if (pairinst->Instruction.DstReg.File == PROGRAM_TEMPORARY) {
 					int j;
 					for(j = 0; j < 4; ++j) {
-						if (!GET_BIT(inst->DstReg.WriteMask, j))
+						if (!GET_BIT(pairinst->Instruction.DstReg.WriteMask, j))
 							continue;
 
-						struct reg_value* v = &s->ValuePool[s->ValuePoolUsed++];
-						v->IP = ip;
+						struct reg_value* v = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*v));
+						memset(v, 0, sizeof(struct reg_value));
+						v->Writer = pairinst;
 						if (t->Value[j]) {
 							pairinst->NumDependencies++;
 							t->Value[j]->Next = v;
@@ -437,7 +415,7 @@ static void scan_instructions(struct pair_state *s)
 			_mesa_printf("scan(%i): NumDeps = %i\n", ip, pairinst->NumDependencies);
 
 		if (!pairinst->NumDependencies)
-			instruction_ready(s, ip);
+			instruction_ready(s, pairinst);
 	}
 
 	/* Clear the PROGRAM_TEMPORARY state */
@@ -449,70 +427,23 @@ static void scan_instructions(struct pair_state *s)
 }
 
 
-/**
- * Reserve hardware temporary registers for the program inputs.
- *
- * @note This allocation is performed explicitly, because the order of inputs
- * is determined by the RS hardware.
- */
-static void allocate_input_registers(struct pair_state *s)
-{
-	GLuint InputsRead = s->Program->InputsRead;
-	int i;
-	GLuint hwindex = 0;
-
-	/* Primary colour */
-	if (InputsRead & FRAG_BIT_COL0)
-		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_COL0, hwindex++);
-	InputsRead &= ~FRAG_BIT_COL0;
-
-	/* Secondary color */
-	if (InputsRead & FRAG_BIT_COL1)
-		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_COL1, hwindex++);
-	InputsRead &= ~FRAG_BIT_COL1;
-
-	/* Texcoords */
-	for (i = 0; i < s->Ctx->Const.MaxTextureUnits; i++) {
-		if (InputsRead & (FRAG_BIT_TEX0 << i))
-			alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_TEX0+i, hwindex++);
-	}
-	InputsRead &= ~FRAG_BITS_TEX_ANY;
-
-	/* Fogcoords treated as a texcoord */
-	if (InputsRead & FRAG_BIT_FOGC)
-		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_FOGC, hwindex++);
-	InputsRead &= ~FRAG_BIT_FOGC;
-
-	/* fragment position treated as a texcoord */
-	if (InputsRead & FRAG_BIT_WPOS)
-		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_WPOS, hwindex++);
-	InputsRead &= ~FRAG_BIT_WPOS;
-
-	/* Anything else */
-	if (InputsRead)
-		error("Don't know how to handle inputs 0x%x\n", InputsRead);
-}
-
-
-static void decrement_dependencies(struct pair_state *s, int ip)
+static void decrement_dependencies(struct pair_state *s, struct pair_state_instruction *pairinst)
 {
-	struct pair_state_instruction *pairinst = s->Instructions + ip;
 	ASSERT(pairinst->NumDependencies > 0);
 	if (!--pairinst->NumDependencies)
-		instruction_ready(s, ip);
+		instruction_ready(s, pairinst);
 }
 
 /**
  * Update the dependency tracking state based on what the instruction
  * at the given IP does.
  */
-static void commit_instruction(struct pair_state *s, int ip)
+static void commit_instruction(struct pair_state *s, struct pair_state_instruction *pairinst)
 {
-	struct prog_instruction *inst = s->Program->Instructions + ip;
-	struct pair_state_instruction *pairinst = s->Instructions + ip;
+	struct prog_instruction *inst = &pairinst->Instruction;
 
 	if (s->Verbose)
-		_mesa_printf("commit_instruction(%i)\n", ip);
+		_mesa_printf("commit_instruction(%i)\n", pairinst->IP);
 
 	if (inst->DstReg.File == PROGRAM_TEMPORARY) {
 		struct pair_register_translation *t = &s->Temps[inst->DstReg.Index];
@@ -527,11 +458,11 @@ static void commit_instruction(struct pair_state *s, int ip)
 			if (t->Value[i]->NumReaders) {
 				struct reg_value_reader *r;
 				for(r = pairinst->Values[i]->Readers; r; r = r->Next)
-					decrement_dependencies(s, r->IP);
+					decrement_dependencies(s, r->Reader);
 			} else if (t->Value[i]->Next) {
 				/* This happens when the only reader writes
 				 * the register at the same time */
-				decrement_dependencies(s, t->Value[i]->Next->IP);
+				decrement_dependencies(s, t->Value[i]->Next->Writer);
 			}
 		}
 	}
@@ -565,7 +496,7 @@ static void commit_instruction(struct pair_state *s, int ip)
 
 			if (!--t->Value[swz]->NumReaders) {
 				if (t->Value[swz]->Next)
-					decrement_dependencies(s, t->Value[swz]->Next->IP);
+					decrement_dependencies(s, t->Value[swz]->Next->Writer);
 			}
 		}
 	}
@@ -596,38 +527,52 @@ static void emit_all_tex(struct pair_state *s)
 
 	// Allocate destination hardware registers in one block to avoid conflicts.
 	for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) {
-		int ip = pairinst - s->Instructions;
-		struct prog_instruction *inst = s->Program->Instructions + ip;
+		struct prog_instruction *inst = &pairinst->Instruction;
 		if (inst->Opcode != OPCODE_KIL)
 			get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index);
 	}
 
-	if (s->Debug)
+	if (s->Compiler->Base.Debug)
 		_mesa_printf(" BEGIN_TEX\n");
 
 	if (s->Handler->BeginTexBlock)
-		s->Error = s->Error || !s->Handler->BeginTexBlock(s->UserData);
+		s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->BeginTexBlock(s->UserData);
 
 	for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) {
-		int ip = pairinst - s->Instructions;
-		struct prog_instruction *inst = s->Program->Instructions + ip;
-		commit_instruction(s, ip);
+		struct prog_instruction *inst = &pairinst->Instruction;
+		commit_instruction(s, pairinst);
 
-		if (inst->Opcode == OPCODE_KIL)
-			--s->NumKillInsts;
-		else
+		if (inst->Opcode != OPCODE_KIL)
 			inst->DstReg.Index = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index);
-
 		inst->SrcReg[0].Index = get_hw_reg(s, inst->SrcReg[0].File, inst->SrcReg[0].Index);
 
-		if (s->Debug) {
+		if (s->Compiler->Base.Debug) {
 			_mesa_printf("   ");
 			_mesa_print_instruction(inst);
+			fflush(stderr);
+		}
+
+		struct radeon_pair_texture_instruction rpti;
+
+		switch(inst->Opcode) {
+		case OPCODE_TEX: rpti.Opcode = RADEON_OPCODE_TEX; break;
+		case OPCODE_TXB: rpti.Opcode = RADEON_OPCODE_TXB; break;
+		case OPCODE_TXP: rpti.Opcode = RADEON_OPCODE_TXP; break;
+		default:
+		case OPCODE_KIL: rpti.Opcode = RADEON_OPCODE_KIL; break;
 		}
-		s->Error = s->Error || !s->Handler->EmitTex(s->UserData, inst);
+
+		rpti.DestIndex = inst->DstReg.Index;
+		rpti.WriteMask = inst->DstReg.WriteMask;
+		rpti.TexSrcUnit = inst->TexSrcUnit;
+		rpti.TexSrcTarget = inst->TexSrcTarget;
+		rpti.SrcIndex = inst->SrcReg[0].Index;
+		rpti.SrcSwizzle = inst->SrcReg[0].Swizzle;
+
+		s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitTex(s->UserData, &rpti);
 	}
 
-	if (s->Debug)
+	if (s->Compiler->Base.Debug)
 		_mesa_printf(" END_TEX\n");
 }
 
@@ -650,7 +595,7 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio
 		index = get_hw_reg(s, src.File, src.Index);
 	} else {
 		constant = 1;
-		s->Error |= !s->Handler->EmitConst(s->UserData, src.File, src.Index, &index);
+		index = src.Index;
 	}
 
 	for(i = 0; i < 3; ++i) {
@@ -697,10 +642,12 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio
  * Fill the given ALU instruction's opcodes and source operands into the given pair,
  * if possible.
  */
-static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_pair_instruction *pair, int ip)
+static GLboolean fill_instruction_into_pair(
+	struct pair_state *s,
+	struct radeon_pair_instruction *pair,
+	struct pair_state_instruction *pairinst)
 {
-	struct pair_state_instruction *pairinst = s->Instructions + ip;
-	struct prog_instruction *inst = s->Program->Instructions + ip;
+	struct prog_instruction *inst = &pairinst->Instruction;
 
 	ASSERT(!pairinst->NeedRGB || pair->RGB.Opcode == OPCODE_NOP);
 	ASSERT(!pairinst->NeedAlpha || pair->Alpha.Opcode == OPCODE_NOP);
@@ -781,16 +728,18 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_
  * we are absolutely certain that we're going to emit a certain
  * instruction pairing.
  */
-static void fill_dest_into_pair(struct pair_state *s, struct radeon_pair_instruction *pair, int ip)
+static void fill_dest_into_pair(
+	struct pair_state *s,
+	struct radeon_pair_instruction *pair,
+	struct pair_state_instruction *pairinst)
 {
-	struct pair_state_instruction *pairinst = s->Instructions + ip;
-	struct prog_instruction *inst = s->Program->Instructions + ip;
+	struct prog_instruction *inst = &pairinst->Instruction;
 
 	if (inst->DstReg.File == PROGRAM_OUTPUT) {
-		if (inst->DstReg.Index == FRAG_RESULT_COLOR) {
+		if (inst->DstReg.Index == s->Compiler->OutputColor) {
 			pair->RGB.OutputWriteMask |= inst->DstReg.WriteMask & WRITEMASK_XYZ;
 			pair->Alpha.OutputWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3);
-		} else if (inst->DstReg.Index == FRAG_RESULT_DEPTH) {
+		} else if (inst->DstReg.Index == s->Compiler->OutputDepth) {
 			pair->Alpha.DepthWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3);
 		}
 	} else {
@@ -817,24 +766,24 @@ static void fill_dest_into_pair(struct pair_state *s, struct radeon_pair_instruc
 static void emit_alu(struct pair_state *s)
 {
 	struct radeon_pair_instruction pair;
+	struct pair_state_instruction *psi;
 
 	if (s->ReadyFullALU || !(s->ReadyRGB && s->ReadyAlpha)) {
-		int ip;
 		if (s->ReadyFullALU) {
-			ip = s->ReadyFullALU - s->Instructions;
+			psi = s->ReadyFullALU;
 			s->ReadyFullALU = s->ReadyFullALU->NextReady;
 		} else if (s->ReadyRGB) {
-			ip = s->ReadyRGB - s->Instructions;
+			psi = s->ReadyRGB;
 			s->ReadyRGB = s->ReadyRGB->NextReady;
 		} else {
-			ip = s->ReadyAlpha - s->Instructions;
+			psi = s->ReadyAlpha;
 			s->ReadyAlpha = s->ReadyAlpha->NextReady;
 		}
 
 		_mesa_bzero(&pair, sizeof(pair));
-		fill_instruction_into_pair(s, &pair, ip);
-		fill_dest_into_pair(s, &pair, ip);
-		commit_instruction(s, ip);
+		fill_instruction_into_pair(s, &pair, psi);
+		fill_dest_into_pair(s, &pair, psi);
+		commit_instruction(s, psi);
 	} else {
 		struct pair_state_instruction **prgb;
 		struct pair_state_instruction **palpha;
@@ -843,108 +792,75 @@ static void emit_alu(struct pair_state *s)
 		 * many source slots; try all possible pairings if necessary */
 		for(prgb = &s->ReadyRGB; *prgb; prgb = &(*prgb)->NextReady) {
 			for(palpha = &s->ReadyAlpha; *palpha; palpha = &(*palpha)->NextReady) {
-				int rgbip = *prgb - s->Instructions;
-				int alphaip = *palpha - s->Instructions;
+				struct pair_state_instruction * psirgb = *prgb;
+				struct pair_state_instruction * psialpha = *palpha;
 				_mesa_bzero(&pair, sizeof(pair));
-				fill_instruction_into_pair(s, &pair, rgbip);
-				if (!fill_instruction_into_pair(s, &pair, alphaip))
+				fill_instruction_into_pair(s, &pair, psirgb);
+				if (!fill_instruction_into_pair(s, &pair, psialpha))
 					continue;
 				*prgb = (*prgb)->NextReady;
 				*palpha = (*palpha)->NextReady;
-				fill_dest_into_pair(s, &pair, rgbip);
-				fill_dest_into_pair(s, &pair, alphaip);
-				commit_instruction(s, rgbip);
-				commit_instruction(s, alphaip);
+				fill_dest_into_pair(s, &pair, psirgb);
+				fill_dest_into_pair(s, &pair, psialpha);
+				commit_instruction(s, psirgb);
+				commit_instruction(s, psialpha);
 				goto success;
 			}
 		}
 
 		/* No success in pairing; just take the first RGB instruction */
-		int ip = s->ReadyRGB - s->Instructions;
+		psi = s->ReadyRGB;
 		s->ReadyRGB = s->ReadyRGB->NextReady;
+
 		_mesa_bzero(&pair, sizeof(pair));
-		fill_instruction_into_pair(s, &pair, ip);
-		fill_dest_into_pair(s, &pair, ip);
-		commit_instruction(s, ip);
+		fill_instruction_into_pair(s, &pair, psi);
+		fill_dest_into_pair(s, &pair, psi);
+		commit_instruction(s, psi);
 	success: ;
 	}
 
-	if (s->Debug)
+	if (s->Compiler->Base.Debug)
 		radeonPrintPairInstruction(&pair);
 
-	s->Error = s->Error || !s->Handler->EmitPaired(s->UserData, &pair);
+	s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitPaired(s->UserData, &pair);
 }
 
-static GLubyte countKillInsts(struct gl_program *prog)
+/* Callback function for assigning input registers to hardware registers */
+static void alloc_helper(void * data, unsigned input, unsigned hwreg)
 {
-	GLubyte i, count = 0;
-
-	for (i = 0; i < prog->NumInstructions; ++i) {
-		if (prog->Instructions[i].Opcode == OPCODE_KIL)
-			++count;
-	}
-
-	return count;
+	struct pair_state * s = data;
+	alloc_hw_reg(s, PROGRAM_INPUT, input, hwreg);
 }
 
-GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
+void radeonPairProgram(
+	struct r300_fragment_program_compiler * compiler,
 	const struct radeon_pair_handler* handler, void *userdata)
 {
 	struct pair_state s;
 
 	_mesa_bzero(&s, sizeof(s));
-	s.Ctx = ctx;
-	s.Program = program;
+	s.Compiler = compiler;
 	s.Handler = handler;
 	s.UserData = userdata;
-	s.Debug = (RADEON_DEBUG & DEBUG_PIXEL) ? GL_TRUE : GL_FALSE;
-	s.Verbose = GL_FALSE && s.Debug;
-	s.NumKillInsts = countKillInsts(program);
-
-	s.Instructions = (struct pair_state_instruction*)_mesa_calloc(
-		sizeof(struct pair_state_instruction)*s.Program->NumInstructions);
-	s.ValuePool = (struct reg_value*)_mesa_calloc(sizeof(struct reg_value)*s.Program->NumInstructions*4);
-	s.ReaderPool = (struct reg_value_reader*)_mesa_calloc(
-		sizeof(struct reg_value_reader)*s.Program->NumInstructions*12);
+	s.Verbose = GL_FALSE && s.Compiler->Base.Debug;
 
-	if (s.Debug)
+	if (s.Compiler->Base.Debug)
 		_mesa_printf("Emit paired program\n");
 
 	scan_instructions(&s);
-	allocate_input_registers(&s);
+	s.Compiler->AllocateHwInputs(s.Compiler, &alloc_helper, &s);
 
-	while(!s.Error &&
+	while(!s.Compiler->Base.Error &&
 	      (s.ReadyTEX || s.ReadyRGB || s.ReadyAlpha || s.ReadyFullALU)) {
 		if (s.ReadyTEX)
 			emit_all_tex(&s);
 
-		if (!s.NumKillInsts) {
-			struct pair_state_instruction *pairinst = s.DeferredInsts;
-			while (pairinst) {
-				if (!pairinst->NeedAlpha)
-					add_pairinst_to_list(&s.ReadyRGB, pairinst);
-				else if (!pairinst->NeedRGB)
-					add_pairinst_to_list(&s.ReadyAlpha, pairinst);
-				else
-					add_pairinst_to_list(&s.ReadyFullALU, pairinst);
-
-				pairinst = pairinst->NextReady;
-			}
-			s.DeferredInsts = NULL;
-		}
-
 		while(s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha)
 			emit_alu(&s);
 	}
 
-	if (s.Debug)
+	if (s.Compiler->Base.Debug)
 		_mesa_printf(" END\n");
-
-	_mesa_free(s.Instructions);
-	_mesa_free(s.ValuePool);
-	_mesa_free(s.ReaderPool);
-
-	return !s.Error;
 }
 
 
diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h
index 4624a24629..ff76178551 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h
@@ -30,6 +30,8 @@
 
 #include "radeon_program.h"
 
+struct r300_fragment_program_compiler;
+
 
 /**
  * Represents a paired instruction, as found in R300 and R500
@@ -82,18 +84,32 @@ struct radeon_pair_instruction {
 };
 
 
+enum {
+	RADEON_OPCODE_TEX = 0,
+	RADEON_OPCODE_TXB,
+	RADEON_OPCODE_TXP,
+	RADEON_OPCODE_KIL
+};
+
+struct radeon_pair_texture_instruction {
+	GLuint Opcode:2; /**< one of RADEON_OPCODE_xxx */
+
+	GLuint DestIndex:8;
+	GLuint WriteMask:4;
+
+	GLuint TexSrcUnit:5;
+	GLuint TexSrcTarget:3;
+
+	GLuint SrcIndex:8;
+	GLuint SrcSwizzle:12;
+};
+
+
 /**
  *
  */
 struct radeon_pair_handler {
 	/**
-	 * Fill in the proper hardware index for the given constant register.
-	 *
-	 * @return GL_FALSE on error.
-	 */
-	GLboolean (*EmitConst)(void*, GLuint file, GLuint index, GLuint *hwindex);
-
-	/**
 	 * Write a paired instruction to the hardware.
 	 *
 	 * @return GL_FALSE on error.
@@ -107,7 +123,7 @@ struct radeon_pair_handler {
 	 *
 	 * @return GL_FALSE on error.
 	 */
-	GLboolean (*EmitTex)(void*, struct prog_instruction*);
+	GLboolean (*EmitTex)(void*, struct radeon_pair_texture_instruction*);
 
 	/**
 	 * Called before a block of contiguous, independent texture
@@ -115,10 +131,11 @@ struct radeon_pair_handler {
 	 */
 	GLboolean (*BeginTexBlock)(void*);
 
-	GLuint MaxHwTemps;
+	unsigned MaxHwTemps;
 };
 
-GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
+void radeonPairProgram(
+	struct r300_fragment_program_compiler * compiler,
 	const struct radeon_pair_handler*, void *userdata);
 
 void radeonPrintPairInstruction(struct radeon_pair_instruction *inst);
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index f4472756f1..0fe32a5443 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -44,247 +44,425 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "radeon_drm.h"
 
-#include "radeon_ioctl.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "radeon_reg.h"
 #include "r300_reg.h"
 #include "r300_cmdbuf.h"
 #include "r300_emit.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_state.h"
+#include "radeon_reg.h"
+#include "radeon_queryobj.h"
 
-// Set this to 1 for extremely verbose debugging of command buffers
-#define DEBUG_CMDBUF		0
-
-/**
- * Send the current command buffer via ioctl to the hardware.
+/** # of dwords reserved for additional instructions that may need to be written
+ * during flushing.
  */
-int r300FlushCmdBufLocked(r300ContextPtr r300, const char *caller)
-{
-	int ret;
-	int i;
-	drm_radeon_cmd_buffer_t cmd;
-	int start;
-
-	if (r300->radeon.lost_context) {
-		start = 0;
-		r300->radeon.lost_context = GL_FALSE;
-	} else
-		start = r300->cmdbuf.count_reemit;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "%s from %s - %i cliprects\n",
-			__FUNCTION__, caller, r300->radeon.numClipRects);
-
-		if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_VERBOSE)
-			for (i = start; i < r300->cmdbuf.count_used; ++i)
-				fprintf(stderr, "%d: %08x\n", i,
-					r300->cmdbuf.cmd_buf[i]);
-	}
+#define SPACE_FOR_FLUSHING	4
 
-	cmd.buf = (char *)(r300->cmdbuf.cmd_buf + start);
-	cmd.bufsz = (r300->cmdbuf.count_used - start) * 4;
+static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt)
+{
+    if (r300->radeon.radeonScreen->kernel_mm) {
+        return ((((*pkt) >> 16) & 0x3FFF) + 1);
+    } else {
+        drm_r300_cmd_header_t *t = (drm_r300_cmd_header_t*)pkt;
+        return t->packet0.count;
+    }
+}
 
-	if (r300->radeon.state.scissor.enabled) {
-		cmd.nbox = r300->radeon.state.scissor.numClipRects;
-		cmd.boxes =
-		    (drm_clip_rect_t *) r300->radeon.state.scissor.pClipRects;
-	} else {
-		cmd.nbox = r300->radeon.numClipRects;
-		cmd.boxes = (drm_clip_rect_t *) r300->radeon.pClipRects;
-	}
+#define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
+#define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count)
 
-	ret = drmCommandWrite(r300->radeon.dri.fd,
-			      DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
+int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	int cnt;
+	int extra = 1;
+	cnt = vpu_count(atom->cmd);
 
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "Syncing in %s (from %s)\n\n",
-			__FUNCTION__, caller);
-		radeonWaitForIdleLocked(&r300->radeon);
+	if (r300->radeon.radeonScreen->kernel_mm) {
+		extra = 5;
 	}
 
-	r300->dma.nr_released_bufs = 0;
-	r300->cmdbuf.count_used = 0;
-	r300->cmdbuf.count_reemit = 0;
-
-	return ret;
+	return cnt ? (cnt * 4) + extra : 0;
 }
 
-int r300FlushCmdBuf(r300ContextPtr r300, const char *caller)
-{
-	int ret;
 
-	LOCK_HARDWARE(&r300->radeon);
+void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	drm_r300_cmd_header_t cmd;
+	uint32_t addr, ndw;
 
-	ret = r300FlushCmdBufLocked(r300, caller);
+	cmd.u = atom->cmd[0];
+	addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo;
+	ndw = atom->check(ctx, atom);
 
-	UNLOCK_HARDWARE(&r300->radeon);
+	BEGIN_BATCH_NO_AUTOSTATE(ndw);
 
-	if (ret) {
-		fprintf(stderr, "drmRadeonCmdBuffer: %d\n", ret);
-		_mesa_exit(ret);
-	}
-
-	return ret;
+	ndw -= 5;
+	OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr);
+	OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR);
+	OUT_BATCH_TABLE(&atom->cmd[1], ndw);
+	OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0);
+	END_BATCH();
 }
 
-static void r300PrintStateAtom(r300ContextPtr r300, struct r300_state_atom *state)
+void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom)
 {
-	int i, j, reg;
-	int dwords = (*state->check) (r300, state);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
 	drm_r300_cmd_header_t cmd;
+	uint32_t addr, ndw, sz;
+	int type, clamp;
+
+	ndw = atom->check(ctx, atom);
+
+	cmd.u = atom->cmd[0];
+	sz = cmd.r500fp.count;
+	addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo;
+	type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE);
+	clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP);
+
+	addr |= (type << 16);
+	addr |= (clamp << 17);
+
+	BEGIN_BATCH_NO_AUTOSTATE(ndw);
+	OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0));
+	OUT_BATCH(addr);
+	ndw-=3;
+	OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR);
+	OUT_BATCH_TABLE(&atom->cmd[1], ndw);
+	END_BATCH();
+}
 
-	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords,
-		state->cmd_size);
-
-	if (RADEON_DEBUG & DEBUG_VERBOSE) {
-		for (i = 0; i < dwords;) {
-			cmd = *((drm_r300_cmd_header_t *) &state->cmd[i]);
-			reg = (cmd.packet0.reghi << 8) | cmd.packet0.reglo;
-			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
-					state->name, i, reg, cmd.packet0.count);
-			++i;
-			for (j = 0; j < cmd.packet0.count; j++) {
-				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
-					state->name, i, reg, state->cmd[i]);
-				reg += 4;
-				++i;
-			}
-		}
+static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	int numtmus = packet0_count(r300, r300->hw.tex.offset.cmd);
+	int dw = 0, i;
+	if (atom->cmd[0] == CP_PACKET2) {
+		return dw;
+	}
+	for(i = 0; i < numtmus; ++i) {
+		radeonTexObj *t = r300->hw.textures[i];
+		if (!t && !r300->radeon.radeonScreen->kernel_mm) {
+			dw += 0;
+		} else if (t && t->image_override && !t->bo) {
+			if (!r300->radeon.radeonScreen->kernel_mm)
+				dw += 2;
+		} else
+			dw += 4;
 	}
+	return dw;
 }
 
-/**
- * Emit all atoms with a dirty field equal to dirty.
- *
- * The caller must have ensured that there is enough space in the command
- * buffer.
- */
-static INLINE void r300EmitAtoms(r300ContextPtr r300, GLboolean dirty)
+static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom)
 {
-	struct r300_state_atom *atom;
-	uint32_t *dest;
-	int dwords;
-
-	dest = r300->cmdbuf.cmd_buf + r300->cmdbuf.count_used;
-
-	/* Emit WAIT */
-	*dest = cmdwait(R300_WAIT_3D | R300_WAIT_3D_CLEAN);
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit cache flush */
-	*dest = cmdpacket0(R300_TX_INVALTAGS, 1);
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	*dest = R300_TX_FLUSH;
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit END3D */
-	*dest = cmdpacify();
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit actual atoms */
-
-	foreach(atom, &r300->hw.atomlist) {
-		if ((atom->dirty || r300->hw.all_dirty) == dirty) {
-			dwords = (*atom->check) (r300, atom);
-			if (dwords) {
-				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
-					r300PrintStateAtom(r300, atom);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	int numtmus = packet0_count(r300, r300->hw.tex.offset.cmd);
+	int i;
+
+	for(i = 0; i < numtmus; ++i) {
+		radeonTexObj *t = r300->hw.textures[i];
+		if (t && !t->image_override) {
+			BEGIN_BATCH_NO_AUTOSTATE(4);
+			OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+			OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+					RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+			END_BATCH();
+		} else if (!t) {
+			/* Texture unit hasn't a texture bound.
+			 * We assign the current color buffer as a fakery to make
+			 * KIL work on KMS (without it, the CS checker will complain).
+			 */
+			if (r300->radeon.radeonScreen->kernel_mm) {
+				struct radeon_renderbuffer *rrb = radeon_get_colorbuffer(&r300->radeon);
+				if (rrb && rrb->bo) {
+					BEGIN_BATCH_NO_AUTOSTATE(4);
+					OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+					OUT_BATCH_RELOC(0, rrb->bo, 0,
+							RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+					END_BATCH();
 				}
-				memcpy(dest, atom->cmd, dwords * 4);
-				dest += dwords;
-				r300->cmdbuf.count_used += dwords;
-				atom->dirty = GL_FALSE;
+			}
+		} else { /* override cases */
+			if (t->bo) {
+				BEGIN_BATCH_NO_AUTOSTATE(4);
+				OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+				OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+						RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+				END_BATCH();
+			} else if (!r300->radeon.radeonScreen->kernel_mm) {
+				BEGIN_BATCH_NO_AUTOSTATE(2);
+				OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+				OUT_BATCH(t->override_offset);
+				END_BATCH();
 			} else {
-				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
-					fprintf(stderr, "  skip state %s\n",
-						atom->name);
-				}
+				/* Texture unit hasn't a texture bound nothings to do */
 			}
 		}
 	}
 }
 
-/**
- * Copy dirty hardware state atoms into the command buffer.
- *
- * We also copy out clean state if we're at the start of a buffer. That makes
- * it easy to recover from lost contexts.
- */
-void r300EmitState(r300ContextPtr r300)
+void r300_emit_scissor(GLcontext *ctx)
 {
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_PRIMS))
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (r300->cmdbuf.count_used && !r300->hw.is_dirty
-	    && !r300->hw.all_dirty)
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    unsigned x1, y1, x2, y2;
+	struct radeon_renderbuffer *rrb;
+
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        return;
+    }
+	rrb = radeon_get_colorbuffer(&r300->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
 		return;
+	}
+    if (r300->radeon.state.scissor.enabled) {
+        x1 = r300->radeon.state.scissor.rect.x1;
+        y1 = r300->radeon.state.scissor.rect.y1;
+        x2 = r300->radeon.state.scissor.rect.x2;
+        y2 = r300->radeon.state.scissor.rect.y2;
+    } else {
+        x1 = 0;
+        y1 = 0;
+        x2 = rrb->base.Width - 1;
+        y2 = rrb->base.Height - 1;
+    }
+    if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
+        x1 += R300_SCISSORS_OFFSET;
+        y1 += R300_SCISSORS_OFFSET;
+        x2 += R300_SCISSORS_OFFSET;
+        y2 += R300_SCISSORS_OFFSET;
+    }
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+    OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT));
+    OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT));
+    END_BATCH();
+}
+static int check_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	uint32_t dw = 6 + 3 + 16;
+	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+		dw -= 3 + 16;
+	}
+	return dw;
+}
 
-	/* To avoid going across the entire set of states multiple times, just check
-	 * for enough space for the case of emitting all state, and inline the
-	 * r300AllocCmdBuf code here without all the checks.
-	 */
-	r300EnsureCmdBufSpace(r300, r300->hw.max_state_size, __FUNCTION__);
+static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t cbpitch;
+	uint32_t offset = r300->radeon.state.color.draw_offset;
+	uint32_t dw = 6;
+	int i;
 
-	if (!r300->cmdbuf.count_used) {
-		if (RADEON_DEBUG & DEBUG_STATE)
-			fprintf(stderr, "Begin reemit state\n");
+	rrb = radeon_get_colorbuffer(&r300->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
+	}
 
-		r300EmitAtoms(r300, GL_FALSE);
-		r300->cmdbuf.count_reemit = r300->cmdbuf.count_used;
+        if (RADEON_DEBUG & RADEON_STATE)
+           fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height);
+	cbpitch = (rrb->pitch / rrb->cpp);
+	if (rrb->cpp == 4)
+		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
+	else switch (rrb->base._ActualFormat) {
+	case GL_RGB5:
+		cbpitch |= R300_COLOR_FORMAT_RGB565;
+		break;
+	case GL_RGBA4:
+		cbpitch |= R300_COLOR_FORMAT_ARGB4444;
+		break;
+	case GL_RGB5_A1:
+		cbpitch |= R300_COLOR_FORMAT_ARGB1555;
+		break;
 	}
 
-	if (RADEON_DEBUG & DEBUG_STATE)
-		fprintf(stderr, "Begin dirty state\n");
+	if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+		cbpitch |= R300_COLOR_TILE_ENABLE;
+
+    	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	BEGIN_BATCH_NO_AUTOSTATE(dw);
+	OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1);
+	OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1);
+    	if (!r300->radeon.radeonScreen->kernel_mm)
+		OUT_BATCH(cbpitch);
+	else
+		OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	END_BATCH();
+    if (r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+            BEGIN_BATCH_NO_AUTOSTATE(3);
+            OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+            OUT_BATCH(0);
+            OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
+            END_BATCH();
+        } else {
+            BEGIN_BATCH_NO_AUTOSTATE(3);
+            OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+            OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) |
+                    (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) |
+                          ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
+            END_BATCH();
+        }
+    }
+}
+
+static int check_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	uint32_t dw;
+	dw = 6;
+	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	return dw;
+}
+
+static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t zbpitch;
+	uint32_t dw = atom->check(ctx, atom);
+
+	rrb = radeon_get_depthbuffer(&r300->radeon);
+	if (!rrb)
+		return;
+
+	zbpitch = (rrb->pitch / rrb->cpp);
+	if (!r300->radeon.radeonScreen->kernel_mm) {
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+	        zbpitch |= R300_DEPTHMACROTILE_ENABLE;
+	   }
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
+	        zbpitch |= R300_DEPTHMICROTILE_TILED;
+	    }
+	}
 
-	r300EmitAtoms(r300, GL_TRUE);
+	BEGIN_BATCH_NO_AUTOSTATE(dw);
+	OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
+	OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1);
+    	if (!r300->radeon.radeonScreen->kernel_mm)
+	    OUT_BATCH(zbpitch);
+	else
+	    OUT_BATCH_RELOC(cbpitch, rrb->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	END_BATCH();
+}
 
-	assert(r300->cmdbuf.count_used < r300->cmdbuf.size);
+static void emit_zstencil_format(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t format = 0;
+
+	rrb = radeon_get_depthbuffer(&r300->radeon);
+	if (!rrb)
+	  format = 0;
+	else {
+	  if (rrb->cpp == 2)
+	    format = R300_DEPTHFORMAT_16BIT_INT_Z;
+	  else if (rrb->cpp == 4)
+	    format = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
+	}
 
-	r300->hw.is_dirty = GL_FALSE;
-	r300->hw.all_dirty = GL_FALSE;
+	BEGIN_BATCH_NO_AUTOSTATE(atom->cmd_size);
+	OUT_BATCH(atom->cmd[0]);
+	atom->cmd[1] &= ~0xf;
+	atom->cmd[1] |= format;
+	OUT_BATCH(atom->cmd[1]);
+	OUT_BATCH(atom->cmd[2]);
+	OUT_BATCH(atom->cmd[3]);
+	OUT_BATCH(atom->cmd[4]);
+	END_BATCH();
 }
 
-#define packet0_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->packet0.count)
-#define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
-#define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count)
+static int check_never(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   return 0;
+}
 
-static int check_always(r300ContextPtr r300, struct r300_state_atom *atom)
+static int check_always(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	return atom->cmd_size;
 }
 
-static int check_variable(r300ContextPtr r300, struct r300_state_atom *atom)
+static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int cnt;
-	cnt = packet0_count(atom->cmd);
+	if (atom->cmd[0] == CP_PACKET2) {
+		return 0;
+	}
+	cnt = packet0_count(r300, atom->cmd);
 	return cnt ? cnt + 1 : 0;
 }
 
-static int check_vpu(r300ContextPtr r300, struct r300_state_atom *atom)
+int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	int cnt;
-	cnt = vpu_count(atom->cmd);
-	return cnt ? (cnt * 4) + 1 : 0;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	int extra = 1;
+	cnt = r500fp_count(atom->cmd);
+	if (r300->radeon.radeonScreen->kernel_mm)
+		extra = 3;
+
+	return cnt ? (cnt * 6) + extra : 0;
 }
 
-static int check_r500fp(r300ContextPtr r300, struct r300_state_atom *atom)
+int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	int cnt;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	int extra = 1;
 	cnt = r500fp_count(atom->cmd);
-	return cnt ? (cnt * 6) + 1 : 0;
-}
+	if (r300->radeon.radeonScreen->kernel_mm)
+		extra = 3;
 
-static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
-{
-	int cnt;
 	cnt = r500fp_count(atom->cmd);
-	return cnt ? (cnt * 4) + 1 : 0;
+	return cnt ? (cnt * 4) + extra : 0;
 }
 
 #define ALLOC_STATE( ATOM, CHK, SZ, IDX )				\
@@ -295,8 +473,8 @@ static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
       r300->hw.ATOM.idx = (IDX);					\
       r300->hw.ATOM.check = check_##CHK;				\
       r300->hw.ATOM.dirty = GL_FALSE;					\
-      r300->hw.max_state_size += (SZ);					\
-      insert_at_tail(&r300->hw.atomlist, &r300->hw.ATOM);		\
+      r300->radeon.hw.max_state_size += (SZ);					\
+      insert_at_tail(&r300->radeon.hw.atomlist, &r300->hw.ATOM);		\
    } while (0)
 /**
  * Allocate memory for the command buffer and initialize the state atom
@@ -304,251 +482,306 @@ static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
  */
 void r300InitCmdBuf(r300ContextPtr r300)
 {
-	int size, mtu;
-	int has_tcl = 1;
+	int mtu;
+	int has_tcl;
 	int is_r500 = 0;
-	int i;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
 		is_r500 = 1;
 
-	r300->hw.max_state_size = 2 + 2;	/* reserve extra space for WAIT_IDLE and tex cache flush */
+	r300->radeon.hw.max_state_size = 2 + 2;	/* reserve extra space for WAIT_IDLE and tex cache flush */
 
 	mtu = r300->radeon.glCtx->Const.MaxTextureUnits;
-	if (RADEON_DEBUG & DEBUG_TEXTURE) {
+	if (RADEON_DEBUG & RADEON_TEXTURE) {
 		fprintf(stderr, "Using %d maximum texture units..\n", mtu);
 	}
 
 	/* Setup the atom linked list */
-	make_empty_list(&r300->hw.atomlist);
-	r300->hw.atomlist.name = "atom-list";
+	make_empty_list(&r300->radeon.hw.atomlist);
+	r300->radeon.hw.atomlist.name = "atom-list";
 
 	/* Initialize state atoms */
 	ALLOC_STATE(vpt, always, R300_VPT_CMDSIZE, 0);
-	r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmdpacket0(R300_SE_VPORT_XSCALE, 6);
+	r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_SE_VPORT_XSCALE, 6);
 	ALLOC_STATE(vap_cntl, always, R300_VAP_CNTL_SIZE, 0);
-	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH] = cmdpacket0(R300_VAP_PVS_STATE_FLUSH_REG, 1);
+	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_STATE_FLUSH_REG, 1);
 	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH_1] = 0;
-	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_CMD] = cmdpacket0(R300_VAP_CNTL, 1);
-	if (is_r500) {
+	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_CMD] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CNTL, 1);
+	if (is_r500 && !r300->radeon.radeonScreen->kernel_mm) {
 	    ALLOC_STATE(vap_index_offset, always, 2, 0);
-	    r300->hw.vap_index_offset.cmd[0] = cmdpacket0(R500_VAP_INDEX_OFFSET, 1);
+	    r300->hw.vap_index_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_VAP_INDEX_OFFSET, 1);
 	    r300->hw.vap_index_offset.cmd[1] = 0;
 	}
 	ALLOC_STATE(vte, always, 3, 0);
-	r300->hw.vte.cmd[0] = cmdpacket0(R300_SE_VTE_CNTL, 2);
+	r300->hw.vte.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SE_VTE_CNTL, 2);
 	ALLOC_STATE(vap_vf_max_vtx_indx, always, 3, 0);
-	r300->hw.vap_vf_max_vtx_indx.cmd[0] = cmdpacket0(R300_VAP_VF_MAX_VTX_INDX, 2);
+	r300->hw.vap_vf_max_vtx_indx.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_VF_MAX_VTX_INDX, 2);
 	ALLOC_STATE(vap_cntl_status, always, 2, 0);
-	r300->hw.vap_cntl_status.cmd[0] = cmdpacket0(R300_VAP_CNTL_STATUS, 1);
+	r300->hw.vap_cntl_status.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CNTL_STATUS, 1);
 	ALLOC_STATE(vir[0], variable, R300_VIR_CMDSIZE, 0);
 	r300->hw.vir[0].cmd[R300_VIR_CMD_0] =
-	    cmdpacket0(R300_VAP_PROG_STREAM_CNTL_0, 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PROG_STREAM_CNTL_0, 1);
 	ALLOC_STATE(vir[1], variable, R300_VIR_CMDSIZE, 1);
 	r300->hw.vir[1].cmd[R300_VIR_CMD_0] =
-	    cmdpacket0(R300_VAP_PROG_STREAM_CNTL_EXT_0, 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PROG_STREAM_CNTL_EXT_0, 1);
 	ALLOC_STATE(vic, always, R300_VIC_CMDSIZE, 0);
-	r300->hw.vic.cmd[R300_VIC_CMD_0] = cmdpacket0(R300_VAP_VTX_STATE_CNTL, 2);
+	r300->hw.vic.cmd[R300_VIC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_VTX_STATE_CNTL, 2);
 	ALLOC_STATE(vap_psc_sgn_norm_cntl, always, 2, 0);
-	r300->hw.vap_psc_sgn_norm_cntl.cmd[0] = cmdpacket0(R300_VAP_PSC_SGN_NORM_CNTL, SGN_NORM_ZERO_CLAMP_MINUS_ONE);
+	r300->hw.vap_psc_sgn_norm_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PSC_SGN_NORM_CNTL, SGN_NORM_ZERO_CLAMP_MINUS_ONE);
 
 	if (has_tcl) {
 		ALLOC_STATE(vap_clip_cntl, always, 2, 0);
-		r300->hw.vap_clip_cntl.cmd[0] = cmdpacket0(R300_VAP_CLIP_CNTL, 1);
+		r300->hw.vap_clip_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CLIP_CNTL, 1);
 		ALLOC_STATE(vap_clip, always, 5, 0);
-		r300->hw.vap_clip.cmd[0] = cmdpacket0(R300_VAP_GB_VERT_CLIP_ADJ, 4);
+		r300->hw.vap_clip.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_GB_VERT_CLIP_ADJ, 4);
 		ALLOC_STATE(vap_pvs_vtx_timeout_reg, always, 2, 0);
-		r300->hw.vap_pvs_vtx_timeout_reg.cmd[0] = cmdpacket0(VAP_PVS_VTX_TIMEOUT_REG, 1);
+		r300->hw.vap_pvs_vtx_timeout_reg.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, VAP_PVS_VTX_TIMEOUT_REG, 1);
 	}
 
 	ALLOC_STATE(vof, always, R300_VOF_CMDSIZE, 0);
 	r300->hw.vof.cmd[R300_VOF_CMD_0] =
-	    cmdpacket0(R300_VAP_OUTPUT_VTX_FMT_0, 2);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_OUTPUT_VTX_FMT_0, 2);
 
 	if (has_tcl) {
 		ALLOC_STATE(pvs, always, R300_PVS_CMDSIZE, 0);
 		r300->hw.pvs.cmd[R300_PVS_CMD_0] =
-		    cmdpacket0(R300_VAP_PVS_CODE_CNTL_0, 3);
+		    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_CODE_CNTL_0, 3);
 	}
 
 	ALLOC_STATE(gb_enable, always, 2, 0);
-	r300->hw.gb_enable.cmd[0] = cmdpacket0(R300_GB_ENABLE, 1);
-	ALLOC_STATE(gb_misc, always, R300_GB_MISC_CMDSIZE, 0);
-	r300->hw.gb_misc.cmd[0] = cmdpacket0(R300_GB_MSPOS0, 5);
+	r300->hw.gb_enable.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_ENABLE, 1);
+	if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+		ALLOC_STATE(gb_misc, always, R300_GB_MISC_CMDSIZE, 0);
+	} else {
+		ALLOC_STATE(gb_misc, never, R300_GB_MISC_CMDSIZE, 0);
+	}
+	r300->hw.gb_misc.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_MSPOS0, 3);
+	ALLOC_STATE(gb_misc2, always, R300_GB_MISC2_CMDSIZE, 0);
+	r300->hw.gb_misc2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x401C, 2);
 	ALLOC_STATE(txe, always, R300_TXE_CMDSIZE, 0);
-	r300->hw.txe.cmd[R300_TXE_CMD_0] = cmdpacket0(R300_TX_ENABLE, 1);
+	r300->hw.txe.cmd[R300_TXE_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_ENABLE, 1);
 	ALLOC_STATE(ga_point_s0, always, 5, 0);
-	r300->hw.ga_point_s0.cmd[0] = cmdpacket0(R300_GA_POINT_S0, 4);
+	r300->hw.ga_point_s0.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_S0, 4);
 	ALLOC_STATE(ga_triangle_stipple, always, 2, 0);
-	r300->hw.ga_triangle_stipple.cmd[0] = cmdpacket0(R300_GA_TRIANGLE_STIPPLE, 1);
+	r300->hw.ga_triangle_stipple.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_TRIANGLE_STIPPLE, 1);
 	ALLOC_STATE(ps, always, R300_PS_CMDSIZE, 0);
-	r300->hw.ps.cmd[0] = cmdpacket0(R300_GA_POINT_SIZE, 1);
+	r300->hw.ps.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_SIZE, 1);
 	ALLOC_STATE(ga_point_minmax, always, 4, 0);
-	r300->hw.ga_point_minmax.cmd[0] = cmdpacket0(R300_GA_POINT_MINMAX, 3);
+	r300->hw.ga_point_minmax.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_MINMAX, 3);
 	ALLOC_STATE(lcntl, always, 2, 0);
-	r300->hw.lcntl.cmd[0] = cmdpacket0(R300_GA_LINE_CNTL, 1);
+	r300->hw.lcntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_CNTL, 1);
 	ALLOC_STATE(ga_line_stipple, always, 4, 0);
-	r300->hw.ga_line_stipple.cmd[0] = cmdpacket0(R300_GA_LINE_STIPPLE_VALUE, 3);
-	ALLOC_STATE(shade, always, 5, 0);
-	r300->hw.shade.cmd[0] = cmdpacket0(R300_GA_ENHANCE, 4);
+	r300->hw.ga_line_stipple.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_STIPPLE_VALUE, 3);
+        if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+		ALLOC_STATE(shade, always, 2, 0);
+        } else {
+		ALLOC_STATE(shade, never, 2, 0);
+        }
+	r300->hw.shade.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_ENHANCE, 1);
+	ALLOC_STATE(shade2, always, 4, 0);
+	r300->hw.shade2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x4278, 3);
 	ALLOC_STATE(polygon_mode, always, 4, 0);
-	r300->hw.polygon_mode.cmd[0] = cmdpacket0(R300_GA_POLY_MODE, 3);
+	r300->hw.polygon_mode.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POLY_MODE, 3);
 	ALLOC_STATE(fogp, always, 3, 0);
-	r300->hw.fogp.cmd[0] = cmdpacket0(R300_GA_FOG_SCALE, 2);
+	r300->hw.fogp.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_FOG_SCALE, 2);
 	ALLOC_STATE(zbias_cntl, always, 2, 0);
-	r300->hw.zbias_cntl.cmd[0] = cmdpacket0(R300_SU_TEX_WRAP, 1);
+	r300->hw.zbias_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_TEX_WRAP, 1);
 	ALLOC_STATE(zbs, always, R300_ZBS_CMDSIZE, 0);
 	r300->hw.zbs.cmd[R300_ZBS_CMD_0] =
-	    cmdpacket0(R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
 	ALLOC_STATE(occlusion_cntl, always, 2, 0);
-	r300->hw.occlusion_cntl.cmd[0] = cmdpacket0(R300_SU_POLY_OFFSET_ENABLE, 1);
+	r300->hw.occlusion_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_POLY_OFFSET_ENABLE, 1);
 	ALLOC_STATE(cul, always, R300_CUL_CMDSIZE, 0);
-	r300->hw.cul.cmd[R300_CUL_CMD_0] = cmdpacket0(R300_SU_CULL_MODE, 1);
+	r300->hw.cul.cmd[R300_CUL_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_CULL_MODE, 1);
 	ALLOC_STATE(su_depth_scale, always, 3, 0);
-	r300->hw.su_depth_scale.cmd[0] = cmdpacket0(R300_SU_DEPTH_SCALE, 2);
+	r300->hw.su_depth_scale.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_DEPTH_SCALE, 2);
 	ALLOC_STATE(rc, always, R300_RC_CMDSIZE, 0);
-	r300->hw.rc.cmd[R300_RC_CMD_0] = cmdpacket0(R300_RS_COUNT, 2);
+	r300->hw.rc.cmd[R300_RC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_COUNT, 2);
 	if (is_r500) {
-		ALLOC_STATE(ri, always, R500_RI_CMDSIZE, 0);
-		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(R500_RS_IP_0, 16);
-		for (i = 0; i < 8; i++) {
-			r300->hw.ri.cmd[R300_RI_CMD_0 + i +1] =
-			  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-                          (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT);
-		}
+		ALLOC_STATE(ri, variable, R500_RI_CMDSIZE, 0);
+		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, 16);
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
-		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R500_RS_INST_0, 1);
+		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, 1);
 	} else {
-		ALLOC_STATE(ri, always, R300_RI_CMDSIZE, 0);
-		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(R300_RS_IP_0, 8);
+		ALLOC_STATE(ri, variable, R300_RI_CMDSIZE, 0);
+		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, 8);
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
-		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R300_RS_INST_0, 1);
+		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, 1);
 	}
 	ALLOC_STATE(sc_hyperz, always, 3, 0);
-	r300->hw.sc_hyperz.cmd[0] = cmdpacket0(R300_SC_HYPERZ, 2);
+	r300->hw.sc_hyperz.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_HYPERZ, 2);
 	ALLOC_STATE(sc_screendoor, always, 2, 0);
-	r300->hw.sc_screendoor.cmd[0] = cmdpacket0(R300_SC_SCREENDOOR, 1);
+	r300->hw.sc_screendoor.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_SCREENDOOR, 1);
 	ALLOC_STATE(us_out_fmt, always, 6, 0);
-	r300->hw.us_out_fmt.cmd[0] = cmdpacket0(R300_US_OUT_FMT, 5);
+	r300->hw.us_out_fmt.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_OUT_FMT, 5);
 
 	if (is_r500) {
 		ALLOC_STATE(fp, always, R500_FP_CMDSIZE, 0);
-		r300->hw.fp.cmd[R500_FP_CMD_0] = cmdpacket0(R500_US_CONFIG, 2);
+		r300->hw.fp.cmd[R500_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_US_CONFIG, 2);
 		r300->hw.fp.cmd[R500_FP_CNTL] = R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO;
-		r300->hw.fp.cmd[R500_FP_CMD_1] = cmdpacket0(R500_US_CODE_ADDR, 3);
-		r300->hw.fp.cmd[R500_FP_CMD_2] = cmdpacket0(R500_US_FC_CTRL, 1);
+		r300->hw.fp.cmd[R500_FP_CMD_1] = cmdpacket0(r300->radeon.radeonScreen, R500_US_CODE_ADDR, 3);
+		r300->hw.fp.cmd[R500_FP_CMD_2] = cmdpacket0(r300->radeon.radeonScreen, R500_US_FC_CTRL, 1);
 		r300->hw.fp.cmd[R500_FP_FC_CNTL] = 0; /* FIXME when we add flow control */
 
 		ALLOC_STATE(r500fp, r500fp, R500_FPI_CMDSIZE, 0);
-		r300->hw.r500fp.cmd[R300_FPI_CMD_0] = cmdr500fp(0, 0, 0, 0);
+		r300->hw.r500fp.cmd[R300_FPI_CMD_0] =
+			cmdr500fp(r300->radeon.radeonScreen, 0, 0, 0, 0);
+		if (r300->radeon.radeonScreen->kernel_mm)
+			r300->hw.r500fp.emit = emit_r500fp;
+
 		ALLOC_STATE(r500fp_const, r500fp_const, R500_FPP_CMDSIZE, 0);
-		r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] = cmdr500fp(0, 0, 1, 0);
+		r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] =
+			cmdr500fp(r300->radeon.radeonScreen, 0, 0, 1, 0);
+		if (r300->radeon.radeonScreen->kernel_mm)
+			r300->hw.r500fp_const.emit = emit_r500fp;
 	} else {
 		ALLOC_STATE(fp, always, R300_FP_CMDSIZE, 0);
-		r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(R300_US_CONFIG, 3);
-		r300->hw.fp.cmd[R300_FP_CMD_1] = cmdpacket0(R300_US_CODE_ADDR_0, 4);
+		r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CONFIG, 3);
+		r300->hw.fp.cmd[R300_FP_CMD_1] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CODE_ADDR_0, 4);
+
 		ALLOC_STATE(fpt, variable, R300_FPT_CMDSIZE, 0);
-		r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmdpacket0(R300_US_TEX_INST_0, 0);
+		r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_TEX_INST_0, 0);
 
 		ALLOC_STATE(fpi[0], variable, R300_FPI_CMDSIZE, 0);
-		r300->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_INST_0, 1);
+		r300->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_RGB_INST_0, 1);
 		ALLOC_STATE(fpi[1], variable, R300_FPI_CMDSIZE, 1);
-		r300->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_ADDR_0, 1);
+		r300->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_RGB_ADDR_0, 1);
 		ALLOC_STATE(fpi[2], variable, R300_FPI_CMDSIZE, 2);
-		r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_INST_0, 1);
+		r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_ALPHA_INST_0, 1);
 		ALLOC_STATE(fpi[3], variable, R300_FPI_CMDSIZE, 3);
-		r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_ADDR_0, 1);
+		r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_ALPHA_ADDR_0, 1);
 		ALLOC_STATE(fpp, variable, R300_FPP_CMDSIZE, 0);
-		r300->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, 0);
+		r300->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_PFS_PARAM_0_X, 0);
 	}
 	ALLOC_STATE(fogs, always, R300_FOGS_CMDSIZE, 0);
-	r300->hw.fogs.cmd[R300_FOGS_CMD_0] = cmdpacket0(R300_FG_FOG_BLEND, 1);
+	r300->hw.fogs.cmd[R300_FOGS_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_FOG_BLEND, 1);
 	ALLOC_STATE(fogc, always, R300_FOGC_CMDSIZE, 0);
-	r300->hw.fogc.cmd[R300_FOGC_CMD_0] = cmdpacket0(R300_FG_FOG_COLOR_R, 3);
+	r300->hw.fogc.cmd[R300_FOGC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_FOG_COLOR_R, 3);
 	ALLOC_STATE(at, always, R300_AT_CMDSIZE, 0);
-	r300->hw.at.cmd[R300_AT_CMD_0] = cmdpacket0(R300_FG_ALPHA_FUNC, 2);
+	r300->hw.at.cmd[R300_AT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_ALPHA_FUNC, 2);
 	ALLOC_STATE(fg_depth_src, always, 2, 0);
-	r300->hw.fg_depth_src.cmd[0] = cmdpacket0(R300_FG_DEPTH_SRC, 1);
+	r300->hw.fg_depth_src.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_DEPTH_SRC, 1);
 	ALLOC_STATE(rb3d_cctl, always, 2, 0);
-	r300->hw.rb3d_cctl.cmd[0] = cmdpacket0(R300_RB3D_CCTL, 1);
+	r300->hw.rb3d_cctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_CCTL, 1);
 	ALLOC_STATE(bld, always, R300_BLD_CMDSIZE, 0);
-	r300->hw.bld.cmd[R300_BLD_CMD_0] = cmdpacket0(R300_RB3D_CBLEND, 2);
+	r300->hw.bld.cmd[R300_BLD_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_CBLEND, 2);
 	ALLOC_STATE(cmk, always, R300_CMK_CMDSIZE, 0);
-	r300->hw.cmk.cmd[R300_CMK_CMD_0] = cmdpacket0(RB3D_COLOR_CHANNEL_MASK, 1);
+	r300->hw.cmk.cmd[R300_CMK_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, RB3D_COLOR_CHANNEL_MASK, 1);
 	if (is_r500) {
 		ALLOC_STATE(blend_color, always, 3, 0);
-		r300->hw.blend_color.cmd[0] = cmdpacket0(R500_RB3D_CONSTANT_COLOR_AR, 2);
+		r300->hw.blend_color.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_CONSTANT_COLOR_AR, 2);
 	} else {
 		ALLOC_STATE(blend_color, always, 2, 0);
-		r300->hw.blend_color.cmd[0] = cmdpacket0(R300_RB3D_BLEND_COLOR, 1);
+		r300->hw.blend_color.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_BLEND_COLOR, 1);
 	}
 	ALLOC_STATE(rop, always, 2, 0);
-	r300->hw.rop.cmd[0] = cmdpacket0(R300_RB3D_ROPCNTL, 1);
-	ALLOC_STATE(cb, always, R300_CB_CMDSIZE, 0);
-	r300->hw.cb.cmd[R300_CB_CMD_0] = cmdpacket0(R300_RB3D_COLOROFFSET0, 1);
-	r300->hw.cb.cmd[R300_CB_CMD_1] = cmdpacket0(R300_RB3D_COLORPITCH0, 1);
+	r300->hw.rop.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_ROPCNTL, 1);
+	ALLOC_STATE(cb, cb_offset, R300_CB_CMDSIZE, 0);
+	r300->hw.cb.emit = &emit_cb_offset;
 	ALLOC_STATE(rb3d_dither_ctl, always, 10, 0);
-	r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(R300_RB3D_DITHER_CTL, 9);
+	r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9);
 	ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0);
-	r300->hw.rb3d_aaresolve_ctl.cmd[0] = cmdpacket0(R300_RB3D_AARESOLVE_CTL, 1);
-	ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
+	r300->hw.rb3d_aaresolve_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_AARESOLVE_CTL, 1);
+	if ((r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) ||
+	      ( !r300->radeon.radeonScreen->kernel_mm && (
+	    (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS400) ||
+	    (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) ||
+	    (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420) ) ) ) {
+		ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
+	} else {
+		ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, never, 3, 0);
+	}
+	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
 	ALLOC_STATE(zs, always, R300_ZS_CMDSIZE, 0);
 	r300->hw.zs.cmd[R300_ZS_CMD_0] =
-	    cmdpacket0(R300_ZB_CNTL, 3);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_ZB_CNTL, 3);
+
 	ALLOC_STATE(zstencil_format, always, 5, 0);
 	r300->hw.zstencil_format.cmd[0] =
-	    cmdpacket0(R300_ZB_FORMAT, 4);
-	ALLOC_STATE(zb, always, R300_ZB_CMDSIZE, 0);
-	r300->hw.zb.cmd[R300_ZB_CMD_0] = cmdpacket0(R300_ZB_DEPTHOFFSET, 2);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_ZB_FORMAT, 4);
+	r300->hw.zstencil_format.emit = emit_zstencil_format;
+
+	ALLOC_STATE(zb, zb_offset, R300_ZB_CMDSIZE, 0);
+	r300->hw.zb.emit = emit_zb_offset;
 	ALLOC_STATE(zb_depthclearvalue, always, 2, 0);
-	r300->hw.zb_depthclearvalue.cmd[0] = cmdpacket0(R300_ZB_DEPTHCLEARVALUE, 1);
-	ALLOC_STATE(unk4F30, always, 3, 0);
-	r300->hw.unk4F30.cmd[0] = cmdpacket0(0x4F30, 2);
+	r300->hw.zb_depthclearvalue.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_DEPTHCLEARVALUE, 1);
+	ALLOC_STATE(zb_zmask, always, 3, 0);
+	r300->hw.zb_zmask.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_ZMASK_OFFSET, 2);
 	ALLOC_STATE(zb_hiz_offset, always, 2, 0);
-	r300->hw.zb_hiz_offset.cmd[0] = cmdpacket0(R300_ZB_HIZ_OFFSET, 1);
+	r300->hw.zb_hiz_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_HIZ_OFFSET, 1);
 	ALLOC_STATE(zb_hiz_pitch, always, 2, 0);
-	r300->hw.zb_hiz_pitch.cmd[0] = cmdpacket0(R300_ZB_HIZ_PITCH, 1);
+	r300->hw.zb_hiz_pitch.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_HIZ_PITCH, 1);
 
 	/* VPU only on TCL */
 	if (has_tcl) {
-   	        int i;
+		int i;
+		if (r300->radeon.radeonScreen->kernel_mm) {
+			ALLOC_STATE(vap_flush, always, 10, 0);
+			/* flush processing vertices */
+			r300->hw.vap_flush.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_SCREENDOOR, 1);
+			r300->hw.vap_flush.cmd[1] = 0;
+			r300->hw.vap_flush.cmd[2] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DSTCACHE_CTLSTAT, 1);
+			r300->hw.vap_flush.cmd[3] = R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D;
+			r300->hw.vap_flush.cmd[4] = cmdpacket0(r300->radeon.radeonScreen, RADEON_WAIT_UNTIL, 1);
+			r300->hw.vap_flush.cmd[5] = RADEON_WAIT_3D_IDLECLEAN;
+			r300->hw.vap_flush.cmd[6] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_SCREENDOOR, 1);
+			r300->hw.vap_flush.cmd[7] = 0xffffff;
+			r300->hw.vap_flush.cmd[8] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_STATE_FLUSH_REG, 1);
+			r300->hw.vap_flush.cmd[9] = 0;
+		} else {
+			ALLOC_STATE(vap_flush, never, 10, 0);
+		}
+
+
 		ALLOC_STATE(vpi, vpu, R300_VPI_CMDSIZE, 0);
-		r300->hw.vpi.cmd[R300_VPI_CMD_0] =
-		    cmdvpu(R300_PVS_CODE_START, 0);
+		r300->hw.vpi.cmd[0] =
+			cmdvpu(r300->radeon.radeonScreen, R300_PVS_CODE_START, 0);
+		if (r300->radeon.radeonScreen->kernel_mm)
+			r300->hw.vpi.emit = emit_vpu;
 
 		if (is_r500) {
-		    ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
-		    r300->hw.vpp.cmd[R300_VPP_CMD_0] =
-			cmdvpu(R500_PVS_CONST_START, 0);
-
-		    ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
-		    r300->hw.vps.cmd[R300_VPS_CMD_0] =
-			cmdvpu(R500_POINT_VPORT_SCALE_OFFSET, 1);
+			ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
+			r300->hw.vpp.cmd[0] =
+				cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0);
+			if (r300->radeon.radeonScreen->kernel_mm)
+				r300->hw.vpp.emit = emit_vpu;
+
+			ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
+			r300->hw.vps.cmd[0] =
+				cmdvpu(r300->radeon.radeonScreen, R500_POINT_VPORT_SCALE_OFFSET, 1);
+			if (r300->radeon.radeonScreen->kernel_mm)
+				r300->hw.vps.emit = emit_vpu;
 
 			for (i = 0; i < 6; i++) {
 				ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0);
-				r300->hw.vpucp[i].cmd[R300_VPUCP_CMD_0] =
-					cmdvpu(R500_PVS_UCP_START + i, 1);
+				r300->hw.vpucp[i].cmd[0] =
+					cmdvpu(r300->radeon.radeonScreen,
+							R500_PVS_UCP_START + i, 1);
+				if (r300->radeon.radeonScreen->kernel_mm)
+					r300->hw.vpucp[i].emit = emit_vpu;
 			}
 		} else {
-		    ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
-		    r300->hw.vpp.cmd[R300_VPP_CMD_0] =
-			cmdvpu(R300_PVS_CONST_START, 0);
-
-		    ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
-		    r300->hw.vps.cmd[R300_VPS_CMD_0] =
-			cmdvpu(R300_POINT_VPORT_SCALE_OFFSET, 1);
+			ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
+			r300->hw.vpp.cmd[0] =
+				cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0);
+			if (r300->radeon.radeonScreen->kernel_mm)
+				r300->hw.vpp.emit = emit_vpu;
+
+			ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
+			r300->hw.vps.cmd[0] =
+				cmdvpu(r300->radeon.radeonScreen, R300_POINT_VPORT_SCALE_OFFSET, 1);
+			if (r300->radeon.radeonScreen->kernel_mm)
+				r300->hw.vps.emit = emit_vpu;
 
 			for (i = 0; i < 6; i++) {
 				ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0);
-				r300->hw.vpucp[i].cmd[R300_VPUCP_CMD_0] =
-					cmdvpu(R300_PVS_UCP_START + i, 1);
+				r300->hw.vpucp[i].cmd[0] =
+					cmdvpu(r300->radeon.radeonScreen,
+							R300_PVS_UCP_START + i, 1);
+				if (r300->radeon.radeonScreen->kernel_mm)
+					r300->hw.vpucp[i].emit = emit_vpu;
 			}
 		}
 	}
@@ -556,130 +789,48 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	/* Textures */
 	ALLOC_STATE(tex.filter, variable, mtu + 1, 0);
 	r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER0_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, 0);
 
 	ALLOC_STATE(tex.filter_1, variable, mtu + 1, 0);
 	r300->hw.tex.filter_1.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER1_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER1_0, 0);
 
 	ALLOC_STATE(tex.size, variable, mtu + 1, 0);
-	r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmdpacket0(R300_TX_SIZE_0, 0);
+	r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_SIZE_0, 0);
 
 	ALLOC_STATE(tex.format, variable, mtu + 1, 0);
 	r300->hw.tex.format.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT_0, 0);
 
 	ALLOC_STATE(tex.pitch, variable, mtu + 1, 0);
-	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] = cmdpacket0(R300_TX_FORMAT2_0, 0);
+	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT2_0, 0);
 
-	ALLOC_STATE(tex.offset, variable, mtu + 1, 0);
+	ALLOC_STATE(tex.offset, tex_offsets, 1, 0);
 	r300->hw.tex.offset.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_OFFSET_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_OFFSET_0, 0);
+	r300->hw.tex.offset.emit = &emit_tex_offsets;
 
 	ALLOC_STATE(tex.chroma_key, variable, mtu + 1, 0);
 	r300->hw.tex.chroma_key.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_CHROMA_KEY_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_CHROMA_KEY_0, 0);
 
 	ALLOC_STATE(tex.border_color, variable, mtu + 1, 0);
 	r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_BORDER_COLOR_0, 0);
-
-	r300->hw.is_dirty = GL_TRUE;
-	r300->hw.all_dirty = GL_TRUE;
-
-	/* Initialize command buffer */
-	size =
-	    256 * driQueryOptioni(&r300->radeon.optionCache,
-				  "command_buffer_size");
-	if (size < 2 * r300->hw.max_state_size) {
-		size = 2 * r300->hw.max_state_size + 65535;
-	}
-	if (size > 64 * 256)
-		size = 64 * 256;
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) {
-		fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n",
-			sizeof(drm_r300_cmd_header_t));
-		fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n",
-			sizeof(drm_radeon_cmd_buffer_t));
-		fprintf(stderr,
-			"Allocating %d bytes command buffer (max state is %d bytes)\n",
-			size * 4, r300->hw.max_state_size * 4);
-	}
-
-	r300->cmdbuf.size = size;
-	r300->cmdbuf.cmd_buf = (uint32_t *) CALLOC(size * 4);
-	r300->cmdbuf.count_used = 0;
-	r300->cmdbuf.count_reemit = 0;
-}
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, 0);
 
-/**
- * Destroy the command buffer and state atoms.
- */
-void r300DestroyCmdBuf(r300ContextPtr r300)
-{
-	struct r300_state_atom *atom;
-
-	FREE(r300->cmdbuf.cmd_buf);
-
-	foreach(atom, &r300->hw.atomlist) {
-		FREE(atom->cmd);
+	radeon_init_query_stateobj(&r300->radeon, R300_QUERYOBJ_CMDSIZE);
+	if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) {
+		r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, RV530_FG_ZBREG_DEST, 1);
+		r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_DATA_0] = RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL;
+	} else {
+		r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_REG_DEST, 1);
+		r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_DATA_0] = R300_RASTER_PIPE_SELECT_ALL;
 	}
-}
-
-void r300EmitBlit(r300ContextPtr rmesa,
-		  GLuint color_fmt,
-		  GLuint src_pitch,
-		  GLuint src_offset,
-		  GLuint dst_pitch,
-		  GLuint dst_offset,
-		  GLint srcx, GLint srcy,
-		  GLint dstx, GLint dsty, GLuint w, GLuint h)
-{
-	drm_r300_cmd_header_t *cmd;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr,
-			"%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-			__FUNCTION__, src_pitch, src_offset, srcx, srcy,
-			dst_pitch, dst_offset, dstx, dsty, w, h);
-
-	assert((src_pitch & 63) == 0);
-	assert((dst_pitch & 63) == 0);
-	assert((src_offset & 1023) == 0);
-	assert((dst_offset & 1023) == 0);
-	assert(w < (1 << 16));
-	assert(h < (1 << 16));
-
-	cmd = (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa, 8, __FUNCTION__);
-
-	cmd[0].header.cmd_type = R300_CMD_PACKET3;
-	cmd[0].header.pad0 = R300_CMD_PACKET3_RAW;
-	cmd[1].u = R300_CP_CMD_BITBLT_MULTI | (5 << 16);
-	cmd[2].u = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-		    RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-		    RADEON_GMC_BRUSH_NONE |
-		    (color_fmt << 8) |
-		    RADEON_GMC_SRC_DATATYPE_COLOR |
-		    RADEON_ROP3_S |
-		    RADEON_DP_SRC_SOURCE_MEMORY |
-		    RADEON_GMC_CLR_CMP_CNTL_DIS | RADEON_GMC_WR_MSK_DIS);
-
-	cmd[3].u = ((src_pitch / 64) << 22) | (src_offset >> 10);
-	cmd[4].u = ((dst_pitch / 64) << 22) | (dst_offset >> 10);
-	cmd[5].u = (srcx << 16) | srcy;
-	cmd[6].u = (dstx << 16) | dsty;	/* dst */
-	cmd[7].u = (w << 16) | h;
-}
-
-void r300EmitWait(r300ContextPtr rmesa, GLuint flags)
-{
-	drm_r300_cmd_header_t *cmd;
+	r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_CMD_1] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_ZPASS_DATA, 1);
+	r300->radeon.query.queryobj.cmd[R300_QUERYOBJ_DATA_1] = 0;
 
-	assert(!(flags & ~(R300_WAIT_2D | R300_WAIT_3D)));
+	r300->radeon.hw.is_dirty = GL_TRUE;
+	r300->radeon.hw.all_dirty = GL_TRUE;
 
-	cmd = (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].u = 0;
-	cmd[0].wait.cmd_type = R300_CMD_WAIT;
-	cmd[0].wait.flags = flags;
+	rcommonInitCmdBuf(&r300->radeon);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
index a8eaa580bd..1b703e518a 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
@@ -38,79 +38,20 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "r300_context.h"
 
-extern int r300FlushCmdBufLocked(r300ContextPtr r300, const char *caller);
-extern int r300FlushCmdBuf(r300ContextPtr r300, const char *caller);
-
-extern void r300EmitState(r300ContextPtr r300);
+#define CACHE_FLUSH_BUFSZ      (4*2)
+#define PRE_EMIT_STATE_BUFSZ   (2+2)
+#define AOS_BUFSZ(nr)          (3+(nr >>1)*3 + (nr&1)*2 + (nr*2))
+#define FIREAOS_BUFSZ          (3)
+#define SCISSORS_BUFSZ         (3)
 
 extern void r300InitCmdBuf(r300ContextPtr r300);
-extern void r300DestroyCmdBuf(r300ContextPtr r300);
-
-/**
- * Make sure that enough space is available in the command buffer
- * by flushing if necessary.
- *
- * \param dwords The number of dwords we need to be free on the command buffer
- */
-static INLINE void r300EnsureCmdBufSpace(r300ContextPtr r300,
-					     int dwords, const char *caller)
-{
-	assert(dwords < r300->cmdbuf.size);
-
-	if (r300->cmdbuf.count_used + dwords > r300->cmdbuf.size)
-		r300FlushCmdBuf(r300, caller);
-}
-
-/**
- * Allocate the given number of dwords in the command buffer and return
- * a pointer to the allocated area.
- * When necessary, these functions cause a flush. r300AllocCmdBuf() also
- * causes state reemission after a flush. This is necessary to ensure
- * correct hardware state after an unlock.
- */
-static INLINE uint32_t *r300RawAllocCmdBuf(r300ContextPtr r300,
-					       int dwords, const char *caller)
-{
-	uint32_t *ptr;
-
-	r300EnsureCmdBufSpace(r300, dwords, caller);
-
-	ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used];
-	r300->cmdbuf.count_used += dwords;
-	return ptr;
-}
-
-static INLINE uint32_t *r300AllocCmdBuf(r300ContextPtr r300,
-					    int dwords, const char *caller)
-{
-	uint32_t *ptr;
-
-	r300EnsureCmdBufSpace(r300, dwords, caller);
-
-	if (!r300->cmdbuf.count_used) {
-		if (RADEON_DEBUG & DEBUG_IOCTL)
-			fprintf(stderr,
-				"Reemit state after flush (from %s)\n", caller);
-		r300EmitState(r300);
-	}
-
-	ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used];
-	r300->cmdbuf.count_used += dwords;
-	return ptr;
-}
+void r300_emit_scissor(GLcontext *ctx);
 
-extern void r300EmitBlit(r300ContextPtr rmesa,
-			 GLuint color_fmt,
-			 GLuint src_pitch,
-			 GLuint src_offset,
-			 GLuint dst_pitch,
-			 GLuint dst_offset,
-			 GLint srcx, GLint srcy,
-			 GLint dstx, GLint dsty, GLuint w, GLuint h);
+void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom);
+int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom);
 
-extern void r300EmitWait(r300ContextPtr rmesa, GLuint flags);
-extern void r300EmitLOAD_VBPNTR(r300ContextPtr rmesa, int start);
-extern void r300EmitVertexShader(r300ContextPtr rmesa);
-extern void r300EmitPixelShader(r300ContextPtr rmesa);
+void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom);
+int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom);
+int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom);
 
 #endif				/* __R300_CMDBUF_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 7d6705058f..2ea1b826de 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -43,8 +43,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/matrix.h"
 #include "main/extensions.h"
 #include "main/state.h"
-#include "main/texobj.h"
 #include "main/bufferobj.h"
+#include "main/texobj.h"
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
@@ -56,46 +56,47 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "drivers/common/driverfuncs.h"
 
-#include "radeon_ioctl.h"
-#include "radeon_span.h"
 #include "r300_context.h"
+#include "radeon_context.h"
+#include "radeon_span.h"
 #include "r300_cmdbuf.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
 #include "r300_swtcl.h"
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
+#include "radeon_bocs_wrapper.h"
+#include "radeon_buffer_objects.h"
+#include "radeon_queryobj.h"
 
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h"		/* for symbolic values of enum-type options */
 
-/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
-int future_hw_tcl_on = 1;
-int hw_tcl_on = 1;
-
 #define need_GL_VERSION_2_0
+#define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_vertex_program
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
 #define need_GL_EXT_blend_minmax
+#define need_GL_EXT_framebuffer_blit
+#define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_stencil_two_side
 #define need_GL_ATI_separate_stencil
 #define need_GL_NV_vertex_program
+
 #include "extension_helper.h"
 
+
 const struct dri_extension card_extensions[] = {
   /* *INDENT-OFF* */
   {"GL_ARB_depth_texture",		NULL},
   {"GL_ARB_fragment_program",		NULL},
+  {"GL_ARB_occlusion_query",		GL_ARB_occlusion_query_functions},
   {"GL_ARB_multitexture",		NULL},
   {"GL_ARB_point_parameters",		GL_ARB_point_parameters_functions},
   {"GL_ARB_shadow",			NULL},
@@ -112,6 +113,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
   {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
   {"GL_EXT_blend_subtract",		NULL},
+  {"GL_EXT_packed_depth_stencil",	NULL},
   {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
   {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
   {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
@@ -125,6 +127,8 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_texture_lod_bias",		NULL},
   {"GL_EXT_texture_mirror_clamp",	NULL},
   {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_EXT_texture_sRGB",		NULL},
+  {"GL_EXT_vertex_array_bgra",		NULL},
   {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
   {"GL_ATI_texture_env_combine3",	NULL},
   {"GL_ATI_texture_mirror_once",	NULL},
@@ -139,6 +143,12 @@ const struct dri_extension card_extensions[] = {
 };
 
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_blit",	GL_EXT_framebuffer_blit_functions },
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 /**
  * The GL 2.0 functions are needed to make display lists work with
  * functions added by GL_ATI_separate_stencil.
@@ -147,16 +157,7 @@ const struct dri_extension gl_20_extension[] = {
   {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
 };
 
-
-extern struct tnl_pipeline_stage _r300_render_stage;
-extern const struct tnl_pipeline_stage _r300_tcl_stage;
-
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
-
-	/* Try and go straight to t&l
-	 */
-	&_r300_tcl_stage,
-
 	/* Catch any t&l fallbacks
 	 */
 	&_tnl_vertex_transform_stage,
@@ -165,147 +166,188 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	&_tnl_fog_coordinate_stage,
 	&_tnl_texgen_stage,
 	&_tnl_texture_transform_stage,
+	&_tnl_point_attenuation_stage,
 	&_tnl_vertex_program_stage,
-
-	/* Try again to go to tcl?
-	 *     - no good for asymmetric-twoside (do with multipass)
-	 *     - no good for asymmetric-unfilled (do with multipass)
-	 *     - good for material
-	 *     - good for texgen
-	 *     - need to manipulate a bit of state
-	 *
-	 * - worth it/not worth it?
-	 */
-
-	/* Else do them here.
-	 */
-	&_r300_render_stage,
-	&_tnl_render_stage,	/* FALLBACK  */
+	&_tnl_render_stage,
 	0,
 };
 
-/* Create the device specific rendering context.
- */
-GLboolean r300CreateContext(const __GLcontextModes * glVisual,
-			    __DRIcontextPrivate * driContextPriv,
-			    void *sharedContextPrivate)
+static void r300_get_lock(radeonContextPtr rmesa)
 {
-	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
-	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
-	struct dd_function_table functions;
-	r300ContextPtr r300;
-	GLcontext *ctx;
-	int tcl_mode, i;
-
-	assert(glVisual);
-	assert(driContextPriv);
-	assert(screen);
+	drm_radeon_sarea_t *sarea = rmesa->sarea;
 
-	/* Allocate the R300 context */
-	r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
-	if (!r300)
-		return GL_FALSE;
+	if (sarea->ctx_owner != rmesa->dri.hwContext) {
+		sarea->ctx_owner = rmesa->dri.hwContext;
+		if (!rmesa->radeonScreen->kernel_mm)
+			radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
+	}
+}
 
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-		hw_tcl_on = future_hw_tcl_on = 0;
+static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+    /* please flush pipe do all pending work */
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_SCREENDOOR, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_SCREENDOOR, 1));
+    radeon_cs_write_dword(cs, 0x00FFFFFF);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_HYPERZ, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_US_CONFIG, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_ZB_CNTL, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen, R300_WAIT_3D));
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_RB3D_DSTCACHE_CTLSTAT, 1));
+    radeon_cs_write_dword(cs, R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_ZB_ZCACHE_CTLSTAT, 1));
+    radeon_cs_write_dword(cs, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE);
+    radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen,
+                               R300_WAIT_3D | R300_WAIT_3D_CLEAN));
+}
 
-	/* Parse configuration files.
-	 * Do this here so that initialMaxAnisotropy is set before we create
-	 * the default textures.
-	 */
-	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
-			    screen->driScreen->myNum, "r300");
-	r300->initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
-						     "def_max_anisotropy");
+static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon)
+{
+	BATCH_LOCALS(radeon);
 
-	/* Init default driver functions then plug in our R300-specific functions
-	 * (the texture functions are especially important)
-	 */
-	_mesa_init_driver_functions(&functions);
-	r300InitIoctlFuncs(&functions);
-	r300InitStateFuncs(&functions);
-	r300InitTextureFuncs(&functions);
-	r300InitShaderFuncs(&functions);
+	cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	BEGIN_BATCH_NO_AUTOSTATE(2);
+	OUT_BATCH_REGVAL(R300_TX_INVALTAGS, R300_TX_FLUSH);
+	END_BATCH();
+	end_3d(radeon);
+}
 
-#ifdef USER_BUFFERS
-	r300_mem_init(r300);
-#endif
+static void r300_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	if (mode)
+		r300->radeon.Fallback |= bit;
+	else
+		r300->radeon.Fallback &= ~bit;
+}
 
-	if (!radeonInitContext(&r300->radeon, &functions,
-			       glVisual, driContextPriv,
-			       sharedContextPrivate)) {
-		FREE(r300);
-		return GL_FALSE;
+static void r300_emit_query_finish(radeonContextPtr radeon)
+{
+	r300ContextPtr r300 = (r300ContextPtr)radeon;
+	struct radeon_query_object *query = radeon->query.current;
+	BATCH_LOCALS(radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(3 * 2 *r300->radeon.radeonScreen->num_gb_pipes + 2);
+	switch (r300->radeon.radeonScreen->num_gb_pipes) {
+	case 4:
+		OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_3);
+		OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+		OUT_BATCH_RELOC(0, query->bo, query->curr_offset+3*sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+	case 3:
+		OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_2);
+		OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+		OUT_BATCH_RELOC(0, query->bo, query->curr_offset+2*sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+	case 2:
+		if (r300->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV380) {
+			OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_3);
+		} else {
+			OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_1);
+		}
+		OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+		OUT_BATCH_RELOC(0, query->bo, query->curr_offset+1*sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+	case 1:
+	default:
+		OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_0);
+		OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+		OUT_BATCH_RELOC(0, query->bo, query->curr_offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+		break;
 	}
+	OUT_BATCH_REGVAL(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL);
+	END_BATCH();
+	query->curr_offset += r300->radeon.radeonScreen->num_gb_pipes * sizeof(uint32_t);
+	assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
+	query->emitted_begin = GL_FALSE;
+}
 
-	/* Init r300 context data */
-	r300->dma.buf0_address =
-	    r300->radeon.radeonScreen->buffers->list[0].address;
-
-	(void)memset(r300->texture_heaps, 0, sizeof(r300->texture_heaps));
-	make_empty_list(&r300->swapped);
-
-	r300->nr_heaps = 1 /* screen->numTexHeaps */ ;
-	assert(r300->nr_heaps < RADEON_NR_TEX_HEAPS);
-	for (i = 0; i < r300->nr_heaps; i++) {
-		/* *INDENT-OFF* */
-		r300->texture_heaps[i] = driCreateTextureHeap(i, r300,
-							       screen->
-							       texSize[i], 12,
-							       RADEON_NR_TEX_REGIONS,
-							       (drmTextureRegionPtr)
-							       r300->radeon.sarea->
-							       tex_list[i],
-							       &r300->radeon.sarea->
-							       tex_age[i],
-							       &r300->swapped,
-							       sizeof
-							       (r300TexObj),
-							       (destroy_texture_object_t
-								*)
-							       r300DestroyTexObj);
-		/* *INDENT-ON* */
-	}
-	r300->texture_depth = driQueryOptioni(&r300->radeon.optionCache,
-					      "texture_depth");
-	if (r300->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-		r300->texture_depth = (screen->cpp == 4) ?
-		    DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
-
-	/* Set the maximum texture size small enough that we can guarentee that
-	 * all texture units can bind a maximal texture and have them both in
-	 * texturable memory at once.
-	 */
+static void rv530_emit_query_finish_single_z(radeonContextPtr radeon)
+{
+	BATCH_LOCALS(radeon);
+	struct radeon_query_object *query = radeon->query.current;
+
+	BEGIN_BATCH_NO_AUTOSTATE(8);
+	OUT_BATCH_REGVAL(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
+	OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+	OUT_BATCH_RELOC(0, query->bo, query->curr_offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+	OUT_BATCH_REGVAL(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
+	END_BATCH();
+
+	query->curr_offset += sizeof(uint32_t);
+	assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
+	query->emitted_begin = GL_FALSE;
+}
 
-	ctx = r300->radeon.glCtx;
+static void rv530_emit_query_finish_double_z(radeonContextPtr radeon)
+{
+	BATCH_LOCALS(radeon);
+	struct radeon_query_object *query = radeon->query.current;
+
+	BEGIN_BATCH_NO_AUTOSTATE(14);
+	OUT_BATCH_REGVAL(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
+	OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+	OUT_BATCH_RELOC(0, query->bo, query->curr_offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+	OUT_BATCH_REGVAL(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_1);
+	OUT_BATCH_REGSEQ(R300_ZB_ZPASS_ADDR, 1);
+	OUT_BATCH_RELOC(0, query->bo, query->curr_offset + sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+	OUT_BATCH_REGVAL(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
+	END_BATCH();
+
+	query->curr_offset += 2 * sizeof(uint32_t);
+	assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
+	query->emitted_begin = GL_FALSE;
+}
+
+static void r300_init_vtbl(radeonContextPtr radeon)
+{
+	radeon->vtbl.get_lock = r300_get_lock;
+	radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset;
+	radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
+	radeon->vtbl.swtcl_flush = r300_swtcl_flush;
+	radeon->vtbl.pre_emit_atoms = r300_vtbl_pre_emit_atoms;
+	radeon->vtbl.fallback = r300_fallback;
+	if (radeon->radeonScreen->chip_family == CHIP_FAMILY_RV530) {
+		if (radeon->radeonScreen->num_z_pipes == 2)
+			radeon->vtbl.emit_query_finish = rv530_emit_query_finish_double_z;
+		else
+			radeon->vtbl.emit_query_finish = rv530_emit_query_finish_single_z;
+	} else
+		radeon->vtbl.emit_query_finish = r300_emit_query_finish;
+}
+
+static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
 	ctx->Const.MaxTextureImageUnits =
 	    driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
 	ctx->Const.MaxTextureCoordUnits =
 	    driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
-	ctx->Const.MaxTextureUnits =
-	    MIN2(ctx->Const.MaxTextureImageUnits,
+	ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureImageUnits,
 		 ctx->Const.MaxTextureCoordUnits);
+
 	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 	ctx->Const.MaxTextureLodBias = 16.0;
 
-	if (screen->chip_family >= CHIP_FAMILY_RV515)
-	    ctx->Const.MaxTextureLevels = 13;
-	else
-	    ctx->Const.MaxTextureLevels = 12;
-
-        driCalculateMaxTextureLevels( r300->texture_heaps,
-                                      r300->nr_heaps,
-                                      & ctx->Const,
-                                      4,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      MIN2(ctx->Const.MaxTextureLevels,
-                                           MAX_3D_TEXTURE_LEVELS) - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      GL_FALSE,
-                                      2 );
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
+		ctx->Const.MaxTextureLevels = 13;
+		ctx->Const.MaxCubeTextureLevels = 13;
+		ctx->Const.MaxTextureRectSize = 4096;
+	}
+	else {
+		ctx->Const.MaxTextureLevels = 12;
+		ctx->Const.MaxCubeTextureLevels = 12;
+		ctx->Const.MaxTextureRectSize = 2048;
+	}
 
 	ctx->Const.MinPointSize = 1.0;
 	ctx->Const.MinPointSizeAA = 1.0;
@@ -317,263 +359,173 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
 	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
 
-#ifdef USER_BUFFERS
-	/* Needs further modifications */
-#if 0
-	ctx->Const.MaxArrayLockSize =
-	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
-#endif
-#endif
-
 	ctx->Const.MaxDrawBuffers = 1;
 
-	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
-
-	/* Initialize the software rasterizer and helper modules.
-	 */
-	_swrast_CreateContext(ctx);
-	_vbo_CreateContext(ctx);
-	_tnl_CreateContext(ctx);
-	_swsetup_CreateContext(ctx);
-	_swsetup_Wakeup(ctx);
-	_ae_create_context(ctx);
-
-	/* Install the customized pipeline:
-	 */
-	_tnl_destroy_pipeline(ctx);
-	_tnl_install_pipeline(ctx, r300_pipeline);
-
-	/* Try and keep materials and vertices separate:
-	 */
-/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
-
-	/* Configure swrast and TNL to match hardware characteristics:
-	 */
-	_swrast_allow_pixel_fog(ctx, GL_FALSE);
-	_swrast_allow_vertex_fog(ctx, GL_TRUE);
-	_tnl_allow_pixel_fog(ctx, GL_FALSE);
-	_tnl_allow_vertex_fog(ctx, GL_TRUE);
-
 	/* currently bogus data */
-	if (screen->chip_flags & RADEON_CHIPSET_TCL) {
-	        ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
-		ctx->Const.VertexProgram.MaxNativeInstructions =
-		  VSF_MAX_FRAGMENT_LENGTH / 4;
+	if (r300->options.hw_tcl_enabled) {
+		ctx->Const.VertexProgram.MaxNativeInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
+		ctx->Const.VertexProgram.MaxNativeAluInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
 		ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
-		ctx->Const.VertexProgram.MaxTemps = 32;
-		ctx->Const.VertexProgram.MaxNativeTemps =
-		  /*VSF_MAX_FRAGMENT_TEMPS */ 32;
+		ctx->Const.VertexProgram.MaxNativeTemps = 32;
 		ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
 		ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
 	}
 
-	ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
-	ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeInstructions =
-	    PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexIndirections =
-	    PFS_MAX_TEX_INDIRECT;
-	ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;	/* and these are?? */
-	ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
-	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R500_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R500_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	} else {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R300_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R300_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R300_PFS_MAX_ALU_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R300_PFS_MAX_ALU_INST + R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R300_PFS_MAX_TEX_INDIRECT;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	}
 
-	driInitExtensions(ctx, card_extensions, GL_TRUE);
+}
 
-	if (driQueryOptionb
-	    (&r300->radeon.optionCache, "disable_stencil_two_side"))
-		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
+static void r300ParseOptions(r300ContextPtr r300, radeonScreenPtr screen)
+{
+	struct r300_options options = { 0 };
 
-	if (r300->radeon.glCtx->Mesa_DXTn
-	    && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
-		_mesa_enable_extension(ctx, "GL_S3_s3tc");
-	} else
-	    if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
-	{
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
-	}
+	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
+			    screen->driScreen->myNum, "r300");
 
-	r300->disable_lowimpact_fallback =
-	    driQueryOptionb(&r300->radeon.optionCache,
-			    "disable_lowimpact_fallback");
+	r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache, "def_max_anisotropy");
 
-	radeonInitSpanFuncs(ctx);
-	r300InitCmdBuf(r300);
-	r300InitState(r300);
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-	        r300InitSwtcl(ctx);
+	options.stencil_two_side_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side");
+	options.s3tc_force_enabled = driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable");
+	options.s3tc_force_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc");
 
-	TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
+	if (!(screen->chip_flags & RADEON_CHIPSET_TCL) || driQueryOptioni(&r300->radeon.optionCache, "tcl_mode") == DRI_CONF_TCL_SW)
+		options.hw_tcl_enabled = 0;
+	else
+		options.hw_tcl_enabled = 1;
 
-	tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
-	if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
-		fprintf(stderr, "disabling 3D acceleration\n");
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
-#endif
-	}
-	if (tcl_mode == DRI_CONF_TCL_SW ||
-	    !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
-		if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-			r300->radeon.radeonScreen->chip_flags &=
-			    ~RADEON_CHIPSET_TCL;
-			fprintf(stderr, "Disabling HW TCL support\n");
-		}
-		TCL_FALLBACK(r300->radeon.glCtx,
-			     RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
-	}
+	options.conformance_mode = !driQueryOptionb(&r300->radeon.optionCache, "disable_lowimpact_fallback");
 
-	return GL_TRUE;
+	r300->options = options;
 }
 
-static void r300FreeGartAllocations(r300ContextPtr r300)
+static void r300InitGLExtensions(GLcontext *ctx)
 {
-	int i, ret, tries = 0, done_age, in_use = 0;
-	drm_radeon_mem_free_t memfree;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
-	memfree.region = RADEON_MEM_REGION_GART;
+	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r300->radeon.radeonScreen->kernel_mm)
+		driInitExtensions(ctx, mm_extensions, GL_FALSE);
 
-#ifdef USER_BUFFERS
-	for (i = r300->rmm->u_last; i > 0; i--) {
-		if (r300->rmm->u_list[i].ptr == NULL) {
-			continue;
-		}
+	if (r300->options.stencil_two_side_disabled)
+		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
 
-		/* check whether this buffer is still in use */
-		if (r300->rmm->u_list[i].pending) {
-			in_use++;
-		}
+	if (r300->options.s3tc_force_enabled) {
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+		_mesa_enable_extension(ctx, "GL_S3_s3tc");
+	} else if (r300->options.s3tc_force_disabled) {
+		_mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc");
 	}
-	/* Cannot flush/lock if no context exists. */
-	if (in_use)
-		r300FlushCmdBuf(r300, __FUNCTION__);
 
-	done_age = radeonGetAge((radeonContextPtr) r300);
+	if (!r300->radeon.radeonScreen->drmSupportsOcclusionQueries) {
+		_mesa_disable_extension(ctx, "GL_ARB_occlusion_query");
+	}
+}
 
-	for (i = r300->rmm->u_last; i > 0; i--) {
-		if (r300->rmm->u_list[i].ptr == NULL) {
-			continue;
-		}
+/* Create the device specific rendering context.
+ */
+GLboolean r300CreateContext(const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+	struct dd_function_table functions;
+	r300ContextPtr r300;
+	GLcontext *ctx;
 
-		/* check whether this buffer is still in use */
-		if (!r300->rmm->u_list[i].pending) {
-			continue;
-		}
+	assert(glVisual);
+	assert(driContextPriv);
+	assert(screen);
 
-		assert(r300->rmm->u_list[i].h_pending == 0);
+	r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
+	if (!r300)
+		return GL_FALSE;
 
-		tries = 0;
-		while (r300->rmm->u_list[i].age > done_age && tries++ < 1000) {
-			usleep(10);
-			done_age = radeonGetAge((radeonContextPtr) r300);
-		}
-		if (tries >= 1000) {
-			WARN_ONCE("Failed to idle region!");
-		}
+	r300ParseOptions(r300, screen);
 
-		memfree.region_offset = (char *)r300->rmm->u_list[i].ptr -
-		    (char *)r300->radeon.radeonScreen->gartTextures.map;
+	r300->radeon.radeonScreen = screen;
+	r300_init_vtbl(&r300->radeon);
 
-		ret = drmCommandWrite(r300->radeon.radeonScreen->driScreen->fd,
-				      DRM_RADEON_FREE, &memfree,
-				      sizeof(memfree));
-		if (ret) {
-			fprintf(stderr, "Failed to free at %p\nret = %s\n",
-				r300->rmm->u_list[i].ptr, strerror(-ret));
-		} else {
-			if (i == r300->rmm->u_last)
-				r300->rmm->u_last--;
+	_mesa_init_driver_functions(&functions);
+	r300InitIoctlFuncs(&functions);
+	r300InitStateFuncs(&functions);
+	r300InitTextureFuncs(&functions);
+	r300InitShaderFuncs(&functions);
+	radeonInitQueryObjFunctions(&functions);
+	radeonInitBufferObjectFuncs(&functions);
 
-			r300->rmm->u_list[i].pending = 0;
-			r300->rmm->u_list[i].ptr = NULL;
-		}
+	if (!radeonInitContext(&r300->radeon, &functions,
+			       glVisual, driContextPriv,
+			       sharedContextPrivate)) {
+		FREE(r300);
+		return GL_FALSE;
 	}
-	r300->rmm->u_head = i;
-#endif				/* USER_BUFFERS */
-}
 
-/* Destroy the device specific context.
- */
-void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
-{
-	GET_CURRENT_CONTEXT(ctx);
-	r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
-	radeonContextPtr radeon = (radeonContextPtr) r300;
-	radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
-	int i;
-
-	if (RADEON_DEBUG & DEBUG_DRI) {
-		fprintf(stderr, "Destroying context !\n");
-	}
+	ctx = r300->radeon.glCtx;
 
-	/* check if we're deleting the currently bound context */
-	if (&r300->radeon == current) {
-		radeonFlush(r300->radeon.glCtx);
-		_mesa_make_current(NULL, NULL, NULL);
-	}
+	r300->fallback = 0;
+	if (r300->options.hw_tcl_enabled)
+		ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
 
-	/* Free r300 context resources */
-	assert(r300);		/* should never be null */
-
-	if (r300) {
-		GLboolean release_texture_heaps;
-
-		release_texture_heaps =
-		    (r300->radeon.glCtx->Shared->RefCount == 1);
-		_swsetup_DestroyContext(r300->radeon.glCtx);
-		_tnl_DestroyContext(r300->radeon.glCtx);
-		_vbo_DestroyContext(r300->radeon.glCtx);
-		_swrast_DestroyContext(r300->radeon.glCtx);
-
-		if (r300->dma.current.buf) {
-			r300ReleaseDmaRegion(r300, &r300->dma.current,
-					     __FUNCTION__);
-#ifndef USER_BUFFERS
-			r300FlushCmdBuf(r300, __FUNCTION__);
-#endif
-		}
-		r300FreeGartAllocations(r300);
-		r300DestroyCmdBuf(r300);
+	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
-		if (radeon->state.scissor.pClipRects) {
-			FREE(radeon->state.scissor.pClipRects);
-			radeon->state.scissor.pClipRects = NULL;
-		}
+	r300InitConstValues(ctx, screen);
 
-		if (release_texture_heaps) {
-			/* This share group is about to go away, free our private
-			 * texture object data.
-			 */
-			int i;
+	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
 
-			for (i = 0; i < r300->nr_heaps; i++) {
-				driDestroyTextureHeap(r300->texture_heaps[i]);
-				r300->texture_heaps[i] = NULL;
-			}
+	/* Initialize the software rasterizer and helper modules.
+	 */
+	_swrast_CreateContext(ctx);
+	_vbo_CreateContext(ctx);
+	_tnl_CreateContext(ctx);
+	_swsetup_CreateContext(ctx);
+	_swsetup_Wakeup(ctx);
 
-			assert(is_empty_list(&r300->swapped));
-		}
+	/* Install the customized pipeline:
+	 */
+	_tnl_destroy_pipeline(ctx);
+	_tnl_install_pipeline(ctx, r300_pipeline);
+	TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
 
-                /* Drop texture object references from current hardware state */
-		for (i = 0; i < 8; i++) {
-			_mesa_reference_texobj(&r300->state.texture.unit[i].texobj, NULL);
-		}
+	/* Configure swrast and TNL to match hardware characteristics:
+	 */
+	_swrast_allow_pixel_fog(ctx, GL_FALSE);
+	_swrast_allow_vertex_fog(ctx, GL_TRUE);
+	_tnl_allow_pixel_fog(ctx, GL_FALSE);
+	_tnl_allow_vertex_fog(ctx, GL_TRUE);
 
-		radeonCleanupContext(&r300->radeon);
+	if (r300->options.hw_tcl_enabled) {
+		r300InitDraw(ctx);
+	} else {
+		r300InitSwtcl(ctx);
+	}
 
-#ifdef USER_BUFFERS
-		/* the memory manager might be accessed when Mesa frees the shared
-		 * state, so don't destroy it earlier
-		 */
-		r300_mem_destroy(r300);
-#endif
+	radeon_fbo_init(&r300->radeon);
+	radeonInitSpanFuncs( ctx );
+	r300InitCmdBuf(r300);
+	r300InitState(r300);
+	r300InitShaderFunctions(r300);
 
-		/* free the option cache */
-		driDestroyOptionCache(&r300->radeon.optionCache);
+	r300InitGLExtensions(ctx);
 
-		FREE(r300);
-	}
+	return GL_TRUE;
 }
+
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 96a3205f1a..1dadcc0a69 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -37,216 +37,29 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __R300_CONTEXT_H__
 #define __R300_CONTEXT_H__
 
-#include "tnl/t_vertex.h"
 #include "drm.h"
 #include "radeon_drm.h"
 #include "dri_util.h"
-#include "texmem.h"
+#include "radeon_common.h"
 
-#include "main/macros.h"
 #include "main/mtypes.h"
-#include "main/colormac.h"
-
-#define USER_BUFFERS
+#include "shader/prog_instruction.h"
+#include "compiler/radeon_code.h"
 
 struct r300_context;
 typedef struct r300_context r300ContextRec;
 typedef struct r300_context *r300ContextPtr;
 
-#include "radeon_lock.h"
-#include "main/mm.h"
-
-/* From http://gcc.gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
-   I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble
-   with other compilers ... GLUE!
-*/
-#define WARN_ONCE(a, ...)	{ \
-	static int warn##__LINE__=1; \
-	if(warn##__LINE__){ \
-		fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \
-		fprintf(stderr, "File %s function %s line %d\n", \
-			__FILE__, __FUNCTION__, __LINE__); \
-		fprintf(stderr,  a, ## __VA_ARGS__);\
-		fprintf(stderr, "***************************************************************************\n"); \
-		warn##__LINE__=0;\
-		} \
-	}
 
 #include "r300_vertprog.h"
-#include "r500_fragprog.h"
-
-/**
- * This function takes a float and packs it into a uint32_t
- */
-static INLINE uint32_t r300PackFloat32(float fl)
-{
-	union {
-		float fl;
-		uint32_t u;
-	} u;
-
-	u.fl = fl;
-	return u.u;
-}
-
-/* This is probably wrong for some values, I need to test this
- * some more.  Range checking would be a good idea also..
- *
- * But it works for most things.  I'll fix it later if someone
- * else with a better clue doesn't
- */
-static INLINE uint32_t r300PackFloat24(float f)
-{
-	float mantissa;
-	int exponent;
-	uint32_t float24 = 0;
-
-	if (f == 0.0)
-		return 0;
-
-	mantissa = frexpf(f, &exponent);
-
-	/* Handle -ve */
-	if (mantissa < 0) {
-		float24 |= (1 << 23);
-		mantissa = mantissa * -1.0;
-	}
-	/* Handle exponent, bias of 63 */
-	exponent += 62;
-	float24 |= (exponent << 16);
-	/* Kill 7 LSB of mantissa */
-	float24 |= (r300PackFloat32(mantissa) & 0x7FFFFF) >> 7;
-
-	return float24;
-}
-
-/************ DMA BUFFERS **************/
-
-/* Need refcounting on dma buffers:
- */
-struct r300_dma_buffer {
-	int refcount;		/**< the number of retained regions in buf */
-	drmBufPtr buf;
-	int id;
-};
-#undef GET_START
-#ifdef USER_BUFFERS
-#define GET_START(rvb) (r300GartOffsetFromVirtual(rmesa, (rvb)->address+(rvb)->start))
-#else
-#define GET_START(rvb) (rmesa->radeon.radeonScreen->gart_buffer_offset +		\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-#endif
-/* A retained region, eg vertices for indexed vertices.
- */
-struct r300_dma_region {
-	struct r300_dma_buffer *buf;
-	char *address;		/* == buf->address */
-	int start, end, ptr;	/* offsets from start of buf */
-
-	int aos_offset;		/* address in GART memory */
-	int aos_stride;		/* distance between elements, in dwords */
-	int aos_size;		/* number of components (1-4) */
-};
-
-struct r300_dma {
-	/* Active dma region.  Allocations for vertices and retained
-	 * regions come from here.  Also used for emitting random vertices,
-	 * these may be flushed by calling flush_current();
-	 */
-	struct r300_dma_region current;
 
-	void (*flush) (r300ContextPtr);
-
-	char *buf0_address;	/* start of buf[0], for index calcs */
-
-	/* Number of "in-flight" DMA buffers, i.e. the number of buffers
-	 * for which a DISCARD command is currently queued in the command buffer.
-	 */
-	GLuint nr_released_bufs;
-};
-
-       /* Texture related */
-
-typedef struct r300_tex_obj r300TexObj, *r300TexObjPtr;
-
-/* Maximum number of mipmap levels supported by any supported GPU
- */
-#define R300_MAX_TEXTURE_LEVELS 13
-
-/* Texture object in locally shared texture space.
- */
-struct r300_tex_obj {
-	driTextureObject base;
-
-	GLuint bufAddr;		/* Offset to start of locally
-				   shared texture block */
-
-	drm_radeon_tex_image_t image[6][R300_MAX_TEXTURE_LEVELS];
-	/* Six, for the cube faces */
-
-	GLboolean image_override;	/* Image overridden by GLX_EXT_tfp */
-
-	GLuint pitch;		/* this isn't sent to hardware just used in calculations */
-	/* hardware register values */
-	/* Note that R200 has 8 registers per texture and R300 only 7 */
-	GLuint filter;
-	GLuint filter_1;
-	GLuint pitch_reg;
-	GLuint size;		/* npot only */
-	GLuint format;
-	GLuint offset;		/* Image location in the card's address space.
-				   All cube faces follow. */
-	GLuint unknown4;
-	GLuint unknown5;
-	/* end hardware registers */
-
-	/* registers computed by r200 code - keep them here to
-	   compare against what is actually written.
-
-	   to be removed later.. */
-	GLuint pp_border_color;
-	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
-	GLuint format_x;
-
-	GLboolean border_fallback;
-
-	GLuint tile_bits;	/* hw texture tile bits used on this texture */
-};
-
-struct r300_texture_env_state {
-	struct gl_texture_object *texobj;
-	GLenum format;
-	GLenum envMode;
-};
 
 /* The blit width for texture uploads
  */
 #define R300_BLIT_WIDTH_BYTES 1024
 #define R300_MAX_TEXTURE_UNITS 8
 
-struct r300_texture_state {
-	struct r300_texture_env_state unit[R300_MAX_TEXTURE_UNITS];
-	int tc_count;		/* number of incoming texture coordinates from VAP */
-};
 
-/**
- * A block of hardware state.
- *
- * When check returns non-zero, the returned number of dwords must be
- * copied verbatim into the command buffer in order to update a state atom
- * when it is dirty.
- */
-struct r300_state_atom {
-	struct r300_state_atom *next, *prev;
-	const char *name;	/* for debug */
-	int cmd_size;		/* maximum size in dwords */
-	GLuint idx;		/* index in an array (e.g. textures) */
-	uint32_t *cmd;
-	GLboolean dirty;
-
-	int (*check) (r300ContextPtr, struct r300_state_atom * atom);
-};
 
 #define R300_VPT_CMD_0		0
 #define R300_VPT_XSCALE		1
@@ -288,9 +101,11 @@ struct r300_state_atom {
 #define R300_GB_MISC_MSPOS_0		1
 #define R300_GB_MISC_MSPOS_1		2
 #define R300_GB_MISC_TILE_CONFIG	3
-#define R300_GB_MISC_SELECT		4
-#define R300_GB_MISC_AA_CONFIG		5
-#define R300_GB_MISC_CMDSIZE		6
+#define R300_GB_MISC_CMDSIZE		4
+#define R300_GB_MISC2_CMD_0		    0
+#define R300_GB_MISC2_SELECT		1
+#define R300_GB_MISC2_AA_CONFIG		2
+#define R300_GB_MISC2_CMDSIZE		3
 
 #define R300_TXE_CMD_0		0
 #define R300_TXE_ENABLE		1
@@ -459,423 +274,180 @@ struct r300_state_atom {
 #define R300_TEX_CMDSIZE	(MAX_TEXTURE_UNITS+1)
 */
 
+#define R300_QUERYOBJ_CMD_0  0
+#define R300_QUERYOBJ_DATA_0 1
+#define R300_QUERYOBJ_CMD_1  2
+#define R300_QUERYOBJ_DATA_1  3
+#define R300_QUERYOBJ_CMDSIZE  4
+
 /**
  * Cache for hardware register state.
  */
 struct r300_hw_state {
-	struct r300_state_atom atomlist;
-
-	GLboolean is_dirty;
-	GLboolean all_dirty;
-	int max_state_size;	/* in dwords */
-
-	struct r300_state_atom vpt;	/* viewport (1D98) */
-	struct r300_state_atom vap_cntl;
-        struct r300_state_atom vap_index_offset; /* 0x208c r5xx only */
-	struct r300_state_atom vof;	/* VAP output format register 0x2090 */
-	struct r300_state_atom vte;	/* (20B0) */
-	struct r300_state_atom vap_vf_max_vtx_indx;	/* Maximum Vertex Indx Clamp (2134) */
-	struct r300_state_atom vap_cntl_status;
-	struct r300_state_atom vir[2];	/* vap input route (2150/21E0) */
-	struct r300_state_atom vic;	/* vap input control (2180) */
-	struct r300_state_atom vap_psc_sgn_norm_cntl; /* Programmable Stream Control Signed Normalize Control (21DC) */
-	struct r300_state_atom vap_clip_cntl;
-	struct r300_state_atom vap_clip;
-	struct r300_state_atom vap_pvs_vtx_timeout_reg;	/* Vertex timeout register (2288) */
-	struct r300_state_atom pvs;	/* pvs_cntl (22D0) */
-	struct r300_state_atom gb_enable;	/* (4008) */
-	struct r300_state_atom gb_misc;	/* Multisampling position shifts ? (4010) */
-	struct r300_state_atom ga_point_s0;	/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) (4200) */
-	struct r300_state_atom ga_triangle_stipple;	/* (4214) */
-	struct r300_state_atom ps;	/* pointsize (421C) */
-	struct r300_state_atom ga_point_minmax;	/* (4230) */
-	struct r300_state_atom lcntl;	/* line control */
-	struct r300_state_atom ga_line_stipple;	/* (4260) */
-	struct r300_state_atom shade;
-	struct r300_state_atom polygon_mode;
-	struct r300_state_atom fogp;	/* fog parameters (4294) */
-	struct r300_state_atom ga_soft_reset;	/* (429C) */
-	struct r300_state_atom zbias_cntl;
-	struct r300_state_atom zbs;	/* zbias (42A4) */
-	struct r300_state_atom occlusion_cntl;
-	struct r300_state_atom cul;	/* cull cntl (42B8) */
-	struct r300_state_atom su_depth_scale;	/* (42C0) */
-	struct r300_state_atom rc;	/* rs control (4300) */
-	struct r300_state_atom ri;	/* rs interpolators (4310) */
-	struct r300_state_atom rr;	/* rs route (4330) */
-	struct r300_state_atom sc_hyperz;	/* (43A4) */
-	struct r300_state_atom sc_screendoor;	/* (43E8) */
-	struct r300_state_atom fp;	/* fragment program cntl + nodes (4600) */
-	struct r300_state_atom fpt;	/* texi - (4620) */
-	struct r300_state_atom us_out_fmt;	/* (46A4) */
-	struct r300_state_atom r500fp;	/* r500 fp instructions */
-	struct r300_state_atom r500fp_const;	/* r500 fp constants */
-	struct r300_state_atom fpi[4];	/* fp instructions (46C0/47C0/48C0/49C0) */
-	struct r300_state_atom fogs;	/* fog state (4BC0) */
-	struct r300_state_atom fogc;	/* fog color (4BC8) */
-	struct r300_state_atom at;	/* alpha test (4BD4) */
-	struct r300_state_atom fg_depth_src;	/* (4BD8) */
-	struct r300_state_atom fpp;	/* 0x4C00 and following */
-	struct r300_state_atom rb3d_cctl;	/* (4E00) */
-	struct r300_state_atom bld;	/* blending (4E04) */
-	struct r300_state_atom cmk;	/* colormask (4E0C) */
-	struct r300_state_atom blend_color;	/* constant blend color */
-	struct r300_state_atom rop;	/* ropcntl */
-	struct r300_state_atom cb;	/* colorbuffer (4E28) */
-	struct r300_state_atom rb3d_dither_ctl;	/* (4E50) */
-	struct r300_state_atom rb3d_aaresolve_ctl;	/* (4E88) */
-	struct r300_state_atom rb3d_discard_src_pixel_lte_threshold;	/* (4E88) I saw it only written on RV350 hardware..  */
-	struct r300_state_atom zs;	/* zstencil control (4F00) */
-	struct r300_state_atom zstencil_format;
-	struct r300_state_atom zb;	/* z buffer (4F20) */
-	struct r300_state_atom zb_depthclearvalue;	/* (4F28) */
-	struct r300_state_atom unk4F30;	/* (4F30) */
-	struct r300_state_atom zb_hiz_offset;	/* (4F44) */
-	struct r300_state_atom zb_hiz_pitch;	/* (4F54) */
-
-	struct r300_state_atom vpi;	/* vp instructions */
-	struct r300_state_atom vpp;	/* vp parameters */
-	struct r300_state_atom vps;	/* vertex point size (?) */
-	struct r300_state_atom vpucp[6];	/* vp user clip plane - 6 */
+	struct radeon_state_atom vpt;	/* viewport (1D98) */
+	struct radeon_state_atom vap_cntl;
+	struct radeon_state_atom vap_index_offset; /* 0x208c r5xx only */
+	struct radeon_state_atom vof;	/* VAP output format register 0x2090 */
+	struct radeon_state_atom vte;	/* (20B0) */
+	struct radeon_state_atom vap_vf_max_vtx_indx;	/* Maximum Vertex Indx Clamp (2134) */
+	struct radeon_state_atom vap_cntl_status;
+	struct radeon_state_atom vir[2];	/* vap input route (2150/21E0) */
+	struct radeon_state_atom vic;	/* vap input control (2180) */
+	struct radeon_state_atom vap_psc_sgn_norm_cntl; /* Programmable Stream Control Signed Normalize Control (21DC) */
+	struct radeon_state_atom vap_clip_cntl;
+	struct radeon_state_atom vap_clip;
+	struct radeon_state_atom vap_pvs_vtx_timeout_reg;	/* Vertex timeout register (2288) */
+	struct radeon_state_atom pvs;	/* pvs_cntl (22D0) */
+	struct radeon_state_atom gb_enable;	/* (4008) */
+	struct radeon_state_atom gb_misc;	/* Multisampling position shifts ? (4010) */
+	struct radeon_state_atom gb_misc2;	/* Multisampling position shifts ? (4010) */
+	struct radeon_state_atom ga_point_s0;	/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) (4200) */
+	struct radeon_state_atom ga_triangle_stipple;	/* (4214) */
+	struct radeon_state_atom ps;	/* pointsize (421C) */
+	struct radeon_state_atom ga_point_minmax;	/* (4230) */
+	struct radeon_state_atom lcntl;	/* line control */
+	struct radeon_state_atom ga_line_stipple;	/* (4260) */
+	struct radeon_state_atom shade;
+	struct radeon_state_atom shade2;
+	struct radeon_state_atom polygon_mode;
+	struct radeon_state_atom fogp;	/* fog parameters (4294) */
+	struct radeon_state_atom ga_soft_reset;	/* (429C) */
+	struct radeon_state_atom zbias_cntl;
+	struct radeon_state_atom zbs;	/* zbias (42A4) */
+	struct radeon_state_atom occlusion_cntl;
+	struct radeon_state_atom cul;	/* cull cntl (42B8) */
+	struct radeon_state_atom su_depth_scale;	/* (42C0) */
+	struct radeon_state_atom rc;	/* rs control (4300) */
+	struct radeon_state_atom ri;	/* rs interpolators (4310) */
+	struct radeon_state_atom rr;	/* rs route (4330) */
+	struct radeon_state_atom sc_hyperz;	/* (43A4) */
+	struct radeon_state_atom sc_screendoor;	/* (43E8) */
+	struct radeon_state_atom fp;	/* fragment program cntl + nodes (4600) */
+	struct radeon_state_atom fpt;	/* texi - (4620) */
+	struct radeon_state_atom us_out_fmt;	/* (46A4) */
+	struct radeon_state_atom r500fp;	/* r500 fp instructions */
+	struct radeon_state_atom r500fp_const;	/* r500 fp constants */
+	struct radeon_state_atom fpi[4];	/* fp instructions (46C0/47C0/48C0/49C0) */
+	struct radeon_state_atom fogs;	/* fog state (4BC0) */
+	struct radeon_state_atom fogc;	/* fog color (4BC8) */
+	struct radeon_state_atom at;	/* alpha test (4BD4) */
+	struct radeon_state_atom fg_depth_src;	/* (4BD8) */
+	struct radeon_state_atom fpp;	/* 0x4C00 and following */
+	struct radeon_state_atom rb3d_cctl;	/* (4E00) */
+	struct radeon_state_atom bld;	/* blending (4E04) */
+	struct radeon_state_atom cmk;	/* colormask (4E0C) */
+	struct radeon_state_atom blend_color;	/* constant blend color */
+	struct radeon_state_atom rop;	/* ropcntl */
+	struct radeon_state_atom cb;	/* colorbuffer (4E28) */
+	struct radeon_state_atom rb3d_dither_ctl;	/* (4E50) */
+	struct radeon_state_atom rb3d_aaresolve_ctl;	/* (4E88) */
+	struct radeon_state_atom rb3d_discard_src_pixel_lte_threshold;	/* (4E88) I saw it only written on RV350 hardware..  */
+	struct radeon_state_atom zs;	/* zstencil control (4F00) */
+	struct radeon_state_atom zstencil_format;
+	struct radeon_state_atom zb;	/* z buffer (4F20) */
+	struct radeon_state_atom zb_depthclearvalue;	/* (4F28) */
+	struct radeon_state_atom zb_zmask;	/* (4F30) */
+	struct radeon_state_atom zb_hiz_offset;	/* (4F44) */
+	struct radeon_state_atom zb_hiz_pitch;	/* (4F54) */
+
+	struct radeon_state_atom vap_flush;
+	struct radeon_state_atom vpi;	/* vp instructions */
+	struct radeon_state_atom vpp;	/* vp parameters */
+	struct radeon_state_atom vps;	/* vertex point size (?) */
+	struct radeon_state_atom vpucp[6];	/* vp user clip plane - 6 */
 	/* 8 texture units */
 	/* the state is grouped by function and not by
 	   texture unit. This makes single unit updates
 	   really awkward - we are much better off
 	   updating the whole thing at once */
 	struct {
-		struct r300_state_atom filter;
-		struct r300_state_atom filter_1;
-		struct r300_state_atom size;
-		struct r300_state_atom format;
-		struct r300_state_atom pitch;
-		struct r300_state_atom offset;
-		struct r300_state_atom chroma_key;
-		struct r300_state_atom border_color;
+		struct radeon_state_atom filter;
+		struct radeon_state_atom filter_1;
+		struct radeon_state_atom size;
+		struct radeon_state_atom format;
+		struct radeon_state_atom pitch;
+		struct radeon_state_atom offset;
+		struct radeon_state_atom chroma_key;
+		struct radeon_state_atom border_color;
 	} tex;
-	struct r300_state_atom txe;	/* tex enable (4104) */
-};
-
-/**
- * This structure holds the command buffer while it is being constructed.
- *
- * The first batch of commands in the buffer is always the state that needs
- * to be re-emitted when the context is lost. This batch can be skipped
- * otherwise.
- */
-struct r300_cmdbuf {
-	int size;		/* DWORDs allocated for buffer */
-	uint32_t *cmd_buf;
-	int count_used;		/* DWORDs filled so far */
-	int count_reemit;	/* size of re-emission batch */
+	struct radeon_state_atom txe;	/* tex enable (4104) */
+	radeonTexObj *textures[R300_MAX_TEXTURE_UNITS];
 };
 
 /**
  * State cache
  */
 
-struct r300_depthbuffer_state {
-	GLfloat scale;
-};
-
-struct r300_stencilbuffer_state {
-	GLboolean hw_stencil;
-};
-
 /* Vertex shader state */
 
-/* Perhaps more if we store programs in vmem? */
-/* drm_r300_cmd_header_t->vpu->count is unsigned char */
-#define VSF_MAX_FRAGMENT_LENGTH (255*4)
-
-/* Can be tested with colormat currently. */
-#define VSF_MAX_FRAGMENT_TEMPS (14)
-
-#define STATE_R300_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
-#define STATE_R300_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
-
-struct r300_vertex_shader_fragment {
-	int length;
-	union {
-		GLuint d[VSF_MAX_FRAGMENT_LENGTH];
-		float f[VSF_MAX_FRAGMENT_LENGTH];
-		GLuint i[VSF_MAX_FRAGMENT_LENGTH];
-	} body;
-};
-
-struct r300_vertex_shader_state {
-	struct r300_vertex_shader_fragment program;
-};
-
-extern int hw_tcl_on;
-
 #define COLOR_IS_RGBA
 #define TAG(x) r300##x
 #include "tnl_dd/t_dd_vertex.h"
 #undef TAG
 
-//#define CURRENT_VERTEX_SHADER(ctx) (ctx->VertexProgram._Current)
-#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->selected_vp)
-
-/* Should but doesnt work */
-//#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->curr_vp)
-
-/* r300_vertex_shader_state and r300_vertex_program should probably be merged together someday.
- * Keeping them them seperate for now should ensure fixed pipeline keeps functioning properly.
- */
-
 struct r300_vertex_program_key {
-	GLuint InputsRead;
-	GLuint OutputsWritten;
-	GLuint OutputsAdded;
+	GLbitfield FpReads;
+	GLuint FogAttr;
+	GLuint WPosAttr;
 };
 
 struct r300_vertex_program {
+	struct gl_vertex_program *Base;
 	struct r300_vertex_program *next;
+
 	struct r300_vertex_program_key key;
-	int translated;
-
-	struct r300_vertex_shader_fragment program;
-
-	int pos_end;
-	int num_temporaries;	/* Number of temp vars used by program */
-	int wpos_idx;
-	int inputs[VERT_ATTRIB_MAX];
-	int outputs[VERT_RESULT_MAX];
-	int native;
-	int ref_count;
-	int use_ref_count;
+	struct r300_vertex_program_code code;
+
+	GLboolean error;
 };
 
 struct r300_vertex_program_cont {
-	struct gl_vertex_program mesa_program;	/* Must be first */
-	struct r300_vertex_shader_fragment params;
+	/* This is the unmodified vertex program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific vertex program based on this.
+	 */
+	struct gl_vertex_program mesa_program;
+	/* This is the list of hw specific vertex programs derived from mesa_program */
 	struct r300_vertex_program *progs;
 };
 
-#define PFS_MAX_ALU_INST	64
-#define PFS_MAX_TEX_INST	64
-#define PFS_MAX_TEX_INDIRECT 4
-#define PFS_NUM_TEMP_REGS	32
-#define PFS_NUM_CONST_REGS	16
-
-struct r300_pfs_compile_state;
-
 
 /**
- * Stores state that influences the compilation of a fragment program.
- */
-struct r300_fragment_program_external_state {
-	struct {
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is:
-		 *  0 - GL_LUMINANCE
-		 *  1 - GL_INTENSITY
-		 *  2 - GL_ALPHA
-		 * depending on the depth texture mode.
-		 */
-		GLuint depth_texture_mode : 2;
-
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is (texture_compare_func - GL_NEVER).
-		 * [e.g. if compare function is GL_LEQUAL, this field is 3]
-		 *
-		 * Otherwise, this field is 0.
-		 */
-		GLuint texture_compare_func : 3;
-	} unit[16];
-};
-
-
-struct r300_fragment_program_node {
-	int tex_offset; /**< first tex instruction */
-	int tex_end; /**< last tex instruction, relative to tex_offset */
-	int alu_offset; /**< first ALU instruction */
-	int alu_end; /**< last ALU instruction, relative to alu_offset */
-	int flags;
-};
-
-/**
- * Stores an R300 fragment program in its compiled-to-hardware form.
- */
-struct r300_fragment_program_code {
-	struct {
-		int length; /**< total # of texture instructions used */
-		GLuint inst[PFS_MAX_TEX_INST];
-	} tex;
-
-	struct {
-		int length; /**< total # of ALU instructions used */
-		struct {
-			GLuint inst0;
-			GLuint inst1;
-			GLuint inst2;
-			GLuint inst3;
-		} inst[PFS_MAX_ALU_INST];
-	} alu;
-
-	struct r300_fragment_program_node node[4];
-	int cur_node;
-	int first_node_has_tex;
-
-	/**
-	 * Remember which program register a given hardware constant
-	 * belongs to.
-	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
-	int const_nr;
-
-	int max_temp_idx;
-};
-
-/**
- * Store everything about a fragment program that is needed
- * to render with that program.
- */
+* Store everything about a fragment program that is needed
+* to render with that program.
+*/
 struct r300_fragment_program {
-	struct gl_fragment_program mesa_program;
-
-	GLboolean translated;
 	GLboolean error;
-
+	struct r300_fragment_program *next;
 	struct r300_fragment_program_external_state state;
-	struct r300_fragment_program_code code;
 
-	GLboolean WritesDepth;
-	GLuint optimization;
-};
-
-struct r500_pfs_compile_state;
+	struct rX00_fragment_program_code code;
+	GLbitfield InputsRead;
 
-struct r500_fragment_program_external_state {
-	struct {
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is:
-		 *  0 - GL_LUMINANCE
-		 *  1 - GL_INTENSITY
-		 *  2 - GL_ALPHA
-		 * depending on the depth texture mode.
-		 */
-		GLuint depth_texture_mode : 2;
-
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is (texture_compare_func - GL_NEVER).
-		 * [e.g. if compare function is GL_LEQUAL, this field is 3]
-		 *
-		 * Otherwise, this field is 0.
-		 */
-		GLuint texture_compare_func : 3;
-	} unit[16];
+	/* attribute that we are sending the WPOS in */
+	gl_frag_attrib wpos_attr;
+	/* attribute that we are sending the fog coordinate in */
+	gl_frag_attrib fog_attr;
 };
 
-struct r500_fragment_program_code {
-	struct {
-		GLuint inst0;
-		GLuint inst1;
-		GLuint inst2;
-		GLuint inst3;
-		GLuint inst4;
-		GLuint inst5;
-	} inst[512];
-
-	int inst_offset;
-	int inst_end;
-
-	/**
-	 * Remember which program register a given hardware constant
-	 * belongs to.
+struct r300_fragment_program_cont {
+	/* This is the unmodified fragment program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific fragment program based on this.
 	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
-	int const_nr;
-
-	int max_temp_idx;
+	struct gl_fragment_program Base;
+	/* This is the list of hw specific fragment programs derived from Base */
+	struct r300_fragment_program *progs;
 };
 
-struct r500_fragment_program {
-	struct gl_fragment_program mesa_program;
-
-	GLcontext *ctx;
-	GLboolean translated;
-	GLboolean error;
-
-	struct r500_fragment_program_external_state state;
-	struct r500_fragment_program_code code;
-
-	GLboolean writes_depth;
-
-	GLuint optimization;
-};
 
 #define R300_MAX_AOS_ARRAYS		16
 
-#define REG_COORDS	0
-#define REG_COLOR0	1
-#define REG_TEX0	2
-
-struct r300_state {
-	struct r300_depthbuffer_state depth;
-	struct r300_texture_state texture;
-	int sw_tcl_inputs[VERT_ATTRIB_MAX];
-	struct r300_vertex_shader_state vertex_shader;
-	struct r300_dma_region aos[R300_MAX_AOS_ARRAYS];
-	int aos_count;
-
-	GLuint *Elts;
-	struct r300_dma_region elt_dma;
-
-	struct r300_dma_region swtcl_dma;
-	DECLARE_RENDERINPUTS(render_inputs_bitset);	/* actual render inputs that R300 was configured for.
-							   They are the same as tnl->render_inputs for fixed pipeline */
-
-	struct r300_stencilbuffer_state stencil;
-
-};
-
-#define R300_FALLBACK_NONE 0
-#define R300_FALLBACK_TCL 1
-#define R300_FALLBACK_RAST 2
 
 /* r300_swtcl.c
  */
 struct r300_swtcl_info {
-   GLuint RenderIndex;
-
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
-
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
-
-   /* Fallback rasterization functions
-    */
-  //   r200_point_func draw_point;
-  //   r200_line_func draw_line;
-  //   r200_tri_func draw_tri;
-
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
-
-   /**
+  /*
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
    GLuint coloroffset;
@@ -884,13 +456,43 @@ struct r300_swtcl_info {
     * Offset of the 3UB specular color data within a hardware (swtcl) vertex.
     */
    GLuint specoffset;
+};
 
-   /**
-    * Should Mesa project vertex data or will the hardware do it?
-    */
-   GLboolean needproj;
+struct r300_vtable {
+	void (* SetupRSUnit)(GLcontext *ctx);
+	void (* SetupFragmentShaderTextures)(GLcontext *ctx, int *tmu_mappings);
+	void (* SetupPixelShader)(GLcontext *ctx);
+};
 
-   struct r300_dma_region indexed_verts;
+struct r300_vertex_buffer {
+	struct vertex_attribute {
+		/* generic */
+		GLubyte element;
+		GLuint stride;
+		GLuint dwords;
+		GLubyte size; /* number of components */
+		GLboolean is_named_bo;
+		struct radeon_bo *bo;
+		GLint bo_offset;
+
+		/* hw specific */
+		uint32_t data_type:4;
+		uint32_t dst_loc:5;
+		uint32_t _signed:1;
+		uint32_t normalize:1;
+		uint32_t swizzle:12;
+		uint32_t write_mask:4;
+	} attribs[VERT_ATTRIB_MAX];
+
+	GLubyte num_attribs;
+};
+
+struct r300_index_buffer {
+	struct radeon_bo *bo;
+	int bo_offset;
+
+	GLboolean is_32bit;
+	GLuint count;
 };
 
 
@@ -900,46 +502,33 @@ struct r300_swtcl_info {
 struct r300_context {
 	struct radeon_context radeon;	/* parent class, must be first */
 
+	struct r300_vtable vtbl;
+
 	struct r300_hw_state hw;
-	struct r300_cmdbuf cmdbuf;
-	struct r300_state state;
-	struct gl_vertex_program *curr_vp;
+
 	struct r300_vertex_program *selected_vp;
+	struct r300_fragment_program *selected_fp;
 
 	/* Vertex buffers
 	 */
-	struct r300_dma dma;
-	GLboolean save_on_next_unlock;
-	GLuint NewGLState;
-
-	/* Texture object bookkeeping
-	 */
-	unsigned nr_heaps;
-	driTexHeap *texture_heaps[RADEON_NR_TEX_HEAPS];
-	driTextureObject swapped;
-	int texture_depth;
-	float initialMaxAnisotropy;
-
-	/* Clientdata textures;
-	 */
-	GLuint prefer_gart_client_texturing;
-
-#ifdef USER_BUFFERS
-	struct r300_memory_manager *rmm;
-#endif
-
 	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
 	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
 
-	GLboolean disable_lowimpact_fallback;
+	struct r300_options {
+		uint32_t conformance_mode:1;
+		uint32_t hw_tcl_enabled:1;
+		uint32_t s3tc_force_enabled:1;
+		uint32_t s3tc_force_disabled:1;
+		uint32_t stencil_two_side_disabled:1;
+	} options;
 
-	DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
 	struct r300_swtcl_info swtcl;
-};
+	struct r300_vertex_buffer vbuf;
+	struct r300_index_buffer ind_buf;
 
-struct r300_buffer_object {
-	struct gl_buffer_object mesa_obj;
-	int id;
+	uint32_t fallback;
+
+	DECLARE_RENDERINPUTS(render_inputs_bitset);
 };
 
 #define R300_CONTEXT(ctx)		((r300ContextPtr)(ctx->DriverCtx))
@@ -949,15 +538,13 @@ extern GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 				   __DRIcontextPrivate * driContextPriv,
 				   void *sharedContextPrivate);
 
-extern void r300SelectVertexShader(r300ContextPtr r300);
 extern void r300InitShaderFuncs(struct dd_function_table *functions);
-extern int r300VertexProgUpdateParams(GLcontext * ctx,
-				      struct r300_vertex_program_cont *vp,
-				      float *dst);
-
-#define RADEON_D_CAPTURE 0
-#define RADEON_D_PLAYBACK 1
-#define RADEON_D_PLAYBACK_RAW 2
-#define RADEON_D_T 3
+
+extern void r300InitShaderFunctions(r300ContextPtr r300);
+
+extern void r300InitDraw(GLcontext *ctx);
+
+#define r300PackFloat32 radeonPackFloat32
+#define r300PackFloat24 radeonPackFloat24
 
 #endif				/* __R300_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
new file mode 100644
index 0000000000..e9968f9ffe
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -0,0 +1,718 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Maciej Cencora
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHOR(S) AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <stdlib.h>
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/state.h"
+#include "main/api_validate.h"
+#include "main/enums.h"
+#include "main/simple_list.h"
+
+#include "r300_reg.h"
+#include "r300_context.h"
+#include "r300_emit.h"
+#include "r300_render.h"
+#include "r300_state.h"
+#include "r300_tex.h"
+#include "r300_cmdbuf.h"
+
+#include "radeon_buffer_objects.h"
+#include "radeon_common_context.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "vbo/vbo_context.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+
+
+static int getTypeSize(GLenum type)
+{
+	switch (type) {
+		case GL_DOUBLE:
+			return sizeof(GLdouble);
+		case GL_FLOAT:
+			return sizeof(GLfloat);
+		case GL_INT:
+			return sizeof(GLint);
+		case GL_UNSIGNED_INT:
+			return sizeof(GLuint);
+		case GL_SHORT:
+			return sizeof(GLshort);
+		case GL_UNSIGNED_SHORT:
+			return sizeof(GLushort);
+		case GL_BYTE:
+			return sizeof(GLbyte);
+		case GL_UNSIGNED_BYTE:
+			return sizeof(GLubyte);
+		default:
+			assert(0);
+			return 0;
+	}
+}
+
+static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLvoid *src_ptr;
+	GLuint *out;
+	int i;
+	GLboolean mapped_named_bo = GL_FALSE;
+
+	if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+		ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+		mapped_named_bo = GL_TRUE;
+		assert(mesa_ind_buf->obj->Pointer != NULL);
+	}
+	src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+	radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT,
+			"%s: Fixing index buffer format. type %d\n",
+			__func__, mesa_ind_buf->type);
+
+	if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
+		GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1);
+		GLubyte *in = (GLubyte *)src_ptr;
+
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4);
+
+		assert(r300->ind_buf.bo->ptr != NULL);
+		out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+
+#if MESA_BIG_ENDIAN
+	} else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */
+		GLushort *in = (GLushort *)src_ptr;
+		GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1);
+
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo,
+				     &r300->ind_buf.bo_offset, size, 4);
+
+		assert(r300->ind_buf.bo->ptr != NULL);
+		out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+#endif
+	}
+
+	r300->ind_buf.is_32bit = GL_FALSE;
+	r300->ind_buf.count = mesa_ind_buf->count;
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj);
+	}
+}
+
+
+static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	if (!mesa_ind_buf) {
+		r300->ind_buf.bo = NULL;
+		return;
+	}
+	radeon_print(RADEON_RENDER, RADEON_TRACE, "%s\n", __func__);
+
+#if MESA_BIG_ENDIAN
+	if (mesa_ind_buf->type == GL_UNSIGNED_INT) {
+#else
+	if (mesa_ind_buf->type != GL_UNSIGNED_BYTE) {
+#endif
+		const GLvoid *src_ptr;
+		GLvoid *dst_ptr;
+		GLboolean mapped_named_bo = GL_FALSE;
+
+		if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+			ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+			assert(mesa_ind_buf->obj->Pointer != NULL);
+			mapped_named_bo = GL_TRUE;
+		}
+
+		src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+		const GLuint size = mesa_ind_buf->count * getTypeSize(mesa_ind_buf->type);
+
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4);
+
+		assert(r300->ind_buf.bo->ptr != NULL);
+		dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
+		_mesa_memcpy(dst_ptr, src_ptr, size);
+
+		r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT);
+		r300->ind_buf.count = mesa_ind_buf->count;
+
+		if (mapped_named_bo) {
+			ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj);
+		}
+	} else {
+		r300FixupIndexBuffer(ctx, mesa_ind_buf);
+	}
+}
+
+#define CONVERT( TYPE, MACRO ) do {		\
+	GLuint i, j, sz;				\
+	sz = input->Size;				\
+	if (input->Normalized) {			\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = MACRO(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	} else {					\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = (GLfloat)(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	}						\
+} while (0)
+
+/**
+ * Convert attribute data type to float
+ * If the attribute uses named buffer object replace the bo with newly allocated bo
+ */
+static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_array *input, struct vertex_attribute *attr)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	const GLvoid *src_ptr;
+	GLboolean mapped_named_bo = GL_FALSE;
+	GLfloat *dst_ptr;
+	GLuint stride;
+
+	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
+
+	/* Convert value for first element only */
+	if (input->StrideB == 0)
+		count = 1;
+
+	if (input->BufferObj->Name) {
+		if (!input->BufferObj->Pointer) {
+			ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+			mapped_named_bo = GL_TRUE;
+		}
+
+		src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+	} else {
+		src_ptr = input->Ptr;
+	}
+
+	radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32);
+	dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset);
+
+	radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT,
+			"%s: Converting vertex attributes, attribute data format %x,"
+			"stride %d, components %d\n"
+			, __FUNCTION__, input->Type
+			, stride, input->Size);
+
+	assert(src_ptr != NULL);
+
+	switch (input->Type) {
+		case GL_DOUBLE:
+			CONVERT(GLdouble, (GLfloat));
+			break;
+		case GL_UNSIGNED_INT:
+			CONVERT(GLuint, UINT_TO_FLOAT);
+			break;
+		case GL_INT:
+			CONVERT(GLint, INT_TO_FLOAT);
+			break;
+		case GL_UNSIGNED_SHORT:
+			CONVERT(GLushort, USHORT_TO_FLOAT);
+			break;
+		case GL_SHORT:
+			CONVERT(GLshort, SHORT_TO_FLOAT);
+			break;
+		case GL_UNSIGNED_BYTE:
+			assert(input->Format != GL_BGRA);
+			CONVERT(GLubyte, UBYTE_TO_FLOAT);
+			break;
+		case GL_BYTE:
+			CONVERT(GLbyte, BYTE_TO_FLOAT);
+			break;
+		default:
+			assert(0);
+			break;
+	}
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj);
+	}
+}
+
+static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *input, int count, struct vertex_attribute *attr)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	const int dst_stride = (input->StrideB + 3) & ~3;
+	const int size = getTypeSize(input->Type) * input->Size * count;
+	GLboolean mapped_named_bo = GL_FALSE;
+
+	radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32);
+
+	if (!input->BufferObj->Pointer) {
+		ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+		mapped_named_bo = GL_TRUE;
+	}
+
+	radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, "%s. Vertex alignment doesn't match hw requirements.\n", __func__);
+
+	{
+		GLvoid *src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+		GLvoid *dst_ptr = ADD_POINTERS(attr->bo->ptr, attr->bo_offset);
+		int i;
+
+		for (i = 0; i < count; ++i) {
+			_mesa_memcpy(dst_ptr, src_ptr, input->StrideB);
+			src_ptr += input->StrideB;
+			dst_ptr += dst_stride;
+		}
+	}
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj);
+	}
+
+	attr->stride = dst_stride;
+}
+
+static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	struct vertex_attribute r300_attr;
+	GLenum type;
+	GLuint stride;
+
+	radeon_print(RADEON_RENDER, RADEON_TRACE, "%s\n", __func__);
+	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
+
+	if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT ||
+#if MESA_BIG_ENDIAN
+	    getTypeSize(input->Type) != 4 ||
+#endif
+	    stride < 4) {
+
+		type = GL_FLOAT;
+
+		if (input->StrideB == 0) {
+			r300_attr.stride = 0;
+		} else {
+			r300_attr.stride = sizeof(GLfloat) * input->Size;
+		}
+		r300_attr.dwords = input->Size;
+		r300_attr.is_named_bo = GL_FALSE;
+	} else {
+		type = input->Type;
+		r300_attr.dwords = (getTypeSize(type) * input->Size + 3)/ 4;
+		if (!input->BufferObj->Name) {
+
+			if (input->StrideB == 0) {
+				r300_attr.stride = 0;
+			} else {
+				r300_attr.stride = (getTypeSize(type) * input->Size + 3) & ~3;
+			}
+
+			r300_attr.is_named_bo = GL_FALSE;
+		}
+	}
+
+	r300_attr.size = input->Size;
+	r300_attr.element = attr;
+	r300_attr.dst_loc = vbuf->num_attribs;
+
+	switch (type) {
+		case GL_FLOAT:
+			switch (input->Size) {
+				case 1: r300_attr.data_type = R300_DATA_TYPE_FLOAT_1; break;
+				case 2: r300_attr.data_type = R300_DATA_TYPE_FLOAT_2; break;
+				case 3: r300_attr.data_type = R300_DATA_TYPE_FLOAT_3; break;
+				case 4: r300_attr.data_type = R300_DATA_TYPE_FLOAT_4; break;
+			}
+			r300_attr._signed = 0;
+			r300_attr.normalize = 0;
+			break;
+		case GL_SHORT:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_BYTE:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+		case GL_UNSIGNED_SHORT:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_UNSIGNED_BYTE:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			if (input->Format == GL_BGRA)
+				r300_attr.data_type = R300_DATA_TYPE_D3DCOLOR;
+			else
+				r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+
+		default:
+		case GL_DOUBLE:
+		case GL_INT:
+		case GL_UNSIGNED_INT:
+			assert(0);
+			break;
+	}
+
+	switch (input->Size) {
+		case 4:
+			r300_attr.swizzle = SWIZZLE_XYZW;
+			break;
+		case 3:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			break;
+		case 2:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+		case 1:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+	}
+
+	r300_attr.write_mask = MASK_XYZW;
+
+	vbuf->attribs[vbuf->num_attribs] = r300_attr;
+	++vbuf->num_attribs;
+}
+
+static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s\n", __func__);
+	{
+		int i, tmp;
+
+		tmp = r300->selected_vp->code.InputsRead;
+		i = 0;
+		vbuf->num_attribs = 0;
+		while (tmp) {
+			/* find first enabled bit */
+			while (!(tmp & 1)) {
+				tmp >>= 1;
+				++i;
+			}
+
+			r300TranslateAttrib(ctx, i, count, arrays[i]);
+
+			tmp >>= 1;
+			++i;
+		}
+	}
+
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback)
+		return;
+}
+
+static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *input[], int count)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	GLuint stride;
+	int ret;
+	int i, index;
+	radeon_print(RADEON_RENDER, RADEON_VERBOSE,
+			"%s: count %d num_attribs %d\n",
+			__func__, count, vbuf->num_attribs);
+
+	for (index = 0; index < vbuf->num_attribs; index++) {
+		struct radeon_aos *aos = &r300->radeon.tcl.aos[index];
+		i = vbuf->attribs[index].element;
+
+		stride = (input[i]->StrideB == 0) ? getTypeSize(input[i]->Type) * input[i]->Size : input[i]->StrideB;
+
+		if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT ||
+#if MESA_BIG_ENDIAN
+				getTypeSize(input[i]->Type) != 4 ||
+#endif
+				stride < 4) {
+
+			r300ConvertAttrib(ctx, count, input[i], &vbuf->attribs[index]);
+		} else {
+			if (input[i]->BufferObj->Name) {
+				if (stride % 4 != 0) {
+					assert(((intptr_t) input[i]->Ptr) % input[i]->StrideB == 0);
+					r300AlignDataToDword(ctx, input[i], count, &vbuf->attribs[index]);
+					vbuf->attribs[index].is_named_bo = GL_FALSE;
+				} else {
+					vbuf->attribs[index].stride = input[i]->StrideB;
+					vbuf->attribs[index].bo_offset = (intptr_t) input[i]->Ptr;
+					vbuf->attribs[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo;
+					vbuf->attribs[index].is_named_bo = GL_TRUE;
+				}
+			} else {
+
+				int size;
+				int local_count = count;
+				uint32_t *dst;
+
+				if (input[i]->StrideB == 0) {
+					size = getTypeSize(input[i]->Type) * input[i]->Size;
+					local_count = 1;
+				} else {
+					size = getTypeSize(input[i]->Type) * input[i]->Size * local_count;
+				}
+
+				radeonAllocDmaRegion(&r300->radeon, &vbuf->attribs[index].bo, &vbuf->attribs[index].bo_offset, size, 32);
+				assert(vbuf->attribs[index].bo->ptr != NULL);
+				dst = (uint32_t *)ADD_POINTERS(vbuf->attribs[index].bo->ptr, vbuf->attribs[index].bo_offset);
+				switch (vbuf->attribs[index].dwords) {
+					case 1: radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); break;
+					case 2: radeonEmitVec8(dst, input[i]->Ptr, input[i]->StrideB, local_count); break;
+					case 3: radeonEmitVec12(dst, input[i]->Ptr, input[i]->StrideB, local_count); break;
+					case 4: radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); break;
+					default: assert(0); break;
+				}
+
+			}
+		}
+
+		aos->count = vbuf->attribs[index].stride == 0 ? 1 : count;
+		aos->stride = vbuf->attribs[index].stride / sizeof(float);
+		aos->components = vbuf->attribs[index].dwords;
+		aos->bo = vbuf->attribs[index].bo;
+		aos->offset = vbuf->attribs[index].bo_offset;
+
+		if (vbuf->attribs[index].is_named_bo) {
+			radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, r300->vbuf.attribs[index].bo, RADEON_GEM_DOMAIN_GTT, 0);
+		}
+	}
+
+	r300->radeon.tcl.aos_count = vbuf->num_attribs;
+	ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, first_elem(&r300->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0);
+	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, ret);
+
+}
+
+static void r300FreeData(GLcontext *ctx)
+{
+	/* Need to zero tcl.aos[n].bo and tcl.elt_dma_bo
+	 * to prevent double unref in radeonReleaseArrays
+	 * called during context destroy
+	 */
+	radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s\n", __func__);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	{
+		int i;
+
+		for (i = 0; i < r300->vbuf.num_attribs; i++) {
+			if (!r300->vbuf.attribs[i].is_named_bo) {
+				radeon_bo_unref(r300->vbuf.attribs[i].bo);
+			}
+			r300->radeon.tcl.aos[i].bo = NULL;
+		}
+	}
+
+	{
+		if (r300->ind_buf.bo != NULL) {
+			radeon_bo_unref(r300->ind_buf.bo);
+		}
+	}
+}
+
+static GLuint r300PredictTryDrawPrimsSize(GLcontext *ctx, GLuint nr_prims)
+{
+	struct r300_context *r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	GLboolean flushed;
+	GLuint dwords;
+	GLuint state_size;
+
+	dwords = 2*CACHE_FLUSH_BUFSZ;
+	dwords += PRE_EMIT_STATE_BUFSZ;
+	dwords += (AOS_BUFSZ(vbuf->num_attribs)
+		+ SCISSORS_BUFSZ*2
+		+ FIREAOS_BUFSZ )*nr_prims;
+
+	state_size = radeonCountStateEmitSize(&r300->radeon);
+	flushed = rcommonEnsureCmdBufSpace(&r300->radeon,
+			dwords + state_size,
+			__FUNCTION__);
+	if (flushed)
+		dwords += radeonCountStateEmitSize(&r300->radeon);
+	else
+		dwords += state_size;
+
+	radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s: total prediction size is %d.\n", __FUNCTION__, dwords);
+	return dwords;
+}
+
+static GLboolean r300TryDrawPrims(GLcontext *ctx,
+					 const struct gl_client_array *arrays[],
+					 const struct _mesa_prim *prim,
+					 GLuint nr_prims,
+					 const struct _mesa_index_buffer *ib,
+					 GLuint min_index,
+					 GLuint max_index )
+{
+	struct r300_context *r300 = R300_CONTEXT(ctx);
+	GLuint i;
+
+	radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s: %u (%d-%d) cs begin at %d\n",
+				__FUNCTION__, nr_prims, min_index, max_index, r300->radeon.cmdbuf.cs->cdw );
+
+	if (ctx->NewState)
+		_mesa_update_state( ctx );
+
+	if (r300->options.hw_tcl_enabled)
+		_tnl_UpdateFixedFunctionProgram(ctx);
+
+	r300UpdateShaders(r300);
+
+	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
+
+	r300SetVertexFormat(ctx, arrays, max_index + 1);
+
+	if (r300->fallback)
+		return GL_FALSE;
+
+	r300SetupVAP(ctx, r300->selected_vp->code.InputsRead, r300->selected_vp->code.OutputsWritten);
+
+	r300UpdateShaderStates(r300);
+
+	/* ensure we have the cmd buf space in advance to cover
+	 * the state + DMA AOS pointers */
+	GLuint emit_end = r300PredictTryDrawPrimsSize(ctx, nr_prims)
+		+ r300->radeon.cmdbuf.cs->cdw;
+
+	r300SetupIndexBuffer(ctx, ib);
+
+	r300AllocDmaRegions(ctx, arrays, max_index + 1);
+
+	if (r300->fallback)
+		return GL_FALSE;
+
+	r300EmitCacheFlush(r300);
+	radeonEmitState(&r300->radeon);
+
+	for (i = 0; i < nr_prims; ++i) {
+		r300RunRenderPrimitive(ctx, prim[i].start, prim[i].start + prim[i].count, prim[i].mode);
+	}
+
+	r300EmitCacheFlush(r300);
+
+	r300FreeData(ctx);
+
+	radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s: %u (%d-%d) cs ending at %d\n",
+			__FUNCTION__, nr_prims, min_index, max_index, r300->radeon.cmdbuf.cs->cdw );
+
+	if (emit_end < r300->radeon.cmdbuf.cs->cdw)
+		WARN_ONCE("Rendering was %d commands larger than predicted size."
+				" We might overflow  command buffer.\n", r300->radeon.cmdbuf.cs->cdw - emit_end);
+
+	return GL_TRUE;
+}
+
+static void r300DrawPrims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLboolean index_bounds_valid,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+	GLboolean retval;
+
+	/* This check should get folded into just the places that
+	 * min/max index are really needed.
+	 */
+	if (!index_bounds_valid) {
+		vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+	}
+
+	if (min_index) {
+		radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT,
+				"%s: Rebasing primitives. %p nr_prims %d min_index %u max_index %u\n",
+				__func__, prim, nr_prims, min_index, max_index);
+		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
+		return;
+	}
+
+	/* Make an attempt at drawing */
+	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+
+	/* If failed run tnl pipeline - it should take care of fallbacks */
+	if (!retval)
+		_tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
+
+void r300InitDraw(GLcontext *ctx)
+{
+	struct vbo_context *vbo = vbo_context(ctx);
+
+	vbo->draw_prims = r300DrawPrims;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 28c3157427..07e6223087 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -31,6 +31,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \file
  *
  * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Maciej Cencora <m.cencora@gmail.com>
  */
 
 #include "main/glheader.h"
@@ -46,222 +47,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_context.h"
 
 #include "r300_context.h"
-#include "radeon_ioctl.h"
 #include "r300_state.h"
 #include "r300_emit.h"
 #include "r300_ioctl.h"
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
-
-#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
-    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
-    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
-    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
-    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
-    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
-#error Cannot change these!
-#endif
-
-#define DEBUG_ALL DEBUG_VERTS
-
-#if defined(USE_X86_ASM)
-#define COPY_DWORDS( dst, src, nr )					\
-do {									\
-	int __tmp;							\
-	__asm__ __volatile__( "rep ; movsl"				\
-			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
-			      : "0" (nr),				\
-			        "D" ((long)dst),			\
-			        "S" ((long)src) );			\
-} while (0)
-#else
-#define COPY_DWORDS( dst, src, nr )		\
-do {						\
-   int j;					\
-   for ( j = 0 ; j < nr ; j++ )			\
-      dst[j] = ((int *)src)[j];			\
-   dst += nr;					\
-} while (0)
-#endif
-
-static void r300EmitVec4(GLcontext * ctx, struct r300_dma_region *rvb,
-			 GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 4)
-		COPY_DWORDS(out, data, count);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out++;
-			data += stride;
-		}
-}
-
-static void r300EmitVec8(GLcontext * ctx, struct r300_dma_region *rvb,
-			 GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 8)
-		COPY_DWORDS(out, data, count * 2);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out += 2;
-			data += stride;
-		}
-}
-
-static void r300EmitVec12(GLcontext * ctx, struct r300_dma_region *rvb,
-			  GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 12)
-		COPY_DWORDS(out, data, count * 3);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out[2] = *(int *)(data + 8);
-			out += 3;
-			data += stride;
-		}
-}
-
-static void r300EmitVec16(GLcontext * ctx, struct r300_dma_region *rvb,
-			  GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 16)
-		COPY_DWORDS(out, data, count * 4);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out[2] = *(int *)(data + 8);
-			out[3] = *(int *)(data + 12);
-			out += 4;
-			data += stride;
-		}
-}
-
-static void r300EmitVec(GLcontext * ctx, struct r300_dma_region *rvb,
-			GLvoid * data, int size, int stride, int count)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (stride == 0) {
-		r300AllocDmaRegion(rmesa, rvb, size * 4, 4);
-		count = 1;
-		rvb->aos_offset = GET_START(rvb);
-		rvb->aos_stride = 0;
-	} else {
-		r300AllocDmaRegion(rmesa, rvb, size * count * 4, 4);
-		rvb->aos_offset = GET_START(rvb);
-		rvb->aos_stride = size;
-	}
-
-	switch (size) {
-	case 1:
-		r300EmitVec4(ctx, rvb, data, stride, count);
-		break;
-	case 2:
-		r300EmitVec8(ctx, rvb, data, stride, count);
-		break;
-	case 3:
-		r300EmitVec12(ctx, rvb, data, stride, count);
-		break;
-	case 4:
-		r300EmitVec16(ctx, rvb, data, stride, count);
-		break;
-	default:
-		assert(0);
-		break;
-	}
-}
-
-#define DW_SIZE(x) ((inputs[tab[(x)]] << R300_DST_VEC_LOC_SHIFT) |	\
-		    (attribptr[tab[(x)]]->size - 1) << R300_DATA_TYPE_0_SHIFT)
-
-GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr)
-{
-	GLuint i, dw;
-
-	/* type, inputs, stop bit, size */
-	for (i = 0; i < nr; i += 2) {
-		/* make sure input is valid, would lockup the gpu */
-		assert(inputs[tab[i]] != -1);
-		dw = (R300_SIGNED | DW_SIZE(i));
-		if (i + 1 == nr) {
-			dw |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
-		} else {
-			assert(inputs[tab[i + 1]] != -1);
-			dw |= (R300_SIGNED |
-			       DW_SIZE(i + 1)) << R300_DATA_TYPE_1_SHIFT;
-			if (i + 2 == nr) {
-				dw |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
-			}
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
-
-static GLuint r300VAPInputRoute1Swizzle(int swizzle[4])
-{
-	return (swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
-	    (swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
-	    (swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
-	    (swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT);
-}
-
-GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr)
-{
-	GLuint i, dw;
-
-	for (i = 0; i < nr; i += 2) {
-		dw = (r300VAPInputRoute1Swizzle(swizzle[i]) |
-		      ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-			R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE0_SHIFT;
-		if (i + 1 < nr) {
-			dw |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) |
-			       ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-				 R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
+#include "r300_render.h"
+#include "r300_swtcl.h"
 
 GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 {
@@ -272,7 +62,6 @@ GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 
 GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	GLuint i, vic_1 = 0;
 
 	if (InputsRead & (1 << VERT_ATTRIB_POS))
@@ -284,281 +73,68 @@ GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 	if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
 		vic_1 |= R300_INPUT_CNTL_COLOR;
 
-	rmesa->state.texture.tc_count = 0;
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
 		if (InputsRead & (1 << (VERT_ATTRIB_TEX0 + i))) {
-			rmesa->state.texture.tc_count++;
 			vic_1 |= R300_INPUT_CNTL_TC0 << i;
 		}
 
 	return vic_1;
 }
 
-GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint ret = 0;
 
-	if (OutputsWritten & (1 << VERT_RESULT_HPOS))
+	if (vp_writes & (1 << VERT_RESULT_HPOS))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL0))
+	if (vp_writes & (1 << VERT_RESULT_COL0))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL1))
+	if (vp_writes & (1 << VERT_RESULT_COL1))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_BFC0)
-	    || OutputsWritten & (1 << VERT_RESULT_BFC1))
-		ret |=
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
+	/* Two sided lighting works only if all 4 colors are written */
+	if (vp_writes & (1 << VERT_RESULT_BFC0) || vp_writes & (1 << VERT_RESULT_BFC1))
+		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
+			   R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_PSIZ))
+	if (vp_writes & (1 << VERT_RESULT_PSIZ))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
 
 	return ret;
 }
 
-GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint i, ret = 0, first_free_texcoord = 0;
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (OutputsWritten & (1 << (VERT_RESULT_TEX0 + i))) {
-			ret |= (4 << (3 * i));
+		if (vp_writes & (1 << (VERT_RESULT_TEX0 + i))) {
+			ret |= (4 << (3 * first_free_texcoord));
 			++first_free_texcoord;
 		}
 	}
 
-	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
-		if (first_free_texcoord > 8) {
-			fprintf(stderr, "\tout of free texcoords to write fog coord\n");
-			_mesa_exit(-1);
-		}
-		ret |= 4 << (3 * first_free_texcoord);
+	if (first_free_texcoord > 8) {
+		fprintf(stderr, "\tout of free texcoords\n");
+		_mesa_exit(-1);
 	}
 
 	return ret;
 }
 
-/* Emit vertex data to GART memory
- * Route inputs to the vertex processor
- * This function should never return R300_FALLBACK_TCL when using software tcl.
- */
-int r300EmitArrays(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-	GLuint nr;
-	GLuint count = vb->Count;
-	GLuint i;
-	GLuint InputsRead = 0, OutputsWritten = 0;
-	int *inputs = NULL;
-	int vir_inputs[VERT_ATTRIB_MAX];
-	GLint tab[VERT_ATTRIB_MAX];
-	int swizzle[VERT_ATTRIB_MAX][4];
-	struct r300_vertex_program *prog =
-	    (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-
-	if (hw_tcl_on) {
-		inputs = prog->inputs;
-		InputsRead = prog->key.InputsRead;
-		OutputsWritten = prog->key.OutputsWritten;
-	} else {
-		inputs = rmesa->state.sw_tcl_inputs;
-
-		DECLARE_RENDERINPUTS(render_inputs_bitset);
-		RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
-
-		vb->AttribPtr[VERT_ATTRIB_POS] = vb->ClipPtr;
-
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS));
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_NORMAL) == 0);
-		//assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0));
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS)) {
-			InputsRead |= 1 << VERT_ATTRIB_POS;
-			OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		}
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR0;
-			OutputsWritten |= 1 << VERT_RESULT_COL0;
-		}
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR1)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR1;
-			OutputsWritten |= 1 << VERT_RESULT_COL1;
-		}
-
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_TEX(i))) {
-				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
-				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-			}
-		}
-
-		for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-			if (InputsRead & (1 << i)) {
-				inputs[i] = nr++;
-			} else {
-				inputs[i] = -1;
-			}
-		}
-
-		/* Fixed, apply to vir0 only */
-		memcpy(vir_inputs, inputs, VERT_ATTRIB_MAX * sizeof(int));
-		inputs = vir_inputs;
-		if (InputsRead & VERT_ATTRIB_POS)
-			inputs[VERT_ATTRIB_POS] = 0;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
-			inputs[VERT_ATTRIB_COLOR0] = 2;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
-			inputs[VERT_ATTRIB_COLOR1] = 3;
-		for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
-			if (InputsRead & (1 << i))
-				inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
-
-		RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
-	}
-
-	assert(InputsRead);
-	assert(OutputsWritten);
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			tab[nr++] = i;
-		}
-	}
-
-	if (nr > R300_MAX_AOS_ARRAYS) {
-		return R300_FALLBACK_TCL;
-	}
-
-	for (i = 0; i < nr; i++) {
-		int ci, fix, found = 0;
-
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
-
-		for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
-			swizzle[i][ci] = ci;
-		}
-
-		if (r300IsGartMemory(rmesa, vb->AttribPtr[tab[i]]->data, 4)) {
-			if (vb->AttribPtr[tab[i]]->stride % 4) {
-				return R300_FALLBACK_TCL;
-			}
-			rmesa->state.aos[i].address = (void *)(vb->AttribPtr[tab[i]]->data);
-			rmesa->state.aos[i].start = 0;
-			rmesa->state.aos[i].aos_offset = r300GartOffsetFromVirtual(rmesa, vb->AttribPtr[tab[i]]->data);
-			rmesa->state.aos[i].aos_stride = vb->AttribPtr[tab[i]]->stride / 4;
-			rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
-		} else {
-			r300EmitVec(ctx, &rmesa->state.aos[i],
-				    vb->AttribPtr[tab[i]]->data,
-				    vb->AttribPtr[tab[i]]->size,
-				    vb->AttribPtr[tab[i]]->stride, count);
-		}
-
-		rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
-
-		for (fix = 0; fix <= 4 - vb->AttribPtr[tab[i]]->size; fix++) {
-			if ((rmesa->state.aos[i].aos_offset - _mesa_sizeof_type(GL_FLOAT) * fix) % 4) {
-				continue;
-			}
-			found = 1;
-			break;
-		}
-
-		if (found) {
-			if (fix > 0) {
-				WARN_ONCE("Feeling lucky?\n");
-			}
-			rmesa->state.aos[i].aos_offset -= _mesa_sizeof_type(GL_FLOAT) * fix;
-			for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
-				swizzle[i][ci] += fix;
-			}
-		} else {
-			WARN_ONCE
-			    ("Cannot handle offset %x with stride %d, comp %d\n",
-			     rmesa->state.aos[i].aos_offset,
-			     rmesa->state.aos[i].aos_stride,
-			     vb->AttribPtr[tab[i]]->size);
-			return R300_FALLBACK_TCL;
-		}
-	}
-
-	/* Setup INPUT_ROUTE. */
-	R300_STATECHANGE(rmesa, vir[0]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
-	    r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-			       vb->AttribPtr, inputs, tab, nr);
-	R300_STATECHANGE(rmesa, vir[1]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
-	    r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-			       nr);
-
-	/* Setup INPUT_CNTL. */
-	R300_STATECHANGE(rmesa, vic);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-
-	/* Setup OUTPUT_VTX_FMT. */
-	R300_STATECHANGE(rmesa, vof);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] =
-	    r300VAPOutputCntl0(ctx, OutputsWritten);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] =
-	    r300VAPOutputCntl1(ctx, OutputsWritten);
-
-	rmesa->state.aos_count = nr;
-
-	return R300_FALLBACK_NONE;
-}
-
-#ifdef USER_BUFFERS
-void r300UseArrays(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-
-	if (rmesa->state.elt_dma.buf)
-		r300_mem_use(rmesa, rmesa->state.elt_dma.buf->id);
-
-	for (i = 0; i < rmesa->state.aos_count; i++) {
-		if (rmesa->state.aos[i].buf)
-			r300_mem_use(rmesa, rmesa->state.aos[i].buf->id);
-	}
-}
-#endif
-
-void r300ReleaseArrays(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-
-	r300ReleaseDmaRegion(rmesa, &rmesa->state.elt_dma, __FUNCTION__);
-	for (i = 0; i < rmesa->state.aos_count; i++) {
-		r300ReleaseDmaRegion(rmesa, &rmesa->state.aos[i], __FUNCTION__);
-	}
-}
-
 void r300EmitCacheFlush(r300ContextPtr rmesa)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	reg_start(R300_RB3D_DSTCACHE_CTLSTAT, 0);
-	e32(R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
-	    R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
-
-	reg_start(R300_ZB_ZCACHE_CTLSTAT, 0);
-	e32(R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
-	    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+	BATCH_LOCALS(&rmesa->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(4);
+	OUT_BATCH_REGVAL(R300_RB3D_DSTCACHE_CTLSTAT,
+		R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
+		R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+	OUT_BATCH_REGVAL(R300_ZB_ZCACHE_CTLSTAT,
+		R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+		R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+	END_BATCH();
+	COMMIT_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h
index 89d738339f..8e57e354d1 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -44,28 +44,31 @@
 #include "r300_cmdbuf.h"
 #include "radeon_reg.h"
 
-/* TODO: move these defines (and the ones from DRM) into r300_reg.h and sync up
- * with DRM */
-#define CP_PACKET0(reg, n)	(RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2))
-#define CP_PACKET3( pkt, n )						\
-	(RADEON_CP_PACKET3 | (pkt) | ((n) << 16))
-
-static INLINE uint32_t cmdpacket0(int reg, int count)
+static INLINE uint32_t cmdpacket0(struct radeon_screen *rscrn,
+                                  int reg, int count)
 {
-	drm_r300_cmd_header_t cmd;
-
-	cmd.packet0.cmd_type = R300_CMD_PACKET0;
-	cmd.packet0.count = count;
-	cmd.packet0.reghi = ((unsigned int)reg & 0xFF00) >> 8;
-	cmd.packet0.reglo = ((unsigned int)reg & 0x00FF);
-
-	return cmd.u;
+    if (!rscrn->kernel_mm) {
+	    drm_r300_cmd_header_t cmd;
+
+	cmd.u = 0;
+    	cmd.packet0.cmd_type = R300_CMD_PACKET0;
+	    cmd.packet0.count = count;
+    	cmd.packet0.reghi = ((unsigned int)reg & 0xFF00) >> 8;
+	    cmd.packet0.reglo = ((unsigned int)reg & 0x00FF);
+
+    	return cmd.u;
+    }
+    if (count) {
+        return CP_PACKET0(reg, count - 1);
+    }
+    return CP_PACKET2;
 }
 
-static INLINE uint32_t cmdvpu(int addr, int count)
+static INLINE uint32_t cmdvpu(struct radeon_screen *rscrn, int addr, int count)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.vpu.cmd_type = R300_CMD_VPU;
 	cmd.vpu.count = count;
 	cmd.vpu.adrhi = ((unsigned int)addr & 0xFF00) >> 8;
@@ -74,10 +77,12 @@ static INLINE uint32_t cmdvpu(int addr, int count)
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdr500fp(int addr, int count, int type, int clamp)
+static INLINE uint32_t cmdr500fp(struct radeon_screen *rscrn,
+                                 int addr, int count, int type, int clamp)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.r500fp.cmd_type = R300_CMD_R500FP;
 	cmd.r500fp.count = count;
 	cmd.r500fp.adrhi_flags = ((unsigned int)addr & 0x100) >> 8;
@@ -88,181 +93,137 @@ static INLINE uint32_t cmdr500fp(int addr, int count, int type, int clamp)
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdpacket3(int packet)
+static INLINE uint32_t cmdpacket3(struct radeon_screen *rscrn, int packet)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.packet3.cmd_type = R300_CMD_PACKET3;
 	cmd.packet3.packet = packet;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdcpdelay(unsigned short count)
+static INLINE uint32_t cmdcpdelay(struct radeon_screen *rscrn,
+                                  unsigned short count)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
+
 	cmd.delay.cmd_type = R300_CMD_CP_DELAY;
 	cmd.delay.count = count;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdwait(unsigned char flags)
+static INLINE uint32_t cmdwait(struct radeon_screen *rscrn,
+                               unsigned char flags)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.wait.cmd_type = R300_CMD_WAIT;
 	cmd.wait.flags = flags;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdpacify(void)
+static INLINE uint32_t cmdpacify(struct radeon_screen *rscrn)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.header.cmd_type = R300_CMD_END3D;
 
 	return cmd.u;
 }
 
 /**
- * Prepare to write a register value to register at address reg.
- * If num_extra > 0 then the following extra values are written
- * to registers with address +4, +8 and so on..
- */
-#define reg_start(reg, num_extra)					\
-	do {								\
-		int _n;							\
-		_n=(num_extra);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+2),				\
-					__FUNCTION__);			\
-		cmd_reserved=_n+2;					\
-		cmd_written=1;						\
-		cmd[0].i=cmdpacket0((reg), _n+1);			\
-	} while (0);
-
-/**
- * Emit GLuint freestyle
+ * Write the header of a packet3 to the command buffer.
+ * Outputs 2 dwords and expects (num_extra+1) additional dwords afterwards.
  */
-#define e32(dword)							\
-	do {								\
-		if(cmd_written<cmd_reserved) {				\
-			cmd[cmd_written].i=(dword);			\
-			cmd_written++;					\
-		} else {						\
-			fprintf(stderr,					\
-				"e32 but no previous packet "		\
-				"declaration.\n"			\
-				"Aborting! in %s::%s at line %d, "	\
-				"cmd_written=%d cmd_reserved=%d\n",	\
-				__FILE__, __FUNCTION__, __LINE__,	\
-				cmd_written, cmd_reserved);		\
-			_mesa_exit(-1);					\
-		}							\
+#define OUT_BATCH_PACKET3(packet, num_extra) do {\
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		\
+    	OUT_BATCH(cmdpacket3(b_l_rmesa->radeonScreen,\
+                  R300_CMD_PACKET3_RAW)); \
+    } else b_l_rmesa->cmdbuf.cs->section_cdw++;\
+	OUT_BATCH(CP_PACKET3((packet), (num_extra))); \
 	} while(0)
 
-#define	efloat(f) e32(r300PackFloat32(f))
-
-#define vsf_start_fragment(dest, length)				\
-	do {								\
-		int _n;							\
-		_n = (length);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+1),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+2;					\
-		cmd_written =1;						\
-		cmd[0].i = cmdvpu((dest), _n/4);			\
-	} while (0);
-
-#define r500fp_start_fragment(dest, length)				\
-	do {								\
-		int _n;							\
-		_n = (length);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+1),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+1;					\
-		cmd_written =1;						\
-		cmd[0].i = cmdr500fp((dest), _n/6, 0, 0);		\
-	} while (0);
-
-#define start_packet3(packet, count)					\
-	{								\
-		int _n;							\
-		GLuint _p;						\
-		_n = (count);						\
-		_p = (packet);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+3),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+3;					\
-		cmd_written = 2;					\
-		if(_n > 0x3fff) {					\
-			fprintf(stderr,"Too big packet3 %08x: cannot "	\
-				"store %d dwords\n",			\
-				_p, _n);				\
-			_mesa_exit(-1);					\
-		}							\
-		cmd[0].i = cmdpacket3(R300_CMD_PACKET3_RAW);		\
-		cmd[1].i = _p | ((_n & 0x3fff)<<16);			\
-	}
-
 /**
  * Must be sent to switch to 2d commands
  */
-void static INLINE end_3d(r300ContextPtr rmesa)
+void static INLINE end_3d(radeonContextPtr radeon)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(radeon);
 
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].header.cmd_type = R300_CMD_END3D;
+	if (!radeon->radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdpacify(radeon->radeonScreen));
+		END_BATCH();
+	}
 }
 
 void static INLINE cp_delay(r300ContextPtr rmesa, unsigned short count)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(&rmesa->radeon);
 
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].i = cmdcpdelay(count);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdcpdelay(rmesa->radeon.radeonScreen, count));
+		END_BATCH();
+	}
 }
 
-void static INLINE cp_wait(r300ContextPtr rmesa, unsigned char flags)
+void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].i = cmdwait(flags);
+	BATCH_LOCALS(radeon);
+	uint32_t wait_until;
+
+	if (!radeon->radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdwait(radeon->radeonScreen, flags));
+		END_BATCH();
+	} else {
+		switch(flags) {
+		case R300_WAIT_2D:
+			wait_until = (1 << 14);
+			break;
+		case R300_WAIT_3D:
+			wait_until = (1 << 15);
+			break;
+		case R300_NEW_WAIT_2D_3D:
+			wait_until = (1 << 14) | (1 << 15);
+			break;
+		case R300_NEW_WAIT_2D_2D_CLEAN:
+			wait_until = (1 << 14) | (1 << 16) | (1 << 18);
+			break;
+		case R300_NEW_WAIT_3D_3D_CLEAN:
+			wait_until = (1 << 15) | (1 << 17) | (1 << 18);
+			break;
+		case R300_NEW_WAIT_2D_2D_CLEAN_3D_3D_CLEAN:
+			wait_until  = (1 << 14) | (1 << 16) | (1 << 18);
+			wait_until |= (1 << 15) | (1 << 17) | (1 << 18);
+			break;
+		default:
+			return;
+		}
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		OUT_BATCH(CP_PACKET0(RADEON_WAIT_UNTIL, 0));
+		OUT_BATCH(wait_until);
+		END_BATCH();
+	}
 }
 
-extern int r300EmitArrays(GLcontext * ctx);
-
-#ifdef USER_BUFFERS
-void r300UseArrays(GLcontext * ctx);
-#endif
-
-extern void r300ReleaseArrays(GLcontext * ctx);
 extern int r300PrimitiveType(r300ContextPtr rmesa, int prim);
 extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim);
 
 extern void r300EmitCacheFlush(r300ContextPtr rmesa);
 
-extern GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr);
-extern GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr);
 extern GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead);
 extern GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead);
-extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten);
-extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten);
+extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes);
+extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
deleted file mode 100644
index 873cde4414..0000000000
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ /dev/null
@@ -1,699 +0,0 @@
-/*
- * Copyright (C) 2005 Ben Skeggs.
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-/**
- * \file
- *
- * Fragment program compiler. Perform transformations on the intermediate
- * representation until the program is in a form where we can translate
- * it more or less directly into machine-readable form.
- *
- * \author Ben Skeggs <darktama@iinet.net.au>
- * \author Jerome Glisse <j.glisse@gmail.com>
- */
-
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "shader/prog_instruction.h"
-#include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-
-#include "r300_context.h"
-#include "r300_fragprog.h"
-#include "r300_fragprog_swizzle.h"
-#include "r300_state.h"
-
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
-
-static void reset_srcreg(struct prog_src_register* reg)
-{
-	_mesa_bzero(reg, sizeof(*reg));
-	reg->Swizzle = SWIZZLE_NOOP;
-}
-
-static struct prog_src_register shadow_ambient(struct gl_program *program, int tmu)
-{
-	gl_state_index fail_value_tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_SHADOW_AMBIENT, 0, 0, 0
-	};
-	struct prog_src_register reg = { 0, };
-
-	fail_value_tokens[2] = tmu;
-	reg.File = PROGRAM_STATE_VAR;
-	reg.Index = _mesa_add_state_reference(program->Parameters, fail_value_tokens);
-	reg.Swizzle = SWIZZLE_WWWW;
-	return reg;
-}
-
-/**
- * Transform TEX, TXP, TXB, and KIL instructions in the following way:
- *  - premultiply texture coordinates for RECT
- *  - extract operand swizzles
- *  - introduce a temporary register when write masks are needed
- *
- * \todo If/when r5xx uses the radeon_program architecture, this can probably
- * be reused.
- */
-static GLboolean transform_TEX(
-	struct radeon_transform_context *t,
-	struct prog_instruction* orig_inst, void* data)
-{
-	struct r300_fragment_program_compiler *compiler =
-		(struct r300_fragment_program_compiler*)data;
-	struct prog_instruction inst = *orig_inst;
-	struct prog_instruction* tgt;
-	GLboolean destredirect = GL_FALSE;
-
-	if (inst.Opcode != OPCODE_TEX &&
-	    inst.Opcode != OPCODE_TXB &&
-	    inst.Opcode != OPCODE_TXP &&
-	    inst.Opcode != OPCODE_KIL)
-		return GL_FALSE;
-
-	if (inst.Opcode != OPCODE_KIL &&
-	    t->Program->ShadowSamplers & (1 << inst.TexSrcUnit)) {
-		GLuint comparefunc = GL_NEVER + compiler->fp->state.unit[inst.TexSrcUnit].texture_compare_func;
-
-		if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) {
-			tgt = radeonAppendInstructions(t->Program, 1);
-
-			tgt->Opcode = OPCODE_MOV;
-			tgt->DstReg = inst.DstReg;
-			if (comparefunc == GL_ALWAYS) {
-				tgt->SrcReg[0].File = PROGRAM_BUILTIN;
-				tgt->SrcReg[0].Swizzle = SWIZZLE_1111;
-			} else {
-				tgt->SrcReg[0] = shadow_ambient(t->Program, inst.TexSrcUnit);
-			}
-			return GL_TRUE;
-		}
-
-		inst.DstReg.File = PROGRAM_TEMPORARY;
-		inst.DstReg.Index = radeonFindFreeTemporary(t);
-		inst.DstReg.WriteMask = WRITEMASK_XYZW;
-	}
-
-
-	/* Hardware uses [0..1]x[0..1] range for rectangle textures
-	 * instead of [0..Width]x[0..Height].
-	 * Add a scaling instruction.
-	 */
-	if (inst.Opcode != OPCODE_KIL && inst.TexSrcTarget == TEXTURE_RECT_INDEX) {
-		gl_state_index tokens[STATE_LENGTH] = {
-			STATE_INTERNAL, STATE_R300_TEXRECT_FACTOR, 0, 0,
-			0
-		};
-
-		int tempreg = radeonFindFreeTemporary(t);
-		int factor_index;
-
-		tokens[2] = inst.TexSrcUnit;
-		factor_index = _mesa_add_state_reference(t->Program->Parameters, tokens);
-
-		tgt = radeonAppendInstructions(t->Program, 1);
-
-		tgt->Opcode = OPCODE_MUL;
-		tgt->DstReg.File = PROGRAM_TEMPORARY;
-		tgt->DstReg.Index = tempreg;
-		tgt->SrcReg[0] = inst.SrcReg[0];
-		tgt->SrcReg[1].File = PROGRAM_STATE_VAR;
-		tgt->SrcReg[1].Index = factor_index;
-
-		reset_srcreg(&inst.SrcReg[0]);
-		inst.SrcReg[0].File = PROGRAM_TEMPORARY;
-		inst.SrcReg[0].Index = tempreg;
-	}
-
-	if (inst.Opcode != OPCODE_KIL) {
-		if (inst.DstReg.File != PROGRAM_TEMPORARY ||
-		    inst.DstReg.WriteMask != WRITEMASK_XYZW) {
-			int tempreg = radeonFindFreeTemporary(t);
-
-			inst.DstReg.File = PROGRAM_TEMPORARY;
-			inst.DstReg.Index = tempreg;
-			inst.DstReg.WriteMask = WRITEMASK_XYZW;
-			destredirect = GL_TRUE;
-		}
-	}
-
-	if (inst.SrcReg[0].File != PROGRAM_TEMPORARY && inst.SrcReg[0].File != PROGRAM_INPUT) {
-		int tmpreg = radeonFindFreeTemporary(t);
-		tgt = radeonAppendInstructions(t->Program, 1);
-		tgt->Opcode = OPCODE_MOV;
-		tgt->DstReg.File = PROGRAM_TEMPORARY;
-		tgt->DstReg.Index = tmpreg;
-		tgt->SrcReg[0] = inst.SrcReg[0];
-
-		reset_srcreg(&inst.SrcReg[0]);
-		inst.SrcReg[0].File = PROGRAM_TEMPORARY;
-		inst.SrcReg[0].Index = tmpreg;
-	}
-	
-	tgt = radeonAppendInstructions(t->Program, 1);
-	_mesa_copy_instructions(tgt, &inst, 1);
-
-	if (inst.Opcode != OPCODE_KIL &&
-	    t->Program->ShadowSamplers & (1 << inst.TexSrcUnit)) {
-		GLuint comparefunc = GL_NEVER + compiler->fp->state.unit[inst.TexSrcUnit].texture_compare_func;
-		GLuint depthmode = compiler->fp->state.unit[inst.TexSrcUnit].depth_texture_mode;
-		int rcptemp = radeonFindFreeTemporary(t);
-		int pass, fail;
-
-		tgt = radeonAppendInstructions(t->Program, 3);
-
-		tgt[0].Opcode = OPCODE_RCP;
-		tgt[0].DstReg.File = PROGRAM_TEMPORARY;
-		tgt[0].DstReg.Index = rcptemp;
-		tgt[0].DstReg.WriteMask = WRITEMASK_W;
-		tgt[0].SrcReg[0] = inst.SrcReg[0];
-		tgt[0].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-
-		tgt[1].Opcode = OPCODE_MAD;
-		tgt[1].DstReg = inst.DstReg;
-		tgt[1].DstReg.WriteMask = orig_inst->DstReg.WriteMask;
-		tgt[1].SrcReg[0] = inst.SrcReg[0];
-		tgt[1].SrcReg[0].Swizzle = SWIZZLE_ZZZZ;
-		tgt[1].SrcReg[1].File = PROGRAM_TEMPORARY;
-		tgt[1].SrcReg[1].Index = rcptemp;
-		tgt[1].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-		tgt[1].SrcReg[2].File = PROGRAM_TEMPORARY;
-		tgt[1].SrcReg[2].Index = inst.DstReg.Index;
-		if (depthmode == 0) /* GL_LUMINANCE */
-			tgt[1].SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z);
-		else if (depthmode == 2) /* GL_ALPHA */
-			tgt[1].SrcReg[2].Swizzle = SWIZZLE_WWWW;
-
-		/* Recall that SrcReg[0] is tex, SrcReg[2] is r and:
-		 *   r  < tex  <=>      -tex+r < 0
-		 *   r >= tex  <=> not (-tex+r < 0 */
-		if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
-			tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW;
-		else
-			tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW;
-
-		tgt[2].Opcode = OPCODE_CMP;
-		tgt[2].DstReg = orig_inst->DstReg;
-		tgt[2].SrcReg[0].File = PROGRAM_TEMPORARY;
-		tgt[2].SrcReg[0].Index = tgt[1].DstReg.Index;
-
-		if (comparefunc == GL_LESS || comparefunc == GL_GREATER) {
-			pass = 1;
-			fail = 2;
-		} else {
-			pass = 2;
-			fail = 1;
-		}
-
-		tgt[2].SrcReg[pass].File = PROGRAM_BUILTIN;
-		tgt[2].SrcReg[pass].Swizzle = SWIZZLE_1111;
-		tgt[2].SrcReg[fail] = shadow_ambient(t->Program, inst.TexSrcUnit);
-	} else if (destredirect) {
-		tgt = radeonAppendInstructions(t->Program, 1);
-
-		tgt->Opcode = OPCODE_MOV;
-		tgt->DstReg = orig_inst->DstReg;
-		tgt->SrcReg[0].File = PROGRAM_TEMPORARY;
-		tgt->SrcReg[0].Index = inst.DstReg.Index;
-	}
-
-	return GL_TRUE;
-}
-
-
-static void update_params(r300ContextPtr r300, struct r300_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-
-void r300TranslateFragmentShader(r300ContextPtr r300,
-				 struct r300_fragment_program *fp)
-{
-	struct r300_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r300_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformTrigSimple, 0 }
-		};
-		radeonLocalTransform(
-			r300->radeon.glCtx,
-			compiler.program,
-			3, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: After native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
-			.BuildSwizzle = &r300FPBuildSwizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		if (!r300FragmentProgramEmit(&compiler))
-			fp->error = GL_TRUE;
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, NULL);
-
-		if (!fp->error)
-			fp->translated = GL_TRUE;
-		if (fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
-			r300FragmentProgramDump(fp, &fp->code);
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-	}
-
-	update_params(r300, fp);
-}
-
-/* just some random things... */
-void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code)
-{
-	int n, i, j;
-	static int pc = 0;
-
-	fprintf(stderr, "pc=%d*************************************\n", pc++);
-
-	fprintf(stderr, "Hardware program\n");
-	fprintf(stderr, "----------------\n");
-
-	for (n = 0; n < (code->cur_node + 1); n++) {
-		fprintf(stderr, "NODE %d: alu_offset: %d, tex_offset: %d, "
-			"alu_end: %d, tex_end: %d, flags: %08x\n", n,
-			code->node[n].alu_offset,
-			code->node[n].tex_offset,
-			code->node[n].alu_end, code->node[n].tex_end,
-			code->node[n].flags);
-
-		if (n > 0 || code->first_node_has_tex) {
-			fprintf(stderr, "  TEX:\n");
-			for (i = code->node[n].tex_offset;
-			     i <= code->node[n].tex_offset + code->node[n].tex_end;
-			     ++i) {
-				const char *instr;
-
-				switch ((code->tex.
-					 inst[i] >> R300_TEX_INST_SHIFT) &
-					15) {
-				case R300_TEX_OP_LD:
-					instr = "TEX";
-					break;
-				case R300_TEX_OP_KIL:
-					instr = "KIL";
-					break;
-				case R300_TEX_OP_TXP:
-					instr = "TXP";
-					break;
-				case R300_TEX_OP_TXB:
-					instr = "TXB";
-					break;
-				default:
-					instr = "UNKNOWN";
-				}
-
-				fprintf(stderr,
-					"    %s t%i, %c%i, texture[%i]   (%08x)\n",
-					instr,
-					(code->tex.
-					 inst[i] >> R300_DST_ADDR_SHIFT) & 31,
-					't',
-					(code->tex.
-					 inst[i] >> R300_SRC_ADDR_SHIFT) & 31,
-					(code->tex.
-					 inst[i] & R300_TEX_ID_MASK) >>
-					R300_TEX_ID_SHIFT,
-					code->tex.inst[i]);
-			}
-		}
-
-		for (i = code->node[n].alu_offset;
-		     i <= code->node[n].alu_offset + code->node[n].alu_end; ++i) {
-			char srcc[3][10], dstc[20];
-			char srca[3][10], dsta[20];
-			char argc[3][20];
-			char arga[3][20];
-			char flags[5], tmp[10];
-
-			for (j = 0; j < 3; ++j) {
-				int regc = code->alu.inst[i].inst1 >> (j * 6);
-				int rega = code->alu.inst[i].inst3 >> (j * 6);
-
-				sprintf(srcc[j], "%c%i",
-					(regc & 32) ? 'c' : 't', regc & 31);
-				sprintf(srca[j], "%c%i",
-					(rega & 32) ? 'c' : 't', rega & 31);
-			}
-
-			dstc[0] = 0;
-			sprintf(flags, "%s%s%s",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_REG_X) ? "x" : "",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_REG_Y) ? "y" : "",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_REG_Z) ? "z" : "");
-			if (flags[0] != 0) {
-				sprintf(dstc, "t%i.%s ",
-					(code->alu.inst[i].
-					 inst1 >> R300_ALU_DSTC_SHIFT) & 31,
-					flags);
-			}
-			sprintf(flags, "%s%s%s",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_OUTPUT_X) ? "x" : "",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_OUTPUT_Y) ? "y" : "",
-				(code->alu.inst[i].
-				 inst1 & R300_ALU_DSTC_OUTPUT_Z) ? "z" : "");
-			if (flags[0] != 0) {
-				sprintf(tmp, "o%i.%s",
-					(code->alu.inst[i].
-					 inst1 >> R300_ALU_DSTC_SHIFT) & 31,
-					flags);
-				strcat(dstc, tmp);
-			}
-
-			dsta[0] = 0;
-			if (code->alu.inst[i].inst3 & R300_ALU_DSTA_REG) {
-				sprintf(dsta, "t%i.w ",
-					(code->alu.inst[i].
-					 inst3 >> R300_ALU_DSTA_SHIFT) & 31);
-			}
-			if (code->alu.inst[i].inst3 & R300_ALU_DSTA_OUTPUT) {
-				sprintf(tmp, "o%i.w ",
-					(code->alu.inst[i].
-					 inst3 >> R300_ALU_DSTA_SHIFT) & 31);
-				strcat(dsta, tmp);
-			}
-			if (code->alu.inst[i].inst3 & R300_ALU_DSTA_DEPTH) {
-				strcat(dsta, "Z");
-			}
-
-			fprintf(stderr,
-				"%3i: xyz: %3s %3s %3s -> %-20s (%08x)\n"
-				"       w: %3s %3s %3s -> %-20s (%08x)\n", i,
-				srcc[0], srcc[1], srcc[2], dstc,
-				code->alu.inst[i].inst1, srca[0], srca[1],
-				srca[2], dsta, code->alu.inst[i].inst3);
-
-			for (j = 0; j < 3; ++j) {
-				int regc = code->alu.inst[i].inst0 >> (j * 7);
-				int rega = code->alu.inst[i].inst2 >> (j * 7);
-				int d;
-				char buf[20];
-
-				d = regc & 31;
-				if (d < 12) {
-					switch (d % 4) {
-					case R300_ALU_ARGC_SRC0C_XYZ:
-						sprintf(buf, "%s.xyz",
-							srcc[d / 4]);
-						break;
-					case R300_ALU_ARGC_SRC0C_XXX:
-						sprintf(buf, "%s.xxx",
-							srcc[d / 4]);
-						break;
-					case R300_ALU_ARGC_SRC0C_YYY:
-						sprintf(buf, "%s.yyy",
-							srcc[d / 4]);
-						break;
-					case R300_ALU_ARGC_SRC0C_ZZZ:
-						sprintf(buf, "%s.zzz",
-							srcc[d / 4]);
-						break;
-					}
-				} else if (d < 15) {
-					sprintf(buf, "%s.www", srca[d - 12]);
-				} else if (d == 20) {
-					sprintf(buf, "0.0");
-				} else if (d == 21) {
-					sprintf(buf, "1.0");
-				} else if (d == 22) {
-					sprintf(buf, "0.5");
-				} else if (d >= 23 && d < 32) {
-					d -= 23;
-					switch (d / 3) {
-					case 0:
-						sprintf(buf, "%s.yzx",
-							srcc[d % 3]);
-						break;
-					case 1:
-						sprintf(buf, "%s.zxy",
-							srcc[d % 3]);
-						break;
-					case 2:
-						sprintf(buf, "%s.Wzy",
-							srcc[d % 3]);
-						break;
-					}
-				} else {
-					sprintf(buf, "%i", d);
-				}
-
-				sprintf(argc[j], "%s%s%s%s",
-					(regc & 32) ? "-" : "",
-					(regc & 64) ? "|" : "",
-					buf, (regc & 64) ? "|" : "");
-
-				d = rega & 31;
-				if (d < 9) {
-					sprintf(buf, "%s.%c", srcc[d / 3],
-						'x' + (char)(d % 3));
-				} else if (d < 12) {
-					sprintf(buf, "%s.w", srca[d - 9]);
-				} else if (d == 16) {
-					sprintf(buf, "0.0");
-				} else if (d == 17) {
-					sprintf(buf, "1.0");
-				} else if (d == 18) {
-					sprintf(buf, "0.5");
-				} else {
-					sprintf(buf, "%i", d);
-				}
-
-				sprintf(arga[j], "%s%s%s%s",
-					(rega & 32) ? "-" : "",
-					(rega & 64) ? "|" : "",
-					buf, (rega & 64) ? "|" : "");
-			}
-
-			fprintf(stderr, "     xyz: %8s %8s %8s    op: %08x\n"
-				"       w: %8s %8s %8s    op: %08x\n",
-				argc[0], argc[1], argc[2],
-				code->alu.inst[i].inst0, arga[0], arga[1],
-				arga[2], code->alu.inst[i].inst2);
-		}
-	}
-}
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.h b/src/mesa/drivers/dri/r300/r300_fragprog.h
deleted file mode 100644
index 94fb554fb3..0000000000
--- a/src/mesa/drivers/dri/r300/r300_fragprog.h
+++ /dev/null
@@ -1,132 +0,0 @@
-/*
- * Copyright (C) 2005 Ben Skeggs.
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-/*
- * Authors:
- *   Ben Skeggs <darktama@iinet.net.au>
- *   Jerome Glisse <j.glisse@gmail.com>
- */
-#ifndef __R300_FRAGPROG_H_
-#define __R300_FRAGPROG_H_
-
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "shader/program.h"
-#include "shader/prog_instruction.h"
-
-#include "r300_context.h"
-#include "radeon_program.h"
-
-#define DRI_CONF_FP_OPTIMIZATION_SPEED   0
-#define DRI_CONF_FP_OPTIMIZATION_QUALITY 1
-
-#if 1
-
-/**
- * Fragment program helper macros
- */
-
-/* Produce unshifted source selectors */
-#define FP_TMP(idx) (idx)
-#define FP_CONST(idx) ((idx) | (1 << 5))
-
-/* Produce source/dest selector dword */
-#define FP_SELC_MASK_NO		0
-#define FP_SELC_MASK_X		1
-#define FP_SELC_MASK_Y		2
-#define FP_SELC_MASK_XY		3
-#define FP_SELC_MASK_Z		4
-#define FP_SELC_MASK_XZ		5
-#define FP_SELC_MASK_YZ		6
-#define FP_SELC_MASK_XYZ	7
-
-#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \
-	(((destidx) << R300_ALU_DSTC_SHIFT) |		\
-	 (FP_SELC_MASK_##regmask << 23) |		\
-	 (FP_SELC_MASK_##outmask << 26) |		\
-	 ((src0) << R300_ALU_SRC0C_SHIFT) |		\
-	 ((src1) << R300_ALU_SRC1C_SHIFT) |		\
-	 ((src2) << R300_ALU_SRC2C_SHIFT))
-
-#define FP_SELA_MASK_NO		0
-#define FP_SELA_MASK_W		1
-
-#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \
-	(((destidx) << R300_ALU_DSTA_SHIFT) |		\
-	 (FP_SELA_MASK_##regmask << 23) |		\
-	 (FP_SELA_MASK_##outmask << 24) |		\
-	 ((src0) << R300_ALU_SRC0A_SHIFT) |		\
-	 ((src1) << R300_ALU_SRC1A_SHIFT) |		\
-	 ((src2) << R300_ALU_SRC2A_SHIFT))
-
-/* Produce unshifted argument selectors */
-#define FP_ARGC(source)	R300_ALU_ARGC_##source
-#define FP_ARGA(source) R300_ALU_ARGA_##source
-#define FP_ABS(arg) ((arg) | (1 << 6))
-#define FP_NEG(arg) ((arg) ^ (1 << 5))
-
-/* Produce instruction dword */
-#define FP_INSTRC(opcode,arg0,arg1,arg2) \
-	(R300_ALU_OUTC_##opcode | 		\
-	((arg0) << R300_ALU_ARG0C_SHIFT) |	\
-	((arg1) << R300_ALU_ARG1C_SHIFT) |	\
-	((arg2) << R300_ALU_ARG2C_SHIFT))
-
-#define FP_INSTRA(opcode,arg0,arg1,arg2) \
-	(R300_ALU_OUTA_##opcode | 		\
-	((arg0) << R300_ALU_ARG0A_SHIFT) |	\
-	((arg1) << R300_ALU_ARG1A_SHIFT) |	\
-	((arg2) << R300_ALU_ARG2A_SHIFT))
-
-#endif
-
-struct r300_fragment_program;
-
-extern void r300TranslateFragmentShader(r300ContextPtr r300,
-					struct r300_fragment_program *fp);
-
-
-/**
- * Used internally by the r300 fragment program code to store compile-time
- * only data.
- */
-struct r300_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r300_fragment_program *fp;
-	struct r300_fragment_program_code *code;
-	struct gl_program *program;
-};
-
-extern GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler);
-
-
-extern void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code);
-
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
new file mode 100644
index 0000000000..469c278b51
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -0,0 +1,263 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/**
+ * \file
+ *
+ * Fragment program compiler. Perform transformations on the intermediate
+ * representation until the program is in a form where we can translate
+ * it more or less directly into machine-readable form.
+ *
+ * \author Ben Skeggs <darktama@iinet.net.au>
+ * \author Jerome Glisse <j.glisse@gmail.com>
+ */
+
+#include "r300_fragprog_common.h"
+
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "compiler/radeon_compiler.h"
+
+#include "r300_state.h"
+
+
+static GLuint build_dtm(GLuint depthmode)
+{
+	switch(depthmode) {
+	default:
+	case GL_LUMINANCE: return 0;
+	case GL_INTENSITY: return 1;
+	case GL_ALPHA: return 2;
+	}
+}
+
+static GLuint build_func(GLuint comparefunc)
+{
+	return comparefunc - GL_NEVER;
+}
+
+/**
+ * Collect all external state that is relevant for compiling the given
+ * fragment program.
+ */
+static void build_state(
+	r300ContextPtr r300,
+	struct gl_fragment_program *fp,
+	struct r300_fragment_program_external_state *state)
+{
+	int unit;
+
+	_mesa_bzero(state, sizeof(*state));
+
+	for(unit = 0; unit < 16; ++unit) {
+		if (fp->Base.ShadowSamplers & (1 << unit)) {
+			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
+
+			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
+			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
+		}
+	}
+}
+
+
+/**
+ * Transform the program to support fragment.position.
+ *
+ * Introduce a small fragment at the start of the program that will be
+ * the only code that directly reads the FRAG_ATTRIB_WPOS input.
+ * All other code pieces that reference that input will be rewritten
+ * to read from a newly allocated temporary.
+ *
+ */
+static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler, struct r300_fragment_program * fp)
+{
+	int i;
+
+	if (!(compiler->Base.Program.InputsRead & FRAG_BIT_WPOS)) {
+		fp->wpos_attr = FRAG_ATTRIB_MAX;
+		return;
+	}
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(compiler->Base.Program.InputsRead & (1 << i))) {
+			fp->wpos_attr = i;
+			break;
+		}
+	}
+
+	rc_transform_fragment_wpos(&compiler->Base, FRAG_ATTRIB_WPOS, fp->wpos_attr);
+}
+
+/**
+ * Rewrite fragment.fogcoord to use a texture coordinate slot.
+ * Note that fogcoord is forced into an X001 pattern, and this enforcement
+ * is done here.
+ *
+ * See also the counterpart rewriting for vertex programs.
+ */
+static void rewriteFog(struct r300_fragment_program_compiler *compiler, struct r300_fragment_program * fp)
+{
+	struct prog_src_register src;
+	int i;
+
+	if (!(compiler->Base.Program.InputsRead & FRAG_BIT_FOGC)) {
+		fp->fog_attr = FRAG_ATTRIB_MAX;
+		return;
+	}
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(compiler->Base.Program.InputsRead & (1 << i))) {
+			fp->fog_attr = i;
+			break;
+		}
+	}
+
+	memset(&src, 0, sizeof(src));
+	src.File = PROGRAM_INPUT;
+	src.Index = fp->fog_attr;
+	src.Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+	rc_move_input(&compiler->Base, FRAG_ATTRIB_FOGC, src);
+}
+
+
+/**
+ * Reserve hardware temporary registers for the program inputs.
+ *
+ * @note This allocation is performed explicitly, because the order of inputs
+ * is determined by the RS hardware.
+ */
+static void allocate_hw_inputs(
+	struct r300_fragment_program_compiler * c,
+	void (*allocate)(void * data, unsigned input, unsigned hwreg),
+	void * mydata)
+{
+	GLuint InputsRead = c->Base.Program.InputsRead;
+	int i;
+	GLuint hwindex = 0;
+
+	/* Primary colour */
+	if (InputsRead & FRAG_BIT_COL0)
+		allocate(mydata, FRAG_ATTRIB_COL0, hwindex++);
+	InputsRead &= ~FRAG_BIT_COL0;
+
+	/* Secondary color */
+	if (InputsRead & FRAG_BIT_COL1)
+		allocate(mydata, FRAG_ATTRIB_COL1, hwindex++);
+	InputsRead &= ~FRAG_BIT_COL1;
+
+	/* Texcoords */
+	for (i = 0; i < 8; i++) {
+		if (InputsRead & (FRAG_BIT_TEX0 << i))
+			allocate(mydata, FRAG_ATTRIB_TEX0+i, hwindex++);
+	}
+	InputsRead &= ~FRAG_BITS_TEX_ANY;
+
+	/* Fogcoords treated as a texcoord */
+	if (InputsRead & FRAG_BIT_FOGC)
+		allocate(mydata, FRAG_ATTRIB_FOGC, hwindex++);
+	InputsRead &= ~FRAG_BIT_FOGC;
+
+	/* fragment position treated as a texcoord */
+	if (InputsRead & FRAG_BIT_WPOS)
+		allocate(mydata, FRAG_ATTRIB_WPOS, hwindex++);
+	InputsRead &= ~FRAG_BIT_WPOS;
+
+	/* Anything else */
+	if (InputsRead)
+		rc_error(&c->Base, "Don't know how to handle inputs 0x%x\n", InputsRead);
+}
+
+
+static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_program_cont *cont, struct r300_fragment_program *fp)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program_compiler compiler;
+
+	rc_init(&compiler.Base);
+	compiler.Base.Debug = (RADEON_DEBUG & RADEON_PIXEL) ? GL_TRUE : GL_FALSE;
+
+	compiler.code = &fp->code;
+	compiler.state = fp->state;
+	compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) ? GL_TRUE : GL_FALSE;
+	compiler.OutputDepth = FRAG_RESULT_DEPTH;
+	compiler.OutputColor = FRAG_RESULT_COLOR;
+	compiler.AllocateHwInputs = &allocate_hw_inputs;
+
+	if (compiler.Base.Debug) {
+		fflush(stderr);
+		_mesa_printf("Fragment Program: Initial program:\n");
+		_mesa_print_program(&cont->Base.Base);
+		fflush(stderr);
+	}
+
+	rc_mesa_to_rc_program(&compiler.Base, &cont->Base.Base);
+
+	insert_WPOS_trailer(&compiler, fp);
+
+	rewriteFog(&compiler, fp);
+
+	r3xx_compile_fragment_program(&compiler);
+	fp->error = compiler.Base.Error;
+
+	fp->InputsRead = compiler.Base.Program.InputsRead;
+
+	rc_destroy(&compiler.Base);
+}
+
+struct r300_fragment_program *r300SelectAndTranslateFragmentShader(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program_cont *fp_list;
+	struct r300_fragment_program *fp;
+	struct r300_fragment_program_external_state state;
+
+	fp_list = (struct r300_fragment_program_cont *)ctx->FragmentProgram._Current;
+	build_state(r300, ctx->FragmentProgram._Current, &state);
+
+	fp = fp_list->progs;
+	while (fp) {
+		if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
+			return r300->selected_fp = fp;
+		}
+		fp = fp->next;
+	}
+
+	fp = _mesa_calloc(sizeof(struct r300_fragment_program));
+
+	fp->state = state;
+
+	fp->next = fp_list->progs;
+	fp_list->progs = fp;
+
+	translate_fragment_program(ctx, fp_list, fp);
+
+	return r300->selected_fp = fp;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.h b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
new file mode 100644
index 0000000000..3d64c08cee
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __R300_FRAGPROG_COMMON_H_
+#define __R300_FRAGPROG_COMMON_H_
+
+#include "main/mtypes.h"
+
+#include "r300_context.h"
+
+struct r300_fragment_program *r300SelectAndTranslateFragmentShader(GLcontext *ctx);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index ee85e229f0..5cb04e2bb6 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -44,82 +44,160 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/imports.h"
 #include "main/macros.h"
 #include "main/context.h"
+#include "main/simple_list.h"
 #include "swrast/swrast.h"
 
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "r300_context.h"
-#include "radeon_ioctl.h"
 #include "r300_ioctl.h"
 #include "r300_cmdbuf.h"
 #include "r300_state.h"
 #include "r300_vertprog.h"
 #include "radeon_reg.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
+#include "r300_context.h"
 
 #include "vblank.h"
 
+#define R200_3D_DRAW_IMMD_2      0xC0003500
+
 #define CLEARBUFFER_COLOR	0x1
 #define CLEARBUFFER_DEPTH	0x2
 #define CLEARBUFFER_STENCIL	0x4
 
-static void r300ClearBuffer(r300ContextPtr r300, int flags, int buffer)
-{
-	GLcontext *ctx = r300->radeon.glCtx;
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-	GLuint cboffset, cbpitch;
-	drm_r300_cmd_header_t *cmd2;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
-	r300ContextPtr rmesa = r300;
+#if 1
 
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s: %s buffer (%i,%i %ix%i)\n",
-			__FUNCTION__, buffer ? "back" : "front",
-			dPriv->x, dPriv->y, dPriv->w, dPriv->h);
+/**
+ * Fragment program helper macros
+ */
 
-	if (buffer) {
-		cboffset = r300->radeon.radeonScreen->backOffset;
-		cbpitch = r300->radeon.radeonScreen->backPitch;
-	} else {
-		cboffset = r300->radeon.radeonScreen->frontOffset;
-		cbpitch = r300->radeon.radeonScreen->frontPitch;
-	}
+/* Produce unshifted source selectors */
+#define FP_TMP(idx) (idx)
+#define FP_CONST(idx) ((idx) | (1 << 5))
+
+/* Produce source/dest selector dword */
+#define FP_SELC_MASK_NO		0
+#define FP_SELC_MASK_X		1
+#define FP_SELC_MASK_Y		2
+#define FP_SELC_MASK_XY		3
+#define FP_SELC_MASK_Z		4
+#define FP_SELC_MASK_XZ		5
+#define FP_SELC_MASK_YZ		6
+#define FP_SELC_MASK_XYZ	7
+
+#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \
+	(((destidx) << R300_ALU_DSTC_SHIFT) |		\
+	 (FP_SELC_MASK_##regmask << 23) |		\
+	 (FP_SELC_MASK_##outmask << 26) |		\
+	 ((src0) << R300_ALU_SRC0C_SHIFT) |		\
+	 ((src1) << R300_ALU_SRC1C_SHIFT) |		\
+	 ((src2) << R300_ALU_SRC2C_SHIFT))
+
+#define FP_SELA_MASK_NO		0
+#define FP_SELA_MASK_W		1
+
+#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \
+	(((destidx) << R300_ALU_DSTA_SHIFT) |		\
+	 (FP_SELA_MASK_##regmask << 23) |		\
+	 (FP_SELA_MASK_##outmask << 24) |		\
+	 ((src0) << R300_ALU_SRC0A_SHIFT) |		\
+	 ((src1) << R300_ALU_SRC1A_SHIFT) |		\
+	 ((src2) << R300_ALU_SRC2A_SHIFT))
+
+/* Produce unshifted argument selectors */
+#define FP_ARGC(source)	R300_ALU_ARGC_##source
+#define FP_ARGA(source) R300_ALU_ARGA_##source
+#define FP_ABS(arg) ((arg) | (1 << 6))
+#define FP_NEG(arg) ((arg) ^ (1 << 5))
+
+/* Produce instruction dword */
+#define FP_INSTRC(opcode,arg0,arg1,arg2) \
+	(R300_ALU_OUTC_##opcode | 		\
+	((arg0) << R300_ALU_ARG0C_SHIFT) |	\
+	((arg1) << R300_ALU_ARG1C_SHIFT) |	\
+	((arg2) << R300_ALU_ARG2C_SHIFT))
+
+#define FP_INSTRA(opcode,arg0,arg1,arg2) \
+	(R300_ALU_OUTA_##opcode | 		\
+	((arg0) << R300_ALU_ARG0A_SHIFT) |	\
+	((arg1) << R300_ALU_ARG1A_SHIFT) |	\
+	((arg2) << R300_ALU_ARG2A_SHIFT))
 
-	cboffset += r300->radeon.radeonScreen->fbLocation;
+#endif
 
-	cp_wait(r300, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
-	end_3d(rmesa);
+static void r300EmitClearState(GLcontext * ctx);
 
-	R300_STATECHANGE(r300, cb);
-	reg_start(R300_RB3D_COLOROFFSET0, 0);
-	e32(cboffset);
+static void r300ClearBuffer(r300ContextPtr r300, int flags,
+			    struct radeon_renderbuffer *rrb,
+			    struct radeon_renderbuffer *rrbd)
+{
+	BATCH_LOCALS(&r300->radeon);
+	GLcontext *ctx = r300->radeon.glCtx;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	GLuint cbpitch = 0;
+	r300ContextPtr rmesa = r300;
 
-	if (r300->radeon.radeonScreen->cpp == 4)
-		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		cbpitch |= R300_COLOR_FORMAT_RGB565;
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s: buffer %p (%i,%i %ix%i)\n",
+			__FUNCTION__, rrb, dPriv->x, dPriv->y,
+			dPriv->w, dPriv->h);
 
-	if (r300->radeon.sarea->tiling_enabled)
-		cbpitch |= R300_COLOR_TILE_ENABLE;
+	if (rrb) {
+		cbpitch = (rrb->pitch / rrb->cpp);
+		if (rrb->cpp == 4)
+			cbpitch |= R300_COLOR_FORMAT_ARGB8888;
+		else
+			cbpitch |= R300_COLOR_FORMAT_RGB565;
 
-	reg_start(R300_RB3D_COLORPITCH0, 0);
-	e32(cbpitch);
+		if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){
+			cbpitch |= R300_COLOR_TILE_ENABLE;
+        }
+	}
 
-	R300_STATECHANGE(r300, cmk);
-	reg_start(RB3D_COLOR_CHANNEL_MASK, 0);
+	/* TODO in bufmgr */
+	cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	end_3d(&rmesa->radeon);
 
 	if (flags & CLEARBUFFER_COLOR) {
-		e32((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) |
-		    (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) |
-		    (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) |
-		    (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0));
+		assert(rrb != 0);
+		BEGIN_BATCH_NO_AUTOSTATE(6);
+		OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1);
+		OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+		OUT_BATCH_REGVAL(R300_RB3D_COLORPITCH0, cbpitch);
+		END_BATCH();
+	}
+#if 1
+	if (flags & (CLEARBUFFER_DEPTH | CLEARBUFFER_STENCIL)) {
+		uint32_t zbpitch = (rrbd->pitch / rrbd->cpp);
+		if (rrbd->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){
+			zbpitch |= R300_DEPTHMACROTILE_ENABLE;
+        }
+		if (rrbd->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
+            zbpitch |= R300_DEPTHMICROTILE_TILED;
+        }
+		BEGIN_BATCH_NO_AUTOSTATE(6);
+		OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
+		OUT_BATCH_RELOC(0, rrbd->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+		OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1);
+		if (!r300->radeon.radeonScreen->kernel_mm)
+			OUT_BATCH(zbpitch);
+		else
+			OUT_BATCH_RELOC(zbpitch, rrbd->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+		END_BATCH();
+	}
+#endif
+	BEGIN_BATCH_NO_AUTOSTATE(6);
+	OUT_BATCH_REGSEQ(RB3D_COLOR_CHANNEL_MASK, 1);
+	if (flags & CLEARBUFFER_COLOR) {
+		OUT_BATCH((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) |
+			  (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) |
+			  (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) |
+			  (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0));
 	} else {
-		e32(0x0);
+		OUT_BATCH(0);
 	}
 
-	R300_STATECHANGE(r300, zs);
-	reg_start(R300_ZB_CNTL, 2);
 
 	{
 		uint32_t t1, t2;
@@ -146,73 +224,92 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags, int buffer)
 			     R300_S_FRONT_ZFAIL_OP_SHIFT);
 		}
 
-		e32(t1);
-		e32(t2);
-		e32(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) << R300_STENCILWRITEMASK_SHIFT) |
-		    (ctx->Stencil.Clear & R300_STENCILREF_MASK));
+		OUT_BATCH_REGSEQ(R300_ZB_CNTL, 3);
+		OUT_BATCH(t1);
+		OUT_BATCH(t2);
+		OUT_BATCH(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) <<
+                   R300_STENCILWRITEMASK_SHIFT) |
+			  (ctx->Stencil.Clear & R300_STENCILREF_MASK));
+		END_BATCH();
 	}
 
-	cmd2 = (drm_r300_cmd_header_t *) r300AllocCmdBuf(r300, 9, __FUNCTION__);
-	cmd2[0].packet3.cmd_type = R300_CMD_PACKET3;
-	cmd2[0].packet3.packet = R300_CMD_PACKET3_CLEAR;
-	cmd2[1].u = r300PackFloat32(dPriv->w / 2.0);
-	cmd2[2].u = r300PackFloat32(dPriv->h / 2.0);
-	cmd2[3].u = r300PackFloat32(ctx->Depth.Clear);
-	cmd2[4].u = r300PackFloat32(1.0);
-	cmd2[5].u = r300PackFloat32(ctx->Color.ClearColor[0]);
-	cmd2[6].u = r300PackFloat32(ctx->Color.ClearColor[1]);
-	cmd2[7].u = r300PackFloat32(ctx->Color.ClearColor[2]);
-	cmd2[8].u = r300PackFloat32(ctx->Color.ClearColor[3]);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(9);
+		OUT_BATCH(cmdpacket3(r300->radeon.radeonScreen, R300_CMD_PACKET3_CLEAR));
+		OUT_BATCH_FLOAT32(dPriv->w / 2.0);
+		OUT_BATCH_FLOAT32(dPriv->h / 2.0);
+		OUT_BATCH_FLOAT32(ctx->Depth.Clear);
+		OUT_BATCH_FLOAT32(1.0);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]);
+		END_BATCH();
+	} else {
+		OUT_BATCH(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8));
+		OUT_BATCH(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
+			  (1 << R300_PRIM_NUM_VERTICES_SHIFT));
+		OUT_BATCH_FLOAT32(dPriv->w / 2.0);
+		OUT_BATCH_FLOAT32(dPriv->h / 2.0);
+		OUT_BATCH_FLOAT32(ctx->Depth.Clear);
+		OUT_BATCH_FLOAT32(1.0);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]);
+	}
 
 	r300EmitCacheFlush(rmesa);
-	cp_wait(rmesa, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+
+	R300_STATECHANGE(r300, cb);
+	R300_STATECHANGE(r300, cmk);
+	R300_STATECHANGE(r300, zs);
 }
 
 static void r300EmitClearState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	r300ContextPtr rmesa = r300;
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	BATCH_LOCALS(&r300->radeon);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	int i;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
-	int has_tcl = 1;
+	int has_tcl;
 	int is_r500 = 0;
 	GLuint vap_cntl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
-        if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-                is_r500 = 1;
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
+		is_r500 = 1;
 
-
-	/* FIXME: the values written to R300_VAP_INPUT_ROUTE_0_0 and
-	 * R300_VAP_INPUT_ROUTE_0_1 are in fact known, however, the values are
-	 * quite complex; see the functions in r300_emit.c.
+	/* State atom dirty tracking is a little subtle here.
+	 *
+	 * On the one hand, we need to make sure base state is emitted
+	 * here if we start with an empty batch buffer, otherwise clear
+	 * works incorrectly with multiple processes. Therefore, the first
+	 * BEGIN_BATCH cannot be a BEGIN_BATCH_NO_AUTOSTATE.
+	 *
+	 * On the other hand, implicit state emission clears the state atom
+	 * dirty bits, so we have to call R300_STATECHANGE later than the
+	 * first BEGIN_BATCH.
 	 *
-	 * I believe it would be a good idea to extend the functions in
-	 * r300_emit.c so that they can be used to setup the default values for
-	 * these registers, as well as the actual values used for rendering.
+	 * The final trickiness is that, because we change state, we need
+	 * to ensure that any stored swtcl primitives are flushed properly
+	 * before we start changing state. See the R300_NEWPRIM in r300Clear
+	 * for this.
 	 */
-	R300_STATECHANGE(r300, vir[0]);
-	reg_start(R300_VAP_PROG_STREAM_CNTL_0, 0);
+	BEGIN_BATCH(31);
+	OUT_BATCH_REGSEQ(R300_VAP_PROG_STREAM_CNTL_0, 1);
 	if (!has_tcl)
-	    e32(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
+		OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
 		 ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)));
 	else
-	    e32(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
+		OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
 		 ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)));
 
-	/* disable fog */
-	R300_STATECHANGE(r300, fogs);
-	reg_start(R300_FG_FOG_BLEND, 0);
-	e32(0x0);
-
-	R300_STATECHANGE(r300, vir[1]);
-	reg_start(R300_VAP_PROG_STREAM_CNTL_EXT_0, 0);
-	e32(((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) |
+	OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0);
+	OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0,
+	   ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) |
 	       (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) |
 	       (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) |
 	       (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) |
@@ -226,619 +323,460 @@ static void r300EmitClearState(GLcontext * ctx)
 	      << R300_SWIZZLE1_SHIFT)));
 
 	/* R300_VAP_INPUT_CNTL_0, R300_VAP_INPUT_CNTL_1 */
-	R300_STATECHANGE(r300, vic);
-	reg_start(R300_VAP_VTX_STATE_CNTL, 1);
-	e32((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT));
-	e32(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0);
+	OUT_BATCH_REGSEQ(R300_VAP_VTX_STATE_CNTL, 2);
+	OUT_BATCH((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT));
+	OUT_BATCH(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0);
 
-	R300_STATECHANGE(r300, vte);
 	/* comes from fglrx startup of clear */
-	reg_start(R300_SE_VTE_CNTL, 1);
-	e32(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA |
-	    R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
-	    R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
-	    R300_VPORT_Z_OFFSET_ENA);
-	e32(0x8);
+	OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2);
+	OUT_BATCH(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA |
+		  R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
+		  R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
+		  R300_VPORT_Z_OFFSET_ENA);
+	OUT_BATCH(0x8);
 
-	reg_start(R300_VAP_PSC_SGN_NORM_CNTL, 0);
-	e32(0xaaaaaaaa);
+	OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa);
 
-	R300_STATECHANGE(r300, vof);
-	reg_start(R300_VAP_OUTPUT_VTX_FMT_0, 1);
-	e32(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT |
-	    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT);
-	e32(0x0);		/* no textures */
+	OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
+	OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT |
+		  R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT);
+	OUT_BATCH(0); /* no textures */
 
-	R300_STATECHANGE(r300, txe);
-	reg_start(R300_TX_ENABLE, 0);
-	e32(0x0);
+	OUT_BATCH_REGVAL(R300_TX_ENABLE, 0);
 
-	R300_STATECHANGE(r300, vpt);
-	reg_start(R300_SE_VPORT_XSCALE, 5);
-	efloat(1.0);
-	efloat(dPriv->x);
-	efloat(1.0);
-	efloat(dPriv->y);
-	efloat(1.0);
-	efloat(0.0);
+	OUT_BATCH_REGSEQ(R300_SE_VPORT_XSCALE, 6);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(dPriv->x);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(dPriv->y);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(0.0);
 
-	R300_STATECHANGE(r300, at);
-	reg_start(R300_FG_ALPHA_FUNC, 0);
-	e32(0x0);
+	OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0);
+
+	OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2);
+	OUT_BATCH(0x0);
+	OUT_BATCH(0x0);
+	END_BATCH();
 
+	R300_STATECHANGE(r300, vir[0]);
+	R300_STATECHANGE(r300, fogs);
+	R300_STATECHANGE(r300, vir[1]);
+	R300_STATECHANGE(r300, vic);
+	R300_STATECHANGE(r300, vte);
+	R300_STATECHANGE(r300, vof);
+	R300_STATECHANGE(r300, txe);
+	R300_STATECHANGE(r300, vpt);
+	R300_STATECHANGE(r300, at);
 	R300_STATECHANGE(r300, bld);
-	reg_start(R300_RB3D_CBLEND, 1);
-	e32(0x0);
-	e32(0x0);
+	R300_STATECHANGE(r300, ps);
 
 	if (has_tcl) {
-	    R300_STATECHANGE(r300, vap_clip_cntl);
-	    reg_start(R300_VAP_CLIP_CNTL, 0);
-	    e32(R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE);
+		R300_STATECHANGE(r300, vap_clip_cntl);
+
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE);
+		END_BATCH();
         }
 
-	R300_STATECHANGE(r300, ps);
-	reg_start(R300_GA_POINT_SIZE, 0);
-	e32(((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) |
-	    ((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT));
+	BEGIN_BATCH_NO_AUTOSTATE(2);
+	OUT_BATCH_REGVAL(R300_GA_POINT_SIZE,
+		((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) |
+		((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT));
+	END_BATCH();
 
 	if (!is_r500) {
 		R300_STATECHANGE(r300, ri);
-		reg_start(R300_RS_IP_0, 7);
-		for (i = 0; i < 8; ++i) {
-			e32(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3));
-		}
-
 		R300_STATECHANGE(r300, rc);
-		/* The second constant is needed to get glxgears display anything .. */
-		reg_start(R300_RS_COUNT, 1);
-		e32((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
-		e32(0x0);
-
 		R300_STATECHANGE(r300, rr);
-		reg_start(R300_RS_INST_0, 0);
-		e32(R300_RS_INST_COL_CN_WRITE);
+
+		BEGIN_BATCH(14);
+		OUT_BATCH_REGSEQ(R300_RS_IP_0, 8);
+		for (i = 0; i < 8; ++i)
+			OUT_BATCH(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3));
+
+		OUT_BATCH_REGSEQ(R300_RS_COUNT, 2);
+		OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+		OUT_BATCH(0x0);
+
+		OUT_BATCH_REGVAL(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE);
+		END_BATCH();
 	} else {
 		R300_STATECHANGE(r300, ri);
-		reg_start(R500_RS_IP_0, 7);
+		R300_STATECHANGE(r300, rc);
+		R300_STATECHANGE(r300, rr);
+
+		BEGIN_BATCH(14);
+		OUT_BATCH_REGSEQ(R500_RS_IP_0, 8);
 		for (i = 0; i < 8; ++i) {
-			e32((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
-			    (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-			    (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-			    (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
+			OUT_BATCH((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
+				  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
+				  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
+				  (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
 		}
 
-		R300_STATECHANGE(r300, rc);
-		/* The second constant is needed to get glxgears display anything .. */
-		reg_start(R300_RS_COUNT, 1);
-		e32((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
-		e32(0x0);
-
-		R300_STATECHANGE(r300, rr);
-		reg_start(R500_RS_INST_0, 0);
-		e32(R500_RS_INST_COL_CN_WRITE);
+		OUT_BATCH_REGSEQ(R300_RS_COUNT, 2);
+		OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+		OUT_BATCH(0x0);
 
+		OUT_BATCH_REGVAL(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE);
+		END_BATCH();
 	}
 
 	if (!is_r500) {
 		R300_STATECHANGE(r300, fp);
-		reg_start(R300_US_CONFIG, 2);
-		e32(0x0);
-		e32(0x0);
-		e32(0x0);
-		reg_start(R300_US_CODE_ADDR_0, 3);
-		e32(0x0);
-		e32(0x0);
-		e32(0x0);
-		e32(R300_RGBA_OUT);
-
 		R300_STATECHANGE(r300, fpi[0]);
 		R300_STATECHANGE(r300, fpi[1]);
 		R300_STATECHANGE(r300, fpi[2]);
 		R300_STATECHANGE(r300, fpi[3]);
 
-		reg_start(R300_US_ALU_RGB_INST_0, 0);
-		e32(FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)));
-
-		reg_start(R300_US_ALU_RGB_ADDR_0, 0);
-		e32(FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0));
-
-		reg_start(R300_US_ALU_ALPHA_INST_0, 0);
-		e32(FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)));
-
-		reg_start(R300_US_ALU_ALPHA_ADDR_0, 0);
-		e32(FP_SELA(0, NO, W, FP_TMP(0), 0, 0));
+		BEGIN_BATCH(17);
+		OUT_BATCH_REGSEQ(R300_US_CONFIG, 3);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(R300_RGBA_OUT);
+
+		OUT_BATCH_REGVAL(R300_US_ALU_RGB_INST_0,
+			FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)));
+		OUT_BATCH_REGVAL(R300_US_ALU_RGB_ADDR_0,
+			FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0));
+		OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_INST_0,
+			FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)));
+		OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_ADDR_0,
+			FP_SELA(0, NO, W, FP_TMP(0), 0, 0));
+		END_BATCH();
 	} else {
- 		R300_STATECHANGE(r300, fp);
- 		reg_start(R500_US_CONFIG, 1);
- 		e32(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
- 		e32(0x0);
- 		reg_start(R500_US_CODE_ADDR, 2);
- 		e32(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1));
- 		e32(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1));
- 		e32(R500_US_CODE_OFFSET_ADDR(0));
+		struct radeon_state_atom r500fp;
+		uint32_t _cmd[10];
 
+		R300_STATECHANGE(r300, fp);
 		R300_STATECHANGE(r300, r500fp);
-		r500fp_start_fragment(0, 6);
-
-		e32(R500_INST_TYPE_OUT |
-		    R500_INST_TEX_SEM_WAIT |
-		    R500_INST_LAST |
-		    R500_INST_RGB_OMASK_R |
-		    R500_INST_RGB_OMASK_G |
-		    R500_INST_RGB_OMASK_B |
-		    R500_INST_ALPHA_OMASK |
-		    R500_INST_RGB_CLAMP |
-		    R500_INST_ALPHA_CLAMP);
-
-		e32(R500_RGB_ADDR0(0) |
-		    R500_RGB_ADDR1(0) |
-		    R500_RGB_ADDR1_CONST |
-		    R500_RGB_ADDR2(0) |
-		    R500_RGB_ADDR2_CONST);
-
-		e32(R500_ALPHA_ADDR0(0) |
-		    R500_ALPHA_ADDR1(0) |
-		    R500_ALPHA_ADDR1_CONST |
-		    R500_ALPHA_ADDR2(0) |
-		    R500_ALPHA_ADDR2_CONST);
-
-		e32(R500_ALU_RGB_SEL_A_SRC0 |
-		    R500_ALU_RGB_R_SWIZ_A_R |
-		    R500_ALU_RGB_G_SWIZ_A_G |
-		    R500_ALU_RGB_B_SWIZ_A_B |
-		    R500_ALU_RGB_SEL_B_SRC0 |
-		    R500_ALU_RGB_R_SWIZ_B_R |
-		    R500_ALU_RGB_B_SWIZ_B_G |
-		    R500_ALU_RGB_G_SWIZ_B_B);
-
-		e32(R500_ALPHA_OP_CMP |
-		    R500_ALPHA_SWIZ_A_A |
-		    R500_ALPHA_SWIZ_B_A);
-
-		e32(R500_ALU_RGBA_OP_CMP |
-		    R500_ALU_RGBA_R_SWIZ_0 |
-		    R500_ALU_RGBA_G_SWIZ_0 |
-		    R500_ALU_RGBA_B_SWIZ_0 |
-		    R500_ALU_RGBA_A_SWIZ_0);
+
+		BEGIN_BATCH(7);
+		OUT_BATCH_REGSEQ(R500_US_CONFIG, 2);
+		OUT_BATCH(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
+		OUT_BATCH(0x0);
+		OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3);
+		OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1));
+		OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1));
+		OUT_BATCH(R500_US_CODE_OFFSET_ADDR(0));
+		END_BATCH();
+
+		r500fp.check = check_r500fp;
+		r500fp.cmd = _cmd;
+		r500fp.cmd[0] = cmdr500fp(r300->radeon.radeonScreen, 0, 1, 0, 0);
+		r500fp.cmd[1] = R500_INST_TYPE_OUT |
+			R500_INST_TEX_SEM_WAIT |
+			R500_INST_LAST |
+			R500_INST_RGB_OMASK_R |
+			R500_INST_RGB_OMASK_G |
+			R500_INST_RGB_OMASK_B |
+			R500_INST_ALPHA_OMASK |
+			R500_INST_RGB_CLAMP |
+			R500_INST_ALPHA_CLAMP;
+		r500fp.cmd[2] = R500_RGB_ADDR0(0) |
+			R500_RGB_ADDR1(0) |
+			R500_RGB_ADDR1_CONST |
+			R500_RGB_ADDR2(0) |
+			R500_RGB_ADDR2_CONST;
+		r500fp.cmd[3] = R500_ALPHA_ADDR0(0) |
+			R500_ALPHA_ADDR1(0) |
+			R500_ALPHA_ADDR1_CONST |
+			R500_ALPHA_ADDR2(0) |
+			R500_ALPHA_ADDR2_CONST;
+		r500fp.cmd[4] = R500_ALU_RGB_SEL_A_SRC0 |
+			R500_ALU_RGB_R_SWIZ_A_R |
+			R500_ALU_RGB_G_SWIZ_A_G |
+			R500_ALU_RGB_B_SWIZ_A_B |
+			R500_ALU_RGB_SEL_B_SRC0 |
+			R500_ALU_RGB_R_SWIZ_B_R |
+			R500_ALU_RGB_B_SWIZ_B_G |
+			R500_ALU_RGB_G_SWIZ_B_B;
+		r500fp.cmd[5] = R500_ALPHA_OP_CMP |
+			R500_ALPHA_SWIZ_A_A |
+			R500_ALPHA_SWIZ_B_A;
+		r500fp.cmd[6] = R500_ALU_RGBA_OP_CMP |
+			R500_ALU_RGBA_R_SWIZ_0 |
+			R500_ALU_RGBA_G_SWIZ_0 |
+			R500_ALU_RGBA_B_SWIZ_0 |
+			R500_ALU_RGBA_A_SWIZ_0;
+
+		r500fp.cmd[7] = 0;
+		if (r300->radeon.radeonScreen->kernel_mm) {
+			emit_r500fp(ctx, &r500fp);
+		} else {
+			int dwords = r500fp.check(ctx,&r500fp);
+			BEGIN_BATCH_NO_AUTOSTATE(dwords);
+			OUT_BATCH_TABLE(r500fp.cmd, dwords);
+			END_BATCH();
+		}
+
 	}
 
-	reg_start(R300_VAP_PVS_STATE_FLUSH_REG, 0);
-	e32(0x00000000);
+	BEGIN_BATCH(2);
+	OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0);
+	END_BATCH();
+
 	if (has_tcl) {
-	    vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
+		vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
 			(5 << R300_PVS_NUM_CNTLRS_SHIFT) |
 			(12 << R300_VF_MAX_VTX_NUM_SHIFT));
-	    if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		vap_cntl |= R500_TCL_STATE_OPTIMIZATION;
-	} else
-	    vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
+		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
+			vap_cntl |= R500_TCL_STATE_OPTIMIZATION;
+	} else {
+		vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
 			(5 << R300_PVS_NUM_CNTLRS_SHIFT) |
 			(5 << R300_VF_MAX_VTX_NUM_SHIFT));
+	}
 
 	if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV515)
-	    vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV560) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV570))
-	    vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420))
-	    vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R520) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R580))
-	    vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT);
 	else
-	    vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT);
 
-	R300_STATECHANGE(rmesa, vap_cntl);
-	reg_start(R300_VAP_CNTL, 0);
-	e32(vap_cntl);
+	R300_STATECHANGE(r300, vap_cntl);
+
+	BEGIN_BATCH(2);
+	OUT_BATCH_REGVAL(R300_VAP_CNTL, vap_cntl);
+	END_BATCH();
 
 	if (has_tcl) {
+        struct radeon_state_atom vpu;
+        uint32_t _cmd[10];
 		R300_STATECHANGE(r300, pvs);
-		reg_start(R300_VAP_PVS_CODE_CNTL_0, 2);
-
-		e32((0 << R300_PVS_FIRST_INST_SHIFT) |
-		    (0 << R300_PVS_XYZW_VALID_INST_SHIFT) |
-		    (1 << R300_PVS_LAST_INST_SHIFT));
-		e32((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-		    (0 << R300_PVS_MAX_CONST_ADDR_SHIFT));
-		e32(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-
+		R300_STATECHANGE(r300, vap_flush);
 		R300_STATECHANGE(r300, vpi);
-		vsf_start_fragment(0x0, 8);
 
-		e32(PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 0, 0xf, PVS_DST_REG_OUT));
-		e32(PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(0x0);
+		BEGIN_BATCH(4);
+		OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
+		OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) |
+			  (0 << R300_PVS_XYZW_VALID_INST_SHIFT) |
+			  (1 << R300_PVS_LAST_INST_SHIFT));
+		OUT_BATCH((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
+			  (0 << R300_PVS_MAX_CONST_ADDR_SHIFT));
+		OUT_BATCH(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
+		END_BATCH();
+
+		vpu.check = check_vpu;
+		vpu.cmd = _cmd;
+		vpu.cmd[0] = cmdvpu(r300->radeon.radeonScreen, 0, 2);
+
+		vpu.cmd[1] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE,
+                                         0, 0xf, PVS_DST_REG_OUT);
+		vpu.cmd[2] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y,
+                                      PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W,
+                                      PVS_SRC_REG_INPUT, NEGATE_NONE);
+		vpu.cmd[3] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_REG_INPUT, NEGATE_NONE);
+		vpu.cmd[4] = 0x0;
+
+		vpu.cmd[5] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf,
+                                         PVS_DST_REG_OUT);
+		vpu.cmd[6] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X,
+                                      PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z,
+                                      PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT,
+                                      NEGATE_NONE);
+		vpu.cmd[7] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_REG_INPUT, NEGATE_NONE);
+		vpu.cmd[8] = 0x0;
+
+		if (r300->radeon.radeonScreen->kernel_mm) {
+			int dwords = r300->hw.vap_flush.check(ctx,&r300->hw.vap_flush);
+			BEGIN_BATCH_NO_AUTOSTATE(dwords);
+			OUT_BATCH_TABLE(r300->hw.vap_flush.cmd, dwords);
+			END_BATCH();
+			emit_vpu(ctx, &vpu);
+		} else {
+			int dwords = vpu.check(ctx,&vpu);
+			BEGIN_BATCH_NO_AUTOSTATE(dwords);
+			OUT_BATCH_TABLE(vpu.cmd, dwords);
+			END_BATCH();
+		}
 
-		e32(PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf, PVS_DST_REG_OUT));
-		e32(PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(0x0);
 	}
 }
 
-/**
- * Buffer clear
- */
-static void r300Clear(GLcontext * ctx, GLbitfield mask)
+static int r300KernelClear(GLcontext *ctx, GLuint flags)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-	int flags = 0;
-	int bits = 0;
-	int swapped;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "r300Clear\n");
-
-	{
-		LOCK_HARDWARE(&r300->radeon);
-		UNLOCK_HARDWARE(&r300->radeon);
-		if (dPriv->numClipRects == 0)
-			return;
-	}
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+	struct radeon_renderbuffer *rrb;
+	struct radeon_renderbuffer *rrbd;
+	int bits = 0, ret;
 
-	if (mask & BUFFER_BIT_FRONT_LEFT) {
-		flags |= BUFFER_BIT_FRONT_LEFT;
-		mask &= ~BUFFER_BIT_FRONT_LEFT;
-	}
+	/* Make sure it fits there. */
+	radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs);
 
-	if (mask & BUFFER_BIT_BACK_LEFT) {
-		flags |= BUFFER_BIT_BACK_LEFT;
-		mask &= ~BUFFER_BIT_BACK_LEFT;
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (mask & BUFFER_BIT_DEPTH) {
-		bits |= CLEARBUFFER_DEPTH;
-		mask &= ~BUFFER_BIT_DEPTH;
+	if (flags & BUFFER_BIT_FRONT_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if ((mask & BUFFER_BIT_STENCIL) && r300->state.stencil.hw_stencil) {
-		bits |= CLEARBUFFER_STENCIL;
-		mask &= ~BUFFER_BIT_STENCIL;
+	if (flags & BUFFER_BIT_BACK_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (mask) {
-		if (RADEON_DEBUG & DEBUG_FALLBACKS)
-			fprintf(stderr, "%s: swrast clear, mask: %x\n",
-				__FUNCTION__, mask);
-		_swrast_Clear(ctx, mask);
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd) {
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	swapped = r300->radeon.sarea->pfCurrentPage == 1;
+	ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs);
+	if (ret)
+	  return -1;
 
-	/* Make sure it fits there. */
-	r300EnsureCmdBufSpace(r300, 421 * 3, __FUNCTION__);
+	rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__);
 	if (flags || bits)
 		r300EmitClearState(ctx);
 
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd && (flags & BUFFER_BIT_DEPTH))
+		bits |= CLEARBUFFER_DEPTH;
+
+	if (rrbd && (flags & BUFFER_BIT_STENCIL))
+		bits |= CLEARBUFFER_STENCIL;
+
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL);
+		bits = 0;
+	}
+
 	if (flags & BUFFER_BIT_FRONT_LEFT) {
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, swapped);
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
 		bits = 0;
 	}
 
 	if (flags & BUFFER_BIT_BACK_LEFT) {
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, swapped ^ 1);
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
 		bits = 0;
 	}
 
 	if (bits)
-		r300ClearBuffer(r300, bits, 0);
-
-}
-
-void r300Flush(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+		r300ClearBuffer(r300, bits, NULL, rrbd);
 
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush( rmesa );
-
-	if (rmesa->cmdbuf.count_used > rmesa->cmdbuf.count_reemit)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
+	COMMIT_BATCH();
+	return 0;
 }
 
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-
-void r300RefillCurrentDmaRegion(r300ContextPtr rmesa, int size)
+/**
+ * Buffer clear
+ */
+static void r300Clear(GLcontext * ctx, GLbitfield mask)
 {
-	struct r300_dma_buffer *dmabuf;
-	size = MAX2(size, RADEON_BUFFER_SIZE * 16);
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-		fprintf(stderr, "%s\n", __FUNCTION__);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+	GLbitfield swrast_mask = 0, tri_mask = 0;
+	int i, ret;
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
 
-	if (rmesa->dma.flush) {
-		rmesa->dma.flush(rmesa);
-	}
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "r300Clear\n");
 
-	if (rmesa->dma.current.buf) {
-#ifdef USER_BUFFERS
-		r300_mem_use(rmesa, rmesa->dma.current.buf->id);
-#endif
-		r300ReleaseDmaRegion(rmesa, &rmesa->dma.current, __FUNCTION__);
+	if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(&r300->radeon);
+		UNLOCK_HARDWARE(&r300->radeon);
+		if (dPriv->numClipRects == 0)
+			return;
 	}
-	if (rmesa->dma.nr_released_bufs > 4)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
 
-	dmabuf = CALLOC_STRUCT(r300_dma_buffer);
-	dmabuf->buf = (void *)1;	/* hack */
-	dmabuf->refcount = 1;
-
-	dmabuf->id = r300_mem_alloc(rmesa, 4, size);
-	if (dmabuf->id == 0) {
-		LOCK_HARDWARE(&rmesa->radeon);	/* no need to validate */
-
-		r300FlushCmdBufLocked(rmesa, __FUNCTION__);
-		radeonWaitForIdleLocked(&rmesa->radeon);
+	/* Flush swtcl vertices if necessary, because we will change hardware
+	 * state during clear. See also the state-related comment in
+	 * r300EmitClearState.
+	 */
+	R300_NEWPRIM(r300);
 
-		dmabuf->id = r300_mem_alloc(rmesa, 4, size);
+	if (colorMask == ~0)
+	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+	else
+	  tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
 
-		UNLOCK_HARDWARE(&rmesa->radeon);
 
-		if (dmabuf->id == 0) {
-			fprintf(stderr,
-				"Error: Could not get dma buffer... exiting\n");
-			_mesa_exit(-1);
-		}
+	/* HW stencil */
+	if (mask & BUFFER_BIT_STENCIL) {
+		tri_mask |= BUFFER_BIT_STENCIL;
 	}
 
-	rmesa->dma.current.buf = dmabuf;
-	rmesa->dma.current.address = r300_mem_ptr(rmesa, dmabuf->id);
-	rmesa->dma.current.end = size;
-	rmesa->dma.current.start = 0;
-	rmesa->dma.current.ptr = 0;
-}
-
-void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-			  struct r300_dma_region *region, const char *caller)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-	if (!region->buf)
-		return;
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (--region->buf->refcount == 0) {
-		r300_mem_free(rmesa, region->buf->id);
-		FREE(region->buf);
-		rmesa->dma.nr_released_bufs++;
+	/* HW depth */
+	if (mask & BUFFER_BIT_DEPTH) {
+    	        tri_mask |= BUFFER_BIT_DEPTH;
 	}
 
-	region->buf = 0;
-	region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r300AllocDmaRegion(r300ContextPtr rmesa,
-			struct r300_dma_region *region,
-			int bytes, int alignment)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (region->buf)
-		r300ReleaseDmaRegion(rmesa, region, __FUNCTION__);
-
-	alignment--;
-	rmesa->dma.current.start = rmesa->dma.current.ptr =
-	    (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-	if (rmesa->dma.current.ptr + bytes > rmesa->dma.current.end)
-		r300RefillCurrentDmaRegion(rmesa, (bytes + 0x7) & ~0x7);
-
-	region->start = rmesa->dma.current.start;
-	region->ptr = rmesa->dma.current.start;
-	region->end = rmesa->dma.current.start + bytes;
-	region->address = rmesa->dma.current.address;
-	region->buf = rmesa->dma.current.buf;
-	region->buf->refcount++;
-
-	rmesa->dma.current.ptr += bytes;	/* bug - if alignment > 7 */
-	rmesa->dma.current.start =
-	    rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;
-
-	assert(rmesa->dma.current.ptr <= rmesa->dma.current.end);
-}
+	/* If we're doing a tri pass for depth/stencil, include a likely color
+	 * buffer with it.
+	 */
 
-#else
-static void r300RefillCurrentDmaRegion(r300ContextPtr rmesa)
-{
-	struct r300_dma_buffer *dmabuf;
-	int fd = rmesa->radeon.dri.fd;
-	int index = 0;
-	int size = 0;
-	drmDMAReq dma;
-	int ret;
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->dma.flush) {
-		rmesa->dma.flush(rmesa);
+	for (i = 0; i < BUFFER_COUNT; i++) {
+	  GLuint bufBit = 1 << i;
+	  if ((tri_mask) & bufBit) {
+	    if (!fb->Attachment[i].Renderbuffer->ClassID) {
+	      tri_mask &= ~bufBit;
+	      swrast_mask |= bufBit;
+	    }
+	  }
 	}
 
-	if (rmesa->dma.current.buf)
-		r300ReleaseDmaRegion(rmesa, &rmesa->dma.current, __FUNCTION__);
-
-	if (rmesa->dma.nr_released_bufs > 4)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-	dma.context = rmesa->radeon.dri.hwContext;
-	dma.send_count = 0;
-	dma.send_list = NULL;
-	dma.send_sizes = NULL;
-	dma.flags = 0;
-	dma.request_count = 1;
-	dma.request_size = RADEON_BUFFER_SIZE;
-	dma.request_list = &index;
-	dma.request_sizes = &size;
-	dma.granted_count = 0;
-
-	LOCK_HARDWARE(&rmesa->radeon);	/* no need to validate */
-
-	ret = drmDMA(fd, &dma);
-
-	if (ret != 0) {
-		/* Try to release some buffers and wait until we can't get any more */
-		if (rmesa->dma.nr_released_bufs) {
-			r300FlushCmdBufLocked(rmesa, __FUNCTION__);
-		}
-
-		if (RADEON_DEBUG & DEBUG_DMA)
-			fprintf(stderr, "Waiting for buffers\n");
-
-		radeonWaitForIdleLocked(&rmesa->radeon);
-		ret = drmDMA(fd, &dma);
-
-		if (ret != 0) {
-			UNLOCK_HARDWARE(&rmesa->radeon);
-			fprintf(stderr,
-				"Error: Could not get dma buffer... exiting\n");
-			_mesa_exit(-1);
+	/* SW fallback clearing */
+	swrast_mask = mask & ~tri_mask;
+
+	ret = 0;
+	if (tri_mask) {
+		if (r300->radeon.radeonScreen->kernel_mm)
+			radeonUserClear(ctx, tri_mask);
+		else {
+			/* if kernel clear fails due to size restraints fallback */
+			ret = r300KernelClear(ctx, tri_mask);
+			if (ret < 0)
+				swrast_mask |= tri_mask;
 		}
 	}
 
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	if (RADEON_DEBUG & DEBUG_DMA)
-		fprintf(stderr, "Allocated buffer %d\n", index);
-
-	dmabuf = CALLOC_STRUCT(r300_dma_buffer);
-	dmabuf->buf = &rmesa->radeon.radeonScreen->buffers->list[index];
-	dmabuf->refcount = 1;
-
-	rmesa->dma.current.buf = dmabuf;
-	rmesa->dma.current.address = dmabuf->buf->address;
-	rmesa->dma.current.end = dmabuf->buf->total;
-	rmesa->dma.current.start = 0;
-	rmesa->dma.current.ptr = 0;
-}
-
-void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-			  struct r300_dma_region *region, const char *caller)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-	if (!region->buf)
-		return;
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (--region->buf->refcount == 0) {
-		drm_radeon_cmd_header_t *cmd;
-
-		if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-			fprintf(stderr, "%s -- DISCARD BUF %d\n",
-				__FUNCTION__, region->buf->buf->idx);
-		cmd =
-		    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa,
-								sizeof
-								(*cmd) / 4,
-								__FUNCTION__);
-		cmd->dma.cmd_type = R300_CMD_DMA_DISCARD;
-		cmd->dma.buf_idx = region->buf->buf->idx;
-
-		FREE(region->buf);
-		rmesa->dma.nr_released_bufs++;
+	if (swrast_mask) {
+		if (RADEON_DEBUG & RADEON_FALLBACKS)
+			fprintf(stderr, "%s: swrast clear, mask: %x\n",
+				__FUNCTION__, swrast_mask);
+		_swrast_Clear(ctx, swrast_mask);
 	}
-
-	region->buf = 0;
-	region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r300AllocDmaRegion(r300ContextPtr rmesa,
-			struct r300_dma_region *region,
-			int bytes, int alignment)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (region->buf)
-		r300ReleaseDmaRegion(rmesa, region, __FUNCTION__);
-
-	alignment--;
-	rmesa->dma.current.start = rmesa->dma.current.ptr =
-	    (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-	if (rmesa->dma.current.ptr + bytes > rmesa->dma.current.end)
-		r300RefillCurrentDmaRegion(rmesa);
-
-	region->start = rmesa->dma.current.start;
-	region->ptr = rmesa->dma.current.start;
-	region->end = rmesa->dma.current.start + bytes;
-	region->address = rmesa->dma.current.address;
-	region->buf = rmesa->dma.current.buf;
-	region->buf->refcount++;
-
-	rmesa->dma.current.ptr += bytes;	/* bug - if alignment > 7 */
-	rmesa->dma.current.start =
-	    rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;
-
-	assert(rmesa->dma.current.ptr <= rmesa->dma.current.end);
-}
-
-#endif
-
-GLboolean r300IsGartMemory(r300ContextPtr rmesa, const GLvoid * pointer,
-			   GLint size)
-{
-	int offset =
-	    (char *)pointer -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-	int valid = (size >= 0 && offset >= 0
-		     && offset + size <
-		     rmesa->radeon.radeonScreen->gartTextures.size);
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "r300IsGartMemory( %p ) : %d\n", pointer,
-			valid);
-
-	return valid;
-}
-
-GLuint r300GartOffsetFromVirtual(r300ContextPtr rmesa, const GLvoid * pointer)
-{
-	int offset =
-	    (char *)pointer -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-
-	//fprintf(stderr, "offset=%08x\n", offset);
-
-	if (offset < 0
-	    || offset > rmesa->radeon.radeonScreen->gartTextures.size)
-		return ~0;
-	else
-		return rmesa->radeon.radeonScreen->gart_texture_offset + offset;
 }
 
 void r300InitIoctlFuncs(struct dd_function_table *functions)
 {
 	functions->Clear = r300Clear;
 	functions->Finish = radeonFinish;
-	functions->Flush = r300Flush;
+	functions->Flush = radeonFlush;
 }
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h
index e1143fb6c3..3abfa71a6e 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.h
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.h
@@ -39,22 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 #include "radeon_drm.h"
 
-extern GLboolean r300IsGartMemory(r300ContextPtr rmesa,
-				  const GLvoid * pointer, GLint size);
-
-extern GLuint r300GartOffsetFromVirtual(r300ContextPtr rmesa,
-					const GLvoid * pointer);
-
-extern void r300Flush(GLcontext * ctx);
-
-extern void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-				 struct r300_dma_region *region,
-				 const char *caller);
-extern void r300AllocDmaRegion(r300ContextPtr rmesa,
-			       struct r300_dma_region *region, int bytes,
-			       int alignment);
-
 extern void r300InitIoctlFuncs(struct dd_function_table *functions);
 
-extern void r300RefillCurrentDmaRegion(r300ContextPtr rmesa, int size);
 #endif				/* __R300_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_mem.c b/src/mesa/drivers/dri/r300/r300_mem.c
deleted file mode 100644
index f8f9d4fcdf..0000000000
--- a/src/mesa/drivers/dri/r300/r300_mem.c
+++ /dev/null
@@ -1,385 +0,0 @@
-/*
- * Copyright (C) 2005 Aapo Tahkola.
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-/**
- * \file
- *
- * \author Aapo Tahkola <aet@rasterburn.org>
- */
-
-#include <unistd.h>
-
-#include "r300_context.h"
-#include "r300_cmdbuf.h"
-#include "r300_ioctl.h"
-#include "r300_mem.h"
-#include "radeon_ioctl.h"
-
-#ifdef USER_BUFFERS
-
-static void resize_u_list(r300ContextPtr rmesa)
-{
-	void *temp;
-	int nsize;
-
-	temp = rmesa->rmm->u_list;
-	nsize = rmesa->rmm->u_size * 2;
-
-	rmesa->rmm->u_list = _mesa_malloc(nsize * sizeof(*rmesa->rmm->u_list));
-	_mesa_memset(rmesa->rmm->u_list, 0,
-		     nsize * sizeof(*rmesa->rmm->u_list));
-
-	if (temp) {
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-		_mesa_memcpy(rmesa->rmm->u_list, temp,
-			     rmesa->rmm->u_size * sizeof(*rmesa->rmm->u_list));
-		_mesa_free(temp);
-	}
-
-	rmesa->rmm->u_size = nsize;
-}
-
-void r300_mem_init(r300ContextPtr rmesa)
-{
-	rmesa->rmm = malloc(sizeof(struct r300_memory_manager));
-	memset(rmesa->rmm, 0, sizeof(struct r300_memory_manager));
-
-	rmesa->rmm->u_size = 128;
-	resize_u_list(rmesa);
-}
-
-void r300_mem_destroy(r300ContextPtr rmesa)
-{
-	_mesa_free(rmesa->rmm->u_list);
-	rmesa->rmm->u_list = NULL;
-
-	_mesa_free(rmesa->rmm);
-	rmesa->rmm = NULL;
-}
-
-void *r300_mem_ptr(r300ContextPtr rmesa, int id)
-{
-	assert(id <= rmesa->rmm->u_last);
-	return rmesa->rmm->u_list[id].ptr;
-}
-
-int r300_mem_find(r300ContextPtr rmesa, void *ptr)
-{
-	int i;
-
-	for (i = 1; i < rmesa->rmm->u_size + 1; i++)
-		if (rmesa->rmm->u_list[i].ptr &&
-		    ptr >= rmesa->rmm->u_list[i].ptr &&
-		    ptr <
-		    rmesa->rmm->u_list[i].ptr + rmesa->rmm->u_list[i].size)
-			break;
-
-	if (i < rmesa->rmm->u_size + 1)
-		return i;
-
-	fprintf(stderr, "%p failed\n", ptr);
-	return 0;
-}
-
-//#define MM_DEBUG
-int r300_mem_alloc(r300ContextPtr rmesa, int alignment, int size)
-{
-	drm_radeon_mem_alloc_t alloc;
-	int offset = 0, ret;
-	int i, free = -1;
-	int done_age;
-	drm_radeon_mem_free_t memfree;
-	int tries = 0;
-	static int bytes_wasted = 0, allocated = 0;
-
-	if (size < 4096)
-		bytes_wasted += 4096 - size;
-
-	allocated += size;
-
-#if 0
-	static int t = 0;
-	if (t != time(NULL)) {
-		t = time(NULL);
-		fprintf(stderr, "slots used %d, wasted %d kb, allocated %d\n",
-			rmesa->rmm->u_last, bytes_wasted / 1024,
-			allocated / 1024);
-	}
-#endif
-
-	memfree.region = RADEON_MEM_REGION_GART;
-
-      again:
-
-	done_age = radeonGetAge((radeonContextPtr) rmesa);
-
-	if (rmesa->rmm->u_last + 1 >= rmesa->rmm->u_size)
-		resize_u_list(rmesa);
-
-	for (i = rmesa->rmm->u_last + 1; i > 0; i--) {
-		if (rmesa->rmm->u_list[i].ptr == NULL) {
-			free = i;
-			continue;
-		}
-
-		if (rmesa->rmm->u_list[i].h_pending == 0 &&
-		    rmesa->rmm->u_list[i].pending
-		    && rmesa->rmm->u_list[i].age <= done_age) {
-			memfree.region_offset =
-			    (char *)rmesa->rmm->u_list[i].ptr -
-			    (char *)rmesa->radeon.radeonScreen->gartTextures.
-			    map;
-
-			ret =
-			    drmCommandWrite(rmesa->radeon.radeonScreen->
-					    driScreen->fd, DRM_RADEON_FREE,
-					    &memfree, sizeof(memfree));
-
-			if (ret) {
-				fprintf(stderr, "Failed to free at %p\n",
-					rmesa->rmm->u_list[i].ptr);
-				fprintf(stderr, "ret = %s\n", strerror(-ret));
-				exit(1);
-			} else {
-#ifdef MM_DEBUG
-				fprintf(stderr, "really freed %d at age %x\n",
-					i,
-					radeonGetAge((radeonContextPtr) rmesa));
-#endif
-				if (i == rmesa->rmm->u_last)
-					rmesa->rmm->u_last--;
-
-				if (rmesa->rmm->u_list[i].size < 4096)
-					bytes_wasted -=
-					    4096 - rmesa->rmm->u_list[i].size;
-
-				allocated -= rmesa->rmm->u_list[i].size;
-				rmesa->rmm->u_list[i].pending = 0;
-				rmesa->rmm->u_list[i].ptr = NULL;
-				free = i;
-			}
-		}
-	}
-	rmesa->rmm->u_head = i;
-
-	if (free == -1) {
-		WARN_ONCE("Ran out of slots!\n");
-		//usleep(100);
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-		tries++;
-		if (tries > 100) {
-			WARN_ONCE("Ran out of slots!\n");
-			exit(1);
-		}
-		goto again;
-	}
-
-	alloc.region = RADEON_MEM_REGION_GART;
-	alloc.alignment = alignment;
-	alloc.size = size;
-	alloc.region_offset = &offset;
-
-	ret =
-	    drmCommandWriteRead(rmesa->radeon.dri.fd, DRM_RADEON_ALLOC, &alloc,
-				sizeof(alloc));
-	if (ret) {
-#if 0
-		WARN_ONCE("Ran out of mem!\n");
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-		//usleep(100);
-		tries2++;
-		tries = 0;
-		if (tries2 > 100) {
-			WARN_ONCE("Ran out of GART memory!\n");
-			exit(1);
-		}
-		goto again;
-#else
-		WARN_ONCE
-		    ("Ran out of GART memory (for %d)!\nPlease consider adjusting GARTSize option.\n",
-		     size);
-		return 0;
-#endif
-	}
-
-	i = free;
-
-	if (i > rmesa->rmm->u_last)
-		rmesa->rmm->u_last = i;
-
-	rmesa->rmm->u_list[i].ptr =
-	    ((GLubyte *) rmesa->radeon.radeonScreen->gartTextures.map) + offset;
-	rmesa->rmm->u_list[i].size = size;
-	rmesa->rmm->u_list[i].age = 0;
-	//fprintf(stderr, "alloc %p at id %d\n", rmesa->rmm->u_list[i].ptr, i);
-
-#ifdef MM_DEBUG
-	fprintf(stderr, "allocated %d at age %x\n", i,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	return i;
-}
-
-void r300_mem_use(r300ContextPtr rmesa, int id)
-{
-	uint64_t ull;
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-	drm_r300_cmd_header_t *cmd;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (id == 0)
-		return;
-
-	cmd =
-	    (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa,
-						      2 + sizeof(ull) / 4,
-						      __FUNCTION__);
-	cmd[0].scratch.cmd_type = R300_CMD_SCRATCH;
-	cmd[0].scratch.reg = R300_MEM_SCRATCH;
-	cmd[0].scratch.n_bufs = 1;
-	cmd[0].scratch.flags = 0;
-	cmd++;
-
-	ull = (uint64_t) (intptr_t) & rmesa->rmm->u_list[id].age;
-	_mesa_memcpy(cmd, &ull, sizeof(ull));
-	cmd += sizeof(ull) / 4;
-
-	cmd[0].u = /*id */ 0;
-
-	LOCK_HARDWARE(&rmesa->radeon);	/* Protect from DRM. */
-	rmesa->rmm->u_list[id].h_pending++;
-	UNLOCK_HARDWARE(&rmesa->radeon);
-}
-
-unsigned long r300_mem_offset(r300ContextPtr rmesa, int id)
-{
-	unsigned long offset;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	offset = (char *)rmesa->rmm->u_list[id].ptr -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-	offset += rmesa->radeon.radeonScreen->gart_texture_offset;
-
-	return offset;
-}
-
-void *r300_mem_map(r300ContextPtr rmesa, int id, int access)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-	void *ptr;
-	int tries = 0;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (access == R300_MEM_R) {
-
-		if (rmesa->rmm->u_list[id].mapped == 1)
-			WARN_ONCE("buffer %d already mapped\n", id);
-
-		rmesa->rmm->u_list[id].mapped = 1;
-		ptr = r300_mem_ptr(rmesa, id);
-
-		return ptr;
-	}
-
-	if (rmesa->rmm->u_list[id].h_pending)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-	if (rmesa->rmm->u_list[id].h_pending) {
-		return NULL;
-	}
-
-	while (rmesa->rmm->u_list[id].age >
-	       radeonGetAge((radeonContextPtr) rmesa) && tries++ < 1000)
-		usleep(10);
-
-	if (tries >= 1000) {
-		fprintf(stderr, "Idling failed (%x vs %x)\n",
-			rmesa->rmm->u_list[id].age,
-			radeonGetAge((radeonContextPtr) rmesa));
-		return NULL;
-	}
-
-	if (rmesa->rmm->u_list[id].mapped == 1)
-		WARN_ONCE("buffer %d already mapped\n", id);
-
-	rmesa->rmm->u_list[id].mapped = 1;
-	ptr = r300_mem_ptr(rmesa, id);
-
-	return ptr;
-}
-
-void r300_mem_unmap(r300ContextPtr rmesa, int id)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (rmesa->rmm->u_list[id].mapped == 0)
-		WARN_ONCE("buffer %d not mapped\n", id);
-
-	rmesa->rmm->u_list[id].mapped = 0;
-}
-
-void r300_mem_free(r300ContextPtr rmesa, int id)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (id == 0)
-		return;
-
-	if (rmesa->rmm->u_list[id].ptr == NULL) {
-		WARN_ONCE("Not allocated!\n");
-		return;
-	}
-
-	if (rmesa->rmm->u_list[id].pending) {
-		WARN_ONCE("%p already pended!\n", rmesa->rmm->u_list[id].ptr);
-		return;
-	}
-
-	rmesa->rmm->u_list[id].pending = 1;
-}
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_mem.h b/src/mesa/drivers/dri/r300/r300_mem.h
deleted file mode 100644
index 625a7f6d8d..0000000000
--- a/src/mesa/drivers/dri/r300/r300_mem.h
+++ /dev/null
@@ -1,37 +0,0 @@
-#ifndef __R300_MEM_H__
-#define __R300_MEM_H__
-
-//#define R300_MEM_PDL 0
-#define R300_MEM_UL 1
-
-#define R300_MEM_R 1
-#define R300_MEM_W 2
-#define R300_MEM_RW (R300_MEM_R | R300_MEM_W)
-
-#define R300_MEM_SCRATCH 2
-
-struct r300_memory_manager {
-	struct {
-		void *ptr;
-		uint32_t size;
-		uint32_t age;
-		uint32_t h_pending;
-		int pending;
-		int mapped;
-	} *u_list;
-	int u_head, u_size, u_last;
-
-};
-
-extern void r300_mem_init(r300ContextPtr rmesa);
-extern void r300_mem_destroy(r300ContextPtr rmesa);
-extern void *r300_mem_ptr(r300ContextPtr rmesa, int id);
-extern int r300_mem_find(r300ContextPtr rmesa, void *ptr);
-extern int r300_mem_alloc(r300ContextPtr rmesa, int alignment, int size);
-extern void r300_mem_use(r300ContextPtr rmesa, int id);
-extern unsigned long r300_mem_offset(r300ContextPtr rmesa, int id);
-extern void *r300_mem_map(r300ContextPtr rmesa, int id, int access);
-extern void r300_mem_unmap(r300ContextPtr rmesa, int id);
-extern void r300_mem_free(r300ContextPtr rmesa, int id);
-
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index 8f1a6630d5..39b4b61a10 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -1128,6 +1128,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* SU Depth Offset value */
 #define R300_SU_DEPTH_OFFSET                0x42c4
 
+#define R300_SU_REG_DEST		    0x42c8
+#	define R300_RASTER_PIPE_SELECT_0	(1 << 0)
+#	define R300_RASTER_PIPE_SELECT_1	(1 << 1)
+#	define R300_RASTER_PIPE_SELECT_2	(1 << 2)
+#	define R300_RASTER_PIPE_SELECT_3	(1 << 3)
+#	define R300_RASTER_PIPE_SELECT_ALL	0xf
+
 
 /* BEGIN: Rasterization / Interpolators - many guesses */
 
@@ -1467,6 +1474,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_3D		   (1 << 25)
 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
 
+#	define R300_TX_FORMAT_GAMMA			(1 << 21)
+
 	/* gap */
 	/* Floating point formats */
 	/* Note - hardware supports both 16 and 32 bit floating point */
@@ -1531,6 +1540,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R500_SEL_FILTER4_TC3		 (3 << 18)
 
 #define R300_TX_OFFSET_0                    0x4540
+#define R300_TX_OFFSET_1                    0x4544
+#define R300_TX_OFFSET_2                    0x4548
+#define R300_TX_OFFSET_3                    0x454C
+#define R300_TX_OFFSET_4                    0x4550
+#define R300_TX_OFFSET_5                    0x4554
+#define R300_TX_OFFSET_6                    0x4558
+#define R300_TX_OFFSET_7                    0x455C
 	/* BEGIN: Guess from R200 */
 #       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
 #       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
@@ -2005,6 +2021,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R500_FG_ALPHA_VALUE                0x4be0
 #	define R500_FG_ALPHA_VALUE_MASK 0x0000ffff
 
+#define RV530_FG_ZBREG_DEST                 0x4be8
+#	define RV530_FG_ZBREG_DEST_PIPE_SELECT_0             (1 << 0)
+#	define RV530_FG_ZBREG_DEST_PIPE_SELECT_1             (1 << 1)
+#	define RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL           (3 << 0)
+
 /* gap */
 
 /* Fragment program parameters in 7.16 floating point */
@@ -2425,6 +2446,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* Z Buffer Clear Value */
 #define R300_ZB_DEPTHCLEARVALUE                  0x4f28
 
+#define R300_ZB_ZMASK_OFFSET                     0x4f30
+#define R300_ZB_ZMASK_PITCH                      0x4f34
+#define R300_ZB_ZMASK_WRINDEX                    0x4f38
+#define R300_ZB_ZMASK_DWORD                      0x4f3c
+#define R300_ZB_ZMASK_RDINDEX                    0x4f40
+
 /* Hierarchical Z Memory Offset */
 #define R300_ZB_HIZ_OFFSET                       0x4f44
 
@@ -2652,6 +2679,24 @@ enum {
 	PVS_SRC_ADDR_MODE_1_SHIFT	= 32,
 };
 
+
+#define PVS_OP_DST_OPERAND(opcode, math_inst, macro_inst, reg_index, reg_writemask, reg_class)	\
+	 (((opcode & PVS_DST_OPCODE_MASK) << PVS_DST_OPCODE_SHIFT)	\
+	 | ((math_inst & PVS_DST_MATH_INST_MASK) << PVS_DST_MATH_INST_SHIFT)	\
+	 | ((macro_inst & PVS_DST_MACRO_INST_MASK) << PVS_DST_MACRO_INST_SHIFT)	\
+	 | ((reg_index & PVS_DST_OFFSET_MASK) << PVS_DST_OFFSET_SHIFT)	\
+	 | ((reg_writemask & 0xf) << PVS_DST_WE_X_SHIFT)	/* X Y Z W */	\
+	 | ((reg_class & PVS_DST_REG_TYPE_MASK) << PVS_DST_REG_TYPE_SHIFT))
+
+#define PVS_SRC_OPERAND(in_reg_index, comp_x, comp_y, comp_z, comp_w, reg_class, negate)	\
+	(((in_reg_index & PVS_SRC_OFFSET_MASK) << PVS_SRC_OFFSET_SHIFT)				\
+	 | ((comp_x & PVS_SRC_SWIZZLE_X_MASK) << PVS_SRC_SWIZZLE_X_SHIFT)			\
+	 | ((comp_y & PVS_SRC_SWIZZLE_Y_MASK) << PVS_SRC_SWIZZLE_Y_SHIFT)			\
+	 | ((comp_z & PVS_SRC_SWIZZLE_Z_MASK) << PVS_SRC_SWIZZLE_Z_SHIFT)			\
+	 | ((comp_w & PVS_SRC_SWIZZLE_W_MASK) << PVS_SRC_SWIZZLE_W_SHIFT)			\
+	 | ((negate & 0xf) << PVS_SRC_MODIFIER_X_SHIFT)	/* X Y Z W */				\
+	 | ((reg_class & PVS_SRC_REG_TYPE_MASK) << PVS_SRC_REG_TYPE_SHIFT))
+
 /*\}*/
 
 /* BEGIN: Packet 3 commands */
@@ -3165,6 +3210,9 @@ enum {
 #   define R300_W_SRC_RAS				(1 << 2)
 
 
+/* Packet0 field ordering to write all values to the same reg */
+#define RADEON_ONE_REG_WR        (1 << 15)
+
 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
  * Two parameter dwords:
  * 0. VAP_VTX_FMT: The first parameter is not written to hardware
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 16ce4a1199..b5ddfdc9f8 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -50,6 +50,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * no bugs...
  */
 
+#include "r300_render.h"
+
 #include "main/glheader.h"
 #include "main/state.h"
 #include "main/imports.h"
@@ -62,20 +64,19 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "vbo/vbo.h"
+#include "vbo/vbo_split.h"
 #include "tnl/tnl.h"
 #include "tnl/t_vp_build.h"
 #include "radeon_reg.h"
 #include "radeon_macros.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
-extern int future_hw_tcl_on;
+#include "r300_fragprog_common.h"
+#include "r300_swtcl.h"
 
 /**
  * \brief Convert a OpenGL primitive type into a R300 primitive type.
@@ -172,96 +173,167 @@ int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts)
+static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type, int offset)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_dma_region *rvb = &rmesa->state.elt_dma;
-	void *out;
-
-	if (r300IsGartMemory(rmesa, elts, n_elts * 4)) {
-		rvb->address = rmesa->radeon.radeonScreen->gartTextures.map;
-		rvb->start = ((char *)elts) - rvb->address;
-		rvb->aos_offset =
-		    rmesa->radeon.radeonScreen->gart_texture_offset +
-		    rvb->start;
-		return;
-	} else if (r300IsGartMemory(rmesa, elts, 1)) {
-		WARN_ONCE("Pointer not within GART memory!\n");
-		_mesa_exit(-1);
+	BATCH_LOCALS(&rmesa->radeon);
+	int size;
+
+	/* offset is in indices */
+	BEGIN_BATCH(10);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
+	if (rmesa->ind_buf.is_32bit) {
+		/* convert to bytes */
+		offset *= 4;
+		size = vertex_count;
+		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+		  (vertex_count << 16) | type |
+		  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
+	} else {
+		/* convert to bytes */
+		offset *= 2;
+		size = (vertex_count + 1) >> 1;
+		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+		   (vertex_count << 16) | type);
 	}
 
-	r300AllocDmaRegion(rmesa, rvb, n_elts * 4, 4);
-	rvb->aos_offset = GET_START(rvb);
-
-	out = rvb->address + rvb->start;
-	memcpy(out, elts, n_elts * 4);
-}
-
-static void r300FireEB(r300ContextPtr rmesa, unsigned long addr,
-		       int vertex_count, int type)
-{
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count << 16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-
-	start_packet3(CP_PACKET3(R300_PACKET3_INDX_BUFFER, 2), 2);
-	e32(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
-	    (R300_VAP_PORT_IDX0 >> 2));
-	e32(addr);
-	e32(vertex_count);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+		OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+				 (R300_VAP_PORT_IDX0 >> 2));
+		OUT_BATCH_RELOC(0, rmesa->ind_buf.bo, rmesa->ind_buf.bo_offset + offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+		OUT_BATCH(size);
+	} else {
+		OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+		OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+				 (R300_VAP_PORT_IDX0 >> 2));
+		OUT_BATCH(rmesa->ind_buf.bo_offset + offset);
+		OUT_BATCH(size);
+		radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+				      rmesa->ind_buf.bo, RADEON_GEM_DOMAIN_GTT, 0, 0);
+	}
+	END_BATCH();
 }
 
 static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 {
+	BATCH_LOCALS(&rmesa->radeon);
+	uint32_t voffset;
 	int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
 	int i;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
 
-	if (RADEON_DEBUG & DEBUG_VERTS)
+	if (RADEON_DEBUG & RADEON_VERTS)
 		fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
 			offset);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1), sz - 1);
-	e32(nr);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH(sz+2+(nr * 2));
+		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+		OUT_BATCH(nr);
+
+		for (i = 0; i + 1 < nr; i += 2) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+				  (rmesa->radeon.tcl.aos[i].stride << 8) |
+				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[i].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+			  offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[i+1].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+		}
 
-	for (i = 0; i + 1 < nr; i += 2) {
-		e32((rmesa->state.aos[i].aos_size << 0) |
-		    (rmesa->state.aos[i].aos_stride << 8) |
-		    (rmesa->state.aos[i + 1].aos_size << 16) |
-		    (rmesa->state.aos[i + 1].aos_stride << 24));
+		if (nr & 1) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+				  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[nr - 1].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+		}
+		END_BATCH();
+	} else {
 
-		e32(rmesa->state.aos[i].aos_offset + offset * 4 * rmesa->state.aos[i].aos_stride);
-		e32(rmesa->state.aos[i + 1].aos_offset + offset * 4 * rmesa->state.aos[i + 1].aos_stride);
-	}
+		BEGIN_BATCH(sz+2+(nr * 2));
+		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+		OUT_BATCH(nr);
+
+		for (i = 0; i + 1 < nr; i += 2) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+				  (rmesa->radeon.tcl.aos[i].stride << 8) |
+				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			OUT_BATCH(voffset);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			OUT_BATCH(voffset);
+		}
 
-	if (nr & 1) {
-		e32((rmesa->state.aos[nr - 1].aos_size << 0) |
-		    (rmesa->state.aos[nr - 1].aos_stride << 8));
-		e32(rmesa->state.aos[nr - 1].aos_offset + offset * 4 * rmesa->state.aos[nr - 1].aos_stride);
+		if (nr & 1) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+			  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			OUT_BATCH(voffset);
+		}
+		for (i = 0; i + 1 < nr; i += 2) {
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[i+0].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[i+1].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+		}
+		if (nr & 1) {
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[nr-1].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+		}
+		END_BATCH();
 	}
+
 }
 
 static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(&rmesa->radeon);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
+        r300_emit_scissor(rmesa->radeon.glCtx);
+	BEGIN_BATCH(3);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
+	END_BATCH();
 }
 
-static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
-				   int start, int end, int prim)
+void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 {
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
 
 	type = r300PrimitiveType(rmesa, prim);
 	num_verts = r300NumVerts(rmesa, end - start, prim);
@@ -269,194 +341,161 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	if (type < 0 || num_verts <= 0)
 		return;
 
-	if (vb->Elts) {
-		if (num_verts > 65535) {
-			/* not implemented yet */
-			WARN_ONCE("Too many elts\n");
+	if (rmesa->ind_buf.bo) {
+		GLuint first, incr, offset = 0;
+
+		if (!split_prim_inplace(prim & PRIM_MODE_MASK, &first, &incr) &&
+			num_verts > 65500) {
+			WARN_ONCE("Fixme: can't handle spliting prim %d\n", prim);
 			return;
 		}
-		/* Note: The following is incorrect, but it's the best I can do
-		 * without a major refactoring of how DMA memory is handled.
-		 * The problem: Ensuring that both vertex arrays *and* index
-		 * arrays are at the right position, and then ensuring that
-		 * the LOAD_VBPNTR, DRAW_INDX and INDX_BUFFER packets are emitted
-		 * at once.
-		 *
-		 * So why is the following incorrect? Well, it seems like
-		 * allocating the index array might actually evict the vertex
-		 * arrays. *sigh*
-		 */
-		r300EmitElts(ctx, vb->Elts, num_verts);
-		r300EmitAOS(rmesa, rmesa->state.aos_count, start);
-		r300FireEB(rmesa, rmesa->state.elt_dma.aos_offset, num_verts, type);
-	} else {
-		r300EmitAOS(rmesa, rmesa->state.aos_count, start);
-		r300FireAOS(rmesa, num_verts, type);
-	}
-}
-
-static GLboolean r300RunRender(GLcontext * ctx,
-			       struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	r300UpdateShaders(rmesa);
-	if (r300EmitArrays(ctx))
-		return GL_TRUE;
 
-	r300UpdateShaderStates(rmesa);
 
-	r300EmitCacheFlush(rmesa);
-	r300EmitState(rmesa);
-
-	for (i = 0; i < vb->PrimitiveCount; i++) {
-		GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
-		GLuint start = vb->Primitive[i].start;
-		GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
-		r300RunRenderPrimitive(rmesa, ctx, start, end, prim);
-	}
-
-	r300EmitCacheFlush(rmesa);
+		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, 0);
+		if (rmesa->radeon.radeonScreen->kernel_mm) {
+			BEGIN_BATCH_NO_AUTOSTATE(2);
+			OUT_BATCH_REGSEQ(R300_VAP_VF_MAX_VTX_INDX, 1);
+			OUT_BATCH(rmesa->radeon.tcl.aos[0].count);
+			END_BATCH();
+		}
 
-#ifdef USER_BUFFERS
-	r300UseArrays(ctx);
-#endif
+		r300_emit_scissor(rmesa->radeon.glCtx);
+		while (num_verts > 0) {
+			int nr;
+			int align;
+
+			nr = MIN2(num_verts, 65535);
+			nr -= (nr - first) % incr;
+
+			/* get alignment for IB correct */
+			if (nr != num_verts) {
+				do {
+				    align = nr * (rmesa->ind_buf.is_32bit ? 4 : 2);
+				    if (align % 4)
+					nr -= incr;
+				} while(align % 4);
+				if (nr <= 0) {
+					WARN_ONCE("did the impossible happen? we never aligned nr to dword\n");
+					return;
+				}
+					
+			}
+			r300FireEB(rmesa, nr, type, offset);
 
-	r300ReleaseArrays(ctx);
+			num_verts -= nr;
+			offset += nr;
+		}
 
-	return GL_FALSE;
-}
+	} else {
+		GLuint first, incr, offset = 0;
 
-#define FALLBACK_IF(expr)						\
-	do {								\
-		if (expr) {						\
-			if (1 || RADEON_DEBUG & DEBUG_FALLBACKS)	\
-				WARN_ONCE("Software fallback:%s\n",	\
-					  #expr);			\
-			return R300_FALLBACK_RAST;			\
-		}							\
-	} while(0)
-
-static int r300Fallback(GLcontext * ctx)
-{
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	const unsigned back = ctx->Stencil._BackFace;
-
-	/* Do we need to use new-style shaders?
-	 * Also is there a better way to do this? */
-	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-		struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r500TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
-			}
+		if (!split_prim_inplace(prim & PRIM_MODE_MASK, &first, &incr) &&
+			num_verts > 65535) {
+			WARN_ONCE("Fixme: can't handle spliting prim %d\n", prim);
+			return;
 		}
-	} else {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r300TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
-			}
+		r300_emit_scissor(rmesa->radeon.glCtx);
+		while (num_verts > 0) {
+			int nr;
+			nr = MIN2(num_verts, 65535);
+			nr -= (nr - first) % incr;
+			r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start + offset);
+			r300FireAOS(rmesa, nr, type);
+			num_verts -= nr;
+			offset += nr;
 		}
 	}
-
-	FALLBACK_IF(ctx->RenderMode != GL_RENDER);
-
-	/* If GL_EXT_stencil_two_side is disabled, this fallback check can
-	 * be removed.
-	 */
-	FALLBACK_IF(ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
-		    || ctx->Stencil.ValueMask[0] !=
-		    ctx->Stencil.ValueMask[back]
-		    || ctx->Stencil.WriteMask[0] !=
-		    ctx->Stencil.WriteMask[back]);
-
-	if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite)
-		FALLBACK_IF(ctx->Point.PointSprite);
-
-	if (!r300->disable_lowimpact_fallback) {
-		FALLBACK_IF(ctx->Polygon.StippleFlag);
-		FALLBACK_IF(ctx->Multisample._Enabled);
-		FALLBACK_IF(ctx->Line.StippleFlag);
-		FALLBACK_IF(ctx->Line.SmoothFlag);
-		FALLBACK_IF(ctx->Point.SmoothFlag);
-	}
-
-	return R300_FALLBACK_NONE;
+	COMMIT_BATCH();
 }
 
-static GLboolean r300RunNonTCLRender(GLcontext * ctx,
-				     struct tnl_pipeline_stage *stage)
+static const char *getFallbackString(uint32_t bit)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (r300Fallback(ctx) >= R300_FALLBACK_RAST)
-		return GL_TRUE;
-
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
- 	        return GL_TRUE;
-
-	return r300RunRender(ctx, stage);
+	switch (bit) {
+		case R300_FALLBACK_VERTEX_PROGRAM :
+			return "vertex program";
+		case R300_FALLBACK_LINE_SMOOTH:
+			return "smooth lines";
+		case R300_FALLBACK_POINT_SMOOTH:
+			return "smooth points";
+		case R300_FALLBACK_POLYGON_SMOOTH:
+			return "smooth polygons";
+		case R300_FALLBACK_LINE_STIPPLE:
+			return "line stipple";
+		case R300_FALLBACK_POLYGON_STIPPLE:
+			return "polygon stipple";
+		case R300_FALLBACK_STENCIL_TWOSIDE:
+			return "two-sided stencil";
+		case R300_FALLBACK_RENDER_MODE:
+			return "render mode != GL_RENDER";
+		case R300_FALLBACK_FRAGMENT_PROGRAM:
+			return "fragment program";
+		case R300_FALLBACK_AOS_LIMIT:
+			return "aos limit";
+		case R300_FALLBACK_INVALID_BUFFERS:
+			return "invalid buffers";
+		default:
+			return "unknown";
+	}
 }
 
-static GLboolean r300RunTCLRender(GLcontext * ctx,
-				  struct tnl_pipeline_stage *stage)
+void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 {
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_vertex_program *vp;
+	uint32_t old_fallback = rmesa->fallback;
+	static uint32_t fallback_warn = 0;
+
+	if (mode) {
+		if ((fallback_warn & bit) == 0) {
+			if (RADEON_DEBUG & RADEON_FALLBACKS)
+				_mesa_fprintf(stderr, "WARNING! Falling back to software for %s\n", getFallbackString(bit));
+			fallback_warn |= bit;
+		}
+		rmesa->fallback |= bit;
+
+		/* update only if we change from no tcl fallbacks to some tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if (((old_fallback & R300_TCL_FALLBACK_MASK) == 0) &&
+				((bit & R300_TCL_FALLBACK_MASK) > 0)) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] |= R300_VAP_TCL_BYPASS;
+			}
+		}
 
-	hw_tcl_on = future_hw_tcl_on;
+		/* update only if we change from no raster fallbacks to some raster fallbacks */
+		if (((old_fallback & R300_RASTER_FALLBACK_MASK) == 0) &&
+			((bit & R300_RASTER_FALLBACK_MASK) > 0)) {
 
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
+			radeon_firevertices(&rmesa->radeon);
+			rmesa->radeon.swtcl.RenderIndex = ~0;
+			_swsetup_Wakeup( ctx );
+		}
+	} else {
+		rmesa->fallback &= ~bit;
 
-	if (hw_tcl_on == GL_FALSE)
-		return GL_TRUE;
+		/* update only if we have disabled all tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] &= ~R300_VAP_TCL_BYPASS;
+			}
+		}
 
-	if (r300Fallback(ctx) >= R300_FALLBACK_TCL) {
-		hw_tcl_on = GL_FALSE;
-		return GL_TRUE;
-	}
+		/* update only if we have disabled all raster fallbacks */
+		if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+			_swrast_flush( ctx );
 
-	r300UpdateShaders(rmesa);
+			tnl->Driver.Render.Start = r300RenderStart;
+			tnl->Driver.Render.Finish = r300RenderFinish;
+			tnl->Driver.Render.PrimitiveNotify = r300RenderPrimitive;
+			tnl->Driver.Render.ResetLineStipple = r300ResetLineStipple;
+			tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+			tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+			tnl->Driver.Render.Interp = _tnl_interp;
 
-	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	if (vp->native == GL_FALSE) {
-		hw_tcl_on = GL_FALSE;
-		return GL_TRUE;
+			_tnl_invalidate_vertex_state( ctx, ~0 );
+			_tnl_invalidate_vertices( ctx, ~0 );
+		}
 	}
 
-	return r300RunRender(ctx, stage);
 }
-
-const struct tnl_pipeline_stage _r300_render_stage = {
-	"r300 Hardware Rasterization",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunNonTCLRender
-};
-
-const struct tnl_pipeline_stage _r300_tcl_stage = {
-	"r300 Hardware Transform, Clipping and Lighting",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunTCLRender
-};
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
new file mode 100644
index 0000000000..ec785474a6
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -0,0 +1,69 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __R300_RENDER_H__
+#define __R300_RENDER_H__
+
+#include "main/mtypes.h"
+
+#define R300_FALLBACK_VERTEX_PROGRAM    (1 << 0)
+#define R300_TCL_FALLBACK_MASK           0x0000ffff
+
+#define R300_FALLBACK_LINE_SMOOTH       (1 << 16)
+#define R300_FALLBACK_POINT_SMOOTH      (1 << 17)
+#define R300_FALLBACK_POLYGON_SMOOTH    (1 << 18)
+#define R300_FALLBACK_LINE_STIPPLE      (1 << 19)
+#define R300_FALLBACK_POLYGON_STIPPLE   (1 << 20)
+#define R300_FALLBACK_STENCIL_TWOSIDE   (1 << 21)
+#define R300_FALLBACK_RENDER_MODE       (1 << 22)
+#define R300_FALLBACK_FRAGMENT_PROGRAM  (1 << 23)
+#define R300_FALLBACK_AOS_LIMIT         (1 << 30)
+#define R300_FALLBACK_INVALID_BUFFERS   (1 << 31)
+#define R300_RASTER_FALLBACK_MASK        0xffff0000
+
+#define MASK_XYZW (R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W)
+#define MASK_X R300_WRITE_ENA_X
+#define MASK_Y R300_WRITE_ENA_Y
+#define MASK_Z R300_WRITE_ENA_Z
+#define MASK_W R300_WRITE_ENA_W
+
+#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
+    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
+    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
+    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
+    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
+    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
+#error Cannot change these!
+#endif
+
+extern const struct tnl_pipeline_stage _r300_render_stage;
+
+extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
+
+extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index f30fd986e0..a4f9db13ec 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -1,47 +1,79 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
 
 #include "main/glheader.h"
 
 #include "shader/program.h"
 #include "tnl/tnl.h"
 #include "r300_context.h"
-#include "r300_fragprog.h"
+#include "r300_fragprog_common.h"
+
+static void freeFragProgCache(GLcontext *ctx, struct r300_fragment_program_cont *cache)
+{
+	struct r300_fragment_program *tmp, *fp = cache->progs;
+
+	while (fp) {
+		tmp = fp->next;
+		rc_constants_destroy(&fp->code.constants);
+		_mesa_free(fp);
+		fp = tmp;
+	}
+}
+
+static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *cache)
+{
+	struct r300_vertex_program *tmp, *vp = cache->progs;
+
+	while (vp) {
+		tmp = vp->next;
+		rc_constants_destroy(&vp->code.constants);
+		_mesa_reference_vertprog(ctx, &vp->Base, NULL);
+		_mesa_free(vp);
+		vp = tmp;
+	}
+}
 
 static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 					 GLuint id)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program_cont *vp;
-	struct r300_fragment_program *r300_fp;
-	struct r500_fragment_program *r500_fp;
+	struct r300_fragment_program_cont *fp;
 
 	switch (target) {
 	case GL_VERTEX_STATE_PROGRAM_NV:
 	case GL_VERTEX_PROGRAM_ARB:
 		vp = CALLOC_STRUCT(r300_vertex_program_cont);
-		return _mesa_init_vertex_program(ctx, &vp->mesa_program,
-						 target, id);
-	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			r500_fp->ctx = ctx;
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
+		return _mesa_init_vertex_program(ctx, &vp->mesa_program, target, id);
 
 	case GL_FRAGMENT_PROGRAM_NV:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
+	case GL_FRAGMENT_PROGRAM_ARB:
+		fp = CALLOC_STRUCT(r300_fragment_program_cont);
+		return _mesa_init_fragment_program(ctx, &fp->Base, target, id);
+
 	default:
 		_mesa_problem(ctx, "Bad target in r300NewProgram");
 	}
@@ -51,26 +83,35 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 
 static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog)
 {
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
+
+	switch (prog->Target) {
+		case GL_VERTEX_PROGRAM_ARB:
+			freeVertProgCache(ctx, vp);
+			break;
+		case GL_FRAGMENT_PROGRAM_ARB:
+			freeFragProgCache(ctx, fp);
+			break;
+	}
+
 	_mesa_delete_program(ctx, prog);
 }
 
 static void
 r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_vertex_program_cont *vp = (void *)prog;
-	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)prog;
-	struct r500_fragment_program *r500_fp = (struct r500_fragment_program *)prog;
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
 
 	switch (target) {
 	case GL_VERTEX_PROGRAM_ARB:
+		freeVertProgCache(ctx, vp);
 		vp->progs = NULL;
 		break;
 	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-			r500_fp->translated = GL_FALSE;
-		else
-			r300_fp->translated = GL_FALSE;
+		freeFragProgCache(ctx, fp);
+		fp->progs = NULL;
 		break;
 	}
 
@@ -81,7 +122,15 @@ r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 static GLboolean
 r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	return GL_TRUE;
+	if (target == GL_FRAGMENT_PROGRAM_ARB) {
+		struct r300_fragment_program *fp = r300SelectAndTranslateFragmentShader(ctx);
+
+		return !fp->error;
+	} else {
+		struct r300_vertex_program *vp = r300SelectAndTranslateVertexShader(ctx);
+
+		return !vp->error;
+	}
 }
 
 void r300InitShaderFuncs(struct dd_function_table *functions)
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 79f0b3625c..9301543d38 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/context.h"
 #include "main/dd.h"
+#include "main/framebuffer.h"
 #include "main/simple_list.h"
 #include "main/api_arrayelt.h"
 #include "main/texformat.h"
@@ -52,22 +53,20 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "shader/prog_statevars.h"
 #include "vbo/vbo.h"
 #include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
 
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
 #include "r300_tex.h"
+#include "r300_fragprog_common.h"
+#include "r300_render.h"
+#include "r300_vertprog.h"
 
 #include "drirenderbuffer.h"
 
-extern int future_hw_tcl_on;
-extern void _tnl_UpdateFixedFunctionProgram(GLcontext * ctx);
-
 static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4])
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
@@ -366,12 +365,13 @@ static void r300ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 	GLint *ip;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!rmesa->options.hw_tcl_enabled)
 			return;
 
 	p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
 	ip = (GLint *)ctx->Transform._ClipUserPlane[p];
 
+	R300_STATECHANGE( rmesa, vap_flush );
 	R300_STATECHANGE( rmesa, vpucp[p] );
 	rmesa->hw.vpucp[p].cmd[R300_VPUCP_X] = ip[0];
 	rmesa->hw.vpucp[p].cmd[R300_VPUCP_Y] = ip[1];
@@ -385,7 +385,7 @@ static void r300SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state)
 	GLuint p;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!r300->options.hw_tcl_enabled)
 		return;
 
 	p = cap - GL_CLIP_PLANE0;
@@ -433,6 +433,10 @@ static void r300UpdateCulling(GLcontext * ctx)
 		break;
 	}
 
+	/* Winding is inverted when rendering to FBO */
+	if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+		val ^= R300_FRONT_FACE_CW;
+
 	R300_STATECHANGE(r300, cul);
 	r300->hw.cul.cmd[R300_CUL_CULL] = val;
 }
@@ -453,22 +457,14 @@ static GLboolean current_fragment_program_writes_depth(GLcontext* ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
-	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-			(char *)ctx->FragmentProgram._Current;
-		return (fp && fp->WritesDepth);
-	} else {
-		struct r500_fragment_program* fp =
-			(struct r500_fragment_program*)(char*)
-			ctx->FragmentProgram._Current;
-		return (fp && fp->writes_depth);
-	}
+	return ctx->FragmentProgram._Current && r300->selected_fp->code.writes_depth;
 }
 
 static void r300SetEarlyZState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	GLuint topZ = R300_ZTOP_ENABLE;
+	GLuint w_fmt, fgdepthsrc;
 
 	if (ctx->Color.AlphaEnabled && ctx->Color.AlphaFunc != GL_ALWAYS)
 		topZ = R300_ZTOP_DISABLE;
@@ -476,6 +472,8 @@ static void r300SetEarlyZState(GLcontext * ctx)
 		topZ = R300_ZTOP_DISABLE;
 	else if (ctx->FragmentProgram._Current && ctx->FragmentProgram._Current->UsesKill)
 		topZ = R300_ZTOP_DISABLE;
+	else if (r300->radeon.query.current)
+		topZ = R300_ZTOP_DISABLE;
 
 	if (topZ != r300->hw.zstencil_format.cmd[2]) {
 		/* Note: This completely reemits the stencil format.
@@ -485,6 +483,26 @@ static void r300SetEarlyZState(GLcontext * ctx)
 		R300_STATECHANGE(r300, zstencil_format);
 		r300->hw.zstencil_format.cmd[2] = topZ;
 	}
+
+	/* w_fmt value is set to get best performance
+	* see p.130 R5xx 3D acceleration guide v1.3 */
+	if (current_fragment_program_writes_depth(ctx)) {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
+		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
+	} else {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
+		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
+	}
+
+	if (w_fmt != r300->hw.us_out_fmt.cmd[5]) {
+		R300_STATECHANGE(r300, us_out_fmt);
+		r300->hw.us_out_fmt.cmd[5] = w_fmt;
+	}
+
+	if (fgdepthsrc != r300->hw.fg_depth_src.cmd[1]) {
+		R300_STATECHANGE(r300, fg_depth_src);
+		r300->hw.fg_depth_src.cmd[1] = fgdepthsrc;
+	}
 }
 
 static void r300SetAlphaState(GLcontext * ctx)
@@ -535,8 +553,6 @@ static void r300SetAlphaState(GLcontext * ctx)
 	R300_STATECHANGE(r300, at);
 	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc;
 	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
-
-	r300SetEarlyZState(ctx);
 }
 
 static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref)
@@ -584,15 +600,35 @@ static void r300SetDepthState(GLcontext * ctx)
 		r300->hw.zs.cmd[R300_ZS_CNTL_1] |=
 		    translate_func(ctx->Depth.Func) << R300_Z_FUNC_SHIFT;
 	}
+}
 
-	r300SetEarlyZState(ctx);
+static void r300CatchStencilFallback(GLcontext *ctx)
+{
+	const unsigned back = ctx->Stencil._BackFace;
+
+	if (ctx->Stencil._Enabled && (ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
+		|| ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[back]
+		|| ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[back])) {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_TRUE);
+	} else {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_FALSE);
+	}
 }
 
 static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLboolean hw_stencil = GL_FALSE;
 
-	if (r300->state.stencil.hw_stencil) {
+	r300CatchStencilFallback(ctx);
+
+	if (ctx->DrawBuffer) {
+		struct radeon_renderbuffer *rrbStencil
+			= radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+		hw_stencil = (rrbStencil && rrbStencil->bo);
+	}
+
+	if (hw_stencil) {
 		R300_STATECHANGE(r300, zs);
 		if (state) {
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] |=
@@ -601,10 +637,6 @@ static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] &=
 			    ~R300_STENCIL_ENABLE;
 		}
-	} else {
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_STENCIL, state);
-#endif
 	}
 }
 
@@ -737,7 +769,12 @@ static void r300ColorMask(GLcontext * ctx,
 static void r300PointSize(GLcontext * ctx, GLfloat size)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        /* same size limits for AA, non-AA points */
+
+	/* We need to clamp to user defined range here, because
+	 * the HW clamping happens only for per vertex point size. */
+	size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize);
+
+	/* same size limits for AA, non-AA points */
 	size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize);
 
 	R300_STATECHANGE(r300, ps);
@@ -830,29 +867,33 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode)
 
 	R300_STATECHANGE(rmesa, shade);
 	rmesa->hw.shade.cmd[1] = 0x00000002;
+	R300_STATECHANGE(rmesa, shade2);
 	switch (mode) {
 	case GL_FLAT:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_FLAT;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_FLAT;
 		break;
 	case GL_SMOOTH:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_SMOOTH;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_SMOOTH;
 		break;
 	default:
 		return;
 	}
-	rmesa->hw.shade.cmd[3] = 0x00000000;
-	rmesa->hw.shade.cmd[4] = 0x00000000;
+	rmesa->hw.shade2.cmd[2] = 0x00000000;
+	rmesa->hw.shade2.cmd[3] = 0x00000000;
 }
 
 static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 				    GLenum func, GLint ref, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint refmask =
-	    ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
-	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
-	const unsigned back = ctx->Stencil._BackFace;
+	GLuint refmask;
 	GLuint flag;
+	const unsigned back = ctx->Stencil._BackFace;
+
+	r300CatchStencilFallback(ctx);
+
+	refmask = ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
+	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
 
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_0] |= R300_STENCIL_FRONT_BACK;
@@ -880,6 +921,8 @@ static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &=
 	    ~(R300_STENCILREF_MASK <<
@@ -896,6 +939,8 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	const unsigned back = ctx->Stencil._BackFace;
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	/* It is easier to mask what's left.. */
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &=
@@ -924,28 +969,32 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
  * Window position and viewport transformation
  */
 
-/*
- * To correctly position primitives:
- */
-#define SUBPIXEL_X 0.125
-#define SUBPIXEL_Y 0.125
-
 static void r300UpdateWindow(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
 	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+	const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
+	GLfloat y_scale, y_bias;
+
+	if (render_to_fbo) {
+		y_scale = 1.0;
+		y_bias = 0;
+	} else {
+		y_scale = -1.0;
+		y_bias = yoffset;
+	}
 
 	GLfloat sx = v[MAT_SX];
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat sy = -v[MAT_SY];
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
-	GLfloat sz = v[MAT_SZ] * rmesa->state.depth.scale;
-	GLfloat tz = v[MAT_TZ] * rmesa->state.depth.scale;
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat sy = v[MAT_SY] * y_scale;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias;
+	GLfloat sz = v[MAT_SZ] * depthScale;
+	GLfloat tz = v[MAT_TZ] * depthScale;
 
-	R300_FIREVERTICES(rmesa);
 	R300_STATECHANGE(rmesa, vpt);
 
 	rmesa->hw.vpt.cmd[R300_VPT_XSCALE] = r300PackFloat32(sx);
@@ -964,6 +1013,8 @@ static void r300Viewport(GLcontext * ctx, GLint x, GLint y,
 	 * values, or keep the originals hanging around.
 	 */
 	r300UpdateWindow(ctx);
+
+	radeon_viewport(ctx, x, y, width, height);
 }
 
 static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
@@ -974,13 +1025,13 @@ static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
 void r300UpdateViewportOffset(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = ((radeonContextPtr) rmesa)->dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = (GLfloat) dPriv->x;
 	GLfloat yoffset = (GLfloat) dPriv->y + dPriv->h;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
 
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat ty = (-v[MAT_TY]) + yoffset;
 
 	if (rmesa->hw.vpt.cmd[R300_VPT_XOFFSET] != r300PackFloat32(tx) ||
 	    rmesa->hw.vpt.cmd[R300_VPT_YOFFSET] != r300PackFloat32(ty)) {
@@ -997,138 +1048,26 @@ void r300UpdateViewportOffset(GLcontext * ctx)
 }
 
 /**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void r300UpdateDrawBuffer(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	r300ContextPtr r300 = rmesa;
-	struct gl_framebuffer *fb = ctx->DrawBuffer;
-	driRenderbuffer *drb;
-
-	if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-		/* draw to front */
-		drb =
-		    (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].
-		    Renderbuffer;
-	} else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-		/* draw to back */
-		drb =
-		    (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].
-		    Renderbuffer;
-	} else {
-		/* drawing to multiple buffers, or none */
-		return;
-	}
-
-	assert(drb);
-	assert(drb->flippedPitch);
-
-	R300_STATECHANGE(rmesa, cb);
-
-	r300->hw.cb.cmd[R300_CB_OFFSET] = drb->flippedOffset +	//r300->radeon.state.color.drawOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.cb.cmd[R300_CB_PITCH] = drb->flippedPitch;	//r300->radeon.state.color.drawPitch;
-
-	if (r300->radeon.radeonScreen->cpp == 4)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-
-	if (r300->radeon.sarea->tiling_enabled)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-#if 0
-	R200_STATECHANGE(rmesa, ctx);
-
-	/* Note: we used the (possibly) page-flipped values */
-	rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-	    = ((drb->flippedOffset + rmesa->r200Screen->fbLocation)
-	       & R200_COLOROFFSET_MASK);
-	rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-
-	if (rmesa->sarea->tiling_enabled) {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
-		    R200_COLOR_TILE_ENABLE;
-	}
-#endif
-}
-
-static void
-r300FetchStateParameter(GLcontext * ctx,
-			const gl_state_index state[STATE_LENGTH],
-			GLfloat * value)
-{
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
-
-	switch (state[0]) {
-	case STATE_INTERNAL:
-		switch (state[1]) {
-		case STATE_R300_WINDOW_DIMENSION:
-			value[0] = r300->radeon.dri.drawable->w * 0.5f;	/* width*0.5 */
-			value[1] = r300->radeon.dri.drawable->h * 0.5f;	/* height*0.5 */
-			value[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
-			value[3] = 1.0F;	/* not used */
-			break;
-
-		case STATE_R300_TEXRECT_FACTOR:{
-				struct gl_texture_object *t =
-				    ctx->Texture.Unit[state[2]].CurrentTex[TEXTURE_RECT_INDEX];
-
-				if (t && t->Image[0][t->BaseLevel]) {
-					struct gl_texture_image *image =
-					    t->Image[0][t->BaseLevel];
-					value[0] = 1.0 / image->Width2;
-					value[1] = 1.0 / image->Height2;
-				} else {
-					value[0] = 1.0;
-					value[1] = 1.0;
-				}
-				value[2] = 1.0;
-				value[3] = 1.0;
-				break;
-			}
-
-		default:
-			break;
-		}
-		break;
-
-	default:
-		break;
-	}
-}
-
-/**
  * Update R300's own internal state parameters.
  * For now just STATE_R300_WINDOW_DIMENSION
  */
-void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
+static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 {
-	struct r300_fragment_program *fp;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct gl_program_parameter_list *paramList;
-	GLuint i;
 
-	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
-	fp = (struct r300_fragment_program *)ctx->FragmentProgram._Current;
-	if (!fp)
+	if (!ctx->FragmentProgram._Current || !rmesa->selected_fp)
 		return;
 
-	paramList = fp->mesa_program.Base.Parameters;
+	paramList = ctx->FragmentProgram._Current->Base.Parameters;
 
 	if (!paramList)
 		return;
 
-	for (i = 0; i < paramList->NumParameters; i++) {
-		if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) {
-			r300FetchStateParameter(ctx,
-						paramList->Parameters[i].
-						StateIndexes,
-						paramList->ParameterValues[i]);
-		}
-	}
+	_mesa_load_state_parameters(ctx, paramList);
 }
 
 /* =============================================================
@@ -1235,9 +1174,7 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r300_fragment_program_code *code = &fp->code;
+	struct r300_fragment_program_code *code = &r300->selected_fp->code.code.r300;
 
 	R300_STATECHANGE(r300, fpt);
 
@@ -1271,15 +1208,15 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 	}
 
 	r300->hw.fpt.cmd[R300_FPT_CMD_0] =
-		cmdpacket0(R300_US_TEX_INST_0, code->tex.length);
+		cmdpacket0(r300->radeon.radeonScreen,
+                   R300_US_TEX_INST_0, code->tex.length);
 }
 
 static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r500_fragment_program_code *code = &fp->code;
+	struct r500_fragment_program_code *code = &r300->selected_fp->code.code.r500;
 
 	/* find all the texture instructions and relocate the texture units */
 	for (i = 0; i < code->inst_end + 1; i++) {
@@ -1319,16 +1256,15 @@ static GLuint translate_lod_bias(GLfloat bias)
 	return (((GLuint)b) << R300_LOD_BIAS_SHIFT) & R300_LOD_BIAS_MASK;
 }
 
+
 static void r300SetupTextures(GLcontext * ctx)
 {
 	int i, mtu;
-	struct r300_tex_obj *t;
+	struct radeon_tex_obj *t;
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int hw_tmu = 0;
 	int last_hw_tmu = -1;	/* -1 translates into no setup costs for fields */
 	int tmu_mappings[R300_MAX_TEXTURE_UNITS] = { -1, };
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
 
 	R300_STATECHANGE(r300, txe);
 	R300_STATECHANGE(r300, tex.filter);
@@ -1343,7 +1279,7 @@ static void r300SetupTextures(GLcontext * ctx)
 	r300->hw.txe.cmd[R300_TXE_ENABLE] = 0x0;
 
 	mtu = r300->radeon.glCtx->Const.MaxTextureUnits;
-	if (RADEON_DEBUG & DEBUG_STATE)
+	if (RADEON_DEBUG & RADEON_STATE)
 		fprintf(stderr, "mtu=%d\n", mtu);
 
 	if (mtu > R300_MAX_TEXTURE_UNITS) {
@@ -1356,24 +1292,19 @@ static void r300SetupTextures(GLcontext * ctx)
 	/* We cannot let disabled tmu offsets pass DRM */
 	for (i = 0; i < mtu; i++) {
 		if (ctx->Texture.Unit[i]._ReallyEnabled) {
-
-#if 0				/* Enables old behaviour */
-			hw_tmu = i;
-#endif
 			tmu_mappings[i] = hw_tmu;
 
-			t = (r300TexObjPtr) r300->state.texture.unit[i].texobj->DriverData;
-			/* XXX questionable fix for bug 9170: */
+			t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
 			if (!t)
 				continue;
 
-			if ((t->format & 0xffffff00) == 0xffffff00) {
+			if ((t->pp_txformat & 0xffffff00) == 0xffffff00) {
 				WARN_ONCE
 				    ("unknown texture format (entry %x) encountered. Help me !\n",
-				     t->format & 0xff);
+				     t->pp_txformat & 0xff);
 			}
 
-			if (RADEON_DEBUG & DEBUG_STATE)
+			if (RADEON_DEBUG & RADEON_STATE)
 				fprintf(stderr,
 					"Activating texture unit %d\n", i);
 
@@ -1381,29 +1312,28 @@ static void r300SetupTextures(GLcontext * ctx)
 
 			r300->hw.tex.filter.cmd[R300_TEX_VALUE_0 +
 						hw_tmu] =
-			    gen_fixed_filter(t->filter) | (hw_tmu << 28);
+			    gen_fixed_filter(t->pp_txfilter) | (hw_tmu << 28);
 			/* Note: There is a LOD bias per texture unit and a LOD bias
 			 * per texture object. We add them here to get the correct behaviour.
 			 * (The per-texture object LOD bias was introduced in OpenGL 1.4
 			 * and is not present in the EXT_texture_object extension).
 			 */
 			r300->hw.tex.filter_1.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-				t->filter_1 |
-				translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.tObj->LodBias);
+				t->pp_txfilter_1 |
+				translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.LodBias);
 			r300->hw.tex.size.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-			    t->size;
+			    t->pp_txsize;
 			r300->hw.tex.format.cmd[R300_TEX_VALUE_0 +
-						hw_tmu] = t->format;
+						hw_tmu] = t->pp_txformat;
 			r300->hw.tex.pitch.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-			    t->pitch_reg;
-			r300->hw.tex.offset.cmd[R300_TEX_VALUE_0 +
-						hw_tmu] = t->offset;
+			  t->pp_txpitch;
+			r300->hw.textures[hw_tmu] = t;
 
-			if (t->offset & R300_TXO_MACRO_TILE) {
+			if (t->tile_bits & R300_TXO_MACRO_TILE) {
 				WARN_ONCE("macro tiling enabled!\n");
 			}
 
-			if (t->offset & R300_TXO_MICRO_TILE) {
+			if (t->tile_bits & R300_TXO_MICRO_TILE) {
 				WARN_ONCE("micro tiling enabled!\n");
 			}
 
@@ -1419,40 +1349,48 @@ static void r300SetupTextures(GLcontext * ctx)
 		}
 	}
 
+	/* R3xx and R4xx chips require that the texture unit corresponding to
+	 * KIL instructions is really enabled.
+	 *
+	 * We do some fakery here and in the state atom emit logic to enable
+	 * the texture without tripping up the CS checker in the kernel.
+	 */
+	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
+		if (ctx->FragmentProgram._Current->UsesKill && last_hw_tmu < 0) {
+			last_hw_tmu++;
+
+			r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1;
+
+			r300->hw.tex.border_color.cmd[R300_TEX_VALUE_0] = 0;
+			r300->hw.tex.chroma_key.cmd[R300_TEX_VALUE_0] = 0;
+			r300->hw.tex.filter.cmd[R300_TEX_VALUE_0] = 0;
+			r300->hw.tex.filter_1.cmd[R300_TEX_VALUE_0] = 0;
+			r300->hw.tex.size.cmd[R300_TEX_VALUE_0] = 0; /* 1x1 texture */
+			r300->hw.tex.format.cmd[R300_TEX_VALUE_0] = 0; /* A8 format */
+			r300->hw.tex.pitch.cmd[R300_TEX_VALUE_0] = 0;
+		}
+	}
+
 	r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER0_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, last_hw_tmu + 1);
 	r300->hw.tex.filter_1.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER1_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER1_0, last_hw_tmu + 1);
 	r300->hw.tex.size.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_SIZE_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_SIZE_0, last_hw_tmu + 1);
 	r300->hw.tex.format.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT_0, last_hw_tmu + 1);
 	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT2_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT2_0, last_hw_tmu + 1);
 	r300->hw.tex.offset.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_OFFSET_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_OFFSET_0, last_hw_tmu + 1);
 	r300->hw.tex.chroma_key.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_CHROMA_KEY_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_CHROMA_KEY_0, last_hw_tmu + 1);
 	r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
+	r300->vtbl.SetupFragmentShaderTextures(ctx, tmu_mappings);
 
-	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		if (fp->mesa_program.UsesKill && last_hw_tmu < 0) {
-			// The KILL operation requires the first texture unit
-			// to be enabled.
-			r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1;
-			r300->hw.tex.filter.cmd[R300_TEX_VALUE_0] = 0;
-			r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-				cmdpacket0(R300_TX_FILTER0_0, 1);
-		}
-		r300SetupFragmentShaderTextures(ctx, tmu_mappings);
-	} else
-		r500SetupFragmentShaderTextures(ctx, tmu_mappings);
-
-	if (RADEON_DEBUG & DEBUG_STATE)
+	if (RADEON_DEBUG & RADEON_STATE)
 		fprintf(stderr, "TX_ENABLE: %08x  last_hw_tmu=%d\n",
 			r300->hw.txe.cmd[R300_TXE_ENABLE], last_hw_tmu);
 }
@@ -1469,26 +1407,21 @@ union r300_outputs_written {
 static void r300SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->code.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = r300->selected_fp->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1507,15 +1440,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1527,15 +1452,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1545,6 +1462,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1554,64 +1472,27 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-		  count = 4;
-		else
-		  count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		switch(count) {
-		case 4: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3); break;
-		case 3: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		default:
-		case 1:
-		case 2: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		};
-
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz | R300_RS_TEX_PTR(rs_tex_count);
+		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count);
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
 		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |=  R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) |  R300_RS_TEX_PTR(rs_tex_count);
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
-	if (InputsRead & FRAG_BIT_WPOS) {
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |=  R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) |  R300_RS_TEX_PTR(rs_tex_count);
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
 		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
 	}
-	InputsRead &= ~FRAG_BIT_WPOS;
 
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R300_RS_COL_PTR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
 	high_rr = (col_ip > tex_ip) ? col_ip : tex_ip;
-	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT)  | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
+	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
 	r300->hw.rc.cmd[2] |= high_rr - 1;
 
-	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R300_RS_INST_0, high_rr);
+	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
@@ -1620,26 +1501,21 @@ static void r300SetupRSUnit(GLcontext * ctx)
 static void r500SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->code.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = r300->selected_fp->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1658,15 +1534,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1678,15 +1546,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1696,7 +1556,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
-
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1706,74 +1566,13 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz = 0;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-		  count = 4;
-		else
-		  count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		if (count == 4) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= (rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 3) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 2) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 1) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else {
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		}
-
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz;
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
-
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
-	if (InputsRead & FRAG_BIT_WPOS) {
 		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
+			((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
+			((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
+			((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
 
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
+		InputsRead &= ~(FRAG_BIT_TEX0 << i);
 		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
@@ -1781,72 +1580,25 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] |= R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R500_RS_COL_PTR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
 	high_rr = (col_ip > tex_ip) ? col_ip : tex_ip;
-	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT)  | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
-	r300->hw.rc.cmd[2] |= 0xC0 | (high_rr - 1);
+	r300->hw.rc.cmd[1] = (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
+	r300->hw.rc.cmd[2] = 0xC0 | (high_rr - 1);
 
-	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R500_RS_INST_0, high_rr);
+	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
 }
 
-
-
-
-#define bump_vpu_count(ptr, new_count)   do{\
-	drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\
-	int _nc=(new_count)/4; \
-	assert(_nc < 256); \
-	if(_nc>_p->vpu.count)_p->vpu.count=_nc;\
-	}while(0)
-
-static INLINE void r300SetupVertexProgramFragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf)
-{
-	int i;
-
-	if (vsf->length == 0)
-		return;
-
-	if (vsf->length & 0x3) {
-		fprintf(stderr, "VERTEX_SHADER_FRAGMENT must have length divisible by 4\n");
-		_mesa_exit(-1);
-	}
-
-	switch ((dest >> 8) & 0xf) {
-	case 0:
-		R300_STATECHANGE(r300, vpi);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpi.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-
-	case 2:
-		R300_STATECHANGE(r300, vpp);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpp.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	case 4:
-		R300_STATECHANGE(r300, vps);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vps.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	default:
-		fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
-		_mesa_exit(-1);
-	}
-}
-
 #define MIN3(a, b, c)	((a) < (b) ? MIN2(a, c) : MIN2(b, c))
 
-
-static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 			GLuint output_count, GLuint temp_count)
 {
     int vtx_mem_size;
@@ -1870,7 +1622,7 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
     pvs_num_cntrls = MIN2(6, vtx_mem_size/temp_count);
 
     R300_STATECHANGE(rmesa, vap_cntl);
-    if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+    if (rmesa->options.hw_tcl_enabled) {
 	rmesa->hw.vap_cntl.cmd[R300_VAP_CNTL_INSTR] =
 	    (pvs_num_slots << R300_PVS_NUM_SLOTS_SHIFT) |
 	    (pvs_num_cntrls << R300_PVS_NUM_CNTLRS_SHIFT) |
@@ -1900,114 +1652,6 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 
 }
 
-static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
-{
-	struct r300_vertex_shader_state *prog = &(rmesa->state.vertex_shader);
-	GLuint o_reg = 0;
-	GLuint i_reg = 0;
-	int i;
-	int inst_count = 0;
-	int param_count = 0;
-	int program_end = 0;
-
-	for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) {
-		if (rmesa->state.sw_tcl_inputs[i] != -1) {
-			prog->program.body.i[program_end + 0] = PVS_OP_DST_OPERAND(VE_MULTIPLY, GL_FALSE, GL_FALSE, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT);
-			prog->program.body.i[program_end + 1] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 2] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 3] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			program_end += 4;
-			i_reg++;
-		}
-	}
-
-	prog->program.length = program_end;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START,
-				       &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, i_reg, o_reg, 0);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	    (0 << R300_PVS_FIRST_INST_SHIFT) |
-	    (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	    (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	    (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	    (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	    (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-static int bit_count (int x)
-{
-    x = ((x & 0xaaaaaaaaU) >> 1) + (x & 0x55555555U);
-    x = ((x & 0xccccccccU) >> 2) + (x & 0x33333333U);
-    x = (x >> 16) + (x & 0xffff);
-    x = ((x & 0xf0f0) >> 4) + (x & 0x0f0f);
-    return (x >> 8) + (x & 0x00ff);
-}
-
-static void r300SetupRealVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_vertex_program *prog = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	int inst_count = 0;
-	int param_count = 0;
-
-	/* FIXME: r300SetupVertexProgramFragment */
-	R300_STATECHANGE(rmesa, vpp);
-	param_count =
-	    r300VertexProgUpdateParams(ctx,
-				       (struct r300_vertex_program_cont *)
-				       ctx->VertexProgram._Current,
-				       (float *)&rmesa->hw.vpp.
-				       cmd[R300_VPP_PARAM_0]);
-	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
-	param_count /= 4;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START, &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, bit_count(prog->key.InputsRead),
-		    bit_count(prog->key.OutputsWritten), prog->num_temporaries);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	  (0 << R300_PVS_FIRST_INST_SHIFT) |
-	  (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	  (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	  (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	  (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	  (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-static void r300SetupVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-
-	/* Reset state, in case we don't use something */
-	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
-
-	/* Not sure why this doesnt work...
-	   0x400 area might have something to do with pixel shaders as it appears right after pfs programming.
-	   0x406 is set to { 0.0, 0.0, 1.0, 0.0 } most of the time but should change with smooth points and in other rare cases. */
-	//setup_vertex_shader_fragment(rmesa, 0x406, &unk4);
-	if (hw_tcl_on && ((struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx))->translated) {
-		r300SetupRealVertexProgram(rmesa);
-	} else {
-		/* FIXME: This needs to be replaced by vertex shader generation code. */
-		r300SetupDefaultVertexProgram(rmesa);
-	}
-
-}
-
 /**
  * Enable/Disable states.
  *
@@ -2015,20 +1659,13 @@ static void r300SetupVertexProgram(r300ContextPtr rmesa)
  */
 static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 {
-	if (RADEON_DEBUG & DEBUG_STATE)
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	if (RADEON_DEBUG & RADEON_STATE)
 		fprintf(stderr, "%s( %s = %s )\n", __FUNCTION__,
 			_mesa_lookup_enum_by_nr(cap),
 			state ? "GL_TRUE" : "GL_FALSE");
 
 	switch (cap) {
-	case GL_TEXTURE_1D:
-	case GL_TEXTURE_2D:
-	case GL_TEXTURE_3D:
-		/* empty */
-		break;
-	case GL_FOG:
-		/* empty */
-		break;
 	case GL_ALPHA_TEST:
 		r300SetAlphaState(ctx);
 		break;
@@ -2046,22 +1683,46 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 	case GL_CLIP_PLANE5:
 		r300SetClipPlaneState(ctx, cap, state);
 		break;
+	case GL_CULL_FACE:
+		r300UpdateCulling(ctx);
+		break;
 	case GL_DEPTH_TEST:
 		r300SetDepthState(ctx);
 		break;
-	case GL_STENCIL_TEST:
-		r300SetStencilState(ctx, state);
+	case GL_LINE_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_SMOOTH, ctx->Line.SmoothFlag);
 		break;
-	case GL_CULL_FACE:
-		r300UpdateCulling(ctx);
+	case GL_LINE_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_STIPPLE, ctx->Line.StippleFlag);
+		break;
+	case GL_POINT_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POINT_SMOOTH, ctx->Point.SmoothFlag);
+		break;
+	case GL_POLYGON_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_SMOOTH, ctx->Polygon.SmoothFlag);
+		break;
+	case GL_POLYGON_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_STIPPLE, ctx->Polygon.StippleFlag);
 		break;
 	case GL_POLYGON_OFFSET_POINT:
 	case GL_POLYGON_OFFSET_LINE:
 	case GL_POLYGON_OFFSET_FILL:
 		r300SetPolygonOffsetState(ctx, state);
 		break;
+	case GL_SCISSOR_TEST:
+		radeon_firevertices(&rmesa->radeon);
+		rmesa->radeon.state.scissor.enabled = state;
+		radeonUpdateScissor( ctx );
+		break;
+	case GL_STENCIL_TEST:
+		r300SetStencilState(ctx, state);
+		break;
 	default:
-		radeonEnable(ctx, cap, state);
 		break;
 	}
 }
@@ -2072,15 +1733,14 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 static void r300ResetHwState(r300ContextPtr r300)
 {
 	GLcontext *ctx = r300->radeon.glCtx;
-	int has_tcl = 1;
+	int has_tcl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
-	if (RADEON_DEBUG & DEBUG_STATE)
+	if (RADEON_DEBUG & RADEON_STATE)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
-	r300UpdateWindow(ctx);
+	radeon_firevertices(&r300->radeon);
 
 	r300ColorMask(ctx,
 		      ctx->Color.ColorMask[RCOMP],
@@ -2102,8 +1762,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300UpdateCulling(ctx);
 
-	r300UpdateTextureState(ctx);
-
 	r300SetBlendState(ctx);
 	r300SetLogicOpState(ctx);
 
@@ -2182,8 +1840,8 @@ static void r300ResetHwState(r300ContextPtr r300)
 	}
 
 	/* XXX: Enable anti-aliasing? */
-	r300->hw.gb_misc.cmd[R300_GB_MISC_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
-	r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = 0;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_SELECT] = 0;
 
 	r300->hw.ga_point_s0.cmd[1] = r300PackFloat32(0.0);
 	r300->hw.ga_point_s0.cmd[2] = r300PackFloat32(0.0);
@@ -2242,20 +1900,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300BlendColor(ctx, ctx->Color.BlendColor);
 
-	/* Again, r300ClearBuffer uses this */
-	r300->hw.cb.cmd[R300_CB_OFFSET] =
-	    r300->radeon.state.color.drawOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.state.color.drawPitch;
-
-	if (r300->radeon.radeonScreen->cpp == 4)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-
-	if (r300->radeon.sarea->tiling_enabled)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-
 	r300->hw.rb3d_dither_ctl.cmd[1] = 0;
 	r300->hw.rb3d_dither_ctl.cmd[2] = 0;
 	r300->hw.rb3d_dither_ctl.cmd[3] = 0;
@@ -2268,44 +1912,18 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300->hw.rb3d_aaresolve_ctl.cmd[1] = 0;
 
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
-
-	r300->hw.zb.cmd[R300_ZB_OFFSET] =
-	    r300->radeon.radeonScreen->depthOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.zb.cmd[R300_ZB_PITCH] = r300->radeon.radeonScreen->depthPitch;
-
-	if (r300->radeon.sarea->tiling_enabled) {
-		/* XXX: Turn off when clearing buffers ? */
-		r300->hw.zb.cmd[R300_ZB_PITCH] |= R300_DEPTHMACROTILE_ENABLE;
-
-		if (ctx->Visual.depthBits == 24)
-			r300->hw.zb.cmd[R300_ZB_PITCH] |=
-			    R300_DEPTHMICROTILE_TILED;
-	}
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
 
 	r300->hw.zb_depthclearvalue.cmd[1] = 0;
 
-	switch (ctx->Visual.depthBits) {
-	case 16:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_16BIT_INT_Z;
-		break;
-	case 24:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
-		break;
-	default:
-		fprintf(stderr, "Error: Unsupported depth %d... exiting\n", ctx->Visual.depthBits);
-		_mesa_exit(-1);
-	}
-
 	r300->hw.zstencil_format.cmd[2] = R300_ZTOP_DISABLE;
 	r300->hw.zstencil_format.cmd[3] = 0x00000003;
 	r300->hw.zstencil_format.cmd[4] = 0x00000000;
 	r300SetEarlyZState(ctx);
 
-	r300->hw.unk4F30.cmd[1] = 0;
-	r300->hw.unk4F30.cmd[2] = 0;
+	r300->hw.zb_zmask.cmd[1] = 0;
+	r300->hw.zb_zmask.cmd[2] = 0;
 
 	r300->hw.zb_hiz_offset.cmd[1] = 0;
 
@@ -2319,135 +1937,151 @@ static void r300ResetHwState(r300ContextPtr r300)
 		r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0;
 	}
 
-	r300->hw.all_dirty = GL_TRUE;
+	r300->radeon.hw.all_dirty = GL_TRUE;
 }
 
 void r300UpdateShaders(r300ContextPtr rmesa)
 {
-	GLcontext *ctx;
-	struct r300_vertex_program *vp;
-	int i;
+	GLcontext *ctx = rmesa->radeon.glCtx;
 
-	ctx = rmesa->radeon.glCtx;
+	/* should only happenen once, just after context is created */
+	/* TODO: shouldn't we fallback to sw here? */
+	if (!ctx->FragmentProgram._Current) {
+		_mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
+		return;
+	}
 
-	if (rmesa->NewGLState && hw_tcl_on) {
-		rmesa->NewGLState = 0;
+	{
+		struct r300_fragment_program *fp;
 
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			rmesa->temp_attrib[i] =
-			    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    &rmesa->dummy_attrib[i];
-		}
+		fp = r300SelectAndTranslateFragmentShader(ctx);
 
-		_tnl_UpdateFixedFunctionProgram(ctx);
+		r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
+	}
 
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    rmesa->temp_attrib[i];
-		}
+	if (rmesa->options.hw_tcl_enabled) {
+		struct r300_vertex_program *vp;
 
-		r300SelectVertexShader(rmesa);
-		vp = (struct r300_vertex_program *)
-		    CURRENT_VERTEX_SHADER(ctx);
-		/*if (vp->translated == GL_FALSE)
-		   r300TranslateVertexShader(vp); */
-		if (vp->translated == GL_FALSE) {
-			fprintf(stderr, "Failing back to sw-tcl\n");
-			hw_tcl_on = future_hw_tcl_on = 0;
-			r300ResetHwState(rmesa);
-
-			r300UpdateStateParameters(ctx, _NEW_PROGRAM);
-			return;
+		if (rmesa->radeon.NewGLState) {
+			int i;
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				rmesa->temp_attrib[i] =
+				    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    &rmesa->dummy_attrib[i];
+			}
+
+			_tnl_UpdateFixedFunctionProgram(ctx);
+
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    rmesa->temp_attrib[i];
+			}
 		}
+
+		vp = r300SelectAndTranslateVertexShader(ctx);
+
+		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, vp->error);
 	}
-	r300UpdateStateParameters(ctx, _NEW_PROGRAM);
+
+	r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+	rmesa->radeon.NewGLState = 0;
 }
 
-static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
-	struct gl_program *program, struct prog_src_register srcreg)
+static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx, GLuint index, GLfloat * buffer)
 {
 	static const GLfloat dummy[4] = { 0, 0, 0, 0 };
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct rc_constant * rcc = &rmesa->selected_fp->code.constants.Constants[index];
+
+	switch(rcc->Type) {
+	case RC_CONSTANT_EXTERNAL:
+		return ctx->FragmentProgram._Current->Base.Parameters->ParameterValues[rcc->u.External];
+	case RC_CONSTANT_IMMEDIATE:
+		return rcc->u.Immediate;
+	case RC_CONSTANT_STATE:
+		switch(rcc->u.State[0]) {
+		case RC_STATE_SHADOW_AMBIENT: {
+			const int unit = (int) rcc->u.State[1];
+			const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
+			if (texObj) {
+				buffer[0] =
+				buffer[1] =
+				buffer[2] =
+				buffer[3] = texObj->CompareFailValue;
+			}
+			return buffer;
+		}
 
-	switch(srcreg.File) {
-	case PROGRAM_LOCAL_PARAM:
-		return program->LocalParams[srcreg.Index];
-	case PROGRAM_ENV_PARAM:
-		return ctx->FragmentProgram.Parameters[srcreg.Index];
-	case PROGRAM_STATE_VAR:
-	case PROGRAM_NAMED_PARAM:
-	case PROGRAM_CONSTANT:
-		return program->Parameters->ParameterValues[srcreg.Index];
-	default:
-		_mesa_problem(ctx, "get_fragmentprogram_constant: Unknown\n");
-		return dummy;
+		case RC_STATE_R300_WINDOW_DIMENSION: {
+			__DRIdrawablePrivate * drawable = radeon_get_drawable(&rmesa->radeon);
+			buffer[0] = drawable->w * 0.5f;	/* width*0.5 */
+			buffer[1] = drawable->h * 0.5f;	/* height*0.5 */
+			buffer[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
+			buffer[3] = 1.0F;	/* not used */
+			return buffer;
+		}
+
+		case RC_STATE_R300_TEXRECT_FACTOR: {
+			struct gl_texture_object *t =
+				ctx->Texture.Unit[rcc->u.State[1]].CurrentTex[TEXTURE_RECT_INDEX];
+
+			if (t && t->Image[0][t->BaseLevel]) {
+				struct gl_texture_image *image =
+					t->Image[0][t->BaseLevel];
+				buffer[0] = 1.0 / image->Width2;
+				buffer[1] = 1.0 / image->Height2;
+			} else {
+				buffer[0] = 1.0;
+				buffer[1] = 1.0;
+			}
+			buffer[2] = 1.0;
+			buffer[3] = 1.0;
+			return buffer;
+		}
+		}
 	}
+
+	return dummy;
 }
 
 
-static void r300SetupPixelShader(r300ContextPtr rmesa)
+static void r300SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	struct r300_fragment_program_code *code;
-	int i, k;
-
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
-	r300TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
+	int i;
 
-	r300SetupTextures(ctx);
+	code = &fp->code.code.r300;
 
 	R300_STATECHANGE(rmesa, fpi[0]);
 	R300_STATECHANGE(rmesa, fpi[1]);
 	R300_STATECHANGE(rmesa, fpi[2]);
 	R300_STATECHANGE(rmesa, fpi[3]);
-	rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_INST_0, code->alu.length);
-	rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_ADDR_0, code->alu.length);
-	rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_INST_0, code->alu.length);
-	rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
+	rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_INST_0, code->alu.length);
+	rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_ADDR_0, code->alu.length);
+	rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_INST_0, code->alu.length);
+	rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
 	for (i = 0; i < code->alu.length; i++) {
-		rmesa->hw.fpi[0].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst0;
-		rmesa->hw.fpi[1].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst1;
-		rmesa->hw.fpi[2].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst2;
-		rmesa->hw.fpi[3].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst3;
+		rmesa->hw.fpi[0].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].rgb_inst;
+		rmesa->hw.fpi[1].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].rgb_addr;
+		rmesa->hw.fpi[2].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].alpha_inst;
+		rmesa->hw.fpi[3].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].alpha_addr;
 	}
 
 	R300_STATECHANGE(rmesa, fp);
-	rmesa->hw.fp.cmd[R300_FP_CNTL0] = code->cur_node | (code->first_node_has_tex << 3);
-	rmesa->hw.fp.cmd[R300_FP_CNTL1] = code->max_temp_idx;
-	rmesa->hw.fp.cmd[R300_FP_CNTL2] =
-	  (0 << R300_PFS_CNTL_ALU_OFFSET_SHIFT) |
-	  ((code->alu.length-1) << R300_PFS_CNTL_ALU_END_SHIFT) |
-	  (0 << R300_PFS_CNTL_TEX_OFFSET_SHIFT) |
-	  ((code->tex.length ? code->tex.length-1 : 0) << R300_PFS_CNTL_TEX_END_SHIFT);
-	/* I just want to say, the way these nodes are stored.. weird.. */
-	for (i = 0, k = (4 - (code->cur_node + 1)); i < 4; i++, k++) {
-		if (i < (code->cur_node + 1)) {
-			rmesa->hw.fp.cmd[R300_FP_NODE0 + k] =
-			  (code->node[i].alu_offset << R300_ALU_START_SHIFT) |
-			  (code->node[i].alu_end << R300_ALU_SIZE_SHIFT) |
-			  (code->node[i].tex_offset << R300_TEX_START_SHIFT) |
-			  (code->node[i].tex_end << R300_TEX_SIZE_SHIFT) |
-			  code->node[i].flags;
-		} else {
-			rmesa->hw.fp.cmd[R300_FP_NODE0 + (3 - i)] = 0;
-		}
-	}
+	rmesa->hw.fp.cmd[R300_FP_CNTL0] = code->config;
+	rmesa->hw.fp.cmd[R300_FP_CNTL1] = code->pixsize;
+	rmesa->hw.fp.cmd[R300_FP_CNTL2] = code->code_offset;
+	for (i = 0; i < 4; i++)
+		rmesa->hw.fp.cmd[R300_FP_NODE0 + i] = code->code_addr[i];
 
 	R300_STATECHANGE(rmesa, fpp);
-	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, code->const_nr * 4);
-	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_PFS_PARAM_0_X, fp->code.constants.Count * 4);
+	for (i = 0; i < fp->code.constants.Count; i++) {
+		GLfloat buffer[4];
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, i, buffer);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(constant[0]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(constant[1]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(constant[2]);
@@ -2469,41 +2103,29 @@ static void r300SetupPixelShader(r300ContextPtr rmesa)
 	if(_nc>_p->r500fp.count)_p->r500fp.count=_nc;\
 } while(0)
 
-static void r500SetupPixelShader(r300ContextPtr rmesa)
+static void r500SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	int i;
 	struct r500_fragment_program_code *code;
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp.cmd)->r500fp.count = 0;
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp_const.cmd)->r500fp.count = 0;
 
-	r500TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
-
-	r300SetupTextures(ctx);
+	code = &fp->code.code.r500;
 
 	R300_STATECHANGE(rmesa, fp);
 	rmesa->hw.fp.cmd[R500_FP_PIXSIZE] = code->max_temp_idx;
 
 	rmesa->hw.fp.cmd[R500_FP_CODE_ADDR] =
-	    R500_US_CODE_START_ADDR(code->inst_offset) |
+	    R500_US_CODE_START_ADDR(0) |
 	    R500_US_CODE_END_ADDR(code->inst_end);
 	rmesa->hw.fp.cmd[R500_FP_CODE_RANGE] =
-	    R500_US_CODE_RANGE_ADDR(code->inst_offset) |
+	    R500_US_CODE_RANGE_ADDR(0) |
 	    R500_US_CODE_RANGE_SIZE(code->inst_end);
 	rmesa->hw.fp.cmd[R500_FP_CODE_OFFSET] =
-	    R500_US_CODE_OFFSET_ADDR(0); /* FIXME when we add flow control */
+	    R500_US_CODE_OFFSET_ADDR(0);
 
 	R300_STATECHANGE(rmesa, r500fp);
 	/* Emit our shader... */
@@ -2519,60 +2141,95 @@ static void r500SetupPixelShader(r300ContextPtr rmesa)
 	bump_r500fp_count(rmesa->hw.r500fp.cmd, (code->inst_end + 1) * 6);
 
 	R300_STATECHANGE(rmesa, r500fp_const);
-	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+	for (i = 0; i < fp->code.constants.Count; i++) {
+		GLfloat buffer[4];
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, i, buffer);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat32(constant[0]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat32(constant[1]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat32(constant[2]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = r300PackFloat32(constant[3]);
 	}
-	bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, code->const_nr * 4);
-
+	bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, fp->code.constants.Count * 4);
 }
 
-void r300UpdateShaderStates(r300ContextPtr rmesa)
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten)
 {
-	GLcontext *ctx;
-	ctx = rmesa->radeon.glCtx;
+	r300ContextPtr rmesa = R300_CONTEXT( ctx );
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
+	int i, j, reg_count;
+	uint32_t *vir0 = &rmesa->hw.vir[0].cmd[1];
+	uint32_t *vir1 = &rmesa->hw.vir[1].cmd[1];
 
-	r300UpdateTextureState(ctx);
-	r300SetEarlyZState(ctx);
+	for (i = 0; i < R300_VIR_CMDSIZE-1; ++i)
+		vir0[i] = vir1[i] = 0;
 
-	/* w_fmt value is set to get best performance
-	 * see p.130 R5xx 3D acceleration guide v1.3 */
-	GLuint w_fmt, fgdepthsrc;
-	if (current_fragment_program_writes_depth(ctx)) {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
-		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
-	} else {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
-		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
+	for (i = 0, j = 0; i < rmesa->vbuf.num_attribs; ++i) {
+		int tmp;
+
+		tmp = attrs[i].data_type | (attrs[i].dst_loc << R300_DST_VEC_LOC_SHIFT);
+		if (attrs[i]._signed)
+			tmp |= R300_SIGNED;
+		if (attrs[i].normalize)
+			tmp |= R300_NORMALIZE;
+
+		if (i % 2 == 0) {
+			vir0[j] = tmp << R300_DATA_TYPE_0_SHIFT;
+			vir1[j] = attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT);
+		} else {
+			vir0[j] |= tmp << R300_DATA_TYPE_1_SHIFT;
+			vir1[j] |= (attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
+			++j;
+		}
 	}
 
-	if (w_fmt != rmesa->hw.us_out_fmt.cmd[5]) {
-		R300_STATECHANGE(rmesa, us_out_fmt);
-		rmesa->hw.us_out_fmt.cmd[5] = w_fmt;
+	reg_count = (rmesa->vbuf.num_attribs + 1) >> 1;
+	if (rmesa->vbuf.num_attribs % 2 != 0) {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
+	} else {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
 	}
 
-	if (fgdepthsrc != rmesa->hw.fg_depth_src.cmd[1]) {
-		R300_STATECHANGE(rmesa, fg_depth_src);
-		rmesa->hw.fg_depth_src.cmd[1] = fgdepthsrc;
+	R300_STATECHANGE(rmesa, vir[0]);
+	R300_STATECHANGE(rmesa, vir[1]);
+	R300_STATECHANGE(rmesa, vof);
+	R300_STATECHANGE(rmesa, vic);
+
+	if (rmesa->radeon.radeonScreen->kernel_mm) {
+		rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[0].cmd[0] |= (reg_count & 0x3FFF) << 16;
+		rmesa->hw.vir[1].cmd[0] |= (reg_count & 0x3FFF) << 16;
+	} else {
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = reg_count;
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = reg_count;
 	}
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupPixelShader(rmesa);
-	else
-		r300SetupPixelShader(rmesa);
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten);
+}
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupRSUnit(ctx);
-	else
-		r300SetupRSUnit(ctx);
+void r300UpdateShaderStates(r300ContextPtr rmesa)
+{
+	GLcontext *ctx;
+	ctx = rmesa->radeon.glCtx;
 
-	if ((rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		r300SetupVertexProgram(rmesa);
+	/* should only happenen once, just after context is created */
+	if (!ctx->FragmentProgram._Current)
+		return;
 
+	r300SetEarlyZState(ctx);
+
+	r300SetupTextures(ctx);
+
+	rmesa->vtbl.SetupPixelShader(ctx);
+
+	rmesa->vtbl.SetupRSUnit(ctx);
+
+	if (rmesa->options.hw_tcl_enabled) {
+		r300SetupVertexProgram(rmesa);
+	}
 }
 
 /**
@@ -2586,15 +2243,17 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 	_swsetup_InvalidateState(ctx, new_state);
 	_vbo_InvalidateState(ctx, new_state);
 	_tnl_InvalidateState(ctx, new_state);
-	_ae_invalidate_state(ctx, new_state);
 
-	if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-		r300UpdateDrawBuffer(ctx);
-	}
+	if (new_state & _NEW_BUFFERS) {
+		_mesa_update_framebuffer(ctx);
+		/* this updates the DrawBuffer's Width/Height if it's a FBO */
+		_mesa_update_draw_buffer_bounds(ctx);
 
-	r300UpdateStateParameters(ctx, new_state);
+		R300_STATECHANGE(r300, cb);
+		R300_STATECHANGE(r300, zb);
+	}
 
-	r300->NewGLState |= new_state;
+	r300->radeon.NewGLState |= new_state;
 }
 
 /**
@@ -2604,58 +2263,12 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
  */
 void r300InitState(r300ContextPtr r300)
 {
-	GLcontext *ctx = r300->radeon.glCtx;
-	GLuint depth_fmt;
-
-	radeonInitState(&r300->radeon);
-
-	switch (ctx->Visual.depthBits) {
-	case 16:
-		r300->state.depth.scale = 1.0 / (GLfloat) 0xffff;
-		depth_fmt = R300_DEPTHFORMAT_16BIT_INT_Z;
-		break;
-	case 24:
-		r300->state.depth.scale = 1.0 / (GLfloat) 0xffffff;
-		depth_fmt = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
-		break;
-	default:
-		fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
-			ctx->Visual.depthBits);
-		_mesa_exit(-1);
-	}
-
-	/* Only have hw stencil when depth buffer is 24 bits deep */
-	r300->state.stencil.hw_stencil = (ctx->Visual.stencilBits > 0 &&
-					  ctx->Visual.depthBits == 24);
-
-	memset(&(r300->state.texture), 0, sizeof(r300->state.texture));
-
 	r300ResetHwState(r300);
 }
 
 static void r300RenderMode(GLcontext * ctx, GLenum mode)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	(void)rmesa;
-	(void)mode;
-}
-
-void r300UpdateClipPlanes( GLcontext *ctx )
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint p;
-
-	for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
-		if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
-			GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p];
-
-			R300_STATECHANGE( rmesa, vpucp[p] );
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_X] = ip[0];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Y] = ip[1];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Z] = ip[2];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_W] = ip[3];
-		}
-	}
+	r300SwitchFallback(ctx, R300_FALLBACK_RENDER_MODE, ctx->RenderMode != GL_RENDER);
 }
 
 /**
@@ -2663,7 +2276,6 @@ void r300UpdateClipPlanes( GLcontext *ctx )
  */
 void r300InitStateFuncs(struct dd_function_table *functions)
 {
-	radeonInitStateFuncs(functions);
 
 	functions->UpdateState = r300InvalidateState;
 	functions->AlphaFunc = r300AlphaFunc;
@@ -2699,4 +2311,21 @@ void r300InitStateFuncs(struct dd_function_table *functions)
 	functions->RenderMode = r300RenderMode;
 
 	functions->ClipPlane = r300ClipPlane;
+	functions->Scissor = radeonScissor;
+
+	functions->DrawBuffer		= radeonDrawBuffer;
+	functions->ReadBuffer		= radeonReadBuffer;
+}
+
+void r300InitShaderFunctions(r300ContextPtr r300)
+{
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		r300->vtbl.SetupRSUnit = r500SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r500SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r500SetupFragmentShaderTextures;
+	} else {
+		r300->vtbl.SetupRSUnit = r300SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r300SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r300SetupFragmentShaderTextures;
+	}
 }
diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h
index 0589ab7cad..d46bf9f179 100644
--- a/src/mesa/drivers/dri/r300/r300_state.h
+++ b/src/mesa/drivers/dri/r300/r300_state.h
@@ -39,42 +39,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define R300_NEWPRIM( rmesa )			\
   do {						\
-    if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );		\
+  if ( rmesa->radeon.dma.flush )			\
+    rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
   } while (0)
 
 #define R300_STATECHANGE(r300, atom) \
 	do {						\
 	  R300_NEWPRIM(r300);				\
 		r300->hw.atom.dirty = GL_TRUE;		\
-		r300->hw.is_dirty = GL_TRUE;		\
+		r300->radeon.hw.is_dirty = GL_TRUE;		\
 	} while(0)
 
-#define R300_PRINT_STATE(r300, atom) \
-		r300PrintStateAtom(r300, &r300->hw.atom)
-
-/* Fire the buffered vertices no matter what.
-   TODO: This has not been implemented yet
- */
-#define R300_FIREVERTICES( r300 )			\
-do {							\
-    \
-   if ( (r300)->cmdbuf.count_used || (r300)->dma.flush ) {	\
-      r300Flush( (r300)->radeon.glCtx );		\
-   }							\
-    \
-} while (0)
-
-// r300_state.c
-extern int future_hw_tcl_on;
-void _tnl_UpdateFixedFunctionProgram (GLcontext * ctx);
 void r300UpdateViewportOffset (GLcontext * ctx);
 void r300UpdateDrawBuffer (GLcontext * ctx);
-void r300UpdateStateParameters (GLcontext * ctx, GLuint new_state);
 void r300UpdateShaders (r300ContextPtr rmesa);
 void r300UpdateShaderStates (r300ContextPtr rmesa);
 void r300InitState (r300ContextPtr r300);
-void r300UpdateClipPlanes (GLcontext * ctx);
 void r300InitStateFuncs (struct dd_function_table *functions);
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, GLuint output_count, GLuint temp_count);
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten);
 
 #endif				/* __R300_STATE_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index ba3621b16b..ee2c71e1a7 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -28,362 +28,264 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Dave Airlie <airlied@linux.ie>
+ *   Maciej Cencora <m.cencora@gmail.com>
  */
 
-/* derived from r200 swtcl path */
-
-
-
-#include "main/glheader.h"
-#include "main/mtypes.h"
-#include "main/colormac.h"
-#include "main/enums.h"
-#include "main/image.h"
-#include "main/imports.h"
-#include "main/light.h"
-#include "main/macros.h"
-
-#include "swrast/s_context.h"
-#include "swrast/s_fog.h"
-#include "swrast_setup/swrast_setup.h"
-#include "math/m_translate.h"
 #include "tnl/tnl.h"
-#include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
 
-#include "r300_context.h"
-#include "r300_swtcl.h"
 #include "r300_state.h"
-#include "r300_ioctl.h"
+#include "r300_swtcl.h"
 #include "r300_emit.h"
-#include "r300_mem.h"
-
-static void flush_last_swtcl_prim( r300ContextPtr rmesa  );
-
+#include "r300_tex.h"
+#include "r300_render.h"
+#include "main/simple_list.h"
 
-void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset);
-void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr);
 #define EMIT_ATTR( ATTR, STYLE )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+	rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+	rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+	rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
-static void r300SetVertexFormat( GLcontext *ctx )
+#define ADD_ATTR(_attr, _format, _dst_loc, _swizzle, _write_mask, _normalize) \
+do { \
+	attrs[num_attrs].element = (_attr); \
+	attrs[num_attrs].data_type = (_format); \
+	attrs[num_attrs].dst_loc = (_dst_loc); \
+	attrs[num_attrs].swizzle = (_swizzle); \
+	attrs[num_attrs].write_mask = (_write_mask); \
+	attrs[num_attrs]._signed = 0; \
+	attrs[num_attrs].normalize = (_normalize); \
+	++num_attrs; \
+} while (0)
+
+void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_OutputsWritten)
 {
 	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct vertex_buffer *VB = &tnl->vb;
-	DECLARE_RENDERINPUTS(index_bitset);
-	GLuint InputsRead = 0, OutputsWritten = 0;
-	int vap_fmt_1 = 0;
-	int offset = 0;
-	int vte = 0;
-	int fog_id;
-	GLint inputs[VERT_ATTRIB_MAX];
-	GLint tab[VERT_ATTRIB_MAX];
-	int swizzle[VERT_ATTRIB_MAX][4];
-	GLuint i, nr;
-	GLuint sz;
-
-	DECLARE_RENDERINPUTS(render_inputs_bitset);
-	RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
-	RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
-	RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
-
-	vte = rmesa->hw.vte.cmd[1];
-	vte &= ~(R300_VTX_XY_FMT | R300_VTX_Z_FMT | R300_VTX_W0_FMT);
-	/* Important:
-	 */
-	if ( VB->NdcPtr != NULL ) {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
-		vte |= R300_VTX_XY_FMT | R300_VTX_Z_FMT;
-	}
-	else {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
-		vte |= R300_VTX_W0_FMT;
-	}
-
-	assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-	rmesa->swtcl.vertex_attr_count = 0;
-
-	/* EMIT_ATTR's must be in order as they tell t_vertex.c how to
-	 * build up a hardware vertex.
-	 */
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POS)) {
-		sz = VB->AttribPtr[VERT_ATTRIB_POS]->size;
-		InputsRead |= 1 << VERT_ATTRIB_POS;
-		OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_1F + sz - 1 );
-		offset = sz;
-	} else {
-		offset = 4;
-		EMIT_PAD(4 * sizeof(float));
-	}
-/*
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) {
-		EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F );
-		offset += 1;
-	}
-*/
-	if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR0)) {
-		sz = VB->AttribPtr[VERT_ATTRIB_COLOR0]->size;
-	        rmesa->swtcl.coloroffset = offset;
+	int first_free_tex = 0;
+	GLuint InputsRead = 0;
+	GLuint OutputsWritten = 0;
+	int num_attrs = 0;
+	GLuint fp_reads = rmesa->selected_fp->InputsRead;
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
+
+	radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __func__);
+	rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0;
+	rmesa->radeon.swtcl.vertex_attr_count = 0;
+
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s\n", __func__);
+
+	/* We always want non Ndc coords format */
+	VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
+
+	/* Always write position vector */
+	InputsRead |= 1 << VERT_ATTRIB_POS;
+	OutputsWritten |= 1 << VERT_RESULT_HPOS;
+	EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
+	ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_POS, SWIZZLE_XYZW, MASK_XYZW, 0);
+	rmesa->swtcl.coloroffset = 4;
+
+	if (fp_reads & FRAG_BIT_COL0) {
 		InputsRead |= 1 << VERT_ATTRIB_COLOR0;
 		OutputsWritten |= 1 << VERT_RESULT_COL0;
-		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_1F + sz - 1 );
-		offset += sz;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
+#endif
 	}
 
-	rmesa->swtcl.specoffset = 0;
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
-		sz = VB->AttribPtr[VERT_ATTRIB_COLOR1]->size;
-		rmesa->swtcl.specoffset = offset;
-		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_1F + sz - 1 );
+	if (fp_reads & FRAG_BIT_COL1) {
+		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 		InputsRead |= 1 << VERT_ATTRIB_COLOR1;
 		OutputsWritten |= 1 << VERT_RESULT_COL1;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
+#endif
+		rmesa->swtcl.specoffset = rmesa->swtcl.coloroffset + 1;
 	}
 
-	fog_id = -1;
-	if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) {
-		/* find first free tex coord slot */
-		if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-			int i;
-			for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-				if (!RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-					fog_id = i;
-					break;
-				}
-			}
-		} else {
-			fog_id = 0;
+	if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
+		VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1];
+		OutputsWritten |= 1 << VERT_RESULT_BFC0;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#endif
+		if (fp_reads & FRAG_BIT_COL1) {
+			VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1];
+			GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			OutputsWritten |= 1 << VERT_RESULT_BFC1;
+#if MESA_LITTLE_ENDIAN
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_RGBA );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#else
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_ABGR );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#endif
 		}
+	}
 
-		if (fog_id == -1) {
-			fprintf(stderr, "\tout of free texcoords to do fog\n");
-			_mesa_exit(-1);
-		}
+	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_POINTSIZE )) {
+		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO);
+		InputsRead |= 1 << VERT_ATTRIB_POINT_SIZE;
+		OutputsWritten |= 1 << VERT_RESULT_PSIZ;
+		EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F );
+		ADD_ATTR(VERT_ATTRIB_POINT_SIZE, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_POINT_SIZE, swiz, MASK_X, 0);
+	}
 
-		sz = VB->AttribPtr[VERT_ATTRIB_FOG]->size;
-		EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F + sz - 1);
-		InputsRead |= 1 << VERT_ATTRIB_FOG;
-		OutputsWritten |= 1 << VERT_RESULT_FOGC;
-		vap_fmt_1 |= sz << (3 * fog_id);
+	if (rmesa->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0;
+
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
 	}
 
-	if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-		int i;
+	if (rmesa->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0;
 
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-				sz = VB->TexCoordPtr[i]->size;
-				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
-				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-				EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1 );
-				vap_fmt_1 |= sz << (3 * i);
-			}
-		}
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
 	}
 
-	/* RS can't put fragment position on the pixel stack, so stuff it in texcoord if needed */
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POS) && (ctx->FragmentProgram._Current->Base.InputsRead & FRAG_BIT_WPOS)) {
-		int first_free_tex = -1;
-		if (fog_id >= 0) {
-			first_free_tex = fog_id+1;
-		} else {
-			if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-				int i;
-				for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-					if (!RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-						first_free_tex = i;
+	/**
+	 *  Sending only one texcoord component may lead to lock up,
+	 *  so for all textures always output 4 texcoord components to RS.
+	 */
+	{
+		int i;
+		GLuint swiz, format, hw_format;
+		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+			if (fp_reads & FRAG_BIT_TEX(i)) {
+				switch (VB->TexCoordPtr[i]->size) {
+					case 1:
+						format = EMIT_1F;
+						hw_format = R300_DATA_TYPE_FLOAT_1;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+						break;
+					case 2:
+						format = EMIT_2F;
+						hw_format = R300_DATA_TYPE_FLOAT_2;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+						break;
+					case 3:
+						format = EMIT_3F;
+						hw_format = R300_DATA_TYPE_FLOAT_3;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+						break;
+					case 4:
+						format = EMIT_4F;
+						hw_format = R300_DATA_TYPE_FLOAT_4;
+						swiz = SWIZZLE_XYZW;
 						break;
-					}
+					default:
+						continue;
 				}
-			} else {
-				first_free_tex = 0;
+				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
+				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
+				EMIT_ATTR(_TNL_ATTRIB_TEX(i), format);
+				ADD_ATTR(VERT_ATTRIB_TEX0 + i, hw_format, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_XYZW, 0);
+				++first_free_tex;
 			}
 		}
-
-		if (first_free_tex == -1) {
-			fprintf(stderr, "\tout of free texcoords to write w pos\n");
-			_mesa_exit(-1);
-		}
-
-		sz = VB->AttribPtr[VERT_ATTRIB_POS]->size;
-		InputsRead |= 1 << (VERT_ATTRIB_TEX0 + first_free_tex);
-		OutputsWritten |= 1 << (VERT_RESULT_TEX0 + first_free_tex);
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_1F + sz - 1 );
-		vap_fmt_1 |= sz << (3 * first_free_tex);
-	}
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			inputs[i] = nr++;
-		} else {
-			inputs[i] = -1;
-		}
-	}
-
-	/* Fixed, apply to vir0 only */
-	if (InputsRead & (1 << VERT_ATTRIB_POS))
-		inputs[VERT_ATTRIB_POS] = 0;
-	if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
-		inputs[VERT_ATTRIB_COLOR0] = 2;
-	if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
-		inputs[VERT_ATTRIB_COLOR1] = 3;
-	if (InputsRead & (1 << VERT_ATTRIB_FOG))
-		inputs[VERT_ATTRIB_FOG] = 6 + fog_id;
-	for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
-		if (InputsRead & (1 << i))
-			inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			tab[nr++] = i;
-		}
 	}
 
-	for (i = 0; i < nr; i++) {
-		int ci;
-
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
-
-		for (ci = 0; ci < VB->AttribPtr[tab[i]]->size; ci++) {
-			swizzle[i][ci] = ci;
-		}
+	if (first_free_tex >= ctx->Const.MaxTextureUnits) {
+		fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
+		_mesa_exit(-1);
 	}
 
 	R300_NEWPRIM(rmesa);
-	R300_STATECHANGE(rmesa, vir[0]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
-		r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-				   VB->AttribPtr, inputs, tab, nr);
-	R300_STATECHANGE(rmesa, vir[1]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
-		r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-				   nr);
-
-	R300_STATECHANGE(rmesa, vic);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-
-	R300_STATECHANGE(rmesa, vof);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = vap_fmt_1;
-
-	rmesa->swtcl.vertex_size =
-		_tnl_install_attrs( ctx,
-				    rmesa->swtcl.vertex_attrs,
-				    rmesa->swtcl.vertex_attr_count,
-				    NULL, 0 );
+	rmesa->vbuf.num_attribs = num_attrs;
+	*_InputsRead = InputsRead;
+	*_OutputsWritten = OutputsWritten;
 
-	rmesa->swtcl.vertex_size /= 4;
-
-	RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
-
-
-	R300_STATECHANGE(rmesa, vte);
-	rmesa->hw.vte.cmd[1] = vte;
-	rmesa->hw.vte.cmd[2] = rmesa->swtcl.vertex_size;
+	RENDERINPUTS_COPY(rmesa->render_inputs_bitset, tnl->render_inputs_bitset);
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( r300ContextPtr rmesa  )
+static void r300PrepareVertices(GLcontext *ctx)
 {
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	rmesa->dma.flush = NULL;
-
-	if (rmesa->dma.current.buf) {
-		struct r300_dma_region *current = &rmesa->dma.current;
-		GLuint current_offset = GET_START(current);
-
-		assert (current->start +
-			rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-			current->ptr);
-
-		if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-
-			r300EnsureCmdBufSpace( rmesa, rmesa->hw.max_state_size + (12*sizeof(int)), __FUNCTION__);
-
-			r300EmitState(rmesa);
-
-			r300EmitVertexAOS( rmesa,
-					   rmesa->swtcl.vertex_size,
-					   current_offset);
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	GLuint InputsRead, OutputsWritten;
+	radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__);
 
-			r300EmitVbufPrim( rmesa,
-					  rmesa->swtcl.hw_primitive,
-					  rmesa->swtcl.numverts);
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
-			r300EmitCacheFlush(rmesa);
-		}
+	rmesa->radeon.swtcl.vertex_size =
+		_tnl_install_attrs( ctx,
+				    rmesa->radeon.swtcl.vertex_attrs,
+				    rmesa->radeon.swtcl.vertex_attr_count,
+				    NULL, 0 );
 
-		rmesa->swtcl.numverts = 0;
-		current->start = current->ptr;
-	}
+	rmesa->radeon.swtcl.vertex_size /= 4;
 }
 
-/* Alloc space in the current dma region.
- */
-static void *
-r300AllocDmaLowVerts( r300ContextPtr rmesa, int nverts, int vsize )
+static void r300_predict_emit_size( r300ContextPtr rmesa )
 {
-	GLuint bytes = vsize * nverts;
-
-	if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end )
-		r300RefillCurrentDmaRegion( rmesa, bytes);
-
-	if (!rmesa->dma.flush) {
-		rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-		rmesa->dma.flush = flush_last_swtcl_prim;
-	}
-
-	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-	ASSERT( rmesa->dma.flush == flush_last_swtcl_prim );
-	ASSERT( rmesa->dma.current.start +
-		rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-		rmesa->dma.current.ptr );
-
-	{
-		GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr);
-		rmesa->dma.current.ptr += bytes;
-		rmesa->swtcl.numverts += nverts;
-		return head;
+	if (!rmesa->radeon.swtcl.emit_prediction) {
+		const int vertex_size = 7;
+		const int prim_size = 3;
+		const int cache_flush_size = 4;
+		const int pre_emit_state = 4;
+		const int scissor_size = 3;
+		const int state_size = radeonCountStateEmitSize(&rmesa->radeon);
+
+		if (rcommonEnsureCmdBufSpace(&rmesa->radeon,
+					state_size + pre_emit_state + scissor_size
+					+ vertex_size + prim_size + cache_flush_size * 2,
+					__FUNCTION__))
+			rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon);
+		else
+			rmesa->radeon.swtcl.emit_prediction = state_size;
+
+		rmesa->radeon.swtcl.emit_prediction += rmesa->radeon.cmdbuf.cs->cdw
+			+ vertex_size + scissor_size + prim_size + cache_flush_size * 2 + pre_emit_state;
+		radeon_print(RADEON_SWRENDER, RADEON_VERBOSE,
+				"%s, size %d\n",
+				__func__, rmesa->radeon.cmdbuf.cs->cdw
+				+ vertex_size + scissor_size + prim_size + cache_flush_size * 2 + pre_emit_state);
 	}
 }
 
+
 static GLuint reduced_prim[] = {
-  GL_POINTS,
-  GL_LINES,
-  GL_LINES,
-  GL_LINES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
+	GL_POINTS,
+	GL_LINES,
+	GL_LINES,
+	GL_LINES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
 };
 
 static void r300RasterPrimitive( GLcontext *ctx, GLuint prim );
-static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
-//static void r300ResetLineStipple( GLcontext *ctx );
 
 /***********************************************************************
  *                    Emit primitives as inline vertices               *
@@ -402,18 +304,26 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
 #define HAVE_POLYGONS    1
 #define HAVE_ELTS        1
 
+static void* r300_alloc_verts(r300ContextPtr rmesa, GLuint n, GLuint size)
+{
+	void *rv;
+	do {
+		r300_predict_emit_size( rmesa );
+		rv = rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 );
+	} while (!rv);
+	return rv;
+}
+
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r300ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r300AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) r300_alloc_verts(rmesa, n, size);
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (r300Vertex *)(r300verts + ((x) * vertsize * sizeof(int)))
 #define VERTEX r300Vertex
-#define DO_DEBUG_VERTS (1 && (RADEON_DEBUG & DEBUG_VERTS))
-#define PRINT_VERTEX(x)
 #undef TAG
 #define TAG(x) r300_##x
 #include "tnl_dd/t_dd_triemit.h"
@@ -433,9 +343,8 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
  *              Build render functions from dd templates               *
  ***********************************************************************/
 
-#define R300_TWOSIDE_BIT	0x01
-#define R300_UNFILLED_BIT	0x02
-#define R300_MAX_TRIFUNC	0x04
+#define R300_UNFILLED_BIT	0x01
+#define R300_MAX_TRIFUNC	0x02
 
 static struct {
    tnl_points_func	        points;
@@ -446,9 +355,9 @@ static struct {
 
 #define DO_FALLBACK  0
 #define DO_UNFILLED (IND & R300_UNFILLED_BIT)
-#define DO_TWOSIDE  (IND & R300_TWOSIDE_BIT)
+#define DO_TWOSIDE   0
 #define DO_FLAT      0
-#define DO_OFFSET     0
+#define DO_OFFSET    0
 #define DO_TRI       1
 #define DO_QUAD      1
 #define DO_LINE      1
@@ -468,33 +377,39 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
-
-/* Only used to pull back colors into vertices (ie, we know color is
- * floating point).
- */
-#define R300_COLOR( dst, src )				\
-do {							\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);	\
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
+
+#define VERT_SET_RGBA( v, c ) \
+do { \
+   r300_color_t *color = (r300_color_t *)&((v)->ui[coloroffset]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]); \
 } while (0)
 
-#define VERT_SET_RGBA( v, c )    if (coloroffset) R300_COLOR( v->ub4[coloroffset], c )
-#define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
-#define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
-#define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
+#define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
+
+#define VERT_SET_SPEC( v0, c ) \
+do { \
+   if (specoffset) { \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.red, (c)[0]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.green, (c)[1]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.blue, (c)[2]); \
+   } \
+} while (0)
 
-#define R300_SPEC( dst, src )				\
-do {							\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);	\
+#define VERT_COPY_SPEC( v0, v1 ) \
+do { \
+   if (specoffset) { \
+       v0->v.specular.red = v1->v.specular.red; \
+       v0->v.specular.green = v1->v.specular.green; \
+       v0->v.specular.blue = v1->v.specular.blue; \
+   } \
 } while (0)
 
-#define VERT_SET_SPEC( v, c )    if (specoffset) R300_SPEC( v->ub4[specoffset], c )
-#define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
+#define VERT_SAVE_RGBA( idx )    color[idx] = v[idx]->ui[coloroffset]
+#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 
@@ -514,7 +429,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r300RasterPrimitive( ctx, reduced_prim[x] )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -530,26 +445,15 @@ do {							\
 #define TAG(x) x
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT)
-#define TAG(x) x##_twoside
-#include "tnl_dd/t_dd_tritmp.h"
-
 #define IND (R300_UNFILLED_BIT)
 #define TAG(x) x##_unfilled
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT|R300_UNFILLED_BIT)
-#define TAG(x) x##_twoside_unfilled
-#include "tnl_dd/t_dd_tritmp.h"
-
-
 
 static void init_rast_tab( void )
 {
    init();
-   init_twoside();
    init_unfilled();
-   init_twoside_unfilled();
 }
 
 /**********************************************************************/
@@ -571,8 +475,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -600,11 +504,11 @@ static void r300ChooseRenderState( GLcontext *ctx )
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	GLuint index = 0;
 	GLuint flags = ctx->_TriangleCaps;
+	radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __func__);
 
-	if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT;
 	if (flags & DD_TRI_UNFILLED)      index |= R300_UNFILLED_BIT;
 
-	if (index != rmesa->swtcl.RenderIndex) {
+	if (index != rmesa->radeon.swtcl.RenderIndex) {
 		tnl->Driver.Render.Points = rast_tab[index].points;
 		tnl->Driver.Render.Line = rast_tab[index].line;
 		tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -621,59 +525,64 @@ static void r300ChooseRenderState( GLcontext *ctx )
 			tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
 		}
 
-		rmesa->swtcl.RenderIndex = index;
+		rmesa->radeon.swtcl.RenderIndex = index;
 	}
 }
 
-
-static void r300RenderStart(GLcontext *ctx)
+void r300RenderStart(GLcontext *ctx)
 {
-        r300ContextPtr rmesa = R300_CONTEXT( ctx );
+	radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __func__);
+	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 
 	r300ChooseRenderState(ctx);
-	r300SetVertexFormat(ctx);
 
 	r300UpdateShaders(rmesa);
-	r300UpdateShaderStates(rmesa);
 
-	r300EmitCacheFlush(rmesa);
+	r300PrepareVertices(ctx);
+
+	r300ValidateBuffers(ctx);
+
+	r300UpdateShaderStates(rmesa);
 
-	if (rmesa->dma.flush != 0 &&
-	    rmesa->dma.flush != flush_last_swtcl_prim)
-		rmesa->dma.flush( rmesa );
 
+	/* investigate if we can put back flush optimisation if needed */
+	if (rmesa->radeon.dma.flush != NULL) {
+		rmesa->radeon.dma.flush(ctx);
+	}
 }
 
-static void r300RenderFinish(GLcontext *ctx)
+void r300RenderFinish(GLcontext *ctx)
 {
 }
 
 static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__);
 
-	if (rmesa->swtcl.hw_primitive != hwprim) {
-	        R300_NEWPRIM( rmesa );
-		rmesa->swtcl.hw_primitive = hwprim;
+	if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
+		R300_NEWPRIM( rmesa );
+		rmesa->radeon.swtcl.hw_primitive = hwprim;
 	}
 }
 
-static void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
+void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
 {
 
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	rmesa->swtcl.render_primitive = prim;
+	rmesa->radeon.swtcl.render_primitive = prim;
+	radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__);
 
 	if ((prim == GL_TRIANGLES) && (ctx->_TriangleCaps & DD_TRI_UNFILLED))
-	  return;
+		return;
 
 	r300RasterPrimitive( ctx, reduced_prim[prim] );
 }
 
-static void r300ResetLineStipple(GLcontext *ctx)
+void r300ResetLineStipple(GLcontext *ctx)
 {
-
-
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s\n", __func__);
 }
 
 void r300InitSwtcl(GLcontext *ctx)
@@ -681,11 +590,13 @@ void r300InitSwtcl(GLcontext *ctx)
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	static int firsttime = 1;
+	radeon_print(RADEON_SWRENDER, RADEON_NORMAL, "%s\n", __func__);
 
 	if (firsttime) {
 		init_rast_tab();
 		firsttime = 0;
 	}
+	rmesa->radeon.swtcl.emit_prediction = 0;
 
 	tnl->Driver.Render.Start = r300RenderStart;
 	tnl->Driver.Render.Finish = r300RenderFinish;
@@ -699,50 +610,75 @@ void r300InitSwtcl(GLcontext *ctx)
 	_tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
 			    48 * sizeof(GLfloat) );
 
-	rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-	rmesa->swtcl.RenderIndex = ~0;
-	rmesa->swtcl.render_primitive = GL_TRIANGLES;
-	rmesa->swtcl.hw_primitive = 0;
+	rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+	rmesa->radeon.swtcl.RenderIndex = ~0;
+	rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+	rmesa->radeon.swtcl.hw_primitive = 0;
 
 	_tnl_invalidate_vertex_state( ctx, ~0 );
 	_tnl_invalidate_vertices( ctx, ~0 );
-	RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
 
 	_tnl_need_projected_coords( ctx, GL_FALSE );
-	r300ChooseRenderState(ctx);
 }
 
 void r300DestroySwtcl(GLcontext *ctx)
 {
 }
 
-void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset)
+static void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-
-	drm_radeon_cmd_header_t *cmd = NULL;
-	if (RADEON_DEBUG & DEBUG_VERTS)
-	  fprintf(stderr, "%s:  vertex_size %d, offset 0x%x \n",
-		  __FUNCTION__, vertex_size, offset);
-
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2), 2);
-	e32(1);
-	e32(vertex_size | (vertex_size << 8));
-	e32(offset);
+	BATCH_LOCALS(&rmesa->radeon);
+
+	radeon_print(RADEON_SWRENDER, RADEON_TRACE,
+		"%s:  vertex_size %d, offset 0x%x \n",
+			__FUNCTION__, vertex_size, offset);
+
+	BEGIN_BATCH(7);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2);
+	OUT_BATCH(1);
+	OUT_BATCH(vertex_size | (vertex_size << 8));
+	OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+	END_BATCH();
 }
 
-void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
+static void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
 {
-
-	int cmd_reserved = 0;
-	int cmd_written = 0;
+	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
-	drm_radeon_cmd_header_t *cmd = NULL;
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s\n", __func__);
 
 	type = r300PrimitiveType(rmesa, primitive);
 	num_verts = r300NumVerts(rmesa, vertex_nr, primitive);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
+	BEGIN_BATCH(3);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
+	END_BATCH();
+}
+
+void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
+{
+	radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__);
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+	r300EmitCacheFlush(rmesa);
+
+	radeonEmitState(&rmesa->radeon);
+    r300_emit_scissor(ctx);
+	r300EmitVertexAOS(rmesa,
+			rmesa->radeon.swtcl.vertex_size,
+			first_elem(&rmesa->radeon.dma.reserved)->bo,
+			current_offset);
+
+	r300EmitVbufPrim(rmesa,
+		   rmesa->radeon.swtcl.hw_primitive,
+		   rmesa->radeon.swtcl.numverts);
+	r300EmitCacheFlush(rmesa);
+	if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw )
+		WARN_ONCE("Rendering was %d commands larger than predicted size."
+			" We might overflow  command buffer.\n",
+			rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction );
+	rmesa->radeon.swtcl.emit_prediction = 0;
+	COMMIT_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h
index 55df53c1ad..c271d26546 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.h
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.h
@@ -39,7 +39,27 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "r300_context.h"
 
+/*
+ * Here are definitions of OVM locations of vertex attributes for non TCL hw
+ */
+#define SWTCL_OVM_POS 0
+#define SWTCL_OVM_COLOR0 2
+#define SWTCL_OVM_COLOR1 3
+#define SWTCL_OVM_COLOR2 4
+#define SWTCL_OVM_COLOR3 5
+#define SWTCL_OVM_TEX(n) ((n) + 6)
+#define SWTCL_OVM_POINT_SIZE 15
+
+extern void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *InputsRead,  GLuint *OutputsWritten);
+
 extern void r300InitSwtcl( GLcontext *ctx );
 extern void r300DestroySwtcl( GLcontext *ctx );
 
+extern void r300RenderStart(GLcontext *ctx);
+extern void r300RenderFinish(GLcontext *ctx);
+extern void r300RenderPrimitive(GLcontext *ctx, GLenum prim);
+extern void r300ResetLineStipple(GLcontext *ctx);
+
+extern void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
+
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c
index 7c699ec572..433e5a87d4 100644
--- a/src/mesa/drivers/dri/r300/r300_tex.c
+++ b/src/mesa/drivers/dri/r300/r300_tex.c
@@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/enums.h"
 #include "main/image.h"
+#include "main/mipmap.h"
 #include "main/simple_list.h"
 #include "main/texformat.h"
 #include "main/texstore.h"
@@ -49,6 +50,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_tex.h"
 
 #include "xmlpool.h"
@@ -77,20 +79,20 @@ static unsigned int translate_wrap_mode(GLenum wrapmode)
  *
  * \param t Texture object whose wrap modes are to be set
  */
-static void r300UpdateTexWrap(r300TexObjPtr t)
+static void r300UpdateTexWrap(radeonTexObjPtr t)
 {
-	struct gl_texture_object *tObj = t->base.tObj;
+	struct gl_texture_object *tObj = &t->base;
 
-	t->filter &=
+	t->pp_txfilter &=
 	    ~(R300_TX_WRAP_S_MASK | R300_TX_WRAP_T_MASK | R300_TX_WRAP_R_MASK);
 
-	t->filter |= translate_wrap_mode(tObj->WrapS) << R300_TX_WRAP_S_SHIFT;
+	t->pp_txfilter |= translate_wrap_mode(tObj->WrapS) << R300_TX_WRAP_S_SHIFT;
 
 	if (tObj->Target != GL_TEXTURE_1D) {
-		t->filter |= translate_wrap_mode(tObj->WrapT) << R300_TX_WRAP_T_SHIFT;
+		t->pp_txfilter |= translate_wrap_mode(tObj->WrapT) << R300_TX_WRAP_T_SHIFT;
 
 		if (tObj->Target == GL_TEXTURE_3D)
-			t->filter |= translate_wrap_mode(tObj->WrapR) << R300_TX_WRAP_R_SHIFT;
+			t->pp_txfilter |= translate_wrap_mode(tObj->WrapR) << R300_TX_WRAP_R_SHIFT;
 	}
 }
 
@@ -117,10 +119,13 @@ static GLuint aniso_filter(GLfloat anisotropy)
  * \param magf Texture magnification mode
  * \param anisotropy Maximum anisotropy level
  */
-static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
+static void r300SetTexFilter(radeonTexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
 {
-	t->filter &= ~(R300_TX_MIN_FILTER_MASK | R300_TX_MIN_FILTER_MIP_MASK | R300_TX_MAG_FILTER_MASK | R300_TX_MAX_ANISO_MASK);
-	t->filter_1 &= ~R300_EDGE_ANISO_EDGE_ONLY;
+	/* Force revalidation to account for switches from/to mipmapping. */
+	t->validated = GL_FALSE;
+
+	t->pp_txfilter &= ~(R300_TX_MIN_FILTER_MASK | R300_TX_MIN_FILTER_MIP_MASK | R300_TX_MAG_FILTER_MASK | R300_TX_MAX_ANISO_MASK);
+	t->pp_txfilter_1 &= ~R300_EDGE_ANISO_EDGE_ONLY;
 
 	/* Note that EXT_texture_filter_anisotropic is extremely vague about
 	 * how anisotropic filtering interacts with the "normal" filter modes.
@@ -128,33 +133,33 @@ static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat
 	 * filter settings completely. This includes driconf's settings.
 	 */
 	if (anisotropy >= 2.0 && (minf != GL_NEAREST) && (magf != GL_NEAREST)) {
-		t->filter |= R300_TX_MAG_FILTER_ANISO
+		t->pp_txfilter |= R300_TX_MAG_FILTER_ANISO
 			| R300_TX_MIN_FILTER_ANISO
 			| R300_TX_MIN_FILTER_MIP_LINEAR
 			| aniso_filter(anisotropy);
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
+		if (RADEON_DEBUG & RADEON_TEXTURE)
 			fprintf(stderr, "Using maximum anisotropy of %f\n", anisotropy);
 		return;
 	}
 
 	switch (minf) {
 	case GL_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST;
 		break;
 	case GL_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR;
 		break;
 	case GL_NEAREST_MIPMAP_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_NEAREST;
 		break;
 	case GL_NEAREST_MIPMAP_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_LINEAR;
 		break;
 	case GL_LINEAR_MIPMAP_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_NEAREST;
 		break;
 	case GL_LINEAR_MIPMAP_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_LINEAR;
 		break;
 	}
 
@@ -163,15 +168,15 @@ static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat
 	 */
 	switch (magf) {
 	case GL_NEAREST:
-		t->filter |= R300_TX_MAG_FILTER_NEAREST;
+		t->pp_txfilter |= R300_TX_MAG_FILTER_NEAREST;
 		break;
 	case GL_LINEAR:
-		t->filter |= R300_TX_MAG_FILTER_LINEAR;
+		t->pp_txfilter |= R300_TX_MAG_FILTER_LINEAR;
 		break;
 	}
 }
 
-static void r300SetTexBorderColor(r300TexObjPtr t, const GLfloat color[4])
+static void r300SetTexBorderColor(radeonTexObjPtr t, const GLfloat color[4])
 {
 	GLubyte c[4];
 	CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
@@ -182,729 +187,6 @@ static void r300SetTexBorderColor(r300TexObjPtr t, const GLfloat color[4])
 }
 
 /**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static r300TexObjPtr r300AllocTexObj(struct gl_texture_object *texObj)
-{
-	r300TexObjPtr t;
-
-	t = CALLOC_STRUCT(r300_tex_obj);
-	texObj->DriverData = t;
-	if (t != NULL) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE) {
-			fprintf(stderr, "%s( %p, %p )\n", __FUNCTION__,
-				(void *)texObj, (void *)t);
-		}
-
-		/* Initialize non-image-dependent parts of the state:
-		 */
-		t->base.tObj = texObj;
-		t->border_fallback = GL_FALSE;
-
-		make_empty_list(&t->base);
-
-		r300UpdateTexWrap(t);
-		r300SetTexFilter(t, texObj->MinFilter, texObj->MagFilter, texObj->MaxAnisotropy);
-		r300SetTexBorderColor(t, texObj->BorderColor);
-	}
-
-	return t;
-}
-
-/* try to find a format which will only need a memcopy */
-static const struct gl_texture_format *r300Choose8888TexFormat(GLenum srcFormat,
-							       GLenum srcType)
-{
-	const GLuint ui = 1;
-	const GLubyte littleEndian = *((const GLubyte *)&ui);
-
-	if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-	    (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
-		return &_mesa_texformat_rgba8888;
-	} else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-		   (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
-		return &_mesa_texformat_rgba8888_rev;
-	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-					    srcType == GL_UNSIGNED_INT_8_8_8_8)) {
-		return &_mesa_texformat_argb8888_rev;
-	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-					    srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) {
-		return &_mesa_texformat_argb8888;
-	} else
-		return _dri_texformat_argb8888;
-}
-
-static const struct gl_texture_format *r300ChooseTextureFormat(GLcontext * ctx,
-							       GLint
-							       internalFormat,
-							       GLenum format,
-							       GLenum type)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	const GLboolean do32bpt =
-	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32);
-	const GLboolean force16bpt =
-	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16);
-	(void)format;
-
-#if 0
-	fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n",
-		_mesa_lookup_enum_by_nr(internalFormat), internalFormat,
-		_mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
-	fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt);
-#endif
-
-	switch (internalFormat) {
-	case 4:
-	case GL_RGBA:
-	case GL_COMPRESSED_RGBA:
-		switch (type) {
-		case GL_UNSIGNED_INT_10_10_10_2:
-		case GL_UNSIGNED_INT_2_10_10_10_REV:
-			return do32bpt ? _dri_texformat_argb8888 :
-			    _dri_texformat_argb1555;
-		case GL_UNSIGNED_SHORT_4_4_4_4:
-		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-			return _dri_texformat_argb4444;
-		case GL_UNSIGNED_SHORT_5_5_5_1:
-		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-			return _dri_texformat_argb1555;
-		default:
-			return do32bpt ? r300Choose8888TexFormat(format, type) :
-			    _dri_texformat_argb4444;
-		}
-
-	case 3:
-	case GL_RGB:
-	case GL_COMPRESSED_RGB:
-		switch (type) {
-		case GL_UNSIGNED_SHORT_4_4_4_4:
-		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-			return _dri_texformat_argb4444;
-		case GL_UNSIGNED_SHORT_5_5_5_1:
-		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-			return _dri_texformat_argb1555;
-		case GL_UNSIGNED_SHORT_5_6_5:
-		case GL_UNSIGNED_SHORT_5_6_5_REV:
-			return _dri_texformat_rgb565;
-		default:
-			return do32bpt ? _dri_texformat_argb8888 :
-			    _dri_texformat_rgb565;
-		}
-
-	case GL_RGBA8:
-	case GL_RGB10_A2:
-	case GL_RGBA12:
-	case GL_RGBA16:
-		return !force16bpt ?
-		    r300Choose8888TexFormat(format,
-					    type) : _dri_texformat_argb4444;
-
-	case GL_RGBA4:
-	case GL_RGBA2:
-		return _dri_texformat_argb4444;
-
-	case GL_RGB5_A1:
-		return _dri_texformat_argb1555;
-
-	case GL_RGB8:
-	case GL_RGB10:
-	case GL_RGB12:
-	case GL_RGB16:
-		return !force16bpt ? _dri_texformat_argb8888 :
-		    _dri_texformat_rgb565;
-
-	case GL_RGB5:
-	case GL_RGB4:
-	case GL_R3_G3_B2:
-		return _dri_texformat_rgb565;
-
-	case GL_ALPHA:
-	case GL_ALPHA4:
-	case GL_ALPHA8:
-	case GL_ALPHA12:
-	case GL_ALPHA16:
-	case GL_COMPRESSED_ALPHA:
-		return _dri_texformat_a8;
-
-	case 1:
-	case GL_LUMINANCE:
-	case GL_LUMINANCE4:
-	case GL_LUMINANCE8:
-	case GL_LUMINANCE12:
-	case GL_LUMINANCE16:
-	case GL_COMPRESSED_LUMINANCE:
-		return _dri_texformat_l8;
-
-	case 2:
-	case GL_LUMINANCE_ALPHA:
-	case GL_LUMINANCE4_ALPHA4:
-	case GL_LUMINANCE6_ALPHA2:
-	case GL_LUMINANCE8_ALPHA8:
-	case GL_LUMINANCE12_ALPHA4:
-	case GL_LUMINANCE12_ALPHA12:
-	case GL_LUMINANCE16_ALPHA16:
-	case GL_COMPRESSED_LUMINANCE_ALPHA:
-		return _dri_texformat_al88;
-
-	case GL_INTENSITY:
-	case GL_INTENSITY4:
-	case GL_INTENSITY8:
-	case GL_INTENSITY12:
-	case GL_INTENSITY16:
-	case GL_COMPRESSED_INTENSITY:
-		return _dri_texformat_i8;
-
-	case GL_YCBCR_MESA:
-		if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-		    type == GL_UNSIGNED_BYTE)
-			return &_mesa_texformat_ycbcr;
-		else
-			return &_mesa_texformat_ycbcr_rev;
-
-	case GL_RGB_S3TC:
-	case GL_RGB4_S3TC:
-	case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-		return &_mesa_texformat_rgb_dxt1;
-
-	case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-		return &_mesa_texformat_rgba_dxt1;
-
-	case GL_RGBA_S3TC:
-	case GL_RGBA4_S3TC:
-	case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-		return &_mesa_texformat_rgba_dxt3;
-
-	case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-		return &_mesa_texformat_rgba_dxt5;
-
-	case GL_ALPHA16F_ARB:
-		return &_mesa_texformat_alpha_float16;
-	case GL_ALPHA32F_ARB:
-		return &_mesa_texformat_alpha_float32;
-	case GL_LUMINANCE16F_ARB:
-		return &_mesa_texformat_luminance_float16;
-	case GL_LUMINANCE32F_ARB:
-		return &_mesa_texformat_luminance_float32;
-	case GL_LUMINANCE_ALPHA16F_ARB:
-		return &_mesa_texformat_luminance_alpha_float16;
-	case GL_LUMINANCE_ALPHA32F_ARB:
-		return &_mesa_texformat_luminance_alpha_float32;
-	case GL_INTENSITY16F_ARB:
-		return &_mesa_texformat_intensity_float16;
-	case GL_INTENSITY32F_ARB:
-		return &_mesa_texformat_intensity_float32;
-	case GL_RGB16F_ARB:
-		return &_mesa_texformat_rgba_float16;
-	case GL_RGB32F_ARB:
-		return &_mesa_texformat_rgba_float32;
-	case GL_RGBA16F_ARB:
-		return &_mesa_texformat_rgba_float16;
-	case GL_RGBA32F_ARB:
-		return &_mesa_texformat_rgba_float32;
-
-	case GL_DEPTH_COMPONENT:
-	case GL_DEPTH_COMPONENT16:
-	case GL_DEPTH_COMPONENT24:
-	case GL_DEPTH_COMPONENT32:
-#if 0
-		switch (type) {
-		case GL_UNSIGNED_BYTE:
-		case GL_UNSIGNED_SHORT:
-			return &_mesa_texformat_z16;
-		case GL_UNSIGNED_INT:
-			return &_mesa_texformat_z32;
-		case GL_UNSIGNED_INT_24_8_EXT:
-		default:
-			return &_mesa_texformat_z24_s8;
-		}
-#else
-		return &_mesa_texformat_z16;
-#endif
-
-	default:
-		_mesa_problem(ctx,
-			      "unexpected internalFormat 0x%x in r300ChooseTextureFormat",
-			      (int)internalFormat);
-		return NULL;
-	}
-
-	return NULL;		/* never get here */
-}
-
-static GLboolean
-r300ValidateClientStorage(GLcontext * ctx, GLenum target,
-			  GLint internalFormat,
-			  GLint srcWidth, GLint srcHeight,
-			  GLenum format, GLenum type, const void *pixels,
-			  const struct gl_pixelstore_attrib *packing,
-			  struct gl_texture_object *texObj,
-			  struct gl_texture_image *texImage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "intformat %s format %s type %s\n",
-			_mesa_lookup_enum_by_nr(internalFormat),
-			_mesa_lookup_enum_by_nr(format),
-			_mesa_lookup_enum_by_nr(type));
-
-	if (!ctx->Unpack.ClientStorage)
-		return 0;
-
-	if (ctx->_ImageTransferState ||
-	    texImage->IsCompressed || texObj->GenerateMipmap)
-		return 0;
-
-	/* This list is incomplete, may be different on ppc???
-	 */
-	switch (internalFormat) {
-	case GL_RGBA:
-		if (format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV) {
-			texImage->TexFormat = _dri_texformat_argb8888;
-		} else
-			return 0;
-		break;
-
-	case GL_RGB:
-		if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) {
-			texImage->TexFormat = _dri_texformat_rgb565;
-		} else
-			return 0;
-		break;
-
-	case GL_YCBCR_MESA:
-		if (format == GL_YCBCR_MESA &&
-		    type == GL_UNSIGNED_SHORT_8_8_REV_APPLE) {
-			texImage->TexFormat = &_mesa_texformat_ycbcr_rev;
-		} else if (format == GL_YCBCR_MESA &&
-			   (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-			    type == GL_UNSIGNED_BYTE)) {
-			texImage->TexFormat = &_mesa_texformat_ycbcr;
-		} else
-			return 0;
-		break;
-
-	default:
-		return 0;
-	}
-
-	/* Could deal with these packing issues, but currently don't:
-	 */
-	if (packing->SkipPixels ||
-	    packing->SkipRows || packing->SwapBytes || packing->LsbFirst) {
-		return 0;
-	}
-
-	GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth,
-						    format, type);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "%s: srcRowStride %d/%x\n",
-			__FUNCTION__, srcRowStride, srcRowStride);
-
-	/* Could check this later in upload, pitch restrictions could be
-	 * relaxed, but would need to store the image pitch somewhere,
-	 * as packing details might change before image is uploaded:
-	 */
-	if (!r300IsGartMemory(rmesa, pixels, srcHeight * srcRowStride)
-	    || (srcRowStride & 63))
-		return 0;
-
-	/* Have validated that _mesa_transfer_teximage would be a straight
-	 * memcpy at this point.  NOTE: future calls to TexSubImage will
-	 * overwrite the client data.  This is explicitly mentioned in the
-	 * extension spec.
-	 */
-	texImage->Data = (void *)pixels;
-	texImage->IsClientData = GL_TRUE;
-	texImage->RowStride = srcRowStride / texImage->TexFormat->TexelBytes;
-
-	return 1;
-}
-
-static void r300TexImage1D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-			return;
-		}
-	}
-
-	/* Note, this will call ChooseTextureFormat */
-	_mesa_store_teximage1d(ctx, target, level, internalFormat,
-			       width, border, format, type, pixels,
-			       &ctx->Unpack, texObj, texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-static void r300TexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
-			      GLint xoffset,
-			      GLsizei width,
-			      GLenum format, GLenum type,
-			      const GLvoid * pixels,
-			      const struct gl_pixelstore_attrib *packing,
-			      struct gl_texture_object *texObj,
-			      struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-			return;
-		}
-	}
-
-	_mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-				  format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-static void r300TexImage2D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint height, GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	if (t != NULL) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_teximage2d(ctx, target, level, internalFormat,
-				       width, height, border, format, type,
-				       pixels, &ctx->Unpack, texObj, texImage);
-
-		t->dirty_images[face] |= (1 << level);
-	}
-}
-
-static void r300TexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
-			      GLint xoffset, GLint yoffset,
-			      GLsizei width, GLsizei height,
-			      GLenum format, GLenum type,
-			      const GLvoid * pixels,
-			      const struct gl_pixelstore_attrib *packing,
-			      struct gl_texture_object *texObj,
-			      struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-			return;
-		}
-	}
-
-	_mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-				  height, format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[face] |= (1 << level);
-}
-
-static void r300CompressedTexImage2D(GLcontext * ctx, GLenum target,
-				     GLint level, GLint internalFormat,
-				     GLint width, GLint height, GLint border,
-				     GLsizei imageSize, const GLvoid * data,
-				     struct gl_texture_object *texObj,
-				     struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	if (t != NULL) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY,
-				    "glCompressedTexImage2D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-	/* can't call this, different parameters. Would never evaluate to true anyway currently */
-#if 0
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else
-#endif
-	{
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_compressed_teximage2d(ctx, target, level,
-						  internalFormat, width, height,
-						  border, imageSize, data,
-						  texObj, texImage);
-
-		t->dirty_images[face] |= (1 << level);
-	}
-}
-
-static void r300CompressedTexSubImage2D(GLcontext * ctx, GLenum target,
-					GLint level, GLint xoffset,
-					GLint yoffset, GLsizei width,
-					GLsizei height, GLenum format,
-					GLsizei imageSize, const GLvoid * data,
-					struct gl_texture_object *texObj,
-					struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY,
-				    "glCompressedTexSubImage3D");
-			return;
-		}
-	}
-
-	_mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset,
-					     yoffset, width, height, format,
-					     imageSize, data, texObj, texImage);
-
-	t->dirty_images[face] |= (1 << level);
-}
-
-static void r300TexImage3D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint height, GLint depth,
-			   GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-#if 0
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else
-#endif
-	{
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_teximage3d(ctx, target, level, internalFormat,
-				       width, height, depth, border,
-				       format, type, pixels,
-				       &ctx->Unpack, texObj, texImage);
-
-		t->dirty_images[0] |= (1 << level);
-	}
-}
-
-static void
-r300TexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
-		  GLint xoffset, GLint yoffset, GLint zoffset,
-		  GLsizei width, GLsizei height, GLsizei depth,
-		  GLenum format, GLenum type,
-		  const GLvoid * pixels,
-		  const struct gl_pixelstore_attrib *packing,
-		  struct gl_texture_object *texObj,
-		  struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-/*     fprintf(stderr, "%s\n", __FUNCTION__); */
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D");
-			return;
-		}
-		texObj->DriverData = t;
-	}
-
-	_mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
-				  width, height, depth,
-				  format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-/**
  * Changes variables and flags for a state update, which will happen at the
  * next UpdateTextureState
  */
@@ -913,9 +195,9 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 			     struct gl_texture_object *texObj,
 			     GLenum pname, const GLfloat * params)
 {
-	r300TexObjPtr t = (r300TexObjPtr) texObj->DriverData;
+	radeonTexObj* t = radeon_tex_obj(texObj);
 
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+	if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) {
 		fprintf(stderr, "%s( %s )\n", __FUNCTION__,
 			_mesa_lookup_enum_by_nr(pname));
 	}
@@ -946,7 +228,11 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 		 * we just have to rely on loading the right subset of mipmap levels
 		 * to simulate a clamped LOD.
 		 */
-		driSwapOutTextureObject((driTextureObject *) t);
+		if (t->mt) {
+			radeon_miptree_unreference(t->mt);
+			t->mt = 0;
+			t->validated = GL_FALSE;
+		}
 		break;
 
 	case GL_DEPTH_TEXTURE_MODE:
@@ -969,42 +255,35 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 	}
 }
 
-static void r300BindTexture(GLcontext * ctx, GLenum target,
-			    struct gl_texture_object *texObj)
-{
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
-		fprintf(stderr, "%s( %p ) unit=%d\n", __FUNCTION__,
-			(void *)texObj, ctx->Texture.CurrentUnit);
-	}
-
-	if ((target == GL_TEXTURE_1D)
-	    || (target == GL_TEXTURE_2D)
-	    || (target == GL_TEXTURE_3D)
-	    || (target == GL_TEXTURE_CUBE_MAP)
-	    || (target == GL_TEXTURE_RECTANGLE_NV)) {
-		assert(texObj->DriverData != NULL);
-	}
-}
-
 static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
+	radeonTexObj* t = radeon_tex_obj(texObj);
 
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+	if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) {
 		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
 			(void *)texObj,
 			_mesa_lookup_enum_by_nr(texObj->Target));
 	}
 
-	if (t != NULL) {
-		if (rmesa) {
-			R300_FIREVERTICES(rmesa);
-		}
+	if (rmesa) {
+		int i;
+		radeon_firevertices(&rmesa->radeon);
+
+		for(i = 0; i < R300_MAX_TEXTURE_UNITS; ++i)
+			if (rmesa->hw.textures[i] == t)
+				rmesa->hw.textures[i] = 0;
+	}
 
-		driDestroyTextureObject(t);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
 	}
-	/* Free mipmap images and the texture object itself */
 	_mesa_delete_texture_object(ctx, texObj);
 }
 
@@ -1013,8 +292,6 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
  * Called via ctx->Driver.NewTextureObject.
  * Note: this function will be called during context creation to
  * allocate the default texture objects.
- * Note: we could use containment here to 'derive' the driver-specific
- * texture object from the core mesa gl_texture_object.  Not done at this time.
  * Fixup MaxAnisotropy according to user preference.
  */
 static struct gl_texture_object *r300NewTextureObject(GLcontext * ctx,
@@ -1022,14 +299,23 @@ static struct gl_texture_object *r300NewTextureObject(GLcontext * ctx,
 						      GLenum target)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_object *obj;
-	obj = _mesa_new_texture_object(ctx, name, target);
-	if (!obj)
-		return NULL;
-	obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
+	radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
 
-	r300AllocTexObj(obj);
-	return obj;
+	if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) {
+		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+			t, _mesa_lookup_enum_by_nr(target));
+	}
+
+	_mesa_initialize_texture_object(&t->base, name, target);
+	t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+	/* Initialize hardware state */
+	r300UpdateTexWrap(t);
+	r300SetTexFilter(t, t->base.MinFilter, t->base.MagFilter, t->base.MaxAnisotropy);
+	r300SetTexBorderColor(t, t->base.BorderColor);
+
+	return &t->base;
 }
 
 void r300InitTextureFuncs(struct dd_function_table *functions)
@@ -1037,22 +323,30 @@ void r300InitTextureFuncs(struct dd_function_table *functions)
 	/* Note: we only plug in the functions we implement in the driver
 	 * since _mesa_init_driver_functions() was already called.
 	 */
-	functions->ChooseTextureFormat = r300ChooseTextureFormat;
-	functions->TexImage1D = r300TexImage1D;
-	functions->TexImage2D = r300TexImage2D;
-	functions->TexImage3D = r300TexImage3D;
-	functions->TexSubImage1D = r300TexSubImage1D;
-	functions->TexSubImage2D = r300TexSubImage2D;
-	functions->TexSubImage3D = r300TexSubImage3D;
+	functions->NewTextureImage = radeonNewTextureImage;
+	functions->FreeTexImageData = radeonFreeTexImageData;
+	functions->MapTexture = radeonMapTexture;
+	functions->UnmapTexture = radeonUnmapTexture;
+
+	functions->ChooseTextureFormat = radeonChooseTextureFormat_mesa;
+	functions->TexImage1D = radeonTexImage1D;
+	functions->TexImage2D = radeonTexImage2D;
+	functions->TexImage3D = radeonTexImage3D;
+	functions->TexSubImage1D = radeonTexSubImage1D;
+	functions->TexSubImage2D = radeonTexSubImage2D;
+	functions->TexSubImage3D = radeonTexSubImage3D;
+	functions->GetTexImage = radeonGetTexImage;
+	functions->GetCompressedTexImage = radeonGetCompressedTexImage;
 	functions->NewTextureObject = r300NewTextureObject;
-	functions->BindTexture = r300BindTexture;
 	functions->DeleteTexture = r300DeleteTexture;
 	functions->IsTextureResident = driIsTextureResident;
 
 	functions->TexParameter = r300TexParameter;
 
-	functions->CompressedTexImage2D = r300CompressedTexImage2D;
-	functions->CompressedTexSubImage2D = r300CompressedTexSubImage2D;
+	functions->CompressedTexImage2D = radeonCompressedTexImage2D;
+	functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D;
+
+	functions->GenerateMipmap = radeonGenerateMipmap;
 
 	driInitTextureFormats();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h
index b86d45bfe0..8a653ea2d1 100644
--- a/src/mesa/drivers/dri/r300/r300_tex.h
+++ b/src/mesa/drivers/dri/r300/r300_tex.h
@@ -37,16 +37,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 extern void r300SetDepthTexMode(struct gl_texture_object *tObj);
 
+extern void r300SetTexBuffer(__DRIcontext *pDRICtx, GLint target,
+			     __DRIdrawable *dPriv);
+
+extern void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
+			      GLint format, __DRIdrawable *dPriv);
+
 extern void r300SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 			     unsigned long long offset, GLint depth,
 			     GLuint pitch);
 
-extern void r300UpdateTextureState(GLcontext * ctx);
-
-extern int r300UploadTexImages(r300ContextPtr rmesa, r300TexObjPtr t,
-			       GLuint face);
-
-extern void r300DestroyTexObj(r300ContextPtr rmesa, r300TexObjPtr t);
+extern GLboolean r300ValidateBuffers(GLcontext * ctx);
 
 extern void r300InitTextureFuncs(struct dd_function_table *functions);
 
diff --git a/src/mesa/drivers/dri/r300/r300_texmem.c b/src/mesa/drivers/dri/r300/r300_texmem.c
deleted file mode 100644
index a89ab83d94..0000000000
--- a/src/mesa/drivers/dri/r300/r300_texmem.c
+++ /dev/null
@@ -1,568 +0,0 @@
-/**************************************************************************
-
-Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.
-The Weather Channel, Inc. funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86
-license. This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/**
- * \file
- *
- * \author Gareth Hughes <gareth@valinux.com>
- *
- * \author Kevin E. Martin <martin@valinux.com>
- */
-
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "main/simple_list.h"
-#include "main/texobj.h"
-#include "radeon_reg.h"		/* gets definition for usleep */
-#include "r300_context.h"
-#include "r300_state.h"
-#include "r300_cmdbuf.h"
-#include "radeon_ioctl.h"
-#include "r300_tex.h"
-#include "r300_ioctl.h"
-#include <unistd.h>		/* for usleep() */
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void r300DestroyTexObj(r300ContextPtr rmesa, r300TexObjPtr t)
-{
-	int i;
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE) {
-		fprintf(stderr, "%s( %p, %p )\n", __FUNCTION__,
-			(void *)t, (void *)t->base.tObj);
-	}
-
-	for (i = 0; i < rmesa->radeon.glCtx->Const.MaxTextureUnits; i++) {
-		if (rmesa->state.texture.unit[i].texobj == t->base.tObj) {
-			_mesa_reference_texobj(&rmesa->state.texture.unit[i].texobj, NULL);
-		}
-	}
-}
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-static void r300UploadGARTClientSubImage(r300ContextPtr rmesa,
-					 r300TexObjPtr t,
-					 struct gl_texture_image *texImage,
-					 GLint hwlevel,
-					 GLint x, GLint y,
-					 GLint width, GLint height)
-{
-	const struct gl_texture_format *texFormat = texImage->TexFormat;
-	GLuint srcPitch, dstPitch;
-	int blit_format;
-	int srcOffset;
-
-	/*
-	 * XXX it appears that we always upload the full image, not a subimage.
-	 * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
-	 * changed, the src pitch will have to change.
-	 */
-	switch (texFormat->TexelBytes) {
-	case 1:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 2:
-		blit_format = R300_CP_COLOR_FORMAT_RGB565;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 4:
-		blit_format = R300_CP_COLOR_FORMAT_ARGB8888;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 8:
-	case 16:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	default:
-		return;
-	}
-
-	t->image[0][hwlevel].data = texImage->Data;
-	srcOffset = r300GartOffsetFromVirtual(rmesa, texImage->Data);
-
-	assert(srcOffset != ~0);
-
-	/* Don't currently need to cope with small pitches?
-	 */
-	width = texImage->Width;
-	height = texImage->Height;
-
-	if (texFormat->TexelBytes > 4) {
-		width *= texFormat->TexelBytes;
-	}
-
-	r300EmitWait(rmesa, R300_WAIT_3D);
-
-	r300EmitBlit(rmesa, blit_format,
-		     srcPitch,
-		     srcOffset,
-		     dstPitch,
-		     t->bufAddr,
-		     x,
-		     y,
-		     t->image[0][hwlevel].x + x,
-		     t->image[0][hwlevel].y + y, width, height);
-
-	r300EmitWait(rmesa, R300_WAIT_2D);
-}
-
-static void r300UploadRectSubImage(r300ContextPtr rmesa,
-				   r300TexObjPtr t,
-				   struct gl_texture_image *texImage,
-				   GLint x, GLint y, GLint width, GLint height)
-{
-	const struct gl_texture_format *texFormat = texImage->TexFormat;
-	int blit_format, dstPitch, done;
-
-	switch (texFormat->TexelBytes) {
-	case 1:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		break;
-	case 2:
-		blit_format = R300_CP_COLOR_FORMAT_RGB565;
-		break;
-	case 4:
-		blit_format = R300_CP_COLOR_FORMAT_ARGB8888;
-		break;
-	case 8:
-	case 16:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		break;
-	default:
-		return;
-	}
-
-	t->image[0][0].data = texImage->Data;
-
-	/* Currently don't need to cope with small pitches.
-	 */
-	width = texImage->Width;
-	height = texImage->Height;
-	dstPitch = t->pitch;
-
-	if (texFormat->TexelBytes > 4) {
-		width *= texFormat->TexelBytes;
-	}
-
-	if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
-		/* In this case, could also use GART texturing.  This is
-		 * currently disabled, but has been tested & works.
-		 */
-		t->offset = r300GartOffsetFromVirtual(rmesa, texImage->Data);
-		t->pitch = texImage->RowStride * texFormat->TexelBytes - 32;
-
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr,
-				"Using GART texturing for rectangular client texture\n");
-
-		/* Release FB memory allocated for this image:
-		 */
-		/* FIXME This may not be correct as driSwapOutTextureObject sets
-		 * FIXME dirty_images.  It may be fine, though.
-		 */
-		if (t->base.memBlock) {
-			driSwapOutTextureObject((driTextureObject *) t);
-		}
-	} else if (texImage->IsClientData) {
-		/* Data already in GART memory, with usable pitch.
-		 */
-		GLuint srcPitch;
-		srcPitch = texImage->RowStride * texFormat->TexelBytes;
-		r300EmitBlit(rmesa,
-			     blit_format,
-			     srcPitch,
-			     r300GartOffsetFromVirtual(rmesa, texImage->Data),
-			     dstPitch, t->bufAddr, 0, 0, 0, 0, width, height);
-	} else {
-		/* Data not in GART memory, or bad pitch.
-		 */
-		for (done = 0; done < height;) {
-			struct r300_dma_region region;
-			int lines =
-			    MIN2(height - done, RADEON_BUFFER_SIZE / dstPitch);
-			int src_pitch;
-			char *tex;
-
-			src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-			tex = (char *)texImage->Data + done * src_pitch;
-
-			memset(&region, 0, sizeof(region));
-			r300AllocDmaRegion(rmesa, &region, lines * dstPitch,
-					   1024);
-
-			/* Copy texdata to dma:
-			 */
-			if (RADEON_DEBUG & DEBUG_TEXTURE)
-				fprintf(stderr,
-					"%s: src_pitch %d dst_pitch %d\n",
-					__FUNCTION__, src_pitch, dstPitch);
-
-			if (src_pitch == dstPitch) {
-				memcpy(region.address + region.start, tex,
-				       lines * src_pitch);
-			} else {
-				char *buf = region.address + region.start;
-				int i;
-				for (i = 0; i < lines; i++) {
-					memcpy(buf, tex, src_pitch);
-					buf += dstPitch;
-					tex += src_pitch;
-				}
-			}
-
-			r300EmitWait(rmesa, R300_WAIT_3D);
-
-			/* Blit to framebuffer
-			 */
-			r300EmitBlit(rmesa,
-				     blit_format,
-				     dstPitch, GET_START(&region),
-				     dstPitch | (t->tile_bits >> 16),
-				     t->bufAddr, 0, 0, 0, done, width, lines);
-
-			r300EmitWait(rmesa, R300_WAIT_2D);
-#ifdef USER_BUFFERS
-			r300_mem_use(rmesa, region.buf->id);
-#endif
-
-			r300ReleaseDmaRegion(rmesa, &region, __FUNCTION__);
-			done += lines;
-		}
-	}
-}
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void r300UploadSubImage(r300ContextPtr rmesa, r300TexObjPtr t,
-			       GLint hwlevel,
-			       GLint x, GLint y, GLint width, GLint height,
-			       GLuint face)
-{
-	struct gl_texture_image *texImage = NULL;
-	GLuint offset;
-	GLint imageWidth, imageHeight;
-	GLint ret;
-	drm_radeon_texture_t tex;
-	drm_radeon_tex_image_t tmp;
-	const int level = hwlevel + t->base.firstLevel;
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE) {
-		fprintf(stderr,
-			"%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n",
-			__FUNCTION__, (void *)t, (void *)t->base.tObj, level,
-			width, height, face);
-	}
-
-	ASSERT(face < 6);
-
-	/* Ensure we have a valid texture to upload */
-	if ((hwlevel < 0) || (hwlevel >= R300_MAX_TEXTURE_LEVELS)) {
-		_mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-		return;
-	}
-
-	texImage = t->base.tObj->Image[face][level];
-
-	if (!texImage) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: texImage %d is NULL!\n",
-				__FUNCTION__, level);
-		return;
-	}
-	if (!texImage->Data) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: image data is NULL!\n",
-				__FUNCTION__);
-		return;
-	}
-
-	if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		assert(level == 0);
-		assert(hwlevel == 0);
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: image data is rectangular\n",
-				__FUNCTION__);
-		r300UploadRectSubImage(rmesa, t, texImage, x, y, width, height);
-		return;
-	} else if (texImage->IsClientData) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr,
-				"%s: image data is in GART client storage\n",
-				__FUNCTION__);
-		r300UploadGARTClientSubImage(rmesa, t, texImage, hwlevel, x, y,
-					     width, height);
-		return;
-	} else if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "%s: image data is in normal memory\n",
-			__FUNCTION__);
-
-	imageWidth = texImage->Width;
-	imageHeight = texImage->Height;
-
-	offset = t->bufAddr;
-
-	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
-		GLint imageX = 0;
-		GLint imageY = 0;
-		GLint blitX = t->image[face][hwlevel].x;
-		GLint blitY = t->image[face][hwlevel].y;
-		GLint blitWidth = t->image[face][hwlevel].width;
-		GLint blitHeight = t->image[face][hwlevel].height;
-		fprintf(stderr, "   upload image: %d,%d at %d,%d\n",
-			imageWidth, imageHeight, imageX, imageY);
-		fprintf(stderr, "   upload  blit: %d,%d at %d,%d\n",
-			blitWidth, blitHeight, blitX, blitY);
-		fprintf(stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-			(GLuint) offset, hwlevel, level);
-	}
-
-	t->image[face][hwlevel].data = texImage->Data;
-
-	/* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-	 * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-	 * We used to use 1, 2 and 4-byte texels and used to use the texture
-	 * width to dictate the blit width - but that won't work for compressed
-	 * textures. (Brian)
-	 * NOTE: can't do that with texture tiling. (sroland)
-	 */
-	tex.offset = offset;
-	tex.image = &tmp;
-	/* copy (x,y,width,height,data) */
-	memcpy(&tmp, &t->image[face][hwlevel], sizeof(tmp));
-
-	if (texImage->TexFormat->TexelBytes > 4) {
-		const int log2TexelBytes =
-		    (3 + (texImage->TexFormat->TexelBytes >> 4));
-		tex.format = RADEON_TXFORMAT_I8;	/* any 1-byte texel format */
-		tex.pitch =
-		    MAX2((texImage->Width * texImage->TexFormat->TexelBytes) /
-			 64, 1);
-		tex.height = imageHeight;
-		tex.width = imageWidth << log2TexelBytes;
-		tex.offset += (tmp.x << log2TexelBytes) & ~1023;
-		tmp.x = tmp.x % (1024 >> log2TexelBytes);
-		tmp.width = tmp.width << log2TexelBytes;
-	} else if (texImage->TexFormat->TexelBytes) {
-		/* use multi-byte upload scheme */
-		tex.height = imageHeight;
-		tex.width = imageWidth;
-		switch (texImage->TexFormat->TexelBytes) {
-		case 1:
-			tex.format = RADEON_TXFORMAT_I8;
-			break;
-		case 2:
-			tex.format = RADEON_TXFORMAT_AI88;
-			break;
-		case 4:
-			tex.format = RADEON_TXFORMAT_ARGB8888;
-			break;
-		}
-		tex.pitch =
-		    MAX2((texImage->Width * texImage->TexFormat->TexelBytes) /
-			 64, 1);
-		tex.offset += tmp.x & ~1023;
-		tmp.x = tmp.x % 1024;
-
-		if (t->tile_bits & R300_TXO_MICRO_TILE) {
-			/* need something like "tiled coordinates" ? */
-			tmp.y = tmp.x / (tex.pitch * 128) * 2;
-			tmp.x =
-			    tmp.x % (tex.pitch * 128) / 2 /
-			    texImage->TexFormat->TexelBytes;
-			tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-		} else {
-			tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-		}
-#if 1
-		if ((t->tile_bits & R300_TXO_MACRO_TILE) &&
-		    (texImage->Width * texImage->TexFormat->TexelBytes >= 256)
-		    && ((!(t->tile_bits & R300_TXO_MICRO_TILE)
-			 && (texImage->Height >= 8))
-			|| (texImage->Height >= 16))) {
-			/* weird: R200 disables macro tiling if mip width is smaller than 256 bytes,
-			   OR if height is smaller than 8 automatically, but if micro tiling is active
-			   the limit is height 16 instead ? */
-			tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-		}
-#endif
-	} else {
-		/* In case of for instance 8x8 texture (2x2 dxt blocks),
-		   padding after the first two blocks is needed (only
-		   with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-		/* set tex.height to 1/4 since 1 "macropixel" (dxt-block)
-		   has 4 real pixels. Needed so the kernel module reads
-		   the right amount of data. */
-		tex.format = RADEON_TXFORMAT_I8;	/* any 1-byte texel format */
-		tex.pitch = (R300_BLIT_WIDTH_BYTES / 64);
-		tex.height = (imageHeight + 3) / 4;
-		tex.width = (imageWidth + 3) / 4;
-		if ((t->format & R300_TX_FORMAT_DXT1) == R300_TX_FORMAT_DXT1) {
-			tex.width *= 8;
-		} else {
-			tex.width *= 16;
-		}
-	}
-
-	LOCK_HARDWARE(&rmesa->radeon);
-	do {
-		ret =
-		    drmCommandWriteRead(rmesa->radeon.dri.fd,
-					DRM_RADEON_TEXTURE, &tex,
-					sizeof(drm_radeon_texture_t));
-		if (ret) {
-			if (RADEON_DEBUG & DEBUG_IOCTL)
-				fprintf(stderr,
-					"DRM_RADEON_TEXTURE:  again!\n");
-			usleep(1);
-		}
-	} while (ret == -EAGAIN);
-
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	if (ret) {
-		fprintf(stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret);
-		fprintf(stderr, "   offset=0x%08x\n", offset);
-		fprintf(stderr, "   image width=%d height=%d\n",
-			imageWidth, imageHeight);
-		fprintf(stderr, "    blit width=%d height=%d data=%p\n",
-			t->image[face][hwlevel].width,
-			t->image[face][hwlevel].height,
-			t->image[face][hwlevel].data);
-		_mesa_exit(-1);
-	}
-}
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- *
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int r300UploadTexImages(r300ContextPtr rmesa, r300TexObjPtr t, GLuint face)
-{
-	const int numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-	if (t->image_override)
-		return 0;
-
-	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
-		fprintf(stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-			(void *)rmesa->radeon.glCtx, (void *)t->base.tObj,
-			t->base.totalSize, t->base.firstLevel,
-			t->base.lastLevel);
-	}
-
-	if (t->base.totalSize == 0)
-		return 0;
-
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-		radeonFinish(rmesa->radeon.glCtx);
-	}
-
-	LOCK_HARDWARE(&rmesa->radeon);
-
-	if (t->base.memBlock == NULL) {
-		int heap;
-
-		heap = driAllocateTexture(rmesa->texture_heaps, rmesa->nr_heaps,
-					  (driTextureObject *) t);
-		if (heap == -1) {
-			UNLOCK_HARDWARE(&rmesa->radeon);
-			return -1;
-		}
-
-		/* Set the base offset of the texture image */
-		t->bufAddr = rmesa->radeon.radeonScreen->texOffset[heap]
-		    + t->base.memBlock->ofs;
-		t->offset = t->bufAddr;
-
-		if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-			/* hope it's safe to add that here... */
-			t->offset |= t->tile_bits;
-		}
-	}
-
-	/* Let the world know we've used this memory recently.
-	 */
-	driUpdateTextureLRU((driTextureObject *) t);
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	/* Upload any images that are new */
-	if (t->base.dirty_images[face]) {
-		int i;
-		for (i = 0; i < numLevels; i++) {
-			if ((t->base.
-			     dirty_images[face] & (1 <<
-						   (i + t->base.firstLevel))) !=
-			    0) {
-				r300UploadSubImage(rmesa, t, i, 0, 0,
-						   t->image[face][i].width,
-						   t->image[face][i].height,
-						   face);
-			}
-		}
-		t->base.dirty_images[face] = 0;
-	}
-
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-		radeonFinish(rmesa->radeon.glCtx);
-	}
-
-	return 0;
-}
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index f6ae4b675b..f030451b28 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -43,11 +43,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/teximage.h"
 #include "main/texobj.h"
 #include "main/enums.h"
+#include "main/simple_list.h"
 
 #include "r300_context.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
-#include "radeon_ioctl.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_tex.h"
 #include "r300_reg.h"
 
@@ -117,7 +118,12 @@ static const struct tx_table {
 	_ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)),
 	_ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)),
 	_ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)),
+	_ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)),
 	_ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)),
+	/* EXT_texture_sRGB */
+	_ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA),
 	/* *INDENT-ON* */
 };
 
@@ -143,13 +149,12 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 		},
 	};
 	const GLuint *format;
-	r300TexObjPtr t;
+	radeonTexObjPtr t;
 
 	if (!tObj)
 		return;
 
-	t = (r300TexObjPtr) tObj->DriverData;
-
+	t = radeon_tex_obj(tObj);
 
 	switch (tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat) {
 	case MESA_FORMAT_Z16:
@@ -171,13 +176,13 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 
 	switch (tObj->DepthMode) {
 	case GL_LUMINANCE:
-		t->format = format[0];
+		t->pp_txformat = format[0];
 		break;
 	case GL_INTENSITY:
-		t->format = format[1];
+		t->pp_txformat = format[1];
 		break;
 	case GL_ALPHA:
-		t->format = format[2];
+		t->pp_txformat = format[2];
 		break;
 	default:
 		/* Error...which should have already been caught by higher
@@ -190,486 +195,307 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 
 
 /**
- * Compute sizes and fill in offset and blit information for the given
- * image (determined by \p face and \p level).
- *
- * \param curOffset points to the offset at which the image is to be stored
- * and is updated by this function according to the size of the image.
- */
-static void compute_tex_image_offset(
-	struct gl_texture_object *tObj,
-	GLuint face,
-	GLint level,
-	GLint* curOffset)
-{
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	const struct gl_texture_image* texImage;
-	GLuint blitWidth = R300_BLIT_WIDTH_BYTES;
-	GLuint texelBytes;
-	GLuint size;
-
-	texImage = tObj->Image[0][level + t->base.firstLevel];
-	if (!texImage)
-		return;
-
-	texelBytes = texImage->TexFormat->TexelBytes;
-
-	/* find image size in bytes */
-	if (texImage->IsCompressed) {
-		if ((t->format & R300_TX_FORMAT_DXT1) ==
-			R300_TX_FORMAT_DXT1) {
-			// fprintf(stderr,"DXT 1 %d %08X\n", texImage->Width, t->format);
-			if ((texImage->Width + 3) < 8)	/* width one block */
-				size = texImage->CompressedSize * 4;
-			else if ((texImage->Width + 3) < 16)
-				size = texImage->CompressedSize * 2;
-			else
-				size = texImage->CompressedSize;
-		} else {
-			/* DXT3/5, 16 bytes per block */
-			WARN_ONCE
-				("DXT 3/5 suffers from multitexturing problems!\n");
-			// fprintf(stderr,"DXT 3/5 %d\n", texImage->Width);
-			if ((texImage->Width + 3) < 8)
-				size = texImage->CompressedSize * 2;
-			else
-				size = texImage->CompressedSize;
-		}
-	} else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		size =
-			((texImage->Width * texelBytes +
-			63) & ~63) * texImage->Height;
-		blitWidth = 64 / texelBytes;
-	} else if (t->tile_bits & R300_TXO_MICRO_TILE) {
-		/* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-			though the actual offset may be different (if texture is less than
-			32 bytes width) to the untiled case */
-		int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-		size =
-			(w * ((texImage->Height + 1) / 2)) *
-			texImage->Depth;
-		blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-	} else {
-		int w = (texImage->Width * texelBytes + 31) & ~31;
-		size = w * texImage->Height * texImage->Depth;
-		blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-	}
-	assert(size > 0);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "w=%d h=%d d=%d tb=%d intFormat=%d\n",
-			texImage->Width, texImage->Height,
-			texImage->Depth,
-			texImage->TexFormat->TexelBytes,
-			texImage->InternalFormat);
-
-	/* All images are aligned to a 32-byte offset */
-	*curOffset = (*curOffset + 0x1f) & ~0x1f;
-
-	if (texelBytes) {
-		/* fix x and y coords up later together with offset */
-		t->image[face][level].x = *curOffset;
-		t->image[face][level].y = 0;
-		t->image[face][level].width =
-			MIN2(size / texelBytes, blitWidth);
-		t->image[face][level].height =
-			(size / texelBytes) / t->image[face][level].width;
-	} else {
-		t->image[face][level].x = *curOffset % R300_BLIT_WIDTH_BYTES;
-		t->image[face][level].y = *curOffset / R300_BLIT_WIDTH_BYTES;
-		t->image[face][level].width =
-			MIN2(size, R300_BLIT_WIDTH_BYTES);
-		t->image[face][level].height = size / t->image[face][level].width;
-	}
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr,
-			"level %d, face %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-			level, face, texImage->Width, texImage->Height,
-			t->image[face][level].x, t->image[face][level].y,
-			t->image[face][level].width, t->image[face][level].height,
-			size, *curOffset);
-
-	*curOffset += size;
-}
-
-
-
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c filter, \c format, etc. will be set here
- * too.
+ * Compute the cached hardware register values for the given texture object.
  *
  * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
+ * \param t the r300 texture object
  */
-static void r300SetTexImages(r300ContextPtr rmesa,
-			     struct gl_texture_object *tObj)
+static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t)
 {
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	const struct gl_texture_image *baseImage =
-	    tObj->Image[0][tObj->BaseLevel];
-	GLint curOffset;
-	GLint i, texelBytes;
-	GLint numLevels;
-	GLint log2Width, log2Height, log2Depth;
-
-	/* Set the hardware texture format
-	 */
+	const struct gl_texture_image *firstImage;
+	int firstlevel = t->mt ? t->mt->firstLevel : 0;
+	    
+	firstImage = t->base.Image[0][firstlevel];
+
 	if (!t->image_override
-	    && VALID_FORMAT(baseImage->TexFormat->MesaFormat)) {
-		if (baseImage->TexFormat->BaseFormat == GL_DEPTH_COMPONENT) {
-			r300SetDepthTexMode(tObj);
+	    && VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+		if (firstImage->TexFormat->BaseFormat == GL_DEPTH_COMPONENT) {
+			r300SetDepthTexMode(&t->base);
 		} else {
-			t->format = tx_table[baseImage->TexFormat->MesaFormat].format;
+			t->pp_txformat = tx_table[firstImage->TexFormat->MesaFormat].format;
 		}
 
-		t->filter |= tx_table[baseImage->TexFormat->MesaFormat].filter;
+		t->pp_txfilter |= tx_table[firstImage->TexFormat->MesaFormat].filter;
 	} else if (!t->image_override) {
 		_mesa_problem(NULL, "unexpected texture format in %s",
 			      __FUNCTION__);
 		return;
 	}
 
-	texelBytes = baseImage->TexFormat->TexelBytes;
-
-	/* Compute which mipmap levels we really want to send to the hardware.
-	 */
-	driCalculateTextureFirstLastLevel((driTextureObject *) t);
-	log2Width = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-	log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-	log2Depth = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-	numLevels = t->base.lastLevel - t->base.firstLevel + 1;
+	if (t->image_override && t->bo)
+		return;
 
-	assert(numLevels <= R300_MAX_TEXTURE_LEVELS);
+	t->pp_txsize = (((firstImage->Width - 1) << R300_TX_WIDTHMASK_SHIFT)
+			| ((firstImage->Height - 1) << R300_TX_HEIGHTMASK_SHIFT)
+			| ((firstImage->DepthLog2) << R300_TX_DEPTHMASK_SHIFT)
+			| ((t->mt->lastLevel - t->mt->firstLevel) << R300_TX_MAX_MIP_LEVEL_SHIFT));
 
-	/* Calculate mipmap offsets and dimensions for blitting (uploading)
-	 * The idea is that we lay out the mipmap levels within a block of
-	 * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-	 */
 	t->tile_bits = 0;
 
-	/* figure out if this texture is suitable for tiling. */
-#if 0				/* Disabled for now */
-	if (texelBytes) {
-		if ((tObj->Target != GL_TEXTURE_RECTANGLE_NV) &&
-		    /* texrect might be able to use micro tiling too in theory? */
-		    (baseImage->Height > 1)) {
-
-			/* allow 32 (bytes) x 1 mip (which will use two times the space
-			   the non-tiled version would use) max if base texture is large enough */
-			if ((numLevels == 1) ||
-			    (((baseImage->Width * texelBytes /
-			       baseImage->Height) <= 32)
-			     && (baseImage->Width * texelBytes > 64))
-			    ||
-			    ((baseImage->Width * texelBytes /
-			      baseImage->Height) <= 16)) {
-				t->tile_bits |= R300_TXO_MICRO_TILE;
-			}
-		}
-
-		if (tObj->Target != GL_TEXTURE_RECTANGLE_NV) {
-			/* we can set macro tiling even for small textures, they will be untiled anyway */
-			t->tile_bits |= R300_TXO_MACRO_TILE;
-		}
-	}
-#endif
-
-	curOffset = 0;
+	if (t->base.Target == GL_TEXTURE_CUBE_MAP)
+		t->pp_txformat |= R300_TX_FORMAT_CUBIC_MAP;
+	if (t->base.Target == GL_TEXTURE_3D)
+		t->pp_txformat |= R300_TX_FORMAT_3D;
 
-	if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-		ASSERT(log2Width == log2Height);
-		t->format |= R300_TX_FORMAT_CUBIC_MAP;
 
-		for(i = 0; i < numLevels; i++) {
-			GLuint face;
-			for(face = 0; face < 6; face++)
-				compute_tex_image_offset(tObj, face, i, &curOffset);
-		}
-	} else {
-		if (tObj->Target == GL_TEXTURE_3D)
-                	t->format |= R300_TX_FORMAT_3D;
-
-		for (i = 0; i < numLevels; i++)
-			compute_tex_image_offset(tObj, 0, i, &curOffset);
-	}
-
-	/* Align the total size of texture memory block.
-	 */
-	t->base.totalSize =
-	    (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-	t->size =
-	    (((tObj->Image[0][t->base.firstLevel]->Width -
-	       1) << R300_TX_WIDTHMASK_SHIFT)
-	     | ((tObj->Image[0][t->base.firstLevel]->Height - 1) <<
-		R300_TX_HEIGHTMASK_SHIFT)
-	     | ((tObj->Image[0][t->base.firstLevel]->DepthLog2) <<
-		R300_TX_DEPTHMASK_SHIFT))
-	    | ((numLevels - 1) << R300_TX_MAX_MIP_LEVEL_SHIFT);
-
-	t->pitch = 0;
-
-	/* Only need to round to nearest 32 for textures, but the blitter
-	 * requires 64-byte aligned pitches, and we may/may not need the
-	 * blitter.   NPOT only!
-	 */
-	if (baseImage->IsCompressed) {
-		t->pitch |=
-		    (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-	} else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		unsigned int align = (64 / texelBytes) - 1;
-		t->pitch |= ((tObj->Image[0][t->base.firstLevel]->Width *
-			     texelBytes) + 63) & ~(63);
-		t->size |= R300_TX_SIZE_TXPITCH_EN;
+	if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
+		unsigned int align = (64 / t->mt->bpp) - 1;
+		t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN;
 		if (!t->image_override)
-			t->pitch_reg =
-			    (((tObj->Image[0][t->base.firstLevel]->Width) +
-			      align) & ~align) - 1;
-	} else {
-		t->pitch |=
-		    ((tObj->Image[0][t->base.firstLevel]->Width *
-		      texelBytes) + 63) & ~(63);
+			t->pp_txpitch = ((firstImage->Width + align) & ~align) - 1;
 	}
 
 	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-	    if (tObj->Image[0][t->base.firstLevel]->Width > 2048)
-		t->pitch_reg |= R500_TXWIDTH_BIT11;
-	    if (tObj->Image[0][t->base.firstLevel]->Height > 2048)
-		t->pitch_reg |= R500_TXHEIGHT_BIT11;
+	    if (firstImage->Width > 2048)
+		t->pp_txpitch |= R500_TXWIDTH_BIT11;
+	    if (firstImage->Height > 2048)
+		t->pp_txpitch |= R500_TXHEIGHT_BIT11;
 	}
 }
 
-/* ================================================================
- * Texture unit state management
+/**
+ * Ensure the given texture is ready for rendering.
+ *
+ * Mostly this means populating the texture object's mipmap tree.
  */
-
-static GLboolean r300EnableTexture2D(GLcontext * ctx, int unit)
+static GLboolean r300_validate_texture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
+	radeonTexObj *t = radeon_tex_obj(texObj);
 
-	ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
-
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+		return GL_FALSE;
 
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock && !t->image_override)
-			return GL_FALSE;
-	}
+	/* Configure the hardware registers (more precisely, the cached version
+	 * of the hardware registers). */
+	setup_hardware_state(rmesa, t);
 
+	t->validated = GL_TRUE;
 	return GL_TRUE;
 }
 
-static GLboolean r300EnableTexture3D(GLcontext * ctx, int unit)
+/**
+ * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+ */
+GLboolean r300ValidateBuffers(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
+	struct radeon_renderbuffer *rrb;
+	int i;
+	int ret;
 
-	ASSERT(tObj->Target == GL_TEXTURE_3D);
+	radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
 
-	/* r300 does not support mipmaps for 3D textures. */
-	if ((tObj->MinFilter != GL_NEAREST) && (tObj->MinFilter != GL_LINEAR)) {
-		return GL_FALSE;
+	rrb = radeon_get_colorbuffer(&rmesa->radeon);
+	/* color buffer */
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock)
-			return GL_FALSE;
+	/* depth buffer */
+	rrb = radeon_get_depthbuffer(&rmesa->radeon);
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
+	
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+		radeonTexObj *t;
 
-	return GL_TRUE;
-}
-
-static GLboolean r300EnableTextureCube(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	GLuint face;
-
-	ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
-
-	if (t->base.dirty_images[0] || t->base.dirty_images[1] ||
-	    t->base.dirty_images[2] || t->base.dirty_images[3] ||
-	    t->base.dirty_images[4] || t->base.dirty_images[5]) {
-		/* flush */
-		R300_FIREVERTICES(rmesa);
-		/* layout memory space, once for all faces */
-		r300SetTexImages(rmesa, tObj);
-	}
+		if (!ctx->Texture.Unit[i]._ReallyEnabled)
+			continue;
 
-	/* upload (per face) */
-	for (face = 0; face < 6; face++) {
-		if (t->base.dirty_images[face]) {
-			r300UploadTexImages(rmesa,
-					    (r300TexObjPtr) tObj->DriverData,
-					    face);
+		if (!r300_validate_texture(ctx, ctx->Texture.Unit[i]._Current)) {
+			_mesa_warning(ctx,
+				      "failed to validate texture for unit %d.\n",
+				      i);
 		}
+		t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+		if (t->image_override && t->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+		else if (t->mt->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->mt->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
 	}
 
-	if (!t->base.memBlock) {
-		/* texmem alloc failed, use s/w fallback */
+	ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0);
+	if (ret)
 		return GL_FALSE;
-	}
-
 	return GL_TRUE;
 }
 
-static GLboolean r300EnableTextureRect(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-
-	ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
-
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock && !t->image_override &&
-		    !rmesa->prefer_gart_client_texturing)
-			return GL_FALSE;
-	}
-
-	return GL_TRUE;
-}
-
-static GLboolean r300UpdateTexture(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_ReallyEnabled ?
-		texUnit->_Current : NULL;
-	r300TexObjPtr t = tObj ? (r300TexObjPtr) tObj->DriverData : NULL;
-
-	/* Fallback if there's a texture border */
-	if (tObj && tObj->Image[0][tObj->BaseLevel]->Border > 0) {
-		tObj = NULL;
-		t = NULL;
-	}
-
-	/* Update state if this is a different texture object to last
-	 * time.
-	 */
-	if (rmesa->state.texture.unit[unit].texobj != tObj) {
-		if (rmesa->state.texture.unit[unit].texobj != NULL) {
-			r300TexObjPtr t_old = (r300TexObjPtr) rmesa->state.texture.unit[unit].texobj->DriverData;
-
-			/* The old texture is no longer bound to this texture unit.
-			 * Mark it as such.
-			 */
-
-			t_old->base.bound &= ~(1 << unit);
-		}
-
-		_mesa_reference_texobj(&rmesa->state.texture.unit[unit].texobj, tObj);
-
-		if (t) {
-			t->base.bound |= (1 << unit);
-			driUpdateTextureLRU(&t->base);	/* XXX: should be locked! */
-		}
-	}
-
-	return !t || !t->border_fallback;
-}
-
 void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 		      unsigned long long offset, GLint depth, GLuint pitch)
 {
 	r300ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
 	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
-	r300TexObjPtr t;
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 	uint32_t pitch_val;
 
 	if (!tObj)
 		return;
 
-	t = (r300TexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
 
-	t->offset = offset;
-	t->pitch_reg &= (1 << 13) -1;
+	t->bo = NULL;
+	t->override_offset = offset;
+	t->pp_txpitch &= (1 << 13) -1;
 	pitch_val = pitch;
 
 	switch (depth) {
 	case 32:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
-		t->filter |= tx_table[2].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[2].filter;
 		pitch_val /= 4;
 		break;
 	case 24:
 	default:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
-		t->filter |= tx_table[4].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[4].filter;
 		pitch_val /= 4;
 		break;
 	case 16:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
-		t->filter |= tx_table[5].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
+		t->pp_txfilter |= tx_table[5].filter;
 		pitch_val /= 2;
 		break;
 	}
 	pitch_val--;
 
-	t->pitch_reg |= pitch_val;
+	t->pp_txpitch |= pitch_val;
 }
 
-static GLboolean r300UpdateTextureUnit(GLcontext * ctx, int unit)
+void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format, __DRIdrawable *dPriv)
 {
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-
-	if (texUnit->_ReallyEnabled & (TEXTURE_RECT_BIT)) {
-		return (r300EnableTextureRect(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_1D_BIT | TEXTURE_2D_BIT)) {
-		return (r300EnableTexture2D(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_3D_BIT)) {
-		return (r300EnableTexture3D(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_CUBE_BIT)) {
-		return (r300EnableTextureCube(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled) {
-		return GL_FALSE;
-	} else {
-		return r300UpdateTexture(ctx, unit);
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r300ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
 	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		else
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[2].filter;
+		pitch_val /= 4;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[4].filter;
+		pitch_val /= 4;
+		break;
+	case 2:
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
+		t->pp_txfilter |= tx_table[5].filter;
+		pitch_val /= 2;
+		break;
+	}
+	pitch_val--;
+	t->pp_txsize = ((rb->base.Width - 1) << R300_TX_WIDTHMASK_SHIFT) |
+              ((rb->base.Height - 1) << R300_TX_HEIGHTMASK_SHIFT);
+	t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN;
+	t->pp_txpitch |= pitch_val;
+
+	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+	    if (rb->base.Width > 2048)
+		t->pp_txpitch |= R500_TXWIDTH_BIT11;
+	    if (rb->base.Height > 2048)
+		t->pp_txpitch |= R500_TXHEIGHT_BIT11;
+	}
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
 }
 
-void r300UpdateTextureState(GLcontext * ctx)
+void r300SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
 {
-	int i;
-
-	for (i = 0; i < 8; i++) {
-		if (!r300UpdateTextureUnit(ctx, i)) {
-			_mesa_warning(ctx,
-				      "failed to update texture state for unit %d.\n",
-				      i);
-		}
-	}
+        r300SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 146daa367c..dd0f27f9cb 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -32,1437 +32,327 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/program.h"
+#include "shader/programopt.h"
 #include "shader/prog_instruction.h"
+#include "shader/prog_optimize.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "tnl/tnl.h"
 
+#include "compiler/radeon_compiler.h"
+#include "compiler/radeon_nqssadce.h"
 #include "r300_context.h"
+#include "r300_state.h"
 
-/* TODO: Get rid of t_src_class call */
-#define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \
-		       ((t_src_class(a.File) == PVS_SRC_REG_CONSTANT && \
-			 t_src_class(b.File) == PVS_SRC_REG_CONSTANT) || \
-			(t_src_class(a.File) == PVS_SRC_REG_INPUT && \
-			 t_src_class(b.File) == PVS_SRC_REG_INPUT)))) \
-
-/*
- * Take an already-setup and valid source then swizzle it appropriately to
- * obtain a constant ZERO or ONE source.
+/**
+ * Write parameter array for the given vertex program into dst.
+ * Return the total number of components written.
  */
-#define __CONST(x, y)	\
-	(PVS_SRC_OPERAND(t_src_index(vp, &src[x]),	\
-			   t_swizzle(y),	\
-			   t_swizzle(y),	\
-			   t_swizzle(y),	\
-			   t_swizzle(y),	\
-			   t_src_class(src[x].File), \
-			   VSF_FLAG_NONE) | (src[x].RelAddr << 4))
-
-#define FREE_TEMPS() \
-	do { \
-		int u_temp_used = (VSF_MAX_FRAGMENT_TEMPS - 1) - u_temp_i; \
-		if((vp->num_temporaries + u_temp_used) > VSF_MAX_FRAGMENT_TEMPS) { \
-			WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_used); \
-			vp->native = GL_FALSE; \
-		} \
-		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
-	} while (0)
-
-int r300VertexProgUpdateParams(GLcontext * ctx,
-			       struct r300_vertex_program_cont *vp, float *dst)
+static int r300VertexProgUpdateParams(GLcontext * ctx, struct r300_vertex_program *vp, float *dst)
 {
-	int pi;
-	struct gl_vertex_program *mesa_vp = &vp->mesa_program;
-	float *dst_o = dst;
-	struct gl_program_parameter_list *paramList;
+	int i;
 
-	if (mesa_vp->IsNVProgram) {
+	if (vp->Base->IsNVProgram) {
 		_mesa_load_tracked_matrices(ctx);
-
-		for (pi = 0; pi < MAX_NV_VERTEX_PROGRAM_PARAMS; pi++) {
-			*dst++ = ctx->VertexProgram.Parameters[pi][0];
-			*dst++ = ctx->VertexProgram.Parameters[pi][1];
-			*dst++ = ctx->VertexProgram.Parameters[pi][2];
-			*dst++ = ctx->VertexProgram.Parameters[pi][3];
+	} else {
+		if (vp->Base->Base.Parameters) {
+			_mesa_load_state_parameters(ctx, vp->Base->Base.Parameters);
 		}
-		return dst - dst_o;
 	}
 
-	assert(mesa_vp->Base.Parameters);
-	_mesa_load_state_parameters(ctx, mesa_vp->Base.Parameters);
-
-	if (mesa_vp->Base.Parameters->NumParameters * 4 >
-	    VSF_MAX_FRAGMENT_LENGTH) {
+	if (vp->code.constants.Count * 4 > VSF_MAX_FRAGMENT_LENGTH) {
+		/* Should have checked this earlier... */
 		fprintf(stderr, "%s:Params exhausted\n", __FUNCTION__);
 		_mesa_exit(-1);
 	}
 
-	paramList = mesa_vp->Base.Parameters;
-	for (pi = 0; pi < paramList->NumParameters; pi++) {
-		switch (paramList->Parameters[pi].Type) {
-		case PROGRAM_STATE_VAR:
-		case PROGRAM_NAMED_PARAM:
-			//fprintf(stderr, "%s", vp->Parameters->Parameters[pi].Name);
-		case PROGRAM_CONSTANT:
-			*dst++ = paramList->ParameterValues[pi][0];
-			*dst++ = paramList->ParameterValues[pi][1];
-			*dst++ = paramList->ParameterValues[pi][2];
-			*dst++ = paramList->ParameterValues[pi][3];
-			break;
-		default:
-			_mesa_problem(NULL, "Bad param type in %s",
-				      __FUNCTION__);
-		}
-
-	}
-
-	return dst - dst_o;
-}
-
-static unsigned long t_dst_mask(GLuint mask)
-{
-	/* WRITEMASK_* is equivalent to VSF_FLAG_* */
-	return mask & VSF_FLAG_ALL;
-}
-
-static unsigned long t_dst_class(gl_register_file file)
-{
-
-	switch (file) {
-	case PROGRAM_TEMPORARY:
-		return PVS_DST_REG_TEMPORARY;
-	case PROGRAM_OUTPUT:
-		return PVS_DST_REG_OUT;
-	case PROGRAM_ADDRESS:
-		return PVS_DST_REG_A0;
-		/*
-		   case PROGRAM_INPUT:
-		   case PROGRAM_LOCAL_PARAM:
-		   case PROGRAM_ENV_PARAM:
-		   case PROGRAM_NAMED_PARAM:
-		   case PROGRAM_STATE_VAR:
-		   case PROGRAM_WRITE_ONLY:
-		   case PROGRAM_ADDRESS:
-		 */
-	default:
-		fprintf(stderr, "problem in %s", __FUNCTION__);
-		_mesa_exit(-1);
-		return -1;
-	}
-}
-
-static unsigned long t_dst_index(struct r300_vertex_program *vp,
-				 struct prog_dst_register *dst)
-{
-	if (dst->File == PROGRAM_OUTPUT)
-		return vp->outputs[dst->Index];
-
-	return dst->Index;
-}
-
-static unsigned long t_src_class(gl_register_file file)
-{
-	switch (file) {
-	case PROGRAM_TEMPORARY:
-		return PVS_SRC_REG_TEMPORARY;
-	case PROGRAM_INPUT:
-		return PVS_SRC_REG_INPUT;
-	case PROGRAM_LOCAL_PARAM:
-	case PROGRAM_ENV_PARAM:
-	case PROGRAM_NAMED_PARAM:
-	case PROGRAM_CONSTANT:
-	case PROGRAM_STATE_VAR:
-		return PVS_SRC_REG_CONSTANT;
-		/*
-		   case PROGRAM_OUTPUT:
-		   case PROGRAM_WRITE_ONLY:
-		   case PROGRAM_ADDRESS:
-		 */
-	default:
-		fprintf(stderr, "problem in %s", __FUNCTION__);
-		_mesa_exit(-1);
-		return -1;
-	}
-}
+	for(i = 0; i < vp->code.constants.Count; ++i) {
+		const float * src = 0;
+		const struct rc_constant * constant = &vp->code.constants.Constants[i];
 
-static INLINE unsigned long t_swizzle(GLubyte swizzle)
-{
-/* this is in fact a NOP as the Mesa SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */
-	return swizzle;
-}
+		switch(constant->Type) {
+		case RC_CONSTANT_EXTERNAL:
+			if (vp->Base->IsNVProgram) {
+				src = ctx->VertexProgram.Parameters[constant->u.External];
+			} else {
+				src = vp->Base->Base.Parameters->ParameterValues[constant->u.External];
+			}
+			break;
 
-#if 0
-static void vp_dump_inputs(struct r300_vertex_program *vp, char *caller)
-{
-	int i;
+		case RC_CONSTANT_IMMEDIATE:
+			src = constant->u.Immediate;
+			break;
+		}
 
-	if (vp == NULL) {
-		fprintf(stderr, "vp null in call to %s from %s\n", __FUNCTION__,
-			caller);
-		return;
+		dst[4*i] = src[0];
+		dst[4*i + 1] = src[1];
+		dst[4*i + 2] = src[2];
+		dst[4*i + 3] = src[3];
 	}
 
-	fprintf(stderr, "%s:<", caller);
-	for (i = 0; i < VERT_ATTRIB_MAX; i++)
-		fprintf(stderr, "%d ", vp->inputs[i]);
-	fprintf(stderr, ">\n");
-
+	return 4 * vp->code.constants.Count;
 }
-#endif
 
-static unsigned long t_src_index(struct r300_vertex_program *vp,
-				 struct prog_src_register *src)
+static GLbitfield compute_required_outputs(struct gl_vertex_program * vp, GLbitfield fpreads)
 {
+	GLbitfield outputs = 0;
 	int i;
-	int max_reg = -1;
-
-	if (src->File == PROGRAM_INPUT) {
-		if (vp->inputs[src->Index] != -1)
-			return vp->inputs[src->Index];
-
-		for (i = 0; i < VERT_ATTRIB_MAX; i++)
-			if (vp->inputs[i] > max_reg)
-				max_reg = vp->inputs[i];
-
-		vp->inputs[src->Index] = max_reg + 1;
-
-		//vp_dump_inputs(vp, __FUNCTION__);
-
-		return vp->inputs[src->Index];
-	} else {
-		if (src->Index < 0) {
-			fprintf(stderr,
-				"negative offsets for indirect addressing do not work.\n");
-			return 0;
-		}
-		return src->Index;
-	}
-}
 
-/* these two functions should probably be merged... */
-
-static unsigned long t_src(struct r300_vertex_program *vp,
-			   struct prog_src_register *src)
-{
-	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
-	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
-	 */
-	return PVS_SRC_OPERAND(t_src_index(vp, src),
-			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 1)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 2)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 3)),
-			       t_src_class(src->File),
-			       src->Negate) | (src->RelAddr << 4);
-}
+#define ADD_OUTPUT(fp_attr, vp_result) \
+	do { \
+		if (fpreads & (1 << (fp_attr))) \
+			outputs |= (1 << (vp_result)); \
+	} while (0)
 
-static unsigned long t_src_scalar(struct r300_vertex_program *vp,
-				  struct prog_src_register *src)
-{
-	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
-	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
-	 */
-	return PVS_SRC_OPERAND(t_src_index(vp, src),
-			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
-			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
-			       t_src_class(src->File),
-			       src->Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src->RelAddr << 4);
-}
+	ADD_OUTPUT(FRAG_ATTRIB_COL0, VERT_RESULT_COL0);
+	ADD_OUTPUT(FRAG_ATTRIB_COL1, VERT_RESULT_COL1);
 
-static GLboolean valid_dst(struct r300_vertex_program *vp,
-			   struct prog_dst_register *dst)
-{
-	if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) {
-		return GL_FALSE;
-	} else if (dst->File == PROGRAM_ADDRESS) {
-		assert(dst->Index == 0);
+	for (i = 0; i <= 7; ++i) {
+		ADD_OUTPUT(FRAG_ATTRIB_TEX0 + i, VERT_RESULT_TEX0 + i);
 	}
 
-	return GL_TRUE;
-}
-
-static GLuint *r300TranslateOpcodeABS(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//MAX RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_MAXIMUM,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),
-				  t_src_class(src[0].File),
-				  (!src[0].
-				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[3] = 0;
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeADD(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
+#undef ADD_OUTPUT
 
-static GLuint *r300TranslateOpcodeARL(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_FLT2FIX_DX,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
+	if ((fpreads & (1 << FRAG_ATTRIB_COL0)) &&
+	    (vp->Base.OutputsWritten & (1 << VERT_RESULT_BFC0)))
+		outputs |= 1 << VERT_RESULT_BFC0;
+	if ((fpreads & (1 << FRAG_ATTRIB_COL1)) &&
+	    (vp->Base.OutputsWritten & (1 << VERT_RESULT_BFC1)))
+		outputs |= 1 << VERT_RESULT_BFC1;
 
-static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO}
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
-				  SWIZZLE_ZERO,
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[2] =
-	    PVS_SRC_OPERAND(t_src_index(vp, &src[1]),
-			    t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
-			    t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
-			    t_swizzle(GET_SWZ(src[1].Swizzle, 2)), SWIZZLE_ZERO,
-			    t_src_class(src[1].File),
-			    src[1].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
-	    (src[1].RelAddr << 4);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
+	outputs |= 1 << VERT_RESULT_HPOS;
+	if (vp->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ))
+		outputs |= 1 << VERT_RESULT_PSIZ;
 
-static GLuint *r300TranslateOpcodeDP4(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
+	return outputs;
 }
 
-static GLuint *r300TranslateOpcodeDPH(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W}
-	inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
-				  PVS_SRC_SELECT_FORCE_1,
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeDST(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_DISTANCE_VECTOR,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
 
-static GLuint *r300TranslateOpcodeEX2(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_EXP_BASE2_FULL_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeEXP(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_EXP_BASE2_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeFLR(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3],
-				      int *u_temp_i)
-{
-	/* FRC TMP 0.X Y Z W PARAM 0{} {X Y Z W}
-	   ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} TMP 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W */
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_FRACTION,
-				     GL_FALSE,
-				     GL_FALSE,
-				     *u_temp_i,
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     PVS_DST_REG_TEMPORARY);
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-	inst += 4;
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = PVS_SRC_OPERAND(*u_temp_i,
-				  PVS_SRC_SELECT_X,
-				  PVS_SRC_SELECT_Y,
-				  PVS_SRC_SELECT_Z,
-				  PVS_SRC_SELECT_W, PVS_SRC_REG_TEMPORARY,
-				  /* Not 100% sure about this */
-				  (!src[0].
-				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE
-				  /*VSF_FLAG_ALL */ );
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-	(*u_temp_i)--;
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeFRC(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_FRACTION,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeLG2(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	// LG2 RESULT 1.X Y Z W PARAM 0{} {X X X X}
-
-	inst[0] = PVS_OP_DST_OPERAND(ME_LOG_BASE2_FULL_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeLIT(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W}
-
-	inst[0] = PVS_OP_DST_OPERAND(ME_LIGHT_COEFF_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	/* NOTE: Users swizzling might not work. */
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
-				  PVS_SRC_SELECT_FORCE_0,	// Z
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
-				  PVS_SRC_SELECT_FORCE_0,	// Z
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
-				  PVS_SRC_SELECT_FORCE_0,	// Z
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeLOG(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_LOG_BASE2_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeMAD(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(PVS_MACRO_OP_2CLK_MADD,
-				     GL_FALSE,
-				     GL_TRUE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = t_src(vp, &src[2]);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeMAX(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_MAXIMUM,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeMIN(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_MINIMUM,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeMOV(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO}
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeMUL(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodePOW(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_POWER_FUNC_FF,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = t_src_scalar(vp, &src[1]);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeRCP(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_RECIP_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeRSQ(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(ME_RECIP_SQRT_DX,
-				     GL_TRUE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src_scalar(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeSGE(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_SET_GREATER_THAN_EQUAL,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeSLT(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	inst[0] = PVS_OP_DST_OPERAND(VE_SET_LESS_THAN,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = t_src(vp, &src[1]);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//ADD RESULT 1.X Y Z W TMP 0{} {X Y Z W} PARAM 1{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
-
-#if 0
-	inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 2)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
-				  t_src_class(src[1].File),
-				  (!src[1].
-				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[1].RelAddr << 4);
-	inst[3] = 0;
-#else
-	inst[0] =
-	    PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD,
-			       GL_FALSE,
-			       GL_FALSE,
-			       t_dst_index(vp, &vpi->DstReg),
-			       t_dst_mask(vpi->DstReg.WriteMask),
-			       t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ONE);
-	inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 2)),
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
-				  t_src_class(src[1].File),
-				  (!src[1].
-				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[1].RelAddr << 4);
-#endif
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeSWZ(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3])
-{
-	//ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO}
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = t_src(vp, &src[0]);
-	inst[2] = __CONST(0, SWIZZLE_ZERO);
-	inst[3] = __CONST(0, SWIZZLE_ZERO);
-
-	return inst;
-}
-
-static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi,
-				      GLuint * inst,
-				      struct prog_src_register src[3],
-				      int *u_temp_i)
-{
-	/* mul r0, r1.yzxw, r2.zxyw
-	   mad r0, -r2.yzxw, r1.zxyw, r0
-	 */
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     *u_temp_i,
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     PVS_DST_REG_TEMPORARY);
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// Z
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// Z
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// X
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// W
-				  t_src_class(src[1].File),
-				  src[1].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[1].RelAddr << 4);
-	inst[3] = __CONST(1, SWIZZLE_ZERO);
-	inst += 4;
-
-	inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD,
-				     GL_FALSE,
-				     GL_FALSE,
-				     t_dst_index(vp, &vpi->DstReg),
-				     t_dst_mask(vpi->DstReg.WriteMask),
-				     t_dst_class(vpi->DstReg.File));
-	inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// Z
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// X
-				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// W
-				  t_src_class(src[1].File),
-				  (!src[1].
-				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[1].RelAddr << 4);
-	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// Z
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
-				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
-				  t_src_class(src[0].File),
-				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
-	    (src[0].RelAddr << 4);
-	inst[3] =
-	    PVS_SRC_OPERAND(*u_temp_i, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y,
-			    PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W,
-			    PVS_SRC_REG_TEMPORARY, VSF_FLAG_NONE);
-
-	(*u_temp_i)--;
-
-	return inst;
-}
-
-static void t_inputs_outputs(struct r300_vertex_program *vp)
+static void t_inputs_outputs(struct r300_vertex_program_compiler * c)
 {
 	int i;
-	int cur_reg = 0;
+	int cur_reg;
+	GLuint OutputsWritten, InputsRead;
+
+	OutputsWritten = c->Base.Program.OutputsWritten;
+	InputsRead = c->Base.Program.InputsRead;
 
-	for (i = 0; i < VERT_ATTRIB_MAX; i++)
-		vp->inputs[i] = -1;
+	cur_reg = -1;
+	for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+		if (InputsRead & (1 << i))
+			c->code->inputs[i] = ++cur_reg;
+		else
+			c->code->inputs[i] = -1;
+	}
 
+	cur_reg = 0;
 	for (i = 0; i < VERT_RESULT_MAX; i++)
-		vp->outputs[i] = -1;
+		c->code->outputs[i] = -1;
 
-	assert(vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS));
+	assert(OutputsWritten & (1 << VERT_RESULT_HPOS));
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) {
-		vp->outputs[VERT_RESULT_HPOS] = cur_reg++;
+	if (OutputsWritten & (1 << VERT_RESULT_HPOS)) {
+		c->code->outputs[VERT_RESULT_HPOS] = cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
-		vp->outputs[VERT_RESULT_PSIZ] = cur_reg++;
+	if (OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
+		c->code->outputs[VERT_RESULT_PSIZ] = cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) {
-		vp->outputs[VERT_RESULT_COL0] = cur_reg++;
+	/* If we're writing back facing colors we need to send
+	 * four colors to make front/back face colors selection work.
+	 * If the vertex program doesn't write all 4 colors, lets
+	 * pretend it does by skipping output index reg so the colors
+	 * get written into appropriate output vectors.
+	 */
+	if (OutputsWritten & (1 << VERT_RESULT_COL0)) {
+		c->code->outputs[VERT_RESULT_COL0] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) {
-		vp->outputs[VERT_RESULT_COL1] =
-		    vp->outputs[VERT_RESULT_COL0] + 1;
-		cur_reg = vp->outputs[VERT_RESULT_COL1] + 1;
+	if (OutputsWritten & (1 << VERT_RESULT_COL1)) {
+		c->code->outputs[VERT_RESULT_COL1] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
-		vp->outputs[VERT_RESULT_BFC0] =
-		    vp->outputs[VERT_RESULT_COL0] + 2;
-		cur_reg = vp->outputs[VERT_RESULT_BFC0] + 2;
+	if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		c->code->outputs[VERT_RESULT_BFC0] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
-		vp->outputs[VERT_RESULT_BFC1] =
-		    vp->outputs[VERT_RESULT_COL0] + 3;
-		cur_reg = vp->outputs[VERT_RESULT_BFC1] + 1;
+	if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		c->code->outputs[VERT_RESULT_BFC1] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		cur_reg++;
 	}
 
 	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) {
-		if (vp->key.OutputsWritten & (1 << i)) {
-			vp->outputs[i] = cur_reg++;
+		if (OutputsWritten & (1 << i)) {
+			c->code->outputs[i] = cur_reg++;
 		}
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) {
-		vp->outputs[VERT_RESULT_FOGC] = cur_reg++;
+	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
+		c->code->outputs[VERT_RESULT_FOGC] = cur_reg++;
 	}
 }
 
-static void r300TranslateVertexShader(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi)
-{
-	int i;
-	GLuint *inst;
-	unsigned long num_operands;
-	/* Initial value should be last tmp reg that hw supports.
-	   Strangely enough r300 doesnt mind even though these would be out of range.
-	   Smart enough to realize that it doesnt need it? */
-	int u_temp_i = VSF_MAX_FRAGMENT_TEMPS - 1;
-	struct prog_src_register src[3];
-
-	vp->pos_end = 0;	/* Not supported yet */
-	vp->program.length = 0;
-	/*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */
-	vp->translated = GL_TRUE;
-	vp->native = GL_TRUE;
-
-	t_inputs_outputs(vp);
-
-	for (inst = vp->program.body.i; vpi->Opcode != OPCODE_END;
-	     vpi++, inst += 4) {
-
-		FREE_TEMPS();
-
-		if (!valid_dst(vp, &vpi->DstReg)) {
-			/* redirect result to unused temp */
-			vpi->DstReg.File = PROGRAM_TEMPORARY;
-			vpi->DstReg.Index = u_temp_i;
-		}
 
-		num_operands = _mesa_num_inst_src_regs(vpi->Opcode);
+static struct r300_vertex_program *build_program(GLcontext *ctx,
+						 struct r300_vertex_program_key *wanted_key,
+						 const struct gl_vertex_program *mesa_vp)
+{
+	struct r300_vertex_program *vp;
+	struct r300_vertex_program_compiler compiler;
 
-		/* copy the sources (src) from mesa into a local variable... is this needed? */
-		for (i = 0; i < num_operands; i++) {
-			src[i] = vpi->SrcReg[i];
-		}
+	vp = _mesa_calloc(sizeof(*vp));
+	vp->Base = (struct gl_vertex_program *) _mesa_clone_program(ctx, &mesa_vp->Base);
+	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
 
-		if (num_operands == 3) {	/* TODO: scalars */
-			if (CMP_SRCS(src[1], src[2])
-			    || CMP_SRCS(src[0], src[2])) {
-				inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-							     GL_FALSE,
-							     GL_FALSE,
-							     u_temp_i,
-							     VSF_FLAG_ALL,
-							     PVS_DST_REG_TEMPORARY);
-				inst[1] =
-				    PVS_SRC_OPERAND(t_src_index(vp, &src[2]),
-						    SWIZZLE_X,
-						    SWIZZLE_Y,
-						    SWIZZLE_Z,
-						    SWIZZLE_W,
-						    t_src_class(src[2].File),
-						    VSF_FLAG_NONE) | (src[2].
-								      RelAddr <<
-								      4);
-				inst[2] = __CONST(2, SWIZZLE_ZERO);
-				inst[3] = __CONST(2, SWIZZLE_ZERO);
-				inst += 4;
-
-				src[2].File = PROGRAM_TEMPORARY;
-				src[2].Index = u_temp_i;
-				src[2].RelAddr = 0;
-				u_temp_i--;
-			}
-		}
+	rc_init(&compiler.Base);
+	compiler.Base.Debug = (RADEON_DEBUG & RADEON_VERTS) ? GL_TRUE : GL_FALSE;
 
-		if (num_operands >= 2) {
-			if (CMP_SRCS(src[1], src[0])) {
-				inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-							     GL_FALSE,
-							     GL_FALSE,
-							     u_temp_i,
-							     VSF_FLAG_ALL,
-							     PVS_DST_REG_TEMPORARY);
-				inst[1] =
-				    PVS_SRC_OPERAND(t_src_index(vp, &src[0]),
-						    SWIZZLE_X,
-						    SWIZZLE_Y,
-						    SWIZZLE_Z,
-						    SWIZZLE_W,
-						    t_src_class(src[0].File),
-						    VSF_FLAG_NONE) | (src[0].
-								      RelAddr <<
-								      4);
-				inst[2] = __CONST(0, SWIZZLE_ZERO);
-				inst[3] = __CONST(0, SWIZZLE_ZERO);
-				inst += 4;
-
-				src[0].File = PROGRAM_TEMPORARY;
-				src[0].Index = u_temp_i;
-				src[0].RelAddr = 0;
-				u_temp_i--;
-			}
-		}
+	compiler.code = &vp->code;
+	compiler.RequiredOutputs = compute_required_outputs(vp->Base, vp->key.FpReads);
+	compiler.SetHwInputOutput = &t_inputs_outputs;
 
-		switch (vpi->Opcode) {
-		case OPCODE_ABS:
-			inst = r300TranslateOpcodeABS(vp, vpi, inst, src);
-			break;
-		case OPCODE_ADD:
-			inst = r300TranslateOpcodeADD(vp, vpi, inst, src);
-			break;
-		case OPCODE_ARL:
-			inst = r300TranslateOpcodeARL(vp, vpi, inst, src);
-			break;
-		case OPCODE_DP3:
-			inst = r300TranslateOpcodeDP3(vp, vpi, inst, src);
-			break;
-		case OPCODE_DP4:
-			inst = r300TranslateOpcodeDP4(vp, vpi, inst, src);
-			break;
-		case OPCODE_DPH:
-			inst = r300TranslateOpcodeDPH(vp, vpi, inst, src);
-			break;
-		case OPCODE_DST:
-			inst = r300TranslateOpcodeDST(vp, vpi, inst, src);
-			break;
-		case OPCODE_EX2:
-			inst = r300TranslateOpcodeEX2(vp, vpi, inst, src);
-			break;
-		case OPCODE_EXP:
-			inst = r300TranslateOpcodeEXP(vp, vpi, inst, src);
-			break;
-		case OPCODE_FLR:
-			inst = r300TranslateOpcodeFLR(vp, vpi, inst, src,	/* FIXME */
-						      &u_temp_i);
-			break;
-		case OPCODE_FRC:
-			inst = r300TranslateOpcodeFRC(vp, vpi, inst, src);
-			break;
-		case OPCODE_LG2:
-			inst = r300TranslateOpcodeLG2(vp, vpi, inst, src);
-			break;
-		case OPCODE_LIT:
-			inst = r300TranslateOpcodeLIT(vp, vpi, inst, src);
-			break;
-		case OPCODE_LOG:
-			inst = r300TranslateOpcodeLOG(vp, vpi, inst, src);
-			break;
-		case OPCODE_MAD:
-			inst = r300TranslateOpcodeMAD(vp, vpi, inst, src);
-			break;
-		case OPCODE_MAX:
-			inst = r300TranslateOpcodeMAX(vp, vpi, inst, src);
-			break;
-		case OPCODE_MIN:
-			inst = r300TranslateOpcodeMIN(vp, vpi, inst, src);
-			break;
-		case OPCODE_MOV:
-			inst = r300TranslateOpcodeMOV(vp, vpi, inst, src);
-			break;
-		case OPCODE_MUL:
-			inst = r300TranslateOpcodeMUL(vp, vpi, inst, src);
-			break;
-		case OPCODE_POW:
-			inst = r300TranslateOpcodePOW(vp, vpi, inst, src);
-			break;
-		case OPCODE_RCP:
-			inst = r300TranslateOpcodeRCP(vp, vpi, inst, src);
-			break;
-		case OPCODE_RSQ:
-			inst = r300TranslateOpcodeRSQ(vp, vpi, inst, src);
-			break;
-		case OPCODE_SGE:
-			inst = r300TranslateOpcodeSGE(vp, vpi, inst, src);
-			break;
-		case OPCODE_SLT:
-			inst = r300TranslateOpcodeSLT(vp, vpi, inst, src);
-			break;
-		case OPCODE_SUB:
-			inst = r300TranslateOpcodeSUB(vp, vpi, inst, src);
-			break;
-		case OPCODE_SWZ:
-			inst = r300TranslateOpcodeSWZ(vp, vpi, inst, src);
-			break;
-		case OPCODE_XPD:
-			inst = r300TranslateOpcodeXPD(vp, vpi, inst, src,	/* FIXME */
-						      &u_temp_i);
-			break;
-		default:
-			assert(0);
-			break;
-		}
-	}
-
-	/* Some outputs may be artificially added, to match the inputs
-	   of the fragment program. Blank the outputs here. */
-	for (i = 0; i < VERT_RESULT_MAX; i++) {
-		if (vp->key.OutputsAdded & (1 << i)) {
-			inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-						     GL_FALSE,
-						     GL_FALSE,
-						     vp->outputs[i],
-						     VSF_FLAG_ALL,
-						     PVS_DST_REG_OUT);
-			inst[1] = __CONST(0, SWIZZLE_ZERO);
-			inst[2] = __CONST(0, SWIZZLE_ZERO);
-			inst[3] = __CONST(0, SWIZZLE_ZERO);
-			inst += 4;
-		}
+	if (compiler.Base.Debug) {
+		fprintf(stderr, "Initial vertex program:\n");
+		_mesa_print_program(&vp->Base->Base);
+		fflush(stderr);
 	}
 
-	vp->program.length = (inst - vp->program.body.i);
-	if (vp->program.length >= VSF_MAX_FRAGMENT_LENGTH) {
-		vp->program.length = 0;
-		vp->native = GL_FALSE;
+	if (mesa_vp->IsPositionInvariant) {
+		_mesa_insert_mvp_code(ctx, vp->Base);
 	}
-#if 0
-	fprintf(stderr, "hw program:\n");
-	for (i = 0; i < vp->program.length; i++)
-		fprintf(stderr, "%08x\n", vp->program.body.d[i]);
-#endif
-}
 
-/* DP4 version seems to trigger some hw peculiarity */
-//#define PREFER_DP4
+	rc_mesa_to_rc_program(&compiler.Base, &vp->Base->Base);
 
-static void position_invariant(struct gl_program *prog)
-{
-	struct prog_instruction *vpi;
-	struct gl_program_parameter_list *paramList;
-	int i;
-
-	gl_state_index tokens[STATE_LENGTH] = { STATE_MVP_MATRIX, 0, 0, 0, 0 };
-
-	/* tokens[4] = matrix modifier */
-#ifdef PREFER_DP4
-	tokens[4] = 0;		/* not transposed or inverted */
-#else
-	tokens[4] = STATE_MATRIX_TRANSPOSE;
-#endif
-	paramList = prog->Parameters;
-
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 4);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 4);
-
-	for (i = 0; i < 4; i++) {
-		GLint idx;
-		tokens[2] = tokens[3] = i;	/* matrix row[i]..row[i] */
-		idx = _mesa_add_state_reference(paramList, tokens);
-#ifdef PREFER_DP4
-		vpi[i].Opcode = OPCODE_DP4;
-		vpi[i].StringPos = 0;
-		vpi[i].Data = 0;
-
-		vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		vpi[i].DstReg.Index = VERT_RESULT_HPOS;
-		vpi[i].DstReg.WriteMask = 1 << i;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = SWIZZLE_XYZW;
-#else
-		if (i == 0)
-			vpi[i].Opcode = OPCODE_MUL;
-		else
-			vpi[i].Opcode = OPCODE_MAD;
+	rc_move_output(&compiler.Base, VERT_RESULT_PSIZ, VERT_RESULT_PSIZ, WRITEMASK_X);
 
-		vpi[i].Data = 0;
-
-		if (i == 3)
-			vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		else
-			vpi[i].DstReg.File = PROGRAM_TEMPORARY;
-		vpi[i].DstReg.Index = 0;
-		vpi[i].DstReg.WriteMask = 0xf;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(i, i, i, i);
-
-		if (i > 0) {
-			vpi[i].SrcReg[2].File = PROGRAM_TEMPORARY;
-			vpi[i].SrcReg[2].Index = 0;
-			vpi[i].SrcReg[2].Swizzle = SWIZZLE_XYZW;
-		}
-#endif
+	if (vp->key.WPosAttr != FRAG_ATTRIB_MAX) {
+		rc_copy_output(&compiler.Base,
+			VERT_RESULT_HPOS,
+			vp->key.WPosAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0);
 	}
 
-	_mesa_copy_instructions(&vpi[i], prog->Instructions,
-				prog->NumInstructions);
-
-	free(prog->Instructions);
-
-	prog->Instructions = vpi;
-
-	prog->NumInstructions += 4;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
-
-	assert(vpi->Opcode == OPCODE_END);
-}
-
-static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
-			GLuint temp_index)
-{
-	struct prog_instruction *vpi;
-	struct prog_instruction *vpi_insert;
-	int i = 0;
-
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 2);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 2);
-	/* all but END */
-	_mesa_copy_instructions(vpi, prog->Instructions,
-				prog->NumInstructions - 1);
-	/* END */
-	_mesa_copy_instructions(&vpi[prog->NumInstructions + 1],
-				&prog->Instructions[prog->NumInstructions - 1],
-				1);
-	vpi_insert = &vpi[prog->NumInstructions - 1];
-
-	vpi_insert[i].Opcode = OPCODE_MOV;
-
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_HPOS;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
-
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
-
-	vpi_insert[i].Opcode = OPCODE_MOV;
-
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_TEX0 + vp->wpos_idx;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
-
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
-
-	free(prog->Instructions);
+	if (vp->key.FogAttr != FRAG_ATTRIB_MAX) {
+		rc_move_output(&compiler.Base,
+			VERT_RESULT_FOGC,
+			vp->key.FogAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0, WRITEMASK_X);
+	}
 
-	prog->Instructions = vpi;
+	r3xx_compile_vertex_program(&compiler);
+	vp->error = compiler.Base.Error;
 
-	prog->NumInstructions += i;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
+	vp->Base->Base.InputsRead = vp->code.InputsRead;
+	vp->Base->Base.OutputsWritten = vp->code.OutputsWritten;
 
-	assert(vpi->Opcode == OPCODE_END);
-}
+	rc_destroy(&compiler.Base);
 
-static void pos_as_texcoord(struct r300_vertex_program *vp,
-			    struct gl_program *prog)
-{
-	struct prog_instruction *vpi;
-	GLuint tempregi = prog->NumTemporaries;
-	/* should do something else if no temps left... */
-	prog->NumTemporaries++;
-
-	for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) {
-		if (vpi->DstReg.File == PROGRAM_OUTPUT
-		    && vpi->DstReg.Index == VERT_RESULT_HPOS) {
-			vpi->DstReg.File = PROGRAM_TEMPORARY;
-			vpi->DstReg.Index = tempregi;
-		}
-	}
-	insert_wpos(vp, prog, tempregi);
+	return vp;
 }
 
-static struct r300_vertex_program *build_program(struct r300_vertex_program_key
-						 *wanted_key, struct gl_vertex_program
-						 *mesa_vp, GLint wpos_idx)
+struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_program_key wanted_key = { 0 };
+	struct r300_vertex_program_cont *vpc;
 	struct r300_vertex_program *vp;
 
-	vp = _mesa_calloc(sizeof(*vp));
-	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
-	vp->wpos_idx = wpos_idx;
-
-	if (mesa_vp->IsPositionInvariant) {
-		position_invariant(&mesa_vp->Base);
-	}
+	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
+	wanted_key.FpReads = r300->selected_fp->InputsRead;
+	wanted_key.FogAttr = r300->selected_fp->fog_attr;
+	wanted_key.WPosAttr = r300->selected_fp->wpos_attr;
 
-	if (wpos_idx > -1) {
-		pos_as_texcoord(vp, &mesa_vp->Base);
+	for (vp = vpc->progs; vp; vp = vp->next) {
+		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
+		    == 0) {
+			return r300->selected_vp = vp;
+		}
 	}
 
-	assert(mesa_vp->Base.NumInstructions);
-	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
-	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
+	vp = build_program(ctx, &wanted_key, &vpc->mesa_program);
+	vp->next = vpc->progs;
+	vpc->progs = vp;
 
-	return vp;
+	return r300->selected_vp = vp;
 }
 
-static void add_outputs(struct r300_vertex_program_key *key, GLint vert)
+#define bump_vpu_count(ptr, new_count)   do { \
+		drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr)); \
+		int _nc=(new_count)/4; \
+		assert(_nc < 256); \
+		if(_nc>_p->vpu.count)_p->vpu.count=_nc; \
+	} while(0)
+
+static void r300EmitVertexProgram(r300ContextPtr r300, int dest, struct r300_vertex_program_code *code)
 {
-	if (key->OutputsWritten & (1 << vert))
-		return;
+	int i;
 
-	key->OutputsWritten |= 1 << vert;
-	key->OutputsAdded |= 1 << vert;
-}
+	assert((code->length > 0) && (code->length % 4 == 0));
 
-void r300SelectVertexShader(r300ContextPtr r300)
-{
-	GLcontext *ctx = ctx = r300->radeon.glCtx;
-	GLuint InputsRead;
-	struct r300_vertex_program_key wanted_key = { 0 };
-	GLint i;
-	struct r300_vertex_program_cont *vpc;
-	struct r300_vertex_program *vp;
-	GLint wpos_idx;
+	R300_STATECHANGE( r300, vap_flush );
 
-	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
-	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
-	wanted_key.OutputsWritten = vpc->mesa_program.Base.OutputsWritten;
-	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-
-	wpos_idx = -1;
-	if (InputsRead & FRAG_BIT_WPOS) {
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-			if (!(InputsRead & (FRAG_BIT_TEX0 << i)))
-				break;
-
-		if (i == ctx->Const.MaxTextureUnits) {
-			fprintf(stderr, "\tno free texcoord found\n");
+	switch ((dest >> 8) & 0xf) {
+		case 0:
+			R300_STATECHANGE(r300, vpi);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpi.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 2:
+			R300_STATECHANGE(r300, vpp);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpp.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 4:
+			R300_STATECHANGE(r300, vps);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vps.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		default:
+			fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
 			_mesa_exit(-1);
-		}
-
-		wanted_key.OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-		wpos_idx = i;
 	}
+}
 
-	add_outputs(&wanted_key, VERT_RESULT_HPOS);
+void r300SetupVertexProgram(r300ContextPtr rmesa)
+{
+	GLcontext *ctx = rmesa->radeon.glCtx;
+	struct r300_vertex_program *prog = rmesa->selected_vp;
+	int inst_count = 0;
+	int param_count = 0;
 
-	if (InputsRead & FRAG_BIT_COL0) {
-		add_outputs(&wanted_key, VERT_RESULT_COL0);
-	}
+	/* Reset state, in case we don't use something */
+	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
 
-	if (InputsRead & FRAG_BIT_COL1) {
-		add_outputs(&wanted_key, VERT_RESULT_COL1);
-	}
+	R300_STATECHANGE(rmesa, vap_flush);
+	R300_STATECHANGE(rmesa, vpp);
+	param_count = r300VertexProgUpdateParams(ctx, prog, (float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
+	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
+	param_count /= 4;
 
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (InputsRead & (FRAG_BIT_TEX0 << i)) {
-			add_outputs(&wanted_key, VERT_RESULT_TEX0 + i);
-		}
-	}
+	r300EmitVertexProgram(rmesa, R300_PVS_CODE_START, &(prog->code));
+	inst_count = (prog->code.length / 4) - 1;
 
-	if (vpc->mesa_program.IsPositionInvariant) {
-		/* we wan't position don't we ? */
-		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
-	}
+	r300VapCntl(rmesa, _mesa_bitcount(prog->code.InputsRead),
+				 _mesa_bitcount(prog->code.OutputsWritten), prog->code.num_temporaries);
 
-	for (vp = vpc->progs; vp; vp = vp->next)
-		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
-		    == 0) {
-			r300->selected_vp = vp;
-			return;
-		}
-	//_mesa_print_program(&vpc->mesa_program.Base);
+	R300_STATECHANGE(rmesa, pvs);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = (0 << R300_PVS_FIRST_INST_SHIFT) | (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
+				(inst_count << R300_PVS_LAST_INST_SHIFT);
 
-	vp = build_program(&wanted_key, &vpc->mesa_program, wpos_idx);
-	vp->next = vpc->progs;
-	vpc->progs = vp;
-	r300->selected_vp = vp;
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h
index 2f35f02bc8..ccec896be4 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.h
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.h
@@ -3,33 +3,9 @@
 
 #include "r300_reg.h"
 
-#define PVS_OP_DST_OPERAND(opcode, math_inst, macro_inst, reg_index, reg_writemask, reg_class)	\
-	 (((opcode & PVS_DST_OPCODE_MASK) << PVS_DST_OPCODE_SHIFT)	\
-	 | ((math_inst & PVS_DST_MATH_INST_MASK) << PVS_DST_MATH_INST_SHIFT)	\
-	 | ((macro_inst & PVS_DST_MACRO_INST_MASK) << PVS_DST_MACRO_INST_SHIFT)	\
-	 | ((reg_index & PVS_DST_OFFSET_MASK) << PVS_DST_OFFSET_SHIFT)	\
-	 | ((reg_writemask & 0xf) << PVS_DST_WE_X_SHIFT)	/* X Y Z W */	\
-	 | ((reg_class & PVS_DST_REG_TYPE_MASK) << PVS_DST_REG_TYPE_SHIFT))
 
-#define PVS_SRC_OPERAND(in_reg_index, comp_x, comp_y, comp_z, comp_w, reg_class, negate)	\
-	(((in_reg_index & PVS_SRC_OFFSET_MASK) << PVS_SRC_OFFSET_SHIFT)				\
-	 | ((comp_x & PVS_SRC_SWIZZLE_X_MASK) << PVS_SRC_SWIZZLE_X_SHIFT)			\
-	 | ((comp_y & PVS_SRC_SWIZZLE_Y_MASK) << PVS_SRC_SWIZZLE_Y_SHIFT)			\
-	 | ((comp_z & PVS_SRC_SWIZZLE_Z_MASK) << PVS_SRC_SWIZZLE_Z_SHIFT)			\
-	 | ((comp_w & PVS_SRC_SWIZZLE_W_MASK) << PVS_SRC_SWIZZLE_W_SHIFT)			\
-	 | ((negate & 0xf) << PVS_SRC_MODIFIER_X_SHIFT)	/* X Y Z W */				\
-	 | ((reg_class & PVS_SRC_REG_TYPE_MASK) << PVS_SRC_REG_TYPE_SHIFT))
+void r300SetupVertexProgram(r300ContextPtr rmesa);
 
-#if 1
-
-#define VSF_FLAG_X	1
-#define VSF_FLAG_Y	2
-#define VSF_FLAG_Z	4
-#define VSF_FLAG_W	8
-#define VSF_FLAG_XYZ	(VSF_FLAG_X | VSF_FLAG_Y | VSF_FLAG_Z)
-#define VSF_FLAG_ALL  0xf
-#define VSF_FLAG_NONE  0
-
-#endif
+struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.c b/src/mesa/drivers/dri/r300/r500_fragprog.c
deleted file mode 100644
index 292573de89..0000000000
--- a/src/mesa/drivers/dri/r300/r500_fragprog.c
+++ /dev/null
@@ -1,718 +0,0 @@
-/*
- * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-#include "r500_fragprog.h"
-
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
-
-static void reset_srcreg(struct prog_src_register* reg)
-{
-	_mesa_bzero(reg, sizeof(*reg));
-	reg->Swizzle = SWIZZLE_NOOP;
-}
-
-static struct prog_src_register shadow_ambient(struct gl_program *program, int tmu)
-{
-	gl_state_index fail_value_tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_SHADOW_AMBIENT, 0, 0, 0
-	};
-	struct prog_src_register reg = { 0, };
-
-	fail_value_tokens[2] = tmu;
-	reg.File = PROGRAM_STATE_VAR;
-	reg.Index = _mesa_add_state_reference(program->Parameters, fail_value_tokens);
-	reg.Swizzle = SWIZZLE_WWWW;
-	return reg;
-}
-
-/**
- * Transform TEX, TXP, TXB, and KIL instructions in the following way:
- *  - premultiply texture coordinates for RECT
- *  - extract operand swizzles
- *  - introduce a temporary register when write masks are needed
- *
- */
-static GLboolean transform_TEX(
-	struct radeon_transform_context *t,
-	struct prog_instruction* orig_inst, void* data)
-{
-	struct r500_fragment_program_compiler *compiler =
-		(struct r500_fragment_program_compiler*)data;
-	struct prog_instruction inst = *orig_inst;
-	struct prog_instruction* tgt;
-	GLboolean destredirect = GL_FALSE;
-
-	if (inst.Opcode != OPCODE_TEX &&
-	    inst.Opcode != OPCODE_TXB &&
-	    inst.Opcode != OPCODE_TXP &&
-	    inst.Opcode != OPCODE_KIL)
-		return GL_FALSE;
-
-	/* ARB_shadow & EXT_shadow_funcs */
-	if (inst.Opcode != OPCODE_KIL &&
-	    t->Program->ShadowSamplers & (1 << inst.TexSrcUnit)) {
-		GLuint comparefunc = GL_NEVER + compiler->fp->state.unit[inst.TexSrcUnit].texture_compare_func;
-
-		if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) {
-			tgt = radeonAppendInstructions(t->Program, 1);
-
-			tgt->Opcode = OPCODE_MOV;
-			tgt->DstReg = inst.DstReg;
-			if (comparefunc == GL_ALWAYS) {
-				tgt->SrcReg[0].File = PROGRAM_BUILTIN;
-				tgt->SrcReg[0].Swizzle = SWIZZLE_1111;
-			} else {
-				tgt->SrcReg[0] = shadow_ambient(t->Program, inst.TexSrcUnit);
-			}
-			return GL_TRUE;
-		}
-
-		inst.DstReg.File = PROGRAM_TEMPORARY;
-		inst.DstReg.Index = radeonFindFreeTemporary(t);
-		inst.DstReg.WriteMask = WRITEMASK_XYZW;
-	} else if (inst.Opcode != OPCODE_KIL && inst.DstReg.File != PROGRAM_TEMPORARY) {
-		int tempreg = radeonFindFreeTemporary(t);
-
-		inst.DstReg.File = PROGRAM_TEMPORARY;
-		inst.DstReg.Index = tempreg;
-		inst.DstReg.WriteMask = WRITEMASK_XYZW;
-		destredirect = GL_TRUE;
-	}
-
-	if (inst.SrcReg[0].File != PROGRAM_TEMPORARY && inst.SrcReg[0].File != PROGRAM_INPUT) {
-		int tmpreg = radeonFindFreeTemporary(t);
-		tgt = radeonAppendInstructions(t->Program, 1);
-		tgt->Opcode = OPCODE_MOV;
-		tgt->DstReg.File = PROGRAM_TEMPORARY;
-		tgt->DstReg.Index = tmpreg;
-		tgt->SrcReg[0] = inst.SrcReg[0];
-
-		reset_srcreg(&inst.SrcReg[0]);
-		inst.SrcReg[0].File = PROGRAM_TEMPORARY;
-		inst.SrcReg[0].Index = tmpreg;
-	}
-
-	tgt = radeonAppendInstructions(t->Program, 1);
-	_mesa_copy_instructions(tgt, &inst, 1);
-
-	if (inst.Opcode != OPCODE_KIL &&
-	    t->Program->ShadowSamplers & (1 << inst.TexSrcUnit)) {
-		GLuint comparefunc = GL_NEVER + compiler->fp->state.unit[inst.TexSrcUnit].texture_compare_func;
-		GLuint depthmode = compiler->fp->state.unit[inst.TexSrcUnit].depth_texture_mode;
-		int rcptemp = radeonFindFreeTemporary(t);
-		int pass, fail;
-
-		tgt = radeonAppendInstructions(t->Program, 3);
-
-		tgt[0].Opcode = OPCODE_RCP;
-		tgt[0].DstReg.File = PROGRAM_TEMPORARY;
-		tgt[0].DstReg.Index = rcptemp;
-		tgt[0].DstReg.WriteMask = WRITEMASK_W;
-		tgt[0].SrcReg[0] = inst.SrcReg[0];
-		tgt[0].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-
-		tgt[1].Opcode = OPCODE_MAD;
-		tgt[1].DstReg = inst.DstReg;
-		tgt[1].DstReg.WriteMask = orig_inst->DstReg.WriteMask;
-		tgt[1].SrcReg[0] = inst.SrcReg[0];
-		tgt[1].SrcReg[0].Swizzle = SWIZZLE_ZZZZ;
-		tgt[1].SrcReg[1].File = PROGRAM_TEMPORARY;
-		tgt[1].SrcReg[1].Index = rcptemp;
-		tgt[1].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-		tgt[1].SrcReg[2].File = PROGRAM_TEMPORARY;
-		tgt[1].SrcReg[2].Index = inst.DstReg.Index;
-		if (depthmode == 0) /* GL_LUMINANCE */
-			tgt[1].SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z);
-		else if (depthmode == 2) /* GL_ALPHA */
-			tgt[1].SrcReg[2].Swizzle = SWIZZLE_WWWW;
-
-		/* Recall that SrcReg[0] is tex, SrcReg[2] is r and:
-		 *   r  < tex  <=>      -tex+r < 0
-		 *   r >= tex  <=> not (-tex+r < 0 */
-		if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
-			tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW;
-		else
-			tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW;
-
-		tgt[2].Opcode = OPCODE_CMP;
-		tgt[2].DstReg = orig_inst->DstReg;
-		tgt[2].SrcReg[0].File = PROGRAM_TEMPORARY;
-		tgt[2].SrcReg[0].Index = tgt[1].DstReg.Index;
-
-		if (comparefunc == GL_LESS || comparefunc == GL_GREATER) {
-			pass = 1;
-			fail = 2;
-		} else {
-			pass = 2;
-			fail = 1;
-		}
-
-		tgt[2].SrcReg[pass].File = PROGRAM_BUILTIN;
-		tgt[2].SrcReg[pass].Swizzle = SWIZZLE_1111;
-		tgt[2].SrcReg[fail] = shadow_ambient(t->Program, inst.TexSrcUnit);
-	} else if (destredirect) {
-		tgt = radeonAppendInstructions(t->Program, 1);
-
-		tgt->Opcode = OPCODE_MOV;
-		tgt->DstReg = orig_inst->DstReg;
-		tgt->SrcReg[0].File = PROGRAM_TEMPORARY;
-		tgt->SrcReg[0].Index = inst.DstReg.Index;
-	}
-
-	return GL_TRUE;
-}
-
-
-static void update_params(r300ContextPtr r300, struct r500_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r500_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
-{
-	GLuint relevant;
-	int i;
-
-	if (opcode == OPCODE_TEX ||
-	    opcode == OPCODE_TXB ||
-	    opcode == OPCODE_TXP ||
-	    opcode == OPCODE_KIL) {
-		if (reg.Abs)
-			return GL_FALSE;
-
-		if (reg.Negate)
-			reg.Negate ^= NEGATE_XYZW;
-
-		if (opcode == OPCODE_KIL) {
-			if (reg.Swizzle != SWIZZLE_NOOP)
-				return GL_FALSE;
-		} else {
-			for(i = 0; i < 4; ++i) {
-				GLuint swz = GET_SWZ(reg.Swizzle, i);
-				if (swz == SWIZZLE_NIL) {
-					reg.Negate &= ~(1 << i);
-					continue;
-				}
-				if (swz >= 4)
-					return GL_FALSE;
-			}
-		}
-
-		if (reg.Negate)
-			return GL_FALSE;
-
-		return GL_TRUE;
-	} else if (opcode == OPCODE_DDX || opcode == OPCODE_DDY) {
-		/* DDX/MDH and DDY/MDV explicitly ignore incoming swizzles;
-		 * if it doesn't fit perfectly into a .xyzw case... */
-		if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs && !reg.Negate)
-			return GL_TRUE;
-
-		return GL_FALSE;
-	} else {
-		/* ALU instructions support almost everything */
-		if (reg.Abs)
-			return GL_TRUE;
-
-		relevant = 0;
-		for(i = 0; i < 3; ++i) {
-			GLuint swz = GET_SWZ(reg.Swizzle, i);
-			if (swz != SWIZZLE_NIL && swz != SWIZZLE_ZERO)
-				relevant |= 1 << i;
-		}
-		if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant))
-			return GL_FALSE;
-
-		return GL_TRUE;
-	}
-}
-
-/**
- * Implement a MOV with a potentially non-native swizzle.
- *
- * The only thing we *cannot* do in an ALU instruction is per-component
- * negation. Therefore, we split the MOV into two instructions when necessary.
- */
-static void nqssadce_build_swizzle(struct nqssadce_state *s,
-	struct prog_dst_register dst, struct prog_src_register src)
-{
-	struct prog_instruction *inst;
-	GLuint negatebase[2] = { 0, 0 };
-	int i;
-
-	for(i = 0; i < 4; ++i) {
-		GLuint swz = GET_SWZ(src.Swizzle, i);
-		if (swz == SWIZZLE_NIL)
-			continue;
-		negatebase[GET_BIT(src.Negate, i)] |= 1 << i;
-	}
-
-	_mesa_insert_instructions(s->Program, s->IP, (negatebase[0] ? 1 : 0) + (negatebase[1] ? 1 : 0));
-	inst = s->Program->Instructions + s->IP;
-
-	for(i = 0; i <= 1; ++i) {
-		if (!negatebase[i])
-			continue;
-
-		inst->Opcode = OPCODE_MOV;
-		inst->DstReg = dst;
-		inst->DstReg.WriteMask = negatebase[i];
-		inst->SrcReg[0] = src;
-		inst++;
-		s->IP++;
-	}
-}
-
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r500_fragment_program *fp,
-	struct r500_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-static void dump_program(struct r500_fragment_program_code *code);
-
-void r500TranslateFragmentShader(r300ContextPtr r300,
-				 struct r500_fragment_program *fp)
-{
-	struct r500_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r500_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformDeriv, 0 },
-			{ &radeonTransformTrigScale, 0 }
-		};
-		radeonLocalTransform(r300->radeon.glCtx, compiler.program,
-			4, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &is_native_swizzle,
-			.BuildSwizzle = &nqssadce_build_swizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		fp->translated = r500FragmentProgramEmit(&compiler);
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, 0);
-
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			if (fp->translated) {
-				_mesa_printf("Machine-readable code:\n");
-				dump_program(&fp->code);
-			}
-		}
-
-	}
-
-	update_params(r300, fp);
-
-}
-
-static char *toswiz(int swiz_val) {
-  switch(swiz_val) {
-  case 0: return "R";
-  case 1: return "G";
-  case 2: return "B";
-  case 3: return "A";
-  case 4: return "0";
-  case 5: return "1/2";
-  case 6: return "1";
-  case 7: return "U";
-  }
-  return NULL;
-}
-
-static char *toop(int op_val)
-{
-  char *str = NULL;
-  switch (op_val) {
-  case 0: str = "MAD"; break;
-  case 1: str = "DP3"; break;
-  case 2: str = "DP4"; break;
-  case 3: str = "D2A"; break;
-  case 4: str = "MIN"; break;
-  case 5: str = "MAX"; break;
-  case 6: str = "Reserved"; break;
-  case 7: str = "CND"; break;
-  case 8: str = "CMP"; break;
-  case 9: str = "FRC"; break;
-  case 10: str = "SOP"; break;
-  case 11: str = "MDH"; break;
-  case 12: str = "MDV"; break;
-  }
-  return str;
-}
-
-static char *to_alpha_op(int op_val)
-{
-  char *str = NULL;
-  switch (op_val) {
-  case 0: str = "MAD"; break;
-  case 1: str = "DP"; break;
-  case 2: str = "MIN"; break;
-  case 3: str = "MAX"; break;
-  case 4: str = "Reserved"; break;
-  case 5: str = "CND"; break;
-  case 6: str = "CMP"; break;
-  case 7: str = "FRC"; break;
-  case 8: str = "EX2"; break;
-  case 9: str = "LN2"; break;
-  case 10: str = "RCP"; break;
-  case 11: str = "RSQ"; break;
-  case 12: str = "SIN"; break;
-  case 13: str = "COS"; break;
-  case 14: str = "MDH"; break;
-  case 15: str = "MDV"; break;
-  }
-  return str;
-}
-
-static char *to_mask(int val)
-{
-  char *str = NULL;
-  switch(val) {
-  case 0: str = "NONE"; break;
-  case 1: str = "R"; break;
-  case 2: str = "G"; break;
-  case 3: str = "RG"; break;
-  case 4: str = "B"; break;
-  case 5: str = "RB"; break;
-  case 6: str = "GB"; break;
-  case 7: str = "RGB"; break;
-  case 8: str = "A"; break;
-  case 9: str = "AR"; break;
-  case 10: str = "AG"; break;
-  case 11: str = "ARG"; break;
-  case 12: str = "AB"; break;
-  case 13: str = "ARB"; break;
-  case 14: str = "AGB"; break;
-  case 15: str = "ARGB"; break;
-  }
-  return str;
-}
-
-static char *to_texop(int val)
-{
-  switch(val) {
-  case 0: return "NOP";
-  case 1: return "LD";
-  case 2: return "TEXKILL";
-  case 3: return "PROJ";
-  case 4: return "LODBIAS";
-  case 5: return "LOD";
-  case 6: return "DXDY";
-  }
-  return NULL;
-}
-
-static void dump_program(struct r500_fragment_program_code *code)
-{
-
-  fprintf(stderr, "R500 Fragment Program:\n--------\n");
-
-  int n;
-  uint32_t inst;
-  uint32_t inst0;
-  char *str = NULL;
-
-  if (code->const_nr) {
-    fprintf(stderr, "--------\nConstants:\n");
-    for (n = 0; n < code->const_nr; n++) {
-      fprintf(stderr, "Constant %d: %i[%i]\n", n,
-        code->constant[n].File, code->constant[n].Index);
-    }
-    fprintf(stderr, "--------\n");
-  }
-
-  for (n = 0; n < code->inst_end+1; n++) {
-    inst0 = inst = code->inst[n].inst0;
-    fprintf(stderr,"%d\t0:CMN_INST   0x%08x:", n, inst);
-    switch(inst & 0x3) {
-    case R500_INST_TYPE_ALU: str = "ALU"; break;
-    case R500_INST_TYPE_OUT: str = "OUT"; break;
-    case R500_INST_TYPE_FC: str = "FC"; break;
-    case R500_INST_TYPE_TEX: str = "TEX"; break;
-    };
-    fprintf(stderr,"%s %s %s %s %s ", str,
-	    inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "",
-	    inst & R500_INST_LAST ? "LAST" : "",
-	    inst & R500_INST_NOP ? "NOP" : "",
-	    inst & R500_INST_ALU_WAIT ? "ALU WAIT" : "");
-    fprintf(stderr,"wmask: %s omask: %s\n", to_mask((inst >> 11) & 0xf),
-	    to_mask((inst >> 15) & 0xf));
-
-    switch(inst0 & 0x3) {
-    case 0:
-    case 1:
-      fprintf(stderr,"\t1:RGB_ADDR   0x%08x:", code->inst[n].inst1);
-      inst = code->inst[n].inst1;
-
-      fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
-	      inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
-	      (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
-	      (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
-	      (inst >> 30));
-
-      fprintf(stderr,"\t2:ALPHA_ADDR 0x%08x:", code->inst[n].inst2);
-      inst = code->inst[n].inst2;
-      fprintf(stderr,"Addr0: %d%c, Addr1: %d%c, Addr2: %d%c, srcp:%d\n",
-	      inst & 0xff, (inst & (1<<8)) ? 'c' : 't',
-	      (inst >> 10) & 0xff, (inst & (1<<18)) ? 'c' : 't',
-	      (inst >> 20) & 0xff, (inst & (1<<28)) ? 'c' : 't',
-	      (inst >> 30));
-      fprintf(stderr,"\t3 RGB_INST:  0x%08x:", code->inst[n].inst3);
-      inst = code->inst[n].inst3;
-      fprintf(stderr,"rgb_A_src:%d %s/%s/%s %d rgb_B_src:%d %s/%s/%s %d\n",
-	      (inst) & 0x3, toswiz((inst >> 2) & 0x7), toswiz((inst >> 5) & 0x7), toswiz((inst >> 8) & 0x7),
-	      (inst >> 11) & 0x3,
-	      (inst >> 13) & 0x3, toswiz((inst >> 15) & 0x7), toswiz((inst >> 18) & 0x7), toswiz((inst >> 21) & 0x7),
-	      (inst >> 24) & 0x3);
-
-
-      fprintf(stderr,"\t4 ALPHA_INST:0x%08x:", code->inst[n].inst4);
-      inst = code->inst[n].inst4;
-      fprintf(stderr,"%s dest:%d%s alp_A_src:%d %s %d alp_B_src:%d %s %d w:%d\n", to_alpha_op(inst & 0xf),
-	      (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
-	      (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), (inst >> 17) & 0x3,
-	      (inst >> 19) & 0x3, toswiz((inst >> 21) & 0x7), (inst >> 24) & 0x3,
-	      (inst >> 31) & 0x1);
-
-      fprintf(stderr,"\t5 RGBA_INST: 0x%08x:", code->inst[n].inst5);
-      inst = code->inst[n].inst5;
-      fprintf(stderr,"%s dest:%d%s rgb_C_src:%d %s/%s/%s %d alp_C_src:%d %s %d\n", toop(inst & 0xf),
-	      (inst >> 4) & 0x7f, inst & (1<<11) ? "(rel)":"",
-	      (inst >> 12) & 0x3, toswiz((inst >> 14) & 0x7), toswiz((inst >> 17) & 0x7), toswiz((inst >> 20) & 0x7),
-	      (inst >> 23) & 0x3,
-	      (inst >> 25) & 0x3, toswiz((inst >> 27) & 0x7), (inst >> 30) & 0x3);
-      break;
-    case 2:
-      break;
-    case 3:
-      inst = code->inst[n].inst1;
-      fprintf(stderr,"\t1:TEX_INST:  0x%08x: id: %d op:%s, %s, %s %s\n", inst, (inst >> 16) & 0xf,
-	      to_texop((inst >> 22) & 0x7), (inst & (1<<25)) ? "ACQ" : "",
-	      (inst & (1<<26)) ? "IGNUNC" : "", (inst & (1<<27)) ? "UNSCALED" : "SCALED");
-      inst = code->inst[n].inst2;
-      fprintf(stderr,"\t2:TEX_ADDR:  0x%08x: src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n", inst,
-	      inst & 127, inst & (1<<7) ? "(rel)" : "",
-	      toswiz((inst >> 8) & 0x3), toswiz((inst >> 10) & 0x3),
-	      toswiz((inst >> 12) & 0x3), toswiz((inst >> 14) & 0x3),
-	      (inst >> 16) & 127, inst & (1<<23) ? "(rel)" : "",
-	      toswiz((inst >> 24) & 0x3), toswiz((inst >> 26) & 0x3),
-	      toswiz((inst >> 28) & 0x3), toswiz((inst >> 30) & 0x3));
-
-      fprintf(stderr,"\t3:TEX_DXDY:  0x%08x\n", code->inst[n].inst3);
-      break;
-    }
-    fprintf(stderr,"\n");
-  }
-
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_bo_legacy.c b/src/mesa/drivers/dri/r300/radeon_bo_legacy.c
new file mode 120000
index 0000000000..79ad050e6b
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_bo_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_bo_legacy.h b/src/mesa/drivers/dri/r300/radeon_bo_legacy.h
new file mode 120000
index 0000000000..83b0f7ffab
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_bo_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/r300/radeon_bocs_wrapper.h
new file mode 120000
index 0000000000..ca894b2443
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_bocs_wrapper.h
@@ -0,0 +1 @@
+../radeon/radeon_bocs_wrapper.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_buffer_objects.c b/src/mesa/drivers/dri/r300/radeon_buffer_objects.c
new file mode 120000
index 0000000000..f6a5f66470
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_buffer_objects.c
@@ -0,0 +1 @@
+../radeon/radeon_buffer_objects.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_buffer_objects.h b/src/mesa/drivers/dri/r300/radeon_buffer_objects.h
new file mode 120000
index 0000000000..2f134fd17b
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_buffer_objects.h
@@ -0,0 +1 @@
+../radeon/radeon_buffer_objects.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_chipset.h b/src/mesa/drivers/dri/r300/radeon_chipset.h
new file mode 120000
index 0000000000..eba99001ff
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_chipset.h
@@ -0,0 +1 @@
+../radeon/radeon_chipset.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_cmdbuf.h b/src/mesa/drivers/dri/r300/radeon_cmdbuf.h
new file mode 120000
index 0000000000..a799e1dc6d
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_cmdbuf.h
@@ -0,0 +1 @@
+../radeon/radeon_cmdbuf.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_common.c b/src/mesa/drivers/dri/r300/radeon_common.c
new file mode 120000
index 0000000000..67b19ba940
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_common.c
@@ -0,0 +1 @@
+../radeon/radeon_common.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_common.h b/src/mesa/drivers/dri/r300/radeon_common.h
new file mode 120000
index 0000000000..5bcb696a9f
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_common.h
@@ -0,0 +1 @@
+../radeon/radeon_common.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_common_context.c b/src/mesa/drivers/dri/r300/radeon_common_context.c
new file mode 120000
index 0000000000..86800f3819
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_common_context.c
@@ -0,0 +1 @@
+../radeon/radeon_common_context.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_common_context.h b/src/mesa/drivers/dri/r300/radeon_common_context.h
new file mode 120000
index 0000000000..4d66312550
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_common_context.h
@@ -0,0 +1 @@
+../radeon/radeon_common_context.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_context.c b/src/mesa/drivers/dri/r300/radeon_context.c
deleted file mode 100644
index 5267fe9a77..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_context.c
+++ /dev/null
@@ -1,330 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/**
- * \file radeon_context.c
- * Common context initialization.
- *
- * \author Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include <dlfcn.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/state.h"
-#include "main/matrix.h"
-#include "main/framebuffer.h"
-
-#include "drivers/common/driverfuncs.h"
-#include "swrast/swrast.h"
-
-#include "radeon_screen.h"
-#include "radeon_ioctl.h"
-#include "radeon_macros.h"
-#include "radeon_reg.h"
-
-#include "radeon_state.h"
-#include "r300_state.h"
-
-#include "utils.h"
-#include "vblank.h"
-#include "xmlpool.h"		/* for symbolic values of enum-type options */
-
-#define DRIVER_DATE "20060815"
-
-
-/* Return various strings for glGetString().
- */
-static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-	static char buffer[128];
-
-	switch (name) {
-	case GL_VENDOR:
-		if (IS_R300_CLASS(radeon->radeonScreen))
-			return (GLubyte *) "DRI R300 Project";
-		else
-			return (GLubyte *) "Tungsten Graphics, Inc.";
-
-	case GL_RENDERER:
-	{
-		unsigned offset;
-		GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
-			radeon->radeonScreen->AGPMode;
-		const char* chipname;
-
-		if (IS_R300_CLASS(radeon->radeonScreen))
-			chipname = "R300";
-		else
-			chipname = "R200";
-
-		offset = driGetRendererString(buffer, chipname, DRIVER_DATE,
-					      agp_mode);
-
-		if (IS_R300_CLASS(radeon->radeonScreen)) {
-		sprintf(&buffer[offset], " %sTCL",
-			(radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)
-			? "" : "NO-");
-		} else {
-			sprintf(&buffer[offset], " %sTCL",
-			!(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
-			? "" : "NO-");
-		}
-
-		return (GLubyte *) buffer;
-	}
-
-	default:
-		return NULL;
-	}
-}
-
-/* Initialize the driver's misc functions.
- */
-static void radeonInitDriverFuncs(struct dd_function_table *functions)
-{
-	functions->GetString = radeonGetString;
-}
-
-
-/**
- * Create and initialize all common fields of the context,
- * including the Mesa context itself.
- */
-GLboolean radeonInitContext(radeonContextPtr radeon,
-			    struct dd_function_table* functions,
-			    const __GLcontextModes * glVisual,
-			    __DRIcontextPrivate * driContextPriv,
-			    void *sharedContextPrivate)
-{
-	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
-	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
-	GLcontext* ctx;
-	GLcontext* shareCtx;
-	int fthrottle_mode;
-
-	/* Fill in additional standard functions. */
-	radeonInitDriverFuncs(functions);
-
-	radeon->radeonScreen = screen;
-	/* Allocate and initialize the Mesa context */
-	if (sharedContextPrivate)
-		shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx;
-	else
-		shareCtx = NULL;
-	radeon->glCtx = _mesa_create_context(glVisual, shareCtx,
-					    functions, (void *)radeon);
-	if (!radeon->glCtx)
-		return GL_FALSE;
-
-	ctx = radeon->glCtx;
-	driContextPriv->driverPrivate = radeon;
-
-	/* DRI fields */
-	radeon->dri.context = driContextPriv;
-	radeon->dri.screen = sPriv;
-	radeon->dri.drawable = NULL;
-	radeon->dri.readable = NULL;
-	radeon->dri.hwContext = driContextPriv->hHWContext;
-	radeon->dri.hwLock = &sPriv->pSAREA->lock;
-	radeon->dri.fd = sPriv->fd;
-	radeon->dri.drmMinor = sPriv->drm_version.minor;
-
-	radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
-					       screen->sarea_priv_offset);
-
-	/* Setup IRQs */
-	fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode");
-	radeon->iw.irq_seq = -1;
-	radeon->irqsEmitted = 0;
-	radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
-			  radeon->radeonScreen->irq);
-
-	radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
-
-	if (!radeon->do_irqs)
-		fprintf(stderr,
-			"IRQ's not enabled, falling back to %s: %d %d\n",
-			radeon->do_usleeps ? "usleeps" : "busy waits",
-			fthrottle_mode, radeon->radeonScreen->irq);
-
-	(*sPriv->systemTime->getUST) (&radeon->swap_ust);
-
-	return GL_TRUE;
-}
-
-
-/**
- * Cleanup common context fields.
- * Called by r200DestroyContext/r300DestroyContext
- */
-void radeonCleanupContext(radeonContextPtr radeon)
-{
-	/* _mesa_destroy_context() might result in calls to functions that
-	 * depend on the DriverCtx, so don't set it to NULL before.
-	 *
-	 * radeon->glCtx->DriverCtx = NULL;
-	 */
-
-	/* free the Mesa context */
-	_mesa_destroy_context(radeon->glCtx);
-
-	if (radeon->state.scissor.pClipRects) {
-		FREE(radeon->state.scissor.pClipRects);
-		radeon->state.scissor.pClipRects = 0;
-	}
-}
-
-
-/**
- * Swap front and back buffer.
- */
-void radeonSwapBuffers(__DRIdrawablePrivate * dPriv)
-{
-	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-		radeonContextPtr radeon;
-		GLcontext *ctx;
-
-		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-		ctx = radeon->glCtx;
-
-		if (ctx->Visual.doubleBufferMode) {
-			_mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-			if (radeon->doPageFlip) {
-				radeonPageFlip(dPriv);
-			} else {
-			    radeonCopyBuffer(dPriv, NULL);
-			}
-		}
-	} else {
-		/* XXX this shouldn't be an error but we can't handle it for now */
-		_mesa_problem(NULL, "%s: drawable has no context!",
-			      __FUNCTION__);
-	}
-}
-
-void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-			 int x, int y, int w, int h )
-{
-    if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-	radeonContextPtr radeon;
-	GLcontext *ctx;
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-	ctx = radeon->glCtx;
-
-	if (ctx->Visual.doubleBufferMode) {
-	    drm_clip_rect_t rect;
-	    rect.x1 = x + dPriv->x;
-	    rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	    rect.x2 = rect.x1 + w;
-	    rect.y2 = rect.y1 + h;
-	    _mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-	    radeonCopyBuffer(dPriv, &rect);
-	}
-    } else {
-	/* XXX this shouldn't be an error but we can't handle it for now */
-	_mesa_problem(NULL, "%s: drawable has no context!",
-		      __FUNCTION__);
-    }
-}
-
-/* Force the context `c' to be the current context and associate with it
- * buffer `b'.
- */
-GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-			    __DRIdrawablePrivate * driDrawPriv,
-			    __DRIdrawablePrivate * driReadPriv)
-{
-	if (driContextPriv) {
-		radeonContextPtr radeon =
-			(radeonContextPtr) driContextPriv->driverPrivate;
-
-		if (RADEON_DEBUG & DEBUG_DRI)
-			fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
-				radeon->glCtx);
-
-		if (radeon->dri.drawable != driDrawPriv) {
-			if (driDrawPriv->swap_interval == (unsigned)-1) {
-				driDrawPriv->vblFlags =
-					(radeon->radeonScreen->irq != 0)
-					? driGetDefaultVBlankFlags(&radeon->
-								   optionCache)
-					: VBLANK_FLAG_NO_IRQ;
-
-				driDrawableInitVBlank(driDrawPriv);
-			}
-		}
-
-		radeon->dri.readable = driReadPriv;
-
-		if (radeon->dri.drawable != driDrawPriv ||
-		    radeon->lastStamp != driDrawPriv->lastStamp) {
-			radeon->dri.drawable = driDrawPriv;
-
-			radeonSetCliprects(radeon);
-			r300UpdateViewportOffset(radeon->glCtx);
-		}
-
-		_mesa_make_current(radeon->glCtx,
-				    (GLframebuffer *) driDrawPriv->
-				    driverPrivate,
-				    (GLframebuffer *) driReadPriv->
-				    driverPrivate);
-
-		_mesa_update_state(radeon->glCtx);		
-
-		radeonUpdatePageFlipping(radeon);
-	} else {
-		if (RADEON_DEBUG & DEBUG_DRI)
-			fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-		_mesa_make_current(0, 0, 0);
-	}
-
-	if (RADEON_DEBUG & DEBUG_DRI)
-		fprintf(stderr, "End %s\n", __FUNCTION__);
-	return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv)
-{
-	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_DRI)
-		fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
-			radeon->glCtx);
-
-	return GL_TRUE;
-}
-
diff --git a/src/mesa/drivers/dri/r300/radeon_context.h b/src/mesa/drivers/dri/r300/radeon_context.h
index 47cbc22a72..250570f6b8 100644
--- a/src/mesa/drivers/dri/r300/radeon_context.h
+++ b/src/mesa/drivers/dri/r300/radeon_context.h
@@ -49,20 +49,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "dri_util.h"
 
-struct radeon_context;
-typedef struct radeon_context radeonContextRec;
-typedef struct radeon_context *radeonContextPtr;
-
-/* Rasterizing fallbacks */
-/* See correponding strings in r200_swtcl.c */
-#define RADEON_FALLBACK_TEXTURE		0x0001
-#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
-#define RADEON_FALLBACK_STENCIL		0x0004
-#define RADEON_FALLBACK_RENDER_MODE	0x0008
-#define RADEON_FALLBACK_BLEND_EQ	0x0010
-#define RADEON_FALLBACK_BLEND_FUNC	0x0020
-#define RADEON_FALLBACK_DISABLE		0x0040
-#define RADEON_FALLBACK_BORDER_MODE	0x0080
+#include "radeon_screen.h"
 
 #if R200_MERGED
 extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
@@ -79,155 +66,11 @@ extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 /* TCL fallbacks */
 extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 
-#define RADEON_TCL_FALLBACK_RASTER		0x0001	/* rasterization */
-#define RADEON_TCL_FALLBACK_UNFILLED		0x0002	/* unfilled tris */
-#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE	0x0004	/* twoside tris */
-#define RADEON_TCL_FALLBACK_MATERIAL		0x0008	/* material in vb */
-#define RADEON_TCL_FALLBACK_TEXGEN_0		0x0010	/* texgen, unit 0 */
-#define RADEON_TCL_FALLBACK_TEXGEN_1		0x0020	/* texgen, unit 1 */
-#define RADEON_TCL_FALLBACK_TEXGEN_2		0x0040	/* texgen, unit 2 */
-#define RADEON_TCL_FALLBACK_TEXGEN_3		0x0080	/* texgen, unit 3 */
-#define RADEON_TCL_FALLBACK_TEXGEN_4		0x0100	/* texgen, unit 4 */
-#define RADEON_TCL_FALLBACK_TEXGEN_5		0x0200	/* texgen, unit 5 */
-#define RADEON_TCL_FALLBACK_TCL_DISABLE		0x0400	/* user disable */
-#define RADEON_TCL_FALLBACK_BITMAP		0x0800	/* draw bitmap with points */
-#define RADEON_TCL_FALLBACK_VERTEX_PROGRAM	0x1000	/* vertex program active */
-
 #if R200_MERGED
 #define TCL_FALLBACK( ctx, bit, mode )	radeonTclFallback( ctx, bit, mode )
 #else
 #define TCL_FALLBACK( ctx, bit, mode )	;
 #endif
 
-struct radeon_dri_mirror {
-	__DRIcontextPrivate *context;	/* DRI context */
-	__DRIscreenPrivate *screen;	/* DRI screen */
-	/**
-	 * DRI drawable bound to this context for drawing.
-	 */
-	__DRIdrawablePrivate *drawable;
-
-	/**
-	 * DRI drawable bound to this context for reading.
-	 */
-	__DRIdrawablePrivate *readable;
-
-	drm_context_t hwContext;
-	drm_hw_lock_t *hwLock;
-	int fd;
-	int drmMinor;
-};
-
-/**
- * Derived state for internal purposes.
- */
-struct radeon_scissor_state {
-	drm_clip_rect_t rect;
-	GLboolean enabled;
-
-	GLuint numClipRects;	/* Cliprects active */
-	GLuint numAllocedClipRects;	/* Cliprects available */
-	drm_clip_rect_t *pClipRects;
-};
-
-struct radeon_colorbuffer_state {
-	GLuint clear;
-	GLint drawOffset, drawPitch;
-};
-
-struct radeon_state {
-	struct radeon_colorbuffer_state color;
-	struct radeon_scissor_state scissor;
-};
-
-/**
- * Common per-context variables shared by R200 and R300.
- * R200- and R300-specific code "derive" their own context from this
- * structure.
- */
-struct radeon_context {
-	GLcontext *glCtx;	/* Mesa context */
-	radeonScreenPtr radeonScreen;	/* Screen private DRI data */
-
-	/* Fallback state */
-	GLuint Fallback;
-	GLuint TclFallback;
-
-	/* Page flipping */
-	GLuint doPageFlip;
-
-	/* Drawable, cliprect and scissor information */
-	GLuint numClipRects;	/* Cliprects for the draw buffer */
-	drm_clip_rect_t *pClipRects;
-	unsigned int lastStamp;
-	GLboolean lost_context;
-	drm_radeon_sarea_t *sarea;	/* Private SAREA data */
-
-	/* Mirrors of some DRI state */
-	struct radeon_dri_mirror dri;
-
-	/* Busy waiting */
-	GLuint do_usleeps;
-	GLuint do_irqs;
-	GLuint irqsEmitted;
-	drm_radeon_irq_wait_t iw;
-
-	/* buffer swap */
-	int64_t swap_ust;
-	int64_t swap_missed_ust;
-
-	GLuint swap_count;
-	GLuint swap_missed_count;
-
-	/* Derived state */
-	struct radeon_state state;
-
-	/* Configuration cache
-	 */
-	driOptionCache optionCache;
-};
-
-#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx))
-
-extern void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
-extern void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-				int x, int y, int w, int h);
-extern GLboolean radeonInitContext(radeonContextPtr radeon,
-				   struct dd_function_table *functions,
-				   const __GLcontextModes * glVisual,
-				   __DRIcontextPrivate * driContextPriv,
-				   void *sharedContextPrivate);
-extern void radeonCleanupContext(radeonContextPtr radeon);
-extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-				   __DRIdrawablePrivate * driDrawPriv,
-				   __DRIdrawablePrivate * driReadPriv);
-extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
-
-/* ================================================================
- * Debugging:
- */
-#define DO_DEBUG		1
-
-#if DO_DEBUG
-extern int RADEON_DEBUG;
-#else
-#define RADEON_DEBUG		0
-#endif
-
-#define DEBUG_TEXTURE	0x0001
-#define DEBUG_STATE	0x0002
-#define DEBUG_IOCTL	0x0004
-#define DEBUG_PRIMS	0x0008
-#define DEBUG_VERTS	0x0010
-#define DEBUG_FALLBACKS	0x0020
-#define DEBUG_VFMT	0x0040
-#define DEBUG_CODEGEN	0x0080
-#define DEBUG_VERBOSE	0x0100
-#define DEBUG_DRI       0x0200
-#define DEBUG_DMA       0x0400
-#define DEBUG_SANITY    0x0800
-#define DEBUG_SYNC      0x1000
-#define DEBUG_PIXEL     0x2000
-#define DEBUG_MEMORY    0x4000
 
 #endif				/* __RADEON_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r300/radeon_cs_legacy.c b/src/mesa/drivers/dri/r300/radeon_cs_legacy.c
new file mode 120000
index 0000000000..006720f8a4
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_cs_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_cs_legacy.h b/src/mesa/drivers/dri/r300/radeon_cs_legacy.h
new file mode 120000
index 0000000000..a5f95e0a3d
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_cs_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_cs_space_drm.c b/src/mesa/drivers/dri/r300/radeon_cs_space_drm.c
new file mode 120000
index 0000000000..c248ea7d1a
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_cs_space_drm.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_space_drm.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_debug.c b/src/mesa/drivers/dri/r300/radeon_debug.c
new file mode 120000
index 0000000000..c98c2e074c
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_debug.c
@@ -0,0 +1 @@
+../radeon/radeon_debug.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_debug.h b/src/mesa/drivers/dri/r300/radeon_debug.h
new file mode 120000
index 0000000000..bd8aa28e89
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_debug.h
@@ -0,0 +1 @@
+../radeon/radeon_debug.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_dma.c b/src/mesa/drivers/dri/r300/radeon_dma.c
new file mode 120000
index 0000000000..43be000625
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_dma.c
@@ -0,0 +1 @@
+../radeon/radeon_dma.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_dma.h b/src/mesa/drivers/dri/r300/radeon_dma.h
new file mode 120000
index 0000000000..82e50634e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_dma.h
@@ -0,0 +1 @@
+../radeon/radeon_dma.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_fbo.c b/src/mesa/drivers/dri/r300/radeon_fbo.c
new file mode 120000
index 0000000000..0d738d8d78
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_fbo.c
@@ -0,0 +1 @@
+../radeon/radeon_fbo.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.c b/src/mesa/drivers/dri/r300/radeon_ioctl.c
deleted file mode 100644
index f042a7b943..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_ioctl.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include <sched.h>
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/macros.h"
-#include "main/context.h"
-#include "swrast/swrast.h"
-#include "r300_context.h"
-#include "radeon_ioctl.h"
-#include "r300_ioctl.h"
-#include "r300_state.h"
-#include "radeon_reg.h"
-
-#include "drirenderbuffer.h"
-#include "vblank.h"
-
-static void radeonWaitForIdle(radeonContextPtr radeon);
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t radeonGetLastFrame(radeonContextPtr radeon)
-{
-	drm_radeon_getparam_t gp;
-	int ret;
-	uint32_t frame = 0;
-
-	gp.param = RADEON_PARAM_LAST_FRAME;
-	gp.value = (int *)&frame;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
-				  &gp, sizeof(gp));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-
-	return frame;
-}
-
-uint32_t radeonGetAge(radeonContextPtr radeon)
-{
-	drm_radeon_getparam_t gp;
-	int ret;
-	uint32_t age = 0;
-
-	gp.param = RADEON_PARAM_LAST_CLEAR;
-	gp.value = (int *)&age;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
-				  &gp, sizeof(gp));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-
-	return age;
-}
-
-static void radeonEmitIrqLocked(radeonContextPtr radeon)
-{
-	drm_radeon_irq_emit_t ie;
-	int ret;
-
-	ie.irq_seq = &radeon->iw.irq_seq;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT,
-				  &ie, sizeof(ie));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-}
-
-static void radeonWaitIrq(radeonContextPtr radeon)
-{
-	int ret;
-
-	do {
-		ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT,
-				      &radeon->iw, sizeof(radeon->iw));
-	} while (ret && (errno == EINTR || errno == EBUSY));
-
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-}
-
-static void radeonWaitForFrameCompletion(radeonContextPtr radeon)
-{
-	drm_radeon_sarea_t *sarea = radeon->sarea;
-
-	if (radeon->do_irqs) {
-		if (radeonGetLastFrame(radeon) < sarea->last_frame) {
-			if (!radeon->irqsEmitted) {
-				while (radeonGetLastFrame(radeon) <
-				       sarea->last_frame) ;
-			} else {
-				UNLOCK_HARDWARE(radeon);
-				radeonWaitIrq(radeon);
-				LOCK_HARDWARE(radeon);
-			}
-			radeon->irqsEmitted = 10;
-		}
-
-		if (radeon->irqsEmitted) {
-			radeonEmitIrqLocked(radeon);
-			radeon->irqsEmitted--;
-		}
-	} else {
-		while (radeonGetLastFrame(radeon) < sarea->last_frame) {
-			UNLOCK_HARDWARE(radeon);
-			if (radeon->do_usleeps)
-				DO_USLEEP(1);
-			LOCK_HARDWARE(radeon);
-		}
-	}
-}
-
-/* Copy the back color buffer to the front color buffer.
- */
-void radeonCopyBuffer(__DRIdrawablePrivate * dPriv,
-		      const drm_clip_rect_t	 * rect)
-{
-	radeonContextPtr radeon;
-	GLint nbox, i, ret;
-	GLboolean missed_target;
-	int64_t ust;
-	__DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-	assert(dPriv);
-	assert(dPriv->driContextPriv);
-	assert(dPriv->driContextPriv->driverPrivate);
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "\n%s( %p )\n\n", __FUNCTION__,
-			(void *)radeon->glCtx);
-	}
-
-	r300Flush(radeon->glCtx);
-
-	LOCK_HARDWARE(radeon);
-
-	/* Throttle the frame rate -- only allow one pending swap buffers
-	 * request at a time.
-	 */
-	radeonWaitForFrameCompletion(radeon);
-	if (!rect)
-	{
-	    UNLOCK_HARDWARE(radeon);
-	    driWaitForVBlank(dPriv, &missed_target);
-	    LOCK_HARDWARE(radeon);
-	}
-
-	nbox = dPriv->numClipRects;	/* must be in locked region */
-
-	for (i = 0; i < nbox;) {
-		GLint nr = MIN2(i + RADEON_NR_SAREA_CLIPRECTS, nbox);
-		drm_clip_rect_t *box = dPriv->pClipRects;
-		drm_clip_rect_t *b = radeon->sarea->boxes;
-		GLint n = 0;
-
-		for ( ; i < nr ; i++ ) {
-
-		    *b = box[i];
-
-		    if (rect)
-		    {
-			if (rect->x1 > b->x1)
-			    b->x1 = rect->x1;
-			if (rect->y1 > b->y1)
-			    b->y1 = rect->y1;
-			if (rect->x2 < b->x2)
-			    b->x2 = rect->x2;
-			if (rect->y2 < b->y2)
-			    b->y2 = rect->y2;
-
-			if (b->x1 >= b->x2 || b->y1 >= b->y2)
-			    continue;
-		    }
-
-		    b++;
-		    n++;
-		}
-		radeon->sarea->nbox = n;
-
-		if (!n)
-		   continue;
-
-		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_SWAP);
-
-		if (ret) {
-			fprintf(stderr, "DRM_RADEON_SWAP: return = %d\n",
-				ret);
-			UNLOCK_HARDWARE(radeon);
-			exit(1);
-		}
-	}
-
-	UNLOCK_HARDWARE(radeon);
-	if (!rect)
-	{
-	    ((r300ContextPtr)radeon)->hw.all_dirty = GL_TRUE;
-
-	    radeon->swap_count++;
-	    (*psp->systemTime->getUST) (&ust);
-	    if (missed_target) {
-		radeon->swap_missed_count++;
-		radeon->swap_missed_ust = ust - radeon->swap_ust;
-	    }
-
-	    radeon->swap_ust = ust;
-
-	    sched_yield();
-	}
-}
-
-void radeonPageFlip(__DRIdrawablePrivate * dPriv)
-{
-	radeonContextPtr radeon;
-	GLint ret;
-	GLboolean missed_target;
-	__DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-	assert(dPriv);
-	assert(dPriv->driContextPriv);
-	assert(dPriv->driContextPriv->driverPrivate);
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-			radeon->sarea->pfCurrentPage);
-	}
-
-	r300Flush(radeon->glCtx);
-	LOCK_HARDWARE(radeon);
-
-	if (!dPriv->numClipRects) {
-		UNLOCK_HARDWARE(radeon);
-		usleep(10000);	/* throttle invisible client 10ms */
-		return;
-	}
-
-	/* Need to do this for the perf box placement:
-	 */
-	{
-		drm_clip_rect_t *box = dPriv->pClipRects;
-		drm_clip_rect_t *b = radeon->sarea->boxes;
-		b[0] = box[0];
-		radeon->sarea->nbox = 1;
-	}
-
-	/* Throttle the frame rate -- only allow a few pending swap buffers
-	 * request at a time.
-	 */
-	radeonWaitForFrameCompletion(radeon);
-	UNLOCK_HARDWARE(radeon);
-	driWaitForVBlank(dPriv, &missed_target);
-	if (missed_target) {
-		radeon->swap_missed_count++;
-		(void)(*psp->systemTime->getUST) (&radeon->swap_missed_ust);
-	}
-	LOCK_HARDWARE(radeon);
-
-	ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_FLIP);
-
-	UNLOCK_HARDWARE(radeon);
-
-	if (ret) {
-		fprintf(stderr, "DRM_RADEON_FLIP: return = %d\n", ret);
-		exit(1);
-	}
-
-	radeon->swap_count++;
-	(void)(*psp->systemTime->getUST) (&radeon->swap_ust);
-
-        driFlipRenderbuffers(radeon->glCtx->WinSysDrawBuffer, 
-                             radeon->sarea->pfCurrentPage);
-
-	if (radeon->sarea->pfCurrentPage == 1) {
-		radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch;
-	} else {
-		radeon->state.color.drawOffset = radeon->radeonScreen->backOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->backPitch;
-	}
-
-	if (IS_R300_CLASS(radeon->radeonScreen)) {
-		r300ContextPtr r300 = (r300ContextPtr)radeon;
-		R300_STATECHANGE(r300, cb);
-		r300->hw.cb.cmd[R300_CB_OFFSET] = r300->radeon.state.color.drawOffset + 
-						r300->radeon.radeonScreen->fbLocation;
-		r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.state.color.drawPitch;
-		
-		if (r300->radeon.radeonScreen->cpp == 4)
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-		else
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-	
-		if (r300->radeon.sarea->tiling_enabled)
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-	}
-}
-
-void radeonWaitForIdleLocked(radeonContextPtr radeon)
-{
-	int ret;
-	int i = 0;
-
-	do {
-		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE);
-		if (ret)
-			DO_USLEEP(1);
-	} while (ret && ++i < 100);
-
-	if (ret < 0) {
-		UNLOCK_HARDWARE(radeon);
-		fprintf(stderr, "Error: R300 timed out... exiting\n");
-		exit(-1);
-	}
-}
-
-static void radeonWaitForIdle(radeonContextPtr radeon)
-{
-	LOCK_HARDWARE(radeon);
-	radeonWaitForIdleLocked(radeon);
-	UNLOCK_HARDWARE(radeon);
-}
-
-void radeonFlush(GLcontext * ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	if (IS_R300_CLASS(radeon->radeonScreen))
-		r300Flush(ctx);
-}
-
-
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void radeonFinish(GLcontext * ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	radeonFlush(ctx);
-
-	if (radeon->do_irqs) {
-		LOCK_HARDWARE(radeon);
-		radeonEmitIrqLocked(radeon);
-		UNLOCK_HARDWARE(radeon);
-		radeonWaitIrq(radeon);
-	} else
-		radeonWaitForIdle(radeon);
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.h b/src/mesa/drivers/dri/r300/radeon_ioctl.h
deleted file mode 100644
index 3add775b82..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_ioctl.h
+++ /dev/null
@@ -1,57 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __RADEON_IOCTL_H__
-#define __RADEON_IOCTL_H__
-
-#include "main/simple_list.h"
-#include "radeon_dri.h"
-#include "radeon_lock.h"
-
-#include "xf86drm.h"
-#include "drm.h"
-#if 0
-#include "r200context.h"
-#endif
-#include "radeon_drm.h"
-
-extern void radeonCopyBuffer(__DRIdrawablePrivate * drawable,
-			     const drm_clip_rect_t	* rect);
-extern void radeonPageFlip(__DRIdrawablePrivate * drawable);
-extern void radeonFlush(GLcontext * ctx);
-extern void radeonFinish(GLcontext * ctx);
-extern void radeonWaitForIdleLocked(radeonContextPtr radeon);
-extern uint32_t radeonGetAge(radeonContextPtr radeon);
-
-#endif				/* __RADEON_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r300/radeon_lock.c b/src/mesa/drivers/dri/r300/radeon_lock.c
index 4f47afd5dc..af4108a8e3 100644..120000
--- a/src/mesa/drivers/dri/r300/radeon_lock.c
+++ b/src/mesa/drivers/dri/r300/radeon_lock.c
@@ -1,137 +1 @@
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Gareth Hughes <gareth@valinux.com>
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Kevin E. Martin <martin@valinux.com>
- */
-
-#include "radeon_lock.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
-#include "r300_context.h"
-#include "r300_state.h"
-
-#include "main/framebuffer.h"
-
-#include "drirenderbuffer.h"
-
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-void radeonUpdatePageFlipping(radeonContextPtr rmesa)
-{
-	int use_back;
-
-	rmesa->doPageFlip = rmesa->sarea->pfState;
-	if (rmesa->glCtx->WinSysDrawBuffer) {
-		driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-				     rmesa->sarea->pfCurrentPage);
-		r300UpdateDrawBuffer(rmesa->glCtx);
-	}
-
-	use_back = rmesa->glCtx->DrawBuffer ?
-	    (rmesa->glCtx->DrawBuffer->_ColorDrawBufferIndexes[0] ==
-	     BUFFER_BACK_LEFT) : 1;
-	use_back ^= (rmesa->sarea->pfCurrentPage == 1);
-
-	if (use_back) {
-		rmesa->state.color.drawOffset =
-		    rmesa->radeonScreen->backOffset;
-		rmesa->state.color.drawPitch = rmesa->radeonScreen->backPitch;
-	} else {
-		rmesa->state.color.drawOffset =
-		    rmesa->radeonScreen->frontOffset;
-		rmesa->state.color.drawPitch =
-		    rmesa->radeonScreen->frontPitch;
-	}
-}
-
-/* Update the hardware state.  This is called if another context has
- * grabbed the hardware lock, which includes the X server.  This
- * function also updates the driver's window state after the X server
- * moves, resizes or restacks a window -- the change will be reflected
- * in the drawable position and clip rects.  Since the X server grabs
- * the hardware lock when it changes the window state, this routine will
- * automatically be called after such a change.
- */
-void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
-{
-	__DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-	__DRIdrawablePrivate *const readable = rmesa->dri.readable;
-	__DRIscreenPrivate *sPriv = rmesa->dri.screen;
-	drm_radeon_sarea_t *sarea = rmesa->sarea;
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-
-	assert(drawable != NULL);
-
-	drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags);
-
-	/* The window might have moved, so we might need to get new clip
-	 * rects.
-	 *
-	 * NOTE: This releases and regrabs the hw lock to allow the X server
-	 * to respond to the DRI protocol request for new drawable info.
-	 * Since the hardware state depends on having the latest drawable
-	 * clip rects, all state checking must be done _after_ this call.
-	 */
-	DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable);
-	if (drawable != readable) {
-		DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable);
-	}
-
-	if (rmesa->lastStamp != drawable->lastStamp) {
-		radeonUpdatePageFlipping(rmesa);
-		radeonSetCliprects(rmesa);
-		r300UpdateViewportOffset(rmesa->glCtx);
-		driUpdateFramebufferSize(rmesa->glCtx, drawable);
-	}
-
-	if (sarea->ctx_owner != rmesa->dri.hwContext) {
-		int i;
-
-		sarea->ctx_owner = rmesa->dri.hwContext;
-		for (i = 0; i < r300->nr_heaps; i++) {
-			DRI_AGE_TEXTURES(r300->texture_heaps[i]);
-		}
-	}
-
-	rmesa->lost_context = GL_TRUE;
-}
+../radeon/radeon_lock.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_lock.h b/src/mesa/drivers/dri/r300/radeon_lock.h
index a344837f47..64bdf94ee7 100644..120000
--- a/src/mesa/drivers/dri/r300/radeon_lock.h
+++ b/src/mesa/drivers/dri/r300/radeon_lock.h
@@ -1,115 +1 @@
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Gareth Hughes <gareth@valinux.com>
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Kevin E. Martin <martin@valinux.com>
- */
-
-#ifndef __RADEON_LOCK_H__
-#define __RADEON_LOCK_H__
-
-#include "radeon_context.h"
-
-extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
-extern void radeonUpdatePageFlipping(radeonContextPtr rmesa);
-
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if (prevLockFile) {						\
-	 fprintf(stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__);	\
-	 exit(1);							\
-      }									\
-   } while (0)
-
-#else
-
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
-
-#endif
-
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
-
-/* Lock the hardware and validate our state.
- */
-#define LOCK_HARDWARE( rmesa )						\
-	do {								\
-		char __ret = 0;						\
-		DEBUG_CHECK_LOCK();					\
-		DRM_CAS((rmesa)->dri.hwLock, (rmesa)->dri.hwContext,	\
-			(DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret); \
-		if (__ret)						\
-			radeonGetLock((rmesa), 0);			\
-		DEBUG_LOCK();						\
-	} while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-	do {								\
-		DRM_UNLOCK((rmesa)->dri.fd,				\
-			(rmesa)->dri.hwLock,				\
-			(rmesa)->dri.hwContext);			\
-		DEBUG_RESET();						\
-	} while (0)
-
-#endif				/* __RADEON_LOCK_H__ */
+../radeon/radeon_lock.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_mipmap_tree.c b/src/mesa/drivers/dri/r300/radeon_mipmap_tree.c
new file mode 120000
index 0000000000..31c0cfbe94
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_mipmap_tree.c
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_mipmap_tree.h b/src/mesa/drivers/dri/r300/radeon_mipmap_tree.h
new file mode 120000
index 0000000000..254d50cf8c
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_mipmap_tree.h
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_program.c b/src/mesa/drivers/dri/r300/radeon_program.c
deleted file mode 100644
index da5e7aefce..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_program.c
+++ /dev/null
@@ -1,128 +0,0 @@
-/*
- * Copyright (C) 2008 Nicolai Haehnle.
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-#include "radeon_program.h"
-
-#include "shader/prog_print.h"
-
-
-/**
- * Transform the given clause in the following way:
- *  1. Replace it with an empty clause
- *  2. For every instruction in the original clause, try the given
- *     transformations in order.
- *  3. If one of the transformations returns GL_TRUE, assume that it
- *     has emitted the appropriate instruction(s) into the new clause;
- *     otherwise, copy the instruction verbatim.
- *
- * \note The transformation is currently not recursive; in other words,
- * instructions emitted by transformations are not transformed.
- *
- * \note The transform is called 'local' because it can only look at
- * one instruction at a time.
- */
-void radeonLocalTransform(
-	GLcontext *Ctx,
-	struct gl_program *program,
-	int num_transformations,
-	struct radeon_program_transformation* transformations)
-{
-	struct radeon_transform_context ctx;
-	int ip;
-
-	ctx.Ctx = Ctx;
-	ctx.Program = program;
-	ctx.OldInstructions = program->Instructions;
-	ctx.OldNumInstructions = program->NumInstructions;
-
-	program->Instructions = 0;
-	program->NumInstructions = 0;
-
-	for(ip = 0; ip < ctx.OldNumInstructions; ++ip) {
-		struct prog_instruction *instr = ctx.OldInstructions + ip;
-		int i;
-
-		for(i = 0; i < num_transformations; ++i) {
-			struct radeon_program_transformation* t = transformations + i;
-
-			if (t->function(&ctx, instr, t->userData))
-				break;
-		}
-
-		if (i >= num_transformations) {
-			struct prog_instruction* dest = radeonAppendInstructions(program, 1);
-			_mesa_copy_instructions(dest, instr, 1);
-		}
-	}
-
-	_mesa_free_instructions(ctx.OldInstructions, ctx.OldNumInstructions);
-}
-
-
-static void scan_instructions(GLboolean* used, const struct prog_instruction* insts, GLuint count)
-{
-	GLuint i;
-	for (i = 0; i < count; i++) {
-		const struct prog_instruction *inst = insts + i;
-		const GLuint n = _mesa_num_inst_src_regs(inst->Opcode);
-		GLuint k;
-
-		for (k = 0; k < n; k++) {
-			if (inst->SrcReg[k].File == PROGRAM_TEMPORARY)
-				used[inst->SrcReg[k].Index] = GL_TRUE;
-		}
-	}
-}
-
-GLint radeonFindFreeTemporary(struct radeon_transform_context *t)
-{
-	GLboolean used[MAX_PROGRAM_TEMPS];
-	GLuint i;
-
-	_mesa_memset(used, 0, sizeof(used));
-	scan_instructions(used, t->Program->Instructions, t->Program->NumInstructions);
-	scan_instructions(used, t->OldInstructions, t->OldNumInstructions);
-
-	for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
-		if (!used[i])
-			return i;
-	}
-
-	return -1;
-}
-
-
-/**
- * Append the given number of instructions to the program and return a
- * pointer to the first new instruction.
- */
-struct prog_instruction *radeonAppendInstructions(struct gl_program *program, int count)
-{
-	int oldnum = program->NumInstructions;
-	_mesa_insert_instructions(program, oldnum, count);
-	return program->Instructions + oldnum;
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_queryobj.c b/src/mesa/drivers/dri/r300/radeon_queryobj.c
new file mode 120000
index 0000000000..1d6ebc1c48
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_queryobj.c
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_queryobj.h b/src/mesa/drivers/dri/r300/radeon_queryobj.h
new file mode 120000
index 0000000000..8f6f842b0a
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_queryobj.h
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_screen.c b/src/mesa/drivers/dri/r300/radeon_screen.c
new file mode 120000
index 0000000000..86161118dd
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_screen.c
@@ -0,0 +1 @@
+../radeon/radeon_screen.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_screen.h b/src/mesa/drivers/dri/r300/radeon_screen.h
new file mode 120000
index 0000000000..23bb6bd459
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_screen.h
@@ -0,0 +1 @@
+../radeon/radeon_screen.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_span.c b/src/mesa/drivers/dri/r300/radeon_span.c
index 16f9fb99e6..232868c4c9 100644..120000
--- a/src/mesa/drivers/dri/r300/radeon_span.c
+++ b/src/mesa/drivers/dri/r300/radeon_span.c
@@ -1,349 +1 @@
-/**************************************************************************
-
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#include "main/glheader.h"
-#include "swrast/swrast.h"
-
-#include "r300_state.h"
-#include "radeon_ioctl.h"
-#include "r300_ioctl.h"
-#include "radeon_span.h"
-
-#include "drirenderbuffer.h"
-
-#define DBG 0
-
-/*
- * Note that all information needed to access pixels in a renderbuffer
- * should be obtained through the gl_renderbuffer parameter, not per-context
- * information.
- */
-#define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
-
-#define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
-
-#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
-
-#define Y_FLIP(Y) (bottom - (Y))
-
-#define HW_LOCK()
-
-#define HW_UNLOCK()
-
-/* ================================================================
- * Color buffer
- */
-
-/* 16 bit, RGB565 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    radeon##x##_RGB565
-#define TAG2(x,y) radeon##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
-#include "spantmp2.h"
-
-/* 32 bit, ARGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    radeon##x##_ARGB8888
-#define TAG2(x,y) radeon##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
-#include "spantmp2.h"
-
-/* ================================================================
- * Depth buffer
- */
-
-/* The Radeon family has depth tiling on all the time, so we have to convert
- * the x,y coordinates into the memory bus address (mba) in the same
- * manner as the engine.  In each case, the linear block address (ba)
- * is calculated, and then wired with x and y to produce the final
- * memory address.
- * The chip will do address translation on its own if the surface registers
- * are set up correctly. It is not quite enough to get it working with hyperz
- * too...
- */
-
-static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 4 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0..1] = 0           */
-
-#ifdef COMPILE_R300
-		ba = (y / 8) * (pitch / 8) + (x / 8);
-#else
-		ba = (y / 16) * (pitch / 16) + (x / 16);
-#endif
-
-		address |= (x & 0x7) << 2;	/* a[2..4] = x[0..2]     */
-		address |= (y & 0x3) << 5;	/* a[5..6] = y[0..1]     */
-		address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5;	/* a[7]    = x[4] ^ y[2] */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7;	/* a[11]   = x[3] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-static INLINE GLuint
-radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 2 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0]    = 0           */
-
-		ba = (y / 16) * (pitch / 32) + (x / 32);
-
-		address |= (x & 0x7) << 1;	/* a[1..3] = x[0..2]     */
-		address |= (y & 0x7) << 4;	/* a[4..6] = y[0..2]     */
-		address |= (x & 0x8) << 4;	/* a[7]    = x[3]        */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= ((x & 0x10) ^ (y & 0x10)) << 7;	/* a[11]   = x[4] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-/* 16-bit depth buffer functions
- */
-#define VALUE_TYPE GLushort
-
-#define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d;
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo ));
-
-#define TAG(x) radeon##x##_z16
-#include "depthtmp.h"
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- *
- * Careful: It looks like the R300 uses ZZZS byte order while the R200
- * uses SZZZ for 24 bit depth, 8 bit stencil mode.
- */
-#define VALUE_TYPE GLuint
-
-#ifdef COMPILE_R300
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x000000ff;							\
-   tmp |= ((d << 8) & 0xffffff00);					\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#else
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#endif
-
-#ifdef COMPILE_R300
-#define READ_DEPTH( d, _x, _y )						\
-  do { \
-    d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,		\
-					 _y + yo )) & 0xffffff00) >> 8; \
-  }while(0)
-#else
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
-#endif
-
-#define TAG(x) radeon##x##_z24_s8
-#include "depthtmp.h"
-
-/* ================================================================
- * Stencil buffer
- */
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#ifdef COMPILE_R300
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xffffff00;							\
-   tmp |= (d) & 0xff;							\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#else
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x00ffffff;							\
-   tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#endif
-
-#ifdef COMPILE_R300
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   d = tmp & 0x000000ff;						\
-} while (0)
-#else
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   d = (tmp & 0xff000000) >> 24;					\
-} while (0)
-#endif
-
-#define TAG(x) radeon##x##_z24_s8
-#include "stenciltmp.h"
-
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
-
-static void radeonSpanRenderStart(GLcontext * ctx)
-{
-	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-#ifdef COMPILE_R300
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-	R300_FIREVERTICES(r300);
-#else
-	RADEON_FIREVERTICES(rmesa);
-#endif
-	LOCK_HARDWARE(rmesa);
-	radeonWaitForIdleLocked(rmesa);
-
-	/* Read the first pixel in the frame buffer.  This should
-	 * be a noop, right?  In fact without this conform fails as reading
-	 * from the framebuffer sometimes produces old results -- the
-	 * on-card read cache gets mixed up and doesn't notice that the
-	 * framebuffer has been updated.
-	 *
-	 * Note that we should probably be reading some otherwise unused
-	 * region of VRAM, otherwise we might get incorrect results when
-	 * reading pixels from the top left of the screen.
-	 *
-	 * I found this problem on an R420 with glean's texCube test.
-	 * Note that the R200 span code also *writes* the first pixel in the
-	 * framebuffer, but I've found this to be unnecessary.
-	 *  -- Nicolai Hähnle, June 2008
-	 */
-	{
-		int p;
-		driRenderbuffer *drb =
-			(driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0];
-		volatile int *buf =
-			(volatile int *)(rmesa->dri.screen->pFB + drb->offset);
-		p = *buf;
-	}
-}
-
-static void radeonSpanRenderFinish(GLcontext * ctx)
-{
-	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-	_swrast_flush(ctx);
-	UNLOCK_HARDWARE(rmesa);
-}
-
-void radeonInitSpanFuncs(GLcontext * ctx)
-{
-	struct swrast_device_driver *swdd =
-	    _swrast_GetDeviceDriverReference(ctx);
-	swdd->SpanRenderStart = radeonSpanRenderStart;
-	swdd->SpanRenderFinish = radeonSpanRenderFinish;
-}
-
-/**
- * Plug in the Get/Put routines for the given driRenderbuffer.
- */
-void radeonSetSpanFunctions(driRenderbuffer * drb, const GLvisual * vis)
-{
-	if (drb->Base.InternalFormat == GL_RGBA) {
-		if (vis->redBits == 5 && vis->greenBits == 6
-		    && vis->blueBits == 5) {
-			radeonInitPointers_RGB565(&drb->Base);
-		} else {
-			radeonInitPointers_ARGB8888(&drb->Base);
-		}
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-		radeonInitDepthPointers_z16(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-		radeonInitDepthPointers_z24_s8(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-		radeonInitStencilPointers_z24_s8(&drb->Base);
-	}
-}
+../radeon/radeon_span.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_span.h b/src/mesa/drivers/dri/r300/radeon_span.h
new file mode 120000
index 0000000000..f9d634508c
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_span.h
@@ -0,0 +1 @@
+../radeon/radeon_span.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_state.c b/src/mesa/drivers/dri/r300/radeon_state.c
deleted file mode 100644
index c401da6c54..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_state.c
+++ /dev/null
@@ -1,244 +0,0 @@
-/**************************************************************************
-
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/api_arrayelt.h"
-#include "main/enums.h"
-#include "main/framebuffer.h"
-#include "main/colormac.h"
-#include "main/light.h"
-
-#include "swrast/swrast.h"
-#include "vbo/vbo.h"
-#include "tnl/tnl.h"
-#include "tnl/t_pipeline.h"
-#include "swrast_setup/swrast_setup.h"
-
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
-#include "r300_ioctl.h"
-
-
-/* =============================================================
- * Scissoring
- */
-
-static GLboolean intersect_rect(drm_clip_rect_t * out,
-				drm_clip_rect_t * a, drm_clip_rect_t * b)
-{
-	*out = *a;
-	if (b->x1 > out->x1)
-		out->x1 = b->x1;
-	if (b->y1 > out->y1)
-		out->y1 = b->y1;
-	if (b->x2 < out->x2)
-		out->x2 = b->x2;
-	if (b->y2 < out->y2)
-		out->y2 = b->y2;
-	if (out->x1 >= out->x2)
-		return GL_FALSE;
-	if (out->y1 >= out->y2)
-		return GL_FALSE;
-	return GL_TRUE;
-}
-
-void radeonRecalcScissorRects(radeonContextPtr radeon)
-{
-	drm_clip_rect_t *out;
-	int i;
-
-	/* Grow cliprect store?
-	 */
-	if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) {
-		while (radeon->state.scissor.numAllocedClipRects <
-		       radeon->numClipRects) {
-			radeon->state.scissor.numAllocedClipRects += 1;	/* zero case */
-			radeon->state.scissor.numAllocedClipRects *= 2;
-		}
-
-		if (radeon->state.scissor.pClipRects)
-			FREE(radeon->state.scissor.pClipRects);
-
-		radeon->state.scissor.pClipRects =
-		    MALLOC(radeon->state.scissor.numAllocedClipRects *
-			   sizeof(drm_clip_rect_t));
-
-		if (radeon->state.scissor.pClipRects == NULL) {
-			radeon->state.scissor.numAllocedClipRects = 0;
-			return;
-		}
-	}
-
-	out = radeon->state.scissor.pClipRects;
-	radeon->state.scissor.numClipRects = 0;
-
-	for (i = 0; i < radeon->numClipRects; i++) {
-		if (intersect_rect(out,
-				   &radeon->pClipRects[i],
-				   &radeon->state.scissor.rect)) {
-			radeon->state.scissor.numClipRects++;
-			out++;
-		}
-	}
-}
-
-void radeonUpdateScissor(GLcontext* ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	if (radeon->dri.drawable) {
-		__DRIdrawablePrivate *dPriv = radeon->dri.drawable;
-		int x1 = dPriv->x + ctx->Scissor.X;
-		int y1 = dPriv->y + dPriv->h - (ctx->Scissor.Y + ctx->Scissor.Height);
-
-		radeon->state.scissor.rect.x1 = x1;
-		radeon->state.scissor.rect.y1 = y1;
-		radeon->state.scissor.rect.x2 = x1 + ctx->Scissor.Width;
-		radeon->state.scissor.rect.y2 = y1 + ctx->Scissor.Height;
-
-		radeonRecalcScissorRects(radeon);
-	}
-}
-
-static void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h)
-{
-	if (ctx->Scissor.Enabled) {
-		/* We don't pipeline cliprect changes */
-		r300Flush(ctx);
-		radeonUpdateScissor(ctx);
-	}
-}
-
-
-/**
- * Update cliprects and scissors.
- */
-void radeonSetCliprects(radeonContextPtr radeon)
-{
-	__DRIdrawablePrivate *const drawable = radeon->dri.drawable;
-	__DRIdrawablePrivate *const readable = radeon->dri.readable;
-	GLframebuffer *const draw_fb = (GLframebuffer*)drawable->driverPrivate;
-	GLframebuffer *const read_fb = (GLframebuffer*)readable->driverPrivate;
-
-	if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-		/* Can't ignore 2d windows if we are page flipping. */
-		if (drawable->numBackClipRects == 0 || radeon->doPageFlip ||
-		    radeon->sarea->pfCurrentPage == 1) {
-			radeon->numClipRects = drawable->numClipRects;
-			radeon->pClipRects = drawable->pClipRects;
-		} else {
-			radeon->numClipRects = drawable->numBackClipRects;
-			radeon->pClipRects = drawable->pBackClipRects;
-		}
-	} else {
-		/* front buffer (or none, or multiple buffers */
-		radeon->numClipRects = drawable->numClipRects;
-		radeon->pClipRects = drawable->pClipRects;
-	}
-
-	if ((draw_fb->Width != drawable->w) ||
-	    (draw_fb->Height != drawable->h)) {
-		_mesa_resize_framebuffer(radeon->glCtx, draw_fb,
-					 drawable->w, drawable->h);
-		draw_fb->Initialized = GL_TRUE;
-	}
-
-	if (drawable != readable) {
-		if ((read_fb->Width != readable->w) ||
-		    (read_fb->Height != readable->h)) {
-			_mesa_resize_framebuffer(radeon->glCtx, read_fb,
-						 readable->w, readable->h);
-			read_fb->Initialized = GL_TRUE;
-		}
-	}
-
-	if (radeon->state.scissor.enabled)
-		radeonRecalcScissorRects(radeon);
-
-	radeon->lastStamp = drawable->lastStamp;
-}
-
-
-/**
- * Handle common enable bits.
- * Called as a fallback by r200Enable/r300Enable.
- */
-void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	switch(cap) {
-	case GL_SCISSOR_TEST:
-		/* We don't pipeline cliprect & scissor changes */
-		r300Flush(ctx);
-
-		radeon->state.scissor.enabled = state;
-		radeonUpdateScissor(ctx);
-		break;
-
-	default:
-		return;
-	}
-}
-
-
-/**
- * Initialize default state.
- * This function is called once at context init time from
- * r200InitState/r300InitState
- */
-void radeonInitState(radeonContextPtr radeon)
-{
-	radeon->Fallback = 0;
-
-	if (radeon->glCtx->Visual.doubleBufferMode && radeon->sarea->pfCurrentPage == 0) {
-		radeon->state.color.drawOffset = radeon->radeonScreen->backOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->backPitch;
-	} else {
-		radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch;
-	}
-}
-
-
-/**
- * Initialize common state functions.
- * Called by r200InitStateFuncs/r300InitStateFuncs
- */
-void radeonInitStateFuncs(struct dd_function_table *functions)
-{
-	functions->Scissor = radeonScissor;
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_texture.c b/src/mesa/drivers/dri/r300/radeon_texture.c
new file mode 120000
index 0000000000..a822710915
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_texture.c
@@ -0,0 +1 @@
+../radeon/radeon_texture.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_texture.h b/src/mesa/drivers/dri/r300/radeon_texture.h
new file mode 120000
index 0000000000..17fac3d5ea
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_texture.h
@@ -0,0 +1 @@
+../radeon/radeon_texture.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon.h b/src/mesa/drivers/dri/r300/server/radeon.h
new file mode 120000
index 0000000000..81274a54f1
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon.h
@@ -0,0 +1 @@
+../../radeon/server/radeon.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon_dri.c b/src/mesa/drivers/dri/r300/server/radeon_dri.c
new file mode 120000
index 0000000000..d05847d650
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon_dri.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon_dri.h b/src/mesa/drivers/dri/r300/server/radeon_dri.h
new file mode 120000
index 0000000000..27c591d3c9
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon_dri.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon_egl.c b/src/mesa/drivers/dri/r300/server/radeon_egl.c
new file mode 120000
index 0000000000..d7735a7643
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon_egl.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_egl.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon_macros.h b/src/mesa/drivers/dri/r300/server/radeon_macros.h
new file mode 120000
index 0000000000..c56cd735b8
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon_macros.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_macros.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/server/radeon_reg.h b/src/mesa/drivers/dri/r300/server/radeon_reg.h
new file mode 120000
index 0000000000..e2349dcb68
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/server/radeon_reg.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_reg.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/Lindent b/src/mesa/drivers/dri/r600/Lindent
new file mode 100755
index 0000000000..7d8d8896e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/Lindent
@@ -0,0 +1,2 @@
+#!/bin/sh
+indent -npro -kr -i8 -ts8 -sob -l80 -ss -ncs "$@"
diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile
new file mode 100644
index 0000000000..d925a2dfe3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/Makefile
@@ -0,0 +1,79 @@
+# src/mesa/drivers/dri/r300/Makefile
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+CFLAGS += $(RADEON_CFLAGS)
+
+LIBNAME = r600_dri.so
+
+MINIGLX_SOURCES = server/radeon_dri.c
+
+ifeq ($(USING_EGL), 1)
+EGL_SOURCES = server/radeon_egl.c
+endif
+
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+COMMON_SOURCES = \
+	../../common/driverfuncs.c \
+	../common/mm.c \
+	../common/utils.c \
+	../common/texmem.c \
+	../common/vblank.c \
+	../common/xmlconfig.c \
+	../common/dri_util.c
+
+RADEON_COMMON_SOURCES = \
+	radeon_bo_legacy.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_cs_legacy.c \
+	radeon_dma.c \
+	radeon_debug.c \
+	radeon_fbo.c \
+	radeon_lock.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_texture.c \
+	radeon_queryobj.c
+
+DRIVER_SOURCES = \
+		 radeon_screen.c \
+		 r600_context.c \
+		 r600_cmdbuf.c \
+		 r600_emit.c       \
+		 r700_assembler.c  \
+		 r700_fragprog.c \
+		 r700_vertprog.c \
+		 r700_shader.c \
+		 r700_shaderinst.c \
+		 r700_ioctl.c \
+		 r700_oglprog.c \
+		 r700_chip.c     \
+		 r700_state.c    \
+		 r700_clear.c    \
+		 r700_render.c   \
+		 r600_tex.c      \
+		 r600_texstate.c      \
+		 r700_debug.c    \
+		 $(RADEON_COMMON_SOURCES) \
+		 $(EGL_SOURCES) \
+		 $(CS_SOURCES)
+
+C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
+
+DRIVER_DEFINES = -DCOMPILE_R600 -DR200_MERGED=0 \
+	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R600 \
+#	-DRADEON_BO_TRACK \
+	-Wall
+
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
+
+##### TARGETS #####
+
+include ../Makefile.template
+
+symlinks:
diff --git a/src/mesa/drivers/dri/r600/defaultendian.h b/src/mesa/drivers/dri/r600/defaultendian.h
new file mode 100644
index 0000000000..32caf32cd2
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/defaultendian.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _DEFINEENDIAN_H_
+#define _DEFINEENDIAN_H_
+
+//We have to choose a reg bits orientation if there is no compile flag for it.
+#if defined(LITTLEENDIAN_CPU)
+#elif defined(BIGENDIAN_CPU)
+#else
+#define LITTLEENDIAN_CPU
+#endif
+
+#endif //_DEFINEENDIAN_H_
diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c
new file mode 100644
index 0000000000..3cfe03a45f
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c
@@ -0,0 +1,514 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * Mostly coppied from \radeon\radeon_cs_legacy.c
+ */
+
+#include <errno.h>
+
+#include "main/glheader.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/simple_list.h"
+#include "swrast/swrast.h"
+
+#include "drm.h"
+#include "radeon_drm.h"
+
+#include "r600_context.h"
+#include "radeon_reg.h"
+#include "r600_cmdbuf.h"
+#include "r600_emit.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_mipmap_tree.h"
+#include "radeon_reg.h"
+
+
+
+static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm,
+                                   uint32_t ndw)
+{
+    struct radeon_cs *cs;
+
+    cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs));
+    if (cs == NULL) {
+        return NULL;
+    }
+    cs->csm = csm;
+    cs->ndw = (ndw + 0x3FF) & (~0x3FF);
+    cs->packets = (uint32_t*)malloc(4*cs->ndw);
+    if (cs->packets == NULL) {
+        free(cs);
+        return NULL;
+    }
+    cs->relocs_total_size = 0;
+    return cs;
+}
+
+static int r600_cs_write_reloc(struct radeon_cs *cs,
+			       struct radeon_bo *bo,
+			       uint32_t read_domain,
+			       uint32_t write_domain,
+			       uint32_t flags)
+{
+    struct r600_cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+    /* check domains */
+    if ((read_domain && write_domain) || (!read_domain && !write_domain)) {
+        /* in one CS a bo can only be in read or write domain but not
+         * in read & write domain at the same sime
+         */
+        return -EINVAL;
+    }
+    if (read_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    if (write_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    /* check if bo is already referenced */
+    for(i = 0; i < cs->crelocs; i++) {
+        uint32_t *indices;
+        uint32_t *reloc_indices;
+
+        if (relocs[i].base.bo->handle == bo->handle) {
+            /* Check domains must be in read or write. As we check already
+             * checked that in argument one of the read or write domain was
+             * set we only need to check that if previous reloc as the read
+             * domain set then the read_domain should also be set for this
+             * new relocation.
+             */
+            if (relocs[i].base.read_domain && !read_domain) {
+                return -EINVAL;
+            }
+            if (relocs[i].base.write_domain && !write_domain) {
+                return -EINVAL;
+            }
+            relocs[i].base.read_domain |= read_domain;
+            relocs[i].base.write_domain |= write_domain;
+            /* save indice */
+            relocs[i].cindices++;
+            indices = (uint32_t*)realloc(relocs[i].indices,
+                                         relocs[i].cindices * 4);
+            reloc_indices = (uint32_t*)realloc(relocs[i].reloc_indices,
+                                               relocs[i].cindices * 4);
+            if ( (indices == NULL) || (reloc_indices == NULL) ) {
+                relocs[i].cindices -= 1;
+                return -ENOMEM;
+            }
+            relocs[i].indices = indices;
+            relocs[i].reloc_indices = reloc_indices;
+            relocs[i].indices[relocs[i].cindices - 1] = cs->cdw;
+            relocs[i].reloc_indices[relocs[i].cindices - 1] = cs->cdw;
+            cs->section_cdw += 2;
+	    cs->cdw += 2;
+
+            return 0;
+        }
+    }
+    /* add bo to reloc */
+    relocs = (struct r600_cs_reloc_legacy*)
+             realloc(cs->relocs,
+                     sizeof(struct r600_cs_reloc_legacy) * (cs->crelocs + 1));
+    if (relocs == NULL) {
+        return -ENOMEM;
+    }
+    cs->relocs = relocs;
+    relocs[cs->crelocs].base.bo = bo;
+    relocs[cs->crelocs].base.read_domain = read_domain;
+    relocs[cs->crelocs].base.write_domain = write_domain;
+    relocs[cs->crelocs].base.flags = flags;
+    relocs[cs->crelocs].indices = (uint32_t*)malloc(4);
+    relocs[cs->crelocs].reloc_indices = (uint32_t*)malloc(4);
+    if ( (relocs[cs->crelocs].indices == NULL) || (relocs[cs->crelocs].reloc_indices == NULL) )
+    {
+        return -ENOMEM;
+    }
+
+    relocs[cs->crelocs].indices[0] = cs->cdw;
+    relocs[cs->crelocs].reloc_indices[0] = cs->cdw;
+    cs->section_cdw += 2;
+    cs->cdw += 2;
+    relocs[cs->crelocs].cindices = 1;
+    cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo);
+    cs->crelocs++;
+
+    radeon_bo_ref(bo);
+
+    return 0;
+}
+
+static int r600_cs_begin(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line)
+{
+    if (cs->section) {
+        fprintf(stderr, "CS already in a section(%s,%s,%d)\n",
+                cs->section_file, cs->section_func, cs->section_line);
+        fprintf(stderr, "CS can't start section(%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+
+    cs->section = 1;
+    cs->section_ndw = ndw;
+    cs->section_cdw = 0;
+    cs->section_file = file;
+    cs->section_func = func;
+    cs->section_line = line;
+
+    if (cs->cdw + ndw > cs->ndw) {
+        uint32_t tmp, *ptr;
+	int num = (ndw > 0x400) ? ndw : 0x400;
+
+        tmp = (cs->cdw + num + 0x3FF) & (~0x3FF);
+        ptr = (uint32_t*)realloc(cs->packets, 4 * tmp);
+        if (ptr == NULL) {
+            return -ENOMEM;
+        }
+        cs->packets = ptr;
+        cs->ndw = tmp;
+    }
+
+    return 0;
+}
+
+static int r600_cs_end(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line)
+
+{
+    if (!cs->section) {
+        fprintf(stderr, "CS no section to end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 0;
+
+    if ( cs->section_ndw != cs->section_cdw ) {
+        fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n",
+                cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw);
+        fprintf(stderr, "cs->section_ndw = %d, cs->cdw = %d, cs->section_cdw = %d \n",
+                cs->section_ndw, cs->cdw, cs->section_cdw);
+        fprintf(stderr, "CS section end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+
+    if (cs->cdw > cs->ndw) {
+	    fprintf(stderr, "CS section overflow at (%s,%s,%d) cdw %d ndw %d\n",
+		    cs->section_file, cs->section_func, cs->section_line,cs->cdw,cs->ndw);
+	    fprintf(stderr, "CS section end at (%s,%s,%d)\n",
+		    file, func, line);
+	    assert(0);
+    }
+
+    return 0;
+}
+
+static int r600_cs_process_relocs(struct radeon_cs *cs, 
+                                  uint32_t * reloc_chunk,
+                                  uint32_t * length_dw_reloc_chunk) 
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct r600_cs_reloc_legacy *relocs;
+    int i, j, r;
+
+    uint32_t offset_dw = 0;
+
+    csm = (struct r600_cs_manager_legacy*)cs->csm;
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+restart:
+    for (i = 0; i < cs->crelocs; i++) {
+            uint32_t soffset, eoffset, asicoffset;
+
+            r = radeon_bo_legacy_validate(relocs[i].base.bo,
+					  &soffset, &eoffset);
+	    if (r == -EAGAIN) {
+		    goto restart;
+            }
+            if (r) {
+		    fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+			    relocs[i].base.bo, soffset, eoffset);
+		    return r;
+            }
+            asicoffset = soffset;
+
+	    for (j = 0; j < relocs[i].cindices; j++) {
+		    if (asicoffset >= eoffset) {
+			    /*                radeon_bo_debug(relocs[i].base.bo, 12); */
+			    fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+				    relocs[i].base.bo, soffset, eoffset);
+			    fprintf(stderr, "above end: %p 0x%08X 0x%08X\n",
+				    relocs[i].base.bo,
+				    cs->packets[relocs[i].indices[j]],
+				    eoffset);
+			    exit(0);
+			    return -EINVAL;
+		    }
+		    /* pkt3 nop header in ib chunk */
+		    cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000;
+		    /* reloc index in ib chunk */
+		    cs->packets[relocs[i].reloc_indices[j] + 1] = offset_dw;
+	    }
+
+	    /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */
+	    reloc_chunk[offset_dw] = asicoffset;
+	    reloc_chunk[offset_dw + 3] = 0;
+
+	    offset_dw += 4;
+    }
+
+    *length_dw_reloc_chunk = offset_dw;
+
+    return 0;
+}
+
+static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct r600_cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+    for (i = 0; i < cs->crelocs; i++) {
+        radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age);
+        radeon_bo_unref(relocs[i].base.bo);
+    }
+    return 0;
+}
+
+#if 0
+static void dump_cmdbuf(struct radeon_cs *cs)
+{
+	int i;
+	fprintf(stderr,"--start--\n");
+	for (i = 0; i < cs->cdw; i++){
+		fprintf(stderr,"0x%08x\n", cs->packets[i]);
+	}
+	fprintf(stderr,"--end--\n");
+
+}
+#endif
+
+static int r600_cs_emit(struct radeon_cs *cs)
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct drm_radeon_cs       cs_cmd;
+    struct drm_radeon_cs_chunk cs_chunk[2];
+    uint32_t length_dw_reloc_chunk;
+    uint64_t chunk_ptrs[2];
+    uint32_t *reloc_chunk;
+    int r;
+    int retry = 0;
+
+    /* TODO : put chip level things here if need. */
+    /* csm->ctx->vtbl.emit_cs_header(cs, csm->ctx); */
+
+    csm->pending_count = 1;
+
+    reloc_chunk = (uint32_t*)calloc(1, cs->crelocs * 4 * 4);
+
+    r = r600_cs_process_relocs(cs, reloc_chunk, &length_dw_reloc_chunk);
+    if (r) {
+	free(reloc_chunk);
+        return 0;
+    }
+
+    /* raw ib chunk */
+    cs_chunk[0].chunk_id   = RADEON_CHUNK_ID_IB;
+    cs_chunk[0].length_dw  = cs->cdw;
+    cs_chunk[0].chunk_data = (unsigned long)(cs->packets);
+
+    /* reloc chaunk */
+    cs_chunk[1].chunk_id   = RADEON_CHUNK_ID_RELOCS;
+    cs_chunk[1].length_dw  = length_dw_reloc_chunk;
+    cs_chunk[1].chunk_data = (unsigned long)reloc_chunk;
+
+    chunk_ptrs[0] = (uint64_t)(unsigned long)&(cs_chunk[0]);
+    chunk_ptrs[1] = (uint64_t)(unsigned long)&(cs_chunk[1]);
+
+    cs_cmd.num_chunks = 2;
+    /* cs_cmd.cs_id      = 0; */
+    cs_cmd.chunks     = (uint64_t)(unsigned long)chunk_ptrs;
+
+    //dump_cmdbuf(cs);
+
+    do 
+    {
+        r = drmCommandWriteRead(cs->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd));
+        retry++;
+    } while (r == -EAGAIN && retry < 1000);
+
+    if (r) {
+	free(reloc_chunk);
+        return r;
+    }
+
+    csm->pending_age = cs_cmd.cs_id;
+
+    r600_cs_set_age(cs);
+
+    cs->csm->read_used = 0;
+    cs->csm->vram_write_used = 0;
+    cs->csm->gart_write_used = 0;
+
+    free(reloc_chunk);
+
+    return 0;
+}
+
+static void inline r600_cs_free_reloc(void *relocs_p, int crelocs)
+{
+    struct r600_cs_reloc_legacy *relocs = relocs_p;
+    int i;
+    if (!relocs_p)
+      return;
+    for (i = 0; i < crelocs; i++)
+    {
+        free(relocs[i].indices);
+        free(relocs[i].reloc_indices);
+    }
+}
+
+static int r600_cs_destroy(struct radeon_cs *cs)
+{
+    r600_cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    free(cs->packets);
+    free(cs);
+    return 0;
+}
+
+static int r600_cs_erase(struct radeon_cs *cs)
+{
+    r600_cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    cs->relocs_total_size = 0;
+    cs->relocs = NULL;
+    cs->crelocs = 0;
+    cs->cdw = 0;
+    cs->section = 0;
+    return 0;
+}
+
+static int r600_cs_need_flush(struct radeon_cs *cs)
+{
+    /* this function used to flush when the BO usage got to
+     * a certain size, now the higher levels handle this better */
+    return 0;
+}
+
+static void r600_cs_print(struct radeon_cs *cs, FILE *file)
+{
+}
+
+static struct radeon_cs_funcs  r600_cs_funcs = {
+    r600_cs_create,
+    r600_cs_write_reloc,
+    r600_cs_begin,
+    r600_cs_end,
+    r600_cs_emit,
+    r600_cs_destroy,
+    r600_cs_erase,
+    r600_cs_need_flush,
+    r600_cs_print
+};
+
+struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx)
+{
+    struct r600_cs_manager_legacy *csm;
+
+    csm = (struct r600_cs_manager_legacy*)
+          calloc(1, sizeof(struct r600_cs_manager_legacy));
+    if (csm == NULL) {
+        return NULL;
+    }
+    csm->base.funcs = &r600_cs_funcs;
+    csm->base.fd = ctx->dri.fd;
+    csm->ctx = ctx;
+    csm->pending_age = 1;
+    return (struct radeon_cs_manager*)csm;
+}
+
+void r600InitCmdBuf(context_t *r600) /* from rcommonInitCmdBuf */
+{
+	radeonContextPtr rmesa = &r600->radeon;
+	GLuint size;
+
+	r600InitAtoms(r600);
+
+	/* Initialize command buffer */
+	size = 256 * driQueryOptioni(&rmesa->optionCache,
+				     "command_buffer_size");
+	if (size < 2 * rmesa->hw.max_state_size) {
+		size = 2 * rmesa->hw.max_state_size + 65535;
+	}
+	if (size > 64 * 256)
+		size = 64 * 256;
+
+	if (rmesa->radeonScreen->kernel_mm) {
+		int fd = rmesa->radeonScreen->driScreen->fd;
+		rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd);
+	} else {
+		rmesa->cmdbuf.csm = r600_radeon_cs_manager_legacy_ctor(rmesa);
+	}
+	if (rmesa->cmdbuf.csm == NULL) {
+		/* FIXME: fatal error */
+		return;
+	}
+	rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size);
+	assert(rmesa->cmdbuf.cs != NULL);
+	rmesa->cmdbuf.size = size;
+
+	radeon_cs_space_set_flush(rmesa->cmdbuf.cs,
+				  (void (*)(void *))rmesa->glCtx->Driver.Flush, rmesa->glCtx);
+
+	if (!rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]);
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size);
+	} else {
+		struct drm_radeon_gem_info mminfo;
+
+		if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo)))
+		{
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_visible);
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size);
+		}
+	}
+}
+
diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h
new file mode 100644
index 0000000000..eba43d37b6
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h
@@ -0,0 +1,212 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#ifndef __R600_CMDBUF_H__
+#define __R600_CMDBUF_H__
+
+#include "r600_context.h"
+#include "r600_emit.h"
+
+#define RADEON_CP_PACKET3_NOP                       0xC0001000
+#define RADEON_CP_PACKET3_NEXT_CHAR                 0xC0001900
+#define RADEON_CP_PACKET3_PLY_NEXTSCAN              0xC0001D00
+#define RADEON_CP_PACKET3_SET_SCISSORS              0xC0001E00
+#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM     0xC0002300
+#define RADEON_CP_PACKET3_LOAD_MICROCODE            0xC0002400
+#define RADEON_CP_PACKET3_WAIT_FOR_IDLE             0xC0002600
+#define RADEON_CP_PACKET3_3D_DRAW_VBUF              0xC0002800
+#define RADEON_CP_PACKET3_3D_DRAW_IMMD              0xC0002900
+#define RADEON_CP_PACKET3_3D_DRAW_INDX              0xC0002A00
+#define RADEON_CP_PACKET3_LOAD_PALETTE              0xC0002C00
+#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR            0xC0002F00
+#define RADEON_CP_PACKET3_CNTL_PAINT                0xC0009100
+#define RADEON_CP_PACKET3_CNTL_BITBLT               0xC0009200
+#define RADEON_CP_PACKET3_CNTL_SMALLTEXT            0xC0009300
+#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT         0xC0009400
+#define RADEON_CP_PACKET3_CNTL_POLYLINE             0xC0009500
+#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES        0xC0009800
+#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI          0xC0009A00
+#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI         0xC0009B00
+#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT         0xC0009C00
+
+/* r6xx/r7xx packet 3 type offsets */
+#define R600_SET_CONFIG_REG_OFFSET                  0x00008000
+#define R600_SET_CONFIG_REG_END                     0x0000ac00
+#define R600_SET_CONTEXT_REG_OFFSET                 0x00028000
+#define R600_SET_CONTEXT_REG_END                    0x00029000
+#define R600_SET_ALU_CONST_OFFSET                   0x00030000
+#define R600_SET_ALU_CONST_END                      0x00032000
+#define R600_SET_RESOURCE_OFFSET                    0x00038000
+#define R600_SET_RESOURCE_END                       0x0003c000
+#define R600_SET_SAMPLER_OFFSET                     0x0003c000
+#define R600_SET_SAMPLER_END                        0x0003cff0
+#define R600_SET_CTL_CONST_OFFSET                   0x0003cff0
+#define R600_SET_CTL_CONST_END                      0x0003e200
+#define R600_SET_LOOP_CONST_OFFSET                  0x0003e200
+#define R600_SET_LOOP_CONST_END                     0x0003e380
+#define R600_SET_BOOL_CONST_OFFSET                  0x0003e380
+#define R600_SET_BOOL_CONST_END                     0x00040000
+
+/* r6xx/r7xx packet 3 types */
+#define R600_IT_INDIRECT_BUFFER_END               0x00001700
+#define R600_IT_SET_PREDICATION                   0x00002000
+#define R600_IT_REG_RMW                           0x00002100
+#define R600_IT_COND_EXEC                         0x00002200
+#define R600_IT_PRED_EXEC                         0x00002300
+#define R600_IT_START_3D_CMDBUF                   0x00002400
+#define R600_IT_DRAW_INDEX_2                      0x00002700
+#define R600_IT_CONTEXT_CONTROL                   0x00002800
+#define R600_IT_DRAW_INDEX_IMMD_BE                0x00002900
+#define R600_IT_INDEX_TYPE                        0x00002A00
+#define R600_IT_DRAW_INDEX                        0x00002B00
+#define R600_IT_DRAW_INDEX_AUTO                   0x00002D00
+#define R600_IT_DRAW_INDEX_IMMD                   0x00002E00
+#define R600_IT_NUM_INSTANCES                     0x00002F00
+#define R600_IT_STRMOUT_BUFFER_UPDATE             0x00003400
+#define R600_IT_INDIRECT_BUFFER_MP                0x00003800
+#define R600_IT_MEM_SEMAPHORE                     0x00003900
+#define R600_IT_MPEG_INDEX                        0x00003A00
+#define R600_IT_WAIT_REG_MEM                      0x00003C00
+#define R600_IT_MEM_WRITE                         0x00003D00
+#define R600_IT_INDIRECT_BUFFER                   0x00003200
+#define R600_IT_CP_INTERRUPT                      0x00004000
+#define R600_IT_SURFACE_SYNC                      0x00004300
+#define R600_IT_ME_INITIALIZE                     0x00004400
+#define R600_IT_COND_WRITE                        0x00004500
+#define R600_IT_EVENT_WRITE                       0x00004600
+#define R600_IT_EVENT_WRITE_EOP                   0x00004700
+#define R600_IT_ONE_REG_WRITE                     0x00005700
+#define R600_IT_SET_CONFIG_REG                    0x00006800
+#define R600_IT_SET_CONTEXT_REG                   0x00006900
+#define R600_IT_SET_ALU_CONST                     0x00006A00
+#define R600_IT_SET_BOOL_CONST                    0x00006B00
+#define R600_IT_SET_LOOP_CONST                    0x00006C00
+#define R600_IT_SET_RESOURCE                      0x00006D00
+#define R600_IT_SET_SAMPLER                       0x00006E00
+#define R600_IT_SET_CTL_CONST                     0x00006F00
+#define R600_IT_SURFACE_BASE_UPDATE               0x00007300
+
+struct r600_cs_manager_legacy
+{
+    struct radeon_cs_manager    base;
+    struct radeon_context       *ctx;
+    /* hack for scratch stuff */
+    uint32_t                    pending_age;
+    uint32_t                    pending_count;
+};
+
+struct r600_cs_reloc_legacy {
+    struct radeon_cs_reloc  base;
+    uint32_t                cindices;
+    uint32_t                *indices;
+    uint32_t                *reloc_indices;
+};
+
+struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx);
+
+/**
+ * Write one dword to the command buffer.
+ */
+#define R600_OUT_BATCH(data)				\
+do {							\
+        radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);	\
+} while(0)
+
+/**
+ * Write n dwords from ptr to the command buffer.
+ */
+#define R600_OUT_BATCH_TABLE(ptr,n)		\
+do {						     \
+	radeon_cs_write_table(b_l_rmesa->cmdbuf.cs, ptr, n);	\
+} while(0)
+
+/**
+ * Write a relocated dword to the command buffer.
+ */
+#define R600_OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags) 	\
+	do { 							\
+        if (0 && offset) {					\
+            fprintf(stderr, "(%s:%s:%d) offset : %d\n",		\
+            __FILE__, __FUNCTION__, __LINE__, offset);		\
+        }							\
+        radeon_cs_write_reloc(b_l_rmesa->cmdbuf.cs, 		\
+                              bo, rd, wd, flags);		\
+	} while(0)
+
+/* R600/R700 */
+#define R600_OUT_BATCH_REGS(reg, num)					\
+do {								\
+	if ((reg) >= R600_SET_CONFIG_REG_OFFSET && (reg) < R600_SET_CONFIG_REG_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_CONTEXT_REG_OFFSET && (reg) < R600_SET_CONTEXT_REG_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_ALU_CONST_OFFSET && (reg) < R600_SET_ALU_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_RESOURCE_OFFSET && (reg) < R600_SET_RESOURCE_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_SAMPLER_OFFSET && (reg) < R600_SET_SAMPLER_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_CTL_CONST_OFFSET && (reg) < R600_SET_CTL_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_LOOP_CONST_OFFSET && (reg) < R600_SET_LOOP_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_BOOL_CONST_OFFSET && (reg) < R600_SET_BOOL_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2);	\
+	} else {							\
+		R600_OUT_BATCH(CP_PACKET0((reg), (num))); \
+	}								\
+} while (0)
+
+/** Single register write to command buffer; requires 3 dwords for most things. */
+#define R600_OUT_BATCH_REGVAL(reg, val)		\
+	R600_OUT_BATCH_REGS((reg), 1);		\
+	R600_OUT_BATCH((val))
+
+/** Continuous register range write to command buffer; requires 1 dword,
+ * expects count dwords afterwards for register contents. */
+#define R600_OUT_BATCH_REGSEQ(reg, count)	\
+	R600_OUT_BATCH_REGS((reg), (count))
+
+extern void r600InitCmdBuf(context_t *r600);
+
+#endif				/* __R600_CMDBUF_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
new file mode 100644
index 0000000000..251c124cbf
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -0,0 +1,393 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ *
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#include "main/glheader.h"
+#include "main/api_arrayelt.h"
+#include "main/context.h"
+#include "main/simple_list.h"
+#include "main/imports.h"
+#include "main/matrix.h"
+#include "main/extensions.h"
+#include "main/state.h"
+#include "main/bufferobj.h"
+#include "main/texobj.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "vbo/vbo.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_vp_build.h"
+
+#include "drivers/common/driverfuncs.h"
+
+#include "radeon_debug.h"
+#include "r600_context.h"
+#include "radeon_common_context.h"
+#include "radeon_span.h"
+#include "r600_cmdbuf.h"
+#include "r600_emit.h"
+#include "radeon_bocs_wrapper.h"
+
+#include "r700_state.h"
+#include "r700_ioctl.h"
+
+
+#include "vblank.h"
+#include "utils.h"
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+
+/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
+int future_hw_tcl_on = 1;
+int hw_tcl_on = 1;
+
+#define need_GL_VERSION_2_0
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_vertex_program
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_gpu_program_parameters
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_stencil_two_side
+#define need_GL_ATI_separate_stencil
+#define need_GL_NV_vertex_program
+
+#include "extension_helper.h"
+
+extern const struct tnl_pipeline_stage *r700_pipeline[];
+
+const struct dri_extension card_extensions[] = {
+  /* *INDENT-OFF* */
+  {"GL_ARB_depth_texture",		NULL},
+  {"GL_ARB_fragment_program",		NULL},
+  {"GL_ARB_multitexture",		NULL},
+  {"GL_ARB_point_parameters",		GL_ARB_point_parameters_functions},
+  {"GL_ARB_shadow",			NULL},
+  {"GL_ARB_shadow_ambient",		NULL},
+  {"GL_ARB_texture_border_clamp",	NULL},
+  {"GL_ARB_texture_cube_map",		NULL},
+  {"GL_ARB_texture_env_add",		NULL},
+  {"GL_ARB_texture_env_combine",	NULL},
+  {"GL_ARB_texture_env_crossbar",	NULL},
+  {"GL_ARB_texture_env_dot3",		NULL},
+  {"GL_ARB_texture_mirrored_repeat",	NULL},
+  {"GL_ARB_vertex_program",		GL_ARB_vertex_program_functions},
+  {"GL_EXT_blend_equation_separate",	GL_EXT_blend_equation_separate_functions},
+  {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
+  {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
+  {"GL_EXT_blend_subtract",		NULL},
+  {"GL_EXT_packed_depth_stencil",	NULL},
+  {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
+  {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
+  {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
+  {"GL_EXT_shadow_funcs",		NULL},
+  {"GL_EXT_stencil_two_side",		GL_EXT_stencil_two_side_functions},
+  {"GL_EXT_stencil_wrap",		NULL},
+  {"GL_EXT_texture_edge_clamp",		NULL},
+  {"GL_EXT_texture_env_combine", 	NULL},
+  {"GL_EXT_texture_env_dot3", 		NULL},
+  {"GL_EXT_texture_filter_anisotropic",	NULL},
+  {"GL_EXT_texture_lod_bias",		NULL},
+  {"GL_EXT_texture_mirror_clamp",	NULL},
+  {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
+  {"GL_ATI_texture_env_combine3",	NULL},
+  {"GL_ATI_texture_mirror_once",	NULL},
+  {"GL_MESA_pack_invert",		NULL},
+  {"GL_MESA_ycbcr_texture",		NULL},
+  {"GL_MESAX_texture_float",		NULL},
+  {"GL_NV_blend_square",		NULL},
+  {"GL_NV_vertex_program",		GL_NV_vertex_program_functions},
+  {"GL_SGIS_generate_mipmap",		NULL},
+  {NULL,				NULL}
+  /* *INDENT-ON* */
+};
+
+
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
+/**
+ * The GL 2.0 functions are needed to make display lists work with
+ * functions added by GL_ATI_separate_stencil.
+ */
+const struct dri_extension gl_20_extension[] = {
+  {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
+};
+
+
+static void r600RunPipeline(GLcontext * ctx)
+{
+    _mesa_lock_context_textures(ctx);
+
+    if (ctx->NewState)
+        _mesa_update_state_locked(ctx);
+    
+    _tnl_run_pipeline(ctx);
+    _mesa_unlock_context_textures(ctx);
+}
+
+static void r600_get_lock(radeonContextPtr rmesa)
+{
+	drm_radeon_sarea_t *sarea = rmesa->sarea;
+
+	if (sarea->ctx_owner != rmesa->dri.hwContext) {
+		sarea->ctx_owner = rmesa->dri.hwContext;
+		if (!rmesa->radeonScreen->kernel_mm)
+			radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
+	}
+}		  
+
+static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+    /* please flush pipe do all pending work */
+    /* to be enabled */
+}
+
+static void r600_vtbl_pre_emit_atoms(radeonContextPtr radeon)
+{
+	r700Start3D((context_t *)radeon);
+}
+
+static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	if (mode)
+		context->radeon.Fallback |= bit;
+	else
+		context->radeon.Fallback &= ~bit;
+}
+
+static void r600_init_vtbl(radeonContextPtr radeon)
+{
+	radeon->vtbl.get_lock = r600_get_lock;
+	radeon->vtbl.update_viewport_offset = r700UpdateViewportOffset;
+	radeon->vtbl.emit_cs_header = r600_vtbl_emit_cs_header;
+	radeon->vtbl.swtcl_flush = NULL;
+	radeon->vtbl.pre_emit_atoms = r600_vtbl_pre_emit_atoms;
+	radeon->vtbl.fallback = r600_fallback;
+}
+
+/* Create the device specific rendering context.
+ */
+GLboolean r600CreateContext(const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+	struct dd_function_table functions;
+	context_t *r600;
+	GLcontext *ctx;
+
+	assert(glVisual);
+	assert(driContextPriv);
+	assert(screen);
+
+	/* Allocate the R600 context */
+	r600 = (context_t*) CALLOC(sizeof(*r600));
+	if (!r600) {
+		radeon_error("Failed to allocate memory for context.\n");
+		return GL_FALSE;
+	}
+
+	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
+		hw_tcl_on = future_hw_tcl_on = 0;
+
+	r600_init_vtbl(&r600->radeon);
+	/* Parse configuration files.
+	 * Do this here so that initialMaxAnisotropy is set before we create
+	 * the default textures.
+	 */
+	driParseConfigFiles(&r600->radeon.optionCache, &screen->optionCache,
+			    screen->driScreen->myNum, "r600");
+
+	r600->radeon.initialMaxAnisotropy = driQueryOptionf(&r600->radeon.optionCache,
+						     "def_max_anisotropy");
+
+	/* Init default driver functions then plug in our R600-specific functions
+	 * (the texture functions are especially important)
+	 */
+	_mesa_init_driver_functions(&functions);
+
+	r700InitStateFuncs(&functions);
+	r600InitTextureFuncs(&functions);
+	r700InitShaderFuncs(&functions);
+	r700InitIoctlFuncs(&functions);
+
+	if (!radeonInitContext(&r600->radeon, &functions,
+			       glVisual, driContextPriv,
+			       sharedContextPrivate)) {
+		radeon_error("Initializing context failed.\n");
+		FREE(r600);
+		return GL_FALSE;
+	}
+
+	/* Init r600 context data */
+	/* Set the maximum texture size small enough that we can guarentee that
+	 * all texture units can bind a maximal texture and have them both in
+	 * texturable memory at once.
+	 */
+
+	ctx = r600->radeon.glCtx;
+
+	ctx->Const.MaxTextureImageUnits =
+	    driQueryOptioni(&r600->radeon.optionCache, "texture_image_units");
+	ctx->Const.MaxTextureCoordUnits =
+	    driQueryOptioni(&r600->radeon.optionCache, "texture_coord_units");
+	ctx->Const.MaxTextureUnits =
+	    MIN2(ctx->Const.MaxTextureImageUnits,
+		 ctx->Const.MaxTextureCoordUnits);
+	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+	ctx->Const.MaxTextureLodBias = 16.0;
+
+	ctx->Const.MaxTextureLevels = 13;
+	ctx->Const.MaxTextureRectSize = 4096;
+
+	ctx->Const.MinPointSize   = 0x0001 / 8.0;
+	ctx->Const.MinPointSizeAA = 0x0001 / 8.0;
+	ctx->Const.MaxPointSize   = 0xffff / 8.0;
+	ctx->Const.MaxPointSizeAA = 0xffff / 8.0;
+
+	ctx->Const.MinLineWidth   = 0x0001 / 8.0;
+	ctx->Const.MinLineWidthAA = 0x0001 / 8.0;
+	ctx->Const.MaxLineWidth   = 0xffff / 8.0;
+	ctx->Const.MaxLineWidthAA = 0xffff / 8.0;
+
+	/* Needs further modifications */
+#if 0
+	ctx->Const.MaxArrayLockSize =
+	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
+#endif
+
+	ctx->Const.MaxDrawBuffers = 1;
+
+	/* Initialize the software rasterizer and helper modules.
+	 */
+	_swrast_CreateContext(ctx);
+	_vbo_CreateContext(ctx);
+	_tnl_CreateContext(ctx);
+	_swsetup_CreateContext(ctx);
+	_swsetup_Wakeup(ctx);
+	_ae_create_context(ctx);
+
+	/* Install the customized pipeline:
+	 */
+	_tnl_destroy_pipeline(ctx);
+	_tnl_install_pipeline(ctx, r700_pipeline);
+
+	/* Try and keep materials and vertices separate:
+	 */
+/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
+
+	/* Configure swrast and TNL to match hardware characteristics:
+	 */
+	_swrast_allow_pixel_fog(ctx, GL_FALSE);
+	_swrast_allow_vertex_fog(ctx, GL_TRUE);
+	_tnl_allow_pixel_fog(ctx, GL_FALSE);
+	_tnl_allow_vertex_fog(ctx, GL_TRUE);
+
+	/* currently bogus data */
+	ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
+	ctx->Const.VertexProgram.MaxNativeInstructions =
+		VSF_MAX_FRAGMENT_LENGTH / 4;
+	ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
+	ctx->Const.VertexProgram.MaxTemps = 32;
+	ctx->Const.VertexProgram.MaxNativeTemps =
+		/*VSF_MAX_FRAGMENT_TEMPS */ 32;
+	ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
+	ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+
+	ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
+	ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+	ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
+	ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
+	ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
+	ctx->Const.FragmentProgram.MaxNativeInstructions =
+	    PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
+	ctx->Const.FragmentProgram.MaxNativeTexIndirections =
+	    PFS_MAX_TEX_INDIRECT;
+	ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;	/* and these are?? */
+	ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
+	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
+
+	radeon_init_debug();
+
+	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r600->radeon.radeonScreen->kernel_mm)
+	  driInitExtensions(ctx, mm_extensions, GL_FALSE);
+
+	if (driQueryOptionb
+	    (&r600->radeon.optionCache, "disable_stencil_two_side"))
+		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
+
+	if (r600->radeon.glCtx->Mesa_DXTn
+	    && !driQueryOptionb(&r600->radeon.optionCache, "disable_s3tc")) {
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+		_mesa_enable_extension(ctx, "GL_S3_s3tc");
+	} else
+	    if (driQueryOptionb(&r600->radeon.optionCache, "force_s3tc_enable"))
+	{
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+	}
+
+	radeon_fbo_init(&r600->radeon);
+   	radeonInitSpanFuncs( ctx );
+
+	r600InitCmdBuf(r600);
+
+	r700InitState(r600->radeon.glCtx);
+
+	TNL_CONTEXT(ctx)->Driver.RunPipeline = r600RunPipeline;
+
+	if (driQueryOptionb(&r600->radeon.optionCache, "no_rast")) {
+		radeon_warning("disabling 3D acceleration\n");
+#if R200_MERGED
+		FALLBACK(&r600->radeon, RADEON_FALLBACK_DISABLE, 1);
+#endif
+	}
+
+	return GL_TRUE;
+}
+
+
diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h
new file mode 100644
index 0000000000..8ae05a301c
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_context.h
@@ -0,0 +1,205 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#ifndef __R600_CONTEXT_H__
+#define __R600_CONTEXT_H__
+
+#include "tnl/t_vertex.h"
+#include "drm.h"
+#include "radeon_drm.h"
+#include "dri_util.h"
+#include "texmem.h"
+#include "radeon_common.h"
+
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+
+#include "r700_chip.h"
+#include "r600_tex.h"
+#include "r700_oglprog.h"
+
+struct r600_context;
+typedef struct r600_context context_t;
+
+#include "main/mm.h"
+
+/************ DMA BUFFERS **************/
+
+/* The blit width for texture uploads
+ */
+#define R600_BLIT_WIDTH_BYTES 1024
+#define R600_MAX_TEXTURE_UNITS 8
+
+struct r600_texture_state {
+	int tc_count;		/* number of incoming texture coordinates from VAP */
+};
+
+/* Perhaps more if we store programs in vmem? */
+/* drm_r600_cmd_header_t->vpu->count is unsigned char */
+#define VSF_MAX_FRAGMENT_LENGTH (255*4)
+
+/* Can be tested with colormat currently. */
+#define VSF_MAX_FRAGMENT_TEMPS (14)
+
+#define STATE_R600_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
+#define STATE_R600_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
+
+extern int hw_tcl_on;
+
+#define COLOR_IS_RGBA
+#define TAG(x) r600##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+#define PFS_MAX_ALU_INST	64
+#define PFS_MAX_TEX_INST	64
+#define PFS_MAX_TEX_INDIRECT 4
+#define PFS_NUM_TEMP_REGS	32
+#define PFS_NUM_CONST_REGS	16
+
+#define R600_MAX_AOS_ARRAYS		16
+
+#define REG_COORDS	0
+#define REG_COLOR0	1
+#define REG_TEX0	2
+
+#define R600_FALLBACK_NONE 0
+#define R600_FALLBACK_TCL 1
+#define R600_FALLBACK_RAST 2
+
+enum 
+{
+    NO_SHIFT    = 0,
+    LEFT_SHIFT  = 1,
+    RIGHT_SHIFT = 2,
+};
+
+struct r600_hw_state {
+	struct radeon_state_atom sq;
+	struct radeon_state_atom db;
+	struct radeon_state_atom stencil;
+	struct radeon_state_atom db_target;
+	struct radeon_state_atom sc;
+	struct radeon_state_atom scissor;
+	struct radeon_state_atom aa;
+	struct radeon_state_atom cl;
+	struct radeon_state_atom gb;
+	struct radeon_state_atom ucp;
+	struct radeon_state_atom su;
+	struct radeon_state_atom poly;
+	struct radeon_state_atom cb;
+	struct radeon_state_atom clrcmp;
+	struct radeon_state_atom blnd;
+	struct radeon_state_atom blnd_clr;
+	struct radeon_state_atom cb_target;
+	struct radeon_state_atom sx;
+	struct radeon_state_atom vgt;
+	struct radeon_state_atom spi;
+	struct radeon_state_atom vpt;
+
+	struct radeon_state_atom fs;
+	struct radeon_state_atom vs;
+	struct radeon_state_atom ps;
+
+	struct radeon_state_atom vs_consts;
+	struct radeon_state_atom ps_consts;
+
+	struct radeon_state_atom vtx;
+	struct radeon_state_atom tx;
+	struct radeon_state_atom tx_smplr;
+	struct radeon_state_atom tx_brdr_clr;
+};
+
+/**
+ * \brief R600 context structure.
+ */
+struct r600_context {
+	struct radeon_context radeon;	/* parent class, must be first */
+
+	/* ------ */
+	R700_CHIP_CONTEXT hw;
+
+	struct r600_hw_state atoms;
+
+	/* Vertex buffers
+	 */
+	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
+	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
+
+};
+
+#define R700_CONTEXT(ctx)		((context_t *)(ctx->DriverCtx))
+#define GL_CONTEXT(context)     ((GLcontext *)(context->radeon.glCtx))
+
+extern GLboolean r600CreateContext(const __GLcontextModes * glVisual,
+				   __DRIcontextPrivate * driContextPriv,
+				   void *sharedContextPrivate);
+
+#define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(&context->hw))
+
+#define R600_NEWPRIM( rmesa )			\
+do {						\
+	if ( rmesa->radeon.dma.flush )			\
+		rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
+} while (0)
+
+#define R600_STATECHANGE(r600, ATOM)			\
+do {							\
+	R600_NEWPRIM(r600);					\
+	r600->atoms.ATOM.dirty = GL_TRUE;					\
+	r600->radeon.hw.is_dirty = GL_TRUE;			\
+} while(0)
+
+extern GLboolean r700SyncSurf(context_t *context,
+			      struct radeon_bo *pbo,
+			      uint32_t read_domain,
+			      uint32_t write_domain,
+			      uint32_t sync_type);
+
+extern void r700SetupStreams(GLcontext * ctx);
+extern void r700Start3D(context_t *context);
+extern void r600InitAtoms(context_t *context);
+
+#define RADEON_D_CAPTURE 0
+#define RADEON_D_PLAYBACK 1
+#define RADEON_D_PLAYBACK_RAW 2
+#define RADEON_D_T 3
+
+#define r600PackFloat32 radeonPackFloat32
+#define r600PackFloat24 radeonPackFloat24
+
+#endif				/* __R600_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c
new file mode 100644
index 0000000000..5c250c2418
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_emit.c
@@ -0,0 +1,117 @@
+/**************************************************************************
+
+Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD)
+
+Copyright (C) Advanced Micro Devices Inc. (AMD)  2009.  All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/image.h"
+
+#include "swrast_setup/swrast_setup.h"
+#include "math/m_translate.h"
+#include "tnl/tnl.h"
+#include "tnl/t_context.h"
+
+#include "r600_context.h"
+#include "r600_emit.h"
+
+void r600EmitCacheFlush(context_t *rmesa)
+{
+}
+
+GLboolean r600EmitShader(GLcontext * ctx,
+                         void ** shaderbo,
+			 GLvoid * data,
+                         int sizeinDWORD,
+                         char * szShaderUsage)
+{
+	radeonContextPtr radeonctx = RADEON_CONTEXT(ctx);
+	struct radeon_bo * pbo;
+	uint32_t *out;
+shader_again_alloc:
+	pbo = radeon_bo_open(radeonctx->radeonScreen->bom,
+			0,
+			sizeinDWORD * 4,
+			256,
+			RADEON_GEM_DOMAIN_GTT,
+			0);
+
+	radeon_print(RADEON_SHADER, RADEON_NORMAL, "%s %p size %d: %s\n", __func__, pbo, sizeinDWORD, szShaderUsage);
+
+	if (!pbo) {
+		radeon_print(RADEON_MEMORY | RADEON_CS, RADEON_IMPORTANT, "No memory for buffer object. Flushing command buffer.\n");
+		rcommonFlushCmdBuf(radeonctx, __FUNCTION__);
+		goto shader_again_alloc;
+	}
+
+	radeon_cs_space_add_persistent_bo(radeonctx->cmdbuf.cs,
+			pbo,
+			RADEON_GEM_DOMAIN_GTT, 0);
+
+	if (radeon_cs_space_check_with_bo(radeonctx->cmdbuf.cs,
+				pbo,
+				RADEON_GEM_DOMAIN_GTT, 0)) {
+		radeon_error("failure to revalidate BOs - badness\n");
+		return GL_FALSE;
+	}
+
+	radeon_bo_map(pbo, 1);
+
+	out = (uint32_t*)(pbo->ptr);
+
+	memcpy(out, data, sizeinDWORD * 4);
+
+	radeon_bo_unmap(pbo);
+
+	*shaderbo = (void*)pbo;
+
+	return GL_TRUE;
+}
+
+GLboolean r600DeleteShader(GLcontext * ctx,
+                           void * shaderbo)
+{
+    struct radeon_bo * pbo = (struct radeon_bo *)shaderbo;
+
+    radeon_print(RADEON_SHADER, RADEON_NORMAL, "%s: %p\n", __func__, pbo);
+
+    if (pbo) {
+	    if (pbo->ptr)
+		radeon_bo_unmap(pbo);
+	    radeon_bo_unref(pbo); /* when bo->cref <= 0, bo will be bo_free */
+    }
+
+    return GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/r300/radeon_state.h b/src/mesa/drivers/dri/r600/r600_emit.h
index 821cb40c7e..661774d11e 100644
--- a/src/mesa/drivers/dri/r300/radeon_state.h
+++ b/src/mesa/drivers/dri/r600/r600_emit.h
@@ -1,5 +1,8 @@
-/*
-Copyright (C) 2004 Nicolai Haehnle.  All Rights Reserved.
+/**************************************************************************
+
+Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD)
+
+Copyright (C) Advanced Micro Devices Inc. (AMD)  2009.  All Rights Reserved.
 
 Permission is hereby granted, free of charge, to any person obtaining
 a copy of this software and associated documentation files (the
@@ -25,19 +28,28 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /*
  * Authors:
- *   Nicolai Haehnle <prefect_@gmx.net>
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
  */
 
-#ifndef __RADEON_STATE_H__
-#define __RADEON_STATE_H__
 
-extern void radeonRecalcScissorRects(radeonContextPtr radeon);
-extern void radeonSetCliprects(radeonContextPtr radeon);
-extern void radeonUpdateScissor(GLcontext* ctx);
+#ifndef __R600_EMIT_H__
+#define __R600_EMIT_H__
+
+#include "main/glheader.h"
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+#include "radeon_reg.h"
+
+void r600EmitCacheFlush(context_t *rmesa);
 
-extern void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state);
+extern GLboolean r600EmitShader(GLcontext * ctx, 
+                                void ** shaderbo,
+			                    GLvoid * data, 
+                                int sizeinDWORD,
+                                char * szShaderUsage); 
 
-extern void radeonInitState(radeonContextPtr radeon);
-extern void radeonInitStateFuncs(struct dd_function_table* functions);
+extern GLboolean r600DeleteShader(GLcontext * ctx, 
+                                 void * shaderbo);
 
 #endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg.h b/src/mesa/drivers/dri/r600/r600_reg.h
new file mode 100644
index 0000000000..ffe5ee4f74
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg.h
@@ -0,0 +1,121 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_H_
+#define _R600_REG_H_
+
+/*
+ * Register definitions
+ */
+
+#include "r600_reg_auto_r6xx.h"
+#include "r600_reg_r6xx.h"
+#include "r600_reg_r7xx.h"
+
+
+/* SET_*_REG offsets + ends */
+enum 
+{
+    SET_CONFIG_REG_offset                = 0x00008000,
+    SET_CONFIG_REG_end                   = 0x0000ac00,
+    SET_CONTEXT_REG_offset               = 0x00028000,
+    SET_CONTEXT_REG_end                  = 0x00029000,
+    SET_ALU_CONST_offset                 = 0x00030000,
+    SET_ALU_CONST_end                    = 0x00032000,
+    SET_RESOURCE_offset                  = 0x00038000,
+    SET_RESOURCE_end                     = 0x0003c000,
+    SET_SAMPLER_offset                   = 0x0003c000,
+    SET_SAMPLER_end                      = 0x0003cff0,
+    SET_CTL_CONST_offset                 = 0x0003cff0,
+    SET_CTL_CONST_end                    = 0x0003e200,
+    SET_LOOP_CONST_offset                = 0x0003e200,
+    SET_LOOP_CONST_end                   = 0x0003e380,
+    SET_BOOL_CONST_offset                = 0x0003e380,
+    SET_BOOL_CONST_end                   = 0x00040000,
+};
+
+/* packet3 IT_SURFACE_BASE_UPDATE bits */
+enum 
+{
+    DEPTH_BASE                           = (1 << 0),
+    COLOR0_BASE                          = (1 << 1),
+    COLOR1_BASE                          = (1 << 2),
+    COLOR2_BASE                          = (1 << 3),
+    COLOR3_BASE                          = (1 << 4),
+    COLOR4_BASE                          = (1 << 5),
+    COLOR5_BASE                          = (1 << 6),
+    COLOR6_BASE                          = (1 << 7),
+    COLOR7_BASE                          = (1 << 8),
+    STRMOUT_BASE0                        = (1 << 9),
+    STRMOUT_BASE1                        = (1 << 10),
+    STRMOUT_BASE2                        = (1 << 11),
+    STRMOUT_BASE3                        = (1 << 12),
+    COHER_BASE0                          = (1 << 13),
+    COHER_BASE1                          = (1 << 14),
+};
+
+/* Packet3 commands */
+enum 
+{
+    IT_NOP                               = 0x10,
+    IT_INDIRECT_BUFFER_END               = 0x17,
+    IT_SET_PREDICATION                   = 0x20,
+    IT_REG_RMW                           = 0x21,
+    IT_COND_EXEC                         = 0x22,
+    IT_PRED_EXEC                         = 0x23,
+    IT_START_3D_CMDBUF                   = 0x24,
+    IT_DRAW_INDEX_2                      = 0x27,
+    IT_CONTEXT_CONTROL                   = 0x28,
+    IT_DRAW_INDEX_IMMD_BE                = 0x29,
+    IT_INDEX_TYPE                        = 0x2A,
+    IT_DRAW_INDEX                        = 0x2B,
+    IT_DRAW_INDEX_AUTO                   = 0x2D,
+    IT_DRAW_INDEX_IMMD                   = 0x2E,
+    IT_NUM_INSTANCES                     = 0x2F,
+    IT_STRMOUT_BUFFER_UPDATE             = 0x34,
+    IT_INDIRECT_BUFFER_MP                = 0x38,
+    IT_MEM_SEMAPHORE                     = 0x39,
+    IT_MPEG_INDEX                        = 0x3A,
+    IT_WAIT_REG_MEM                      = 0x3C,
+    IT_MEM_WRITE                         = 0x3D,
+    IT_INDIRECT_BUFFER                   = 0x32,
+    IT_CP_INTERRUPT                      = 0x40,
+    IT_SURFACE_SYNC                      = 0x43,
+    IT_ME_INITIALIZE                     = 0x44,
+    IT_COND_WRITE                        = 0x45,
+    IT_EVENT_WRITE                       = 0x46,
+    IT_EVENT_WRITE_EOP                   = 0x47,
+    IT_ONE_REG_WRITE                     = 0x57,
+    IT_SET_CONFIG_REG                    = 0x68,
+    IT_SET_CONTEXT_REG                   = 0x69,
+    IT_SET_ALU_CONST                     = 0x6A,
+    IT_SET_BOOL_CONST                    = 0x6B,
+    IT_SET_LOOP_CONST                    = 0x6C,
+    IT_SET_RESOURCE                      = 0x6D,
+    IT_SET_SAMPLER                       = 0x6E,
+    IT_SET_CTL_CONST                     = 0x6F,
+    IT_SURFACE_BASE_UPDATE               = 0x73,
+};
+
+#endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h b/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h
new file mode 100644
index 0000000000..9d5aa3c7e4
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h
@@ -0,0 +1,3087 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _AUTOREGS
+#define _AUTOREGS
+
+enum {
+
+    VGT_VTX_VECT_EJECT_REG                                = 0x000088b0,
+	PRIM_COUNT_mask                                   = 0x3ff << 0,
+	PRIM_COUNT_shift                                  = 0,
+    VGT_LAST_COPY_STATE                                   = 0x000088c0,
+	SRC_STATE_ID_mask                                 = 0x07 << 0,
+	SRC_STATE_ID_shift                                = 0,
+	DST_STATE_ID_mask                                 = 0x07 << 16,
+	DST_STATE_ID_shift                                = 16,
+    VGT_CACHE_INVALIDATION                                = 0x000088c4,
+	CACHE_INVALIDATION_mask                           = 0x03 << 0,
+	CACHE_INVALIDATION_shift                          = 0,
+	    VC_ONLY                                       = 0x00,
+	    TC_ONLY                                       = 0x01,
+	    VC_AND_TC                                     = 0x02,
+	VS_NO_EXTRA_BUFFER_bit                            = 1 << 5,
+    VGT_GS_PER_ES                                         = 0x000088c8,
+    VGT_ES_PER_GS                                         = 0x000088cc,
+    VGT_GS_VERTEX_REUSE                                   = 0x000088d4,
+	VERT_REUSE_mask                                   = 0x1f << 0,
+	VERT_REUSE_shift                                  = 0,
+    VGT_MC_LAT_CNTL                                       = 0x000088d8,
+	MC_TIME_STAMP_RES_mask                            = 0x03 << 0,
+	MC_TIME_STAMP_RES_shift                           = 0,
+	    X_0_992_MAX_LATENCY                           = 0x00,
+	    X_0_496_MAX_LATENCY                           = 0x01,
+	    X_0_248_MAX_LATENCY                           = 0x02,
+	    X_0_124_MAX_LATENCY                           = 0x03,
+    VGT_GS_PER_VS                                         = 0x000088e8,
+	GS_PER_VS_mask                                    = 0x0f << 0,
+	GS_PER_VS_shift                                   = 0,
+    VGT_CNTL_STATUS                                       = 0x000088f0,
+	VGT_OUT_INDX_BUSY_bit                             = 1 << 0,
+	VGT_OUT_BUSY_bit                                  = 1 << 1,
+	VGT_PT_BUSY_bit                                   = 1 << 2,
+	VGT_TE_BUSY_bit                                   = 1 << 3,
+	VGT_VR_BUSY_bit                                   = 1 << 4,
+	VGT_GRP_BUSY_bit                                  = 1 << 5,
+	VGT_DMA_REQ_BUSY_bit                              = 1 << 6,
+	VGT_DMA_BUSY_bit                                  = 1 << 7,
+	VGT_GS_BUSY_bit                                   = 1 << 8,
+	VGT_BUSY_bit                                      = 1 << 9,
+    VGT_PRIMITIVE_TYPE                                    = 0x00008958,
+	VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask                = 0x3f << 0,
+	VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift               = 0,
+	    DI_PT_NONE                                    = 0x00,
+	    DI_PT_POINTLIST                               = 0x01,
+	    DI_PT_LINELIST                                = 0x02,
+	    DI_PT_LINESTRIP                               = 0x03,
+	    DI_PT_TRILIST                                 = 0x04,
+	    DI_PT_TRIFAN                                  = 0x05,
+	    DI_PT_TRISTRIP                                = 0x06,
+	    DI_PT_UNUSED_0                                = 0x07,
+	    DI_PT_UNUSED_1                                = 0x08,
+	    DI_PT_UNUSED_2                                = 0x09,
+	    DI_PT_LINELIST_ADJ                            = 0x0a,
+	    DI_PT_LINESTRIP_ADJ                           = 0x0b,
+	    DI_PT_TRILIST_ADJ                             = 0x0c,
+	    DI_PT_TRISTRIP_ADJ                            = 0x0d,
+	    DI_PT_UNUSED_3                                = 0x0e,
+	    DI_PT_UNUSED_4                                = 0x0f,
+	    DI_PT_TRI_WITH_WFLAGS                         = 0x10,
+	    DI_PT_RECTLIST                                = 0x11,
+	    DI_PT_LINELOOP                                = 0x12,
+	    DI_PT_QUADLIST                                = 0x13,
+	    DI_PT_QUADSTRIP                               = 0x14,
+	    DI_PT_POLYGON                                 = 0x15,
+	    DI_PT_2D_COPY_RECT_LIST_V0                    = 0x16,
+	    DI_PT_2D_COPY_RECT_LIST_V1                    = 0x17,
+	    DI_PT_2D_COPY_RECT_LIST_V2                    = 0x18,
+	    DI_PT_2D_COPY_RECT_LIST_V3                    = 0x19,
+	    DI_PT_2D_FILL_RECT_LIST                       = 0x1a,
+	    DI_PT_2D_LINE_STRIP                           = 0x1b,
+	    DI_PT_2D_TRI_STRIP                            = 0x1c,
+    VGT_INDEX_TYPE                                        = 0x0000895c,
+	INDEX_TYPE_mask                                   = 0x03 << 0,
+	INDEX_TYPE_shift                                  = 0,
+	    DI_INDEX_SIZE_16_BIT                          = 0x00,
+	    DI_INDEX_SIZE_32_BIT                          = 0x01,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_0                      = 0x00008960,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_1                      = 0x00008964,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_2                      = 0x00008968,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_3                      = 0x0000896c,
+    VGT_NUM_INDICES                                       = 0x00008970,
+    VGT_NUM_INSTANCES                                     = 0x00008974,
+    PA_CL_CNTL_STATUS                                     = 0x00008a10,
+	CL_BUSY_bit                                       = 1 << 31,
+    PA_CL_ENHANCE                                         = 0x00008a14,
+	CLIP_VTX_REORDER_ENA_bit                          = 1 << 0,
+	NUM_CLIP_SEQ_mask                                 = 0x03 << 1,
+	NUM_CLIP_SEQ_shift                                = 1,
+	CLIPPED_PRIM_SEQ_STALL_bit                        = 1 << 3,
+	VE_NAN_PROC_DISABLE_bit                           = 1 << 4,
+    PA_SU_CNTL_STATUS                                     = 0x00008a50,
+	SU_BUSY_bit                                       = 1 << 31,
+    PA_SC_LINE_STIPPLE_STATE                              = 0x00008b10,
+	CURRENT_PTR_mask                                  = 0x0f << 0,
+	CURRENT_PTR_shift                                 = 0,
+	CURRENT_COUNT_mask                                = 0xff << 8,
+	CURRENT_COUNT_shift                               = 8,
+    PA_SC_MULTI_CHIP_CNTL                                 = 0x00008b20,
+	LOG2_NUM_CHIPS_mask                               = 0x07 << 0,
+	LOG2_NUM_CHIPS_shift                              = 0,
+	MULTI_CHIP_TILE_SIZE_mask                         = 0x03 << 3,
+	MULTI_CHIP_TILE_SIZE_shift                        = 3,
+	    X_16_X_16_PIXEL_TILE_PER_CHIP                 = 0x00,
+	    X_32_X_32_PIXEL_TILE_PER_CHIP                 = 0x01,
+	    X_64_X_64_PIXEL_TILE_PER_CHIP                 = 0x02,
+	    X_128X128_PIXEL_TILE_PER_CHIP                 = 0x03,
+	CHIP_TILE_X_LOC_mask                              = 0x07 << 5,
+	CHIP_TILE_X_LOC_shift                             = 5,
+	CHIP_TILE_Y_LOC_mask                              = 0x07 << 8,
+	CHIP_TILE_Y_LOC_shift                             = 8,
+	CHIP_SUPER_TILE_B_bit                             = 1 << 11,
+    PA_SC_AA_SAMPLE_LOCS_2S                               = 0x00008b40,
+	S0_X_mask                                         = 0x0f << 0,
+	S0_X_shift                                        = 0,
+	S0_Y_mask                                         = 0x0f << 4,
+	S0_Y_shift                                        = 4,
+	S1_X_mask                                         = 0x0f << 8,
+	S1_X_shift                                        = 8,
+	S1_Y_mask                                         = 0x0f << 12,
+	S1_Y_shift                                        = 12,
+    PA_SC_AA_SAMPLE_LOCS_4S                               = 0x00008b44,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+	S2_X_mask                                         = 0x0f << 16,
+	S2_X_shift                                        = 16,
+	S2_Y_mask                                         = 0x0f << 20,
+	S2_Y_shift                                        = 20,
+	S3_X_mask                                         = 0x0f << 24,
+	S3_X_shift                                        = 24,
+	S3_Y_mask                                         = 0x0f << 28,
+	S3_Y_shift                                        = 28,
+    PA_SC_AA_SAMPLE_LOCS_8S_WD0                           = 0x00008b48,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+/* 	S2_X_mask                                         = 0x0f << 16, */
+/* 	S2_X_shift                                        = 16, */
+/* 	S2_Y_mask                                         = 0x0f << 20, */
+/* 	S2_Y_shift                                        = 20, */
+/* 	S3_X_mask                                         = 0x0f << 24, */
+/* 	S3_X_shift                                        = 24, */
+/* 	S3_Y_mask                                         = 0x0f << 28, */
+/* 	S3_Y_shift                                        = 28, */
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1                           = 0x00008b4c,
+	S4_X_mask                                         = 0x0f << 0,
+	S4_X_shift                                        = 0,
+	S4_Y_mask                                         = 0x0f << 4,
+	S4_Y_shift                                        = 4,
+	S5_X_mask                                         = 0x0f << 8,
+	S5_X_shift                                        = 8,
+	S5_Y_mask                                         = 0x0f << 12,
+	S5_Y_shift                                        = 12,
+	S6_X_mask                                         = 0x0f << 16,
+	S6_X_shift                                        = 16,
+	S6_Y_mask                                         = 0x0f << 20,
+	S6_Y_shift                                        = 20,
+	S7_X_mask                                         = 0x0f << 24,
+	S7_X_shift                                        = 24,
+	S7_Y_mask                                         = 0x0f << 28,
+	S7_Y_shift                                        = 28,
+    PA_SC_CNTL_STATUS                                     = 0x00008be0,
+	MPASS_OVERFLOW_bit                                = 1 << 30,
+    PA_SC_ENHANCE                                         = 0x00008bf0,
+	FORCE_EOV_MAX_CLK_CNT_mask                        = 0xfff << 0,
+	FORCE_EOV_MAX_CLK_CNT_shift                       = 0,
+	FORCE_EOV_MAX_TILE_CNT_mask                       = 0xfff << 12,
+	FORCE_EOV_MAX_TILE_CNT_shift                      = 12,
+    SQ_CONFIG                                             = 0x00008c00,
+	VC_ENABLE_bit                                     = 1 << 0,
+	EXPORT_SRC_C_bit                                  = 1 << 1,
+	DX9_CONSTS_bit                                    = 1 << 2,
+	ALU_INST_PREFER_VECTOR_bit                        = 1 << 3,
+	SQ_CONFIG__DX10_CLAMP_bit                         = 1 << 4,
+	ALU_PREFER_ONE_WATERFALL_bit                      = 1 << 5,
+	ALU_MAX_ONE_WATERFALL_bit                         = 1 << 6,
+	CLAUSE_SEQ_PRIO_mask                              = 0x03 << 8,
+	CLAUSE_SEQ_PRIO_shift                             = 8,
+	    SQ_CL_PRIO_RND_ROBIN                          = 0x00,
+	    SQ_CL_PRIO_MACRO_SEQ                          = 0x01,
+	    SQ_CL_PRIO_NONE                               = 0x02,
+	PS_PRIO_mask                                      = 0x03 << 24,
+	PS_PRIO_shift                                     = 24,
+	VS_PRIO_mask                                      = 0x03 << 26,
+	VS_PRIO_shift                                     = 26,
+	GS_PRIO_mask                                      = 0x03 << 28,
+	GS_PRIO_shift                                     = 28,
+	ES_PRIO_mask                                      = 0x03 << 30,
+	ES_PRIO_shift                                     = 30,
+    SQ_GPR_RESOURCE_MGMT_1                                = 0x00008c04,
+	NUM_PS_GPRS_mask                                  = 0xff << 0,
+	NUM_PS_GPRS_shift                                 = 0,
+	NUM_VS_GPRS_mask                                  = 0xff << 16,
+	NUM_VS_GPRS_shift                                 = 16,
+	NUM_CLAUSE_TEMP_GPRS_mask                         = 0x0f << 28,
+	NUM_CLAUSE_TEMP_GPRS_shift                        = 28,
+    SQ_GPR_RESOURCE_MGMT_2                                = 0x00008c08,
+	NUM_GS_GPRS_mask                                  = 0xff << 0,
+	NUM_GS_GPRS_shift                                 = 0,
+	NUM_ES_GPRS_mask                                  = 0xff << 16,
+	NUM_ES_GPRS_shift                                 = 16,
+    SQ_THREAD_RESOURCE_MGMT                               = 0x00008c0c,
+	NUM_PS_THREADS_mask                               = 0xff << 0,
+	NUM_PS_THREADS_shift                              = 0,
+	NUM_VS_THREADS_mask                               = 0xff << 8,
+	NUM_VS_THREADS_shift                              = 8,
+	NUM_GS_THREADS_mask                               = 0xff << 16,
+	NUM_GS_THREADS_shift                              = 16,
+	NUM_ES_THREADS_mask                               = 0xff << 24,
+	NUM_ES_THREADS_shift                              = 24,
+    SQ_STACK_RESOURCE_MGMT_1                              = 0x00008c10,
+	NUM_PS_STACK_ENTRIES_mask                         = 0xfff << 0,
+	NUM_PS_STACK_ENTRIES_shift                        = 0,
+	NUM_VS_STACK_ENTRIES_mask                         = 0xfff << 16,
+	NUM_VS_STACK_ENTRIES_shift                        = 16,
+    SQ_STACK_RESOURCE_MGMT_2                              = 0x00008c14,
+	NUM_GS_STACK_ENTRIES_mask                         = 0xfff << 0,
+	NUM_GS_STACK_ENTRIES_shift                        = 0,
+	NUM_ES_STACK_ENTRIES_mask                         = 0xfff << 16,
+	NUM_ES_STACK_ENTRIES_shift                        = 16,
+    SQ_ESGS_RING_BASE                                     = 0x00008c40,
+    SQ_ESGS_RING_SIZE                                     = 0x00008c44,
+    SQ_GSVS_RING_BASE                                     = 0x00008c48,
+    SQ_GSVS_RING_SIZE                                     = 0x00008c4c,
+    SQ_ESTMP_RING_BASE                                    = 0x00008c50,
+    SQ_ESTMP_RING_SIZE                                    = 0x00008c54,
+    SQ_GSTMP_RING_BASE                                    = 0x00008c58,
+    SQ_GSTMP_RING_SIZE                                    = 0x00008c5c,
+    SQ_VSTMP_RING_BASE                                    = 0x00008c60,
+    SQ_VSTMP_RING_SIZE                                    = 0x00008c64,
+    SQ_PSTMP_RING_BASE                                    = 0x00008c68,
+    SQ_PSTMP_RING_SIZE                                    = 0x00008c6c,
+    SQ_FBUF_RING_BASE                                     = 0x00008c70,
+    SQ_FBUF_RING_SIZE                                     = 0x00008c74,
+    SQ_REDUC_RING_BASE                                    = 0x00008c78,
+    SQ_REDUC_RING_SIZE                                    = 0x00008c7c,
+    SQ_ALU_WORD1_OP3                                      = 0x00008dfc,
+	SRC2_SEL_mask                                     = 0x1ff << 0,
+	SRC2_SEL_shift                                    = 0,
+	    SQ_ALU_SRC_0                                  = 0xf8,
+	    SQ_ALU_SRC_1                                  = 0xf9,
+	    SQ_ALU_SRC_1_INT                              = 0xfa,
+	    SQ_ALU_SRC_M_1_INT                            = 0xfb,
+	    SQ_ALU_SRC_0_5                                = 0xfc,
+	    SQ_ALU_SRC_LITERAL                            = 0xfd,
+	    SQ_ALU_SRC_PV                                 = 0xfe,
+	    SQ_ALU_SRC_PS                                 = 0xff,
+	SRC2_REL_bit                                      = 1 << 9,
+	SRC2_CHAN_mask                                    = 0x03 << 10,
+	SRC2_CHAN_shift                                   = 10,
+	    SQ_CHAN_X                                     = 0x00,
+	    SQ_CHAN_Y                                     = 0x01,
+	    SQ_CHAN_Z                                     = 0x02,
+	    SQ_CHAN_W                                     = 0x03,
+	SRC2_NEG_bit                                      = 1 << 12,
+	SQ_ALU_WORD1_OP3__ALU_INST_mask                   = 0x1f << 13,
+	SQ_ALU_WORD1_OP3__ALU_INST_shift                  = 13,
+	    SQ_OP3_INST_MUL_LIT                           = 0x0c,
+	    SQ_OP3_INST_MUL_LIT_M2                        = 0x0d,
+	    SQ_OP3_INST_MUL_LIT_M4                        = 0x0e,
+	    SQ_OP3_INST_MUL_LIT_D2                        = 0x0f,
+	    SQ_OP3_INST_MULADD                            = 0x10,
+	    SQ_OP3_INST_MULADD_M2                         = 0x11,
+	    SQ_OP3_INST_MULADD_M4                         = 0x12,
+	    SQ_OP3_INST_MULADD_D2                         = 0x13,
+	    SQ_OP3_INST_MULADD_IEEE                       = 0x14,
+	    SQ_OP3_INST_MULADD_IEEE_M2                    = 0x15,
+	    SQ_OP3_INST_MULADD_IEEE_M4                    = 0x16,
+	    SQ_OP3_INST_MULADD_IEEE_D2                    = 0x17,
+	    SQ_OP3_INST_CNDE                              = 0x18,
+	    SQ_OP3_INST_CNDGT                             = 0x19,
+	    SQ_OP3_INST_CNDGE                             = 0x1a,
+	    SQ_OP3_INST_CNDE_INT                          = 0x1c,
+	    SQ_OP3_INST_CNDGT_INT                         = 0x1d,
+	    SQ_OP3_INST_CNDGE_INT                         = 0x1e,
+    SQ_TEX_WORD2                                          = 0x00008dfc,
+	OFFSET_X_mask                                     = 0x1f << 0,
+	OFFSET_X_shift                                    = 0,
+	OFFSET_Y_mask                                     = 0x1f << 5,
+	OFFSET_Y_shift                                    = 5,
+	OFFSET_Z_mask                                     = 0x1f << 10,
+	OFFSET_Z_shift                                    = 10,
+	SAMPLER_ID_mask                                   = 0x1f << 15,
+	SAMPLER_ID_shift                                  = 15,
+	SQ_TEX_WORD2__SRC_SEL_X_mask                      = 0x07 << 20,
+	SQ_TEX_WORD2__SRC_SEL_X_shift                     = 20,
+	    SQ_SEL_X                                      = 0x00,
+	    SQ_SEL_Y                                      = 0x01,
+	    SQ_SEL_Z                                      = 0x02,
+	    SQ_SEL_W                                      = 0x03,
+	    SQ_SEL_0                                      = 0x04,
+	    SQ_SEL_1                                      = 0x05,
+	SRC_SEL_Y_mask                                    = 0x07 << 23,
+	SRC_SEL_Y_shift                                   = 23,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SRC_SEL_Z_mask                                    = 0x07 << 26,
+	SRC_SEL_Z_shift                                   = 26,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SRC_SEL_W_mask                                    = 0x07 << 29,
+	SRC_SEL_W_shift                                   = 29,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+    SQ_CF_ALLOC_EXPORT_WORD1                              = 0x00008dfc,
+	BURST_COUNT_mask                                  = 0x0f << 17,
+	BURST_COUNT_shift                                 = 17,
+	END_OF_PROGRAM_bit                                = 1 << 21,
+	VALID_PIXEL_MODE_bit                              = 1 << 22,
+	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_mask            = 0x7f << 23,
+	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_shift           = 23,
+	    SQ_CF_INST_MEM_STREAM0                        = 0x20,
+	    SQ_CF_INST_MEM_STREAM1                        = 0x21,
+	    SQ_CF_INST_MEM_STREAM2                        = 0x22,
+	    SQ_CF_INST_MEM_STREAM3                        = 0x23,
+	    SQ_CF_INST_MEM_SCRATCH                        = 0x24,
+	    SQ_CF_INST_MEM_REDUCTION                      = 0x25,
+	    SQ_CF_INST_MEM_RING                           = 0x26,
+	    SQ_CF_INST_EXPORT                             = 0x27,
+	    SQ_CF_INST_EXPORT_DONE                        = 0x28,
+	WHOLE_QUAD_MODE_bit                               = 1 << 30,
+	BARRIER_bit                                       = 1 << 31,
+    SQ_CF_ALU_WORD1                                       = 0x00008dfc,
+	KCACHE_MODE1_mask                                 = 0x03 << 0,
+	KCACHE_MODE1_shift                                = 0,
+	    SQ_CF_KCACHE_NOP                              = 0x00,
+	    SQ_CF_KCACHE_LOCK_1                           = 0x01,
+	    SQ_CF_KCACHE_LOCK_2                           = 0x02,
+	    SQ_CF_KCACHE_LOCK_LOOP_INDEX                  = 0x03,
+	KCACHE_ADDR0_mask                                 = 0xff << 2,
+	KCACHE_ADDR0_shift                                = 2,
+	KCACHE_ADDR1_mask                                 = 0xff << 10,
+	KCACHE_ADDR1_shift                                = 10,
+	SQ_CF_ALU_WORD1__COUNT_mask                       = 0x7f << 18,
+	SQ_CF_ALU_WORD1__COUNT_shift                      = 18,
+	SQ_CF_ALU_WORD1__ALT_CONST_bit                    = 1 << 25,
+	SQ_CF_ALU_WORD1__CF_INST_mask                     = 0x0f << 26,
+	SQ_CF_ALU_WORD1__CF_INST_shift                    = 26,
+	    SQ_CF_INST_ALU                                = 0x08,
+	    SQ_CF_INST_ALU_PUSH_BEFORE                    = 0x09,
+	    SQ_CF_INST_ALU_POP_AFTER                      = 0x0a,
+	    SQ_CF_INST_ALU_POP2_AFTER                     = 0x0b,
+	    SQ_CF_INST_ALU_CONTINUE                       = 0x0d,
+	    SQ_CF_INST_ALU_BREAK                          = 0x0e,
+	    SQ_CF_INST_ALU_ELSE_AFTER                     = 0x0f,
+/* 	WHOLE_QUAD_MODE_bit                               = 1 << 30, */
+/* 	BARRIER_bit                                       = 1 << 31, */
+    SQ_TEX_WORD1                                          = 0x00008dfc,
+	SQ_TEX_WORD1__DST_GPR_mask                        = 0x7f << 0,
+	SQ_TEX_WORD1__DST_GPR_shift                       = 0,
+	SQ_TEX_WORD1__DST_REL_bit                         = 1 << 7,
+	SQ_TEX_WORD1__DST_SEL_X_mask                      = 0x07 << 9,
+	SQ_TEX_WORD1__DST_SEL_X_shift                     = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	    SQ_SEL_MASK                                   = 0x07,
+	SQ_TEX_WORD1__DST_SEL_Y_mask                      = 0x07 << 12,
+	SQ_TEX_WORD1__DST_SEL_Y_shift                     = 12,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__DST_SEL_Z_mask                      = 0x07 << 15,
+	SQ_TEX_WORD1__DST_SEL_Z_shift                     = 15,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__DST_SEL_W_mask                      = 0x07 << 18,
+	SQ_TEX_WORD1__DST_SEL_W_shift                     = 18,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__LOD_BIAS_mask                       = 0x7f << 21,
+	SQ_TEX_WORD1__LOD_BIAS_shift                      = 21,
+	COORD_TYPE_X_bit                                  = 1 << 28,
+	COORD_TYPE_Y_bit                                  = 1 << 29,
+	COORD_TYPE_Z_bit                                  = 1 << 30,
+	COORD_TYPE_W_bit                                  = 1 << 31,
+    SQ_VTX_WORD0                                          = 0x00008dfc,
+	VTX_INST_mask                                     = 0x1f << 0,
+	VTX_INST_shift                                    = 0,
+	    SQ_VTX_INST_FETCH                             = 0x00,
+	    SQ_VTX_INST_SEMANTIC                          = 0x01,
+	FETCH_TYPE_mask                                   = 0x03 << 5,
+	FETCH_TYPE_shift                                  = 5,
+	    SQ_VTX_FETCH_VERTEX_DATA                      = 0x00,
+	    SQ_VTX_FETCH_INSTANCE_DATA                    = 0x01,
+	    SQ_VTX_FETCH_NO_INDEX_OFFSET                  = 0x02,
+	FETCH_WHOLE_QUAD_bit                              = 1 << 7,
+	BUFFER_ID_mask                                    = 0xff << 8,
+	BUFFER_ID_shift                                   = 8,
+	SRC_GPR_mask                                      = 0x7f << 16,
+	SRC_GPR_shift                                     = 16,
+	SRC_REL_bit                                       = 1 << 23,
+	SQ_VTX_WORD0__SRC_SEL_X_mask                      = 0x03 << 24,
+	SQ_VTX_WORD0__SRC_SEL_X_shift                     = 24,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+	MEGA_FETCH_COUNT_mask                             = 0x3f << 26,
+	MEGA_FETCH_COUNT_shift                            = 26,
+    SQ_CF_ALLOC_EXPORT_WORD1_SWIZ                         = 0x00008dfc,
+	SEL_X_mask                                        = 0x07 << 0,
+	SEL_X_shift                                       = 0,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_Y_mask                                        = 0x07 << 3,
+	SEL_Y_shift                                       = 3,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_Z_mask                                        = 0x07 << 6,
+	SEL_Z_shift                                       = 6,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_W_mask                                        = 0x07 << 9,
+	SEL_W_shift                                       = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+    SQ_ALU_WORD1                                          = 0x00008dfc,
+	ENCODING_mask                                     = 0x07 << 15,
+	ENCODING_shift                                    = 15,
+	BANK_SWIZZLE_mask                                 = 0x07 << 18,
+	BANK_SWIZZLE_shift                                = 18,
+	    SQ_ALU_VEC_012                                = 0x00,
+	    SQ_ALU_VEC_021                                = 0x01,
+	    SQ_ALU_VEC_120                                = 0x02,
+	    SQ_ALU_VEC_102                                = 0x03,
+	    SQ_ALU_VEC_201                                = 0x04,
+	    SQ_ALU_VEC_210                                = 0x05,
+	SQ_ALU_WORD1__DST_GPR_mask                        = 0x7f << 21,
+	SQ_ALU_WORD1__DST_GPR_shift                       = 21,
+	SQ_ALU_WORD1__DST_REL_bit                         = 1 << 28,
+	DST_CHAN_mask                                     = 0x03 << 29,
+	DST_CHAN_shift                                    = 29,
+	    CHAN_X                                        = 0x00,
+	    CHAN_Y                                        = 0x01,
+	    CHAN_Z                                        = 0x02,
+	    CHAN_W                                        = 0x03,
+	SQ_ALU_WORD1__CLAMP_bit                           = 1 << 31,
+    SQ_CF_ALU_WORD0                                       = 0x00008dfc,
+	SQ_CF_ALU_WORD0__ADDR_mask                        = 0x3fffff << 0,
+	SQ_CF_ALU_WORD0__ADDR_shift                       = 0,
+	KCACHE_BANK0_mask                                 = 0x0f << 22,
+	KCACHE_BANK0_shift                                = 22,
+	KCACHE_BANK1_mask                                 = 0x0f << 26,
+	KCACHE_BANK1_shift                                = 26,
+	KCACHE_MODE0_mask                                 = 0x03 << 30,
+	KCACHE_MODE0_shift                                = 30,
+/* 	    SQ_CF_KCACHE_NOP                              = 0x00, */
+/* 	    SQ_CF_KCACHE_LOCK_1                           = 0x01, */
+/* 	    SQ_CF_KCACHE_LOCK_2                           = 0x02, */
+/* 	    SQ_CF_KCACHE_LOCK_LOOP_INDEX                  = 0x03, */
+    SQ_VTX_WORD2                                          = 0x00008dfc,
+	SQ_VTX_WORD2__OFFSET_mask                         = 0xffff << 0,
+	SQ_VTX_WORD2__OFFSET_shift                        = 0,
+	SQ_VTX_WORD2__ENDIAN_SWAP_mask                    = 0x03 << 16,
+	SQ_VTX_WORD2__ENDIAN_SWAP_shift                   = 16,
+	    SQ_ENDIAN_NONE                                = 0x00,
+	    SQ_ENDIAN_8IN16                               = 0x01,
+	    SQ_ENDIAN_8IN32                               = 0x02,
+	CONST_BUF_NO_STRIDE_bit                           = 1 << 18,
+	MEGA_FETCH_bit                                    = 1 << 19,
+	SQ_VTX_WORD2__ALT_CONST_bit                       = 1 << 20,
+    SQ_ALU_WORD1_OP2_V2                                   = 0x00008dfc,
+	SRC0_ABS_bit                                      = 1 << 0,
+	SRC1_ABS_bit                                      = 1 << 1,
+	UPDATE_EXECUTE_MASK_bit                           = 1 << 2,
+	UPDATE_PRED_bit                                   = 1 << 3,
+	WRITE_MASK_bit                                    = 1 << 4,
+	SQ_ALU_WORD1_OP2_V2__OMOD_mask                    = 0x03 << 5,
+	SQ_ALU_WORD1_OP2_V2__OMOD_shift                   = 5,
+	    SQ_ALU_OMOD_OFF                               = 0x00,
+	    SQ_ALU_OMOD_M2                                = 0x01,
+	    SQ_ALU_OMOD_M4                                = 0x02,
+	    SQ_ALU_OMOD_D2                                = 0x03,
+	SQ_ALU_WORD1_OP2_V2__ALU_INST_mask                = 0x7ff << 7,
+	SQ_ALU_WORD1_OP2_V2__ALU_INST_shift               = 7,
+	    SQ_OP2_INST_ADD                               = 0x00,
+	    SQ_OP2_INST_MUL                               = 0x01,
+	    SQ_OP2_INST_MUL_IEEE                          = 0x02,
+	    SQ_OP2_INST_MAX                               = 0x03,
+	    SQ_OP2_INST_MIN                               = 0x04,
+	    SQ_OP2_INST_MAX_DX10                          = 0x05,
+	    SQ_OP2_INST_MIN_DX10                          = 0x06,
+	    SQ_OP2_INST_SETE                              = 0x08,
+	    SQ_OP2_INST_SETGT                             = 0x09,
+	    SQ_OP2_INST_SETGE                             = 0x0a,
+	    SQ_OP2_INST_SETNE                             = 0x0b,
+	    SQ_OP2_INST_SETE_DX10                         = 0x0c,
+	    SQ_OP2_INST_SETGT_DX10                        = 0x0d,
+	    SQ_OP2_INST_SETGE_DX10                        = 0x0e,
+	    SQ_OP2_INST_SETNE_DX10                        = 0x0f,
+	    SQ_OP2_INST_FRACT                             = 0x10,
+	    SQ_OP2_INST_TRUNC                             = 0x11,
+	    SQ_OP2_INST_CEIL                              = 0x12,
+	    SQ_OP2_INST_RNDNE                             = 0x13,
+	    SQ_OP2_INST_FLOOR                             = 0x14,
+	    SQ_OP2_INST_MOVA                              = 0x15,
+	    SQ_OP2_INST_MOVA_FLOOR                        = 0x16,
+	    SQ_OP2_INST_MOVA_INT                          = 0x18,
+	    SQ_OP2_INST_MOV                               = 0x19,
+	    SQ_OP2_INST_NOP                               = 0x1a,
+	    SQ_OP2_INST_PRED_SETGT_UINT                   = 0x1e,
+	    SQ_OP2_INST_PRED_SETGE_UINT                   = 0x1f,
+	    SQ_OP2_INST_PRED_SETE                         = 0x20,
+	    SQ_OP2_INST_PRED_SETGT                        = 0x21,
+	    SQ_OP2_INST_PRED_SETGE                        = 0x22,
+	    SQ_OP2_INST_PRED_SETNE                        = 0x23,
+	    SQ_OP2_INST_PRED_SET_INV                      = 0x24,
+	    SQ_OP2_INST_PRED_SET_POP                      = 0x25,
+	    SQ_OP2_INST_PRED_SET_CLR                      = 0x26,
+	    SQ_OP2_INST_PRED_SET_RESTORE                  = 0x27,
+	    SQ_OP2_INST_PRED_SETE_PUSH                    = 0x28,
+	    SQ_OP2_INST_PRED_SETGT_PUSH                   = 0x29,
+	    SQ_OP2_INST_PRED_SETGE_PUSH                   = 0x2a,
+	    SQ_OP2_INST_PRED_SETNE_PUSH                   = 0x2b,
+	    SQ_OP2_INST_KILLE                             = 0x2c,
+	    SQ_OP2_INST_KILLGT                            = 0x2d,
+	    SQ_OP2_INST_KILLGE                            = 0x2e,
+	    SQ_OP2_INST_KILLNE                            = 0x2f,
+	    SQ_OP2_INST_AND_INT                           = 0x30,
+	    SQ_OP2_INST_OR_INT                            = 0x31,
+	    SQ_OP2_INST_XOR_INT                           = 0x32,
+	    SQ_OP2_INST_NOT_INT                           = 0x33,
+	    SQ_OP2_INST_ADD_INT                           = 0x34,
+	    SQ_OP2_INST_SUB_INT                           = 0x35,
+	    SQ_OP2_INST_MAX_INT                           = 0x36,
+	    SQ_OP2_INST_MIN_INT                           = 0x37,
+	    SQ_OP2_INST_MAX_UINT                          = 0x38,
+	    SQ_OP2_INST_MIN_UINT                          = 0x39,
+	    SQ_OP2_INST_SETE_INT                          = 0x3a,
+	    SQ_OP2_INST_SETGT_INT                         = 0x3b,
+	    SQ_OP2_INST_SETGE_INT                         = 0x3c,
+	    SQ_OP2_INST_SETNE_INT                         = 0x3d,
+	    SQ_OP2_INST_SETGT_UINT                        = 0x3e,
+	    SQ_OP2_INST_SETGE_UINT                        = 0x3f,
+	    SQ_OP2_INST_KILLGT_UINT                       = 0x40,
+	    SQ_OP2_INST_KILLGE_UINT                       = 0x41,
+	    SQ_OP2_INST_PRED_SETE_INT                     = 0x42,
+	    SQ_OP2_INST_PRED_SETGT_INT                    = 0x43,
+	    SQ_OP2_INST_PRED_SETGE_INT                    = 0x44,
+	    SQ_OP2_INST_PRED_SETNE_INT                    = 0x45,
+	    SQ_OP2_INST_KILLE_INT                         = 0x46,
+	    SQ_OP2_INST_KILLGT_INT                        = 0x47,
+	    SQ_OP2_INST_KILLGE_INT                        = 0x48,
+	    SQ_OP2_INST_KILLNE_INT                        = 0x49,
+	    SQ_OP2_INST_PRED_SETE_PUSH_INT                = 0x4a,
+	    SQ_OP2_INST_PRED_SETGT_PUSH_INT               = 0x4b,
+	    SQ_OP2_INST_PRED_SETGE_PUSH_INT               = 0x4c,
+	    SQ_OP2_INST_PRED_SETNE_PUSH_INT               = 0x4d,
+	    SQ_OP2_INST_PRED_SETLT_PUSH_INT               = 0x4e,
+	    SQ_OP2_INST_PRED_SETLE_PUSH_INT               = 0x4f,
+	    SQ_OP2_INST_DOT4                              = 0x50,
+	    SQ_OP2_INST_DOT4_IEEE                         = 0x51,
+	    SQ_OP2_INST_CUBE                              = 0x52,
+	    SQ_OP2_INST_MAX4                              = 0x53,
+	    SQ_OP2_INST_MOVA_GPR_INT                      = 0x60,
+	    SQ_OP2_INST_EXP_IEEE                          = 0x61,
+	    SQ_OP2_INST_LOG_CLAMPED                       = 0x62,
+	    SQ_OP2_INST_LOG_IEEE                          = 0x63,
+	    SQ_OP2_INST_RECIP_CLAMPED                     = 0x64,
+	    SQ_OP2_INST_RECIP_FF                          = 0x65,
+	    SQ_OP2_INST_RECIP_IEEE                        = 0x66,
+	    SQ_OP2_INST_RECIPSQRT_CLAMPED                 = 0x67,
+	    SQ_OP2_INST_RECIPSQRT_FF                      = 0x68,
+	    SQ_OP2_INST_RECIPSQRT_IEEE                    = 0x69,
+	    SQ_OP2_INST_SQRT_IEEE                         = 0x6a,
+	    SQ_OP2_INST_FLT_TO_INT                        = 0x6b,
+	    SQ_OP2_INST_INT_TO_FLT                        = 0x6c,
+	    SQ_OP2_INST_UINT_TO_FLT                       = 0x6d,
+	    SQ_OP2_INST_SIN                               = 0x6e,
+	    SQ_OP2_INST_COS                               = 0x6f,
+	    SQ_OP2_INST_ASHR_INT                          = 0x70,
+	    SQ_OP2_INST_LSHR_INT                          = 0x71,
+	    SQ_OP2_INST_LSHL_INT                          = 0x72,
+	    SQ_OP2_INST_MULLO_INT                         = 0x73,
+	    SQ_OP2_INST_MULHI_INT                         = 0x74,
+	    SQ_OP2_INST_MULLO_UINT                        = 0x75,
+	    SQ_OP2_INST_MULHI_UINT                        = 0x76,
+	    SQ_OP2_INST_RECIP_INT                         = 0x77,
+	    SQ_OP2_INST_RECIP_UINT                        = 0x78,
+	    SQ_OP2_INST_FLT_TO_UINT                       = 0x79,
+    SQ_CF_ALLOC_EXPORT_WORD1_BUF                          = 0x00008dfc,
+	ARRAY_SIZE_mask                                   = 0xfff << 0,
+	ARRAY_SIZE_shift                                  = 0,
+	COMP_MASK_mask                                    = 0x0f << 12,
+	COMP_MASK_shift                                   = 12,
+    SQ_CF_WORD0                                           = 0x00008dfc,
+    SQ_CF_ALLOC_EXPORT_WORD0                              = 0x00008dfc,
+	ARRAY_BASE_mask                                   = 0x1fff << 0,
+	ARRAY_BASE_shift                                  = 0,
+	SQ_CF_ALLOC_EXPORT_WORD0__TYPE_mask               = 0x03 << 13,
+	SQ_CF_ALLOC_EXPORT_WORD0__TYPE_shift              = 13,
+	    SQ_EXPORT_PIXEL                               = 0x00,
+	    SQ_EXPORT_POS                                 = 0x01,
+	    SQ_EXPORT_PARAM                               = 0x02,
+	    X_UNUSED_FOR_SX_EXPORTS                       = 0x03,
+	RW_GPR_mask                                       = 0x7f << 15,
+	RW_GPR_shift                                      = 15,
+	RW_REL_bit                                        = 1 << 22,
+	INDEX_GPR_mask                                    = 0x7f << 23,
+	INDEX_GPR_shift                                   = 23,
+	ELEM_SIZE_mask                                    = 0x03 << 30,
+	ELEM_SIZE_shift                                   = 30,
+    SQ_VTX_WORD1                                          = 0x00008dfc,
+	SQ_VTX_WORD1__DST_SEL_X_mask                      = 0x07 << 9,
+	SQ_VTX_WORD1__DST_SEL_X_shift                     = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_Y_mask                      = 0x07 << 12,
+	SQ_VTX_WORD1__DST_SEL_Y_shift                     = 12,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_Z_mask                      = 0x07 << 15,
+	SQ_VTX_WORD1__DST_SEL_Z_shift                     = 15,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_W_mask                      = 0x07 << 18,
+	SQ_VTX_WORD1__DST_SEL_W_shift                     = 18,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	USE_CONST_FIELDS_bit                              = 1 << 21,
+	SQ_VTX_WORD1__DATA_FORMAT_mask                    = 0x3f << 22,
+	SQ_VTX_WORD1__DATA_FORMAT_shift                   = 22,
+	SQ_VTX_WORD1__NUM_FORMAT_ALL_mask                 = 0x03 << 28,
+	SQ_VTX_WORD1__NUM_FORMAT_ALL_shift                = 28,
+	    SQ_NUM_FORMAT_NORM                            = 0x00,
+	    SQ_NUM_FORMAT_INT                             = 0x01,
+	    SQ_NUM_FORMAT_SCALED                          = 0x02,
+	SQ_VTX_WORD1__FORMAT_COMP_ALL_bit                 = 1 << 30,
+	SQ_VTX_WORD1__SRF_MODE_ALL_bit                    = 1 << 31,
+    SQ_ALU_WORD1_OP2                                      = 0x00008dfc,
+/* 	SRC0_ABS_bit                                      = 1 << 0, */
+/* 	SRC1_ABS_bit                                      = 1 << 1, */
+/* 	UPDATE_EXECUTE_MASK_bit                           = 1 << 2, */
+/* 	UPDATE_PRED_bit                                   = 1 << 3, */
+/* 	WRITE_MASK_bit                                    = 1 << 4, */
+	FOG_MERGE_bit                                     = 1 << 5,
+	SQ_ALU_WORD1_OP2__OMOD_mask                       = 0x03 << 6,
+	SQ_ALU_WORD1_OP2__OMOD_shift                      = 6,
+/* 	    SQ_ALU_OMOD_OFF                               = 0x00, */
+/* 	    SQ_ALU_OMOD_M2                                = 0x01, */
+/* 	    SQ_ALU_OMOD_M4                                = 0x02, */
+/* 	    SQ_ALU_OMOD_D2                                = 0x03, */
+	SQ_ALU_WORD1_OP2__ALU_INST_mask                   = 0x3ff << 8,
+	SQ_ALU_WORD1_OP2__ALU_INST_shift                  = 8,
+/* 	    SQ_OP2_INST_ADD                               = 0x00, */
+/* 	    SQ_OP2_INST_MUL                               = 0x01, */
+/* 	    SQ_OP2_INST_MUL_IEEE                          = 0x02, */
+/* 	    SQ_OP2_INST_MAX                               = 0x03, */
+/* 	    SQ_OP2_INST_MIN                               = 0x04, */
+/* 	    SQ_OP2_INST_MAX_DX10                          = 0x05, */
+/* 	    SQ_OP2_INST_MIN_DX10                          = 0x06, */
+/* 	    SQ_OP2_INST_SETE                              = 0x08, */
+/* 	    SQ_OP2_INST_SETGT                             = 0x09, */
+/* 	    SQ_OP2_INST_SETGE                             = 0x0a, */
+/* 	    SQ_OP2_INST_SETNE                             = 0x0b, */
+/* 	    SQ_OP2_INST_SETE_DX10                         = 0x0c, */
+/* 	    SQ_OP2_INST_SETGT_DX10                        = 0x0d, */
+/* 	    SQ_OP2_INST_SETGE_DX10                        = 0x0e, */
+/* 	    SQ_OP2_INST_SETNE_DX10                        = 0x0f, */
+/* 	    SQ_OP2_INST_FRACT                             = 0x10, */
+/* 	    SQ_OP2_INST_TRUNC                             = 0x11, */
+/* 	    SQ_OP2_INST_CEIL                              = 0x12, */
+/* 	    SQ_OP2_INST_RNDNE                             = 0x13, */
+/* 	    SQ_OP2_INST_FLOOR                             = 0x14, */
+/* 	    SQ_OP2_INST_MOVA                              = 0x15, */
+/* 	    SQ_OP2_INST_MOVA_FLOOR                        = 0x16, */
+/* 	    SQ_OP2_INST_MOVA_INT                          = 0x18, */
+/* 	    SQ_OP2_INST_MOV                               = 0x19, */
+/* 	    SQ_OP2_INST_NOP                               = 0x1a, */
+/* 	    SQ_OP2_INST_PRED_SETGT_UINT                   = 0x1e, */
+/* 	    SQ_OP2_INST_PRED_SETGE_UINT                   = 0x1f, */
+/* 	    SQ_OP2_INST_PRED_SETE                         = 0x20, */
+/* 	    SQ_OP2_INST_PRED_SETGT                        = 0x21, */
+/* 	    SQ_OP2_INST_PRED_SETGE                        = 0x22, */
+/* 	    SQ_OP2_INST_PRED_SETNE                        = 0x23, */
+/* 	    SQ_OP2_INST_PRED_SET_INV                      = 0x24, */
+/* 	    SQ_OP2_INST_PRED_SET_POP                      = 0x25, */
+/* 	    SQ_OP2_INST_PRED_SET_CLR                      = 0x26, */
+/* 	    SQ_OP2_INST_PRED_SET_RESTORE                  = 0x27, */
+/* 	    SQ_OP2_INST_PRED_SETE_PUSH                    = 0x28, */
+/* 	    SQ_OP2_INST_PRED_SETGT_PUSH                   = 0x29, */
+/* 	    SQ_OP2_INST_PRED_SETGE_PUSH                   = 0x2a, */
+/* 	    SQ_OP2_INST_PRED_SETNE_PUSH                   = 0x2b, */
+/* 	    SQ_OP2_INST_KILLE                             = 0x2c, */
+/* 	    SQ_OP2_INST_KILLGT                            = 0x2d, */
+/* 	    SQ_OP2_INST_KILLGE                            = 0x2e, */
+/* 	    SQ_OP2_INST_KILLNE                            = 0x2f, */
+/* 	    SQ_OP2_INST_AND_INT                           = 0x30, */
+/* 	    SQ_OP2_INST_OR_INT                            = 0x31, */
+/* 	    SQ_OP2_INST_XOR_INT                           = 0x32, */
+/* 	    SQ_OP2_INST_NOT_INT                           = 0x33, */
+/* 	    SQ_OP2_INST_ADD_INT                           = 0x34, */
+/* 	    SQ_OP2_INST_SUB_INT                           = 0x35, */
+/* 	    SQ_OP2_INST_MAX_INT                           = 0x36, */
+/* 	    SQ_OP2_INST_MIN_INT                           = 0x37, */
+/* 	    SQ_OP2_INST_MAX_UINT                          = 0x38, */
+/* 	    SQ_OP2_INST_MIN_UINT                          = 0x39, */
+/* 	    SQ_OP2_INST_SETE_INT                          = 0x3a, */
+/* 	    SQ_OP2_INST_SETGT_INT                         = 0x3b, */
+/* 	    SQ_OP2_INST_SETGE_INT                         = 0x3c, */
+/* 	    SQ_OP2_INST_SETNE_INT                         = 0x3d, */
+/* 	    SQ_OP2_INST_SETGT_UINT                        = 0x3e, */
+/* 	    SQ_OP2_INST_SETGE_UINT                        = 0x3f, */
+/* 	    SQ_OP2_INST_KILLGT_UINT                       = 0x40, */
+/* 	    SQ_OP2_INST_KILLGE_UINT                       = 0x41, */
+/* 	    SQ_OP2_INST_PRED_SETE_INT                     = 0x42, */
+/* 	    SQ_OP2_INST_PRED_SETGT_INT                    = 0x43, */
+/* 	    SQ_OP2_INST_PRED_SETGE_INT                    = 0x44, */
+/* 	    SQ_OP2_INST_PRED_SETNE_INT                    = 0x45, */
+/* 	    SQ_OP2_INST_KILLE_INT                         = 0x46, */
+/* 	    SQ_OP2_INST_KILLGT_INT                        = 0x47, */
+/* 	    SQ_OP2_INST_KILLGE_INT                        = 0x48, */
+/* 	    SQ_OP2_INST_KILLNE_INT                        = 0x49, */
+/* 	    SQ_OP2_INST_PRED_SETE_PUSH_INT                = 0x4a, */
+/* 	    SQ_OP2_INST_PRED_SETGT_PUSH_INT               = 0x4b, */
+/* 	    SQ_OP2_INST_PRED_SETGE_PUSH_INT               = 0x4c, */
+/* 	    SQ_OP2_INST_PRED_SETNE_PUSH_INT               = 0x4d, */
+/* 	    SQ_OP2_INST_PRED_SETLT_PUSH_INT               = 0x4e, */
+/* 	    SQ_OP2_INST_PRED_SETLE_PUSH_INT               = 0x4f, */
+/* 	    SQ_OP2_INST_DOT4                              = 0x50, */
+/* 	    SQ_OP2_INST_DOT4_IEEE                         = 0x51, */
+/* 	    SQ_OP2_INST_CUBE                              = 0x52, */
+/* 	    SQ_OP2_INST_MAX4                              = 0x53, */
+/* 	    SQ_OP2_INST_MOVA_GPR_INT                      = 0x60, */
+/* 	    SQ_OP2_INST_EXP_IEEE                          = 0x61, */
+/* 	    SQ_OP2_INST_LOG_CLAMPED                       = 0x62, */
+/* 	    SQ_OP2_INST_LOG_IEEE                          = 0x63, */
+/* 	    SQ_OP2_INST_RECIP_CLAMPED                     = 0x64, */
+/* 	    SQ_OP2_INST_RECIP_FF                          = 0x65, */
+/* 	    SQ_OP2_INST_RECIP_IEEE                        = 0x66, */
+/* 	    SQ_OP2_INST_RECIPSQRT_CLAMPED                 = 0x67, */
+/* 	    SQ_OP2_INST_RECIPSQRT_FF                      = 0x68, */
+/* 	    SQ_OP2_INST_RECIPSQRT_IEEE                    = 0x69, */
+/* 	    SQ_OP2_INST_SQRT_IEEE                         = 0x6a, */
+/* 	    SQ_OP2_INST_FLT_TO_INT                        = 0x6b, */
+/* 	    SQ_OP2_INST_INT_TO_FLT                        = 0x6c, */
+/* 	    SQ_OP2_INST_UINT_TO_FLT                       = 0x6d, */
+/* 	    SQ_OP2_INST_SIN                               = 0x6e, */
+/* 	    SQ_OP2_INST_COS                               = 0x6f, */
+/* 	    SQ_OP2_INST_ASHR_INT                          = 0x70, */
+/* 	    SQ_OP2_INST_LSHR_INT                          = 0x71, */
+/* 	    SQ_OP2_INST_LSHL_INT                          = 0x72, */
+/* 	    SQ_OP2_INST_MULLO_INT                         = 0x73, */
+/* 	    SQ_OP2_INST_MULHI_INT                         = 0x74, */
+/* 	    SQ_OP2_INST_MULLO_UINT                        = 0x75, */
+/* 	    SQ_OP2_INST_MULHI_UINT                        = 0x76, */
+/* 	    SQ_OP2_INST_RECIP_INT                         = 0x77, */
+/* 	    SQ_OP2_INST_RECIP_UINT                        = 0x78, */
+/* 	    SQ_OP2_INST_FLT_TO_UINT                       = 0x79, */
+    SQ_CF_WORD1                                           = 0x00008dfc,
+	POP_COUNT_mask                                    = 0x07 << 0,
+	POP_COUNT_shift                                   = 0,
+	CF_CONST_mask                                     = 0x1f << 3,
+	CF_CONST_shift                                    = 3,
+	COND_mask                                         = 0x03 << 8,
+	COND_shift                                        = 8,
+	    SQ_CF_COND_ACTIVE                             = 0x00,
+	    SQ_CF_COND_FALSE                              = 0x01,
+	    SQ_CF_COND_BOOL                               = 0x02,
+	    SQ_CF_COND_NOT_BOOL                           = 0x03,
+	SQ_CF_WORD1__COUNT_mask                           = 0x07 << 10,
+	SQ_CF_WORD1__COUNT_shift                          = 10,
+	CALL_COUNT_mask                                   = 0x3f << 13,
+	CALL_COUNT_shift                                  = 13,
+	COUNT_3_bit                                       = 1 << 19,
+/* 	END_OF_PROGRAM_bit                                = 1 << 21, */
+/* 	VALID_PIXEL_MODE_bit                              = 1 << 22, */
+	SQ_CF_WORD1__CF_INST_mask                         = 0x7f << 23,
+	SQ_CF_WORD1__CF_INST_shift                        = 23,
+	    SQ_CF_INST_NOP                                = 0x00,
+	    SQ_CF_INST_TEX                                = 0x01,
+	    SQ_CF_INST_VTX                                = 0x02,
+	    SQ_CF_INST_VTX_TC                             = 0x03,
+	    SQ_CF_INST_LOOP_START                         = 0x04,
+	    SQ_CF_INST_LOOP_END                           = 0x05,
+	    SQ_CF_INST_LOOP_START_DX10                    = 0x06,
+	    SQ_CF_INST_LOOP_START_NO_AL                   = 0x07,
+	    SQ_CF_INST_LOOP_CONTINUE                      = 0x08,
+	    SQ_CF_INST_LOOP_BREAK                         = 0x09,
+	    SQ_CF_INST_JUMP                               = 0x0a,
+	    SQ_CF_INST_PUSH                               = 0x0b,
+	    SQ_CF_INST_PUSH_ELSE                          = 0x0c,
+	    SQ_CF_INST_ELSE                               = 0x0d,
+	    SQ_CF_INST_POP                                = 0x0e,
+	    SQ_CF_INST_POP_JUMP                           = 0x0f,
+	    SQ_CF_INST_POP_PUSH                           = 0x10,
+	    SQ_CF_INST_POP_PUSH_ELSE                      = 0x11,
+	    SQ_CF_INST_CALL                               = 0x12,
+	    SQ_CF_INST_CALL_FS                            = 0x13,
+	    SQ_CF_INST_RETURN                             = 0x14,
+	    SQ_CF_INST_EMIT_VERTEX                        = 0x15,
+	    SQ_CF_INST_EMIT_CUT_VERTEX                    = 0x16,
+	    SQ_CF_INST_CUT_VERTEX                         = 0x17,
+	    SQ_CF_INST_KILL                               = 0x18,
+/* 	WHOLE_QUAD_MODE_bit                               = 1 << 30, */
+/* 	BARRIER_bit                                       = 1 << 31, */
+    SQ_VTX_WORD1_SEM                                      = 0x00008dfc,
+	SEMANTIC_ID_mask                                  = 0xff << 0,
+	SEMANTIC_ID_shift                                 = 0,
+    SQ_TEX_WORD0                                          = 0x00008dfc,
+	TEX_INST_mask                                     = 0x1f << 0,
+	TEX_INST_shift                                    = 0,
+	    SQ_TEX_INST_VTX_FETCH                         = 0x00,
+	    SQ_TEX_INST_VTX_SEMANTIC                      = 0x01,
+	    SQ_TEX_INST_LD                                = 0x03,
+	    SQ_TEX_INST_GET_TEXTURE_RESINFO               = 0x04,
+	    SQ_TEX_INST_GET_NUMBER_OF_SAMPLES             = 0x05,
+	    SQ_TEX_INST_GET_LOD                           = 0x06,
+	    SQ_TEX_INST_GET_GRADIENTS_H                   = 0x07,
+	    SQ_TEX_INST_GET_GRADIENTS_V                   = 0x08,
+	    SQ_TEX_INST_GET_LERP                          = 0x09,
+	    SQ_TEX_INST_RESERVED_10                       = 0x0a,
+	    SQ_TEX_INST_SET_GRADIENTS_H                   = 0x0b,
+	    SQ_TEX_INST_SET_GRADIENTS_V                   = 0x0c,
+	    SQ_TEX_INST_PASS                              = 0x0d,
+	    X_Z_SET_INDEX_FOR_ARRAY_OF_CUBEMAPS           = 0x0e,
+	    SQ_TEX_INST_SAMPLE                            = 0x10,
+	    SQ_TEX_INST_SAMPLE_L                          = 0x11,
+	    SQ_TEX_INST_SAMPLE_LB                         = 0x12,
+	    SQ_TEX_INST_SAMPLE_LZ                         = 0x13,
+	    SQ_TEX_INST_SAMPLE_G                          = 0x14,
+	    SQ_TEX_INST_SAMPLE_G_L                        = 0x15,
+	    SQ_TEX_INST_SAMPLE_G_LB                       = 0x16,
+	    SQ_TEX_INST_SAMPLE_G_LZ                       = 0x17,
+	    SQ_TEX_INST_SAMPLE_C                          = 0x18,
+	    SQ_TEX_INST_SAMPLE_C_L                        = 0x19,
+	    SQ_TEX_INST_SAMPLE_C_LB                       = 0x1a,
+	    SQ_TEX_INST_SAMPLE_C_LZ                       = 0x1b,
+	    SQ_TEX_INST_SAMPLE_C_G                        = 0x1c,
+	    SQ_TEX_INST_SAMPLE_C_G_L                      = 0x1d,
+	    SQ_TEX_INST_SAMPLE_C_G_LB                     = 0x1e,
+	    SQ_TEX_INST_SAMPLE_C_G_LZ                     = 0x1f,
+	BC_FRAC_MODE_bit                                  = 1 << 5,
+/* 	FETCH_WHOLE_QUAD_bit                              = 1 << 7, */
+	RESOURCE_ID_mask                                  = 0xff << 8,
+	RESOURCE_ID_shift                                 = 8,
+/* 	SRC_GPR_mask                                      = 0x7f << 16, */
+/* 	SRC_GPR_shift                                     = 16, */
+/* 	SRC_REL_bit                                       = 1 << 23, */
+	SQ_TEX_WORD0__ALT_CONST_bit                       = 1 << 24,
+    SQ_VTX_WORD1_GPR                                      = 0x00008dfc,
+	SQ_VTX_WORD1_GPR__DST_GPR_mask                    = 0x7f << 0,
+	SQ_VTX_WORD1_GPR__DST_GPR_shift                   = 0,
+	SQ_VTX_WORD1_GPR__DST_REL_bit                     = 1 << 7,
+    SQ_ALU_WORD0                                          = 0x00008dfc,
+	SRC0_SEL_mask                                     = 0x1ff << 0,
+	SRC0_SEL_shift                                    = 0,
+/* 	    SQ_ALU_SRC_0                                  = 0xf8, */
+/* 	    SQ_ALU_SRC_1                                  = 0xf9, */
+/* 	    SQ_ALU_SRC_1_INT                              = 0xfa, */
+/* 	    SQ_ALU_SRC_M_1_INT                            = 0xfb, */
+/* 	    SQ_ALU_SRC_0_5                                = 0xfc, */
+/* 	    SQ_ALU_SRC_LITERAL                            = 0xfd, */
+/* 	    SQ_ALU_SRC_PV                                 = 0xfe, */
+/* 	    SQ_ALU_SRC_PS                                 = 0xff, */
+	SRC0_REL_bit                                      = 1 << 9,
+	SRC0_CHAN_mask                                    = 0x03 << 10,
+	SRC0_CHAN_shift                                   = 10,
+/* 	    SQ_CHAN_X                                     = 0x00, */
+/* 	    SQ_CHAN_Y                                     = 0x01, */
+/* 	    SQ_CHAN_Z                                     = 0x02, */
+/* 	    SQ_CHAN_W                                     = 0x03, */
+	SRC0_NEG_bit                                      = 1 << 12,
+	SRC1_SEL_mask                                     = 0x1ff << 13,
+	SRC1_SEL_shift                                    = 13,
+/* 	    SQ_ALU_SRC_0                                  = 0xf8, */
+/* 	    SQ_ALU_SRC_1                                  = 0xf9, */
+/* 	    SQ_ALU_SRC_1_INT                              = 0xfa, */
+/* 	    SQ_ALU_SRC_M_1_INT                            = 0xfb, */
+/* 	    SQ_ALU_SRC_0_5                                = 0xfc, */
+/* 	    SQ_ALU_SRC_LITERAL                            = 0xfd, */
+/* 	    SQ_ALU_SRC_PV                                 = 0xfe, */
+/* 	    SQ_ALU_SRC_PS                                 = 0xff, */
+	SRC1_REL_bit                                      = 1 << 22,
+	SRC1_CHAN_mask                                    = 0x03 << 23,
+	SRC1_CHAN_shift                                   = 23,
+/* 	    SQ_CHAN_X                                     = 0x00, */
+/* 	    SQ_CHAN_Y                                     = 0x01, */
+/* 	    SQ_CHAN_Z                                     = 0x02, */
+/* 	    SQ_CHAN_W                                     = 0x03, */
+	SRC1_NEG_bit                                      = 1 << 25,
+	INDEX_MODE_mask                                   = 0x07 << 26,
+	INDEX_MODE_shift                                  = 26,
+	    SQ_INDEX_AR_X                                 = 0x00,
+	    SQ_INDEX_AR_Y                                 = 0x01,
+	    SQ_INDEX_AR_Z                                 = 0x02,
+	    SQ_INDEX_AR_W                                 = 0x03,
+	    SQ_INDEX_LOOP                                 = 0x04,
+	PRED_SEL_mask                                     = 0x03 << 29,
+	PRED_SEL_shift                                    = 29,
+	    SQ_PRED_SEL_OFF                               = 0x00,
+	    SQ_PRED_SEL_ZERO                              = 0x02,
+	    SQ_PRED_SEL_ONE                               = 0x03,
+	LAST_bit                                          = 1 << 31,
+    SX_EXPORT_BUFFER_SIZES                                = 0x0000900c,
+	COLOR_BUFFER_SIZE_mask                            = 0xff << 0,
+	COLOR_BUFFER_SIZE_shift                           = 0,
+	POSITION_BUFFER_SIZE_mask                         = 0xff << 8,
+	POSITION_BUFFER_SIZE_shift                        = 8,
+	SMX_BUFFER_SIZE_mask                              = 0xff << 16,
+	SMX_BUFFER_SIZE_shift                             = 16,
+    SX_MEMORY_EXPORT_BASE                                 = 0x00009010,
+    SX_MEMORY_EXPORT_SIZE                                 = 0x00009014,
+    SPI_CONFIG_CNTL                                       = 0x00009100,
+	GPR_WRITE_PRIORITY_mask                           = 0x1f << 0,
+	GPR_WRITE_PRIORITY_shift                          = 0,
+	    X_PRIORITY_ORDER                              = 0x00,
+	    X_PRIORITY_ORDER_VS                           = 0x01,
+	DISABLE_INTERP_1_bit                              = 1 << 5,
+	DEBUG_THREAD_TYPE_SEL_mask                        = 0x03 << 6,
+	DEBUG_THREAD_TYPE_SEL_shift                       = 6,
+	DEBUG_GROUP_SEL_mask                              = 0x1f << 8,
+	DEBUG_GROUP_SEL_shift                             = 8,
+	DEBUG_GRBM_OVERRIDE_bit                           = 1 << 13,
+    SPI_CONFIG_CNTL_1                                     = 0x0000913c,
+	VTX_DONE_DELAY_mask                               = 0x0f << 0,
+	VTX_DONE_DELAY_shift                              = 0,
+	    X_DELAY_10_CLKS                               = 0x00,
+	    X_DELAY_11_CLKS                               = 0x01,
+	    X_DELAY_12_CLKS                               = 0x02,
+	    X_DELAY_13_CLKS                               = 0x03,
+	    X_DELAY_14_CLKS                               = 0x04,
+	    X_DELAY_15_CLKS                               = 0x05,
+	    X_DELAY_16_CLKS                               = 0x06,
+	    X_DELAY_17_CLKS                               = 0x07,
+	    X_DELAY_2_CLKS                                = 0x08,
+	    X_DELAY_3_CLKS                                = 0x09,
+	    X_DELAY_4_CLKS                                = 0x0a,
+	    X_DELAY_5_CLKS                                = 0x0b,
+	    X_DELAY_6_CLKS                                = 0x0c,
+	    X_DELAY_7_CLKS                                = 0x0d,
+	    X_DELAY_8_CLKS                                = 0x0e,
+	    X_DELAY_9_CLKS                                = 0x0f,
+	INTERP_ONE_PRIM_PER_ROW_bit                       = 1 << 4,
+    TD_FILTER4                                            = 0x00009400,
+	WEIGHT_1_mask                                     = 0x7ff << 0,
+	WEIGHT_1_shift                                    = 0,
+	WEIGHT_0_mask                                     = 0x7ff << 11,
+	WEIGHT_0_shift                                    = 11,
+	WEIGHT_PAIR_bit                                   = 1 << 22,
+	PHASE_mask                                        = 0x0f << 23,
+	PHASE_shift                                       = 23,
+	DIRECTION_bit                                     = 1 << 27,
+    TD_FILTER4_1                                          = 0x00009404,
+	TD_FILTER4_1_num                                  = 35,
+/* 	WEIGHT_1_mask                                     = 0x7ff << 0, */
+/* 	WEIGHT_1_shift                                    = 0, */
+/* 	WEIGHT_0_mask                                     = 0x7ff << 11, */
+/* 	WEIGHT_0_shift                                    = 11, */
+    TD_CNTL                                               = 0x00009490,
+	SYNC_PHASE_SH_mask                                = 0x03 << 0,
+	SYNC_PHASE_SH_shift                               = 0,
+	SYNC_PHASE_VC_SMX_mask                            = 0x03 << 4,
+	SYNC_PHASE_VC_SMX_shift                           = 4,
+    TD0_CNTL                                              = 0x00009494,
+	TD0_CNTL_num                                      = 4,
+	ID_OVERRIDE_mask                                  = 0x03 << 28,
+	ID_OVERRIDE_shift                                 = 28,
+    TD0_STATUS                                            = 0x000094a4,
+	TD0_STATUS_num                                    = 4,
+	BUSY_bit                                          = 1 << 31,
+    TA_CNTL                                               = 0x00009504,
+	GRADIENT_CREDIT_mask                              = 0x1f << 0,
+	GRADIENT_CREDIT_shift                             = 0,
+	WALKER_CREDIT_mask                                = 0x1f << 8,
+	WALKER_CREDIT_shift                               = 8,
+	ALIGNER_CREDIT_mask                               = 0x1f << 16,
+	ALIGNER_CREDIT_shift                              = 16,
+	TD_FIFO_CREDIT_mask                               = 0x3ff << 22,
+	TD_FIFO_CREDIT_shift                              = 22,
+    TA_CNTL_AUX                                           = 0x00009508,
+	DISABLE_CUBE_WRAP_bit                             = 1 << 0,
+	SYNC_GRADIENT_bit                                 = 1 << 24,
+	SYNC_WALKER_bit                                   = 1 << 25,
+	SYNC_ALIGNER_bit                                  = 1 << 26,
+	BILINEAR_PRECISION_bit                            = 1 << 31,
+    TA0_CNTL                                              = 0x00009510,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA1_CNTL                                              = 0x00009514,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA2_CNTL                                              = 0x00009518,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA3_CNTL                                              = 0x0000951c,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA0_STATUS                                            = 0x00009520,
+	FG_PFIFO_EMPTYB_bit                               = 1 << 12,
+	FG_LFIFO_EMPTYB_bit                               = 1 << 13,
+	FG_SFIFO_EMPTYB_bit                               = 1 << 14,
+	FL_PFIFO_EMPTYB_bit                               = 1 << 16,
+	FL_LFIFO_EMPTYB_bit                               = 1 << 17,
+	FL_SFIFO_EMPTYB_bit                               = 1 << 18,
+	FA_PFIFO_EMPTYB_bit                               = 1 << 20,
+	FA_LFIFO_EMPTYB_bit                               = 1 << 21,
+	FA_SFIFO_EMPTYB_bit                               = 1 << 22,
+	IN_BUSY_bit                                       = 1 << 24,
+	FG_BUSY_bit                                       = 1 << 25,
+	FL_BUSY_bit                                       = 1 << 27,
+	TA_BUSY_bit                                       = 1 << 28,
+	FA_BUSY_bit                                       = 1 << 29,
+	AL_BUSY_bit                                       = 1 << 30,
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA1_STATUS                                            = 0x00009524,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA2_STATUS                                            = 0x00009528,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA3_STATUS                                            = 0x0000952c,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TC_STATUS                                             = 0x00009600,
+	TC_BUSY_bit                                       = 1 << 0,
+    TC_INVALIDATE                                         = 0x00009604,
+	START_bit                                         = 1 << 0,
+    TC_CNTL                                               = 0x00009608,
+	FORCE_HIT_bit                                     = 1 << 0,
+	FORCE_MISS_bit                                    = 1 << 1,
+	L2_SIZE_mask                                      = 0x0f << 5,
+	L2_SIZE_shift                                     = 5,
+	    _256K                                         = 0x00,
+	    _224K                                         = 0x01,
+	    _192K                                         = 0x02,
+	    _160K                                         = 0x03,
+	    _128K                                         = 0x04,
+	    _96K                                          = 0x05,
+	    _64K                                          = 0x06,
+	    _32K                                          = 0x07,
+	L2_DISABLE_LATE_HIT_bit                           = 1 << 9,
+	DISABLE_VERT_PERF_bit                             = 1 << 10,
+	DISABLE_INVAL_BUSY_bit                            = 1 << 11,
+	DISABLE_INVAL_SAME_SURFACE_bit                    = 1 << 12,
+	PARTITION_MODE_mask                               = 0x03 << 13,
+	PARTITION_MODE_shift                              = 13,
+	    X_VERTEX                                      = 0x00,
+	MISS_ARB_MODE_bit                                 = 1 << 15,
+	HIT_ARB_MODE_bit                                  = 1 << 16,
+	DISABLE_WRITE_DELAY_bit                           = 1 << 17,
+	HIT_FIFO_DEPTH_bit                                = 1 << 18,
+    VC_CNTL                                               = 0x00009700,
+	L2_INVALIDATE_bit                                 = 1 << 0,
+	RESERVED_bit                                      = 1 << 1,
+	CC_FORCE_MISS_bit                                 = 1 << 2,
+	MI_CHAN_SEL_mask                                  = 0x03 << 3,
+	MI_CHAN_SEL_shift                                 = 3,
+	    X_MC0_USES_CH_0_1                             = 0x00,
+	    X_MC0_USES_CH_0_3                             = 0x01,
+	    X_VC_MC0_IS_ACTIVE                            = 0x02,
+	    X_VC_MC1_IS_DISABLED                          = 0x03,
+	MI_STEER_DISABLE_bit                              = 1 << 5,
+	MI_CREDIT_CTR_mask                                = 0x0f << 6,
+	MI_CREDIT_CTR_shift                               = 6,
+	MI_CREDIT_WE_bit                                  = 1 << 10,
+	MI_REQ_STALL_THLD_mask                            = 0x07 << 11,
+	MI_REQ_STALL_THLD_shift                           = 11,
+	    X_LATENCY_EXCEEDS_399_CLOCKS                  = 0x00,
+	    X_LATENCY_EXCEEDS_415_CLOCKS                  = 0x01,
+	    X_LATENCY_EXCEEDS_431_CLOCKS                  = 0x02,
+	    X_LATENCY_EXCEEDS_447_CLOCKS                  = 0x03,
+	    X_LATENCY_EXCEEDS_463_CLOCKS                  = 0x04,
+	    X_LATENCY_EXCEEDS_479_CLOCKS                  = 0x05,
+	    X_LATENCY_EXCEEDS_495_CLOCKS                  = 0x06,
+	    X_LATENCY_EXCEEDS_511_CLOCKS                  = 0x07,
+	VC_CNTL__MI_TIMESTAMP_RES_mask                    = 0x1f << 14,
+	VC_CNTL__MI_TIMESTAMP_RES_shift                   = 14,
+	    X_1X_SYSTEM_CLOCK                             = 0x00,
+	    X_2X_SYSTEM_CLOCK                             = 0x01,
+	    X_4X_SYSTEM_CLOCK                             = 0x02,
+	    X_8X_SYSTEM_CLOCK                             = 0x03,
+	    X_16X_SYSTEM_CLOCK                            = 0x04,
+	    X_32X_SYSTEM_CLOCK                            = 0x05,
+	    X_64X_SYSTEM_CLOCK                            = 0x06,
+	    X_128X_SYSTEM_CLOCK                           = 0x07,
+	    X_256X_SYSTEM_CLOCK                           = 0x08,
+	    X_512X_SYSTEM_CLOCK                           = 0x09,
+	    X_1024X_SYSTEM_CLOCK                          = 0x0a,
+	    X_2048X_SYSTEM_CLOCK                          = 0x0b,
+	    X_4092X_SYSTEM_CLOCK                          = 0x0c,
+	    X_8192X_SYSTEM_CLOCK                          = 0x0d,
+	    X_16384X_SYSTEM_CLOCK                         = 0x0e,
+	    X_32768X_SYSTEM_CLOCK                         = 0x0f,
+    VC_CNTL_STATUS                                        = 0x00009704,
+	RP_BUSY_bit                                       = 1 << 0,
+	RG_BUSY_bit                                       = 1 << 1,
+	VC_BUSY_bit                                       = 1 << 2,
+	CLAMP_DETECT_bit                                  = 1 << 3,
+    VC_CONFIG                                             = 0x00009718,
+	WRITE_DIS_bit                                     = 1 << 0,
+	GPR_DATA_PHASE_ADJ_mask                           = 0x07 << 1,
+	GPR_DATA_PHASE_ADJ_shift                          = 1,
+	    X_LATENCY_BASE_0_CYCLES                       = 0x00,
+	    X_LATENCY_BASE_1_CYCLES                       = 0x01,
+	    X_LATENCY_BASE_2_CYCLES                       = 0x02,
+	    X_LATENCY_BASE_3_CYCLES                       = 0x03,
+	TD_SIMD_SYNC_ADJ_mask                             = 0x07 << 4,
+	TD_SIMD_SYNC_ADJ_shift                            = 4,
+	    X_0_CYCLES_DELAY                              = 0x00,
+	    X_1_CYCLES_DELAY                              = 0x01,
+	    X_2_CYCLES_DELAY                              = 0x02,
+	    X_3_CYCLES_DELAY                              = 0x03,
+	    X_4_CYCLES_DELAY                              = 0x04,
+	    X_5_CYCLES_DELAY                              = 0x05,
+	    X_6_CYCLES_DELAY                              = 0x06,
+	    X_7_CYCLES_DELAY                              = 0x07,
+    SMX_DC_CTL0                                           = 0x0000a020,
+	WR_GATHER_STREAM0_bit                             = 1 << 0,
+	WR_GATHER_STREAM1_bit                             = 1 << 1,
+	WR_GATHER_STREAM2_bit                             = 1 << 2,
+	WR_GATHER_STREAM3_bit                             = 1 << 3,
+	WR_GATHER_SCRATCH_bit                             = 1 << 4,
+	WR_GATHER_REDUC_BUF_bit                           = 1 << 5,
+	WR_GATHER_RING_BUF_bit                            = 1 << 6,
+	WR_GATHER_F_BUF_bit                               = 1 << 7,
+	DISABLE_CACHES_bit                                = 1 << 8,
+	AUTO_FLUSH_INVAL_EN_bit                           = 1 << 10,
+	AUTO_FLUSH_EN_bit                                 = 1 << 11,
+	AUTO_FLUSH_CNT_mask                               = 0xffff << 12,
+	AUTO_FLUSH_CNT_shift                              = 12,
+	MC_RD_STALL_FACTOR_mask                           = 0x03 << 28,
+	MC_RD_STALL_FACTOR_shift                          = 28,
+	MC_WR_STALL_FACTOR_mask                           = 0x03 << 30,
+	MC_WR_STALL_FACTOR_shift                          = 30,
+    SMX_DC_CTL1                                           = 0x0000a024,
+	OP_FIFO_SKID_mask                                 = 0x7f << 0,
+	OP_FIFO_SKID_shift                                = 0,
+	CACHE_LINE_SIZE_bit                               = 1 << 8,
+	MULTI_FLUSH_MODE_bit                              = 1 << 9,
+	MULTI_FLUSH_REQ_ABORT_IDX_FIFO_SKID_mask          = 0x0f << 10,
+	MULTI_FLUSH_REQ_ABORT_IDX_FIFO_SKID_shift         = 10,
+	DISABLE_WR_GATHER_RD_HIT_FORCE_EVICT_bit          = 1 << 16,
+	DISABLE_WR_GATHER_RD_HIT_COMP_VLDS_CHECK_bit      = 1 << 17,
+	DISABLE_FLUSH_ES_ALSO_INVALS_bit                  = 1 << 18,
+	DISABLE_FLUSH_GS_ALSO_INVALS_bit                  = 1 << 19,
+    SMX_DC_CTL2                                           = 0x0000a028,
+	INVALIDATE_CACHES_bit                             = 1 << 0,
+	CACHES_INVALID_bit                                = 1 << 1,
+	CACHES_DIRTY_bit                                  = 1 << 2,
+	FLUSH_ALL_bit                                     = 1 << 4,
+	FLUSH_GS_THREADS_bit                              = 1 << 8,
+	FLUSH_ES_THREADS_bit                              = 1 << 9,
+    SMX_DC_MC_INTF_CTL                                    = 0x0000a02c,
+	MC_RD_REQ_CRED_mask                               = 0xff << 0,
+	MC_RD_REQ_CRED_shift                              = 0,
+	MC_WR_REQ_CRED_mask                               = 0xff << 16,
+	MC_WR_REQ_CRED_shift                              = 16,
+    TD_PS_SAMPLER0_BORDER_RED                             = 0x0000a400,
+	TD_PS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_PS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_PS_SAMPLER0_BORDER_GREEN                           = 0x0000a404,
+	TD_PS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_PS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_PS_SAMPLER0_BORDER_BLUE                            = 0x0000a408,
+	TD_PS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_PS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_PS_SAMPLER0_BORDER_ALPHA                           = 0x0000a40c,
+	TD_PS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_PS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_VS_SAMPLER0_BORDER_RED                             = 0x0000a600,
+	TD_VS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_VS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_VS_SAMPLER0_BORDER_GREEN                           = 0x0000a604,
+	TD_VS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_VS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_VS_SAMPLER0_BORDER_BLUE                            = 0x0000a608,
+	TD_VS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_VS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_VS_SAMPLER0_BORDER_ALPHA                           = 0x0000a60c,
+	TD_VS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_VS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_GS_SAMPLER0_BORDER_RED                             = 0x0000a800,
+	TD_GS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_GS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_GS_SAMPLER0_BORDER_GREEN                           = 0x0000a804,
+	TD_GS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_GS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_GS_SAMPLER0_BORDER_BLUE                            = 0x0000a808,
+	TD_GS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_GS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_GS_SAMPLER0_BORDER_ALPHA                           = 0x0000a80c,
+	TD_GS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_GS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_PS_SAMPLER0_CLEARTYPE_KERNEL                       = 0x0000aa00,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL_num               = 18,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__WIDTH_mask       = 0x07 << 0,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__WIDTH_shift      = 0,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__HEIGHT_mask      = 0x07 << 3,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__HEIGHT_shift     = 3,
+    DB_DEPTH_SIZE                                         = 0x00028000,
+	PITCH_TILE_MAX_mask                               = 0x3ff << 0,
+	PITCH_TILE_MAX_shift                              = 0,
+	SLICE_TILE_MAX_mask                               = 0xfffff << 10,
+	SLICE_TILE_MAX_shift                              = 10,
+    DB_DEPTH_VIEW                                         = 0x00028004,
+	SLICE_START_mask                                  = 0x7ff << 0,
+	SLICE_START_shift                                 = 0,
+	SLICE_MAX_mask                                    = 0x7ff << 13,
+	SLICE_MAX_shift                                   = 13,
+    DB_DEPTH_BASE                                         = 0x0002800c,
+    DB_DEPTH_INFO                                         = 0x00028010,
+	DB_DEPTH_INFO__FORMAT_mask                        = 0x07 << 0,
+	DB_DEPTH_INFO__FORMAT_shift                       = 0,
+	    DEPTH_INVALID                                 = 0x00,
+	    DEPTH_16                                      = 0x01,
+	    DEPTH_X8_24                                   = 0x02,
+	    DEPTH_8_24                                    = 0x03,
+	    DEPTH_X8_24_FLOAT                             = 0x04,
+	    DEPTH_8_24_FLOAT                              = 0x05,
+	    DEPTH_32_FLOAT                                = 0x06,
+	    DEPTH_X24_8_32_FLOAT                          = 0x07,
+	DB_DEPTH_INFO__READ_SIZE_bit                      = 1 << 3,
+	DB_DEPTH_INFO__ARRAY_MODE_mask                    = 0x0f << 15,
+	DB_DEPTH_INFO__ARRAY_MODE_shift                   = 15,
+	    ARRAY_2D_TILED_THIN1                          = 0x04,
+	TILE_SURFACE_ENABLE_bit                           = 1 << 25,
+	TILE_COMPACT_bit                                  = 1 << 26,
+	ZRANGE_PRECISION_bit                              = 1 << 31,
+    DB_HTILE_DATA_BASE                                    = 0x00028014,
+    DB_STENCIL_CLEAR                                      = 0x00028028,
+	DB_STENCIL_CLEAR__CLEAR_mask                      = 0xff << 0,
+	DB_STENCIL_CLEAR__CLEAR_shift                     = 0,
+	MIN_mask                                          = 0xff << 16,
+	MIN_shift                                         = 16,
+    DB_DEPTH_CLEAR                                        = 0x0002802c,
+    PA_SC_SCREEN_SCISSOR_TL                               = 0x00028030,
+	PA_SC_SCREEN_SCISSOR_TL__TL_X_mask                = 0x7fff << 0,
+	PA_SC_SCREEN_SCISSOR_TL__TL_X_shift               = 0,
+	PA_SC_SCREEN_SCISSOR_TL__TL_Y_mask                = 0x7fff << 16,
+	PA_SC_SCREEN_SCISSOR_TL__TL_Y_shift               = 16,
+    PA_SC_SCREEN_SCISSOR_BR                               = 0x00028034,
+	PA_SC_SCREEN_SCISSOR_BR__BR_X_mask                = 0x7fff << 0,
+	PA_SC_SCREEN_SCISSOR_BR__BR_X_shift               = 0,
+	PA_SC_SCREEN_SCISSOR_BR__BR_Y_mask                = 0x7fff << 16,
+	PA_SC_SCREEN_SCISSOR_BR__BR_Y_shift               = 16,
+    CB_COLOR0_BASE                                        = 0x00028040,
+	CB_COLOR0_BASE_num                                = 8,
+    CB_COLOR0_SIZE                                        = 0x00028060,
+	CB_COLOR0_SIZE_num                                = 8,
+/* 	PITCH_TILE_MAX_mask                               = 0x3ff << 0, */
+/* 	PITCH_TILE_MAX_shift                              = 0, */
+/* 	SLICE_TILE_MAX_mask                               = 0xfffff << 10, */
+/* 	SLICE_TILE_MAX_shift                              = 10, */
+    CB_COLOR0_VIEW                                        = 0x00028080,
+	CB_COLOR0_VIEW_num                                = 8,
+/* 	SLICE_START_mask                                  = 0x7ff << 0, */
+/* 	SLICE_START_shift                                 = 0, */
+/* 	SLICE_MAX_mask                                    = 0x7ff << 13, */
+/* 	SLICE_MAX_shift                                   = 13, */
+    CB_COLOR0_INFO                                        = 0x000280a0,
+	CB_COLOR0_INFO_num                                = 8,
+	ENDIAN_mask                                       = 0x03 << 0,
+	ENDIAN_shift                                      = 0,
+	    ENDIAN_NONE                                   = 0x00,
+	    ENDIAN_8IN16                                  = 0x01,
+	    ENDIAN_8IN32                                  = 0x02,
+	    ENDIAN_8IN64                                  = 0x03,
+	CB_COLOR0_INFO__FORMAT_mask                       = 0x3f << 2,
+	CB_COLOR0_INFO__FORMAT_shift                      = 2,
+	    COLOR_INVALID                                 = 0x00,
+	    COLOR_8                                       = 0x01,
+	    COLOR_4_4                                     = 0x02,
+	    COLOR_3_3_2                                   = 0x03,
+	    COLOR_16                                      = 0x05,
+	    COLOR_16_FLOAT                                = 0x06,
+	    COLOR_8_8                                     = 0x07,
+	    COLOR_5_6_5                                   = 0x08,
+	    COLOR_6_5_5                                   = 0x09,
+	    COLOR_1_5_5_5                                 = 0x0a,
+	    COLOR_4_4_4_4                                 = 0x0b,
+	    COLOR_5_5_5_1                                 = 0x0c,
+	    COLOR_32                                      = 0x0d,
+	    COLOR_32_FLOAT                                = 0x0e,
+	    COLOR_16_16                                   = 0x0f,
+	    COLOR_16_16_FLOAT                             = 0x10,
+	    COLOR_8_24                                    = 0x11,
+	    COLOR_8_24_FLOAT                              = 0x12,
+	    COLOR_24_8                                    = 0x13,
+	    COLOR_24_8_FLOAT                              = 0x14,
+	    COLOR_10_11_11                                = 0x15,
+	    COLOR_10_11_11_FLOAT                          = 0x16,
+	    COLOR_11_11_10                                = 0x17,
+	    COLOR_11_11_10_FLOAT                          = 0x18,
+	    COLOR_2_10_10_10                              = 0x19,
+	    COLOR_8_8_8_8                                 = 0x1a,
+	    COLOR_10_10_10_2                              = 0x1b,
+	    COLOR_X24_8_32_FLOAT                          = 0x1c,
+	    COLOR_32_32                                   = 0x1d,
+	    COLOR_32_32_FLOAT                             = 0x1e,
+	    COLOR_16_16_16_16                             = 0x1f,
+	    COLOR_16_16_16_16_FLOAT                       = 0x20,
+	    COLOR_32_32_32_32                             = 0x22,
+	    COLOR_32_32_32_32_FLOAT                       = 0x23,
+	CB_COLOR0_INFO__ARRAY_MODE_mask                   = 0x0f << 8,
+	CB_COLOR0_INFO__ARRAY_MODE_shift                  = 8,
+	    ARRAY_LINEAR_GENERAL                          = 0x00,
+	    ARRAY_LINEAR_ALIGNED                          = 0x01,
+/* 	    ARRAY_2D_TILED_THIN1                          = 0x04, */
+	NUMBER_TYPE_mask                                  = 0x07 << 12,
+	NUMBER_TYPE_shift                                 = 12,
+	    NUMBER_UNORM                                  = 0x00,
+	    NUMBER_SNORM                                  = 0x01,
+	    NUMBER_USCALED                                = 0x02,
+	    NUMBER_SSCALED                                = 0x03,
+	    NUMBER_UINT                                   = 0x04,
+	    NUMBER_SINT                                   = 0x05,
+	    NUMBER_SRGB                                   = 0x06,
+	    NUMBER_FLOAT                                  = 0x07,
+	CB_COLOR0_INFO__READ_SIZE_bit                     = 1 << 15,
+	COMP_SWAP_mask                                    = 0x03 << 16,
+	COMP_SWAP_shift                                   = 16,
+	    SWAP_STD                                      = 0x00,
+	    SWAP_ALT                                      = 0x01,
+	    SWAP_STD_REV                                  = 0x02,
+	    SWAP_ALT_REV                                  = 0x03,
+	CB_COLOR0_INFO__TILE_MODE_mask                    = 0x03 << 18,
+	CB_COLOR0_INFO__TILE_MODE_shift                   = 18,
+	    TILE_DISABLE                                  = 0x00,
+	    TILE_CLEAR_ENABLE                             = 0x01,
+	    TILE_FRAG_ENABLE                              = 0x02,
+	BLEND_CLAMP_bit                                   = 1 << 20,
+	CLEAR_COLOR_bit                                   = 1 << 21,
+	BLEND_BYPASS_bit                                  = 1 << 22,
+	BLEND_FLOAT32_bit                                 = 1 << 23,
+	SIMPLE_FLOAT_bit                                  = 1 << 24,
+	CB_COLOR0_INFO__ROUND_MODE_bit                    = 1 << 25,
+/* 	TILE_COMPACT_bit                                  = 1 << 26, */
+	SOURCE_FORMAT_bit                                 = 1 << 27,
+    CB_COLOR0_TILE                                        = 0x000280c0,
+	CB_COLOR0_TILE_num                                = 8,
+    CB_COLOR0_FRAG                                        = 0x000280e0,
+	CB_COLOR0_FRAG_num                                = 8,
+    CB_COLOR0_MASK                                        = 0x00028100,
+	CB_COLOR0_MASK_num                                = 8,
+	CMASK_BLOCK_MAX_mask                              = 0xfff << 0,
+	CMASK_BLOCK_MAX_shift                             = 0,
+	FMASK_TILE_MAX_mask                               = 0xfffff << 12,
+	FMASK_TILE_MAX_shift                              = 12,
+    CB_CLEAR_RED                                          = 0x00028120,
+    CB_CLEAR_GREEN                                        = 0x00028124,
+    CB_CLEAR_BLUE                                         = 0x00028128,
+    CB_CLEAR_ALPHA                                        = 0x0002812c,
+    SQ_ALU_CONST_BUFFER_SIZE_PS_0                         = 0x00028140,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0__DATA_shift         = 0,
+    SQ_ALU_CONST_BUFFER_SIZE_VS_0                         = 0x00028180,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0__DATA_shift         = 0,
+    SQ_ALU_CONST_BUFFER_SIZE_GS_0                         = 0x000281c0,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0__DATA_shift         = 0,
+    PA_SC_WINDOW_OFFSET                                   = 0x00028200,
+	WINDOW_X_OFFSET_mask                              = 0x7fff << 0,
+	WINDOW_X_OFFSET_shift                             = 0,
+	WINDOW_Y_OFFSET_mask                              = 0x7fff << 16,
+	WINDOW_Y_OFFSET_shift                             = 16,
+    PA_SC_WINDOW_SCISSOR_TL                               = 0x00028204,
+	PA_SC_WINDOW_SCISSOR_TL__TL_X_mask                = 0x3fff << 0,
+	PA_SC_WINDOW_SCISSOR_TL__TL_X_shift               = 0,
+	PA_SC_WINDOW_SCISSOR_TL__TL_Y_mask                = 0x3fff << 16,
+	PA_SC_WINDOW_SCISSOR_TL__TL_Y_shift               = 16,
+	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31,
+    PA_SC_WINDOW_SCISSOR_BR                               = 0x00028208,
+	PA_SC_WINDOW_SCISSOR_BR__BR_X_mask                = 0x3fff << 0,
+	PA_SC_WINDOW_SCISSOR_BR__BR_X_shift               = 0,
+	PA_SC_WINDOW_SCISSOR_BR__BR_Y_mask                = 0x3fff << 16,
+	PA_SC_WINDOW_SCISSOR_BR__BR_Y_shift               = 16,
+    PA_SC_CLIPRECT_RULE                                   = 0x0002820c,
+	CLIP_RULE_mask                                    = 0xffff << 0,
+	CLIP_RULE_shift                                   = 0,
+    PA_SC_CLIPRECT_0_TL                                   = 0x00028210,
+	PA_SC_CLIPRECT_0_TL_num                           = 4,
+	PA_SC_CLIPRECT_0_TL_offset                        = 8,
+	PA_SC_CLIPRECT_0_TL__TL_X_mask                    = 0x3fff << 0,
+	PA_SC_CLIPRECT_0_TL__TL_X_shift                   = 0,
+	PA_SC_CLIPRECT_0_TL__TL_Y_mask                    = 0x3fff << 16,
+	PA_SC_CLIPRECT_0_TL__TL_Y_shift                   = 16,
+    PA_SC_CLIPRECT_0_BR                                   = 0x00028214,
+	PA_SC_CLIPRECT_0_BR_num                           = 4,
+	PA_SC_CLIPRECT_0_BR_offset                        = 8,
+	PA_SC_CLIPRECT_0_BR__BR_X_mask                    = 0x3fff << 0,
+	PA_SC_CLIPRECT_0_BR__BR_X_shift                   = 0,
+	PA_SC_CLIPRECT_0_BR__BR_Y_mask                    = 0x3fff << 16,
+	PA_SC_CLIPRECT_0_BR__BR_Y_shift                   = 16,
+    CB_TARGET_MASK                                        = 0x00028238,
+	TARGET0_ENABLE_mask                               = 0x0f << 0,
+	TARGET0_ENABLE_shift                              = 0,
+	TARGET1_ENABLE_mask                               = 0x0f << 4,
+	TARGET1_ENABLE_shift                              = 4,
+	TARGET2_ENABLE_mask                               = 0x0f << 8,
+	TARGET2_ENABLE_shift                              = 8,
+	TARGET3_ENABLE_mask                               = 0x0f << 12,
+	TARGET3_ENABLE_shift                              = 12,
+	TARGET4_ENABLE_mask                               = 0x0f << 16,
+	TARGET4_ENABLE_shift                              = 16,
+	TARGET5_ENABLE_mask                               = 0x0f << 20,
+	TARGET5_ENABLE_shift                              = 20,
+	TARGET6_ENABLE_mask                               = 0x0f << 24,
+	TARGET6_ENABLE_shift                              = 24,
+	TARGET7_ENABLE_mask                               = 0x0f << 28,
+	TARGET7_ENABLE_shift                              = 28,
+    CB_SHADER_MASK                                        = 0x0002823c,
+	OUTPUT0_ENABLE_mask                               = 0x0f << 0,
+	OUTPUT0_ENABLE_shift                              = 0,
+	OUTPUT1_ENABLE_mask                               = 0x0f << 4,
+	OUTPUT1_ENABLE_shift                              = 4,
+	OUTPUT2_ENABLE_mask                               = 0x0f << 8,
+	OUTPUT2_ENABLE_shift                              = 8,
+	OUTPUT3_ENABLE_mask                               = 0x0f << 12,
+	OUTPUT3_ENABLE_shift                              = 12,
+	OUTPUT4_ENABLE_mask                               = 0x0f << 16,
+	OUTPUT4_ENABLE_shift                              = 16,
+	OUTPUT5_ENABLE_mask                               = 0x0f << 20,
+	OUTPUT5_ENABLE_shift                              = 20,
+	OUTPUT6_ENABLE_mask                               = 0x0f << 24,
+	OUTPUT6_ENABLE_shift                              = 24,
+	OUTPUT7_ENABLE_mask                               = 0x0f << 28,
+	OUTPUT7_ENABLE_shift                              = 28,
+    PA_SC_GENERIC_SCISSOR_TL                              = 0x00028240,
+	PA_SC_GENERIC_SCISSOR_TL__TL_X_mask               = 0x3fff << 0,
+	PA_SC_GENERIC_SCISSOR_TL__TL_X_shift              = 0,
+	PA_SC_GENERIC_SCISSOR_TL__TL_Y_mask               = 0x3fff << 16,
+	PA_SC_GENERIC_SCISSOR_TL__TL_Y_shift              = 16,
+/* 	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31, */
+    PA_SC_GENERIC_SCISSOR_BR                              = 0x00028244,
+	PA_SC_GENERIC_SCISSOR_BR__BR_X_mask               = 0x3fff << 0,
+	PA_SC_GENERIC_SCISSOR_BR__BR_X_shift              = 0,
+	PA_SC_GENERIC_SCISSOR_BR__BR_Y_mask               = 0x3fff << 16,
+	PA_SC_GENERIC_SCISSOR_BR__BR_Y_shift              = 16,
+    PA_SC_VPORT_SCISSOR_0_TL                              = 0x00028250,
+	PA_SC_VPORT_SCISSOR_0_TL_num                      = 16,
+	PA_SC_VPORT_SCISSOR_0_TL_offset                   = 8,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask               = 0x3fff << 0,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift              = 0,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask               = 0x3fff << 16,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift              = 16,
+/* 	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31, */
+    PA_SC_VPORT_SCISSOR_0_BR                              = 0x00028254,
+	PA_SC_VPORT_SCISSOR_0_BR_num                      = 16,
+	PA_SC_VPORT_SCISSOR_0_BR_offset                   = 8,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask               = 0x3fff << 0,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift              = 0,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask               = 0x3fff << 16,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift              = 16,
+    PA_SC_VPORT_ZMIN_0                                    = 0x000282d0,
+	PA_SC_VPORT_ZMIN_0_num                            = 16,
+	PA_SC_VPORT_ZMIN_0_offset                         = 8,
+    PA_SC_VPORT_ZMAX_0                                    = 0x000282d4,
+	PA_SC_VPORT_ZMAX_0_num                            = 16,
+	PA_SC_VPORT_ZMAX_0_offset                         = 8,
+    SX_MISC                                               = 0x00028350,
+	MULTIPASS_bit                                     = 1 << 0,
+    SQ_VTX_SEMANTIC_0                                     = 0x00028380,
+	SQ_VTX_SEMANTIC_0_num                             = 32,
+/* 	SEMANTIC_ID_mask                                  = 0xff << 0, */
+/* 	SEMANTIC_ID_shift                                 = 0, */
+    VGT_MAX_VTX_INDX                                      = 0x00028400,
+    VGT_MIN_VTX_INDX                                      = 0x00028404,
+    VGT_INDX_OFFSET                                       = 0x00028408,
+    VGT_MULTI_PRIM_IB_RESET_INDX                          = 0x0002840c,
+    SX_ALPHA_TEST_CONTROL                                 = 0x00028410,
+	ALPHA_FUNC_mask                                   = 0x07 << 0,
+	ALPHA_FUNC_shift                                  = 0,
+	    REF_NEVER                                     = 0x00,
+	    REF_LESS                                      = 0x01,
+	    REF_EQUAL                                     = 0x02,
+	    REF_LEQUAL                                    = 0x03,
+	    REF_GREATER                                   = 0x04,
+	    REF_NOTEQUAL                                  = 0x05,
+	    REF_GEQUAL                                    = 0x06,
+	    REF_ALWAYS                                    = 0x07,
+	ALPHA_TEST_ENABLE_bit                             = 1 << 3,
+	ALPHA_TEST_BYPASS_bit                             = 1 << 8,
+    CB_BLEND_RED                                          = 0x00028414,
+    CB_BLEND_GREEN                                        = 0x00028418,
+    CB_BLEND_BLUE                                         = 0x0002841c,
+    CB_BLEND_ALPHA                                        = 0x00028420,
+    CB_FOG_RED                                            = 0x00028424,
+    CB_FOG_GREEN                                          = 0x00028428,
+    CB_FOG_BLUE                                           = 0x0002842c,
+    DB_STENCILREFMASK                                     = 0x00028430,
+	STENCILREF_mask                                   = 0xff << 0,
+	STENCILREF_shift                                  = 0,
+	STENCILMASK_mask                                  = 0xff << 8,
+	STENCILMASK_shift                                 = 8,
+	STENCILWRITEMASK_mask                             = 0xff << 16,
+	STENCILWRITEMASK_shift                            = 16,
+    DB_STENCILREFMASK_BF                                  = 0x00028434,
+	STENCILREF_BF_mask                                = 0xff << 0,
+	STENCILREF_BF_shift                               = 0,
+	STENCILMASK_BF_mask                               = 0xff << 8,
+	STENCILMASK_BF_shift                              = 8,
+	STENCILWRITEMASK_BF_mask                          = 0xff << 16,
+	STENCILWRITEMASK_BF_shift                         = 16,
+    SX_ALPHA_REF                                          = 0x00028438,
+    PA_CL_VPORT_XSCALE_0                                  = 0x0002843c,
+	PA_CL_VPORT_XSCALE_0_num                          = 16,
+	PA_CL_VPORT_XSCALE_0_offset                       = 24,
+    PA_CL_VPORT_XOFFSET_0                                 = 0x00028440,
+	PA_CL_VPORT_XOFFSET_0_num                         = 16,
+	PA_CL_VPORT_XOFFSET_0_offset                      = 24,
+    PA_CL_VPORT_YSCALE_0                                  = 0x00028444,
+	PA_CL_VPORT_YSCALE_0_num                          = 16,
+	PA_CL_VPORT_YSCALE_0_offset                       = 24,
+    PA_CL_VPORT_YOFFSET_0                                 = 0x00028448,
+	PA_CL_VPORT_YOFFSET_0_num                         = 16,
+	PA_CL_VPORT_YOFFSET_0_offset                      = 24,
+    PA_CL_VPORT_ZSCALE_0                                  = 0x0002844c,
+	PA_CL_VPORT_ZSCALE_0_num                          = 16,
+	PA_CL_VPORT_ZSCALE_0_offset                       = 24,
+    PA_CL_VPORT_ZOFFSET_0                                 = 0x00028450,
+	PA_CL_VPORT_ZOFFSET_0_num                         = 16,
+	PA_CL_VPORT_ZOFFSET_0_offset                      = 24,
+    SPI_VS_OUT_ID_0                                       = 0x00028614,
+	SPI_VS_OUT_ID_0_num                               = 10,
+	SEMANTIC_0_mask                                   = 0xff << 0,
+	SEMANTIC_0_shift                                  = 0,
+	SEMANTIC_1_mask                                   = 0xff << 8,
+	SEMANTIC_1_shift                                  = 8,
+	SEMANTIC_2_mask                                   = 0xff << 16,
+	SEMANTIC_2_shift                                  = 16,
+	SEMANTIC_3_mask                                   = 0xff << 24,
+	SEMANTIC_3_shift                                  = 24,
+    SPI_PS_INPUT_CNTL_0                                   = 0x00028644,
+	SPI_PS_INPUT_CNTL_0_num                           = 32,
+	SEMANTIC_mask                                     = 0xff << 0,
+	SEMANTIC_shift                                    = 0,
+	DEFAULT_VAL_mask                                  = 0x03 << 8,
+	DEFAULT_VAL_shift                                 = 8,
+	    X_0_0F                                        = 0x00,
+	FLAT_SHADE_bit                                    = 1 << 10,
+	SEL_CENTROID_bit                                  = 1 << 11,
+	SEL_LINEAR_bit                                    = 1 << 12,
+	CYL_WRAP_mask                                     = 0x0f << 13,
+	CYL_WRAP_shift                                    = 13,
+	PT_SPRITE_TEX_bit                                 = 1 << 17,
+	SEL_SAMPLE_bit                                    = 1 << 18,
+    SPI_VS_OUT_CONFIG                                     = 0x000286c4,
+	VS_PER_COMPONENT_bit                              = 1 << 0,
+	VS_EXPORT_COUNT_mask                              = 0x1f << 1,
+	VS_EXPORT_COUNT_shift                             = 1,
+	VS_EXPORTS_FOG_bit                                = 1 << 8,
+	VS_OUT_FOG_VEC_ADDR_mask                          = 0x1f << 9,
+	VS_OUT_FOG_VEC_ADDR_shift                         = 9,
+    SPI_PS_IN_CONTROL_0                                   = 0x000286cc,
+	NUM_INTERP_mask                                   = 0x3f << 0,
+	NUM_INTERP_shift                                  = 0,
+	POSITION_ENA_bit                                  = 1 << 8,
+	POSITION_CENTROID_bit                             = 1 << 9,
+	POSITION_ADDR_mask                                = 0x1f << 10,
+	POSITION_ADDR_shift                               = 10,
+	PARAM_GEN_mask                                    = 0x0f << 15,
+	PARAM_GEN_shift                                   = 15,
+	PARAM_GEN_ADDR_mask                               = 0x7f << 19,
+	PARAM_GEN_ADDR_shift                              = 19,
+	BARYC_SAMPLE_CNTL_mask                            = 0x03 << 26,
+	BARYC_SAMPLE_CNTL_shift                           = 26,
+	    CENTROIDS_ONLY                                = 0x00,
+	    CENTERS_ONLY                                  = 0x01,
+	    CENTROIDS_AND_CENTERS                         = 0x02,
+	    UNDEF                                         = 0x03,
+	PERSP_GRADIENT_ENA_bit                            = 1 << 28,
+	LINEAR_GRADIENT_ENA_bit                           = 1 << 29,
+	POSITION_SAMPLE_bit                               = 1 << 30,
+	BARYC_AT_SAMPLE_ENA_bit                           = 1 << 31,
+    SPI_PS_IN_CONTROL_1                                   = 0x000286d0,
+	GEN_INDEX_PIX_bit                                 = 1 << 0,
+	GEN_INDEX_PIX_ADDR_mask                           = 0x7f << 1,
+	GEN_INDEX_PIX_ADDR_shift                          = 1,
+	FRONT_FACE_ENA_bit                                = 1 << 8,
+	FRONT_FACE_CHAN_mask                              = 0x03 << 9,
+	FRONT_FACE_CHAN_shift                             = 9,
+	FRONT_FACE_ALL_BITS_bit                           = 1 << 11,
+	FRONT_FACE_ADDR_mask                              = 0x1f << 12,
+	FRONT_FACE_ADDR_shift                             = 12,
+	FOG_ADDR_mask                                     = 0x7f << 17,
+	FOG_ADDR_shift                                    = 17,
+	FIXED_PT_POSITION_ENA_bit                         = 1 << 24,
+	FIXED_PT_POSITION_ADDR_mask                       = 0x1f << 25,
+	FIXED_PT_POSITION_ADDR_shift                      = 25,
+    SPI_INTERP_CONTROL_0                                  = 0x000286d4,
+	FLAT_SHADE_ENA_bit                                = 1 << 0,
+	PNT_SPRITE_ENA_bit                                = 1 << 1,
+	PNT_SPRITE_OVRD_X_mask                            = 0x07 << 2,
+	PNT_SPRITE_OVRD_X_shift                           = 2,
+	    SPI_PNT_SPRITE_SEL_0                          = 0x00,
+	    SPI_PNT_SPRITE_SEL_1                          = 0x01,
+	    SPI_PNT_SPRITE_SEL_S                          = 0x02,
+	    SPI_PNT_SPRITE_SEL_T                          = 0x03,
+	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04,
+	PNT_SPRITE_OVRD_Y_mask                            = 0x07 << 5,
+	PNT_SPRITE_OVRD_Y_shift                           = 5,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_OVRD_Z_mask                            = 0x07 << 8,
+	PNT_SPRITE_OVRD_Z_shift                           = 8,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_OVRD_W_mask                            = 0x07 << 11,
+	PNT_SPRITE_OVRD_W_shift                           = 11,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_TOP_1_bit                              = 1 << 14,
+    SPI_INPUT_Z                                           = 0x000286d8,
+	PROVIDE_Z_TO_SPI_bit                              = 1 << 0,
+    SPI_FOG_CNTL                                          = 0x000286dc,
+	PASS_FOG_THROUGH_PS_bit                           = 1 << 0,
+	PIXEL_FOG_FUNC_mask                               = 0x03 << 1,
+	PIXEL_FOG_FUNC_shift                              = 1,
+	    SPI_FOG_NONE                                  = 0x00,
+	    SPI_FOG_EXP                                   = 0x01,
+	    SPI_FOG_EXP2                                  = 0x02,
+	    SPI_FOG_LINEAR                                = 0x03,
+	PIXEL_FOG_SRC_SEL_bit                             = 1 << 3,
+	VS_FOG_CLAMP_DISABLE_bit                          = 1 << 4,
+    SPI_FOG_FUNC_SCALE                                    = 0x000286e0,
+    SPI_FOG_FUNC_BIAS                                     = 0x000286e4,
+    CB_BLEND0_CONTROL                                     = 0x00028780,
+	CB_BLEND0_CONTROL_num                             = 8,
+	COLOR_SRCBLEND_mask                               = 0x1f << 0,
+	COLOR_SRCBLEND_shift                              = 0,
+	COLOR_COMB_FCN_mask                               = 0x07 << 5,
+	COLOR_COMB_FCN_shift                              = 5,
+	COLOR_DESTBLEND_mask                              = 0x1f << 8,
+	COLOR_DESTBLEND_shift                             = 8,
+	OPACITY_WEIGHT_bit                                = 1 << 13,
+	ALPHA_SRCBLEND_mask                               = 0x1f << 16,
+	ALPHA_SRCBLEND_shift                              = 16,
+	ALPHA_COMB_FCN_mask                               = 0x07 << 21,
+	ALPHA_COMB_FCN_shift                              = 21,
+	ALPHA_DESTBLEND_mask                              = 0x1f << 24,
+	ALPHA_DESTBLEND_shift                             = 24,
+	SEPARATE_ALPHA_BLEND_bit                          = 1 << 29,
+    VGT_DMA_BASE_HI                                       = 0x000287e4,
+	VGT_DMA_BASE_HI__BASE_ADDR_mask                   = 0xff << 0,
+	VGT_DMA_BASE_HI__BASE_ADDR_shift                  = 0,
+    VGT_DMA_BASE                                          = 0x000287e8,
+    VGT_DRAW_INITIATOR                                    = 0x000287f0,
+	SOURCE_SELECT_mask                                = 0x03 << 0,
+	SOURCE_SELECT_shift                               = 0,
+	    DI_SRC_SEL_DMA                                = 0x00,
+	    DI_SRC_SEL_IMMEDIATE                          = 0x01,
+	    DI_SRC_SEL_AUTO_INDEX                         = 0x02,
+	    DI_SRC_SEL_RESERVED                           = 0x03,
+	MAJOR_MODE_mask                                   = 0x03 << 2,
+	MAJOR_MODE_shift                                  = 2,
+	    DI_MAJOR_MODE_0                               = 0x00,
+	    DI_MAJOR_MODE_1                               = 0x01,
+	SPRITE_EN_bit                                     = 1 << 4,
+	NOT_EOP_bit                                       = 1 << 5,
+	USE_OPAQUE_bit                                    = 1 << 6,
+    VGT_IMMED_DATA                                        = 0x000287f4,
+    VGT_EVENT_ADDRESS_REG                                 = 0x000287f8,
+	ADDRESS_LOW_mask                                  = 0xfffffff << 0,
+	ADDRESS_LOW_shift                                 = 0,
+    DB_DEPTH_CONTROL                                      = 0x00028800,
+	STENCIL_ENABLE_bit                                = 1 << 0,
+	Z_ENABLE_bit                                      = 1 << 1,
+	Z_WRITE_ENABLE_bit                                = 1 << 2,
+	ZFUNC_mask                                        = 0x07 << 4,
+	ZFUNC_shift                                       = 4,
+	    FRAG_NEVER                                    = 0x00,
+	    FRAG_LESS                                     = 0x01,
+	    FRAG_EQUAL                                    = 0x02,
+	    FRAG_LEQUAL                                   = 0x03,
+	    FRAG_GREATER                                  = 0x04,
+	    FRAG_NOTEQUAL                                 = 0x05,
+	    FRAG_GEQUAL                                   = 0x06,
+	    FRAG_ALWAYS                                   = 0x07,
+	BACKFACE_ENABLE_bit                               = 1 << 7,
+	STENCILFUNC_mask                                  = 0x07 << 8,
+	STENCILFUNC_shift                                 = 8,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	STENCILFAIL_mask                                  = 0x07 << 11,
+	STENCILFAIL_shift                                 = 11,
+	    STENCIL_KEEP                                  = 0x00,
+	    STENCIL_ZERO                                  = 0x01,
+	    STENCIL_REPLACE                               = 0x02,
+	    STENCIL_INCR_CLAMP                            = 0x03,
+	    STENCIL_DECR_CLAMP                            = 0x04,
+	    STENCIL_INVERT                                = 0x05,
+	    STENCIL_INCR_WRAP                             = 0x06,
+	    STENCIL_DECR_WRAP                             = 0x07,
+	STENCILZPASS_mask                                 = 0x07 << 14,
+	STENCILZPASS_shift                                = 14,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZFAIL_mask                                 = 0x07 << 17,
+	STENCILZFAIL_shift                                = 17,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILFUNC_BF_mask                               = 0x07 << 20,
+	STENCILFUNC_BF_shift                              = 20,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	STENCILFAIL_BF_mask                               = 0x07 << 23,
+	STENCILFAIL_BF_shift                              = 23,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZPASS_BF_mask                              = 0x07 << 26,
+	STENCILZPASS_BF_shift                             = 26,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZFAIL_BF_mask                              = 0x07 << 29,
+	STENCILZFAIL_BF_shift                             = 29,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+    CB_BLEND_CONTROL                                      = 0x00028804,
+/* 	COLOR_SRCBLEND_mask                               = 0x1f << 0, */
+/* 	COLOR_SRCBLEND_shift                              = 0, */
+	    BLEND_ZERO                                    = 0x00,
+	    BLEND_ONE                                     = 0x01,
+	    BLEND_SRC_COLOR                               = 0x02,
+	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03,
+	    BLEND_SRC_ALPHA                               = 0x04,
+	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05,
+	    BLEND_DST_ALPHA                               = 0x06,
+	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07,
+	    BLEND_DST_COLOR                               = 0x08,
+	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09,
+	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a,
+	    BLEND_BOTH_SRC_ALPHA                          = 0x0b,
+	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c,
+	    BLEND_CONSTANT_COLOR                          = 0x0d,
+	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e,
+	    BLEND_SRC1_COLOR                              = 0x0f,
+	    BLEND_INV_SRC1_COLOR                          = 0x10,
+	    BLEND_SRC1_ALPHA                              = 0x11,
+	    BLEND_INV_SRC1_ALPHA                          = 0x12,
+	    BLEND_CONSTANT_ALPHA                          = 0x13,
+	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14,
+/* 	COLOR_COMB_FCN_mask                               = 0x07 << 5, */
+/* 	COLOR_COMB_FCN_shift                              = 5, */
+	    COMB_DST_PLUS_SRC                             = 0x00,
+	    COMB_SRC_MINUS_DST                            = 0x01,
+	    COMB_MIN_DST_SRC                              = 0x02,
+	    COMB_MAX_DST_SRC                              = 0x03,
+	    COMB_DST_MINUS_SRC                            = 0x04,
+/* 	COLOR_DESTBLEND_mask                              = 0x1f << 8, */
+/* 	COLOR_DESTBLEND_shift                             = 8, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	OPACITY_WEIGHT_bit                                = 1 << 13, */
+/* 	ALPHA_SRCBLEND_mask                               = 0x1f << 16, */
+/* 	ALPHA_SRCBLEND_shift                              = 16, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	ALPHA_COMB_FCN_mask                               = 0x07 << 21, */
+/* 	ALPHA_COMB_FCN_shift                              = 21, */
+/* 	    COMB_DST_PLUS_SRC                             = 0x00, */
+/* 	    COMB_SRC_MINUS_DST                            = 0x01, */
+/* 	    COMB_MIN_DST_SRC                              = 0x02, */
+/* 	    COMB_MAX_DST_SRC                              = 0x03, */
+/* 	    COMB_DST_MINUS_SRC                            = 0x04, */
+/* 	ALPHA_DESTBLEND_mask                              = 0x1f << 24, */
+/* 	ALPHA_DESTBLEND_shift                             = 24, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	SEPARATE_ALPHA_BLEND_bit                          = 1 << 29, */
+    CB_COLOR_CONTROL                                      = 0x00028808,
+	FOG_ENABLE_bit                                    = 1 << 0,
+	MULTIWRITE_ENABLE_bit                             = 1 << 1,
+	DITHER_ENABLE_bit                                 = 1 << 2,
+	DEGAMMA_ENABLE_bit                                = 1 << 3,
+	SPECIAL_OP_mask                                   = 0x07 << 4,
+	SPECIAL_OP_shift                                  = 4,
+	    SPECIAL_NORMAL                                = 0x00,
+	    SPECIAL_DISABLE                               = 0x01,
+	    SPECIAL_FAST_CLEAR                            = 0x02,
+	    SPECIAL_FORCE_CLEAR                           = 0x03,
+	    SPECIAL_EXPAND_COLOR                          = 0x04,
+	    SPECIAL_EXPAND_TEXTURE                        = 0x05,
+	    SPECIAL_EXPAND_SAMPLES                        = 0x06,
+	    SPECIAL_RESOLVE_BOX                           = 0x07,
+	PER_MRT_BLEND_bit                                 = 1 << 7,
+	TARGET_BLEND_ENABLE_mask                          = 0xff << 8,
+	TARGET_BLEND_ENABLE_shift                         = 8,
+	ROP3_mask                                         = 0xff << 16,
+	ROP3_shift                                        = 16,
+    DB_SHADER_CONTROL                                     = 0x0002880c,
+	Z_EXPORT_ENABLE_bit                               = 1 << 0,
+	STENCIL_REF_EXPORT_ENABLE_bit                     = 1 << 1,
+	Z_ORDER_mask                                      = 0x03 << 4,
+	Z_ORDER_shift                                     = 4,
+	    LATE_Z                                        = 0x00,
+	    EARLY_Z_THEN_LATE_Z                           = 0x01,
+	    RE_Z                                          = 0x02,
+	    EARLY_Z_THEN_RE_Z                             = 0x03,
+	KILL_ENABLE_bit                                   = 1 << 6,
+	COVERAGE_TO_MASK_ENABLE_bit                       = 1 << 7,
+	MASK_EXPORT_ENABLE_bit                            = 1 << 8,
+	DUAL_EXPORT_ENABLE_bit                            = 1 << 9,
+	EXEC_ON_HIER_FAIL_bit                             = 1 << 10,
+	EXEC_ON_NOOP_bit                                  = 1 << 11,
+    PA_CL_CLIP_CNTL                                       = 0x00028810,
+	UCP_ENA_0_bit                                     = 1 << 0,
+	UCP_ENA_1_bit                                     = 1 << 1,
+	UCP_ENA_2_bit                                     = 1 << 2,
+	UCP_ENA_3_bit                                     = 1 << 3,
+	UCP_ENA_4_bit                                     = 1 << 4,
+	UCP_ENA_5_bit                                     = 1 << 5,
+	PS_UCP_Y_SCALE_NEG_bit                            = 1 << 13,
+	PS_UCP_MODE_mask                                  = 0x03 << 14,
+	PS_UCP_MODE_shift                                 = 14,
+	CLIP_DISABLE_bit                                  = 1 << 16,
+	UCP_CULL_ONLY_ENA_bit                             = 1 << 17,
+	BOUNDARY_EDGE_FLAG_ENA_bit                        = 1 << 18,
+	DX_CLIP_SPACE_DEF_bit                             = 1 << 19,
+	DIS_CLIP_ERR_DETECT_bit                           = 1 << 20,
+	VTX_KILL_OR_bit                                   = 1 << 21,
+	DX_LINEAR_ATTR_CLIP_ENA_bit                       = 1 << 24,
+	VTE_VPORT_PROVOKE_DISABLE_bit                     = 1 << 25,
+	ZCLIP_NEAR_DISABLE_bit                            = 1 << 26,
+	ZCLIP_FAR_DISABLE_bit                             = 1 << 27,
+    PA_SU_SC_MODE_CNTL                                    = 0x00028814,
+	CULL_FRONT_bit                                    = 1 << 0,
+	CULL_BACK_bit                                     = 1 << 1,
+	FACE_bit                                          = 1 << 2,
+	POLY_MODE_mask                                    = 0x03 << 3,
+	POLY_MODE_shift                                   = 3,
+	    X_DISABLE_POLY_MODE                           = 0x00,
+	    X_DUAL_MODE                                   = 0x01,
+	POLYMODE_FRONT_PTYPE_mask                         = 0x07 << 5,
+	POLYMODE_FRONT_PTYPE_shift                        = 5,
+	    X_DRAW_POINTS                                 = 0x00,
+	    X_DRAW_LINES                                  = 0x01,
+	    X_DRAW_TRIANGLES                              = 0x02,
+	POLYMODE_BACK_PTYPE_mask                          = 0x07 << 8,
+	POLYMODE_BACK_PTYPE_shift                         = 8,
+/* 	    X_DRAW_POINTS                                 = 0x00, */
+/* 	    X_DRAW_LINES                                  = 0x01, */
+/* 	    X_DRAW_TRIANGLES                              = 0x02, */
+	POLY_OFFSET_FRONT_ENABLE_bit                      = 1 << 11,
+	POLY_OFFSET_BACK_ENABLE_bit                       = 1 << 12,
+	POLY_OFFSET_PARA_ENABLE_bit                       = 1 << 13,
+	VTX_WINDOW_OFFSET_ENABLE_bit                      = 1 << 16,
+	PROVOKING_VTX_LAST_bit                            = 1 << 19,
+	PERSP_CORR_DIS_bit                                = 1 << 20,
+	MULTI_PRIM_IB_ENA_bit                             = 1 << 21,
+    PA_CL_VTE_CNTL                                        = 0x00028818,
+	VPORT_X_SCALE_ENA_bit                             = 1 << 0,
+	VPORT_X_OFFSET_ENA_bit                            = 1 << 1,
+	VPORT_Y_SCALE_ENA_bit                             = 1 << 2,
+	VPORT_Y_OFFSET_ENA_bit                            = 1 << 3,
+	VPORT_Z_SCALE_ENA_bit                             = 1 << 4,
+	VPORT_Z_OFFSET_ENA_bit                            = 1 << 5,
+	VTX_XY_FMT_bit                                    = 1 << 8,
+	VTX_Z_FMT_bit                                     = 1 << 9,
+	VTX_W0_FMT_bit                                    = 1 << 10,
+	PERFCOUNTER_REF_bit                               = 1 << 11,
+    PA_CL_VS_OUT_CNTL                                     = 0x0002881c,
+	CLIP_DIST_ENA_0_bit                               = 1 << 0,
+	CLIP_DIST_ENA_1_bit                               = 1 << 1,
+	CLIP_DIST_ENA_2_bit                               = 1 << 2,
+	CLIP_DIST_ENA_3_bit                               = 1 << 3,
+	CLIP_DIST_ENA_4_bit                               = 1 << 4,
+	CLIP_DIST_ENA_5_bit                               = 1 << 5,
+	CLIP_DIST_ENA_6_bit                               = 1 << 6,
+	CLIP_DIST_ENA_7_bit                               = 1 << 7,
+	CULL_DIST_ENA_0_bit                               = 1 << 8,
+	CULL_DIST_ENA_1_bit                               = 1 << 9,
+	CULL_DIST_ENA_2_bit                               = 1 << 10,
+	CULL_DIST_ENA_3_bit                               = 1 << 11,
+	CULL_DIST_ENA_4_bit                               = 1 << 12,
+	CULL_DIST_ENA_5_bit                               = 1 << 13,
+	CULL_DIST_ENA_6_bit                               = 1 << 14,
+	CULL_DIST_ENA_7_bit                               = 1 << 15,
+	USE_VTX_POINT_SIZE_bit                            = 1 << 16,
+	USE_VTX_EDGE_FLAG_bit                             = 1 << 17,
+	USE_VTX_RENDER_TARGET_INDX_bit                    = 1 << 18,
+	USE_VTX_VIEWPORT_INDX_bit                         = 1 << 19,
+	USE_VTX_KILL_FLAG_bit                             = 1 << 20,
+	VS_OUT_MISC_VEC_ENA_bit                           = 1 << 21,
+	VS_OUT_CCDIST0_VEC_ENA_bit                        = 1 << 22,
+	VS_OUT_CCDIST1_VEC_ENA_bit                        = 1 << 23,
+    PA_CL_NANINF_CNTL                                     = 0x00028820,
+	VTE_XY_INF_DISCARD_bit                            = 1 << 0,
+	VTE_Z_INF_DISCARD_bit                             = 1 << 1,
+	VTE_W_INF_DISCARD_bit                             = 1 << 2,
+	VTE_0XNANINF_IS_0_bit                             = 1 << 3,
+	VTE_XY_NAN_RETAIN_bit                             = 1 << 4,
+	VTE_Z_NAN_RETAIN_bit                              = 1 << 5,
+	VTE_W_NAN_RETAIN_bit                              = 1 << 6,
+	VTE_W_RECIP_NAN_IS_0_bit                          = 1 << 7,
+	VS_XY_NAN_TO_INF_bit                              = 1 << 8,
+	VS_XY_INF_RETAIN_bit                              = 1 << 9,
+	VS_Z_NAN_TO_INF_bit                               = 1 << 10,
+	VS_Z_INF_RETAIN_bit                               = 1 << 11,
+	VS_W_NAN_TO_INF_bit                               = 1 << 12,
+	VS_W_INF_RETAIN_bit                               = 1 << 13,
+	VS_CLIP_DIST_INF_DISCARD_bit                      = 1 << 14,
+	VTE_NO_OUTPUT_NEG_0_bit                           = 1 << 20,
+    SQ_PGM_START_PS                                       = 0x00028840,
+    SQ_PGM_RESOURCES_PS                                   = 0x00028850,
+	NUM_GPRS_mask                                     = 0xff << 0,
+	NUM_GPRS_shift                                    = 0,
+	STACK_SIZE_mask                                   = 0xff << 8,
+	STACK_SIZE_shift                                  = 8,
+	SQ_PGM_RESOURCES_PS__DX10_CLAMP_bit               = 1 << 21,
+	FETCH_CACHE_LINES_mask                            = 0x07 << 24,
+	FETCH_CACHE_LINES_shift                           = 24,
+	UNCACHED_FIRST_INST_bit                           = 1 << 28,
+	CLAMP_CONSTS_bit                                  = 1 << 31,
+    SQ_PGM_EXPORTS_PS                                     = 0x00028854,
+	EXPORT_MODE_mask                                  = 0x1f << 0,
+	EXPORT_MODE_shift                                 = 0,
+    SQ_PGM_START_VS                                       = 0x00028858,
+    SQ_PGM_RESOURCES_VS                                   = 0x00028868,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_VS__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_GS                                       = 0x0002886c,
+    SQ_PGM_RESOURCES_GS                                   = 0x0002887c,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_GS__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_ES                                       = 0x00028880,
+    SQ_PGM_RESOURCES_ES                                   = 0x00028890,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_ES__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_FS                                       = 0x00028894,
+    SQ_PGM_RESOURCES_FS                                   = 0x000288a4,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_FS__DX10_CLAMP_bit               = 1 << 21,
+    SQ_ESGS_RING_ITEMSIZE                                 = 0x000288a8,
+	ITEMSIZE_mask                                     = 0x7fff << 0,
+	ITEMSIZE_shift                                    = 0,
+    SQ_GSVS_RING_ITEMSIZE                                 = 0x000288ac,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_ESTMP_RING_ITEMSIZE                                = 0x000288b0,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_GSTMP_RING_ITEMSIZE                                = 0x000288b4,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_VSTMP_RING_ITEMSIZE                                = 0x000288b8,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_PSTMP_RING_ITEMSIZE                                = 0x000288bc,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_FBUF_RING_ITEMSIZE                                 = 0x000288c0,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_REDUC_RING_ITEMSIZE                                = 0x000288c4,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_GS_VERT_ITEMSIZE                                   = 0x000288c8,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_PGM_CF_OFFSET_PS                                   = 0x000288cc,
+	PGM_CF_OFFSET_mask                                = 0xfffff << 0,
+	PGM_CF_OFFSET_shift                               = 0,
+    SQ_PGM_CF_OFFSET_VS                                   = 0x000288d0,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_GS                                   = 0x000288d4,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_ES                                   = 0x000288d8,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_FS                                   = 0x000288dc,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_VTX_SEMANTIC_CLEAR                                 = 0x000288e0,
+    SQ_ALU_CONST_CACHE_PS_0                               = 0x00028940,
+	SQ_ALU_CONST_CACHE_PS_0_num                       = 16,
+    SQ_ALU_CONST_CACHE_VS_0                               = 0x00028980,
+	SQ_ALU_CONST_CACHE_VS_0_num                       = 16,
+    SQ_ALU_CONST_CACHE_GS_0                               = 0x000289c0,
+	SQ_ALU_CONST_CACHE_GS_0_num                       = 16,
+    PA_SU_POINT_SIZE                                      = 0x00028a00,
+	PA_SU_POINT_SIZE__HEIGHT_mask                     = 0xffff << 0,
+	PA_SU_POINT_SIZE__HEIGHT_shift                    = 0,
+	PA_SU_POINT_SIZE__WIDTH_mask                      = 0xffff << 16,
+	PA_SU_POINT_SIZE__WIDTH_shift                     = 16,
+    PA_SU_POINT_MINMAX                                    = 0x00028a04,
+	MIN_SIZE_mask                                     = 0xffff << 0,
+	MIN_SIZE_shift                                    = 0,
+	MAX_SIZE_mask                                     = 0xffff << 16,
+	MAX_SIZE_shift                                    = 16,
+    PA_SU_LINE_CNTL                                       = 0x00028a08,
+	PA_SU_LINE_CNTL__WIDTH_mask                       = 0xffff << 0,
+	PA_SU_LINE_CNTL__WIDTH_shift                      = 0,
+    PA_SC_LINE_STIPPLE                                    = 0x00028a0c,
+	LINE_PATTERN_mask                                 = 0xffff << 0,
+	LINE_PATTERN_shift                                = 0,
+	REPEAT_COUNT_mask                                 = 0xff << 16,
+	REPEAT_COUNT_shift                                = 16,
+	PATTERN_BIT_ORDER_bit                             = 1 << 28,
+	AUTO_RESET_CNTL_mask                              = 0x03 << 29,
+	AUTO_RESET_CNTL_shift                             = 29,
+    VGT_OUTPUT_PATH_CNTL                                  = 0x00028a10,
+	PATH_SELECT_mask                                  = 0x03 << 0,
+	PATH_SELECT_shift                                 = 0,
+	    VGT_OUTPATH_VTX_REUSE                         = 0x00,
+	    VGT_OUTPATH_TESS_EN                           = 0x01,
+	    VGT_OUTPATH_PASSTHRU                          = 0x02,
+	    VGT_OUTPATH_GS_BLOCK                          = 0x03,
+    VGT_HOS_CNTL                                          = 0x00028a14,
+	TESS_MODE_mask                                    = 0x03 << 0,
+	TESS_MODE_shift                                   = 0,
+    VGT_HOS_MAX_TESS_LEVEL                                = 0x00028a18,
+    VGT_HOS_MIN_TESS_LEVEL                                = 0x00028a1c,
+    VGT_HOS_REUSE_DEPTH                                   = 0x00028a20,
+	REUSE_DEPTH_mask                                  = 0xff << 0,
+	REUSE_DEPTH_shift                                 = 0,
+    VGT_GROUP_PRIM_TYPE                                   = 0x00028a24,
+	VGT_GROUP_PRIM_TYPE__PRIM_TYPE_mask               = 0x1f << 0,
+	VGT_GROUP_PRIM_TYPE__PRIM_TYPE_shift              = 0,
+	    VGT_GRP_3D_POINT                              = 0x00,
+	    VGT_GRP_3D_LINE                               = 0x01,
+	    VGT_GRP_3D_TRI                                = 0x02,
+	    VGT_GRP_3D_RECT                               = 0x03,
+	    VGT_GRP_3D_QUAD                               = 0x04,
+	    VGT_GRP_2D_COPY_RECT_V0                       = 0x05,
+	    VGT_GRP_2D_COPY_RECT_V1                       = 0x06,
+	    VGT_GRP_2D_COPY_RECT_V2                       = 0x07,
+	    VGT_GRP_2D_COPY_RECT_V3                       = 0x08,
+	    VGT_GRP_2D_FILL_RECT                          = 0x09,
+	    VGT_GRP_2D_LINE                               = 0x0a,
+	    VGT_GRP_2D_TRI                                = 0x0b,
+	    VGT_GRP_PRIM_INDEX_LINE                       = 0x0c,
+	    VGT_GRP_PRIM_INDEX_TRI                        = 0x0d,
+	    VGT_GRP_PRIM_INDEX_QUAD                       = 0x0e,
+	    VGT_GRP_3D_LINE_ADJ                           = 0x0f,
+	    VGT_GRP_3D_TRI_ADJ                            = 0x10,
+	RETAIN_ORDER_bit                                  = 1 << 14,
+	RETAIN_QUADS_bit                                  = 1 << 15,
+	PRIM_ORDER_mask                                   = 0x07 << 16,
+	PRIM_ORDER_shift                                  = 16,
+	    VGT_GRP_LIST                                  = 0x00,
+	    VGT_GRP_STRIP                                 = 0x01,
+	    VGT_GRP_FAN                                   = 0x02,
+	    VGT_GRP_LOOP                                  = 0x03,
+	    VGT_GRP_POLYGON                               = 0x04,
+    VGT_GROUP_FIRST_DECR                                  = 0x00028a28,
+	FIRST_DECR_mask                                   = 0x0f << 0,
+	FIRST_DECR_shift                                  = 0,
+    VGT_GROUP_DECR                                        = 0x00028a2c,
+	DECR_mask                                         = 0x0f << 0,
+	DECR_shift                                        = 0,
+    VGT_GROUP_VECT_0_CNTL                                 = 0x00028a30,
+	COMP_X_EN_bit                                     = 1 << 0,
+	COMP_Y_EN_bit                                     = 1 << 1,
+	COMP_Z_EN_bit                                     = 1 << 2,
+	COMP_W_EN_bit                                     = 1 << 3,
+	VGT_GROUP_VECT_0_CNTL__STRIDE_mask                = 0xff << 8,
+	VGT_GROUP_VECT_0_CNTL__STRIDE_shift               = 8,
+	SHIFT_mask                                        = 0xff << 16,
+	SHIFT_shift                                       = 16,
+    VGT_GROUP_VECT_1_CNTL                                 = 0x00028a34,
+/* 	COMP_X_EN_bit                                     = 1 << 0, */
+/* 	COMP_Y_EN_bit                                     = 1 << 1, */
+/* 	COMP_Z_EN_bit                                     = 1 << 2, */
+/* 	COMP_W_EN_bit                                     = 1 << 3, */
+	VGT_GROUP_VECT_1_CNTL__STRIDE_mask                = 0xff << 8,
+	VGT_GROUP_VECT_1_CNTL__STRIDE_shift               = 8,
+/* 	SHIFT_mask                                        = 0xff << 16, */
+/* 	SHIFT_shift                                       = 16, */
+    VGT_GROUP_VECT_0_FMT_CNTL                             = 0x00028a38,
+	X_CONV_mask                                       = 0x0f << 0,
+	X_CONV_shift                                      = 0,
+	    VGT_GRP_INDEX_16                              = 0x00,
+	    VGT_GRP_INDEX_32                              = 0x01,
+	    VGT_GRP_UINT_16                               = 0x02,
+	    VGT_GRP_UINT_32                               = 0x03,
+	    VGT_GRP_SINT_16                               = 0x04,
+	    VGT_GRP_SINT_32                               = 0x05,
+	    VGT_GRP_FLOAT_32                              = 0x06,
+	    VGT_GRP_AUTO_PRIM                             = 0x07,
+	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08,
+	X_OFFSET_mask                                     = 0x0f << 4,
+	X_OFFSET_shift                                    = 4,
+	Y_CONV_mask                                       = 0x0f << 8,
+	Y_CONV_shift                                      = 8,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	Y_OFFSET_mask                                     = 0x0f << 12,
+	Y_OFFSET_shift                                    = 12,
+	Z_CONV_mask                                       = 0x0f << 16,
+	Z_CONV_shift                                      = 16,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	Z_OFFSET_mask                                     = 0x0f << 20,
+	Z_OFFSET_shift                                    = 20,
+	W_CONV_mask                                       = 0x0f << 24,
+	W_CONV_shift                                      = 24,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	W_OFFSET_mask                                     = 0x0f << 28,
+	W_OFFSET_shift                                    = 28,
+    VGT_GROUP_VECT_1_FMT_CNTL                             = 0x00028a3c,
+/* 	X_CONV_mask                                       = 0x0f << 0, */
+/* 	X_CONV_shift                                      = 0, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	X_OFFSET_mask                                     = 0x0f << 4, */
+/* 	X_OFFSET_shift                                    = 4, */
+/* 	Y_CONV_mask                                       = 0x0f << 8, */
+/* 	Y_CONV_shift                                      = 8, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	Y_OFFSET_mask                                     = 0x0f << 12, */
+/* 	Y_OFFSET_shift                                    = 12, */
+/* 	Z_CONV_mask                                       = 0x0f << 16, */
+/* 	Z_CONV_shift                                      = 16, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	Z_OFFSET_mask                                     = 0x0f << 20, */
+/* 	Z_OFFSET_shift                                    = 20, */
+/* 	W_CONV_mask                                       = 0x0f << 24, */
+/* 	W_CONV_shift                                      = 24, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	W_OFFSET_mask                                     = 0x0f << 28, */
+/* 	W_OFFSET_shift                                    = 28, */
+    VGT_GS_MODE                                           = 0x00028a40,
+	MODE_mask                                         = 0x03 << 0,
+	MODE_shift                                        = 0,
+	    GS_OFF                                        = 0x00,
+	    GS_SCENARIO_A                                 = 0x01,
+	    GS_SCENARIO_B                                 = 0x02,
+	    GS_SCENARIO_G                                 = 0x03,
+	ES_PASSTHRU_bit                                   = 1 << 2,
+	CUT_MODE_mask                                     = 0x03 << 3,
+	CUT_MODE_shift                                    = 3,
+	    GS_CUT_1024                                   = 0x00,
+	    GS_CUT_512                                    = 0x01,
+	    GS_CUT_256                                    = 0x02,
+	    GS_CUT_128                                    = 0x03,
+    PA_SC_MPASS_PS_CNTL                                   = 0x00028a48,
+	MPASS_PIX_VEC_PER_PASS_mask                       = 0xfffff << 0,
+	MPASS_PIX_VEC_PER_PASS_shift                      = 0,
+	MPASS_PS_ENA_bit                                  = 1 << 31,
+    PA_SC_MODE_CNTL                                       = 0x00028a4c,
+	MSAA_ENABLE_bit                                   = 1 << 0,
+	CLIPRECT_ENABLE_bit                               = 1 << 1,
+	LINE_STIPPLE_ENABLE_bit                           = 1 << 2,
+	MULTI_CHIP_PRIM_DISCARD_ENAB_bit                  = 1 << 3,
+	WALK_ORDER_ENABLE_bit                             = 1 << 4,
+	HALVE_DETAIL_SAMPLE_PERF_bit                      = 1 << 5,
+	WALK_SIZE_bit                                     = 1 << 6,
+	WALK_ALIGNMENT_bit                                = 1 << 7,
+	WALK_ALIGN8_PRIM_FITS_ST_bit                      = 1 << 8,
+	TILE_COVER_NO_SCISSOR_bit                         = 1 << 9,
+	KILL_PIX_POST_HI_Z_bit                            = 1 << 10,
+	KILL_PIX_POST_DETAIL_MASK_bit                     = 1 << 11,
+	MULTI_CHIP_SUPERTILE_ENABLE_bit                   = 1 << 12,
+	TILE_COVER_DISABLE_bit                            = 1 << 13,
+	FORCE_EOV_CNTDWN_ENABLE_bit                       = 1 << 14,
+	FORCE_EOV_TILE_ENABLE_bit                         = 1 << 15,
+	FORCE_EOV_REZ_ENABLE_bit                          = 1 << 16,
+	PS_ITER_SAMPLE_bit                                = 1 << 17,
+    VGT_ENHANCE                                           = 0x00028a50,
+	VGT_ENHANCE__MI_TIMESTAMP_RES_mask                = 0x03 << 0,
+	VGT_ENHANCE__MI_TIMESTAMP_RES_shift               = 0,
+	    X_0_992_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_32   = 0x00,
+	    X_0_496_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_16   = 0x01,
+	    X_0_248_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_8    = 0x02,
+	    X_0_124_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_4    = 0x03,
+	MISC_mask                                         = 0x3fffffff << 2,
+	MISC_shift                                        = 2,
+    VGT_GS_OUT_PRIM_TYPE                                  = 0x00028a6c,
+	OUTPRIM_TYPE_mask                                 = 0x3f << 0,
+	OUTPRIM_TYPE_shift                                = 0,
+	    POINTLIST                                     = 0x00,
+	    LINESTRIP                                     = 0x01,
+	    TRISTRIP                                      = 0x02,
+    VGT_DMA_SIZE                                          = 0x00028a74,
+    VGT_DMA_INDEX_TYPE                                    = 0x00028a7c,
+/* 	INDEX_TYPE_mask                                   = 0x03 << 0, */
+/* 	INDEX_TYPE_shift                                  = 0, */
+	    VGT_INDEX_16                                  = 0x00,
+	    VGT_INDEX_32                                  = 0x01,
+	SWAP_MODE_mask                                    = 0x03 << 2,
+	SWAP_MODE_shift                                   = 2,
+	    VGT_DMA_SWAP_NONE                             = 0x00,
+	    VGT_DMA_SWAP_16_BIT                           = 0x01,
+	    VGT_DMA_SWAP_32_BIT                           = 0x02,
+	    VGT_DMA_SWAP_WORD                             = 0x03,
+    VGT_PRIMITIVEID_EN                                    = 0x00028a84,
+	PRIMITIVEID_EN_bit                                = 1 << 0,
+    VGT_DMA_NUM_INSTANCES                                 = 0x00028a88,
+    VGT_EVENT_INITIATOR                                   = 0x00028a90,
+	EVENT_TYPE_mask                                   = 0x3f << 0,
+	EVENT_TYPE_shift                                  = 0,
+	    CACHE_FLUSH_TS                                = 0x04,
+	    CONTEXT_DONE                                  = 0x05,
+	    CACHE_FLUSH                                   = 0x06,
+	    VIZQUERY_START                                = 0x07,
+	    VIZQUERY_END                                  = 0x08,
+	    SC_WAIT_WC                                    = 0x09,
+	    MPASS_PS_CP_REFETCH                           = 0x0a,
+	    MPASS_PS_RST_START                            = 0x0b,
+	    MPASS_PS_INCR_START                           = 0x0c,
+	    RST_PIX_CNT                                   = 0x0d,
+	    RST_VTX_CNT                                   = 0x0e,
+	    VS_PARTIAL_FLUSH                              = 0x0f,
+	    PS_PARTIAL_FLUSH                              = 0x10,
+	    CACHE_FLUSH_AND_INV_TS_EVENT                  = 0x14,
+	    ZPASS_DONE                                    = 0x15,
+	    CACHE_FLUSH_AND_INV_EVENT                     = 0x16,
+	    PERFCOUNTER_START                             = 0x17,
+	    PERFCOUNTER_STOP                              = 0x18,
+	    PIPELINESTAT_START                            = 0x19,
+	    PIPELINESTAT_STOP                             = 0x1a,
+	    PERFCOUNTER_SAMPLE                            = 0x1b,
+	    FLUSH_ES_OUTPUT                               = 0x1c,
+	    FLUSH_GS_OUTPUT                               = 0x1d,
+	    SAMPLE_PIPELINESTAT                           = 0x1e,
+	    SO_VGTSTREAMOUT_FLUSH                         = 0x1f,
+	    SAMPLE_STREAMOUTSTATS                         = 0x20,
+	    RESET_VTX_CNT                                 = 0x21,
+	    BLOCK_CONTEXT_DONE                            = 0x22,
+	    CR_CONTEXT_DONE                               = 0x23,
+	    VGT_FLUSH                                     = 0x24,
+	    CR_DONE_TS                                    = 0x25,
+	    SQ_NON_EVENT                                  = 0x26,
+	    SC_SEND_DB_VPZ                                = 0x27,
+	    BOTTOM_OF_PIPE_TS                             = 0x28,
+	    DB_CACHE_FLUSH_AND_INV                        = 0x2a,
+	ADDRESS_HI_mask                                   = 0xff << 19,
+	ADDRESS_HI_shift                                  = 19,
+	EXTENDED_EVENT_bit                                = 1 << 27,
+    VGT_MULTI_PRIM_IB_RESET_EN                            = 0x00028a94,
+	RESET_EN_bit                                      = 1 << 0,
+    VGT_INSTANCE_STEP_RATE_0                              = 0x00028aa0,
+    VGT_INSTANCE_STEP_RATE_1                              = 0x00028aa4,
+    VGT_STRMOUT_EN                                        = 0x00028ab0,
+	STREAMOUT_bit                                     = 1 << 0,
+    VGT_REUSE_OFF                                         = 0x00028ab4,
+	REUSE_OFF_bit                                     = 1 << 0,
+    VGT_VTX_CNT_EN                                        = 0x00028ab8,
+	VTX_CNT_EN_bit                                    = 1 << 0,
+    VGT_STRMOUT_BUFFER_SIZE_0                             = 0x00028ad0,
+    VGT_STRMOUT_VTX_STRIDE_0                              = 0x00028ad4,
+	VGT_STRMOUT_VTX_STRIDE_0__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_0__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_0                             = 0x00028ad8,
+    VGT_STRMOUT_BUFFER_OFFSET_0                           = 0x00028adc,
+    VGT_STRMOUT_BUFFER_SIZE_1                             = 0x00028ae0,
+    VGT_STRMOUT_VTX_STRIDE_1                              = 0x00028ae4,
+	VGT_STRMOUT_VTX_STRIDE_1__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_1__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_1                             = 0x00028ae8,
+    VGT_STRMOUT_BUFFER_OFFSET_1                           = 0x00028aec,
+    VGT_STRMOUT_BUFFER_SIZE_2                             = 0x00028af0,
+    VGT_STRMOUT_VTX_STRIDE_2                              = 0x00028af4,
+	VGT_STRMOUT_VTX_STRIDE_2__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_2__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_2                             = 0x00028af8,
+    VGT_STRMOUT_BUFFER_OFFSET_2                           = 0x00028afc,
+    VGT_STRMOUT_BUFFER_SIZE_3                             = 0x00028b00,
+    VGT_STRMOUT_VTX_STRIDE_3                              = 0x00028b04,
+	VGT_STRMOUT_VTX_STRIDE_3__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_3__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_3                             = 0x00028b08,
+    VGT_STRMOUT_BUFFER_OFFSET_3                           = 0x00028b0c,
+    VGT_STRMOUT_BASE_OFFSET_0                             = 0x00028b10,
+    VGT_STRMOUT_BASE_OFFSET_1                             = 0x00028b14,
+    VGT_STRMOUT_BASE_OFFSET_2                             = 0x00028b18,
+    VGT_STRMOUT_BASE_OFFSET_3                             = 0x00028b1c,
+    VGT_STRMOUT_BUFFER_EN                                 = 0x00028b20,
+	BUFFER_0_EN_bit                                   = 1 << 0,
+	BUFFER_1_EN_bit                                   = 1 << 1,
+	BUFFER_2_EN_bit                                   = 1 << 2,
+	BUFFER_3_EN_bit                                   = 1 << 3,
+    VGT_STRMOUT_DRAW_OPAQUE_OFFSET                        = 0x00028b28,
+    VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE            = 0x00028b2c,
+    VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE                 = 0x00028b30,
+    VGT_STRMOUT_BASE_OFFSET_HI_0                          = 0x00028b44,
+	VGT_STRMOUT_BASE_OFFSET_HI_0__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_0__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_1                          = 0x00028b48,
+	VGT_STRMOUT_BASE_OFFSET_HI_1__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_1__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_2                          = 0x00028b4c,
+	VGT_STRMOUT_BASE_OFFSET_HI_2__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_2__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_3                          = 0x00028b50,
+	VGT_STRMOUT_BASE_OFFSET_HI_3__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_3__BASE_OFFSET_shift   = 0,
+    PA_SC_LINE_CNTL                                       = 0x00028c00,
+	BRES_CNTL_mask                                    = 0xff << 0,
+	BRES_CNTL_shift                                   = 0,
+	USE_BRES_CNTL_bit                                 = 1 << 8,
+	EXPAND_LINE_WIDTH_bit                             = 1 << 9,
+	LAST_PIXEL_bit                                    = 1 << 10,
+    PA_SC_AA_CONFIG                                       = 0x00028c04,
+	MSAA_NUM_SAMPLES_mask                             = 0x03 << 0,
+	MSAA_NUM_SAMPLES_shift                            = 0,
+	AA_MASK_CENTROID_DTMN_bit                         = 1 << 4,
+	MAX_SAMPLE_DIST_mask                              = 0x0f << 13,
+	MAX_SAMPLE_DIST_shift                             = 13,
+    PA_SU_VTX_CNTL                                        = 0x00028c08,
+	PIX_CENTER_bit                                    = 1 << 0,
+	PA_SU_VTX_CNTL__ROUND_MODE_mask                   = 0x03 << 1,
+	PA_SU_VTX_CNTL__ROUND_MODE_shift                  = 1,
+	    X_TRUNCATE                                    = 0x00,
+	    X_ROUND                                       = 0x01,
+	    X_ROUND_TO_EVEN                               = 0x02,
+	    X_ROUND_TO_ODD                                = 0x03,
+	QUANT_MODE_mask                                   = 0x07 << 3,
+	QUANT_MODE_shift                                  = 3,
+	    X_1_16TH                                      = 0x00,
+	    X_1_8TH                                       = 0x01,
+	    X_1_4TH                                       = 0x02,
+	    X_1_2                                         = 0x03,
+	    X_1                                           = 0x04,
+	    X_1_256TH                                     = 0x05,
+    PA_CL_GB_VERT_CLIP_ADJ                                = 0x00028c0c,
+    PA_CL_GB_VERT_DISC_ADJ                                = 0x00028c10,
+    PA_CL_GB_HORZ_CLIP_ADJ                                = 0x00028c14,
+    PA_CL_GB_HORZ_DISC_ADJ                                = 0x00028c18,
+    PA_SC_AA_SAMPLE_LOCS_MCTX                             = 0x00028c1c,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+/* 	S2_X_mask                                         = 0x0f << 16, */
+/* 	S2_X_shift                                        = 16, */
+/* 	S2_Y_mask                                         = 0x0f << 20, */
+/* 	S2_Y_shift                                        = 20, */
+/* 	S3_X_mask                                         = 0x0f << 24, */
+/* 	S3_X_shift                                        = 24, */
+/* 	S3_Y_mask                                         = 0x0f << 28, */
+/* 	S3_Y_shift                                        = 28, */
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX                      = 0x00028c20,
+/* 	S4_X_mask                                         = 0x0f << 0, */
+/* 	S4_X_shift                                        = 0, */
+/* 	S4_Y_mask                                         = 0x0f << 4, */
+/* 	S4_Y_shift                                        = 4, */
+/* 	S5_X_mask                                         = 0x0f << 8, */
+/* 	S5_X_shift                                        = 8, */
+/* 	S5_Y_mask                                         = 0x0f << 12, */
+/* 	S5_Y_shift                                        = 12, */
+/* 	S6_X_mask                                         = 0x0f << 16, */
+/* 	S6_X_shift                                        = 16, */
+/* 	S6_Y_mask                                         = 0x0f << 20, */
+/* 	S6_Y_shift                                        = 20, */
+/* 	S7_X_mask                                         = 0x0f << 24, */
+/* 	S7_X_shift                                        = 24, */
+/* 	S7_Y_mask                                         = 0x0f << 28, */
+/* 	S7_Y_shift                                        = 28, */
+    CB_CLRCMP_CONTROL                                     = 0x00028c30,
+	CLRCMP_FCN_SRC_mask                               = 0x07 << 0,
+	CLRCMP_FCN_SRC_shift                              = 0,
+	    CLRCMP_DRAW_ALWAYS                            = 0x00,
+	    CLRCMP_DRAW_NEVER                             = 0x01,
+	    CLRCMP_DRAW_ON_NEQ                            = 0x04,
+	    CLRCMP_DRAW_ON_EQ                             = 0x05,
+	CLRCMP_FCN_DST_mask                               = 0x07 << 8,
+	CLRCMP_FCN_DST_shift                              = 8,
+/* 	    CLRCMP_DRAW_ALWAYS                            = 0x00, */
+/* 	    CLRCMP_DRAW_NEVER                             = 0x01, */
+/* 	    CLRCMP_DRAW_ON_NEQ                            = 0x04, */
+/* 	    CLRCMP_DRAW_ON_EQ                             = 0x05, */
+	CLRCMP_FCN_SEL_mask                               = 0x03 << 24,
+	CLRCMP_FCN_SEL_shift                              = 24,
+	    CLRCMP_SEL_DST                                = 0x00,
+	    CLRCMP_SEL_SRC                                = 0x01,
+	    CLRCMP_SEL_AND                                = 0x02,
+    CB_CLRCMP_SRC                                         = 0x00028c34,
+    CB_CLRCMP_DST                                         = 0x00028c38,
+    CB_CLRCMP_MSK                                         = 0x00028c3c,
+    PA_SC_AA_MASK                                         = 0x00028c48,
+    VGT_VERTEX_REUSE_BLOCK_CNTL                           = 0x00028c58,
+	VTX_REUSE_DEPTH_mask                              = 0xff << 0,
+	VTX_REUSE_DEPTH_shift                             = 0,
+    VGT_OUT_DEALLOC_CNTL                                  = 0x00028c5c,
+	DEALLOC_DIST_mask                                 = 0x7f << 0,
+	DEALLOC_DIST_shift                                = 0,
+    DB_RENDER_CONTROL                                     = 0x00028d0c,
+	DEPTH_CLEAR_ENABLE_bit                            = 1 << 0,
+	STENCIL_CLEAR_ENABLE_bit                          = 1 << 1,
+	DEPTH_COPY_bit                                    = 1 << 2,
+	STENCIL_COPY_bit                                  = 1 << 3,
+	RESUMMARIZE_ENABLE_bit                            = 1 << 4,
+	STENCIL_COMPRESS_DISABLE_bit                      = 1 << 5,
+	DEPTH_COMPRESS_DISABLE_bit                        = 1 << 6,
+	COPY_CENTROID_bit                                 = 1 << 7,
+	COPY_SAMPLE_mask                                  = 0x07 << 8,
+	COPY_SAMPLE_shift                                 = 8,
+	ZPASS_INCREMENT_DISABLE_bit                       = 1 << 11,
+    DB_RENDER_OVERRIDE                                    = 0x00028d10,
+	FORCE_HIZ_ENABLE_mask                             = 0x03 << 0,
+	FORCE_HIZ_ENABLE_shift                            = 0,
+	    FORCE_OFF                                     = 0x00,
+	    FORCE_ENABLE                                  = 0x01,
+	    FORCE_DISABLE                                 = 0x02,
+	    FORCE_RESERVED                                = 0x03,
+	FORCE_HIS_ENABLE0_mask                            = 0x03 << 2,
+	FORCE_HIS_ENABLE0_shift                           = 2,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_HIS_ENABLE1_mask                            = 0x03 << 4,
+	FORCE_HIS_ENABLE1_shift                           = 4,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_SHADER_Z_ORDER_bit                          = 1 << 6,
+	FAST_Z_DISABLE_bit                                = 1 << 7,
+	FAST_STENCIL_DISABLE_bit                          = 1 << 8,
+	NOOP_CULL_DISABLE_bit                             = 1 << 9,
+	FORCE_COLOR_KILL_bit                              = 1 << 10,
+	FORCE_Z_READ_bit                                  = 1 << 11,
+	FORCE_STENCIL_READ_bit                            = 1 << 12,
+	FORCE_FULL_Z_RANGE_mask                           = 0x03 << 13,
+	FORCE_FULL_Z_RANGE_shift                          = 13,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_QC_SMASK_CONFLICT_bit                       = 1 << 15,
+	DISABLE_VIEWPORT_CLAMP_bit                        = 1 << 16,
+	IGNORE_SC_ZRANGE_bit                              = 1 << 17,
+    DB_HTILE_SURFACE                                      = 0x00028d24,
+	HTILE_WIDTH_bit                                   = 1 << 0,
+	HTILE_HEIGHT_bit                                  = 1 << 1,
+	LINEAR_bit                                        = 1 << 2,
+	FULL_CACHE_bit                                    = 1 << 3,
+	HTILE_USES_PRELOAD_WIN_bit                        = 1 << 4,
+	PRELOAD_bit                                       = 1 << 5,
+	PREFETCH_WIDTH_mask                               = 0x3f << 6,
+	PREFETCH_WIDTH_shift                              = 6,
+	PREFETCH_HEIGHT_mask                              = 0x3f << 12,
+	PREFETCH_HEIGHT_shift                             = 12,
+    DB_SRESULTS_COMPARE_STATE1                            = 0x00028d2c,
+	COMPAREFUNC1_mask                                 = 0x07 << 0,
+	COMPAREFUNC1_shift                                = 0,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	COMPAREVALUE1_mask                                = 0xff << 4,
+	COMPAREVALUE1_shift                               = 4,
+	COMPAREMASK1_mask                                 = 0xff << 12,
+	COMPAREMASK1_shift                                = 12,
+	ENABLE1_bit                                       = 1 << 24,
+    DB_PRELOAD_CONTROL                                    = 0x00028d30,
+	START_X_mask                                      = 0xff << 0,
+	START_X_shift                                     = 0,
+	START_Y_mask                                      = 0xff << 8,
+	START_Y_shift                                     = 8,
+	MAX_X_mask                                        = 0xff << 16,
+	MAX_X_shift                                       = 16,
+	MAX_Y_mask                                        = 0xff << 24,
+	MAX_Y_shift                                       = 24,
+    DB_PREFETCH_LIMIT                                     = 0x00028d34,
+	DEPTH_HEIGHT_TILE_MAX_mask                        = 0x3ff << 0,
+	DEPTH_HEIGHT_TILE_MAX_shift                       = 0,
+    PA_SU_POLY_OFFSET_DB_FMT_CNTL                         = 0x00028df8,
+	POLY_OFFSET_NEG_NUM_DB_BITS_mask                  = 0xff << 0,
+	POLY_OFFSET_NEG_NUM_DB_BITS_shift                 = 0,
+	POLY_OFFSET_DB_IS_FLOAT_FMT_bit                   = 1 << 8,
+    PA_SU_POLY_OFFSET_CLAMP                               = 0x00028dfc,
+    PA_SU_POLY_OFFSET_FRONT_SCALE                         = 0x00028e00,
+    PA_SU_POLY_OFFSET_FRONT_OFFSET                        = 0x00028e04,
+    PA_SU_POLY_OFFSET_BACK_SCALE                          = 0x00028e08,
+    PA_SU_POLY_OFFSET_BACK_OFFSET                         = 0x00028e0c,
+    PA_CL_POINT_X_RAD                                     = 0x00028e10,
+    PA_CL_POINT_Y_RAD                                     = 0x00028e14,
+    PA_CL_POINT_SIZE                                      = 0x00028e18,
+    PA_CL_POINT_CULL_RAD                                  = 0x00028e1c,
+    PA_CL_UCP_0_X                                         = 0x00028e20,
+	PA_CL_UCP_0_X_num                                 = 6,
+	PA_CL_UCP_0_X_offset                              = 16,
+    PA_CL_UCP_0_Y                                         = 0x00028e24,
+	PA_CL_UCP_0_Y_num                                 = 6,
+	PA_CL_UCP_0_Y_offset                              = 16,
+    PA_CL_UCP_0_Z                                         = 0x00028e28,
+	PA_CL_UCP_0_Z_num                                 = 6,
+	PA_CL_UCP_0_Z_offset                              = 16,
+    SQ_ALU_CONSTANT0_0                                    = 0x00030000,
+    SQ_ALU_CONSTANT1_0                                    = 0x00030004,
+    SQ_ALU_CONSTANT2_0                                    = 0x00030008,
+    SQ_ALU_CONSTANT3_0                                    = 0x0003000c,
+    SQ_VTX_CONSTANT_WORD0_0                               = 0x00038000,
+    SQ_TEX_RESOURCE_WORD0_0                               = 0x00038000,
+	DIM_mask                                          = 0x07 << 0,
+	DIM_shift                                         = 0,
+	    SQ_TEX_DIM_1D                                 = 0x00,
+	    SQ_TEX_DIM_2D                                 = 0x01,
+	    SQ_TEX_DIM_3D                                 = 0x02,
+	    SQ_TEX_DIM_CUBEMAP                            = 0x03,
+	    SQ_TEX_DIM_1D_ARRAY                           = 0x04,
+	    SQ_TEX_DIM_2D_ARRAY                           = 0x05,
+	    SQ_TEX_DIM_2D_MSAA                            = 0x06,
+	    SQ_TEX_DIM_2D_ARRAY_MSAA                      = 0x07,
+	SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask           = 0x0f << 3,
+	SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift          = 3,
+	TILE_TYPE_bit                                     = 1 << 7,
+	PITCH_mask                                        = 0x7ff << 8,
+	PITCH_shift                                       = 8,
+	TEX_WIDTH_mask                                    = 0x1fff << 19,
+	TEX_WIDTH_shift                                   = 19,
+    SQ_VTX_CONSTANT_WORD1_0                               = 0x00038004,
+    SQ_TEX_RESOURCE_WORD1_0                               = 0x00038004,
+	TEX_HEIGHT_mask                                   = 0x1fff << 0,
+	TEX_HEIGHT_shift                                  = 0,
+	TEX_DEPTH_mask                                    = 0x1fff << 13,
+	TEX_DEPTH_shift                                   = 13,
+	SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask         = 0x3f << 26,
+	SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift        = 26,
+    SQ_VTX_CONSTANT_WORD2_0                               = 0x00038008,
+	BASE_ADDRESS_HI_mask                              = 0xff << 0,
+	BASE_ADDRESS_HI_shift                             = 0,
+	SQ_VTX_CONSTANT_WORD2_0__STRIDE_mask              = 0x7ff << 8,
+	SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift             = 8,
+	SQ_VTX_CONSTANT_WORD2_0__CLAMP_X_bit              = 1 << 19,
+	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask         = 0x3f << 20,
+	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift        = 20,
+	SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask      = 0x03 << 26,
+	SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift     = 26,
+/* 	    SQ_NUM_FORMAT_NORM                            = 0x00, */
+/* 	    SQ_NUM_FORMAT_INT                             = 0x01, */
+/* 	    SQ_NUM_FORMAT_SCALED                          = 0x02, */
+	SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit      = 1 << 28,
+	SQ_VTX_CONSTANT_WORD2_0__SRF_MODE_ALL_bit         = 1 << 29,
+	SQ_VTX_CONSTANT_WORD2_0__ENDIAN_SWAP_mask         = 0x03 << 30,
+	SQ_VTX_CONSTANT_WORD2_0__ENDIAN_SWAP_shift        = 30,
+/* 	    SQ_ENDIAN_NONE                                = 0x00, */
+/* 	    SQ_ENDIAN_8IN16                               = 0x01, */
+/* 	    SQ_ENDIAN_8IN32                               = 0x02, */
+    SQ_TEX_RESOURCE_WORD2_0                               = 0x00038008,
+    SQ_VTX_CONSTANT_WORD3_0                               = 0x0003800c,
+	MEM_REQUEST_SIZE_mask                             = 0x03 << 0,
+	MEM_REQUEST_SIZE_shift                            = 0,
+    SQ_TEX_RESOURCE_WORD3_0                               = 0x0003800c,
+    SQ_TEX_RESOURCE_WORD4_0                               = 0x00038010,
+	FORMAT_COMP_X_mask                                = 0x03 << 0,
+	FORMAT_COMP_X_shift                               = 0,
+	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00,
+	    SQ_FORMAT_COMP_SIGNED                         = 0x01,
+	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02,
+	FORMAT_COMP_Y_mask                                = 0x03 << 2,
+	FORMAT_COMP_Y_shift                               = 2,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	FORMAT_COMP_Z_mask                                = 0x03 << 4,
+	FORMAT_COMP_Z_shift                               = 4,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	FORMAT_COMP_W_mask                                = 0x03 << 6,
+	FORMAT_COMP_W_shift                               = 6,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_mask      = 0x03 << 8,
+	SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_shift     = 8,
+/* 	    SQ_NUM_FORMAT_NORM                            = 0x00, */
+/* 	    SQ_NUM_FORMAT_INT                             = 0x01, */
+/* 	    SQ_NUM_FORMAT_SCALED                          = 0x02, */
+	SQ_TEX_RESOURCE_WORD4_0__SRF_MODE_ALL_bit         = 1 << 10,
+	SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit        = 1 << 11,
+	SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_mask         = 0x03 << 12,
+	SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_shift        = 12,
+/* 	    SQ_ENDIAN_NONE                                = 0x00, */
+/* 	    SQ_ENDIAN_8IN16                               = 0x01, */
+/* 	    SQ_ENDIAN_8IN32                               = 0x02, */
+	REQUEST_SIZE_mask                                 = 0x03 << 14,
+	REQUEST_SIZE_shift                                = 14,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask           = 0x07 << 16,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift          = 16,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask           = 0x07 << 19,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift          = 19,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask           = 0x07 << 22,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift          = 22,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask           = 0x07 << 25,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift          = 25,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	BASE_LEVEL_mask                                   = 0x0f << 28,
+	BASE_LEVEL_shift                                  = 28,
+    SQ_TEX_RESOURCE_WORD5_0                               = 0x00038014,
+	LAST_LEVEL_mask                                   = 0x0f << 0,
+	LAST_LEVEL_shift                                  = 0,
+	BASE_ARRAY_mask                                   = 0x1fff << 4,
+	BASE_ARRAY_shift                                  = 4,
+	LAST_ARRAY_mask                                   = 0x1fff << 17,
+	LAST_ARRAY_shift                                  = 17,
+    SQ_TEX_RESOURCE_WORD6_0                               = 0x00038018,
+	MPEG_CLAMP_mask                                   = 0x03 << 0,
+	MPEG_CLAMP_shift                                  = 0,
+	    SQ_TEX_MPEG_CLAMP_OFF                         = 0x00,
+	    SQ_TEX_MPEG_9                                 = 0x01,
+	    SQ_TEX_MPEG_10                                = 0x02,
+	PERF_MODULATION_mask                              = 0x07 << 5,
+	PERF_MODULATION_shift                             = 5,
+	INTERLACED_bit                                    = 1 << 8,
+	SQ_TEX_RESOURCE_WORD6_0__TYPE_mask                = 0x03 << 30,
+	SQ_TEX_RESOURCE_WORD6_0__TYPE_shift               = 30,
+	    SQ_TEX_VTX_INVALID_TEXTURE                    = 0x00,
+	    SQ_TEX_VTX_INVALID_BUFFER                     = 0x01,
+	    SQ_TEX_VTX_VALID_TEXTURE                      = 0x02,
+	    SQ_TEX_VTX_VALID_BUFFER                       = 0x03,
+    SQ_VTX_CONSTANT_WORD6_0                               = 0x00038018,
+	SQ_VTX_CONSTANT_WORD6_0__TYPE_mask                = 0x03 << 30,
+	SQ_VTX_CONSTANT_WORD6_0__TYPE_shift               = 30,
+/* 	    SQ_TEX_VTX_INVALID_TEXTURE                    = 0x00, */
+/* 	    SQ_TEX_VTX_INVALID_BUFFER                     = 0x01, */
+/* 	    SQ_TEX_VTX_VALID_TEXTURE                      = 0x02, */
+/* 	    SQ_TEX_VTX_VALID_BUFFER                       = 0x03, */
+    SQ_TEX_SAMPLER_WORD0_0                                = 0x0003c000,
+	SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_mask              = 0x07 << 0,
+	SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift             = 0,
+	    SQ_TEX_WRAP                                   = 0x00,
+	    SQ_TEX_MIRROR                                 = 0x01,
+	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02,
+	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03,
+	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04,
+	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05,
+	    SQ_TEX_CLAMP_BORDER                           = 0x06,
+	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07,
+	CLAMP_Y_mask                                      = 0x07 << 3,
+	CLAMP_Y_shift                                     = 3,
+/* 	    SQ_TEX_WRAP                                   = 0x00, */
+/* 	    SQ_TEX_MIRROR                                 = 0x01, */
+/* 	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02, */
+/* 	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03, */
+/* 	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04, */
+/* 	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05, */
+/* 	    SQ_TEX_CLAMP_BORDER                           = 0x06, */
+/* 	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07, */
+	CLAMP_Z_mask                                      = 0x07 << 6,
+	CLAMP_Z_shift                                     = 6,
+/* 	    SQ_TEX_WRAP                                   = 0x00, */
+/* 	    SQ_TEX_MIRROR                                 = 0x01, */
+/* 	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02, */
+/* 	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03, */
+/* 	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04, */
+/* 	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05, */
+/* 	    SQ_TEX_CLAMP_BORDER                           = 0x06, */
+/* 	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07, */
+	XY_MAG_FILTER_mask                                = 0x07 << 9,
+	XY_MAG_FILTER_shift                               = 9,
+	    SQ_TEX_XY_FILTER_POINT                        = 0x00,
+	    SQ_TEX_XY_FILTER_BILINEAR                     = 0x01,
+	    SQ_TEX_XY_FILTER_BICUBIC                      = 0x02,
+	XY_MIN_FILTER_mask                                = 0x07 << 12,
+	XY_MIN_FILTER_shift                               = 12,
+/* 	    SQ_TEX_XY_FILTER_POINT                        = 0x00, */
+/* 	    SQ_TEX_XY_FILTER_BILINEAR                     = 0x01, */
+/* 	    SQ_TEX_XY_FILTER_BICUBIC                      = 0x02, */
+	Z_FILTER_mask                                     = 0x03 << 15,
+	Z_FILTER_shift                                    = 15,
+	    SQ_TEX_Z_FILTER_NONE                          = 0x00,
+	    SQ_TEX_Z_FILTER_POINT                         = 0x01,
+	    SQ_TEX_Z_FILTER_LINEAR                        = 0x02,
+	MIP_FILTER_mask                                   = 0x03 << 17,
+	MIP_FILTER_shift                                  = 17,
+/* 	    SQ_TEX_Z_FILTER_NONE                          = 0x00, */
+/* 	    SQ_TEX_Z_FILTER_POINT                         = 0x01, */
+/* 	    SQ_TEX_Z_FILTER_LINEAR                        = 0x02, */
+	BORDER_COLOR_TYPE_mask                            = 0x03 << 22,
+	BORDER_COLOR_TYPE_shift                           = 22,
+	    SQ_TEX_BORDER_COLOR_TRANS_BLACK               = 0x00,
+	    SQ_TEX_BORDER_COLOR_OPAQUE_BLACK              = 0x01,
+	    SQ_TEX_BORDER_COLOR_OPAQUE_WHITE              = 0x02,
+	    SQ_TEX_BORDER_COLOR_REGISTER                  = 0x03,
+	POINT_SAMPLING_CLAMP_bit                          = 1 << 24,
+	TEX_ARRAY_OVERRIDE_bit                            = 1 << 25,
+	DEPTH_COMPARE_FUNCTION_mask                       = 0x07 << 26,
+	DEPTH_COMPARE_FUNCTION_shift                      = 26,
+	    SQ_TEX_DEPTH_COMPARE_NEVER                    = 0x00,
+	    SQ_TEX_DEPTH_COMPARE_LESS                     = 0x01,
+	    SQ_TEX_DEPTH_COMPARE_EQUAL                    = 0x02,
+	    SQ_TEX_DEPTH_COMPARE_LESSEQUAL                = 0x03,
+	    SQ_TEX_DEPTH_COMPARE_GREATER                  = 0x04,
+	    SQ_TEX_DEPTH_COMPARE_NOTEQUAL                 = 0x05,
+	    SQ_TEX_DEPTH_COMPARE_GREATEREQUAL             = 0x06,
+	    SQ_TEX_DEPTH_COMPARE_ALWAYS                   = 0x07,
+	CHROMA_KEY_mask                                   = 0x03 << 29,
+	CHROMA_KEY_shift                                  = 29,
+	    SQ_TEX_CHROMA_KEY_DISABLED                    = 0x00,
+	    SQ_TEX_CHROMA_KEY_KILL                        = 0x01,
+	    SQ_TEX_CHROMA_KEY_BLEND                       = 0x02,
+	LOD_USES_MINOR_AXIS_bit                           = 1 << 31,
+    SQ_TEX_SAMPLER_WORD1_0                                = 0x0003c004,
+	MIN_LOD_mask                                      = 0x3ff << 0,
+	MIN_LOD_shift                                     = 0,
+	MAX_LOD_mask                                      = 0x3ff << 10,
+	MAX_LOD_shift                                     = 10,
+	SQ_TEX_SAMPLER_WORD1_0__LOD_BIAS_mask             = 0xfff << 20,
+	SQ_TEX_SAMPLER_WORD1_0__LOD_BIAS_shift            = 20,
+    SQ_TEX_SAMPLER_WORD2_0                                = 0x0003c008,
+	LOD_BIAS_SEC_mask                                 = 0xfff << 0,
+	LOD_BIAS_SEC_shift                                = 0,
+	MC_COORD_TRUNCATE_bit                             = 1 << 12,
+	SQ_TEX_SAMPLER_WORD2_0__FORCE_DEGAMMA_bit         = 1 << 13,
+	HIGH_PRECISION_FILTER_bit                         = 1 << 14,
+	PERF_MIP_mask                                     = 0x07 << 15,
+	PERF_MIP_shift                                    = 15,
+	PERF_Z_mask                                       = 0x03 << 18,
+	PERF_Z_shift                                      = 18,
+	FETCH_4_bit                                       = 1 << 26,
+	SAMPLE_IS_PCF_bit                                 = 1 << 27,
+	SQ_TEX_SAMPLER_WORD2_0__TYPE_bit                  = 1 << 31,
+    SQ_VTX_BASE_VTX_LOC                                   = 0x0003cff0,
+    SQ_VTX_START_INST_LOC                                 = 0x0003cff4,
+    SQ_LOOP_CONST_DX10_0                                  = 0x0003e200,
+    SQ_LOOP_CONST_0                                       = 0x0003e200,
+	SQ_LOOP_CONST_0__COUNT_mask                       = 0xfff << 0,
+	SQ_LOOP_CONST_0__COUNT_shift                      = 0,
+	INIT_mask                                         = 0xfff << 12,
+	INIT_shift                                        = 12,
+	INC_mask                                          = 0xff << 24,
+	INC_shift                                         = 24,
+    SQ_BOOL_CONST_0                                       = 0x0003e380,
+	SQ_BOOL_CONST_0_num                               = 3,
+
+} ;
+
+#endif /* _AUTOREGS */
+
diff --git a/src/mesa/drivers/dri/r600/r600_reg_r6xx.h b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h
new file mode 100644
index 0000000000..f7702c46de
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h
@@ -0,0 +1,492 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_R6xx_H_
+#define _R600_REG_R6xx_H_
+
+/*
+ * Registers for R6xx chips that are not documented yet
+ */
+
+enum {
+
+    MM_INDEX                                              = 0x0000,
+    MM_DATA                                               = 0x0004,
+
+    SRBM_STATUS                                           = 0x0e50,
+	RLC_RQ_PENDING_bit                                = 1 << 3,
+	RCU_RQ_PENDING_bit                                = 1 << 4,
+	GRBM_RQ_PENDING_bit                               = 1 << 5,
+	HI_RQ_PENDING_bit                                 = 1 << 6,
+	IO_EXTERN_SIGNAL_bit                              = 1 << 7,
+	VMC_BUSY_bit                                      = 1 << 8,
+	MCB_BUSY_bit                                      = 1 << 9,
+	MCDZ_BUSY_bit                                     = 1 << 10,
+	MCDY_BUSY_bit                                     = 1 << 11,
+	MCDX_BUSY_bit                                     = 1 << 12,
+	MCDW_BUSY_bit                                     = 1 << 13,
+	SEM_BUSY_bit                                      = 1 << 14,
+	SRBM_STATUS__RLC_BUSY_bit                         = 1 << 15,
+	PDMA_BUSY_bit                                     = 1 << 16,
+	IH_BUSY_bit                                       = 1 << 17,
+	CSC_BUSY_bit                                      = 1 << 20,
+	CMC7_BUSY_bit                                     = 1 << 21,
+	CMC6_BUSY_bit                                     = 1 << 22,
+	CMC5_BUSY_bit                                     = 1 << 23,
+	CMC4_BUSY_bit                                     = 1 << 24,
+	CMC3_BUSY_bit                                     = 1 << 25,
+	CMC2_BUSY_bit                                     = 1 << 26,
+	CMC1_BUSY_bit                                     = 1 << 27,
+	CMC0_BUSY_bit                                     = 1 << 28,
+	BIF_BUSY_bit                                      = 1 << 29,
+	IDCT_BUSY_bit                                     = 1 << 30,
+
+    SRBM_READ_ERROR                                       = 0x0e98,
+	READ_ADDRESS_mask                                 = 0xffff << 2,
+	READ_ADDRESS_shift                                = 2,
+	READ_REQUESTER_HI_bit                             = 1 << 24,
+	READ_REQUESTER_GRBM_bit                           = 1 << 25,
+	READ_REQUESTER_RCU_bit                            = 1 << 26,
+	READ_REQUESTER_RLC_bit                            = 1 << 27,
+	READ_ERROR_bit                                    = 1 << 31,
+
+    SRBM_INT_STATUS                                       = 0x0ea4,
+	RDERR_INT_STAT_bit                                = 1 << 0,
+	GFX_CNTX_SWITCH_INT_STAT_bit                      = 1 << 1,
+    SRBM_INT_ACK                                          = 0x0ea8,
+	RDERR_INT_ACK_bit                                 = 1 << 0,
+	GFX_CNTX_SWITCH_INT_ACK_bit                       = 1 << 1,
+
+    R6XX_MC_VM_FB_LOCATION                                = 0x2180,
+
+    VENDOR_DEVICE_ID                                      = 0x4000,
+
+    D1GRPH_PRIMARY_SURFACE_ADDRESS                        = 0x6110,
+    D1GRPH_PITCH                                          = 0x6120,
+    D1GRPH_Y_END                                          = 0x6138,
+
+    GRBM_STATUS                                           = 0x8010,
+	CMDFIFO_AVAIL_mask                                = 0x1f << 0,
+	CMDFIFO_AVAIL_shift                               = 0,
+	SRBM_RQ_PENDING_bit                               = 1 << 5,
+	CP_RQ_PENDING_bit                                 = 1 << 6,
+	CF_RQ_PENDING_bit                                 = 1 << 7,
+	PF_RQ_PENDING_bit                                 = 1 << 8,
+	GRBM_EE_BUSY_bit                                  = 1 << 10,
+	GRBM_STATUS__VC_BUSY_bit                          = 1 << 11,
+	DB03_CLEAN_bit                                    = 1 << 12,
+	CB03_CLEAN_bit                                    = 1 << 13,
+	VGT_BUSY_NO_DMA_bit                               = 1 << 16,
+	GRBM_STATUS__VGT_BUSY_bit                         = 1 << 17,
+	TA03_BUSY_bit                                     = 1 << 18,
+	GRBM_STATUS__TC_BUSY_bit                          = 1 << 19,
+	SX_BUSY_bit                                       = 1 << 20,
+	SH_BUSY_bit                                       = 1 << 21,
+	SPI03_BUSY_bit                                    = 1 << 22,
+	SMX_BUSY_bit                                      = 1 << 23,
+	SC_BUSY_bit                                       = 1 << 24,
+	PA_BUSY_bit                                       = 1 << 25,
+	DB03_BUSY_bit                                     = 1 << 26,
+	CR_BUSY_bit                                       = 1 << 27,
+	CP_COHERENCY_BUSY_bit                             = 1 << 28,
+	GRBM_STATUS__CP_BUSY_bit                          = 1 << 29,
+	CB03_BUSY_bit                                     = 1 << 30,
+	GUI_ACTIVE_bit                                    = 1 << 31,
+    GRBM_STATUS2                                          = 0x8014,
+	CR_CLEAN_bit                                      = 1 << 0,
+	SMX_CLEAN_bit                                     = 1 << 1,
+	SPI0_BUSY_bit                                     = 1 << 8,
+	SPI1_BUSY_bit                                     = 1 << 9,
+	SPI2_BUSY_bit                                     = 1 << 10,
+	SPI3_BUSY_bit                                     = 1 << 11,
+	TA0_BUSY_bit                                      = 1 << 12,
+	TA1_BUSY_bit                                      = 1 << 13,
+	TA2_BUSY_bit                                      = 1 << 14,
+	TA3_BUSY_bit                                      = 1 << 15,
+	DB0_BUSY_bit                                      = 1 << 16,
+	DB1_BUSY_bit                                      = 1 << 17,
+	DB2_BUSY_bit                                      = 1 << 18,
+	DB3_BUSY_bit                                      = 1 << 19,
+	CB0_BUSY_bit                                      = 1 << 20,
+	CB1_BUSY_bit                                      = 1 << 21,
+	CB2_BUSY_bit                                      = 1 << 22,
+	CB3_BUSY_bit                                      = 1 << 23,
+    GRBM_SOFT_RESET                                       = 0x8020,
+	SOFT_RESET_CP_bit                                 = 1 << 0,
+	SOFT_RESET_CB_bit                                 = 1 << 1,
+	SOFT_RESET_CR_bit                                 = 1 << 2,
+	SOFT_RESET_DB_bit                                 = 1 << 3,
+	SOFT_RESET_PA_bit                                 = 1 << 5,
+	SOFT_RESET_SC_bit                                 = 1 << 6,
+	SOFT_RESET_SMX_bit                                = 1 << 7,
+	SOFT_RESET_SPI_bit                                = 1 << 8,
+	SOFT_RESET_SH_bit                                 = 1 << 9,
+	SOFT_RESET_SX_bit                                 = 1 << 10,
+	SOFT_RESET_TC_bit                                 = 1 << 11,
+	SOFT_RESET_TA_bit                                 = 1 << 12,
+	SOFT_RESET_VC_bit                                 = 1 << 13,
+	SOFT_RESET_VGT_bit                                = 1 << 14,
+	SOFT_RESET_GRBM_GCA_bit                           = 1 << 15,
+
+    WAIT_UNTIL                                            = 0x8040,
+	WAIT_CP_DMA_IDLE_bit                              = 1 << 8,
+	WAIT_CMDFIFO_bit                                  = 1 << 10,
+	WAIT_2D_IDLE_bit                                  = 1 << 14,
+	WAIT_3D_IDLE_bit                                  = 1 << 15,
+	WAIT_2D_IDLECLEAN_bit                             = 1 << 16,
+	WAIT_3D_IDLECLEAN_bit                             = 1 << 17,
+	WAIT_EXTERN_SIG_bit                               = 1 << 19,
+	CMDFIFO_ENTRIES_mask                              = 0x1f << 20,
+	CMDFIFO_ENTRIES_shift                             = 20,
+
+    GRBM_READ_ERROR                                       = 0x8058,
+/* 	READ_ADDRESS_mask                                 = 0xffff << 2, */
+/* 	READ_ADDRESS_shift                                = 2, */
+	READ_REQUESTER_SRBM_bit                           = 1 << 28,
+	READ_REQUESTER_CP_bit                             = 1 << 29,
+	READ_REQUESTER_WU_POLL_bit                        = 1 << 30,
+/* 	READ_ERROR_bit                                    = 1 << 31, */
+
+    SCRATCH_REG0		                          = 0x8500,
+    SCRATCH_REG1		                          = 0x8504,
+    SCRATCH_REG2		                          = 0x8508,
+    SCRATCH_REG3		                          = 0x850c,
+    SCRATCH_REG4		                          = 0x8510,
+    SCRATCH_REG5		                          = 0x8514,
+    SCRATCH_REG6		                          = 0x8518,
+    SCRATCH_REG7		                          = 0x851c,
+    SCRATCH_UMSK		                          = 0x8540,
+    SCRATCH_ADDR		                          = 0x8544,
+
+    CP_COHER_CNTL                                         = 0x85f0,
+	DEST_BASE_0_ENA_bit                               = 1 << 0,
+	DEST_BASE_1_ENA_bit                               = 1 << 1,
+	SO0_DEST_BASE_ENA_bit                             = 1 << 2,
+	SO1_DEST_BASE_ENA_bit                             = 1 << 3,
+	SO2_DEST_BASE_ENA_bit                             = 1 << 4,
+	SO3_DEST_BASE_ENA_bit                             = 1 << 5,
+	CB0_DEST_BASE_ENA_bit                             = 1 << 6,
+	CB1_DEST_BASE_ENA_bit                             = 1 << 7,
+	CB2_DEST_BASE_ENA_bit                             = 1 << 8,
+	CB3_DEST_BASE_ENA_bit                             = 1 << 9,
+	CB4_DEST_BASE_ENA_bit                             = 1 << 10,
+	CB5_DEST_BASE_ENA_bit                             = 1 << 11,
+	CB6_DEST_BASE_ENA_bit                             = 1 << 12,
+	CB7_DEST_BASE_ENA_bit                             = 1 << 13,
+	DB_DEST_BASE_ENA_bit                              = 1 << 14,
+	CR_DEST_BASE_ENA_bit                              = 1 << 15,
+	TC_ACTION_ENA_bit                                 = 1 << 23,
+	VC_ACTION_ENA_bit                                 = 1 << 24,
+	CB_ACTION_ENA_bit                                 = 1 << 25,
+	DB_ACTION_ENA_bit                                 = 1 << 26,
+	SH_ACTION_ENA_bit                                 = 1 << 27,
+	SMX_ACTION_ENA_bit                                = 1 << 28,
+	CR0_ACTION_ENA_bit                                = 1 << 29,
+	CR1_ACTION_ENA_bit                                = 1 << 30,
+	CR2_ACTION_ENA_bit                                = 1 << 31,
+    CP_COHER_SIZE                                         = 0x85f4,
+    CP_COHER_BASE                                         = 0x85f8,
+    CP_COHER_STATUS                                       = 0x85fc,
+	MATCHING_GFX_CNTX_mask                            = 0xff << 0,
+	MATCHING_GFX_CNTX_shift                           = 0,
+	MATCHING_CR_CNTX_mask                             = 0xffff << 8,
+	MATCHING_CR_CNTX_shift                            = 8,
+	STATUS_bit                                        = 1 << 31,
+
+    CP_STALLED_STAT1                                      = 0x8674,
+	RBIU_TO_DMA_NOT_RDY_TO_RCV_bit                    = 1 << 0,
+	RBIU_TO_IBS_NOT_RDY_TO_RCV_bit                    = 1 << 1,
+	RBIU_TO_SEM_NOT_RDY_TO_RCV_bit                    = 1 << 2,
+	RBIU_TO_2DREGS_NOT_RDY_TO_RCV_bit                 = 1 << 3,
+	RBIU_TO_MEMWR_NOT_RDY_TO_RCV_bit                  = 1 << 4,
+	RBIU_TO_MEMRD_NOT_RDY_TO_RCV_bit                  = 1 << 5,
+	RBIU_TO_EOPD_NOT_RDY_TO_RCV_bit                   = 1 << 6,
+	RBIU_TO_RECT_NOT_RDY_TO_RCV_bit                   = 1 << 7,
+	RBIU_TO_STRMO_NOT_RDY_TO_RCV_bit                  = 1 << 8,
+	RBIU_TO_PSTAT_NOT_RDY_TO_RCV_bit                  = 1 << 9,
+	MIU_WAITING_ON_RDREQ_FREE_bit                     = 1 << 16,
+	MIU_WAITING_ON_WRREQ_FREE_bit                     = 1 << 17,
+	MIU_NEEDS_AVAIL_WRREQ_PHASE_bit                   = 1 << 18,
+	RCIU_WAITING_ON_GRBM_FREE_bit                     = 1 << 24,
+	RCIU_WAITING_ON_VGT_FREE_bit                      = 1 << 25,
+	RCIU_STALLED_ON_ME_READ_bit                       = 1 << 26,
+	RCIU_STALLED_ON_DMA_READ_bit                      = 1 << 27,
+	RCIU_HALTED_BY_REG_VIOLATION_bit                  = 1 << 28,
+    CP_STALLED_STAT2                                      = 0x8678,
+	PFP_TO_CSF_NOT_RDY_TO_RCV_bit                     = 1 << 0,
+	PFP_TO_MEQ_NOT_RDY_TO_RCV_bit                     = 1 << 1,
+	PFP_TO_VGT_NOT_RDY_TO_RCV_bit                     = 1 << 2,
+	PFP_HALTED_BY_INSTR_VIOLATION_bit                 = 1 << 3,
+	MULTIPASS_IB_PENDING_IN_PFP_bit                   = 1 << 4,
+	ME_BRUSH_WC_NOT_RDY_TO_RCV_bit                    = 1 << 8,
+	ME_STALLED_ON_BRUSH_LOGIC_bit                     = 1 << 9,
+	CR_CNTX_NOT_AVAIL_TO_ME_bit                       = 1 << 10,
+	GFX_CNTX_NOT_AVAIL_TO_ME_bit                      = 1 << 11,
+	ME_RCIU_NOT_RDY_TO_RCV_bit                        = 1 << 12,
+	ME_TO_CONST_NOT_RDY_TO_RCV_bit                    = 1 << 13,
+	ME_WAITING_DATA_FROM_PFP_bit                      = 1 << 14,
+	ME_WAITING_ON_PARTIAL_FLUSH_bit                   = 1 << 15,
+	RECT_FIFO_NEEDS_CR_RECT_DONE_bit                  = 1 << 16,
+	RECT_FIFO_NEEDS_WR_CONFIRM_bit                    = 1 << 17,
+	EOPD_FIFO_NEEDS_SC_EOP_DONE_bit                   = 1 << 18,
+	EOPD_FIFO_NEEDS_SMX_EOP_DONE_bit                  = 1 << 19,
+	EOPD_FIFO_NEEDS_WR_CONFIRM_bit                    = 1 << 20,
+	EOPD_FIFO_NEEDS_SIGNAL_SEM_bit                    = 1 << 21,
+	SO_NUMPRIM_FIFO_NEEDS_SOADDR_bit                  = 1 << 22,
+	SO_NUMPRIM_FIFO_NEEDS_NUMPRIM_bit                 = 1 << 23,
+	PIPE_STATS_FIFO_NEEDS_SAMPLE_bit                  = 1 << 24,
+	SURF_SYNC_NEEDS_IDLE_CNTXS_bit                    = 1 << 30,
+	SURF_SYNC_NEEDS_ALL_CLEAN_bit                     = 1 << 31,
+    CP_BUSY_STAT                                          = 0x867c,
+	REG_BUS_FIFO_BUSY_bit                             = 1 << 0,
+	RING_FETCHING_DATA_bit                            = 1 << 1,
+	INDR1_FETCHING_DATA_bit                           = 1 << 2,
+	INDR2_FETCHING_DATA_bit                           = 1 << 3,
+	STATE_FETCHING_DATA_bit                           = 1 << 4,
+	PRED_FETCHING_DATA_bit                            = 1 << 5,
+	COHER_CNTR_NEQ_ZERO_bit                           = 1 << 6,
+	PFP_PARSING_PACKETS_bit                           = 1 << 7,
+	ME_PARSING_PACKETS_bit                            = 1 << 8,
+	RCIU_PFP_BUSY_bit                                 = 1 << 9,
+	RCIU_ME_BUSY_bit                                  = 1 << 10,
+	OUTSTANDING_READ_TAGS_bit                         = 1 << 11,
+	SEM_CMDFIFO_NOT_EMPTY_bit                         = 1 << 12,
+	SEM_FAILED_AND_HOLDING_bit                        = 1 << 13,
+	SEM_POLLING_FOR_PASS_bit                          = 1 << 14,
+	_3D_BUSY_bit                                      = 1 << 15,
+	_2D_BUSY_bit                                      = 1 << 16,
+    CP_STAT                                               = 0x8680,
+	CSF_RING_BUSY_bit                                 = 1 << 0,
+	CSF_WPTR_POLL_BUSY_bit                            = 1 << 1,
+	CSF_INDIRECT1_BUSY_bit                            = 1 << 2,
+	CSF_INDIRECT2_BUSY_bit                            = 1 << 3,
+	CSF_STATE_BUSY_bit                                = 1 << 4,
+	CSF_PREDICATE_BUSY_bit                            = 1 << 5,
+	CSF_BUSY_bit                                      = 1 << 6,
+	MIU_RDREQ_BUSY_bit                                = 1 << 7,
+	MIU_WRREQ_BUSY_bit                                = 1 << 8,
+	ROQ_RING_BUSY_bit                                 = 1 << 9,
+	ROQ_INDIRECT1_BUSY_bit                            = 1 << 10,
+	ROQ_INDIRECT2_BUSY_bit                            = 1 << 11,
+	ROQ_STATE_BUSY_bit                                = 1 << 12,
+	ROQ_PREDICATE_BUSY_bit                            = 1 << 13,
+	ROQ_ALIGN_BUSY_bit                                = 1 << 14,
+	PFP_BUSY_bit                                      = 1 << 15,
+	MEQ_BUSY_bit                                      = 1 << 16,
+	ME_BUSY_bit                                       = 1 << 17,
+	QUERY_BUSY_bit                                    = 1 << 18,
+	SEMAPHORE_BUSY_bit                                = 1 << 19,
+	INTERRUPT_BUSY_bit                                = 1 << 20,
+	SURFACE_SYNC_BUSY_bit                             = 1 << 21,
+	DMA_BUSY_bit                                      = 1 << 22,
+	RCIU_BUSY_bit                                     = 1 << 23,
+	CP_STAT__CP_BUSY_bit                              = 1 << 31,
+
+    CP_ME_CNTL                                            = 0x86d8,
+	ME_STATMUX_mask                                   = 0xff << 0,
+	ME_STATMUX_shift                                  = 0,
+	ME_HALT_bit                                       = 1 << 28,
+    CP_ME_STATUS                                          = 0x86dc,
+
+    CP_RB_RPTR                                            = 0x8700,
+	RB_RPTR_mask                                      = 0xfffff << 0,
+	RB_RPTR_shift                                     = 0,
+    CP_RB_WPTR_DELAY                                      = 0x8704,
+	PRE_WRITE_TIMER_mask                              = 0xfffffff << 0,
+	PRE_WRITE_TIMER_shift                             = 0,
+	PRE_WRITE_LIMIT_mask                              = 0x0f << 28,
+	PRE_WRITE_LIMIT_shift                             = 28,
+
+    CP_ROQ_RB_STAT                                        = 0x8780,
+	ROQ_RPTR_PRIMARY_mask                             = 0x3ff << 0,
+	ROQ_RPTR_PRIMARY_shift                            = 0,
+	ROQ_WPTR_PRIMARY_mask                             = 0x3ff << 16,
+	ROQ_WPTR_PRIMARY_shift                            = 16,
+    CP_ROQ_IB1_STAT                                       = 0x8784,
+	ROQ_RPTR_INDIRECT1_mask                           = 0x3ff << 0,
+	ROQ_RPTR_INDIRECT1_shift                          = 0,
+	ROQ_WPTR_INDIRECT1_mask                           = 0x3ff << 16,
+	ROQ_WPTR_INDIRECT1_shift                          = 16,
+    CP_ROQ_IB2_STAT                                       = 0x8788,
+	ROQ_RPTR_INDIRECT2_mask                           = 0x3ff << 0,
+	ROQ_RPTR_INDIRECT2_shift                          = 0,
+	ROQ_WPTR_INDIRECT2_mask                           = 0x3ff << 16,
+	ROQ_WPTR_INDIRECT2_shift                          = 16,
+
+    CP_MEQ_STAT                                           = 0x8794,
+	MEQ_RPTR_mask                                     = 0x3ff << 0,
+	MEQ_RPTR_shift                                    = 0,
+	MEQ_WPTR_mask                                     = 0x3ff << 16,
+	MEQ_WPTR_shift                                    = 16,
+
+    CC_GC_SHADER_PIPE_CONFIG                              = 0x8950,
+	INACTIVE_QD_PIPES_mask                            = 0xff << 8,
+	INACTIVE_QD_PIPES_shift                           = 8,
+	    R6XX_MAX_QD_PIPES                             = 8,
+	INACTIVE_SIMDS_mask                               = 0xff << 16,
+	INACTIVE_SIMDS_shift                              = 16,
+	    R6XX_MAX_SIMDS                                = 8,
+    GC_USER_SHADER_PIPE_CONFIG                            = 0x8954,
+
+    VC_ENHANCE                                            = 0x9714,
+    DB_DEBUG                                              = 0x9830,
+        PREZ_MUST_WAIT_FOR_POSTZ_DONE                     = 1 << 31,
+
+    DB_WATERMARKS                                         = 0x00009838,
+	DEPTH_FREE_mask                                   = 0x1f << 0,
+	DEPTH_FREE_shift                                  = 0,
+	DEPTH_FLUSH_mask                                  = 0x3f << 5,
+	DEPTH_FLUSH_shift                                 = 5,
+	FORCE_SUMMARIZE_mask                              = 0x0f << 11,
+	FORCE_SUMMARIZE_shift                             = 11,
+	DEPTH_PENDING_FREE_mask                           = 0x1f << 15,
+	DEPTH_PENDING_FREE_shift                          = 15,
+	DEPTH_CACHELINE_FREE_mask                         = 0x1f << 20,
+	DEPTH_CACHELINE_FREE_shift                        = 20,
+	EARLY_Z_PANIC_DISABLE_bit                         = 1 << 25,
+	LATE_Z_PANIC_DISABLE_bit                          = 1 << 26,
+	RE_Z_PANIC_DISABLE_bit                            = 1 << 27,
+	DB_EXTRA_DEBUG_mask                               = 0x0f << 28,
+	DB_EXTRA_DEBUG_shift                              = 28,
+
+    CP_RB_BASE                                            = 0xc100,
+    CP_RB_CNTL                                            = 0xc104,
+        RB_BUFSZ_mask                                     = 0x3f << 0,
+    CP_RB_WPTR                                            = 0xc114,
+	RB_WPTR_mask                                      = 0xfffff << 0,
+	RB_WPTR_shift                                     = 0,
+    CP_RB_RPTR_WR                                         = 0xc108,
+	RB_RPTR_WR_mask                                   = 0xfffff << 0,
+	RB_RPTR_WR_shift                                  = 0,
+
+    CP_INT_STATUS                                         = 0xc128,
+	DISABLE_CNTX_SWITCH_INT_STAT_bit                  = 1 << 0,
+	ENABLE_CNTX_SWITCH_INT_STAT_bit                   = 1 << 1,
+	SEM_SIGNAL_INT_STAT_bit                           = 1 << 18,
+	CNTX_BUSY_INT_STAT_bit                            = 1 << 19,
+	CNTX_EMPTY_INT_STAT_bit                           = 1 << 20,
+	WAITMEM_SEM_INT_STAT_bit                          = 1 << 21,
+	PRIV_INSTR_INT_STAT_bit                           = 1 << 22,
+	PRIV_REG_INT_STAT_bit                             = 1 << 23,
+	OPCODE_ERROR_INT_STAT_bit                         = 1 << 24,
+	SCRATCH_INT_STAT_bit                              = 1 << 25,
+	TIME_STAMP_INT_STAT_bit                           = 1 << 26,
+	RESERVED_BIT_ERROR_INT_STAT_bit                   = 1 << 27,
+	DMA_INT_STAT_bit                                  = 1 << 28,
+	IB2_INT_STAT_bit                                  = 1 << 29,
+	IB1_INT_STAT_bit                                  = 1 << 30,
+	RB_INT_STAT_bit                                   = 1 << 31,
+
+//  SX_ALPHA_TEST_CONTROL                                 = 0x00028410,
+	ALPHA_FUNC__REF_NEVER                             = 0,
+	ALPHA_FUNC__REF_ALWAYS                            = 7,
+//  DB_SHADER_CONTROL                                     = 0x0002880c,
+	Z_ORDER__EARLY_Z_THEN_LATE_Z                      = 2,
+//  PA_SU_SC_MODE_CNTL                                    = 0x00028814,
+//	POLY_MODE_mask                                    = 0x03 << 3,
+	POLY_MODE__TRIANGLES = 0, POLY_MODE__DUAL_MODE,
+//	POLYMODE_FRONT_PTYPE_mask                         = 0x07 << 5,
+	POLYMODE_PTYPE__POINTS = 0, POLYMODE_PTYPE__LINES, POLYMODE_PTYPE__TRIANGLES,
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1_M                         = 0x00028c20,
+    DB_SRESULTS_COMPARE_STATE0                            = 0x00028d28,	/* See autoregs: DB_SRESULTS_COMPARE_STATE1 */
+//  DB_SRESULTS_COMPARE_STATE1                            = 0x00028d2c,
+    DB_ALPHA_TO_MASK                                      = 0x00028d44,
+	ALPHA_TO_MASK_ENABLE                              = 1 << 0,
+	ALPHA_TO_MASK_OFFSET0_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET0_shift                       = 8,
+	ALPHA_TO_MASK_OFFSET1_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET1_shift                       = 10,
+	ALPHA_TO_MASK_OFFSET2_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET2_shift                       = 12,
+	ALPHA_TO_MASK_OFFSET3_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET3_shift                       = 14,
+
+//  SQ_VTX_CONSTANT_WORD2_0                               = 0x00038008,
+//    	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask         = 0x3f << 20,
+	FMT_INVALID=0,      FMT_8,          FMT_4_4,            FMT_3_3_2,
+	                    FMT_16=5,       FMT_16_FLOAT,       FMT_8_8,
+	FMT_5_6_5,          FMT_6_5_5,      FMT_1_5_5_5,        FMT_4_4_4_4,
+	FMT_5_5_5_1,        FMT_32,         FMT_32_FLOAT,       FMT_16_16,
+	FMT_16_16_FLOAT=16, FMT_8_24,       FMT_8_24_FLOAT,     FMT_24_8,
+	FMT_24_8_FLOAT,     FMT_10_11_11,   FMT_10_11_11_FLOAT, FMT_11_11_10,
+	FMT_11_11_10_FLOAT, FMT_2_10_10_10, FMT_8_8_8_8,        FMT_10_10_10_2,
+	FMT_X24_8_32_FLOAT, FMT_32_32,      FMT_32_32_FLOAT,    FMT_16_16_16_16,
+	FMT_16_16_16_16_FLOAT=32,           FMT_32_32_32_32=34, FMT_32_32_32_32_FLOAT,
+	                    FMT_1 = 37,                         FMT_GB_GR=39,
+	FMT_BG_RG,          FMT_32_AS_8,    FMT_32_AS_8_8,      FMT_5_9_9_9_SHAREDEXP,
+	FMT_8_8_8,          FMT_16_16_16,   FMT_16_16_16_FLOAT, FMT_32_32_32,
+	FMT_32_32_32_FLOAT=48,
+
+//  High level register file lengths
+    SQ_ALU_CONSTANT                                       = SQ_ALU_CONSTANT0_0,	/* 256 PS, 256 VS */
+    SQ_ALU_CONSTANT_ps_num                                = 256,
+    SQ_ALU_CONSTANT_vs_num                                = 256,
+    SQ_ALU_CONSTANT_all_num                               = 512,
+    SQ_ALU_CONSTANT_offset                                = 16,
+    SQ_ALU_CONSTANT_ps                                    = 0,
+    SQ_ALU_CONSTANT_vs                                    = SQ_ALU_CONSTANT_ps + SQ_ALU_CONSTANT_ps_num,
+    SQ_TEX_RESOURCE                                       = SQ_TEX_RESOURCE_WORD0_0,	/* 160 PS, 160 VS, 16 FS, 160 GS */
+    SQ_TEX_RESOURCE_ps_num                                = 160,
+    SQ_TEX_RESOURCE_vs_num                                = 160,
+    SQ_TEX_RESOURCE_fs_num                                = 16,
+    SQ_TEX_RESOURCE_gs_num                                = 160,
+    SQ_TEX_RESOURCE_all_num                               = 496,
+    SQ_TEX_RESOURCE_offset                                = 28,
+    SQ_TEX_RESOURCE_ps                                    = 0,
+    SQ_TEX_RESOURCE_vs                                    = SQ_TEX_RESOURCE_ps + SQ_TEX_RESOURCE_ps_num,
+    SQ_TEX_RESOURCE_fs                                    = SQ_TEX_RESOURCE_vs + SQ_TEX_RESOURCE_vs_num,
+    SQ_TEX_RESOURCE_gs                                    = SQ_TEX_RESOURCE_fs + SQ_TEX_RESOURCE_fs_num,
+    SQ_VTX_RESOURCE                                       = SQ_VTX_CONSTANT_WORD0_0,	/* 160 PS, 160 VS, 16 FS, 160 GS */
+    SQ_VTX_RESOURCE_ps_num                                = 160,
+    SQ_VTX_RESOURCE_vs_num                                = 160,
+    SQ_VTX_RESOURCE_fs_num                                = 16,
+    SQ_VTX_RESOURCE_gs_num                                = 160,
+    SQ_VTX_RESOURCE_all_num                               = 496,
+    SQ_VTX_RESOURCE_offset                                = 28,
+    SQ_VTX_RESOURCE_ps                                    = 0,
+    SQ_VTX_RESOURCE_vs                                    = SQ_VTX_RESOURCE_ps + SQ_VTX_RESOURCE_ps_num,
+    SQ_VTX_RESOURCE_fs                                    = SQ_VTX_RESOURCE_vs + SQ_VTX_RESOURCE_vs_num,
+    SQ_VTX_RESOURCE_gs                                    = SQ_VTX_RESOURCE_fs + SQ_VTX_RESOURCE_fs_num,
+    SQ_TEX_SAMPLER_WORD                                   = SQ_TEX_SAMPLER_WORD0_0,	/* 18 per PS, VS, GS */
+    SQ_TEX_SAMPLER_WORD_ps_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_vs_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_gs_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_all_num                           = 54,
+    SQ_TEX_SAMPLER_WORD_offset                            = 12,
+    SQ_TEX_SAMPLER_WORD_ps                                = 0,
+    SQ_TEX_SAMPLER_WORD_vs                                = SQ_TEX_SAMPLER_WORD_ps + SQ_TEX_SAMPLER_WORD_ps_num,
+    SQ_TEX_SAMPLER_WORD_gs                                = SQ_TEX_SAMPLER_WORD_vs + SQ_TEX_SAMPLER_WORD_vs_num,
+    SQ_LOOP_CONST                                         = SQ_LOOP_CONST_0,		/* 32 per PS, VS, GS */
+    SQ_LOOP_CONST_ps_num                                  = 32,
+    SQ_LOOP_CONST_vs_num                                  = 32,
+    SQ_LOOP_CONST_gs_num                                  = 32,
+    SQ_LOOP_CONST_all_num                                 = 96,
+    SQ_LOOP_CONST_offset                                  = 4,
+    SQ_LOOP_CONST_ps                                      = 0,
+    SQ_LOOP_CONST_vs                                      = SQ_LOOP_CONST_ps + SQ_LOOP_CONST_ps_num,
+    SQ_LOOP_CONST_gs                                      = SQ_LOOP_CONST_vs + SQ_LOOP_CONST_vs_num,
+} ;
+
+
+#endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg_r7xx.h b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h
new file mode 100644
index 0000000000..e5c01c861a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h
@@ -0,0 +1,149 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_R7xx_H_
+#define _R600_REG_R7xx_H_
+
+/*
+ * Register update for R7xx chips
+ */
+
+enum {
+
+    R7XX_MC_VM_FB_LOCATION                                = 0x00002024,
+
+//  GRBM_STATUS                                           = 0x00008010,
+	R7XX_TA_BUSY_bit                                  = 1 << 14,
+
+    R7xx_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ                     = 0x00008d8c,
+	RING0_OFFSET_mask                                 = 0xff << 0,
+	RING0_OFFSET_shift                                = 0,
+	ISOLATE_ES_ENABLE_bit                             = 1 << 12,
+	ISOLATE_GS_ENABLE_bit                             = 1 << 13,
+	VS_PC_LIMIT_ENABLE_bit                            = 1 << 14,
+
+//  SQ_ALU_WORD0                                          = 0x00008dfc,
+//	SRC0_SEL_mask                                     = 0x1ff << 0,
+// 	SRC1_SEL_mask                                     = 0x1ff << 13,
+	    R7xx_SQ_ALU_SRC_1_DBL_L                       = 0xf4,
+	    R7xx_SQ_ALU_SRC_1_DBL_M                       = 0xf5,
+	    R7xx_SQ_ALU_SRC_0_5_DBL_L                     = 0xf6,
+	    R7xx_SQ_ALU_SRC_0_5_DBL_M                     = 0xf7,
+// 	INDEX_MODE_mask                                   = 0x07 << 26,
+	    R7xx_SQ_INDEX_GLOBAL                          = 0x05,
+	    R7xx_SQ_INDEX_GLOBAL_AR_X                     = 0x06,
+    R6xx_SQ_ALU_WORD1_OP2                                 = 0x00008dfc,
+    R7xx_SQ_ALU_WORD1_OP2_V2                              = 0x00008dfc,
+	R6xx_FOG_MERGE_bit                                = 1 << 5,
+	R6xx_OMOD_mask                                    = 0x03 << 6,
+	R7xx_OMOD_mask                                    = 0x03 << 5,
+	R6xx_OMOD_shift                                   = 6,
+	R7xx_OMOD_shift                                   = 5,
+	R6xx_SQ_ALU_WORD1_OP2__ALU_INST_mask              = 0x3ff << 8,
+	R7xx_SQ_ALU_WORD1_OP2_V2__ALU_INST_mask           = 0x7ff << 7,
+	R6xx_SQ_ALU_WORD1_OP2__ALU_INST_shift             = 8,
+	R7xx_SQ_ALU_WORD1_OP2_V2__ALU_INST_shift          = 7,
+	    R7xx_SQ_OP2_INST_FREXP_64                     = 0x07,
+	    R7xx_SQ_OP2_INST_ADD_64                       = 0x17,
+	    R7xx_SQ_OP2_INST_MUL_64                       = 0x1b,
+	    R7xx_SQ_OP2_INST_FLT64_TO_FLT32               = 0x1c,
+	    R7xx_SQ_OP2_INST_FLT32_TO_FLT64               = 0x1d,
+	    R7xx_SQ_OP2_INST_LDEXP_64                     = 0x7a,
+	    R7xx_SQ_OP2_INST_FRACT_64                     = 0x7b,
+	    R7xx_SQ_OP2_INST_PRED_SETGT_64                = 0x7c,
+	    R7xx_SQ_OP2_INST_PRED_SETE_64                 = 0x7d,
+	    R7xx_SQ_OP2_INST_PRED_SETGE_64                = 0x7e,
+//  SQ_ALU_WORD1_OP3                                      = 0x00008dfc,
+//	SRC2_SEL_mask                                     = 0x1ff << 0,
+//	    R7xx_SQ_ALU_SRC_1_DBL_L                       = 0xf4,
+//	    R7xx_SQ_ALU_SRC_1_DBL_M                       = 0xf5,
+//	    R7xx_SQ_ALU_SRC_0_5_DBL_L                     = 0xf6,
+//	    R7xx_SQ_ALU_SRC_0_5_DBL_M                     = 0xf7,
+// 	SQ_ALU_WORD1_OP3__ALU_INST_mask                   = 0x1f << 13,
+	    R7xx_SQ_OP3_INST_MULADD_64                    = 0x08,
+	    R7xx_SQ_OP3_INST_MULADD_64_M2                 = 0x09,
+	    R7xx_SQ_OP3_INST_MULADD_64_M4                 = 0x0a,
+	    R7xx_SQ_OP3_INST_MULADD_64_D2                 = 0x0b,
+//  SQ_CF_ALU_WORD1                                       = 0x00008dfc,
+	R6xx_USES_WATERFALL_bit                           = 1 << 25,
+	R7xx_SQ_CF_ALU_WORD1__ALT_CONST_bit               = 1 << 25,
+//  SQ_CF_ALLOC_EXPORT_WORD0                              = 0x00008dfc,
+//	ARRAY_BASE_mask                                   = 0x1fff << 0,
+//	TYPE_mask                                         = 0x03 << 13,
+//	    SQ_EXPORT_PARAM                               = 0x02,
+//	    X_UNUSED_FOR_SX_EXPORTS                       = 0x03,
+//	ELEM_SIZE_mask                                    = 0x03 << 30,
+//  SQ_CF_ALLOC_EXPORT_WORD1                              = 0x00008dfc,
+//	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_mask            = 0x7f << 23,
+	    R7xx_SQ_CF_INST_MEM_EXPORT                    = 0x3a,
+//  SQ_CF_WORD1                                           = 0x00008dfc,
+//	SQ_CF_WORD1__COUNT_mask                           = 0x07 << 10,
+	R7xx_COUNT_3_bit                                  = 1 << 19,
+//	SQ_CF_WORD1__CF_INST_mask                         = 0x7f << 23,
+	    R7xx_SQ_CF_INST_END_PROGRAM                   = 0x19,
+	    R7xx_SQ_CF_INST_WAIT_ACK                      = 0x1a,
+	    R7xx_SQ_CF_INST_TEX_ACK                       = 0x1b,
+	    R7xx_SQ_CF_INST_VTX_ACK                       = 0x1c,
+	    R7xx_SQ_CF_INST_VTX_TC_ACK                    = 0x1d,
+//  SQ_VTX_WORD0                                          = 0x00008dfc,
+//	VTX_INST_mask                                     = 0x1f << 0,
+	    R7xx_SQ_VTX_INST_MEM                          = 0x02,
+//  SQ_VTX_WORD2                                          = 0x00008dfc,
+	R7xx_SQ_VTX_WORD2__ALT_CONST_bit                  = 1 << 20,
+
+//  SQ_TEX_WORD0                                          = 0x00008dfc,
+//	TEX_INST_mask                                     = 0x1f << 0,
+	    R7xx_X_MEMORY_READ                            = 0x02,
+	    R7xx_SQ_TEX_INST_KEEP_GRADIENTS               = 0x0a,
+	    R7xx_X_FETCH4_LOAD4_INSTRUCTION_FOR_DX10_1    = 0x0f,
+	R7xx_SQ_TEX_WORD0__ALT_CONST_bit                  = 1 << 24,
+
+    R7xx_PA_SC_EDGERULE                                   = 0x00028230,
+    R7xx_SPI_THREAD_GROUPING                              = 0x000286c8,
+	PS_GROUPING_mask                                  = 0x1f << 0,
+	PS_GROUPING_shift                                 = 0,
+	VS_GROUPING_mask                                  = 0x1f << 8,
+	VS_GROUPING_shift                                 = 8,
+	GS_GROUPING_mask                                  = 0x1f << 16,
+	GS_GROUPING_shift                                 = 16,
+	ES_GROUPING_mask                                  = 0x1f << 24,
+	ES_GROUPING_shift                                 = 24,
+    R7xx_CB_SHADER_CONTROL                                = 0x000287a0,
+	RT0_ENABLE_bit                                    = 1 << 0,
+	RT1_ENABLE_bit                                    = 1 << 1,
+	RT2_ENABLE_bit                                    = 1 << 2,
+	RT3_ENABLE_bit                                    = 1 << 3,
+	RT4_ENABLE_bit                                    = 1 << 4,
+	RT5_ENABLE_bit                                    = 1 << 5,
+	RT6_ENABLE_bit                                    = 1 << 6,
+	RT7_ENABLE_bit                                    = 1 << 7,
+//  DB_ALPHA_TO_MASK                                      = 0x00028d44,
+	R7xx_OFFSET_ROUND_bit                             = 1 << 16,
+//  SQ_TEX_SAMPLER_MISC_0                                 = 0x0003d03c,
+	R7xx_TRUNCATE_COORD_bit                           = 1 << 9,
+	R7xx_DISABLE_CUBE_WRAP_bit                        = 1 << 10,
+
+} ;
+
+#endif /* _R600_REG_R7xx_H_ */
diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c
new file mode 100644
index 0000000000..d105b90cd1
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_tex.c
@@ -0,0 +1,440 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mipmap.h"
+#include "main/simple_list.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+
+#include "texmem.h"
+
+#include "r600_context.h"
+#include "r700_state.h"
+#include "radeon_mipmap_tree.h"
+#include "r600_tex.h"
+
+#include "xmlpool.h"
+
+
+static unsigned int translate_wrap_mode(GLenum wrapmode)
+{
+	switch(wrapmode) {
+	case GL_REPEAT: return SQ_TEX_WRAP;
+	case GL_CLAMP: return SQ_TEX_CLAMP_HALF_BORDER;
+	case GL_CLAMP_TO_EDGE: return SQ_TEX_CLAMP_LAST_TEXEL;
+	case GL_CLAMP_TO_BORDER: return SQ_TEX_CLAMP_BORDER;
+	case GL_MIRRORED_REPEAT: return SQ_TEX_MIRROR;
+	case GL_MIRROR_CLAMP_EXT: return SQ_TEX_MIRROR_ONCE_HALF_BORDER;
+	case GL_MIRROR_CLAMP_TO_EDGE_EXT: return SQ_TEX_MIRROR_ONCE_LAST_TEXEL;
+	case GL_MIRROR_CLAMP_TO_BORDER_EXT: return SQ_TEX_MIRROR_ONCE_BORDER;
+	default:
+		radeon_error("bad wrap mode in %s", __FUNCTION__);
+		return 0;
+	}
+}
+
+
+/**
+ * Update the cached hardware registers based on the current texture wrap modes.
+ *
+ * \param t Texture object whose wrap modes are to be set
+ */
+static void r600UpdateTexWrap(radeonTexObjPtr t)
+{
+	struct gl_texture_object *tObj = &t->base;
+
+        SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapS),
+                 SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift, SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_mask);
+
+	if (tObj->Target != GL_TEXTURE_1D) {
+		SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapT),
+			 CLAMP_Y_shift, CLAMP_Y_mask);
+
+		if (tObj->Target == GL_TEXTURE_3D)
+			SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapR),
+				 CLAMP_Z_shift, CLAMP_Z_mask);
+	}
+}
+
+static void r600SetTexDefaultState(radeonTexObjPtr t)
+{
+        /* Init text object to default states. */
+        t->SQ_TEX_RESOURCE0              = 0;
+        SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_2D, DIM_shift, DIM_mask);
+        SETfield(t->SQ_TEX_RESOURCE0, ARRAY_LINEAR_GENERAL,
+                 SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift, SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask);
+        CLEARbit(t->SQ_TEX_RESOURCE0, TILE_TYPE_bit);
+
+        t->SQ_TEX_RESOURCE1                = 0;
+        SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+                 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+        t->SQ_TEX_RESOURCE2                = 0;
+        t->SQ_TEX_RESOURCE3                = 0;
+
+        t->SQ_TEX_RESOURCE4                   = 0;
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_X_shift, FORMAT_COMP_X_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_Y_shift, FORMAT_COMP_Y_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_Z_shift, FORMAT_COMP_Z_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_W_shift, FORMAT_COMP_W_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_NUM_FORMAT_NORM,
+                 SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_shift, SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_mask);
+        CLEARbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__SRF_MODE_ALL_bit);
+        CLEARbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_ENDIAN_NONE,
+                 SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_shift, SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, 1, REQUEST_SIZE_shift, REQUEST_SIZE_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+	SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+	SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+	SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift,
+		 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, 0, BASE_LEVEL_shift, BASE_LEVEL_mask); /* mip-maps */
+
+        t->SQ_TEX_RESOURCE5 = 0;
+        t->SQ_TEX_RESOURCE6 = 0;
+
+        SETfield(t->SQ_TEX_RESOURCE6, SQ_TEX_VTX_VALID_TEXTURE,
+                 SQ_TEX_RESOURCE_WORD6_0__TYPE_shift, SQ_TEX_RESOURCE_WORD6_0__TYPE_mask);
+
+        /* Initialize sampler registers */
+        t->SQ_TEX_SAMPLER0                           = 0;
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_WRAP, SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift,
+		 SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_WRAP, CLAMP_Y_shift, CLAMP_Y_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_WRAP, CLAMP_Z_shift, CLAMP_Z_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_XY_FILTER_POINT, XY_MAG_FILTER_shift, XY_MAG_FILTER_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_XY_FILTER_POINT, XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_Z_FILTER_NONE, Z_FILTER_shift, Z_FILTER_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_Z_FILTER_NONE, MIP_FILTER_shift, MIP_FILTER_mask);
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_BORDER_COLOR_TRANS_BLACK, BORDER_COLOR_TYPE_shift, BORDER_COLOR_TYPE_mask);
+
+        t->SQ_TEX_SAMPLER1                           = 0;
+        SETfield(t->SQ_TEX_SAMPLER1, 0x3ff, MAX_LOD_shift, MAX_LOD_mask);
+
+        t->SQ_TEX_SAMPLER2                          = 0;
+        SETbit(t->SQ_TEX_SAMPLER2, SQ_TEX_SAMPLER_WORD2_0__TYPE_bit);
+}
+
+
+#if 0
+static GLuint aniso_filter(GLfloat anisotropy)
+{
+	if (anisotropy >= 16.0) {
+		return R300_TX_MAX_ANISO_16_TO_1;
+	} else if (anisotropy >= 8.0) {
+		return R300_TX_MAX_ANISO_8_TO_1;
+	} else if (anisotropy >= 4.0) {
+		return R300_TX_MAX_ANISO_4_TO_1;
+	} else if (anisotropy >= 2.0) {
+		return R300_TX_MAX_ANISO_2_TO_1;
+	} else {
+		return R300_TX_MAX_ANISO_1_TO_1;
+	}
+	return 0;
+}
+#endif
+
+/**
+ * Set the texture magnification and minification modes.
+ *
+ * \param t Texture whose filter modes are to be set
+ * \param minf Texture minification mode
+ * \param magf Texture magnification mode
+ * \param anisotropy Maximum anisotropy level
+ */
+static void r600SetTexFilter(radeonTexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
+{
+	/* Force revalidation to account for switches from/to mipmapping. */
+	t->validated = GL_FALSE;
+
+	/* Note that EXT_texture_filter_anisotropic is extremely vague about
+	 * how anisotropic filtering interacts with the "normal" filter modes.
+	 * When anisotropic filtering is enabled, we override min and mag
+	 * filter settings completely. This includes driconf's settings.
+	 */
+	if (anisotropy >= 2.0 && (minf != GL_NEAREST) && (magf != GL_NEAREST)) {
+		/*t->pp_txfilter |= R300_TX_MAG_FILTER_ANISO
+			| R300_TX_MIN_FILTER_ANISO
+			| R300_TX_MIN_FILTER_MIP_LINEAR
+			| aniso_filter(anisotropy);*/
+		radeon_print(RADEON_TEXTURE, RADEON_NORMAL, "Using maximum anisotropy of %f\n", anisotropy);
+		return;
+	}
+
+	switch (minf) {
+	case GL_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_None,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_None,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_NEAREST_MIPMAP_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Point,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_NEAREST_MIPMAP_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Linear,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR_MIPMAP_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Point,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR_MIPMAP_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Linear,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	}
+
+	/* Note we don't have 3D mipmaps so only use the mag filter setting
+	 * to set the 3D texture filter mode.
+	 */
+	switch (magf) {
+	case GL_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MAG_FILTER_shift, XY_MAG_FILTER_mask);
+		break;
+	case GL_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MAG_FILTER_shift, XY_MAG_FILTER_mask);
+		break;
+	}
+}
+
+static void r600SetTexBorderColor(radeonTexObjPtr t, const GLfloat color[4])
+{
+	t->TD_PS_SAMPLER0_BORDER_ALPHA = *((uint32_t*)&(color[3]));
+	t->TD_PS_SAMPLER0_BORDER_RED = *((uint32_t*)&(color[2]));
+	t->TD_PS_SAMPLER0_BORDER_GREEN = *((uint32_t*)&(color[1]));
+	t->TD_PS_SAMPLER0_BORDER_BLUE = *((uint32_t*)&(color[0]));
+        SETfield(t->SQ_TEX_SAMPLER0, SQ_TEX_BORDER_COLOR_REGISTER,
+		 BORDER_COLOR_TYPE_shift, BORDER_COLOR_TYPE_mask);
+}
+
+/**
+ * Changes variables and flags for a state update, which will happen at the
+ * next UpdateTextureState
+ */
+
+static void r600TexParameter(GLcontext * ctx, GLenum target,
+			     struct gl_texture_object *texObj,
+			     GLenum pname, const GLfloat * params)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+
+	radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_VERBOSE,
+			"%s( %s )\n", __FUNCTION__,
+			_mesa_lookup_enum_by_nr(pname));
+
+	switch (pname) {
+	case GL_TEXTURE_MIN_FILTER:
+	case GL_TEXTURE_MAG_FILTER:
+	case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+		r600SetTexFilter(t, texObj->MinFilter, texObj->MagFilter, texObj->MaxAnisotropy);
+		break;
+
+	case GL_TEXTURE_WRAP_S:
+	case GL_TEXTURE_WRAP_T:
+	case GL_TEXTURE_WRAP_R:
+		r600UpdateTexWrap(t);
+		break;
+
+	case GL_TEXTURE_BORDER_COLOR:
+		r600SetTexBorderColor(t, texObj->BorderColor);
+		break;
+
+	case GL_TEXTURE_BASE_LEVEL:
+	case GL_TEXTURE_MAX_LEVEL:
+	case GL_TEXTURE_MIN_LOD:
+	case GL_TEXTURE_MAX_LOD:
+		/* This isn't the most efficient solution but there doesn't appear to
+		 * be a nice alternative.  Since there's no LOD clamping,
+		 * we just have to rely on loading the right subset of mipmap levels
+		 * to simulate a clamped LOD.
+		 */
+		if (t->mt) {
+			radeon_miptree_unreference(t->mt);
+			t->mt = 0;
+			t->validated = GL_FALSE;
+		}
+		break;
+
+	case GL_DEPTH_TEXTURE_MODE:
+		if (!texObj->Image[0][texObj->BaseLevel])
+			return;
+		if (texObj->Image[0][texObj->BaseLevel]->TexFormat->BaseFormat
+		    == GL_DEPTH_COMPONENT) {
+			r600SetDepthTexMode(texObj);
+			break;
+		} else {
+			/* If the texture isn't a depth texture, changing this
+			 * state won't cause any changes to the hardware.
+			 * Don't force a flush of texture state.
+			 */
+			return;
+		}
+
+	default:
+		return;
+	}
+}
+
+static void r600DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	context_t* rmesa = R700_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+
+	radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_NORMAL,
+		"%s( %p (target = %s) )\n", __FUNCTION__,
+			(void *)texObj,
+			_mesa_lookup_enum_by_nr(texObj->Target));
+
+	if (rmesa) {
+		int i;
+		radeon_firevertices(&rmesa->radeon);
+
+		for(i = 0; i < R700_MAX_TEXTURE_UNITS; ++i)
+			if (rmesa->hw.textures[i] == t)
+				rmesa->hw.textures[i] = 0;
+	}
+
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
+	}
+	_mesa_delete_texture_object(ctx, texObj);
+}
+
+/**
+ * Allocate a new texture object.
+ * Called via ctx->Driver.NewTextureObject.
+ * Note: this function will be called during context creation to
+ * allocate the default texture objects.
+ * Fixup MaxAnisotropy according to user preference.
+ */
+static struct gl_texture_object *r600NewTextureObject(GLcontext * ctx,
+						      GLuint name,
+						      GLenum target)
+{
+	context_t* rmesa = R700_CONTEXT(ctx);
+	radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+
+	radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_NORMAL,
+		"%s( %p (target = %s) )\n", __FUNCTION__,
+			t, _mesa_lookup_enum_by_nr(target));
+
+	_mesa_initialize_texture_object(&t->base, name, target);
+	t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+	/* Initialize hardware state */
+	r600SetTexDefaultState(t);
+	r600UpdateTexWrap(t);
+	r600SetTexFilter(t, t->base.MinFilter, t->base.MagFilter, t->base.MaxAnisotropy);
+	r600SetTexBorderColor(t, t->base.BorderColor);
+
+	return &t->base;
+}
+
+void r600InitTextureFuncs(struct dd_function_table *functions)
+{
+	/* Note: we only plug in the functions we implement in the driver
+	 * since _mesa_init_driver_functions() was already called.
+	 */
+	functions->NewTextureImage = radeonNewTextureImage;
+	functions->FreeTexImageData = radeonFreeTexImageData;
+	functions->MapTexture = radeonMapTexture;
+	functions->UnmapTexture = radeonUnmapTexture;
+
+	functions->ChooseTextureFormat = radeonChooseTextureFormat_mesa;
+	functions->TexImage1D = radeonTexImage1D;
+	functions->TexImage2D = radeonTexImage2D;
+	functions->TexImage3D = radeonTexImage3D;
+	functions->TexSubImage1D = radeonTexSubImage1D;
+	functions->TexSubImage2D = radeonTexSubImage2D;
+	functions->TexSubImage3D = radeonTexSubImage3D;
+	functions->GetTexImage = radeonGetTexImage;
+	functions->GetCompressedTexImage = radeonGetCompressedTexImage;
+	functions->NewTextureObject = r600NewTextureObject;
+	functions->DeleteTexture = r600DeleteTexture;
+	functions->IsTextureResident = driIsTextureResident;
+
+	functions->TexParameter = r600TexParameter;
+
+	functions->CompressedTexImage2D = radeonCompressedTexImage2D;
+	functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D;
+
+	functions->GenerateMipmap = radeonGenerateMipmap;
+
+	driInitTextureFormats();
+}
diff --git a/src/mesa/drivers/dri/r200/r200_span.h b/src/mesa/drivers/dri/r600/r600_tex.h
index bae5644309..fb0e1a023e 100644
--- a/src/mesa/drivers/dri/r200/r200_span.h
+++ b/src/mesa/drivers/dri/r600/r600_tex.h
@@ -32,14 +32,32 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Keith Whitwell <keith@tungstengraphics.com>
  */
 
-#ifndef __R200_SPAN_H__
-#define __R200_SPAN_H__
+#ifndef __r600_TEX_H__
+#define __r600_TEX_H__
 
-#include "drirenderbuffer.h"
+/* TODO : review this after texture load code. */
+#define R700_BLIT_WIDTH_BYTES 1024
+/* The BASE_ADDRESS and MIP_ADDRESS fields are 256-byte-aligned */
+#define R700_TEXTURE_ALIGNMENT_MASK     0x255
+/* Texel pitch is 8 alignment. */
+#define R700_TEXEL_PITCH_ALIGNMENT_MASK 0x7
 
-extern void r200InitSpanFuncs( GLcontext *ctx );
+#define R700_MAX_TEXTURE_UNITS 8 /* TODO : should be 16, lets make it work, review later */
 
-extern void
-radeonSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis);
+extern void r600SetDepthTexMode(struct gl_texture_object *tObj);
 
-#endif
+extern void r600SetTexBuffer(__DRIcontext *pDRICtx, GLint target,
+			     __DRIdrawable *dPriv);
+
+extern void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
+			      GLint format, __DRIdrawable *dPriv);
+
+extern void r600SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
+			     unsigned long long offset, GLint depth,
+			     GLuint pitch);
+
+extern GLboolean r600ValidateBuffers(GLcontext * ctx);
+
+extern void r600InitTextureFuncs(struct dd_function_table *functions);
+
+#endif				/* __r600_TEX_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c
new file mode 100644
index 0000000000..49b603b65e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_texstate.c
@@ -0,0 +1,955 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ *
+ * \todo Enable R300 texture tiling code?
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/enums.h"
+#include "main/simple_list.h"
+
+#include "r600_context.h"
+#include "r700_state.h"
+#include "radeon_mipmap_tree.h"
+#include "r600_tex.h"
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+
+void r600UpdateTextureState(GLcontext * ctx);
+
+void r600UpdateTextureState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	struct gl_texture_unit *texUnit;
+	struct radeon_tex_obj *t;
+	GLuint    unit;
+
+	R600_STATECHANGE(context, tx);
+	R600_STATECHANGE(context, tx_smplr);
+	R600_STATECHANGE(context, tx_brdr_clr);
+
+	for (unit = 0; unit < R700_MAX_TEXTURE_UNITS; unit++) {
+		texUnit = &ctx->Texture.Unit[unit];
+		t = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+
+		if (texUnit->_ReallyEnabled) {
+			if (!t)
+				continue;
+			r700->textures[unit] = t;
+		}
+	}
+}
+
+static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_format)
+{
+	radeonTexObj *t = radeon_tex_obj(tObj);
+
+	CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+	CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+	CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+	CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+
+	switch (mesa_format) /* This is mesa format. */
+	{
+	case MESA_FORMAT_RGBA8888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGBA8888_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB8888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB8888_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB565:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB565_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB4444:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_4_4_4_4,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB4444_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_4_4_4_4,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB1555:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_1_5_5_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ARGB1555_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_1_5_5_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_AL88:
+	case MESA_FORMAT_AL88_REV: /* TODO : Check this. */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB332:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_3_3_2,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_A8: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_L8: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_I8: /* X, X, X, X */
+	case MESA_FORMAT_CI8:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+		/* YUV422 TODO conversion */  /* X, Y, Z, ONE, G8R8_G8B8 */
+		/*
+		  case MESA_FORMAT_YCBCR:
+		  t->SQ_TEX_RESOURCE1.bitfields.DATA_FORMAT = ;
+		  break;
+		*/
+		/* VUY422 TODO conversion */  /* X, Y, Z, ONE, G8R8_G8B8 */
+		/*
+		  case MESA_FORMAT_YCBCR_REV:
+		  t->SQ_TEX_RESOURCE1.bitfields.DATA_FORMAT = ;
+		  break;
+		*/
+	case MESA_FORMAT_RGB_DXT1: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT1: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT3: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT5: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_FLOAT32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGBA_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB_FLOAT32: /* X, Y, Z, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_RGB_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ALPHA_FLOAT32: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_ALPHA_FLOAT16: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_LUMINANCE_FLOAT32: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_LUMINANCE_FLOAT16: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_INTENSITY_FLOAT32: /* X, X, X, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_INTENSITY_FLOAT16: /* X, X, X, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		break;
+	case MESA_FORMAT_Z16:
+	case MESA_FORMAT_Z24_S8:
+	case MESA_FORMAT_Z32:
+		switch (mesa_format) {
+		case MESA_FORMAT_Z16:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_16,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		case MESA_FORMAT_Z24_S8:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_24_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		case MESA_FORMAT_Z32:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_32,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		};
+		switch (tObj->DepthMode) {
+		case GL_LUMINANCE:  /* X, X, X, ONE */
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+			break;
+		case GL_INTENSITY:  /* X, X, X, X */
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+			break;
+		case GL_ALPHA:     /* ZERO, ZERO, ZERO, X */
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_0,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+			break;
+		default:
+			return GL_FALSE;
+		}
+		break;
+	default:
+		/* Not supported format */
+		return GL_FALSE;
+	};
+
+	return GL_TRUE;
+}
+
+void r600SetDepthTexMode(struct gl_texture_object *tObj)
+{
+	radeonTexObjPtr t;
+
+	if (!tObj)
+		return;
+
+	t = radeon_tex_obj(tObj);
+
+	r600GetTexFormat(tObj, tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat);
+
+}
+
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *texObj)
+{
+	radeonTexObj *t = radeon_tex_obj(texObj);
+	const struct gl_texture_image *firstImage;
+	int firstlevel = t->mt ? t->mt->firstLevel : 0;
+	GLuint uTexelPitch, row_align;
+
+	firstImage = t->base.Image[0][firstlevel];
+
+	if (!t->image_override) {
+		if (!r600GetTexFormat(texObj, firstImage->TexFormat->MesaFormat)) {
+			radeon_error("unexpected texture format in %s\n",
+				      __FUNCTION__);
+			return;
+		}
+	}
+
+	switch (texObj->Target) {
+        case GL_TEXTURE_1D:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_1D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_2D:
+        case GL_TEXTURE_RECTANGLE_NV:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_2D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_3D:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_3D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, firstImage->Depth - 1, // ???
+			 TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_CUBE_MAP:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_CUBEMAP, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        default:
+		radeon_error("unexpected texture target type in %s\n", __FUNCTION__);
+		return;
+	}
+
+	row_align = rmesa->radeon.texture_row_align - 1;
+	uTexelPitch = ((firstImage->Width * t->mt->bpp + row_align) & ~row_align) / t->mt->bpp;
+	uTexelPitch = (uTexelPitch + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+
+	/* min pitch is 8 */
+	if (uTexelPitch < 8)
+		uTexelPitch = 8;
+
+	SETfield(t->SQ_TEX_RESOURCE0, (uTexelPitch/8)-1, PITCH_shift, PITCH_mask);
+	SETfield(t->SQ_TEX_RESOURCE0, firstImage->Width - 1,
+		 TEX_WIDTH_shift, TEX_WIDTH_mask);
+	SETfield(t->SQ_TEX_RESOURCE1, firstImage->Height - 1,
+		 TEX_HEIGHT_shift, TEX_HEIGHT_mask);
+
+	if ((t->mt->lastLevel - t->mt->firstLevel) > 0) {
+		t->SQ_TEX_RESOURCE3 = t->mt->levels[0].size / 256;
+		SETfield(t->SQ_TEX_RESOURCE4, t->mt->firstLevel, BASE_LEVEL_shift, BASE_LEVEL_mask);
+		SETfield(t->SQ_TEX_RESOURCE5, t->mt->lastLevel, LAST_LEVEL_shift, LAST_LEVEL_mask);
+	}
+}
+
+/**
+ * Ensure the given texture is ready for rendering.
+ *
+ * Mostly this means populating the texture object's mipmap tree.
+ */
+static GLboolean r600_validate_texture(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	context_t *rmesa = R700_CONTEXT(ctx);
+	radeonTexObj *t = radeon_tex_obj(texObj);
+
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+		return GL_FALSE;
+
+	/* Configure the hardware registers (more precisely, the cached version
+	 * of the hardware registers). */
+	setup_hardware_state(rmesa, texObj);
+
+	t->validated = GL_TRUE;
+	return GL_TRUE;
+}
+
+/**
+ * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+ */
+GLboolean r600ValidateBuffers(GLcontext * ctx)
+{
+	context_t *rmesa = R700_CONTEXT(ctx);
+	struct radeon_renderbuffer *rrb;
+	struct radeon_bo *pbo;
+	int i;
+	int ret;
+
+	radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+	rrb = radeon_get_colorbuffer(&rmesa->radeon);
+	/* color buffer */
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	/* depth buffer */
+	rrb = radeon_get_depthbuffer(&rmesa->radeon);
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
+	}
+	
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+		radeonTexObj *t;
+
+		if (!ctx->Texture.Unit[i]._ReallyEnabled)
+			continue;
+
+		if (!r600_validate_texture(ctx, ctx->Texture.Unit[i]._Current)) {
+			radeon_warning("failed to validate texture for unit %d.\n", i);
+		}
+		t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+		if (t->image_override && t->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+		else if (t->mt->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->mt->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+	}
+
+	pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(ctx);
+	if (pbo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, pbo,
+						  RADEON_GEM_DOMAIN_GTT, 0);
+	}
+
+	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(ctx);
+	if (pbo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, pbo,
+						  RADEON_GEM_DOMAIN_GTT, 0);
+	}
+
+	ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0);
+	if (ret)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
+		      unsigned long long offset, GLint depth, GLuint pitch)
+{
+	context_t *rmesa = pDRICtx->driverPrivate;
+	struct gl_texture_object *tObj =
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
+	uint32_t pitch_val, size;
+
+	if (!tObj)
+		return;
+
+	t->image_override = GL_TRUE;
+
+	if (!offset)
+		return;
+
+	size = pitch;//h * w * (depth / 8);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	t->bo = radeon_legacy_bo_alloc_fake(rmesa->radeon.radeonScreen->bom, size, offset);
+	t->override_offset = offset;
+	pitch_val = pitch;
+	switch (depth) {
+	case 32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		pitch_val /= 4;
+		break;
+	case 24:
+	default:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		pitch_val /= 4;
+		break;
+	case 16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		pitch_val /= 2;
+		break;
+	}
+
+	pitch_val = (pitch_val + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+
+	/* min pitch is 8 */
+	if (pitch_val < 8)
+		pitch_val = 8;
+
+	SETfield(t->SQ_TEX_RESOURCE0, (pitch_val/8)-1, PITCH_shift, PITCH_mask);
+}
+
+void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format, __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	context_t *rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) {
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		} else {
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+				 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		}
+		pitch_val /= 4;
+		break;
+	case 3:
+	default:
+		// FMT_8_8_8 ???
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		pitch_val /= 4;
+		break;
+	case 2:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+		pitch_val /= 2;
+		break;
+	}
+
+	pitch_val = (pitch_val + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+
+	/* min pitch is 8 */
+	if (pitch_val < 8)
+		pitch_val = 8;
+
+	SETfield(t->SQ_TEX_RESOURCE0, (pitch_val/8)-1, PITCH_shift, PITCH_mask);
+	SETfield(t->SQ_TEX_RESOURCE0, rb->base.Width - 1,
+		 TEX_WIDTH_shift, TEX_WIDTH_mask);
+	SETfield(t->SQ_TEX_RESOURCE1, rb->base.Height - 1,
+		 TEX_HEIGHT_shift, TEX_HEIGHT_mask);
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+void r600SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        r600SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c
new file mode 100644
index 0000000000..2d8480daaf
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_assembler.c
@@ -0,0 +1,4106 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/mtypes.h"
+#include "main/imports.h"
+
+#include "radeon_debug.h"
+#include "r600_context.h"
+
+#include "r700_assembler.h"
+
+BITS addrmode_PVSDST(PVSDST * pPVSDST)
+{
+	return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1);
+}
+
+void setaddrmode_PVSDST(PVSDST * pPVSDST, BITS addrmode) 
+{
+	pPVSDST->addrmode0 = addrmode & 1;
+	pPVSDST->addrmode1 = (addrmode >> 1) & 1;
+}
+
+void nomask_PVSDST(PVSDST * pPVSDST) 
+{
+	pPVSDST->writex = pPVSDST->writey = pPVSDST->writez = pPVSDST->writew = 1;
+}
+
+BITS addrmode_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	return pPVSSRC->addrmode0 | ((BITS)pPVSSRC->addrmode1 << 1);
+}
+
+void setaddrmode_PVSSRC(PVSSRC* pPVSSRC, BITS addrmode) 
+{
+	pPVSSRC->addrmode0 = addrmode & 1;
+	pPVSSRC->addrmode1 = (addrmode >> 1) & 1;
+}
+
+
+void setswizzle_PVSSRC(PVSSRC* pPVSSRC, BITS swz) 
+{
+	pPVSSRC->swizzlex = 
+	pPVSSRC->swizzley = 
+	pPVSSRC->swizzlez = 
+	pPVSSRC->swizzlew = swz;
+}
+
+void noswizzle_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->swizzlex = SQ_SEL_X;
+	pPVSSRC->swizzley = SQ_SEL_Y;
+	pPVSSRC->swizzlez = SQ_SEL_Z;
+	pPVSSRC->swizzlew = SQ_SEL_W;
+}
+
+void
+swizzleagain_PVSSRC(PVSSRC * pPVSSRC, BITS x, BITS y, BITS z, BITS w)
+{
+    switch (x) 
+    {
+        case SQ_SEL_X: x = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: x = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: x = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: x = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (y) 
+    {
+        case SQ_SEL_X: y = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: y = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: y = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: y = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (z) 
+    {
+        case SQ_SEL_X: z = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: z = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: z = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: z = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (w) 
+    {
+        case SQ_SEL_X: w = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: w = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: w = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: w = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    pPVSSRC->swizzlex = x;
+    pPVSSRC->swizzley = y;
+    pPVSSRC->swizzlez = z;
+    pPVSSRC->swizzlew = w;
+}
+
+void neg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = 1;
+	pPVSSRC->negy = 1;
+	pPVSSRC->negz = 1;
+	pPVSSRC->negw = 1;
+}
+
+void noneg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = 0;
+	pPVSSRC->negy = 0;
+	pPVSSRC->negz = 0;
+	pPVSSRC->negw = 0;
+}
+
+// negate argument (for SUB instead of ADD and alike)
+void flipneg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = !pPVSSRC->negx;
+	pPVSSRC->negy = !pPVSSRC->negy;
+	pPVSSRC->negz = !pPVSSRC->negz;
+	pPVSSRC->negw = !pPVSSRC->negw;
+}
+
+void zerocomp_PVSSRC(PVSSRC* pPVSSRC, int c) 
+{
+	switch (c) 
+	{
+		case 0: pPVSSRC->swizzlex = SQ_SEL_0; pPVSSRC->negx = 0; break;
+		case 1: pPVSSRC->swizzley = SQ_SEL_0; pPVSSRC->negy = 0; break;
+		case 2: pPVSSRC->swizzlez = SQ_SEL_0; pPVSSRC->negz = 0; break;
+		case 3: pPVSSRC->swizzlew = SQ_SEL_0; pPVSSRC->negw = 0; break;
+		default:;
+	} 
+}
+
+void onecomp_PVSSRC(PVSSRC* pPVSSRC, int c) 
+{
+	switch (c) 
+	{
+		case 0: pPVSSRC->swizzlex = SQ_SEL_1; pPVSSRC->negx = 0; break;
+		case 1: pPVSSRC->swizzley = SQ_SEL_1; pPVSSRC->negy = 0; break;
+		case 2: pPVSSRC->swizzlez = SQ_SEL_1; pPVSSRC->negz = 0; break;
+		case 3: pPVSSRC->swizzlew = SQ_SEL_1; pPVSSRC->negw = 0; break;
+		default:;
+	} 
+}
+
+BITS is_misc_component_exported(VAP_OUT_VTX_FMT_0* pOutVTXFmt0)  
+{
+	  return (pOutVTXFmt0->point_size            |
+			  pOutVTXFmt0->edge_flag             |
+			  pOutVTXFmt0->rta_index             |
+			  pOutVTXFmt0->kill_flag             |
+			  pOutVTXFmt0->viewport_index);
+}
+
+BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) 
+{
+	  return (pFPOutFmt->depth            | 
+			  pFPOutFmt->stencil_ref      | 
+			  pFPOutFmt->mask             | 
+			  pFPOutFmt->coverage_to_mask);
+}
+
+GLboolean is_reduction_opcode(PVSDWORD* dest)
+{
+    if (dest->dst.op3 == 0) 
+    {
+        if ( (dest->dst.opcode == SQ_OP2_INST_DOT4 || dest->dst.opcode == SQ_OP2_INST_DOT4_IEEE) ) 
+        {
+            return GL_TRUE;
+        }
+    }
+    return GL_FALSE;
+}
+
+GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size)
+{
+    GLuint format = FMT_INVALID;
+	GLuint uiElemSize = 0;
+
+    switch (eType)
+    {
+        case GL_BYTE:
+        case GL_UNSIGNED_BYTE:
+			uiElemSize = 1;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_8; break;
+                case 2:
+                    format = FMT_8_8; break;
+                case 3:
+                    format = FMT_8_8_8; break;
+                case 4:
+                    format = FMT_8_8_8_8; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_UNSIGNED_SHORT:
+        case GL_SHORT:
+			uiElemSize = 2;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_16; break;
+                case 2:
+                    format = FMT_16_16; break;
+                case 3:
+                    format = FMT_16_16_16; break;
+                case 4:
+                    format = FMT_16_16_16_16; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_UNSIGNED_INT:
+        case GL_INT:
+			uiElemSize = 4;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32; break;
+                case 2:
+                    format = FMT_32_32; break;
+                case 3:
+                    format = FMT_32_32_32; break;
+                case 4:
+                    format = FMT_32_32_32_32; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_FLOAT:
+			uiElemSize = 4;
+			switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32_FLOAT; break;
+                case 2:
+                    format = FMT_32_32_FLOAT; break;
+                case 3:
+                    format = FMT_32_32_32_FLOAT; break;
+                case 4:
+                    format = FMT_32_32_32_32_FLOAT; break;
+                default:
+                    break;
+            }
+			break;
+        case GL_DOUBLE:
+			uiElemSize = 8;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32_FLOAT; break;
+                case 2:
+                    format = FMT_32_32_FLOAT; break;
+                case 3:
+                    format = FMT_32_32_32_FLOAT; break;
+                case 4:
+                    format = FMT_32_32_32_32_FLOAT; break;
+                default:
+                    break;
+            }
+            break;
+        default:
+			;
+            //GL_ASSERT_NO_CASE();
+    }
+
+    if(NULL != pClient_size)
+    {
+	    *pClient_size = uiElemSize * nChannels;
+    }
+
+    return(format);
+}
+
+unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm)
+{
+    if(pAsm->D.dst.op3)
+    {
+        return 3;
+    }
+
+    switch (pAsm->D.dst.opcode)
+    {
+    case SQ_OP2_INST_ADD:                          
+    case SQ_OP2_INST_MUL: 
+    case SQ_OP2_INST_MAX:
+    case SQ_OP2_INST_MIN:
+    //case SQ_OP2_INST_MAX_DX10:
+    //case SQ_OP2_INST_MIN_DX10:
+    case SQ_OP2_INST_SETGT:
+    case SQ_OP2_INST_SETGE:
+    case SQ_OP2_INST_PRED_SETE:
+    case SQ_OP2_INST_PRED_SETGT:
+    case SQ_OP2_INST_PRED_SETGE:
+    case SQ_OP2_INST_PRED_SETNE:
+    case SQ_OP2_INST_DOT4:
+    case SQ_OP2_INST_DOT4_IEEE:
+        return 2;  
+
+    case SQ_OP2_INST_MOV: 
+    case SQ_OP2_INST_FRACT:
+    case SQ_OP2_INST_FLOOR:
+    case SQ_OP2_INST_KILLGT:
+    case SQ_OP2_INST_EXP_IEEE:
+    case SQ_OP2_INST_LOG_CLAMPED:
+    case SQ_OP2_INST_LOG_IEEE:
+    case SQ_OP2_INST_RECIP_IEEE:
+    case SQ_OP2_INST_RECIPSQRT_IEEE:
+    case SQ_OP2_INST_FLT_TO_INT:
+    case SQ_OP2_INST_SIN:
+    case SQ_OP2_INST_COS:
+        return 1;
+        
+    default: radeon_error(
+		    "Need instruction operand number for %x.\n", pAsm->D.dst.opcode);
+    };
+
+    return 3;
+}
+
+int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader)
+{
+    GLuint i;
+
+    Init_R700_Shader(pShader);
+    pAsm->pR700Shader = pShader;
+    pAsm->currentShaderType = spt;
+
+	pAsm->cf_last_export_ptr   = NULL;
+
+	pAsm->cf_current_export_clause_ptr = NULL;
+	pAsm->cf_current_alu_clause_ptr    = NULL;
+	pAsm->cf_current_tex_clause_ptr    = NULL;
+	pAsm->cf_current_vtx_clause_ptr    = NULL;
+	pAsm->cf_current_cf_clause_ptr     = NULL;
+
+	// No clause has been created yet
+	pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE;
+
+	pAsm->number_of_colorandz_exports = 0;
+	pAsm->number_of_exports           = 0;
+	pAsm->number_of_export_opcodes    = 0;
+
+
+	pAsm->D.bits = 0;
+	pAsm->S[0].bits = 0;
+	pAsm->S[1].bits = 0;
+	pAsm->S[2].bits = 0;
+
+	pAsm->uLastPosUpdate = 0; 
+	
+	*(BITS *) &pAsm->fp_stOutFmt0 = 0;
+
+	pAsm->uIIns = 0;
+	pAsm->uOIns = 0;
+	pAsm->number_used_registers = 0;
+	pAsm->uUsedConsts = 256; 
+
+
+	// Fragment programs
+	pAsm->uBoolConsts = 0;
+	pAsm->uIntConsts = 0;
+	pAsm->uInsts = 0;
+	pAsm->uConsts = 0;
+
+	pAsm->FCSP = 0;
+	pAsm->fc_stack[0].type = FC_NONE;
+
+	pAsm->branch_depth     = 0;
+	pAsm->max_branch_depth = 0;
+
+	pAsm->aArgSubst[0] =
+	pAsm->aArgSubst[1] =
+	pAsm->aArgSubst[2] =
+	pAsm->aArgSubst[3] = (-1);
+
+	pAsm->uOutputs = 0;
+
+	for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) 
+	{
+		pAsm->color_export_register_number[i] = (-1);
+	}
+
+
+	pAsm->depth_export_register_number = (-1);
+	pAsm->stencil_export_register_number = (-1);
+	pAsm->coverage_to_mask_export_register_number = (-1);
+	pAsm->mask_export_register_number = (-1);
+
+	pAsm->starting_export_register_number = 0;
+	pAsm->starting_vfetch_register_number = 0;
+	pAsm->starting_temp_register_number   = 0;
+	pAsm->uFirstHelpReg = 0;
+
+
+	pAsm->input_position_is_used = GL_FALSE;
+	pAsm->input_normal_is_used   = GL_FALSE;
+
+
+	for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) 
+	{
+		pAsm->input_color_is_used[ i ] = GL_FALSE;
+	}
+
+	for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) 
+	{
+		pAsm->input_texture_unit_is_used[ i ] = GL_FALSE;
+	}
+
+	for (i=0; i<VERT_ATTRIB_MAX; i++) 
+	{
+		pAsm->vfetch_instruction_ptr_array[ i ] = NULL;
+	}
+
+	pAsm->number_of_inputs = 0;
+
+	return 0;
+}
+
+GLboolean IsTex(gl_inst_opcode Opcode)
+{
+    if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) )
+    {
+        return GL_TRUE;
+    }
+    return GL_FALSE;
+}
+
+GLboolean IsAlu(gl_inst_opcode Opcode)
+{
+    //TODO : more for fc and ex for higher spec.
+    if( IsTex(Opcode) )
+    {
+        return GL_FALSE;
+    }
+    return GL_TRUE;
+}
+
+int check_current_clause(r700_AssemblerBase* pAsm,
+					     CF_CLAUSE_TYPE      new_clause_type)
+{
+	if (pAsm->cf_current_clause_type != new_clause_type) 
+	{	//Close last open clause
+		switch (pAsm->cf_current_clause_type) 
+		{
+		case CF_ALU_CLAUSE:
+			if ( pAsm->cf_current_alu_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_alu_clause_ptr = NULL;
+            }
+			break;
+		case CF_VTX_CLAUSE:
+			if ( pAsm->cf_current_vtx_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_vtx_clause_ptr = NULL;
+            }
+			break;
+		case CF_TEX_CLAUSE:
+			if ( pAsm->cf_current_tex_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_tex_clause_ptr = NULL;
+            }
+			break;
+		case CF_EXPORT_CLAUSE:
+			if ( pAsm->cf_current_export_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_export_clause_ptr = NULL;
+            }
+			break;
+		case CF_OTHER_CLAUSE:
+			if ( pAsm->cf_current_cf_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_cf_clause_ptr = NULL;
+            }
+			break;
+		case CF_EMPTY_CLAUSE:
+			break;
+		default:
+            radeon_error(
+                       "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type);
+			return GL_FALSE;
+		}
+
+        pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE;
+
+		// Create new clause
+        switch (new_clause_type) 
+	    {
+        case CF_ALU_CLAUSE:
+            pAsm->cf_current_clause_type = CF_ALU_CLAUSE;
+            break;
+        case CF_VTX_CLAUSE:
+            pAsm->cf_current_clause_type = CF_VTX_CLAUSE;
+            break;
+        case CF_TEX_CLAUSE:        
+            pAsm->cf_current_clause_type = CF_TEX_CLAUSE;
+            break;
+        case CF_EXPORT_CLAUSE:
+            {
+                R700ControlFlowSXClause* pR700ControlFlowSXClause 
+                            = (R700ControlFlowSXClause*) CALLOC_STRUCT(R700ControlFlowSXClause); 
+            
+                // Add new export instruction to control flow program        
+                if (pR700ControlFlowSXClause != 0) 
+                {
+                    pAsm->cf_current_export_clause_ptr = pR700ControlFlowSXClause;
+                    Init_R700ControlFlowSXClause(pR700ControlFlowSXClause);
+                    AddCFInstruction( pAsm->pR700Shader, 
+                                      (R700ControlFlowInstruction *)pR700ControlFlowSXClause );
+                }
+                else 
+                {
+                    radeon_error(
+                               "Error allocating new EXPORT CF instruction in check_current_clause. \n");
+                    return GL_FALSE;
+                }
+                pAsm->cf_current_clause_type = CF_EXPORT_CLAUSE;
+            }
+            break;
+        case CF_EMPTY_CLAUSE:
+            break;
+        case CF_OTHER_CLAUSE:
+            pAsm->cf_current_clause_type = CF_OTHER_CLAUSE;
+            break;
+        default:
+            radeon_error(
+                       "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type);
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean add_vfetch_instruction(r700_AssemblerBase*     pAsm,
+								 R700VertexInstruction*  vertex_instruction_ptr)
+{
+	if( GL_FALSE == check_current_clause(pAsm,  CF_VTX_CLAUSE) )
+	{
+		return GL_FALSE;
+	}
+
+    if( pAsm->cf_current_vtx_clause_ptr == NULL ||
+        ( (pAsm->cf_current_vtx_clause_ptr != NULL) && 
+         (pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count >= GetCFMaxInstructions(pAsm->cf_current_vtx_clause_ptr->m_ShaderInstType)-1) 
+        ) ) 
+    { 
+		// Create new Vfetch control flow instruction for this new clause
+		pAsm->cf_current_vtx_clause_ptr = (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause);
+
+		if (pAsm->cf_current_vtx_clause_ptr != NULL) 
+		{
+			Init_R700ControlFlowGenericClause(pAsm->cf_current_vtx_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_vtx_clause_ptr );
+		}
+		else 
+		{
+            radeon_error("Could not allocate a new VFetch CF instruction.\n");
+			return GL_FALSE;
+		}
+
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.pop_count        = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cf_const         = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cond             = SQ_CF_COND_ACTIVE;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count            = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_VTX;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.barrier          = 0x1;
+
+		LinkVertexInstruction(pAsm->cf_current_vtx_clause_ptr, vertex_instruction_ptr );
+	}
+	else
+	{
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count++;
+	}
+
+	AddVTXInstruction(pAsm->pR700Shader, vertex_instruction_ptr);
+
+	return GL_TRUE;
+}
+
+GLboolean add_tex_instruction(r700_AssemblerBase*     pAsm,
+                              R700TextureInstruction* tex_instruction_ptr)
+{ 
+    if ( GL_FALSE == check_current_clause(pAsm, CF_TEX_CLAUSE) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( pAsm->cf_current_tex_clause_ptr == NULL ||
+         ( (pAsm->cf_current_tex_clause_ptr != NULL) && 
+           (pAsm->cf_current_tex_clause_ptr->m_Word1.f.count >= GetCFMaxInstructions(pAsm->cf_current_tex_clause_ptr->m_ShaderInstType)-1) 
+         ) ) 
+    {
+        // new tex cf instruction for this new clause  
+        pAsm->cf_current_tex_clause_ptr = (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause);
+
+		if (pAsm->cf_current_tex_clause_ptr != NULL) 
+		{
+			Init_R700ControlFlowGenericClause(pAsm->cf_current_tex_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_tex_clause_ptr );
+		}
+		else 
+		{
+            radeon_error("Could not allocate a new TEX CF instruction.\n");
+			return GL_FALSE;
+		}
+        
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.pop_count        = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cf_const         = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cond             = SQ_CF_COND_ACTIVE;
+
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_TEX;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.barrier          = 0x0;   //0x1;
+    }
+    else 
+    {        
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.count++;
+    }
+
+    // If this clause constains any TEX instruction that is dependent on a previous instruction, 
+    // set the barrier bit
+    if( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) )
+    {
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.barrier = 0x1;  
+    }
+
+    if(NULL == pAsm->cf_current_tex_clause_ptr->m_pLinkedTEXInstruction)
+    {
+        pAsm->cf_current_tex_clause_ptr->m_pLinkedTEXInstruction = tex_instruction_ptr;
+        tex_instruction_ptr->m_pLinkedGenericClause = pAsm->cf_current_tex_clause_ptr;
+    }
+
+    AddTEXInstruction(pAsm->pR700Shader, tex_instruction_ptr);
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm,
+								GLuint gl_client_id,
+                                GLuint destination_register,
+								GLuint number_of_elements,
+                                GLenum dataElementType,
+								VTX_FETCH_METHOD* pFetchMethod)
+{
+    GLuint client_size_inbyte;
+	GLuint data_format;
+    GLuint mega_fetch_count;
+	GLuint is_mega_fetch_flag;
+
+	R700VertexGenericFetch*   vfetch_instruction_ptr;
+	R700VertexGenericFetch*   assembled_vfetch_instruction_ptr = pAsm->vfetch_instruction_ptr_array[ gl_client_id ];
+
+	if (assembled_vfetch_instruction_ptr == NULL) 
+	{
+		vfetch_instruction_ptr = (R700VertexGenericFetch*) CALLOC_STRUCT(R700VertexGenericFetch);
+		if (vfetch_instruction_ptr == NULL) 
+		{
+			return GL_FALSE;
+		}
+        Init_R700VertexGenericFetch(vfetch_instruction_ptr);
+    }
+	else 
+	{
+		vfetch_instruction_ptr = assembled_vfetch_instruction_ptr;
+	}
+
+	data_format = GetSurfaceFormat(dataElementType, number_of_elements, &client_size_inbyte);
+
+	if(GL_TRUE == pFetchMethod->bEnableMini) //More conditions here
+	{
+		//TODO : mini fetch
+	}
+	else
+	{
+		mega_fetch_count = MEGA_FETCH_BYTES - 1;
+		is_mega_fetch_flag       = 0x1;
+		pFetchMethod->mega_fetch_remainder = MEGA_FETCH_BYTES - client_size_inbyte;
+	}
+
+	vfetch_instruction_ptr->m_Word0.f.vtx_inst         = SQ_VTX_INST_FETCH;
+	vfetch_instruction_ptr->m_Word0.f.fetch_type       = SQ_VTX_FETCH_VERTEX_DATA;
+	vfetch_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0;
+
+	vfetch_instruction_ptr->m_Word0.f.buffer_id        = gl_client_id;
+	vfetch_instruction_ptr->m_Word0.f.src_gpr          = 0x0; 
+	vfetch_instruction_ptr->m_Word0.f.src_rel          = SQ_ABSOLUTE;
+	vfetch_instruction_ptr->m_Word0.f.src_sel_x        = SQ_SEL_X;
+	vfetch_instruction_ptr->m_Word0.f.mega_fetch_count = mega_fetch_count;
+
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_x        = (number_of_elements < 1) ? SQ_SEL_0 : SQ_SEL_X;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_y        = (number_of_elements < 2) ? SQ_SEL_0 : SQ_SEL_Y;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_z        = (number_of_elements < 3) ? SQ_SEL_0 : SQ_SEL_Z;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_w        = (number_of_elements < 4) ? SQ_SEL_1 : SQ_SEL_W;
+
+	vfetch_instruction_ptr->m_Word1.f.use_const_fields = 1;
+
+	// Destination register
+	vfetch_instruction_ptr->m_Word1_GPR.f.dst_gpr = destination_register; 
+	vfetch_instruction_ptr->m_Word1_GPR.f.dst_rel = SQ_ABSOLUTE;
+
+	vfetch_instruction_ptr->m_Word2.f.offset              = 0;
+	vfetch_instruction_ptr->m_Word2.f.const_buf_no_stride = 0x0;
+
+	vfetch_instruction_ptr->m_Word2.f.mega_fetch          = is_mega_fetch_flag;
+
+	if (assembled_vfetch_instruction_ptr == NULL) 
+	{
+		if ( GL_FALSE == add_vfetch_instruction(pAsm, (R700VertexInstruction *)vfetch_instruction_ptr) ) 
+        {   
+			return GL_FALSE;
+		}
+
+		if (pAsm->vfetch_instruction_ptr_array[ gl_client_id ] != NULL) 
+		{
+			return GL_FALSE;
+		}
+		else 
+		{
+			pAsm->vfetch_instruction_ptr_array[ gl_client_id ] = vfetch_instruction_ptr;
+		}
+	}
+
+	return GL_TRUE;
+}
+
+GLuint gethelpr(r700_AssemblerBase* pAsm) 
+{
+    GLuint r = pAsm->uHelpReg;
+    pAsm->uHelpReg++;
+    if (pAsm->uHelpReg > pAsm->number_used_registers)
+    {
+        pAsm->number_used_registers = pAsm->uHelpReg;
+	}
+    return r;
+}
+void resethelpr(r700_AssemblerBase* pAsm) 
+{
+    pAsm->uHelpReg = pAsm->uFirstHelpReg;
+}
+
+void checkop_init(r700_AssemblerBase* pAsm)
+{
+    resethelpr(pAsm);
+    pAsm->aArgSubst[0] =
+    pAsm->aArgSubst[1] =
+    pAsm->aArgSubst[2] =
+    pAsm->aArgSubst[3] = -1;
+}
+
+GLboolean mov_temp(r700_AssemblerBase* pAsm, int src)
+{
+    GLuint tmp = gethelpr(pAsm);
+
+    //mov src to temp helper gpr.
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+  
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    if( GL_FALSE == assemble_src(pAsm, src, 0) )
+    {
+        return GL_FALSE;
+    }
+
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+    noneg_PVSSRC(&(pAsm->S[0].src));
+   
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->aArgSubst[1 + src] = tmp;
+
+    return GL_TRUE;
+}
+
+GLboolean checkop1(r700_AssemblerBase* pAsm)
+{
+    checkop_init(pAsm);
+    return GL_TRUE;
+}
+
+GLboolean checkop2(r700_AssemblerBase* pAsm)
+{
+    GLboolean bSrcConst[2];
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    checkop_init(pAsm);
+
+    if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[0] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[0] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[1] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[1] = GL_FALSE;
+    }
+
+    if( (bSrcConst[0] == GL_TRUE) && (bSrcConst[1] == GL_TRUE) )
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[1].Index)
+        {
+            if( GL_FALSE == mov_temp(pAsm, 1) )
+            {
+                return GL_FALSE;
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean checkop3(r700_AssemblerBase* pAsm)
+{
+    GLboolean bSrcConst[3];
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    checkop_init(pAsm);
+
+    if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[0] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[0] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[1] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[1] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[2].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[2].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[2].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[2].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[2] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[2] = GL_FALSE;
+    }
+
+    if( (GL_TRUE == bSrcConst[0]) && 
+        (GL_TRUE == bSrcConst[1]) && 
+        (GL_TRUE == bSrcConst[2]) ) 
+    {
+        if( GL_FALSE == mov_temp(pAsm, 1) )
+        {
+            return GL_FALSE;
+        }
+        if( GL_FALSE == mov_temp(pAsm, 2) )
+        {
+            return GL_FALSE;
+        }
+
+        return GL_TRUE;
+    }
+    else if( (GL_TRUE == bSrcConst[0]) && 
+             (GL_TRUE == bSrcConst[1]) ) 
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[1].Index)    
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 1) )
+            {
+                return 1;
+            }
+        }
+
+        return GL_TRUE;
+    }
+    else if ( (GL_TRUE == bSrcConst[0]) && 
+              (GL_TRUE == bSrcConst[2]) )  
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[2].Index)     
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 2) )
+            {
+                return GL_FALSE;
+            }
+        }
+
+        return GL_TRUE;
+    }
+    else if( (GL_TRUE == bSrcConst[1]) && 
+             (GL_TRUE == bSrcConst[2]) ) 
+    {
+        if(pILInst->SrcReg[1].Index != pILInst->SrcReg[2].Index)
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 2) )
+            {
+                return GL_FALSE;
+            }
+        }
+
+        return GL_TRUE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_src(r700_AssemblerBase *pAsm,
+                       int src, 
+                       int fld)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if (fld == -1)
+    {
+        fld = src;
+    }
+
+    if(pAsm->aArgSubst[1+src] >= 0) 
+    {
+        setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+        pAsm->S[fld].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[fld].src.reg   = pAsm->aArgSubst[1+src];
+    }
+    else 
+    {
+        switch (pILInst->SrcReg[src].File)
+        {
+        case PROGRAM_TEMPORARY:
+            setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+            pAsm->S[fld].src.rtype = SRC_REG_TEMPORARY;
+            pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index + pAsm->starting_temp_register_number;
+            break;
+        case PROGRAM_CONSTANT:
+        case PROGRAM_LOCAL_PARAM:
+        case PROGRAM_ENV_PARAM:
+        case PROGRAM_STATE_VAR:
+            if (1 == pILInst->SrcReg[src].RelAddr)
+            {
+                setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0);
+            }
+            else
+            {
+                setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);              
+            }
+
+            pAsm->S[fld].src.rtype = SRC_REG_CONSTANT;
+            pAsm->S[fld].src.reg   = pILInst->SrcReg[src].Index;
+            break;      
+        case PROGRAM_INPUT:
+            setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+            pAsm->S[fld].src.rtype = SRC_REG_INPUT;
+            switch (pAsm->currentShaderType)
+            {
+            case SPT_FP:
+                pAsm->S[fld].src.reg = pAsm->uiFP_AttributeMap[pILInst->SrcReg[src].Index];
+                break;
+            case SPT_VP:
+                pAsm->S[fld].src.reg = pAsm->ucVP_AttributeMap[pILInst->SrcReg[src].Index];
+                break;
+            }
+            break;      
+        default:
+            radeon_error("Invalid source argument type\n");
+            return GL_FALSE;
+        }
+    } 
+
+    pAsm->S[fld].src.swizzlex = pILInst->SrcReg[src].Swizzle & 0x7;
+    pAsm->S[fld].src.swizzley = (pILInst->SrcReg[src].Swizzle >> 3) & 0x7;
+    pAsm->S[fld].src.swizzlez = (pILInst->SrcReg[src].Swizzle >> 6) & 0x7;
+    pAsm->S[fld].src.swizzlew = (pILInst->SrcReg[src].Swizzle >> 9) & 0x7;
+
+    pAsm->S[fld].src.negx = pILInst->SrcReg[src].Negate & 0x1;
+    pAsm->S[fld].src.negy = (pILInst->SrcReg[src].Negate >> 1) & 0x1;
+    pAsm->S[fld].src.negz = (pILInst->SrcReg[src].Negate >> 2) & 0x1;
+    pAsm->S[fld].src.negw = (pILInst->SrcReg[src].Negate >> 3) & 0x1;
+     
+    return GL_TRUE;
+}
+
+GLboolean assemble_dst(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+    switch (pILInst->DstReg.File) 
+    {
+    case PROGRAM_TEMPORARY:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg = pILInst->DstReg.Index + pAsm->starting_temp_register_number;
+        break;
+    case PROGRAM_ADDRESS:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_A0;
+        pAsm->D.dst.reg = 0;
+        break;
+    case PROGRAM_OUTPUT:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_OUT;
+        switch (pAsm->currentShaderType)
+        {
+        case SPT_FP:
+            pAsm->D.dst.reg = pAsm->uiFP_OutputMap[pILInst->DstReg.Index];
+            break;
+        case SPT_VP:
+            pAsm->D.dst.reg = pAsm->ucVP_OutputMap[pILInst->DstReg.Index];
+            break;
+        }
+        break;   
+    default:
+        radeon_error("Invalid destination output argument type\n");
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = pILInst->DstReg.WriteMask & 0x1;
+    pAsm->D.dst.writey = (pILInst->DstReg.WriteMask >> 1) & 0x1;
+    pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1;
+    pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1;
+  
+    return GL_TRUE;
+}
+
+GLboolean tex_dst(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if(PROGRAM_TEMPORARY == pILInst->DstReg.File)
+    {
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = pAsm->pILInst[pAsm->uiCurInst].DstReg.Index + pAsm->starting_temp_register_number;
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    }
+    else if(PROGRAM_OUTPUT == pILInst->DstReg.File)
+    {
+        pAsm->D.dst.rtype = DST_REG_OUT;
+        switch (pAsm->currentShaderType)
+        {
+        case SPT_FP:
+            pAsm->D.dst.reg = pAsm->uiFP_OutputMap[pILInst->DstReg.Index];
+            break;
+        case SPT_VP:
+            pAsm->D.dst.reg = pAsm->ucVP_OutputMap[pILInst->DstReg.Index];
+            break;
+        }
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    }
+    else 
+    {
+        radeon_error("Invalid destination output argument type\n");
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = pILInst->DstReg.WriteMask & 0x1;
+    pAsm->D.dst.writey = (pILInst->DstReg.WriteMask >> 1) & 0x1;
+    pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1;
+    pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1;
+  
+    return GL_TRUE;
+}
+
+GLboolean tex_src(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+ 
+    GLboolean bValidTexCoord = GL_FALSE;
+
+    switch (pILInst->SrcReg[0].File)
+    {
+    case PROGRAM_TEMPORARY:
+        bValidTexCoord = GL_TRUE;
+
+        pAsm->S[0].src.reg   = pILInst->SrcReg[0].Index + pAsm->starting_temp_register_number;
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+
+        break;
+    case PROGRAM_INPUT:
+        switch (pILInst->SrcReg[0].Index)
+        {
+        case FRAG_ATTRIB_COL0:
+        case FRAG_ATTRIB_COL1:
+        case FRAG_ATTRIB_TEX0:
+        case FRAG_ATTRIB_TEX1:
+        case FRAG_ATTRIB_TEX2:
+        case FRAG_ATTRIB_TEX3:
+        case FRAG_ATTRIB_TEX4:
+        case FRAG_ATTRIB_TEX5:
+        case FRAG_ATTRIB_TEX6:
+        case FRAG_ATTRIB_TEX7:
+            bValidTexCoord = GL_TRUE;
+
+            pAsm->S[0].src.reg   = pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index];
+            pAsm->S[0].src.rtype = SRC_REG_INPUT;
+        }
+        break;
+    }
+
+    if(GL_TRUE == bValidTexCoord)
+    { 
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    }
+    else
+    {
+        radeon_error("Invalid source texcoord for TEX instruction\n");
+        return GL_FALSE;
+    }
+
+    pAsm->S[0].src.swizzlex = pILInst->SrcReg[0].Swizzle & 0x7;
+    pAsm->S[0].src.swizzley = (pILInst->SrcReg[0].Swizzle >> 3) & 0x7;
+    pAsm->S[0].src.swizzlez = (pILInst->SrcReg[0].Swizzle >> 6) & 0x7;
+    pAsm->S[0].src.swizzlew = (pILInst->SrcReg[0].Swizzle >> 9) & 0x7;
+
+    pAsm->S[0].src.negx = pILInst->SrcReg[0].Negate & 0x1;
+    pAsm->S[0].src.negy = (pILInst->SrcReg[0].Negate >> 1) & 0x1;
+    pAsm->S[0].src.negz = (pILInst->SrcReg[0].Negate >> 2) & 0x1;
+    pAsm->S[0].src.negw = (pILInst->SrcReg[0].Negate >> 3) & 0x1;
+     
+    return GL_TRUE;
+}
+
+GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalized)
+{
+    PVSSRC *   texture_coordinate_source;
+    PVSSRC *   texture_unit_source;
+    
+    R700TextureInstruction* tex_instruction_ptr = (R700TextureInstruction*) CALLOC_STRUCT(R700TextureInstruction);
+	if (tex_instruction_ptr == NULL) 
+	{
+		return GL_FALSE;
+	}
+    Init_R700TextureInstruction(tex_instruction_ptr);
+
+    texture_coordinate_source = &(pAsm->S[0].src);
+    texture_unit_source       = &(pAsm->S[1].src);
+
+    tex_instruction_ptr->m_Word0.f.tex_inst         = pAsm->D.dst.opcode;
+    tex_instruction_ptr->m_Word0.f.bc_frac_mode     = 0x0;
+    tex_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0;
+
+    tex_instruction_ptr->m_Word0.f.resource_id      = texture_unit_source->reg;
+
+    tex_instruction_ptr->m_Word1.f.lod_bias     = 0x0;
+    if (normalized) {
+	    tex_instruction_ptr->m_Word1.f.coord_type_x = SQ_TEX_NORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_y = SQ_TEX_NORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_z = SQ_TEX_NORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_w = SQ_TEX_NORMALIZED;
+    } else {
+	    /* XXX: UNNORMALIZED tex coords have limited wrap modes */
+	    tex_instruction_ptr->m_Word1.f.coord_type_x = SQ_TEX_UNNORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_y = SQ_TEX_UNNORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_z = SQ_TEX_UNNORMALIZED;
+	    tex_instruction_ptr->m_Word1.f.coord_type_w = SQ_TEX_UNNORMALIZED;
+    }
+
+    tex_instruction_ptr->m_Word2.f.offset_x   = 0x0;
+    tex_instruction_ptr->m_Word2.f.offset_y   = 0x0;
+    tex_instruction_ptr->m_Word2.f.offset_z   = 0x0;
+
+    tex_instruction_ptr->m_Word2.f.sampler_id = texture_unit_source->reg;
+
+    // dst
+    if ( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || 
+         (pAsm->D.dst.rtype == DST_REG_OUT) ) 
+    {
+        tex_instruction_ptr->m_Word0.f.src_gpr    = texture_coordinate_source->reg;
+        tex_instruction_ptr->m_Word0.f.src_rel    = SQ_ABSOLUTE;
+
+        tex_instruction_ptr->m_Word1.f.dst_gpr    = pAsm->D.dst.reg;
+        tex_instruction_ptr->m_Word1.f.dst_rel    = SQ_ABSOLUTE;
+
+        tex_instruction_ptr->m_Word1.f.dst_sel_x  = (pAsm->D.dst.writex ? texture_unit_source->swizzlex : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_y  = (pAsm->D.dst.writey ? texture_unit_source->swizzley : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_z  = (pAsm->D.dst.writez ? texture_unit_source->swizzlez : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_w  = (pAsm->D.dst.writew ? texture_unit_source->swizzlew : SQ_SEL_MASK);
+
+
+        tex_instruction_ptr->m_Word2.f.src_sel_x  = texture_coordinate_source->swizzlex;
+        tex_instruction_ptr->m_Word2.f.src_sel_y  = texture_coordinate_source->swizzley;
+        tex_instruction_ptr->m_Word2.f.src_sel_z  = texture_coordinate_source->swizzlez;
+        tex_instruction_ptr->m_Word2.f.src_sel_w  = texture_coordinate_source->swizzlew;
+    }
+    else 
+    {
+        radeon_error("Only temp destination registers supported for TEX dest regs.\n");
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == add_tex_instruction(pAsm, tex_instruction_ptr) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+void initialize(r700_AssemblerBase *pAsm)
+{
+    GLuint cycle, component;
+
+    for (cycle=0; cycle<NUMBER_OF_CYCLES; cycle++) 
+    {
+        for (component=0; component<NUMBER_OF_COMPONENTS; component++) 
+        {
+            pAsm->hw_gpr[cycle][component] = (-1);
+        }
+    }
+    for (component=0; component<NUMBER_OF_COMPONENTS; component++) 
+    {
+        pAsm->hw_cfile_addr[component] = (-1);
+        pAsm->hw_cfile_chan[component] = (-1);
+    }
+}
+
+GLboolean assemble_alu_src(R700ALUInstruction*  alu_instruction_ptr,
+                           int                  source_index,
+                           PVSSRC*              pSource,
+                           BITS                 scalar_channel_index)
+{
+    BITS src_sel;
+    BITS src_rel;
+    BITS src_chan;
+    BITS src_neg;
+
+    //--------------------------------------------------------------------------
+    // Source for operands src0, src1. 
+    // Values [0,127] correspond to GPR[0..127]. 
+    // Values [256,511] correspond to cfile constants c[0..255]. 
+
+    //--------------------------------------------------------------------------
+    // Other special values are shown in the list below.
+
+    // 248	SQ_ALU_SRC_0: special constant 0.0.
+    // 249	SQ_ALU_SRC_1: special constant 1.0 float.
+
+    // 250	SQ_ALU_SRC_1_INT: special constant 1 integer.
+    // 251	SQ_ALU_SRC_M_1_INT: special constant -1 integer.
+
+    // 252	SQ_ALU_SRC_0_5: special constant 0.5 float.
+    // 253	SQ_ALU_SRC_LITERAL: literal constant.
+
+    // 254	SQ_ALU_SRC_PV: previous vector result.
+    // 255	SQ_ALU_SRC_PS: previous scalar result.
+    //--------------------------------------------------------------------------
+
+    BITS channel_swizzle;
+    switch (scalar_channel_index) 
+    {
+        case 0: channel_swizzle = pSource->swizzlex; break;
+        case 1: channel_swizzle = pSource->swizzley; break;
+        case 2: channel_swizzle = pSource->swizzlez; break;
+        case 3: channel_swizzle = pSource->swizzlew; break;
+        default: channel_swizzle = SQ_SEL_MASK; break;
+    }
+
+    if(channel_swizzle == SQ_SEL_0) 
+    {
+        src_sel = SQ_ALU_SRC_0; 
+    }
+    else if (channel_swizzle == SQ_SEL_1) 
+    {
+        src_sel = SQ_ALU_SRC_1; 
+    }
+    else 
+    {
+        if ( (pSource->rtype == SRC_REG_TEMPORARY) || 
+             (pSource->rtype == SRC_REG_INPUT)
+        ) 
+        {
+            src_sel = pSource->reg;
+        }
+        else if (pSource->rtype == SRC_REG_CONSTANT)
+        {
+            src_sel = pSource->reg + CFILE_REGISTER_OFFSET;            
+        }
+        else
+        {
+            radeon_error("Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.\n",
+                     source_index, pSource->rtype);
+            return GL_FALSE;
+        }
+    }
+
+    if( ADDR_ABSOLUTE == addrmode_PVSSRC(pSource) ) 
+    {
+        src_rel = SQ_ABSOLUTE;
+    }
+    else 
+    {
+        src_rel = SQ_RELATIVE;
+    }
+
+    switch (channel_swizzle) 
+    {
+        case SQ_SEL_X: 
+            src_chan = SQ_CHAN_X; 
+            break;
+        case SQ_SEL_Y: 
+            src_chan = SQ_CHAN_Y; 
+            break;
+        case SQ_SEL_Z: 
+            src_chan = SQ_CHAN_Z; 
+            break;
+        case SQ_SEL_W: 
+            src_chan = SQ_CHAN_W; 
+            break;
+        case SQ_SEL_0:
+        case SQ_SEL_1:
+            // Does not matter since src_sel controls
+            src_chan = SQ_CHAN_X; 
+            break;
+        default:
+            radeon_error("Unknown source select value (%d) in assemble_alu_src().\n", channel_swizzle);
+            return GL_FALSE;
+            break;
+    }
+
+    switch (scalar_channel_index) 
+    {
+        case 0: src_neg = pSource->negx; break;
+        case 1: src_neg = pSource->negy; break;
+        case 2: src_neg = pSource->negz; break;
+        case 3: src_neg = pSource->negw; break;
+        default: src_neg = 0; break;
+    }
+
+    switch (source_index) 
+    {
+        case 0:
+            alu_instruction_ptr->m_Word0.f.src0_sel  = src_sel;
+            alu_instruction_ptr->m_Word0.f.src0_rel  = src_rel;
+            alu_instruction_ptr->m_Word0.f.src0_chan = src_chan;
+            alu_instruction_ptr->m_Word0.f.src0_neg  = src_neg;
+            break;
+        case 1:
+            alu_instruction_ptr->m_Word0.f.src1_sel  = src_sel;
+            alu_instruction_ptr->m_Word0.f.src1_rel  = src_rel;
+            alu_instruction_ptr->m_Word0.f.src1_chan = src_chan;
+            alu_instruction_ptr->m_Word0.f.src1_neg  = src_neg;
+            break;
+        case 2:
+            alu_instruction_ptr->m_Word1_OP3.f.src2_sel  = src_sel;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_rel  = src_rel;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_chan = src_chan;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_neg  = src_neg;
+            break;
+        default:
+            radeon_error("Only three sources allowed in ALU opcodes.\n");
+          return GL_FALSE;
+          break;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean add_alu_instruction(r700_AssemblerBase* pAsm,
+                              R700ALUInstruction* alu_instruction_ptr,
+                              GLuint              contiguous_slots_needed)
+{
+    if( GL_FALSE == check_current_clause(pAsm, CF_ALU_CLAUSE) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( pAsm->cf_current_alu_clause_ptr == NULL ||
+         ( (pAsm->cf_current_alu_clause_ptr != NULL) && 
+           (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-contiguous_slots_needed-1) )
+         ) ) 
+    {
+
+        //new cf inst for this clause
+        pAsm->cf_current_alu_clause_ptr = (R700ControlFlowALUClause*) CALLOC_STRUCT(R700ControlFlowALUClause);
+            
+        // link the new cf to cf segment    
+        if(NULL != pAsm->cf_current_alu_clause_ptr) 
+        {
+            Init_R700ControlFlowALUClause(pAsm->cf_current_alu_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_alu_clause_ptr );            
+        }
+        else 
+        {
+            radeon_error("Could not allocate a new ALU CF instruction.\n");
+            return GL_FALSE;
+        }
+
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_bank0 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_bank1 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_mode0 = SQ_CF_KCACHE_NOP;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_mode1 = SQ_CF_KCACHE_NOP;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0;
+
+        //cf_current_alu_clause_ptr->m_Word1.f.count           = number_of_scalar_operations - 1;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.count           = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst         = SQ_CF_INST_ALU;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.barrier         = 0x1;
+    }
+    else 
+    {
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.count++;
+    }
+
+    // If this clause constains any instruction that is forward dependent on a TEX instruction, 
+    // set the whole_quad_mode for this clause
+    if ( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) ) 
+    {
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x1;   
+    }
+
+    if (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-1) ) 
+    {
+        alu_instruction_ptr->m_Word0.f.last = 1;
+    }
+
+    if(NULL == pAsm->cf_current_alu_clause_ptr->m_pLinkedALUInstruction)
+    {
+        pAsm->cf_current_alu_clause_ptr->m_pLinkedALUInstruction = alu_instruction_ptr;
+        alu_instruction_ptr->m_pLinkedALUClause = pAsm->cf_current_alu_clause_ptr;
+    }
+    
+    AddALUInstruction(pAsm->pR700Shader, alu_instruction_ptr);
+
+    return GL_TRUE;
+}
+
+void get_src_properties(R700ALUInstruction*  alu_instruction_ptr,
+                        int                  source_index,
+                        BITS*                psrc_sel,
+                        BITS*                psrc_rel,
+                        BITS*                psrc_chan,
+                        BITS*                psrc_neg)
+{
+    switch (source_index) 
+    {
+        case 0:
+            *psrc_sel  = alu_instruction_ptr->m_Word0.f.src0_sel ;
+            *psrc_rel  = alu_instruction_ptr->m_Word0.f.src0_rel ;
+            *psrc_chan = alu_instruction_ptr->m_Word0.f.src0_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word0.f.src0_neg ;
+            break;
+
+        case 1:
+            *psrc_sel  = alu_instruction_ptr->m_Word0.f.src1_sel ;
+            *psrc_rel  = alu_instruction_ptr->m_Word0.f.src1_rel ;
+            *psrc_chan = alu_instruction_ptr->m_Word0.f.src1_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word0.f.src1_neg ;
+            break;
+
+        case 2:
+            *psrc_sel  = alu_instruction_ptr->m_Word1_OP3.f.src2_sel;
+            *psrc_rel  = alu_instruction_ptr->m_Word1_OP3.f.src2_rel;
+            *psrc_chan = alu_instruction_ptr->m_Word1_OP3.f.src2_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word1_OP3.f.src2_neg;
+            break;
+    }
+}
+
+int is_cfile(BITS sel) 
+{
+    if (sel > 255 && sel < 512) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+int is_const(BITS sel) 
+{
+    if (is_cfile(sel)) 
+    {
+        return 1;
+    }
+    else if(sel >= SQ_ALU_SRC_0 && sel <= SQ_ALU_SRC_LITERAL) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+int is_gpr(BITS sel) 
+{
+    if (sel >= 0 && sel < 128) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+const GLuint BANK_SWIZZLE_VEC[8] = {SQ_ALU_VEC_210,  //000
+                                    SQ_ALU_VEC_120,  //001
+                                    SQ_ALU_VEC_102,  //010
+
+                                    SQ_ALU_VEC_201,  //011
+                                    SQ_ALU_VEC_012,  //100
+                                    SQ_ALU_VEC_021,  //101
+
+                                    SQ_ALU_VEC_012,  //110
+                                    SQ_ALU_VEC_012}; //111
+
+const GLuint BANK_SWIZZLE_SCL[8] = {SQ_ALU_SCL_210,  //000
+                                    SQ_ALU_SCL_122,  //001 
+                                    SQ_ALU_SCL_122,  //010
+
+                                    SQ_ALU_SCL_221,  //011
+                                    SQ_ALU_SCL_212,  //100
+                                    SQ_ALU_SCL_122,  //101
+
+                                    SQ_ALU_SCL_122,  //110
+                                    SQ_ALU_SCL_122}; //111
+
+GLboolean reserve_cfile(r700_AssemblerBase* pAsm, 
+                        GLuint sel, 
+                        GLuint chan)
+{
+    int res_match = (-1);
+    int res_empty = (-1);
+
+    GLint res;
+
+    for (res=3; res>=0; res--) 
+    {
+        if(pAsm->hw_cfile_addr[ res] < 0)  
+        {
+            res_empty = res;
+        }
+        else if( (pAsm->hw_cfile_addr[res] == (int)sel)
+                 &&
+                 (pAsm->hw_cfile_chan[ res ] == (int) chan) ) 
+        {
+            res_match = res;
+        }
+    }
+
+    if(res_match >= 0) 
+    {
+        // Read for this scalar component already reserved, nothing to do here.
+        ;
+    }
+    else if(res_empty >= 0) 
+    {
+        pAsm->hw_cfile_addr[ res_empty ] = sel;
+        pAsm->hw_cfile_chan[ res_empty ] = chan;
+    }
+    else 
+    {
+        radeon_error("All cfile read ports are used, cannot reference C$sel, channel $chan.\n");
+        return GL_FALSE;
+    }
+    return GL_TRUE;
+}
+
+GLboolean reserve_gpr(r700_AssemblerBase* pAsm, GLuint sel, GLuint chan, GLuint cycle)
+{
+    if(pAsm->hw_gpr[cycle][chan] < 0) 
+    {
+        pAsm->hw_gpr[cycle][chan] = sel;
+    }
+    else if(pAsm->hw_gpr[cycle][chan] != (int)sel) 
+    {
+        radeon_error("Another scalar operation has already used GPR read port for given channel\n");
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean cycle_for_scalar_bank_swizzle(const int swiz, const int sel, GLuint* pCycle)
+{
+    switch (swiz) 
+    {
+        case SQ_ALU_SCL_210:
+            {
+                int table[3] = {2,	1,	0};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_122:
+            {
+                int table[3] = {1,	2,	2};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_212:
+            {	
+                int table[3] = {2,	1,	2};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_221:
+            {
+                int table[3] = {2, 2, 1};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        default:
+            radeon_error("Bad Scalar bank swizzle value\n");
+            break;
+    }
+
+    return GL_FALSE;
+}
+
+GLboolean cycle_for_vector_bank_swizzle(const int swiz, const int sel, GLuint* pCycle)
+{
+    switch (swiz) 
+    {
+        case SQ_ALU_VEC_012:
+            {
+                int table[3] = {0, 1, 2};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_021:
+            {
+                int table[3] = {0, 2,	1};
+                *pCycle = table[sel];
+            }
+            break;        
+        case SQ_ALU_VEC_120:
+            {
+                int table[3] = {1, 2,	0};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_102:
+            {
+                int table[3] = {1, 0,	2};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_201:
+            {
+                int table[3] = {2, 0,	1};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_210:
+            {
+                int table[3] = {2, 1,	0};
+                *pCycle = table[sel];
+            }
+            break;
+        default:
+            radeon_error("Bad Vec bank swizzle value\n");
+            return GL_FALSE;
+            break;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean check_scalar(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr)
+{
+    GLuint cycle;
+    GLuint bank_swizzle;
+    GLuint const_count = 0;
+
+    BITS sel;
+    BITS chan;
+    BITS rel;
+    BITS neg;
+
+    GLuint src;
+
+    BITS src_sel [3] = {0,0,0};
+    BITS src_chan[3] = {0,0,0};
+    BITS src_rel [3] = {0,0,0};
+    BITS src_neg [3] = {0,0,0};
+
+    GLuint swizzle_key;
+
+    GLuint number_of_operands = r700GetNumOperands(pAsm);
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        get_src_properties(alu_instruction_ptr,
+                           src,
+                           &(src_sel[src]), 
+                           &(src_rel[src]), 
+                           &(src_chan[src]), 
+                           &(src_neg[src]) );
+    }
+
+
+    swizzle_key = ( (is_const( src_sel[0] ) ? 4 : 0) + 
+                    (is_const( src_sel[1] ) ? 2 : 0) + 
+                    (is_const( src_sel[2] ) ? 1 : 0) );
+  
+    alu_instruction_ptr->m_Word1.f.bank_swizzle = BANK_SWIZZLE_SCL[ swizzle_key ];
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+        if (is_const( sel )) 
+        {
+            // Any constant, including literal and inline constants
+            const_count++;
+
+            if (is_cfile( sel )) 
+            {
+                reserve_cfile(pAsm, sel, chan);
+            }
+
+        }
+    }
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+        if( is_gpr(sel) ) 
+        {
+            bank_swizzle = alu_instruction_ptr->m_Word1.f.bank_swizzle;
+
+            if( GL_FALSE == cycle_for_scalar_bank_swizzle(bank_swizzle, src, &cycle) )
+            {
+                return GL_FALSE;
+            }
+
+            if(cycle < const_count) 
+            {
+                if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean check_vector(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr)
+{
+    GLuint cycle;
+    GLuint bank_swizzle;
+    GLuint const_count = 0;
+
+    GLuint src;
+
+    BITS sel;
+    BITS chan;
+    BITS rel;
+    BITS neg;
+
+    BITS src_sel [3] = {0,0,0};
+    BITS src_chan[3] = {0,0,0};
+    BITS src_rel [3] = {0,0,0};
+    BITS src_neg [3] = {0,0,0};
+
+    GLuint swizzle_key;
+
+    GLuint number_of_operands = r700GetNumOperands(pAsm);
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        get_src_properties(alu_instruction_ptr,
+                           src,
+                           &(src_sel[src]), 
+                           &(src_rel[src]), 
+                           &(src_chan[src]), 
+                           &(src_neg[src]) );
+    }
+
+
+    swizzle_key = ( (is_const( src_sel[0] ) ? 4 : 0) + 
+                           (is_const( src_sel[1] ) ? 2 : 0) + 
+                           (is_const( src_sel[2] ) ? 1 : 0) 
+                         );
+
+    alu_instruction_ptr->m_Word1.f.bank_swizzle = BANK_SWIZZLE_VEC[swizzle_key];
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+
+        bank_swizzle = alu_instruction_ptr->m_Word1.f.bank_swizzle;
+
+        if( is_gpr(sel) ) 
+        {
+            if( GL_FALSE == cycle_for_vector_bank_swizzle(bank_swizzle, src, &cycle) )
+            {
+                return GL_FALSE;
+            }
+
+            if ( (src  == 1)          && 
+                 (sel  == src_sel[0]) &&
+                 (chan == src_chan[0]) ) 
+            {        
+            }
+            else 
+            {
+                if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+        else if( is_const(sel) ) 
+        {                  
+            const_count++;
+
+            if( is_cfile(sel) ) 
+            {        
+                if( GL_FALSE == reserve_cfile(pAsm, sel, chan) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm)
+{
+    GLuint    number_of_scalar_operations;
+    GLboolean is_single_scalar_operation;
+    GLuint    scalar_channel_index;
+
+    PVSSRC * pcurrent_source;
+    int    current_source_index;
+    GLuint contiguous_slots_needed;
+
+    GLuint    uNumSrc = r700GetNumOperands(pAsm);
+    GLuint    channel_swizzle, j;
+    GLuint    chan_counter[4] = {0, 0, 0, 0};
+    PVSSRC *  pSource[3];
+    GLboolean bSplitInst = GL_FALSE;
+
+    if (1 == pAsm->D.dst.math) 
+    {
+        is_single_scalar_operation = GL_TRUE;
+        number_of_scalar_operations = 1;
+    }
+    else 
+    {
+        is_single_scalar_operation = GL_FALSE;
+        number_of_scalar_operations = 4;
+        
+        /* check read port, only very preliminary algorithm, not count in 
+           src0/1 same comp case and prev slot repeat case; also not count relative
+           addressing. TODO: improve performance. */
+        for(j=0; j<uNumSrc; j++)
+        {
+            pSource[j] = &(pAsm->S[j].src);
+        }
+        for(scalar_channel_index=0; scalar_channel_index<4; scalar_channel_index++) 
+        {
+            for(j=0; j<uNumSrc; j++) 
+            {
+                switch (scalar_channel_index) 
+                {
+                    case 0: channel_swizzle = pSource[j]->swizzlex; break;
+                    case 1: channel_swizzle = pSource[j]->swizzley; break;
+                    case 2: channel_swizzle = pSource[j]->swizzlez; break;
+                    case 3: channel_swizzle = pSource[j]->swizzlew; break;
+                    default: channel_swizzle = SQ_SEL_MASK; break;
+                }
+                if ( ((pSource[j]->rtype == SRC_REG_TEMPORARY) || 
+                     (pSource[j]->rtype == SRC_REG_INPUT))
+                     && (channel_swizzle <= SQ_SEL_W) )
+                {                    
+                    chan_counter[channel_swizzle]++;                        
+                }
+            }
+        }
+        if(   (chan_counter[SQ_SEL_X] > 3)
+           || (chan_counter[SQ_SEL_Y] > 3)
+           || (chan_counter[SQ_SEL_Z] > 3)
+           || (chan_counter[SQ_SEL_W] > 3) ) /* each chan bank has only 3 ports. */
+        {
+            bSplitInst = GL_TRUE;
+        }
+    }
+
+    contiguous_slots_needed = 0;
+
+    if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) 
+    {
+        contiguous_slots_needed = 4;
+    }
+
+    initialize(pAsm);    
+
+    for (scalar_channel_index=0;
+            scalar_channel_index < number_of_scalar_operations; 
+                scalar_channel_index++) 
+    {
+        R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction);
+        if (alu_instruction_ptr == NULL) 
+		{
+			return GL_FALSE;
+		}
+        Init_R700ALUInstruction(alu_instruction_ptr);
+        
+        //src 0
+        current_source_index = 0;
+        pcurrent_source = &(pAsm->S[0].src);
+
+        if (GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                         current_source_index,
+                                         pcurrent_source, 
+                                         scalar_channel_index) )     
+        {
+            return GL_FALSE;
+        }
+   
+        if (pAsm->D.dst.math == 0) 
+        {            
+            // Process source 1            
+            current_source_index = 1;
+            pcurrent_source = &(pAsm->S[current_source_index].src);
+
+            if (GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                             current_source_index,
+                                             pcurrent_source, 
+                                             scalar_channel_index) ) 
+            {
+                return GL_FALSE;
+            }
+        }
+
+        //other bits
+        alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP;
+
+        if(   (is_single_scalar_operation == GL_TRUE) 
+           || (GL_TRUE == bSplitInst) )
+        {
+            alu_instruction_ptr->m_Word0.f.last = 1;
+        }
+        else 
+        {
+            alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ?  1 : 0;
+        }
+
+        alu_instruction_ptr->m_Word0.f.pred_sel                = 0x0;
+        alu_instruction_ptr->m_Word1_OP2.f.update_pred         = 0x0;  
+        alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0;
+
+        // dst
+        if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || 
+            (pAsm->D.dst.rtype == DST_REG_OUT) ) 
+        {
+            alu_instruction_ptr->m_Word1.f.dst_gpr  = pAsm->D.dst.reg;
+        }
+        else 
+        {
+            radeon_error("Only temp destination registers supported for ALU dest regs.\n");
+            return GL_FALSE;
+        }
+
+        alu_instruction_ptr->m_Word1.f.dst_rel  = SQ_ABSOLUTE;  //D.rtype
+
+        if ( is_single_scalar_operation == GL_TRUE ) 
+        {
+            // Override scalar_channel_index since only one scalar value will be written
+            if(pAsm->D.dst.writex) 
+            {
+                scalar_channel_index = 0;
+            }
+            else if(pAsm->D.dst.writey) 
+            {
+                scalar_channel_index = 1;
+            }
+            else if(pAsm->D.dst.writez) 
+            {
+                scalar_channel_index = 2;
+            }
+            else if(pAsm->D.dst.writew) 
+            {
+                scalar_channel_index = 3;
+            }
+        }
+
+        alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index;
+
+        alu_instruction_ptr->m_Word1.f.clamp    = pAsm->pILInst[pAsm->uiCurInst].SaturateMode;
+
+        if (pAsm->D.dst.op3) 
+        {            
+            //op3
+
+            alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode;
+
+            //There's 3rd src for op3
+            current_source_index = 2;
+            pcurrent_source = &(pAsm->S[current_source_index].src);
+
+            if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                              current_source_index,
+                                              pcurrent_source, 
+                                              scalar_channel_index) ) 
+            {
+                return GL_FALSE;
+            }
+        }
+        else 
+        {
+            //op2
+            if (pAsm->bR6xx)
+            {
+                alu_instruction_ptr->m_Word1_OP2.f6.alu_inst           = pAsm->D.dst.opcode;
+
+                alu_instruction_ptr->m_Word1_OP2.f6.src0_abs           = 0x0;
+                alu_instruction_ptr->m_Word1_OP2.f6.src1_abs           = 0x0;
+
+                //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0;
+                //alu_instruction_ptr->m_Word1_OP2.f6.update_pred         = 0x0;
+                switch (scalar_channel_index) 
+                {
+                    case 0: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writex; 
+                        break;
+                    case 1: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writey; 
+                        break;
+                    case 2: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writez; 
+                        break;
+                    case 3: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writew; 
+                        break;
+                    default: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = 1; //SQ_SEL_MASK;
+                        break;
+                }            
+                alu_instruction_ptr->m_Word1_OP2.f6.omod               = SQ_ALU_OMOD_OFF;
+            }
+            else
+            {
+                alu_instruction_ptr->m_Word1_OP2.f.alu_inst           = pAsm->D.dst.opcode;
+
+                alu_instruction_ptr->m_Word1_OP2.f.src0_abs           = 0x0;
+                alu_instruction_ptr->m_Word1_OP2.f.src1_abs           = 0x0;
+
+                //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0;
+                //alu_instruction_ptr->m_Word1_OP2.f.update_pred         = 0x0;
+                switch (scalar_channel_index) 
+                {
+                    case 0: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writex; 
+                        break;
+                    case 1: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writey; 
+                        break;
+                    case 2: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writez; 
+                        break;
+                    case 3: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writew; 
+                        break;
+                    default: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = 1; //SQ_SEL_MASK;
+                        break;
+                }            
+                alu_instruction_ptr->m_Word1_OP2.f.omod               = SQ_ALU_OMOD_OFF;
+            }
+        }
+
+        if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) )
+        {
+            return GL_FALSE;
+        }
+
+        /*
+         * Judge the type of current instruction, is it vector or scalar 
+         * instruction.
+         */        
+        if (is_single_scalar_operation) 
+        {
+            if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) )
+            {
+                return GL_FALSE;
+            }
+        }
+        else 
+        {
+            if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) )
+            {
+                return 1;
+            }
+        }
+
+        contiguous_slots_needed = 0;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean next_ins(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if( GL_TRUE == IsTex(pILInst->Opcode) )
+    {
+	    if (pILInst->TexSrcTarget == TEXTURE_RECT_INDEX) {
+		    if( GL_FALSE == assemble_tex_instruction(pAsm, GL_FALSE) ) 
+		    {
+			    radeon_error("Error assembling TEX instruction\n");
+			    return GL_FALSE;
+		    }
+	    } else {
+		    if( GL_FALSE == assemble_tex_instruction(pAsm, GL_TRUE) ) 
+		    {
+			    radeon_error("Error assembling TEX instruction\n");
+			    return GL_FALSE;
+		    }
+	    }
+    }
+    else 
+    {   //ALU      
+        if( GL_FALSE == assemble_alu_instruction(pAsm) ) 
+        {
+            radeon_error("Error assembling ALU instruction\n");
+            return GL_FALSE;
+        }
+    } 
+      
+    if(pAsm->D.dst.rtype == DST_REG_OUT) 
+    {
+        if(pAsm->D.dst.op3) 
+        {        
+            // There is no mask for OP3 instructions, so all channels are written        
+            pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] = 0xF;
+        }
+        else 
+        {
+            pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] 
+               |= (unsigned char)pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask;
+        }
+    }
+    
+    //reset for next inst.
+    pAsm->D.bits    = 0;
+    pAsm->S[0].bits = 0;
+    pAsm->S[1].bits = 0;
+    pAsm->S[2].bits = 0;
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode)
+{
+    BITS tmp;
+
+    checkop1(pAsm);
+
+    tmp = gethelpr(pAsm);
+
+    // opcode  tmp.x,    a.x
+    // MOV     dst,      tmp.x
+
+    pAsm->D.dst.opcode = opcode;
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+    pAsm->D.dst.writex = 1;
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // Now replicate result to all necessary channels in destination
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_ABS(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MAX;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    pAsm->S[1].bits = pAsm->S[0].bits;
+    flipneg_PVSSRC(&(pAsm->S[1].src));
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_ADD(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_ADD;
+ 
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_SUB)
+    {
+        flipneg_PVSSRC(&(pAsm->S[1].src));
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_BAD(char *opcode_str) 
+{
+    radeon_error("Not yet implemented instruction (%s)\n", opcode_str);
+    return GL_FALSE;
+}
+
+GLboolean assemble_CMP(r700_AssemblerBase *pAsm)
+{
+    int tmp;
+
+    if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_CNDGE;
+    pAsm->D.dst.op3     = 1;  
+
+    tmp = (-1);
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask)
+    {
+        //OP3 has no support for write mask
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+              
+    if( GL_FALSE == assemble_src(pAsm, 2, 1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, 2) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if (0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        //tmp for source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_COS(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_COS);
+}
+ 
+GLboolean assemble_DOT(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    pAsm->D.dst.opcode = SQ_OP2_INST_DOT4;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if(OPCODE_DP3 == pAsm->pILInst[pAsm->uiCurInst].Opcode)
+    {
+        zerocomp_PVSSRC(&(pAsm->S[0].src), 3);
+        zerocomp_PVSSRC(&(pAsm->S[1].src), 3);
+    }
+    else if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_DPH) 
+    {
+        onecomp_PVSSRC(&(pAsm->S[1].src), 3);
+    } 
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_DST(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    onecomp_PVSSRC(&(pAsm->S[0].src), 0);
+    onecomp_PVSSRC(&(pAsm->S[0].src), 3);
+
+    onecomp_PVSSRC(&(pAsm->S[1].src), 0);
+    onecomp_PVSSRC(&(pAsm->S[1].src), 2);
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_EX2(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_EXP_IEEE);
+}
+ 
+GLboolean assemble_FLR(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FLOOR;  
+
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_FLT_TO_INT);
+}
+
+GLboolean assemble_FRC(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FRACT; 
+
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_KIL(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT;  
+  
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = 0;
+    pAsm->D.dst.writey = 0;
+    pAsm->D.dst.writez = 0;
+    pAsm->D.dst.writew = 0;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = 0;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+
+    if(PROGRAM_TEMPORARY == pAsm->pILInst[pAsm->uiCurInst].DstReg.File)
+    {
+        pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].DstReg.Index + pAsm->starting_temp_register_number;
+    }
+    else
+    {   //PROGRAM_OUTPUT
+        pAsm->S[1].src.reg = pAsm->uiFP_OutputMap[pAsm->pILInst[pAsm->uiCurInst].DstReg.Index];
+    }
+  
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noswizzle_PVSSRC(&(pAsm->S[1].src));
+  
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->pR700Shader->killIsUsed = GL_TRUE;
+    
+    return GL_TRUE;
+}
+
+GLboolean assemble_LG2(r700_AssemblerBase *pAsm) 
+{ 
+    return assemble_math_function(pAsm, SQ_OP2_INST_LOG_IEEE);
+}
+
+GLboolean assemble_LRP(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    tmp = gethelpr(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_ADD;
+
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    nomask_PVSDST(&(pAsm->D.dst));
+
+          
+    if( GL_FALSE == assemble_src(pAsm, 1, 0) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 2, 1) )   
+    {
+	    return GL_FALSE;
+    }
+
+    neg_PVSSRC(&(pAsm->S[1].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+
+    if( GL_FALSE == assemble_src(pAsm, 0, 1) ) 
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 2, -1) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_MAD(struct r700_AssemblerBase *pAsm) 
+{
+    int tmp, ii;
+    GLboolean bReplaceDst = GL_FALSE;
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+	if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+	pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;  
+	pAsm->D.dst.op3     = 1; 
+
+	tmp = (-1);
+
+    if(PROGRAM_TEMPORARY == pILInst->DstReg.File)
+    {   /* TODO : more investigation on MAD src and dst using same register */
+        for(ii=0; ii<3; ii++)
+        {
+            if(   (PROGRAM_TEMPORARY == pILInst->SrcReg[ii].File)
+               && (pILInst->DstReg.Index == pILInst->SrcReg[ii].Index) )
+            {
+                bReplaceDst = GL_TRUE;
+                break;
+            }
+        }
+    }
+    if(0xF != pILInst->DstReg.WriteMask)
+    {   /* OP3 has no support for write mask */
+        bReplaceDst = GL_TRUE;
+    }
+
+	if(GL_TRUE == bReplaceDst)
+    {
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+              
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 2, -1) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+	if (GL_TRUE == bReplaceDst) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        //tmp for source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+/* LIT dst, src */
+GLboolean assemble_LIT(r700_AssemblerBase *pAsm)
+{
+    unsigned int dstReg;
+    unsigned int dstType;
+    unsigned int srcReg;
+    unsigned int srcType;
+    checkop1(pAsm);
+    int tmp = gethelpr(pAsm);
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+    dstReg  = pAsm->D.dst.reg;
+    dstType = pAsm->D.dst.rtype;
+    srcReg  = pAsm->S[0].src.reg;
+    srcType = pAsm->S[0].src.rtype;
+
+    /* dst.xw, <- 1.0  */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_MOV;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 1;
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_1;
+    pAsm->S[0].src.swizzley = SQ_SEL_1;
+    pAsm->S[0].src.swizzlez = SQ_SEL_1;
+    pAsm->S[0].src.swizzlew = SQ_SEL_1;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* dst.y = max(src.x, 0.0) */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_MAX;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 0;
+    pAsm->D.dst.writey   = 1;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_X;
+    pAsm->S[0].src.swizzley = SQ_SEL_X;
+    pAsm->S[0].src.swizzlez = SQ_SEL_X;
+    pAsm->S[0].src.swizzlew = SQ_SEL_X;
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[1].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[1].src));
+    pAsm->S[1].src.swizzlex = SQ_SEL_0;
+    pAsm->S[1].src.swizzley = SQ_SEL_0;
+    pAsm->S[1].src.swizzlez = SQ_SEL_0;
+    pAsm->S[1].src.swizzlew = SQ_SEL_0;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* before: dst.w = log(src.y)
+     * after : dst.x = log(src.y)
+     * why change dest register is that dst.w has been initialized as 1 before
+     */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_LOG_CLAMPED;
+    pAsm->D.dst.math     = 1;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_Y;
+    pAsm->S[0].src.swizzley = SQ_SEL_Y;
+    pAsm->S[0].src.swizzlez = SQ_SEL_Y;
+    pAsm->S[0].src.swizzlew = SQ_SEL_Y;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* before: tmp.x = amd MUL_LIT(src.w, dst.w, src.x ) */
+    /* after : tmp.x = amd MUL_LIT(src.w, dst.x, src.x ) */
+    pAsm->D.dst.opcode   = SQ_OP3_INST_MUL_LIT;
+    pAsm->D.dst.op3      = 1;
+    pAsm->D.dst.rtype    = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg      = tmp;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_W;
+    pAsm->S[0].src.swizzley = SQ_SEL_W;
+    pAsm->S[0].src.swizzlez = SQ_SEL_W;
+    pAsm->S[0].src.swizzlew = SQ_SEL_W;
+
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[1].src.reg   = dstReg;
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[1].src));
+    pAsm->S[1].src.swizzlex = SQ_SEL_X;
+    pAsm->S[1].src.swizzley = SQ_SEL_X;
+    pAsm->S[1].src.swizzlez = SQ_SEL_X;
+    pAsm->S[1].src.swizzlew = SQ_SEL_X;
+
+    pAsm->S[2].src.rtype = srcType;
+    pAsm->S[2].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[2].src));
+    pAsm->S[2].src.swizzlex = SQ_SEL_X;
+    pAsm->S[2].src.swizzley = SQ_SEL_X;
+    pAsm->S[2].src.swizzlez = SQ_SEL_X;
+    pAsm->S[2].src.swizzlew = SQ_SEL_X;
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* dst.z = exp(tmp.x) */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_EXP_IEEE;
+    pAsm->D.dst.math     = 1;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 0;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 1;
+    pAsm->D.dst.writew   = 0;
+
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_X;
+    pAsm->S[0].src.swizzley = SQ_SEL_X;
+    pAsm->S[0].src.swizzlez = SQ_SEL_X;
+    pAsm->S[0].src.swizzlew = SQ_SEL_X;
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MAX(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MAX; 
+	
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MIN(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MIN;  
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+ 
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MOV(r700_AssemblerBase *pAsm) 
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if (GL_FALSE == assemble_dst(pAsm))
+    {
+        return GL_FALSE;
+    }
+
+    if (GL_FALSE == assemble_src(pAsm, 0, -1))
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MUL(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) ) 
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_POW(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    checkop1(pAsm);
+
+    tmp = gethelpr(pAsm);
+
+    // LG2 tmp.x,     a.swizzle
+    pAsm->D.dst.opcode = SQ_OP2_INST_LOG_IEEE;  
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // MUL tmp.x,     tmp.x, b.swizzle
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // EX2 dst.mask,          tmp.x
+    // EX2 tmp.x,             tmp.x
+    pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE;
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // Now replicate result to all necessary channels in destination
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_RCP(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_RECIP_IEEE);
+}
+ 
+GLboolean assemble_RSQ(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_RECIPSQRT_IEEE);
+}
+ 
+GLboolean assemble_SIN(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_SIN);
+}
+ 
+GLboolean assemble_SCS(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+	checkop1(pAsm);
+
+	tmp = gethelpr(pAsm);
+
+	// COS tmp.x,    a.x
+	pAsm->D.dst.opcode = SQ_OP2_INST_COS;
+	pAsm->D.dst.math = 1;
+
+	setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+	pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+	pAsm->D.dst.reg = tmp;
+	pAsm->D.dst.writex = 1;
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if ( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	// SIN tmp.y,    a.x
+	pAsm->D.dst.opcode = SQ_OP2_INST_SIN;
+	pAsm->D.dst.math = 1;
+
+	setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+	pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+	pAsm->D.dst.reg = tmp;
+	pAsm->D.dst.writey = 1;
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	// MOV dst.mask,     tmp
+	pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+	pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+	pAsm->S[0].src.reg = tmp;
+
+	noswizzle_PVSSRC(&(pAsm->S[0].src));
+	pAsm->S[0].src.swizzlez = SQ_SEL_0;
+	pAsm->S[0].src.swizzlew = SQ_SEL_0;
+
+	if ( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_SGE(r700_AssemblerBase *pAsm) 
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_SETGE;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_SLT(r700_AssemblerBase *pAsm) 
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_SETGT;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+                
+    if( GL_FALSE == assemble_src(pAsm, 0, 1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, 0) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_STP(r700_AssemblerBase *pAsm) 
+{
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_TEX(r700_AssemblerBase *pAsm) 
+{
+    GLboolean src_const;
+
+    switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File)
+    {
+    case PROGRAM_CONSTANT:
+    case PROGRAM_LOCAL_PARAM:
+    case PROGRAM_ENV_PARAM:
+    case PROGRAM_STATE_VAR:
+        src_const = GL_TRUE;
+    case PROGRAM_TEMPORARY:
+    case PROGRAM_INPUT:
+        src_const = GL_FALSE;
+    }
+
+    if (GL_TRUE == src_const) 
+    {
+        radeon_error("TODO: Texture coordinates from a constant register not supported.\n");
+        return GL_FALSE;
+    }
+
+    switch (pAsm->pILInst[pAsm->uiCurInst].Opcode) 
+    {
+        case OPCODE_TEX:
+            pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;            
+            break;
+        case OPCODE_TXB:            
+            radeon_error("do not support TXB yet\n");
+            return GL_FALSE;
+            break;
+        case OPCODE_TXP:            
+            /* TODO : tex proj version : divid first 3 components by 4th */ 
+            pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
+            break;
+        default:
+            radeon_error("Internal error: bad texture op (not TEX)\n");
+            return GL_FALSE;
+            break;
+    }
+
+    // Set src1 to tex unit id
+    pAsm->S[1].src.reg   = pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit;
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+
+    //No sw info from mesa compiler, so hard code here.
+    pAsm->S[1].src.swizzlex = SQ_SEL_X;
+    pAsm->S[1].src.swizzley = SQ_SEL_Y;
+    pAsm->S[1].src.swizzlez = SQ_SEL_Z;
+    pAsm->S[1].src.swizzlew = SQ_SEL_W;
+
+    if( GL_FALSE == tex_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    if( GL_FALSE == tex_src(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_XPD(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    tmp = gethelpr(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+  
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y, SQ_SEL_0);
+    swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Y, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_0);
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask)
+    {
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Y, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_0);
+    swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y, SQ_SEL_0);
+
+    // result1 + (neg) result0
+    setaddrmode_PVSSRC(&(pAsm->S[2].src),ADDR_ABSOLUTE);
+    pAsm->S[2].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[2].src.reg   = tmp;
+
+    neg_PVSSRC(&(pAsm->S[2].src));
+    noswizzle_PVSSRC(&(pAsm->S[2].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        // Use tmp as source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean assemble_IF(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean AssembleInstr(GLuint uiNumberInsts,
+                        struct prog_instruction *pILInst, 
+						r700_AssemblerBase *pR700AsmCode)
+{
+    GLuint i;
+
+    pR700AsmCode->pILInst = pILInst;
+	for(i=0; i<uiNumberInsts; i++)
+    {
+        pR700AsmCode->uiCurInst = i;
+
+        switch (pILInst[i].Opcode)
+        {
+        case OPCODE_ABS: 
+            if ( GL_FALSE == assemble_ABS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_ADD: 
+        case OPCODE_SUB: 
+            if ( GL_FALSE == assemble_ADD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_ARL: 
+            radeon_error("Not yet implemented instruction OPCODE_ARL \n");
+            //if ( GL_FALSE == assemble_BAD("ARL") ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ARR: 
+            radeon_error("Not yet implemented instruction OPCODE_ARR \n");
+            //if ( GL_FALSE == assemble_BAD("ARR") ) 
+                return GL_FALSE;
+            break;
+
+        case OPCODE_CMP: 
+            if ( GL_FALSE == assemble_CMP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_COS: 
+            if ( GL_FALSE == assemble_COS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_DP3: 
+        case OPCODE_DP4: 
+        case OPCODE_DPH: 
+            if ( GL_FALSE == assemble_DOT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_DST: 
+            if ( GL_FALSE == assemble_DST(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_EX2: 
+            if ( GL_FALSE == assemble_EX2(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_EXP: 
+            radeon_error("Not yet implemented instruction OPCODE_EXP \n");
+            //if ( GL_FALSE == assemble_BAD("EXP") ) 
+                return GL_FALSE;
+            break; // approx of EX2
+
+        case OPCODE_FLR:     
+            if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        //case OP_FLR_INT: 
+        //    if ( GL_FALSE == assemble_FLR_INT() ) 
+        //        return GL_FALSE;
+        //    break;  
+
+        case OPCODE_FRC: 
+            if ( GL_FALSE == assemble_FRC(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_KIL: 
+            if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_LG2: 
+            if ( GL_FALSE == assemble_LG2(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_LIT:
+            if ( GL_FALSE == assemble_LIT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_LRP: 
+            if ( GL_FALSE == assemble_LRP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_LOG: 
+            radeon_error("Not yet implemented instruction OPCODE_LOG \n");
+            //if ( GL_FALSE == assemble_BAD("LOG") ) 
+                return GL_FALSE;
+            break; // approx of LG2
+
+        case OPCODE_MAD: 
+            if ( GL_FALSE == assemble_MAD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MAX: 
+            if ( GL_FALSE == assemble_MAX(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MIN: 
+            if ( GL_FALSE == assemble_MIN(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_MOV: 
+            if ( GL_FALSE == assemble_MOV(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MUL: 
+            if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+
+        case OPCODE_POW: 
+            if ( GL_FALSE == assemble_POW(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_RCP: 
+            if ( GL_FALSE == assemble_RCP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_RSQ: 
+            if ( GL_FALSE == assemble_RSQ(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_SIN: 
+            if ( GL_FALSE == assemble_SIN(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_SCS: 
+            if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_SGE: 
+            if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+        case OPCODE_SLT: 
+            if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+
+        //case OP_STP: 
+        //    if ( GL_FALSE == assemble_STP(pR700AsmCode) ) 
+        //        return GL_FALSE;
+        //    break;
+
+        case OPCODE_SWZ: 
+            if ( GL_FALSE == assemble_MOV(pR700AsmCode) ) 
+            {
+                return GL_FALSE; 
+            }
+            else
+            {
+                if( (i+1)<uiNumberInsts )
+                {
+                    if(OPCODE_END != pILInst[i+1].Opcode)
+                    {
+                        if( GL_TRUE == IsTex(pILInst[i+1].Opcode) )
+                        {
+                            pR700AsmCode->pInstDeps[i+1].nDstDep = i+1; //=1?
+                        }
+                    }
+                }
+            }
+            break;
+
+        case OPCODE_TEX: 
+        case OPCODE_TXB:  
+        case OPCODE_TXP: 
+            if ( GL_FALSE == assemble_TEX(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+
+        case OPCODE_XPD: 
+            if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_IF   : 
+            if ( GL_FALSE == assemble_IF(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ELSE : 
+            radeon_error("Not yet implemented instruction OPCODE_ELSE \n");
+            //if ( GL_FALSE == assemble_BAD("ELSE") ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ENDIF: 
+            if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+
+        //case OPCODE_EXPORT: 
+        //    if ( GL_FALSE == assemble_EXPORT() ) 
+        //        return GL_FALSE;
+        //    break;
+
+        case OPCODE_END: 
+			//pR700AsmCode->uiCurInst = i;
+			//This is to remaind that if in later exoort there is depth/stencil
+			//export, we need a mov to re-arrange DST channel, where using a
+			//psuedo inst, we will use this end inst to do it.
+            return GL_TRUE;
+
+        default:
+            radeon_error("internal: unknown instruction\n");
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean Process_Export(r700_AssemblerBase* pAsm,
+                         GLuint type,
+                         GLuint export_starting_index,
+                         GLuint export_count, 
+                         GLuint starting_register_number,
+                         GLboolean is_depth_export)
+{
+    unsigned char ucWriteMask;
+
+    check_current_clause(pAsm, CF_EMPTY_CLAUSE);
+    check_current_clause(pAsm, CF_EXPORT_CLAUSE); //alloc the cf_current_export_clause_ptr
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.type = type;
+
+    switch (type) 
+    {
+        case SQ_EXPORT_PIXEL:
+            if(GL_TRUE == is_depth_export) 
+            {
+                pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_PIXEL_Z;
+            }
+            else 
+            {
+                pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_PIXEL_MRT0 + export_starting_index;
+            }
+            break;
+
+        case SQ_EXPORT_POS:
+            pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_POS_0 + export_starting_index; 
+            break;
+
+        case SQ_EXPORT_PARAM:
+            pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = 0x0 + export_starting_index; 
+            break;
+
+        default:
+            radeon_error("Unknown export type: %d\n", type);
+            return GL_FALSE;
+            break;
+    }
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.rw_gpr      = starting_register_number;
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.rw_rel      = SQ_ABSOLUTE;
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.index_gpr   = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.elem_size   = 0x3; 
+
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.burst_count      = (export_count - 1);
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_EXPORT;  // _DONE
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.barrier          = 0x1;
+
+    if (export_count == 1) 
+    {
+        ucWriteMask = pAsm->pucOutMask[starting_register_number - pAsm->starting_export_register_number];
+	/* exports Z as a float into Red channel */
+	if (GL_TRUE == is_depth_export)
+	    ucWriteMask = 0x1;
+
+        if( (ucWriteMask & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_X;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>1) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_Y;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>2) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_Z;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>3) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_W;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_MASK;
+        }
+    }
+    else 
+    {
+        // This should only be used if all components for all registers have been written
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_X;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_Y;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_Z;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_W;
+    }
+
+    pAsm->cf_last_export_ptr = pAsm->cf_current_export_clause_ptr;
+
+    return GL_TRUE;
+}
+
+GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select)
+{
+	gl_inst_opcode Opcode_save = pAsm->pILInst[pAsm->uiCurInst].Opcode; //Should be OPCODE_END
+    pAsm->pILInst[pAsm->uiCurInst].Opcode = OPCODE_MOV;
+
+    // MOV depth_export_register.hw_depth_channel, depth_export_register.depth_channel_select
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = pAsm->depth_export_register_number;
+
+    pAsm->D.dst.writex = 1;   // depth          goes in R channel for HW                       
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = pAsm->depth_export_register_number;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), depth_channel_select);
+
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->pILInst[pAsm->uiCurInst].Opcode = Opcode_save;
+
+    return GL_TRUE;
+}
+ 
+GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode,
+                                   GLbitfield          OutputsWritten)  
+{ 
+    unsigned int unBit;
+
+    if(pR700AsmCode->depth_export_register_number >= 0) 
+    {
+        if( GL_FALSE == Move_Depth_Exports_To_Correct_Channels(pR700AsmCode, SQ_SEL_Z) )  // depth
+		{
+			return GL_FALSE;
+		}
+    }
+
+    unBit = 1 << FRAG_RESULT_COLOR;
+	if(OutputsWritten & unBit)
+	{
+		if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PIXEL, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->uiFP_OutputMap[FRAG_RESULT_COLOR], 
+                                       GL_FALSE) ) 
+        {
+            return GL_FALSE;
+        }
+	}
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PIXEL, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->uiFP_OutputMap[FRAG_RESULT_DEPTH], 
+                                       GL_TRUE)) 
+        {
+            return GL_FALSE;
+        }
+	}
+
+    if(pR700AsmCode->cf_last_export_ptr != NULL) 
+    {
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst        = SQ_CF_INST_EXPORT_DONE;
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.end_of_program = 0x1;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode,
+                                 GLbitfield          OutputsWritten)  
+{
+    unsigned int unBit;
+    unsigned int i;
+
+    GLuint export_starting_index  = 0;
+    GLuint export_count           = pR700AsmCode->number_of_exports;
+
+    unBit = 1 << VERT_RESULT_HPOS;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_POS, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_HPOS],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_count--;
+
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;
+	}
+
+    pR700AsmCode->number_of_exports = export_count;
+
+	unBit = 1 << VERT_RESULT_COL0;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_PARAM, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_COL0],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_starting_index++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL1;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_PARAM, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_COL1],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_starting_index++;
+	}
+
+        unBit = 1 << VERT_RESULT_FOGC;
+        if(OutputsWritten & unBit)
+        {
+        if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PARAM,
+                                       export_starting_index,
+                                       1,
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_FOGC],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_starting_index++;
+        }
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (VERT_RESULT_TEX0 + i);
+		if(OutputsWritten & unBit)
+		{
+            if( GL_FALSE == Process_Export(pR700AsmCode,
+                                          SQ_EXPORT_PARAM, 
+                                          export_starting_index, 
+                                          1, 
+                                          pR700AsmCode->ucVP_OutputMap[VERT_RESULT_TEX0 + i],
+                                          GL_FALSE) )
+            {
+                return GL_FALSE;
+            }
+
+            export_starting_index++;
+		}
+	}
+
+    // At least one param should be exported
+    if (export_count) 
+    {
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;    
+    }
+    else
+    {
+        if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PARAM, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->starting_export_register_number,
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+      
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_1;
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;
+    }
+
+    pR700AsmCode->cf_last_export_ptr->m_Word1.f.end_of_program = 0x1;
+
+    return GL_TRUE;
+}
+
+GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode)
+{
+    FREE(pR700AsmCode->pucOutMask);
+    FREE(pR700AsmCode->pInstDeps);
+    return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h
new file mode 100644
index 0000000000..f9c4d849c6
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_assembler.h
@@ -0,0 +1,512 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_ASSEMBLER_H_
+#define _R700_ASSEMBLER_H_
+
+#include "main/mtypes.h"
+#include "shader/prog_instruction.h"
+
+#include "r700_chip.h"
+#include "r700_shaderinst.h"
+#include "r700_shader.h"
+
+typedef enum SHADER_PIPE_TYPE 
+{
+    SPT_VP = 0,
+    SPT_FP = 1
+} SHADER_PIPE_TYPE;
+
+typedef enum ConstantCycles 
+{
+    NUMBER_OF_CYCLES     = 3,
+    NUMBER_OF_COMPONENTS = 4
+} ConstantCycles;
+
+typedef enum  HARDWARE_LIMIT_VALUES  
+{
+   TEMPORARY_REGISTER_OFFSET = SQ_ALU_SRC_GPR_BASE,
+   MAX_TEMPORARY_REGISTERS   = SQ_ALU_SRC_GPR_SIZE,
+   MAX_CONSTANT_REGISTERS    = SQ_ALU_SRC_CFILE_SIZE,
+   CFILE_REGISTER_OFFSET     = SQ_ALU_SRC_CFILE_BASE,
+   NUMBER_OF_INPUT_COLORS    = 2,
+   NUMBER_OF_OUTPUT_COLORS   = 8,
+   NUMBER_OF_TEXTURE_UNITS   = 16,
+   MEGA_FETCH_BYTES          = 32
+} HARDWARE_LIMIT_VALUES;
+
+typedef enum AddressMode 
+{
+    ADDR_ABSOLUTE          = 0,
+    ADDR_RELATIVE_A0       = 1,
+    ADDR_RELATIVE_FLI_0    = 2,
+    NUMBER_OF_ADDR_MOD     = 3
+} AddressMode;
+
+typedef enum SrcRegisterType 
+{
+    SRC_REG_TEMPORARY      = 0,
+    SRC_REG_INPUT          = 1,
+    SRC_REG_CONSTANT       = 2,
+    SRC_REG_ALT_TEMPORARY  = 3,
+    NUMBER_OF_SRC_REG_TYPE = 4
+} SrcRegisterType;
+
+typedef enum DstRegisterType 
+{
+    DST_REG_TEMPORARY      = 0,
+    DST_REG_A0             = 1,
+    DST_REG_OUT            = 2,
+    DST_REG_OUT_X_REPL     = 3,
+    DST_REG_ALT_TEMPORARY  = 4,
+    DST_REG_INPUT          = 5,
+    NUMBER_OF_DST_REG_TYPE = 6
+} DstRegisterType;
+
+typedef unsigned int BITS;
+
+typedef struct PVSDSTtag 
+{
+	BITS opcode:8;     //(:6)  //@@@ really should be 10 bits for OP2
+	BITS math:1;
+	BITS predicated:1; //10   //8
+	BITS pred_inv  :1; //11   //8
+
+	BITS rtype:3;
+	BITS reg:10;       //24   //20
+
+	BITS writex:1;
+	BITS writey:1;
+	BITS writez:1;
+	BITS writew:1;     //28
+
+	BITS op3:1;       // 29  Represents *_OP3_* ALU opcode
+
+	BITS dualop:1;    // 30  //26
+
+	BITS addrmode0:1; //31   //29
+	BITS addrmode1:1; //32
+} PVSDST;
+
+typedef struct PVSSRCtag 
+{
+	BITS rtype:4;            
+	BITS addrmode0:1;        
+	BITS reg:10;      //15     (8)
+	BITS swizzlex:3;
+	BITS swizzley:3;
+	BITS swizzlez:3;
+	BITS swizzlew:3;  //27        
+
+	BITS negx:1;
+	BITS negy:1;
+	BITS negz:1;
+	BITS negw:1;      //31
+	//BITS addrsel:2;
+	BITS addrmode1:1; //32
+} PVSSRC;
+
+typedef struct PVSMATHtag 
+{
+	BITS rtype:4;
+	BITS spare:1;
+	BITS reg:8;
+	BITS swizzlex:3;
+	BITS swizzley:3;
+	BITS dstoff:2; // 2 bits of dest offset into alt ram
+	BITS opcode:4;
+	BITS negx:1;
+	BITS negy:1;
+	BITS dstcomp:2; // select dest component
+	BITS spare2:3;
+} PVSMATH;
+
+typedef union PVSDWORDtag 
+{
+	BITS    bits;
+	PVSDST  dst;
+	PVSSRC  src;
+	PVSMATH math;
+	float   f;
+} PVSDWORD;
+
+typedef struct VAP_OUT_VTX_FMT_0tag 
+{
+	BITS pos:1;      // 0
+	BITS misc:1;
+	BITS clip_dist0:1;
+	BITS clip_dist1:1;
+	BITS pos_param:1; // 4
+
+	BITS color0:1;    // 5
+	BITS color1:1;
+	BITS color2:1;
+	BITS color3:1;
+	BITS color4:1;
+	BITS color5:1;
+	BITS color6:1;
+	BITS color7:1;
+
+	BITS normal:1;    
+
+	BITS depth:1;          // 14
+
+	BITS point_size:1;     // 15   
+	BITS edge_flag:1;      
+	BITS rta_index:1;      //     shares same channel as kill_flag
+	BITS kill_flag:1;
+	BITS viewport_index:1; // 19   
+
+	BITS resvd1:12;        // 20
+} VAP_OUT_VTX_FMT_0;
+
+typedef struct VAP_OUT_VTX_FMT_1tag 
+{
+	BITS tex0comp:3;
+	BITS tex1comp:3;
+	BITS tex2comp:3;
+	BITS tex3comp:3;
+	BITS tex4comp:3;
+	BITS tex5comp:3;
+	BITS tex6comp:3;
+	BITS tex7comp:3;
+
+	BITS resvd:8;
+} VAP_OUT_VTX_FMT_1;
+
+typedef struct VAP_OUT_VTX_FMT_2tag 
+{
+	BITS tex8comp :3;
+	BITS tex9comp :3;
+	BITS tex10comp:3;
+	BITS tex11comp:3;
+	BITS tex12comp:3;
+	BITS tex13comp:3;
+	BITS tex14comp:3;
+	BITS tex15comp:3;
+
+	BITS resvd:8;
+} VAP_OUT_VTX_FMT_2;
+
+typedef struct OUT_FRAGMENT_FMT_0tag 
+{
+	BITS color0:1;
+	BITS color1:1;
+	BITS color2:1;
+	BITS color3:1;
+	BITS color4:1;
+	BITS color5:1;
+	BITS color6:1;
+	BITS color7:1;
+
+	BITS depth:1;
+	BITS stencil_ref:1;
+	BITS coverage_to_mask:1;
+	BITS mask:1;
+
+	BITS resvd1:20;
+} OUT_FRAGMENT_FMT_0;
+
+typedef enum  CF_CLAUSE_TYPE 
+{
+   CF_EXPORT_CLAUSE,
+   CF_ALU_CLAUSE,
+   CF_TEX_CLAUSE,
+   CF_VTX_CLAUSE,
+   CF_OTHER_CLAUSE,
+   CF_EMPTY_CLAUSE,
+   NUMBER_CF_CLAUSE_TYPES
+} CF_CLAUSE_TYPE;
+
+enum 
+{
+    MAX_BOOL_CONSTANTS   = 32,
+    MAX_INT_CONSTANTS    = 32,
+    MAX_FLOAT_CONSTANTS  = 256,
+
+    FC_NONE = 0,
+    FC_IF = 1,
+    FC_LOOP = 2,
+    FC_REP = 3,
+
+    COND_NONE = 0,
+    COND_BOOL = 1,
+    COND_PRED = 2,
+    COND_ALU = 3,
+
+    SAFEDIST_TEX = 6, ///< safe distance for using result of texture lookup in alu or another tex lookup
+    SAFEDIST_ALU = 6 ///< the same for alu->fc
+};
+
+typedef struct FC_LEVEL 
+{
+	unsigned int           first; ///< first fc instruction on level (if, rep, loop)
+	unsigned int*          mid; ///< middle instructions - else or all breaks on this level
+	unsigned int           midLen;
+	unsigned int           type;
+	unsigned int           cond;
+	unsigned int           inv;
+	unsigned int           bpush; ///< 1 if first instruction does branch stack push
+			 int           id; ///< id of bool or int variable
+} FC_LEVEL;
+
+typedef struct VTX_FETCH_METHOD 
+{
+	GLboolean bEnableMini;
+	GLuint mega_fetch_remainder;
+} VTX_FETCH_METHOD;
+
+typedef struct r700_AssemblerBase 
+{
+	R700ControlFlowSXClause*      cf_last_export_ptr;
+	R700ControlFlowSXClause*      cf_current_export_clause_ptr;
+	R700ControlFlowALUClause*     cf_current_alu_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_tex_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_vtx_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_cf_clause_ptr;
+
+    //Result shader
+    R700_Shader * pR700Shader;
+
+	// No clause has been created yet
+	CF_CLAUSE_TYPE cf_current_clause_type;
+
+	GLuint number_of_exports;
+	GLuint number_of_colorandz_exports;
+	GLuint number_of_export_opcodes;
+
+	PVSDWORD D;
+	PVSDWORD S[3];
+
+	unsigned int uLastPosUpdate;
+
+	OUT_FRAGMENT_FMT_0     fp_stOutFmt0;
+
+	unsigned int uIIns;
+	unsigned int uOIns;
+	unsigned int number_used_registers;
+	unsigned int uUsedConsts; 
+
+	// Fragment programs
+	unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX];
+	unsigned int uiFP_OutputMap[FRAG_RESULT_MAX];
+	unsigned int uBoolConsts;
+	unsigned int uIntConsts;
+	unsigned int uInsts;
+	unsigned int uConsts;
+
+	// Vertex programs
+	unsigned char ucVP_AttributeMap[VERT_ATTRIB_MAX];
+	unsigned char ucVP_OutputMap[VERT_RESULT_MAX];
+
+    unsigned char * pucOutMask;
+
+	//-----------------------------------------------------------------------------------
+	// flow control members
+	//-----------------------------------------------------------------------------------
+	unsigned int FCSP;
+	FC_LEVEL fc_stack[32];
+
+	unsigned int branch_depth;
+	unsigned int max_branch_depth;
+
+	//-----------------------------------------------------------------------------------
+	// ArgSubst used in Assemble_Source() function
+	//-----------------------------------------------------------------------------------
+	int aArgSubst[4];
+
+    GLint hw_gpr[ NUMBER_OF_CYCLES ][ NUMBER_OF_COMPONENTS ];
+    GLint hw_cfile_addr[ NUMBER_OF_COMPONENTS ];
+    GLint hw_cfile_chan[ NUMBER_OF_COMPONENTS ];
+
+    GLuint uOutputs;
+  
+    GLint color_export_register_number[NUMBER_OF_OUTPUT_COLORS];
+	GLint depth_export_register_number;
+
+	GLint stencil_export_register_number;
+	GLint coverage_to_mask_export_register_number;
+	GLint mask_export_register_number;
+
+	GLuint starting_export_register_number;
+	GLuint starting_vfetch_register_number;
+	GLuint starting_temp_register_number;
+	GLuint uHelpReg;
+	GLuint uFirstHelpReg;
+
+	GLboolean input_position_is_used;
+	GLboolean input_normal_is_used;
+
+	GLboolean input_color_is_used[NUMBER_OF_INPUT_COLORS];
+  
+	GLboolean input_texture_unit_is_used[NUMBER_OF_TEXTURE_UNITS];
+  
+    R700VertexGenericFetch* vfetch_instruction_ptr_array[VERT_ATTRIB_MAX];
+  
+	GLuint number_of_inputs;
+
+    InstDeps *pInstDeps;
+
+    SHADER_PIPE_TYPE currentShaderType;
+    struct prog_instruction * pILInst;
+    GLuint             uiCurInst;
+    GLboolean   bR6xx;
+} r700_AssemblerBase;
+
+//Internal use
+BITS addrmode_PVSDST(PVSDST * pPVSDST);
+void setaddrmode_PVSDST(PVSDST * pPVSDST, BITS addrmode);
+void nomask_PVSDST(PVSDST * pPVSDST);
+BITS addrmode_PVSSRC(PVSSRC* pPVSSRC);
+void setaddrmode_PVSSRC(PVSSRC* pPVSSRC, BITS addrmode);
+void setswizzle_PVSSRC(PVSSRC* pPVSSRC, BITS swz);
+void noswizzle_PVSSRC(PVSSRC* pPVSSRC);
+void swizzleagain_PVSSRC(PVSSRC * pPVSSRC, BITS x, BITS y, BITS z, BITS w);
+void neg_PVSSRC(PVSSRC* pPVSSRC);
+void noneg_PVSSRC(PVSSRC* pPVSSRC);
+void flipneg_PVSSRC(PVSSRC* pPVSSRC);
+void zerocomp_PVSSRC(PVSSRC* pPVSSRC, int c);
+void onecomp_PVSSRC(PVSSRC* pPVSSRC, int c);
+BITS is_misc_component_exported(VAP_OUT_VTX_FMT_0* pOutVTXFmt0);
+BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ;
+GLboolean is_reduction_opcode(PVSDWORD * dest);
+GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size);
+
+unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm);
+
+GLboolean IsTex(gl_inst_opcode Opcode);
+GLboolean IsAlu(gl_inst_opcode Opcode);
+int check_current_clause(r700_AssemblerBase* pAsm,
+					     CF_CLAUSE_TYPE      new_clause_type);
+GLboolean add_vfetch_instruction(r700_AssemblerBase*     pAsm,
+								 R700VertexInstruction*  vertex_instruction_ptr);
+GLboolean add_tex_instruction(r700_AssemblerBase*     pAsm,
+                              R700TextureInstruction* tex_instruction_ptr);
+GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm,
+								GLuint gl_client_id,
+                                GLuint destination_register,
+								GLuint number_of_elements,
+                                GLenum dataElementType,
+								VTX_FETCH_METHOD* pFetchMethod);
+GLuint gethelpr(r700_AssemblerBase* pAsm);
+void resethelpr(r700_AssemblerBase* pAsm);
+void checkop_init(r700_AssemblerBase* pAsm);
+GLboolean mov_temp(r700_AssemblerBase* pAsm, int src);
+GLboolean checkop1(r700_AssemblerBase* pAsm);
+GLboolean checkop2(r700_AssemblerBase* pAsm);
+GLboolean checkop3(r700_AssemblerBase* pAsm);
+GLboolean assemble_src(r700_AssemblerBase *pAsm,
+                       int src, 
+                       int fld);
+GLboolean assemble_dst(r700_AssemblerBase *pAsm);
+GLboolean tex_dst(r700_AssemblerBase *pAsm);
+GLboolean tex_src(r700_AssemblerBase *pAsm);
+GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalized);
+void initialize(r700_AssemblerBase *pAsm);
+GLboolean assemble_alu_src(R700ALUInstruction*  alu_instruction_ptr,
+                           int                  source_index,
+                           PVSSRC*              pSource,
+                           BITS                 scalar_channel_index);
+GLboolean add_alu_instruction(r700_AssemblerBase* pAsm,
+                              R700ALUInstruction* alu_instruction_ptr,
+                              GLuint              contiguous_slots_needed);
+void get_src_properties(R700ALUInstruction*  alu_instruction_ptr,
+                        int                  source_index,
+                        BITS*                psrc_sel,
+                        BITS*                psrc_rel,
+                        BITS*                psrc_chan,
+                        BITS*                psrc_neg);
+int is_cfile(BITS sel);
+int is_const(BITS sel);
+int is_gpr(BITS sel);
+GLboolean reserve_cfile(r700_AssemblerBase* pAsm, 
+                        GLuint sel, 
+                        GLuint chan);
+GLboolean reserve_gpr(r700_AssemblerBase* pAsm, GLuint sel, GLuint chan, GLuint cycle);
+GLboolean cycle_for_scalar_bank_swizzle(const int swiz, const int sel, GLuint* pCycle);
+GLboolean cycle_for_vector_bank_swizzle(const int swiz, const int sel, GLuint* pCycle);
+GLboolean check_scalar(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr);
+GLboolean check_vector(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr);
+GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm);
+GLboolean next_ins(r700_AssemblerBase *pAsm);
+GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode);
+GLboolean assemble_ABS(r700_AssemblerBase *pAsm);
+GLboolean assemble_ADD(r700_AssemblerBase *pAsm);
+GLboolean assemble_BAD(char *opcode_str);
+GLboolean assemble_CMP(r700_AssemblerBase *pAsm);
+GLboolean assemble_COS(r700_AssemblerBase *pAsm);
+GLboolean assemble_DOT(r700_AssemblerBase *pAsm);
+GLboolean assemble_DST(r700_AssemblerBase *pAsm);
+GLboolean assemble_EX2(r700_AssemblerBase *pAsm);
+GLboolean assemble_FLR(r700_AssemblerBase *pAsm);
+GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm);
+GLboolean assemble_FRC(r700_AssemblerBase *pAsm);
+GLboolean assemble_KIL(r700_AssemblerBase *pAsm);
+GLboolean assemble_LG2(r700_AssemblerBase *pAsm);
+GLboolean assemble_LRP(r700_AssemblerBase *pAsm);
+GLboolean assemble_MAD(r700_AssemblerBase *pAsm);
+GLboolean assemble_LIT(r700_AssemblerBase *pAsm);
+GLboolean assemble_MAX(r700_AssemblerBase *pAsm);
+GLboolean assemble_MIN(r700_AssemblerBase *pAsm);
+GLboolean assemble_MOV(r700_AssemblerBase *pAsm);
+GLboolean assemble_MUL(r700_AssemblerBase *pAsm);
+GLboolean assemble_POW(r700_AssemblerBase *pAsm);
+GLboolean assemble_RCP(r700_AssemblerBase *pAsm);
+GLboolean assemble_RSQ(r700_AssemblerBase *pAsm);
+GLboolean assemble_SIN(r700_AssemblerBase *pAsm);
+GLboolean assemble_SCS(r700_AssemblerBase *pAsm);
+GLboolean assemble_SGE(r700_AssemblerBase *pAsm);
+GLboolean assemble_SLT(r700_AssemblerBase *pAsm);
+GLboolean assemble_STP(r700_AssemblerBase *pAsm);
+GLboolean assemble_TEX(r700_AssemblerBase *pAsm);
+GLboolean assemble_XPD(r700_AssemblerBase *pAsm);
+GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm);
+GLboolean assemble_IF(r700_AssemblerBase *pAsm);
+GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm);
+
+GLboolean Process_Export(r700_AssemblerBase* pAsm,
+                         GLuint type, 
+                         GLuint export_starting_index,
+                         GLuint export_count, 
+                         GLuint starting_register_number,
+                         GLboolean is_depth_export);
+GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, 
+                                                 BITS depth_channel_select);
+
+
+//Interface
+GLboolean AssembleInstr(GLuint uiNumberInsts,
+                        struct prog_instruction *pILInst, 
+						r700_AssemblerBase *pR700AsmCode);
+GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten);  
+GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten);
+
+int       Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader);
+GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode);
+
+#endif //_R700_ASSEMBLER_H_
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
new file mode 100644
index 0000000000..37bff56f5a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -0,0 +1,1268 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/imports.h"
+#include "main/glheader.h"
+#include "main/simple_list.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_state.h"
+#include "r600_tex.h"
+#include "r700_oglprog.h"
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+#include "r700_ioctl.h"
+
+#include "radeon_mipmap_tree.h"
+
+static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t         *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	struct radeon_bo *bo = NULL;
+	unsigned int i;
+	BATCH_LOCALS(&context->radeon);
+
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
+		radeonTexObj *t = r700->textures[i];
+		if (t) {
+			if (!t->image_override)
+				bo = t->mt->bo;
+			else
+				bo = t->bo;
+			if (bo) {
+
+				r700SyncSurf(context, bo,
+					     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM,
+					     0, TC_ACTION_ENA_bit);
+
+				BEGIN_BATCH_NO_AUTOSTATE(9 + 4);
+				R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
+				R600_OUT_BATCH(i * 7);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE3);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE4);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE5);
+				R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6);
+				R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2,
+						     bo,
+						     0,
+						     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+				R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3,
+						     bo,
+						     r700->textures[i]->SQ_TEX_RESOURCE3,
+						     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+				END_BATCH();
+				COMMIT_BATCH();
+			}
+		}
+	}
+}
+
+static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t         *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned int i;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
+		radeonTexObj *t = r700->textures[i];
+		if (t) {
+			BEGIN_BATCH_NO_AUTOSTATE(5);
+			R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3));
+			R600_OUT_BATCH(i * 3);
+			R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0);
+			R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1);
+			R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2);
+			END_BATCH();
+			COMMIT_BATCH();
+		}
+	}
+}
+
+static void r700SendTexBorderColorState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t         *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned int i;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
+		radeonTexObj *t = r700->textures[i];
+		if (t) {
+			BEGIN_BATCH_NO_AUTOSTATE(2 + 4);
+			R600_OUT_BATCH_REGSEQ((TD_PS_SAMPLER0_BORDER_RED + (i * 16)), 4);
+			R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_RED);
+			R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_GREEN);
+			R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_BLUE);
+			R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_ALPHA);
+			END_BATCH();
+			COMMIT_BATCH();
+		}
+	}
+}
+
+static void r700SetupVTXConstants(GLcontext  * ctx,
+				  unsigned int nStreamID,
+				  void *       pAos,
+				  unsigned int size,      /* number of elements in vector */
+				  unsigned int stride,
+				  unsigned int count)     /* number of vectors in stream */
+{
+    context_t *context = R700_CONTEXT(ctx);
+    struct radeon_aos * paos = (struct radeon_aos *)pAos;
+    BATCH_LOCALS(&context->radeon);
+    radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+    unsigned int uSQ_VTX_CONSTANT_WORD0_0;
+    unsigned int uSQ_VTX_CONSTANT_WORD1_0;
+    unsigned int uSQ_VTX_CONSTANT_WORD2_0 = 0;
+    unsigned int uSQ_VTX_CONSTANT_WORD3_0 = 0;
+    unsigned int uSQ_VTX_CONSTANT_WORD6_0 = 0;
+
+    if (!paos->bo)
+	    return;
+
+    if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS880) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710))
+	    r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, TC_ACTION_ENA_bit);
+    else
+	    r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, VC_ACTION_ENA_bit);
+
+    uSQ_VTX_CONSTANT_WORD0_0 = paos->offset;
+    uSQ_VTX_CONSTANT_WORD1_0 = count * (size * 4) - 1;
+
+    SETfield(uSQ_VTX_CONSTANT_WORD2_0, 0, BASE_ADDRESS_HI_shift, BASE_ADDRESS_HI_mask); /* TODO */
+    SETfield(uSQ_VTX_CONSTANT_WORD2_0, stride, SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift,
+	     SQ_VTX_CONSTANT_WORD2_0__STRIDE_mask);
+    SETfield(uSQ_VTX_CONSTANT_WORD2_0, GetSurfaceFormat(GL_FLOAT, size, NULL),
+	     SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift,
+	     SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask); /* TODO : trace back api for initial data type, not only GL_FLOAT */
+    SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_SCALED,
+	     SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask);
+    SETbit(uSQ_VTX_CONSTANT_WORD2_0, SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit);
+
+    SETfield(uSQ_VTX_CONSTANT_WORD3_0, 1, MEM_REQUEST_SIZE_shift, MEM_REQUEST_SIZE_mask);
+    SETfield(uSQ_VTX_CONSTANT_WORD6_0, SQ_TEX_VTX_VALID_BUFFER,
+	     SQ_TEX_RESOURCE_WORD6_0__TYPE_shift, SQ_TEX_RESOURCE_WORD6_0__TYPE_mask);
+
+    BEGIN_BATCH_NO_AUTOSTATE(9 + 2);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
+    R600_OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD0_0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0);
+    R600_OUT_BATCH(0);
+    R600_OUT_BATCH(0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0);
+    R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD0_0,
+                         paos->bo,
+                         uSQ_VTX_CONSTANT_WORD0_0,
+                         RADEON_GEM_DOMAIN_GTT, 0, 0);
+    END_BATCH();
+    COMMIT_BATCH();
+
+}
+
+void r700SetupStreams(GLcontext *ctx)
+{
+    context_t         *context = R700_CONTEXT(ctx);
+     struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *vb = &tnl->vb;
+    unsigned int i, j = 0;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+    R600_STATECHANGE(context, vtx);
+
+    for(i=0; i<VERT_ATTRIB_MAX; i++) {
+	    if(vpc->mesa_program.Base.InputsRead & (1 << i)) {
+		    rcommon_emit_vector(ctx,
+					&context->radeon.tcl.aos[j],
+					vb->AttribPtr[i]->data,
+					vb->AttribPtr[i]->size,
+					vb->AttribPtr[i]->stride,
+					vb->Count);
+		    j++;
+	    }
+    }
+    context->radeon.tcl.aos_count = j;
+}
+
+static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+    context_t         *context = R700_CONTEXT(ctx);
+    struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    unsigned int i, j = 0;
+    BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+    if (context->radeon.tcl.aos_count == 0)
+	    return;
+
+    BEGIN_BATCH_NO_AUTOSTATE(6);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
+    R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(0);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
+    R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(0);
+    END_BATCH();
+    COMMIT_BATCH();
+
+    for(i=0; i<VERT_ATTRIB_MAX; i++) {
+	    if(vpc->mesa_program.Base.InputsRead & (1 << i)) {
+		    /* currently aos are packed */
+		    r700SetupVTXConstants(ctx,
+					  i,
+					  (void*)(&context->radeon.tcl.aos[j]),
+					  (unsigned int)context->radeon.tcl.aos[j].components,
+					  (unsigned int)context->radeon.tcl.aos[j].stride * 4,
+					  (unsigned int)context->radeon.tcl.aos[j].count);
+		    j++;
+	    }
+    }
+}
+
+static void r700SetRenderTarget(context_t *context, int id)
+{
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct radeon_renderbuffer *rrb;
+    unsigned int nPitchInPixel;
+
+    rrb = radeon_get_colorbuffer(&context->radeon);
+    if (!rrb || !rrb->bo) {
+	    return;
+    }
+
+    R600_STATECHANGE(context, cb_target);
+
+    /* color buffer */
+    r700->render_target[id].CB_COLOR0_BASE.u32All = context->radeon.state.color.draw_offset;
+
+    nPitchInPixel = rrb->pitch/rrb->cpp;
+    SETfield(r700->render_target[id].CB_COLOR0_SIZE.u32All, (nPitchInPixel/8)-1,
+             PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask);
+    SETfield(r700->render_target[id].CB_COLOR0_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1,
+             SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask);
+    r700->render_target[id].CB_COLOR0_BASE.u32All = 0;
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, ENDIAN_NONE, ENDIAN_shift, ENDIAN_mask);
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, ARRAY_LINEAR_GENERAL,
+             CB_COLOR0_INFO__ARRAY_MODE_shift, CB_COLOR0_INFO__ARRAY_MODE_mask);
+    if(4 == rrb->cpp)
+    {
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, COLOR_8_8_8_8,
+                 CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask);
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, SWAP_ALT, COMP_SWAP_shift, COMP_SWAP_mask);
+    }
+    else
+    {
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, COLOR_5_6_5,
+                 CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask);
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, SWAP_ALT_REV,
+                 COMP_SWAP_shift, COMP_SWAP_mask);
+    }
+    SETbit(r700->render_target[id].CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit);
+    SETbit(r700->render_target[id].CB_COLOR0_INFO.u32All, BLEND_CLAMP_bit);
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, NUMBER_UNORM, NUMBER_TYPE_shift, NUMBER_TYPE_mask);
+
+    r700->render_target[id].enabled = GL_TRUE;
+}
+
+static void r700SetDepthTarget(context_t *context)
+{
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct radeon_renderbuffer *rrb;
+    unsigned int nPitchInPixel;
+
+    rrb = radeon_get_depthbuffer(&context->radeon);
+    if (!rrb)
+	    return;
+
+    R600_STATECHANGE(context, db_target);
+
+    /* depth buf */
+    r700->DB_DEPTH_SIZE.u32All = 0;
+    r700->DB_DEPTH_BASE.u32All = 0;
+    r700->DB_DEPTH_INFO.u32All = 0;
+    r700->DB_DEPTH_VIEW.u32All = 0;
+
+    nPitchInPixel = rrb->pitch/rrb->cpp;
+
+    SETfield(r700->DB_DEPTH_SIZE.u32All, (nPitchInPixel/8)-1,
+             PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask);
+    SETfield(r700->DB_DEPTH_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1,
+             SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); /* size in pixel / 64 - 1 */
+
+    if(4 == rrb->cpp)
+    {
+        SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_8_24,
+                 DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask);
+    }
+    else
+    {
+        SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_16,
+                     DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask);
+    }
+    SETfield(r700->DB_DEPTH_INFO.u32All, ARRAY_2D_TILED_THIN1,
+             DB_DEPTH_INFO__ARRAY_MODE_shift, DB_DEPTH_INFO__ARRAY_MODE_mask);
+    /* r700->DB_PREFETCH_LIMIT.bits.DEPTH_HEIGHT_TILE_MAX = (context->currentDraw->h >> 3) - 1; */ /* z buffer sie may much bigger than what need, so use actual used h. */
+}
+
+static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	rrb = radeon_get_depthbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
+	}
+
+	r700SetDepthTarget(context);
+
+        BEGIN_BATCH_NO_AUTOSTATE(8 + 2);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_SIZE, 2);
+	R600_OUT_BATCH(r700->DB_DEPTH_SIZE.u32All);
+	R600_OUT_BATCH(r700->DB_DEPTH_VIEW.u32All);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 2);
+	R600_OUT_BATCH(r700->DB_DEPTH_BASE.u32All);
+	R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All);
+	R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All,
+			     rrb->bo,
+			     r700->DB_DEPTH_BASE.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0);
+        END_BATCH();
+
+	if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
+	    (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
+		R600_OUT_BATCH(1 << 0);
+		END_BATCH();
+	}
+
+	COMMIT_BATCH();
+
+}
+
+static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	BATCH_LOCALS(&context->radeon);
+	int id = 0;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	rrb = radeon_get_colorbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
+	}
+
+	r700SetRenderTarget(context, 0);
+
+	if (id > R700_MAX_RENDER_TARGETS)
+		return;
+
+	if (!r700->render_target[id].enabled)
+		return;
+
+        BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGSEQ(CB_COLOR0_BASE + (4 * id), 1);
+	R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_BASE.u32All);
+	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     rrb->bo,
+			     r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0);
+        END_BATCH();
+
+	if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
+	    (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
+		R600_OUT_BATCH((2 << id));
+		END_BATCH();
+	}
+
+        BEGIN_BATCH_NO_AUTOSTATE(18);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), r700->render_target[id].CB_COLOR0_SIZE.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), r700->render_target[id].CB_COLOR0_VIEW.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_TILE + (4 * id), r700->render_target[id].CB_COLOR0_TILE.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_FRAG + (4 * id), r700->render_target[id].CB_COLOR0_FRAG.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), r700->render_target[id].CB_COLOR0_MASK.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+}
+
+static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_bo * pbo;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context));
+
+	if (!pbo)
+		return;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1);
+	R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All);
+	R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All,
+			     pbo,
+			     r700->ps.SQ_PGM_START_PS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(9);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+}
+
+static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_bo * pbo;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
+
+	if (!pbo)
+		return;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1);
+	R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All);
+	R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All,
+			     pbo,
+			     r700->vs.SQ_PGM_START_VS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+}
+
+static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_bo * pbo;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	/* XXX fixme
+	 * R6xx chips require a FS be emitted, even if it's not used.
+	 * since we aren't using FS yet, just send the VS address to make
+	 * the kernel command checker happy
+	 */
+	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
+	r700->fs.SQ_PGM_START_FS.u32All = r700->vs.SQ_PGM_START_VS.u32All;
+	r700->fs.SQ_PGM_RESOURCES_FS.u32All = 0;
+	r700->fs.SQ_PGM_CF_OFFSET_FS.u32All = 0;
+	/* XXX */
+
+	if (!pbo)
+		return;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_FS, 1);
+	R600_OUT_BATCH(r700->fs.SQ_PGM_START_FS.u32All);
+	R600_OUT_BATCH_RELOC(r700->fs.SQ_PGM_START_FS.u32All,
+			     pbo,
+			     r700->fs.SQ_PGM_START_FS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_FS, r700->fs.SQ_PGM_RESOURCES_FS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_FS, r700->fs.SQ_PGM_CF_OFFSET_FS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+}
+
+static void r700SendViewportState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	int id = 0;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	if (id > R700_MAX_VIEWPORTS)
+		return;
+
+	if (!r700->viewport[id].enabled)
+		return;
+
+        BEGIN_BATCH_NO_AUTOSTATE(16);
+	R600_OUT_BATCH_REGSEQ(PA_SC_VPORT_SCISSOR_0_TL + (8 * id), 2);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_SC_VPORT_ZMIN_0 + (8 * id), 2);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_ZMIN_0.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_ZMAX_0.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_CL_VPORT_XSCALE_0 + (24 * id), 6);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_XSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_XOFFSET.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_YSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_YOFFSET.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_ZSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_ZOFFSET.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+}
+
+static void r700SendSQConfig(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+        BEGIN_BATCH_NO_AUTOSTATE(34);
+	R600_OUT_BATCH_REGSEQ(SQ_CONFIG, 6);
+	R600_OUT_BATCH(r700->sq_config.SQ_CONFIG.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All);
+
+	R600_OUT_BATCH_REGVAL(TA_CNTL_AUX, r700->TA_CNTL_AUX.u32All);
+	R600_OUT_BATCH_REGVAL(VC_ENHANCE, r700->VC_ENHANCE.u32All);
+	R600_OUT_BATCH_REGVAL(R7xx_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, r700->SQ_DYN_GPR_CNTL_PS_FLUSH_REQ.u32All);
+	R600_OUT_BATCH_REGVAL(DB_DEBUG, r700->DB_DEBUG.u32All);
+	R600_OUT_BATCH_REGVAL(DB_WATERMARKS, r700->DB_WATERMARKS.u32All);
+
+	R600_OUT_BATCH_REGSEQ(SQ_ESGS_RING_ITEMSIZE, 9);
+	R600_OUT_BATCH(r700->SQ_ESGS_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_GSVS_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_ESTMP_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_GSTMP_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_VSTMP_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_PSTMP_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_FBUF_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_REDUC_RING_ITEMSIZE.u32All);
+	R600_OUT_BATCH(r700->SQ_GS_VERT_ITEMSIZE.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+}
+
+static void r700SendUCPState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	int i;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	for (i = 0; i < R700_MAX_UCP; i++) {
+		if (r700->ucp[i].enabled) {
+			BEGIN_BATCH_NO_AUTOSTATE(6);
+			R600_OUT_BATCH_REGSEQ(PA_CL_UCP_0_X + (16 * i), 4);
+			R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_X.u32All);
+			R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_Y.u32All);
+			R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_Z.u32All);
+			R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_W.u32All);
+			END_BATCH();
+			COMMIT_BATCH();
+		}
+	}
+}
+
+static void r700SendSPIState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	unsigned int ui;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	BEGIN_BATCH_NO_AUTOSTATE(59 + R700_MAX_SHADER_EXPORTS);
+
+	R600_OUT_BATCH_REGSEQ(SQ_VTX_SEMANTIC_0, 32);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_0.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_1.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_2.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_3.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_4.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_5.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_6.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_7.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_8.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_9.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_10.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_11.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_12.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_13.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_14.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_15.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_16.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_17.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_18.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_19.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_20.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_21.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_22.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_23.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_24.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_25.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_26.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_27.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_28.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_29.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_30.u32All);
+	R600_OUT_BATCH(r700->SQ_VTX_SEMANTIC_31.u32All);
+
+	R600_OUT_BATCH_REGSEQ(SPI_VS_OUT_ID_0, 10);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_0.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_1.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_2.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_3.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_4.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_5.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_6.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_7.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_8.u32All);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_ID_9.u32All);
+
+	R600_OUT_BATCH_REGSEQ(SPI_VS_OUT_CONFIG, 9);
+	R600_OUT_BATCH(r700->SPI_VS_OUT_CONFIG.u32All);
+	R600_OUT_BATCH(r700->SPI_THREAD_GROUPING.u32All);
+	R600_OUT_BATCH(r700->SPI_PS_IN_CONTROL_0.u32All);
+	R600_OUT_BATCH(r700->SPI_PS_IN_CONTROL_1.u32All);
+	R600_OUT_BATCH(r700->SPI_INTERP_CONTROL_0.u32All);
+	R600_OUT_BATCH(r700->SPI_INPUT_Z.u32All);
+	R600_OUT_BATCH(r700->SPI_FOG_CNTL.u32All);
+	R600_OUT_BATCH(r700->SPI_FOG_FUNC_SCALE.u32All);
+	R600_OUT_BATCH(r700->SPI_FOG_FUNC_BIAS.u32All);
+
+	R600_OUT_BATCH_REGSEQ(SPI_PS_INPUT_CNTL_0, R700_MAX_SHADER_EXPORTS);
+	for(ui = 0; ui < R700_MAX_SHADER_EXPORTS; ui++)
+		R600_OUT_BATCH(r700->SPI_PS_INPUT_CNTL[ui].u32All);
+
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendVGTState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+        BEGIN_BATCH_NO_AUTOSTATE(41);
+
+	R600_OUT_BATCH_REGSEQ(VGT_MAX_VTX_INDX, 4);
+	R600_OUT_BATCH(r700->VGT_MAX_VTX_INDX.u32All);
+	R600_OUT_BATCH(r700->VGT_MIN_VTX_INDX.u32All);
+	R600_OUT_BATCH(r700->VGT_INDX_OFFSET.u32All);
+	R600_OUT_BATCH(r700->VGT_MULTI_PRIM_IB_RESET_INDX.u32All);
+
+	R600_OUT_BATCH_REGSEQ(VGT_OUTPUT_PATH_CNTL, 13);
+	R600_OUT_BATCH(r700->VGT_OUTPUT_PATH_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_HOS_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_HOS_MAX_TESS_LEVEL.u32All);
+	R600_OUT_BATCH(r700->VGT_HOS_MIN_TESS_LEVEL.u32All);
+	R600_OUT_BATCH(r700->VGT_HOS_REUSE_DEPTH.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_PRIM_TYPE.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_FIRST_DECR.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_DECR.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_VECT_0_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_VECT_1_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_VECT_0_FMT_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_GROUP_VECT_1_FMT_CNTL.u32All);
+	R600_OUT_BATCH(r700->VGT_GS_MODE.u32All);
+
+	R600_OUT_BATCH_REGVAL(VGT_PRIMITIVEID_EN, r700->VGT_PRIMITIVEID_EN.u32All);
+	R600_OUT_BATCH_REGVAL(VGT_MULTI_PRIM_IB_RESET_EN, r700->VGT_MULTI_PRIM_IB_RESET_EN.u32All);
+	R600_OUT_BATCH_REGVAL(VGT_INSTANCE_STEP_RATE_0, r700->VGT_INSTANCE_STEP_RATE_0.u32All);
+	R600_OUT_BATCH_REGVAL(VGT_INSTANCE_STEP_RATE_1, r700->VGT_INSTANCE_STEP_RATE_1.u32All);
+
+	R600_OUT_BATCH_REGSEQ(VGT_STRMOUT_EN, 3);
+	R600_OUT_BATCH(r700->VGT_STRMOUT_EN.u32All);
+	R600_OUT_BATCH(r700->VGT_REUSE_OFF.u32All);
+	R600_OUT_BATCH(r700->VGT_VTX_CNT_EN.u32All);
+
+	R600_OUT_BATCH_REGVAL(VGT_STRMOUT_BUFFER_EN, r700->VGT_STRMOUT_BUFFER_EN.u32All);
+
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendSXState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+        BEGIN_BATCH_NO_AUTOSTATE(9);
+	R600_OUT_BATCH_REGVAL(SX_MISC, r700->SX_MISC.u32All);
+	R600_OUT_BATCH_REGVAL(SX_ALPHA_TEST_CONTROL, r700->SX_ALPHA_TEST_CONTROL.u32All);
+	R600_OUT_BATCH_REGVAL(SX_ALPHA_REF, r700->SX_ALPHA_REF.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+        BEGIN_BATCH_NO_AUTOSTATE(23);
+	R600_OUT_BATCH_REGVAL(DB_HTILE_DATA_BASE, r700->DB_HTILE_DATA_BASE.u32All);
+
+	R600_OUT_BATCH_REGSEQ(DB_STENCIL_CLEAR, 2);
+	R600_OUT_BATCH(r700->DB_STENCIL_CLEAR.u32All);
+	R600_OUT_BATCH(r700->DB_DEPTH_CLEAR.u32All);
+
+	R600_OUT_BATCH_REGVAL(DB_DEPTH_CONTROL, r700->DB_DEPTH_CONTROL.u32All);
+	R600_OUT_BATCH_REGVAL(DB_SHADER_CONTROL, r700->DB_SHADER_CONTROL.u32All);
+
+	R600_OUT_BATCH_REGSEQ(DB_RENDER_CONTROL, 2);
+	R600_OUT_BATCH(r700->DB_RENDER_CONTROL.u32All);
+	R600_OUT_BATCH(r700->DB_RENDER_OVERRIDE.u32All);
+
+	R600_OUT_BATCH_REGVAL(DB_HTILE_SURFACE, r700->DB_HTILE_SURFACE.u32All);
+	R600_OUT_BATCH_REGVAL(DB_ALPHA_TO_MASK, r700->DB_ALPHA_TO_MASK.u32All);
+
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendStencilState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+        BEGIN_BATCH_NO_AUTOSTATE(4);
+	R600_OUT_BATCH_REGSEQ(DB_STENCILREFMASK, 2);
+	R600_OUT_BATCH(r700->DB_STENCILREFMASK.u32All);
+	R600_OUT_BATCH(r700->DB_STENCILREFMASK_BF.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendCBState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+		BEGIN_BATCH_NO_AUTOSTATE(11);
+		R600_OUT_BATCH_REGSEQ(CB_CLEAR_RED, 4);
+		R600_OUT_BATCH(r700->CB_CLEAR_RED_R6XX.u32All);
+		R600_OUT_BATCH(r700->CB_CLEAR_GREEN_R6XX.u32All);
+		R600_OUT_BATCH(r700->CB_CLEAR_BLUE_R6XX.u32All);
+		R600_OUT_BATCH(r700->CB_CLEAR_ALPHA_R6XX.u32All);
+		R600_OUT_BATCH_REGSEQ(CB_FOG_RED, 3);
+		R600_OUT_BATCH(r700->CB_FOG_RED_R6XX.u32All);
+		R600_OUT_BATCH(r700->CB_FOG_GREEN_R6XX.u32All);
+		R600_OUT_BATCH(r700->CB_FOG_BLUE_R6XX.u32All);
+		END_BATCH();
+	}
+
+	BEGIN_BATCH_NO_AUTOSTATE(7);
+	R600_OUT_BATCH_REGSEQ(CB_TARGET_MASK, 2);
+	R600_OUT_BATCH(r700->CB_TARGET_MASK.u32All);
+	R600_OUT_BATCH(r700->CB_SHADER_MASK.u32All);
+	R600_OUT_BATCH_REGVAL(R7xx_CB_SHADER_CONTROL, r700->CB_SHADER_CONTROL.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendCBCLRCMPState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGSEQ(CB_CLRCMP_CONTROL, 4);
+	R600_OUT_BATCH(r700->CB_CLRCMP_CONTROL.u32All);
+	R600_OUT_BATCH(r700->CB_CLRCMP_SRC.u32All);
+	R600_OUT_BATCH(r700->CB_CLRCMP_DST.u32All);
+	R600_OUT_BATCH(r700->CB_CLRCMP_MSK.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendCBBlendState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	unsigned int ui;
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+		BEGIN_BATCH_NO_AUTOSTATE(3);
+		R600_OUT_BATCH_REGVAL(CB_BLEND_CONTROL, r700->CB_BLEND_CONTROL.u32All);
+		END_BATCH();
+	}
+
+	BEGIN_BATCH_NO_AUTOSTATE(3);
+	R600_OUT_BATCH_REGVAL(CB_COLOR_CONTROL, r700->CB_COLOR_CONTROL.u32All);
+	END_BATCH();
+
+	if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) {
+		for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) {
+			if (r700->render_target[ui].enabled) {
+				BEGIN_BATCH_NO_AUTOSTATE(3);
+				R600_OUT_BATCH_REGVAL(CB_BLEND0_CONTROL + (4 * ui),
+						      r700->render_target[ui].CB_BLEND0_CONTROL.u32All);
+				END_BATCH();
+			}
+		}
+	}
+
+	COMMIT_BATCH();
+}
+
+static void r700SendCBBlendColorState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGSEQ(CB_BLEND_RED, 4);
+	R600_OUT_BATCH(r700->CB_BLEND_RED.u32All);
+	R600_OUT_BATCH(r700->CB_BLEND_GREEN.u32All);
+	R600_OUT_BATCH(r700->CB_BLEND_BLUE.u32All);
+	R600_OUT_BATCH(r700->CB_BLEND_ALPHA.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendSUState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(9);
+	R600_OUT_BATCH_REGVAL(PA_SU_SC_MODE_CNTL, r700->PA_SU_SC_MODE_CNTL.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_SU_POINT_SIZE, 4);
+	R600_OUT_BATCH(r700->PA_SU_POINT_SIZE.u32All);
+	R600_OUT_BATCH(r700->PA_SU_POINT_MINMAX.u32All);
+	R600_OUT_BATCH(r700->PA_SU_LINE_CNTL.u32All);
+	R600_OUT_BATCH(r700->PA_SU_VTX_CNTL.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+
+}
+
+static void r700SendPolyState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(10);
+	R600_OUT_BATCH_REGSEQ(PA_SU_POLY_OFFSET_DB_FMT_CNTL, 2);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_DB_FMT_CNTL.u32All);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_CLAMP.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_SU_POLY_OFFSET_FRONT_SCALE, 4);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_FRONT_SCALE.u32All);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.u32All);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_BACK_SCALE.u32All);
+	R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_BACK_OFFSET.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+
+}
+
+static void r700SendCLState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	BEGIN_BATCH_NO_AUTOSTATE(12);
+	R600_OUT_BATCH_REGVAL(PA_CL_CLIP_CNTL, r700->PA_CL_CLIP_CNTL.u32All);
+	R600_OUT_BATCH_REGVAL(PA_CL_VTE_CNTL, r700->PA_CL_VTE_CNTL.u32All);
+	R600_OUT_BATCH_REGVAL(PA_CL_VS_OUT_CNTL, r700->PA_CL_VS_OUT_CNTL.u32All);
+	R600_OUT_BATCH_REGVAL(PA_CL_NANINF_CNTL, r700->PA_CL_NANINF_CNTL.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendGBState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGSEQ(PA_CL_GB_VERT_CLIP_ADJ, 4);
+	R600_OUT_BATCH(r700->PA_CL_GB_VERT_CLIP_ADJ.u32All);
+	R600_OUT_BATCH(r700->PA_CL_GB_VERT_DISC_ADJ.u32All);
+	R600_OUT_BATCH(r700->PA_CL_GB_HORZ_CLIP_ADJ.u32All);
+	R600_OUT_BATCH(r700->PA_CL_GB_HORZ_DISC_ADJ.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendScissorState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	BEGIN_BATCH_NO_AUTOSTATE(22);
+	R600_OUT_BATCH_REGSEQ(PA_SC_SCREEN_SCISSOR_TL, 2);
+	R600_OUT_BATCH(r700->PA_SC_SCREEN_SCISSOR_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_SCREEN_SCISSOR_BR.u32All);
+
+	R600_OUT_BATCH_REGSEQ(PA_SC_WINDOW_OFFSET, 12);
+	R600_OUT_BATCH(r700->PA_SC_WINDOW_OFFSET.u32All);
+	R600_OUT_BATCH(r700->PA_SC_WINDOW_SCISSOR_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_WINDOW_SCISSOR_BR.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_RULE.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_0_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_0_BR.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_1_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_1_BR.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_2_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_2_BR.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_3_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_CLIPRECT_3_BR.u32All);
+
+	R600_OUT_BATCH_REGSEQ(PA_SC_GENERIC_SCISSOR_TL, 2);
+	R600_OUT_BATCH(r700->PA_SC_GENERIC_SCISSOR_TL.u32All);
+	R600_OUT_BATCH(r700->PA_SC_GENERIC_SCISSOR_BR.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	BEGIN_BATCH_NO_AUTOSTATE(15);
+	R600_OUT_BATCH_REGVAL(R7xx_PA_SC_EDGERULE, r700->PA_SC_EDGERULE.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_LINE_STIPPLE, r700->PA_SC_LINE_STIPPLE.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_MPASS_PS_CNTL, r700->PA_SC_MPASS_PS_CNTL.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_MODE_CNTL, r700->PA_SC_MODE_CNTL.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_LINE_CNTL, r700->PA_SC_LINE_CNTL.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendAAState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(12);
+	R600_OUT_BATCH_REGVAL(PA_SC_AA_CONFIG, r700->PA_SC_AA_CONFIG.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_AA_SAMPLE_LOCS_MCTX, r700->PA_SC_AA_SAMPLE_LOCS_MCTX.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX, r700->PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX.u32All);
+	R600_OUT_BATCH_REGVAL(PA_SC_AA_MASK, r700->PA_SC_AA_MASK.u32All);
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendPSConsts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	int i;
+	BATCH_LOCALS(&context->radeon);
+
+	if (r700->ps.num_consts == 0)
+		return;
+
+	BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->ps.num_consts * 4));
+	R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->ps.num_consts * 4)));
+	/* assembler map const from very beginning. */
+	R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4);
+	for (i = 0; i < r700->ps.num_consts; i++) {
+		R600_OUT_BATCH(r700->ps.consts[i][0].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][1].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][2].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][3].u32All);
+	}
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendVSConsts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	int i;
+	BATCH_LOCALS(&context->radeon);
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
+
+	if (r700->vs.num_consts == 0)
+		return;
+
+	BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->vs.num_consts * 4));
+	R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->vs.num_consts * 4)));
+	/* assembler map const from very beginning. */
+	R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4);
+	for (i = 0; i < r700->vs.num_consts; i++) {
+		R600_OUT_BATCH(r700->vs.consts[i][0].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][1].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][2].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][3].u32All);
+	}
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static int check_always(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	return atom->cmd_size;
+}
+
+static int check_cb(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	int count = 7;
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+		count += 11;
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+
+	return count;
+}
+
+static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned int ui;
+	int count = 3;
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+		count += 3;
+
+	if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) {
+		for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) {
+                        if (r700->render_target[ui].enabled)
+				count += 3;
+		}
+	}
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+
+	return count;
+}
+
+static int check_ucp(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int i;
+	int count = 0;
+
+	for (i = 0; i < R700_MAX_UCP; i++) {
+		if (r700->ucp[i].enabled)
+			count += 6;
+	}
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+	return count;
+}
+
+static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	int count = context->radeon.tcl.aos_count * 18;
+
+	if (count)
+		count += 6;
+
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+	return count;
+}
+
+static int check_tx(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	unsigned int i, count = 0;
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
+		radeonTexObj *t = r700->textures[i];
+		if (t)
+			count++;
+	}
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+	return count * 31;
+}
+
+static int check_ps_consts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int count = r700->ps.num_consts * 4;
+
+	if (count)
+		count += 2;
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+
+	return count;
+}
+
+static int check_vs_consts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int count = r700->vs.num_consts * 4;
+
+	if (count)
+		count += 2;
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
+
+	return count;
+}
+
+#define ALLOC_STATE( ATOM, CHK, SZ, EMIT )				\
+do {									\
+	context->atoms.ATOM.cmd_size = (SZ);				\
+	context->atoms.ATOM.cmd = NULL;					\
+	context->atoms.ATOM.name = #ATOM;				\
+	context->atoms.ATOM.idx = 0;					\
+	context->atoms.ATOM.check = check_##CHK;			\
+	context->atoms.ATOM.dirty = GL_FALSE;				\
+	context->atoms.ATOM.emit = (EMIT);				\
+	context->radeon.hw.max_state_size += (SZ);			\
+	insert_at_tail(&context->radeon.hw.atomlist, &context->atoms.ATOM); \
+} while (0)
+
+void r600InitAtoms(context_t *context)
+{
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s %p\n", __func__, context);
+	context->radeon.hw.max_state_size = 10 + 5 + 14; /* start 3d, idle, cb/db flush */
+
+	/* Setup the atom linked list */
+	make_empty_list(&context->radeon.hw.atomlist);
+	context->radeon.hw.atomlist.name = "atom-list";
+
+	ALLOC_STATE(sq, always, 34, r700SendSQConfig);
+	ALLOC_STATE(db, always, 23, r700SendDBState);
+	ALLOC_STATE(stencil, always, 4, r700SendStencilState);
+	ALLOC_STATE(db_target, always, 12, r700SendDepthTargetState);
+	ALLOC_STATE(sc, always, 15, r700SendSCState);
+	ALLOC_STATE(scissor, always, 22, r700SendScissorState);
+	ALLOC_STATE(aa, always, 12, r700SendAAState);
+	ALLOC_STATE(cl, always, 12, r700SendCLState);
+	ALLOC_STATE(gb, always, 6, r700SendGBState);
+	ALLOC_STATE(ucp, ucp, (R700_MAX_UCP * 6), r700SendUCPState);
+	ALLOC_STATE(su, always, 9, r700SendSUState);
+	ALLOC_STATE(poly, always, 10, r700SendPolyState);
+	ALLOC_STATE(cb, cb, 18, r700SendCBState);
+	ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState);
+	ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState);
+	ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState);
+	ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState);
+	ALLOC_STATE(sx, always, 9, r700SendSXState);
+	ALLOC_STATE(vgt, always, 41, r700SendVGTState);
+	ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState);
+	ALLOC_STATE(vpt, always, 16, r700SendViewportState);
+	ALLOC_STATE(fs, always, 18, r700SendFSState);
+	ALLOC_STATE(vs, always, 18, r700SendVSState);
+	ALLOC_STATE(ps, always, 21, r700SendPSState);
+	ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts);
+	ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts);
+	ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState);
+	ALLOC_STATE(tx, tx, (R700_TEXTURE_NUMBERUNITS * 20), r700SendTexState);
+	ALLOC_STATE(tx_smplr, tx, (R700_TEXTURE_NUMBERUNITS * 5), r700SendTexSamplerState);
+	ALLOC_STATE(tx_brdr_clr, tx, (R700_TEXTURE_NUMBERUNITS * 6), r700SendTexBorderColorState);
+
+	context->radeon.hw.is_dirty = GL_TRUE;
+	context->radeon.hw.all_dirty = GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h
new file mode 100644
index 0000000000..ae249e15fd
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chip.h
@@ -0,0 +1,503 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_CHIP_H_
+#define _R700_CHIP_H_
+
+#include "r600_context.h"
+
+#include "r600_reg.h"
+#include "r600_reg_auto_r6xx.h"
+#include "r600_reg_r6xx.h"
+#include "r600_reg_r7xx.h"
+
+#include "r700_chipoffset.h"
+
+#define SETfield(x, val, shift, mask)  ( (x) = ((x) & ~(mask)) | ((val) << (shift)) ) /* u32All */
+#define CLEARfield(x, mask)            ( (x) &= ~(mask) )
+#define SETbit(x, bit)                 ( (x) |= (bit) )
+#define CLEARbit(x, bit)               ( (x) &= ~(bit) )
+
+#define R700_TEXTURE_NUMBERUNITS 16
+#define R700_MAX_RENDER_TARGETS  8
+#define R700_MAX_VIEWPORTS       16
+#define R700_MAX_SHADER_EXPORTS  32
+#define R700_MAX_UCP             6
+#define R700_MAX_DX9_CONSTS      256
+
+/* Enum not show in r600_*.h */
+
+#define FETCH_RESOURCE_STRIDE 7
+
+#define ASIC_CONFIG_BASE_INDEX    0x2000
+#define ASIC_CONTEXT_BASE_INDEX   0xA000
+#define ASIC_CTL_CONST_BASE_INDEX 0xF3FC
+
+
+enum 
+{
+    SQ_ABSOLUTE                              = 0x00000000,
+    SQ_RELATIVE                              = 0x00000001,
+};
+
+enum 
+{
+    SQ_ALU_SCL_210                           = 0x00000000,
+    SQ_ALU_SCL_122                           = 0x00000001,
+    SQ_ALU_SCL_212                           = 0x00000002,
+    SQ_ALU_SCL_221                           = 0x00000003,
+};
+
+enum 
+{
+    SQ_TEX_UNNORMALIZED                      = 0x00000000,
+    SQ_TEX_NORMALIZED                        = 0x00000001,
+};
+
+enum 
+{
+    SQ_CF_PIXEL_MRT0                         = 0x00000000,
+    SQ_CF_PIXEL_MRT1                         = 0x00000001,
+    SQ_CF_PIXEL_MRT2                         = 0x00000002,
+    SQ_CF_PIXEL_MRT3                         = 0x00000003,
+    SQ_CF_PIXEL_MRT4                         = 0x00000004,
+    SQ_CF_PIXEL_MRT5                         = 0x00000005,
+    SQ_CF_PIXEL_MRT6                         = 0x00000006,
+    SQ_CF_PIXEL_MRT7                         = 0x00000007,
+    SQ_CF_PIXEL_Z                            = 0x0000003d,
+};
+
+typedef enum ENUM_SQ_CF_ARRAY_BASE_POS {
+SQ_CF_POS_0                              = 0x0000003c,
+SQ_CF_POS_1                              = 0x0000003d,
+SQ_CF_POS_2                              = 0x0000003e,
+SQ_CF_POS_3                              = 0x0000003f,
+} ENUM_SQ_CF_ARRAY_BASE_POS;
+
+enum
+{
+    PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit = 23,
+};
+
+enum 
+{
+    TEX_XYFilter_Point                       = 0x00000000,
+    TEX_XYFilter_Linear                      = 0x00000001,
+    TEX_XYFilter_Cubic                       = 0x00000002,
+    TEX_XYFilter_Cleartype                   = 0x00000003,
+
+    TEX_MipFilter_None                       = 0x00000000,
+    TEX_MipFilter_Point                      = 0x00000001,
+    TEX_MipFilter_Linear                     = 0x00000002,
+};
+
+enum 
+{
+    SQ_EXPORT_WRITE                          = 0x00000000,
+    SQ_EXPORT_WRITE_IND                      = 0x00000001,
+    SQ_EXPORT_WRITE_ACK                      = 0x00000002,
+    SQ_EXPORT_WRITE_IND_ACK                  = 0x00000003,
+};
+
+/* --------------------------------- */
+
+enum
+{
+    R700_PM4_PACKET0_NOP = 0x00000000,
+    R700_PM4_PACKET1_NOP = 0x40000000,
+    R700_PM4_PACKET2_NOP = 0x80000000,
+    R700_PM4_PACKET3_NOP = 0xC0000000,
+};
+
+#define  PM4_OPCODE_SET_INDEX_TYPE      (R700_PM4_PACKET3_NOP | (IT_INDEX_TYPE << 8))
+
+#define  PM4_OPCODE_DRAW_INDEX_AUTO     (R700_PM4_PACKET3_NOP | (IT_DRAW_INDEX_AUTO << 8))
+#define  PM4_OPCODE_DRAW_INDEX_IMMD     (R700_PM4_PACKET3_NOP | (IT_DRAW_INDEX_IMMD << 8))
+#define  PM4_OPCODE_WAIT_REG_MEM        (R700_PM4_PACKET3_NOP | (IT_WAIT_REG_MEM << 8))
+#define  PM4_OPCODE_SET_CONTEXT_REG     (R700_PM4_PACKET3_NOP | (IT_SET_CONTEXT_REG << 8))
+#define  PM4_OPCODE_SET_CONFIG_REG      (R700_PM4_PACKET3_NOP | (IT_SET_CONFIG_REG << 8))
+#define  PM4_OPCODE_SET_ALU_CONST       (R700_PM4_PACKET3_NOP | (IT_SET_ALU_CONST << 8))
+#define  PM4_OPCODE_SET_RESOURCE        (R700_PM4_PACKET3_NOP | (IT_SET_RESOURCE << 8))
+#define  PM4_OPCODE_SET_SAMPLER         (R700_PM4_PACKET3_NOP | (IT_SET_SAMPLER << 8))
+#define  PM4_OPCODE_CONTEXT_CONTROL     (R700_PM4_PACKET3_NOP | (IT_CONTEXT_CONTROL << 8))
+
+union UINT_FLOAT 
+{
+    unsigned int u32All;
+    float	f32All;
+};
+
+#if 0
+typedef struct _TEXTURE_STATE_STRUCT
+{
+    union UINT_FLOAT     SQ_TEX_RESOURCE0;
+    union UINT_FLOAT     SQ_TEX_RESOURCE1;
+    union UINT_FLOAT     SQ_TEX_RESOURCE2;
+    union UINT_FLOAT     SQ_TEX_RESOURCE3;
+    union UINT_FLOAT     SQ_TEX_RESOURCE4;
+    union UINT_FLOAT     SQ_TEX_RESOURCE5;
+    union UINT_FLOAT     SQ_TEX_RESOURCE6;
+    GLboolean                         enabled;
+} TEXTURE_STATE_STRUCT;
+
+typedef struct _SAMPLER_STATE_STRUCT
+{
+    union UINT_FLOAT      SQ_TEX_SAMPLER0;
+    union UINT_FLOAT      SQ_TEX_SAMPLER1;
+    union UINT_FLOAT      SQ_TEX_SAMPLER2;
+    GLboolean                         enabled;
+} SAMPLER_STATE_STRUCT;
+
+typedef struct _R700_TEXTURE_STATES
+{
+    TEXTURE_STATE_STRUCT *textures[R700_TEXTURE_NUMBERUNITS];
+    SAMPLER_STATE_STRUCT *samplers[R700_TEXTURE_NUMBERUNITS];
+} R700_TEXTURE_STATES;
+#endif
+
+typedef struct _RENDER_TARGET_STATE_STRUCT
+{
+	union UINT_FLOAT            	CB_COLOR0_BASE;  /* 0xA010 */
+	union UINT_FLOAT            	CB_COLOR0_SIZE;  /* 0xA018 */
+	union UINT_FLOAT            	CB_COLOR0_VIEW;  /* 0xA020 */
+	union UINT_FLOAT            	CB_COLOR0_INFO;  /* 0xA028 */
+	union UINT_FLOAT            	CB_COLOR0_TILE;  /* 0xA030 */
+	union UINT_FLOAT            	CB_COLOR0_FRAG;  /* 0xA038 */
+	union UINT_FLOAT            	CB_COLOR0_MASK;  /* 0xA040 */
+	union UINT_FLOAT         	CB_BLEND0_CONTROL;  /* 0xA1E0 */
+	GLboolean                         enabled;
+	GLboolean                         dirty;
+} RENDER_TARGET_STATE_STRUCT;
+
+typedef struct _VIEWPORT_STATE_STRUCT
+{
+	union UINT_FLOAT  	PA_SC_VPORT_SCISSOR_0_TL;  /* 0xA094 */
+	union UINT_FLOAT  	PA_SC_VPORT_SCISSOR_0_BR;  /* 0xA095 */
+	union UINT_FLOAT        PA_SC_VPORT_ZMIN_0;        /* 0xA0B4 */
+	union UINT_FLOAT        PA_SC_VPORT_ZMAX_0;        /* 0xA0B5 */
+	union UINT_FLOAT        PA_CL_VPORT_XSCALE;        /* 0xA10F */
+	union UINT_FLOAT       	PA_CL_VPORT_XOFFSET;       /* 0xA110 */
+	union UINT_FLOAT        PA_CL_VPORT_YSCALE;        /* 0xA111 */
+	union UINT_FLOAT       	PA_CL_VPORT_YOFFSET;       /* 0xA112 */
+	union UINT_FLOAT        PA_CL_VPORT_ZSCALE;        /* 0xA113 */
+	union UINT_FLOAT       	PA_CL_VPORT_ZOFFSET;       /* 0xA114 */
+	GLboolean                         enabled;
+	GLboolean                         dirty;
+} VIEWPORT_STATE_STRUCT;
+
+typedef struct _UCP_STATE_STRUCT
+{
+	union UINT_FLOAT        PA_CL_UCP_0_X;
+	union UINT_FLOAT        PA_CL_UCP_0_Y;
+	union UINT_FLOAT        PA_CL_UCP_0_Z;
+	union UINT_FLOAT        PA_CL_UCP_0_W;
+	GLboolean                         enabled;
+	GLboolean                         dirty;
+} UCP_STATE_STRUCT;
+
+typedef struct _PS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_PS           ;  /* 0xA210 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_PS       ;  /* 0xA214 */
+	union UINT_FLOAT         	SQ_PGM_EXPORTS_PS         ;  /* 0xA215 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_PS       ;  /* 0xA233 */
+	GLboolean                         dirty;
+	int                             num_consts;
+	union UINT_FLOAT                consts[R700_MAX_DX9_CONSTS][4];
+} PS_STATE_STRUCT;
+
+typedef struct _VS_STATE_STRUCT
+{
+ 	union UINT_FLOAT           	SQ_PGM_START_VS           ;  /* 0xA216 */
+	union UINT_FLOAT  		SQ_PGM_RESOURCES_VS       ;  /* 0xA21A */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_VS       ;  /* 0xA234 */
+	GLboolean                         dirty;
+	int                             num_consts;
+	union UINT_FLOAT                consts[R700_MAX_DX9_CONSTS][4];
+} VS_STATE_STRUCT;
+
+typedef struct _GS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_GS           ;  /* 0xA21B */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_GS       ;  /* 0xA21F */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_GS       ;  /* 0xA235 */
+	GLboolean                         dirty;
+} GS_STATE_STRUCT;
+
+typedef struct _ES_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_ES           ;  /* 0xA220 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_ES       ;  /* 0xA224 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_ES       ;  /* 0xA236 */
+	GLboolean                         dirty;
+} ES_STATE_STRUCT;
+
+typedef struct _FS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_FS           ;  /* 0xA225 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_FS       ;  /* 0xA229 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_FS       ;  /* 0xA237 */
+	GLboolean                         dirty;
+} FS_STATE_STRUCT;
+
+typedef struct _SQ_CONFIG_STRUCT
+{
+	union UINT_FLOAT     	        SQ_CONFIG                 ;  /* 0x2300 */
+	union UINT_FLOAT     	        SQ_GPR_RESOURCE_MGMT_1    ;  /* 0x2301 */
+	union UINT_FLOAT     	        SQ_GPR_RESOURCE_MGMT_2    ;  /* 0x2302 */
+	union UINT_FLOAT     	        SQ_THREAD_RESOURCE_MGMT   ;  /* 0x2303 */
+	union UINT_FLOAT     	        SQ_STACK_RESOURCE_MGMT_1  ;  /* 0x2304 */
+	union UINT_FLOAT     	        SQ_STACK_RESOURCE_MGMT_2  ;  /* 0x2305 */
+} SQ_CONFIG_STRUCT;
+
+typedef struct _R700_CHIP_CONTEXT
+{
+	// DB
+	union UINT_FLOAT             	DB_DEPTH_SIZE             ;  /* 0xA000 */
+	union UINT_FLOAT             	DB_DEPTH_VIEW             ;  /* 0xA001 */
+	union UINT_FLOAT             	DB_DEPTH_BASE             ;  /* 0xA003 */
+	union UINT_FLOAT             	DB_DEPTH_INFO             ;  /* 0xA004 */
+	GLboolean                       db_target_dirty;
+	union UINT_FLOAT                DB_HTILE_DATA_BASE        ;  /* 0xA005 */
+	union UINT_FLOAT          	DB_STENCIL_CLEAR          ;  /* 0xA00A */
+	union UINT_FLOAT            	DB_DEPTH_CLEAR            ;  /* 0xA00B */
+	union UINT_FLOAT            	DB_STENCILREFMASK         ;  /* 0xA10C */
+	union UINT_FLOAT            	DB_STENCILREFMASK_BF      ;  /* 0xA10D */
+	union UINT_FLOAT         	DB_RENDER_CONTROL         ;  /* 0xA343 */
+	union UINT_FLOAT        	DB_RENDER_OVERRIDE        ;  /* 0xA344 */
+	union UINT_FLOAT          	DB_HTILE_SURFACE          ;  /* 0xA349 */
+	union UINT_FLOAT          	DB_ALPHA_TO_MASK          ;  /* 0xA351 */
+	union UINT_FLOAT          	DB_DEPTH_CONTROL          ;  /* 0xA200 */
+	union UINT_FLOAT         	DB_SHADER_CONTROL         ;  /* 0xA203 */
+	GLboolean                       db_dirty;
+
+	// SC
+	union UINT_FLOAT   	        PA_SC_SCREEN_SCISSOR_TL   ;  /* 0xA00C */
+	union UINT_FLOAT   	        PA_SC_SCREEN_SCISSOR_BR   ;  /* 0xA00D */
+	union UINT_FLOAT       	        PA_SC_WINDOW_OFFSET       ;  /* 0xA080 */
+	union UINT_FLOAT   	        PA_SC_WINDOW_SCISSOR_TL   ;  /* 0xA081 */
+	union UINT_FLOAT   	        PA_SC_WINDOW_SCISSOR_BR   ;  /* 0xA082 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_RULE       ;  /* 0xA083 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_0_TL       ;  /* 0xA084 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_0_BR       ;  /* 0xA085 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_1_TL       ;  /* 0xA086 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_1_BR       ;  /* 0xA087 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_2_TL       ;  /* 0xA088 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_2_BR       ;  /* 0xA089 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_3_TL       ;  /* 0xA08A */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_3_BR       ;  /* 0xA08B */
+	union UINT_FLOAT            	PA_SC_EDGERULE            ;  /* 0xA08C */
+	union UINT_FLOAT  	        PA_SC_GENERIC_SCISSOR_TL  ;  /* 0xA090 */
+	union UINT_FLOAT  	        PA_SC_GENERIC_SCISSOR_BR  ;  /* 0xA091 */
+	GLboolean                       scissor_dirty;
+
+	union UINT_FLOAT        	PA_SC_LINE_STIPPLE        ;  /* 0xA283 */
+	union UINT_FLOAT           	PA_SC_LINE_CNTL           ;  /* 0xA300 */
+	union UINT_FLOAT           	PA_SC_AA_CONFIG           ;  /* 0xA301 */
+	union UINT_FLOAT       	        PA_SC_MPASS_PS_CNTL       ;  /* 0xA292 */
+	union UINT_FLOAT           	PA_SC_MODE_CNTL           ;  /* 0xA293 */
+	union UINT_FLOAT 	        PA_SC_AA_SAMPLE_LOCS_MCTX ;  /* 0xA307 */
+	union UINT_FLOAT                PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX; /* 0xA308 */
+	union UINT_FLOAT             	PA_SC_AA_MASK             ;  /* 0xA312 */
+	GLboolean                       sc_dirty;
+
+	// CL
+	union UINT_FLOAT           	PA_CL_CLIP_CNTL           ;  /* 0xA204 */
+	union UINT_FLOAT            	PA_CL_VTE_CNTL            ;  /* 0xA206 */
+	union UINT_FLOAT         	PA_CL_VS_OUT_CNTL         ;  /* 0xA207 */
+	union UINT_FLOAT         	PA_CL_NANINF_CNTL         ;  /* 0xA208 */
+	union UINT_FLOAT    	        PA_CL_GB_VERT_CLIP_ADJ    ;  /* 0xA303 */
+	union UINT_FLOAT    	        PA_CL_GB_VERT_DISC_ADJ    ;  /* 0xA304 */
+	union UINT_FLOAT    	        PA_CL_GB_HORZ_CLIP_ADJ    ;  /* 0xA305 */
+	union UINT_FLOAT    	        PA_CL_GB_HORZ_DISC_ADJ    ;  /* 0xA306 */
+	GLboolean                       cl_dirty;
+
+	// SU
+	union UINT_FLOAT        	PA_SU_SC_MODE_CNTL        ;  /* 0xA205 */
+	union UINT_FLOAT          	PA_SU_POINT_SIZE          ;  /* 0xA280 */
+	union UINT_FLOAT        	PA_SU_POINT_MINMAX        ;  /* 0xA281 */
+	union UINT_FLOAT           	PA_SU_LINE_CNTL           ;  /* 0xA282 */
+	union UINT_FLOAT            	PA_SU_VTX_CNTL            ;  /* 0xA302 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_DB_FMT_CNTL;   /* 0xA37E */
+	union UINT_FLOAT   	        PA_SU_POLY_OFFSET_CLAMP   ;      /* 0xA37F */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_FRONT_SCALE;   /* 0xA380 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_FRONT_OFFSET; /* 0xA381 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_BACK_SCALE;    /* 0xA382 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_BACK_OFFSET;   /* 0xA383 */
+	GLboolean                       su_dirty;
+
+	VIEWPORT_STATE_STRUCT           viewport[R700_MAX_VIEWPORTS];
+	UCP_STATE_STRUCT                ucp[R700_MAX_UCP];
+
+	// CB
+	union UINT_FLOAT              	CB_CLEAR_RED_R6XX         ;  /* 0xA048 */
+	union UINT_FLOAT            	CB_CLEAR_GREEN_R6XX       ;  /* 0xA049 */
+	union UINT_FLOAT             	CB_CLEAR_BLUE_R6XX        ;  /* 0xA04A */
+	union UINT_FLOAT            	CB_CLEAR_ALPHA_R6XX       ;  /* 0xA04B */
+	union UINT_FLOAT            	CB_TARGET_MASK            ;  /* 0xA08E */
+	union UINT_FLOAT            	CB_SHADER_MASK            ;  /* 0xA08F */
+	union UINT_FLOAT              	CB_BLEND_RED              ;  /* 0xA105 */
+	union UINT_FLOAT            	CB_BLEND_GREEN            ;  /* 0xA106 */
+	union UINT_FLOAT             	CB_BLEND_BLUE             ;  /* 0xA107 */
+	union UINT_FLOAT            	CB_BLEND_ALPHA            ;  /* 0xA108 */
+	union UINT_FLOAT              	CB_FOG_RED_R6XX           ;  /* 0xA109 */
+	union UINT_FLOAT            	CB_FOG_GREEN_R6XX         ;  /* 0xA10A */
+	union UINT_FLOAT             	CB_FOG_BLUE_R6XX          ;  /* 0xA10B */
+	union UINT_FLOAT         	CB_SHADER_CONTROL         ;  /* 0xA1E8 */
+	union UINT_FLOAT          	CB_COLOR_CONTROL          ;  /* 0xA202 */
+	union UINT_FLOAT         	CB_CLRCMP_CONTROL         ;  /* 0xA30C */
+	union UINT_FLOAT             	CB_CLRCMP_SRC             ;  /* 0xA30D */
+	union UINT_FLOAT             	CB_CLRCMP_DST             ;  /* 0xA30E */
+	union UINT_FLOAT             	CB_CLRCMP_MSK             ;  /* 0xA30F */
+	union UINT_FLOAT             	CB_BLEND_CONTROL          ;  /* 0xABD0 */
+	GLboolean                       cb_dirty;
+	RENDER_TARGET_STATE_STRUCT      render_target[R700_MAX_RENDER_TARGETS];
+
+	// SX
+	union UINT_FLOAT                SX_MISC                   ;  /* 0xA0D4 */
+	union UINT_FLOAT     	        SX_ALPHA_TEST_CONTROL     ;  /* 0xA104 */
+	union UINT_FLOAT     	        SX_ALPHA_REF              ;  /* 0xA10E */
+	GLboolean                       sx_dirty;
+
+	// VGT
+	union UINT_FLOAT          	VGT_MAX_VTX_INDX          ;  /* 0xA100 */
+	union UINT_FLOAT          	VGT_MIN_VTX_INDX          ;  /* 0xA101 */
+	union UINT_FLOAT           	VGT_INDX_OFFSET           ;  /* 0xA102 */
+	union UINT_FLOAT                VGT_MULTI_PRIM_IB_RESET_INDX;  /* 0xA103 */
+	union UINT_FLOAT      	        VGT_OUTPUT_PATH_CNTL      ;  /* 0xA284 */
+	union UINT_FLOAT      	        VGT_HOS_CNTL              ;  /* 0xA285 */
+	union UINT_FLOAT      	        VGT_HOS_MAX_TESS_LEVEL    ;  /* 0xA286 */
+	union UINT_FLOAT      	        VGT_HOS_MIN_TESS_LEVEL    ;  /* 0xA287 */
+	union UINT_FLOAT      	        VGT_HOS_REUSE_DEPTH       ;  /* 0xA288 */
+	union UINT_FLOAT      	        VGT_GROUP_PRIM_TYPE       ;  /* 0xA289 */
+	union UINT_FLOAT      	        VGT_GROUP_FIRST_DECR      ;  /* 0xA28A */
+	union UINT_FLOAT      	        VGT_GROUP_DECR            ;  /* 0xA28B */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_0_CNTL     ;  /* 0xA28C */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_1_CNTL     ;  /* 0xA28D */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_0_FMT_CNTL ;  /* 0xA28E */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_1_FMT_CNTL ;  /* 0xA28F */
+	union UINT_FLOAT               	VGT_GS_MODE               ;  /* 0xA290 */
+	union UINT_FLOAT        	VGT_PRIMITIVEID_EN        ;  /* 0xA2A1 */
+	union UINT_FLOAT	        VGT_MULTI_PRIM_IB_RESET_EN;  /* 0xA2A5 */
+	union UINT_FLOAT  	        VGT_INSTANCE_STEP_RATE_0  ;  /* 0xA2A8 */
+	union UINT_FLOAT  	        VGT_INSTANCE_STEP_RATE_1  ;  /* 0xA2A9 */
+	union UINT_FLOAT            	VGT_STRMOUT_EN            ;  /* 0xA2AC */
+	union UINT_FLOAT             	VGT_REUSE_OFF             ;  /* 0xA2AD */
+	union UINT_FLOAT             	VGT_VTX_CNT_EN            ;  /* 0xA2AE */
+	union UINT_FLOAT            	VGT_STRMOUT_BUFFER_EN     ;  /* 0xA2C8 */
+	GLboolean                       vgt_dirty;
+
+	// SPI
+	union UINT_FLOAT           	SPI_VS_OUT_ID_0           ;  /* 0xA185 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_1           ;  /* 0xA186 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_2           ;  /* 0xA187 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_3           ;  /* 0xA188 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_4           ;  /* 0xA189 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_5           ;  /* 0xA18A */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_6           ;  /* 0xA18B */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_7           ;  /* 0xA18C */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_8           ;  /* 0xA18D */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_9           ;  /* 0xA18E */
+	union UINT_FLOAT                SPI_VS_OUT_CONFIG         ;  /* 0xA1B1 */
+	union UINT_FLOAT       	        SPI_THREAD_GROUPING       ;  /* 0xA1B2 */
+	union UINT_FLOAT       	        SPI_PS_IN_CONTROL_0       ;  /* 0xA1B3 */
+	union UINT_FLOAT       	        SPI_PS_IN_CONTROL_1       ;  /* 0xA1B4 */
+	union UINT_FLOAT       	        SPI_INTERP_CONTROL_0      ;  /* 0xA1B5 */
+ 	union UINT_FLOAT               	SPI_INPUT_Z               ;  /* 0xA1B6 */
+	union UINT_FLOAT              	SPI_FOG_CNTL              ;  /* 0xA1B7 */
+	union UINT_FLOAT              	SPI_FOG_FUNC_SCALE        ;  /* 0xA1B8 */
+	union UINT_FLOAT              	SPI_FOG_FUNC_BIAS         ;  /* 0xA1B9 */
+
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_0         ;  /* 0xA0E0 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_1         ;  /* 0xA0E1 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_2         ;  /* 0xA0E2 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_3         ;  /* 0xA0E3 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_4         ;  /* 0xA0E4 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_5         ;  /* 0xA0E5 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_6         ;  /* 0xA0E6 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_7         ;  /* 0xA0E7 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_8         ;  /* 0xA0E8 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_9         ;  /* 0xA0E9 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_10        ;  /* 0xA0EA */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_11        ;  /* 0xA0EB */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_12        ;  /* 0xA0EC */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_13        ;  /* 0xA0ED */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_14        ;  /* 0xA0EE */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_15        ;  /* 0xA0EF */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_16        ;  /* 0xA0F0 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_17        ;  /* 0xA0F1 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_18        ;  /* 0xA0F2 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_19        ;  /* 0xA0F3 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_20        ;  /* 0xA0F4 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_21        ;  /* 0xA0F5 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_22        ;  /* 0xA0F6 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_23        ;  /* 0xA0F7 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_24        ;  /* 0xA0F8 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_25        ;  /* 0xA0F9 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_26        ;  /* 0xA0FA */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_27        ;  /* 0xA0FB */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_28        ;  /* 0xA0FC */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_29        ;  /* 0xA0FD */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_30        ;  /* 0xA0FE */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_31        ;  /* 0xA0FF */
+	union UINT_FLOAT       	        SPI_PS_INPUT_CNTL[R700_MAX_SHADER_EXPORTS];
+	GLboolean                       spi_dirty;
+
+	// shaders
+	PS_STATE_STRUCT                 ps;
+	VS_STATE_STRUCT                 vs;
+	GS_STATE_STRUCT                 gs;
+	ES_STATE_STRUCT                 es;
+	FS_STATE_STRUCT                 fs;
+
+	// SQ CONFIG
+	SQ_CONFIG_STRUCT                sq_config;
+	// misc
+	union UINT_FLOAT             	TA_CNTL_AUX               ;  /* 0x2542 */
+	union UINT_FLOAT             	VC_ENHANCE                ;  /* 0x25C5 */
+	union UINT_FLOAT             	SQ_DYN_GPR_CNTL_PS_FLUSH_REQ;  /* 0x2363 */
+	union UINT_FLOAT             	DB_DEBUG                  ;  /* 0x260C */
+	union UINT_FLOAT             	DB_WATERMARKS             ;  /* 0x260E */
+	// SQ
+	union UINT_FLOAT     	        SQ_ESGS_RING_ITEMSIZE     ;  /* 0xA22A */
+	union UINT_FLOAT     	        SQ_GSVS_RING_ITEMSIZE     ;  /* 0xA22B */
+	union UINT_FLOAT    	        SQ_ESTMP_RING_ITEMSIZE    ;  /* 0xA22C */
+	union UINT_FLOAT    	        SQ_GSTMP_RING_ITEMSIZE    ;  /* 0xA22D */
+	union UINT_FLOAT    	        SQ_VSTMP_RING_ITEMSIZE    ;  /* 0xA22E */
+	union UINT_FLOAT    	        SQ_PSTMP_RING_ITEMSIZE    ;  /* 0xA22F */
+	union UINT_FLOAT     	        SQ_FBUF_RING_ITEMSIZE     ;  /* 0xA230 */
+	union UINT_FLOAT    	        SQ_REDUC_RING_ITEMSIZE    ;  /* 0xA231 */
+	union UINT_FLOAT       	        SQ_GS_VERT_ITEMSIZE       ;  /* 0xA232 */
+	GLboolean                       sq_dirty;
+
+	radeonTexObj*                   textures[R700_TEXTURE_NUMBERUNITS];
+
+	GLboolean                       bEnablePerspective;
+
+} R700_CHIP_CONTEXT;
+
+#endif /* _R700_CHIP_H_ */
+
diff --git a/src/mesa/drivers/dri/r600/r700_chipoffset.h b/src/mesa/drivers/dri/r600/r700_chipoffset.h
new file mode 100644
index 0000000000..4d73fb99a7
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chipoffset.h
@@ -0,0 +1,693 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_CHIPOFFSET_H_
+#define _R700_CHIPOFFSET_H_
+
+#define mmWAIT_UNTIL                                    0x2010
+#define mmSCRATCH_REG0                                  0x2140
+#define mmGUI_SCRATCH_REG0                              0x2140
+#define mmSCRATCH_REG1                                  0x2141
+#define mmGUI_SCRATCH_REG1                              0x2141
+#define mmSCRATCH_REG2                                  0x2142
+#define mmGUI_SCRATCH_REG2                              0x2142
+#define mmSCRATCH_REG3                                  0x2143
+#define mmGUI_SCRATCH_REG3                              0x2143
+#define mmSCRATCH_REG4                                  0x2144
+#define mmGUI_SCRATCH_REG4                              0x2144
+#define mmSCRATCH_REG5                                  0x2145
+#define mmGUI_SCRATCH_REG5                              0x2145
+#define mmSCRATCH_REG6                                  0x2146
+#define mmGUI_SCRATCH_REG6                              0x2146
+#define mmSCRATCH_REG7                                  0x2147
+#define mmGUI_SCRATCH_REG7                              0x2147
+
+#define mmCP_COHER_CNTL                                 0x217C
+#define mmCP_COHER_SIZE                                 0x217D
+#define mmCP_COHER_BASE                                 0x217E
+#define mmCP_COHER_STATUS                               0x217F
+
+#define mmTA_CNTL_AUX                                   0x2542
+#define mmVC_ENHANCE                                    0x25C5
+#define mmSQ_DYN_GPR_CNTL_PS_FLUSH_REQ                  0x2363
+#define mmDB_DEBUG                                      0x260C
+#define mmDB_WATERMARKS                                 0x260E
+
+#define mmPA_CL_VPORT_XSCALE                            0xA10F
+#define mmPA_CL_VPORT_XOFFSET                           0xA110
+#define mmPA_CL_VPORT_YSCALE                            0xA111
+#define mmPA_CL_VPORT_YOFFSET                           0xA112
+#define mmPA_CL_VPORT_ZSCALE                            0xA113
+#define mmPA_CL_VPORT_ZOFFSET                           0xA114
+#define mmPA_CL_VPORT_XSCALE_1                          0xA115
+#define mmPA_CL_VPORT_XSCALE_2                          0xA11B
+#define mmPA_CL_VPORT_XSCALE_3                          0xA121
+#define mmPA_CL_VPORT_XSCALE_4                          0xA127
+#define mmPA_CL_VPORT_XSCALE_5                          0xA12D
+#define mmPA_CL_VPORT_XSCALE_6                          0xA133
+#define mmPA_CL_VPORT_XSCALE_7                          0xA139
+#define mmPA_CL_VPORT_XSCALE_8                          0xA13F
+#define mmPA_CL_VPORT_XSCALE_9                          0xA145
+#define mmPA_CL_VPORT_XSCALE_10                         0xA14B
+#define mmPA_CL_VPORT_XSCALE_11                         0xA151
+#define mmPA_CL_VPORT_XSCALE_12                         0xA157
+#define mmPA_CL_VPORT_XSCALE_13                         0xA15D
+#define mmPA_CL_VPORT_XSCALE_14                         0xA163
+#define mmPA_CL_VPORT_XSCALE_15                         0xA169
+#define mmPA_CL_VPORT_XOFFSET_1                         0xA116
+#define mmPA_CL_VPORT_XOFFSET_2                         0xA11C
+#define mmPA_CL_VPORT_XOFFSET_3                         0xA122
+#define mmPA_CL_VPORT_XOFFSET_4                         0xA128
+#define mmPA_CL_VPORT_XOFFSET_5                         0xA12E
+#define mmPA_CL_VPORT_XOFFSET_6                         0xA134
+#define mmPA_CL_VPORT_XOFFSET_7                         0xA13A
+#define mmPA_CL_VPORT_XOFFSET_8                         0xA140
+#define mmPA_CL_VPORT_XOFFSET_9                         0xA146
+#define mmPA_CL_VPORT_XOFFSET_10                        0xA14C
+#define mmPA_CL_VPORT_XOFFSET_11                        0xA152
+#define mmPA_CL_VPORT_XOFFSET_12                        0xA158
+#define mmPA_CL_VPORT_XOFFSET_13                        0xA15E
+#define mmPA_CL_VPORT_XOFFSET_14                        0xA164
+#define mmPA_CL_VPORT_XOFFSET_15                        0xA16A
+#define mmPA_CL_VPORT_YSCALE_1                          0xA117
+#define mmPA_CL_VPORT_YSCALE_2                          0xA11D
+#define mmPA_CL_VPORT_YSCALE_3                          0xA123
+#define mmPA_CL_VPORT_YSCALE_4                          0xA129
+#define mmPA_CL_VPORT_YSCALE_5                          0xA12F
+#define mmPA_CL_VPORT_YSCALE_6                          0xA135
+#define mmPA_CL_VPORT_YSCALE_7                          0xA13B
+#define mmPA_CL_VPORT_YSCALE_8                          0xA141
+#define mmPA_CL_VPORT_YSCALE_9                          0xA147
+#define mmPA_CL_VPORT_YSCALE_10                         0xA14D
+#define mmPA_CL_VPORT_YSCALE_11                         0xA153
+#define mmPA_CL_VPORT_YSCALE_12                         0xA159
+#define mmPA_CL_VPORT_YSCALE_13                         0xA15F
+#define mmPA_CL_VPORT_YSCALE_14                         0xA165
+#define mmPA_CL_VPORT_YSCALE_15                         0xA16B
+#define mmPA_CL_VPORT_YOFFSET_1                         0xA118
+#define mmPA_CL_VPORT_YOFFSET_2                         0xA11E
+#define mmPA_CL_VPORT_YOFFSET_3                         0xA124
+#define mmPA_CL_VPORT_YOFFSET_4                         0xA12A
+#define mmPA_CL_VPORT_YOFFSET_5                         0xA130
+#define mmPA_CL_VPORT_YOFFSET_6                         0xA136
+#define mmPA_CL_VPORT_YOFFSET_7                         0xA13C
+#define mmPA_CL_VPORT_YOFFSET_8                         0xA142
+#define mmPA_CL_VPORT_YOFFSET_9                         0xA148
+#define mmPA_CL_VPORT_YOFFSET_10                        0xA14E
+#define mmPA_CL_VPORT_YOFFSET_11                        0xA154
+#define mmPA_CL_VPORT_YOFFSET_12                        0xA15A
+#define mmPA_CL_VPORT_YOFFSET_13                        0xA160
+#define mmPA_CL_VPORT_YOFFSET_14                        0xA166
+#define mmPA_CL_VPORT_YOFFSET_15                        0xA16C
+#define mmPA_CL_VPORT_ZSCALE_1                          0xA119
+#define mmPA_CL_VPORT_ZSCALE_2                          0xA11F
+#define mmPA_CL_VPORT_ZSCALE_3                          0xA125
+#define mmPA_CL_VPORT_ZSCALE_4                          0xA12B
+#define mmPA_CL_VPORT_ZSCALE_5                          0xA131
+#define mmPA_CL_VPORT_ZSCALE_6                          0xA137
+#define mmPA_CL_VPORT_ZSCALE_7                          0xA13D
+#define mmPA_CL_VPORT_ZSCALE_8                          0xA143
+#define mmPA_CL_VPORT_ZSCALE_9                          0xA149
+#define mmPA_CL_VPORT_ZSCALE_10                         0xA14F
+#define mmPA_CL_VPORT_ZSCALE_11                         0xA155
+#define mmPA_CL_VPORT_ZSCALE_12                         0xA15B
+#define mmPA_CL_VPORT_ZSCALE_13                         0xA161
+#define mmPA_CL_VPORT_ZSCALE_14                         0xA167
+#define mmPA_CL_VPORT_ZSCALE_15                         0xA16D
+#define mmPA_CL_VPORT_ZOFFSET_1                         0xA11A
+#define mmPA_CL_VPORT_ZOFFSET_2                         0xA120
+#define mmPA_CL_VPORT_ZOFFSET_3                         0xA126
+#define mmPA_CL_VPORT_ZOFFSET_4                         0xA12C
+#define mmPA_CL_VPORT_ZOFFSET_5                         0xA132
+#define mmPA_CL_VPORT_ZOFFSET_6                         0xA138
+#define mmPA_CL_VPORT_ZOFFSET_7                         0xA13E
+#define mmPA_CL_VPORT_ZOFFSET_8                         0xA144
+#define mmPA_CL_VPORT_ZOFFSET_9                         0xA14A
+#define mmPA_CL_VPORT_ZOFFSET_10                        0xA150
+#define mmPA_CL_VPORT_ZOFFSET_11                        0xA156
+#define mmPA_CL_VPORT_ZOFFSET_12                        0xA15C
+#define mmPA_CL_VPORT_ZOFFSET_13                        0xA162
+#define mmPA_CL_VPORT_ZOFFSET_14                        0xA168
+#define mmPA_CL_VPORT_ZOFFSET_15                        0xA16E
+#define mmPA_CL_VTE_CNTL                                0xA206
+#define mmPA_CL_VS_OUT_CNTL                             0xA207
+#define mmPA_CL_NANINF_CNTL                             0xA208
+#define mmPA_CL_CLIP_CNTL                               0xA204
+#define mmPA_CL_GB_VERT_CLIP_ADJ                        0xA303
+#define mmPA_CL_GB_VERT_DISC_ADJ                        0xA304
+#define mmPA_CL_GB_HORZ_CLIP_ADJ                        0xA305
+#define mmPA_CL_GB_HORZ_DISC_ADJ                        0xA306
+#define mmPA_CL_UCP_0_X                                 0xA388
+#define mmPA_CL_UCP_0_Y                                 0xA389
+#define mmPA_CL_UCP_0_Z                                 0xA38A
+#define mmPA_CL_UCP_0_W                                 0xA38B
+#define mmPA_CL_UCP_1_X                                 0xA38C
+#define mmPA_CL_UCP_1_Y                                 0xA38D
+#define mmPA_CL_UCP_1_Z                                 0xA38E
+#define mmPA_CL_UCP_1_W                                 0xA38F
+#define mmPA_CL_UCP_2_X                                 0xA390
+#define mmPA_CL_UCP_2_Y                                 0xA391
+#define mmPA_CL_UCP_2_Z                                 0xA392
+#define mmPA_CL_UCP_2_W                                 0xA393
+#define mmPA_CL_UCP_3_X                                 0xA394
+#define mmPA_CL_UCP_3_Y                                 0xA395
+#define mmPA_CL_UCP_3_Z                                 0xA396
+#define mmPA_CL_UCP_3_W                                 0xA397
+#define mmPA_CL_UCP_4_X                                 0xA398
+#define mmPA_CL_UCP_4_Y                                 0xA399
+#define mmPA_CL_UCP_4_Z                                 0xA39A
+#define mmPA_CL_UCP_4_W                                 0xA39B
+#define mmPA_CL_UCP_5_X                                 0xA39C
+#define mmPA_CL_UCP_5_Y                                 0xA39D
+#define mmPA_CL_UCP_5_Z                                 0xA39E
+#define mmPA_CL_UCP_5_W                                 0xA39F
+#define mmPA_CL_POINT_X_RAD                             0xA384
+#define mmPA_CL_POINT_Y_RAD                             0xA385
+#define mmPA_CL_POINT_SIZE                              0xA386
+#define mmPA_CL_POINT_CULL_RAD                          0xA387
+
+#define mmPA_SU_VTX_CNTL                                0xA302
+#define mmPA_SU_POINT_SIZE                              0xA280
+#define mmPA_SU_POINT_MINMAX                            0xA281
+#define mmPA_SU_LINE_CNTL                               0xA282
+#define mmPA_SU_SC_MODE_CNTL                            0xA205
+#define mmPA_SU_POLY_OFFSET_DB_FMT_CNTL                 0xA37E
+#define mmPA_SU_POLY_OFFSET_CLAMP                       0xA37F
+#define mmPA_SU_POLY_OFFSET_FRONT_SCALE                 0xA380
+#define mmPA_SU_POLY_OFFSET_FRONT_OFFSET                0xA381
+#define mmPA_SU_POLY_OFFSET_BACK_SCALE                  0xA382
+#define mmPA_SU_POLY_OFFSET_BACK_OFFSET                 0xA383
+
+#define mmPA_SC_WINDOW_OFFSET                           0xA080
+#define mmPA_SC_AA_CONFIG                               0xA301
+#define mmPA_SC_AA_MASK                                 0xA312
+#define mmPA_SC_AA_SAMPLE_LOCS_MCTX                     0xA307
+#define mmPA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX              0xA308
+#define mmPA_SC_LINE_STIPPLE                            0xA283
+#define mmPA_SC_LINE_CNTL                               0xA300
+#define mmPA_SC_SCREEN_SCISSOR_TL                       0xA00C
+#define mmPA_SC_SCREEN_SCISSOR_BR                       0xA00D
+#define mmPA_SC_WINDOW_SCISSOR_TL                       0xA081
+#define mmPA_SC_WINDOW_SCISSOR_BR                       0xA082
+#define mmPA_SC_CLIPRECT_RULE                           0xA083
+#define mmPA_SC_CLIPRECT_0_TL                           0xA084
+#define mmPA_SC_CLIPRECT_0_BR                           0xA085
+#define mmPA_SC_CLIPRECT_1_TL                           0xA086
+#define mmPA_SC_CLIPRECT_1_BR                           0xA087
+#define mmPA_SC_CLIPRECT_2_TL                           0xA088
+#define mmPA_SC_CLIPRECT_2_BR                           0xA089
+#define mmPA_SC_CLIPRECT_3_TL                           0xA08A
+#define mmPA_SC_CLIPRECT_3_BR                           0xA08B
+#define mmPA_SC_EDGERULE                                0xA08C
+#define mmPA_SC_GENERIC_SCISSOR_TL                      0xA090
+#define mmPA_SC_GENERIC_SCISSOR_BR                      0xA091
+#define mmPA_SC_VPORT_SCISSOR_0_TL                      0xA094
+#define mmPA_SC_VPORT_SCISSOR_1_TL                      0xA096
+#define mmPA_SC_VPORT_SCISSOR_2_TL                      0xA098
+#define mmPA_SC_VPORT_SCISSOR_3_TL                      0xA09A
+#define mmPA_SC_VPORT_SCISSOR_4_TL                      0xA09C
+#define mmPA_SC_VPORT_SCISSOR_5_TL                      0xA09E
+#define mmPA_SC_VPORT_SCISSOR_6_TL                      0xA0A0
+#define mmPA_SC_VPORT_SCISSOR_7_TL                      0xA0A2
+#define mmPA_SC_VPORT_SCISSOR_8_TL                      0xA0A4
+#define mmPA_SC_VPORT_SCISSOR_9_TL                      0xA0A6
+#define mmPA_SC_VPORT_SCISSOR_10_TL                     0xA0A8
+#define mmPA_SC_VPORT_SCISSOR_11_TL                     0xA0AA
+#define mmPA_SC_VPORT_SCISSOR_12_TL                     0xA0AC
+#define mmPA_SC_VPORT_SCISSOR_13_TL                     0xA0AE
+#define mmPA_SC_VPORT_SCISSOR_14_TL                     0xA0B0
+#define mmPA_SC_VPORT_SCISSOR_15_TL                     0xA0B2
+#define mmPA_SC_VPORT_SCISSOR_0_BR                      0xA095
+#define mmPA_SC_VPORT_SCISSOR_1_BR                      0xA097
+#define mmPA_SC_VPORT_SCISSOR_2_BR                      0xA099
+#define mmPA_SC_VPORT_SCISSOR_3_BR                      0xA09B
+#define mmPA_SC_VPORT_SCISSOR_4_BR                      0xA09D
+#define mmPA_SC_VPORT_SCISSOR_5_BR                      0xA09F
+#define mmPA_SC_VPORT_SCISSOR_6_BR                      0xA0A1
+#define mmPA_SC_VPORT_SCISSOR_7_BR                      0xA0A3
+#define mmPA_SC_VPORT_SCISSOR_8_BR                      0xA0A5
+#define mmPA_SC_VPORT_SCISSOR_9_BR                      0xA0A7
+#define mmPA_SC_VPORT_SCISSOR_10_BR                     0xA0A9
+#define mmPA_SC_VPORT_SCISSOR_11_BR                     0xA0AB
+#define mmPA_SC_VPORT_SCISSOR_12_BR                     0xA0AD
+#define mmPA_SC_VPORT_SCISSOR_13_BR                     0xA0AF
+#define mmPA_SC_VPORT_SCISSOR_14_BR                     0xA0B1
+#define mmPA_SC_VPORT_SCISSOR_15_BR                     0xA0B3
+#define mmPA_SC_VPORT_ZMIN_0                            0xA0B4
+#define mmPA_SC_VPORT_ZMIN_1                            0xA0B6
+#define mmPA_SC_VPORT_ZMIN_2                            0xA0B8
+#define mmPA_SC_VPORT_ZMIN_3                            0xA0BA
+#define mmPA_SC_VPORT_ZMIN_4                            0xA0BC
+#define mmPA_SC_VPORT_ZMIN_5                            0xA0BE
+#define mmPA_SC_VPORT_ZMIN_6                            0xA0C0
+#define mmPA_SC_VPORT_ZMIN_7                            0xA0C2
+#define mmPA_SC_VPORT_ZMIN_8                            0xA0C4
+#define mmPA_SC_VPORT_ZMIN_9                            0xA0C6
+#define mmPA_SC_VPORT_ZMIN_10                           0xA0C8
+#define mmPA_SC_VPORT_ZMIN_11                           0xA0CA
+#define mmPA_SC_VPORT_ZMIN_12                           0xA0CC
+#define mmPA_SC_VPORT_ZMIN_13                           0xA0CE
+#define mmPA_SC_VPORT_ZMIN_14                           0xA0D0
+#define mmPA_SC_VPORT_ZMIN_15                           0xA0D2
+#define mmPA_SC_VPORT_ZMAX_0                            0xA0B5
+#define mmPA_SC_VPORT_ZMAX_1                            0xA0B7
+#define mmPA_SC_VPORT_ZMAX_2                            0xA0B9
+#define mmPA_SC_VPORT_ZMAX_3                            0xA0BB
+#define mmPA_SC_VPORT_ZMAX_4                            0xA0BD
+#define mmPA_SC_VPORT_ZMAX_5                            0xA0BF
+#define mmPA_SC_VPORT_ZMAX_6                            0xA0C1
+#define mmPA_SC_VPORT_ZMAX_7                            0xA0C3
+#define mmPA_SC_VPORT_ZMAX_8                            0xA0C5
+#define mmPA_SC_VPORT_ZMAX_9                            0xA0C7
+#define mmPA_SC_VPORT_ZMAX_10                           0xA0C9
+#define mmPA_SC_VPORT_ZMAX_11                           0xA0CB
+#define mmPA_SC_VPORT_ZMAX_12                           0xA0CD
+#define mmPA_SC_VPORT_ZMAX_13                           0xA0CF
+#define mmPA_SC_VPORT_ZMAX_14                           0xA0D1
+#define mmPA_SC_VPORT_ZMAX_15                           0xA0D3
+#define mmPA_SC_MODE_CNTL                               0xA293
+#define mmPA_SC_MPASS_PS_CNTL                           0xA292
+
+#define mmVGT_DRAW_INITIATOR                            0xA1FC
+#define mmVGT_EVENT_INITIATOR                           0xA2A4
+#define mmVGT_EVENT_ADDRESS_REG                         0xA1FE
+#define mmVGT_DMA_BASE_HI                               0xA1F9
+#define mmVGT_DMA_BASE                                  0xA1FA
+#define mmVGT_DMA_INDEX_TYPE                            0xA29F
+#define mmVGT_DMA_NUM_INSTANCES                         0xA2A2
+#define mmVGT_DMA_SIZE                                  0xA29D
+
+#define mmVGT_IMMED_DATA                                0xA1FD
+#define mmVGT_INDEX_TYPE                                0x2257
+#define mmVGT_NUM_INDICES                               0x225C
+#define mmVGT_NUM_INSTANCES                             0x225D
+#define mmVGT_PRIMITIVE_TYPE                            0x2256
+#define mmVGT_PRIMITIVEID_EN                            0xA2A1
+#define mmVGT_VTX_CNT_EN                                0xA2AE
+#define mmVGT_REUSE_OFF                                 0xA2AD
+#define mmVGT_INSTANCE_STEP_RATE_0                      0xA2A8
+#define mmVGT_INSTANCE_STEP_RATE_1                      0xA2A9
+#define mmVGT_MAX_VTX_INDX                              0xA100
+#define mmVGT_MIN_VTX_INDX                              0xA101
+#define mmVGT_INDX_OFFSET                               0xA102
+#define mmVGT_VERTEX_REUSE_BLOCK_CNTL                   0xA316
+#define mmVGT_OUT_DEALLOC_CNTL                          0xA317
+#define mmVGT_MULTI_PRIM_IB_RESET_INDX                  0xA103
+#define mmVGT_MULTI_PRIM_IB_RESET_EN                    0xA2A5
+#define mmVGT_ENHANCE                                   0xA294
+#define mmVGT_OUTPUT_PATH_CNTL                          0xA284
+#define mmVGT_HOS_CNTL                                  0xA285
+#define mmVGT_HOS_MAX_TESS_LEVEL                        0xA286
+#define mmVGT_HOS_MIN_TESS_LEVEL                        0xA287
+#define mmVGT_HOS_REUSE_DEPTH                           0xA288
+#define mmVGT_GROUP_PRIM_TYPE                           0xA289
+#define mmVGT_GROUP_FIRST_DECR                          0xA28A
+#define mmVGT_GROUP_DECR                                0xA28B
+#define mmVGT_GROUP_VECT_0_CNTL                         0xA28C
+#define mmVGT_GROUP_VECT_1_CNTL                         0xA28D
+#define mmVGT_GROUP_VECT_0_FMT_CNTL                     0xA28E
+#define mmVGT_GROUP_VECT_1_FMT_CNTL                     0xA28F
+#define mmVGT_GS_MODE                                   0xA290
+#define mmVGT_GS_OUT_PRIM_TYPE                          0xA29B
+
+#define mmVGT_STRMOUT_EN                                0xA2AC
+#define mmVGT_STRMOUT_BUFFER_SIZE_0                     0xA2B4
+#define mmVGT_STRMOUT_BUFFER_SIZE_1                     0xA2B8
+#define mmVGT_STRMOUT_BUFFER_SIZE_2                     0xA2BC
+#define mmVGT_STRMOUT_BUFFER_SIZE_3                     0xA2C0
+#define mmVGT_STRMOUT_BUFFER_OFFSET_0                   0xA2B7
+#define mmVGT_STRMOUT_BUFFER_OFFSET_1                   0xA2BB
+#define mmVGT_STRMOUT_BUFFER_OFFSET_2                   0xA2BF
+#define mmVGT_STRMOUT_BUFFER_OFFSET_3                   0xA2C3
+#define mmVGT_STRMOUT_VTX_STRIDE_0                      0xA2B5
+#define mmVGT_STRMOUT_VTX_STRIDE_1                      0xA2B9
+#define mmVGT_STRMOUT_VTX_STRIDE_2                      0xA2BD
+#define mmVGT_STRMOUT_VTX_STRIDE_3                      0xA2C1
+#define mmVGT_STRMOUT_BUFFER_BASE_0                     0xA2B6
+#define mmVGT_STRMOUT_BUFFER_BASE_1                     0xA2BA
+#define mmVGT_STRMOUT_BUFFER_BASE_2                     0xA2BE
+#define mmVGT_STRMOUT_BUFFER_BASE_3                     0xA2C2
+#define mmVGT_STRMOUT_BUFFER_EN                         0xA2C8
+#define mmVGT_STRMOUT_BASE_OFFSET_0                     0xA2C4
+#define mmVGT_STRMOUT_BASE_OFFSET_1                     0xA2C5
+#define mmVGT_STRMOUT_BASE_OFFSET_2                     0xA2C6
+#define mmVGT_STRMOUT_BASE_OFFSET_3                     0xA2C7
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_0                  0xA2D1
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_1                  0xA2D2
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_2                  0xA2D3
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_3                  0xA2D4
+#define mmVGT_STRMOUT_DRAW_OPAQUE_OFFSET                0xA2CA
+#define mmVGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE    0xA2CB
+#define mmVGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE         0xA2CC
+
+#define mmSQ_PGM_START_PS                               0xA210
+#define mmSQ_PGM_CF_OFFSET_PS                           0xA233
+#define mmSQ_PGM_RESOURCES_PS                           0xA214
+#define mmSQ_PGM_EXPORTS_PS                             0xA215
+#define mmSQ_PGM_START_VS                               0xA216
+#define mmSQ_PGM_CF_OFFSET_VS                           0xA234
+#define mmSQ_PGM_RESOURCES_VS                           0xA21A
+#define mmSQ_PGM_START_GS                               0xA21B
+#define mmSQ_PGM_CF_OFFSET_GS                           0xA235
+#define mmSQ_PGM_RESOURCES_GS                           0xA21F
+#define mmSQ_PGM_START_ES                               0xA220
+#define mmSQ_PGM_CF_OFFSET_ES                           0xA236
+#define mmSQ_PGM_RESOURCES_ES                           0xA224
+#define mmSQ_PGM_START_FS                               0xA225
+#define mmSQ_PGM_CF_OFFSET_FS                           0xA237
+#define mmSQ_PGM_RESOURCES_FS                           0xA229
+#define mmSQ_ESGS_RING_ITEMSIZE                         0xA22A
+#define mmSQ_GSVS_RING_ITEMSIZE                         0xA22B
+#define mmSQ_ESTMP_RING_ITEMSIZE                        0xA22C
+#define mmSQ_GSTMP_RING_ITEMSIZE                        0xA22D
+#define mmSQ_VSTMP_RING_ITEMSIZE                        0xA22E
+#define mmSQ_PSTMP_RING_ITEMSIZE                        0xA22F
+#define mmSQ_FBUF_RING_ITEMSIZE                         0xA230
+#define mmSQ_REDUC_RING_ITEMSIZE                        0xA231
+#define mmSQ_GS_VERT_ITEMSIZE                           0xA232
+#define mmSQ_VTX_SEMANTIC_CLEAR                         0xA238
+
+#define mmSQ_VTX_SEMANTIC_0                             0xA0E0
+#define mmSQ_VTX_SEMANTIC_1                             0xA0E1
+#define mmSQ_VTX_SEMANTIC_2                             0xA0E2
+#define mmSQ_VTX_SEMANTIC_3                             0xA0E3
+#define mmSQ_VTX_SEMANTIC_4                             0xA0E4
+#define mmSQ_VTX_SEMANTIC_5                             0xA0E5
+#define mmSQ_VTX_SEMANTIC_6                             0xA0E6
+#define mmSQ_VTX_SEMANTIC_7                             0xA0E7
+#define mmSQ_VTX_SEMANTIC_8                             0xA0E8
+#define mmSQ_VTX_SEMANTIC_9                             0xA0E9
+#define mmSQ_VTX_SEMANTIC_10                            0xA0EA
+#define mmSQ_VTX_SEMANTIC_11                            0xA0EB
+#define mmSQ_VTX_SEMANTIC_12                            0xA0EC
+#define mmSQ_VTX_SEMANTIC_13                            0xA0ED
+#define mmSQ_VTX_SEMANTIC_14                            0xA0EE
+#define mmSQ_VTX_SEMANTIC_15                            0xA0EF
+#define mmSQ_VTX_SEMANTIC_16                            0xA0F0
+#define mmSQ_VTX_SEMANTIC_17                            0xA0F1
+#define mmSQ_VTX_SEMANTIC_18                            0xA0F2
+#define mmSQ_VTX_SEMANTIC_19                            0xA0F3
+#define mmSQ_VTX_SEMANTIC_20                            0xA0F4
+#define mmSQ_VTX_SEMANTIC_21                            0xA0F5
+#define mmSQ_VTX_SEMANTIC_22                            0xA0F6
+#define mmSQ_VTX_SEMANTIC_23                            0xA0F7
+#define mmSQ_VTX_SEMANTIC_24                            0xA0F8
+#define mmSQ_VTX_SEMANTIC_25                            0xA0F9
+#define mmSQ_VTX_SEMANTIC_26                            0xA0FA
+#define mmSQ_VTX_SEMANTIC_27                            0xA0FB
+#define mmSQ_VTX_SEMANTIC_28                            0xA0FC
+#define mmSQ_VTX_SEMANTIC_29                            0xA0FD
+#define mmSQ_VTX_SEMANTIC_30                            0xA0FE
+#define mmSQ_VTX_SEMANTIC_31                            0xA0FF
+
+#define mmSQ_ALU_CONST_CACHE_PS_0                       0xA250
+#define mmSQ_ALU_CONST_CACHE_PS_1                       0xA251
+#define mmSQ_ALU_CONST_CACHE_PS_2                       0xA252
+#define mmSQ_ALU_CONST_CACHE_PS_3                       0xA253
+#define mmSQ_ALU_CONST_CACHE_PS_4                       0xA254
+#define mmSQ_ALU_CONST_CACHE_PS_5                       0xA255
+#define mmSQ_ALU_CONST_CACHE_PS_6                       0xA256
+#define mmSQ_ALU_CONST_CACHE_PS_7                       0xA257
+#define mmSQ_ALU_CONST_CACHE_PS_8                       0xA258
+#define mmSQ_ALU_CONST_CACHE_PS_9                       0xA259
+#define mmSQ_ALU_CONST_CACHE_PS_10                      0xA25A
+#define mmSQ_ALU_CONST_CACHE_PS_11                      0xA25B
+#define mmSQ_ALU_CONST_CACHE_PS_12                      0xA25C
+#define mmSQ_ALU_CONST_CACHE_PS_13                      0xA25D
+#define mmSQ_ALU_CONST_CACHE_PS_14                      0xA25E
+#define mmSQ_ALU_CONST_CACHE_PS_15                      0xA25F
+#define mmSQ_ALU_CONST_CACHE_VS_0                       0xA260
+#define mmSQ_ALU_CONST_CACHE_VS_1                       0xA261
+#define mmSQ_ALU_CONST_CACHE_VS_2                       0xA262
+#define mmSQ_ALU_CONST_CACHE_VS_3                       0xA263
+#define mmSQ_ALU_CONST_CACHE_VS_4                       0xA264
+#define mmSQ_ALU_CONST_CACHE_VS_5                       0xA265
+#define mmSQ_ALU_CONST_CACHE_VS_6                       0xA266
+#define mmSQ_ALU_CONST_CACHE_VS_7                       0xA267
+#define mmSQ_ALU_CONST_CACHE_VS_8                       0xA268
+#define mmSQ_ALU_CONST_CACHE_VS_9                       0xA269
+#define mmSQ_ALU_CONST_CACHE_VS_10                      0xA26A
+#define mmSQ_ALU_CONST_CACHE_VS_11                      0xA26B
+#define mmSQ_ALU_CONST_CACHE_VS_12                      0xA26C
+#define mmSQ_ALU_CONST_CACHE_VS_13                      0xA26D
+#define mmSQ_ALU_CONST_CACHE_VS_14                      0xA26E
+#define mmSQ_ALU_CONST_CACHE_VS_15                      0xA26F
+#define mmSQ_ALU_CONST_CACHE_GS_0                       0xA270
+#define mmSQ_ALU_CONST_CACHE_GS_1                       0xA271
+#define mmSQ_ALU_CONST_CACHE_GS_2                       0xA272
+#define mmSQ_ALU_CONST_CACHE_GS_3                       0xA273
+#define mmSQ_ALU_CONST_CACHE_GS_4                       0xA274
+#define mmSQ_ALU_CONST_CACHE_GS_5                       0xA275
+#define mmSQ_ALU_CONST_CACHE_GS_6                       0xA276
+#define mmSQ_ALU_CONST_CACHE_GS_7                       0xA277
+#define mmSQ_ALU_CONST_CACHE_GS_8                       0xA278
+#define mmSQ_ALU_CONST_CACHE_GS_9                       0xA279
+#define mmSQ_ALU_CONST_CACHE_GS_10                      0xA27A
+#define mmSQ_ALU_CONST_CACHE_GS_11                      0xA27B
+#define mmSQ_ALU_CONST_CACHE_GS_12                      0xA27C
+#define mmSQ_ALU_CONST_CACHE_GS_13                      0xA27D
+#define mmSQ_ALU_CONST_CACHE_GS_14                      0xA27E
+#define mmSQ_ALU_CONST_CACHE_GS_15                      0xA27F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_0                 0xA050
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_1                 0xA051
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_2                 0xA052
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_3                 0xA053
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_4                 0xA054
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_5                 0xA055
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_6                 0xA056
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_7                 0xA057
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_8                 0xA058
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_9                 0xA059
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_10                0xA05A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_11                0xA05B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_12                0xA05C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_13                0xA05D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_14                0xA05E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_15                0xA05F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_0                 0xA060
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_1                 0xA061
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_2                 0xA062
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_3                 0xA063
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_4                 0xA064
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_5                 0xA065
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_6                 0xA066
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_7                 0xA067
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_8                 0xA068
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_9                 0xA069
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_10                0xA06A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_11                0xA06B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_12                0xA06C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_13                0xA06D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_14                0xA06E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_15                0xA06F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_0                 0xA070
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_1                 0xA071
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_2                 0xA072
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_3                 0xA073
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_4                 0xA074
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_5                 0xA075
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_6                 0xA076
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_7                 0xA077
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_8                 0xA078
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_9                 0xA079
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_10                0xA07A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_11                0xA07B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_12                0xA07C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_13                0xA07D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_14                0xA07E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_15                0xA07F
+
+#define mmSPI_VS_OUT_ID_0                               0xA185
+#define mmSPI_VS_OUT_ID_1                               0xA186
+#define mmSPI_VS_OUT_ID_2                               0xA187
+#define mmSPI_VS_OUT_ID_3                               0xA188
+#define mmSPI_VS_OUT_ID_4                               0xA189
+#define mmSPI_VS_OUT_ID_5                               0xA18A
+#define mmSPI_VS_OUT_ID_6                               0xA18B
+#define mmSPI_VS_OUT_ID_7                               0xA18C
+#define mmSPI_VS_OUT_ID_8                               0xA18D
+#define mmSPI_VS_OUT_ID_9                               0xA18E
+#define mmSPI_PS_INPUT_CNTL_0                           0xA191
+#define mmSPI_PS_INPUT_CNTL_1                           0xA192
+#define mmSPI_PS_INPUT_CNTL_2                           0xA193
+#define mmSPI_PS_INPUT_CNTL_3                           0xA194
+#define mmSPI_PS_INPUT_CNTL_4                           0xA195
+#define mmSPI_PS_INPUT_CNTL_5                           0xA196
+#define mmSPI_PS_INPUT_CNTL_6                           0xA197
+#define mmSPI_PS_INPUT_CNTL_7                           0xA198
+#define mmSPI_PS_INPUT_CNTL_8                           0xA199
+#define mmSPI_PS_INPUT_CNTL_9                           0xA19A
+#define mmSPI_PS_INPUT_CNTL_10                          0xA19B
+#define mmSPI_PS_INPUT_CNTL_11                          0xA19C
+#define mmSPI_PS_INPUT_CNTL_12                          0xA19D
+#define mmSPI_PS_INPUT_CNTL_13                          0xA19E
+#define mmSPI_PS_INPUT_CNTL_14                          0xA19F
+#define mmSPI_PS_INPUT_CNTL_15                          0xA1A0
+#define mmSPI_PS_INPUT_CNTL_16                          0xA1A1
+#define mmSPI_PS_INPUT_CNTL_17                          0xA1A2
+#define mmSPI_PS_INPUT_CNTL_18                          0xA1A3
+#define mmSPI_PS_INPUT_CNTL_19                          0xA1A4
+#define mmSPI_PS_INPUT_CNTL_20                          0xA1A5
+#define mmSPI_PS_INPUT_CNTL_21                          0xA1A6
+#define mmSPI_PS_INPUT_CNTL_22                          0xA1A7
+#define mmSPI_PS_INPUT_CNTL_23                          0xA1A8
+#define mmSPI_PS_INPUT_CNTL_24                          0xA1A9
+#define mmSPI_PS_INPUT_CNTL_25                          0xA1AA
+#define mmSPI_PS_INPUT_CNTL_26                          0xA1AB
+#define mmSPI_PS_INPUT_CNTL_27                          0xA1AC
+#define mmSPI_PS_INPUT_CNTL_28                          0xA1AD
+#define mmSPI_PS_INPUT_CNTL_29                          0xA1AE
+#define mmSPI_PS_INPUT_CNTL_30                          0xA1AF
+#define mmSPI_PS_INPUT_CNTL_31                          0xA1B0
+#define mmSPI_VS_OUT_CONFIG                             0xA1B1
+#define mmSPI_THREAD_GROUPING                           0xA1B2
+#define mmSPI_PS_IN_CONTROL_0                           0xA1B3
+#define mmSPI_PS_IN_CONTROL_1                           0xA1B4
+#define mmSPI_INTERP_CONTROL_0                          0xA1B5
+#define mmSPI_INPUT_Z                                   0xA1B6
+#define mmSPI_FOG_CNTL                                  0xA1B7
+#define mmSPI_FOG_FUNC_SCALE                            0xA1B8
+#define mmSPI_FOG_FUNC_BIAS                             0xA1B9
+
+#define mmSX_MISC                                       0xA0D4
+#define mmSX_ALPHA_TEST_CONTROL                         0xA104
+#define mmSX_ALPHA_REF                                  0xA10E
+
+#define mmDB_DEPTH_BASE                                 0xA003
+#define mmDB_DEPTH_INFO                                 0xA004
+#define mmDB_HTILE_DATA_BASE                            0xA005
+#define mmDB_DEPTH_SIZE                                 0xA000
+#define mmDB_DEPTH_VIEW                                 0xA001
+#define mmDB_RENDER_CONTROL                             0xA343
+#define mmDB_RENDER_OVERRIDE                            0xA344
+#define mmDB_SHADER_CONTROL                             0xA203
+#define mmDB_STENCIL_CLEAR                              0xA00A
+#define mmDB_DEPTH_CLEAR                                0xA00B
+#define mmDB_HTILE_SURFACE                              0xA349
+#define mmDB_PRELOAD_CONTROL                            0xA34C
+#define mmDB_PREFETCH_LIMIT                             0xA34D
+#define mmDB_STENCILREFMASK                             0xA10C
+#define mmDB_STENCILREFMASK_BF                          0xA10D
+#define mmDB_SRESULTS_COMPARE_STATE0                    0xA34A
+#define mmDB_SRESULTS_COMPARE_STATE1                    0xA34B
+#define mmDB_DEPTH_CONTROL                              0xA200
+#define mmDB_ALPHA_TO_MASK                              0xA351
+
+#define mmCB_CLEAR_RED_R6XX                             0xA048
+#define mmCB_CLEAR_GREEN_R6XX                           0xA049
+#define mmCB_CLEAR_BLUE_R6XX                            0xA04A
+#define mmCB_CLEAR_ALPHA_R6XX                           0xA04B
+#define mmCB_BLEND_RED                                  0xA105
+#define mmCB_BLEND_GREEN                                0xA106
+#define mmCB_BLEND_BLUE                                 0xA107
+#define mmCB_BLEND_ALPHA                                0xA108
+#define mmCB_FOG_RED_R6XX                               0xA109
+#define mmCB_FOG_GREEN_R6XX                             0xA10A
+#define mmCB_FOG_BLUE_R6XX                              0xA10B
+#define mmCB_BLEND_CONTROL                              0xA201
+#define mmCB_COLOR_CONTROL                              0xA202
+#define mmCB_BLEND0_CONTROL                             0xA1E0
+#define mmCB_BLEND1_CONTROL                             0xA1E1
+#define mmCB_BLEND2_CONTROL                             0xA1E2
+#define mmCB_BLEND3_CONTROL                             0xA1E3
+#define mmCB_BLEND4_CONTROL                             0xA1E4
+#define mmCB_BLEND5_CONTROL                             0xA1E5
+#define mmCB_BLEND6_CONTROL                             0xA1E6
+#define mmCB_BLEND7_CONTROL                             0xA1E7
+#define mmCB_CLRCMP_CONTROL                             0xA30C
+#define mmCB_CLRCMP_SRC                                 0xA30D
+#define mmCB_CLRCMP_DST                                 0xA30E
+#define mmCB_CLRCMP_MSK                                 0xA30F
+#define mmCB_COLOR0_BASE                                0xA010
+#define mmCB_COLOR1_BASE                                0xA011
+#define mmCB_COLOR2_BASE                                0xA012
+#define mmCB_COLOR3_BASE                                0xA013
+#define mmCB_COLOR4_BASE                                0xA014
+#define mmCB_COLOR5_BASE                                0xA015
+#define mmCB_COLOR6_BASE                                0xA016
+#define mmCB_COLOR7_BASE                                0xA017
+#define mmCB_COLOR0_SIZE                                0xA018
+#define mmCB_COLOR1_SIZE                                0xA019
+#define mmCB_COLOR2_SIZE                                0xA01A
+#define mmCB_COLOR3_SIZE                                0xA01B
+#define mmCB_COLOR4_SIZE                                0xA01C
+#define mmCB_COLOR5_SIZE                                0xA01D
+#define mmCB_COLOR6_SIZE                                0xA01E
+#define mmCB_COLOR7_SIZE                                0xA01F
+#define mmCB_COLOR0_VIEW                                0xA020
+#define mmCB_COLOR1_VIEW                                0xA021
+#define mmCB_COLOR2_VIEW                                0xA022
+#define mmCB_COLOR3_VIEW                                0xA023
+#define mmCB_COLOR4_VIEW                                0xA024
+#define mmCB_COLOR5_VIEW                                0xA025
+#define mmCB_COLOR6_VIEW                                0xA026
+#define mmCB_COLOR7_VIEW                                0xA027
+#define mmCB_COLOR0_INFO                                0xA028
+#define mmCB_COLOR1_INFO                                0xA029
+#define mmCB_COLOR2_INFO                                0xA02A
+#define mmCB_COLOR3_INFO                                0xA02B
+#define mmCB_COLOR4_INFO                                0xA02C
+#define mmCB_COLOR5_INFO                                0xA02D
+#define mmCB_COLOR6_INFO                                0xA02E
+#define mmCB_COLOR7_INFO                                0xA02F
+#define mmCB_COLOR0_TILE                                0xA030
+#define mmCB_COLOR1_TILE                                0xA031
+#define mmCB_COLOR2_TILE                                0xA032
+#define mmCB_COLOR3_TILE                                0xA033
+#define mmCB_COLOR4_TILE                                0xA034
+#define mmCB_COLOR5_TILE                                0xA035
+#define mmCB_COLOR6_TILE                                0xA036
+#define mmCB_COLOR7_TILE                                0xA037
+#define mmCB_COLOR0_FRAG                                0xA038
+#define mmCB_COLOR1_FRAG                                0xA039
+#define mmCB_COLOR2_FRAG                                0xA03A
+#define mmCB_COLOR3_FRAG                                0xA03B
+#define mmCB_COLOR4_FRAG                                0xA03C
+#define mmCB_COLOR5_FRAG                                0xA03D
+#define mmCB_COLOR6_FRAG                                0xA03E
+#define mmCB_COLOR7_FRAG                                0xA03F
+#define mmCB_COLOR0_MASK                                0xA040
+#define mmCB_COLOR1_MASK                                0xA041
+#define mmCB_COLOR2_MASK                                0xA042
+#define mmCB_COLOR3_MASK                                0xA043
+#define mmCB_COLOR4_MASK                                0xA044
+#define mmCB_COLOR5_MASK                                0xA045
+#define mmCB_COLOR6_MASK                                0xA046
+#define mmCB_COLOR7_MASK                                0xA047
+#define mmCB_CLEAR_RED_R6XX                             0xA048
+#define mmCB_CLEAR_GREEN_R6XX                           0xA049
+#define mmCB_CLEAR_BLUE_R6XX                            0xA04A
+#define mmCB_CLEAR_ALPHA_R6XX                           0xA04B
+#define mmCB_TARGET_MASK                                0xA08E
+#define mmCB_SHADER_MASK                                0xA08F
+#define mmCB_SHADER_CONTROL                             0xA1E8
+
+#define mmSQ_VTX_BASE_VTX_LOC                           0xF3FC
+#define mmSQ_VTX_START_INST_LOC                         0xF3FD
+
+#endif /* _R700_CHIPOFFSET_H_ */
+
diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c
new file mode 100644
index 0000000000..c6546ab00c
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_clear.c
@@ -0,0 +1,118 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+ 
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/enums.h"
+#include "swrast/swrast.h"
+
+#include "radeon_lock.h"
+#include "r600_context.h"
+
+#include "r700_shaderinst.h"
+#include "r600_emit.h"
+#include "r700_clear.h"
+
+static GLboolean r700ClearFast(context_t *context, GLbitfield mask)
+{
+    /* TODO, fast clear need implementation */
+    return GL_FALSE;
+}
+
+void r700Clear(GLcontext * ctx, GLbitfield mask)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    __DRIdrawablePrivate *dPriv = radeon_get_drawable(&context->radeon);
+    const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+    GLbitfield swrast_mask = 0, tri_mask = 0;
+    int i;
+    struct gl_framebuffer *fb = ctx->DrawBuffer;
+
+    radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x\n", __func__, mask);
+
+    if( GL_TRUE == r700ClearFast(context, mask) )
+    {
+        return;
+    }
+	if (!context->radeon.radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(&context->radeon);
+		UNLOCK_HARDWARE(&context->radeon);
+		if (dPriv->numClipRects == 0)
+			return;
+	}
+
+	R600_NEWPRIM(context);
+
+	if (colorMask == ~0)
+	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+	else
+	  tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
+
+
+	/* HW stencil */
+	if (mask & BUFFER_BIT_STENCIL) {
+		tri_mask |= BUFFER_BIT_STENCIL;
+	}
+
+	/* HW depth */
+	if (mask & BUFFER_BIT_DEPTH) {
+    	        tri_mask |= BUFFER_BIT_DEPTH;
+	}
+
+	/* If we're doing a tri pass for depth/stencil, include a likely color
+	 * buffer with it.
+	 */
+
+	for (i = 0; i < BUFFER_COUNT; i++) {
+	  GLuint bufBit = 1 << i;
+	  if ((tri_mask) & bufBit) {
+	    if (!fb->Attachment[i].Renderbuffer->ClassID) {
+	      tri_mask &= ~bufBit;
+	      swrast_mask |= bufBit;
+	    }
+	  }
+	}
+
+	/* SW fallback clearing */
+	swrast_mask = mask & ~tri_mask;
+
+	if (tri_mask) {
+		radeonUserClear(ctx, tri_mask);
+	}
+
+	if (swrast_mask) {
+		radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, "%s: swrast clear, mask: %x\n",
+				__FUNCTION__, swrast_mask);
+		_swrast_Clear(ctx, swrast_mask);
+	}
+
+}
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_clear.h b/src/mesa/drivers/dri/r600/r700_clear.h
new file mode 100644
index 0000000000..bed1d3a90e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_clear.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef __r700_CLEAR_H__
+#define __r700_CLEAR_H__
+
+extern void r700Clear(GLcontext * ctx, GLbitfield mask);
+
+#endif /* __r700_CLEAR_H__ */
diff --git a/src/mesa/drivers/dri/r600/r700_debug.c b/src/mesa/drivers/dri/r600/r700_debug.c
new file mode 100644
index 0000000000..cd1ba9eca3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_debug.c
@@ -0,0 +1,60 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "r700_debug.h"
+#include "radeon_debug.h"
+
+void DumpHwBinary(int type, void *addr, int size)
+{
+    int i;
+    unsigned int *pHw = (unsigned int *)addr;
+
+    return;
+
+    switch (type)
+    {
+        case DUMP_PIXEL_SHADER:
+            radeon_print(RADEON_SHADER, RADEON_TRACE, "Pixel Shader\n");
+        break;
+        case DUMP_VERTEX_SHADER:
+            radeon_print(RADEON_SHADER, RADEON_TRACE, "Vertex Shader\n");
+        break;
+        case DUMP_FETCH_SHADER:
+            radeon_print(RADEON_SHADER, RADEON_TRACE, "Fetch Shader\n");
+        break;
+    }
+
+    for (i = 0; i < size; i++)
+    {
+        radeon_print(RADEON_SHADER, RADEON_TRACE, "0x%08x,\t", *pHw);
+        if (i%4 == 3)
+            radeon_print(RADEON_SHADER, RADEON_TRACE, "0x%08x\n", *pHw);
+        pHw++;
+
+    }
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_debug.h b/src/mesa/drivers/dri/r600/r700_debug.h
new file mode 100644
index 0000000000..c0921bf610
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_debug.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_DEBUG_H_
+#define _R700_DEBUG_H_
+enum R700_DUMP_TYPE
+{
+	DUMP_VERTEX_SHADER      = 0x1,
+	DUMP_PIXEL_SHADER       = 0x2,
+	DUMP_FETCH_SHADER       = 0x4,
+};
+
+extern void DumpHwBinary(int, void *, int);
+
+#endif /*_R700_DEBUG_H_*/
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.h b/src/mesa/drivers/dri/r600/r700_driconf.h
index e5fd960072..a9e2152344 100644
--- a/src/gallium/state_trackers/glx/xlib/fakeglx.h
+++ b/src/mesa/drivers/dri/r600/r700_driconf.h
@@ -1,9 +1,5 @@
-
 /*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -18,24 +14,20 @@
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
 
-#ifndef FAKEGLX_H
-#define FAKEGLX_H
-
-
-#include <X11/Xlib.h>
-
-struct _glxapi_table;
-
-extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void);
-
-extern void Fake_glXUseXFont( Font font, int first, int count, int listbase );
-
+#ifndef _R700_DRICONF_H_
+#define _R700_DRICONF_H_
 
-#endif
+#define DRI_CONF_FP_OPTIMIZATION_SPEED   0
+#define DRI_CONF_FP_OPTIMIZATION_QUALITY 1
 
+#endif /* _R700_DRICONF_H_ */
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c
new file mode 100644
index 0000000000..78ce3ae436
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.c
@@ -0,0 +1,476 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_fragprog.h"
+
+#include "r700_debug.h"
+
+//TODO : Validate FP input with VP output.
+void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
+						  struct gl_fragment_program *mesa_fp)
+{
+	unsigned int unBit;
+    unsigned int i;
+    GLuint       ui;
+
+	pAsm->number_used_registers = 0;
+
+//Input mapping : mesa_fp->Base.InputsRead set the flag, set in 
+	//The flags parsed in parse_attrib_binding. FRAG_ATTRIB_COLx, FRAG_ATTRIB_TEXx, ...
+	//MUST match order in Map_Vertex_Output
+	unBit = 1 << FRAG_ATTRIB_WPOS;
+	if(mesa_fp->Base.InputsRead & unBit)
+	{
+		pAsm->uiFP_AttributeMap[FRAG_ATTRIB_WPOS] = pAsm->number_used_registers++;
+	}
+
+	unBit = 1 << FRAG_ATTRIB_COL0;
+	if(mesa_fp->Base.InputsRead & unBit)
+	{
+		pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0] = pAsm->number_used_registers++;
+	}
+
+	unBit = 1 << FRAG_ATTRIB_COL1;
+	if(mesa_fp->Base.InputsRead & unBit)
+	{
+		pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++;
+	}
+
+        unBit = 1 << FRAG_ATTRIB_FOGC;
+        if(mesa_fp->Base.InputsRead & unBit)
+        {
+                pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++;
+        }
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (FRAG_ATTRIB_TEX0 + i);
+		if(mesa_fp->Base.InputsRead & unBit)
+		{
+			pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i] = pAsm->number_used_registers++;
+		}
+	}
+
+/* Map temporary registers (GPRs) */
+    pAsm->starting_temp_register_number = pAsm->number_used_registers;
+
+    if(mesa_fp->Base.NumNativeTemporaries >= mesa_fp->Base.NumTemporaries)
+    {
+	    pAsm->number_used_registers += mesa_fp->Base.NumNativeTemporaries;
+    }
+    else
+    {
+        pAsm->number_used_registers += mesa_fp->Base.NumTemporaries;
+    }
+
+/* Output mapping */
+	pAsm->number_of_exports = 0;
+	pAsm->number_of_colorandz_exports = 0; /* don't include stencil and mask out. */
+	pAsm->starting_export_register_number = pAsm->number_used_registers;
+	unBit = 1 << FRAG_RESULT_COLOR;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+		pAsm->uiFP_OutputMap[FRAG_RESULT_COLOR] = pAsm->number_used_registers++;
+		pAsm->number_of_exports++;
+		pAsm->number_of_colorandz_exports++;
+	}
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+        pAsm->depth_export_register_number = pAsm->number_used_registers;
+		pAsm->uiFP_OutputMap[FRAG_RESULT_DEPTH] = pAsm->number_used_registers++;
+		pAsm->number_of_exports++;
+		pAsm->number_of_colorandz_exports++;
+		pAsm->pR700Shader->depthIsExported = 1;
+	}
+
+    pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports);
+    for(ui=0; ui<pAsm->number_of_exports; ui++)
+    {
+        pAsm->pucOutMask[ui] = 0x0;
+    }
+
+    pAsm->uFirstHelpReg = pAsm->number_used_registers;
+}
+
+GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
+					                	struct gl_fragment_program   *mesa_fp)
+{
+    GLuint i, j;
+    GLint * puiTEMPwrites;
+    struct prog_instruction * pILInst;
+    InstDeps         *pInstDeps;
+    struct prog_instruction * texcoord_DepInst;
+    GLint              nDepInstID;
+
+    puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_fp->Base.NumTemporaries);
+    for(i=0; i<mesa_fp->Base.NumTemporaries; i++)
+    {
+        puiTEMPwrites[i] = -1;
+    }
+
+    pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_fp->Base.NumInstructions);
+
+    for(i=0; i<mesa_fp->Base.NumInstructions; i++)
+    {
+        pInstDeps[i].nDstDep = -1;
+        pILInst = &(mesa_fp->Base.Instructions[i]);
+
+        //Dst
+        if(pILInst->DstReg.File == PROGRAM_TEMPORARY)
+        {
+            //Set lastwrite for the temp
+            puiTEMPwrites[pILInst->DstReg.Index] = i;
+        }
+
+        //Src
+        for(j=0; j<3; j++)
+        {
+            if(pILInst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            {
+                //Set dep.
+                pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index];
+            }
+            else
+            {
+                pInstDeps[i].nSrcDeps[j] = -1;
+            }
+        }
+    }
+
+    fp->r700AsmCode.pInstDeps = pInstDeps;
+
+    FREE(puiTEMPwrites);
+
+    //Find dep for tex inst    
+    for(i=0; i<mesa_fp->Base.NumInstructions; i++)
+    {
+        pILInst = &(mesa_fp->Base.Instructions[i]);
+
+        if(GL_TRUE == IsTex(pILInst->Opcode))
+        {   //src0 is the tex coord register, src1 is texunit, src2 is textype
+            nDepInstID = pInstDeps[i].nSrcDeps[0];
+            if(nDepInstID >= 0)
+            {
+                texcoord_DepInst = &(mesa_fp->Base.Instructions[nDepInstID]);
+                if(GL_TRUE == IsAlu(texcoord_DepInst->Opcode) )
+                {
+                    pInstDeps[nDepInstID].nDstDep = i;
+                    pInstDeps[i].nDstDep = i;
+                }
+                else if(GL_TRUE == IsTex(texcoord_DepInst->Opcode) )
+                {
+                    pInstDeps[i].nDstDep = i;
+                }
+                else
+                {   //... other deps?
+                }
+            }
+        }
+	}
+
+    return GL_TRUE;
+}
+
+GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
+							     struct gl_fragment_program   *mesa_fp)
+{
+	GLuint    number_of_colors_exported;
+	GLboolean z_enabled = GL_FALSE;
+	GLuint    unBit;
+
+    //Init_Program
+	Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) );
+	Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp);
+
+    if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) )
+	{
+		return GL_FALSE;
+    }
+	
+	if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions,
+                                  &(mesa_fp->Base.Instructions[0]), 
+                                  &(fp->r700AsmCode)) )
+	{
+		return GL_FALSE;
+	}
+
+    if(GL_FALSE == Process_Fragment_Exports(&(fp->r700AsmCode), mesa_fp->Base.OutputsWritten) )
+    {
+        return GL_FALSE;
+    }
+
+    fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 
+                         : (fp->r700AsmCode.number_used_registers - 1);
+
+	fp->r700Shader.nParamExports = fp->r700AsmCode.number_of_exports;
+
+	number_of_colors_exported = fp->r700AsmCode.number_of_colorandz_exports;
+
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+		z_enabled = GL_TRUE;
+		number_of_colors_exported--;
+	}
+
+	fp->r700Shader.exportMode = number_of_colors_exported << 1 | z_enabled;
+
+    fp->translated = GL_TRUE;
+
+	return GL_TRUE;
+}
+
+void r700SelectFragmentShader(GLcontext *ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	    (ctx->FragmentProgram._Current);
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+	    fp->r700AsmCode.bR6xx = 1;
+    }
+
+    if (GL_FALSE == fp->translated)
+	    r700TranslateFragmentShader(fp, &(fp->mesa_program));
+}
+
+void * r700GetActiveFpShaderBo(GLcontext * ctx)
+{
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	                                   (ctx->FragmentProgram._Current);
+
+    return fp->shaderbo;
+}
+
+GLboolean r700SetupFragmentProgram(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	                                   (ctx->FragmentProgram._Current);
+    r700_AssemblerBase         *pAsm = &(fp->r700AsmCode);
+    struct gl_fragment_program *mesa_fp = &(fp->mesa_program);
+    struct gl_program_parameter_list *paramList;
+    unsigned int unNumParamData;
+    unsigned int ui, i;
+    unsigned int unNumOfReg;
+    unsigned int unBit;
+    GLuint exportCount;
+
+    if(GL_FALSE == fp->loaded)
+    {
+	    if(fp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    {
+		    Assemble( &(fp->r700Shader) );
+	    }
+
+        /* Load fp to gpu */
+        r600EmitShader(ctx,
+                       &(fp->shaderbo),
+                       (GLvoid *)(fp->r700Shader.pProgram),
+                       fp->r700Shader.uShaderBinaryDWORDSize,
+                       "FS");
+
+        fp->loaded = GL_TRUE;
+    }
+
+    DumpHwBinary(DUMP_PIXEL_SHADER, (GLvoid *)(fp->r700Shader.pProgram),
+                 fp->r700Shader.uShaderBinaryDWORDSize);
+
+    /* TODO : enable this after MemUse fixed *=
+    (context->chipobj.MemUse)(context, fp->shadercode.buf->id);
+    */
+
+    R600_STATECHANGE(context, ps);
+
+    r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0;
+    SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
+
+    r700->ps.SQ_PGM_START_PS.u32All = 0; /* set from buffer obj */
+
+    R600_STATECHANGE(context, spi);
+
+    unNumOfReg = fp->r700Shader.nRegs + 1;
+
+    ui = (r700->SPI_PS_IN_CONTROL_0.u32All & NUM_INTERP_mask) / (1 << NUM_INTERP_shift);
+
+    /* PS uses fragment.position */
+    if (mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS))
+    {
+        ui += 1;
+        SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask);
+        SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, CENTERS_ONLY, BARYC_SAMPLE_CNTL_shift, BARYC_SAMPLE_CNTL_mask);
+        SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, POSITION_ENA_bit);
+        SETbit(r700->SPI_INPUT_Z.u32All, PROVIDE_Z_TO_SPI_bit);
+    }
+
+    ui = (unNumOfReg < ui) ? ui : unNumOfReg;
+
+    SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask);
+
+    CLEARbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, UNCACHED_FIRST_INST_bit);
+
+    if(fp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
+	{
+        SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, fp->r700Shader.uStackSize,
+                 STACK_SIZE_shift, STACK_SIZE_mask);
+    }
+
+    SETfield(r700->ps.SQ_PGM_EXPORTS_PS.u32All, fp->r700Shader.exportMode,
+             EXPORT_MODE_shift, EXPORT_MODE_mask);
+
+    R600_STATECHANGE(context, db);
+
+    if(fp->r700Shader.killIsUsed)
+    {
+	    SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit);
+    }
+    else
+    {
+        CLEARbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit);
+    }
+
+    if(fp->r700Shader.depthIsExported)
+    {
+	    SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit);
+    }
+    else
+    {
+        CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit);
+    }
+
+    // emit ps input map
+    unBit = 1 << FRAG_ATTRIB_WPOS;
+    if(mesa_fp->Base.InputsRead & unBit)
+    {
+            ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_WPOS];
+            SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit);
+            SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui,
+                     SEMANTIC_shift, SEMANTIC_mask);
+            if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit)
+                    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+            else
+                    CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+    }
+
+    unBit = 1 << FRAG_ATTRIB_COL0;
+    if(mesa_fp->Base.InputsRead & unBit)
+    {
+	    ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0];
+	    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit);
+	    SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui,
+		     SEMANTIC_shift, SEMANTIC_mask);
+	    if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit)
+		    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+	    else
+		    CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+    }
+
+    unBit = 1 << FRAG_ATTRIB_COL1;
+    if(mesa_fp->Base.InputsRead & unBit)
+    {
+	    ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1];
+	    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit);
+	    SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui,
+		     SEMANTIC_shift, SEMANTIC_mask);
+	    if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit)
+		    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+	    else
+		    CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+    }
+
+    unBit = 1 << FRAG_ATTRIB_FOGC;
+    if(mesa_fp->Base.InputsRead & unBit)
+    {
+            ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC];
+            SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit);
+            SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui,
+                     SEMANTIC_shift, SEMANTIC_mask);
+            if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit)
+                    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+            else
+                    CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+    }
+
+    for(i=0; i<8; i++)
+    {
+	    unBit = 1 << (FRAG_ATTRIB_TEX0 + i);
+	    if(mesa_fp->Base.InputsRead & unBit)
+	    {
+		    ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i];
+		    SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit);
+		    SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui,
+			     SEMANTIC_shift, SEMANTIC_mask);
+		    CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit);
+	    }
+    }
+
+    R600_STATECHANGE(context, cb);
+    exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
+    r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
+
+    /* sent out shader constants. */
+    paramList = fp->mesa_program.Base.Parameters;
+
+    if(NULL != paramList) {
+	    _mesa_load_state_parameters(ctx, paramList);
+
+	    if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
+		    return GL_FALSE;
+
+	    R600_STATECHANGE(context, ps_consts);
+
+	    r700->ps.num_consts = paramList->NumParameters;
+
+	    unNumParamData = paramList->NumParameters;
+
+	    for(ui=0; ui<unNumParamData; ui++) {
+		    r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0];
+		    r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1];
+		    r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2];
+		    r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3];
+	    }
+    } else
+	    r700->ps.num_consts = 0;
+
+    return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h
new file mode 100644
index 0000000000..cbb108d212
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.h
@@ -0,0 +1,66 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_FRAGPROG_H_
+#define _R700_FRAGPROG_H_
+
+#include "r600_context.h"
+#include "r700_assembler.h"
+
+struct r700_fragment_program
+{
+	struct gl_fragment_program mesa_program;
+
+    r700_AssemblerBase r700AsmCode;
+	R700_Shader        r700Shader;
+
+	GLboolean translated;
+    GLboolean loaded;
+	GLboolean error;
+
+    void * shaderbo;
+
+	GLboolean WritesDepth;
+	GLuint optimization;
+};
+
+/* Internal */
+void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
+			  struct gl_fragment_program *mesa_fp);
+GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
+					   struct gl_fragment_program   *mesa_fp);
+
+GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
+				      struct gl_fragment_program   *mesa_vp);
+
+/* Interface */
+extern void r700SelectFragmentShader(GLcontext *ctx);
+
+extern GLboolean r700SetupFragmentProgram(GLcontext * ctx);
+
+extern void *    r700GetActiveFpShaderBo(GLcontext * ctx);
+
+#endif /*_R700_FRAGPROG_H_*/
diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c
new file mode 100644
index 0000000000..72a8978976
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_ioctl.c
@@ -0,0 +1,50 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <sched.h>
+#include <errno.h>
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/simple_list.h"
+#include "swrast/swrast.h"
+
+#include "radeon_common.h"
+#include "radeon_lock.h"
+#include "r600_context.h"
+
+#include "r700_ioctl.h"
+#include "r700_clear.h"
+
+
+void r700InitIoctlFuncs(struct dd_function_table *functions)
+{
+	functions->Clear = r700Clear;
+	functions->Finish = radeonFinish;
+	functions->Flush = radeonFlush;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.h b/src/mesa/drivers/dri/r600/r700_ioctl.h
new file mode 100644
index 0000000000..414dc3e23e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_ioctl.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef __R700_IOCTL_H__
+#define __R700_IOCTL_H__
+
+#include "r600_context.h"
+#include "radeon_drm.h"
+
+extern void r700InitIoctlFuncs(struct dd_function_table *functions);
+
+#endif				/* __R700_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c
new file mode 100644
index 0000000000..3c8c1fd7a3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_oglprog.c
@@ -0,0 +1,147 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <string.h>
+
+#include "main/glheader.h"
+#include "main/imports.h"
+
+#include "shader/program.h"
+#include "tnl/tnl.h"
+
+#include "r600_context.h"
+#include "r600_emit.h"
+
+#include "r700_oglprog.h"
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+
+
+static struct gl_program *r700NewProgram(GLcontext * ctx, 
+                                         GLenum target,
+					                     GLuint id)
+{
+	struct gl_program *pProgram = NULL;
+
+    struct r700_vertex_program *vp;
+	struct r700_fragment_program *fp;
+
+	radeon_print(RADEON_SHADER, RADEON_VERBOSE,
+			"%s %u, %u\n", __func__, target, id);
+
+    switch (target) 
+    {
+    case GL_VERTEX_STATE_PROGRAM_NV:
+    case GL_VERTEX_PROGRAM_ARB:	    
+        vp       = CALLOC_STRUCT(r700_vertex_program);
+	    pProgram = _mesa_init_vertex_program(ctx, 
+                                             &vp->mesa_program,
+					                         target, 
+                                             id);
+        vp->translated = GL_FALSE;
+        vp->loaded     = GL_FALSE;
+ 
+        vp->shaderbo   = NULL;
+
+	    break;
+    case GL_FRAGMENT_PROGRAM_NV:
+    case GL_FRAGMENT_PROGRAM_ARB:
+		fp       = CALLOC_STRUCT(r700_fragment_program);
+		pProgram = _mesa_init_fragment_program(ctx, 
+                                               &fp->mesa_program,
+						                       target, 
+                                               id);
+        fp->translated = GL_FALSE;
+        fp->loaded     = GL_FALSE;
+
+        fp->shaderbo   = NULL;
+
+	    break;
+    default:
+	    _mesa_problem(ctx, "Bad target in r700NewProgram");
+    }
+
+	return pProgram;
+}
+
+static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog)
+{
+    struct r700_vertex_program   * vp;
+    struct r700_fragment_program * fp;
+
+	radeon_print(RADEON_SHADER, RADEON_VERBOSE,
+			"%s %p\n", __func__, prog);
+
+    switch (prog->Target) 
+    {
+    case GL_VERTEX_STATE_PROGRAM_NV:
+    case GL_VERTEX_PROGRAM_ARB:	    
+        vp = (struct r700_vertex_program*)prog;
+        /* Release DMA region */
+
+        r600DeleteShader(ctx, vp->shaderbo);
+
+        /* Clean up */
+        Clean_Up_Assembler(&(vp->r700AsmCode));
+        Clean_Up_Shader(&(vp->r700Shader));
+	    break;
+    case GL_FRAGMENT_PROGRAM_NV:
+    case GL_FRAGMENT_PROGRAM_ARB:
+		fp = (struct r700_fragment_program*)prog;
+        /* Release DMA region */
+
+        r600DeleteShader(ctx, fp->shaderbo);
+
+        /* Clean up */
+        Clean_Up_Assembler(&(fp->r700AsmCode));
+        Clean_Up_Shader(&(fp->r700Shader));
+	    break;
+    default:
+	    _mesa_problem(ctx, "Bad target in r700NewProgram");
+    }
+
+	_mesa_delete_program(ctx, prog);
+}
+
+static void
+r700ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
+{
+
+}
+
+static GLboolean r700IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
+{
+
+	return GL_TRUE;
+}
+
+void r700InitShaderFuncs(struct dd_function_table *functions)
+{
+	functions->NewProgram = r700NewProgram;
+	functions->DeleteProgram = r700DeleteProgram;
+	functions->ProgramStringNotify = r700ProgramStringNotify;
+	functions->IsProgramNative = r700IsProgramNative;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.h b/src/mesa/drivers/dri/r600/r700_oglprog.h
new file mode 100644
index 0000000000..fe2e9d1974
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_oglprog.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_OGLPROG_H_
+#define _R700_OGLPROG_H_
+#include "r600_context.h"
+
+extern void r700InitShaderFuncs(struct dd_function_table *functions);
+
+#endif /*_R700_OGLPROG_H_*/
diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c
new file mode 100644
index 0000000000..3566bf3ca7
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_render.c
@@ -0,0 +1,479 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/enums.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/dd.h"
+#include "main/simple_list.h"
+#include "main/api_arrayelt.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "vbo/vbo.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "tnl/t_context.h"
+#include "tnl/t_vertex.h"
+#include "tnl/t_pipeline.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r600_tex.h"
+
+#include "r700_vertprog.h"
+#include "r700_fragprog.h"
+#include "r700_state.h"
+
+#include "radeon_common_context.h"
+
+void r700WaitForIdle(context_t *context);
+void r700WaitForIdleClean(context_t *context);
+GLboolean r700SendTextureState(context_t *context);
+static unsigned int r700PrimitiveType(int prim);
+void r600UpdateTextureState(GLcontext * ctx);
+GLboolean r700SyncSurf(context_t *context,
+		       struct radeon_bo *pbo,
+		       uint32_t read_domain,
+		       uint32_t write_domain,
+		       uint32_t sync_type);
+
+void r700WaitForIdle(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    radeon_print(RADEON_RENDER | RADEON_STATE, RADEON_TRACE, "%s\n", __func__);
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+    R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
+    R600_OUT_BATCH(WAIT_3D_IDLE_bit);
+
+    END_BATCH();
+    COMMIT_BATCH();
+}
+
+void r700WaitForIdleClean(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    radeon_print(RADEON_RENDER | RADEON_STATE, RADEON_TRACE, "%s\n", __func__);
+    BEGIN_BATCH_NO_AUTOSTATE(5);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
+    R600_OUT_BATCH(CACHE_FLUSH_AND_INV_EVENT);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+    R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
+    R600_OUT_BATCH(WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
+
+    END_BATCH();
+    COMMIT_BATCH();
+}
+
+void r700Start3D(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    radeon_print(RADEON_RENDER | RADEON_STATE, RADEON_TRACE, "%s\n", __func__);
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+        BEGIN_BATCH_NO_AUTOSTATE(2);
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 0));
+        R600_OUT_BATCH(0);
+        END_BATCH();
+    }
+
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1));
+    R600_OUT_BATCH(0x80000000);
+    R600_OUT_BATCH(0x80000000);
+    END_BATCH();
+
+    COMMIT_BATCH();
+
+    r700WaitForIdleClean(context);
+}
+
+GLboolean r700SyncSurf(context_t *context,
+		       struct radeon_bo *pbo,
+		       uint32_t read_domain,
+		       uint32_t write_domain,
+		       uint32_t sync_type)
+{
+    BATCH_LOCALS(&context->radeon);
+    radeon_print(RADEON_RENDER | RADEON_STATE, RADEON_TRACE, "%s\n", __func__);
+    uint32_t cp_coher_size;
+
+    if (!pbo)
+	    return GL_FALSE;
+
+    if (pbo->size == 0xffffffff)
+	    cp_coher_size = 0xffffffff;
+    else
+	    cp_coher_size = ((pbo->size + 255) >> 8);
+
+    BEGIN_BATCH_NO_AUTOSTATE(5 + 2);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_SYNC, 3));
+    R600_OUT_BATCH(sync_type);
+    R600_OUT_BATCH(cp_coher_size);
+    R600_OUT_BATCH(0);
+    R600_OUT_BATCH(10);
+    R600_OUT_BATCH_RELOC(0,
+			 pbo,
+			 0,
+			 read_domain, write_domain, 0);
+    END_BATCH();
+    COMMIT_BATCH();
+
+    return GL_TRUE;
+}
+
+static unsigned int r700PrimitiveType(int prim)
+{
+    switch (prim & PRIM_MODE_MASK)
+    {
+    case GL_POINTS:
+        return DI_PT_POINTLIST;
+        break;
+    case GL_LINES:
+        return DI_PT_LINELIST;
+        break;
+    case GL_LINE_STRIP:
+        return DI_PT_LINESTRIP;
+        break;
+    case GL_LINE_LOOP:
+        return DI_PT_LINELOOP;
+        break;
+    case GL_TRIANGLES:
+        return DI_PT_TRILIST;
+        break;
+    case GL_TRIANGLE_STRIP:
+        return DI_PT_TRISTRIP;
+        break;
+    case GL_TRIANGLE_FAN:
+        return DI_PT_TRIFAN;
+        break;
+    case GL_QUADS:
+        return DI_PT_QUADLIST;
+        break;
+    case GL_QUAD_STRIP:
+        return DI_PT_QUADSTRIP;
+        break;
+    case GL_POLYGON:
+        return DI_PT_POLYGON;
+        break;
+    default:
+        assert(0);
+        return -1;
+        break;
+    }
+}
+
+static int r700NumVerts(int num_verts, int prim)
+{
+	int verts_off = 0;
+
+	switch (prim & PRIM_MODE_MASK) {
+	case GL_POINTS:
+		verts_off = 0;
+		break;
+	case GL_LINES:
+		verts_off = num_verts % 2;
+		break;
+	case GL_LINE_STRIP:
+		if (num_verts < 2)
+			verts_off = num_verts;
+		break;
+	case GL_LINE_LOOP:
+		if (num_verts < 2)
+			verts_off = num_verts;
+		break;
+	case GL_TRIANGLES:
+		verts_off = num_verts % 3;
+		break;
+	case GL_TRIANGLE_STRIP:
+		if (num_verts < 3)
+			verts_off = num_verts;
+		break;
+	case GL_TRIANGLE_FAN:
+		if (num_verts < 3)
+			verts_off = num_verts;
+		break;
+	case GL_QUADS:
+		verts_off = num_verts % 4;
+		break;
+	case GL_QUAD_STRIP:
+		if (num_verts < 4)
+			verts_off = num_verts;
+		else
+			verts_off = num_verts % 2;
+		break;
+	case GL_POLYGON:
+		if (num_verts < 3)
+			verts_off = num_verts;
+		break;
+	default:
+		assert(0);
+		return -1;
+		break;
+	}
+
+	return num_verts - verts_off;
+}
+
+static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	BATCH_LOCALS(&context->radeon);
+	int type, i, total_emit;
+	int num_indices;
+	uint32_t vgt_draw_initiator = 0;
+	uint32_t vgt_index_type     = 0;
+	uint32_t vgt_primitive_type = 0;
+	uint32_t vgt_num_indices    = 0;
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct vertex_buffer *vb = &tnl->vb;
+
+	type = r700PrimitiveType(prim);
+	num_indices = r700NumVerts(end - start, prim);
+
+	radeon_print(RADEON_RENDER, RADEON_TRACE,
+		"%s type %x num_indices %d\n",
+		__func__, type, num_indices);
+
+	if (type < 0 || num_indices <= 0)
+		return;
+
+        total_emit =   3 /* VGT_PRIMITIVE_TYPE */
+		     + 2 /* VGT_INDEX_TYPE */
+		     + 2 /* NUM_INSTANCES */
+                     + num_indices + 3; /* DRAW_INDEX_IMMD */
+
+        BEGIN_BATCH_NO_AUTOSTATE(total_emit);
+	// prim
+        SETfield(vgt_primitive_type, type,
+		 VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask);
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+        R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX);
+        R600_OUT_BATCH(vgt_primitive_type);
+
+	// index type
+        SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask);
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
+        R600_OUT_BATCH(vgt_index_type);
+
+	// num instances
+	R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
+        R600_OUT_BATCH(1);
+
+	// draw packet
+        vgt_num_indices = num_indices;
+        SETfield(vgt_draw_initiator, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask);
+	SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask);
+
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (num_indices + 1)));
+        R600_OUT_BATCH(vgt_num_indices);
+        R600_OUT_BATCH(vgt_draw_initiator);
+
+        for (i = start; i < (start + num_indices); i++) {
+		if(vb->Elts)
+			R600_OUT_BATCH(vb->Elts[i]);
+		else
+			R600_OUT_BATCH(i);
+        }
+        END_BATCH();
+        COMMIT_BATCH();
+
+}
+
+/* start 3d, idle, cb/db flush */
+#define PRE_EMIT_STATE_BUFSZ 10 + 5 + 14
+
+static GLuint r700PredictRenderSize(GLcontext* ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct r700_vertex_program *vpc
+        = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    struct vertex_buffer *vb = &tnl->vb;
+    GLboolean flushed;
+    GLuint dwords, i;
+    GLuint state_size;
+    /* pre calculate aos count so state prediction works */
+    context->radeon.tcl.aos_count = _mesa_bitcount(vpc->mesa_program.Base.InputsRead);
+
+    dwords = PRE_EMIT_STATE_BUFSZ;
+    for (i = 0; i < vb->PrimitiveCount; i++)
+        dwords += vb->Primitive[i].count + 10;
+    state_size = radeonCountStateEmitSize(&context->radeon);
+    flushed = rcommonEnsureCmdBufSpace(&context->radeon,
+            dwords + state_size, __FUNCTION__);
+
+    if (flushed)
+        dwords += radeonCountStateEmitSize(&context->radeon);
+    else
+        dwords += state_size;
+
+    radeon_print(RADEON_RENDER, RADEON_VERBOSE,
+	"%s: total prediction size is %d.\n", __FUNCTION__, dwords);
+    return dwords;
+}
+
+static GLboolean r700RunRender(GLcontext * ctx,
+			       struct tnl_pipeline_stage *stage)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    radeonContextPtr radeon = &context->radeon;
+    unsigned int i, id = 0;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *vb = &tnl->vb;
+    struct radeon_renderbuffer *rrb;
+
+    radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s: cs begin at %d\n",
+                __func__, context->radeon.cmdbuf.cs->cdw);
+
+    /* always emit CB base to prevent
+     * lock ups on some chips.
+     */
+    R600_STATECHANGE(context, cb_target);
+    /* mark vtx as dirty since it changes per-draw */
+    R600_STATECHANGE(context, vtx);
+
+    r700UpdateShaders(ctx);
+    r700SetScissor(context);
+    r700SetupVertexProgram(ctx);
+    r700SetupFragmentProgram(ctx);
+    r600UpdateTextureState(ctx);
+
+    GLuint emit_end = r700PredictRenderSize(ctx) 
+        + context->radeon.cmdbuf.cs->cdw;
+    r700SetupStreams(ctx);
+
+    radeonEmitState(radeon);
+
+    radeon_debug_add_indent();
+    /* richard test code */
+    for (i = 0; i < vb->PrimitiveCount; i++) {
+        GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
+        GLuint start = vb->Primitive[i].start;
+        GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
+        r700RunRenderPrimitive(ctx, start, end, prim);
+    }
+    radeon_debug_remove_indent();
+
+    /* Flush render op cached for last several quads. */
+    r700WaitForIdleClean(context);
+
+    rrb = radeon_get_colorbuffer(&context->radeon);
+    if (rrb && rrb->bo)
+	    r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
+			 CB_ACTION_ENA_bit | (1 << (id + 6)));
+
+    rrb = radeon_get_depthbuffer(&context->radeon);
+    if (rrb && rrb->bo)
+	    r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
+			 DB_ACTION_ENA_bit | DB_DEST_BASE_ENA_bit);
+
+    radeonReleaseArrays(ctx, ~0);
+
+    radeon_print(RADEON_RENDER, RADEON_TRACE, "%s: cs end at %d\n",
+                __func__, context->radeon.cmdbuf.cs->cdw);
+
+    if ( emit_end < context->radeon.cmdbuf.cs->cdw )
+       WARN_ONCE("Rendering was %d commands larger than predicted size."
+	       " We might overflow  command buffer.\n", context->radeon.cmdbuf.cs->cdw - emit_end);
+
+    return GL_FALSE;
+}
+
+static GLboolean r700RunNonTCLRender(GLcontext * ctx,
+				     struct tnl_pipeline_stage *stage) /* -------------------- */
+{
+	GLboolean bRet = GL_TRUE;
+	
+	return bRet;
+}
+
+static GLboolean r700RunTCLRender(GLcontext * ctx,  /*----------------------*/
+				  struct tnl_pipeline_stage *stage)
+{
+	GLboolean bRet = GL_FALSE;
+
+    /* TODO : sw fallback */
+
+    /**
+    * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+    */
+    if(!r600ValidateBuffers(ctx))
+    {
+        return GL_TRUE;
+    }
+
+    bRet = r700RunRender(ctx, stage);
+
+    return bRet;
+	//GL_FALSE will stop to do other pipe stage in _tnl_run_pipeline
+    //The render here DOES finish the whole pipe, so GL_FALSE should be returned for success.
+}
+
+const struct tnl_pipeline_stage _r700_render_stage = {
+	"r700 Hardware Rasterization",
+	NULL,
+	NULL,
+	NULL,
+	NULL,
+	r700RunNonTCLRender
+};
+
+const struct tnl_pipeline_stage _r700_tcl_stage = {
+	"r700 Hardware Transform, Clipping and Lighting",
+	NULL,
+	NULL,
+	NULL,
+	NULL,
+	r700RunTCLRender
+};
+
+const struct tnl_pipeline_stage *r700_pipeline[] = 
+{
+    &_r700_tcl_stage,
+    &_tnl_vertex_transform_stage,
+	&_tnl_normal_transform_stage,
+	&_tnl_lighting_stage,
+	&_tnl_fog_coordinate_stage,
+	&_tnl_texgen_stage,
+	&_tnl_texture_transform_stage,
+	&_tnl_vertex_program_stage,
+
+    &_r700_render_stage,
+    &_tnl_render_stage,
+    0,
+};
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c
new file mode 100644
index 0000000000..b4fd51c137
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shader.c
@@ -0,0 +1,527 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+
+#include "main/glheader.h"
+
+#include "r600_context.h"
+#include "r700_debug.h"
+
+#include "r700_shader.h"
+
+void r700ShaderInit(GLcontext * ctx)
+{
+}
+
+void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst)
+{
+	if(NULL == plstCFInstructions->pTail)
+	{	//first
+		plstCFInstructions->pHead = pInst;
+		plstCFInstructions->pTail = pInst;
+	}
+	else
+	{
+		plstCFInstructions->pTail->pNextInst = pInst;
+		plstCFInstructions->pTail = pInst;
+	}
+	pInst->pNextInst = NULL;
+
+	plstCFInstructions->uNumOfNode++;
+}
+
+void Init_R700_Shader(R700_Shader * pShader)
+{
+	pShader->Type = R700_SHADER_INVALID;
+	pShader->pProgram = NULL;
+	pShader->bBinaryShader = GL_FALSE;
+	pShader->bFetchShaderRequired = GL_FALSE;
+	pShader->bNeedsAssembly = GL_FALSE;
+	pShader->bLinksDirty = GL_TRUE;
+	pShader->uShaderBinaryDWORDSize = 0;
+	pShader->nRegs = 0;
+	pShader->nParamExports = 0;
+	pShader->nMemExports = 0;
+	pShader->resource = 0;
+
+	pShader->exportMode = 0;
+	pShader->depthIsImported = GL_FALSE;
+
+	pShader->positionVectorIsExported = GL_FALSE;
+	pShader->miscVectorIsExported = GL_FALSE;
+	pShader->renderTargetArrayIndexIsExported = GL_FALSE;
+	pShader->ccDist0VectorIsExported = GL_FALSE;
+	pShader->ccDist1VectorIsExported = GL_FALSE; 
+
+
+	pShader->depthIsExported = GL_FALSE;
+	pShader->stencilRefIsExported = GL_FALSE;
+	pShader->coverageToMaskIsExported = GL_FALSE;
+	pShader->maskIsExported = GL_FALSE;
+	pShader->killIsUsed = GL_FALSE;
+
+	pShader->uCFOffset = 0;
+	pShader->uStackSize = 0;
+	pShader->uMaxCallDepth = 0;
+
+	pShader->bSurfAllocated = GL_FALSE;
+	
+	pShader->lstCFInstructions.pHead=NULL;  
+	pShader->lstCFInstructions.pTail=NULL;  
+	pShader->lstCFInstructions.uNumOfNode=0;
+	pShader->lstALUInstructions.pHead=NULL; 
+	pShader->lstALUInstructions.pTail=NULL; 
+	pShader->lstALUInstructions.uNumOfNode=0;
+	pShader->lstTEXInstructions.pHead=NULL; 
+	pShader->lstTEXInstructions.pTail=NULL; 
+	pShader->lstTEXInstructions.uNumOfNode=0;
+	pShader->lstVTXInstructions.pHead=NULL; 
+	pShader->lstVTXInstructions.pTail=NULL; 
+	pShader->lstVTXInstructions.uNumOfNode=0;
+}
+
+void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst)
+{
+    R700ControlFlowSXClause*  pSXClause; 
+    R700ControlFlowSMXClause* pSMXClause;
+
+    pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode;
+    AddInstToList(&(pShader->lstCFInstructions), 
+                  (R700ShaderInstruction*)pCFInst);
+    pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType);
+
+    pSXClause = NULL;
+    pSMXClause = NULL; 
+	switch (pCFInst->m_ShaderInstType)
+	{
+	case SIT_CF_ALL_EXP_SX:
+		pSXClause =  (R700ControlFlowSXClause*)pCFInst;
+		break;
+	case SIT_CF_ALL_EXP_SMX:
+		pSMXClause = (R700ControlFlowSMXClause*)pCFInst;
+		break;
+	default:
+		break;
+	};
+
+    if((pSXClause != NULL) && (pSXClause->m_Word0.f.type == SQ_EXPORT_PARAM))
+    {
+        pShader->nParamExports += pSXClause->m_Word1.f.burst_count + 1;
+    }
+    else if ((pSMXClause != NULL) && (pSMXClause->m_Word1.f.cf_inst == SQ_CF_INST_MEM_RING) &&
+            (pSMXClause->m_Word0.f.type == SQ_EXPORT_WRITE || pSMXClause->m_Word0.f.type == SQ_EXPORT_WRITE_IND))
+    {
+        pShader->nMemExports += pSMXClause->m_Word1.f.burst_count + 1;
+    }
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pCFInst->useCount++;
+}
+
+void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst)
+{
+    pVTXInst->m_uIndex = pShader->lstVTXInstructions.uNumOfNode;
+	AddInstToList(&(pShader->lstVTXInstructions), 
+                  (R700ShaderInstruction*)pVTXInst);
+	pShader->uShaderBinaryDWORDSize += GetInstructionSize(pVTXInst->m_ShaderInstType);
+
+	if(pVTXInst->m_ShaderInstType == SIT_VTX_GENERIC)
+	{
+		R700VertexGenericFetch* pVTXGenericClause = (R700VertexGenericFetch*)pVTXInst;	
+		pShader->nRegs = (pShader->nRegs < pVTXGenericClause->m_Word1_GPR.f.dst_gpr) ? pVTXGenericClause->m_Word1_GPR.f.dst_gpr : pShader->nRegs;
+	}
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pVTXInst->useCount++;
+}
+
+void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst)
+{
+    pTEXInst->m_uIndex = pShader->lstTEXInstructions.uNumOfNode;
+	AddInstToList(&(pShader->lstTEXInstructions), 
+                  (R700ShaderInstruction*)pTEXInst);
+	pShader->uShaderBinaryDWORDSize += GetInstructionSize(pTEXInst->m_ShaderInstType);
+
+    pShader->nRegs = (pShader->nRegs < pTEXInst->m_Word1.f.dst_gpr) ? pTEXInst->m_Word1.f.dst_gpr : pShader->nRegs;
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pTEXInst->useCount++;
+}
+
+void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst)
+{
+    pALUInst->m_uIndex = pShader->lstALUInstructions.uNumOfNode;
+    AddInstToList(&(pShader->lstALUInstructions), 
+                  (R700ShaderInstruction*)pALUInst);
+    pShader->uShaderBinaryDWORDSize += GetInstructionSize(pALUInst->m_ShaderInstType);
+
+    pShader->nRegs = (pShader->nRegs < pALUInst->m_Word1.f.dst_gpr) ? pALUInst->m_Word1.f.dst_gpr : pShader->nRegs;
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pALUInst->useCount++;
+}
+
+void ResolveLinks(R700_Shader *pShader)
+{
+    GLuint uiSize;
+    R700ShaderInstruction  *pInst;
+    R700ALUInstruction     *pALUinst;
+    R700TextureInstruction *pTEXinst;
+    R700VertexInstruction  *pVTXinst; 
+
+    GLuint vtxOffset;
+
+	GLuint cfOffset = 0x0;  
+
+    GLuint aluOffset = cfOffset + pShader->lstCFInstructions.uNumOfNode * GetInstructionSize(SIT_CF);
+
+    GLuint texOffset = aluOffset;  // + m_lstALUInstructions.size() * R700ALUInstruction::SIZE,
+
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        texOffset += GetInstructionSize(pInst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+  
+    vtxOffset = texOffset + pShader->lstTEXInstructions.uNumOfNode * GetInstructionSize(SIT_TEX);
+
+    if ( ((pShader->lstTEXInstructions.uNumOfNode > 0) && (texOffset % 4 != 0)) || 
+         ((pShader->lstVTXInstructions.uNumOfNode > 0) && (vtxOffset % 4 != 0))    )
+    {
+        pALUinst = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction);
+        Init_R700ALUInstruction(pALUinst);
+        AddALUInstruction(pShader, pALUinst);
+        texOffset += GetInstructionSize(SIT_ALU);
+        vtxOffset += GetInstructionSize(SIT_ALU);
+    }
+
+    pInst  = pShader->lstALUInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pALUinst = (R700ALUInstruction*)pInst;
+
+        if(pALUinst->m_pLinkedALUClause != NULL)
+        {
+            // This address is quad-word aligned
+            pALUinst->m_pLinkedALUClause->m_Word0.f.addr = (aluOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pALUinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pInst  = pShader->lstTEXInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pTEXinst = (R700TextureInstruction*)pInst;
+
+        if (pTEXinst->m_pLinkedGenericClause != NULL)
+        {
+            pTEXinst->m_pLinkedGenericClause->m_Word0.f.addr = (texOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pTEXinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pInst  = pShader->lstVTXInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pVTXinst = (R700VertexInstruction*)pInst;
+
+        if (pVTXinst->m_pLinkedGenericClause != NULL)
+        {
+            pVTXinst->m_pLinkedGenericClause->m_Word0.f.addr = (vtxOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pVTXinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pShader->bLinksDirty = GL_FALSE;
+}
+
+void Assemble(R700_Shader *pShader)
+{
+	GLuint i;
+    GLuint *pShaderBinary;
+    GLuint size_of_program;
+    GLuint *pCurrPos;
+
+    GLuint end_of_cf_instructions;
+    GLuint number_of_alu_dwords;
+
+    R700ShaderInstruction  *pInst;
+
+    if(GL_TRUE == pShader->bBinaryShader)
+    {
+        return;
+    }
+
+    if(pShader->bLinksDirty == GL_TRUE) 
+    {
+        ResolveLinks(pShader);
+    }
+
+    size_of_program = pShader->uShaderBinaryDWORDSize;
+    
+    pShaderBinary = (GLuint*) MALLOC(sizeof(GLuint)*size_of_program);
+ 
+    pCurrPos = pShaderBinary;
+
+    for (i = 0; i < size_of_program; i++)
+    {
+        pShaderBinary[i] = 0;
+    }
+
+    pInst = pShader->lstCFInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_CF_GENERIC: 
+            {
+                R700ControlFlowGenericClause* pCFgeneric = (R700ControlFlowGenericClause*)pInst;
+                *pCurrPos++ = pCFgeneric->m_Word0.val;
+                *pCurrPos++ = pCFgeneric->m_Word1.val;
+            }
+            break;
+        case SIT_CF_ALU: 
+            {
+                R700ControlFlowALUClause* pCFalu = (R700ControlFlowALUClause*)pInst;
+                *pCurrPos++ = pCFalu->m_Word0.val;
+                *pCurrPos++ = pCFalu->m_Word1.val;
+            }
+            break;
+        case SIT_CF_ALL_EXP_SX: 
+            {
+                R700ControlFlowSXClause* pCFsx = (R700ControlFlowSXClause*)pInst;
+                *pCurrPos++ = pCFsx->m_Word0.val;
+                *pCurrPos++ = (pCFsx->m_Word1.val | pCFsx->m_Word1_SWIZ.val);
+            }
+            break;
+        case SIT_CF_ALL_EXP_SMX: 
+            {
+                R700ControlFlowSMXClause* pCFsmx = (R700ControlFlowSMXClause*)pInst;
+                *pCurrPos++ = pCFsmx->m_Word0.val;
+                *pCurrPos++ = (pCFsmx->m_Word1.val | pCFsmx->m_Word1_BUF.val);
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+    
+    number_of_alu_dwords = 0;
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_ALU: 
+            {
+                R700ALUInstruction* pALU = (R700ALUInstruction*)pInst;
+
+                *pCurrPos++ = pALU->m_Word0.val;
+                *pCurrPos++ = (pALU->m_Word1.val | pALU->m_Word1_OP2.val | pALU->m_Word1_OP3.val);
+
+                number_of_alu_dwords += 2;
+            }
+            break;
+        case SIT_ALU_HALF_LIT: 
+            {
+                R700ALUInstructionHalfLiteral* pALUhalf = (R700ALUInstructionHalfLiteral*)pInst;
+
+                *pCurrPos++ = pALUhalf->m_Word0.val;
+                *pCurrPos++ = (pALUhalf->m_Word1.val | pALUhalf->m_Word1_OP2.val | pALUhalf->m_Word1_OP3.val);
+                *pCurrPos++ = *((GLuint*)&(pALUhalf->m_fLiteralX));
+                *pCurrPos++ = *((GLuint*)&(pALUhalf->m_fLiteralY));
+
+                number_of_alu_dwords += 4;
+            }
+            break;
+        case SIT_ALU_FALL_LIT: 
+            {
+                R700ALUInstructionFullLiteral* pALUfull = (R700ALUInstructionFullLiteral*)pInst;
+
+                *pCurrPos++ = pALUfull->m_Word0.val;
+                *pCurrPos++ = (pALUfull->m_Word1.val | pALUfull->m_Word1_OP2.val | pALUfull->m_Word1_OP3.val);
+
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralX));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralY));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralZ));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralW));
+
+                number_of_alu_dwords += 6;
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+    
+    pInst = pShader->lstTEXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        R700TextureInstruction* pTEX = (R700TextureInstruction*)pInst;
+
+        *pCurrPos++ = pTEX->m_Word0.val;
+        *pCurrPos++ = pTEX->m_Word1.val;
+        *pCurrPos++ = pTEX->m_Word2.val;
+        *pCurrPos++ = 0x0beadeaf;
+
+        pInst = pInst->pNextInst;
+    };
+    
+    pInst = pShader->lstVTXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_VTX_SEM: //
+            {
+                R700VertexSemanticFetch* pVTXsem = (R700VertexSemanticFetch*)pInst;
+
+                *pCurrPos++ = pVTXsem->m_Word0.val;
+                *pCurrPos++ = (pVTXsem->m_Word1.val | pVTXsem->m_Word1_SEM.val);
+                *pCurrPos++ = pVTXsem->m_Word2.val;
+                *pCurrPos++ = 0x0beadeaf;
+            }
+            break;
+        case SIT_VTX_GENERIC: //
+            {
+                R700VertexGenericFetch* pVTXgeneric = (R700VertexGenericFetch*)pInst;
+
+                *pCurrPos++ = pVTXgeneric->m_Word0.val;
+                *pCurrPos++ = (pVTXgeneric->m_Word1.val | pVTXgeneric->m_Word1_GPR.val);
+                *pCurrPos++ = pVTXgeneric->m_Word2.val;
+                *pCurrPos++ = 0x0beadeaf;
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+
+    if(NULL != pShader->pProgram)
+    {
+        FREE(pShader->pProgram);
+    }
+    pShader->pProgram = (GLubyte*)pShaderBinary;
+
+    end_of_cf_instructions = pShader->uCFOffset + pShader->lstCFInstructions.uNumOfNode * GetInstructionSize(SIT_CF);
+    
+    pShader->uEndOfCF = end_of_cf_instructions >> 1;
+
+    pShader->uEndOfALU = (end_of_cf_instructions + number_of_alu_dwords) >> 1;
+
+    pShader->uEndOfFetch = (pShader->uCFOffset + pShader->uShaderBinaryDWORDSize) >> 1;
+
+    pShader->bNeedsAssembly = GL_FALSE;
+}
+
+void LoadProgram(R700_Shader *pShader) //context
+{
+}
+
+void UpdateShaderRegisters(R700_Shader *pShader) //context
+{
+}
+
+void DeleteInstructions(R700_Shader *pShader)
+{
+}
+
+void DebugPrint(void)
+{
+}
+
+void Clean_Up_Shader(R700_Shader *pShader)
+{
+    FREE(pShader->pProgram);
+
+    R700ShaderInstruction  *pInst;
+    R700ShaderInstruction  *pInstToFree;
+
+    pInst = pShader->lstCFInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstTEXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstVTXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h
new file mode 100644
index 0000000000..bfd01e1a93
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shader.h
@@ -0,0 +1,150 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef __R700_SHADER_H__
+#define __R700_SHADER_H__
+
+#include "main/mtypes.h"
+
+#include "r700_shaderinst.h"
+
+
+void r700ShaderInit(GLcontext * ctx);
+
+typedef enum R700ShaderType
+{
+    R700_SHADER_FS      = 0x0,
+    R700_SHADER_ES      = 0x1,
+    R700_SHADER_GS      = 0x2,
+    R700_SHADER_VS      = 0x3,
+    R700_SHADER_PS      = 0x4,
+    R700_SHADER_INVALID = 0x5,
+} R700ShaderType;
+
+typedef struct TypedShaderList 
+{
+	R700ShaderInstruction * pHead;
+	R700ShaderInstruction * pTail;
+	GLuint  uNumOfNode;
+} TypedShaderList;
+
+typedef struct RealRegister 
+{
+    GLuint uAddr;
+    GLuint uValue;
+} RealRegister;
+
+typedef struct InstDeps
+{
+    GLint nDstDep;
+    GLint nSrcDeps[3];
+} InstDeps;
+
+typedef struct R700_Shader 
+{
+	R700ShaderType   Type;
+
+    GLubyte*  pProgram;
+
+    GLboolean bBinaryShader;
+    GLboolean bFetchShaderRequired;
+    GLboolean bNeedsAssembly;
+    GLboolean bLinksDirty;
+
+    GLuint  uShaderBinaryDWORDSize; // in DWORDS
+    GLuint  nRegs;      
+    GLuint  nParamExports;   // VS_ EXPORT_COUNT (1 based, the actual register is 0 based!)
+    GLuint  nMemExports; 
+    GLuint  resource;     // VS and PS _RESOURCE
+    GLuint  exportMode;   // VS and PS _EXPORT_MODE
+
+    GLboolean  depthIsImported;             
+
+    // Vertex program exports
+    GLboolean  positionVectorIsExported;          
+
+    GLboolean  miscVectorIsExported;               
+    GLboolean  renderTargetArrayIndexIsExported;  
+
+    GLboolean  ccDist0VectorIsExported;  
+    GLboolean  ccDist1VectorIsExported;  
+
+    // Pixel program exports
+    GLboolean  depthIsExported;             
+    GLboolean  stencilRefIsExported;        
+    GLboolean  coverageToMaskIsExported;    
+    GLboolean  maskIsExported;              
+
+    GLboolean  killIsUsed;                  
+
+    GLuint  uStartAddr;
+    GLuint  uCFOffset;
+    GLuint  uEndOfCF;
+    GLuint  uEndOfALU;
+    GLuint  uEndOfFetch;
+    GLuint  uStackSize;
+    GLuint  uMaxCallDepth;
+
+	TypedShaderList lstCFInstructions;
+	TypedShaderList lstALUInstructions;
+	TypedShaderList lstTEXInstructions;
+	TypedShaderList lstVTXInstructions;
+
+    RealRegister RegStartAddr;
+    RealRegister RegCFOffset;
+    RealRegister RegEndCF;
+    RealRegister RegEndALU;
+    RealRegister egEndFetcg;
+
+	// -------- constants
+	GLfloat   ConstantArray[SQ_ALU_CONSTANT_PS_COUNT * 4];
+	
+	GLboolean bSurfAllocated;
+} R700_Shader;
+
+//Internal
+void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst);
+void ResolveLinks(R700_Shader *pShader);
+void Assemble(R700_Shader *pShader);
+
+
+//Interface
+void Init_R700_Shader(R700_Shader * pShader);
+void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst);
+void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst);
+void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst);
+void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst);
+
+void LoadProgram(R700_Shader *pShader);
+void UpdateShaderRegisters(R700_Shader *pShader);
+void DeleteInstructions(R700_Shader *pShader);
+void DebugPrint(void);
+
+void Clean_Up_Shader(R700_Shader *pShader);
+
+#endif /*__R700_SHADER_H__*/
+
diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.c b/src/mesa/drivers/dri/r600/r700_shaderinst.c
new file mode 100644
index 0000000000..f120d9f941
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shaderinst.c
@@ -0,0 +1,224 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#include "main/mtypes.h"
+
+#include "radeon_debug.h"
+#include "r700_shaderinst.h"
+
+void Init_R700ControlFlowGenericClause(R700ControlFlowGenericClause* pInst)
+{
+    pInst->m_Word0.val = 0x00000000;
+    pInst->m_Word1.val = 0x00000000;
+
+    pInst->m_pLinkedVTXInstruction = 0;
+    pInst->m_pLinkedTEXInstruction = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_GENERIC;
+}
+
+void Init_R700ControlFlowALUClause(R700ControlFlowALUClause* pInst)
+{
+    pInst->m_Word0.val = 0x00000000;
+    pInst->m_Word1.val = 0x00000000;
+
+    pInst->m_pLinkedALUInstruction = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALU;
+}
+
+void Init_R700ControlFlowSXClause(R700ControlFlowSXClause* pInst)
+{
+    pInst->m_Word0.val      = 0x00000000;
+    pInst->m_Word1.val      = 0x00000000;
+    pInst->m_Word1_SWIZ.val = 0x00000000;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALL_EXP_SX;
+}
+
+void Init_R700ControlFlowSMXClause(R700ControlFlowSMXClause* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_BUF.val = 0x00000000;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALL_EXP_SMX;
+}
+
+void Init_R700ALUInstruction(R700ALUInstruction* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+    pInst->m_pLinkedALUClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU;
+}
+
+void Init_R700ALUInstructionHalfLiteral(R700ALUInstructionHalfLiteral* pInst, GLfloat x, GLfloat y)
+{
+	pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+	pInst->m_pLinkedALUClause = 0;
+
+    pInst->m_fLiteralX = x;
+    pInst->m_fLiteralY = y;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU_HALF_LIT;
+}
+
+void Init_R700ALUInstructionFullLiteral(R700ALUInstructionFullLiteral* pInst, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
+{
+	pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+	pInst->m_pLinkedALUClause = 0;
+
+    pInst->m_fLiteralX = x;
+    pInst->m_fLiteralY = y;
+    pInst->m_fLiteralZ = z;
+    pInst->m_fLiteralW = w;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU_FALL_LIT;
+}
+
+void Init_R700TextureInstruction(R700TextureInstruction* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_TEX;
+}
+
+void Init_R700VertexSemanticFetch(R700VertexSemanticFetch* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_SEM.val = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_VTX_SEM;
+}
+
+void Init_R700VertexGenericFetch(R700VertexGenericFetch* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_GPR.val = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_VTX_GENERIC;
+}
+
+unsigned int GetInstructionSize(ShaderInstType instType)
+{
+    switch(instType)
+    {
+    case SIT_ALU_HALF_LIT:  
+    case SIT_TEX:           
+    case SIT_VTX:           
+    case SIT_VTX_GENERIC:   
+    case SIT_VTX_SEM:       
+        return 4;
+    case SIT_ALU_FALL_LIT:
+        return 6;
+    default:
+        break;
+    }
+
+    return 2;
+}
+
+unsigned int GetCFMaxInstructions(ShaderInstType instType)
+{
+    switch (instType)
+    {
+    case SIT_CF_ALL_EXP:    
+    case SIT_CF_ALL_EXP_SX: 
+    case SIT_CF_ALL_EXP_SMX:  
+        return 0x10;
+    case SIT_CF_GENERIC:
+        return 0x8;  //For tex and vtx
+    case SIT_CF_ALU:
+        return 0x80;
+    default:
+        break;
+    }
+    return 0x10;
+}
+
+GLboolean LinkVertexInstruction(R700ControlFlowGenericClause *pCFGeneric,
+								R700VertexInstruction *pVTXInstruction)
+{
+    if (pCFGeneric->m_pLinkedTEXInstruction != 0)
+    {
+	radeon_error("This instruction is already linked to a texture instruction.\n");
+	return GL_FALSE;
+    }
+
+    pCFGeneric->m_pLinkedVTXInstruction     = pVTXInstruction;
+    pVTXInstruction->m_pLinkedGenericClause = pCFGeneric;
+
+    return GL_TRUE;
+}
+
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.h b/src/mesa/drivers/dri/r600/r700_shaderinst.h
new file mode 100644
index 0000000000..2829cca0a3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shaderinst.h
@@ -0,0 +1,321 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _R700_SHADERINST_H_
+#define _R700_SHADERINST_H_
+
+#include "main/glheader.h"
+
+#include "defaultendian.h" 
+#include "sq_micro_reg.h"
+
+#define SQ_ALU_CONSTANT_PS_OFFSET      0x00000000
+#define SQ_ALU_CONSTANT_PS_COUNT       0x00000100
+#define SQ_ALU_CONSTANT_VS_OFFSET      0x00000100
+#define SQ_ALU_CONSTANT_VS_COUNT       0x00000100
+#define SQ_FETCH_RESOURCE_PS_OFFSET    0x00000000
+#define SQ_FETCH_RESOURCE_PS_COUNT     0x000000a0
+#define SQ_FETCH_RESOURCE_VS_OFFSET    0x000000a0
+#define SQ_FETCH_RESOURCE_VS_COUNT     0x000000b0
+
+#define SHADERINST_TYPEMASK_CF  0x10
+#define SHADERINST_TYPEMASK_ALU 0x20
+#define SHADERINST_TYPEMASK_TEX 0x40
+#define SHADERINST_TYPEMASK_VTX 0x80
+
+typedef enum ShaderInstType 
+{
+    SIT_CF = 0x10,            /*SIZE = 0x2*/
+        SIT_CF_ALL_EXP = 0x14,    /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+            SIT_CF_ALL_EXP_SX = 0x15, /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+            SIT_CF_ALL_EXP_SMX= 0x16, /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+        SIT_CF_GENERIC = 0x18,    /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x8;  //For tex and vtx*/
+        SIT_CF_ALU = 0x19,        /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x80;*/
+    SIT_ALU = 0x20,           /*SIZE = 0x2,*/
+        SIT_ALU_HALF_LIT = 0x21,  /*SIZE = 0x4,*/
+        SIT_ALU_FALL_LIT = 0x22,  /*SIZE = 0x6,*/
+    SIT_TEX = 0x40,           /*SIZE = 0x4,*/
+    SIT_VTX = 0x80,           /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+        SIT_VTX_GENERIC = 0x81,   /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+        SIT_VTX_SEM = 0x82       /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+} ShaderInstType;
+
+typedef struct R700ShaderInstruction 
+{
+    ShaderInstType m_ShaderInstType;
+    struct R700ShaderInstruction *pNextInst;
+    GLuint m_uIndex;
+    GLuint useCount;
+} R700ShaderInstruction;
+
+// ------------------ CF insts ---------------------------
+
+typedef R700ShaderInstruction R700ControlFlowInstruction;
+
+typedef struct R700ControlFlowAllocExportClause  
+{
+    ShaderInstType          m_ShaderInstType;
+    R700ShaderInstruction * pNextInst;    
+    GLuint m_uIndex;
+    GLuint useCount;
+		
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+} R700ControlFlowAllocExportClause;
+
+typedef struct R700ControlFlowSXClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ControlFlowAllocExportClause
+		//R700ControlFlowInstruction 
+			//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+			//---------------------
+		//---------------------------
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+	//-------------------------------------
+
+    sq_cf_alloc_export_word1_swiz_u m_Word1_SWIZ;
+} R700ControlFlowSXClause;
+
+typedef struct R700ControlFlowSMXClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+    //R700ControlFlowAllocExportClause
+		//R700ControlFlowInstruction 
+			//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+			//---------------------
+		//---------------------------
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+	//-------------------------------
+
+    sq_cf_alloc_export_word1_buf_u m_Word1_BUF;
+} R700ControlFlowSMXClause;
+
+typedef struct R700ControlFlowGenericClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ControlFlowInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	//---------------------
+
+    sq_cf_word0_u m_Word0;
+    sq_cf_word1_u m_Word1;
+
+    struct R700VertexInstruction  *m_pLinkedVTXInstruction;
+    struct R700TextureInstruction *m_pLinkedTEXInstruction;
+} R700ControlFlowGenericClause;
+
+typedef struct R700ControlFlowALUClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+    //R700ControlFlowInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	//---------------------
+
+    sq_cf_alu_word0_u m_Word0;
+    sq_cf_alu_word1_u m_Word1;
+    
+    struct R700ALUInstruction *m_pLinkedALUInstruction;
+} R700ControlFlowALUClause;
+
+// ------------------- End of CF Inst ------------------------
+
+// ------------------- ALU Inst ------------------------------
+typedef struct R700ALUInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+} R700ALUInstruction;
+
+typedef struct R700ALUInstructionHalfLiteral
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ALUInstruction 
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+	//-------------------
+
+    GLfloat m_fLiteralX,
+            m_fLiteralY;
+} R700ALUInstructionHalfLiteral;
+
+typedef struct R700ALUInstructionFullLiteral 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ALUInstruction 
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+	//-------------------
+
+    GLfloat m_fLiteralX,
+            m_fLiteralY,
+            m_fLiteralZ,
+            m_fLiteralW;
+} R700ALUInstructionFullLiteral;
+// ------------------- End of ALU Inst -----------------------
+
+// ------------------- Textuer/Vertex Instruction --------------------
+
+typedef struct R700TextureInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+	
+    sq_tex_word0_u m_Word0;
+    sq_tex_word1_u m_Word1;
+    sq_tex_word2_u m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+} R700TextureInstruction;
+
+typedef struct R700VertexInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+} R700VertexInstruction;
+//
+typedef struct R700VertexSemanticFetch 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700VertexInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+	//---------------------------
+
+    sq_vtx_word1_sem_u m_Word1_SEM;
+} R700VertexSemanticFetch;
+//
+typedef struct R700VertexGenericFetch 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700VertexInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+	//---------------------------
+
+    sq_vtx_word1_gpr_u m_Word1_GPR;
+} R700VertexGenericFetch;
+
+// ------------------- End of Texture Vertex Instruction --------------------
+
+void Init_R700ControlFlowGenericClause(R700ControlFlowGenericClause* pInst);
+void Init_R700ControlFlowALUClause(R700ControlFlowALUClause* pInst);
+void Init_R700ControlFlowSXClause(R700ControlFlowSXClause* pInst);
+void Init_R700ControlFlowSMXClause(R700ControlFlowSMXClause* pInst);
+void Init_R700ALUInstruction(R700ALUInstruction* pInst);
+void Init_R700ALUInstructionHalfLiteral(R700ALUInstructionHalfLiteral* pInst, GLfloat x, GLfloat y);
+void Init_R700ALUInstructionFullLiteral(R700ALUInstructionFullLiteral* pInst, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+void Init_R700TextureInstruction(R700TextureInstruction* pInst);
+void Init_R700VertexSemanticFetch(R700VertexSemanticFetch* pInst);
+void Init_R700VertexGenericFetch(R700VertexGenericFetch* pInst);
+
+unsigned int GetInstructionSize(ShaderInstType instType);
+unsigned int GetCFMaxInstructions(ShaderInstType instType);
+
+GLboolean LinkVertexInstruction(R700ControlFlowGenericClause *pCFGeneric,
+								R700VertexInstruction *pVTXInstruction);
+
+#endif //_R700_SHADERINST_H_
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
new file mode 100644
index 0000000000..1f4724e838
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -0,0 +1,1802 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/enums.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/dd.h"
+#include "main/simple_list.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_vp_build.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "main/api_arrayelt.h"
+#include "main/state.h"
+#include "main/framebuffer.h"
+
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+#include "vbo/vbo.h"
+#include "main/texformat.h"
+
+#include "r600_context.h"
+
+#include "r700_state.h"
+
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+
+
+static void r700SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state);
+static void r700UpdatePolygonMode(GLcontext * ctx);
+static void r700SetPolygonOffsetState(GLcontext * ctx, GLboolean state);
+static void r700SetStencilState(GLcontext * ctx, GLboolean state);
+
+void r700UpdateShaders (GLcontext * ctx)  //----------------------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+    GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
+    GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
+    int i;
+
+    /* should only happenen once, just after context is created */
+    /* TODO: shouldn't we fallback to sw here? */
+    if (!ctx->FragmentProgram._Current) {
+	    _mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
+	    return;
+    }
+
+    r700SelectFragmentShader(ctx);
+
+    if (context->radeon.NewGLState) {
+	    for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+		    /* mat states from state var not array for sw */
+		    dummy_attrib[i].stride = 0;
+	            temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+		    TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &(dummy_attrib[i]);
+	    }
+
+	    _tnl_UpdateFixedFunctionProgram(ctx);
+
+	    for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+		    TNL_CONTEXT(ctx)->vb.AttribPtr[i] = temp_attrib[i];
+	    }
+    }
+
+    r700SelectVertexShader(ctx);
+    r700UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+    context->radeon.NewGLState = 0;
+}
+
+/*
+ * To correctly position primitives:
+ */
+void r700UpdateViewportOffset(GLcontext * ctx) //------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&context->radeon);
+	GLfloat xoffset = (GLfloat) dPriv->x;
+	GLfloat yoffset = (GLfloat) dPriv->y + dPriv->h;
+	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	int id = 0;
+
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat ty = (-v[MAT_TY]) + yoffset;
+
+	if (r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All != tx ||
+	    r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All != ty) {
+		/* Note: this should also modify whatever data the context reset
+		 * code uses...
+		 */
+		R600_STATECHANGE(context, vpt);
+		r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx;
+		r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All = ty;
+	}
+
+	radeonUpdateScissor(ctx);
+}
+
+void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //--------------------
+{
+	struct r700_fragment_program *fp =
+		(struct r700_fragment_program *)ctx->FragmentProgram._Current;
+	struct gl_program_parameter_list *paramList;
+
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
+		return;
+
+	if (!ctx->FragmentProgram._Current || !fp)
+		return;
+
+	paramList = ctx->FragmentProgram._Current->Base.Parameters;
+
+	if (!paramList)
+		return;
+
+	_mesa_load_state_parameters(ctx, paramList);
+
+}
+
+/**
+ * Called by Mesa after an internal state update.
+ */
+static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //-------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    _swrast_InvalidateState(ctx, new_state);
+    _swsetup_InvalidateState(ctx, new_state);
+    _vbo_InvalidateState(ctx, new_state);
+    _tnl_InvalidateState(ctx, new_state);
+    _ae_invalidate_state(ctx, new_state);
+
+    if (new_state & _NEW_BUFFERS) {
+	    _mesa_update_framebuffer(ctx);
+	    /* this updates the DrawBuffer's Width/Height if it's a FBO */
+	    _mesa_update_draw_buffer_bounds(ctx);
+
+	    R600_STATECHANGE(context, cb_target);
+	    R600_STATECHANGE(context, db_target);
+    }
+
+    r700UpdateStateParameters(ctx, new_state);
+
+    R600_STATECHANGE(context, cl);
+    R600_STATECHANGE(context, spi);
+
+    if(GL_TRUE == r700->bEnablePerspective)
+    {
+        /* Do scale XY and Z by 1/W0 for perspective correction on pos. For orthogonal case, set both to one. */
+        CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+        CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+        SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+        CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    }
+    else
+    {
+        /* For orthogonal case. */
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+        CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+        SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    }
+
+    context->radeon.NewGLState |= new_state;
+}
+
+static void r700SetDepthState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, db);
+
+    if (ctx->Depth.Test)
+    {
+        SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit);
+        if (ctx->Depth.Mask)
+        {
+            SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+        }
+        else
+        {
+            CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+        }
+
+        switch (ctx->Depth.Func)
+        {
+        case GL_NEVER:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_NEVER,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_LESS:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_LESS,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_EQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_EQUAL,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_LEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_LEQUAL,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_GREATER:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_GREATER,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_NOTEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_NOTEQUAL,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_GEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_GEQUAL,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_ALWAYS:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        default:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS,
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        }
+    }
+    else
+    {
+        CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit);
+        CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+    }
+}
+
+static void r700SetAlphaState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	uint32_t alpha_func = REF_ALWAYS;
+	GLboolean really_enabled = ctx->Color.AlphaEnabled;
+
+	R600_STATECHANGE(context, sx);
+
+	switch (ctx->Color.AlphaFunc) {
+	case GL_NEVER:
+		alpha_func = REF_NEVER;
+		break;
+	case GL_LESS:
+		alpha_func = REF_LESS;
+		break;
+	case GL_EQUAL:
+		alpha_func = REF_EQUAL;
+		break;
+	case GL_LEQUAL:
+		alpha_func = REF_LEQUAL;
+		break;
+	case GL_GREATER:
+		alpha_func = REF_GREATER;
+		break;
+	case GL_NOTEQUAL:
+		alpha_func = REF_NOTEQUAL;
+		break;
+	case GL_GEQUAL:
+		alpha_func = REF_GEQUAL;
+		break;
+	case GL_ALWAYS:
+		/*alpha_func = REF_ALWAYS; */
+		really_enabled = GL_FALSE;
+		break;
+	}
+
+	if (really_enabled) {
+		SETfield(r700->SX_ALPHA_TEST_CONTROL.u32All, alpha_func,
+			 ALPHA_FUNC_shift, ALPHA_FUNC_mask);
+		SETbit(r700->SX_ALPHA_TEST_CONTROL.u32All, ALPHA_TEST_ENABLE_bit);
+		r700->SX_ALPHA_REF.f32All = ctx->Color.AlphaRef;
+	} else {
+		CLEARbit(r700->SX_ALPHA_TEST_CONTROL.u32All, ALPHA_TEST_ENABLE_bit);
+	}
+
+}
+
+static void r700AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) //---------------
+{
+	(void)func;
+	(void)ref;
+	r700SetAlphaState(ctx);
+}
+
+
+static void r700BlendColor(GLcontext * ctx, const GLfloat cf[4]) //----------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, blnd_clr);
+
+	r700->CB_BLEND_RED.f32All = cf[0];
+	r700->CB_BLEND_GREEN.f32All = cf[1];
+	r700->CB_BLEND_BLUE.f32All = cf[2];
+	r700->CB_BLEND_ALPHA.f32All = cf[3];
+}
+
+static int blend_factor(GLenum factor, GLboolean is_src)
+{
+	switch (factor) {
+	case GL_ZERO:
+		return BLEND_ZERO;
+		break;
+	case GL_ONE:
+		return BLEND_ONE;
+		break;
+	case GL_DST_COLOR:
+		return BLEND_DST_COLOR;
+		break;
+	case GL_ONE_MINUS_DST_COLOR:
+		return BLEND_ONE_MINUS_DST_COLOR;
+		break;
+	case GL_SRC_COLOR:
+		return BLEND_SRC_COLOR;
+		break;
+	case GL_ONE_MINUS_SRC_COLOR:
+		return BLEND_ONE_MINUS_SRC_COLOR;
+		break;
+	case GL_SRC_ALPHA:
+		return BLEND_SRC_ALPHA;
+		break;
+	case GL_ONE_MINUS_SRC_ALPHA:
+		return BLEND_ONE_MINUS_SRC_ALPHA;
+		break;
+	case GL_DST_ALPHA:
+		return BLEND_DST_ALPHA;
+		break;
+	case GL_ONE_MINUS_DST_ALPHA:
+		return BLEND_ONE_MINUS_DST_ALPHA;
+		break;
+	case GL_SRC_ALPHA_SATURATE:
+		return (is_src) ? BLEND_SRC_ALPHA_SATURATE : BLEND_ZERO;
+		break;
+	case GL_CONSTANT_COLOR:
+		return BLEND_CONSTANT_COLOR;
+		break;
+	case GL_ONE_MINUS_CONSTANT_COLOR:
+		return BLEND_ONE_MINUS_CONSTANT_COLOR;
+		break;
+	case GL_CONSTANT_ALPHA:
+		return BLEND_CONSTANT_ALPHA;
+		break;
+	case GL_ONE_MINUS_CONSTANT_ALPHA:
+		return BLEND_ONE_MINUS_CONSTANT_ALPHA;
+		break;
+	default:
+		fprintf(stderr, "unknown blend factor %x\n", factor);
+		return (is_src) ? BLEND_ONE : BLEND_ZERO;
+		break;
+	}
+}
+
+static void r700SetBlendState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int id = 0;
+	uint32_t blend_reg = 0, eqn, eqnA;
+
+	R600_STATECHANGE(context, blnd);
+
+	if (RGBA_LOGICOP_ENABLED(ctx) || !ctx->Color.BlendEnabled) {
+		SETfield(blend_reg,
+			 BLEND_ONE, COLOR_SRCBLEND_shift, COLOR_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ZERO, COLOR_DESTBLEND_shift, COLOR_DESTBLEND_mask);
+		SETfield(blend_reg,
+			 COMB_DST_PLUS_SRC, COLOR_COMB_FCN_shift, COLOR_COMB_FCN_mask);
+		SETfield(blend_reg,
+			 BLEND_ONE, ALPHA_SRCBLEND_shift, ALPHA_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ZERO, ALPHA_DESTBLEND_shift, ALPHA_DESTBLEND_mask);
+		SETfield(blend_reg,
+			 COMB_DST_PLUS_SRC, ALPHA_COMB_FCN_shift, ALPHA_COMB_FCN_mask);
+		if (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_R600)
+			r700->CB_BLEND_CONTROL.u32All = blend_reg;
+		else
+			r700->render_target[id].CB_BLEND0_CONTROL.u32All = blend_reg;
+		return;
+	}
+
+	SETfield(blend_reg,
+		 blend_factor(ctx->Color.BlendSrcRGB, GL_TRUE),
+		 COLOR_SRCBLEND_shift, COLOR_SRCBLEND_mask);
+	SETfield(blend_reg,
+		 blend_factor(ctx->Color.BlendDstRGB, GL_FALSE),
+		 COLOR_DESTBLEND_shift, COLOR_DESTBLEND_mask);
+
+	switch (ctx->Color.BlendEquationRGB) {
+	case GL_FUNC_ADD:
+		eqn = COMB_DST_PLUS_SRC;
+		break;
+	case GL_FUNC_SUBTRACT:
+		eqn = COMB_SRC_MINUS_DST;
+		break;
+	case GL_FUNC_REVERSE_SUBTRACT:
+		eqn = COMB_DST_MINUS_SRC;
+		break;
+	case GL_MIN:
+		eqn = COMB_MIN_DST_SRC;
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 COLOR_SRCBLEND_shift, COLOR_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 COLOR_DESTBLEND_shift, COLOR_DESTBLEND_mask);
+		break;
+	case GL_MAX:
+		eqn = COMB_MAX_DST_SRC;
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 COLOR_SRCBLEND_shift, COLOR_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 COLOR_DESTBLEND_shift, COLOR_DESTBLEND_mask);
+		break;
+
+	default:
+		fprintf(stderr,
+			"[%s:%u] Invalid RGB blend equation (0x%04x).\n",
+			__FUNCTION__, __LINE__, ctx->Color.BlendEquationRGB);
+		return;
+	}
+	SETfield(blend_reg,
+		 eqn, COLOR_COMB_FCN_shift, COLOR_COMB_FCN_mask);
+
+	SETfield(blend_reg,
+		 blend_factor(ctx->Color.BlendSrcRGB, GL_TRUE),
+		 ALPHA_SRCBLEND_shift, ALPHA_SRCBLEND_mask);
+	SETfield(blend_reg,
+		 blend_factor(ctx->Color.BlendDstRGB, GL_FALSE),
+		 ALPHA_DESTBLEND_shift, ALPHA_DESTBLEND_mask);
+
+	switch (ctx->Color.BlendEquationA) {
+	case GL_FUNC_ADD:
+		eqnA = COMB_DST_PLUS_SRC;
+		break;
+	case GL_FUNC_SUBTRACT:
+		eqnA = COMB_SRC_MINUS_DST;
+		break;
+	case GL_FUNC_REVERSE_SUBTRACT:
+		eqnA = COMB_DST_MINUS_SRC;
+		break;
+	case GL_MIN:
+		eqnA = COMB_MIN_DST_SRC;
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 ALPHA_SRCBLEND_shift, ALPHA_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 ALPHA_DESTBLEND_shift, ALPHA_DESTBLEND_mask);
+		break;
+	case GL_MAX:
+		eqnA = COMB_MAX_DST_SRC;
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 ALPHA_SRCBLEND_shift, ALPHA_SRCBLEND_mask);
+		SETfield(blend_reg,
+			 BLEND_ONE,
+			 ALPHA_DESTBLEND_shift, ALPHA_DESTBLEND_mask);
+		break;
+	default:
+		fprintf(stderr,
+			"[%s:%u] Invalid A blend equation (0x%04x).\n",
+			__FUNCTION__, __LINE__, ctx->Color.BlendEquationA);
+		return;
+	}
+
+	SETfield(blend_reg,
+		 eqnA, ALPHA_COMB_FCN_shift, ALPHA_COMB_FCN_mask);
+
+	SETbit(blend_reg, SEPARATE_ALPHA_BLEND_bit);
+
+	if (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_R600)
+		r700->CB_BLEND_CONTROL.u32All = blend_reg;
+	else {
+		r700->render_target[id].CB_BLEND0_CONTROL.u32All = blend_reg;
+		SETbit(r700->CB_COLOR_CONTROL.u32All, PER_MRT_BLEND_bit);
+	}
+	SETfield(r700->CB_COLOR_CONTROL.u32All, (1 << id),
+		 TARGET_BLEND_ENABLE_shift, TARGET_BLEND_ENABLE_mask);
+
+}
+
+static void r700BlendEquationSeparate(GLcontext * ctx,
+				                      GLenum modeRGB, GLenum modeA) //-----------------
+{
+	r700SetBlendState(ctx);
+}
+
+static void r700BlendFuncSeparate(GLcontext * ctx,
+				  GLenum sfactorRGB, GLenum dfactorRGB,
+				  GLenum sfactorA, GLenum dfactorA) //------------------------
+{
+	r700SetBlendState(ctx);
+}
+
+/**
+ * Translate LogicOp enums into hardware representation.
+ */
+static GLuint translate_logicop(GLenum logicop)
+{
+	switch (logicop) {
+	case GL_CLEAR:
+		return 0x00;
+	case GL_SET:
+		return 0xff;
+	case GL_COPY:
+		return 0xcc;
+	case GL_COPY_INVERTED:
+		return 0x33;
+	case GL_NOOP:
+		return 0xaa;
+	case GL_INVERT:
+		return 0x55;
+	case GL_AND:
+		return 0x88;
+	case GL_NAND:
+		return 0x77;
+	case GL_OR:
+		return 0xee;
+	case GL_NOR:
+		return 0x11;
+	case GL_XOR:
+		return 0x66;
+	case GL_EQUIV:
+		return 0xaa;
+	case GL_AND_REVERSE:
+		return 0x44;
+	case GL_AND_INVERTED:
+		return 0x22;
+	case GL_OR_REVERSE:
+		return 0xdd;
+	case GL_OR_INVERTED:
+		return 0xbb;
+	default:
+		fprintf(stderr, "unknown blend logic operation %x\n", logicop);
+		return 0xcc;
+	}
+}
+
+/**
+ * Used internally to update the r300->hw hardware state to match the
+ * current OpenGL state.
+ */
+static void r700SetLogicOpState(GLcontext *ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+
+	R600_STATECHANGE(context, blnd);
+
+	if (RGBA_LOGICOP_ENABLED(ctx))
+		SETfield(r700->CB_COLOR_CONTROL.u32All,
+			 translate_logicop(ctx->Color.LogicOp), ROP3_shift, ROP3_mask);
+	else
+		SETfield(r700->CB_COLOR_CONTROL.u32All, 0xCC, ROP3_shift, ROP3_mask);
+}
+
+/**
+ * Called by Mesa when an application program changes the LogicOp state
+ * via glLogicOp.
+ */
+static void r700LogicOpcode(GLcontext *ctx, GLenum logicop)
+{
+	if (RGBA_LOGICOP_ENABLED(ctx))
+		r700SetLogicOpState(ctx);
+}
+
+static void r700UpdateCulling(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+
+    R600_STATECHANGE(context, su);
+
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit);
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+
+    if (ctx->Polygon.CullFlag)
+    {
+        switch (ctx->Polygon.CullFaceMode)
+        {
+        case GL_FRONT:
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        case GL_BACK:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        case GL_FRONT_AND_BACK:
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        default:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        }
+    }
+
+    switch (ctx->Polygon.FrontFace)
+    {
+        case GL_CW:
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit);
+            break;
+        case GL_CCW:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit);
+            break;
+        default:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); /* default: ccw */
+            break;
+    }
+}
+
+static void r700UpdateLineStipple(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+
+    R600_STATECHANGE(context, sc);
+
+    if (ctx->Line.StippleFlag)
+    {
+	SETbit(r700->PA_SC_MODE_CNTL.u32All, LINE_STIPPLE_ENABLE_bit);
+    }
+    else
+    {
+	CLEARbit(r700->PA_SC_MODE_CNTL.u32All, LINE_STIPPLE_ENABLE_bit);
+    }
+}
+
+static void r700Enable(GLcontext * ctx, GLenum cap, GLboolean state) //------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+
+	switch (cap) {
+	case GL_TEXTURE_1D:
+	case GL_TEXTURE_2D:
+	case GL_TEXTURE_3D:
+		/* empty */
+		break;
+	case GL_FOG:
+		/* empty */
+		break;
+	case GL_ALPHA_TEST:
+		r700SetAlphaState(ctx);
+		break;
+	case GL_COLOR_LOGIC_OP:
+		r700SetLogicOpState(ctx);
+		/* fall-through, because logic op overrides blending */
+	case GL_BLEND:
+		r700SetBlendState(ctx);
+		break;
+	case GL_CLIP_PLANE0:
+	case GL_CLIP_PLANE1:
+	case GL_CLIP_PLANE2:
+	case GL_CLIP_PLANE3:
+	case GL_CLIP_PLANE4:
+	case GL_CLIP_PLANE5:
+		r700SetClipPlaneState(ctx, cap, state);
+		break;
+	case GL_DEPTH_TEST:
+		r700SetDepthState(ctx);
+		break;
+	case GL_STENCIL_TEST:
+		r700SetStencilState(ctx, state);
+		break;
+	case GL_CULL_FACE:
+		r700UpdateCulling(ctx);
+		break;
+	case GL_POLYGON_OFFSET_POINT:
+	case GL_POLYGON_OFFSET_LINE:
+	case GL_POLYGON_OFFSET_FILL:
+		r700SetPolygonOffsetState(ctx, state);
+		break;
+	case GL_SCISSOR_TEST:
+		radeon_firevertices(&context->radeon);
+		context->radeon.state.scissor.enabled = state;
+		radeonUpdateScissor(ctx);
+		break;
+	case GL_LINE_STIPPLE:
+		r700UpdateLineStipple(ctx);
+		break;
+	default:
+		break;
+	}
+
+}
+
+/**
+ * Handle glColorMask()
+ */
+static void r700ColorMask(GLcontext * ctx,
+			  GLboolean r, GLboolean g, GLboolean b, GLboolean a) //------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+	unsigned int mask = ((r ? 1 : 0) |
+			     (g ? 2 : 0) |
+			     (b ? 4 : 0) |
+			     (a ? 8 : 0));
+
+	if (mask != r700->CB_SHADER_MASK.u32All) {
+		R600_STATECHANGE(context, cb);
+		SETfield(r700->CB_SHADER_MASK.u32All, mask, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask);
+	}
+}
+
+/**
+ * Change the depth testing function.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700DepthFunc(GLcontext * ctx, GLenum func) //--------------------
+{
+    r700SetDepthState(ctx);
+}
+
+/**
+ * Enable/Disable depth writing.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700DepthMask(GLcontext * ctx, GLboolean mask) //------------------
+{
+    r700SetDepthState(ctx);
+}
+
+/**
+ * Change the culling mode.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700CullFace(GLcontext * ctx, GLenum mode) //-----------------
+{
+    r700UpdateCulling(ctx);
+}
+
+/* =============================================================
+ * Fog
+ */
+static void r700Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) //--------------
+{
+}
+
+/**
+ * Change the polygon orientation.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700FrontFace(GLcontext * ctx, GLenum mode) //------------------
+{
+    r700UpdateCulling(ctx);
+    r700UpdatePolygonMode(ctx);
+}
+
+static void r700ShadeModel(GLcontext * ctx, GLenum mode) //--------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, spi);
+
+	/* also need to set/clear FLAT_SHADE bit per param in SPI_PS_INPUT_CNTL_[0-31] */
+	switch (mode) {
+	case GL_FLAT:
+		SETbit(r700->SPI_INTERP_CONTROL_0.u32All, FLAT_SHADE_ENA_bit);
+		break;
+	case GL_SMOOTH:
+		CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, FLAT_SHADE_ENA_bit);
+		break;
+	default:
+		return;
+	}
+}
+
+/* =============================================================
+ * Point state
+ */
+static void r700PointSize(GLcontext * ctx, GLfloat size)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, su);
+
+	/* We need to clamp to user defined range here, because
+	 * the HW clamping happens only for per vertex point size. */
+	size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize);
+
+	/* same size limits for AA, non-AA points */
+	size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize);
+
+	/* format is 12.4 fixed point */
+	SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 16),
+		 PA_SU_POINT_SIZE__HEIGHT_shift, PA_SU_POINT_SIZE__HEIGHT_mask);
+	SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 16),
+		 PA_SU_POINT_SIZE__WIDTH_shift, PA_SU_POINT_SIZE__WIDTH_mask);
+
+}
+
+static void r700PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * param) //---------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, su);
+
+	/* format is 12.4 fixed point */
+	switch (pname) {
+	case GL_POINT_SIZE_MIN:
+		SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MinSize * 16.0),
+			 MIN_SIZE_shift, MIN_SIZE_mask);
+		break;
+	case GL_POINT_SIZE_MAX:
+		SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MaxSize * 16.0),
+			 MAX_SIZE_shift, MAX_SIZE_mask);
+		break;
+	case GL_POINT_DISTANCE_ATTENUATION:
+		break;
+	case GL_POINT_FADE_THRESHOLD_SIZE:
+		break;
+	default:
+		break;
+	}
+}
+
+static int translate_stencil_func(int func)
+{
+	switch (func) {
+	case GL_NEVER:
+		return REF_NEVER;
+	case GL_LESS:
+		return REF_LESS;
+	case GL_EQUAL:
+		return REF_EQUAL;
+	case GL_LEQUAL:
+		return REF_LEQUAL;
+	case GL_GREATER:
+		return REF_GREATER;
+	case GL_NOTEQUAL:
+		return REF_NOTEQUAL;
+	case GL_GEQUAL:
+		return REF_GEQUAL;
+	case GL_ALWAYS:
+		return REF_ALWAYS;
+	}
+	return 0;
+}
+
+static int translate_stencil_op(int op)
+{
+	switch (op) {
+	case GL_KEEP:
+		return STENCIL_KEEP;
+	case GL_ZERO:
+		return STENCIL_ZERO;
+	case GL_REPLACE:
+		return STENCIL_REPLACE;
+	case GL_INCR:
+		return STENCIL_INCR_CLAMP;
+	case GL_DECR:
+		return STENCIL_DECR_CLAMP;
+	case GL_INCR_WRAP_EXT:
+		return STENCIL_INCR_WRAP;
+	case GL_DECR_WRAP_EXT:
+		return STENCIL_DECR_WRAP;
+	case GL_INVERT:
+		return STENCIL_INVERT;
+	default:
+		WARN_ONCE("Do not know how to translate stencil op");
+		return STENCIL_KEEP;
+	}
+	return 0;
+}
+
+static void r700SetStencilState(GLcontext * ctx, GLboolean state)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	GLboolean hw_stencil = GL_FALSE;
+
+	if (ctx->DrawBuffer) {
+		struct radeon_renderbuffer *rrbStencil
+			= radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+		hw_stencil = (rrbStencil && rrbStencil->bo);
+	}
+
+	if (hw_stencil) {
+		R600_STATECHANGE(context, db);
+		if (state) {
+			SETbit(r700->DB_DEPTH_CONTROL.u32All, STENCIL_ENABLE_bit);
+			SETbit(r700->DB_DEPTH_CONTROL.u32All, BACKFACE_ENABLE_bit);
+		} else
+			CLEARbit(r700->DB_DEPTH_CONTROL.u32All, STENCIL_ENABLE_bit);
+	}
+}
+
+static void r700StencilFuncSeparate(GLcontext * ctx, GLenum face,
+				    GLenum func, GLint ref, GLuint mask) //---------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	const unsigned back = ctx->Stencil._BackFace;
+
+	R600_STATECHANGE(context, stencil);
+	R600_STATECHANGE(context, db);
+
+	//front
+	SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.Ref[0],
+		 STENCILREF_shift, STENCILREF_mask);
+	SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.ValueMask[0],
+		 STENCILMASK_shift, STENCILMASK_mask);
+
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_func(ctx->Stencil.Function[0]),
+		 STENCILFUNC_shift, STENCILFUNC_mask);
+
+	//back
+	SETfield(r700->DB_STENCILREFMASK_BF.u32All, ctx->Stencil.Ref[back],
+		 STENCILREF_BF_shift, STENCILREF_BF_mask);
+	SETfield(r700->DB_STENCILREFMASK_BF.u32All, ctx->Stencil.ValueMask[back],
+		 STENCILMASK_BF_shift, STENCILMASK_BF_mask);
+
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_func(ctx->Stencil.Function[back]),
+		 STENCILFUNC_BF_shift, STENCILFUNC_BF_mask);
+
+}
+
+static void r700StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) //--------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	const unsigned back = ctx->Stencil._BackFace;
+
+	R600_STATECHANGE(context, stencil);
+
+	// front
+	SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.WriteMask[0],
+		 STENCILWRITEMASK_shift, STENCILWRITEMASK_mask);
+
+	// back
+	SETfield(r700->DB_STENCILREFMASK_BF.u32All, ctx->Stencil.WriteMask[back],
+		 STENCILWRITEMASK_BF_shift, STENCILWRITEMASK_BF_mask);
+
+}
+
+static void r700StencilOpSeparate(GLcontext * ctx, GLenum face,
+				  GLenum fail, GLenum zfail, GLenum zpass) //--------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	const unsigned back = ctx->Stencil._BackFace;
+
+	R600_STATECHANGE(context, db);
+
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.FailFunc[0]),
+		 STENCILFAIL_shift, STENCILFAIL_mask);
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.ZFailFunc[0]),
+		 STENCILZFAIL_shift, STENCILZFAIL_mask);
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.ZPassFunc[0]),
+		 STENCILZPASS_shift, STENCILZPASS_mask);
+
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.FailFunc[back]),
+		 STENCILFAIL_BF_shift, STENCILFAIL_BF_mask);
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.ZFailFunc[back]),
+		 STENCILZFAIL_BF_shift, STENCILZFAIL_BF_mask);
+	SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.ZPassFunc[back]),
+		 STENCILZPASS_BF_shift, STENCILZPASS_BF_mask);
+}
+
+static void r700UpdateWindow(GLcontext * ctx, int id) //--------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&context->radeon);
+	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
+	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+	const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
+	GLfloat y_scale, y_bias;
+
+	if (render_to_fbo) {
+		y_scale = 1.0;
+		y_bias = 0;
+	} else {
+		y_scale = -1.0;
+		y_bias = yoffset;
+	}
+
+	GLfloat sx = v[MAT_SX];
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat sy = v[MAT_SY] * y_scale;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias;
+	GLfloat sz = v[MAT_SZ] * depthScale;
+	GLfloat tz = v[MAT_TZ] * depthScale;
+
+	R600_STATECHANGE(context, vpt);
+
+	r700->viewport[id].PA_CL_VPORT_XSCALE.f32All  = sx;
+	r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx;
+
+	r700->viewport[id].PA_CL_VPORT_YSCALE.f32All  = sy;
+	r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All = ty;
+
+	r700->viewport[id].PA_CL_VPORT_ZSCALE.f32All  = sz;
+	r700->viewport[id].PA_CL_VPORT_ZOFFSET.f32All = tz;
+
+	r700->viewport[id].enabled = GL_TRUE;
+
+	r700SetScissor(context);
+}
+
+
+static void r700Viewport(GLcontext * ctx,
+                         GLint x,
+                         GLint y,
+			 GLsizei width,
+                         GLsizei height) //--------------------
+{
+	r700UpdateWindow(ctx, 0);
+
+	radeon_viewport(ctx, x, y, width, height);
+}
+
+static void r700DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) //-------------
+{
+	r700UpdateWindow(ctx, 0);
+}
+
+static void r700LineWidth(GLcontext * ctx, GLfloat widthf) //---------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    uint32_t lineWidth = (uint32_t)((widthf * 0.5) * (1 << 4));
+
+    R600_STATECHANGE(context, su);
+
+    if (lineWidth > 0xFFFF)
+	    lineWidth = 0xFFFF;
+    SETfield(r700->PA_SU_LINE_CNTL.u32All,(uint16_t)lineWidth,
+	     PA_SU_LINE_CNTL__WIDTH_shift, PA_SU_LINE_CNTL__WIDTH_mask);
+}
+
+static void r700LineStipple(GLcontext *ctx, GLint factor, GLushort pattern)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    R600_STATECHANGE(context, sc);
+
+    SETfield(r700->PA_SC_LINE_STIPPLE.u32All, pattern, LINE_PATTERN_shift, LINE_PATTERN_mask);
+    SETfield(r700->PA_SC_LINE_STIPPLE.u32All, (factor-1), REPEAT_COUNT_shift, REPEAT_COUNT_mask);
+    SETfield(r700->PA_SC_LINE_STIPPLE.u32All, 1, AUTO_RESET_CNTL_shift, AUTO_RESET_CNTL_mask);
+}
+
+static void r700SetPolygonOffsetState(GLcontext * ctx, GLboolean state)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, su);
+
+	if (state) {
+		SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_FRONT_ENABLE_bit);
+		SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_BACK_ENABLE_bit);
+		SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_PARA_ENABLE_bit);
+	} else {
+		CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_FRONT_ENABLE_bit);
+		CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_BACK_ENABLE_bit);
+		CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_PARA_ENABLE_bit);
+	}
+}
+
+static void r700PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) //--------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	GLfloat constant = units;
+
+	switch (ctx->Visual.depthBits) {
+	case 16:
+		constant *= 4.0;
+		break;
+	case 24:
+		constant *= 2.0;
+		break;
+	}
+
+	factor *= 12.0;
+
+	R600_STATECHANGE(context, poly);
+
+	r700->PA_SU_POLY_OFFSET_FRONT_SCALE.f32All = factor;
+	r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.f32All = constant;
+	r700->PA_SU_POLY_OFFSET_BACK_SCALE.f32All = factor;
+	r700->PA_SU_POLY_OFFSET_BACK_OFFSET.f32All = constant;
+}
+
+static void r700UpdatePolygonMode(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	R600_STATECHANGE(context, su);
+
+	SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DISABLE_POLY_MODE, POLY_MODE_shift, POLY_MODE_mask);
+
+	/* Only do something if a polygon mode is wanted, default is GL_FILL */
+	if (ctx->Polygon.FrontMode != GL_FILL ||
+	    ctx->Polygon.BackMode != GL_FILL) {
+		GLenum f, b;
+
+		/* Handle GL_CW (clock wise and GL_CCW (counter clock wise)
+		 * correctly by selecting the correct front and back face
+		 */
+		if (ctx->Polygon.FrontFace == GL_CCW) {
+			f = ctx->Polygon.FrontMode;
+			b = ctx->Polygon.BackMode;
+		} else {
+			f = ctx->Polygon.BackMode;
+			b = ctx->Polygon.FrontMode;
+		}
+
+		/* Enable polygon mode */
+		SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DUAL_MODE, POLY_MODE_shift, POLY_MODE_mask);
+
+		switch (f) {
+		case GL_LINE:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_LINES,
+				 POLYMODE_FRONT_PTYPE_shift, POLYMODE_FRONT_PTYPE_mask);
+			break;
+		case GL_POINT:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_POINTS,
+				 POLYMODE_FRONT_PTYPE_shift, POLYMODE_FRONT_PTYPE_mask);
+			break;
+		case GL_FILL:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_TRIANGLES,
+				 POLYMODE_FRONT_PTYPE_shift, POLYMODE_FRONT_PTYPE_mask);
+			break;
+		}
+
+		switch (b) {
+		case GL_LINE:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_LINES,
+				 POLYMODE_BACK_PTYPE_shift, POLYMODE_BACK_PTYPE_mask);
+			break;
+		case GL_POINT:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_POINTS,
+				 POLYMODE_BACK_PTYPE_shift, POLYMODE_BACK_PTYPE_mask);
+			break;
+		case GL_FILL:
+			SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_TRIANGLES,
+				 POLYMODE_BACK_PTYPE_shift, POLYMODE_BACK_PTYPE_mask);
+			break;
+		}
+	}
+}
+
+static void r700PolygonMode(GLcontext * ctx, GLenum face, GLenum mode) //------------------
+{
+	(void)face;
+	(void)mode;
+
+	r700UpdatePolygonMode(ctx);
+}
+
+static void r700RenderMode(GLcontext * ctx, GLenum mode) //---------------------
+{
+}
+
+static void r700ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	GLint p;
+	GLint *ip;
+
+	p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
+	ip = (GLint *)ctx->Transform._ClipUserPlane[p];
+
+	R600_STATECHANGE(context, ucp);
+
+	r700->ucp[p].PA_CL_UCP_0_X.u32All = ip[0];
+	r700->ucp[p].PA_CL_UCP_0_Y.u32All = ip[1];
+	r700->ucp[p].PA_CL_UCP_0_Z.u32All = ip[2];
+	r700->ucp[p].PA_CL_UCP_0_W.u32All = ip[3];
+}
+
+static void r700SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	GLuint p;
+
+	p = cap - GL_CLIP_PLANE0;
+
+	R600_STATECHANGE(context, cl);
+
+	if (state) {
+		r700->PA_CL_CLIP_CNTL.u32All |= (UCP_ENA_0_bit << p);
+		r700->ucp[p].enabled = GL_TRUE;
+		r700ClipPlane(ctx, cap, NULL);
+	} else {
+		r700->PA_CL_CLIP_CNTL.u32All &= ~(UCP_ENA_0_bit << p);
+		r700->ucp[p].enabled = GL_FALSE;
+	}
+}
+
+void r700SetScissor(context_t *context) //---------------
+{
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned x1, y1, x2, y2;
+	int id = 0;
+	struct radeon_renderbuffer *rrb;
+
+	rrb = radeon_get_colorbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		return;
+	}
+	if (context->radeon.state.scissor.enabled) {
+		x1 = context->radeon.state.scissor.rect.x1;
+		y1 = context->radeon.state.scissor.rect.y1;
+		x2 = context->radeon.state.scissor.rect.x2;
+		y2 = context->radeon.state.scissor.rect.y2;
+	} else {
+		if (context->radeon.radeonScreen->driScreen->dri2.enabled) {
+			x1 = 0;
+			y1 = 0;
+			x2 = rrb->base.Width - 1;
+			y2 = rrb->base.Height - 1;
+		} else {
+			x1 = rrb->dPriv->x;
+			y1 = rrb->dPriv->y;
+			x2 = rrb->dPriv->x + rrb->dPriv->w;
+			y2 = rrb->dPriv->y + rrb->dPriv->h;
+		}
+	}
+
+	R600_STATECHANGE(context, scissor);
+
+	/* screen */
+	SETbit(r700->PA_SC_SCREEN_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->PA_SC_SCREEN_SCISSOR_TL.u32All, x1,
+		 PA_SC_SCREEN_SCISSOR_TL__TL_X_shift, PA_SC_SCREEN_SCISSOR_TL__TL_X_mask);
+	SETfield(r700->PA_SC_SCREEN_SCISSOR_TL.u32All, y1,
+		 PA_SC_SCREEN_SCISSOR_TL__TL_Y_shift, PA_SC_SCREEN_SCISSOR_TL__TL_Y_mask);
+
+	SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, x2,
+		 PA_SC_SCREEN_SCISSOR_BR__BR_X_shift, PA_SC_SCREEN_SCISSOR_BR__BR_X_mask);
+	SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, y2,
+		 PA_SC_SCREEN_SCISSOR_BR__BR_Y_shift, PA_SC_SCREEN_SCISSOR_BR__BR_Y_mask);
+
+	/* window */
+	SETbit(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, x1,
+		 PA_SC_WINDOW_SCISSOR_TL__TL_X_shift, PA_SC_WINDOW_SCISSOR_TL__TL_X_mask);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, y1,
+		 PA_SC_WINDOW_SCISSOR_TL__TL_Y_shift, PA_SC_WINDOW_SCISSOR_TL__TL_Y_mask);
+
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, x2,
+		 PA_SC_WINDOW_SCISSOR_BR__BR_X_shift, PA_SC_WINDOW_SCISSOR_BR__BR_X_mask);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, y2,
+		 PA_SC_WINDOW_SCISSOR_BR__BR_Y_shift, PA_SC_WINDOW_SCISSOR_BR__BR_Y_mask);
+
+
+	SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, x1,
+		 PA_SC_CLIPRECT_0_TL__TL_X_shift, PA_SC_CLIPRECT_0_TL__TL_X_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, y1,
+		 PA_SC_CLIPRECT_0_TL__TL_Y_shift, PA_SC_CLIPRECT_0_TL__TL_Y_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, x2,
+		 PA_SC_CLIPRECT_0_BR__BR_X_shift, PA_SC_CLIPRECT_0_BR__BR_X_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, y2,
+		 PA_SC_CLIPRECT_0_BR__BR_Y_shift, PA_SC_CLIPRECT_0_BR__BR_Y_mask);
+
+	r700->PA_SC_CLIPRECT_1_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_1_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+	r700->PA_SC_CLIPRECT_2_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_2_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+	r700->PA_SC_CLIPRECT_3_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_3_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+
+	/* more....2d clip */
+	SETbit(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, x1,
+		 PA_SC_GENERIC_SCISSOR_TL__TL_X_shift, PA_SC_GENERIC_SCISSOR_TL__TL_X_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, y1,
+		 PA_SC_GENERIC_SCISSOR_TL__TL_Y_shift, PA_SC_GENERIC_SCISSOR_TL__TL_Y_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, x2,
+		 PA_SC_GENERIC_SCISSOR_BR__BR_X_shift, PA_SC_GENERIC_SCISSOR_BR__BR_X_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, y2,
+		 PA_SC_GENERIC_SCISSOR_BR__BR_Y_shift, PA_SC_GENERIC_SCISSOR_BR__BR_Y_mask);
+
+	SETbit(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, x1,
+		 PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, y1,
+		 PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All, x2,
+		 PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All, y2,
+		 PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask);
+
+	r700->viewport[id].PA_SC_VPORT_ZMIN_0.u32All = 0;
+	r700->viewport[id].PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000;
+	r700->viewport[id].enabled = GL_TRUE;
+}
+
+static void r700InitSQConfig(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    int ps_prio;
+    int vs_prio;
+    int gs_prio;
+    int es_prio;
+    int num_ps_gprs;
+    int num_vs_gprs;
+    int num_gs_gprs;
+    int num_es_gprs;
+    int num_temp_gprs;
+    int num_ps_threads;
+    int num_vs_threads;
+    int num_gs_threads;
+    int num_es_threads;
+    int num_ps_stack_entries;
+    int num_vs_stack_entries;
+    int num_gs_stack_entries;
+    int num_es_stack_entries;
+
+    R600_STATECHANGE(context, sq);
+
+    // SQ
+    ps_prio = 0;
+    vs_prio = 1;
+    gs_prio = 2;
+    es_prio = 3;
+    switch (context->radeon.radeonScreen->chip_family) {
+    case CHIP_FAMILY_R600:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV630:
+    case CHIP_FAMILY_RV635:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 144;
+	    num_vs_threads = 40;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV610:
+    case CHIP_FAMILY_RV620:
+    case CHIP_FAMILY_RS780:
+    case CHIP_FAMILY_RS880:
+    default:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV670:
+	    num_ps_gprs = 144;
+	    num_vs_gprs = 40;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV770:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 188;
+	    num_vs_threads = 60;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 256;
+	    num_vs_stack_entries = 256;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV730:
+    case CHIP_FAMILY_RV740:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 188;
+	    num_vs_threads = 60;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV710:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 144;
+	    num_vs_threads = 48;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    }
+
+    r700->sq_config.SQ_CONFIG.u32All = 0;
+    if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) ||
+        (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS880) ||
+        (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710))
+	    CLEARbit(r700->sq_config.SQ_CONFIG.u32All, VC_ENABLE_bit);
+    else
+	    SETbit(r700->sq_config.SQ_CONFIG.u32All, VC_ENABLE_bit);
+    SETbit(r700->sq_config.SQ_CONFIG.u32All, DX9_CONSTS_bit);
+    SETbit(r700->sq_config.SQ_CONFIG.u32All, ALU_INST_PREFER_VECTOR_bit);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, PS_PRIO_shift, PS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, VS_PRIO_shift, VS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, GS_PRIO_shift, GS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, ES_PRIO_shift, ES_PRIO_mask);
+
+    r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All = 0;
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_ps_gprs, NUM_PS_GPRS_shift, NUM_PS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_vs_gprs, NUM_VS_GPRS_shift, NUM_VS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_temp_gprs,
+	     NUM_CLAUSE_TEMP_GPRS_shift, NUM_CLAUSE_TEMP_GPRS_mask);
+
+    r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All = 0;
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All, num_gs_gprs, NUM_GS_GPRS_shift, NUM_GS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All, num_es_gprs, NUM_ES_GPRS_shift, NUM_ES_GPRS_mask);
+
+    r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All = 0;
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_ps_threads,
+	     NUM_PS_THREADS_shift, NUM_PS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_vs_threads,
+	     NUM_VS_THREADS_shift, NUM_VS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_gs_threads,
+	     NUM_GS_THREADS_shift, NUM_GS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_es_threads,
+	     NUM_ES_THREADS_shift, NUM_ES_THREADS_mask);
+
+    r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All = 0;
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All, num_ps_stack_entries,
+	     NUM_PS_STACK_ENTRIES_shift, NUM_PS_STACK_ENTRIES_mask);
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All, num_vs_stack_entries,
+	     NUM_VS_STACK_ENTRIES_shift, NUM_VS_STACK_ENTRIES_mask);
+
+    r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All = 0;
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All, num_gs_stack_entries,
+	     NUM_GS_STACK_ENTRIES_shift, NUM_GS_STACK_ENTRIES_mask);
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All, num_es_stack_entries,
+	     NUM_ES_STACK_ENTRIES_shift, NUM_ES_STACK_ENTRIES_mask);
+
+}
+
+/**
+ * Calculate initial hardware state and register state functions.
+ * Assumes that the command buffer and state atoms have been
+ * initialized already.
+ */
+void r700InitState(GLcontext * ctx) //-------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    int id = 0;
+
+    radeon_firevertices(&context->radeon);
+
+    r700->TA_CNTL_AUX.u32All = 0;
+    SETfield(r700->TA_CNTL_AUX.u32All, 28, TD_FIFO_CREDIT_shift, TD_FIFO_CREDIT_mask);
+    r700->VC_ENHANCE.u32All = 0;
+    r700->DB_WATERMARKS.u32All = 0;
+    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_FREE_shift, DEPTH_FREE_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 16, DEPTH_FLUSH_shift, DEPTH_FLUSH_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 0, FORCE_SUMMARIZE_shift, FORCE_SUMMARIZE_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_PENDING_FREE_shift, DEPTH_PENDING_FREE_mask);
+    r700->SQ_DYN_GPR_CNTL_PS_FLUSH_REQ.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+	    SETfield(r700->TA_CNTL_AUX.u32All, 3, GRADIENT_CREDIT_shift, GRADIENT_CREDIT_mask);
+	    r700->DB_DEBUG.u32All = 0x82000000;
+	    SETfield(r700->DB_WATERMARKS.u32All, 16, DEPTH_CACHELINE_FREE_shift, DEPTH_CACHELINE_FREE_mask);
+    } else {
+	    SETfield(r700->TA_CNTL_AUX.u32All, 2, GRADIENT_CREDIT_shift, GRADIENT_CREDIT_mask);
+	    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_CACHELINE_FREE_shift, DEPTH_CACHELINE_FREE_mask);
+	    SETbit(r700->SQ_DYN_GPR_CNTL_PS_FLUSH_REQ.u32All, VS_PC_LIMIT_ENABLE_bit);
+    }
+
+    /* Turn off vgt reuse */
+    r700->VGT_REUSE_OFF.u32All = 0;
+    SETbit(r700->VGT_REUSE_OFF.u32All, REUSE_OFF_bit);
+
+    /* Specify offsetting and clamp values for vertices */
+    r700->VGT_MAX_VTX_INDX.u32All      = 0xFFFFFF;
+    r700->VGT_MIN_VTX_INDX.u32All      = 0;
+    r700->VGT_INDX_OFFSET.u32All    = 0;
+
+    /* default shader connections. */
+    r700->SPI_VS_OUT_ID_0.u32All  = 0x03020100;
+    r700->SPI_VS_OUT_ID_1.u32All  = 0x07060504;
+    r700->SPI_VS_OUT_ID_2.u32All  = 0x0b0a0908;
+    r700->SPI_VS_OUT_ID_3.u32All  = 0x0f0e0d0c;
+
+    r700->SPI_THREAD_GROUPING.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV770)
+	    SETfield(r700->SPI_THREAD_GROUPING.u32All, 1, PS_GROUPING_shift, PS_GROUPING_mask);
+
+    /* 4 clip rectangles */ /* TODO : set these clip rects according to context->currentDraw->numClipRects */
+    r700->PA_SC_CLIPRECT_RULE.u32All = 0;
+    SETfield(r700->PA_SC_CLIPRECT_RULE.u32All, CLIP_RULE_mask, CLIP_RULE_shift, CLIP_RULE_mask);
+
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+	    r700->PA_SC_EDGERULE.u32All = 0;
+    else
+	    r700->PA_SC_EDGERULE.u32All = 0xAAAAAAAA;
+
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+	    r700->PA_SC_MODE_CNTL.u32All = 0;
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, WALK_ORDER_ENABLE_bit);
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_CNTDWN_ENABLE_bit);
+    } else {
+	    r700->PA_SC_MODE_CNTL.u32All = 0x00500000;
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_REZ_ENABLE_bit);
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_CNTDWN_ENABLE_bit);
+    }
+
+    /* Do scale XY and Z by 1/W0. */
+    r700->bEnablePerspective = GL_TRUE;
+    CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+    CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+    /* Enable viewport scaling for all three axis */
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_X_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_X_OFFSET_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Y_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Y_OFFSET_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Z_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Z_OFFSET_ENA_bit);
+
+    /* GL uses last vtx for flat shading components */
+    SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, PROVOKING_VTX_LAST_bit);
+
+    /* Set up vertex control */
+    r700->PA_SU_VTX_CNTL.u32All = 0;
+    CLEARfield(r700->PA_SU_VTX_CNTL.u32All, QUANT_MODE_mask);
+    SETbit(r700->PA_SU_VTX_CNTL.u32All, PIX_CENTER_bit);
+    SETfield(r700->PA_SU_VTX_CNTL.u32All, X_ROUND_TO_EVEN,
+             PA_SU_VTX_CNTL__ROUND_MODE_shift, PA_SU_VTX_CNTL__ROUND_MODE_mask);
+
+    /* to 1.0 = no guard band */
+    r700->PA_CL_GB_VERT_CLIP_ADJ.u32All  = 0x3F800000;  /* 1.0 */
+    r700->PA_CL_GB_VERT_DISC_ADJ.u32All  = 0x3F800000;
+    r700->PA_CL_GB_HORZ_CLIP_ADJ.u32All  = 0x3F800000;
+    r700->PA_CL_GB_HORZ_DISC_ADJ.u32All  = 0x3F800000;
+
+    /* Enable all samples for multi-sample anti-aliasing */
+    r700->PA_SC_AA_MASK.u32All = 0xFFFFFFFF;
+    /* Turn off AA */
+    r700->PA_SC_AA_CONFIG.u32All = 0;
+
+    r700->SX_MISC.u32All = 0;
+
+    r700InitSQConfig(ctx);
+
+    r700ColorMask(ctx,
+		  ctx->Color.ColorMask[RCOMP],
+		  ctx->Color.ColorMask[GCOMP],
+		  ctx->Color.ColorMask[BCOMP],
+		  ctx->Color.ColorMask[ACOMP]);
+
+    r700Enable(ctx, GL_DEPTH_TEST, ctx->Depth.Test);
+    r700DepthMask(ctx, ctx->Depth.Mask);
+    r700DepthFunc(ctx, ctx->Depth.Func);
+    SETbit(r700->DB_SHADER_CONTROL.u32All, DUAL_EXPORT_ENABLE_bit);
+
+    r700->DB_DEPTH_CLEAR.u32All     = 0x3F800000;
+
+    r700->DB_RENDER_CONTROL.u32All  = 0;
+    SETbit(r700->DB_RENDER_CONTROL.u32All, STENCIL_COMPRESS_DISABLE_bit);
+    SETbit(r700->DB_RENDER_CONTROL.u32All, DEPTH_COMPRESS_DISABLE_bit);
+    r700->DB_RENDER_OVERRIDE.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+	    SETbit(r700->DB_RENDER_OVERRIDE.u32All, FORCE_SHADER_Z_ORDER_bit);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask);
+
+    r700->DB_ALPHA_TO_MASK.u32All = 0;
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET0_shift, ALPHA_TO_MASK_OFFSET0_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET1_shift, ALPHA_TO_MASK_OFFSET1_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET2_shift, ALPHA_TO_MASK_OFFSET2_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET3_shift, ALPHA_TO_MASK_OFFSET3_mask);
+
+    /* stencil */
+    r700Enable(ctx, GL_STENCIL_TEST, ctx->Stencil._Enabled);
+    r700StencilMaskSeparate(ctx, 0, ctx->Stencil.WriteMask[0]);
+    r700StencilFuncSeparate(ctx, 0, ctx->Stencil.Function[0],
+			    ctx->Stencil.Ref[0], ctx->Stencil.ValueMask[0]);
+    r700StencilOpSeparate(ctx, 0, ctx->Stencil.FailFunc[0],
+			  ctx->Stencil.ZFailFunc[0],
+			  ctx->Stencil.ZPassFunc[0]);
+
+    r700UpdateCulling(ctx);
+
+    r700SetBlendState(ctx);
+    r700SetLogicOpState(ctx);
+
+    r700AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef);
+    r700Enable(ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled);
+
+    r700PointSize(ctx, 1.0);
+
+    CLEARfield(r700->PA_SU_POINT_MINMAX.u32All, MIN_SIZE_mask);
+    SETfield(r700->PA_SU_POINT_MINMAX.u32All, 0x8000, MAX_SIZE_shift, MAX_SIZE_mask);
+
+    r700LineWidth(ctx, 1.0);
+
+    r700->PA_SC_LINE_CNTL.u32All = 0;
+    CLEARbit(r700->PA_SC_LINE_CNTL.u32All, EXPAND_LINE_WIDTH_bit);
+    SETbit(r700->PA_SC_LINE_CNTL.u32All, LAST_PIXEL_bit);
+
+    r700ShadeModel(ctx, ctx->Light.ShadeModel);
+    r700PolygonMode(ctx, GL_FRONT, ctx->Polygon.FrontMode);
+    r700PolygonMode(ctx, GL_BACK, ctx->Polygon.BackMode);
+    r700PolygonOffset(ctx, ctx->Polygon.OffsetFactor,
+		      ctx->Polygon.OffsetUnits);
+    r700Enable(ctx, GL_POLYGON_OFFSET_POINT, ctx->Polygon.OffsetPoint);
+    r700Enable(ctx, GL_POLYGON_OFFSET_LINE, ctx->Polygon.OffsetLine);
+    r700Enable(ctx, GL_POLYGON_OFFSET_FILL, ctx->Polygon.OffsetFill);
+
+    /* CB */
+    r700BlendColor(ctx, ctx->Color.BlendColor);
+
+    r700->CB_CLEAR_RED_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_CLEAR_GREEN_R6XX.f32All = 0.0; //r6xx only
+    r700->CB_CLEAR_BLUE_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_CLEAR_ALPHA_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_FOG_RED_R6XX.u32All = 0; //r6xx only
+    r700->CB_FOG_GREEN_R6XX.u32All = 0; //r6xx only
+    r700->CB_FOG_BLUE_R6XX.u32All = 0; //r6xx only
+
+    /* Disable color compares */
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_DRAW_ALWAYS,
+             CLRCMP_FCN_SRC_shift, CLRCMP_FCN_SRC_mask);
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_DRAW_ALWAYS,
+             CLRCMP_FCN_DST_shift, CLRCMP_FCN_DST_mask);
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_SEL_SRC,
+             CLRCMP_FCN_SEL_shift, CLRCMP_FCN_SEL_mask);
+
+    /* Zero out source */
+    r700->CB_CLRCMP_SRC.u32All = 0x00000000;
+
+    /* Put a compare color in for error checking */
+    r700->CB_CLRCMP_DST.u32All = 0x000000FF;
+
+    /* Set up color compare mask */
+    r700->CB_CLRCMP_MSK.u32All = 0xFFFFFFFF;
+
+    /* screen/window/view */
+    SETfield(r700->CB_TARGET_MASK.u32All, 0xF, (4 * id), TARGET0_ENABLE_mask);
+
+    context->radeon.hw.all_dirty = GL_TRUE;
+
+}
+
+void r700InitStateFuncs(struct dd_function_table *functions) //-----------------
+{
+	functions->UpdateState = r700InvalidateState;
+	functions->AlphaFunc = r700AlphaFunc;
+	functions->BlendColor = r700BlendColor;
+	functions->BlendEquationSeparate = r700BlendEquationSeparate;
+	functions->BlendFuncSeparate = r700BlendFuncSeparate;
+	functions->Enable = r700Enable;
+	functions->ColorMask = r700ColorMask;
+	functions->DepthFunc = r700DepthFunc;
+	functions->DepthMask = r700DepthMask;
+	functions->CullFace = r700CullFace;
+	functions->Fogfv = r700Fogfv;
+	functions->FrontFace = r700FrontFace;
+	functions->ShadeModel = r700ShadeModel;
+	functions->LogicOpcode = r700LogicOpcode;
+
+	/* ARB_point_parameters */
+	functions->PointParameterfv = r700PointParameter;
+
+	/* Stencil related */
+	functions->StencilFuncSeparate = r700StencilFuncSeparate;
+	functions->StencilMaskSeparate = r700StencilMaskSeparate;
+	functions->StencilOpSeparate = r700StencilOpSeparate;
+
+	/* Viewport related */
+	functions->Viewport = r700Viewport;
+	functions->DepthRange = r700DepthRange;
+	functions->PointSize = r700PointSize;
+	functions->LineWidth = r700LineWidth;
+	functions->LineStipple = r700LineStipple;
+
+	functions->PolygonOffset = r700PolygonOffset;
+	functions->PolygonMode = r700PolygonMode;
+
+	functions->RenderMode = r700RenderMode;
+
+	functions->ClipPlane = r700ClipPlane;
+
+	functions->Scissor = radeonScissor;
+
+	functions->DrawBuffer		= radeonDrawBuffer;
+	functions->ReadBuffer		= radeonReadBuffer;
+
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_state.h b/src/mesa/drivers/dri/r600/r700_state.h
new file mode 100644
index 0000000000..0f53d5b4c5
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_state.h
@@ -0,0 +1,46 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_STATE_H
+#define _R700_STATE_H
+
+#include "main/mtypes.h"
+
+#include "r600_context.h"
+
+#include "r700_chip.h"
+
+extern void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state);
+extern void r700UpdateShaders (GLcontext * ctx);
+
+extern void r700UpdateViewportOffset(GLcontext * ctx);
+
+extern void r700InitState (GLcontext * ctx);
+extern void r700InitStateFuncs (struct dd_function_table *functions);
+
+extern void r700SetScissor(context_t *context);
+
+#endif	/* _R600_SCREEN_H */
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
new file mode 100644
index 0000000000..d107f99e7b
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -0,0 +1,437 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "tnl/t_context.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+
+#include "radeon_debug.h"
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_debug.h"
+#include "r700_vertprog.h"
+
+unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm, 
+					           struct gl_vertex_program *mesa_vp,
+					           unsigned int unStart)
+{
+    unsigned int i;
+	unsigned int unBit;
+	unsigned int unTotal = unStart;
+
+    //!!!!!!! THE ORDER MATCH FS INPUT
+
+	unBit = 1 << VERT_RESULT_HPOS;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_HPOS] = unTotal++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL0;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_COL0] = unTotal++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL1;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_COL1] = unTotal++;
+	}
+
+	//TODO : dealing back face.
+	unBit = 1 << VERT_RESULT_BFC0;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_BFC0] = unTotal++;
+	}
+
+	unBit = 1 << VERT_RESULT_BFC1;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_BFC1] = unTotal++;
+	}
+
+	//TODO : dealing fog.
+	unBit = 1 << VERT_RESULT_FOGC;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_FOGC] = unTotal++;
+	}
+
+	//TODO : dealing point size.
+	unBit = 1 << VERT_RESULT_PSIZ;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_PSIZ] = unTotal++;
+	}
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (VERT_RESULT_TEX0 + i);
+		if(mesa_vp->Base.OutputsWritten & unBit)
+		{
+			pAsm->ucVP_OutputMap[VERT_RESULT_TEX0 + i] = unTotal++;
+		}
+	}
+
+	return (unTotal - unStart);
+}
+
+unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm, 
+					  struct gl_vertex_program *mesa_vp,
+					  unsigned int unStart)
+{
+	int i;
+	unsigned int unBit;
+	unsigned int unTotal = unStart;
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(mesa_vp->Base.InputsRead & unBit)
+		{
+			pAsm->ucVP_AttributeMap[i] = unTotal++;
+		}
+	}
+	return (unTotal - unStart);
+}
+
+GLboolean Process_Vertex_Program_Vfetch_Instructions(
+						struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp)
+{
+	int i;
+    unsigned int unBit;
+	VTX_FETCH_METHOD vtxFetchMethod;
+	vtxFetchMethod.bEnableMini          = GL_FALSE;
+	vtxFetchMethod.mega_fetch_remainder = 0;
+
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(mesa_vp->Base.InputsRead & unBit)
+		{
+			assemble_vfetch_instruction(&vp->r700AsmCode,
+						    i,
+						    vp->r700AsmCode.ucVP_AttributeMap[i],
+						    vp->aos_desc[i].size,
+						    vp->aos_desc[i].type,
+						    &vtxFetchMethod);
+		}
+	}
+	
+	return GL_TRUE;
+}
+
+void Map_Vertex_Program(struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp)
+{
+    GLuint ui;
+    r700_AssemblerBase *pAsm = &(vp->r700AsmCode);
+	unsigned int num_inputs;
+
+	// R0 will always be used for index into vertex buffer
+	pAsm->number_used_registers = 1;
+	pAsm->starting_vfetch_register_number = pAsm->number_used_registers;
+
+    // Map Inputs: Add 1 to mapping since R0 is used for index
+	num_inputs = Map_Vertex_Input(pAsm, mesa_vp, pAsm->number_used_registers);
+	pAsm->number_used_registers += num_inputs;
+
+	// Create VFETCH instructions for inputs
+	if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions(vp, mesa_vp) ) 
+	{
+		radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions return error. \n");
+		return; //error
+	}
+
+	// Map Outputs
+	pAsm->number_of_exports = Map_Vertex_Output(pAsm, mesa_vp, pAsm->number_used_registers);
+
+	pAsm->starting_export_register_number = pAsm->number_used_registers;
+
+	pAsm->number_used_registers += pAsm->number_of_exports;
+
+    pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports);
+    
+    for(ui=0; ui<pAsm->number_of_exports; ui++)
+    {
+        pAsm->pucOutMask[ui] = 0x0;
+    }
+
+    /* Map temporary registers (GPRs) */
+    pAsm->starting_temp_register_number = pAsm->number_used_registers;
+
+    if(mesa_vp->Base.NumNativeTemporaries >= mesa_vp->Base.NumTemporaries)
+    {   /* arb uses NumNativeTemporaries */
+        pAsm->number_used_registers += mesa_vp->Base.NumNativeTemporaries;
+    }
+    else
+    {   /* fix func t_vp uses NumTemporaries */
+        pAsm->number_used_registers += mesa_vp->Base.NumTemporaries;
+    }
+	
+    pAsm->uFirstHelpReg = pAsm->number_used_registers;
+}
+
+GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
+					                	struct gl_vertex_program   *mesa_vp)
+{
+    GLuint i, j;
+    GLint * puiTEMPwrites;
+    struct prog_instruction *pILInst;
+    InstDeps         *pInstDeps;
+
+    puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_vp->Base.NumTemporaries);
+    for(i=0; i<mesa_vp->Base.NumTemporaries; i++)
+    {
+        puiTEMPwrites[i] = -1;
+    }
+
+    pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_vp->Base.NumInstructions);
+
+    for(i=0; i<mesa_vp->Base.NumInstructions; i++)
+    {
+        pInstDeps[i].nDstDep = -1;
+        pILInst = &(mesa_vp->Base.Instructions[i]);
+
+        //Dst
+        if(pILInst->DstReg.File == PROGRAM_TEMPORARY)
+        {
+            //Set lastwrite for the temp
+            puiTEMPwrites[pILInst->DstReg.Index] = i;
+        }
+
+        //Src
+        for(j=0; j<3; j++)
+        {
+            if(pILInst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            {
+                //Set dep.
+                pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index];
+            }
+            else
+            {
+                pInstDeps[i].nSrcDeps[j] = -1;
+            }
+        }
+    }
+
+    vp->r700AsmCode.pInstDeps = pInstDeps;
+
+    FREE(puiTEMPwrites);
+
+    return GL_TRUE;
+}
+
+GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
+							   struct gl_vertex_program   *mesa_vp)
+{
+	//Init_Program
+	Init_r700_AssemblerBase(SPT_VP, &(vp->r700AsmCode), &(vp->r700Shader) );
+	Map_Vertex_Program( vp, mesa_vp );
+
+	if(GL_FALSE == Find_Instruction_Dependencies_vp(vp, mesa_vp))
+	{
+		return GL_FALSE;
+    }
+
+	if(GL_FALSE == AssembleInstr(mesa_vp->Base.NumInstructions,
+                                 &(mesa_vp->Base.Instructions[0]), 
+                                 &(vp->r700AsmCode)) )
+	{
+		return GL_FALSE;
+	} 
+
+    if(GL_FALSE == Process_Vertex_Exports(&(vp->r700AsmCode), mesa_vp->Base.OutputsWritten) )
+    {
+        return GL_FALSE;
+    }
+
+    vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 
+                         : (vp->r700AsmCode.number_used_registers - 1);
+
+	vp->r700Shader.nParamExports = vp->r700AsmCode.number_of_exports;
+
+    vp->translated = GL_TRUE;
+
+	return GL_TRUE;
+}
+
+void r700SelectVertexShader(GLcontext *ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *vb = &tnl->vb;
+    unsigned int unBit;
+    unsigned int i;
+
+    if (context->radeon.NewGLState & (_NEW_PROGRAM_CONSTANTS|_NEW_PROGRAM))
+    {
+	vpc->needUpdateVF = 1;
+    }
+
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+        vpc->r700AsmCode.bR6xx = 1;
+    }
+
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(vpc->mesa_program.Base.InputsRead & unBit) /* ctx->Array.ArrayObj->xxxxxxx */
+		{
+			vpc->aos_desc[i].size   = vb->AttribPtr[i]->size;
+			vpc->aos_desc[i].stride = vb->AttribPtr[i]->size * sizeof(GL_FLOAT);/* when emit array, data is packed. vb->AttribPtr[i]->stride;*/
+			vpc->aos_desc[i].type   = GL_FLOAT;
+		}
+	}
+
+	if(GL_FALSE == vpc->translated) {
+		r700TranslateVertexShader(vpc, &(vpc->mesa_program) );
+	}
+}
+
+void * r700GetActiveVpShaderBo(GLcontext * ctx)
+{
+    struct r700_vertex_program *vp
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+    return vp->shaderbo;
+}
+
+GLboolean r700SetupVertexProgram(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    struct r700_vertex_program *vp
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+    struct gl_program_parameter_list *paramList;
+    unsigned int unNumParamData;
+    unsigned int ui;
+
+    if (vp->needUpdateVF)
+    {
+	vp->loaded = GL_FALSE;
+	vp->r700Shader.bNeedsAssembly = GL_TRUE;
+	Process_Vertex_Program_Vfetch_Instructions(vp, &(vp->mesa_program));
+	r600DeleteShader(ctx, vp->shaderbo);
+    }
+
+    if(GL_FALSE == vp->loaded)
+    {
+	    if(vp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    {
+		    Assemble( &(vp->r700Shader) );
+	    }
+
+        /* Load vp to gpu */
+        r600EmitShader(ctx,
+                       &(vp->shaderbo),
+                       (GLvoid *)(vp->r700Shader.pProgram),
+                       vp->r700Shader.uShaderBinaryDWORDSize,
+                       "VS");
+
+        vp->loaded = GL_TRUE;
+    }
+
+    DumpHwBinary(DUMP_VERTEX_SHADER, (GLvoid *)(vp->r700Shader.pProgram),
+                 vp->r700Shader.uShaderBinaryDWORDSize);
+
+    /* TODO : enable this after MemUse fixed *=
+    (context->chipobj.MemUse)(context, vp->shadercode.buf->id);
+    */
+
+    R600_STATECHANGE(context, vs);
+    R600_STATECHANGE(context, fs); /* hack */
+
+    r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
+    SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
+
+    r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */
+
+    SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1,
+             NUM_GPRS_shift, NUM_GPRS_mask);
+
+    if(vp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
+	{
+        SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.uStackSize,
+                 STACK_SIZE_shift, STACK_SIZE_mask);
+    }
+
+    R600_STATECHANGE(context, spi);
+
+    SETfield(r700->SPI_VS_OUT_CONFIG.u32All,
+	     vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0,
+             VS_EXPORT_COUNT_shift, VS_EXPORT_COUNT_mask);
+    SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports,
+             NUM_INTERP_shift, NUM_INTERP_mask);
+
+    /*
+    SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+    CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    */
+
+    /* sent out shader constants. */
+    paramList = vp->mesa_program.Base.Parameters;
+
+    if(NULL != paramList) {
+	    _mesa_load_state_parameters(ctx, paramList);
+
+	    if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
+		    return GL_FALSE;
+
+	    R600_STATECHANGE(context, vs_consts);
+
+	    r700->vs.num_consts = paramList->NumParameters;
+
+	    unNumParamData = paramList->NumParameters;
+
+	    for(ui=0; ui<unNumParamData; ui++) {
+		    r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0];
+		    r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1];
+		    r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2];
+		    r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3];
+	    }
+    } else
+	    r700->vs.num_consts = 0;
+
+    return GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.h b/src/mesa/drivers/dri/r600/r700_vertprog.h
new file mode 100644
index 0000000000..e2e65021fd
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.h
@@ -0,0 +1,87 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _R700_VERTPROG_H_
+#define _R700_VERTPROG_H_
+
+#include "main/glheader.h"
+#include "main/mtypes.h" 
+
+#include "r700_shader.h"
+#include "r700_assembler.h"
+
+typedef struct ArrayDesc //TEMP
+{
+	GLint size;   //number of data element
+	GLenum type;  //data element type
+	GLsizei stride;
+} ArrayDesc;
+
+struct r700_vertex_program 
+{
+    struct gl_vertex_program mesa_program; /* Must be first */
+
+    struct r700_vertex_program *next;
+
+    r700_AssemblerBase r700AsmCode;
+    R700_Shader        r700Shader;
+
+    GLboolean translated;
+    GLboolean loaded;
+    GLboolean needUpdateVF;
+	
+    void * shaderbo;
+
+    ArrayDesc              aos_desc[VERT_ATTRIB_MAX];
+};
+
+//Internal
+unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm, 
+			       struct gl_vertex_program *mesa_vp,
+			       unsigned int unStart);
+unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm, 
+			      struct gl_vertex_program *mesa_vp,
+			      unsigned int unStart);
+GLboolean Process_Vertex_Program_Vfetch_Instructions(
+	struct r700_vertex_program *vp,
+	struct gl_vertex_program   *mesa_vp);
+void Map_Vertex_Program(struct r700_vertex_program *vp,
+			struct gl_vertex_program   *mesa_vp);
+GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
+					   struct gl_vertex_program   *mesa_vp);
+
+GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
+				    struct gl_vertex_program   *mesa_vp);
+
+/* Interface */
+extern void r700SelectVertexShader(GLcontext *ctx);
+
+extern GLboolean r700SetupVertexProgram(GLcontext * ctx);
+
+extern void *    r700GetActiveVpShaderBo(GLcontext * ctx);
+
+#endif /* _R700_VERTPROG_H_ */
diff --git a/src/mesa/drivers/dri/r600/radeon_bo_legacy.c b/src/mesa/drivers/dri/r600/radeon_bo_legacy.c
new file mode 120000
index 0000000000..79ad050e6b
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_bo_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_bo_legacy.h b/src/mesa/drivers/dri/r600/radeon_bo_legacy.h
new file mode 120000
index 0000000000..83b0f7ffab
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_bo_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_bo_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/r600/radeon_bocs_wrapper.h
new file mode 120000
index 0000000000..ca894b2443
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_bocs_wrapper.h
@@ -0,0 +1 @@
+../radeon/radeon_bocs_wrapper.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_chipset.h b/src/mesa/drivers/dri/r600/radeon_chipset.h
new file mode 120000
index 0000000000..eba99001ff
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_chipset.h
@@ -0,0 +1 @@
+../radeon/radeon_chipset.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_cmdbuf.h b/src/mesa/drivers/dri/r600/radeon_cmdbuf.h
new file mode 120000
index 0000000000..a799e1dc6d
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_cmdbuf.h
@@ -0,0 +1 @@
+../radeon/radeon_cmdbuf.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_common.c b/src/mesa/drivers/dri/r600/radeon_common.c
new file mode 120000
index 0000000000..67b19ba940
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_common.c
@@ -0,0 +1 @@
+../radeon/radeon_common.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_common.h b/src/mesa/drivers/dri/r600/radeon_common.h
new file mode 120000
index 0000000000..5bcb696a9f
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_common.h
@@ -0,0 +1 @@
+../radeon/radeon_common.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_common_context.c b/src/mesa/drivers/dri/r600/radeon_common_context.c
new file mode 120000
index 0000000000..86800f3819
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_common_context.c
@@ -0,0 +1 @@
+../radeon/radeon_common_context.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_common_context.h b/src/mesa/drivers/dri/r600/radeon_common_context.h
new file mode 120000
index 0000000000..4d66312550
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_common_context.h
@@ -0,0 +1 @@
+../radeon/radeon_common_context.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_cs_legacy.c b/src/mesa/drivers/dri/r600/radeon_cs_legacy.c
new file mode 120000
index 0000000000..006720f8a4
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_cs_legacy.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_cs_legacy.h b/src/mesa/drivers/dri/r600/radeon_cs_legacy.h
new file mode 120000
index 0000000000..a5f95e0a3d
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_cs_legacy.h
@@ -0,0 +1 @@
+../radeon/radeon_cs_legacy.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_cs_space_drm.c b/src/mesa/drivers/dri/r600/radeon_cs_space_drm.c
new file mode 120000
index 0000000000..c248ea7d1a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_cs_space_drm.c
@@ -0,0 +1 @@
+../radeon/radeon_cs_space_drm.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_debug.c b/src/mesa/drivers/dri/r600/radeon_debug.c
new file mode 120000
index 0000000000..c98c2e074c
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_debug.c
@@ -0,0 +1 @@
+../radeon/radeon_debug.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_debug.h b/src/mesa/drivers/dri/r600/radeon_debug.h
new file mode 120000
index 0000000000..bd8aa28e89
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_debug.h
@@ -0,0 +1 @@
+../radeon/radeon_debug.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_dma.c b/src/mesa/drivers/dri/r600/radeon_dma.c
new file mode 120000
index 0000000000..43be000625
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_dma.c
@@ -0,0 +1 @@
+../radeon/radeon_dma.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_dma.h b/src/mesa/drivers/dri/r600/radeon_dma.h
new file mode 120000
index 0000000000..82e50634e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_dma.h
@@ -0,0 +1 @@
+../radeon/radeon_dma.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_fbo.c b/src/mesa/drivers/dri/r600/radeon_fbo.c
new file mode 120000
index 0000000000..0d738d8d78
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_fbo.c
@@ -0,0 +1 @@
+../radeon/radeon_fbo.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_lock.c b/src/mesa/drivers/dri/r600/radeon_lock.c
new file mode 120000
index 0000000000..af4108a8e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_lock.c
@@ -0,0 +1 @@
+../radeon/radeon_lock.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_lock.h b/src/mesa/drivers/dri/r600/radeon_lock.h
new file mode 120000
index 0000000000..64bdf94ee7
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_lock.h
@@ -0,0 +1 @@
+../radeon/radeon_lock.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_mipmap_tree.c b/src/mesa/drivers/dri/r600/radeon_mipmap_tree.c
new file mode 120000
index 0000000000..31c0cfbe94
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_mipmap_tree.c
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_mipmap_tree.h b/src/mesa/drivers/dri/r600/radeon_mipmap_tree.h
new file mode 120000
index 0000000000..254d50cf8c
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_mipmap_tree.h
@@ -0,0 +1 @@
+../radeon/radeon_mipmap_tree.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_queryobj.c b/src/mesa/drivers/dri/r600/radeon_queryobj.c
new file mode 120000
index 0000000000..1d6ebc1c48
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_queryobj.c
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_queryobj.h b/src/mesa/drivers/dri/r600/radeon_queryobj.h
new file mode 120000
index 0000000000..8f6f842b0a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_queryobj.h
@@ -0,0 +1 @@
+../radeon/radeon_queryobj.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_screen.c b/src/mesa/drivers/dri/r600/radeon_screen.c
new file mode 120000
index 0000000000..86161118dd
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_screen.c
@@ -0,0 +1 @@
+../radeon/radeon_screen.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_screen.h b/src/mesa/drivers/dri/r600/radeon_screen.h
new file mode 120000
index 0000000000..23bb6bd459
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_screen.h
@@ -0,0 +1 @@
+../radeon/radeon_screen.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_span.c b/src/mesa/drivers/dri/r600/radeon_span.c
new file mode 120000
index 0000000000..232868c4c9
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_span.c
@@ -0,0 +1 @@
+../radeon/radeon_span.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_span.h b/src/mesa/drivers/dri/r600/radeon_span.h
new file mode 120000
index 0000000000..f9d634508c
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_span.h
@@ -0,0 +1 @@
+../radeon/radeon_span.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_texture.c b/src/mesa/drivers/dri/r600/radeon_texture.c
new file mode 120000
index 0000000000..a822710915
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_texture.c
@@ -0,0 +1 @@
+../radeon/radeon_texture.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/radeon_texture.h b/src/mesa/drivers/dri/r600/radeon_texture.h
new file mode 120000
index 0000000000..17fac3d5ea
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/radeon_texture.h
@@ -0,0 +1 @@
+../radeon/radeon_texture.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon.h b/src/mesa/drivers/dri/r600/server/radeon.h
new file mode 120000
index 0000000000..81274a54f1
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon.h
@@ -0,0 +1 @@
+../../radeon/server/radeon.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon_dri.c b/src/mesa/drivers/dri/r600/server/radeon_dri.c
new file mode 120000
index 0000000000..d05847d650
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon_dri.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon_dri.h b/src/mesa/drivers/dri/r600/server/radeon_dri.h
new file mode 120000
index 0000000000..27c591d3c9
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon_dri.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_dri.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon_egl.c b/src/mesa/drivers/dri/r600/server/radeon_egl.c
new file mode 120000
index 0000000000..d7735a7643
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon_egl.c
@@ -0,0 +1 @@
+../../radeon/server/radeon_egl.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon_macros.h b/src/mesa/drivers/dri/r600/server/radeon_macros.h
new file mode 120000
index 0000000000..c56cd735b8
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon_macros.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_macros.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/server/radeon_reg.h b/src/mesa/drivers/dri/r600/server/radeon_reg.h
new file mode 120000
index 0000000000..e2349dcb68
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/server/radeon_reg.h
@@ -0,0 +1 @@
+../../radeon/server/radeon_reg.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r600/sq_micro_reg.h b/src/mesa/drivers/dri/r600/sq_micro_reg.h
new file mode 100644
index 0000000000..bfd21cef62
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/sq_micro_reg.h
@@ -0,0 +1,2008 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Contacts:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#if !defined (_SQ_MICRO_REG_H)
+#define _SQ_MICRO_REG_H
+
+#if defined(LITTLEENDIAN_CPU)
+#elif defined(BIGENDIAN_CPU)
+#else
+#error "BIGENDIAN_CPU or LITTLEENDIAN_CPU must be defined"
+#endif
+
+/*
+ * SQ_ALU_SRC_GPR_BASE value
+ */
+
+#define SQ_ALU_SRC_GPR_BASE            0x00000000
+
+/*
+ * SQ_ALU_SRC_GPR_SIZE value
+ */
+
+#define SQ_ALU_SRC_GPR_SIZE            0x00000080
+
+/*
+ * SQ_ALU_SRC_KCACHE0_BASE value
+ */
+
+#define SQ_ALU_SRC_KCACHE0_BASE        0x00000080
+
+/*
+ * SQ_ALU_SRC_KCACHE0_SIZE value
+ */
+
+#define SQ_ALU_SRC_KCACHE0_SIZE        0x00000020
+
+/*
+ * SQ_ALU_SRC_KCACHE1_BASE value
+ */
+
+#define SQ_ALU_SRC_KCACHE1_BASE        0x000000a0
+
+/*
+ * SQ_ALU_SRC_KCACHE1_SIZE value
+ */
+
+#define SQ_ALU_SRC_KCACHE1_SIZE        0x00000020
+
+/*
+ * SQ_ALU_SRC_CFILE_BASE value
+ */
+
+#define SQ_ALU_SRC_CFILE_BASE          0x00000100
+
+/*
+ * SQ_ALU_SRC_CFILE_SIZE value
+ */
+
+#define SQ_ALU_SRC_CFILE_SIZE          0x00000100
+
+/*
+ * SQ_SP_OP_REDUC_BEGIN value
+ */
+
+#define SQ_SP_OP_REDUC_BEGIN           0x00000050
+
+/*
+ * SQ_SP_OP_REDUC_END value
+ */
+
+#define SQ_SP_OP_REDUC_END             0x00000053
+
+/*
+ * SQ_SP_OP_TRANS_BEGIN value
+ */
+
+#define SQ_SP_OP_TRANS_BEGIN           0x00000060
+
+/*
+ * SQ_SP_OP_TRANS_END value
+ */
+
+#define SQ_SP_OP_TRANS_END             0x0000007f
+
+/*
+ * SQ_CF_WORD0 struct
+ */
+
+#define SQ_CF_WORD0_ADDR_SIZE          32
+
+#define SQ_CF_WORD0_ADDR_SHIFT         0
+
+#define SQ_CF_WORD0_ADDR_MASK          0xffffffff
+
+#define SQ_CF_WORD0_MASK \
+     (SQ_CF_WORD0_ADDR_MASK)
+
+#define SQ_CF_WORD0_DEFAULT            0xcdcdcdcd
+
+#define SQ_CF_WORD0_GET_ADDR(sq_cf_word0) \
+     ((sq_cf_word0 & SQ_CF_WORD0_ADDR_MASK) >> SQ_CF_WORD0_ADDR_SHIFT)
+
+#define SQ_CF_WORD0_SET_ADDR(sq_cf_word0_reg, addr) \
+     sq_cf_word0_reg = (sq_cf_word0_reg & ~SQ_CF_WORD0_ADDR_MASK) | (addr << SQ_CF_WORD0_ADDR_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_word0_t {
+          unsigned int addr                           : SQ_CF_WORD0_ADDR_SIZE;
+     } sq_cf_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_word0_t {
+          unsigned int addr                           : SQ_CF_WORD0_ADDR_SIZE;
+     } sq_cf_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_word0_t f;
+} sq_cf_word0_u;
+
+
+/*
+ * SQ_CF_WORD1 struct
+ */
+
+#define SQ_CF_WORD1_POP_COUNT_SIZE     3
+#define SQ_CF_WORD1_CF_CONST_SIZE      5
+#define SQ_CF_WORD1_COND_SIZE          2
+#define SQ_CF_WORD1_COUNT_SIZE         3
+#define SQ_CF_WORD1_CALL_COUNT_SIZE    6
+#define SQ_CF_WORD1_COUNT_3_SIZE       1
+#define SQ_CF_WORD1_END_OF_PROGRAM_SIZE 1
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE 1
+#define SQ_CF_WORD1_CF_INST_SIZE       7
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_WORD1_BARRIER_SIZE       1
+
+#define SQ_CF_WORD1_POP_COUNT_SHIFT    0
+#define SQ_CF_WORD1_CF_CONST_SHIFT     3
+#define SQ_CF_WORD1_COND_SHIFT         8
+#define SQ_CF_WORD1_COUNT_SHIFT        10
+#define SQ_CF_WORD1_CALL_COUNT_SHIFT   13
+#define SQ_CF_WORD1_COUNT_3_SHIFT      19
+#define SQ_CF_WORD1_END_OF_PROGRAM_SHIFT 21
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT 22
+#define SQ_CF_WORD1_CF_INST_SHIFT      23
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_WORD1_BARRIER_SHIFT      31
+
+#define SQ_CF_WORD1_POP_COUNT_MASK     0x00000007
+#define SQ_CF_WORD1_CF_CONST_MASK      0x000000f8
+#define SQ_CF_WORD1_COND_MASK          0x00000300
+#define SQ_CF_WORD1_COUNT_MASK         0x00001c00
+#define SQ_CF_WORD1_CALL_COUNT_MASK    0x0007e000
+#define SQ_CF_WORD1_COUNT_3_MASK       0x00080000
+#define SQ_CF_WORD1_END_OF_PROGRAM_MASK 0x00200000
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_MASK 0x00400000
+#define SQ_CF_WORD1_CF_INST_MASK       0x3f800000
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_WORD1_BARRIER_MASK       0x80000000
+
+#define SQ_CF_WORD1_MASK \
+     (SQ_CF_WORD1_POP_COUNT_MASK | \
+      SQ_CF_WORD1_CF_CONST_MASK | \
+      SQ_CF_WORD1_COND_MASK | \
+      SQ_CF_WORD1_COUNT_MASK | \
+      SQ_CF_WORD1_CALL_COUNT_MASK | \
+      SQ_CF_WORD1_COUNT_3_MASK | \
+      SQ_CF_WORD1_END_OF_PROGRAM_MASK | \
+      SQ_CF_WORD1_VALID_PIXEL_MODE_MASK | \
+      SQ_CF_WORD1_CF_INST_MASK | \
+      SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_WORD1_BARRIER_MASK)
+
+#define SQ_CF_WORD1_DEFAULT            0xcdcdcdcd
+
+#define SQ_CF_WORD1_GET_POP_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_POP_COUNT_MASK) >> SQ_CF_WORD1_POP_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_CF_CONST(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CF_CONST_MASK) >> SQ_CF_WORD1_CF_CONST_SHIFT)
+#define SQ_CF_WORD1_GET_COND(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COND_MASK) >> SQ_CF_WORD1_COND_SHIFT)
+#define SQ_CF_WORD1_GET_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COUNT_MASK) >> SQ_CF_WORD1_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_CALL_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CALL_COUNT_MASK) >> SQ_CF_WORD1_CALL_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_COUNT_3(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COUNT_3_MASK) >> SQ_CF_WORD1_COUNT_3_SHIFT)
+#define SQ_CF_WORD1_GET_END_OF_PROGRAM(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_END_OF_PROGRAM_MASK) >> SQ_CF_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_WORD1_GET_VALID_PIXEL_MODE(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_VALID_PIXEL_MODE_MASK) >> SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_WORD1_GET_CF_INST(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CF_INST_MASK) >> SQ_CF_WORD1_CF_INST_SHIFT)
+#define SQ_CF_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_WORD1_GET_BARRIER(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_BARRIER_MASK) >> SQ_CF_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_WORD1_SET_POP_COUNT(sq_cf_word1_reg, pop_count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_POP_COUNT_MASK) | (pop_count << SQ_CF_WORD1_POP_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_CF_CONST(sq_cf_word1_reg, cf_const) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CF_CONST_MASK) | (cf_const << SQ_CF_WORD1_CF_CONST_SHIFT)
+#define SQ_CF_WORD1_SET_COND(sq_cf_word1_reg, cond) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COND_MASK) | (cond << SQ_CF_WORD1_COND_SHIFT)
+#define SQ_CF_WORD1_SET_COUNT(sq_cf_word1_reg, count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COUNT_MASK) | (count << SQ_CF_WORD1_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_CALL_COUNT(sq_cf_word1_reg, call_count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CALL_COUNT_MASK) | (call_count << SQ_CF_WORD1_CALL_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_COUNT_3(sq_cf_word1_reg, count_3) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COUNT_3_MASK) | (count_3 << SQ_CF_WORD1_COUNT_3_SHIFT)
+#define SQ_CF_WORD1_SET_END_OF_PROGRAM(sq_cf_word1_reg, end_of_program) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_END_OF_PROGRAM_MASK) | (end_of_program << SQ_CF_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_WORD1_SET_VALID_PIXEL_MODE(sq_cf_word1_reg, valid_pixel_mode) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_VALID_PIXEL_MODE_MASK) | (valid_pixel_mode << SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_WORD1_SET_CF_INST(sq_cf_word1_reg, cf_inst) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_WORD1_CF_INST_SHIFT)
+#define SQ_CF_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_word1_reg, whole_quad_mode) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_WORD1_SET_BARRIER(sq_cf_word1_reg, barrier) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_BARRIER_MASK) | (barrier << SQ_CF_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_word1_t {
+          unsigned int pop_count                      : SQ_CF_WORD1_POP_COUNT_SIZE;
+          unsigned int cf_const                       : SQ_CF_WORD1_CF_CONST_SIZE;
+          unsigned int cond                           : SQ_CF_WORD1_COND_SIZE;
+          unsigned int count                          : SQ_CF_WORD1_COUNT_SIZE;
+          unsigned int call_count                     : SQ_CF_WORD1_CALL_COUNT_SIZE;
+          unsigned int count_3                        : SQ_CF_WORD1_COUNT_3_SIZE;
+          unsigned int                                : 1;
+          unsigned int end_of_program                 : SQ_CF_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_WORD1_BARRIER_SIZE;
+     } sq_cf_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_word1_t {
+          unsigned int barrier                        : SQ_CF_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_WORD1_CF_INST_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int end_of_program                 : SQ_CF_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int                                : 1;
+          unsigned int count_3                        : SQ_CF_WORD1_COUNT_3_SIZE;
+          unsigned int call_count                     : SQ_CF_WORD1_CALL_COUNT_SIZE;
+          unsigned int count                          : SQ_CF_WORD1_COUNT_SIZE;
+          unsigned int cond                           : SQ_CF_WORD1_COND_SIZE;
+          unsigned int cf_const                       : SQ_CF_WORD1_CF_CONST_SIZE;
+          unsigned int pop_count                      : SQ_CF_WORD1_POP_COUNT_SIZE;
+     } sq_cf_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_word1_t f;
+} sq_cf_word1_u;
+
+
+/*
+ * SQ_CF_ALU_WORD0 struct
+ */
+
+#define SQ_CF_ALU_WORD0_ADDR_SIZE      22
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE 4
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE 4
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE 2
+
+#define SQ_CF_ALU_WORD0_ADDR_SHIFT     0
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT 22
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT 26
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT 30
+
+#define SQ_CF_ALU_WORD0_ADDR_MASK      0x003fffff
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK 0x03c00000
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK 0x3c000000
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK 0xc0000000
+
+#define SQ_CF_ALU_WORD0_MASK \
+     (SQ_CF_ALU_WORD0_ADDR_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK)
+
+#define SQ_CF_ALU_WORD0_DEFAULT        0xcdcdcdcd
+
+#define SQ_CF_ALU_WORD0_GET_ADDR(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_ADDR_MASK) >> SQ_CF_ALU_WORD0_ADDR_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_BANK0(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK) >> SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_BANK1(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK) >> SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_MODE0(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK) >> SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT)
+
+#define SQ_CF_ALU_WORD0_SET_ADDR(sq_cf_alu_word0_reg, addr) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_ADDR_MASK) | (addr << SQ_CF_ALU_WORD0_ADDR_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_BANK0(sq_cf_alu_word0_reg, kcache_bank0) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK) | (kcache_bank0 << SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_BANK1(sq_cf_alu_word0_reg, kcache_bank1) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK) | (kcache_bank1 << SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_MODE0(sq_cf_alu_word0_reg, kcache_mode0) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK) | (kcache_mode0 << SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word0_t {
+          unsigned int addr                           : SQ_CF_ALU_WORD0_ADDR_SIZE;
+          unsigned int kcache_bank0                   : SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE;
+          unsigned int kcache_bank1                   : SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE;
+          unsigned int kcache_mode0                   : SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE;
+     } sq_cf_alu_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word0_t {
+          unsigned int kcache_mode0                   : SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE;
+          unsigned int kcache_bank1                   : SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE;
+          unsigned int kcache_bank0                   : SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE;
+          unsigned int addr                           : SQ_CF_ALU_WORD0_ADDR_SIZE;
+     } sq_cf_alu_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alu_word0_t f;
+} sq_cf_alu_word0_u;
+
+
+/*
+ * SQ_CF_ALU_WORD1 struct
+ */
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE 2
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE 8
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE 8
+#define SQ_CF_ALU_WORD1_COUNT_SIZE     7
+#define SQ_CF_ALU_WORD1_ALT_CONST_SIZE 1
+#define SQ_CF_ALU_WORD1_CF_INST_SIZE   4
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_ALU_WORD1_BARRIER_SIZE   1
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT 0
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT 2
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT 10
+#define SQ_CF_ALU_WORD1_COUNT_SHIFT    18
+#define SQ_CF_ALU_WORD1_ALT_CONST_SHIFT 25
+#define SQ_CF_ALU_WORD1_CF_INST_SHIFT  26
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_ALU_WORD1_BARRIER_SHIFT  31
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK 0x00000003
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK 0x000003fc
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK 0x0003fc00
+#define SQ_CF_ALU_WORD1_COUNT_MASK     0x01fc0000
+#define SQ_CF_ALU_WORD1_ALT_CONST_MASK 0x02000000
+#define SQ_CF_ALU_WORD1_CF_INST_MASK   0x3c000000
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_ALU_WORD1_BARRIER_MASK   0x80000000
+
+#define SQ_CF_ALU_WORD1_MASK \
+     (SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK | \
+      SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK | \
+      SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK | \
+      SQ_CF_ALU_WORD1_COUNT_MASK | \
+      SQ_CF_ALU_WORD1_ALT_CONST_MASK | \
+      SQ_CF_ALU_WORD1_CF_INST_MASK | \
+      SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_ALU_WORD1_BARRIER_MASK)
+
+#define SQ_CF_ALU_WORD1_DEFAULT        0xcdcdcdcd
+
+#define SQ_CF_ALU_WORD1_GET_KCACHE_MODE1(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK) >> SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_KCACHE_ADDR0(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK) >> SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_KCACHE_ADDR1(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK) >> SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_COUNT(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_COUNT_MASK) >> SQ_CF_ALU_WORD1_COUNT_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_ALT_CONST(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_ALT_CONST_MASK) >> SQ_CF_ALU_WORD1_ALT_CONST_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_CF_INST(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_CF_INST_MASK) >> SQ_CF_ALU_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_BARRIER(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_BARRIER_MASK) >> SQ_CF_ALU_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_ALU_WORD1_SET_KCACHE_MODE1(sq_cf_alu_word1_reg, kcache_mode1) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK) | (kcache_mode1 << SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_KCACHE_ADDR0(sq_cf_alu_word1_reg, kcache_addr0) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK) | (kcache_addr0 << SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_KCACHE_ADDR1(sq_cf_alu_word1_reg, kcache_addr1) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK) | (kcache_addr1 << SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_COUNT(sq_cf_alu_word1_reg, count) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_COUNT_MASK) | (count << SQ_CF_ALU_WORD1_COUNT_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_ALT_CONST(sq_cf_alu_word1_reg, alt_const) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_ALT_CONST_MASK) | (alt_const << SQ_CF_ALU_WORD1_ALT_CONST_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_CF_INST(sq_cf_alu_word1_reg, cf_inst) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_ALU_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_alu_word1_reg, whole_quad_mode) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_BARRIER(sq_cf_alu_word1_reg, barrier) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_BARRIER_MASK) | (barrier << SQ_CF_ALU_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word1_t {
+          unsigned int kcache_mode1                   : SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE;
+          unsigned int kcache_addr0                   : SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE;
+          unsigned int kcache_addr1                   : SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE;
+          unsigned int count                          : SQ_CF_ALU_WORD1_COUNT_SIZE;
+          unsigned int alt_const                      : SQ_CF_ALU_WORD1_ALT_CONST_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALU_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_ALU_WORD1_BARRIER_SIZE;
+     } sq_cf_alu_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word1_t {
+          unsigned int barrier                        : SQ_CF_ALU_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALU_WORD1_CF_INST_SIZE;
+          unsigned int alt_const                      : SQ_CF_ALU_WORD1_ALT_CONST_SIZE;
+          unsigned int count                          : SQ_CF_ALU_WORD1_COUNT_SIZE;
+          unsigned int kcache_addr1                   : SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE;
+          unsigned int kcache_addr0                   : SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE;
+          unsigned int kcache_mode1                   : SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE;
+     } sq_cf_alu_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alu_word1_t f;
+} sq_cf_alu_word1_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD0 struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE 13
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE 2
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE 2
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT 13
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT 15
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT 22
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT 23
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT 30
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK 0x00001fff
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK 0x00006000
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK 0x003f8000
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK 0x00400000
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK 0x3f800000
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK 0xc0000000
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_DEFAULT 0xcdcdcdcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_ARRAY_BASE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_TYPE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_RW_GPR(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_RW_REL(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_INDEX_GPR(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_ELEM_SIZE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_ARRAY_BASE(sq_cf_alloc_export_word0_reg, array_base) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK) | (array_base << SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_TYPE(sq_cf_alloc_export_word0_reg, type) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK) | (type << SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_RW_GPR(sq_cf_alloc_export_word0_reg, rw_gpr) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK) | (rw_gpr << SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_RW_REL(sq_cf_alloc_export_word0_reg, rw_rel) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK) | (rw_rel << SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_INDEX_GPR(sq_cf_alloc_export_word0_reg, index_gpr) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK) | (index_gpr << SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_ELEM_SIZE(sq_cf_alloc_export_word0_reg, elem_size) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK) | (elem_size << SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word0_t {
+          unsigned int array_base                     : SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE;
+          unsigned int type                           : SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE;
+          unsigned int rw_gpr                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE;
+          unsigned int rw_rel                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE;
+          unsigned int index_gpr                      : SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE;
+          unsigned int elem_size                      : SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE;
+     } sq_cf_alloc_export_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word0_t {
+          unsigned int elem_size                      : SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE;
+          unsigned int index_gpr                      : SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE;
+          unsigned int rw_rel                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE;
+          unsigned int rw_gpr                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE;
+          unsigned int type                           : SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE;
+          unsigned int array_base                     : SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE;
+     } sq_cf_alloc_export_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word0_t f;
+} sq_cf_alloc_export_word0_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1 struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE 4
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE 1
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT 17
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT 21
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT 22
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT 23
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT 31
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK 0x001e0000
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK 0x00200000
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK 0x00400000
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK 0x3f800000
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK 0x80000000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_DEFAULT 0xcdcc0000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_BURST_COUNT(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_END_OF_PROGRAM(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_VALID_PIXEL_MODE(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_CF_INST(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_BARRIER(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_BURST_COUNT(sq_cf_alloc_export_word1_reg, burst_count) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK) | (burst_count << SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_END_OF_PROGRAM(sq_cf_alloc_export_word1_reg, end_of_program) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK) | (end_of_program << SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_VALID_PIXEL_MODE(sq_cf_alloc_export_word1_reg, valid_pixel_mode) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK) | (valid_pixel_mode << SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_CF_INST(sq_cf_alloc_export_word1_reg, cf_inst) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_alloc_export_word1_reg, whole_quad_mode) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_BARRIER(sq_cf_alloc_export_word1_reg, barrier) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK) | (barrier << SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_t {
+          unsigned int                                : 17;
+          unsigned int burst_count                    : SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE;
+          unsigned int end_of_program                 : SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE;
+     } sq_cf_alloc_export_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_t {
+          unsigned int barrier                        : SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int end_of_program                 : SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int burst_count                    : SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE;
+          unsigned int                                : 17;
+     } sq_cf_alloc_export_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_t f;
+} sq_cf_alloc_export_word1_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1_BUF struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE 12
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE 4
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT 12
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK 0x00000fff
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK 0x0000f000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_DEFAULT 0x0000cdcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_GET_ARRAY_SIZE(sq_cf_alloc_export_word1_buf) \
+     ((sq_cf_alloc_export_word1_buf & SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_GET_COMP_MASK(sq_cf_alloc_export_word1_buf) \
+     ((sq_cf_alloc_export_word1_buf & SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_SET_ARRAY_SIZE(sq_cf_alloc_export_word1_buf_reg, array_size) \
+     sq_cf_alloc_export_word1_buf_reg = (sq_cf_alloc_export_word1_buf_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK) | (array_size << SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_SET_COMP_MASK(sq_cf_alloc_export_word1_buf_reg, comp_mask) \
+     sq_cf_alloc_export_word1_buf_reg = (sq_cf_alloc_export_word1_buf_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK) | (comp_mask << SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_buf_t {
+          unsigned int array_size                     : SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE;
+          unsigned int comp_mask                      : SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE;
+          unsigned int                                : 16;
+     } sq_cf_alloc_export_word1_buf_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_buf_t {
+          unsigned int                                : 16;
+          unsigned int comp_mask                      : SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE;
+          unsigned int array_size                     : SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE;
+     } sq_cf_alloc_export_word1_buf_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_buf_t f;
+} sq_cf_alloc_export_word1_buf_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1_SWIZ struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE 3
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT 6
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT 9
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK 0x00000007
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK 0x00000038
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK 0x000001c0
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK 0x00000e00
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_DEFAULT 0x00000dcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_X(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_Y(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_Z(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_W(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_X(sq_cf_alloc_export_word1_swiz_reg, sel_x) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK) | (sel_x << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_Y(sq_cf_alloc_export_word1_swiz_reg, sel_y) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK) | (sel_y << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_Z(sq_cf_alloc_export_word1_swiz_reg, sel_z) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK) | (sel_z << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_W(sq_cf_alloc_export_word1_swiz_reg, sel_w) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK) | (sel_w << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_swiz_t {
+          unsigned int sel_x                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE;
+          unsigned int sel_y                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE;
+          unsigned int sel_z                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE;
+          unsigned int sel_w                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE;
+          unsigned int                                : 20;
+     } sq_cf_alloc_export_word1_swiz_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_swiz_t {
+          unsigned int                                : 20;
+          unsigned int sel_w                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE;
+          unsigned int sel_z                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE;
+          unsigned int sel_y                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE;
+          unsigned int sel_x                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE;
+     } sq_cf_alloc_export_word1_swiz_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_swiz_t f;
+} sq_cf_alloc_export_word1_swiz_u;
+
+
+/*
+ * SQ_ALU_WORD0 struct
+ */
+
+#define SQ_ALU_WORD0_SRC0_SEL_SIZE     9
+#define SQ_ALU_WORD0_SRC0_REL_SIZE     1
+#define SQ_ALU_WORD0_SRC0_CHAN_SIZE    2
+#define SQ_ALU_WORD0_SRC0_NEG_SIZE     1
+#define SQ_ALU_WORD0_SRC1_SEL_SIZE     9
+#define SQ_ALU_WORD0_SRC1_REL_SIZE     1
+#define SQ_ALU_WORD0_SRC1_CHAN_SIZE    2
+#define SQ_ALU_WORD0_SRC1_NEG_SIZE     1
+#define SQ_ALU_WORD0_INDEX_MODE_SIZE   3
+#define SQ_ALU_WORD0_PRED_SEL_SIZE     2
+#define SQ_ALU_WORD0_LAST_SIZE         1
+
+#define SQ_ALU_WORD0_SRC0_SEL_SHIFT    0
+#define SQ_ALU_WORD0_SRC0_REL_SHIFT    9
+#define SQ_ALU_WORD0_SRC0_CHAN_SHIFT   10
+#define SQ_ALU_WORD0_SRC0_NEG_SHIFT    12
+#define SQ_ALU_WORD0_SRC1_SEL_SHIFT    13
+#define SQ_ALU_WORD0_SRC1_REL_SHIFT    22
+#define SQ_ALU_WORD0_SRC1_CHAN_SHIFT   23
+#define SQ_ALU_WORD0_SRC1_NEG_SHIFT    25
+#define SQ_ALU_WORD0_INDEX_MODE_SHIFT  26
+#define SQ_ALU_WORD0_PRED_SEL_SHIFT    29
+#define SQ_ALU_WORD0_LAST_SHIFT        31
+
+#define SQ_ALU_WORD0_SRC0_SEL_MASK     0x000001ff
+#define SQ_ALU_WORD0_SRC0_REL_MASK     0x00000200
+#define SQ_ALU_WORD0_SRC0_CHAN_MASK    0x00000c00
+#define SQ_ALU_WORD0_SRC0_NEG_MASK     0x00001000
+#define SQ_ALU_WORD0_SRC1_SEL_MASK     0x003fe000
+#define SQ_ALU_WORD0_SRC1_REL_MASK     0x00400000
+#define SQ_ALU_WORD0_SRC1_CHAN_MASK    0x01800000
+#define SQ_ALU_WORD0_SRC1_NEG_MASK     0x02000000
+#define SQ_ALU_WORD0_INDEX_MODE_MASK   0x1c000000
+#define SQ_ALU_WORD0_PRED_SEL_MASK     0x60000000
+#define SQ_ALU_WORD0_LAST_MASK         0x80000000
+
+#define SQ_ALU_WORD0_MASK \
+     (SQ_ALU_WORD0_SRC0_SEL_MASK | \
+      SQ_ALU_WORD0_SRC0_REL_MASK | \
+      SQ_ALU_WORD0_SRC0_CHAN_MASK | \
+      SQ_ALU_WORD0_SRC0_NEG_MASK | \
+      SQ_ALU_WORD0_SRC1_SEL_MASK | \
+      SQ_ALU_WORD0_SRC1_REL_MASK | \
+      SQ_ALU_WORD0_SRC1_CHAN_MASK | \
+      SQ_ALU_WORD0_SRC1_NEG_MASK | \
+      SQ_ALU_WORD0_INDEX_MODE_MASK | \
+      SQ_ALU_WORD0_PRED_SEL_MASK | \
+      SQ_ALU_WORD0_LAST_MASK)
+
+#define SQ_ALU_WORD0_DEFAULT           0xcdcdcdcd
+
+#define SQ_ALU_WORD0_GET_SRC0_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_SEL_MASK) >> SQ_ALU_WORD0_SRC0_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_REL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_REL_MASK) >> SQ_ALU_WORD0_SRC0_REL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_CHAN(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_CHAN_MASK) >> SQ_ALU_WORD0_SRC0_CHAN_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_NEG(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_NEG_MASK) >> SQ_ALU_WORD0_SRC0_NEG_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_SEL_MASK) >> SQ_ALU_WORD0_SRC1_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_REL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_REL_MASK) >> SQ_ALU_WORD0_SRC1_REL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_CHAN(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_CHAN_MASK) >> SQ_ALU_WORD0_SRC1_CHAN_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_NEG(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_NEG_MASK) >> SQ_ALU_WORD0_SRC1_NEG_SHIFT)
+#define SQ_ALU_WORD0_GET_INDEX_MODE(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_INDEX_MODE_MASK) >> SQ_ALU_WORD0_INDEX_MODE_SHIFT)
+#define SQ_ALU_WORD0_GET_PRED_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_PRED_SEL_MASK) >> SQ_ALU_WORD0_PRED_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_LAST(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_LAST_MASK) >> SQ_ALU_WORD0_LAST_SHIFT)
+
+#define SQ_ALU_WORD0_SET_SRC0_SEL(sq_alu_word0_reg, src0_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_SEL_MASK) | (src0_sel << SQ_ALU_WORD0_SRC0_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_REL(sq_alu_word0_reg, src0_rel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_REL_MASK) | (src0_rel << SQ_ALU_WORD0_SRC0_REL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_CHAN(sq_alu_word0_reg, src0_chan) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_CHAN_MASK) | (src0_chan << SQ_ALU_WORD0_SRC0_CHAN_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_NEG(sq_alu_word0_reg, src0_neg) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_NEG_MASK) | (src0_neg << SQ_ALU_WORD0_SRC0_NEG_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_SEL(sq_alu_word0_reg, src1_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_SEL_MASK) | (src1_sel << SQ_ALU_WORD0_SRC1_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_REL(sq_alu_word0_reg, src1_rel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_REL_MASK) | (src1_rel << SQ_ALU_WORD0_SRC1_REL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_CHAN(sq_alu_word0_reg, src1_chan) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_CHAN_MASK) | (src1_chan << SQ_ALU_WORD0_SRC1_CHAN_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_NEG(sq_alu_word0_reg, src1_neg) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_NEG_MASK) | (src1_neg << SQ_ALU_WORD0_SRC1_NEG_SHIFT)
+#define SQ_ALU_WORD0_SET_INDEX_MODE(sq_alu_word0_reg, index_mode) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_INDEX_MODE_MASK) | (index_mode << SQ_ALU_WORD0_INDEX_MODE_SHIFT)
+#define SQ_ALU_WORD0_SET_PRED_SEL(sq_alu_word0_reg, pred_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_PRED_SEL_MASK) | (pred_sel << SQ_ALU_WORD0_PRED_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_LAST(sq_alu_word0_reg, last) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_LAST_MASK) | (last << SQ_ALU_WORD0_LAST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word0_t {
+          unsigned int src0_sel                       : SQ_ALU_WORD0_SRC0_SEL_SIZE;
+          unsigned int src0_rel                       : SQ_ALU_WORD0_SRC0_REL_SIZE;
+          unsigned int src0_chan                      : SQ_ALU_WORD0_SRC0_CHAN_SIZE;
+          unsigned int src0_neg                       : SQ_ALU_WORD0_SRC0_NEG_SIZE;
+          unsigned int src1_sel                       : SQ_ALU_WORD0_SRC1_SEL_SIZE;
+          unsigned int src1_rel                       : SQ_ALU_WORD0_SRC1_REL_SIZE;
+          unsigned int src1_chan                      : SQ_ALU_WORD0_SRC1_CHAN_SIZE;
+          unsigned int src1_neg                       : SQ_ALU_WORD0_SRC1_NEG_SIZE;
+          unsigned int index_mode                     : SQ_ALU_WORD0_INDEX_MODE_SIZE;
+          unsigned int pred_sel                       : SQ_ALU_WORD0_PRED_SEL_SIZE;
+          unsigned int last                           : SQ_ALU_WORD0_LAST_SIZE;
+     } sq_alu_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word0_t {
+          unsigned int last                           : SQ_ALU_WORD0_LAST_SIZE;
+          unsigned int pred_sel                       : SQ_ALU_WORD0_PRED_SEL_SIZE;
+          unsigned int index_mode                     : SQ_ALU_WORD0_INDEX_MODE_SIZE;
+          unsigned int src1_neg                       : SQ_ALU_WORD0_SRC1_NEG_SIZE;
+          unsigned int src1_chan                      : SQ_ALU_WORD0_SRC1_CHAN_SIZE;
+          unsigned int src1_rel                       : SQ_ALU_WORD0_SRC1_REL_SIZE;
+          unsigned int src1_sel                       : SQ_ALU_WORD0_SRC1_SEL_SIZE;
+          unsigned int src0_neg                       : SQ_ALU_WORD0_SRC0_NEG_SIZE;
+          unsigned int src0_chan                      : SQ_ALU_WORD0_SRC0_CHAN_SIZE;
+          unsigned int src0_rel                       : SQ_ALU_WORD0_SRC0_REL_SIZE;
+          unsigned int src0_sel                       : SQ_ALU_WORD0_SRC0_SEL_SIZE;
+     } sq_alu_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word0_t f;
+} sq_alu_word0_u;
+
+
+/*
+ * SQ_ALU_WORD1 struct
+ */
+
+#define SQ_ALU_WORD1_ENCODING_SIZE     3
+#define SQ_ALU_WORD1_BANK_SWIZZLE_SIZE 3
+#define SQ_ALU_WORD1_DST_GPR_SIZE      7
+#define SQ_ALU_WORD1_DST_REL_SIZE      1
+#define SQ_ALU_WORD1_DST_CHAN_SIZE     2
+#define SQ_ALU_WORD1_CLAMP_SIZE        1
+
+#define SQ_ALU_WORD1_ENCODING_SHIFT    15
+#define SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT 18
+#define SQ_ALU_WORD1_DST_GPR_SHIFT     21
+#define SQ_ALU_WORD1_DST_REL_SHIFT     28
+#define SQ_ALU_WORD1_DST_CHAN_SHIFT    29
+#define SQ_ALU_WORD1_CLAMP_SHIFT       31
+
+#define SQ_ALU_WORD1_ENCODING_MASK     0x00038000
+#define SQ_ALU_WORD1_BANK_SWIZZLE_MASK 0x001c0000
+#define SQ_ALU_WORD1_DST_GPR_MASK      0x0fe00000
+#define SQ_ALU_WORD1_DST_REL_MASK      0x10000000
+#define SQ_ALU_WORD1_DST_CHAN_MASK     0x60000000
+#define SQ_ALU_WORD1_CLAMP_MASK        0x80000000
+
+#define SQ_ALU_WORD1_MASK \
+     (SQ_ALU_WORD1_ENCODING_MASK | \
+      SQ_ALU_WORD1_BANK_SWIZZLE_MASK | \
+      SQ_ALU_WORD1_DST_GPR_MASK | \
+      SQ_ALU_WORD1_DST_REL_MASK | \
+      SQ_ALU_WORD1_DST_CHAN_MASK | \
+      SQ_ALU_WORD1_CLAMP_MASK)
+
+#define SQ_ALU_WORD1_DEFAULT           0xcdcd8000
+
+#define SQ_ALU_WORD1_GET_ENCODING(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_ENCODING_MASK) >> SQ_ALU_WORD1_ENCODING_SHIFT)
+#define SQ_ALU_WORD1_GET_BANK_SWIZZLE(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_BANK_SWIZZLE_MASK) >> SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_GPR(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_GPR_MASK) >> SQ_ALU_WORD1_DST_GPR_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_REL(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_REL_MASK) >> SQ_ALU_WORD1_DST_REL_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_CHAN(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_CHAN_MASK) >> SQ_ALU_WORD1_DST_CHAN_SHIFT)
+#define SQ_ALU_WORD1_GET_CLAMP(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_CLAMP_MASK) >> SQ_ALU_WORD1_CLAMP_SHIFT)
+
+#define SQ_ALU_WORD1_SET_ENCODING(sq_alu_word1_reg, encoding) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_ENCODING_MASK) | (encoding << SQ_ALU_WORD1_ENCODING_SHIFT)
+#define SQ_ALU_WORD1_SET_BANK_SWIZZLE(sq_alu_word1_reg, bank_swizzle) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_BANK_SWIZZLE_MASK) | (bank_swizzle << SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_GPR(sq_alu_word1_reg, dst_gpr) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_GPR_MASK) | (dst_gpr << SQ_ALU_WORD1_DST_GPR_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_REL(sq_alu_word1_reg, dst_rel) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_REL_MASK) | (dst_rel << SQ_ALU_WORD1_DST_REL_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_CHAN(sq_alu_word1_reg, dst_chan) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_CHAN_MASK) | (dst_chan << SQ_ALU_WORD1_DST_CHAN_SHIFT)
+#define SQ_ALU_WORD1_SET_CLAMP(sq_alu_word1_reg, clamp) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_CLAMP_MASK) | (clamp << SQ_ALU_WORD1_CLAMP_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_t {
+          unsigned int                                : 15;
+          unsigned int encoding                       : SQ_ALU_WORD1_ENCODING_SIZE;
+          unsigned int bank_swizzle                   : SQ_ALU_WORD1_BANK_SWIZZLE_SIZE;
+          unsigned int dst_gpr                        : SQ_ALU_WORD1_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_ALU_WORD1_DST_REL_SIZE;
+          unsigned int dst_chan                       : SQ_ALU_WORD1_DST_CHAN_SIZE;
+          unsigned int clamp                          : SQ_ALU_WORD1_CLAMP_SIZE;
+     } sq_alu_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_t {
+          unsigned int clamp                          : SQ_ALU_WORD1_CLAMP_SIZE;
+          unsigned int dst_chan                       : SQ_ALU_WORD1_DST_CHAN_SIZE;
+          unsigned int dst_rel                        : SQ_ALU_WORD1_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_ALU_WORD1_DST_GPR_SIZE;
+          unsigned int bank_swizzle                   : SQ_ALU_WORD1_BANK_SWIZZLE_SIZE;
+          unsigned int encoding                       : SQ_ALU_WORD1_ENCODING_SIZE;
+          unsigned int                                : 15;
+     } sq_alu_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_t f;
+} sq_alu_word1_u;
+
+
+/*
+ * SQ_ALU_WORD1_OP2_V2 struct
+ */
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_OMOD_SIZE  2
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE 11
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT 0
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT 2
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT 3
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT 4
+#define SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT 5
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT 7
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK 0x00000001
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK 0x00000002
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK 0x00000004
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK 0x00000008
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK 0x00000010
+#define SQ_ALU_WORD1_OP2_V2_OMOD_MASK  0x00000060
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK 0x0003ff80
+
+#define SQ_ALU_WORD1_OP2_V2_MASK \
+     (SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK | \
+      SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK | \
+      SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK | \
+      SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK | \
+      SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK | \
+      SQ_ALU_WORD1_OP2_V2_OMOD_MASK | \
+      SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK)
+
+#define SQ_ALU_WORD1_OP2_V2_DEFAULT    0x0001cdcd
+
+#define SQ_ALU_WORD1_OP2_V2_GET_SRC0_ABS(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK) >> SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_SRC1_ABS(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK) >> SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_UPDATE_EXECUTE_MASK(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK) >> SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_UPDATE_PRED(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK) >> SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_WRITE_MASK(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK) >> SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_OMOD(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_OMOD_MASK) >> SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_ALU_INST(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK) >> SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT)
+
+#define SQ_ALU_WORD1_OP2_V2_SET_SRC0_ABS(sq_alu_word1_op2_v2_reg, src0_abs) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK) | (src0_abs << SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_SRC1_ABS(sq_alu_word1_op2_v2_reg, src1_abs) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK) | (src1_abs << SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_UPDATE_EXECUTE_MASK(sq_alu_word1_op2_v2_reg, update_execute_mask) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK) | (update_execute_mask << SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_UPDATE_PRED(sq_alu_word1_op2_v2_reg, update_pred) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK) | (update_pred << SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_WRITE_MASK(sq_alu_word1_op2_v2_reg, write_mask) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK) | (write_mask << SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_OMOD(sq_alu_word1_op2_v2_reg, omod) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_OMOD_MASK) | (omod << SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_ALU_INST(sq_alu_word1_op2_v2_reg, alu_inst) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK) | (alu_inst << SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_v2_t {
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE;
+          unsigned int                                : 14;
+     } sq_alu_word1_op2_v2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_v2_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+     } sq_alu_word1_op2_v2_t;
+
+#endif
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_r6xx_t {
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int fog_export                     : 1;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int alu_inst                       : 10;
+          unsigned int                                : 14;
+     } sq_alu_word1_op2_v1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_r6xx_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : 10;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int fog_export                     : 1;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+     } sq_alu_word1_op2_v1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_op2_v2_t f;
+     sq_alu_word1_op2_v1_t f6;
+} sq_alu_word1_op2_v2_u;
+
+
+/*
+ * SQ_ALU_WORD1_OP3 struct
+ */
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE 9
+#define SQ_ALU_WORD1_OP3_SRC2_REL_SIZE 1
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE 2
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE 1
+#define SQ_ALU_WORD1_OP3_ALU_INST_SIZE 5
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT 0
+#define SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT 9
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT 10
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT 12
+#define SQ_ALU_WORD1_OP3_ALU_INST_SHIFT 13
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_MASK 0x000001ff
+#define SQ_ALU_WORD1_OP3_SRC2_REL_MASK 0x00000200
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK 0x00000c00
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_MASK 0x00001000
+#define SQ_ALU_WORD1_OP3_ALU_INST_MASK 0x0003e000
+
+#define SQ_ALU_WORD1_OP3_MASK \
+     (SQ_ALU_WORD1_OP3_SRC2_SEL_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_REL_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_NEG_MASK | \
+      SQ_ALU_WORD1_OP3_ALU_INST_MASK)
+
+#define SQ_ALU_WORD1_OP3_DEFAULT       0x0001cdcd
+
+#define SQ_ALU_WORD1_OP3_GET_SRC2_SEL(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_SEL_MASK) >> SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_REL(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_REL_MASK) >> SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_CHAN(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK) >> SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_NEG(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_NEG_MASK) >> SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_ALU_INST(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_ALU_INST_MASK) >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT)
+
+#define SQ_ALU_WORD1_OP3_SET_SRC2_SEL(sq_alu_word1_op3_reg, src2_sel) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_SEL_MASK) | (src2_sel << SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_REL(sq_alu_word1_op3_reg, src2_rel) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_REL_MASK) | (src2_rel << SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_CHAN(sq_alu_word1_op3_reg, src2_chan) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK) | (src2_chan << SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_NEG(sq_alu_word1_op3_reg, src2_neg) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_NEG_MASK) | (src2_neg << SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_ALU_INST(sq_alu_word1_op3_reg, alu_inst) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_ALU_INST_MASK) | (alu_inst << SQ_ALU_WORD1_OP3_ALU_INST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op3_t {
+          unsigned int src2_sel                       : SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE;
+          unsigned int src2_rel                       : SQ_ALU_WORD1_OP3_SRC2_REL_SIZE;
+          unsigned int src2_chan                      : SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE;
+          unsigned int src2_neg                       : SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP3_ALU_INST_SIZE;
+          unsigned int                                : 14;
+     } sq_alu_word1_op3_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op3_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP3_ALU_INST_SIZE;
+          unsigned int src2_neg                       : SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE;
+          unsigned int src2_chan                      : SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE;
+          unsigned int src2_rel                       : SQ_ALU_WORD1_OP3_SRC2_REL_SIZE;
+          unsigned int src2_sel                       : SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE;
+     } sq_alu_word1_op3_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_op3_t f;
+} sq_alu_word1_op3_u;
+
+
+/*
+ * SQ_TEX_WORD0 struct
+ */
+
+#define SQ_TEX_WORD0_TEX_INST_SIZE     5
+#define SQ_TEX_WORD0_BC_FRAC_MODE_SIZE 1
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE 1
+#define SQ_TEX_WORD0_RESOURCE_ID_SIZE  8
+#define SQ_TEX_WORD0_SRC_GPR_SIZE      7
+#define SQ_TEX_WORD0_SRC_REL_SIZE      1
+#define SQ_TEX_WORD0_ALT_CONST_SIZE    1
+
+#define SQ_TEX_WORD0_TEX_INST_SHIFT    0
+#define SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT 5
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT 7
+#define SQ_TEX_WORD0_RESOURCE_ID_SHIFT 8
+#define SQ_TEX_WORD0_SRC_GPR_SHIFT     16
+#define SQ_TEX_WORD0_SRC_REL_SHIFT     23
+#define SQ_TEX_WORD0_ALT_CONST_SHIFT   24
+
+#define SQ_TEX_WORD0_TEX_INST_MASK     0x0000001f
+#define SQ_TEX_WORD0_BC_FRAC_MODE_MASK 0x00000020
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK 0x00000080
+#define SQ_TEX_WORD0_RESOURCE_ID_MASK  0x0000ff00
+#define SQ_TEX_WORD0_SRC_GPR_MASK      0x007f0000
+#define SQ_TEX_WORD0_SRC_REL_MASK      0x00800000
+#define SQ_TEX_WORD0_ALT_CONST_MASK    0x01000000
+
+#define SQ_TEX_WORD0_MASK \
+     (SQ_TEX_WORD0_TEX_INST_MASK | \
+      SQ_TEX_WORD0_BC_FRAC_MODE_MASK | \
+      SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK | \
+      SQ_TEX_WORD0_RESOURCE_ID_MASK | \
+      SQ_TEX_WORD0_SRC_GPR_MASK | \
+      SQ_TEX_WORD0_SRC_REL_MASK | \
+      SQ_TEX_WORD0_ALT_CONST_MASK)
+
+#define SQ_TEX_WORD0_DEFAULT           0x01cdcd8d
+
+#define SQ_TEX_WORD0_GET_TEX_INST(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_TEX_INST_MASK) >> SQ_TEX_WORD0_TEX_INST_SHIFT)
+#define SQ_TEX_WORD0_GET_BC_FRAC_MODE(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_BC_FRAC_MODE_MASK) >> SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT)
+#define SQ_TEX_WORD0_GET_FETCH_WHOLE_QUAD(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK) >> SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_TEX_WORD0_GET_RESOURCE_ID(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_RESOURCE_ID_MASK) >> SQ_TEX_WORD0_RESOURCE_ID_SHIFT)
+#define SQ_TEX_WORD0_GET_SRC_GPR(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_SRC_GPR_MASK) >> SQ_TEX_WORD0_SRC_GPR_SHIFT)
+#define SQ_TEX_WORD0_GET_SRC_REL(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_SRC_REL_MASK) >> SQ_TEX_WORD0_SRC_REL_SHIFT)
+#define SQ_TEX_WORD0_GET_ALT_CONST(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_ALT_CONST_MASK) >> SQ_TEX_WORD0_ALT_CONST_SHIFT)
+
+#define SQ_TEX_WORD0_SET_TEX_INST(sq_tex_word0_reg, tex_inst) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_TEX_INST_MASK) | (tex_inst << SQ_TEX_WORD0_TEX_INST_SHIFT)
+#define SQ_TEX_WORD0_SET_BC_FRAC_MODE(sq_tex_word0_reg, bc_frac_mode) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_BC_FRAC_MODE_MASK) | (bc_frac_mode << SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT)
+#define SQ_TEX_WORD0_SET_FETCH_WHOLE_QUAD(sq_tex_word0_reg, fetch_whole_quad) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK) | (fetch_whole_quad << SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_TEX_WORD0_SET_RESOURCE_ID(sq_tex_word0_reg, resource_id) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_RESOURCE_ID_MASK) | (resource_id << SQ_TEX_WORD0_RESOURCE_ID_SHIFT)
+#define SQ_TEX_WORD0_SET_SRC_GPR(sq_tex_word0_reg, src_gpr) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_SRC_GPR_MASK) | (src_gpr << SQ_TEX_WORD0_SRC_GPR_SHIFT)
+#define SQ_TEX_WORD0_SET_SRC_REL(sq_tex_word0_reg, src_rel) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_SRC_REL_MASK) | (src_rel << SQ_TEX_WORD0_SRC_REL_SHIFT)
+#define SQ_TEX_WORD0_SET_ALT_CONST(sq_tex_word0_reg, alt_const) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_ALT_CONST_MASK) | (alt_const << SQ_TEX_WORD0_ALT_CONST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word0_t {
+          unsigned int tex_inst                       : SQ_TEX_WORD0_TEX_INST_SIZE;
+          unsigned int bc_frac_mode                   : SQ_TEX_WORD0_BC_FRAC_MODE_SIZE;
+          unsigned int                                : 1;
+          unsigned int fetch_whole_quad               : SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int resource_id                    : SQ_TEX_WORD0_RESOURCE_ID_SIZE;
+          unsigned int src_gpr                        : SQ_TEX_WORD0_SRC_GPR_SIZE;
+          unsigned int src_rel                        : SQ_TEX_WORD0_SRC_REL_SIZE;
+          unsigned int alt_const                      : SQ_TEX_WORD0_ALT_CONST_SIZE;
+          unsigned int                                : 7;
+     } sq_tex_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word0_t {
+          unsigned int                                : 7;
+          unsigned int alt_const                      : SQ_TEX_WORD0_ALT_CONST_SIZE;
+          unsigned int src_rel                        : SQ_TEX_WORD0_SRC_REL_SIZE;
+          unsigned int src_gpr                        : SQ_TEX_WORD0_SRC_GPR_SIZE;
+          unsigned int resource_id                    : SQ_TEX_WORD0_RESOURCE_ID_SIZE;
+          unsigned int fetch_whole_quad               : SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int                                : 1;
+          unsigned int bc_frac_mode                   : SQ_TEX_WORD0_BC_FRAC_MODE_SIZE;
+          unsigned int tex_inst                       : SQ_TEX_WORD0_TEX_INST_SIZE;
+     } sq_tex_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word0_t f;
+} sq_tex_word0_u;
+
+
+/*
+ * SQ_TEX_WORD1 struct
+ */
+
+#define SQ_TEX_WORD1_DST_GPR_SIZE      7
+#define SQ_TEX_WORD1_DST_REL_SIZE      1
+#define SQ_TEX_WORD1_DST_SEL_X_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_Y_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_Z_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_W_SIZE    3
+#define SQ_TEX_WORD1_LOD_BIAS_SIZE     7
+#define SQ_TEX_WORD1_COORD_TYPE_X_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_Y_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_Z_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_W_SIZE 1
+
+#define SQ_TEX_WORD1_DST_GPR_SHIFT     0
+#define SQ_TEX_WORD1_DST_REL_SHIFT     7
+#define SQ_TEX_WORD1_DST_SEL_X_SHIFT   9
+#define SQ_TEX_WORD1_DST_SEL_Y_SHIFT   12
+#define SQ_TEX_WORD1_DST_SEL_Z_SHIFT   15
+#define SQ_TEX_WORD1_DST_SEL_W_SHIFT   18
+#define SQ_TEX_WORD1_LOD_BIAS_SHIFT    21
+#define SQ_TEX_WORD1_COORD_TYPE_X_SHIFT 28
+#define SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT 29
+#define SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT 30
+#define SQ_TEX_WORD1_COORD_TYPE_W_SHIFT 31
+
+#define SQ_TEX_WORD1_DST_GPR_MASK      0x0000007f
+#define SQ_TEX_WORD1_DST_REL_MASK      0x00000080
+#define SQ_TEX_WORD1_DST_SEL_X_MASK    0x00000e00
+#define SQ_TEX_WORD1_DST_SEL_Y_MASK    0x00007000
+#define SQ_TEX_WORD1_DST_SEL_Z_MASK    0x00038000
+#define SQ_TEX_WORD1_DST_SEL_W_MASK    0x001c0000
+#define SQ_TEX_WORD1_LOD_BIAS_MASK     0x0fe00000
+#define SQ_TEX_WORD1_COORD_TYPE_X_MASK 0x10000000
+#define SQ_TEX_WORD1_COORD_TYPE_Y_MASK 0x20000000
+#define SQ_TEX_WORD1_COORD_TYPE_Z_MASK 0x40000000
+#define SQ_TEX_WORD1_COORD_TYPE_W_MASK 0x80000000
+
+#define SQ_TEX_WORD1_MASK \
+     (SQ_TEX_WORD1_DST_GPR_MASK | \
+      SQ_TEX_WORD1_DST_REL_MASK | \
+      SQ_TEX_WORD1_DST_SEL_X_MASK | \
+      SQ_TEX_WORD1_DST_SEL_Y_MASK | \
+      SQ_TEX_WORD1_DST_SEL_Z_MASK | \
+      SQ_TEX_WORD1_DST_SEL_W_MASK | \
+      SQ_TEX_WORD1_LOD_BIAS_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_X_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_Y_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_Z_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_W_MASK)
+
+#define SQ_TEX_WORD1_DEFAULT           0xcdcdcccd
+
+#define SQ_TEX_WORD1_GET_DST_GPR(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_GPR_MASK) >> SQ_TEX_WORD1_DST_GPR_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_REL(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_REL_MASK) >> SQ_TEX_WORD1_DST_REL_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_X(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_X_MASK) >> SQ_TEX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_Y(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_Y_MASK) >> SQ_TEX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_Z(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_Z_MASK) >> SQ_TEX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_W(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_W_MASK) >> SQ_TEX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_TEX_WORD1_GET_LOD_BIAS(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_LOD_BIAS_MASK) >> SQ_TEX_WORD1_LOD_BIAS_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_X(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_X_MASK) >> SQ_TEX_WORD1_COORD_TYPE_X_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_Y(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_Y_MASK) >> SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_Z(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_Z_MASK) >> SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_W(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_W_MASK) >> SQ_TEX_WORD1_COORD_TYPE_W_SHIFT)
+
+#define SQ_TEX_WORD1_SET_DST_GPR(sq_tex_word1_reg, dst_gpr) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_GPR_MASK) | (dst_gpr << SQ_TEX_WORD1_DST_GPR_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_REL(sq_tex_word1_reg, dst_rel) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_REL_MASK) | (dst_rel << SQ_TEX_WORD1_DST_REL_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_X(sq_tex_word1_reg, dst_sel_x) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_X_MASK) | (dst_sel_x << SQ_TEX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_Y(sq_tex_word1_reg, dst_sel_y) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_Y_MASK) | (dst_sel_y << SQ_TEX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_Z(sq_tex_word1_reg, dst_sel_z) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_Z_MASK) | (dst_sel_z << SQ_TEX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_W(sq_tex_word1_reg, dst_sel_w) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_W_MASK) | (dst_sel_w << SQ_TEX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_TEX_WORD1_SET_LOD_BIAS(sq_tex_word1_reg, lod_bias) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_LOD_BIAS_MASK) | (lod_bias << SQ_TEX_WORD1_LOD_BIAS_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_X(sq_tex_word1_reg, coord_type_x) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_X_MASK) | (coord_type_x << SQ_TEX_WORD1_COORD_TYPE_X_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_Y(sq_tex_word1_reg, coord_type_y) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_Y_MASK) | (coord_type_y << SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_Z(sq_tex_word1_reg, coord_type_z) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_Z_MASK) | (coord_type_z << SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_W(sq_tex_word1_reg, coord_type_w) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_W_MASK) | (coord_type_w << SQ_TEX_WORD1_COORD_TYPE_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word1_t {
+          unsigned int dst_gpr                        : SQ_TEX_WORD1_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_TEX_WORD1_DST_REL_SIZE;
+          unsigned int                                : 1;
+          unsigned int dst_sel_x                      : SQ_TEX_WORD1_DST_SEL_X_SIZE;
+          unsigned int dst_sel_y                      : SQ_TEX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_z                      : SQ_TEX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_w                      : SQ_TEX_WORD1_DST_SEL_W_SIZE;
+          unsigned int lod_bias                       : SQ_TEX_WORD1_LOD_BIAS_SIZE;
+          unsigned int coord_type_x                   : SQ_TEX_WORD1_COORD_TYPE_X_SIZE;
+          unsigned int coord_type_y                   : SQ_TEX_WORD1_COORD_TYPE_Y_SIZE;
+          unsigned int coord_type_z                   : SQ_TEX_WORD1_COORD_TYPE_Z_SIZE;
+          unsigned int coord_type_w                   : SQ_TEX_WORD1_COORD_TYPE_W_SIZE;
+     } sq_tex_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word1_t {
+          unsigned int coord_type_w                   : SQ_TEX_WORD1_COORD_TYPE_W_SIZE;
+          unsigned int coord_type_z                   : SQ_TEX_WORD1_COORD_TYPE_Z_SIZE;
+          unsigned int coord_type_y                   : SQ_TEX_WORD1_COORD_TYPE_Y_SIZE;
+          unsigned int coord_type_x                   : SQ_TEX_WORD1_COORD_TYPE_X_SIZE;
+          unsigned int lod_bias                       : SQ_TEX_WORD1_LOD_BIAS_SIZE;
+          unsigned int dst_sel_w                      : SQ_TEX_WORD1_DST_SEL_W_SIZE;
+          unsigned int dst_sel_z                      : SQ_TEX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_y                      : SQ_TEX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_x                      : SQ_TEX_WORD1_DST_SEL_X_SIZE;
+          unsigned int                                : 1;
+          unsigned int dst_rel                        : SQ_TEX_WORD1_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_TEX_WORD1_DST_GPR_SIZE;
+     } sq_tex_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word1_t f;
+} sq_tex_word1_u;
+
+
+/*
+ * SQ_TEX_WORD2 struct
+ */
+
+#define SQ_TEX_WORD2_OFFSET_X_SIZE     5
+#define SQ_TEX_WORD2_OFFSET_Y_SIZE     5
+#define SQ_TEX_WORD2_OFFSET_Z_SIZE     5
+#define SQ_TEX_WORD2_SAMPLER_ID_SIZE   5
+#define SQ_TEX_WORD2_SRC_SEL_X_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_Y_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_Z_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_W_SIZE    3
+
+#define SQ_TEX_WORD2_OFFSET_X_SHIFT    0
+#define SQ_TEX_WORD2_OFFSET_Y_SHIFT    5
+#define SQ_TEX_WORD2_OFFSET_Z_SHIFT    10
+#define SQ_TEX_WORD2_SAMPLER_ID_SHIFT  15
+#define SQ_TEX_WORD2_SRC_SEL_X_SHIFT   20
+#define SQ_TEX_WORD2_SRC_SEL_Y_SHIFT   23
+#define SQ_TEX_WORD2_SRC_SEL_Z_SHIFT   26
+#define SQ_TEX_WORD2_SRC_SEL_W_SHIFT   29
+
+#define SQ_TEX_WORD2_OFFSET_X_MASK     0x0000001f
+#define SQ_TEX_WORD2_OFFSET_Y_MASK     0x000003e0
+#define SQ_TEX_WORD2_OFFSET_Z_MASK     0x00007c00
+#define SQ_TEX_WORD2_SAMPLER_ID_MASK   0x000f8000
+#define SQ_TEX_WORD2_SRC_SEL_X_MASK    0x00700000
+#define SQ_TEX_WORD2_SRC_SEL_Y_MASK    0x03800000
+#define SQ_TEX_WORD2_SRC_SEL_Z_MASK    0x1c000000
+#define SQ_TEX_WORD2_SRC_SEL_W_MASK    0xe0000000
+
+#define SQ_TEX_WORD2_MASK \
+     (SQ_TEX_WORD2_OFFSET_X_MASK | \
+      SQ_TEX_WORD2_OFFSET_Y_MASK | \
+      SQ_TEX_WORD2_OFFSET_Z_MASK | \
+      SQ_TEX_WORD2_SAMPLER_ID_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_X_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_Y_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_Z_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_W_MASK)
+
+#define SQ_TEX_WORD2_DEFAULT           0xcdcdcdcd
+
+#define SQ_TEX_WORD2_GET_OFFSET_X(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_X_MASK) >> SQ_TEX_WORD2_OFFSET_X_SHIFT)
+#define SQ_TEX_WORD2_GET_OFFSET_Y(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_Y_MASK) >> SQ_TEX_WORD2_OFFSET_Y_SHIFT)
+#define SQ_TEX_WORD2_GET_OFFSET_Z(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_Z_MASK) >> SQ_TEX_WORD2_OFFSET_Z_SHIFT)
+#define SQ_TEX_WORD2_GET_SAMPLER_ID(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SAMPLER_ID_MASK) >> SQ_TEX_WORD2_SAMPLER_ID_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_X(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_X_MASK) >> SQ_TEX_WORD2_SRC_SEL_X_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_Y(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_Y_MASK) >> SQ_TEX_WORD2_SRC_SEL_Y_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_Z(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_Z_MASK) >> SQ_TEX_WORD2_SRC_SEL_Z_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_W(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_W_MASK) >> SQ_TEX_WORD2_SRC_SEL_W_SHIFT)
+
+#define SQ_TEX_WORD2_SET_OFFSET_X(sq_tex_word2_reg, offset_x) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_X_MASK) | (offset_x << SQ_TEX_WORD2_OFFSET_X_SHIFT)
+#define SQ_TEX_WORD2_SET_OFFSET_Y(sq_tex_word2_reg, offset_y) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_Y_MASK) | (offset_y << SQ_TEX_WORD2_OFFSET_Y_SHIFT)
+#define SQ_TEX_WORD2_SET_OFFSET_Z(sq_tex_word2_reg, offset_z) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_Z_MASK) | (offset_z << SQ_TEX_WORD2_OFFSET_Z_SHIFT)
+#define SQ_TEX_WORD2_SET_SAMPLER_ID(sq_tex_word2_reg, sampler_id) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SAMPLER_ID_MASK) | (sampler_id << SQ_TEX_WORD2_SAMPLER_ID_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_X(sq_tex_word2_reg, src_sel_x) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_X_MASK) | (src_sel_x << SQ_TEX_WORD2_SRC_SEL_X_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_Y(sq_tex_word2_reg, src_sel_y) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_Y_MASK) | (src_sel_y << SQ_TEX_WORD2_SRC_SEL_Y_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_Z(sq_tex_word2_reg, src_sel_z) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_Z_MASK) | (src_sel_z << SQ_TEX_WORD2_SRC_SEL_Z_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_W(sq_tex_word2_reg, src_sel_w) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_W_MASK) | (src_sel_w << SQ_TEX_WORD2_SRC_SEL_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word2_t {
+          unsigned int offset_x                       : SQ_TEX_WORD2_OFFSET_X_SIZE;
+          unsigned int offset_y                       : SQ_TEX_WORD2_OFFSET_Y_SIZE;
+          unsigned int offset_z                       : SQ_TEX_WORD2_OFFSET_Z_SIZE;
+          unsigned int sampler_id                     : SQ_TEX_WORD2_SAMPLER_ID_SIZE;
+          unsigned int src_sel_x                      : SQ_TEX_WORD2_SRC_SEL_X_SIZE;
+          unsigned int src_sel_y                      : SQ_TEX_WORD2_SRC_SEL_Y_SIZE;
+          unsigned int src_sel_z                      : SQ_TEX_WORD2_SRC_SEL_Z_SIZE;
+          unsigned int src_sel_w                      : SQ_TEX_WORD2_SRC_SEL_W_SIZE;
+     } sq_tex_word2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word2_t {
+          unsigned int src_sel_w                      : SQ_TEX_WORD2_SRC_SEL_W_SIZE;
+          unsigned int src_sel_z                      : SQ_TEX_WORD2_SRC_SEL_Z_SIZE;
+          unsigned int src_sel_y                      : SQ_TEX_WORD2_SRC_SEL_Y_SIZE;
+          unsigned int src_sel_x                      : SQ_TEX_WORD2_SRC_SEL_X_SIZE;
+          unsigned int sampler_id                     : SQ_TEX_WORD2_SAMPLER_ID_SIZE;
+          unsigned int offset_z                       : SQ_TEX_WORD2_OFFSET_Z_SIZE;
+          unsigned int offset_y                       : SQ_TEX_WORD2_OFFSET_Y_SIZE;
+          unsigned int offset_x                       : SQ_TEX_WORD2_OFFSET_X_SIZE;
+     } sq_tex_word2_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word2_t f;
+} sq_tex_word2_u;
+
+
+/*
+ * SQ_VTX_WORD0 struct
+ */
+
+#define SQ_VTX_WORD0_VTX_INST_SIZE     5
+#define SQ_VTX_WORD0_FETCH_TYPE_SIZE   2
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE 1
+#define SQ_VTX_WORD0_BUFFER_ID_SIZE    8
+#define SQ_VTX_WORD0_SRC_GPR_SIZE      7
+#define SQ_VTX_WORD0_SRC_REL_SIZE      1
+#define SQ_VTX_WORD0_SRC_SEL_X_SIZE    2
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE 6
+
+#define SQ_VTX_WORD0_VTX_INST_SHIFT    0
+#define SQ_VTX_WORD0_FETCH_TYPE_SHIFT  5
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT 7
+#define SQ_VTX_WORD0_BUFFER_ID_SHIFT   8
+#define SQ_VTX_WORD0_SRC_GPR_SHIFT     16
+#define SQ_VTX_WORD0_SRC_REL_SHIFT     23
+#define SQ_VTX_WORD0_SRC_SEL_X_SHIFT   24
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT 26
+
+#define SQ_VTX_WORD0_VTX_INST_MASK     0x0000001f
+#define SQ_VTX_WORD0_FETCH_TYPE_MASK   0x00000060
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK 0x00000080
+#define SQ_VTX_WORD0_BUFFER_ID_MASK    0x0000ff00
+#define SQ_VTX_WORD0_SRC_GPR_MASK      0x007f0000
+#define SQ_VTX_WORD0_SRC_REL_MASK      0x00800000
+#define SQ_VTX_WORD0_SRC_SEL_X_MASK    0x03000000
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK 0xfc000000
+
+#define SQ_VTX_WORD0_MASK \
+     (SQ_VTX_WORD0_VTX_INST_MASK | \
+      SQ_VTX_WORD0_FETCH_TYPE_MASK | \
+      SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK | \
+      SQ_VTX_WORD0_BUFFER_ID_MASK | \
+      SQ_VTX_WORD0_SRC_GPR_MASK | \
+      SQ_VTX_WORD0_SRC_REL_MASK | \
+      SQ_VTX_WORD0_SRC_SEL_X_MASK | \
+      SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK)
+
+#define SQ_VTX_WORD0_DEFAULT           0xcdcdcdcd
+
+#define SQ_VTX_WORD0_GET_VTX_INST(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_VTX_INST_MASK) >> SQ_VTX_WORD0_VTX_INST_SHIFT)
+#define SQ_VTX_WORD0_GET_FETCH_TYPE(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_FETCH_TYPE_MASK) >> SQ_VTX_WORD0_FETCH_TYPE_SHIFT)
+#define SQ_VTX_WORD0_GET_FETCH_WHOLE_QUAD(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK) >> SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_VTX_WORD0_GET_BUFFER_ID(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_BUFFER_ID_MASK) >> SQ_VTX_WORD0_BUFFER_ID_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_GPR(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_GPR_MASK) >> SQ_VTX_WORD0_SRC_GPR_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_REL(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_REL_MASK) >> SQ_VTX_WORD0_SRC_REL_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_SEL_X(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_SEL_X_MASK) >> SQ_VTX_WORD0_SRC_SEL_X_SHIFT)
+#define SQ_VTX_WORD0_GET_MEGA_FETCH_COUNT(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK) >> SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT)
+
+#define SQ_VTX_WORD0_SET_VTX_INST(sq_vtx_word0_reg, vtx_inst) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_VTX_INST_MASK) | (vtx_inst << SQ_VTX_WORD0_VTX_INST_SHIFT)
+#define SQ_VTX_WORD0_SET_FETCH_TYPE(sq_vtx_word0_reg, fetch_type) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_FETCH_TYPE_MASK) | (fetch_type << SQ_VTX_WORD0_FETCH_TYPE_SHIFT)
+#define SQ_VTX_WORD0_SET_FETCH_WHOLE_QUAD(sq_vtx_word0_reg, fetch_whole_quad) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK) | (fetch_whole_quad << SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_VTX_WORD0_SET_BUFFER_ID(sq_vtx_word0_reg, buffer_id) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_BUFFER_ID_MASK) | (buffer_id << SQ_VTX_WORD0_BUFFER_ID_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_GPR(sq_vtx_word0_reg, src_gpr) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_GPR_MASK) | (src_gpr << SQ_VTX_WORD0_SRC_GPR_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_REL(sq_vtx_word0_reg, src_rel) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_REL_MASK) | (src_rel << SQ_VTX_WORD0_SRC_REL_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_SEL_X(sq_vtx_word0_reg, src_sel_x) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_SEL_X_MASK) | (src_sel_x << SQ_VTX_WORD0_SRC_SEL_X_SHIFT)
+#define SQ_VTX_WORD0_SET_MEGA_FETCH_COUNT(sq_vtx_word0_reg, mega_fetch_count) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK) | (mega_fetch_count << SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word0_t {
+          unsigned int vtx_inst                       : SQ_VTX_WORD0_VTX_INST_SIZE;
+          unsigned int fetch_type                     : SQ_VTX_WORD0_FETCH_TYPE_SIZE;
+          unsigned int fetch_whole_quad               : SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int buffer_id                      : SQ_VTX_WORD0_BUFFER_ID_SIZE;
+          unsigned int src_gpr                        : SQ_VTX_WORD0_SRC_GPR_SIZE;
+          unsigned int src_rel                        : SQ_VTX_WORD0_SRC_REL_SIZE;
+          unsigned int src_sel_x                      : SQ_VTX_WORD0_SRC_SEL_X_SIZE;
+          unsigned int mega_fetch_count               : SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE;
+     } sq_vtx_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word0_t {
+          unsigned int mega_fetch_count               : SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE;
+          unsigned int src_sel_x                      : SQ_VTX_WORD0_SRC_SEL_X_SIZE;
+          unsigned int src_rel                        : SQ_VTX_WORD0_SRC_REL_SIZE;
+          unsigned int src_gpr                        : SQ_VTX_WORD0_SRC_GPR_SIZE;
+          unsigned int buffer_id                      : SQ_VTX_WORD0_BUFFER_ID_SIZE;
+          unsigned int fetch_whole_quad               : SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int fetch_type                     : SQ_VTX_WORD0_FETCH_TYPE_SIZE;
+          unsigned int vtx_inst                       : SQ_VTX_WORD0_VTX_INST_SIZE;
+     } sq_vtx_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word0_t f;
+} sq_vtx_word0_u;
+
+
+/*
+ * SQ_VTX_WORD1 struct
+ */
+
+#define SQ_VTX_WORD1_DST_SEL_X_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_Y_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_Z_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_W_SIZE    3
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE 1
+#define SQ_VTX_WORD1_DATA_FORMAT_SIZE  6
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE 2
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE 1
+#define SQ_VTX_WORD1_SRF_MODE_ALL_SIZE 1
+
+#define SQ_VTX_WORD1_DST_SEL_X_SHIFT   9
+#define SQ_VTX_WORD1_DST_SEL_Y_SHIFT   12
+#define SQ_VTX_WORD1_DST_SEL_Z_SHIFT   15
+#define SQ_VTX_WORD1_DST_SEL_W_SHIFT   18
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT 21
+#define SQ_VTX_WORD1_DATA_FORMAT_SHIFT 22
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT 28
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT 30
+#define SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT 31
+
+#define SQ_VTX_WORD1_DST_SEL_X_MASK    0x00000e00
+#define SQ_VTX_WORD1_DST_SEL_Y_MASK    0x00007000
+#define SQ_VTX_WORD1_DST_SEL_Z_MASK    0x00038000
+#define SQ_VTX_WORD1_DST_SEL_W_MASK    0x001c0000
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_MASK 0x00200000
+#define SQ_VTX_WORD1_DATA_FORMAT_MASK  0x0fc00000
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK 0x30000000
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK 0x40000000
+#define SQ_VTX_WORD1_SRF_MODE_ALL_MASK 0x80000000
+
+#define SQ_VTX_WORD1_MASK \
+     (SQ_VTX_WORD1_DST_SEL_X_MASK | \
+      SQ_VTX_WORD1_DST_SEL_Y_MASK | \
+      SQ_VTX_WORD1_DST_SEL_Z_MASK | \
+      SQ_VTX_WORD1_DST_SEL_W_MASK | \
+      SQ_VTX_WORD1_USE_CONST_FIELDS_MASK | \
+      SQ_VTX_WORD1_DATA_FORMAT_MASK | \
+      SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK | \
+      SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK | \
+      SQ_VTX_WORD1_SRF_MODE_ALL_MASK)
+
+#define SQ_VTX_WORD1_DEFAULT           0xcdcdcc00
+
+#define SQ_VTX_WORD1_GET_DST_SEL_X(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_X_MASK) >> SQ_VTX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_Y(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_Y_MASK) >> SQ_VTX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_Z(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_Z_MASK) >> SQ_VTX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_W(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_W_MASK) >> SQ_VTX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_VTX_WORD1_GET_USE_CONST_FIELDS(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_USE_CONST_FIELDS_MASK) >> SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT)
+#define SQ_VTX_WORD1_GET_DATA_FORMAT(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DATA_FORMAT_MASK) >> SQ_VTX_WORD1_DATA_FORMAT_SHIFT)
+#define SQ_VTX_WORD1_GET_NUM_FORMAT_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK) >> SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT)
+#define SQ_VTX_WORD1_GET_FORMAT_COMP_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK) >> SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT)
+#define SQ_VTX_WORD1_GET_SRF_MODE_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_SRF_MODE_ALL_MASK) >> SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT)
+
+#define SQ_VTX_WORD1_SET_DST_SEL_X(sq_vtx_word1_reg, dst_sel_x) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_X_MASK) | (dst_sel_x << SQ_VTX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_Y(sq_vtx_word1_reg, dst_sel_y) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_Y_MASK) | (dst_sel_y << SQ_VTX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_Z(sq_vtx_word1_reg, dst_sel_z) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_Z_MASK) | (dst_sel_z << SQ_VTX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_W(sq_vtx_word1_reg, dst_sel_w) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_W_MASK) | (dst_sel_w << SQ_VTX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_VTX_WORD1_SET_USE_CONST_FIELDS(sq_vtx_word1_reg, use_const_fields) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_USE_CONST_FIELDS_MASK) | (use_const_fields << SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT)
+#define SQ_VTX_WORD1_SET_DATA_FORMAT(sq_vtx_word1_reg, data_format) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DATA_FORMAT_MASK) | (data_format << SQ_VTX_WORD1_DATA_FORMAT_SHIFT)
+#define SQ_VTX_WORD1_SET_NUM_FORMAT_ALL(sq_vtx_word1_reg, num_format_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK) | (num_format_all << SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT)
+#define SQ_VTX_WORD1_SET_FORMAT_COMP_ALL(sq_vtx_word1_reg, format_comp_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK) | (format_comp_all << SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT)
+#define SQ_VTX_WORD1_SET_SRF_MODE_ALL(sq_vtx_word1_reg, srf_mode_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_SRF_MODE_ALL_MASK) | (srf_mode_all << SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_t {
+          unsigned int                                : 9;
+          unsigned int dst_sel_x                      : SQ_VTX_WORD1_DST_SEL_X_SIZE;
+          unsigned int dst_sel_y                      : SQ_VTX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_z                      : SQ_VTX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_w                      : SQ_VTX_WORD1_DST_SEL_W_SIZE;
+          unsigned int use_const_fields               : SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE;
+          unsigned int data_format                    : SQ_VTX_WORD1_DATA_FORMAT_SIZE;
+          unsigned int num_format_all                 : SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE;
+          unsigned int format_comp_all                : SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE;
+          unsigned int srf_mode_all                   : SQ_VTX_WORD1_SRF_MODE_ALL_SIZE;
+     } sq_vtx_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_t {
+          unsigned int srf_mode_all                   : SQ_VTX_WORD1_SRF_MODE_ALL_SIZE;
+          unsigned int format_comp_all                : SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE;
+          unsigned int num_format_all                 : SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE;
+          unsigned int data_format                    : SQ_VTX_WORD1_DATA_FORMAT_SIZE;
+          unsigned int use_const_fields               : SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE;
+          unsigned int dst_sel_w                      : SQ_VTX_WORD1_DST_SEL_W_SIZE;
+          unsigned int dst_sel_z                      : SQ_VTX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_y                      : SQ_VTX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_x                      : SQ_VTX_WORD1_DST_SEL_X_SIZE;
+          unsigned int                                : 9;
+     } sq_vtx_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_t f;
+} sq_vtx_word1_u;
+
+
+/*
+ * SQ_VTX_WORD1_GPR struct
+ */
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_SIZE  7
+#define SQ_VTX_WORD1_GPR_DST_REL_SIZE  1
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_SHIFT 0
+#define SQ_VTX_WORD1_GPR_DST_REL_SHIFT 7
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_MASK  0x0000007f
+#define SQ_VTX_WORD1_GPR_DST_REL_MASK  0x00000080
+
+#define SQ_VTX_WORD1_GPR_MASK \
+     (SQ_VTX_WORD1_GPR_DST_GPR_MASK | \
+      SQ_VTX_WORD1_GPR_DST_REL_MASK)
+
+#define SQ_VTX_WORD1_GPR_DEFAULT       0x000000cd
+
+#define SQ_VTX_WORD1_GPR_GET_DST_GPR(sq_vtx_word1_gpr) \
+     ((sq_vtx_word1_gpr & SQ_VTX_WORD1_GPR_DST_GPR_MASK) >> SQ_VTX_WORD1_GPR_DST_GPR_SHIFT)
+#define SQ_VTX_WORD1_GPR_GET_DST_REL(sq_vtx_word1_gpr) \
+     ((sq_vtx_word1_gpr & SQ_VTX_WORD1_GPR_DST_REL_MASK) >> SQ_VTX_WORD1_GPR_DST_REL_SHIFT)
+
+#define SQ_VTX_WORD1_GPR_SET_DST_GPR(sq_vtx_word1_gpr_reg, dst_gpr) \
+     sq_vtx_word1_gpr_reg = (sq_vtx_word1_gpr_reg & ~SQ_VTX_WORD1_GPR_DST_GPR_MASK) | (dst_gpr << SQ_VTX_WORD1_GPR_DST_GPR_SHIFT)
+#define SQ_VTX_WORD1_GPR_SET_DST_REL(sq_vtx_word1_gpr_reg, dst_rel) \
+     sq_vtx_word1_gpr_reg = (sq_vtx_word1_gpr_reg & ~SQ_VTX_WORD1_GPR_DST_REL_MASK) | (dst_rel << SQ_VTX_WORD1_GPR_DST_REL_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_gpr_t {
+          unsigned int dst_gpr                        : SQ_VTX_WORD1_GPR_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_VTX_WORD1_GPR_DST_REL_SIZE;
+          unsigned int                                : 24;
+     } sq_vtx_word1_gpr_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_gpr_t {
+          unsigned int                                : 24;
+          unsigned int dst_rel                        : SQ_VTX_WORD1_GPR_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_VTX_WORD1_GPR_DST_GPR_SIZE;
+     } sq_vtx_word1_gpr_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_gpr_t f;
+} sq_vtx_word1_gpr_u;
+
+
+/*
+ * SQ_VTX_WORD1_SEM struct
+ */
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE 8
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT 0
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK 0x000000ff
+
+#define SQ_VTX_WORD1_SEM_MASK \
+     (SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK)
+
+#define SQ_VTX_WORD1_SEM_DEFAULT       0x000000cd
+
+#define SQ_VTX_WORD1_SEM_GET_SEMANTIC_ID(sq_vtx_word1_sem) \
+     ((sq_vtx_word1_sem & SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK) >> SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT)
+
+#define SQ_VTX_WORD1_SEM_SET_SEMANTIC_ID(sq_vtx_word1_sem_reg, semantic_id) \
+     sq_vtx_word1_sem_reg = (sq_vtx_word1_sem_reg & ~SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK) | (semantic_id << SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_sem_t {
+          unsigned int semantic_id                    : SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE;
+          unsigned int                                : 24;
+     } sq_vtx_word1_sem_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_sem_t {
+          unsigned int                                : 24;
+          unsigned int semantic_id                    : SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE;
+     } sq_vtx_word1_sem_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_sem_t f;
+} sq_vtx_word1_sem_u;
+
+
+/*
+ * SQ_VTX_WORD2 struct
+ */
+
+#define SQ_VTX_WORD2_OFFSET_SIZE       16
+#define SQ_VTX_WORD2_ENDIAN_SWAP_SIZE  2
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE 1
+#define SQ_VTX_WORD2_MEGA_FETCH_SIZE   1
+#define SQ_VTX_WORD2_ALT_CONST_SIZE    1
+
+#define SQ_VTX_WORD2_OFFSET_SHIFT      0
+#define SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT 16
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT 18
+#define SQ_VTX_WORD2_MEGA_FETCH_SHIFT  19
+#define SQ_VTX_WORD2_ALT_CONST_SHIFT   20
+
+#define SQ_VTX_WORD2_OFFSET_MASK       0x0000ffff
+#define SQ_VTX_WORD2_ENDIAN_SWAP_MASK  0x00030000
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK 0x00040000
+#define SQ_VTX_WORD2_MEGA_FETCH_MASK   0x00080000
+#define SQ_VTX_WORD2_ALT_CONST_MASK    0x00100000
+
+#define SQ_VTX_WORD2_MASK \
+     (SQ_VTX_WORD2_OFFSET_MASK | \
+      SQ_VTX_WORD2_ENDIAN_SWAP_MASK | \
+      SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK | \
+      SQ_VTX_WORD2_MEGA_FETCH_MASK | \
+      SQ_VTX_WORD2_ALT_CONST_MASK)
+
+#define SQ_VTX_WORD2_DEFAULT           0x000dcdcd
+
+#define SQ_VTX_WORD2_GET_OFFSET(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_OFFSET_MASK) >> SQ_VTX_WORD2_OFFSET_SHIFT)
+#define SQ_VTX_WORD2_GET_ENDIAN_SWAP(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_ENDIAN_SWAP_MASK) >> SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT)
+#define SQ_VTX_WORD2_GET_CONST_BUF_NO_STRIDE(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK) >> SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT)
+#define SQ_VTX_WORD2_GET_MEGA_FETCH(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_MEGA_FETCH_MASK) >> SQ_VTX_WORD2_MEGA_FETCH_SHIFT)
+#define SQ_VTX_WORD2_GET_ALT_CONST(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_ALT_CONST_MASK) >> SQ_VTX_WORD2_ALT_CONST_SHIFT)
+
+#define SQ_VTX_WORD2_SET_OFFSET(sq_vtx_word2_reg, offset) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_OFFSET_MASK) | (offset << SQ_VTX_WORD2_OFFSET_SHIFT)
+#define SQ_VTX_WORD2_SET_ENDIAN_SWAP(sq_vtx_word2_reg, endian_swap) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_ENDIAN_SWAP_MASK) | (endian_swap << SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT)
+#define SQ_VTX_WORD2_SET_CONST_BUF_NO_STRIDE(sq_vtx_word2_reg, const_buf_no_stride) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK) | (const_buf_no_stride << SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT)
+#define SQ_VTX_WORD2_SET_MEGA_FETCH(sq_vtx_word2_reg, mega_fetch) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_MEGA_FETCH_MASK) | (mega_fetch << SQ_VTX_WORD2_MEGA_FETCH_SHIFT)
+#define SQ_VTX_WORD2_SET_ALT_CONST(sq_vtx_word2_reg, alt_const) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_ALT_CONST_MASK) | (alt_const << SQ_VTX_WORD2_ALT_CONST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word2_t {
+          unsigned int offset                         : SQ_VTX_WORD2_OFFSET_SIZE;
+          unsigned int endian_swap                    : SQ_VTX_WORD2_ENDIAN_SWAP_SIZE;
+          unsigned int const_buf_no_stride            : SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE;
+          unsigned int mega_fetch                     : SQ_VTX_WORD2_MEGA_FETCH_SIZE;
+          unsigned int alt_const                      : SQ_VTX_WORD2_ALT_CONST_SIZE;
+          unsigned int                                : 11;
+     } sq_vtx_word2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word2_t {
+          unsigned int                                : 11;
+          unsigned int alt_const                      : SQ_VTX_WORD2_ALT_CONST_SIZE;
+          unsigned int mega_fetch                     : SQ_VTX_WORD2_MEGA_FETCH_SIZE;
+          unsigned int const_buf_no_stride            : SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE;
+          unsigned int endian_swap                    : SQ_VTX_WORD2_ENDIAN_SWAP_SIZE;
+          unsigned int offset                         : SQ_VTX_WORD2_OFFSET_SIZE;
+     } sq_vtx_word2_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word2_t f;
+} sq_vtx_word2_u;
+
+#endif /* _SQ_MICRO_REG_H */
+
+
diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile
index f223b2d922..1f286776b5 100644
--- a/src/mesa/drivers/dri/radeon/Makefile
+++ b/src/mesa/drivers/dri/radeon/Makefile
@@ -4,31 +4,53 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = radeon_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c 
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+RADEON_COMMON_SOURCES = \
+	radeon_bo_legacy.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_cs_legacy.c \
+	radeon_dma.c \
+	radeon_debug.c \
+	radeon_fbo.c \
+	radeon_lock.c \
+	radeon_mipmap_tree.c \
+	radeon_queryobj.c \
+	radeon_span.c \
+	radeon_texture.c
+
 DRIVER_SOURCES = \
 	radeon_context.c \
 	radeon_ioctl.c \
-	radeon_lock.c \
 	radeon_screen.c \
 	radeon_state.c \
 	radeon_state_init.c \
 	radeon_tex.c \
-	radeon_texmem.c \
 	radeon_texstate.c \
 	radeon_tcl.c \
 	radeon_swtcl.c \
-	radeon_span.c \
 	radeon_maos.c \
-	radeon_sanity.c 
+	radeon_sanity.c \
+	$(RADEON_COMMON_SOURCES)
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
-	$(DRIVER_SOURCES) 
+	$(DRIVER_SOURCES) \
+	$(CS_SOURCES)
+
+DRIVER_DEFINES = -DRADEON_COMMON=0 \
+				 -Wall
 
-DRIVER_DEFINES = -DRADEON_COMMON=0
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
 X86_SOURCES = 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
new file mode 100644
index 0000000000..7141371633
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
@@ -0,0 +1,219 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_BO_H
+#define RADEON_BO_H
+
+#include <stdio.h>
+#include <stdint.h>
+//#include "radeon_track.h"
+
+/* bo object */
+#define RADEON_BO_FLAGS_MACRO_TILE  1
+#define RADEON_BO_FLAGS_MICRO_TILE  2
+
+struct radeon_bo_manager;
+
+struct radeon_bo {
+    uint32_t                    alignment;
+    uint32_t                    handle;
+    uint32_t                    size;
+    uint32_t                    domains;
+    uint32_t                    flags;
+    unsigned                    cref;
+#ifdef RADEON_BO_TRACK
+    struct radeon_track         *track;
+#endif
+    void                        *ptr;
+    struct radeon_bo_manager    *bom;
+    uint32_t                    space_accounted;
+};
+
+/* bo functions */
+struct radeon_bo_funcs {
+    struct radeon_bo *(*bo_open)(struct radeon_bo_manager *bom,
+                                 uint32_t handle,
+                                 uint32_t size,
+                                 uint32_t alignment,
+                                 uint32_t domains,
+                                 uint32_t flags);
+    void (*bo_ref)(struct radeon_bo *bo);
+    struct radeon_bo *(*bo_unref)(struct radeon_bo *bo);
+    int (*bo_map)(struct radeon_bo *bo, int write);
+    int (*bo_unmap)(struct radeon_bo *bo);
+    int (*bo_wait)(struct radeon_bo *bo);
+    int (*bo_is_static)(struct radeon_bo *bo);
+    int (*bo_set_tiling)(struct radeon_bo *bo, uint32_t tiling_flags,
+			  uint32_t pitch);
+    int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags,
+			  uint32_t *pitch);
+    int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain);
+};
+
+struct radeon_bo_manager {
+    struct radeon_bo_funcs  *funcs;
+    int                     fd;
+
+#ifdef RADEON_BO_TRACK
+    struct radeon_tracker   tracker;
+#endif
+};
+    
+static inline void _radeon_bo_debug(struct radeon_bo *bo,
+                                    const char *op,
+                                    const char *file,
+                                    const char *func,
+                                    int line)
+{
+    fprintf(stderr, "%s %p 0x%08X 0x%08X 0x%08X [%s %s %d]\n",
+            op, bo, bo->handle, bo->size, bo->cref, file, func, line);
+}
+
+static inline struct radeon_bo *_radeon_bo_open(struct radeon_bo_manager *bom,
+                                                uint32_t handle,
+                                                uint32_t size,
+                                                uint32_t alignment,
+                                                uint32_t domains,
+                                                uint32_t flags,
+                                                const char *file,
+                                                const char *func,
+                                                int line)
+{
+    struct radeon_bo *bo;
+
+    bo = bom->funcs->bo_open(bom, handle, size, alignment, domains, flags);
+
+#ifdef RADEON_BO_TRACK
+    if (bo) {
+        bo->track = radeon_tracker_add_track(&bom->tracker, bo->handle);
+        radeon_track_add_event(bo->track, file, func, "open", line);
+    }
+#endif
+    return bo;
+}
+
+static inline void _radeon_bo_ref(struct radeon_bo *bo,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    bo->cref++;
+#ifdef RADEON_BO_TRACK
+    radeon_track_add_event(bo->track, file, func, "ref", line); 
+#endif
+    bo->bom->funcs->bo_ref(bo);
+}
+
+static inline struct radeon_bo *_radeon_bo_unref(struct radeon_bo *bo,
+                                                 const char *file,
+                                                 const char *func,
+                                                 int line)
+{
+    bo->cref--;
+#ifdef RADEON_BO_TRACK
+    radeon_track_add_event(bo->track, file, func, "unref", line);
+    if (bo->cref <= 0) {
+        radeon_tracker_remove_track(&bo->bom->tracker, bo->track);
+        bo->track = NULL;
+    }
+#endif
+    return bo->bom->funcs->bo_unref(bo);
+}
+
+static inline int _radeon_bo_map(struct radeon_bo *bo,
+                                 int write,
+                                 const char *file,
+                                 const char *func,
+                                 int line)
+{
+    return bo->bom->funcs->bo_map(bo, write);
+}
+
+static inline int _radeon_bo_unmap(struct radeon_bo *bo,
+                                   const char *file,
+                                   const char *func,
+                                   int line)
+{
+    return bo->bom->funcs->bo_unmap(bo);
+}
+
+static inline int _radeon_bo_wait(struct radeon_bo *bo,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    return bo->bom->funcs->bo_wait(bo);
+}
+
+static inline int _radeon_bo_is_busy(struct radeon_bo *bo,
+				     uint32_t *domain,
+                                     const char *file,
+                                     const char *func,
+                                     int line)
+{
+    return bo->bom->funcs->bo_is_busy(bo, domain);
+}
+
+static inline int radeon_bo_set_tiling(struct radeon_bo *bo,
+				       uint32_t tiling_flags, uint32_t pitch)
+{
+    return bo->bom->funcs->bo_set_tiling(bo, tiling_flags, pitch);
+}
+
+static inline int radeon_bo_get_tiling(struct radeon_bo *bo,
+				       uint32_t *tiling_flags, uint32_t *pitch)
+{
+    return bo->bom->funcs->bo_get_tiling(bo, tiling_flags, pitch);
+}
+
+static inline int radeon_bo_is_static(struct radeon_bo *bo)
+{
+	if (bo->bom->funcs->bo_is_static)
+		return bo->bom->funcs->bo_is_static(bo);
+	return 0;
+}
+
+#define radeon_bo_open(bom, h, s, a, d, f)\
+    _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_ref(bo)\
+    _radeon_bo_ref(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_unref(bo)\
+    _radeon_bo_unref(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_map(bo, w)\
+    _radeon_bo_map(bo, w, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_unmap(bo)\
+    _radeon_bo_unmap(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_debug(bo, opcode)\
+    _radeon_bo_debug(bo, opcode, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_wait(bo) \
+    _radeon_bo_wait(bo, __FILE__, __func__, __LINE__)
+#define radeon_bo_is_busy(bo, domain) \
+    _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__)
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
new file mode 100644
index 0000000000..3e7547d2f9
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
@@ -0,0 +1,926 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Dave Airlie
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Dave Airlie
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <stdio.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <sys/mman.h>
+#include <sys/ioctl.h>
+#include "xf86drm.h"
+#include "texmem.h"
+#include "main/simple_list.h"
+
+#include "drm.h"
+#include "radeon_drm.h"
+#include "radeon_common.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_macros.h"
+
+/* no seriously texmem.c is this screwed up */
+struct bo_legacy_texture_object {
+    driTextureObject    base;
+    struct bo_legacy *parent;
+};
+
+struct bo_legacy {
+    struct radeon_bo    base;
+    int                 map_count;
+    uint32_t            pending;
+    int                 is_pending;
+    int                 static_bo;
+    uint32_t            offset;
+    struct bo_legacy_texture_object *tobj;
+    int                 validated;
+    int                 dirty;
+    void                *ptr;
+    struct bo_legacy    *next, *prev;
+    struct bo_legacy    *pnext, *pprev;
+};
+
+struct bo_manager_legacy {
+    struct radeon_bo_manager    base;
+    unsigned                    nhandle;
+    unsigned                    nfree_handles;
+    unsigned                    cfree_handles;
+    uint32_t                    current_age;
+    struct bo_legacy            bos;
+    struct bo_legacy            pending_bos;
+    uint32_t                    fb_location;
+    uint32_t                    texture_offset;
+    unsigned                    dma_alloc_size;
+    uint32_t                    dma_buf_count;
+    unsigned                    cpendings;
+    driTextureObject            texture_swapped;
+    driTexHeap                  *texture_heap;
+    struct radeon_screen        *screen;
+    unsigned                    *free_handles;
+};
+
+static void bo_legacy_tobj_destroy(void *data, driTextureObject *t)
+{
+    struct bo_legacy_texture_object *tobj = (struct bo_legacy_texture_object *)t;
+    
+    if (tobj->parent) {
+        tobj->parent->tobj = NULL;
+        tobj->parent->validated = 0;
+    }
+}
+
+static void inline clean_handles(struct bo_manager_legacy *bom)
+{
+  while (bom->cfree_handles > 0 &&
+	 !bom->free_handles[bom->cfree_handles - 1])
+    bom->cfree_handles--;
+
+}
+static int legacy_new_handle(struct bo_manager_legacy *bom, uint32_t *handle)
+{
+    uint32_t tmp;
+
+    *handle = 0;
+    if (bom->nhandle == 0xFFFFFFFF) {
+        return -EINVAL;
+    }
+    if (bom->cfree_handles > 0) {
+        tmp = bom->free_handles[--bom->cfree_handles];
+	clean_handles(bom);
+    } else {
+        bom->cfree_handles = 0;
+        tmp = bom->nhandle++;
+    }
+    assert(tmp);
+    *handle = tmp;
+    return 0;
+}
+
+static int legacy_free_handle(struct bo_manager_legacy *bom, uint32_t handle)
+{
+    uint32_t *handles;
+
+    if (!handle) {
+        return 0;
+    }
+    if (handle == (bom->nhandle - 1)) {
+        int i;
+
+        bom->nhandle--;
+        for (i = bom->cfree_handles - 1; i >= 0; i--) {
+            if (bom->free_handles[i] == (bom->nhandle - 1)) {
+                bom->nhandle--;
+                bom->free_handles[i] = 0;
+            }
+        }
+        clean_handles(bom);
+        return 0;
+    }
+    if (bom->cfree_handles < bom->nfree_handles) {
+        bom->free_handles[bom->cfree_handles++] = handle;
+        return 0;
+    }
+    bom->nfree_handles += 0x100;
+    handles = (uint32_t*)realloc(bom->free_handles, bom->nfree_handles * 4);
+    if (handles == NULL) {
+        bom->nfree_handles -= 0x100;
+        return -ENOMEM;
+    }
+    bom->free_handles = handles;
+    bom->free_handles[bom->cfree_handles++] = handle;
+    return 0;
+}
+
+static void legacy_get_current_age(struct bo_manager_legacy *boml)
+{
+    drm_radeon_getparam_t gp;
+    unsigned char *RADEONMMIO = NULL;
+    int r;
+
+    if (   IS_R300_CLASS(boml->screen) 
+        || IS_R600_CLASS(boml->screen) ) 
+    {
+    	gp.param = RADEON_PARAM_LAST_CLEAR;
+    	gp.value = (int *)&boml->current_age;
+    	r = drmCommandWriteRead(boml->base.fd, DRM_RADEON_GETPARAM,
+       	                     &gp, sizeof(gp));
+    	if (r) {
+       	 fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, r);
+         exit(1);
+       }
+    } 
+    else {
+        RADEONMMIO = boml->screen->mmio.map;
+        boml->current_age = boml->screen->scratch[3];
+        boml->current_age = INREG(RADEON_GUI_SCRATCH_REG3);
+    }
+}
+
+static int legacy_is_pending(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo_legacy->is_pending <= 0) {
+        bo_legacy->is_pending = 0;
+        return 0;
+    }
+    if (boml->current_age >= bo_legacy->pending) {
+        if (boml->pending_bos.pprev == bo_legacy) {
+            boml->pending_bos.pprev = bo_legacy->pprev;
+        }
+        bo_legacy->pprev->pnext = bo_legacy->pnext;
+        if (bo_legacy->pnext) {
+            bo_legacy->pnext->pprev = bo_legacy->pprev;
+        }
+	assert(bo_legacy->is_pending <= bo->cref);
+        while (bo_legacy->is_pending--) {
+	    bo = radeon_bo_unref(bo);
+	    if (!bo)
+	      break;
+        }
+	if (bo)
+	  bo_legacy->is_pending = 0;
+        boml->cpendings--;
+        return 0;
+    }
+    return 1;
+}
+
+static int legacy_wait_pending(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (!bo_legacy->is_pending) {
+        return 0;
+    }
+    /* FIXME: lockup and userspace busy looping that's all the folks */
+    legacy_get_current_age(boml);
+    while (legacy_is_pending(bo)) {
+        usleep(10);
+        legacy_get_current_age(boml);
+    }
+    return 0;
+}
+
+void legacy_track_pending(struct radeon_bo_manager *bom, int debug)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy*) bom;
+    struct bo_legacy *bo_legacy;
+    struct bo_legacy *next;
+
+    legacy_get_current_age(boml);
+    bo_legacy = boml->pending_bos.pnext;
+    while (bo_legacy) {
+        if (debug)
+            fprintf(stderr,"pending %p %d %d %d\n", bo_legacy, bo_legacy->base.size,
+                    boml->current_age, bo_legacy->pending);
+        next = bo_legacy->pnext;
+        if (legacy_is_pending(&(bo_legacy->base))) {
+        }
+        bo_legacy = next;
+    } 
+}
+
+static int legacy_wait_any_pending(struct bo_manager_legacy *boml)
+{
+    struct bo_legacy *bo_legacy;
+
+    legacy_get_current_age(boml);
+    bo_legacy = boml->pending_bos.pnext;
+    if (!bo_legacy)
+      return -1;
+    legacy_wait_pending(&bo_legacy->base);
+    return 0;
+}
+
+static void legacy_kick_all_buffers(struct bo_manager_legacy *boml)
+{
+    struct bo_legacy *legacy;
+
+    legacy = boml->bos.next;
+    while (legacy != &boml->bos) {
+	if (legacy->tobj) {
+	    if (legacy->validated) {
+		driDestroyTextureObject(&legacy->tobj->base);
+		legacy->tobj = 0;
+		legacy->validated = 0;
+	    }
+	}
+	legacy = legacy->next;
+    }
+}
+
+static struct bo_legacy *bo_allocate(struct bo_manager_legacy *boml,
+                                     uint32_t size,
+                                     uint32_t alignment,
+                                     uint32_t domains,
+                                     uint32_t flags)
+{
+    struct bo_legacy *bo_legacy;
+    static int pgsize;
+
+    if (pgsize == 0)
+        pgsize = getpagesize() - 1;
+
+    size = (size + pgsize) & ~pgsize;
+
+    bo_legacy = (struct bo_legacy*)calloc(1, sizeof(struct bo_legacy));
+    if (bo_legacy == NULL) {
+        return NULL;
+    }
+    bo_legacy->base.bom = (struct radeon_bo_manager*)boml;
+    bo_legacy->base.handle = 0;
+    bo_legacy->base.size = size;
+    bo_legacy->base.alignment = alignment;
+    bo_legacy->base.domains = domains;
+    bo_legacy->base.flags = flags;
+    bo_legacy->base.ptr = NULL;
+    bo_legacy->map_count = 0;
+    bo_legacy->next = NULL;
+    bo_legacy->prev = NULL;
+    bo_legacy->pnext = NULL;
+    bo_legacy->pprev = NULL;
+    bo_legacy->next = boml->bos.next;
+    bo_legacy->prev = &boml->bos;
+    boml->bos.next = bo_legacy;
+    if (bo_legacy->next) {
+        bo_legacy->next->prev = bo_legacy;
+    }
+
+    return bo_legacy;
+}
+
+static int bo_dma_alloc(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    drm_radeon_mem_alloc_t alloc;
+    unsigned size;
+    int base_offset;
+    int r;
+
+    /* align size on 4Kb */
+    size = (((4 * 1024) - 1) + bo->size) & ~((4 * 1024) - 1);
+    alloc.region = RADEON_MEM_REGION_GART;
+    alloc.alignment = bo_legacy->base.alignment;
+    alloc.size = size;
+    alloc.region_offset = &base_offset;
+    r = drmCommandWriteRead(bo->bom->fd,
+                            DRM_RADEON_ALLOC,
+                            &alloc,
+                            sizeof(alloc));
+    if (r) {
+        /* ptr is set to NULL if dma allocation failed */
+        bo_legacy->ptr = NULL;
+        return r;
+    }
+    bo_legacy->ptr = boml->screen->gartTextures.map + base_offset;
+    bo_legacy->offset = boml->screen->gart_texture_offset + base_offset;
+    bo->size = size;
+    boml->dma_alloc_size += size;
+    boml->dma_buf_count++;
+    return 0;
+}
+
+static int bo_dma_free(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    drm_radeon_mem_free_t memfree;
+    int r;
+
+    if (bo_legacy->ptr == NULL) {
+        /* ptr is set to NULL if dma allocation failed */
+        return 0;
+    }
+    legacy_get_current_age(boml);
+    memfree.region = RADEON_MEM_REGION_GART;
+    memfree.region_offset  = bo_legacy->offset;
+    memfree.region_offset -= boml->screen->gart_texture_offset;
+    r = drmCommandWrite(boml->base.fd,
+                        DRM_RADEON_FREE,
+                        &memfree,
+                        sizeof(memfree));
+    if (r) {
+        fprintf(stderr, "Failed to free bo[%p] at %08x\n",
+                &bo_legacy->base, memfree.region_offset);
+        fprintf(stderr, "ret = %s\n", strerror(-r));
+        return r;
+    }
+    boml->dma_alloc_size -= bo_legacy->base.size;
+    boml->dma_buf_count--;
+    return 0;
+}
+
+static void bo_free(struct bo_legacy *bo_legacy)
+{
+    struct bo_manager_legacy *boml;
+
+    if (bo_legacy == NULL) {
+        return;
+    }
+    boml = (struct bo_manager_legacy *)bo_legacy->base.bom;
+    bo_legacy->prev->next = bo_legacy->next;
+    if (bo_legacy->next) {
+        bo_legacy->next->prev = bo_legacy->prev;
+    }
+    if (!bo_legacy->static_bo) {
+        legacy_free_handle(boml, bo_legacy->base.handle);
+        if (bo_legacy->base.domains & RADEON_GEM_DOMAIN_GTT) {
+            /* dma buffers */
+            bo_dma_free(&bo_legacy->base);
+        } else {
+  	    driDestroyTextureObject(&bo_legacy->tobj->base);
+	    bo_legacy->tobj = NULL;
+            /* free backing store */
+            free(bo_legacy->ptr);
+        }
+    }
+    memset(bo_legacy, 0 , sizeof(struct bo_legacy));
+    free(bo_legacy);
+}
+
+static struct radeon_bo *bo_open(struct radeon_bo_manager *bom,
+                                 uint32_t handle,
+                                 uint32_t size,
+                                 uint32_t alignment,
+                                 uint32_t domains,
+                                 uint32_t flags)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    struct bo_legacy *bo_legacy;
+    int r;
+
+    if (handle) {
+        bo_legacy = boml->bos.next;
+        while (bo_legacy) {
+            if (bo_legacy->base.handle == handle) {
+                radeon_bo_ref(&(bo_legacy->base));
+                return (struct radeon_bo*)bo_legacy;
+            }
+            bo_legacy = bo_legacy->next;
+        }
+        return NULL;
+    }
+    bo_legacy = bo_allocate(boml, size, alignment, domains, flags);
+    bo_legacy->static_bo = 0;
+    r = legacy_new_handle(boml, &bo_legacy->base.handle);
+    if (r) {
+        bo_free(bo_legacy);
+        return NULL;
+    }
+    if (bo_legacy->base.domains & RADEON_GEM_DOMAIN_GTT) 
+    {
+retry:
+        legacy_track_pending(&boml->base, 0);
+        /* dma buffers */
+
+        r = bo_dma_alloc(&(bo_legacy->base));
+        if (r) 
+        {
+	         if (legacy_wait_any_pending(boml) == -1) 
+             {
+                  bo_free(bo_legacy);
+	              return NULL;
+             }
+	         goto retry;
+	         return NULL;
+        }
+    } 
+    else 
+    {
+        bo_legacy->ptr = malloc(bo_legacy->base.size);
+        if (bo_legacy->ptr == NULL) {
+            bo_free(bo_legacy);
+            return NULL;
+        }
+    }
+    radeon_bo_ref(&(bo_legacy->base));
+
+    return (struct radeon_bo*)bo_legacy;
+}
+
+static void bo_ref(struct radeon_bo *bo)
+{
+}
+
+static struct radeon_bo *bo_unref(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo->cref <= 0) {
+        bo_legacy->prev->next = bo_legacy->next;
+        if (bo_legacy->next) {
+            bo_legacy->next->prev = bo_legacy->prev;
+        }
+        if (!bo_legacy->is_pending) {
+            bo_free(bo_legacy);
+        }
+        return NULL;
+    }
+    return bo;
+}
+
+static int bo_map(struct radeon_bo *bo, int write)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    legacy_wait_pending(bo);
+    bo_legacy->validated = 0;
+    bo_legacy->dirty = 1;
+    bo_legacy->map_count++;
+    bo->ptr = bo_legacy->ptr;
+    /* Read the first pixel in the frame buffer.  This should
+     * be a noop, right?  In fact without this conform fails as reading
+     * from the framebuffer sometimes produces old results -- the
+     * on-card read cache gets mixed up and doesn't notice that the
+     * framebuffer has been updated.
+     *
+     * Note that we should probably be reading some otherwise unused
+     * region of VRAM, otherwise we might get incorrect results when
+     * reading pixels from the top left of the screen.
+     *
+     * I found this problem on an R420 with glean's texCube test.
+     * Note that the R200 span code also *writes* the first pixel in the
+     * framebuffer, but I've found this to be unnecessary.
+     *  -- Nicolai Hähnle, June 2008
+     */
+    if (!(bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+        int p;
+        volatile int *buf = (int*)boml->screen->driScreen->pFB;
+        p = *buf;
+    }
+
+    return 0;
+}
+
+static int bo_unmap(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (--bo_legacy->map_count > 0) 
+    {
+        return 0;
+    }
+    
+    bo->ptr = NULL;
+
+    return 0;
+}
+
+static int bo_is_busy(struct radeon_bo *bo, uint32_t *domain)
+{
+    *domain = 0;
+    if (bo->domains & RADEON_GEM_DOMAIN_GTT)
+        *domain = RADEON_GEM_DOMAIN_GTT;
+    else
+        *domain = RADEON_GEM_DOMAIN_CPU;
+    if (legacy_is_pending(bo))
+        return -EBUSY;
+    else
+        return 0;
+}
+
+static int bo_is_static(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    return bo_legacy->static_bo;
+}
+
+static struct radeon_bo_funcs bo_legacy_funcs = {
+    bo_open,
+    bo_ref,
+    bo_unref,
+    bo_map,
+    bo_unmap,
+    NULL,
+    bo_is_static,
+    NULL,
+    NULL,
+    bo_is_busy
+};
+
+static int bo_vram_validate(struct radeon_bo *bo,
+                            uint32_t *soffset,
+                            uint32_t *eoffset)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    int r;
+    int retry_count = 0, pending_retry = 0;
+    
+    if (!bo_legacy->tobj) {
+	bo_legacy->tobj = CALLOC(sizeof(struct bo_legacy_texture_object));
+	bo_legacy->tobj->parent = bo_legacy;
+	make_empty_list(&bo_legacy->tobj->base);
+	bo_legacy->tobj->base.totalSize = bo->size;
+    retry:
+        r = driAllocateTexture(&boml->texture_heap, 1,
+                               &bo_legacy->tobj->base);
+        if (r) {
+		pending_retry = 0;
+		while(boml->cpendings && pending_retry++ < 10000) {
+			legacy_track_pending(&boml->base, 0);
+			retry_count++;
+			if (retry_count > 2) {
+				free(bo_legacy->tobj);
+				bo_legacy->tobj = NULL;
+				fprintf(stderr, "Ouch! vram_validate failed %d\n", r);
+				return -1;
+			}
+			goto retry;
+		}
+	}
+        bo_legacy->offset = boml->texture_offset +
+                            bo_legacy->tobj->base.memBlock->ofs;
+        bo_legacy->dirty = 1;
+    }
+
+    assert(bo_legacy->tobj->base.memBlock);
+
+    if (bo_legacy->tobj)
+	driUpdateTextureLRU(&bo_legacy->tobj->base);
+
+    if (bo_legacy->dirty || bo_legacy->tobj->base.dirty_images[0]) {
+	    if (IS_R600_CLASS(boml->screen)) {
+		    drm_radeon_texture_t tex;
+		    drm_radeon_tex_image_t tmp;
+		    int ret;
+
+		    tex.offset = bo_legacy->offset;
+		    tex.image = &tmp;
+		    assert(!(tex.offset & 1023));
+
+		    tmp.x = 0;
+		    tmp.y = 0;
+		    tmp.width = bo->size;
+		    tmp.height = 1;
+		    tmp.data = bo_legacy->ptr;
+		    tex.format = RADEON_TXFORMAT_ARGB8888;
+		    tex.width = tmp.width;
+		    tex.height = tmp.height;
+		    tex.pitch = bo->size;
+		    do {
+			    ret = drmCommandWriteRead(bo->bom->fd,
+						      DRM_RADEON_TEXTURE,
+						      &tex,
+						      sizeof(drm_radeon_texture_t));
+			    if (ret) {
+				    if (RADEON_DEBUG & RADEON_IOCTL)
+					    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
+				    usleep(1);
+			    }
+		    } while (ret == -EAGAIN);
+	    } else {
+		    /* Copy to VRAM using a blit.
+		     * All memory is 4K aligned. We're using 1024 pixels wide blits.
+		     */
+		    drm_radeon_texture_t tex;
+		    drm_radeon_tex_image_t tmp;
+		    int ret;
+
+		    tex.offset = bo_legacy->offset;
+		    tex.image = &tmp;
+		    assert(!(tex.offset & 1023));
+
+		    tmp.x = 0;
+		    tmp.y = 0;
+		    if (bo->size < 4096) {
+			    tmp.width = (bo->size + 3) / 4;
+			    tmp.height = 1;
+		    } else {
+			    tmp.width = 1024;
+			    tmp.height = (bo->size + 4095) / 4096;
+		    }
+		    tmp.data = bo_legacy->ptr;
+		    tex.format = RADEON_TXFORMAT_ARGB8888;
+		    tex.width = tmp.width;
+		    tex.height = tmp.height;
+		    tex.pitch = MAX2(tmp.width / 16, 1);
+		    do {
+			    ret = drmCommandWriteRead(bo->bom->fd,
+						      DRM_RADEON_TEXTURE,
+						      &tex,
+						      sizeof(drm_radeon_texture_t));
+			    if (ret) {
+				    if (RADEON_DEBUG & RADEON_IOCTL)
+					    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
+				    usleep(1);
+			    }
+		    } while (ret == -EAGAIN);
+	    }
+	    bo_legacy->dirty = 0;
+	    bo_legacy->tobj->base.dirty_images[0] = 0;
+    }
+    return 0;
+}
+
+/* 
+ *  radeon_bo_legacy_validate -
+ *  returns:
+ *  0 - all good
+ *  -EINVAL - mapped buffer can't be validated
+ *  -EAGAIN - restart validation we've kicked all the buffers out
+ */
+int radeon_bo_legacy_validate(struct radeon_bo *bo,
+                              uint32_t *soffset,
+                              uint32_t *eoffset)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    int r;
+    int retries = 0;
+
+    if (bo_legacy->map_count) {
+        fprintf(stderr, "bo(%p, %d) is mapped (%d) can't valide it.\n",
+                bo, bo->size, bo_legacy->map_count);
+        return -EINVAL;
+    }
+    if (bo_legacy->static_bo || bo_legacy->validated) {
+        *soffset = bo_legacy->offset;
+        *eoffset = bo_legacy->offset + bo->size;
+
+        return 0;
+    }
+    if (!(bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+
+        r = bo_vram_validate(bo, soffset, eoffset);
+        if (r) {
+	    legacy_track_pending(&boml->base, 0);
+	    legacy_kick_all_buffers(boml);
+	    retries++;
+	    if (retries == 2) {
+		fprintf(stderr,"legacy bo: failed to get relocations into aperture\n");
+		assert(0);
+		exit(-1);
+	    }
+	    return -EAGAIN;
+        }
+    }
+    *soffset = bo_legacy->offset;
+    *eoffset = bo_legacy->offset + bo->size;
+    bo_legacy->validated = 1;
+
+    return 0;
+}
+
+void radeon_bo_legacy_pending(struct radeon_bo *bo, uint32_t pending)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    bo_legacy->pending = pending;
+    bo_legacy->is_pending++;
+    /* add to pending list */
+    radeon_bo_ref(bo);
+    if (bo_legacy->is_pending > 1) {
+        return;    
+    }
+    bo_legacy->pprev = boml->pending_bos.pprev;
+    bo_legacy->pnext = NULL;
+    bo_legacy->pprev->pnext = bo_legacy;
+    boml->pending_bos.pprev = bo_legacy;
+    boml->cpendings++;
+}
+
+void radeon_bo_manager_legacy_dtor(struct radeon_bo_manager *bom)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    struct bo_legacy *bo_legacy;
+
+    if (bom == NULL) {
+        return;
+    }
+    bo_legacy = boml->bos.next;
+    while (bo_legacy) {
+        struct bo_legacy *next;
+
+        next = bo_legacy->next;
+        bo_free(bo_legacy);
+        bo_legacy = next;
+    }
+    driDestroyTextureHeap(boml->texture_heap);
+    free(boml->free_handles);
+    free(boml);
+}
+
+static struct bo_legacy *radeon_legacy_bo_alloc_static(struct bo_manager_legacy *bom,
+						       int size,
+						       uint32_t offset)
+{
+    struct bo_legacy *bo;
+
+    bo = bo_allocate(bom, size, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+    if (bo == NULL)
+	return NULL;
+    bo->static_bo = 1;
+    bo->offset = offset + bom->fb_location;
+    bo->base.handle = bo->offset;
+    bo->ptr = bom->screen->driScreen->pFB + offset;
+    if (bo->base.handle > bom->nhandle) {
+        bom->nhandle = bo->base.handle + 1;
+    }
+    radeon_bo_ref(&(bo->base));
+    return bo;
+}
+
+struct radeon_bo_manager *radeon_bo_manager_legacy_ctor(struct radeon_screen *scrn)
+{
+    struct bo_manager_legacy *bom;
+    struct bo_legacy *bo;
+    unsigned size;
+
+    bom = (struct bo_manager_legacy*)
+          calloc(1, sizeof(struct bo_manager_legacy));
+    if (bom == NULL) {
+        return NULL;
+    }
+
+    make_empty_list(&bom->texture_swapped);
+
+    bom->texture_heap = driCreateTextureHeap(0,
+                                             bom,
+                                             scrn->texSize[0],
+                                             12,
+                                             RADEON_NR_TEX_REGIONS,
+                                             (drmTextureRegionPtr)scrn->sarea->tex_list[0],
+                                             &scrn->sarea->tex_age[0],
+                                             &bom->texture_swapped,
+                                             sizeof(struct bo_legacy_texture_object),
+                                             &bo_legacy_tobj_destroy);
+    bom->texture_offset = scrn->texOffset[0];
+
+    bom->base.funcs = &bo_legacy_funcs;
+    bom->base.fd = scrn->driScreen->fd;
+    bom->bos.next = NULL;
+    bom->bos.prev = NULL;
+    bom->pending_bos.pprev = &bom->pending_bos;
+    bom->pending_bos.pnext = NULL;
+    bom->screen = scrn;
+    bom->fb_location = scrn->fbLocation;
+    bom->nhandle = 1;
+    bom->cfree_handles = 0;
+    bom->nfree_handles = 0x400;
+    bom->free_handles = (uint32_t*)malloc(bom->nfree_handles * 4);
+    if (bom->free_handles == NULL) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+
+    /* biggest framebuffer size */
+    size = 4096*4096*4; 
+
+    /* allocate front */
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->frontOffset);
+
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags = RADEON_BO_FLAGS_MACRO_TILE;
+    }
+
+    /* allocate back */
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->backOffset);
+
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags = RADEON_BO_FLAGS_MACRO_TILE;
+    }
+
+    /* allocate depth */
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->depthOffset);
+
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    bo->base.flags = 0;
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags |= RADEON_BO_FLAGS_MACRO_TILE;
+        bo->base.flags |= RADEON_BO_FLAGS_MICRO_TILE;
+    }
+    return (struct radeon_bo_manager*)bom;
+}
+
+void radeon_bo_legacy_texture_age(struct radeon_bo_manager *bom)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    DRI_AGE_TEXTURES(boml->texture_heap);
+}
+
+unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo_legacy->static_bo || (bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+        return 0;
+    }
+    return bo->size;
+}
+
+/*
+ * Fake up a bo for things like texture image_override.
+ * bo->offset already includes fb_location
+ */
+struct radeon_bo *radeon_legacy_bo_alloc_fake(struct radeon_bo_manager *bom,
+					      int size,
+	                                      uint32_t offset)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    struct bo_legacy *bo;
+
+    bo = bo_allocate(boml, size, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+    if (bo == NULL)
+	return NULL;
+    bo->static_bo = 1;
+    bo->offset = offset;
+    bo->base.handle = bo->offset;
+    bo->ptr = boml->screen->driScreen->pFB + (offset - boml->fb_location);
+    if (bo->base.handle > boml->nhandle) {
+        boml->nhandle = bo->base.handle + 1;
+    }
+    radeon_bo_ref(&(bo->base));
+    return &(bo->base);
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h
new file mode 100644
index 0000000000..2cf15dfaff
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h
@@ -0,0 +1,50 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_BO_LEGACY_H
+#define RADEON_BO_LEGACY_H
+
+#include "radeon_screen.h"
+
+void radeon_bo_legacy_pending(struct radeon_bo *bo, uint32_t pending);
+int radeon_bo_legacy_validate(struct radeon_bo *bo,
+                              uint32_t *soffset,
+                              uint32_t *eoffset);
+struct radeon_bo_manager *radeon_bo_manager_legacy_ctor(struct radeon_screen *scrn);
+void radeon_bo_manager_legacy_dtor(struct radeon_bo_manager *bom);
+void radeon_bo_legacy_texture_age(struct radeon_bo_manager *bom);
+unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo);
+struct radeon_bo *radeon_legacy_bo_alloc_fake(struct radeon_bo_manager *bom,
+					      int size,
+	                                      uint32_t offset);
+void legacy_track_pending(struct radeon_bo_manager *bom, int debug);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
new file mode 100644
index 0000000000..4520a7d7d4
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
@@ -0,0 +1,99 @@
+#ifndef RADEON_CS_WRAPPER_H
+#define RADEON_CS_WRAPPER_H
+
+#ifdef HAVE_LIBDRM_RADEON
+
+#include "radeon_bo.h"
+#include "radeon_bo_gem.h"
+#include "radeon_cs.h"
+#include "radeon_cs_gem.h"
+
+#else
+#include <stdint.h>
+
+#define RADEON_GEM_DOMAIN_CPU 0x1   // Cached CPU domain
+#define RADEON_GEM_DOMAIN_GTT 0x2   // GTT or cache flushed
+#define RADEON_GEM_DOMAIN_VRAM 0x4  // VRAM domain
+
+#define RADEON_TILING_MACRO 0x1
+#define RADEON_TILING_MICRO 0x2
+#define RADEON_TILING_SWAP 0x4
+#define RADEON_TILING_SURFACE 0x8 /* this object requires a surface
+				   * when mapped - i.e. front buffer */
+
+/* to be used to build locally in mesa with no libdrm bits */
+#include "../radeon/radeon_bo_drm.h"
+#include "../radeon/radeon_cs_drm.h"
+
+#ifndef DRM_RADEON_GEM_INFO
+#define DRM_RADEON_GEM_INFO 0x1c
+
+struct drm_radeon_gem_info {
+        uint64_t gart_size;
+        uint64_t vram_size;
+        uint64_t vram_visible;
+};
+
+struct drm_radeon_info {
+	uint32_t request;
+	uint32_t pad;
+	uint32_t value;
+};
+#endif
+
+#ifndef RADEON_PARAM_DEVICE_ID
+#define RADEON_PARAM_DEVICE_ID 16
+#endif
+
+#ifndef RADEON_PARAM_NUM_Z_PIPES
+#define RADEON_PARAM_NUM_Z_PIPES 17
+#endif
+
+#ifndef RADEON_INFO_DEVICE_ID
+#define RADEON_INFO_DEVICE_ID 0
+#endif
+#ifndef RADEON_INFO_NUM_GB_PIPES
+#define RADEON_INFO_NUM_GB_PIPES 0
+#endif
+
+#ifndef RADEON_INFO_NUM_Z_PIPES
+#define RADEON_INFO_NUM_Z_PIPES 0
+#endif
+
+#ifndef DRM_RADEON_INFO
+#define DRM_RADEON_INFO 0x1
+#endif
+
+
+static inline uint32_t radeon_gem_name_bo(struct radeon_bo *dummy)
+{
+  return 0;
+}
+
+static inline void *radeon_bo_manager_gem_ctor(int fd)
+{
+  return NULL;
+}
+
+static inline void radeon_bo_manager_gem_dtor(void *dummy)
+{
+}
+
+static inline void *radeon_cs_manager_gem_ctor(int fd)
+{
+  return NULL;
+}
+
+static inline void radeon_cs_manager_gem_dtor(void *dummy)
+{
+}
+
+static inline void radeon_tracker_print(void *ptr, int io)
+{
+}
+#endif
+
+#include "radeon_bo_legacy.h"
+#include "radeon_cs_legacy.h"
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c
new file mode 100644
index 0000000000..a24b6dac26
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c
@@ -0,0 +1,222 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "radeon_buffer_objects.h"
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/bufferobj.h"
+
+#include "radeon_common.h"
+
+struct radeon_buffer_object *
+get_radeon_buffer_object(struct gl_buffer_object *obj)
+{
+    return (struct radeon_buffer_object *) obj;
+}
+
+static struct gl_buffer_object *
+radeonNewBufferObject(GLcontext * ctx,
+                      GLuint name,
+                      GLenum target)
+{
+    struct radeon_buffer_object *obj = CALLOC_STRUCT(radeon_buffer_object);
+
+    _mesa_initialize_buffer_object(&obj->Base, name, target);
+
+    obj->bo = NULL;
+
+    return &obj->Base;
+}
+
+/**
+ * Called via glDeleteBuffersARB().
+ */
+static void
+radeonDeleteBufferObject(GLcontext * ctx,
+                         struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (obj->Pointer) {
+        radeon_bo_unmap(radeon_obj->bo);
+    }
+
+    if (radeon_obj->bo) {
+        radeon_bo_unref(radeon_obj->bo);
+    }
+
+    _mesa_free(radeon_obj);
+}
+
+
+/**
+ * Allocate space for and store data in a buffer object.  Any data that was
+ * previously stored in the buffer object is lost.  If data is NULL,
+ * memory will be allocated, but no copy will occur.
+ * Called via ctx->Driver.BufferData().
+ * \return GL_TRUE for success, GL_FALSE if out of memory
+ */
+static GLboolean
+radeonBufferData(GLcontext * ctx,
+                 GLenum target,
+                 GLsizeiptrARB size,
+                 const GLvoid * data,
+                 GLenum usage,
+                 struct gl_buffer_object *obj)
+{
+    radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_obj->Base.Size = size;
+    radeon_obj->Base.Usage = usage;
+
+    if (radeon_obj->bo != NULL) {
+        radeon_bo_unref(radeon_obj->bo);
+        radeon_obj->bo = NULL;
+    }
+
+    if (size != 0) {
+        radeon_obj->bo = radeon_bo_open(radeon->radeonScreen->bom,
+                                        0,
+                                        size,
+                                        32,
+                                        RADEON_GEM_DOMAIN_GTT,
+                                        0);
+
+        if (!radeon_obj->bo)
+            return GL_FALSE;
+
+        if (data != NULL) {
+            radeon_bo_map(radeon_obj->bo, GL_TRUE);
+
+            _mesa_memcpy(radeon_obj->bo->ptr, data, size);
+
+            radeon_bo_unmap(radeon_obj->bo);
+        }
+    }
+    return GL_TRUE;
+}
+
+/**
+ * Replace data in a subrange of buffer object.  If the data range
+ * specified by size + offset extends beyond the end of the buffer or
+ * if data is NULL, no copy is performed.
+ * Called via glBufferSubDataARB().
+ */
+static void
+radeonBufferSubData(GLcontext * ctx,
+                    GLenum target,
+                    GLintptrARB offset,
+                    GLsizeiptrARB size,
+                    const GLvoid * data,
+                    struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_bo_map(radeon_obj->bo, GL_TRUE);
+
+    _mesa_memcpy(radeon_obj->bo->ptr + offset, data, size);
+
+    radeon_bo_unmap(radeon_obj->bo);
+}
+
+/**
+ * Called via glGetBufferSubDataARB()
+ */
+static void
+radeonGetBufferSubData(GLcontext * ctx,
+                       GLenum target,
+                       GLintptrARB offset,
+                       GLsizeiptrARB size,
+                       GLvoid * data,
+                       struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_bo_map(radeon_obj->bo, GL_FALSE);
+
+    _mesa_memcpy(data, radeon_obj->bo->ptr + offset, size);
+
+    radeon_bo_unmap(radeon_obj->bo);
+}
+
+/**
+ * Called via glMapBufferARB()
+ */
+static void *
+radeonMapBuffer(GLcontext * ctx,
+                GLenum target,
+                GLenum access,
+                struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (access == GL_WRITE_ONLY_ARB) {
+        ctx->Driver.Flush(ctx);
+    }
+
+    if (radeon_obj->bo == NULL) {
+        obj->Pointer = NULL;
+        return NULL;
+    }
+
+    radeon_bo_map(radeon_obj->bo, access == GL_WRITE_ONLY_ARB);
+
+    return obj->Pointer = radeon_obj->bo->ptr;
+}
+
+
+/**
+ * Called via glUnmapBufferARB()
+ */
+static GLboolean
+radeonUnmapBuffer(GLcontext * ctx,
+                  GLenum target,
+                  struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (radeon_obj->bo != NULL) {
+        radeon_bo_unmap(radeon_obj->bo);
+        obj->Pointer = NULL;
+    }
+
+    return GL_TRUE;
+}
+
+void
+radeonInitBufferObjectFuncs(struct dd_function_table *functions)
+{
+    functions->NewBufferObject = radeonNewBufferObject;
+    functions->DeleteBuffer = radeonDeleteBufferObject;
+    functions->BufferData = radeonBufferData;
+    functions->BufferSubData = radeonBufferSubData;
+    functions->GetBufferSubData = radeonGetBufferSubData;
+    functions->MapBuffer = radeonMapBuffer;
+    functions->UnmapBuffer = radeonUnmapBuffer;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h
new file mode 100644
index 0000000000..d681960825
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef RADEON_BUFFER_OBJECTS_H
+#define RADEON_BUFFER_OBJECTS_H
+
+#include "main/mtypes.h"
+
+struct radeon_bo;
+
+/**
+ * Radeon vertex/pixel buffer object, derived from Mesa's gl_buffer_object.
+ */
+struct radeon_buffer_object
+{
+   struct gl_buffer_object Base;
+   struct radeon_bo *bo;
+};
+
+struct radeon_buffer_object *
+get_radeon_buffer_object(struct gl_buffer_object *obj);
+
+/**
+ * Hook the bufferobject implementation into mesa:
+ */
+void radeonInitBufferObjectFuncs(struct dd_function_table *functions);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_chipset.h b/src/mesa/drivers/dri/radeon/radeon_chipset.h
index f6bd1eb83f..46a9cd5ff8 100644
--- a/src/mesa/drivers/dri/radeon/radeon_chipset.h
+++ b/src/mesa/drivers/dri/radeon/radeon_chipset.h
@@ -255,6 +255,145 @@
 #define PCI_CHIP_RS740_796E             0x796E
 #define PCI_CHIP_RS740_796F             0x796F
 
+#define PCI_CHIP_R600_9400              0x9400
+#define PCI_CHIP_R600_9401              0x9401
+#define PCI_CHIP_R600_9402              0x9402
+#define PCI_CHIP_R600_9403              0x9403
+#define PCI_CHIP_R600_9405              0x9405
+#define PCI_CHIP_R600_940A              0x940A
+#define PCI_CHIP_R600_940B              0x940B
+#define PCI_CHIP_R600_940F              0x940F
+
+#define PCI_CHIP_RV610_94C0             0x94C0
+#define PCI_CHIP_RV610_94C1             0x94C1
+#define PCI_CHIP_RV610_94C3             0x94C3
+#define PCI_CHIP_RV610_94C4             0x94C4
+#define PCI_CHIP_RV610_94C5             0x94C5
+#define PCI_CHIP_RV610_94C6             0x94C6
+#define PCI_CHIP_RV610_94C7             0x94C7
+#define PCI_CHIP_RV610_94C8             0x94C8
+#define PCI_CHIP_RV610_94C9             0x94C9
+#define PCI_CHIP_RV610_94CB             0x94CB
+#define PCI_CHIP_RV610_94CC             0x94CC
+#define PCI_CHIP_RV610_94CD             0x94CD
+
+#define PCI_CHIP_RV630_9580             0x9580
+#define PCI_CHIP_RV630_9581             0x9581
+#define PCI_CHIP_RV630_9583             0x9583
+#define PCI_CHIP_RV630_9586             0x9586
+#define PCI_CHIP_RV630_9587             0x9587
+#define PCI_CHIP_RV630_9588             0x9588
+#define PCI_CHIP_RV630_9589             0x9589
+#define PCI_CHIP_RV630_958A             0x958A
+#define PCI_CHIP_RV630_958B             0x958B
+#define PCI_CHIP_RV630_958C             0x958C
+#define PCI_CHIP_RV630_958D             0x958D
+#define PCI_CHIP_RV630_958E             0x958E
+#define PCI_CHIP_RV630_958F             0x958F
+
+#define PCI_CHIP_RV670_9500             0x9500
+#define PCI_CHIP_RV670_9501             0x9501
+#define PCI_CHIP_RV670_9504             0x9504
+#define PCI_CHIP_RV670_9505             0x9505
+#define PCI_CHIP_RV670_9506             0x9506
+#define PCI_CHIP_RV670_9507             0x9507
+#define PCI_CHIP_RV670_9508             0x9508
+#define PCI_CHIP_RV670_9509             0x9509
+#define PCI_CHIP_RV670_950F             0x950F
+#define PCI_CHIP_RV670_9511             0x9511
+#define PCI_CHIP_RV670_9515             0x9515
+#define PCI_CHIP_RV670_9517             0x9517
+#define PCI_CHIP_RV670_9519             0x9519
+
+#define PCI_CHIP_RV620_95C0             0x95C0
+#define PCI_CHIP_RV620_95C2             0x95C2
+#define PCI_CHIP_RV620_95C4             0x95C4
+#define PCI_CHIP_RV620_95C5             0x95C5
+#define PCI_CHIP_RV620_95C6             0x95C6
+#define PCI_CHIP_RV620_95C7             0x95C7
+#define PCI_CHIP_RV620_95C9             0x95C9
+#define PCI_CHIP_RV620_95CC             0x95CC
+#define PCI_CHIP_RV620_95CD             0x95CD
+#define PCI_CHIP_RV620_95CE             0x95CE
+#define PCI_CHIP_RV620_95CF             0x95CF
+
+#define PCI_CHIP_RV635_9590             0x9590
+#define PCI_CHIP_RV635_9591             0x9591
+#define PCI_CHIP_RV635_9593             0x9593
+#define PCI_CHIP_RV635_9595             0x9595
+#define PCI_CHIP_RV635_9596             0x9596
+#define PCI_CHIP_RV635_9597             0x9597
+#define PCI_CHIP_RV635_9598             0x9598
+#define PCI_CHIP_RV635_9599             0x9599
+#define PCI_CHIP_RV635_959B             0x959B
+
+#define PCI_CHIP_RS780_9610             0x9610
+#define PCI_CHIP_RS780_9611             0x9611
+#define PCI_CHIP_RS780_9612             0x9612
+#define PCI_CHIP_RS780_9613             0x9613
+#define PCI_CHIP_RS780_9614             0x9614
+#define PCI_CHIP_RS780_9615             0x9615
+#define PCI_CHIP_RS780_9616             0x9616
+
+#define PCI_CHIP_RS880_9710             0x9710
+#define PCI_CHIP_RS880_9711             0x9711
+#define PCI_CHIP_RS880_9712             0x9712
+#define PCI_CHIP_RS880_9713             0x9713
+#define PCI_CHIP_RS880_9714             0x9714
+
+#define PCI_CHIP_RV770_9440             0x9440
+#define PCI_CHIP_RV770_9441             0x9441
+#define PCI_CHIP_RV770_9442             0x9442
+#define PCI_CHIP_RV770_9443             0x9443
+#define PCI_CHIP_RV770_9444             0x9444
+#define PCI_CHIP_RV770_9446             0x9446
+#define PCI_CHIP_RV770_944A             0x944A
+#define PCI_CHIP_RV770_944B             0x944B
+#define PCI_CHIP_RV770_944C             0x944C
+#define PCI_CHIP_RV770_944E             0x944E
+#define PCI_CHIP_RV770_9450             0x9450
+#define PCI_CHIP_RV770_9452             0x9452
+#define PCI_CHIP_RV770_9456             0x9456
+#define PCI_CHIP_RV770_945A             0x945A
+#define PCI_CHIP_RV770_945B             0x945B
+#define PCI_CHIP_RV790_9460             0x9460
+#define PCI_CHIP_RV790_9462             0x9462
+#define PCI_CHIP_RV770_946A             0x946A
+#define PCI_CHIP_RV770_946B             0x946B
+#define PCI_CHIP_RV770_947A             0x947A
+#define PCI_CHIP_RV770_947B             0x947B
+
+#define PCI_CHIP_RV730_9480             0x9480
+#define PCI_CHIP_RV730_9487             0x9487
+#define PCI_CHIP_RV730_9488             0x9488
+#define PCI_CHIP_RV730_9489             0x9489
+#define PCI_CHIP_RV730_948F             0x948F
+#define PCI_CHIP_RV730_9490             0x9490
+#define PCI_CHIP_RV730_9491             0x9491
+#define PCI_CHIP_RV730_9495             0x9495
+#define PCI_CHIP_RV730_9498             0x9498
+#define PCI_CHIP_RV730_949C             0x949C
+#define PCI_CHIP_RV730_949E             0x949E
+#define PCI_CHIP_RV730_949F             0x949F
+
+#define PCI_CHIP_RV710_9540             0x9540
+#define PCI_CHIP_RV710_9541             0x9541
+#define PCI_CHIP_RV710_9542             0x9542
+#define PCI_CHIP_RV710_954E             0x954E
+#define PCI_CHIP_RV710_954F             0x954F
+#define PCI_CHIP_RV710_9552             0x9552
+#define PCI_CHIP_RV710_9553             0x9553
+#define PCI_CHIP_RV710_9555             0x9555
+#define PCI_CHIP_RV710_9557             0x9557
+
+#define PCI_CHIP_RV740_94A0             0x94A0
+#define PCI_CHIP_RV740_94A1             0x94A1
+#define PCI_CHIP_RV740_94A3             0x94A3
+#define PCI_CHIP_RV740_94B1             0x94B1
+#define PCI_CHIP_RV740_94B3             0x94B3
+#define PCI_CHIP_RV740_94B4             0x94B4
+#define PCI_CHIP_RV740_94B5             0x94B5
+#define PCI_CHIP_RV740_94B9             0x94B9
 
 enum {
    CHIP_FAMILY_R100,
@@ -282,6 +421,18 @@ enum {
    CHIP_FAMILY_R580,
    CHIP_FAMILY_RV560,
    CHIP_FAMILY_RV570,
+   CHIP_FAMILY_R600,
+   CHIP_FAMILY_RV610,
+   CHIP_FAMILY_RV630,
+   CHIP_FAMILY_RV670,
+   CHIP_FAMILY_RV620,
+   CHIP_FAMILY_RV635,
+   CHIP_FAMILY_RS780,
+   CHIP_FAMILY_RS880,
+   CHIP_FAMILY_RV770,
+   CHIP_FAMILY_RV730,
+   CHIP_FAMILY_RV710,
+   CHIP_FAMILY_RV740,
    CHIP_FAMILY_LAST
 };
 
@@ -289,6 +440,7 @@ enum {
 #define RADEON_CLASS_R100		(0 << 0)
 #define RADEON_CLASS_R200		(1 << 0)
 #define RADEON_CLASS_R300		(2 << 0)
+#define RADEON_CLASS_R600		(3 << 0)
 #define RADEON_CLASS_MASK		(3 << 0)
 
 #define RADEON_CHIPSET_TCL		(1 << 2)	/* tcl support - any radeon */
diff --git a/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h
new file mode 100644
index 0000000000..6fcd1ce7ca
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h
@@ -0,0 +1,121 @@
+#ifndef COMMON_CMDBUF_H
+#define COMMON_CMDBUF_H
+
+#include "radeon_bocs_wrapper.h"
+
+GLboolean rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller);
+int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller);
+int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller);
+void rcommonInitCmdBuf(radeonContextPtr rmesa);
+void rcommonDestroyCmdBuf(radeonContextPtr rmesa);
+
+void rcommonBeginBatch(radeonContextPtr rmesa,
+		       int n,
+		       int dostate,
+		       const char *file,
+		       const char *function,
+		       int line);
+
+/* +r6/r7 : code here moved */
+
+#define CP_PACKET2  (2 << 30)
+#define CP_PACKET0(reg, n)	(RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2))
+#define CP_PACKET0_ONE(reg, n)	(RADEON_CP_PACKET0 | RADEON_CP_PACKET0_ONE_REG_WR | ((n)<<16) | ((reg)>>2))
+#define CP_PACKET3(pkt, n)	(RADEON_CP_PACKET3 | (pkt) | ((n) << 16))
+
+/**
+ * Every function writing to the command buffer needs to declare this
+ * to get the necessary local variables.
+ */
+#define BATCH_LOCALS(rmesa) \
+	const radeonContextPtr b_l_rmesa = rmesa
+
+/**
+ * Prepare writing n dwords to the command buffer,
+ * including producing any necessary state emits on buffer wraparound.
+ */
+#define BEGIN_BATCH(n) rcommonBeginBatch(b_l_rmesa, n, 1, __FILE__, __FUNCTION__, __LINE__)
+
+/**
+ * Same as BEGIN_BATCH, but do not cause automatic state emits.
+ */
+#define BEGIN_BATCH_NO_AUTOSTATE(n) rcommonBeginBatch(b_l_rmesa, n, 0, __FILE__, __FUNCTION__, __LINE__)
+
+/**
+ * Write one dword to the command buffer.
+ */
+#define OUT_BATCH(data) \
+	do { \
+        radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\
+	} while(0)
+
+/**
+ * Write a relocated dword to the command buffer.
+ */
+#define OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags) 	\
+	do { 							\
+	int  __offset = (offset);				\
+        if (0 && __offset) {					\
+            fprintf(stderr, "(%s:%s:%d) offset : %d\n",		\
+            __FILE__, __FUNCTION__, __LINE__, __offset);	\
+        }							\
+        radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, __offset);	\
+        radeon_cs_write_reloc(b_l_rmesa->cmdbuf.cs, 		\
+                              bo, rd, wd, flags);		\
+	if (!b_l_rmesa->radeonScreen->kernel_mm) 		\
+		b_l_rmesa->cmdbuf.cs->section_cdw += 2;		\
+	} while(0)
+
+
+/**
+ * Write n dwords from ptr to the command buffer.
+ */
+#define OUT_BATCH_TABLE(ptr,n) \
+	do { \
+		radeon_cs_write_table(b_l_rmesa->cmdbuf.cs, (ptr), (n));\
+	} while(0)
+
+/**
+ * Finish writing dwords to the command buffer.
+ * The number of (direct or indirect) OUT_BATCH calls between the previous
+ * BEGIN_BATCH and END_BATCH must match the number specified at BEGIN_BATCH time.
+ */
+#define END_BATCH() \
+	do { \
+        radeon_cs_end(b_l_rmesa->cmdbuf.cs, __FILE__, __FUNCTION__, __LINE__);\
+	} while(0)
+
+/**
+ * After the last END_BATCH() of rendering, this indicates that flushing
+ * the command buffer now is okay.
+ */
+#define COMMIT_BATCH() \
+	do { \
+	} while(0)
+
+
+/** Single register write to command buffer; requires 2 dwords. */
+#define OUT_BATCH_REGVAL(reg, val) \
+	OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), 1)); \
+	OUT_BATCH((val))
+
+/** Continuous register range write to command buffer; requires 1 dword,
+ * expects count dwords afterwards for register contents. */
+#define OUT_BATCH_REGSEQ(reg, count) \
+	OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), (count)))
+
+/** Write a 32 bit float to the ring; requires 1 dword. */
+#define OUT_BATCH_FLOAT32(f) \
+	OUT_BATCH(radeonPackFloat32((f)))
+
+/* +r6/r7 : code here moved */
+
+/* Fire the buffered vertices no matter what.
+ */
+static INLINE void radeon_firevertices(radeonContextPtr radeon)
+{
+   if (radeon->cmdbuf.cs->cdw || radeon->dma.flush )
+      radeon->glCtx->Driver.Flush(radeon->glCtx); /* +r6/r7 */
+}
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
new file mode 100644
index 0000000000..a4c7b40798
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -0,0 +1,1349 @@
+/**************************************************************************
+
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/*
+   - Scissor implementation
+   - buffer swap/copy ioctls
+   - finish/flush
+   - state emission
+   - cmdbuffer management
+*/
+
+#include <errno.h>
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "drivers/common/meta.h"
+
+#include "vblank.h"
+
+#include "radeon_common.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_lock.h"
+#include "radeon_drm.h"
+#include "radeon_queryobj.h"
+
+/**
+ * Enable verbose debug output for emit code.
+ * 0 no output
+ * 1 most output
+ * 2 also print state alues
+ */
+#define RADEON_CMDBUF         0
+
+/* =============================================================
+ * Scissoring
+ */
+
+static GLboolean intersect_rect(drm_clip_rect_t * out,
+				drm_clip_rect_t * a, drm_clip_rect_t * b)
+{
+	*out = *a;
+	if (b->x1 > out->x1)
+		out->x1 = b->x1;
+	if (b->y1 > out->y1)
+		out->y1 = b->y1;
+	if (b->x2 < out->x2)
+		out->x2 = b->x2;
+	if (b->y2 < out->y2)
+		out->y2 = b->y2;
+	if (out->x1 >= out->x2)
+		return GL_FALSE;
+	if (out->y1 >= out->y2)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void radeonRecalcScissorRects(radeonContextPtr radeon)
+{
+	drm_clip_rect_t *out;
+	int i;
+
+	/* Grow cliprect store?
+	 */
+	if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) {
+		while (radeon->state.scissor.numAllocedClipRects <
+		       radeon->numClipRects) {
+			radeon->state.scissor.numAllocedClipRects += 1;	/* zero case */
+			radeon->state.scissor.numAllocedClipRects *= 2;
+		}
+
+		if (radeon->state.scissor.pClipRects)
+			FREE(radeon->state.scissor.pClipRects);
+
+		radeon->state.scissor.pClipRects =
+			MALLOC(radeon->state.scissor.numAllocedClipRects *
+			       sizeof(drm_clip_rect_t));
+
+		if (radeon->state.scissor.pClipRects == NULL) {
+			radeon->state.scissor.numAllocedClipRects = 0;
+			return;
+		}
+	}
+
+	out = radeon->state.scissor.pClipRects;
+	radeon->state.scissor.numClipRects = 0;
+
+	for (i = 0; i < radeon->numClipRects; i++) {
+		if (intersect_rect(out,
+				   &radeon->pClipRects[i],
+				   &radeon->state.scissor.rect)) {
+			radeon->state.scissor.numClipRects++;
+			out++;
+		}
+	}
+
+	if (radeon->vtbl.update_scissor)
+	   radeon->vtbl.update_scissor(radeon->glCtx);
+}
+
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off)
+{
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(radeon);
+	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+
+	if (radeon->constant_cliprect) {
+		radeon->fboRect.x1 = 0;
+		radeon->fboRect.y1 = 0;
+		radeon->fboRect.x2 = radeon->glCtx->DrawBuffer->Width;
+		radeon->fboRect.y2 = radeon->glCtx->DrawBuffer->Height;
+
+		*cliprects = &radeon->fboRect;
+		*num_cliprects = 1;
+		*x_off = 0;
+		*y_off = 0;
+	} else if (radeon->front_cliprects ||
+		   rfb->pf_active || dPriv->numBackClipRects == 0) {
+		*cliprects = dPriv->pClipRects;
+		*num_cliprects = dPriv->numClipRects;
+		*x_off = dPriv->x;
+		*y_off = dPriv->y;
+	} else {
+		*num_cliprects = dPriv->numBackClipRects;
+		*cliprects = dPriv->pBackClipRects;
+		*x_off = dPriv->backX;
+		*y_off = dPriv->backY;
+	}
+}
+
+/**
+ * Update cliprects and scissors.
+ */
+void radeonSetCliprects(radeonContextPtr radeon)
+{
+	__DRIdrawablePrivate *const drawable = radeon_get_drawable(radeon);
+	__DRIdrawablePrivate *const readable = radeon_get_readable(radeon);
+	struct radeon_framebuffer *const draw_rfb = drawable->driverPrivate;
+	struct radeon_framebuffer *const read_rfb = readable->driverPrivate;
+	int x_off, y_off;
+
+	radeon_get_cliprects(radeon, &radeon->pClipRects,
+			     &radeon->numClipRects, &x_off, &y_off);
+
+	if ((draw_rfb->base.Width != drawable->w) ||
+	    (draw_rfb->base.Height != drawable->h)) {
+		_mesa_resize_framebuffer(radeon->glCtx, &draw_rfb->base,
+					 drawable->w, drawable->h);
+		draw_rfb->base.Initialized = GL_TRUE;
+	}
+
+	if (drawable != readable) {
+		if ((read_rfb->base.Width != readable->w) ||
+		    (read_rfb->base.Height != readable->h)) {
+			_mesa_resize_framebuffer(radeon->glCtx, &read_rfb->base,
+						 readable->w, readable->h);
+			read_rfb->base.Initialized = GL_TRUE;
+		}
+	}
+
+	if (radeon->state.scissor.enabled)
+		radeonRecalcScissorRects(radeon);
+
+}
+
+
+
+void radeonUpdateScissor( GLcontext *ctx )
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	GLint x = ctx->Scissor.X, y = ctx->Scissor.Y;
+	GLsizei w = ctx->Scissor.Width, h = ctx->Scissor.Height;
+	int x1, y1, x2, y2;
+	int min_x, min_y, max_x, max_y;
+
+	if (!ctx->DrawBuffer)
+	    return;
+	min_x = min_y = 0;
+	max_x = ctx->DrawBuffer->Width - 1;
+	max_y = ctx->DrawBuffer->Height - 1;
+
+	if ( !ctx->DrawBuffer->Name ) {
+		x1 = x;
+		y1 = ctx->DrawBuffer->Height - (y + h);
+		x2 = x + w - 1;
+		y2 = y1 + h - 1;
+	} else {
+		x1 = x;
+		y1 = y;
+		x2 = x + w - 1;
+		y2 = y + h - 1;
+
+	}
+	if (!rmesa->radeonScreen->kernel_mm) {
+	   /* Fix scissors for dri 1 */
+
+	   __DRIdrawablePrivate *dPriv = radeon_get_drawable(rmesa);
+	   x1 += dPriv->x;
+	   x2 += dPriv->x + 1;
+	   min_x += dPriv->x;
+	   max_x += dPriv->x + 1;
+	   y1 += dPriv->y;
+	   y2 += dPriv->y + 1;
+	   min_y += dPriv->y;
+	   max_y += dPriv->y + 1;
+	}
+
+	rmesa->state.scissor.rect.x1 = CLAMP(x1,  min_x, max_x);
+	rmesa->state.scissor.rect.y1 = CLAMP(y1,  min_y, max_y);
+	rmesa->state.scissor.rect.x2 = CLAMP(x2,  min_x, max_x);
+	rmesa->state.scissor.rect.y2 = CLAMP(y2,  min_y, max_y);
+
+	radeonRecalcScissorRects( rmesa );
+}
+
+/* =============================================================
+ * Scissoring
+ */
+
+void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	if (ctx->Scissor.Enabled) {
+		/* We don't pipeline cliprect changes */
+		radeon_firevertices(radeon);
+		radeonUpdateScissor(ctx);
+	}
+}
+
+void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask )
+{
+   radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+   GLuint i;
+   drm_radeon_stipple_t stipple;
+
+   /* Must flip pattern upside down.
+   */
+   for ( i = 0 ; i < 32 ; i++ ) {
+      stipple.mask[31 - i] = ((GLuint *) mask)[i];
+   }
+
+   /* TODO: push this into cmd mechanism
+   */
+   radeon_firevertices(radeon);
+   LOCK_HARDWARE( radeon );
+
+   drmCommandWrite( radeon->dri.fd, DRM_RADEON_STIPPLE,
+	 &stipple, sizeof(stipple) );
+   UNLOCK_HARDWARE( radeon );
+}
+
+
+/* ================================================================
+ * SwapBuffers with client-side throttling
+ */
+
+static uint32_t radeonGetLastFrame(radeonContextPtr radeon)
+{
+	drm_radeon_getparam_t gp;
+	int ret;
+	uint32_t frame = 0;
+
+	gp.param = RADEON_PARAM_LAST_FRAME;
+	gp.value = (int *)&frame;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
+				  &gp, sizeof(gp));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+
+	return frame;
+}
+
+uint32_t radeonGetAge(radeonContextPtr radeon)
+{
+	drm_radeon_getparam_t gp;
+	int ret;
+	uint32_t age;
+
+	gp.param = RADEON_PARAM_LAST_CLEAR;
+	gp.value = (int *)&age;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
+				  &gp, sizeof(gp));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+
+	return age;
+}
+
+static void radeonEmitIrqLocked(radeonContextPtr radeon)
+{
+	drm_radeon_irq_emit_t ie;
+	int ret;
+
+	ie.irq_seq = &radeon->iw.irq_seq;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT,
+				  &ie, sizeof(ie));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+}
+
+static void radeonWaitIrq(radeonContextPtr radeon)
+{
+	int ret;
+
+	do {
+		ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT,
+				      &radeon->iw, sizeof(radeon->iw));
+	} while (ret && (errno == EINTR || errno == EBUSY));
+
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+}
+
+static void radeonWaitForFrameCompletion(radeonContextPtr radeon)
+{
+	drm_radeon_sarea_t *sarea = radeon->sarea;
+
+	if (radeon->do_irqs) {
+		if (radeonGetLastFrame(radeon) < sarea->last_frame) {
+			if (!radeon->irqsEmitted) {
+				while (radeonGetLastFrame(radeon) <
+				       sarea->last_frame) ;
+			} else {
+				UNLOCK_HARDWARE(radeon);
+				radeonWaitIrq(radeon);
+				LOCK_HARDWARE(radeon);
+			}
+			radeon->irqsEmitted = 10;
+		}
+
+		if (radeon->irqsEmitted) {
+			radeonEmitIrqLocked(radeon);
+			radeon->irqsEmitted--;
+		}
+	} else {
+		while (radeonGetLastFrame(radeon) < sarea->last_frame) {
+			UNLOCK_HARDWARE(radeon);
+			if (radeon->do_usleeps)
+				DO_USLEEP(1);
+			LOCK_HARDWARE(radeon);
+		}
+	}
+}
+
+/* wait for idle */
+void radeonWaitForIdleLocked(radeonContextPtr radeon)
+{
+	int ret;
+	int i = 0;
+
+	do {
+		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE);
+		if (ret)
+			DO_USLEEP(1);
+	} while (ret && ++i < 100);
+
+	if (ret < 0) {
+		UNLOCK_HARDWARE(radeon);
+		fprintf(stderr, "Error: R300 timed out... exiting\n");
+		exit(-1);
+	}
+}
+
+static void radeonWaitForIdle(radeonContextPtr radeon)
+{
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+        LOCK_HARDWARE(radeon);
+	    radeonWaitForIdleLocked(radeon);
+	    UNLOCK_HARDWARE(radeon);
+    }
+}
+
+static void radeon_flip_renderbuffers(struct radeon_framebuffer *rfb)
+{
+	int current_page = rfb->pf_current_page;
+	int next_page = (current_page + 1) % rfb->pf_num_pages;
+	struct gl_renderbuffer *tmp_rb;
+
+	/* Exchange renderbuffers if necessary but make sure their
+	 * reference counts are preserved.
+	 */
+	if (rfb->color_rb[current_page] &&
+	    rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer !=
+	    &rfb->color_rb[current_page]->base) {
+		tmp_rb = NULL;
+		_mesa_reference_renderbuffer(&tmp_rb,
+					     rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+		tmp_rb = &rfb->color_rb[current_page]->base;
+		_mesa_reference_renderbuffer(&rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb);
+		_mesa_reference_renderbuffer(&tmp_rb, NULL);
+	}
+
+	if (rfb->color_rb[next_page] &&
+	    rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer !=
+	    &rfb->color_rb[next_page]->base) {
+		tmp_rb = NULL;
+		_mesa_reference_renderbuffer(&tmp_rb,
+					     rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+		tmp_rb = &rfb->color_rb[next_page]->base;
+		_mesa_reference_renderbuffer(&rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb);
+		_mesa_reference_renderbuffer(&tmp_rb, NULL);
+	}
+}
+
+/* Copy the back color buffer to the front color buffer.
+ */
+void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
+		       const drm_clip_rect_t	  *rect)
+{
+	radeonContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	GLint nbox, i, ret;
+
+	assert(dPriv);
+	assert(dPriv->driContextPriv);
+	assert(dPriv->driContextPriv->driverPrivate);
+
+	rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+
+	LOCK_HARDWARE(rmesa);
+
+	rfb = dPriv->driverPrivate;
+
+	if ( RADEON_DEBUG & RADEON_IOCTL ) {
+		fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx );
+	}
+
+	nbox = dPriv->numClipRects; /* must be in locked region */
+
+	for ( i = 0 ; i < nbox ; ) {
+		GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
+		drm_clip_rect_t *box = dPriv->pClipRects;
+		drm_clip_rect_t *b = rmesa->sarea->boxes;
+		GLint n = 0;
+
+		for ( ; i < nr ; i++ ) {
+
+			*b = box[i];
+
+			if (rect)
+			{
+				if (rect->x1 > b->x1)
+					b->x1 = rect->x1;
+				if (rect->y1 > b->y1)
+					b->y1 = rect->y1;
+				if (rect->x2 < b->x2)
+					b->x2 = rect->x2;
+				if (rect->y2 < b->y2)
+					b->y2 = rect->y2;
+
+				if (b->x1 >= b->x2 || b->y1 >= b->y2)
+					continue;
+			}
+
+			b++;
+			n++;
+		}
+		rmesa->sarea->nbox = n;
+
+		if (!n)
+			continue;
+
+		ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
+
+		if ( ret ) {
+			fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret );
+			UNLOCK_HARDWARE( rmesa );
+			exit( 1 );
+		}
+	}
+
+	UNLOCK_HARDWARE( rmesa );
+}
+
+static int radeonScheduleSwap(__DRIdrawablePrivate *dPriv, GLboolean *missed_target)
+{
+	radeonContextPtr rmesa;
+
+	rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+	radeon_firevertices(rmesa);
+
+	LOCK_HARDWARE( rmesa );
+
+	if (!dPriv->numClipRects) {
+		UNLOCK_HARDWARE(rmesa);
+		usleep(10000);	/* throttle invisible client 10ms */
+		return 0;
+	}
+
+	radeonWaitForFrameCompletion(rmesa);
+
+	UNLOCK_HARDWARE(rmesa);
+	driWaitForVBlank(dPriv, missed_target);
+
+	return 0;
+}
+
+static GLboolean radeonPageFlip( __DRIdrawablePrivate *dPriv )
+{
+	radeonContextPtr radeon;
+	GLint ret;
+	__DRIscreenPrivate *psp;
+	struct radeon_renderbuffer *rrb;
+	struct radeon_framebuffer *rfb;
+
+	assert(dPriv);
+	assert(dPriv->driContextPriv);
+	assert(dPriv->driContextPriv->driverPrivate);
+
+	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+	rfb = dPriv->driverPrivate;
+	rrb = (void *)rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+
+	psp = dPriv->driScreenPriv;
+
+	LOCK_HARDWARE(radeon);
+
+	if ( RADEON_DEBUG & RADEON_IOCTL ) {
+		fprintf(stderr, "%s: pfCurrentPage: %d %d\n", __FUNCTION__,
+			radeon->sarea->pfCurrentPage, radeon->sarea->pfState);
+	}
+	drm_clip_rect_t *box = dPriv->pClipRects;
+	drm_clip_rect_t *b = radeon->sarea->boxes;
+	b[0] = box[0];
+	radeon->sarea->nbox = 1;
+
+	ret = drmCommandNone( radeon->dri.fd, DRM_RADEON_FLIP );
+
+	UNLOCK_HARDWARE(radeon);
+
+	if ( ret ) {
+		fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
+		return GL_FALSE;
+	}
+
+	if (!rfb->pf_active)
+		return GL_FALSE;
+
+	rfb->pf_current_page = radeon->sarea->pfCurrentPage;
+	radeon_flip_renderbuffers(rfb);
+	radeon_draw_buffer(radeon->glCtx, &rfb->base);
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Swap front and back buffer.
+ */
+void radeonSwapBuffers(__DRIdrawablePrivate * dPriv)
+{
+	int64_t ust;
+	__DRIscreenPrivate *psp;
+
+	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+		radeonContextPtr radeon;
+		GLcontext *ctx;
+
+		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+		ctx = radeon->glCtx;
+
+		if (ctx->Visual.doubleBufferMode) {
+			GLboolean missed_target;
+			struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+			_mesa_notifySwapBuffers(ctx);/* flush pending rendering comands */
+
+			radeonScheduleSwap(dPriv, &missed_target);
+
+			if (rfb->pf_active) {
+				radeonPageFlip(dPriv);
+			} else {
+				radeonCopyBuffer(dPriv, NULL);
+			}
+
+			psp = dPriv->driScreenPriv;
+
+			rfb->swap_count++;
+			(*psp->systemTime->getUST)( & ust );
+			if ( missed_target ) {
+				rfb->swap_missed_count++;
+				rfb->swap_missed_ust = ust - rfb->swap_ust;
+			}
+
+			rfb->swap_ust = ust;
+			radeon->hw.all_dirty = GL_TRUE;
+		}
+	} else {
+		/* XXX this shouldn't be an error but we can't handle it for now */
+		_mesa_problem(NULL, "%s: drawable has no context!",
+			      __FUNCTION__);
+	}
+}
+
+void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
+			 int x, int y, int w, int h )
+{
+	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+		radeonContextPtr radeon;
+		GLcontext *ctx;
+
+		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+		ctx = radeon->glCtx;
+
+		if (ctx->Visual.doubleBufferMode) {
+			drm_clip_rect_t rect;
+			rect.x1 = x + dPriv->x;
+			rect.y1 = (dPriv->h - y - h) + dPriv->y;
+			rect.x2 = rect.x1 + w;
+			rect.y2 = rect.y1 + h;
+			_mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
+			radeonCopyBuffer(dPriv, &rect);
+		}
+	} else {
+		/* XXX this shouldn't be an error but we can't handle it for now */
+		_mesa_problem(NULL, "%s: drawable has no context!",
+			      __FUNCTION__);
+	}
+}
+
+void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_renderbuffer *rrbDepth = NULL, *rrbStencil = NULL,
+		*rrbColor = NULL;
+	uint32_t offset = 0;
+
+
+	if (!fb) {
+		/* this can happen during the initial context initialization */
+		return;
+	}
+
+	/* radeons only handle 1 color draw so far */
+	if (fb->_NumColorDrawBuffers != 1) {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE);
+		return;
+	}
+
+	/* Do this here, note core Mesa, since this function is called from
+	 * many places within the driver.
+	 */
+	if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+		/* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
+		_mesa_update_framebuffer(ctx);
+		/* this updates the DrawBuffer's Width/Height if it's a FBO */
+		_mesa_update_draw_buffer_bounds(ctx);
+	}
+
+	if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+		/* this may occur when we're called by glBindFrameBuffer() during
+		 * the process of someone setting up renderbuffers, etc.
+		 */
+		/*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/
+		return;
+	}
+
+	if (fb->Name)
+		;/* do something depthy/stencily TODO */
+
+
+		/* none */
+	if (fb->Name == 0) {
+		if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+			radeon->front_cliprects = GL_TRUE;
+			radeon->front_buffer_dirty = GL_TRUE;
+		} else {
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+			radeon->front_cliprects = GL_FALSE;
+		}
+	} else {
+		/* user FBO in theory */
+		struct radeon_renderbuffer *rrb;
+		rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[0]);
+		if (rrb) {
+			offset = rrb->draw_offset;
+			rrbColor = rrb;
+		}
+		radeon->constant_cliprect = GL_TRUE;
+	}
+
+	if (rrbColor == NULL)
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE);
+	else
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_FALSE);
+
+
+	if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) {
+		rrbDepth = radeon_renderbuffer(fb->_DepthBuffer->Wrapped);
+		if (rrbDepth && rrbDepth->bo) {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_FALSE);
+		} else {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_TRUE);
+		}
+	} else {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_FALSE);
+		rrbDepth = NULL;
+	}
+
+	if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) {
+		rrbStencil = radeon_renderbuffer(fb->_StencilBuffer->Wrapped);
+		if (rrbStencil && rrbStencil->bo) {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_FALSE);
+			/* need to re-compute stencil hw state */
+			if (!rrbDepth)
+				rrbDepth = rrbStencil;
+		} else {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_TRUE);
+		}
+	} else {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_FALSE);
+		if (ctx->Driver.Enable != NULL)
+			ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
+		else
+			ctx->NewState |= _NEW_STENCIL;
+	}
+
+	/* Update culling direction which changes depending on the
+	 * orientation of the buffer:
+	 */
+	if (ctx->Driver.FrontFace)
+		ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+	else
+		ctx->NewState |= _NEW_POLYGON;
+
+	/*
+	 * Update depth test state
+	 */
+	if (ctx->Driver.Enable) {
+		ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
+				   (ctx->Depth.Test && fb->Visual.depthBits > 0));
+		/* Need to update the derived ctx->Stencil._Enabled first */
+		ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
+				   (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0));
+	} else {
+		ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
+	}
+
+	_mesa_reference_renderbuffer(&radeon->state.depth.rb, &rrbDepth->base);
+	_mesa_reference_renderbuffer(&radeon->state.color.rb, &rrbColor->base);
+	radeon->state.color.draw_offset = offset;
+
+#if 0
+	/* update viewport since it depends on window size */
+	if (ctx->Driver.Viewport) {
+		ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
+				     ctx->Viewport.Width, ctx->Viewport.Height);
+	} else {
+
+	}
+#endif
+	ctx->NewState |= _NEW_VIEWPORT;
+
+	/* Set state we know depends on drawable parameters:
+	 */
+	radeonUpdateScissor(ctx);
+	radeon->NewGLState |= _NEW_SCISSOR;
+
+	if (ctx->Driver.DepthRange)
+		ctx->Driver.DepthRange(ctx,
+				       ctx->Viewport.Near,
+				       ctx->Viewport.Far);
+
+	/* Update culling direction which changes depending on the
+	 * orientation of the buffer:
+	 */
+	if (ctx->Driver.FrontFace)
+		ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+	else
+		ctx->NewState |= _NEW_POLYGON;
+}
+
+/**
+ * Called via glDrawBuffer.
+ */
+void radeonDrawBuffer( GLcontext *ctx, GLenum mode )
+{
+	if (RADEON_DEBUG & RADEON_DRI)
+		fprintf(stderr, "%s %s\n", __FUNCTION__,
+			_mesa_lookup_enum_by_nr( mode ));
+
+	if (ctx->DrawBuffer->Name == 0) {
+		radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+		const GLboolean was_front_buffer_rendering =
+			radeon->is_front_buffer_rendering;
+
+		radeon->is_front_buffer_rendering = (mode == GL_FRONT_LEFT) ||
+                                            (mode == GL_FRONT);
+
+      /* If we weren't front-buffer rendering before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+		if (!was_front_buffer_rendering && radeon->is_front_buffer_rendering) {
+			radeon_update_renderbuffers(radeon->dri.context,
+				radeon->dri.context->driDrawablePriv);
+      }
+	}
+
+	radeon_draw_buffer(ctx, ctx->DrawBuffer);
+}
+
+void radeonReadBuffer( GLcontext *ctx, GLenum mode )
+{
+	if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
+		struct radeon_context *const rmesa = RADEON_CONTEXT(ctx);
+		const GLboolean was_front_buffer_reading = rmesa->is_front_buffer_reading;
+		rmesa->is_front_buffer_reading = (mode == GL_FRONT_LEFT)
+					|| (mode == GL_FRONT);
+
+		if (!was_front_buffer_reading && rmesa->is_front_buffer_reading) {
+			radeon_update_renderbuffers(rmesa->dri.context,
+						    rmesa->dri.context->driReadablePriv);
+	 	}
+	}
+	/* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
+	if (ctx->ReadBuffer == ctx->DrawBuffer) {
+		/* This will update FBO completeness status.
+		 * A framebuffer will be incomplete if the GL_READ_BUFFER setting
+		 * refers to a missing renderbuffer.  Calling glReadBuffer can set
+		 * that straight and can make the drawing buffer complete.
+		 */
+		radeon_draw_buffer(ctx, ctx->DrawBuffer);
+	}
+}
+
+
+/* Turn on/off page flipping according to the flags in the sarea:
+ */
+void radeonUpdatePageFlipping(radeonContextPtr radeon)
+{
+	struct radeon_framebuffer *rfb = radeon_get_drawable(radeon)->driverPrivate;
+
+	rfb->pf_active = radeon->sarea->pfState;
+	rfb->pf_current_page = radeon->sarea->pfCurrentPage;
+	rfb->pf_num_pages = 2;
+	radeon_flip_renderbuffers(rfb);
+	radeon_draw_buffer(radeon->glCtx, radeon->glCtx->DrawBuffer);
+}
+
+void radeon_window_moved(radeonContextPtr radeon)
+{
+	/* Cliprects has to be updated before doing anything else */
+	radeonSetCliprects(radeon);
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		radeonUpdatePageFlipping(radeon);
+	}
+}
+
+void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	__DRIcontext *driContext = radeon->dri.context;
+	void (*old_viewport)(GLcontext *ctx, GLint x, GLint y,
+			     GLsizei w, GLsizei h);
+
+	if (!driContext->driScreenPriv->dri2.enabled)
+		return;
+
+	if (!radeon->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+		if (radeon->is_front_buffer_rendering) {
+			ctx->Driver.Flush(ctx);
+		}
+		radeon_update_renderbuffers(driContext, driContext->driDrawablePriv);
+		if (driContext->driDrawablePriv != driContext->driReadablePriv)
+			radeon_update_renderbuffers(driContext, driContext->driReadablePriv);
+	}
+
+	old_viewport = ctx->Driver.Viewport;
+	ctx->Driver.Viewport = NULL;
+	radeon_window_moved(radeon);
+	radeon_draw_buffer(ctx, radeon->glCtx->DrawBuffer);
+	ctx->Driver.Viewport = old_viewport;
+}
+
+static void radeon_print_state_atom_prekmm(radeonContextPtr radeon, struct radeon_state_atom *state)
+{
+	int i, j, reg;
+	int dwords = (*state->check) (radeon->glCtx, state);
+	drm_r300_cmd_header_t cmd;
+
+	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords, state->cmd_size);
+
+	if (radeon_is_debug_enabled(RADEON_STATE, RADEON_TRACE)) {
+		if (dwords > state->cmd_size)
+			dwords = state->cmd_size;
+
+		for (i = 0; i < dwords;) {
+			cmd = *((drm_r300_cmd_header_t *) &state->cmd[i]);
+			reg = (cmd.packet0.reghi << 8) | cmd.packet0.reglo;
+			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
+					state->name, i, reg, cmd.packet0.count);
+			++i;
+			for (j = 0; j < cmd.packet0.count && i < dwords; j++) {
+				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
+						state->name, i, reg, state->cmd[i]);
+				reg += 4;
+				++i;
+			}
+		}
+	}
+}
+
+static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state_atom *state)
+{
+	int i, j, reg, count;
+	int dwords;
+	uint32_t packet0;
+	if (!radeon_is_debug_enabled(RADEON_STATE, RADEON_VERBOSE) )
+		return;
+
+	if (!radeon->radeonScreen->kernel_mm) {
+		radeon_print_state_atom_prekmm(radeon, state);
+		return;
+	}
+
+	dwords = (*state->check) (radeon->glCtx, state);
+
+	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords, state->cmd_size);
+
+	if (radeon_is_debug_enabled(RADEON_STATE, RADEON_TRACE)) {
+		if (dwords > state->cmd_size)
+			dwords = state->cmd_size;
+		for (i = 0; i < dwords;) {
+			packet0 = state->cmd[i];
+			reg = (packet0 & 0x1FFF) << 2;
+			count = ((packet0 & 0x3FFF0000) >> 16) + 1;
+			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
+					state->name, i, reg, count);
+			++i;
+			for (j = 0; j < count && i < dwords; j++) {
+				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
+						state->name, i, reg, state->cmd[i]);
+				reg += 4;
+				++i;
+			}
+		}
+	}
+}
+
+/**
+ * Count total size for next state emit.
+ **/
+GLuint radeonCountStateEmitSize(radeonContextPtr radeon)
+{
+	struct radeon_state_atom *atom;
+	GLuint dwords = 0;
+	/* check if we are going to emit full state */
+
+	if (radeon->cmdbuf.cs->cdw && !radeon->hw.all_dirty) {
+		if (!radeon->hw.is_dirty)
+			goto out;
+		foreach(atom, &radeon->hw.atomlist) {
+			if (atom->dirty) {
+				const GLuint atom_size = atom->check(radeon->glCtx, atom);
+				dwords += atom_size;
+				if (RADEON_CMDBUF && atom_size) {
+					radeon_print_state_atom(radeon, atom);
+				}
+			}
+		}
+	} else {
+		foreach(atom, &radeon->hw.atomlist) {
+			const GLuint atom_size = atom->check(radeon->glCtx, atom);
+			dwords += atom_size;
+			if (RADEON_CMDBUF && atom_size) {
+				radeon_print_state_atom(radeon, atom);
+			}
+
+		}
+	}
+out:
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s %u\n", __func__, dwords);
+	return dwords;
+}
+
+static INLINE void radeon_emit_atom(radeonContextPtr radeon, struct radeon_state_atom *atom)
+{
+	BATCH_LOCALS(radeon);
+	int dwords;
+
+	dwords = (*atom->check) (radeon->glCtx, atom);
+	if (dwords) {
+
+		radeon_print_state_atom(radeon, atom);
+
+		if (atom->emit) {
+			(*atom->emit)(radeon->glCtx, atom);
+		} else {
+			BEGIN_BATCH_NO_AUTOSTATE(dwords);
+			OUT_BATCH_TABLE(atom->cmd, dwords);
+			END_BATCH();
+		}
+	} else {
+		radeon_print(RADEON_STATE, RADEON_VERBOSE, "  skip state %s\n", atom->name);
+	}
+	atom->dirty = GL_FALSE;
+
+}
+
+static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean emitAll)
+{
+	struct radeon_state_atom *atom;
+
+	if (radeon->vtbl.pre_emit_atoms)
+		radeon->vtbl.pre_emit_atoms(radeon);
+
+	/* Emit actual atoms */
+	if (radeon->hw.all_dirty || emitAll) {
+		foreach(atom, &radeon->hw.atomlist)
+			radeon_emit_atom( radeon, atom );
+	} else {
+		foreach(atom, &radeon->hw.atomlist) {
+			if ( atom->dirty )
+				radeon_emit_atom( radeon, atom );
+		}
+	}
+
+	COMMIT_BATCH();
+}
+
+static GLboolean radeon_revalidate_bos(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	int ret;
+
+	ret = radeon_cs_space_check(radeon->cmdbuf.cs);
+	if (ret == RADEON_CS_SPACE_FLUSH)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void radeonEmitState(radeonContextPtr radeon)
+{
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s\n", __FUNCTION__);
+
+	if (radeon->vtbl.pre_emit_state)
+		radeon->vtbl.pre_emit_state(radeon);
+
+	/* this code used to return here but now it emits zbs */
+	if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty)
+		return;
+
+	if (!radeon->cmdbuf.cs->cdw) {
+		if (RADEON_DEBUG & RADEON_STATE)
+			fprintf(stderr, "Begin reemit state\n");
+
+		radeonEmitAtoms(radeon, GL_TRUE);
+	} else {
+
+		if (RADEON_DEBUG & RADEON_STATE)
+			fprintf(stderr, "Begin dirty state\n");
+
+		radeonEmitAtoms(radeon, GL_FALSE);
+	}
+
+	radeon->hw.is_dirty = GL_FALSE;
+	radeon->hw.all_dirty = GL_FALSE;
+}
+
+
+void radeonFlush(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, radeon->cmdbuf.cs->cdw);
+
+	/* okay if we have no cmds in the buffer &&
+	   we have no DMA flush &&
+	   we have no DMA buffer allocated.
+	   then no point flushing anything at all.
+	*/
+	if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && is_empty_list(&radeon->dma.reserved))
+		return;
+
+	if (radeon->dma.flush)
+		radeon->dma.flush( ctx );
+
+	radeonEmitState(radeon);
+
+	if (radeon->cmdbuf.cs->cdw)
+		rcommonFlushCmdBuf(radeon, __FUNCTION__);
+
+	if ((ctx->DrawBuffer->Name == 0) && radeon->front_buffer_dirty) {
+		__DRIscreen *const screen = radeon->radeonScreen->driScreen;
+
+		if (screen->dri2.loader && (screen->dri2.loader->base.version >= 2)
+			&& (screen->dri2.loader->flushFrontBuffer != NULL)) {
+			__DRIdrawablePrivate * drawable = radeon_get_drawable(radeon);
+			(*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate);
+
+			/* Only clear the dirty bit if front-buffer rendering is no longer
+			 * enabled.  This is done so that the dirty bit can only be set in
+			 * glDrawBuffer.  Otherwise the dirty bit would have to be set at
+			 * each of N places that do rendering.  This has worse performances,
+			 * but it is much easier to get correct.
+			 */
+			if (!radeon->is_front_buffer_rendering) {
+				radeon->front_buffer_dirty = GL_FALSE;
+			}
+		}
+	}
+
+	make_empty_list(&radeon->query.not_flushed_head);
+
+}
+
+/* Make sure all commands have been sent to the hardware and have
+ * completed processing.
+ */
+void radeonFinish(GLcontext * ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
+	int i;
+
+	if (ctx->Driver.Flush)
+		ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+	if (radeon->radeonScreen->kernel_mm) {
+		for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
+			struct radeon_renderbuffer *rrb;
+			rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[i]);
+			if (rrb && rrb->bo)
+				radeon_bo_wait(rrb->bo);
+		}
+		{
+			struct radeon_renderbuffer *rrb;
+			rrb = radeon_get_depthbuffer(radeon);
+			if (rrb && rrb->bo)
+				radeon_bo_wait(rrb->bo);
+		}
+	} else if (radeon->do_irqs) {
+		LOCK_HARDWARE(radeon);
+		radeonEmitIrqLocked(radeon);
+		UNLOCK_HARDWARE(radeon);
+		radeonWaitIrq(radeon);
+	} else {
+		radeonWaitForIdle(radeon);
+	}
+}
+
+/* cmdbuffer */
+/**
+ * Send the current command buffer via ioctl to the hardware.
+ */
+int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller)
+{
+	int ret = 0;
+
+	if (rmesa->cmdbuf.flushing) {
+		fprintf(stderr, "Recursive call into r300FlushCmdBufLocked!\n");
+		exit(-1);
+	}
+	rmesa->cmdbuf.flushing = 1;
+
+	if (RADEON_DEBUG & RADEON_IOCTL) {
+		fprintf(stderr, "%s from %s - %i cliprects\n",
+			__FUNCTION__, caller, rmesa->numClipRects);
+	}
+
+	radeonEmitQueryEnd(rmesa->glCtx);
+
+	if (rmesa->cmdbuf.cs->cdw) {
+		ret = radeon_cs_emit(rmesa->cmdbuf.cs);
+		rmesa->hw.all_dirty = GL_TRUE;
+	}
+	radeon_cs_erase(rmesa->cmdbuf.cs);
+	rmesa->cmdbuf.flushing = 0;
+
+	if (radeon_revalidate_bos(rmesa->glCtx) == GL_FALSE) {
+		fprintf(stderr,"failed to revalidate buffers\n");
+	}
+
+	return ret;
+}
+
+int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller)
+{
+	int ret;
+
+	radeonReleaseDmaRegions(rmesa);
+
+	LOCK_HARDWARE(rmesa);
+	ret = rcommonFlushCmdBufLocked(rmesa, caller);
+	UNLOCK_HARDWARE(rmesa);
+
+	if (ret) {
+		fprintf(stderr, "drmRadeonCmdBuffer: %d. Kernel failed to "
+				"parse or rejected command stream. See dmesg "
+				"for more info.\n", ret);
+		_mesa_exit(ret);
+	}
+
+	return ret;
+}
+
+/**
+ * Make sure that enough space is available in the command buffer
+ * by flushing if necessary.
+ *
+ * \param dwords The number of dwords we need to be free on the command buffer
+ */
+GLboolean rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller)
+{
+   if ((rmesa->cmdbuf.cs->cdw + dwords + 128) > rmesa->cmdbuf.size
+	 || radeon_cs_need_flush(rmesa->cmdbuf.cs)) {
+      /* If we try to flush empty buffer there is too big rendering operation. */
+      assert(rmesa->cmdbuf.cs->cdw);
+      rcommonFlushCmdBuf(rmesa, caller);
+      return GL_TRUE;
+   }
+   return GL_FALSE;
+}
+
+void rcommonInitCmdBuf(radeonContextPtr rmesa)
+{
+	GLuint size;
+	/* Initialize command buffer */
+	size = 256 * driQueryOptioni(&rmesa->optionCache,
+				     "command_buffer_size");
+	if (size < 2 * rmesa->hw.max_state_size) {
+		size = 2 * rmesa->hw.max_state_size + 65535;
+	}
+	if (size > 64 * 256)
+		size = 64 * 256;
+
+	radeon_print(RADEON_CS, RADEON_VERBOSE,
+			"sizeof(drm_r300_cmd_header_t)=%zd\n", sizeof(drm_r300_cmd_header_t));
+	radeon_print(RADEON_CS, RADEON_VERBOSE,
+			"sizeof(drm_radeon_cmd_buffer_t)=%zd\n", sizeof(drm_radeon_cmd_buffer_t));
+	radeon_print(RADEON_CS, RADEON_VERBOSE,
+			"Allocating %d bytes command buffer (max state is %d bytes)\n",
+			size * 4, rmesa->hw.max_state_size * 4);
+
+	if (rmesa->radeonScreen->kernel_mm) {
+		int fd = rmesa->radeonScreen->driScreen->fd;
+		rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd);
+	} else {
+		rmesa->cmdbuf.csm = radeon_cs_manager_legacy_ctor(rmesa);
+	}
+	if (rmesa->cmdbuf.csm == NULL) {
+		/* FIXME: fatal error */
+		return;
+	}
+	rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size);
+	assert(rmesa->cmdbuf.cs != NULL);
+	rmesa->cmdbuf.size = size;
+
+	radeon_cs_space_set_flush(rmesa->cmdbuf.cs,
+				  (void (*)(void *))rmesa->glCtx->Driver.Flush, rmesa->glCtx);
+
+	if (!rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]);
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size);
+	} else {
+		struct drm_radeon_gem_info mminfo = { 0 };
+
+		if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo)))
+		{
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_visible);
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size);
+		}
+	}
+
+}
+/**
+ * Destroy the command buffer
+ */
+void rcommonDestroyCmdBuf(radeonContextPtr rmesa)
+{
+	radeon_cs_destroy(rmesa->cmdbuf.cs);
+	if (rmesa->radeonScreen->driScreen->dri2.enabled || rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_manager_gem_dtor(rmesa->cmdbuf.csm);
+	} else {
+		radeon_cs_manager_legacy_dtor(rmesa->cmdbuf.csm);
+	}
+}
+
+void rcommonBeginBatch(radeonContextPtr rmesa, int n,
+		       int dostate,
+		       const char *file,
+		       const char *function,
+		       int line)
+{
+	if (!rmesa->cmdbuf.cs->cdw && dostate) {
+		radeon_print(RADEON_STATE, RADEON_NORMAL,
+				"Reemit state after flush (from %s)\n", function);
+		radeonEmitState(rmesa);
+	}
+	radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line);
+
+    radeon_print(RADEON_CS, RADEON_VERBOSE, "BEGIN_BATCH(%d) at %d, from %s:%i\n",
+                        n, rmesa->cmdbuf.cs->cdw, function, line);
+
+}
+
+void radeonUserClear(GLcontext *ctx, GLuint mask)
+{
+   _mesa_meta_clear(ctx, mask);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
new file mode 100644
index 0000000000..f3201911ac
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -0,0 +1,87 @@
+#ifndef COMMON_MISC_H
+#define COMMON_MISC_H
+
+#include "radeon_common_context.h"
+#include "radeon_dma.h"
+#include "radeon_texture.h"
+
+void radeonUserClear(GLcontext *ctx, GLuint mask);
+void radeonRecalcScissorRects(radeonContextPtr radeon);
+void radeonSetCliprects(radeonContextPtr radeon);
+void radeonUpdateScissor( GLcontext *ctx );
+void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h);
+void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask );
+
+void radeonWaitForIdleLocked(radeonContextPtr radeon);
+extern uint32_t radeonGetAge(radeonContextPtr radeon);
+void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
+		       const drm_clip_rect_t	  *rect);
+void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
+void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
+			 int x, int y, int w, int h );
+
+void radeonUpdatePageFlipping(radeonContextPtr rmesa);
+
+void radeonFlush(GLcontext *ctx);
+void radeonFinish(GLcontext * ctx);
+void radeonEmitState(radeonContextPtr radeon);
+GLuint radeonCountStateEmitSize(radeonContextPtr radeon);
+
+void radeon_clear_tris(GLcontext *ctx, GLbitfield mask);
+
+void radeon_window_moved(radeonContextPtr radeon);
+void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb);
+void radeonDrawBuffer( GLcontext *ctx, GLenum mode );
+void radeonReadBuffer( GLcontext *ctx, GLenum mode );
+void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height);
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off);
+void radeon_fbo_init(struct radeon_context *radeon);
+void
+radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+			   struct radeon_bo *bo);
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv);
+static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb)
+{
+	struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb;
+	if (rrb && rrb->base.ClassID == RADEON_RB_CLASS)
+		return rrb;
+	else
+		return NULL;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_renderbuffer(struct gl_framebuffer *fb, int att_index)
+{
+	if (att_index >= 0)
+		return radeon_renderbuffer(fb->Attachment[att_index].Renderbuffer);
+	else
+		return NULL;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPtr rmesa)
+{
+	struct radeon_renderbuffer *rrb;
+	rrb = radeon_renderbuffer(rmesa->state.depth.rb);
+	if (!rrb)
+		return NULL;
+
+	return rrb;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_colorbuffer(radeonContextPtr rmesa)
+{
+	struct radeon_renderbuffer *rrb;
+
+	rrb = radeon_renderbuffer(rmesa->state.color.rb);
+	if (!rrb)
+		return NULL;
+	return rrb;
+}
+
+#include "radeon_cmdbuf.h"
+
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
new file mode 100644
index 0000000000..71ee06d9a7
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -0,0 +1,805 @@
+/**************************************************************************
+
+Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
+                     VA Linux Systems Inc., Fremont, California.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#include "radeon_common.h"
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+#include "utils.h"
+#include "vblank.h"
+#include "drirenderbuffer.h"
+#include "drivers/common/meta.h"
+#include "main/context.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/state.h"
+#include "main/simple_list.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/tnl.h"
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */
+#include "r600_context.h"
+#endif
+
+#define DRIVER_DATE "20090101"
+
+#ifndef RADEON_DEBUG
+int RADEON_DEBUG = (0);
+#endif
+
+
+static const char* get_chip_family_name(int chip_family)
+{
+	switch(chip_family) {
+	case CHIP_FAMILY_R100: return "R100";
+	case CHIP_FAMILY_RV100: return "RV100";
+	case CHIP_FAMILY_RS100: return "RS100";
+	case CHIP_FAMILY_RV200: return "RV200";
+	case CHIP_FAMILY_RS200: return "RS200";
+	case CHIP_FAMILY_R200: return "R200";
+	case CHIP_FAMILY_RV250: return "RV250";
+	case CHIP_FAMILY_RS300: return "RS300";
+	case CHIP_FAMILY_RV280: return "RV280";
+	case CHIP_FAMILY_R300: return "R300";
+	case CHIP_FAMILY_R350: return "R350";
+	case CHIP_FAMILY_RV350: return "RV350";
+	case CHIP_FAMILY_RV380: return "RV380";
+	case CHIP_FAMILY_R420: return "R420";
+	case CHIP_FAMILY_RV410: return "RV410";
+	case CHIP_FAMILY_RS400: return "RS400";
+	case CHIP_FAMILY_RS600: return "RS600";
+	case CHIP_FAMILY_RS690: return "RS690";
+	case CHIP_FAMILY_RS740: return "RS740";
+	case CHIP_FAMILY_RV515: return "RV515";
+	case CHIP_FAMILY_R520: return "R520";
+	case CHIP_FAMILY_RV530: return "RV530";
+	case CHIP_FAMILY_R580: return "R580";
+	case CHIP_FAMILY_RV560: return "RV560";
+	case CHIP_FAMILY_RV570: return "RV570";
+	case CHIP_FAMILY_R600: return "R600";
+	case CHIP_FAMILY_RV610: return "RV610";
+	case CHIP_FAMILY_RV630: return "RV630";
+	case CHIP_FAMILY_RV670: return "RV670";
+	case CHIP_FAMILY_RV620: return "RV620";
+	case CHIP_FAMILY_RV635: return "RV635";
+	case CHIP_FAMILY_RS780: return "RS780";
+	case CHIP_FAMILY_RS880: return "RS880";
+	case CHIP_FAMILY_RV770: return "RV770";
+	case CHIP_FAMILY_RV730: return "RV730";
+	case CHIP_FAMILY_RV710: return "RV710";
+	case CHIP_FAMILY_RV740: return "RV740";
+	default: return "unknown";
+	}
+}
+
+
+/* Return various strings for glGetString().
+ */
+static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	static char buffer[128];
+
+	switch (name) {
+	case GL_VENDOR:
+		if (IS_R600_CLASS(radeon->radeonScreen))
+			return (GLubyte *) "Advanced Micro Devices, Inc.";
+		else if (IS_R300_CLASS(radeon->radeonScreen))
+			return (GLubyte *) "DRI R300 Project";
+		else
+			return (GLubyte *) "Tungsten Graphics, Inc.";
+
+	case GL_RENDERER:
+	{
+		unsigned offset;
+		GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
+			radeon->radeonScreen->AGPMode;
+		const char* chipclass;
+		char hardwarename[32];
+
+		if (IS_R600_CLASS(radeon->radeonScreen))
+			chipclass = "R600";
+		else if (IS_R300_CLASS(radeon->radeonScreen))
+			chipclass = "R300";
+		else if (IS_R200_CLASS(radeon->radeonScreen))
+			chipclass = "R200";
+		else
+			chipclass = "R100";
+
+		sprintf(hardwarename, "%s (%s %04X)",
+		        chipclass,
+		        get_chip_family_name(radeon->radeonScreen->chip_family),
+		        radeon->radeonScreen->device_id);
+
+		offset = driGetRendererString(buffer, hardwarename, DRIVER_DATE,
+					      agp_mode);
+
+		if (IS_R600_CLASS(radeon->radeonScreen)) {
+			sprintf(&buffer[offset], " TCL");
+		} else if (IS_R300_CLASS(radeon->radeonScreen)) {
+			sprintf(&buffer[offset], " %sTCL",
+				(radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)
+				? "" : "NO-");
+		} else {
+			sprintf(&buffer[offset], " %sTCL",
+				!(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
+				? "" : "NO-");
+		}
+
+		if (radeon->radeonScreen->driScreen->dri2.enabled)
+			strcat(buffer, " DRI2");
+
+		return (GLubyte *) buffer;
+	}
+
+	default:
+		return NULL;
+	}
+}
+
+/* Initialize the driver's misc functions.
+ */
+static void radeonInitDriverFuncs(struct dd_function_table *functions)
+{
+	functions->GetString = radeonGetString;
+}
+
+/**
+ * Create and initialize all common fields of the context,
+ * including the Mesa context itself.
+ */
+GLboolean radeonInitContext(radeonContextPtr radeon,
+			    struct dd_function_table* functions,
+			    const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+	GLcontext* ctx;
+	GLcontext* shareCtx;
+	int fthrottle_mode;
+
+	/* Fill in additional standard functions. */
+	radeonInitDriverFuncs(functions);
+
+	radeon->radeonScreen = screen;
+	/* Allocate and initialize the Mesa context */
+	if (sharedContextPrivate)
+		shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx;
+	else
+		shareCtx = NULL;
+	radeon->glCtx = _mesa_create_context(glVisual, shareCtx,
+					    functions, (void *)radeon);
+	if (!radeon->glCtx)
+		return GL_FALSE;
+
+	ctx = radeon->glCtx;
+	driContextPriv->driverPrivate = radeon;
+
+	meta_init_metaops(ctx, &radeon->meta);
+
+	_mesa_meta_init(ctx);
+
+	/* DRI fields */
+	radeon->dri.context = driContextPriv;
+	radeon->dri.screen = sPriv;
+	radeon->dri.hwContext = driContextPriv->hHWContext;
+	radeon->dri.hwLock = &sPriv->pSAREA->lock;
+	radeon->dri.hwLockCount = 0;
+	radeon->dri.fd = sPriv->fd;
+	radeon->dri.drmMinor = sPriv->drm_version.minor;
+
+	radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
+					       screen->sarea_priv_offset);
+
+	/* Setup IRQs */
+	fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode");
+	radeon->iw.irq_seq = -1;
+	radeon->irqsEmitted = 0;
+	if (IS_R600_CLASS(radeon->radeonScreen))
+		radeon->do_irqs = 0;
+	else
+		radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
+				   radeon->radeonScreen->irq);
+
+	radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
+
+	if (!radeon->do_irqs)
+		fprintf(stderr,
+			"IRQ's not enabled, falling back to %s: %d %d\n",
+			radeon->do_usleeps ? "usleeps" : "busy waits",
+			fthrottle_mode, radeon->radeonScreen->irq);
+
+        radeon->texture_depth = driQueryOptioni (&radeon->optionCache,
+					        "texture_depth");
+        if (radeon->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
+                radeon->texture_depth = ( glVisual->rgbBits > 16 ) ?
+	        DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
+
+	if (IS_R600_CLASS(radeon->radeonScreen)) {
+		radeon->texture_row_align = 256;
+		radeon->texture_rect_row_align = 256;
+		radeon->texture_compressed_row_align = 256;
+	} else if (IS_R200_CLASS(radeon->radeonScreen) ||
+		   IS_R100_CLASS(radeon->radeonScreen)) {
+		radeon->texture_row_align = 32;
+		radeon->texture_rect_row_align = 64;
+		radeon->texture_compressed_row_align = 32;
+	} else { /* R300 - not sure this is all correct */
+		int chip_family = radeon->radeonScreen->chip_family;
+		if (chip_family == CHIP_FAMILY_RS600 ||
+		    chip_family == CHIP_FAMILY_RS690 ||
+		    chip_family == CHIP_FAMILY_RS740)
+			radeon->texture_row_align = 64;
+		else
+			radeon->texture_row_align = 32;
+		radeon->texture_rect_row_align = 64;
+		radeon->texture_compressed_row_align = 64;
+	}
+
+	make_empty_list(&radeon->query.not_flushed_head);
+	radeon_init_dma(radeon);
+
+	return GL_TRUE;
+}
+
+
+
+/**
+ * Destroy the command buffer and state atoms.
+ */
+static void radeon_destroy_atom_list(radeonContextPtr radeon)
+{
+	struct radeon_state_atom *atom;
+
+	foreach(atom, &radeon->hw.atomlist) {
+		FREE(atom->cmd);
+		if (atom->lastcmd)
+			FREE(atom->lastcmd);
+	}
+
+}
+
+/**
+ * Cleanup common context fields.
+ * Called by r200DestroyContext/r300DestroyContext
+ */
+void radeonDestroyContext(__DRIcontextPrivate *driContextPriv )
+{
+#ifdef RADEON_BO_TRACK
+	FILE *track;
+#endif
+	GET_CURRENT_CONTEXT(ctx);
+	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+	radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
+
+	assert(radeon);
+
+	_mesa_meta_free(radeon->glCtx);
+
+	if (radeon == current) {
+		radeon_firevertices(radeon);
+		_mesa_make_current(NULL, NULL, NULL);
+	}
+
+	if (!is_empty_list(&radeon->dma.reserved)) {
+		rcommonFlushCmdBuf( radeon, __FUNCTION__ );
+	}
+
+	radeonFreeDmaRegions(radeon);
+	radeonReleaseArrays(radeon->glCtx, ~0);
+	meta_destroy_metaops(&radeon->meta);
+	if (radeon->vtbl.free_context)
+		radeon->vtbl.free_context(radeon->glCtx);
+	_swsetup_DestroyContext( radeon->glCtx );
+	_tnl_DestroyContext( radeon->glCtx );
+	_vbo_DestroyContext( radeon->glCtx );
+	_swrast_DestroyContext( radeon->glCtx );
+
+	/* free atom list */
+	/* free the Mesa context */
+	_mesa_destroy_context(radeon->glCtx);
+
+	/* _mesa_destroy_context() might result in calls to functions that
+	 * depend on the DriverCtx, so don't set it to NULL before.
+	 *
+	 * radeon->glCtx->DriverCtx = NULL;
+	 */
+	/* free the option cache */
+	driDestroyOptionCache(&radeon->optionCache);
+
+	rcommonDestroyCmdBuf(radeon);
+
+	radeon_destroy_atom_list(radeon);
+
+	if (radeon->state.scissor.pClipRects) {
+		FREE(radeon->state.scissor.pClipRects);
+		radeon->state.scissor.pClipRects = 0;
+	}
+#ifdef RADEON_BO_TRACK
+	track = fopen("/tmp/tracklog", "w");
+	if (track) {
+		radeon_tracker_print(&radeon->radeonScreen->bom->tracker, track);
+		fclose(track);
+	}
+#endif
+	FREE(radeon);
+}
+
+/* Force the context `c' to be unbound from its buffer.
+ */
+GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv)
+{
+	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+
+	if (RADEON_DEBUG & RADEON_DRI)
+		fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
+			radeon->glCtx);
+
+	return GL_TRUE;
+}
+
+
+static void
+radeon_make_kernel_renderbuffer_current(radeonContextPtr radeon,
+					struct radeon_framebuffer *draw)
+{
+	/* if radeon->fake */
+	struct radeon_renderbuffer *rb;
+
+	if ((rb = (void *)draw->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->backPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_DEPTH].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_STENCIL].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+}
+
+static void
+radeon_make_renderbuffer_current(radeonContextPtr radeon,
+				 struct radeon_framebuffer *draw)
+{
+	int size = 4096*4096*4;
+	/* if radeon->fake */
+	struct radeon_renderbuffer *rb;
+
+	if (radeon->radeonScreen->kernel_mm) {
+		radeon_make_kernel_renderbuffer_current(radeon, draw);
+		return;
+	}
+
+
+	if ((rb = (void *)draw->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->backPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_DEPTH].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_STENCIL].Renderbuffer)) {
+		if (!rb->bo) {
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+}
+
+static unsigned
+radeon_bits_per_pixel(const struct radeon_renderbuffer *rb)
+{
+   switch (rb->base._ActualFormat) {
+   case GL_RGB5:
+   case GL_DEPTH_COMPONENT16:
+      return 16;
+   case GL_RGB8:
+   case GL_RGBA8:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH24_STENCIL8_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+      return 32;
+   default:
+      return 0;
+   }
+}
+
+void
+radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
+{
+	unsigned int attachments[10];
+	__DRIbuffer *buffers = NULL;
+	__DRIscreen *screen;
+	struct radeon_renderbuffer *rb;
+	int i, count;
+	struct radeon_framebuffer *draw;
+	radeonContextPtr radeon;
+	char *regname;
+	struct radeon_bo *depth_bo = NULL, *bo;
+
+	if (RADEON_DEBUG & RADEON_DRI)
+	    fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
+
+	draw = drawable->driverPrivate;
+	screen = context->driScreenPriv;
+	radeon = (radeonContextPtr) context->driverPrivate;
+
+	if (screen->dri2.loader
+	   && (screen->dri2.loader->base.version > 2)
+	   && (screen->dri2.loader->getBuffersWithFormat != NULL)) {
+		struct radeon_renderbuffer *depth_rb;
+		struct radeon_renderbuffer *stencil_rb;
+
+		i = 0;
+		if ((radeon->is_front_buffer_rendering ||
+		     radeon->is_front_buffer_reading ||
+		     !draw->color_rb[1])
+		    && draw->color_rb[0]) {
+			attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+			attachments[i++] = radeon_bits_per_pixel(draw->color_rb[0]);
+		}
+
+		if (draw->color_rb[1]) {
+			attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+			attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]);
+		}
+
+		depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+		stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+
+		if ((depth_rb != NULL) && (stencil_rb != NULL)) {
+			attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL;
+			attachments[i++] = radeon_bits_per_pixel(depth_rb);
+		} else if (depth_rb != NULL) {
+			attachments[i++] = __DRI_BUFFER_DEPTH;
+			attachments[i++] = radeon_bits_per_pixel(depth_rb);
+		} else if (stencil_rb != NULL) {
+			attachments[i++] = __DRI_BUFFER_STENCIL;
+			attachments[i++] = radeon_bits_per_pixel(stencil_rb);
+		}
+
+		buffers = (*screen->dri2.loader->getBuffersWithFormat)(drawable,
+								&drawable->w,
+								&drawable->h,
+								attachments, i / 2,
+								&count,
+								drawable->loaderPrivate);
+	} else if (screen->dri2.loader) {
+		i = 0;
+		if (draw->color_rb[0])
+			attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+		if (draw->color_rb[1])
+			attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+		if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH))
+			attachments[i++] = __DRI_BUFFER_DEPTH;
+		if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL))
+			attachments[i++] = __DRI_BUFFER_STENCIL;
+
+		buffers = (*screen->dri2.loader->getBuffers)(drawable,
+								 &drawable->w,
+								 &drawable->h,
+								 attachments, i,
+								 &count,
+								 drawable->loaderPrivate);
+	}
+
+	if (buffers == NULL)
+		return;
+
+	/* set one cliprect to cover the whole drawable */
+	drawable->x = 0;
+	drawable->y = 0;
+	drawable->backX = 0;
+	drawable->backY = 0;
+	drawable->numClipRects = 1;
+	drawable->pClipRects[0].x1 = 0;
+	drawable->pClipRects[0].y1 = 0;
+	drawable->pClipRects[0].x2 = drawable->w;
+	drawable->pClipRects[0].y2 = drawable->h;
+	drawable->numBackClipRects = 1;
+	drawable->pBackClipRects[0].x1 = 0;
+	drawable->pBackClipRects[0].y1 = 0;
+	drawable->pBackClipRects[0].x2 = drawable->w;
+	drawable->pBackClipRects[0].y2 = drawable->h;
+	for (i = 0; i < count; i++) {
+		switch (buffers[i].attachment) {
+		case __DRI_BUFFER_FRONT_LEFT:
+			rb = draw->color_rb[0];
+			regname = "dri2 front buffer";
+			break;
+		case __DRI_BUFFER_FAKE_FRONT_LEFT:
+			rb = draw->color_rb[0];
+			regname = "dri2 fake front buffer";
+			break;
+		case __DRI_BUFFER_BACK_LEFT:
+			rb = draw->color_rb[1];
+			regname = "dri2 back buffer";
+			break;
+		case __DRI_BUFFER_DEPTH:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+			regname = "dri2 depth buffer";
+			break;
+		case __DRI_BUFFER_DEPTH_STENCIL:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+			regname = "dri2 depth / stencil buffer";
+			break;
+		case __DRI_BUFFER_STENCIL:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+			regname = "dri2 stencil buffer";
+			break;
+		case __DRI_BUFFER_ACCUM:
+		default:
+			fprintf(stderr,
+				"unhandled buffer attach event, attacment type %d\n",
+				buffers[i].attachment);
+			return;
+		}
+
+		if (rb == NULL)
+			continue;
+
+		if (rb->bo) {
+			uint32_t name = radeon_gem_name_bo(rb->bo);
+			if (name == buffers[i].name)
+				continue;
+		}
+
+		if (RADEON_DEBUG & RADEON_DRI)
+			fprintf(stderr,
+				"attaching buffer %s, %d, at %d, cpp %d, pitch %d\n",
+				regname, buffers[i].name, buffers[i].attachment,
+				buffers[i].cpp, buffers[i].pitch);
+
+		rb->cpp = buffers[i].cpp;
+		rb->pitch = buffers[i].pitch;
+		rb->base.Width = drawable->w;
+		rb->base.Height = drawable->h;
+		rb->has_surface = 0;
+
+		if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) {
+			if (RADEON_DEBUG & RADEON_DRI)
+				fprintf(stderr, "(reusing depth buffer as stencil)\n");
+			bo = depth_bo;
+			radeon_bo_ref(bo);
+		} else {
+			uint32_t tiling_flags = 0, pitch = 0;
+			int ret;
+
+			bo = radeon_bo_open(radeon->radeonScreen->bom,
+						buffers[i].name,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						buffers[i].flags);
+
+			if (bo == NULL) {
+
+				fprintf(stderr, "failed to attach %s %d\n",
+					regname, buffers[i].name);
+
+			}
+
+			ret = radeon_bo_get_tiling(bo, &tiling_flags, &pitch);
+			if (tiling_flags & RADEON_TILING_MACRO)
+				bo->flags |= RADEON_BO_FLAGS_MACRO_TILE;
+			if (tiling_flags & RADEON_TILING_MICRO)
+				bo->flags |= RADEON_BO_FLAGS_MICRO_TILE;
+			
+		}
+
+		if (buffers[i].attachment == __DRI_BUFFER_DEPTH) {
+			if (draw->base.Visual.depthBits == 16)
+				rb->cpp = 2;
+			depth_bo = bo;
+		}
+
+		radeon_renderbuffer_set_bo(rb, bo);
+		radeon_bo_unref(bo);
+
+		if (buffers[i].attachment == __DRI_BUFFER_DEPTH_STENCIL) {
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+			if (rb != NULL) {
+				struct radeon_bo *stencil_bo = NULL;
+
+				if (rb->bo) {
+					uint32_t name = radeon_gem_name_bo(rb->bo);
+					if (name == buffers[i].name)
+						continue;
+				}
+
+				stencil_bo = bo;
+				radeon_bo_ref(stencil_bo);
+				radeon_renderbuffer_set_bo(rb, stencil_bo);
+				radeon_bo_unref(stencil_bo);
+			}
+		}
+	}
+
+	driUpdateFramebufferSize(radeon->glCtx, drawable);
+}
+
+/* Force the context `c' to be the current context and associate with it
+ * buffer `b'.
+ */
+GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
+			    __DRIdrawablePrivate * driDrawPriv,
+			    __DRIdrawablePrivate * driReadPriv)
+{
+	radeonContextPtr radeon;
+	struct radeon_framebuffer *drfb;
+	struct gl_framebuffer *readfb;
+
+	if (!driContextPriv) {
+		if (RADEON_DEBUG & RADEON_DRI)
+			fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
+		_mesa_make_current(NULL, NULL, NULL);
+		return GL_TRUE;
+	}
+
+	radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+	drfb = driDrawPriv->driverPrivate;
+	readfb = driReadPriv->driverPrivate;
+
+	if (driContextPriv->driScreenPriv->dri2.enabled) {
+		radeon_update_renderbuffers(driContextPriv, driDrawPriv);
+		if (driDrawPriv != driReadPriv)
+			radeon_update_renderbuffers(driContextPriv, driReadPriv);
+		_mesa_reference_renderbuffer(&radeon->state.color.rb,
+			&(radeon_get_renderbuffer(&drfb->base, BUFFER_BACK_LEFT)->base));
+		_mesa_reference_renderbuffer(&radeon->state.depth.rb,
+			&(radeon_get_renderbuffer(&drfb->base, BUFFER_DEPTH)->base));
+	} else {
+		radeon_make_renderbuffer_current(radeon, drfb);
+	}
+
+	if (RADEON_DEBUG & RADEON_DRI)
+	     fprintf(stderr, "%s ctx %p dfb %p rfb %p\n", __FUNCTION__, radeon->glCtx, drfb, readfb);
+
+	driUpdateFramebufferSize(radeon->glCtx, driDrawPriv);
+	if (driReadPriv != driDrawPriv)
+		driUpdateFramebufferSize(radeon->glCtx, driReadPriv);
+
+	_mesa_make_current(radeon->glCtx, &drfb->base, readfb);
+
+	_mesa_update_state(radeon->glCtx);
+
+	if (radeon->glCtx->DrawBuffer == &drfb->base) {
+		if (driDrawPriv->swap_interval == (unsigned)-1) {
+			int i;
+			driDrawPriv->vblFlags =
+				(radeon->radeonScreen->irq != 0)
+				? driGetDefaultVBlankFlags(&radeon->
+							   optionCache)
+				: VBLANK_FLAG_NO_IRQ;
+
+			driDrawableInitVBlank(driDrawPriv);
+			drfb->vbl_waited = driDrawPriv->vblSeq;
+
+			for (i = 0; i < 2; i++) {
+				if (drfb->color_rb[i])
+					drfb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq;
+			}
+
+		}
+
+		radeon_window_moved(radeon);
+		radeon_draw_buffer(radeon->glCtx, &drfb->base);
+	}
+
+
+	if (RADEON_DEBUG & RADEON_DRI)
+		fprintf(stderr, "End %s\n", __FUNCTION__);
+
+	return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
new file mode 100644
index 0000000000..0309345393
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -0,0 +1,594 @@
+
+#ifndef COMMON_CONTEXT_H
+#define COMMON_CONTEXT_H
+
+#include "main/mm.h"
+#include "math/m_vector.h"
+#include "texmem.h"
+#include "tnl/t_context.h"
+#include "main/colormac.h"
+
+#include "radeon_debug.h"
+#include "radeon_screen.h"
+#include "radeon_drm.h"
+#include "dri_util.h"
+#include "tnl/t_vertex.h"
+
+#include "dri_metaops.h"
+struct radeon_context;
+
+#include "radeon_bocs_wrapper.h"
+
+/* This union is used to avoid warnings/miscompilation
+   with float to uint32_t casts due to strict-aliasing */
+typedef union { GLfloat f; uint32_t ui32; } float_ui32_type;
+
+struct radeon_context;
+typedef struct radeon_context radeonContextRec;
+typedef struct radeon_context *radeonContextPtr;
+
+
+#define TEX_0   0x1
+#define TEX_1   0x2
+#define TEX_2   0x4
+#define TEX_3	0x8
+#define TEX_4	0x10
+#define TEX_5	0x20
+
+/* Rasterizing fallbacks */
+/* See correponding strings in r200_swtcl.c */
+#define RADEON_FALLBACK_TEXTURE		0x0001
+#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
+#define RADEON_FALLBACK_STENCIL		0x0004
+#define RADEON_FALLBACK_RENDER_MODE	0x0008
+#define RADEON_FALLBACK_BLEND_EQ	0x0010
+#define RADEON_FALLBACK_BLEND_FUNC	0x0020
+#define RADEON_FALLBACK_DISABLE 	0x0040
+#define RADEON_FALLBACK_BORDER_MODE	0x0080
+#define RADEON_FALLBACK_DEPTH_BUFFER	0x0100
+#define RADEON_FALLBACK_STENCIL_BUFFER  0x0200
+
+#define R200_FALLBACK_TEXTURE           0x01
+#define R200_FALLBACK_DRAW_BUFFER       0x02
+#define R200_FALLBACK_STENCIL           0x04
+#define R200_FALLBACK_RENDER_MODE       0x08
+#define R200_FALLBACK_DISABLE           0x10
+#define R200_FALLBACK_BORDER_MODE       0x20
+
+#define RADEON_TCL_FALLBACK_RASTER            0x1 /* rasterization */
+#define RADEON_TCL_FALLBACK_UNFILLED          0x2 /* unfilled tris */
+#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE     0x4 /* twoside tris */
+#define RADEON_TCL_FALLBACK_MATERIAL          0x8 /* material in vb */
+#define RADEON_TCL_FALLBACK_TEXGEN_0          0x10 /* texgen, unit 0 */
+#define RADEON_TCL_FALLBACK_TEXGEN_1          0x20 /* texgen, unit 1 */
+#define RADEON_TCL_FALLBACK_TEXGEN_2          0x40 /* texgen, unit 2 */
+#define RADEON_TCL_FALLBACK_TCL_DISABLE       0x80 /* user disable */
+#define RADEON_TCL_FALLBACK_FOGCOORDSPEC      0x100 /* fogcoord, sep. spec light */
+
+/* The blit width for texture uploads
+ */
+#define BLIT_WIDTH_BYTES 1024
+
+/* Use the templated vertex format:
+ */
+#define COLOR_IS_RGBA
+#define TAG(x) radeon##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+#define RADEON_RB_CLASS 0xdeadbeef
+
+struct radeon_renderbuffer
+{
+	struct gl_renderbuffer base;
+	struct radeon_bo *bo;
+	unsigned int cpp;
+	/* unsigned int offset; */
+	unsigned int pitch;
+
+	uint32_t draw_offset; /* FBO */
+	/* boo Xorg 6.8.2 compat */
+	int has_surface;
+
+	GLuint pf_pending;  /**< sequence number of pending flip */
+	GLuint vbl_pending;   /**< vblank sequence number of pending flip */
+	__DRIdrawablePrivate *dPriv;
+};
+
+struct radeon_framebuffer
+{
+	struct gl_framebuffer base;
+
+	struct radeon_renderbuffer *color_rb[2];
+
+	GLuint vbl_waited;
+
+	/* buffer swap */
+	int64_t swap_ust;
+	int64_t swap_missed_ust;
+
+	GLuint swap_count;
+	GLuint swap_missed_count;
+
+	/* Drawable page flipping state */
+	GLboolean pf_active;
+	GLint pf_current_page;
+	GLint pf_num_pages;
+
+};
+
+
+struct radeon_colorbuffer_state {
+	GLuint clear;
+	int roundEnable;
+	struct gl_renderbuffer *rb;
+	uint32_t draw_offset; /* offset into color renderbuffer - FBOs */
+};
+
+struct radeon_depthbuffer_state {
+	GLuint clear;
+	struct gl_renderbuffer *rb;
+};
+
+struct radeon_scissor_state {
+	drm_clip_rect_t rect;
+	GLboolean enabled;
+
+	GLuint numClipRects;	/* Cliprects active */
+	GLuint numAllocedClipRects;	/* Cliprects available */
+	drm_clip_rect_t *pClipRects;
+};
+
+struct radeon_stencilbuffer_state {
+	GLuint clear;		/* rb3d_stencilrefmask value */
+};
+
+struct radeon_state_atom {
+	struct radeon_state_atom *next, *prev;
+	const char *name;	/* for debug */
+	int cmd_size;		/* size in bytes */
+        GLuint idx;
+	GLuint is_tcl;
+        GLuint *cmd;		/* one or more cmd's */
+	GLuint *lastcmd;		/* one or more cmd's */
+	GLboolean dirty;	/* dirty-mark in emit_state_list */
+        int (*check) (GLcontext *, struct radeon_state_atom *atom); /* is this state active? */
+        void (*emit) (GLcontext *, struct radeon_state_atom *atom);
+};
+
+struct radeon_hw_state {
+  	/* Head of the linked list of state atoms. */
+	struct radeon_state_atom atomlist;
+	int max_state_size;	/* Number of bytes necessary for a full state emit. */
+	int max_post_flush_size; /* Number of bytes necessary for post flushing emits */
+	GLboolean is_dirty, all_dirty;
+};
+
+
+/* Texture related */
+typedef struct _radeon_texture_image radeon_texture_image;
+
+struct _radeon_texture_image {
+	struct gl_texture_image base;
+
+	/**
+	 * If mt != 0, the image is stored in hardware format in the
+	 * given mipmap tree. In this case, base.Data may point into the
+	 * mapping of the buffer object that contains the mipmap tree.
+	 *
+	 * If mt == 0, the image is stored in normal memory pointed to
+	 * by base.Data.
+	 */
+	struct _radeon_mipmap_tree *mt;
+	struct radeon_bo *bo;
+
+	int mtlevel; /** if mt != 0, this is the image's level in the mipmap tree */
+	int mtface; /** if mt != 0, this is the image's face in the mipmap tree */
+};
+
+
+static INLINE radeon_texture_image *get_radeon_texture_image(struct gl_texture_image *image)
+{
+	return (radeon_texture_image*)image;
+}
+
+
+typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr;
+
+#define RADEON_TXO_MICRO_TILE               (1 << 3)
+
+/* Texture object in locally shared texture space.
+ */
+struct radeon_tex_obj {
+	struct gl_texture_object base;
+	struct _radeon_mipmap_tree *mt;
+
+	/**
+	 * This is true if we've verified that the mipmap tree above is complete
+	 * and so on.
+	 */
+	GLboolean validated;
+
+	GLuint override_offset;
+	GLboolean image_override; /* Image overridden by GLX_EXT_tfp */
+	GLuint tile_bits;	/* hw texture tile bits used on this texture */
+        struct radeon_bo *bo;
+
+	GLuint pp_txfilter;	/* hardware register values */
+	GLuint pp_txformat;
+	GLuint pp_txformat_x;
+	GLuint pp_txsize;	/* npot only */
+	GLuint pp_txpitch;	/* npot only */
+	GLuint pp_border_color;
+	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
+
+        GLuint pp_txfilter_1;	/*  r300 */
+
+	/* r700 texture states */
+	GLuint SQ_TEX_RESOURCE0;
+	GLuint SQ_TEX_RESOURCE1;
+	GLuint SQ_TEX_RESOURCE2;
+	GLuint SQ_TEX_RESOURCE3;
+	GLuint SQ_TEX_RESOURCE4;
+	GLuint SQ_TEX_RESOURCE5;
+	GLuint SQ_TEX_RESOURCE6;
+
+	GLuint SQ_TEX_SAMPLER0;
+	GLuint SQ_TEX_SAMPLER1;
+	GLuint SQ_TEX_SAMPLER2;
+
+	GLuint TD_PS_SAMPLER0_BORDER_RED;
+	GLuint TD_PS_SAMPLER0_BORDER_GREEN;
+	GLuint TD_PS_SAMPLER0_BORDER_BLUE;
+	GLuint TD_PS_SAMPLER0_BORDER_ALPHA;
+
+	GLboolean border_fallback;
+
+
+};
+
+static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj)
+{
+	return (radeonTexObj*)texObj;
+}
+
+/* occlusion query */
+struct radeon_query_object {
+	struct gl_query_object Base;
+	struct radeon_bo *bo;
+	int curr_offset;
+	GLboolean emitted_begin;
+
+	/* Double linked list of not flushed query objects */
+	struct radeon_query_object *prev, *next;
+};
+
+/* Need refcounting on dma buffers:
+ */
+struct radeon_dma_buffer {
+	int refcount;		/* the number of retained regions in buf */
+	drmBufPtr buf;
+};
+
+struct radeon_aos {
+	struct radeon_bo *bo; /** Buffer object where vertex data is stored */
+	int offset; /** Offset into buffer object, in bytes */
+	int components; /** Number of components per vertex */
+	int stride; /** Stride in dwords (may be 0 for repeating) */
+	int count; /** Number of vertices */
+};
+
+#define DMA_BO_FREE_TIME 100
+
+struct radeon_dma_bo {
+  struct radeon_dma_bo *next, *prev;
+  struct radeon_bo *bo;
+  int expire_counter;
+};
+
+struct radeon_dma {
+        /* Active dma region.  Allocations for vertices and retained
+         * regions come from here.  Also used for emitting random vertices,
+         * these may be flushed by calling flush_current();
+         */
+	struct radeon_dma_bo free;
+	struct radeon_dma_bo wait;
+	struct radeon_dma_bo reserved;
+        size_t current_used; /** Number of bytes allocated and forgotten about */
+        size_t current_vertexptr; /** End of active vertex region */
+        size_t minimum_size;
+
+        /**
+         * If current_vertexptr != current_used then flush must be non-zero.
+         * flush must be called before non-active vertex allocations can be
+         * performed.
+         */
+        void (*flush) (GLcontext *);
+};
+
+/* radeon_swtcl.c
+ */
+struct radeon_swtcl_info {
+
+	GLuint RenderIndex;
+	GLuint vertex_size;
+	GLubyte *verts;
+
+	/* Fallback rasterization functions
+	 */
+	GLuint hw_primitive;
+	GLenum render_primitive;
+	GLuint numverts;
+
+	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
+	GLuint vertex_attr_count;
+
+	GLuint emit_prediction;
+};
+
+#define RADEON_MAX_AOS_ARRAYS		16
+struct radeon_tcl_info {
+	struct radeon_aos aos[RADEON_MAX_AOS_ARRAYS];
+	GLuint aos_count;
+	struct radeon_bo *elt_dma_bo; /** Buffer object that contains element indices */
+	int elt_dma_offset; /** Offset into this buffer object, in bytes */
+};
+
+struct radeon_ioctl {
+	GLuint vertex_offset;
+	GLuint vertex_max;
+	struct radeon_bo *bo;
+	GLuint vertex_size;
+};
+
+#define RADEON_MAX_PRIMS 64
+
+struct radeon_prim {
+	GLuint start;
+	GLuint end;
+	GLuint prim;
+};
+
+static INLINE GLuint radeonPackColor(GLuint cpp,
+                                     GLubyte r, GLubyte g,
+                                     GLubyte b, GLubyte a)
+{
+	switch (cpp) {
+	case 2:
+		return PACK_COLOR_565(r, g, b);
+	case 4:
+		return PACK_COLOR_8888(a, r, g, b);
+	default:
+		return 0;
+	}
+}
+
+#define MAX_CMD_BUF_SZ (16*1024)
+
+#define MAX_DMA_BUF_SZ (64*1024)
+
+struct radeon_store {
+	GLuint statenr;
+	GLuint primnr;
+	char cmd_buf[MAX_CMD_BUF_SZ];
+	int cmd_used;
+	int elts_start;
+};
+
+struct radeon_dri_mirror {
+	__DRIcontextPrivate *context;	/* DRI context */
+	__DRIscreenPrivate *screen;	/* DRI screen */
+
+	drm_context_t hwContext;
+	drm_hw_lock_t *hwLock;
+	int hwLockCount;
+	int fd;
+	int drmMinor;
+};
+
+typedef void (*radeon_tri_func) (radeonContextPtr,
+				 radeonVertex *,
+				 radeonVertex *, radeonVertex *);
+
+typedef void (*radeon_line_func) (radeonContextPtr,
+				  radeonVertex *, radeonVertex *);
+
+typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *);
+
+#define RADEON_MAX_BOS 32
+struct radeon_state {
+	struct radeon_colorbuffer_state color;
+	struct radeon_depthbuffer_state depth;
+	struct radeon_scissor_state scissor;
+	struct radeon_stencilbuffer_state stencil;
+
+	struct radeon_cs_space_check bos[RADEON_MAX_BOS];
+	int validated_bo_count;
+};
+
+/**
+ * This structure holds the command buffer while it is being constructed.
+ *
+ * The first batch of commands in the buffer is always the state that needs
+ * to be re-emitted when the context is lost. This batch can be skipped
+ * otherwise.
+ */
+struct radeon_cmdbuf {
+	struct radeon_cs_manager    *csm;
+	struct radeon_cs            *cs;
+	int size; /** # of dwords total */
+	unsigned int flushing:1; /** whether we're currently in FlushCmdBufLocked */
+};
+
+struct radeon_context {
+   GLcontext *glCtx;
+   radeonScreenPtr radeonScreen;	/* Screen private DRI data */
+
+   /* Texture object bookkeeping
+    */
+   int                   texture_depth;
+   float                 initialMaxAnisotropy;
+   uint32_t              texture_row_align;
+   uint32_t              texture_rect_row_align;
+   uint32_t              texture_compressed_row_align;
+
+  struct radeon_dma dma;
+  struct radeon_hw_state hw;
+   /* Rasterization and vertex state:
+    */
+   GLuint TclFallback;
+   GLuint Fallback;
+   GLuint NewGLState;
+   DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
+
+   /* Drawable, cliprect and scissor information */
+   GLuint numClipRects;	/* Cliprects for the draw buffer */
+   drm_clip_rect_t *pClipRects;
+   unsigned int lastStamp;
+   drm_radeon_sarea_t *sarea;	/* Private SAREA data */
+
+   /* Mirrors of some DRI state */
+   struct radeon_dri_mirror dri;
+
+   /* Busy waiting */
+   GLuint do_usleeps;
+   GLuint do_irqs;
+   GLuint irqsEmitted;
+   drm_radeon_irq_wait_t iw;
+
+   /* Derived state - for r300 only */
+   struct radeon_state state;
+
+   struct radeon_swtcl_info swtcl;
+   struct radeon_tcl_info tcl;
+   /* Configuration cache
+    */
+   driOptionCache optionCache;
+
+   struct radeon_cmdbuf cmdbuf;
+
+   struct radeon_debug debug;
+
+  drm_clip_rect_t fboRect;
+  GLboolean constant_cliprect; /* use for FBO or DRI2 rendering */
+  GLboolean front_cliprects;
+
+   /**
+    * Set if rendering has occured to the drawable's front buffer.
+    *
+    * This is used in the DRI2 case to detect that glFlush should also copy
+    * the contents of the fake front buffer to the real front buffer.
+    */
+   GLboolean front_buffer_dirty;
+
+   /**
+    * Track whether front-buffer rendering is currently enabled
+    *
+    * A separate flag is used to track this in order to support MRT more
+    * easily.
+    */
+   GLboolean is_front_buffer_rendering;
+
+   /**
+    * Track whether front-buffer is the current read target.
+    *
+    * This is closely associated with is_front_buffer_rendering, but may
+    * be set separately.  The DRI2 fake front buffer must be referenced
+    * either way.
+    */
+   GLboolean is_front_buffer_reading;
+
+   struct dri_metaops meta;
+
+   struct {
+	struct radeon_query_object *current;
+	struct radeon_query_object not_flushed_head;
+	struct radeon_state_atom queryobj;
+   } query;
+
+   struct {
+	   void (*get_lock)(radeonContextPtr radeon);
+	   void (*update_viewport_offset)(GLcontext *ctx);
+	   void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
+	   void (*swtcl_flush)(GLcontext *ctx, uint32_t offset);
+	   void (*pre_emit_atoms)(radeonContextPtr rmesa);
+	   void (*pre_emit_state)(radeonContextPtr rmesa);
+	   void (*fallback)(GLcontext *ctx, GLuint bit, GLboolean mode);
+	   void (*free_context)(GLcontext *ctx);
+	   void (*emit_query_finish)(radeonContextPtr radeon);
+	   void (*update_scissor)(GLcontext *ctx);
+   } vtbl;
+};
+
+#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx))
+
+static inline __DRIdrawablePrivate* radeon_get_drawable(radeonContextPtr radeon)
+{
+	return radeon->dri.context->driDrawablePriv;
+}
+
+static inline __DRIdrawablePrivate* radeon_get_readable(radeonContextPtr radeon)
+{
+	return radeon->dri.context->driReadablePriv;
+}
+
+/**
+ * This function takes a float and packs it into a uint32_t
+ */
+static INLINE uint32_t radeonPackFloat32(float fl)
+{
+	union {
+		float fl;
+		uint32_t u;
+	} u;
+
+	u.fl = fl;
+	return u.u;
+}
+
+/* This is probably wrong for some values, I need to test this
+ * some more.  Range checking would be a good idea also..
+ *
+ * But it works for most things.  I'll fix it later if someone
+ * else with a better clue doesn't
+ */
+static INLINE uint32_t radeonPackFloat24(float f)
+{
+	float mantissa;
+	int exponent;
+	uint32_t float24 = 0;
+
+	if (f == 0.0)
+		return 0;
+
+	mantissa = frexpf(f, &exponent);
+
+	/* Handle -ve */
+	if (mantissa < 0) {
+		float24 |= (1 << 23);
+		mantissa = mantissa * -1.0;
+	}
+	/* Handle exponent, bias of 63 */
+	exponent += 62;
+	float24 |= (exponent << 16);
+	/* Kill 7 LSB of mantissa */
+	float24 |= (radeonPackFloat32(mantissa) & 0x7FFFFF) >> 7;
+
+	return float24;
+}
+
+GLboolean radeonInitContext(radeonContextPtr radeon,
+			    struct dd_function_table* functions,
+			    const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate);
+
+void radeonCleanupContext(radeonContextPtr radeon);
+GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
+void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable);
+GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
+			    __DRIdrawablePrivate * driDrawPriv,
+			    __DRIdrawablePrivate * driReadPriv);
+extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_compat.c b/src/mesa/drivers/dri/radeon/radeon_compat.c
deleted file mode 100644
index 46b490d61f..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_compat.c
+++ /dev/null
@@ -1,301 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 ATI Technologies Inc., Ontario, Canada, and
-               Tungsten Graphics Inc., Austin, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-
-#include "radeon_context.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-
-
-static struct { 
-	int start; 
-	int len; 
-	const char *name;
-} packet[RADEON_MAX_STATE_PACKETS] = {
-	{ RADEON_PP_MISC,7,"RADEON_PP_MISC" },
-	{ RADEON_PP_CNTL,3,"RADEON_PP_CNTL" },
-	{ RADEON_RB3D_COLORPITCH,1,"RADEON_RB3D_COLORPITCH" },
-	{ RADEON_RE_LINE_PATTERN,2,"RADEON_RE_LINE_PATTERN" },
-	{ RADEON_SE_LINE_WIDTH,1,"RADEON_SE_LINE_WIDTH" },
-	{ RADEON_PP_LUM_MATRIX,1,"RADEON_PP_LUM_MATRIX" },
-	{ RADEON_PP_ROT_MATRIX_0,2,"RADEON_PP_ROT_MATRIX_0" },
-	{ RADEON_RB3D_STENCILREFMASK,3,"RADEON_RB3D_STENCILREFMASK" },
-	{ RADEON_SE_VPORT_XSCALE,6,"RADEON_SE_VPORT_XSCALE" },
-	{ RADEON_SE_CNTL,2,"RADEON_SE_CNTL" },
-	{ RADEON_SE_CNTL_STATUS,1,"RADEON_SE_CNTL_STATUS" },
-	{ RADEON_RE_MISC,1,"RADEON_RE_MISC" },
-	{ RADEON_PP_TXFILTER_0,6,"RADEON_PP_TXFILTER_0" },
-	{ RADEON_PP_BORDER_COLOR_0,1,"RADEON_PP_BORDER_COLOR_0" },
-	{ RADEON_PP_TXFILTER_1,6,"RADEON_PP_TXFILTER_1" },
-	{ RADEON_PP_BORDER_COLOR_1,1,"RADEON_PP_BORDER_COLOR_1" },
-	{ RADEON_PP_TXFILTER_2,6,"RADEON_PP_TXFILTER_2" },
-	{ RADEON_PP_BORDER_COLOR_2,1,"RADEON_PP_BORDER_COLOR_2" },
-	{ RADEON_SE_ZBIAS_FACTOR,2,"RADEON_SE_ZBIAS_FACTOR" },
-	{ RADEON_SE_TCL_OUTPUT_VTX_FMT,11,"RADEON_SE_TCL_OUTPUT_VTX_FMT" },
-	{ RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED,17,"RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED" },
-};
-
-
-static void radeonCompatEmitPacket( radeonContextPtr rmesa, 
-				    struct radeon_state_atom *state )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-   drm_radeon_context_regs_t *ctx = &sarea->context_state;
-   drm_radeon_texture_regs_t *tex0 = &sarea->tex_state[0];
-   drm_radeon_texture_regs_t *tex1 = &sarea->tex_state[1];
-   int i;
-   int *buf = state->cmd;
-
-   for ( i = 0 ; i < state->cmd_size ; ) {
-      drm_radeon_cmd_header_t *header = (drm_radeon_cmd_header_t *)&buf[i++];
-
-      if (RADEON_DEBUG & DEBUG_STATE)
-	 fprintf(stderr, "%s %d: %s\n", __FUNCTION__, header->packet.packet_id,
-		 packet[(int)header->packet.packet_id].name);
-
-      switch (header->packet.packet_id) {
-      case RADEON_EMIT_PP_MISC:
-	 ctx->pp_misc = buf[i++]; 
-	 ctx->pp_fog_color = buf[i++];
-	 ctx->re_solid_color = buf[i++];
-	 ctx->rb3d_blendcntl = buf[i++];
-	 ctx->rb3d_depthoffset = buf[i++];
-	 ctx->rb3d_depthpitch = buf[i++];
-	 ctx->rb3d_zstencilcntl = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_PP_CNTL:
-	 ctx->pp_cntl = buf[i++];
-	 ctx->rb3d_cntl = buf[i++];
-	 ctx->rb3d_coloroffset = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_RB3D_COLORPITCH:
-	 ctx->rb3d_colorpitch = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_RE_LINE_PATTERN:
-	 ctx->re_line_pattern = buf[i++];
-	 ctx->re_line_state = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_LINE;
-	 break;
-      case RADEON_EMIT_SE_LINE_WIDTH:
-	 ctx->se_line_width = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_LINE;
-	 break;
-      case RADEON_EMIT_PP_LUM_MATRIX:
-	 ctx->pp_lum_matrix = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_BUMPMAP;
-	 break;
-      case RADEON_EMIT_PP_ROT_MATRIX_0:
-	 ctx->pp_rot_matrix_0 = buf[i++];
-	 ctx->pp_rot_matrix_1 = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_BUMPMAP;
-	 break;
-      case RADEON_EMIT_RB3D_STENCILREFMASK:
-	 ctx->rb3d_stencilrefmask = buf[i++];
-	 ctx->rb3d_ropcntl = buf[i++];
-	 ctx->rb3d_planemask = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_MASKS;
-	 break;
-      case RADEON_EMIT_SE_VPORT_XSCALE:
-	 ctx->se_vport_xscale = buf[i++];
-	 ctx->se_vport_xoffset = buf[i++];
-	 ctx->se_vport_yscale = buf[i++];
-	 ctx->se_vport_yoffset = buf[i++];
-	 ctx->se_vport_zscale = buf[i++];
-	 ctx->se_vport_zoffset = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_VIEWPORT;
-	 break;
-      case RADEON_EMIT_SE_CNTL:
-	 ctx->se_cntl = buf[i++];
-	 ctx->se_coord_fmt = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT | RADEON_UPLOAD_VERTFMT;
-	 break;
-      case RADEON_EMIT_SE_CNTL_STATUS:
-	 ctx->se_cntl_status = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_SETUP;
-	 break;
-      case RADEON_EMIT_RE_MISC:
-	 ctx->re_misc = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_MISC;
-	 break;
-      case RADEON_EMIT_PP_TXFILTER_0:
-	 tex0->pp_txfilter = buf[i++];
-	 tex0->pp_txformat = buf[i++];
-	 tex0->pp_txoffset = buf[i++];
-	 tex0->pp_txcblend = buf[i++];
-	 tex0->pp_txablend = buf[i++];
-	 tex0->pp_tfactor = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX0;
-	 break;
-      case RADEON_EMIT_PP_BORDER_COLOR_0:
-	 tex0->pp_border_color = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX0;
-	 break;
-      case RADEON_EMIT_PP_TXFILTER_1:
-	 tex1->pp_txfilter = buf[i++];
-	 tex1->pp_txformat = buf[i++];
-	 tex1->pp_txoffset = buf[i++];
-	 tex1->pp_txcblend = buf[i++];
-	 tex1->pp_txablend = buf[i++];
-	 tex1->pp_tfactor = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX1;
-	 break;
-      case RADEON_EMIT_PP_BORDER_COLOR_1:
-	 tex1->pp_border_color = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX1;
-	 break;
-
-      case RADEON_EMIT_SE_ZBIAS_FACTOR:
-	 i++;
-	 i++;
-	 break;
-
-      case RADEON_EMIT_PP_TXFILTER_2:
-      case RADEON_EMIT_PP_BORDER_COLOR_2:
-      case RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT:
-      case RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED:
-      default:
-	 /* These states aren't understood by radeon drm 1.1 */
-	 fprintf(stderr, "Tried to emit unsupported state\n");
-	 return;
-      }
-   }
-}
-
-
-
-static void radeonCompatEmitStateLocked( radeonContextPtr rmesa )
-{
-   struct radeon_state_atom *atom;
-
-   if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (!rmesa->hw.is_dirty && !rmesa->hw.all_dirty)
-      return;
-
-   foreach(atom, &rmesa->hw.atomlist) {
-      if (rmesa->hw.all_dirty)
-	 atom->dirty = GL_TRUE;
-      if (atom->is_tcl)
-	 atom->dirty = GL_FALSE;
-      if (atom->dirty)
-	 radeonCompatEmitPacket(rmesa, atom);
-   }
- 
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
-}
-
-
-static void radeonCompatEmitPrimitiveLocked( radeonContextPtr rmesa,
-					     GLuint hw_primitive,
-					     GLuint nverts,
-					     drm_clip_rect_t *pbox,
-					     GLuint nbox )
-{
-   int i;
-
-   for ( i = 0 ; i < nbox ; ) {
-      int nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, nbox );
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      drm_radeon_vertex_t vtx;
-      
-      rmesa->sarea->dirty |= RADEON_UPLOAD_CLIPRECTS;
-      rmesa->sarea->nbox = nr - i;
-
-      for ( ; i < nr ; i++) 
-	 *b++ = pbox[i];
-      
-      if (RADEON_DEBUG & DEBUG_IOCTL)
-	 fprintf(stderr, 
-		 "RadeonFlushVertexBuffer: prim %x buf %d verts %d "
-		 "disc %d nbox %d\n",
-		 hw_primitive, 
-		 rmesa->dma.current.buf->buf->idx, 
-		 nverts, 
-		 nr == nbox,
-		 rmesa->sarea->nbox );
-
-      vtx.prim = hw_primitive;
-      vtx.idx = rmesa->dma.current.buf->buf->idx;
-      vtx.count = nverts;
-      vtx.discard = (nr == nbox);      
-
-      drmCommandWrite( rmesa->dri.fd, 
-		       DRM_RADEON_VERTEX,
-		       &vtx, sizeof(vtx));
-   }
-}
-
-
-
-/* No 'start' for 1.1 vertices ioctl: only one vertex prim/buffer!  
- */
-void radeonCompatEmitPrimitive( radeonContextPtr rmesa,
-				GLuint vertex_format,
-				GLuint hw_primitive,
-				GLuint nrverts )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   LOCK_HARDWARE( rmesa );
-
-   radeonCompatEmitStateLocked( rmesa );
-   rmesa->sarea->vc_format = vertex_format;
-   
-   if (rmesa->state.scissor.enabled) {
-      radeonCompatEmitPrimitiveLocked( rmesa, 
-				       hw_primitive,
-				       nrverts,
-				       rmesa->state.scissor.pClipRects,
-				       rmesa->state.scissor.numClipRects );
-   }
-   else {
-      radeonCompatEmitPrimitiveLocked( rmesa, 
-				       hw_primitive,
-				       nrverts,
-				       rmesa->pClipRects,
-				       rmesa->numClipRects );
-   }
-
-
-   UNLOCK_HARDWARE( rmesa );
-}
-
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index b67bda7d06..8f4485aee7 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -53,6 +53,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "drivers/common/driverfuncs.h"
 
+#include "radeon_common.h"
 #include "radeon_context.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
@@ -61,10 +62,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_swtcl.h"
 #include "radeon_tcl.h"
 #include "radeon_maos.h"
+#include "radeon_queryobj.h"
 
+#define need_GL_ARB_occlusion_query
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20061018"
@@ -72,46 +76,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h" /* for symbolic values of enum-type options */
-#ifndef RADEON_DEBUG
-int RADEON_DEBUG = (0);
-#endif
-
-
-/* Return various strings for glGetString().
- */
-static const GLubyte *radeonGetString( GLcontext *ctx, GLenum name )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   static char buffer[128];
-   unsigned   offset;
-   GLuint agp_mode = (rmesa->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
-      rmesa->radeonScreen->AGPMode;
-
-   switch ( name ) {
-   case GL_VENDOR:
-      return (GLubyte *)"Tungsten Graphics, Inc.";
-
-   case GL_RENDERER:
-      offset = driGetRendererString( buffer, "Radeon", DRIVER_DATE,
-				     agp_mode );
-
-      sprintf( & buffer[ offset ], " %sTCL",
-	       !(rmesa->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
-	       ? "" : "NO-" );
-
-      return (GLubyte *)buffer;
-
-   default:
-      return NULL;
-   }
-}
-
 
 /* Extension strings exported by the R100 driver.
  */
 const struct dri_extension card_extensions[] =
 {
     { "GL_ARB_multitexture",               NULL },
+    { "GL_ARB_occlusion_query",		   GL_ARB_occlusion_query_functions},
     { "GL_ARB_texture_border_clamp",       NULL },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
@@ -121,6 +92,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_logic_op",             NULL },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -137,6 +109,11 @@ const struct dri_extension card_extensions[] =
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _radeon_render_stage;
 extern const struct tnl_pipeline_stage _radeon_tcl_stage;
 
@@ -160,47 +137,85 @@ static const struct tnl_pipeline_stage *radeon_pipeline[] = {
    NULL,
 };
 
+static void r100_get_lock(radeonContextPtr radeon)
+{
+   r100ContextPtr rmesa = (r100ContextPtr)radeon;
+   drm_radeon_sarea_t *sarea = radeon->sarea;
 
+   RADEON_STATECHANGE(rmesa, ctx);
+   if (rmesa->radeon.sarea->tiling_enabled) {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
+	 RADEON_COLOR_TILE_ENABLE;
+   } else {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &=
+	 ~RADEON_COLOR_TILE_ENABLE;
+   }
+   
+   if (sarea->ctx_owner != rmesa->radeon.dri.hwContext) {
+      sarea->ctx_owner = rmesa->radeon.dri.hwContext;
+      
+      if (!radeon->radeonScreen->kernel_mm)
+         radeon_bo_legacy_texture_age(radeon->radeonScreen->bom);
+   }
+}
 
-/* Initialize the driver's misc functions.
- */
-static void radeonInitDriverFuncs( struct dd_function_table *functions )
+static void r100_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
 {
-    functions->GetString	= radeonGetString;
 }
 
-static const struct dri_debug_control debug_control[] =
+static void r100_vtbl_pre_emit_state(radeonContextPtr radeon)
 {
-    { "fall",  DEBUG_FALLBACKS },
-    { "tex",   DEBUG_TEXTURE },
-    { "ioctl", DEBUG_IOCTL },
-    { "prim",  DEBUG_PRIMS },
-    { "vert",  DEBUG_VERTS },
-    { "state", DEBUG_STATE },
-    { "code",  DEBUG_CODEGEN },
-    { "vfmt",  DEBUG_VFMT },
-    { "vtxf",  DEBUG_VFMT },
-    { "verb",  DEBUG_VERBOSE },
-    { "dri",   DEBUG_DRI },
-    { "dma",   DEBUG_DMA },
-    { "san",   DEBUG_SANITY },
-    { "sync",  DEBUG_SYNC },
-    { NULL,    0 }
-};
+   r100ContextPtr rmesa = (r100ContextPtr)radeon;
+   
+   /* r100 always needs to emit ZBS to avoid TCL lockups */
+   rmesa->hw.zbs.dirty = 1;
+   radeon->hw.is_dirty = 1;
+}
+
+static void r100_vtbl_free_context(GLcontext *ctx)
+{
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   _mesa_vector4f_free( &rmesa->tcl.ObjClean );
+}
 
+static void r100_emit_query_finish(radeonContextPtr radeon)
+{
+   BATCH_LOCALS(radeon);
+   struct radeon_query_object *query = radeon->query.current;
+
+   BEGIN_BATCH_NO_AUTOSTATE(4);
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZPASS_ADDR, 0));
+   OUT_BATCH_RELOC(0, query->bo, query->curr_offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+   END_BATCH();
+   query->curr_offset += sizeof(uint32_t);
+   assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
+   query->emitted_begin = GL_FALSE;
+}
+
+static void r100_init_vtbl(radeonContextPtr radeon)
+{
+   radeon->vtbl.get_lock = r100_get_lock;
+   radeon->vtbl.update_viewport_offset = radeonUpdateViewportOffset;
+   radeon->vtbl.emit_cs_header = r100_vtbl_emit_cs_header;
+   radeon->vtbl.swtcl_flush = r100_swtcl_flush;
+   radeon->vtbl.pre_emit_state = r100_vtbl_pre_emit_state;
+   radeon->vtbl.fallback = radeonFallback;
+   radeon->vtbl.free_context = r100_vtbl_free_context;
+   radeon->vtbl.emit_query_finish = r100_emit_query_finish;
+}
 
 /* Create the device specific context.
  */
 GLboolean
-radeonCreateContext( const __GLcontextModes *glVisual,
+r100CreateContext( const __GLcontextModes *glVisual,
                      __DRIcontextPrivate *driContextPriv,
                      void *sharedContextPrivate)
 {
    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
    radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private);
    struct dd_function_table functions;
-   radeonContextPtr rmesa;
-   GLcontext *ctx, *shareCtx;
+   r100ContextPtr rmesa;
+   GLcontext *ctx;
    int i;
    int tcl_mode, fthrottle_mode;
 
@@ -209,10 +224,12 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    assert(screen);
 
    /* Allocate the Radeon context */
-   rmesa = (radeonContextPtr) CALLOC( sizeof(*rmesa) );
+   rmesa = (r100ContextPtr) CALLOC( sizeof(*rmesa) );
    if ( !rmesa )
       return GL_FALSE;
 
+   r100_init_vtbl(&rmesa->radeon);
+
    /* init exp fog table data */
    radeonInitStaticFogData();
    
@@ -220,12 +237,12 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * Do this here so that initialMaxAnisotropy is set before we create
     * the default textures.
     */
-   driParseConfigFiles (&rmesa->optionCache, &screen->optionCache,
+   driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
 			screen->driScreen->myNum, "radeon");
-   rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache,
+   rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
                                                  "def_max_anisotropy");
 
-   if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) {
+   if ( driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) {
       if ( sPriv->drm_version.minor < 13 )
 	 fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
 			  "disabling.\n", sPriv->drm_version.minor );
@@ -240,65 +257,18 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * (the texture functions are especially important)
     */
    _mesa_init_driver_functions( &functions );
-   radeonInitDriverFuncs( &functions );
    radeonInitTextureFuncs( &functions );
+   radeonInitQueryObjFunctions(&functions);
 
-   /* Allocate the Mesa context */
-   if (sharedContextPrivate)
-      shareCtx = ((radeonContextPtr) sharedContextPrivate)->glCtx;
-   else
-      shareCtx = NULL;
-   rmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
-                                       &functions, (void *) rmesa);
-   if (!rmesa->glCtx) {
-      FREE(rmesa);
-      return GL_FALSE;
-   }
-   driContextPriv->driverPrivate = rmesa;
-
-   /* Init radeon context data */
-   rmesa->dri.context = driContextPriv;
-   rmesa->dri.screen = sPriv;
-   rmesa->dri.drawable = NULL;
-   rmesa->dri.readable = NULL;
-   rmesa->dri.hwContext = driContextPriv->hHWContext;
-   rmesa->dri.hwLock = &sPriv->pSAREA->lock;
-   rmesa->dri.fd = sPriv->fd;
-   rmesa->dri.drmMinor = sPriv->drm_version.minor;
-
-   rmesa->radeonScreen = screen;
-   rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA +
-				       screen->sarea_priv_offset);
-
-
-   rmesa->dma.buf0_address = rmesa->radeonScreen->buffers->list[0].address;
-
-   (void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) );
-   make_empty_list( & rmesa->swapped );
-
-   rmesa->nr_heaps = screen->numTexHeaps;
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa,
-	    screen->texSize[i],
-	    12,
-	    RADEON_NR_TEX_REGIONS,
-	    (drmTextureRegionPtr)rmesa->sarea->tex_list[i],
-	    & rmesa->sarea->tex_age[i],
-	    & rmesa->swapped,
-	    sizeof( radeonTexObj ),
-	    (destroy_texture_object_t *) radeonDestroyTexObj );
-
-      driSetTextureSwapCounterLocation( rmesa->texture_heaps[i],
-					& rmesa->c_textureSwaps );
+   if (!radeonInitContext(&rmesa->radeon, &functions,
+			  glVisual, driContextPriv,
+			  sharedContextPrivate)) {
+     FREE(rmesa);
+     return GL_FALSE;
    }
-   rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache,
-					   "texture_depth");
-   if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-      rmesa->texture_depth = ( screen->cpp == 4 ) ?
-	 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
 
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->hw.all_dirty = GL_TRUE;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
 
    /* Set the maximum texture size small enough that we can guarentee that
     * all texture units can bind a maximal texture and have all of them in
@@ -306,26 +276,20 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * setting allow larger textures.
     */
 
-   ctx = rmesa->glCtx;
-   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache,
+   ctx = rmesa->radeon.glCtx;
+   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
 						 "texture_units");
    ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
    ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
 
-   i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures");
-
-   driCalculateMaxTextureLevels( rmesa->texture_heaps,
-				 rmesa->nr_heaps,
-				 & ctx->Const,
-				 4,
-				 11, /* max 2D texture size is 2048x2048 */
-				 8,  /* 256^3 */
-				 9,  /* \todo: max cube texture size seems to be 512x512(x6) */
-				 11, /* max rect texture size is 2048x2048. */
-				 12,
-				 GL_FALSE,
-				 i );
+   i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures");
 
+   /* FIXME: When no memory manager is available we should set this 
+    * to some reasonable value based on texture memory pool size */
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
+   ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 
@@ -390,38 +354,42 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    }
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
-   if (rmesa->radeonScreen->drmSupportsCubeMapsR100)
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100)
       _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
-   if (rmesa->glCtx->Mesa_DXTn) {
+   if (rmesa->radeon.glCtx->Mesa_DXTn) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
       _mesa_enable_extension( ctx, "GL_S3_s3tc" );
    }
-   else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) {
+   else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
    }
 
-   if (rmesa->dri.drmMinor >= 9)
+   if (rmesa->radeon.radeonScreen->kernel_mm || rmesa->radeon.dri.drmMinor >= 9)
       _mesa_enable_extension( ctx, "GL_NV_texture_rectangle");
 
+   if (!rmesa->radeon.radeonScreen->kernel_mm)
+      _mesa_disable_extension(ctx, "GL_ARB_occlusion_query");
+
    /* XXX these should really go right after _mesa_init_driver_functions() */
-   radeonInitIoctlFuncs( ctx );
-   radeonInitStateFuncs( ctx );
+   radeon_fbo_init(&rmesa->radeon);
    radeonInitSpanFuncs( ctx );
+   radeonInitIoctlFuncs( ctx );
+   radeonInitStateFuncs( ctx , rmesa->radeon.radeonScreen->kernel_mm );
    radeonInitState( rmesa );
    radeonInitSwtcl( ctx );
 
    _mesa_vector4f_alloc( &rmesa->tcl.ObjClean, 0, 
 			 ctx->Const.MaxArrayLockSize, 32 );
 
-   fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode");
-   rmesa->iw.irq_seq = -1;
-   rmesa->irqsEmitted = 0;
-   rmesa->do_irqs = (rmesa->radeonScreen->irq != 0 &&
-		     fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
-
-   rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
+   fthrottle_mode = driQueryOptioni(&rmesa->radeon.optionCache, "fthrottle_mode");
+   rmesa->radeon.iw.irq_seq = -1;
+   rmesa->radeon.irqsEmitted = 0;
+   rmesa->radeon.do_irqs = (rmesa->radeon.radeonScreen->irq != 0 &&
+			    fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
 
-   (*sPriv->systemTime->getUST)( & rmesa->swap_ust );
+   rmesa->radeon.do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
 
 
 #if DO_DEBUG
@@ -429,206 +397,21 @@ radeonCreateContext( const __GLcontextModes *glVisual,
 				       debug_control );
 #endif
 
-   tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-   if (driQueryOptionb(&rmesa->optionCache, "no_rast")) {
+   tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
+   if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) {
       fprintf(stderr, "disabling 3D acceleration\n");
       FALLBACK(rmesa, RADEON_FALLBACK_DISABLE, 1);
    } else if (tcl_mode == DRI_CONF_TCL_SW ||
-	      !(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
-      if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-	 rmesa->radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	      !(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+	 rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
 	 fprintf(stderr, "Disabling HW TCL support\n");
       }
-      TCL_FALLBACK(rmesa->glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
+      TCL_FALLBACK(rmesa->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
-   if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+   if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
 /*       _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
    }
    return GL_TRUE;
 }
-
-
-/* Destroy the device specific context.
- */
-/* Destroy the Mesa and driver specific context data.
- */
-void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = (radeonContextPtr) driContextPriv->driverPrivate;
-   radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
-
-   /* check if we're deleting the currently bound context */
-   if (rmesa == current) {
-      RADEON_FIREVERTICES( rmesa );
-      _mesa_make_current(NULL, NULL, NULL);
-   }
-
-   /* Free radeon context resources */
-   assert(rmesa); /* should never be null */
-   if ( rmesa ) {
-      GLboolean   release_texture_heaps;
-
-
-      release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
-      _swsetup_DestroyContext( rmesa->glCtx );
-      _tnl_DestroyContext( rmesa->glCtx );
-      _vbo_DestroyContext( rmesa->glCtx );
-      _swrast_DestroyContext( rmesa->glCtx );
-
-      radeonDestroySwtcl( rmesa->glCtx );
-      radeonReleaseArrays( rmesa->glCtx, ~0 );
-      if (rmesa->dma.current.buf) {
-	 radeonReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-	 radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-      }
-
-      _mesa_vector4f_free( &rmesa->tcl.ObjClean );
-
-      if (rmesa->state.scissor.pClipRects) {
-	 FREE(rmesa->state.scissor.pClipRects);
-	 rmesa->state.scissor.pClipRects = NULL;
-      }
-
-      if ( release_texture_heaps ) {
-         /* This share group is about to go away, free our private
-          * texture object data.
-          */
-         int i;
-
-         for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-	    driDestroyTextureHeap( rmesa->texture_heaps[ i ] );
-	    rmesa->texture_heaps[ i ] = NULL;
-         }
-
-	 assert( is_empty_list( & rmesa->swapped ) );
-      }
-
-      /* free the Mesa context */
-      rmesa->glCtx->DriverCtx = NULL;
-      _mesa_destroy_context( rmesa->glCtx );
-
-      /* free the option cache */
-      driDestroyOptionCache (&rmesa->optionCache);
-
-      FREE( rmesa );
-   }
-}
-
-
-
-
-void
-radeonSwapBuffers( __DRIdrawablePrivate *dPriv )
-{
-
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      radeonContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-
-         if ( rmesa->doPageFlip ) {
-            radeonPageFlip( dPriv );
-         }
-         else {
-	     radeonCopyBuffer( dPriv, NULL );
-         }
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-			 int x, int y, int w, int h )
-{
-    if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-	radeonContextPtr radeon;
-	GLcontext *ctx;
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-	ctx = radeon->glCtx;
-
-	if (ctx->Visual.doubleBufferMode) {
-	    drm_clip_rect_t rect;
-	    rect.x1 = x + dPriv->x;
-	    rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	    rect.x2 = rect.x1 + w;
-	    rect.y2 = rect.y1 + h;
-	    _mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-	    radeonCopyBuffer(dPriv, &rect);
-	}
-    } else {
-	/* XXX this shouldn't be an error but we can't handle it for now */
-	_mesa_problem(NULL, "%s: drawable has no context!",
-		      __FUNCTION__);
-    }
-}
-
-/* Make context `c' the current context and bind it to the given
- * drawing and reading surfaces.
- */
-GLboolean
-radeonMakeCurrent( __DRIcontextPrivate *driContextPriv,
-                   __DRIdrawablePrivate *driDrawPriv,
-                   __DRIdrawablePrivate *driReadPriv )
-{
-   if ( driContextPriv ) {
-      radeonContextPtr newCtx = 
-	 (radeonContextPtr) driContextPriv->driverPrivate;
-
-      if (RADEON_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *) newCtx->glCtx);
-
-      newCtx->dri.readable = driReadPriv;
-
-      if ( (newCtx->dri.drawable != driDrawPriv) ||
-           newCtx->lastStamp != driDrawPriv->lastStamp ) {
-	 if (driDrawPriv->swap_interval == (unsigned)-1) {
-	    driDrawPriv->vblFlags = (newCtx->radeonScreen->irq != 0)
-	       ? driGetDefaultVBlankFlags(&newCtx->optionCache)
-	       : VBLANK_FLAG_NO_IRQ;
-
-	    driDrawableInitVBlank( driDrawPriv );
-	 }
-
-	 newCtx->dri.drawable = driDrawPriv;
-
-	 radeonSetCliprects(newCtx);
-	 radeonUpdateViewportOffset( newCtx->glCtx );
-      }
-
-      _mesa_make_current( newCtx->glCtx,
-			  (GLframebuffer *) driDrawPriv->driverPrivate,
-			  (GLframebuffer *) driReadPriv->driverPrivate );
-
-      _mesa_update_state( newCtx->glCtx );
-   } else {
-      if (RADEON_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-      _mesa_make_current( NULL, NULL, NULL );
-   }
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "End %s\n", __FUNCTION__);
-   return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean
-radeonUnbindContext( __DRIcontextPrivate *driContextPriv )
-{
-   radeonContextPtr rmesa = (radeonContextPtr) driContextPriv->driverPrivate;
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *) rmesa->glCtx);
-
-   return GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h
index 53df766f8c..4e2c52c835 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_context.h
@@ -48,91 +48,23 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "radeon_drm.h"
 #include "texmem.h"
-
 #include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/colormac.h"
-
-struct radeon_context;
-typedef struct radeon_context radeonContextRec;
-typedef struct radeon_context *radeonContextPtr;
-
-/* This union is used to avoid warnings/miscompilation
-   with float to uint32_t casts due to strict-aliasing */
-typedef union {
-	GLfloat f;
-	uint32_t ui32;
-} float_ui32_type;
-
-#include "radeon_lock.h"
 #include "radeon_screen.h"
-#include "main/mm.h"
-
-#include "math/m_vector.h"
-
-#define TEX_0   0x1
-#define TEX_1   0x2
-#define TEX_2   0x4
-#define TEX_ALL 0x7
-
-/* Rasterizing fallbacks */
-/* See correponding strings in r200_swtcl.c */
-#define RADEON_FALLBACK_TEXTURE		0x0001
-#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
-#define RADEON_FALLBACK_STENCIL		0x0004
-#define RADEON_FALLBACK_RENDER_MODE	0x0008
-#define RADEON_FALLBACK_BLEND_EQ	0x0010
-#define RADEON_FALLBACK_BLEND_FUNC	0x0020
-#define RADEON_FALLBACK_DISABLE 	0x0040
-#define RADEON_FALLBACK_BORDER_MODE	0x0080
-
-/* The blit width for texture uploads
- */
-#define BLIT_WIDTH_BYTES 1024
 
-/* Use the templated vertex format:
- */
-#define COLOR_IS_RGBA
-#define TAG(x) radeon##x
-#include "tnl_dd/t_dd_vertex.h"
-#undef TAG
-
-typedef void (*radeon_tri_func) (radeonContextPtr,
-				 radeonVertex *,
-				 radeonVertex *, radeonVertex *);
-
-typedef void (*radeon_line_func) (radeonContextPtr,
-				  radeonVertex *, radeonVertex *);
+#include "radeon_common.h"
 
-typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *);
-
-struct radeon_colorbuffer_state {
-	GLuint clear;
-	int roundEnable;
-};
 
-struct radeon_depthbuffer_state {
-	GLuint clear;
-	GLfloat scale;
-};
+struct r100_context;
+typedef struct r100_context r100ContextRec;
+typedef struct r100_context *r100ContextPtr;
 
-struct radeon_scissor_state {
-	drm_clip_rect_t rect;
-	GLboolean enabled;
+#include "radeon_lock.h"
 
-	GLuint numClipRects;	/* Cliprects active */
-	GLuint numAllocedClipRects;	/* Cliprects available */
-	drm_clip_rect_t *pClipRects;
-};
 
-struct radeon_stencilbuffer_state {
-	GLboolean hwBuffer;
-	GLuint clear;		/* rb3d_stencilrefmask value */
-};
 
-struct radeon_stipple_state {
-	GLuint mask[32];
-};
+#define R100_TEX_ALL 0x7
 
 /* used for both tcl_vtx and vc_frmt tex bits (they are identical) */
 #define RADEON_ST_BIT(unit) \
@@ -141,42 +73,6 @@ struct radeon_stipple_state {
 #define RADEON_Q_BIT(unit) \
 (unit == 0 ? RADEON_CP_VC_FRMT_Q0 : (RADEON_CP_VC_FRMT_Q1 >> 2) << (2 * unit))
 
-typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr;
-
-/* Texture object in locally shared texture space.
- */
-struct radeon_tex_obj {
-	driTextureObject base;
-
-	GLuint bufAddr;		/* Offset to start of locally
-				   shared texture block */
-
-	GLuint dirty_state;	/* Flags (1 per texunit) for
-				   whether or not this texobj
-				   has dirty hardware state
-				   (pp_*) that needs to be
-				   brought into the
-				   texunit. */
-
-	drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
-	/* Six, for the cube faces */
-
-	GLboolean image_override; /* Image overridden by GLX_EXT_tfp */
-
-	GLuint pp_txfilter;	/* hardware register values */
-	GLuint pp_txformat;
-	GLuint pp_txoffset;	/* Image location in texmem.
-				   All cube faces follow. */
-	GLuint pp_txsize;	/* npot only */
-	GLuint pp_txpitch;	/* npot only */
-	GLuint pp_border_color;
-	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
-
-	GLboolean border_fallback;
-
-	GLuint tile_bits;	/* hw texture tile bits used on this texture */
-};
-
 struct radeon_texture_env_state {
 	radeonTexObjPtr texobj;
 	GLenum format;
@@ -187,17 +83,6 @@ struct radeon_texture_state {
 	struct radeon_texture_env_state unit[RADEON_MAX_TEXTURE_UNITS];
 };
 
-struct radeon_state_atom {
-	struct radeon_state_atom *next, *prev;
-	const char *name;	/* for debug */
-	int cmd_size;		/* size in bytes */
-	GLuint is_tcl;
-	int *cmd;		/* one or more cmd's */
-	int *lastcmd;		/* one or more cmd's */
-	GLboolean dirty;	/* dirty-mark in emit_state_list */
-	 GLboolean(*check) (GLcontext *);	/* is this state active? */
-};
-
 /* Trying to keep these relatively short as the variables are becoming
  * extravagently long.  Drop the driver name prefix off the front of
  * everything - I think we know which driver we're in by now, and keep the
@@ -410,10 +295,16 @@ struct radeon_state_atom {
 #define SHN_SHININESS      1
 #define SHN_STATE_SIZE     2
 
-struct radeon_hw_state {
-	/* Head of the linked list of state atoms. */
-	struct radeon_state_atom atomlist;
+#define R100_QUERYOBJ_CMD_0  0
+#define R100_QUERYOBJ_DATA_0 1
+#define R100_QUERYOBJ_CMDSIZE  2
+
+#define STP_CMD_0 0
+#define STP_DATA_0 1
+#define STP_CMD_1 2
+#define STP_STATE_SIZE 35
 
+struct r100_hw_state {
 	/* Hardware state, stored as cmdbuf commands:  
 	 *   -- Need to doublebuffer for
 	 *           - eliding noop statechange loops? (except line stipple count)
@@ -437,90 +328,19 @@ struct radeon_hw_state {
 	struct radeon_state_atom fog;
 	struct radeon_state_atom glt;
 	struct radeon_state_atom txr[3];	/* for NPOT */
-
-	int max_state_size;	/* Number of bytes necessary for a full state emit. */
-	GLboolean is_dirty, all_dirty;
-};
-
-struct radeon_state {
-	/* Derived state for internal purposes:
-	 */
-	struct radeon_colorbuffer_state color;
-	struct radeon_depthbuffer_state depth;
-	struct radeon_scissor_state scissor;
-	struct radeon_stencilbuffer_state stencil;
-	struct radeon_stipple_state stipple;
-	struct radeon_texture_state texture;
-};
-
-/* Need refcounting on dma buffers:
- */
-struct radeon_dma_buffer {
-	int refcount;		/* the number of retained regions in buf */
-	drmBufPtr buf;
-};
-
-#define GET_START(rvb) (rmesa->radeonScreen->gart_buffer_offset +			\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-
-/* A retained region, eg vertices for indexed vertices.
- */
-struct radeon_dma_region {
-	struct radeon_dma_buffer *buf;
-	char *address;		/* == buf->address */
-	int start, end, ptr;	/* offsets from start of buf */
-	int aos_start;
-	int aos_stride;
-	int aos_size;
-};
-
-struct radeon_dma {
-	/* Active dma region.  Allocations for vertices and retained
-	 * regions come from here.  Also used for emitting random vertices,
-	 * these may be flushed by calling flush_current();
-	 */
-	struct radeon_dma_region current;
-
-	void (*flush) (radeonContextPtr);
-
-	char *buf0_address;	/* start of buf[0], for index calcs */
-	GLuint nr_released_bufs;	/* flush after so many buffers released */
+	struct radeon_state_atom stp;
 };
 
-struct radeon_dri_mirror {
-	__DRIcontextPrivate *context;	/* DRI context */
-	__DRIscreenPrivate *screen;	/* DRI screen */
-
-   /**
-    * DRI drawable bound to this context for drawing.
-    */
-	__DRIdrawablePrivate *drawable;
 
-   /**
-    * DRI drawable bound to this context for reading.
-    */
-	__DRIdrawablePrivate *readable;
-
-	drm_context_t hwContext;
-	drm_hw_lock_t *hwLock;
-	int fd;
-	int drmMinor;
+struct r100_state {
+	struct radeon_texture_state texture;
 };
 
 #define RADEON_CMD_BUF_SZ  (8*1024)
-
-struct radeon_store {
-	GLuint statenr;
-	GLuint primnr;
-	char cmd_buf[RADEON_CMD_BUF_SZ];
-	int cmd_used;
-	int elts_start;
-};
-
+#define R200_ELT_BUF_SZ  (8*1024)
 /* radeon_tcl.c
  */
-struct radeon_tcl_info {
+struct r100_tcl_info {
 	GLuint vertex_format;
 	GLuint hw_primitive;
 
@@ -529,30 +349,18 @@ struct radeon_tcl_info {
 	 */
 	GLvector4f ObjClean;
 
-	struct radeon_dma_region *aos_components[8];
-	GLuint nr_aos_components;
-
 	GLuint *Elts;
 
-	struct radeon_dma_region indexed_verts;
-	struct radeon_dma_region obj;
-	struct radeon_dma_region rgba;
-	struct radeon_dma_region spec;
-	struct radeon_dma_region fog;
-	struct radeon_dma_region tex[RADEON_MAX_TEXTURE_UNITS];
-	struct radeon_dma_region norm;
+        int elt_cmd_offset;
+	int elt_cmd_start;
+        int elt_used;
 };
 
 /* radeon_swtcl.c
  */
-struct radeon_swtcl_info {
-	GLuint RenderIndex;
-	GLuint vertex_size;
+struct r100_swtcl_info {
 	GLuint vertex_format;
 
-	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-	GLuint vertex_attr_count;
-
 	GLubyte *verts;
 
 	/* Fallback rasterization functions
@@ -561,10 +369,6 @@ struct radeon_swtcl_info {
 	radeon_line_func draw_line;
 	radeon_tri_func draw_tri;
 
-	GLuint hw_primitive;
-	GLenum render_primitive;
-	GLuint numverts;
-
    /**
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
@@ -576,22 +380,9 @@ struct radeon_swtcl_info {
 	GLuint specoffset;
 
 	GLboolean needproj;
-
-	struct radeon_dma_region indexed_verts;
 };
 
-struct radeon_ioctl {
-	GLuint vertex_offset;
-	GLuint vertex_size;
-};
 
-#define RADEON_MAX_PRIMS 64
-
-struct radeon_prim {
-	GLuint start;
-	GLuint end;
-	GLuint prim;
-};
 
 /* A maximum total of 20 elements per vertex:  3 floats for position, 3
  * floats for normal, 4 floats for color, 4 bytes for secondary color,
@@ -602,59 +393,18 @@ struct radeon_prim {
  */
 #define RADEON_MAX_VERTEX_SIZE 20
 
-struct radeon_context {
-	GLcontext *glCtx;	/* Mesa context */
+struct r100_context {
+        struct radeon_context radeon;
 
 	/* Driver and hardware state management
 	 */
-	struct radeon_hw_state hw;
-	struct radeon_state state;
-
-	/* Texture object bookkeeping
-	 */
-	unsigned nr_heaps;
-	driTexHeap *texture_heaps[RADEON_NR_TEX_HEAPS];
-	driTextureObject swapped;
-	int texture_depth;
-	float initialMaxAnisotropy;
-
-	/* Rasterization and vertex state:
-	 */
-	GLuint TclFallback;
-	GLuint Fallback;
-	GLuint NewGLState;
-	 DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
+	struct r100_hw_state hw;
+	struct r100_state state;
 
 	/* Vertex buffers
 	 */
 	struct radeon_ioctl ioctl;
-	struct radeon_dma dma;
 	struct radeon_store store;
-	/* A full state emit as of the first state emit in the main store, in case
-	 * the context is lost.
-	 */
-	struct radeon_store backup_store;
-
-	/* Page flipping
-	 */
-	GLuint doPageFlip;
-
-	/* Busy waiting
-	 */
-	GLuint do_usleeps;
-	GLuint do_irqs;
-	GLuint irqsEmitted;
-	drm_radeon_irq_wait_t iw;
-
-	/* Drawable, cliprect and scissor information
-	 */
-	GLuint numClipRects;	/* Cliprects for the draw buffer */
-	drm_clip_rect_t *pClipRects;
-	unsigned int lastStamp;
-	GLboolean lost_context;
-	GLboolean save_on_next_emit;
-	radeonScreenPtr radeonScreen;	/* Screen private DRI data */
-	drm_radeon_sarea_t *sarea;	/* Private SAREA data */
 
 	/* TCL stuff
 	 */
@@ -667,29 +417,13 @@ struct radeon_context {
 	GLmatrix tmpmat[RADEON_MAX_TEXTURE_UNITS];
 	GLuint last_ReallyEnabled;
 
-	/* VBI
-	 */
-	int64_t swap_ust;
-	int64_t swap_missed_ust;
-
-	GLuint swap_count;
-	GLuint swap_missed_count;
-
 	/* radeon_tcl.c
 	 */
-	struct radeon_tcl_info tcl;
+	struct r100_tcl_info tcl;
 
 	/* radeon_swtcl.c
 	 */
-	struct radeon_swtcl_info swtcl;
-
-	/* Mirrors of some DRI state
-	 */
-	struct radeon_dri_mirror dri;
-
-	/* Configuration cache
-	 */
-	driOptionCache optionCache;
+	struct r100_swtcl_info swtcl;
 
 	GLboolean using_hyperz;
 	GLboolean texmicrotile;
@@ -703,61 +437,19 @@ struct radeon_context {
 	GLuint c_textureSwaps;
 	GLuint c_textureBytes;
 	GLuint c_vertexBuffers;
+
 };
 
-#define RADEON_CONTEXT(ctx)		((radeonContextPtr)(ctx->DriverCtx))
-
-static INLINE GLuint radeonPackColor(GLuint cpp,
-                                     GLubyte r, GLubyte g,
-                                     GLubyte b, GLubyte a)
-{
-	switch (cpp) {
-	case 2:
-		return PACK_COLOR_565(r, g, b);
-	case 4:
-		return PACK_COLOR_8888(a, r, g, b);
-	default:
-		return 0;
-	}
-}
+
+#define R100_CONTEXT(ctx)		((r100ContextPtr)(ctx->DriverCtx))
+
 
 #define RADEON_OLD_PACKETS 1
 
-extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv);
-extern GLboolean radeonCreateContext(const __GLcontextModes * glVisual,
-				     __DRIcontextPrivate * driContextPriv,
-				     void *sharedContextPrivate);
-extern void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
-extern void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-				int x, int y, int w, int h);
-extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-				   __DRIdrawablePrivate * driDrawPriv,
-				   __DRIdrawablePrivate * driReadPriv);
-extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
-
-/* ================================================================
- * Debugging:
- */
-#define DO_DEBUG		1
-
-#if DO_DEBUG
-extern int RADEON_DEBUG;
-#else
-#define RADEON_DEBUG		0
-#endif
-
-#define DEBUG_TEXTURE	0x0001
-#define DEBUG_STATE	0x0002
-#define DEBUG_IOCTL	0x0004
-#define DEBUG_PRIMS	0x0008
-#define DEBUG_VERTS	0x0010
-#define DEBUG_FALLBACKS	0x0020
-#define DEBUG_VFMT	0x0040
-#define DEBUG_CODEGEN	0x0080
-#define DEBUG_VERBOSE	0x0100
-#define DEBUG_DRI       0x0200
-#define DEBUG_DMA       0x0400
-#define DEBUG_SANITY    0x0800
-#define DEBUG_SYNC      0x1000
+extern GLboolean r100CreateContext( const __GLcontextModes *glVisual,
+				    __DRIcontextPrivate *driContextPriv,
+				    void *sharedContextPrivate);
+  
+
 
 #endif				/* __RADEON_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
new file mode 100644
index 0000000000..ab4eca31a3
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
@@ -0,0 +1,246 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_CS_H
+#define RADEON_CS_H
+
+#include <stdint.h>
+#include <string.h>
+#include "drm.h"
+#include "radeon_drm.h"
+
+struct radeon_cs_reloc {
+    struct radeon_bo    *bo;
+    uint32_t            read_domain;
+    uint32_t            write_domain;
+    uint32_t            flags;
+};
+
+
+#define RADEON_CS_SPACE_OK 0
+#define RADEON_CS_SPACE_OP_TO_BIG 1
+#define RADEON_CS_SPACE_FLUSH 2
+
+struct radeon_cs_space_check {
+    struct radeon_bo *bo;
+    uint32_t read_domains;
+    uint32_t write_domain;
+    uint32_t new_accounted;
+};
+
+#define MAX_SPACE_BOS (32)
+
+struct radeon_cs_manager;
+
+struct radeon_cs {
+    struct radeon_cs_manager    *csm;
+    void                        *relocs;
+    uint32_t                    *packets;
+    unsigned                    crelocs;
+    unsigned                    relocs_total_size;
+    unsigned                    cdw;
+    unsigned                    ndw;
+    int                         section;
+    unsigned                    section_ndw;
+    unsigned                    section_cdw;
+    const char                  *section_file;
+    const char                  *section_func;
+    int                         section_line;
+    struct radeon_cs_space_check bos[MAX_SPACE_BOS];
+    int                         bo_count;
+    void                        (*space_flush_fn)(void *);
+    void                        *space_flush_data;
+};
+
+/* cs functions */
+struct radeon_cs_funcs {
+    struct radeon_cs *(*cs_create)(struct radeon_cs_manager *csm,
+                                   uint32_t ndw);
+    int (*cs_write_reloc)(struct radeon_cs *cs,
+                          struct radeon_bo *bo,
+                          uint32_t read_domain,
+                          uint32_t write_domain,
+                          uint32_t flags);
+    int (*cs_begin)(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line);
+    int (*cs_end)(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line);
+    int (*cs_emit)(struct radeon_cs *cs);
+    int (*cs_destroy)(struct radeon_cs *cs);
+    int (*cs_erase)(struct radeon_cs *cs);
+    int (*cs_need_flush)(struct radeon_cs *cs);
+    void (*cs_print)(struct radeon_cs *cs, FILE *file);
+};
+
+struct radeon_cs_manager {
+    struct radeon_cs_funcs  *funcs;
+    int                     fd;
+    int32_t vram_limit, gart_limit;
+    int32_t vram_write_used, gart_write_used;
+    int32_t read_used;
+};
+
+static inline struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm,
+                                                 uint32_t ndw)
+{
+    return csm->funcs->cs_create(csm, ndw);
+}
+
+static inline int radeon_cs_write_reloc(struct radeon_cs *cs,
+                                        struct radeon_bo *bo,
+                                        uint32_t read_domain,
+                                        uint32_t write_domain,
+                                        uint32_t flags)
+{
+    return cs->csm->funcs->cs_write_reloc(cs,
+                                          bo,
+                                          read_domain,
+                                          write_domain,
+                                          flags);
+}
+
+static inline int radeon_cs_begin(struct radeon_cs *cs,
+                                  uint32_t ndw,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    return cs->csm->funcs->cs_begin(cs, ndw, file, func, line);
+}
+
+static inline int radeon_cs_end(struct radeon_cs *cs,
+                                const char *file,
+                                const char *func,
+                                int line)
+{
+    return cs->csm->funcs->cs_end(cs, file, func, line);
+}
+
+static inline int radeon_cs_emit(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_emit(cs);
+}
+
+static inline int radeon_cs_destroy(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_destroy(cs);
+}
+
+static inline int radeon_cs_erase(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_erase(cs);
+}
+
+static inline int radeon_cs_need_flush(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_need_flush(cs);
+}
+
+static inline void radeon_cs_print(struct radeon_cs *cs, FILE *file)
+{
+    cs->csm->funcs->cs_print(cs, file);
+}
+
+static inline void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit)
+{
+    
+    if (domain == RADEON_GEM_DOMAIN_VRAM)
+	cs->csm->vram_limit = limit;
+    else
+	cs->csm->gart_limit = limit;
+}
+
+static inline void radeon_cs_write_dword(struct radeon_cs *cs, uint32_t dword)
+{
+    cs->packets[cs->cdw++] = dword;
+    if (cs->section) {
+        cs->section_cdw++;
+    }
+}
+
+static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword)
+{
+
+    memcpy(cs->packets + cs->cdw, &qword, sizeof(qword));
+    cs->cdw+=2;
+    if (cs->section) {
+        cs->section_cdw+=2;
+    }
+}
+
+static inline void radeon_cs_write_table(struct radeon_cs *cs, void *data, uint32_t size)
+{
+    memcpy(cs->packets + cs->cdw, data, size * 4);
+    cs->cdw += size;
+    if (cs->section) {
+	    cs->section_cdw += size;
+    }
+}
+
+static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data)
+{
+    cs->space_flush_fn = fn;
+    cs->space_flush_data = data;
+}
+
+
+/*
+ * add a persistent BO to the list
+ * a persistent BO is one that will be referenced across flushes,
+ * i.e. colorbuffer, textures etc.
+ * They get reset when a new "operation" happens, where an operation
+ * is a state emission with a color/textures etc followed by a bunch of vertices.
+ */
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs,
+				       struct radeon_bo *bo,
+				       uint32_t read_domains,
+				       uint32_t write_domain);
+
+/* reset the persistent BO list */
+void radeon_cs_space_reset_bos(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list */
+int radeon_cs_space_check(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list and a temporary BO
+ * a temporary BO is like a DMA buffer, which  gets flushed with the
+ * command buffer */
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains,
+				  uint32_t write_domain);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
new file mode 100644
index 0000000000..f1addb299e
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
@@ -0,0 +1,409 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <errno.h>
+
+#include "radeon_bocs_wrapper.h"
+#include "radeon_common.h"
+
+struct cs_manager_legacy {
+    struct radeon_cs_manager    base;
+    struct radeon_context       *ctx;
+    /* hack for scratch stuff */
+    uint32_t                    pending_age;
+    uint32_t                    pending_count;
+
+
+};
+
+struct cs_reloc_legacy {
+    struct radeon_cs_reloc  base;
+    uint32_t                cindices;
+    uint32_t                *indices;
+};
+
+
+static struct radeon_cs *cs_create(struct radeon_cs_manager *csm,
+                                   uint32_t ndw)
+{
+    struct radeon_cs *cs;
+
+    cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs));
+    if (cs == NULL) {
+        return NULL;
+    }
+    cs->csm = csm;
+    cs->ndw = (ndw + 0x3FF) & (~0x3FF);
+    cs->packets = (uint32_t*)malloc(4*cs->ndw);
+    if (cs->packets == NULL) {
+        free(cs);
+        return NULL;
+    }
+    cs->relocs_total_size = 0;
+    return cs;
+}
+
+static int cs_write_reloc(struct radeon_cs *cs,
+                          struct radeon_bo *bo,
+                          uint32_t read_domain,
+                          uint32_t write_domain,
+                          uint32_t flags)
+{
+    struct cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+    /* check domains */
+    if ((read_domain && write_domain) || (!read_domain && !write_domain)) {
+        /* in one CS a bo can only be in read or write domain but not
+         * in read & write domain at the same sime
+         */
+        return -EINVAL;
+    }
+    if (read_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    if (write_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    /* check if bo is already referenced */
+    for(i = 0; i < cs->crelocs; i++) {
+        uint32_t *indices;
+
+        if (relocs[i].base.bo->handle == bo->handle) {
+            /* Check domains must be in read or write. As we check already
+             * checked that in argument one of the read or write domain was
+             * set we only need to check that if previous reloc as the read
+             * domain set then the read_domain should also be set for this
+             * new relocation.
+             */
+            if (relocs[i].base.read_domain && !read_domain) {
+                return -EINVAL;
+            }
+            if (relocs[i].base.write_domain && !write_domain) {
+                return -EINVAL;
+            }
+            relocs[i].base.read_domain |= read_domain;
+            relocs[i].base.write_domain |= write_domain;
+            /* save indice */
+            relocs[i].cindices++;
+            indices = (uint32_t*)realloc(relocs[i].indices,
+                                         relocs[i].cindices * 4);
+            if (indices == NULL) {
+                relocs[i].cindices -= 1;
+                return -ENOMEM;
+            }
+            relocs[i].indices = indices;
+            relocs[i].indices[relocs[i].cindices - 1] = cs->cdw - 1;
+            return 0;
+        }
+    }
+    /* add bo to reloc */
+    relocs = (struct cs_reloc_legacy*)
+             realloc(cs->relocs,
+                     sizeof(struct cs_reloc_legacy) * (cs->crelocs + 1));
+    if (relocs == NULL) {
+        return -ENOMEM;
+    }
+    cs->relocs = relocs;
+    relocs[cs->crelocs].base.bo = bo;
+    relocs[cs->crelocs].base.read_domain = read_domain;
+    relocs[cs->crelocs].base.write_domain = write_domain;
+    relocs[cs->crelocs].base.flags = flags;
+    relocs[cs->crelocs].indices = (uint32_t*)malloc(4);
+    if (relocs[cs->crelocs].indices == NULL) {
+        return -ENOMEM;
+    }
+    relocs[cs->crelocs].indices[0] = cs->cdw - 1;
+    relocs[cs->crelocs].cindices = 1;
+    cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo);
+    cs->crelocs++;
+    radeon_bo_ref(bo);
+    return 0;
+}
+
+static int cs_begin(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line)
+{
+    if (cs->section) {
+        fprintf(stderr, "CS already in a section(%s,%s,%d)\n",
+                cs->section_file, cs->section_func, cs->section_line);
+        fprintf(stderr, "CS can't start section(%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 1;
+    cs->section_ndw = ndw;
+    cs->section_cdw = 0;
+    cs->section_file = file;
+    cs->section_func = func;
+    cs->section_line = line;
+
+
+    if (cs->cdw + ndw > cs->ndw) {
+        uint32_t tmp, *ptr;
+	int num = (ndw > 0x3FF) ? ndw : 0x3FF;
+
+        tmp = (cs->cdw + 1 + num) & (~num);
+        ptr = (uint32_t*)realloc(cs->packets, 4 * tmp);
+        if (ptr == NULL) {
+            return -ENOMEM;
+        }
+        cs->packets = ptr;
+        cs->ndw = tmp;
+    }
+
+    return 0;
+}
+
+static int cs_end(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line)
+
+{
+    if (!cs->section) {
+        fprintf(stderr, "CS no section to end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 0;
+    if (cs->section_ndw != cs->section_cdw) {
+        fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n",
+                cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw);
+        fprintf(stderr, "CS section end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    return 0;
+}
+
+static int cs_process_relocs(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    struct cs_reloc_legacy *relocs;
+    int i, j, r;
+
+    csm = (struct cs_manager_legacy*)cs->csm;
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+restart:
+    for (i = 0; i < cs->crelocs; i++) 
+    {
+        for (j = 0; j < relocs[i].cindices; j++) 
+        {
+            uint32_t soffset, eoffset;
+
+            r = radeon_bo_legacy_validate(relocs[i].base.bo,
+                                           &soffset, &eoffset);
+	        if (r == -EAGAIN)
+            {
+	             goto restart;
+            }
+            if (r) 
+            {
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                return r;
+            }
+            cs->packets[relocs[i].indices[j]] += soffset;
+            if (cs->packets[relocs[i].indices[j]] >= eoffset) 
+            {
+	      /*                radeon_bo_debug(relocs[i].base.bo, 12); */
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                fprintf(stderr, "above end: %p 0x%08X 0x%08X\n",
+                        relocs[i].base.bo,
+                        cs->packets[relocs[i].indices[j]],
+                        eoffset);
+                exit(0);
+                return -EINVAL;
+            }
+        }
+    }
+    return 0;
+}
+
+static int cs_set_age(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    struct cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+    for (i = 0; i < cs->crelocs; i++) {
+        radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age);
+        radeon_bo_unref(relocs[i].base.bo);
+    }
+    return 0;
+}
+
+static int cs_emit(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    drm_radeon_cmd_buffer_t cmd;
+    drm_r300_cmd_header_t age;
+    uint64_t ull;
+    int r;
+
+    csm->ctx->vtbl.emit_cs_header(cs, csm->ctx);
+
+    /* append buffer age */
+    if ( IS_R300_CLASS(csm->ctx->radeonScreen) )
+    { 
+      age.scratch.cmd_type = R300_CMD_SCRATCH;
+      /* Scratch register 2 corresponds to what radeonGetAge polls */
+      csm->pending_age = 0;
+      csm->pending_count = 1;
+      ull = (uint64_t) (intptr_t) &csm->pending_age;
+      age.scratch.reg = 2;
+      age.scratch.n_bufs = 1;
+      age.scratch.flags = 0;
+      radeon_cs_write_dword(cs, age.u);
+      radeon_cs_write_qword(cs, ull);
+      radeon_cs_write_dword(cs, 0);
+    }
+
+    r = cs_process_relocs(cs);
+    if (r) {
+        return 0;
+    }
+
+    cmd.buf = (char *)cs->packets;
+    cmd.bufsz = cs->cdw * 4;
+    if (csm->ctx->state.scissor.enabled) {
+        cmd.nbox = csm->ctx->state.scissor.numClipRects;
+        cmd.boxes = (drm_clip_rect_t *) csm->ctx->state.scissor.pClipRects;
+    } else {
+        cmd.nbox = csm->ctx->numClipRects;
+        cmd.boxes = (drm_clip_rect_t *) csm->ctx->pClipRects;
+    }
+
+    //dump_cmdbuf(cs);
+
+    r = drmCommandWrite(cs->csm->fd, DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
+    if (r) {
+        return r;
+    }
+    if ((!IS_R300_CLASS(csm->ctx->radeonScreen)) &&
+        (!IS_R600_CLASS(csm->ctx->radeonScreen))) { /* +r6/r7 : No irq for r6/r7 yet. */
+	drm_radeon_irq_emit_t emit_cmd;
+	emit_cmd.irq_seq = (int*)&csm->pending_age;
+	r = drmCommandWrite(cs->csm->fd, DRM_RADEON_IRQ_EMIT, &emit_cmd, sizeof(emit_cmd));
+	if (r) {
+		return r;
+	}
+    }
+    cs_set_age(cs);
+
+    cs->csm->read_used = 0;
+    cs->csm->vram_write_used = 0;
+    cs->csm->gart_write_used = 0;
+    return 0;
+}
+
+static void inline cs_free_reloc(void *relocs_p, int crelocs)
+{
+    struct cs_reloc_legacy *relocs = relocs_p;
+    int i;
+    if (!relocs_p)
+      return;
+    for (i = 0; i < crelocs; i++)
+      free(relocs[i].indices);
+}
+
+static int cs_destroy(struct radeon_cs *cs)
+{
+    cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    free(cs->packets);
+    free(cs);
+    return 0;
+}
+
+static int cs_erase(struct radeon_cs *cs)
+{
+    cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    cs->relocs_total_size = 0;
+    cs->relocs = NULL;
+    cs->crelocs = 0;
+    cs->cdw = 0;
+    cs->section = 0;
+    return 0;
+}
+
+static int cs_need_flush(struct radeon_cs *cs)
+{
+    /* this function used to flush when the BO usage got to
+     * a certain size, now the higher levels handle this better */
+    return 0;
+}
+
+static void cs_print(struct radeon_cs *cs, FILE *file)
+{
+}
+
+static struct radeon_cs_funcs  radeon_cs_legacy_funcs = {
+    cs_create,
+    cs_write_reloc,
+    cs_begin,
+    cs_end,
+    cs_emit,
+    cs_destroy,
+    cs_erase,
+    cs_need_flush,
+    cs_print,
+};
+
+struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx)
+{
+    struct cs_manager_legacy *csm;
+
+    csm = (struct cs_manager_legacy*)
+          calloc(1, sizeof(struct cs_manager_legacy));
+    if (csm == NULL) {
+        return NULL;
+    }
+    csm->base.funcs = &radeon_cs_legacy_funcs;
+    csm->base.fd = ctx->dri.fd;
+    csm->ctx = ctx;
+    csm->pending_age = 1;
+    return (struct radeon_cs_manager*)csm;
+}
+
+void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm)
+{
+    free(csm);
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h
new file mode 100644
index 0000000000..cafbc9e576
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h
@@ -0,0 +1,40 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_CS_LEGACY_H
+#define RADEON_CS_LEGACY_H
+
+struct radeon_context;
+
+struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx);
+void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
new file mode 100644
index 0000000000..89cbbb5a6b
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
@@ -0,0 +1,234 @@
+/* 
+ * Copyright © 2009 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ */
+#include <assert.h>
+#include <errno.h>
+#include <stdlib.h>
+#include "radeon_bocs_wrapper.h"
+
+struct rad_sizes {
+    int32_t op_read;
+    int32_t op_gart_write;
+    int32_t op_vram_write;
+};
+
+static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct rad_sizes *sizes)
+{
+    uint32_t read_domains, write_domain;
+    struct radeon_bo *bo;
+
+    bo = sc->bo;
+    sc->new_accounted = 0;
+    read_domains = sc->read_domains;
+    write_domain = sc->write_domain;
+
+    /* legacy needs a static check */
+    if (radeon_bo_is_static(bo)) {
+	bo->space_accounted = sc->new_accounted = (read_domains << 16) | write_domain;
+	return 0;
+    }
+
+    /* already accounted this bo */
+    if (write_domain && (write_domain == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+    if (read_domains && ((read_domains << 16) == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+
+    if (bo->space_accounted == 0) {
+	if (write_domain == RADEON_GEM_DOMAIN_VRAM)
+	    sizes->op_vram_write += bo->size;
+	else if (write_domain == RADEON_GEM_DOMAIN_GTT)
+	  sizes->op_gart_write += bo->size;
+	else
+	    sizes->op_read += bo->size;
+	sc->new_accounted = (read_domains << 16) | write_domain;
+    } else {
+	uint16_t old_read, old_write;
+	
+	old_read = bo->space_accounted >> 16;
+	old_write = bo->space_accounted & 0xffff;
+	
+	if (write_domain && (old_read & write_domain)) {
+	    sc->new_accounted = write_domain;
+	    /* moving from read to a write domain */
+	    if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
+		sizes->op_read -= bo->size;
+		sizes->op_vram_write += bo->size;
+	    } else if (write_domain == RADEON_GEM_DOMAIN_GTT) {
+		sizes->op_read -= bo->size;
+		sizes->op_gart_write += bo->size;
+	    }
+	} else if (read_domains & old_write) {
+	    sc->new_accounted = bo->space_accounted & 0xffff;
+	} else {
+	    /* rewrite the domains */
+	    if (write_domain != old_write)
+		fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write);
+	    if (read_domains != old_read)
+		fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read);
+	    return RADEON_CS_SPACE_FLUSH;
+	}
+    }
+    return 0;
+}
+
+static int radeon_cs_do_space_check(struct radeon_cs *cs, struct radeon_cs_space_check *new_tmp)
+{
+    struct radeon_cs_manager *csm = cs->csm;
+    int i;
+    struct radeon_bo *bo;
+    struct rad_sizes sizes;
+    int ret;
+
+    /* check the totals for this operation */
+
+    if (cs->bo_count == 0 && !new_tmp)
+	return 0;
+
+    memset(&sizes, 0, sizeof(struct rad_sizes));
+
+    /* prepare */
+    for (i = 0; i < cs->bo_count; i++) {
+	ret = radeon_cs_setup_bo(&cs->bos[i], &sizes);
+	if (ret)
+	    return ret;
+    }
+
+    if (new_tmp) {
+	ret = radeon_cs_setup_bo(new_tmp, &sizes);
+	if (ret)
+	    return ret;
+    }
+	
+    if (sizes.op_read < 0)
+	    sizes.op_read = 0;
+
+    /* check sizes - operation first */
+    if ((sizes.op_read + sizes.op_gart_write > csm->gart_limit) ||
+	(sizes.op_vram_write > csm->vram_limit)) {
+	    return RADEON_CS_SPACE_OP_TO_BIG;
+    }
+    
+    if (((csm->vram_write_used + sizes.op_vram_write) > csm->vram_limit) ||
+	((csm->read_used + csm->gart_write_used + sizes.op_gart_write + sizes.op_read) > csm->gart_limit)) {
+	    return RADEON_CS_SPACE_FLUSH;
+    }
+    
+    csm->gart_write_used += sizes.op_gart_write;
+    csm->vram_write_used += sizes.op_vram_write;
+    csm->read_used += sizes.op_read;
+    /* commit */
+    for (i = 0; i < cs->bo_count; i++) {
+	    bo = cs->bos[i].bo;
+	    bo->space_accounted = cs->bos[i].new_accounted;
+    }
+    if (new_tmp)
+	new_tmp->bo->space_accounted = new_tmp->new_accounted;
+    
+    return RADEON_CS_SPACE_OK;
+}
+
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	if (cs->bos[i].bo == bo &&
+	    cs->bos[i].read_domains == read_domains &&
+	    cs->bos[i].write_domain == write_domain)
+	    return;
+    }
+    radeon_bo_ref(bo);
+    i = cs->bo_count;
+    cs->bos[i].bo = bo;
+    cs->bos[i].read_domains = read_domains;
+    cs->bos[i].write_domain = write_domain;
+    cs->bos[i].new_accounted = 0;
+    cs->bo_count++;
+
+    assert(cs->bo_count < MAX_SPACE_BOS);
+}
+
+static int radeon_cs_check_space_internal(struct radeon_cs *cs, struct radeon_cs_space_check *tmp_bo)
+{
+    int ret;
+    int flushed = 0;
+
+again:
+    ret = radeon_cs_do_space_check(cs, tmp_bo);
+    if (ret == RADEON_CS_SPACE_OP_TO_BIG)
+	return -1;
+    if (ret == RADEON_CS_SPACE_FLUSH) {
+	(*cs->space_flush_fn)(cs->space_flush_data);
+	if (flushed)
+	    return -1;
+	flushed = 1;
+	goto again;
+    }
+    return 0;
+}
+
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains, uint32_t write_domain)
+{									
+    struct radeon_cs_space_check temp_bo;
+    int ret = 0;
+
+    if (bo) {
+	temp_bo.bo = bo;
+	temp_bo.read_domains = read_domains;
+	temp_bo.write_domain = write_domain;
+	temp_bo.new_accounted = 0;
+    }
+
+    ret = radeon_cs_check_space_internal(cs, bo ? &temp_bo : NULL);
+    return ret;
+}
+
+int radeon_cs_space_check(struct radeon_cs *cs)
+{
+    return radeon_cs_check_space_internal(cs, NULL);
+}
+
+void radeon_cs_space_reset_bos(struct radeon_cs *cs)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	radeon_bo_unref(cs->bos[i].bo);
+	cs->bos[i].bo = NULL;
+	cs->bos[i].read_domains = 0;
+	cs->bos[i].write_domain = 0;
+	cs->bos[i].new_accounted = 0;
+    }
+    cs->bo_count = 0;
+}
+
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.c b/src/mesa/drivers/dri/radeon/radeon_debug.c
new file mode 100644
index 0000000000..3b6f003803
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_debug.c
@@ -0,0 +1,107 @@
+/*
+ * Copyright © 2009 Pauli Nieminen
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Pauli Nieminen <suokkos@gmail.com>
+ */
+
+#include "utils.h"
+
+#include "radeon_debug.h"
+#include "radeon_common_context.h"
+
+#include <stdarg.h>
+#include <stdio.h>
+
+static const struct dri_debug_control debug_control[] = {
+	{"fall", RADEON_FALLBACKS},
+	{"tex", RADEON_TEXTURE},
+	{"ioctl", RADEON_IOCTL},
+	{"verts", RADEON_RENDER},
+	{"render", RADEON_RENDER},
+	{"swrender", RADEON_SWRENDER},
+	{"state", RADEON_STATE},
+	{"shader", RADEON_SHADER},
+	{"vfmt", RADEON_VFMT},
+	{"vtxf", RADEON_VFMT},
+	{"dri", RADEON_DRI},
+	{"dma", RADEON_DMA},
+	{"sanity", RADEON_SANITY},
+	{"sync", RADEON_SYNC},
+	{"pixel", RADEON_PIXEL},
+	{"mem", RADEON_MEMORY},
+	{"cs", RADEON_CS},
+	{"allmsg", ~RADEON_SYNC}, /* avoid the term "sync" because the parser uses strstr */
+	{NULL, 0}
+};
+
+radeon_debug_type_t radeon_enabled_debug_types;
+
+void radeon_init_debug(void)
+{
+	radeon_enabled_debug_types = driParseDebugString(getenv("RADEON_DEBUG"), debug_control);
+
+	radeon_enabled_debug_types |= RADEON_GENERAL;
+}
+
+void _radeon_debug_add_indent(void)
+{
+	GET_CURRENT_CONTEXT(ctx);
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	const size_t length = sizeof(radeon->debug.indent)
+		/ sizeof(radeon->debug.indent[0]);
+	if (radeon->debug.indent_depth < length - 1) {
+		radeon->debug.indent[radeon->debug.indent_depth] = '\t';
+		++radeon->debug.indent_depth;
+	};
+}
+
+void _radeon_debug_remove_indent(void)
+{
+	GET_CURRENT_CONTEXT(ctx);
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	if (radeon->debug.indent_depth > 0) {
+		radeon->debug.indent[radeon->debug.indent_depth] = '\0';
+		--radeon->debug.indent_depth;
+	}
+}
+
+void _radeon_print(const radeon_debug_type_t type,
+	   const radeon_debug_level_t level,
+	   const char* message,
+	   ...)
+{
+	GET_CURRENT_CONTEXT(ctx);
+	if (ctx) {
+		radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+		// FIXME: Make this multi thread safe
+		if (radeon->debug.indent_depth)
+			fprintf(stderr, "%s", radeon->debug.indent);
+	}
+	va_list values;
+	va_start( values, message );
+	vfprintf(stderr, message, values);
+	va_end( values );
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.h b/src/mesa/drivers/dri/radeon/radeon_debug.h
new file mode 100644
index 0000000000..2a8302293b
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_debug.h
@@ -0,0 +1,169 @@
+/*
+ * Copyright © 2009 Pauli Nieminen
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Pauli Nieminen <suokkos@gmail.com>
+ */
+
+#ifndef RADEON_DEBUG_H_INCLUDED
+#define RADEON_DEBUG_H_INCLUDED
+
+#include <stdlib.h>
+
+typedef enum radeon_debug_levels {
+	RADEON_CRITICAL  = 0, /* Only errors */
+	RADEON_IMPORTANT = 1, /* Important warnings and messages */
+	RADEON_NORMAL    = 2, /* Normal log messages usefull for debugging */
+	RADEON_VERBOSE   = 3, /* Extra details to debugging */
+	RADEON_TRACE     = 4  /* Log about everything that happens */
+} radeon_debug_level_t;
+
+/**
+ * Compile time option to change level of debugging compiled to dri driver.
+ * Selecting critical level is not recommended because perfromance gains are
+ * going to minimal but you will lose a lot of important warnings in case of
+ * errors.
+ */
+#ifndef RADEON_DEBUG_LEVEL
+#define RADEON_DEBUG_LEVEL RADEON_VERBOSE
+#endif
+
+typedef enum radeon_debug_types {
+	RADEON_TEXTURE   = 0x00001,
+	RADEON_STATE     = 0x00002,
+	RADEON_IOCTL     = 0x00004,
+	RADEON_RENDER    = 0x00008,
+	RADEON_SWRENDER  = 0x00010,
+	RADEON_FALLBACKS = 0x00020,
+	RADEON_VFMT      = 0x00040,
+	RADEON_SHADER    = 0x00080,
+	RADEON_CS        = 0x00100,
+	RADEON_DRI       = 0x00200,
+	RADEON_DMA       = 0x00400,
+	RADEON_SANITY    = 0x00800,
+	RADEON_SYNC      = 0x01000,
+	RADEON_PIXEL     = 0x02000,
+	RADEON_MEMORY    = 0x04000,
+	RADEON_VERTS     = 0x08000,
+	RADEON_GENERAL   = 0x10000   /* Used for errors and warnings */
+} radeon_debug_type_t;
+
+#define RADEON_MAX_INDENT 5
+
+struct radeon_debug {
+       size_t indent_depth;
+       char indent[RADEON_MAX_INDENT];
+};
+
+extern radeon_debug_type_t radeon_enabled_debug_types;
+
+/**
+ * Compabibility layer for old debug code
+ **/
+#define RADEON_DEBUG radeon_enabled_debug_types
+
+static inline int radeon_is_debug_enabled(const radeon_debug_type_t type,
+	   const radeon_debug_level_t level)
+{
+       return RADEON_DEBUG_LEVEL >= level
+		&& (type & radeon_enabled_debug_types);
+}
+/*
+ * define macro for gcc specific __attribute__ if using alternative compiler
+ */
+#ifndef __GNUC__
+#define  __attribute__(x)  /*empty*/
+#endif
+
+
+extern void _radeon_print(const radeon_debug_type_t type,
+	   const radeon_debug_level_t level,
+	   const char* message,
+	   ...)  __attribute__((format(printf,3,4)));
+/**
+ * Print out debug message if channel specified by type is enabled
+ * and compile time debugging level is at least as high as level parameter
+ */
+#define radeon_print(type, level, message, ...) do {		\
+	const radeon_debug_level_t _debug_level = (level);	\
+	const radeon_debug_type_t _debug_type = (type);		\
+	/* Compile out if level of message is too high */	\
+	if (radeon_is_debug_enabled(type, level)) {		\
+		_radeon_print(_debug_type, _debug_level,	\
+			(message), ## __VA_ARGS__);		\
+	}							\
+} while(0)
+
+/**
+ * printf style function for writing error messages.
+ */
+#define radeon_error(message, ...) do {				\
+	radeon_print(RADEON_GENERAL, RADEON_CRITICAL,		\
+		(message), ## __VA_ARGS__);			\
+} while(0)
+
+/**
+ * printf style function for writing warnings.
+ */
+#define radeon_warning(message, ...) do {			\
+	radeon_print(RADEON_GENERAL, RADEON_IMPORTANT,		\
+		(message), ## __VA_ARGS__);			\
+} while(0)
+
+extern void radeon_init_debug(void);
+extern void _radeon_debug_add_indent(void);
+extern void _radeon_debug_remove_indent(void);
+
+static inline void radeon_debug_add_indent(void)
+{
+       if (RADEON_DEBUG_LEVEL >= RADEON_VERBOSE) {
+	      _radeon_debug_add_indent();
+       }
+}
+static inline void radeon_debug_remove_indent(void)
+{
+       if (RADEON_DEBUG_LEVEL >= RADEON_VERBOSE) {
+	      _radeon_debug_remove_indent();
+       }
+}
+
+/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
+   I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble
+   with other compilers ... GLUE!
+*/
+#define WARN_ONCE(a, ...)      { \
+       static int warn##__LINE__=1; \
+       if(warn##__LINE__){ \
+               radeon_warning("*********************************WARN_ONCE*********************************\n"); \
+               radeon_warning("File %s function %s line %d\n", \
+                       __FILE__, __FUNCTION__, __LINE__); \
+               radeon_warning(  (a), ## __VA_ARGS__);\
+               radeon_warning("***************************************************************************\n"); \
+               warn##__LINE__=0;\
+               } \
+       }
+
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
new file mode 100644
index 0000000000..c9a32c808b
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -0,0 +1,476 @@
+/**************************************************************************
+
+Copyright (C) 2004 Nicolai Haehnle.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#include <errno.h>
+#include "radeon_common.h"
+#include "main/simple_list.h"
+
+#if defined(USE_X86_ASM)
+#define COPY_DWORDS( dst, src, nr )					\
+do {									\
+	int __tmp;							\
+	__asm__ __volatile__( "rep ; movsl"				\
+			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
+			      : "0" (nr),				\
+			        "D" ((long)dst),			\
+			        "S" ((long)src) );			\
+} while (0)
+#else
+#define COPY_DWORDS( dst, src, nr )		\
+do {						\
+   int j;					\
+   for ( j = 0 ; j < nr ; j++ )			\
+      dst[j] = ((int *)src)[j];			\
+   dst += nr;					\
+} while (0)
+#endif
+
+void radeonEmitVec4(uint32_t *out, const GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 4)
+		COPY_DWORDS(out, data, count);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out++;
+			data += stride;
+		}
+}
+
+void radeonEmitVec8(uint32_t *out, const GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 8)
+		COPY_DWORDS(out, data, count * 2);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out += 2;
+			data += stride;
+		}
+}
+
+void radeonEmitVec12(uint32_t *out, const GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 12) {
+		COPY_DWORDS(out, data, count * 3);
+    }
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out[2] = *(int *)(data + 8);
+			out += 3;
+			data += stride;
+		}
+}
+
+void radeonEmitVec16(uint32_t *out, const GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & RADEON_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 16)
+		COPY_DWORDS(out, data, count * 4);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out[2] = *(int *)(data + 8);
+			out[3] = *(int *)(data + 12);
+			out += 4;
+			data += stride;
+		}
+}
+
+void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
+			 const GLvoid * data, int size, int stride, int count)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	uint32_t *out;
+
+	if (stride == 0) {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		count = 1;
+		aos->stride = 0;
+	} else {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32);
+		aos->stride = size;
+	}
+
+	aos->components = size;
+	aos->count = count;
+
+	out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
+	switch (size) {
+	case 1: radeonEmitVec4(out, data, stride, count); break;
+	case 2: radeonEmitVec8(out, data, stride, count); break;
+	case 3: radeonEmitVec12(out, data, stride, count); break;
+	case 4: radeonEmitVec16(out, data, stride, count); break;
+	default:
+		assert(0);
+		break;
+	}
+}
+
+void radeon_init_dma(radeonContextPtr rmesa)
+{
+	make_empty_list(&rmesa->dma.free);
+	make_empty_list(&rmesa->dma.wait);
+	make_empty_list(&rmesa->dma.reserved);
+	rmesa->dma.minimum_size = MAX_DMA_BUF_SZ;
+}
+
+void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size)
+{
+	struct radeon_dma_bo *dma_bo = NULL;
+	/* we set minimum sizes to at least requested size
+	   aligned to next 16 bytes. */
+	if (size > rmesa->dma.minimum_size)
+		rmesa->dma.minimum_size = (size + 15) & (~15);
+
+	radeon_print(RADEON_DMA, RADEON_NORMAL, "%s size %d minimum_size %d\n",
+			__FUNCTION__, size, rmesa->dma.minimum_size);
+
+
+	/* unmap old reserved bo */
+	if (!is_empty_list(&rmesa->dma.reserved))
+		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
+
+	if (is_empty_list(&rmesa->dma.free)
+	      || last_elem(&rmesa->dma.free)->bo->size < size) {
+		dma_bo = CALLOC_STRUCT(radeon_dma_bo);
+		assert(dma_bo);
+
+again_alloc:
+		dma_bo->bo = radeon_bo_open(rmesa->radeonScreen->bom,
+					    0, rmesa->dma.minimum_size, 4,
+					    RADEON_GEM_DOMAIN_GTT, 0);
+
+		if (!dma_bo->bo) {
+			rcommonFlushCmdBuf(rmesa, __FUNCTION__);
+			goto again_alloc;
+		}
+		insert_at_head(&rmesa->dma.reserved, dma_bo);
+	} else {
+		/* We push and pop buffers from end of list so we can keep
+		   counter on unused buffers for later freeing them from
+		   begin of list */
+		dma_bo = last_elem(&rmesa->dma.free);
+		assert(dma_bo->bo->cref == 1);
+		remove_from_list(dma_bo);
+		insert_at_head(&rmesa->dma.reserved, dma_bo);
+	}
+
+	rmesa->dma.current_used = 0;
+	rmesa->dma.current_vertexptr = 0;
+	
+	if (radeon_cs_space_check_with_bo(rmesa->cmdbuf.cs,
+					  first_elem(&rmesa->dma.reserved)->bo,
+					  RADEON_GEM_DOMAIN_GTT, 0))
+		fprintf(stderr,"failure to revalidate BOs - badness\n");
+
+	if (is_empty_list(&rmesa->dma.reserved)) {
+        /* Cmd buff have been flushed in radeon_revalidate_bos */
+		goto again_alloc;
+	}
+
+	radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1);
+}
+
+/* Allocates a region from rmesa->dma.current.  If there isn't enough
+ * space in current, grab a new buffer (and discard what was left of current)
+ */
+void radeonAllocDmaRegion(radeonContextPtr rmesa,
+			  struct radeon_bo **pbo, int *poffset,
+			  int bytes, int alignment)
+{
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
+
+	if (rmesa->dma.flush)
+		rmesa->dma.flush(rmesa->glCtx);
+
+	assert(rmesa->dma.current_used == rmesa->dma.current_vertexptr);
+
+	alignment--;
+	rmesa->dma.current_used = (rmesa->dma.current_used + alignment) & ~alignment;
+
+	if (is_empty_list(&rmesa->dma.reserved)
+		|| rmesa->dma.current_used + bytes > first_elem(&rmesa->dma.reserved)->bo->size)
+		radeonRefillCurrentDmaRegion(rmesa, bytes);
+
+	*poffset = rmesa->dma.current_used;
+	*pbo = first_elem(&rmesa->dma.reserved)->bo;
+	radeon_bo_ref(*pbo);
+
+	/* Always align to at least 16 bytes */
+	rmesa->dma.current_used = (rmesa->dma.current_used + bytes + 15) & ~15;
+	rmesa->dma.current_vertexptr = rmesa->dma.current_used;
+
+	assert(rmesa->dma.current_used <= first_elem(&rmesa->dma.reserved)->bo->size);
+}
+
+void radeonFreeDmaRegions(radeonContextPtr rmesa)
+{
+	struct radeon_dma_bo *dma_bo;
+	struct radeon_dma_bo *temp;
+	if (RADEON_DEBUG & RADEON_DMA)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+
+	foreach_s(dma_bo, temp, &rmesa->dma.free) {
+		remove_from_list(dma_bo);
+	        radeon_bo_unref(dma_bo->bo);
+		FREE(dma_bo);
+	}
+
+	foreach_s(dma_bo, temp, &rmesa->dma.wait) {
+		remove_from_list(dma_bo);
+	        radeon_bo_unref(dma_bo->bo);
+		FREE(dma_bo);
+	}
+
+	foreach_s(dma_bo, temp, &rmesa->dma.reserved) {
+		remove_from_list(dma_bo);
+		radeon_bo_unmap(dma_bo->bo);
+	        radeon_bo_unref(dma_bo->bo);
+		FREE(dma_bo);
+	}
+}
+
+void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes)
+{
+	if (is_empty_list(&rmesa->dma.reserved))
+		return;
+
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, return_bytes);
+	rmesa->dma.current_used -= return_bytes;
+	rmesa->dma.current_vertexptr = rmesa->dma.current_used;
+}
+
+static int radeon_bo_is_idle(struct radeon_bo* bo)
+{
+	uint32_t domain;
+	int ret = radeon_bo_is_busy(bo, &domain);
+	if (ret == -EINVAL) {
+		WARN_ONCE("Your libdrm or kernel doesn't have support for busy query.\n"
+			"This may cause small performance drop for you.\n");
+	}
+	return ret != -EBUSY;
+}
+
+void radeonReleaseDmaRegions(radeonContextPtr rmesa)
+{
+	struct radeon_dma_bo *dma_bo;
+	struct radeon_dma_bo *temp;
+	const int expire_at = ++rmesa->dma.free.expire_counter + DMA_BO_FREE_TIME;
+	const int time = rmesa->dma.free.expire_counter;
+
+	if (RADEON_DEBUG & RADEON_DMA) {
+		size_t free = 0,
+		       wait = 0,
+		       reserved = 0;
+		foreach(dma_bo, &rmesa->dma.free)
+			++free;
+
+		foreach(dma_bo, &rmesa->dma.wait)
+			++wait;
+
+		foreach(dma_bo, &rmesa->dma.reserved)
+			++reserved;
+
+		fprintf(stderr, "%s: free %zu, wait %zu, reserved %zu, minimum_size: %zu\n",
+		      __FUNCTION__, free, wait, reserved, rmesa->dma.minimum_size);
+	}
+
+	if (!rmesa->radeonScreen->driScreen->dri2.enabled) {
+		/* request updated cs processing information from kernel */
+		legacy_track_pending(rmesa->radeonScreen->bom, 0);
+	}
+	/* move waiting bos to free list.
+	   wait list provides gpu time to handle data before reuse */
+	foreach_s(dma_bo, temp, &rmesa->dma.wait) {
+		if (dma_bo->expire_counter == time) {
+			WARN_ONCE("Leaking dma buffer object!\n");
+			radeon_bo_unref(dma_bo->bo);
+			remove_from_list(dma_bo);
+			FREE(dma_bo);
+			continue;
+		}
+		/* free objects that are too small to be used because of large request */
+		if (dma_bo->bo->size < rmesa->dma.minimum_size) {
+		   radeon_bo_unref(dma_bo->bo);
+		   remove_from_list(dma_bo);
+		   FREE(dma_bo);
+		   continue;
+		}
+		if (!radeon_bo_is_idle(dma_bo->bo))
+			continue;
+		remove_from_list(dma_bo);
+		dma_bo->expire_counter = expire_at;
+		insert_at_tail(&rmesa->dma.free, dma_bo);
+	}
+
+	/* unmap the last dma region */
+	if (!is_empty_list(&rmesa->dma.reserved))
+		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
+	/* move reserved to wait list */
+	foreach_s(dma_bo, temp, &rmesa->dma.reserved) {
+		/* free objects that are too small to be used because of large request */
+		if (dma_bo->bo->size < rmesa->dma.minimum_size) {
+		   radeon_bo_unref(dma_bo->bo);
+		   remove_from_list(dma_bo);
+		   FREE(dma_bo);
+		   continue;
+		}
+		remove_from_list(dma_bo);
+		dma_bo->expire_counter = expire_at;
+		insert_at_tail(&rmesa->dma.wait, dma_bo);
+	}
+
+	/* free bos that have been unused for some time */
+	foreach_s(dma_bo, temp, &rmesa->dma.free) {
+		if (dma_bo->expire_counter != time)
+			break;
+		remove_from_list(dma_bo);
+	        radeon_bo_unref(dma_bo->bo);
+		FREE(dma_bo);
+	}
+
+}
+
+
+/* Flush vertices in the current dma region.
+ */
+void rcommon_flush_last_swtcl_prim( GLcontext *ctx  )
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	struct radeon_dma *dma = &rmesa->dma;
+		
+
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+	dma->flush = NULL;
+
+	if (!is_empty_list(&dma->reserved)) {
+	    GLuint current_offset = dma->current_used;
+
+	    assert (dma->current_used +
+		    rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
+		    dma->current_vertexptr);
+
+	    if (dma->current_used != dma->current_vertexptr) {
+		    dma->current_used = dma->current_vertexptr;
+
+		    rmesa->vtbl.swtcl_flush(ctx, current_offset);
+	    }
+	    rmesa->swtcl.numverts = 0;
+	}
+}
+/* Alloc space in the current dma region.
+ */
+void *
+rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
+{
+	GLuint bytes = vsize * nverts;
+	void *head;
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+	if(is_empty_list(&rmesa->dma.reserved)
+	      ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) {
+		if (rmesa->dma.flush) {
+			rmesa->dma.flush(rmesa->glCtx);
+		}
+
+                radeonRefillCurrentDmaRegion(rmesa, bytes);
+
+		return NULL;
+	}
+
+        if (!rmesa->dma.flush) {
+		/* if cmdbuf flushed DMA restart */
+                rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+                rmesa->dma.flush = rcommon_flush_last_swtcl_prim;
+        }
+
+	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
+        ASSERT( rmesa->dma.flush == rcommon_flush_last_swtcl_prim );
+        ASSERT( rmesa->dma.current_used +
+                rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
+                rmesa->dma.current_vertexptr );
+
+	head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr);
+	rmesa->dma.current_vertexptr += bytes;
+	rmesa->swtcl.numverts += nverts;
+	return head;
+}
+
+void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
+{
+   radeonContextPtr radeon = RADEON_CONTEXT( ctx );
+   int i;
+	if (RADEON_DEBUG & RADEON_IOCTL)
+		fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (radeon->dma.flush) {
+       radeon->dma.flush(radeon->glCtx);
+   }
+   for (i = 0; i < radeon->tcl.aos_count; i++) {
+      if (radeon->tcl.aos[i].bo) {
+         radeon_bo_unref(radeon->tcl.aos[i].bo);
+         radeon->tcl.aos[i].bo = NULL;
+
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h
new file mode 100644
index 0000000000..74e653fd18
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.h
@@ -0,0 +1,58 @@
+/**************************************************************************
+
+Copyright (C) 2004 Nicolai Haehnle.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#ifndef RADEON_DMA_H
+#define RADEON_DMA_H
+
+void radeonEmitVec4(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec8(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec12(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec16(uint32_t *out, const GLvoid * data, int stride, int count);
+
+void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
+			 const GLvoid * data, int size, int stride, int count);
+
+void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes);
+void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size);
+void radeon_init_dma(radeonContextPtr rmesa);
+void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes);
+void radeonAllocDmaRegion(radeonContextPtr rmesa,
+			  struct radeon_bo **pbo, int *poffset,
+			  int bytes, int alignment);
+void radeonReleaseDmaRegions(radeonContextPtr rmesa);
+
+void rcommon_flush_last_swtcl_prim(GLcontext *ctx);
+
+void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize);
+void radeonFreeDmaRegions(radeonContextPtr rmesa);
+void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs );
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c
new file mode 100644
index 0000000000..d83b166742
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c
@@ -0,0 +1,599 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/context.h"
+#include "main/texformat.h"
+#include "main/texrender.h"
+#include "drivers/common/meta.h"
+
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
+
+#define FILE_DEBUG_FLAG RADEON_TEXTURE
+#define DBG(...) do {                                           \
+        if (RADEON_DEBUG & FILE_DEBUG_FLAG)                      \
+                _mesa_printf(__VA_ARGS__);                      \
+} while(0)
+
+static struct gl_framebuffer *
+radeon_new_framebuffer(GLcontext *ctx, GLuint name)
+{
+  return _mesa_new_framebuffer(ctx, name);
+}
+
+static void
+radeon_delete_renderbuffer(struct gl_renderbuffer *rb)
+{
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+
+  ASSERT(rrb);
+
+  if (rrb && rrb->bo) {
+    radeon_bo_unref(rrb->bo);
+  }
+  _mesa_free(rrb);
+}
+
+static void *
+radeon_get_pointer(GLcontext *ctx, struct gl_renderbuffer *rb,
+		   GLint x, GLint y)
+{
+  return NULL;
+}
+
+/**
+ * Called via glRenderbufferStorageEXT() to set the format and allocate
+ * storage for a user-created renderbuffer.
+ */
+static GLboolean
+radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                                 GLenum internalFormat,
+                                 GLuint width, GLuint height)
+{
+  struct radeon_context *radeon = RADEON_CONTEXT(ctx);
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+  GLboolean software_buffer = GL_FALSE;
+  int cpp;
+
+   ASSERT(rb->Name != 0);
+  switch (internalFormat) {
+   case GL_R3_G3_B2:
+   case GL_RGB4:
+   case GL_RGB5:
+      rb->_ActualFormat = GL_RGB5;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 5;
+      rb->GreenBits = 6;
+      rb->BlueBits = 5;
+      cpp = 2;
+      break;
+   case GL_RGB:
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      rb->_ActualFormat = GL_RGB8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 0;
+      cpp = 4;
+      break;
+   case GL_RGBA:
+   case GL_RGBA2:
+   case GL_RGBA4:
+   case GL_RGB5_A1:
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+   case GL_RGBA16:
+      rb->_ActualFormat = GL_RGBA8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 8;
+      cpp = 4;
+      break;
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+      /* alloc a depth+stencil buffer */
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   case GL_DEPTH_COMPONENT16:
+      rb->_ActualFormat = GL_DEPTH_COMPONENT16;
+      rb->DataType = GL_UNSIGNED_SHORT;
+      rb->DepthBits = 16;
+      cpp = 2;
+      break;
+   case GL_DEPTH_COMPONENT:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH_COMPONENT32:
+      rb->_ActualFormat = GL_DEPTH_COMPONENT24;
+      rb->DataType = GL_UNSIGNED_INT;
+      rb->DepthBits = 24;
+      cpp = 4;
+      break;
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->DepthBits = 24;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   default:
+      _mesa_problem(ctx,
+                    "Unexpected format in intel_alloc_renderbuffer_storage");
+      return GL_FALSE;
+   }
+
+  if (ctx->Driver.Flush)
+	  ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+  if (rrb->bo)
+    radeon_bo_unref(rrb->bo);
+  
+    
+   if (software_buffer) {
+      return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat,
+                                             width, height);
+   }
+   else {
+     uint32_t size;
+     uint32_t pitch = ((cpp * width + 63) & ~63) / cpp;
+
+     fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width,
+	  height, pitch);
+
+     size = pitch * height * cpp;
+     rrb->pitch = pitch * cpp;
+     rrb->cpp = cpp;
+     rrb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+			      0,
+			      size,
+			      0,
+			      RADEON_GEM_DOMAIN_VRAM,
+			      0);
+     rb->Width = width;
+     rb->Height = height;
+       return GL_TRUE;
+   }    
+   
+}
+
+
+/**
+ * Called for each hardware renderbuffer when a _window_ is resized.
+ * Just update fields.
+ * Not used for user-created renderbuffers!
+ */
+static GLboolean
+radeon_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                           GLenum internalFormat, GLuint width, GLuint height)
+{
+   ASSERT(rb->Name == 0);
+   rb->Width = width;
+   rb->Height = height;
+   rb->_ActualFormat = internalFormat;
+
+   return GL_TRUE;
+}
+
+
+static void
+radeon_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
+		     GLuint width, GLuint height)
+{
+     struct radeon_framebuffer *radeon_fb = (struct radeon_framebuffer*)fb;
+   int i;
+
+   _mesa_resize_framebuffer(ctx, fb, width, height);
+
+   fb->Initialized = GL_TRUE; /* XXX remove someday */
+
+   if (fb->Name != 0) {
+      return;
+   }
+
+   /* Make sure all window system renderbuffers are up to date */
+   for (i = 0; i < 2; i++) {
+      struct gl_renderbuffer *rb = &radeon_fb->color_rb[i]->base;
+
+      /* only resize if size is changing */
+      if (rb && (rb->Width != width || rb->Height != height)) {
+	 rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height);
+      }
+   }
+}
+
+
+/** Dummy function for gl_renderbuffer::AllocStorage() */
+static GLboolean
+radeon_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+			 GLenum internalFormat, GLuint width, GLuint height)
+{
+   _mesa_problem(ctx, "radeon_op_alloc_storage should never be called.");
+   return GL_FALSE;
+}
+
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv)
+{
+    struct radeon_renderbuffer *rrb;
+
+    rrb = CALLOC_STRUCT(radeon_renderbuffer);
+    if (!rrb)
+	return NULL;
+
+    _mesa_init_renderbuffer(&rrb->base, 0);
+    rrb->base.ClassID = RADEON_RB_CLASS;
+
+    /* XXX format junk */
+    switch (format) {
+	case GL_RGB5:
+	    rrb->base._ActualFormat = GL_RGB5;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 5;
+	    rrb->base.GreenBits = 6;
+	    rrb->base.BlueBits = 5;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_RGB8:
+	    rrb->base._ActualFormat = GL_RGB8;
+	    rrb->base._BaseFormat = GL_RGB;
+	    rrb->base.RedBits = 8;
+	    rrb->base.GreenBits = 8;
+	    rrb->base.BlueBits = 8;
+	    rrb->base.AlphaBits = 0;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_RGBA8:
+	    rrb->base._ActualFormat = GL_RGBA8;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 8;
+	    rrb->base.GreenBits = 8;
+	    rrb->base.BlueBits = 8;
+	    rrb->base.AlphaBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_STENCIL_INDEX8_EXT:
+	    rrb->base._ActualFormat = GL_STENCIL_INDEX8_EXT;
+	    rrb->base._BaseFormat = GL_STENCIL_INDEX;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_DEPTH_COMPONENT16:
+	    rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 16;
+	    rrb->base.DataType = GL_UNSIGNED_SHORT;
+	    break;
+	case GL_DEPTH_COMPONENT24:
+	    rrb->base._ActualFormat = GL_DEPTH_COMPONENT24;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.DataType = GL_UNSIGNED_INT;
+	    break;
+	case GL_DEPTH24_STENCIL8_EXT:
+	    rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+	    rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+	    break;
+	default:
+	    fprintf(stderr, "%s: Unknown format 0x%04x\n", __FUNCTION__, format);
+	    _mesa_delete_renderbuffer(&rrb->base);
+	    return NULL;
+    }
+
+    rrb->dPriv = driDrawPriv;
+    rrb->base.InternalFormat = format;
+
+    rrb->base.Delete = radeon_delete_renderbuffer;
+    rrb->base.AllocStorage = radeon_alloc_window_storage;
+    rrb->base.GetPointer = radeon_get_pointer;
+
+    rrb->bo = NULL;
+    return rrb;
+}
+
+static struct gl_renderbuffer *
+radeon_new_renderbuffer(GLcontext * ctx, GLuint name)
+{
+  struct radeon_renderbuffer *rrb;
+
+  rrb = CALLOC_STRUCT(radeon_renderbuffer);
+  if (!rrb)
+    return NULL;
+
+  _mesa_init_renderbuffer(&rrb->base, name);
+  rrb->base.ClassID = RADEON_RB_CLASS;
+
+  rrb->base.Delete = radeon_delete_renderbuffer;
+  rrb->base.AllocStorage = radeon_alloc_renderbuffer_storage;
+  rrb->base.GetPointer = radeon_get_pointer;
+
+  return &rrb->base;
+}
+
+static void
+radeon_bind_framebuffer(GLcontext * ctx, GLenum target,
+                       struct gl_framebuffer *fb, struct gl_framebuffer *fbread)
+{
+   if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) {
+      radeon_draw_buffer(ctx, fb);
+   }
+   else {
+      /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */
+   }
+}
+
+static void
+radeon_framebuffer_renderbuffer(GLcontext * ctx,
+                               struct gl_framebuffer *fb,
+                               GLenum attachment, struct gl_renderbuffer *rb)
+{
+
+	if (ctx->Driver.Flush)
+		ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+   radeon_draw_buffer(ctx, fb);
+}
+
+
+static GLboolean
+radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, 
+		     struct gl_texture_image *texImage)
+{
+	int retry = 0;
+restart:
+	if (texImage->TexFormat == &_mesa_texformat_argb8888) {
+		rrb->cpp = 4;
+		rrb->base._ActualFormat = GL_RGBA8;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to RGBA8 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_rgb565) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGB5;
+		rrb->base._BaseFormat = GL_RGB;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to RGB5 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_argb1555) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGB5_A1;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to ARGB1555 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_argb4444) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGBA4;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to ARGB1555 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_z16) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+		rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+		rrb->base.DataType = GL_UNSIGNED_SHORT;
+		DBG("Render to DEPTH16 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_s8_z24) {
+		rrb->cpp = 4;
+		rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+		rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+		rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+		DBG("Render to DEPTH_STENCIL texture OK\n");
+	}
+	else {
+		/* try redoing the FBO */
+		if (retry == 1) {
+			DBG("Render to texture BAD FORMAT %d\n",
+			    texImage->TexFormat->MesaFormat);
+			return GL_FALSE;
+		}
+		texImage->TexFormat = radeonChooseTextureFormat(ctx, texImage->InternalFormat, 0,
+								texImage->TexFormat->DataType,
+								1);
+
+		retry++;
+		goto restart;
+	}
+	
+	rrb->pitch = texImage->Width * rrb->cpp;
+	rrb->base.InternalFormat = rrb->base._ActualFormat;
+	rrb->base.Width = texImage->Width;
+	rrb->base.Height = texImage->Height;
+	rrb->base.RedBits = texImage->TexFormat->RedBits;
+	rrb->base.GreenBits = texImage->TexFormat->GreenBits;
+	rrb->base.BlueBits = texImage->TexFormat->BlueBits;
+	rrb->base.AlphaBits = texImage->TexFormat->AlphaBits;
+	rrb->base.DepthBits = texImage->TexFormat->DepthBits;
+	rrb->base.StencilBits = texImage->TexFormat->StencilBits;
+	
+	rrb->base.Delete = radeon_delete_renderbuffer;
+	rrb->base.AllocStorage = radeon_nop_alloc_storage;
+	
+	return GL_TRUE;
+}
+
+
+static struct radeon_renderbuffer *
+radeon_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage)
+{
+  const GLuint name = ~0;   /* not significant, but distinct for debugging */
+  struct radeon_renderbuffer *rrb;
+
+   /* make an radeon_renderbuffer to wrap the texture image */
+   rrb = CALLOC_STRUCT(radeon_renderbuffer);
+   if (!rrb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&rrb->base, name);
+   rrb->base.ClassID = RADEON_RB_CLASS;
+
+   if (!radeon_update_wrapper(ctx, rrb, texImage)) {
+      _mesa_free(rrb);
+      return NULL;
+   }
+
+   return rrb;
+  
+}
+static void
+radeon_render_texture(GLcontext * ctx,
+                     struct gl_framebuffer *fb,
+                     struct gl_renderbuffer_attachment *att)
+{
+   struct gl_texture_image *newImage
+      = att->Texture->Image[att->CubeMapFace][att->TextureLevel];
+   struct radeon_renderbuffer *rrb = radeon_renderbuffer(att->Renderbuffer);
+   radeon_texture_image *radeon_image;
+   GLuint imageOffset;
+
+   (void) fb;
+
+   ASSERT(newImage);
+
+   if (newImage->Border != 0) {
+      /* Fallback on drawing to a texture with a border, which won't have a
+       * miptree.
+       */
+      _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+      _mesa_render_texture(ctx, fb, att);
+      return;
+   }
+   else if (!rrb) {
+      rrb = radeon_wrap_texture(ctx, newImage);
+      if (rrb) {
+         /* bind the wrapper to the attachment point */
+         _mesa_reference_renderbuffer(&att->Renderbuffer, &rrb->base);
+      }
+      else {
+         /* fallback to software rendering */
+         _mesa_render_texture(ctx, fb, att);
+         return;
+      }
+   }
+
+   if (!radeon_update_wrapper(ctx, rrb, newImage)) {
+       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+       _mesa_render_texture(ctx, fb, att);
+       return;
+   }
+
+   DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n",
+       _glthread_GetID(),
+       att->Texture->Name, newImage->Width, newImage->Height,
+       rrb->base.RefCount);
+
+   /* point the renderbufer's region to the texture image region */
+   radeon_image = (radeon_texture_image *)newImage;
+   if (rrb->bo != radeon_image->mt->bo) {
+      if (rrb->bo)
+  	radeon_bo_unref(rrb->bo);
+      rrb->bo = radeon_image->mt->bo;
+      radeon_bo_ref(rrb->bo);
+   }
+
+   /* compute offset of the particular 2D image within the texture region */
+   imageOffset = radeon_miptree_image_offset(radeon_image->mt,
+                                            att->CubeMapFace,
+                                            att->TextureLevel);
+
+   if (att->Texture->Target == GL_TEXTURE_3D) {
+      GLuint offsets[6];
+      radeon_miptree_depth_offsets(radeon_image->mt, att->TextureLevel,
+				   offsets);
+      imageOffset += offsets[att->Zoffset];
+   }
+
+   /* store that offset in the region */
+   rrb->draw_offset = imageOffset;
+
+   /* update drawing region, etc */
+   radeon_draw_buffer(ctx, fb);
+}
+
+static void
+radeon_finish_render_texture(GLcontext * ctx,
+                            struct gl_renderbuffer_attachment *att)
+{
+
+}
+static void
+radeon_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+}
+
+void radeon_fbo_init(struct radeon_context *radeon)
+{
+  radeon->glCtx->Driver.NewFramebuffer = radeon_new_framebuffer;
+  radeon->glCtx->Driver.NewRenderbuffer = radeon_new_renderbuffer;
+  radeon->glCtx->Driver.BindFramebuffer = radeon_bind_framebuffer;
+  radeon->glCtx->Driver.FramebufferRenderbuffer = radeon_framebuffer_renderbuffer;
+  radeon->glCtx->Driver.RenderTexture = radeon_render_texture;
+  radeon->glCtx->Driver.FinishRenderTexture = radeon_finish_render_texture;
+  radeon->glCtx->Driver.ResizeBuffers = radeon_resize_buffers;
+  radeon->glCtx->Driver.ValidateFramebuffer = radeon_validate_framebuffer;
+  radeon->glCtx->Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer;
+}
+
+  
+void radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+				struct radeon_bo *bo)
+{
+  struct radeon_bo *old;
+  old = rb->bo;
+  rb->bo = bo;
+  radeon_bo_ref(bo);
+  if (old)
+    radeon_bo_unref(old);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
index 09acf6b4f8..a0106d00fa 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
@@ -35,7 +35,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
 #include <sched.h>
-#include <errno.h> 
+#include <errno.h>
+
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/shaders.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+#include "main/stencil.h"
+#include "main/matrix.h"
 
 #include "main/glheader.h"
 #include "main/imports.h"
@@ -43,6 +57,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 
 #include "radeon_context.h"
+#include "radeon_common.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_tcl.h"
@@ -58,75 +73,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define RADEON_IDLE_RETRY           16
 
 
-static void radeonWaitForIdle( radeonContextPtr rmesa );
-static int radeonFlushCmdBufLocked( radeonContextPtr rmesa, 
-				    const char * caller );
-
-static void print_state_atom( struct radeon_state_atom *state )
-{
-   int i;
-
-   fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size);
-
-   if (RADEON_DEBUG & DEBUG_VERBOSE) 
-      for (i = 0 ; i < state->cmd_size ; i++) 
-	 fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]);
-
-}
-
-static void radeonSaveHwState( radeonContextPtr rmesa )
-{
-   struct radeon_state_atom *atom;
-   char * dest = rmesa->backup_store.cmd_buf;
-
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-   
-   rmesa->backup_store.cmd_used = 0;
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( atom->check( rmesa->glCtx ) ) {
-	 int size = atom->cmd_size * 4;
-	 memcpy( dest, atom->cmd, size);
-	 dest += size;
-	 rmesa->backup_store.cmd_used += size;
-	 if (RADEON_DEBUG & DEBUG_STATE)
-	    print_state_atom( atom );
-      }
-   }
-
-   assert( rmesa->backup_store.cmd_used <= RADEON_CMD_BUF_SZ );
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Returning to radeonEmitState\n");
-}
-
-/* At this point we were in FlushCmdBufLocked but we had lost our context, so
- * we need to unwire our current cmdbuf, hook the one with the saved state in
- * it, flush it, and then put the current one back.  This is so commands at the
- * start of a cmdbuf can rely on the state being kept from the previous one.
- */
-static void radeonBackUpAndEmitLostStateLocked( radeonContextPtr rmesa )
-{
-   GLuint nr_released_bufs;
-   struct radeon_store saved_store;
-
-   if (rmesa->backup_store.cmd_used == 0)
-      return;
-
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Emitting backup state on lost context\n");
-
-   rmesa->lost_context = GL_FALSE;
-
-   nr_released_bufs = rmesa->dma.nr_released_bufs;
-   saved_store = rmesa->store;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->store = rmesa->backup_store;
-   radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
-   rmesa->dma.nr_released_bufs = nr_released_bufs;
-   rmesa->store = saved_store;
-}
-
 /* =============================================================
  * Kernel command buffer handling
  */
@@ -134,965 +80,382 @@ static void radeonBackUpAndEmitLostStateLocked( radeonContextPtr rmesa )
 /* The state atoms will be emitted in the order they appear in the atom list,
  * so this step is important.
  */
-void radeonSetUpAtomList( radeonContextPtr rmesa )
+void radeonSetUpAtomList( r100ContextPtr rmesa )
 {
-   int i, mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   make_empty_list(&rmesa->hw.atomlist);
-   rmesa->hw.atomlist.name = "atom-list";
-
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ctx);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.set);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lin);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msk);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.vpt);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tcl);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msc);
+   int i, mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits;
+
+   make_empty_list(&rmesa->radeon.hw.atomlist);
+   rmesa->radeon.hw.atomlist.name = "atom-list";
+
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ctx);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.set);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lin);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msk);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.vpt);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tcl);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msc);
    for (i = 0; i < mtu; ++i) {
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tex[i]);
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.txr[i]);
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.cube[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.txr[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i]);
    }
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.zbs);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mtl);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.zbs);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mtl);
    for (i = 0; i < 3 + mtu; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mat[i]);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i]);
    for (i = 0; i < 8; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lit[i]);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i]);
    for (i = 0; i < 6; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ucp[i]);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.eye);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.grd);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.fog);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.glt);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i]);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.eye);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.grd);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.fog);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.glt);
 }
 
-void radeonEmitState( radeonContextPtr rmesa )
+static void radeonEmitScissor(r100ContextPtr rmesa)
 {
-   struct radeon_state_atom *atom;
-   char *dest;
-
-   if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->save_on_next_emit) {
-      radeonSaveHwState(rmesa);
-      rmesa->save_on_next_emit = GL_FALSE;
-   }
-
-   /* this code used to return here but now it emits zbs */
-
-   /* To avoid going across the entire set of states multiple times, just check
-    * for enough space for the case of emitting all state, and inline the
-    * radeonAllocCmdBuf code here without all the checks.
-    */
-   radeonEnsureCmdBufSpace(rmesa, rmesa->hw.max_state_size);
-   dest = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-
-   /* We always always emit zbs, this is due to a bug found by keithw in
-      the hardware and rediscovered after Erics changes by me.
-      if you ever touch this code make sure you emit zbs otherwise
-      you get tcl lockups on at least M7/7500 class of chips - airlied */
-   rmesa->hw.zbs.dirty=1;
-
-   if (RADEON_DEBUG & DEBUG_STATE) {
-      foreach(atom, &rmesa->hw.atomlist) {
-	 if (atom->dirty || rmesa->hw.all_dirty) {
-	    if (atom->check(rmesa->glCtx))
-	       print_state_atom(atom);
-	    else
-	       fprintf(stderr, "skip state %s\n", atom->name);
-	 }
-      }
-   }
-
-   foreach(atom, &rmesa->hw.atomlist) {
-      if (rmesa->hw.all_dirty)
-	 atom->dirty = GL_TRUE;
-      if (!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) &&
-	   atom->is_tcl)
-	 atom->dirty = GL_FALSE;
-      if (atom->dirty) {
-	 if (atom->check(rmesa->glCtx)) {
-	    int size = atom->cmd_size * 4;
-	    memcpy(dest, atom->cmd, size);
-	    dest += size;
-	    rmesa->store.cmd_used += size;
-	    atom->dirty = GL_FALSE;
-	 }
-      }
-   }
-
-   assert(rmesa->store.cmd_used <= RADEON_CMD_BUF_SZ);
- 
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
+    BATCH_LOCALS(&rmesa->radeon);
+    if (!rmesa->radeon.radeonScreen->kernel_mm) {
+       return;
+    }
+    if (rmesa->radeon.state.scissor.enabled) {
+        BEGIN_BATCH(6);
+        OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 0));
+        OUT_BATCH(rmesa->hw.ctx.cmd[CTX_PP_CNTL] | RADEON_SCISSOR_ENABLE);
+        OUT_BATCH(CP_PACKET0(RADEON_RE_TOP_LEFT, 0));
+        OUT_BATCH((rmesa->radeon.state.scissor.rect.y1 << 16) |
+                  rmesa->radeon.state.scissor.rect.x1);
+        OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
+        OUT_BATCH(((rmesa->radeon.state.scissor.rect.y2) << 16) |
+                  (rmesa->radeon.state.scissor.rect.x2));
+        END_BATCH();
+    } else {
+        BEGIN_BATCH(2);
+        OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 0));
+        OUT_BATCH(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & ~RADEON_SCISSOR_ENABLE);
+        END_BATCH();
+    }
 }
 
 /* Fire a section of the retained (indexed_verts) buffer as a regular
- * primtive.  
+ * primtive.
  */
-extern void radeonEmitVbufPrim( radeonContextPtr rmesa,
+extern void radeonEmitVbufPrim( r100ContextPtr rmesa,
 				GLuint vertex_format,
 				GLuint primitive,
 				GLuint vertex_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
-
+   BATCH_LOCALS(&rmesa->radeon);
 
    assert(!(primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
-   
-   radeonEmitState( rmesa );
-
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s cmd_used/4: %d\n", __FUNCTION__,
-	      rmesa->store.cmd_used/4);
-   
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VBUF_BUFSZ,
-						       __FUNCTION__ );
+
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitScissor(rmesa);
+
 #if RADEON_OLD_PACKETS
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM | (3 << 16);
-   cmd[2].i = rmesa->ioctl.vertex_offset;
-   cmd[3].i = vertex_nr;
-   cmd[4].i = vertex_format;
-   cmd[5].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
-	       (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
-
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x offt 0x%x vfmt 0x%x vfcntl %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, cmd[2].i, cmd[4].i, cmd[5].i);
+   BEGIN_BATCH(8);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 3);
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+     OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   } else {
+     OUT_BATCH(rmesa->ioctl.vertex_offset);
+   }
+
+   OUT_BATCH(vertex_nr);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |  RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
+	     (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			   rmesa->ioctl.bo,
+			   RADEON_GEM_DOMAIN_GTT,
+			   0, 0);
+   }
+
+   END_BATCH();
+
 #else
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_VBUF | (1 << 16);
-   cmd[2].i = vertex_format;
-   cmd[3].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_MAOS_ENABLE |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
-	       (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
-
-
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x vfmt 0x%x vfcntl %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, cmd[2].i, cmd[3].i);
+   BEGIN_BATCH(4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_DRAW_VBUF, 1);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_MAOS_ENABLE |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
+	     (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+   END_BATCH();
 #endif
 }
 
-
-void radeonFlushElts( radeonContextPtr rmesa )
+void radeonFlushElts( GLcontext *ctx )
 {
-   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
-   int dwords;
-#if RADEON_OLD_PACKETS
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 24)) / 2;
-#else
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 16)) / 2;
-#endif
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&rmesa->radeon);
+   int nr;
+   uint32_t *cmd = (uint32_t *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_start);
+   int dwords = (rmesa->radeon.cmdbuf.cs->section_ndw - rmesa->radeon.cmdbuf.cs->section_cdw);
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
+   if (RADEON_DEBUG & RADEON_IOCTL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-   assert( rmesa->dma.flush == radeonFlushElts );
-   rmesa->dma.flush = NULL;
+   assert( rmesa->radeon.dma.flush == radeonFlushElts );
+   rmesa->radeon.dma.flush = NULL;
 
-   /* Cope with odd number of elts:
-    */
-   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
-   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+   nr = rmesa->tcl.elt_used;
 
 #if RADEON_OLD_PACKETS
-   cmd[1] |= (dwords - 3) << 16;
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     dwords -= 2;
+   }
+#endif
+
+#if RADEON_OLD_PACKETS
+   cmd[1] |= (dwords + 3) << 16;
    cmd[5] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT;
 #else
-   cmd[1] |= (dwords - 3) << 16;
+   cmd[1] |= (dwords + 2) << 16;
    cmd[3] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT;
 #endif
 
-   if (RADEON_DEBUG & DEBUG_SYNC) {
+   rmesa->radeon.cmdbuf.cs->cdw += dwords;
+   rmesa->radeon.cmdbuf.cs->section_cdw += dwords;
+
+#if RADEON_OLD_PACKETS
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			    rmesa->ioctl.bo,
+			    RADEON_GEM_DOMAIN_GTT,
+			    0, 0);
+   }
+#endif
+
+   END_BATCH();
+
+   if (RADEON_DEBUG & RADEON_SYNC) {
       fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-      radeonFinish( rmesa->glCtx );
+      radeonFinish( rmesa->radeon.glCtx );
    }
-}
 
+}
 
-GLushort *radeonAllocEltsOpenEnded( radeonContextPtr rmesa,
+GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa,
 				    GLuint vertex_format,
 				    GLuint primitive,
 				    GLuint min_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
    GLushort *retval;
+   int align_min_nr;
+   BATCH_LOCALS(&rmesa->radeon);
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, min_nr);
+   if (RADEON_DEBUG & RADEON_IOCTL)
+      fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
 
    assert((primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
-   
-   radeonEmitState( rmesa );
-   
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa,
-						       ELTS_BUFSZ(min_nr),
-						       __FUNCTION__ );
+
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitScissor(rmesa);
+
+   rmesa->tcl.elt_cmd_start = rmesa->radeon.cmdbuf.cs->cdw;
+
+   /* round up min_nr to align the state */
+   align_min_nr = (min_nr + 1) & ~1;
+
 #if RADEON_OLD_PACKETS
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM;
-   cmd[2].i = rmesa->ioctl.vertex_offset;
-   cmd[3].i = 0xffff;
-   cmd[4].i = vertex_format;
-   cmd[5].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_IND |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
-
-   retval = (GLushort *)(cmd+6);
-#else   
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_INDX;
-   cmd[2].i = vertex_format;
-   cmd[3].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_IND |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_MAOS_ENABLE |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
-
-   retval = (GLushort *)(cmd+4);
+   BEGIN_BATCH_NO_AUTOSTATE(2+ELTS_BUFSZ(align_min_nr)/4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 0);
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+     OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   } else {
+     OUT_BATCH(rmesa->ioctl.vertex_offset);
+   }
+   OUT_BATCH(rmesa->ioctl.vertex_max);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_IND |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
+#else
+   BEGIN_BATCH_NO_AUTOSTATE(ELTS_BUFSZ(align_min_nr)/4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_DRAW_INDX, 0);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_IND |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_MAOS_ENABLE |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
 #endif
 
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x vfmt 0x%x prim %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, vertex_format, primitive);
 
-   assert(!rmesa->dma.flush);
-   rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   rmesa->dma.flush = radeonFlushElts;
+   rmesa->tcl.elt_cmd_offset = rmesa->radeon.cmdbuf.cs->cdw;
+   rmesa->tcl.elt_used = min_nr;
 
-   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+   retval = (GLushort *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_offset);
 
-   return retval;
-}
+   if (RADEON_DEBUG & RADEON_RENDER)
+      fprintf(stderr, "%s: header prim %x \n",
+	      __FUNCTION__, primitive);
 
+   assert(!rmesa->radeon.dma.flush);
+   rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+   rmesa->radeon.dma.flush = radeonFlushElts;
 
+   return retval;
+}
 
-void radeonEmitVertexAOS( radeonContextPtr rmesa,
+void radeonEmitVertexAOS( r100ContextPtr rmesa,
 			  GLuint vertex_size,
+			  struct radeon_bo *bo,
 			  GLuint offset )
 {
 #if RADEON_OLD_PACKETS
-   rmesa->ioctl.vertex_size = vertex_size;
    rmesa->ioctl.vertex_offset = offset;
+   rmesa->ioctl.bo = bo;
 #else
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
-   if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL))
+   if (RADEON_DEBUG & (RADEON_PRIMS|DEBUG_IOCTL))
       fprintf(stderr, "%s:  vertex_size 0x%x offset 0x%x \n",
 	      __FUNCTION__, vertex_size, offset);
 
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VERT_AOS_BUFSZ,
-						  __FUNCTION__ );
+   BEGIN_BATCH(7);
+   OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, 2);
+   OUT_BATCH(1);
+   OUT_BATCH(vertex_size | (vertex_size << 8));
+   OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   END_BATCH();
 
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (2 << 16);
-   cmd[2].i = 1;
-   cmd[3].i = vertex_size | (vertex_size << 8);
-   cmd[4].i = offset;
 #endif
 }
-		       
 
-void radeonEmitAOS( radeonContextPtr rmesa,
-		    struct radeon_dma_region **component,
+
+void radeonEmitAOS( r100ContextPtr rmesa,
 		    GLuint nr,
 		    GLuint offset )
 {
 #if RADEON_OLD_PACKETS
    assert( nr == 1 );
-   assert( component[0]->aos_size == component[0]->aos_stride );
-   rmesa->ioctl.vertex_size = component[0]->aos_size;
-   rmesa->ioctl.vertex_offset = 
-      (component[0]->aos_start + offset * component[0]->aos_stride * 4);
+   rmesa->ioctl.bo = rmesa->radeon.tcl.aos[0].bo;
+   rmesa->ioctl.vertex_offset =
+     (rmesa->radeon.tcl.aos[0].offset + offset * rmesa->radeon.tcl.aos[0].stride * 4);
+   rmesa->ioctl.vertex_max = rmesa->radeon.tcl.aos[0].count;
 #else
-   drm_radeon_cmd_header_t *cmd;
-   int sz = AOS_BUFSZ(nr);
+   BATCH_LOCALS(&rmesa->radeon);
+   uint32_t voffset;
+   //   int sz = AOS_BUFSZ(nr);
+   int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
    int i;
-   int *tmp;
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
+   if (RADEON_DEBUG & RADEON_IOCTL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, sz,
-						  __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (((sz / sizeof(int))-3) << 16);
-   cmd[2].i = nr;
-   tmp = &cmd[0].i;
-   cmd += 3;
-
-   for (i = 0 ; i < nr ; i++) {
-      if (i & 1) {
-	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
-		      (component[i]->aos_size << 16));
-	 cmd[2].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-	 cmd += 3;
-      }
-      else {
-	 cmd[0].i = ((component[i]->aos_stride << 8) | 
-		     (component[i]->aos_size << 0));
-	 cmd[1].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
+   BEGIN_BATCH(sz+2+(nr * 2));
+   OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+   OUT_BATCH(nr);
+
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i+1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-   }
 
-   if (RADEON_DEBUG & DEBUG_VERTS) {
-      fprintf(stderr, "%s:\n", __FUNCTION__);
-      for (i = 0 ; i < sz ; i++)
-	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
-   }
-#endif
-}
-
-/* using already shifted color_fmt! */
-void radeonEmitBlit( radeonContextPtr rmesa, /* FIXME: which drmMinor is required? */
-		   GLuint color_fmt,
-		   GLuint src_pitch,
-		   GLuint src_offset,
-		   GLuint dst_pitch,
-		   GLuint dst_offset,
-		   GLint srcx, GLint srcy,
-		   GLint dstx, GLint dsty,
-		   GLuint w, GLuint h )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-	      __FUNCTION__, 
-	      src_pitch, src_offset, srcx, srcy,
-	      dst_pitch, dst_offset, dstx, dsty,
-	      w, h);
-
-   assert( (src_pitch & 63) == 0 );
-   assert( (dst_pitch & 63) == 0 );
-   assert( (src_offset & 1023) == 0 ); 
-   assert( (dst_offset & 1023) == 0 ); 
-   assert( w < (1<<16) );
-   assert( h < (1<<16) );
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, 8 * sizeof(int),
-						  __FUNCTION__ );
-
-
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_CNTL_BITBLT_MULTI | (5 << 16);
-   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_BRUSH_NONE |
-	       color_fmt |
-	       RADEON_GMC_SRC_DATATYPE_COLOR |
-	       RADEON_ROP3_S |
-	       RADEON_DP_SRC_SOURCE_MEMORY |
-	       RADEON_GMC_CLR_CMP_CNTL_DIS |
-	       RADEON_GMC_WR_MSK_DIS );
-
-   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
-   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
-   cmd[5].i = (srcx << 16) | srcy;
-   cmd[6].i = (dstx << 16) | dsty; /* dst */
-   cmd[7].i = (w << 16) | h;
-}
-
-
-void radeonEmitWait( radeonContextPtr rmesa, GLuint flags )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, 1 * sizeof(int),
-					   __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
-   cmd[0].wait.flags = flags;
-}
-
-
-static int radeonFlushCmdBufLocked( radeonContextPtr rmesa, 
-				    const char * caller )
-{
-   int ret, i;
-   drm_radeon_cmd_buffer_t cmd;
-
-   if (rmesa->lost_context)
-      radeonBackUpAndEmitLostStateLocked(rmesa);
-
-   if (RADEON_DEBUG & DEBUG_IOCTL) {
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-
-      if (RADEON_DEBUG & DEBUG_VERBOSE) 
-	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
-	    fprintf(stderr, "%d: %x\n", i/4, 
-		    *(int *)(&rmesa->store.cmd_buf[i]));
-   }
-
-   if (RADEON_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
-	      rmesa->dma.nr_released_bufs);
-
-
-   if (RADEON_DEBUG & DEBUG_SANITY) {
-      if (rmesa->state.scissor.enabled) 
-	 ret = radeonSanityCmdBuffer( rmesa, 
-				      rmesa->state.scissor.numClipRects,
-				      rmesa->state.scissor.pClipRects);
-      else
-	 ret = radeonSanityCmdBuffer( rmesa, 
-				      rmesa->numClipRects,
-				      rmesa->pClipRects);
-      if (ret) {
-	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
-	 goto out;
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[nr - 1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-   }
-
-
-   cmd.bufsz = rmesa->store.cmd_used;
-   cmd.buf = rmesa->store.cmd_buf;
-
-   if (rmesa->state.scissor.enabled) {
-      cmd.nbox = rmesa->state.scissor.numClipRects;
-      cmd.boxes = rmesa->state.scissor.pClipRects;
    } else {
-      cmd.nbox = rmesa->numClipRects;
-      cmd.boxes = rmesa->pClipRects;
-   }
-
-   ret = drmCommandWrite( rmesa->dri.fd,
-			  DRM_RADEON_CMDBUF,
-			  &cmd, sizeof(cmd) );
-
-   if (ret)
-      fprintf(stderr, "drmCommandWrite: %d\n", ret);
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
-      radeonWaitForIdleLocked( rmesa );
-   }
-
- out:
-   rmesa->store.primnr = 0;
-   rmesa->store.statenr = 0;
-   rmesa->store.cmd_used = 0;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->save_on_next_emit = 1;
-
-   return ret;
-}
-
-
-/* Note: does not emit any commands to avoid recursion on
- * radeonAllocCmdBuf.
- */
-void radeonFlushCmdBuf( radeonContextPtr rmesa, const char *caller )
-{
-   int ret;
-
-	      
-   LOCK_HARDWARE( rmesa );
-
-   ret = radeonFlushCmdBufLocked( rmesa, caller );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if (ret) {
-      fprintf(stderr, "drm_radeon_cmd_buffer_t: %d (exiting)\n", ret);
-      exit(ret);
-   }
-}
-
-/* =============================================================
- * Hardware vertex buffer handling
- */
-
-
-void radeonRefillCurrentDmaRegion( radeonContextPtr rmesa )
-{
-   struct radeon_dma_buffer *dmabuf;
-   int fd = rmesa->dri.fd;
-   int index = 0;
-   int size = 0;
-   drmDMAReq dma;
-   int ret;
-
-   if (RADEON_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-      fprintf(stderr, "%s\n", __FUNCTION__);  
-
-   if (rmesa->dma.flush) {
-      rmesa->dma.flush( rmesa );
-   }
-
-   if (rmesa->dma.current.buf)
-      radeonReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-
-   if (rmesa->dma.nr_released_bufs > 4)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-
-   dma.context = rmesa->dri.hwContext;
-   dma.send_count = 0;
-   dma.send_list = NULL;
-   dma.send_sizes = NULL;
-   dma.flags = 0;
-   dma.request_count = 1;
-   dma.request_size = RADEON_BUFFER_SIZE;
-   dma.request_list = &index;
-   dma.request_sizes = &size;
-   dma.granted_count = 0;
-
-   LOCK_HARDWARE(rmesa);	/* no need to validate */
-
-   ret = drmDMA( fd, &dma );
-      
-   if (ret != 0) {
-      /* Free some up this way?
-       */
-      if (rmesa->dma.nr_released_bufs) {
-	 radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
-      }
-      
-      if (RADEON_DEBUG & DEBUG_DMA)
-	 fprintf(stderr, "Waiting for buffers\n");
-
-      radeonWaitForIdleLocked( rmesa );
-      ret = drmDMA( fd, &dma );
-
-      if ( ret != 0 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 fprintf( stderr, "Error: Could not get dma buffer... exiting\n" );
-	 exit( -1 );
-      }
-   }
-
-   UNLOCK_HARDWARE(rmesa);
-
-   if (RADEON_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "Allocated buffer %d\n", index);
-
-   dmabuf = CALLOC_STRUCT( radeon_dma_buffer );
-   dmabuf->buf = &rmesa->radeonScreen->buffers->list[index];
-   dmabuf->refcount = 1;
-
-   rmesa->dma.current.buf = dmabuf;
-   rmesa->dma.current.address = dmabuf->buf->address;
-   rmesa->dma.current.end = dmabuf->buf->total;
-   rmesa->dma.current.start = 0;
-   rmesa->dma.current.ptr = 0;
-
-   rmesa->c_vertexBuffers++;
-}
-
-void radeonReleaseDmaRegion( radeonContextPtr rmesa,
-			     struct radeon_dma_region *region,
-			     const char *caller )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-   
-   if (!region->buf)
-      return;
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (--region->buf->refcount == 0) {
-      drm_radeon_cmd_header_t *cmd;
-
-      if (RADEON_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
-		 region->buf->buf->idx);  
-      
-      cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, sizeof(*cmd), 
-						     __FUNCTION__ );
-      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
-      cmd->dma.buf_idx = region->buf->buf->idx;
-      FREE(region->buf);
-      rmesa->dma.nr_released_bufs++;
-   }
-
-   region->buf = NULL;
-   region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void radeonAllocDmaRegion( radeonContextPtr rmesa, 
-			   struct radeon_dma_region *region,
-			   int bytes,
-			   int alignment )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (region->buf)
-      radeonReleaseDmaRegion( rmesa, region, __FUNCTION__ );
-
-   alignment--;
-   rmesa->dma.current.start = rmesa->dma.current.ptr = 
-      (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      radeonRefillCurrentDmaRegion( rmesa );
-
-   region->start = rmesa->dma.current.start;
-   region->ptr = rmesa->dma.current.start;
-   region->end = rmesa->dma.current.start + bytes;
-   region->address = rmesa->dma.current.address;
-   region->buf = rmesa->dma.current.buf;
-   region->buf->refcount++;
-
-   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
-   rmesa->dma.current.start = 
-      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
-}
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t radeonGetLastFrame (radeonContextPtr rmesa) 
-{
-   drm_radeon_getparam_t gp;
-   int ret;
-   uint32_t frame;
-
-   gp.param = RADEON_PARAM_LAST_FRAME;
-   gp.value = (int *)&frame;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM,
-			      &gp, sizeof(gp) );
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drm_radeon_getparam_t: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-
-   return frame;
-}
-
-static void radeonEmitIrqLocked( radeonContextPtr rmesa )
-{
-   drm_radeon_irq_emit_t ie;
-   int ret;
-
-   ie.irq_seq = &rmesa->iw.irq_seq;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, 
-			      &ie, sizeof(ie) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drm_radeon_irq_emit_t: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void radeonWaitIrq( radeonContextPtr rmesa )
-{
-   int ret;
-
-   do {
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT,
-			     &rmesa->iw, sizeof(rmesa->iw) );
-   } while (ret && (errno == EINTR || errno == EBUSY));
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void radeonWaitForFrameCompletion( radeonContextPtr rmesa )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-
-   if (rmesa->do_irqs) {
-      if (radeonGetLastFrame(rmesa) < sarea->last_frame) {
-	 if (!rmesa->irqsEmitted) {
-	    while (radeonGetLastFrame (rmesa) < sarea->last_frame)
-	       ;
-	 }
-	 else {
-	    UNLOCK_HARDWARE( rmesa ); 
-	    radeonWaitIrq( rmesa );	
-	    LOCK_HARDWARE( rmesa ); 
-	 }
-	 rmesa->irqsEmitted = 10;
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH(voffset);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH(voffset);
       }
 
-      if (rmesa->irqsEmitted) {
-	 radeonEmitIrqLocked( rmesa );
-	 rmesa->irqsEmitted--;
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH(voffset);
       }
-   } 
-   else {
-      while (radeonGetLastFrame (rmesa) < sarea->last_frame) {
-	 UNLOCK_HARDWARE( rmesa ); 
-	 if (rmesa->do_usleeps) 
-	    DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa ); 
+      for (i = 0; i + 1 < nr; i += 2) {
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+0].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
-   }
-}
-
-/* Copy the back color buffer to the front color buffer.
- */
-void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
-		       const drm_clip_rect_t	  *rect)
-{
-   radeonContextPtr rmesa;
-   GLint nbox, i, ret;
-   GLboolean   missed_target;
-   int64_t ust;
-   __DRIscreenPrivate *psp;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx );
-   }
-
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* Throttle the frame rate -- only allow one pending swap buffers
-    * request at a time.
-    */
-   radeonWaitForFrameCompletion( rmesa );
-   if (!rect)
-   {
-       UNLOCK_HARDWARE( rmesa );
-       driWaitForVBlank( dPriv, & missed_target );
-       LOCK_HARDWARE( rmesa );
-   }
-
-   nbox = dPriv->numClipRects; /* must be in locked region */
-
-   for ( i = 0 ; i < nbox ; ) {
-      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      GLint n = 0;
-
-      for ( ; i < nr ; i++ ) {
-
-	  *b = box[i];
-
-	  if (rect)
-	  {
-	      if (rect->x1 > b->x1)
-		  b->x1 = rect->x1;
-	      if (rect->y1 > b->y1)
-		  b->y1 = rect->y1;
-	      if (rect->x2 < b->x2)
-		  b->x2 = rect->x2;
-	      if (rect->y2 < b->y2)
-		  b->y2 = rect->y2;
-
-	      if (b->x1 >= b->x2 || b->y1 >= b->y2)
-		  continue;
-	  }
-
-	  b++;
-	  n++;
-      }
-      rmesa->sarea->nbox = n;
-
-      if (!n)
-	 continue;
-
-      ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
-
-      if ( ret ) {
-	 fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret );
-	 UNLOCK_HARDWARE( rmesa );
-	 exit( 1 );
+      if (nr & 1) {
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[nr-1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
    }
+   END_BATCH();
 
-   UNLOCK_HARDWARE( rmesa );
-   if (!rect)
-   {
-       psp = dPriv->driScreenPriv;
-       rmesa->swap_count++;
-       (*psp->systemTime->getUST)( & ust );
-       if ( missed_target ) {
-	   rmesa->swap_missed_count++;
-	   rmesa->swap_missed_ust = ust - rmesa->swap_ust;
-       }
-
-       rmesa->swap_ust = ust;
-       rmesa->hw.all_dirty = GL_TRUE;
-   }
-}
-
-void radeonPageFlip( __DRIdrawablePrivate *dPriv )
-{
-   radeonContextPtr rmesa;
-   GLint ret;
-   GLboolean   missed_target;
-   __DRIscreenPrivate *psp;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-   psp = dPriv->driScreenPriv;
-
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-	      rmesa->sarea->pfCurrentPage);
-   }
-
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* Need to do this for the perf box placement:
-    */
-   if (dPriv->numClipRects)
-   {
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      b[0] = box[0];
-      rmesa->sarea->nbox = 1;
-   }
-
-   /* Throttle the frame rate -- only allow a few pending swap buffers
-    * request at a time.
-    */
-   radeonWaitForFrameCompletion( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   driWaitForVBlank( dPriv, & missed_target );
-   if ( missed_target ) {
-      rmesa->swap_missed_count++;
-      (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust );
-   }
-   LOCK_HARDWARE( rmesa );
-
-   ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
-      exit( 1 );
-   }
-
-   rmesa->swap_count++;
-   (void) (*psp->systemTime->getUST)( & rmesa->swap_ust );
-
-   /* Get ready for drawing next frame.  Update the renderbuffers'
-    * flippedOffset/Pitch fields so we draw into the right place.
-    */
-   driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                        rmesa->sarea->pfCurrentPage);
-
-   radeonUpdateDrawBuffer(rmesa->glCtx);
+#endif
 }
 
-
 /* ================================================================
  * Buffer clear
  */
 #define RADEON_MAX_CLEARS	256
 
-static void radeonClear( GLcontext *ctx, GLbitfield mask )
+static void radeonKernelClear(GLcontext *ctx, GLuint flags)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
+     r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   drm_radeon_sarea_t *sarea = rmesa->radeon.sarea;
    uint32_t clear;
-   GLuint flags = 0;
-   GLuint color_mask = 0;
    GLint ret, i;
    GLint cx, cy, cw, ch;
 
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "radeonClear\n");
-   }
-
-   {
-      LOCK_HARDWARE( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      if ( dPriv->numClipRects == 0 ) 
-	 return;
-   }
-   
-   radeonFlush( ctx ); 
-
-   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
-      flags |= RADEON_FRONT;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_FRONT_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_BACK_LEFT ) {
-      flags |= RADEON_BACK;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_BACK_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_DEPTH ) {
-      flags |= RADEON_DEPTH;
-      mask &= ~BUFFER_BIT_DEPTH;
-   }
-
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->state.stencil.hwBuffer ) {
-      flags |= RADEON_STENCIL;
-      mask &= ~BUFFER_BIT_STENCIL;
-   }
-
-   if ( mask ) {
-      if (RADEON_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
-      _swrast_Clear( ctx, mask );
-   }
-
-   if ( !flags ) 
-      return;
-
-   if (rmesa->using_hyperz) {
-      flags |= RADEON_USE_COMP_ZBUF;
-/*      if (rmesa->radeonScreen->chipset & RADEON_CHIPSET_TCL) 
-         flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
-	    ((rmesa->state.stencil.clear & RADEON_STENCIL_WRITE_MASK) == RADEON_STENCIL_WRITE_MASK))) {
-	  flags |= RADEON_CLEAR_FASTZ;
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* compute region after locking: */
    cx = ctx->DrawBuffer->_Xmin;
@@ -1112,7 +475,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 
       gp.param = RADEON_PARAM_LAST_CLEAR;
       gp.value = (int *)&clear;
-      ret = drmCommandWriteRead( rmesa->dri.fd,
+      ret = drmCommandWriteRead( rmesa->radeon.dri.fd,
 				 DRM_RADEON_GETPARAM, &gp, sizeof(gp) );
 
       if ( ret ) {
@@ -1124,20 +487,20 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 	 break;
       }
 
-      if ( rmesa->do_usleeps ) {
-	 UNLOCK_HARDWARE( rmesa );
+      if ( rmesa->radeon.do_usleeps ) {
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
+	 LOCK_HARDWARE( &rmesa->radeon );
       }
    }
 
    /* Send current state to the hardware */
-   radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    for ( i = 0 ; i < dPriv->numClipRects ; ) {
       GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects );
       drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
+      drm_clip_rect_t *b = rmesa->radeon.sarea->boxes;
       drm_radeon_clear_t clear;
       drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
       GLint n = 0;
@@ -1172,105 +535,107 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 	 }
       }
 
-      rmesa->sarea->nbox = n;
+      rmesa->radeon.sarea->nbox = n;
 
       clear.flags       = flags;
-      clear.clear_color = rmesa->state.color.clear;
-      clear.clear_depth = rmesa->state.depth.clear;
+      clear.clear_color = rmesa->radeon.state.color.clear;
+      clear.clear_depth = rmesa->radeon.state.depth.clear;
       clear.color_mask  = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      clear.depth_mask  = rmesa->state.stencil.clear;
+      clear.depth_mask  = rmesa->radeon.state.stencil.clear;
       clear.depth_boxes = depth_boxes;
 
       n--;
-      b = rmesa->sarea->boxes;
+      b = rmesa->radeon.sarea->boxes;
       for ( ; n >= 0 ; n-- ) {
 	 depth_boxes[n].f[CLEAR_X1] = (float)b[n].x1;
 	 depth_boxes[n].f[CLEAR_Y1] = (float)b[n].y1;
 	 depth_boxes[n].f[CLEAR_X2] = (float)b[n].x2;
 	 depth_boxes[n].f[CLEAR_Y2] = (float)b[n].y2;
-	 depth_boxes[n].f[CLEAR_DEPTH] = 
-	    (float)rmesa->state.depth.clear;
+	 depth_boxes[n].f[CLEAR_DEPTH] =
+	    (float)rmesa->radeon.state.depth.clear;
       }
 
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR,
+      ret = drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_CLEAR,
 			     &clear, sizeof(drm_radeon_clear_t));
 
       if ( ret ) {
-	 UNLOCK_HARDWARE( rmesa );
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret );
 	 exit( 1 );
       }
    }
-
-   UNLOCK_HARDWARE( rmesa );
-   rmesa->hw.all_dirty = GL_TRUE;
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
 
-
-void radeonWaitForIdleLocked( radeonContextPtr rmesa )
+static void radeonClear( GLcontext *ctx, GLbitfield mask )
 {
-    int fd = rmesa->dri.fd;
-    int to = 0;
-    int ret, i = 0;
-
-    rmesa->c_drawWaits++;
-
-    do {
-        do {
-            ret = drmCommandNone( fd, DRM_RADEON_CP_IDLE);
-        } while ( ret && errno == EBUSY && i++ < RADEON_IDLE_RETRY );
-    } while ( ( ret == -EBUSY ) && ( to++ < RADEON_TIMEOUT ) );
-
-    if ( ret < 0 ) {
-	UNLOCK_HARDWARE( rmesa );
-	fprintf( stderr, "Error: Radeon timed out... exiting\n" );
-	exit( -1 );
-    }
-}
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLuint flags = 0;
+   GLuint color_mask = 0;
+   GLuint orig_mask = mask;
 
+   if ( RADEON_DEBUG & RADEON_IOCTL ) {
+      fprintf( stderr, "radeonClear\n");
+   }
 
-static void radeonWaitForIdle( radeonContextPtr rmesa )
-{
-   LOCK_HARDWARE(rmesa);
-   radeonWaitForIdleLocked( rmesa );
-   UNLOCK_HARDWARE(rmesa);
-}
+   {
+      LOCK_HARDWARE( &rmesa->radeon );
+      UNLOCK_HARDWARE( &rmesa->radeon );
+      if ( dPriv->numClipRects == 0 )
+	 return;
+   }
 
+   radeon_firevertices(&rmesa->radeon);
 
-void radeonFlush( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
+      flags |= RADEON_FRONT;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_FRONT_LEFT;
+   }
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   if ( mask & BUFFER_BIT_BACK_LEFT ) {
+      flags |= RADEON_BACK;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
 
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+   if ( mask & BUFFER_BIT_DEPTH ) {
+      flags |= RADEON_DEPTH;
+      mask &= ~BUFFER_BIT_DEPTH;
+   }
 
-   radeonEmitState( rmesa );
-   
-   if (rmesa->store.cmd_used)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-}
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
+      flags |= RADEON_STENCIL;
+      mask &= ~BUFFER_BIT_STENCIL;
+   }
 
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void radeonFinish( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeonFlush( ctx );
-
-   if (rmesa->do_irqs) {
-      LOCK_HARDWARE( rmesa );
-      radeonEmitIrqLocked( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      radeonWaitIrq( rmesa );
+   if ( mask ) {
+      if (RADEON_DEBUG & RADEON_FALLBACKS)
+	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
+      _swrast_Clear( ctx, mask );
    }
-   else
-      radeonWaitForIdle( rmesa );
-}
 
+   if ( !flags )
+      return;
+
+   if (rmesa->using_hyperz) {
+      flags |= RADEON_USE_COMP_ZBUF;
+/*      if (rmesa->radeon.radeonScreen->chipset & RADEON_CHIPSET_TCL)
+         flags |= RADEON_USE_HIERZ; */
+      if (((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+	    ((rmesa->radeon.state.stencil.clear & RADEON_STENCIL_WRITE_MASK) == RADEON_STENCIL_WRITE_MASK))) {
+	  flags |= RADEON_CLEAR_FASTZ;
+      }
+   }
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     radeonUserClear(ctx, orig_mask);
+   else {
+      radeonKernelClear(ctx, flags);
+      rmesa->radeon.hw.all_dirty = GL_TRUE;
+   }
+}
 
 void radeonInitIoctlFuncs( GLcontext *ctx )
 {
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
index 4e3a44df07..deb53ae313 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
@@ -38,31 +38,32 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/simple_list.h"
 #include "radeon_lock.h"
+#include "radeon_bocs_wrapper.h"
 
-
-extern void radeonEmitState( radeonContextPtr rmesa );
-extern void radeonEmitVertexAOS( radeonContextPtr rmesa,
+extern void radeonEmitVertexAOS( r100ContextPtr rmesa,
 				 GLuint vertex_size,
+				 struct radeon_bo *bo,
 				 GLuint offset );
 
-extern void radeonEmitVbufPrim( radeonContextPtr rmesa,
+extern void radeonEmitVbufPrim( r100ContextPtr rmesa,
 				GLuint vertex_format,
 				GLuint primitive,
 				GLuint vertex_nr );
 
-extern void radeonFlushElts( radeonContextPtr rmesa );
+extern void radeonFlushElts( GLcontext *ctx );
+			    
 
-extern GLushort *radeonAllocEltsOpenEnded( radeonContextPtr rmesa,
+extern GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa,
 					   GLuint vertex_format,
 					   GLuint primitive,
 					   GLuint min_nr );
 
-extern void radeonEmitAOS( radeonContextPtr rmesa,
-			   struct radeon_dma_region **regions,
+
+extern void radeonEmitAOS( r100ContextPtr rmesa,
 			   GLuint n,
 			   GLuint offset );
 
-extern void radeonEmitBlit( radeonContextPtr rmesa,
+extern void radeonEmitBlit( r100ContextPtr rmesa,
 			    GLuint color_fmt,
 			    GLuint src_pitch,
 			    GLuint src_offset,
@@ -72,30 +73,15 @@ extern void radeonEmitBlit( radeonContextPtr rmesa,
 			    GLint dstx, GLint dsty,
 			    GLuint w, GLuint h );
 
-extern void radeonEmitWait( radeonContextPtr rmesa, GLuint flags );
-
-extern void radeonFlushCmdBuf( radeonContextPtr rmesa, const char * );
-extern void radeonRefillCurrentDmaRegion( radeonContextPtr rmesa );
+extern void radeonEmitWait( r100ContextPtr rmesa, GLuint flags );
 
-extern void radeonAllocDmaRegion( radeonContextPtr rmesa,
-				  struct radeon_dma_region *region,
-				  int bytes, 
-				  int alignment );
+extern void radeonFlushCmdBuf( r100ContextPtr rmesa, const char * );
 
-extern void radeonReleaseDmaRegion( radeonContextPtr rmesa,
-				    struct radeon_dma_region *region,
-				    const char *caller );
-
-extern void radeonCopyBuffer( __DRIdrawablePrivate *drawable,
-			      const drm_clip_rect_t	 *rect);
-extern void radeonPageFlip( __DRIdrawablePrivate *drawable );
 extern void radeonFlush( GLcontext *ctx );
 extern void radeonFinish( GLcontext *ctx );
-extern void radeonWaitForIdleLocked( radeonContextPtr rmesa );
-extern void radeonWaitForVBlank( radeonContextPtr rmesa );
 extern void radeonInitIoctlFuncs( GLcontext *ctx );
-extern void radeonGetAllParams( radeonContextPtr rmesa );
-extern void radeonSetUpAtomList( radeonContextPtr rmesa );
+extern void radeonGetAllParams( r100ContextPtr rmesa );
+extern void radeonSetUpAtomList( r100ContextPtr rmesa );
 
 /* ================================================================
  * Helper macros:
@@ -105,33 +91,33 @@ extern void radeonSetUpAtomList( radeonContextPtr rmesa );
  */
 #define RADEON_NEWPRIM( rmesa )			\
 do {						\
-   if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );	\
+   if ( rmesa->radeon.dma.flush )			\
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
 } while (0)
 
 /* Can accomodate several state changes and primitive changes without
  * actually firing the buffer.
  */
+
 #define RADEON_STATECHANGE( rmesa, ATOM )			\
 do {								\
    RADEON_NEWPRIM( rmesa );					\
    rmesa->hw.ATOM.dirty = GL_TRUE;				\
-   rmesa->hw.is_dirty = GL_TRUE;				\
+   rmesa->radeon.hw.is_dirty = GL_TRUE;				\
 } while (0)
 
-#define RADEON_DB_STATE( ATOM )			        \
+#define RADEON_DB_STATE( ATOM )				\
    memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd,	\
 	   rmesa->hw.ATOM.cmd_size * 4)
 
-static INLINE int RADEON_DB_STATECHANGE( 
-   radeonContextPtr rmesa,
-   struct radeon_state_atom *atom )
+static INLINE int RADEON_DB_STATECHANGE(r100ContextPtr rmesa,
+					struct radeon_state_atom *atom )
 {
    if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
-      int *tmp;
+      GLuint *tmp;
       RADEON_NEWPRIM( rmesa );
       atom->dirty = GL_TRUE;
-      rmesa->hw.is_dirty = GL_TRUE;
+      rmesa->radeon.hw.is_dirty = GL_TRUE;
       tmp = atom->cmd; 
       atom->cmd = atom->lastcmd;
       atom->lastcmd = tmp;
@@ -141,62 +127,55 @@ static INLINE int RADEON_DB_STATECHANGE(
       return 0;
 }
 
-
-/* Fire the buffered vertices no matter what.
- */
-#define RADEON_FIREVERTICES( rmesa )			\
-do {							\
-   if ( rmesa->store.cmd_used || rmesa->dma.flush ) {	\
-      radeonFlush( rmesa->glCtx );			\
-   }							\
-} while (0)
-
 /* Command lengths.  Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ
  * are available, you will also be adding an rmesa->state.max_state_size because
  * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts.
  */
 #if RADEON_OLD_PACKETS
-#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int))
+#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2))+nr*2)
 #define VERT_AOS_BUFSZ	(0)
 #define ELTS_BUFSZ(nr)	(24 + nr * 2)
-#define VBUF_BUFSZ	(6 * sizeof(int))
+#define VBUF_BUFSZ	(8)
 #else
-#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int))
-#define VERT_AOS_BUFSZ	(5 * sizeof(int))
+#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2) + nr*2))
+#define VERT_AOS_BUFSZ	(5)
 #define ELTS_BUFSZ(nr)	(16 + nr * 2)
-#define VBUF_BUFSZ	(4 * sizeof(int))
+#define VBUF_BUFSZ	(4)
 #endif
+#define SCISSOR_BUFSZ	(8)
+#define INDEX_BUFSZ	(7)
 
-/* Ensure that a minimum amount of space is available in the command buffer.
- * This is used to ensure atomicity of state updates with the rendering requests
- * that rely on them.
- *
- * An alternative would be to implement a "soft lock" such that when the buffer
- * wraps at an inopportune time, we grab the lock, flush the current buffer,
- * and hang on to the lock until the critical section is finished and we flush
- * the buffer again and unlock.
- */
-static INLINE void radeonEnsureCmdBufSpace( radeonContextPtr rmesa,
-					      int bytes )
-{
-   if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-   assert( bytes <= RADEON_CMD_BUF_SZ );
-}
 
-/* Alloc space in the command buffer
- */
-static INLINE char *radeonAllocCmdBuf( radeonContextPtr rmesa,
-					 int bytes, const char *where )
+static inline uint32_t cmdpacket3(int cmd_type)
 {
-   if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
+  drm_radeon_cmd_header_t cmd;
+
+  cmd.i = 0;
+  cmd.header.cmd_type = cmd_type;
+
+  return (uint32_t)cmd.i;
 
-   {
-      char *head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-      rmesa->store.cmd_used += bytes;
-      return head;
-   }
 }
 
+#define OUT_BATCH_PACKET3(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3));				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP));	      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+
 #endif /* __RADEON_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_lighting.c b/src/mesa/drivers/dri/radeon/radeon_lighting.c
index ac3b94e4a6..ba444f2b10 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lighting.c
+++ b/src/mesa/drivers/dri/radeon/radeon_lighting.c
@@ -195,7 +195,7 @@ void radeonUpdateMaterial( GLcontext *ctx )
    if (ctx->Light.ColorMaterialEnabled)
       mask &= ~ctx->Light.ColorMaterialBitmask;
 
-   if (RADEON_DEBUG & DEBUG_STATE)
+   if (RADEON_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
       
@@ -234,7 +234,7 @@ void radeonUpdateMaterial( GLcontext *ctx )
       check_twoside_fallback( ctx );
       update_global_ambient( ctx );
    }
-   else if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_STATE))
+   else if (RADEON_DEBUG & (RADEON_PRIMS|DEBUG_STATE))
       fprintf(stderr, "%s: Elided noop material call\n", __FUNCTION__);
 }
 
@@ -624,7 +624,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx )
    GLboolean tmp;
    RADEON_STATECHANGE( rmesa, tcl );
 
-   if (RADEON_DEBUG & DEBUG_STATE)
+   if (RADEON_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s %d\n", __FUNCTION__, ctx->_NeedEyeCoords);
 
    if (ctx->_NeedEyeCoords)
diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c
index 64bb3ca103..02de8e5fd1 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lock.c
+++ b/src/mesa/drivers/dri/radeon/radeon_lock.c
@@ -41,30 +41,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/glheader.h"
 #include "main/mtypes.h"
-#include "radeon_context.h"
+#include "main/colormac.h"
+#include "dri_util.h"
+#include "radeon_screen.h"
+#include "radeon_common.h"
 #include "radeon_lock.h"
-#include "radeon_tex.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-
 #include "drirenderbuffer.h"
 
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-static void radeonUpdatePageFlipping(radeonContextPtr rmesa)
-{
-	rmesa->doPageFlip = rmesa->sarea->pfState;
-	if (rmesa->glCtx->WinSysDrawBuffer) {
-		driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-				     rmesa->sarea->pfCurrentPage);
-	}
-}
-
 /* Update the hardware state.  This is called if another context has
  * grabbed the hardware lock, which includes the X server.  This
  * function also updates the driver's window state after the X server
@@ -75,10 +58,11 @@ static void radeonUpdatePageFlipping(radeonContextPtr rmesa)
  */
 void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 {
-	__DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-	__DRIdrawablePrivate *const readable = rmesa->dri.readable;
+	__DRIdrawablePrivate *const drawable = radeon_get_drawable(rmesa);
+	__DRIdrawablePrivate *const readable = radeon_get_readable(rmesa);
 	__DRIscreenPrivate *sPriv = rmesa->dri.screen;
-	drm_radeon_sarea_t *sarea = rmesa->sarea;
+
+	assert(drawable != NULL);
 
 	drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags);
 
@@ -96,29 +80,89 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 	}
 
 	if (rmesa->lastStamp != drawable->lastStamp) {
-		radeonUpdatePageFlipping(rmesa);
-		radeonSetCliprects(rmesa);
-		radeonUpdateViewportOffset(rmesa->glCtx);
-		driUpdateFramebufferSize(rmesa->glCtx, drawable);
+		radeon_window_moved(rmesa);
+		rmesa->lastStamp = drawable->lastStamp;
 	}
 
-	RADEON_STATECHANGE(rmesa, ctx);
-	if (rmesa->sarea->tiling_enabled) {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
-		    RADEON_COLOR_TILE_ENABLE;
-	} else {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &=
-		    ~RADEON_COLOR_TILE_ENABLE;
-	}
+	rmesa->vtbl.get_lock(rmesa);
+}
+#ifndef NDEBUG
+struct lock_debug {
+	const char* function;
+	const char* file;
+	int line;
+};
+
+static struct lock_debug ldebug = {0};
+#endif
+
+#if 0
+/** TODO: use atomic operations for reference counting **/
+/** gcc 4.2 has builtin functios for this **/
+#define ATOMIC_INC_AND_FETCH(atomic) __sync_add_and_fetch(&atomic, 1)
+#define ATOMIC_DEC_AND_FETCH(atomic) __sync_sub_and_fetch(&atomic, 1)
+#else
+#define ATOMIC_INC_AND_FETCH(atomic) (++atomic)
+#define ATOMIC_DEC_AND_FETCH(atomic) (--atomic)
+#endif
+
 
-	if (sarea->ctx_owner != rmesa->dri.hwContext) {
-		int i;
-		sarea->ctx_owner = rmesa->dri.hwContext;
+void radeon_lock_hardware(radeonContextPtr radeon
+#ifndef NDEBUG
+		,const char* function
+		,const char* file
+		,const int line
+#endif
+		)
+{
+	char ret = 0;
+	struct radeon_framebuffer *rfb = NULL;
+	struct radeon_renderbuffer *rrb = NULL;
+
+	if (radeon_get_drawable(radeon)) {
+		rfb = radeon_get_drawable(radeon)->driverPrivate;
+
+		if (rfb)
+			rrb = radeon_get_renderbuffer(&rfb->base,
+						      rfb->base._ColorDrawBufferIndexes[0]);
+	}
 
-		for (i = 0; i < rmesa->nr_heaps; i++) {
-			DRI_AGE_TEXTURES(rmesa->texture_heaps[i]);
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		if (ATOMIC_INC_AND_FETCH(radeon->dri.hwLockCount) > 1)
+		{
+#ifndef NDEBUG
+			if ( RADEON_DEBUG & RADEON_SANITY )
+				fprintf(stderr, "*** %d times of recursive call to %s ***\n"
+						"Original call was from %s (file: %s line: %d)\n"
+						"Now call is coming from %s (file: %s line: %d)\n"
+						, radeon->dri.hwLockCount, __FUNCTION__
+						, ldebug.function, ldebug.file, ldebug.line
+						, function, file, line
+					   );
+#endif
+			return;
 		}
+		DRM_CAS(radeon->dri.hwLock, radeon->dri.hwContext,
+			 (DRM_LOCK_HELD | radeon->dri.hwContext), ret );
+		if (ret)
+			radeonGetLock(radeon, 0);
+#ifndef NDEBUG
+		ldebug.function = function;
+		ldebug.file = file;
+		ldebug.line = line;
+#endif
 	}
+}
 
-	rmesa->lost_context = GL_TRUE;
+void radeon_unlock_hardware(radeonContextPtr radeon)
+{
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		if (ATOMIC_DEC_AND_FETCH(radeon->dri.hwLockCount) > 0)
+		{
+			return;
+		}
+		DRM_UNLOCK( radeon->dri.fd,
+			    radeon->dri.hwLock,
+			    radeon->dri.hwContext );
+	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.h b/src/mesa/drivers/dri/radeon/radeon_lock.h
index 86e96aa7d2..da5a5b4371 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lock.h
+++ b/src/mesa/drivers/dri/radeon/radeon_lock.h
@@ -39,74 +39,31 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Kevin E. Martin <martin@valinux.com>
  */
 
-#ifndef __RADEON_LOCK_H__
-#define __RADEON_LOCK_H__
+#ifndef COMMON_LOCK_H
+#define COMMON_LOCK_H
 
-extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
-
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if ( prevLockFile ) {						\
-	 fprintf( stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__ );	\
-	 exit( 1 );							\
-      }									\
-   } while (0)
-
-#else
+#include "main/colormac.h"
+#include "radeon_screen.h"
+#include "radeon_common.h"
 
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
+extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
 
+void radeon_lock_hardware(radeonContextPtr rmesa
+#ifndef NDEBUG
+		,const char* function
+		,const char* file
+		,const int line
 #endif
-
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
+		);
+void radeon_unlock_hardware(radeonContextPtr rmesa);
 
 /* Lock the hardware and validate our state.
  */
-#define LOCK_HARDWARE( rmesa )					\
-   do {								\
-      char __ret = 0;						\
-      DEBUG_CHECK_LOCK();					\
-      DRM_CAS( (rmesa)->dri.hwLock, (rmesa)->dri.hwContext,		\
-	       (DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret );	\
-      if ( __ret )						\
-	 radeonGetLock( (rmesa), 0 );				\
-      DEBUG_LOCK();						\
-   } while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-   do {									\
-      DRM_UNLOCK( (rmesa)->dri.fd,					\
-		  (rmesa)->dri.hwLock,					\
-		  (rmesa)->dri.hwContext );				\
-      DEBUG_RESET();							\
-   } while (0)
+#ifdef NDEBUG
+#define LOCK_HARDWARE( rmesa )	radeon_lock_hardware(rmesa)
+#else
+#define LOCK_HARDWARE( rmesa )	radeon_lock_hardware(rmesa, __FUNCTION__, __FILE__, __LINE__)
+#endif
+#define UNLOCK_HARDWARE( rmesa )  radeon_unlock_hardware(rmesa)
 
-#endif				/* __RADEON_LOCK_H__ */
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos.h b/src/mesa/drivers/dri/radeon/radeon_maos.h
index b8935e84a0..b88eb198d5 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos.h
+++ b/src/mesa/drivers/dri/radeon/radeon_maos.h
@@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 extern void radeonEmitArrays( GLcontext *ctx, GLuint inputs );
-extern void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs );
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
index 31eea13f4e..08e1c5d00d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
@@ -48,160 +48,35 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_maos.h"
 #include "radeon_tcl.h"
 
-#if 0
-/* Usage:
- *   - from radeon_tcl_render
- *   - call radeonEmitArrays to ensure uptodate arrays in dma
- *   - emit primitives (new type?) which reference the data
- *       -- need to use elts for lineloop, quads, quadstrip/flat
- *       -- other primitives are all well-formed (need tristrip-1,fake-poly)
- *
- */
-static void emit_ubyte_rgba3( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   radeon_color_t *out = (radeon_color_t *)(rvb->start + rvb->address);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p\n",
-	      __FUNCTION__, count, stride, (void *)out);
-
-   for (i = 0; i < count; i++) {
-      out->red   = *data;
-      out->green = *(data+1);
-      out->blue  = *(data+2);
-      out->alpha = 0xFF;
-      out++;
-      data += stride;
-   }
-}
-
-static void emit_ubyte_rgba4( GLcontext *ctx,
-			      struct radeon_dma_region *rvb,
-			      char *data,
-			      int stride,
-			      int count )
+static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos,
+			GLvoid *data, int stride, int count)
 {
    int i;
-   int *out = (int *)(rvb->address + rvb->start);
+   uint32_t *out;
+   int size = 1;
+   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 
-   if (RADEON_DEBUG & DEBUG_VERTS)
+   if (RADEON_DEBUG & RADEON_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
 	      __FUNCTION__, count, stride);
 
-   if (stride == 4)
-       COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 *out++ = LE32_TO_CPU(*(int *)data);
-	 data += stride;
-      }
-}
-
-
-static void emit_ubyte_rgba( GLcontext *ctx,
-			     struct radeon_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
-
-   assert (!rvb->buf);
-
    if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4, 4 );
+      radeonAllocDmaRegion( rmesa, &aos->bo, &aos->offset, size * 4, 32 );
       count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
+      aos->stride = 0;
    }
    else {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * count, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 3:
-      emit_ubyte_rgba3( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_ubyte_rgba4( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+      aos->stride = size;
    }
-}
-#endif
-
-#if defined(USE_X86_ASM)
-#define COPY_DWORDS( dst, src, nr )					\
-do {									\
-	int __tmp;							\
-	__asm__ __volatile__( "rep ; movsl"				\
-			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
-			      : "0" (nr),				\
-			        "D" ((long)dst),			\
-			        "S" ((long)src) );			\
-} while (0)
-#else
-#define COPY_DWORDS( dst, src, nr )		\
-do {						\
-   int j;					\
-   for ( j = 0 ; j < nr ; j++ )			\
-      dst[j] = ((int *)src)[j];			\
-   dst += nr;					\
-} while (0)
-#endif
-
-static void emit_vecfog( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
-{
-   int i;
-   GLfloat *out;
-
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
+   aos->components = size;
+   aos->count = count;
 
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      radeonAllocDmaRegion( rmesa, rvb, count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
 
    /* Emit the data
     */
-   out = (GLfloat *)(rvb->address + rvb->start);
+   out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
    for (i = 0; i < count; i++) {
       out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data );
       out++;
@@ -209,170 +84,10 @@ static void emit_vecfog( GLcontext *ctx,
    }
 }
 
-static void emit_vec4( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4)
-      COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out++;
-	 data += stride;
-      }
-}
-
-
-static void emit_vec8( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 8)
-      COPY_DWORDS( out, data, count*2 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out += 2;
-	 data += stride;
-      }
-}
-
-static void emit_vec12( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-	      __FUNCTION__, count, stride, (void *)out, (void *)data);
-
-   if (stride == 12)
-      COPY_DWORDS( out, data, count*3 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out += 3;
-	 data += stride;
-      }
-}
-
-static void emit_vec16( GLcontext *ctx,
-			struct radeon_dma_region *rvb,
-			char *data,
-			int stride,
-			int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 16)
-      COPY_DWORDS( out, data, count*4 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out[3] = *(int *)(data+12);
-	 out += 4;
-	 data += stride;
-      }
-}
-
-
-static void emit_vector( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int size,
-			 int stride,
-			 int count )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d size %d stride %d\n",
-	      __FUNCTION__, count, size, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, size * 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = size;
-   }
-   else {
-      radeonAllocDmaRegion( rmesa, rvb, size * count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = size;
-      rvb->aos_size = size;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 1:
-      emit_vec4( ctx, rvb, data, stride, count );
-      break;
-   case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
-      break;
-   case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_vec16( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-
-}
-
-
-
-static void emit_s0_vec( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count)
 {
    int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
+   if (RADEON_DEBUG & RADEON_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
 	      __FUNCTION__, count, stride);
 
@@ -384,16 +99,11 @@ static void emit_s0_vec( GLcontext *ctx,
    }
 }
 
-static void emit_stq_vec( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void emit_stq_vec(uint32_t *out, GLvoid *data, int stride, int count)
 {
    int i;
-   int *out = (int *)(rvb->address + rvb->start);
 
-   if (RADEON_DEBUG & DEBUG_VERTS)
+   if (RADEON_DEBUG & RADEON_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
 	      __FUNCTION__, count, stride);
 
@@ -409,21 +119,16 @@ static void emit_stq_vec( GLcontext *ctx,
 
 
 
-static void emit_tex_vector( GLcontext *ctx,
-			     struct radeon_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
+static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos,
+			    GLvoid *data, int size, int stride, int count)
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
    int emitsize;
+   uint32_t *out;
 
-   if (RADEON_DEBUG & DEBUG_VERTS)
+   if (RADEON_DEBUG & RADEON_VERTS)
       fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
 
-   assert (!rvb->buf);
-
    switch (size) {
    case 4: emitsize = 3; break;
    case 3: emitsize = 3; break;
@@ -432,34 +137,33 @@ static void emit_tex_vector( GLcontext *ctx,
 
 
    if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * emitsize, 4 );
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, emitsize * 4, 32);
       count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = emitsize;
+      aos->stride = 0;
    }
    else {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * emitsize * count, 4 );
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = emitsize;
-      rvb->aos_size = emitsize;
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, emitsize * count * 4, 32);
+      aos->stride = emitsize;
    }
 
+   aos->components = emitsize;
+   aos->count = count;
 
    /* Emit the data
     */
+   out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
    switch (size) {
    case 1:
-      emit_s0_vec( ctx, rvb, data, stride, count ); 
+      emit_s0_vec( out, data, stride, count );
       break;
    case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
+      radeonEmitVec8( out, data, stride, count );
       break;
    case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
+      radeonEmitVec12( out, data, stride, count );
       break;
    case 4:
-      emit_stq_vec( ctx, rvb, data, stride, count );
+      emit_stq_vec( out, data, stride, count );
       break;
    default:
       assert(0);
@@ -476,27 +180,26 @@ static void emit_tex_vector( GLcontext *ctx,
  */
 void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
-   struct radeon_dma_region **component = rmesa->tcl.aos_components;
    GLuint nr = 0;
    GLuint vfmt = 0;
    GLuint count = VB->Count;
    GLuint vtx, unit;
    
 #if 0
-   if (RADEON_DEBUG & DEBUG_VERTS) 
+   if (RADEON_DEBUG & RADEON_VERTS)
       _tnl_print_vert_flags( __FUNCTION__, inputs );
 #endif
 
    if (1) {
       if (!rmesa->tcl.obj.buf) 
-	 emit_vector( ctx, 
-		      &rmesa->tcl.obj, 
-		      (char *)VB->ObjPtr->data,
-		      VB->ObjPtr->size,
-		      VB->ObjPtr->stride,
-		      count);
+	rcommon_emit_vector( ctx, 
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->ObjPtr->data,
+			     VB->ObjPtr->size,
+			     VB->ObjPtr->stride,
+			     count);
 
       switch( VB->ObjPtr->size ) {
       case 4: vfmt |= RADEON_CP_VC_FRMT_W0;
@@ -505,21 +208,21 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       default:
          break;
       }
-      component[nr++] = &rmesa->tcl.obj;
+      nr++;
    }
    
 
    if (inputs & VERT_BIT_NORMAL) {
       if (!rmesa->tcl.norm.buf)
-	 emit_vector( ctx, 
-		      &(rmesa->tcl.norm), 
-		      (char *)VB->NormalPtr->data,
-		      3,
-		      VB->NormalPtr->stride,
-		      count);
+	 rcommon_emit_vector( ctx, 
+			      &(rmesa->tcl.aos[nr]),
+			      (char *)VB->NormalPtr->data,
+			      3,
+			      VB->NormalPtr->stride,
+			      count);
 
       vfmt |= RADEON_CP_VC_FRMT_N0;
-      component[nr++] = &rmesa->tcl.norm;
+      nr++;
    }
 
    if (inputs & VERT_BIT_COLOR0) {
@@ -537,31 +240,30 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       }
 
       if (!rmesa->tcl.rgba.buf)
-	 emit_vector( ctx,
-		      &(rmesa->tcl.rgba),
-		      (char *)VB->ColorPtr[0]->data,
-		      emitsize,
-		      VB->ColorPtr[0]->stride,
-		      count);
-
-
-      component[nr++] = &rmesa->tcl.rgba;
+	rcommon_emit_vector( ctx,
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->ColorPtr[0]->data,
+			     emitsize,
+			     VB->ColorPtr[0]->stride,
+			     count);
+
+      nr++;
    }
 
 
    if (inputs & VERT_BIT_COLOR1) {
       if (!rmesa->tcl.spec.buf) {
 
-	 emit_vector( ctx,
-		      &rmesa->tcl.spec,
-		      (char *)VB->SecondaryColorPtr[0]->data,
-		      3,
-		      VB->SecondaryColorPtr[0]->stride,
-		      count);
+	rcommon_emit_vector( ctx,
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->SecondaryColorPtr[0]->data,
+			     3,
+			     VB->SecondaryColorPtr[0]->stride,
+			     count);
       }
 
       vfmt |= RADEON_CP_VC_FRMT_FPSPEC;
-      component[nr++] = &rmesa->tcl.spec;
+      nr++;
    }
 
 /* FIXME: not sure if this is correct. May need to stitch this together with
@@ -570,13 +272,13 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
    if (inputs & VERT_BIT_FOG) {
       if (!rmesa->tcl.fog.buf)
 	 emit_vecfog( ctx,
-		      &(rmesa->tcl.fog),
+		      &(rmesa->tcl.aos[nr]),
 		      (char *)VB->FogCoordPtr->data,
 		      VB->FogCoordPtr->stride,
 		      count);
 
       vfmt |= RADEON_CP_VC_FRMT_FPFOG;
-      component[nr++] = &rmesa->tcl.fog;
+      nr++;
    }
 
 
@@ -587,11 +289,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       if (inputs & VERT_BIT_TEX(unit)) {
 	 if (!rmesa->tcl.tex[unit].buf)
 	    emit_tex_vector( ctx,
-			     &(rmesa->tcl.tex[unit]),
+			     &(rmesa->tcl.aos[nr]),
 			     (char *)VB->TexCoordPtr[unit]->data,
 			     VB->TexCoordPtr[unit]->size,
 			     VB->TexCoordPtr[unit]->stride,
 			     count );
+	 nr++;
 
 	 vfmt |= RADEON_ST_BIT(unit);
          /* assume we need the 3rd coord if texgen is active for r/q OR at least
@@ -609,7 +312,6 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 		 (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1)))
 	       radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ;
 	 }
-	 component[nr++] = &rmesa->tcl.tex[unit];
       }
    }
 
@@ -622,34 +324,3 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
    rmesa->tcl.vertex_format = vfmt;
 }
 
-
-void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   GLuint unit;
-
-#if 0
-   if (RADEON_DEBUG & DEBUG_VERTS) 
-      _tnl_print_vert_flags( __FUNCTION__, newinputs );
-#endif
-
-   if (newinputs & VERT_BIT_POS) 
-     radeonReleaseDmaRegion( rmesa, &rmesa->tcl.obj, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_NORMAL) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.norm, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR0) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.rgba, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR1) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.spec, __FUNCTION__ );
-      
-   if (newinputs & VERT_BIT_FOG)
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.fog, __FUNCTION__ );
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (newinputs & VERT_BIT_TEX(unit))
-         radeonReleaseDmaRegion( rmesa, &rmesa->tcl.tex[unit], __FUNCTION__ );
-   }
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h
index 034cda8a65..515783135d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h
@@ -54,8 +54,7 @@ static void TAG(emit)( GLcontext *ctx,
 
    union emit_union *v = (union emit_union *)dest;
 
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s\n", __FUNCTION__); 
+   radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__);
 
    coord = (GLuint (*)[4])VB->ObjPtr->data;
    coord_stride = VB->ObjPtr->stride;
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
index 126d0727c6..78ec119302 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
@@ -310,7 +310,7 @@ static void init_tcl_verts( void )
 
 void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
    GLuint req = 0;
    GLuint unit;
@@ -374,14 +374,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 	 break;
 
    if (rmesa->tcl.vertex_format == setup_tab[i].vertex_format &&
-       rmesa->tcl.indexed_verts.buf)
+       rmesa->radeon.tcl.aos[0].bo)
       return;
 
-   if (rmesa->tcl.indexed_verts.buf)
+   if (rmesa->radeon.tcl.aos[0].bo)
       radeonReleaseArrays( ctx, ~0 );
 
-   radeonAllocDmaRegion( rmesa,
-			 &rmesa->tcl.indexed_verts, 
+   radeonAllocDmaRegion( &rmesa->radeon,
+			 &rmesa->radeon.tcl.aos[0].bo,
+			 &rmesa->radeon.tcl.aos[0].offset,
 			 VB->Count * setup_tab[i].vertex_size * 4, 
 			 4);
 
@@ -421,29 +422,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 
 
    setup_tab[i].emit( ctx, 0, VB->Count, 
-		      rmesa->tcl.indexed_verts.address + 
-		      rmesa->tcl.indexed_verts.start );
+		      rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset);
 
+   //   rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size;
+   rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size;
    rmesa->tcl.vertex_format = setup_tab[i].vertex_format;
-   rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts );
-   rmesa->tcl.indexed_verts.aos_size = setup_tab[i].vertex_size;
-   rmesa->tcl.indexed_verts.aos_stride = setup_tab[i].vertex_size;
-
-   rmesa->tcl.aos_components[0] = &rmesa->tcl.indexed_verts;
-   rmesa->tcl.nr_aos_components = 1;
+   rmesa->radeon.tcl.aos_count = 1;
 }
 
 
-
-void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-#if 0
-   if (RADEON_DEBUG & DEBUG_VERTS) 
-      _tnl_print_vert_flags( __FUNCTION__, newinputs );
-#endif
-
-   if (newinputs) 
-     radeonReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ );
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
new file mode 100644
index 0000000000..38db305e2a
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -0,0 +1,421 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "radeon_mipmap_tree.h"
+
+#include <errno.h>
+#include <unistd.h>
+
+#include "main/simple_list.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+
+static GLuint radeon_compressed_texture_size(GLcontext *ctx,
+		GLsizei width, GLsizei height, GLsizei depth,
+		GLuint mesaFormat)
+{
+	GLuint size = _mesa_compressed_texture_size(ctx, width, height, depth, mesaFormat);
+
+	if (mesaFormat == MESA_FORMAT_RGB_DXT1 ||
+	    mesaFormat == MESA_FORMAT_RGBA_DXT1) {
+		if (width + 3 < 8)	/* width one block */
+			size = size * 4;
+		else if (width + 3 < 16)
+			size = size * 2;
+	} else {
+		/* DXT3/5, 16 bytes per block */
+	  //		WARN_ONCE("DXT 3/5 suffers from multitexturing problems!\n");
+		if (width + 3 < 8)
+			size = size * 2;
+	}
+
+	return size;
+}
+
+
+static int radeon_compressed_num_bytes(GLuint mesaFormat)
+{
+   int bytes = 0;
+   switch(mesaFormat) {
+     
+   case MESA_FORMAT_RGB_FXT1:
+   case MESA_FORMAT_RGBA_FXT1:
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_RGBA_DXT1:
+     bytes = 2;
+     break;
+     
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_RGBA_DXT5:
+     bytes = 4;
+   default:
+     break;
+   }
+   
+   return bytes;
+}
+
+/**
+ * Compute sizes and fill in offset and blit information for the given
+ * image (determined by \p face and \p level).
+ *
+ * \param curOffset points to the offset at which the image is to be stored
+ * and is updated by this function according to the size of the image.
+ */
+static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree *mt,
+	GLuint face, GLuint level, GLuint* curOffset)
+{
+	radeon_mipmap_level *lvl = &mt->levels[level];
+	uint32_t row_align;
+
+	/* Find image size in bytes */
+	if (mt->compressed) {
+		/* TODO: Is this correct? Need test cases for compressed textures! */
+		row_align = rmesa->texture_compressed_row_align - 1;
+		lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align;
+		lvl->size = radeon_compressed_texture_size(mt->radeon->glCtx,
+							   lvl->width, lvl->height, lvl->depth, mt->compressed);
+	} else if (mt->target == GL_TEXTURE_RECTANGLE_NV) {
+		row_align = rmesa->texture_rect_row_align - 1;
+		lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align;
+		lvl->size = lvl->rowstride * lvl->height;
+	} else if (mt->tilebits & RADEON_TXO_MICRO_TILE) {
+		/* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
+		 * though the actual offset may be different (if texture is less than
+		 * 32 bytes width) to the untiled case */
+		lvl->rowstride = (lvl->width * mt->bpp * 2 + 31) & ~31;
+		lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth;
+	} else {
+		row_align = rmesa->texture_row_align - 1;
+		lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align;
+		lvl->size = lvl->rowstride * lvl->height * lvl->depth;
+	}
+	assert(lvl->size > 0);
+
+	/* All images are aligned to a 32-byte offset */
+	*curOffset = (*curOffset + 0x1f) & ~0x1f;
+	lvl->faces[face].offset = *curOffset;
+	*curOffset += lvl->size;
+
+	if (RADEON_DEBUG & RADEON_TEXTURE)
+	  fprintf(stderr,
+		  "level %d, face %d: rs:%d %dx%d at %d\n",
+		  level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset);
+}
+
+static GLuint minify(GLuint size, GLuint levels)
+{
+	size = size >> levels;
+	if (size < 1)
+		size = 1;
+	return size;
+}
+
+
+static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
+{
+	GLuint curOffset;
+	GLuint numLevels;
+	GLuint i;
+	GLuint face;
+
+	numLevels = mt->lastLevel - mt->firstLevel + 1;
+	assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
+
+	curOffset = 0;
+	for(face = 0; face < mt->faces; face++) {
+
+		for(i = 0; i < numLevels; i++) {
+			mt->levels[i].width = minify(mt->width0, i);
+			mt->levels[i].height = minify(mt->height0, i);
+			mt->levels[i].depth = minify(mt->depth0, i);
+			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+		}
+	}
+
+	/* Note the required size in memory */
+	mt->totalsize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
+}
+
+static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
+{
+	GLuint curOffset;
+	GLuint numLevels;
+	GLuint i;
+
+	numLevels = mt->lastLevel - mt->firstLevel + 1;
+	assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
+
+	curOffset = 0;
+	for(i = 0; i < numLevels; i++) {
+		GLuint face;
+
+		mt->levels[i].width = minify(mt->width0, i);
+		mt->levels[i].height = minify(mt->height0, i);
+		mt->levels[i].depth = minify(mt->depth0, i);
+
+		for(face = 0; face < mt->faces; face++)
+			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+	}
+
+	/* Note the required size in memory */
+	mt->totalsize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
+}
+
+/**
+ * Create a new mipmap tree, calculate its layout and allocate memory.
+ */
+radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t,
+		GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel,
+		GLuint width0, GLuint height0, GLuint depth0,
+		GLuint bpp, GLuint tilebits, GLuint compressed)
+{
+	radeon_mipmap_tree *mt = CALLOC_STRUCT(_radeon_mipmap_tree);
+
+	mt->radeon = rmesa;
+	mt->internal_format = internal_format;
+	mt->refcount = 1;
+	mt->t = t;
+	mt->target = target;
+	mt->faces = (target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+	mt->firstLevel = firstLevel;
+	mt->lastLevel = lastLevel;
+	mt->width0 = width0;
+	mt->height0 = height0;
+	mt->depth0 = depth0;
+	mt->bpp = compressed ? radeon_compressed_num_bytes(compressed) : bpp;
+	mt->tilebits = tilebits;
+	mt->compressed = compressed;
+
+	if (rmesa->radeonScreen->chip_family >= CHIP_FAMILY_R300)
+		calculate_miptree_layout_r300(rmesa, mt);
+	else
+		calculate_miptree_layout_r100(rmesa, mt);
+
+	mt->bo = radeon_bo_open(rmesa->radeonScreen->bom,
+                            0, mt->totalsize, 1024,
+                            RADEON_GEM_DOMAIN_VRAM,
+                            0);
+
+	return mt;
+}
+
+void radeon_miptree_reference(radeon_mipmap_tree *mt)
+{
+	mt->refcount++;
+	assert(mt->refcount > 0);
+}
+
+void radeon_miptree_unreference(radeon_mipmap_tree *mt)
+{
+	if (!mt)
+		return;
+
+	assert(mt->refcount > 0);
+	mt->refcount--;
+	if (!mt->refcount) {
+		radeon_bo_unref(mt->bo);
+		free(mt);
+	}
+}
+
+
+/**
+ * Calculate first and last mip levels for the given texture object,
+ * where the dimensions are taken from the given texture image at
+ * the given level.
+ *
+ * Note: level is the OpenGL level number, which is not necessarily the same
+ * as the first level that is actually present.
+ *
+ * The base level image of the given texture face must be non-null,
+ * or this will fail.
+ */
+static void calculate_first_last_level(struct gl_texture_object *tObj,
+				       GLuint *pfirstLevel, GLuint *plastLevel,
+				       GLuint face, GLuint level)
+{
+	const struct gl_texture_image * const baseImage =
+		tObj->Image[face][level];
+
+	assert(baseImage);
+	
+	/* These must be signed values.  MinLod and MaxLod can be negative numbers,
+	* and having firstLevel and lastLevel as signed prevents the need for
+	* extra sign checks.
+	*/
+	int   firstLevel;
+	int   lastLevel;
+
+	/* Yes, this looks overly complicated, but it's all needed.
+	*/
+	switch (tObj->Target) {
+	case GL_TEXTURE_1D:
+	case GL_TEXTURE_2D:
+	case GL_TEXTURE_3D:
+	case GL_TEXTURE_CUBE_MAP:
+		if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) {
+			/* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL.
+			*/
+			firstLevel = lastLevel = tObj->BaseLevel;
+		} else {
+			firstLevel = tObj->BaseLevel + (GLint)(tObj->MinLod + 0.5);
+			firstLevel = MAX2(firstLevel, tObj->BaseLevel);
+			firstLevel = MIN2(firstLevel, level + baseImage->MaxLog2);
+			lastLevel = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5);
+			lastLevel = MAX2(lastLevel, tObj->BaseLevel);
+			lastLevel = MIN2(lastLevel, level + baseImage->MaxLog2);
+			lastLevel = MIN2(lastLevel, tObj->MaxLevel);
+			lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */
+		}
+		break;
+	case GL_TEXTURE_RECTANGLE_NV:
+	case GL_TEXTURE_4D_SGIS:
+		firstLevel = lastLevel = 0;
+		break;
+	default:
+		return;
+	}
+
+	/* save these values */
+	*pfirstLevel = firstLevel;
+	*plastLevel = lastLevel;
+}
+
+
+/**
+ * Checks whether the given miptree can hold the given texture image at the
+ * given face and level.
+ */
+GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
+		struct gl_texture_image *texImage, GLuint face, GLuint level)
+{
+	radeon_mipmap_level *lvl;
+
+	if (face >= mt->faces || level < mt->firstLevel || level > mt->lastLevel)
+		return GL_FALSE;
+
+	if (texImage->InternalFormat != mt->internal_format ||
+	    texImage->IsCompressed != mt->compressed)
+		return GL_FALSE;
+
+	if (!texImage->IsCompressed &&
+	    !mt->compressed &&
+	    texImage->TexFormat->TexelBytes != mt->bpp)
+		return GL_FALSE;
+
+	lvl = &mt->levels[level - mt->firstLevel];
+	if (lvl->width != texImage->Width ||
+	    lvl->height != texImage->Height ||
+	    lvl->depth != texImage->Depth)
+		return GL_FALSE;
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Checks whether the given miptree has the right format to store the given texture object.
+ */
+GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj)
+{
+	struct gl_texture_image *firstImage;
+	GLuint compressed;
+	GLuint numfaces = 1;
+	GLuint firstLevel, lastLevel;
+
+	calculate_first_last_level(texObj, &firstLevel, &lastLevel, 0, texObj->BaseLevel);
+	if (texObj->Target == GL_TEXTURE_CUBE_MAP)
+		numfaces = 6;
+
+	firstImage = texObj->Image[0][firstLevel];
+	compressed = firstImage->IsCompressed ? firstImage->TexFormat->MesaFormat : 0;
+
+	return (mt->firstLevel == firstLevel &&
+	        mt->lastLevel == lastLevel &&
+	        mt->width0 == firstImage->Width &&
+	        mt->height0 == firstImage->Height &&
+	        mt->depth0 == firstImage->Depth &&
+	        mt->compressed == compressed &&
+	        (!mt->compressed ? (mt->bpp == firstImage->TexFormat->TexelBytes) : 1));
+}
+
+
+/**
+ * Try to allocate a mipmap tree for the given texture that will fit the
+ * given image in the given position.
+ */
+void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
+		radeon_texture_image *image, GLuint face, GLuint level)
+{
+	GLuint compressed = image->base.IsCompressed ? image->base.TexFormat->MesaFormat : 0;
+	GLuint numfaces = 1;
+	GLuint firstLevel, lastLevel;
+
+	assert(!t->mt);
+
+	calculate_first_last_level(&t->base, &firstLevel, &lastLevel, face, level);
+	if (t->base.Target == GL_TEXTURE_CUBE_MAP)
+		numfaces = 6;
+
+	if (level != firstLevel || face >= numfaces)
+		return;
+
+	t->mt = radeon_miptree_create(rmesa, t, t->base.Target,
+		image->base.InternalFormat,
+		firstLevel, lastLevel,
+		image->base.Width, image->base.Height, image->base.Depth,
+		image->base.TexFormat->TexelBytes, t->tile_bits, compressed);
+}
+
+/* Although we use the image_offset[] array to store relative offsets
+ * to cube faces, Mesa doesn't know anything about this and expects
+ * each cube face to be treated as a separate image.
+ *
+ * These functions present that view to mesa:
+ */
+void
+radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets)
+{
+     if (mt->target != GL_TEXTURE_3D || mt->faces == 1)
+        offsets[0] = 0;
+     else {
+	int i;
+	for (i = 0; i < 6; i++)
+		offsets[i] = mt->levels[level].faces[i].offset;
+     }
+}
+
+GLuint
+radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+			    GLuint face, GLuint level)
+{
+   if (mt->target == GL_TEXTURE_CUBE_MAP_ARB)
+      return (mt->levels[level].faces[face].offset);
+   else
+      return mt->levels[level].faces[0].offset;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
new file mode 100644
index 0000000000..db28252da3
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
@@ -0,0 +1,101 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __RADEON_MIPMAP_TREE_H_
+#define __RADEON_MIPMAP_TREE_H_
+
+#include "radeon_common.h"
+
+typedef struct _radeon_mipmap_tree radeon_mipmap_tree;
+typedef struct _radeon_mipmap_level radeon_mipmap_level;
+typedef struct _radeon_mipmap_image radeon_mipmap_image;
+
+struct _radeon_mipmap_image {
+	GLuint offset; /** Offset of this image from the start of mipmap tree buffer, in bytes */
+};
+
+struct _radeon_mipmap_level {
+	GLuint width;
+	GLuint height;
+	GLuint depth;
+	GLuint size; /** Size of each image, in bytes */
+	GLuint rowstride; /** in bytes */
+	radeon_mipmap_image faces[6];
+};
+
+/* store the max possible in the miptree */
+#define RADEON_MIPTREE_MAX_TEXTURE_LEVELS 13
+
+/**
+ * A mipmap tree contains texture images in the layout that the hardware
+ * expects.
+ *
+ * The meta-data of mipmap trees is immutable, i.e. you cannot change the
+ * layout on-the-fly; however, the texture contents (i.e. texels) can be
+ * changed.
+ */
+struct _radeon_mipmap_tree {
+	radeonContextPtr radeon;
+	radeonTexObj *t;
+	struct radeon_bo *bo;
+	GLuint refcount;
+
+	GLuint totalsize; /** total size of the miptree, in bytes */
+
+	GLenum target; /** GL_TEXTURE_xxx */
+	GLenum internal_format;
+	GLuint faces; /** # of faces: 6 for cubemaps, 1 otherwise */
+	GLuint firstLevel; /** First mip level stored in this mipmap tree */
+	GLuint lastLevel; /** Last mip level stored in this mipmap tree */
+
+	GLuint width0; /** Width of firstLevel image */
+	GLuint height0; /** Height of firstLevel image */
+	GLuint depth0; /** Depth of firstLevel image */
+
+	GLuint bpp; /** Bytes per texel */
+	GLuint tilebits; /** RADEON_TXO_xxx_TILE */
+	GLuint compressed; /** MESA_FORMAT_xxx indicating a compressed format, or 0 if uncompressed */
+
+	radeon_mipmap_level levels[RADEON_MIPTREE_MAX_TEXTURE_LEVELS];
+};
+
+radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t,
+		GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel,
+		GLuint width0, GLuint height0, GLuint depth0,
+		GLuint bpp, GLuint tilebits, GLuint compressed);
+void radeon_miptree_reference(radeon_mipmap_tree *mt);
+void radeon_miptree_unreference(radeon_mipmap_tree *mt);
+
+GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
+		struct gl_texture_image *texImage, GLuint face, GLuint level);
+GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj);
+void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
+			      radeon_texture_image *texImage, GLuint face, GLuint level);
+GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+				   GLuint face, GLuint level);
+void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets);
+#endif /* __RADEON_MIPMAP_TREE_H_ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
new file mode 100644
index 0000000000..b79d864ba2
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
@@ -0,0 +1,235 @@
+/*
+ * Copyright © 2008-2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Maciej Cencora <m.cencora@gmail.com>
+ *
+ */
+#include "radeon_common.h"
+#include "radeon_queryobj.h"
+#include "radeon_debug.h"
+
+#include "main/imports.h"
+#include "main/simple_list.h"
+
+static int radeonQueryIsFlushed(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q;
+
+	foreach(tmp, &radeon->query.not_flushed_head) {
+		if (tmp == query) {
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
+static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q)
+{
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+	uint32_t *result;
+	int i;
+
+	radeon_print(RADEON_STATE, RADEON_VERBOSE,
+			"%s: query id %d, result %d\n",
+			__FUNCTION__, query->Base.Id, (int) query->Base.Result);
+
+	radeon_bo_map(query->bo, GL_FALSE);
+
+	result = query->bo->ptr;
+
+	query->Base.Result = 0;
+	for (i = 0; i < query->curr_offset/sizeof(uint32_t); ++i) {
+		query->Base.Result += result[i];
+		radeon_print(RADEON_STATE, RADEON_TRACE, "result[%d] = %d\n", i, result[i]);
+	}
+
+	radeon_bo_unmap(query->bo);
+}
+
+static struct gl_query_object * radeonNewQueryObject(GLcontext *ctx, GLuint id)
+{
+	struct radeon_query_object *query;
+
+	query = _mesa_calloc(sizeof(struct radeon_query_object));
+
+	query->Base.Id = id;
+	query->Base.Result = 0;
+	query->Base.Active = GL_FALSE;
+	query->Base.Ready = GL_TRUE;
+
+	radeon_print(RADEON_STATE, RADEON_VERBOSE,"%s: query id %d\n", __FUNCTION__, query->Base.Id);
+
+	return &query->Base;
+}
+
+static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	if (query->bo) {
+		radeon_bo_unref(query->bo);
+	}
+
+	_mesa_free(query);
+}
+
+static void radeonWaitQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+
+	/* If the cmdbuf with packets for this query hasn't been flushed yet, do it now */
+	if (!radeonQueryIsFlushed(ctx, q))
+		ctx->Driver.Flush(ctx);
+
+	radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, q->Id, query->bo, query->curr_offset);
+
+	radeonQueryGetResult(ctx, q);
+
+	query->Base.Ready = GL_TRUE;
+}
+
+
+static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	assert(radeon->query.current == NULL);
+
+	if (radeon->dma.flush)
+		radeon->dma.flush(radeon->glCtx);
+
+	if (!query->bo) {
+		query->bo = radeon_bo_open(radeon->radeonScreen->bom, 0, RADEON_QUERY_PAGE_SIZE, RADEON_QUERY_PAGE_SIZE, RADEON_GEM_DOMAIN_GTT, 0);
+	}
+	query->curr_offset = 0;
+
+	radeon->query.current = query;
+
+	radeon->query.queryobj.dirty = GL_TRUE;
+	radeon->hw.is_dirty = GL_TRUE;
+	insert_at_tail(&radeon->query.not_flushed_head, query);
+
+}
+
+void radeonEmitQueryEnd(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = radeon->query.current;
+
+	if (!query)
+		return;
+
+	if (query->emitted_begin == GL_FALSE)
+		return;
+
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, query->Base.Id, query->bo, query->curr_offset);
+
+	radeon_cs_space_check_with_bo(radeon->cmdbuf.cs,
+				      query->bo,
+				      0, RADEON_GEM_DOMAIN_GTT);
+
+	radeon->vtbl.emit_query_finish(radeon);
+}
+
+static void radeonEndQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+	radeon_print(RADEON_STATE, RADEON_NORMAL, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	if (radeon->dma.flush)
+		radeon->dma.flush(radeon->glCtx);
+	radeonEmitQueryEnd(ctx);
+
+	radeon->query.current = NULL;
+}
+
+static void radeonCheckQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeon_print(RADEON_STATE, RADEON_TRACE, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+#ifdef DRM_RADEON_GEM_BUSY
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+	if (radeon->radeonScreen->kernel_mm) {
+		struct radeon_query_object *query = (struct radeon_query_object *)q;
+		uint32_t domain;
+
+		/* Need to perform a flush, as per ARB_occlusion_query spec */
+		if (!radeonQueryIsFlushed(ctx, q)) {
+			ctx->Driver.Flush(ctx);
+		}
+
+		if (radeon_bo_is_busy(query->bo, &domain) == 0) {
+			radeonQueryGetResult(ctx, q);
+			query->Base.Ready = GL_TRUE;
+		}
+	} else {
+		radeonWaitQuery(ctx, q);
+	}
+#else
+	radeonWaitQuery(ctx, q);
+#endif
+}
+
+void radeonInitQueryObjFunctions(struct dd_function_table *functions)
+{
+	functions->NewQueryObject = radeonNewQueryObject;
+	functions->DeleteQuery = radeonDeleteQuery;
+	functions->BeginQuery = radeonBeginQuery;
+	functions->EndQuery = radeonEndQuery;
+	functions->CheckQuery = radeonCheckQuery;
+	functions->WaitQuery = radeonWaitQuery;
+}
+
+int radeon_check_query_active(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = radeon->query.current;
+
+	if (!query || query->emitted_begin)
+		return 0;
+	return atom->cmd_size;
+}
+
+void radeon_emit_queryobj(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	BATCH_LOCALS(radeon);
+	int dwords;
+
+	dwords = (*atom->check) (ctx, atom);
+
+	BEGIN_BATCH_NO_AUTOSTATE(dwords);
+	OUT_BATCH_TABLE(atom->cmd, dwords);
+	END_BATCH();
+
+	radeon->query.current->emitted_begin = GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.h b/src/mesa/drivers/dri/radeon/radeon_queryobj.h
new file mode 100644
index 0000000000..19374dc76b
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.h
@@ -0,0 +1,55 @@
+/*
+ * Copyright © 2008 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Maciej Cencora <m.cencora@gmail.com>
+ *
+ */
+
+#include "main/imports.h"
+#include "main/simple_list.h"
+#include "radeon_common_context.h"
+
+extern void radeonEmitQueryBegin(GLcontext *ctx);
+extern void radeonEmitQueryEnd(GLcontext *ctx);
+
+extern void radeonInitQueryObjFunctions(struct dd_function_table *functions);
+
+#define RADEON_QUERY_PAGE_SIZE 4096
+
+int radeon_check_query_active(GLcontext *ctx, struct radeon_state_atom *atom);
+void radeon_emit_queryobj(GLcontext *ctx, struct radeon_state_atom *atom);
+
+static inline void radeon_init_query_stateobj(radeonContextPtr radeon, int SZ)
+{
+	radeon->query.queryobj.cmd_size = (SZ);
+	radeon->query.queryobj.cmd = (uint32_t*)CALLOC((SZ) * sizeof(uint32_t));
+	radeon->query.queryobj.name = "queryobj";
+	radeon->query.queryobj.idx = 0;
+	radeon->query.queryobj.check = radeon_check_query_active;
+	radeon->query.queryobj.dirty = GL_FALSE;
+	radeon->query.queryobj.emit = radeon_emit_queryobj;
+
+	radeon->hw.max_state_size += (SZ);
+	insert_at_tail(&radeon->hw.atomlist, &radeon->query.queryobj);
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.c b/src/mesa/drivers/dri/radeon/radeon_sanity.c
index 6613757fce..1ab570f507 100644
--- a/src/mesa/drivers/dri/radeon/radeon_sanity.c
+++ b/src/mesa/drivers/dri/radeon/radeon_sanity.c
@@ -44,11 +44,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define MORE_VERBOSE 1
 
 #if MORE_VERBOSE
-#define VERBOSE (RADEON_DEBUG & DEBUG_VERBOSE)
+#define VERBOSE (RADEON_DEBUG & RADEON_VERBOSE)
 #define NORMAL  (1)
 #else
 #define VERBOSE 0
-#define NORMAL  (RADEON_DEBUG & DEBUG_VERBOSE)
+#define NORMAL  (RADEON_DEBUG & RADEON_VERBOSE)
 #endif
 
 
@@ -973,7 +973,7 @@ static int radeon_emit_packet3_cliprect( drm_radeon_cmd_buffer_t *cmdbuf )
 }
 
 
-int radeonSanityCmdBuffer( radeonContextPtr rmesa,
+int radeonSanityCmdBuffer( r100ContextPtr rmesa,
 			   int nbox,
 			   drm_clip_rect_t *boxes )
 {
diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.h b/src/mesa/drivers/dri/radeon/radeon_sanity.h
index 1ec06bc586..f30eb1c4f1 100644
--- a/src/mesa/drivers/dri/radeon/radeon_sanity.h
+++ b/src/mesa/drivers/dri/radeon/radeon_sanity.h
@@ -1,7 +1,7 @@
 #ifndef RADEON_SANITY_H
 #define RADEON_SANITY_H
 
-extern int radeonSanityCmdBuffer( radeonContextPtr rmesa,
+extern int radeonSanityCmdBuffer( r100ContextPtr rmesa,
 				  int nbox,
 				  drm_clip_rect_t *boxes );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index a977bedbc2..5ffb55db5e 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -35,6 +35,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \author  Gareth Hughes <gareth@valinux.com>
  */
 
+#include <errno.h>
 #include "main/glheader.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
@@ -45,32 +46,42 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_chipset.h"
 #include "radeon_macros.h"
 #include "radeon_screen.h"
+#include "radeon_common.h"
+#include "radeon_span.h"
 #if !RADEON_COMMON
 #include "radeon_context.h"
-#include "radeon_span.h"
 #include "radeon_tex.h"
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
 #include "r200_context.h"
 #include "r200_ioctl.h"
-#include "r200_span.h"
 #include "r200_tex.h"
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
 #include "r300_context.h"
-#include "r300_fragprog.h"
 #include "r300_tex.h"
-#include "radeon_span.h"
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+#include "r600_context.h"
+#include "r700_driconf.h" /* +r6/r7 */
+#include "r600_tex.h"     /* +r6/r7 */
 #endif
 
 #include "utils.h"
 #include "vblank.h"
 #include "drirenderbuffer.h"
 
+#include "radeon_bocs_wrapper.h"
+
 #include "GL/internal/dri_interface.h"
 
 /* Radeon configuration
  */
 #include "xmlpool.h"
 
+#define DRI_CONF_COMMAND_BUFFER_SIZE(def,min,max) \
+DRI_CONF_OPT_BEGIN_V(command_buffer_size,int,def, # min ":" # max ) \
+        DRI_CONF_DESC(en,"Size of command buffer (in KB)") \
+        DRI_CONF_DESC(de,"Grösse des Befehlspuffers (in KB)") \
+DRI_CONF_OPT_END
+
 #if !RADEON_COMMON	/* R100 */
 PUBLIC const char __driConfigOptions[] =
 DRI_CONF_BEGIN
@@ -80,6 +91,7 @@ DRI_CONF_BEGIN
         DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
         DRI_CONF_MAX_TEXTURE_UNITS(3,2,3)
         DRI_CONF_HYPERZ(false)
+        DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32)
     DRI_CONF_SECTION_END
     DRI_CONF_SECTION_QUALITY
         DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB)
@@ -95,7 +107,7 @@ DRI_CONF_BEGIN
         DRI_CONF_NO_RAST(false)
     DRI_CONF_SECTION_END
 DRI_CONF_END;
-static const GLuint __driNConfigOptions = 14;
+static const GLuint __driNConfigOptions = 15;
 
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
 
@@ -107,6 +119,7 @@ DRI_CONF_BEGIN
         DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
         DRI_CONF_MAX_TEXTURE_UNITS(6,2,6)
         DRI_CONF_HYPERZ(false)
+        DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32)
     DRI_CONF_SECTION_END
     DRI_CONF_SECTION_QUALITY
         DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB)
@@ -126,7 +139,7 @@ DRI_CONF_BEGIN
         DRI_CONF_NV_VERTEX_PROGRAM(false)
     DRI_CONF_SECTION_END
 DRI_CONF_END;
-static const GLuint __driNConfigOptions = 16;
+static const GLuint __driNConfigOptions = 17;
 
 extern const struct dri_extension blend_extensions[];
 extern const struct dri_extension ARB_vp_extension[];
@@ -134,7 +147,10 @@ extern const struct dri_extension NV_vp_extension[];
 extern const struct dri_extension ATI_fs_extension[];
 extern const struct dri_extension point_extensions[];
 
-#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+#elif RADEON_COMMON && (defined(RADEON_COMMON_FOR_R300) || defined(RADEON_COMMON_FOR_R600))
+
+#define DRI_CONF_FP_OPTIMIZATION_SPEED   0
+#define DRI_CONF_FP_OPTIMIZATION_QUALITY 1
 
 /* TODO: integrate these into xmlpool.h! */
 #define DRI_CONF_MAX_TEXTURE_IMAGE_UNITS(def,min,max) \
@@ -149,11 +165,7 @@ DRI_CONF_OPT_BEGIN_V(texture_coord_units,int,def, # min ":" # max ) \
         DRI_CONF_DESC(de,"Anzahl der Texturkoordinateneinheiten") \
 DRI_CONF_OPT_END
 
-#define DRI_CONF_COMMAND_BUFFER_SIZE(def,min,max) \
-DRI_CONF_OPT_BEGIN_V(command_buffer_size,int,def, # min ":" # max ) \
-        DRI_CONF_DESC(en,"Size of command buffer (in KB)") \
-        DRI_CONF_DESC(de,"Grösse des Befehlspuffers (in KB)") \
-DRI_CONF_OPT_END
+
 
 #define DRI_CONF_DISABLE_S3TC(def) \
 DRI_CONF_OPT_BEGIN(disable_s3tc,bool,def) \
@@ -208,47 +220,42 @@ static const GLuint __driNConfigOptions = 17;
 
 extern const struct dri_extension gl_20_extension[];
 
-#ifndef RADEON_DEBUG
-int RADEON_DEBUG = 0;
-
-static const struct dri_debug_control debug_control[] = {
-	{"fall", DEBUG_FALLBACKS},
-	{"tex", DEBUG_TEXTURE},
-	{"ioctl", DEBUG_IOCTL},
-	{"prim", DEBUG_PRIMS},
-	{"vert", DEBUG_VERTS},
-	{"state", DEBUG_STATE},
-	{"code", DEBUG_CODEGEN},
-	{"vfmt", DEBUG_VFMT},
-	{"vtxf", DEBUG_VFMT},
-	{"verb", DEBUG_VERBOSE},
-	{"dri", DEBUG_DRI},
-	{"dma", DEBUG_DMA},
-	{"san", DEBUG_SANITY},
-	{"sync", DEBUG_SYNC},
-	{"pix", DEBUG_PIXEL},
-	{"mem", DEBUG_MEMORY},
-	{"allmsg", ~DEBUG_SYNC}, /* avoid the term "sync" because the parser uses strstr */
-	{NULL, 0}
-};
-#endif /* RADEON_DEBUG */
-
 #endif /* RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) */
 
 extern const struct dri_extension card_extensions[];
+extern const struct dri_extension mm_extensions[];
 
 static int getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo );
 
 static int
-radeonGetParam(int fd, int param, void *value)
+radeonGetParam(__DRIscreenPrivate *sPriv, int param, void *value)
 {
   int ret;
-  drm_radeon_getparam_t gp;
-
-  gp.param = param;
-  gp.value = value;
+  drm_radeon_getparam_t gp = { 0 };
+  struct drm_radeon_info info = { 0 };
+
+  if (sPriv->drm_version.major >= 2) {
+      info.value = (uint64_t)(uintptr_t)value;
+      switch (param) {
+      case RADEON_PARAM_DEVICE_ID:
+          info.request = RADEON_INFO_DEVICE_ID;
+          break;
+      case RADEON_PARAM_NUM_GB_PIPES:
+          info.request = RADEON_INFO_NUM_GB_PIPES;
+          break;
+      case RADEON_PARAM_NUM_Z_PIPES:
+          info.request = RADEON_INFO_NUM_Z_PIPES;
+          break;
+      default:
+          return -EINVAL;
+      }
+      ret = drmCommandWriteRead(sPriv->fd, DRM_RADEON_INFO, &info, sizeof(info));
+  } else {
+      gp.param = param;
+      gp.value = value;
 
-  ret = drmCommandWriteRead( fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+      ret = drmCommandWriteRead(sPriv->fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+  }
   return ret;
 }
 
@@ -335,6 +342,12 @@ static const __DRItexOffsetExtension radeonTexOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
     radeonSetTexOffset,
 };
+
+static const __DRItexBufferExtension radeonTexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   radeonSetTexBuffer,
+   radeonSetTexBuffer2,
+};
 #endif
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
@@ -349,6 +362,12 @@ static const __DRItexOffsetExtension r200texOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
    r200SetTexOffset,
 };
+
+static const __DRItexBufferExtension r200TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r200SetTexBuffer,
+   r200SetTexBuffer2,
+};
 #endif
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
@@ -356,137 +375,32 @@ static const __DRItexOffsetExtension r300texOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
    r300SetTexOffset,
 };
-#endif
 
-/* Create the device specific screen private data struct.
- */
-static radeonScreenPtr
-radeonCreateScreen( __DRIscreenPrivate *sPriv )
-{
-   radeonScreenPtr screen;
-   RADEONDRIPtr dri_priv = (RADEONDRIPtr)sPriv->pDevPriv;
-   unsigned char *RADEONMMIO;
-   int i;
-   int ret;
-   uint32_t temp = 0;
-
-   if (sPriv->devPrivSize != sizeof(RADEONDRIRec)) {
-      fprintf(stderr,"\nERROR!  sizeof(RADEONDRIRec) does not match passed size from device driver\n");
-      return GL_FALSE;
-   }
-
-   /* Allocate the private area */
-   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
-   if ( !screen ) {
-      __driUtilMessage("%s: Could not allocate memory for screen structure",
-		       __FUNCTION__);
-      return NULL;
-   }
-
-#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
-	RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control);
+static const __DRItexBufferExtension r300TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r300SetTexBuffer,
+   r300SetTexBuffer2,
+};
 #endif
 
-   /* parse information in __driConfigOptions */
-   driParseOptionInfo (&screen->optionCache,
-		       __driConfigOptions, __driNConfigOptions);
-
-   /* This is first since which regions we map depends on whether or
-    * not we are using a PCI card.
-    */
-   screen->card_type = (dri_priv->IsPCI ? RADEON_CARD_PCI : RADEON_CARD_AGP);
-   {
-      int ret;
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_GART_BUFFER_OFFSET,
-			    &screen->gart_buffer_offset);
-
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BUFFER_OFFSET): %d\n", ret);
-	 return NULL;
-      }
-
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_GART_BASE,
-			    &screen->gart_base);
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BASE): %d\n", ret);
-	 return NULL;
-      }
-
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_IRQ_NR,
-			    &screen->irq);
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_IRQ_NR): %d\n", ret);
-	 return NULL;
-      }
-      screen->drmSupportsCubeMapsR200 = (sPriv->drm_version.minor >= 7);
-      screen->drmSupportsBlendColor = (sPriv->drm_version.minor >= 11);
-      screen->drmSupportsTriPerf = (sPriv->drm_version.minor >= 16);
-      screen->drmSupportsFragShader = (sPriv->drm_version.minor >= 18);
-      screen->drmSupportsPointSprites = (sPriv->drm_version.minor >= 13);
-      screen->drmSupportsCubeMapsR100 = (sPriv->drm_version.minor >= 15);
-      screen->drmSupportsVertexProgram = (sPriv->drm_version.minor >= 25);
-   }
-
-   screen->mmio.handle = dri_priv->registerHandle;
-   screen->mmio.size   = dri_priv->registerSize;
-   if ( drmMap( sPriv->fd,
-		screen->mmio.handle,
-		screen->mmio.size,
-		&screen->mmio.map ) ) {
-      FREE( screen );
-      __driUtilMessage("%s: drmMap failed\n", __FUNCTION__ );
-      return NULL;
-   }
-
-   RADEONMMIO = screen->mmio.map;
-
-   screen->status.handle = dri_priv->statusHandle;
-   screen->status.size   = dri_priv->statusSize;
-   if ( drmMap( sPriv->fd,
-		screen->status.handle,
-		screen->status.size,
-		&screen->status.map ) ) {
-      drmUnmap( screen->mmio.map, screen->mmio.size );
-      FREE( screen );
-      __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__ );
-      return NULL;
-   }
-   screen->scratch = (__volatile__ uint32_t *)
-      ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET);
-
-   screen->buffers = drmMapBufs( sPriv->fd );
-   if ( !screen->buffers ) {
-      drmUnmap( screen->status.map, screen->status.size );
-      drmUnmap( screen->mmio.map, screen->mmio.size );
-      FREE( screen );
-      __driUtilMessage("%s: drmMapBufs failed\n", __FUNCTION__ );
-      return NULL;
-   }
-
-   if ( dri_priv->gartTexHandle && dri_priv->gartTexMapSize ) {
-      screen->gartTextures.handle = dri_priv->gartTexHandle;
-      screen->gartTextures.size   = dri_priv->gartTexMapSize;
-      if ( drmMap( sPriv->fd,
-		   screen->gartTextures.handle,
-		   screen->gartTextures.size,
-		   (drmAddressPtr)&screen->gartTextures.map ) ) {
-	 drmUnmapBufs( screen->buffers );
-	 drmUnmap( screen->status.map, screen->status.size );
-	 drmUnmap( screen->mmio.map, screen->mmio.size );
-	 FREE( screen );
-	 __driUtilMessage("%s: drmMap failed for GART texture area\n", __FUNCTION__);
-	 return NULL;
-      }
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+static const __DRItexOffsetExtension r600texOffsetExtension = {
+    { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
+   r600SetTexOffset, /* +r6/r7 */
+};
 
-      screen->gart_texture_offset = dri_priv->gartTexOffset + screen->gart_base;
-   }
+static const __DRItexBufferExtension r600TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r600SetTexBuffer,  /* +r6/r7 */
+   r600SetTexBuffer2, /* +r6/r7 */
+};
+#endif
 
+static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id)
+{
+   screen->device_id = device_id;
    screen->chip_flags = 0;
-   /* XXX: add more chipsets */
-   switch ( dri_priv->deviceID ) {
+   switch ( device_id ) {
    case PCI_CHIP_RADEON_LY:
    case PCI_CHIP_RADEON_LZ:
    case PCI_CHIP_RADEON_QY:
@@ -561,7 +475,6 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
       screen->chip_family = CHIP_FAMILY_RS300;
       break;
 
-
    case PCI_CHIP_R300_AD:
    case PCI_CHIP_R300_AE:
    case PCI_CHIP_R300_AF:
@@ -819,11 +732,325 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
       screen->chip_flags = RADEON_CHIPSET_TCL;
       break;
 
+   case PCI_CHIP_R600_9400:
+   case PCI_CHIP_R600_9401:
+   case PCI_CHIP_R600_9402:
+   case PCI_CHIP_R600_9403:
+   case PCI_CHIP_R600_9405:
+   case PCI_CHIP_R600_940A:
+   case PCI_CHIP_R600_940B:
+   case PCI_CHIP_R600_940F:
+      screen->chip_family = CHIP_FAMILY_R600;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV610_94C0:
+   case PCI_CHIP_RV610_94C1:
+   case PCI_CHIP_RV610_94C3:
+   case PCI_CHIP_RV610_94C4:
+   case PCI_CHIP_RV610_94C5:
+   case PCI_CHIP_RV610_94C6:
+   case PCI_CHIP_RV610_94C7:
+   case PCI_CHIP_RV610_94C8:
+   case PCI_CHIP_RV610_94C9:
+   case PCI_CHIP_RV610_94CB:
+   case PCI_CHIP_RV610_94CC:
+   case PCI_CHIP_RV610_94CD:
+      screen->chip_family = CHIP_FAMILY_RV610;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV630_9580:
+   case PCI_CHIP_RV630_9581:
+   case PCI_CHIP_RV630_9583:
+   case PCI_CHIP_RV630_9586:
+   case PCI_CHIP_RV630_9587:
+   case PCI_CHIP_RV630_9588:
+   case PCI_CHIP_RV630_9589:
+   case PCI_CHIP_RV630_958A:
+   case PCI_CHIP_RV630_958B:
+   case PCI_CHIP_RV630_958C:
+   case PCI_CHIP_RV630_958D:
+   case PCI_CHIP_RV630_958E:
+   case PCI_CHIP_RV630_958F:
+      screen->chip_family = CHIP_FAMILY_RV630;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV670_9500:
+   case PCI_CHIP_RV670_9501:
+   case PCI_CHIP_RV670_9504:
+   case PCI_CHIP_RV670_9505:
+   case PCI_CHIP_RV670_9506:
+   case PCI_CHIP_RV670_9507:
+   case PCI_CHIP_RV670_9508:
+   case PCI_CHIP_RV670_9509:
+   case PCI_CHIP_RV670_950F:
+   case PCI_CHIP_RV670_9511:
+   case PCI_CHIP_RV670_9515:
+   case PCI_CHIP_RV670_9517:
+   case PCI_CHIP_RV670_9519:
+      screen->chip_family = CHIP_FAMILY_RV670;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV620_95C0:
+   case PCI_CHIP_RV620_95C2:
+   case PCI_CHIP_RV620_95C4:
+   case PCI_CHIP_RV620_95C5:
+   case PCI_CHIP_RV620_95C6:
+   case PCI_CHIP_RV620_95C7:
+   case PCI_CHIP_RV620_95C9:
+   case PCI_CHIP_RV620_95CC:
+   case PCI_CHIP_RV620_95CD:
+   case PCI_CHIP_RV620_95CE:
+   case PCI_CHIP_RV620_95CF:
+      screen->chip_family = CHIP_FAMILY_RV620;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV635_9590:
+   case PCI_CHIP_RV635_9591:
+   case PCI_CHIP_RV635_9593:
+   case PCI_CHIP_RV635_9595:
+   case PCI_CHIP_RV635_9596:
+   case PCI_CHIP_RV635_9597:
+   case PCI_CHIP_RV635_9598:
+   case PCI_CHIP_RV635_9599:
+   case PCI_CHIP_RV635_959B:
+      screen->chip_family = CHIP_FAMILY_RV635;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RS780_9610:
+   case PCI_CHIP_RS780_9611:
+   case PCI_CHIP_RS780_9612:
+   case PCI_CHIP_RS780_9613:
+   case PCI_CHIP_RS780_9614:
+   case PCI_CHIP_RS780_9615:
+   case PCI_CHIP_RS780_9616:
+      screen->chip_family = CHIP_FAMILY_RS780;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+   case PCI_CHIP_RS880_9710:
+   case PCI_CHIP_RS880_9711:
+   case PCI_CHIP_RS880_9712:
+   case PCI_CHIP_RS880_9713:
+   case PCI_CHIP_RS880_9714:
+      screen->chip_family = CHIP_FAMILY_RS880;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV770_9440:
+   case PCI_CHIP_RV770_9441:
+   case PCI_CHIP_RV770_9442:
+   case PCI_CHIP_RV770_9443:
+   case PCI_CHIP_RV770_9444:
+   case PCI_CHIP_RV770_9446:
+   case PCI_CHIP_RV770_944A:
+   case PCI_CHIP_RV770_944B:
+   case PCI_CHIP_RV770_944C:
+   case PCI_CHIP_RV770_944E:
+   case PCI_CHIP_RV770_9450:
+   case PCI_CHIP_RV770_9452:
+   case PCI_CHIP_RV770_9456:
+   case PCI_CHIP_RV770_945A:
+   case PCI_CHIP_RV770_945B:
+   case PCI_CHIP_RV790_9460:
+   case PCI_CHIP_RV790_9462:
+   case PCI_CHIP_RV770_946A:
+   case PCI_CHIP_RV770_946B:
+   case PCI_CHIP_RV770_947A:
+   case PCI_CHIP_RV770_947B:
+      screen->chip_family = CHIP_FAMILY_RV770;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV730_9480:
+   case PCI_CHIP_RV730_9487:
+   case PCI_CHIP_RV730_9488:
+   case PCI_CHIP_RV730_9489:
+   case PCI_CHIP_RV730_948F:
+   case PCI_CHIP_RV730_9490:
+   case PCI_CHIP_RV730_9491:
+   case PCI_CHIP_RV730_9495:
+   case PCI_CHIP_RV730_9498:
+   case PCI_CHIP_RV730_949C:
+   case PCI_CHIP_RV730_949E:
+   case PCI_CHIP_RV730_949F:
+      screen->chip_family = CHIP_FAMILY_RV730;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV710_9540:
+   case PCI_CHIP_RV710_9541:
+   case PCI_CHIP_RV710_9542:
+   case PCI_CHIP_RV710_954E:
+   case PCI_CHIP_RV710_954F:
+   case PCI_CHIP_RV710_9552:
+   case PCI_CHIP_RV710_9553:
+   case PCI_CHIP_RV710_9555:
+   case PCI_CHIP_RV710_9557:
+      screen->chip_family = CHIP_FAMILY_RV710;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV740_94A0:
+   case PCI_CHIP_RV740_94A1:
+   case PCI_CHIP_RV740_94A3:
+   case PCI_CHIP_RV740_94B1:
+   case PCI_CHIP_RV740_94B3:
+   case PCI_CHIP_RV740_94B4:
+   case PCI_CHIP_RV740_94B5:
+   case PCI_CHIP_RV740_94B9:
+      screen->chip_family = CHIP_FAMILY_RV740;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
    default:
       fprintf(stderr, "unknown chip id 0x%x, can't guess.\n",
-	      dri_priv->deviceID);
+	      device_id);
+      return -1;
+   }
+
+   return 0;
+}
+
+
+/* Create the device specific screen private data struct.
+ */
+static radeonScreenPtr
+radeonCreateScreen( __DRIscreenPrivate *sPriv )
+{
+   radeonScreenPtr screen;
+   RADEONDRIPtr dri_priv = (RADEONDRIPtr)sPriv->pDevPriv;
+   unsigned char *RADEONMMIO = NULL;
+   int i;
+   int ret;
+   uint32_t temp = 0;
+
+   if (sPriv->devPrivSize != sizeof(RADEONDRIRec)) {
+      fprintf(stderr,"\nERROR!  sizeof(RADEONDRIRec) does not match passed size from device driver\n");
+      return GL_FALSE;
+   }
+
+   /* Allocate the private area */
+   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
+   if ( !screen ) {
+      __driUtilMessage("%s: Could not allocate memory for screen structure",
+		       __FUNCTION__);
       return NULL;
    }
+
+   radeon_init_debug();
+
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo (&screen->optionCache,
+		       __driConfigOptions, __driNConfigOptions);
+
+   /* This is first since which regions we map depends on whether or
+    * not we are using a PCI card.
+    */
+   screen->card_type = (dri_priv->IsPCI ? RADEON_CARD_PCI : RADEON_CARD_AGP);
+   {
+      int ret;
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_GART_BUFFER_OFFSET,
+			    &screen->gart_buffer_offset);
+
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BUFFER_OFFSET): %d\n", ret);
+	 return NULL;
+      }
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_GART_BASE,
+			    &screen->gart_base);
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BASE): %d\n", ret);
+	 return NULL;
+      }
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_IRQ_NR,
+			    &screen->irq);
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_IRQ_NR): %d\n", ret);
+	 return NULL;
+      }
+      screen->drmSupportsCubeMapsR200 = (sPriv->drm_version.minor >= 7);
+      screen->drmSupportsBlendColor = (sPriv->drm_version.minor >= 11);
+      screen->drmSupportsTriPerf = (sPriv->drm_version.minor >= 16);
+      screen->drmSupportsFragShader = (sPriv->drm_version.minor >= 18);
+      screen->drmSupportsPointSprites = (sPriv->drm_version.minor >= 13);
+      screen->drmSupportsCubeMapsR100 = (sPriv->drm_version.minor >= 15);
+      screen->drmSupportsVertexProgram = (sPriv->drm_version.minor >= 25);
+      screen->drmSupportsOcclusionQueries = (sPriv->drm_version.minor >= 30);
+   }
+
+   ret = radeon_set_screen_flags(screen, dri_priv->deviceID);
+   if (ret == -1)
+     return NULL;
+
+   screen->mmio.handle = dri_priv->registerHandle;
+   screen->mmio.size   = dri_priv->registerSize;
+   if ( drmMap( sPriv->fd,
+		screen->mmio.handle,
+		screen->mmio.size,
+		&screen->mmio.map ) ) {
+     FREE( screen );
+     __driUtilMessage("%s: drmMap failed\n", __FUNCTION__ );
+     return NULL;
+   }
+
+   RADEONMMIO = screen->mmio.map;
+
+   screen->status.handle = dri_priv->statusHandle;
+   screen->status.size   = dri_priv->statusSize;
+   if ( drmMap( sPriv->fd,
+		screen->status.handle,
+		screen->status.size,
+		&screen->status.map ) ) {
+     drmUnmap( screen->mmio.map, screen->mmio.size );
+     FREE( screen );
+     __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__ );
+     return NULL;
+   }
+   if (screen->chip_family < CHIP_FAMILY_R600)
+	   screen->scratch = (__volatile__ uint32_t *)
+		   ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET);
+   else
+	   screen->scratch = (__volatile__ uint32_t *)
+		   ((GLubyte *)screen->status.map + R600_SCRATCH_REG_OFFSET);
+
+   screen->buffers = drmMapBufs( sPriv->fd );
+   if ( !screen->buffers ) {
+     drmUnmap( screen->status.map, screen->status.size );
+     drmUnmap( screen->mmio.map, screen->mmio.size );
+     FREE( screen );
+     __driUtilMessage("%s: drmMapBufs failed\n", __FUNCTION__ );
+     return NULL;
+   }
+
+   if ( dri_priv->gartTexHandle && dri_priv->gartTexMapSize ) {
+     screen->gartTextures.handle = dri_priv->gartTexHandle;
+     screen->gartTextures.size   = dri_priv->gartTexMapSize;
+     if ( drmMap( sPriv->fd,
+		  screen->gartTextures.handle,
+		  screen->gartTextures.size,
+		  (drmAddressPtr)&screen->gartTextures.map ) ) {
+       drmUnmapBufs( screen->buffers );
+       drmUnmap( screen->status.map, screen->status.size );
+       drmUnmap( screen->mmio.map, screen->mmio.size );
+       FREE( screen );
+       __driUtilMessage("%s: drmMap failed for GART texture area\n", __FUNCTION__);
+       return NULL;
+    }
+
+     screen->gart_texture_offset = dri_priv->gartTexOffset + screen->gart_base;
+   }
+
    if ((screen->chip_family == CHIP_FAMILY_R350 || screen->chip_family == CHIP_FAMILY_R300) &&
        sPriv->ddx_version.minor < 2) {
       fprintf(stderr, "xf86-video-ati-6.6.2 or newer needed for Radeon 9500/9700/9800 cards.\n");
@@ -836,35 +1063,57 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
    }
 
    if (getenv("R300_NO_TCL"))
-     screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	   screen->chip_flags &= ~RADEON_CHIPSET_TCL;
 
    if (screen->chip_family <= CHIP_FAMILY_RS200)
-      screen->chip_flags |= RADEON_CLASS_R100;
+	   screen->chip_flags |= RADEON_CLASS_R100;
    else if (screen->chip_family <= CHIP_FAMILY_RV280)
-      screen->chip_flags |= RADEON_CLASS_R200;
+	   screen->chip_flags |= RADEON_CLASS_R200;
+   else if (screen->chip_family <= CHIP_FAMILY_RV570)
+	   screen->chip_flags |= RADEON_CLASS_R300;
    else
-      screen->chip_flags |= RADEON_CLASS_R300;
+	   screen->chip_flags |= RADEON_CLASS_R600;
 
    screen->cpp = dri_priv->bpp / 8;
    screen->AGPMode = dri_priv->AGPMode;
 
-   ret = radeonGetParam( sPriv->fd, RADEON_PARAM_FB_LOCATION,
-                         &temp);
-   if (ret) {
-       if (screen->chip_family < CHIP_FAMILY_RS600)
-	   screen->fbLocation      = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16;
-       else {
-           FREE( screen );
-           fprintf(stderr, "Unable to get fb location need newer drm\n");
-           return NULL;
+   ret = radeonGetParam(sPriv, RADEON_PARAM_FB_LOCATION, &temp);
+
+   /* +r6/r7 */
+   if(screen->chip_family >= CHIP_FAMILY_R600)
+   {
+       if (ret)
+       {
+            FREE( screen );
+            fprintf(stderr, "Unable to get fb location need newer drm\n");
+            return NULL;
+       }
+       else
+       {
+            screen->fbLocation = (temp & 0xffff) << 24;
        }
-   } else {
-       screen->fbLocation = (temp & 0xffff) << 16;
+   }
+   else
+   {
+        if (ret)
+        {
+            if (screen->chip_family < CHIP_FAMILY_RS600 && !screen->kernel_mm)
+	            screen->fbLocation      = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16;
+            else
+            {
+                FREE( screen );
+                fprintf(stderr, "Unable to get fb location need newer drm\n");
+                return NULL;
+            }
+        }
+        else
+        {
+            screen->fbLocation = (temp & 0xffff) << 16;
+        }
    }
 
-   if (screen->chip_family >= CHIP_FAMILY_R300) {
-       ret = radeonGetParam( sPriv->fd, RADEON_PARAM_NUM_GB_PIPES,
-			     &temp);
+   if (IS_R300_CLASS(screen)) {
+       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp);
        if (ret) {
 	   fprintf(stderr, "Unable to get num_pipes, need newer drm\n");
 	   switch (screen->chip_family) {
@@ -902,6 +1151,14 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
 	   break;
        }
 
+       if ( sPriv->drm_version.minor >= 31 ) {
+	       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_Z_PIPES, &temp);
+	       if (ret)
+		       screen->num_z_pipes = 2;
+	       else
+		       screen->num_z_pipes = temp;
+       } else
+	       screen->num_z_pipes = 2;
    }
 
    if ( sPriv->drm_version.minor >= 10 ) {
@@ -971,7 +1228,7 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
    if (IS_R200_CLASS(screen))
-       screen->extensions[i++] = &r200AllocateExtension.base;
+      screen->extensions[i++] = &r200AllocateExtension.base;
 
    screen->extensions[i++] = &r200texOffsetExtension.base;
 #endif
@@ -980,11 +1237,173 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
    screen->extensions[i++] = &r300texOffsetExtension.base;
 #endif
 
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   screen->extensions[i++] = &r600texOffsetExtension.base;
+#endif
+
    screen->extensions[i++] = NULL;
    sPriv->extensions = screen->extensions;
 
    screen->driScreen = sPriv;
    screen->sarea_priv_offset = dri_priv->sarea_priv_offset;
+   screen->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
+					       screen->sarea_priv_offset);
+
+   screen->bom = radeon_bo_manager_legacy_ctor(screen);
+   if (screen->bom == NULL) {
+     free(screen);
+     return NULL;
+   }
+
+   return screen;
+}
+
+static radeonScreenPtr
+radeonCreateScreen2(__DRIscreenPrivate *sPriv)
+{
+   radeonScreenPtr screen;
+   int i;
+   int ret;
+   uint32_t device_id = 0;
+   uint32_t temp = 0;
+
+   /* Allocate the private area */
+   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
+   if ( !screen ) {
+      __driUtilMessage("%s: Could not allocate memory for screen structure",
+		       __FUNCTION__);
+      fprintf(stderr, "leaving here\n");
+      return NULL;
+   }
+
+   radeon_init_debug();
+
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo (&screen->optionCache,
+		       __driConfigOptions, __driNConfigOptions);
+
+   screen->kernel_mm = 1;
+   screen->chip_flags = 0;
+
+   /* if we have kms we can support all of these */
+   screen->drmSupportsCubeMapsR200 = 1;
+   screen->drmSupportsBlendColor = 1;
+   screen->drmSupportsTriPerf = 1;
+   screen->drmSupportsFragShader = 1;
+   screen->drmSupportsPointSprites = 1;
+   screen->drmSupportsCubeMapsR100 = 1;
+   screen->drmSupportsVertexProgram = 1;
+   screen->drmSupportsOcclusionQueries = 1;
+   screen->irq = 1;
+
+   ret = radeonGetParam(sPriv, RADEON_PARAM_DEVICE_ID, &device_id);
+   if (ret) {
+     FREE( screen );
+     fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_DEVICE_ID): %d\n", ret);
+     return NULL;
+   }
+
+   ret = radeon_set_screen_flags(screen, device_id);
+   if (ret == -1)
+     return NULL;
+
+   if (getenv("R300_NO_TCL"))
+	   screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+
+   if (screen->chip_family <= CHIP_FAMILY_RS200)
+	   screen->chip_flags |= RADEON_CLASS_R100;
+   else if (screen->chip_family <= CHIP_FAMILY_RV280)
+	   screen->chip_flags |= RADEON_CLASS_R200;
+   else if (screen->chip_family <= CHIP_FAMILY_RV570)
+	   screen->chip_flags |= RADEON_CLASS_R300;
+   else
+	   screen->chip_flags |= RADEON_CLASS_R600;
+
+   if (IS_R300_CLASS(screen)) {
+       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp);
+       if (ret) {
+	   fprintf(stderr, "Unable to get num_pipes, need newer drm\n");
+	   switch (screen->chip_family) {
+	   case CHIP_FAMILY_R300:
+	   case CHIP_FAMILY_R350:
+	       screen->num_gb_pipes = 2;
+	       break;
+	   case CHIP_FAMILY_R420:
+	   case CHIP_FAMILY_R520:
+	   case CHIP_FAMILY_R580:
+	   case CHIP_FAMILY_RV560:
+	   case CHIP_FAMILY_RV570:
+	       screen->num_gb_pipes = 4;
+	       break;
+	   case CHIP_FAMILY_RV350:
+	   case CHIP_FAMILY_RV515:
+	   case CHIP_FAMILY_RV530:
+	   case CHIP_FAMILY_RV410:
+	   default:
+	       screen->num_gb_pipes = 1;
+	       break;
+	   }
+       } else {
+	   screen->num_gb_pipes = temp;
+       }
+
+       /* pipe overrides */
+       switch (device_id) {
+       case PCI_CHIP_R300_AD: /* 9500 with 1 quadpipe verified by: Reid Linnemann <lreid@cs.okstate.edu> */
+       case PCI_CHIP_RV410_5E4C: /* RV410 SE only have 1 quadpipe */
+       case PCI_CHIP_RV410_5E4F: /* RV410 SE only have 1 quadpipe */
+	   screen->num_gb_pipes = 1;
+	   break;
+       default:
+	   break;
+       }
+
+       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_Z_PIPES, &temp);
+       if (ret)
+	       screen->num_z_pipes = 2;
+       else
+	       screen->num_z_pipes = temp;
+
+   }
+
+   i = 0;
+   screen->extensions[i++] = &driCopySubBufferExtension.base;
+   screen->extensions[i++] = &driFrameTrackingExtension.base;
+   screen->extensions[i++] = &driReadDrawableExtension;
+
+   if ( screen->irq != 0 ) {
+       screen->extensions[i++] = &driSwapControlExtension.base;
+       screen->extensions[i++] = &driMediaStreamCounterExtension.base;
+   }
+
+#if !RADEON_COMMON
+   screen->extensions[i++] = &radeonTexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+   if (IS_R200_CLASS(screen))
+       screen->extensions[i++] = &r200AllocateExtension.base;
+
+   screen->extensions[i++] = &r200TexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+   screen->extensions[i++] = &r300TexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   screen->extensions[i++] = &r600TexBufferExtension.base;
+#endif
+
+   screen->extensions[i++] = NULL;
+   sPriv->extensions = screen->extensions;
+
+   screen->driScreen = sPriv;
+   screen->bom = radeon_bo_manager_gem_ctor(sPriv->fd);
+   if (screen->bom == NULL) {
+       free(screen);
+       return NULL;
+   }
    return screen;
 }
 
@@ -993,23 +1412,32 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
 static void
 radeonDestroyScreen( __DRIscreenPrivate *sPriv )
 {
-   radeonScreenPtr screen = (radeonScreenPtr)sPriv->private;
+    radeonScreenPtr screen = (radeonScreenPtr)sPriv->private;
 
-   if (!screen)
-      return;
+    if (!screen)
+        return;
 
-   if ( screen->gartTextures.map ) {
-      drmUnmap( screen->gartTextures.map, screen->gartTextures.size );
-   }
-   drmUnmapBufs( screen->buffers );
-   drmUnmap( screen->status.map, screen->status.size );
-   drmUnmap( screen->mmio.map, screen->mmio.size );
+    if (screen->kernel_mm) {
+#ifdef RADEON_BO_TRACK
+        radeon_tracker_print(&screen->bom->tracker, stderr);
+#endif
+        radeon_bo_manager_gem_dtor(screen->bom);
+    } else {
+        radeon_bo_manager_legacy_dtor(screen->bom);
+
+        if ( screen->gartTextures.map ) {
+            drmUnmap( screen->gartTextures.map, screen->gartTextures.size );
+        }
+        drmUnmapBufs( screen->buffers );
+        drmUnmap( screen->status.map, screen->status.size );
+        drmUnmap( screen->mmio.map, screen->mmio.size );
+    }
 
-   /* free all option information */
-   driDestroyOptionInfo (&screen->optionCache);
+    /* free all option information */
+    driDestroyOptionInfo (&screen->optionCache);
 
-   FREE( screen );
-   sPriv->private = NULL;
+    FREE( screen );
+    sPriv->private = NULL;
 }
 
 
@@ -1018,16 +1446,21 @@ radeonDestroyScreen( __DRIscreenPrivate *sPriv )
 static GLboolean
 radeonInitDriver( __DRIscreenPrivate *sPriv )
 {
-   sPriv->private = (void *) radeonCreateScreen( sPriv );
-   if ( !sPriv->private ) {
-      radeonDestroyScreen( sPriv );
-      return GL_FALSE;
-   }
+    if (sPriv->dri2.enabled) {
+        sPriv->private = (void *) radeonCreateScreen2( sPriv );
+    } else {
+        sPriv->private = (void *) radeonCreateScreen( sPriv );
+    }
+    if ( !sPriv->private ) {
+        radeonDestroyScreen( sPriv );
+        return GL_FALSE;
+    }
 
-   return GL_TRUE;
+    return GL_TRUE;
 }
 
 
+
 /**
  * Create the Mesa framebuffer and renderbuffers for a given window/drawable.
  *
@@ -1040,132 +1473,112 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv,
                     const __GLcontextModes *mesaVis,
                     GLboolean isPixmap )
 {
-   radeonScreenPtr screen = (radeonScreenPtr) driScrnPriv->private;
+    radeonScreenPtr screen = (radeonScreenPtr) driScrnPriv->private;
 
-   if (isPixmap) {
+    const GLboolean swDepth = GL_FALSE;
+    const GLboolean swAlpha = GL_FALSE;
+    const GLboolean swAccum = mesaVis->accumRedBits > 0;
+    const GLboolean swStencil = mesaVis->stencilBits > 0 &&
+	mesaVis->depthBits != 24;
+    GLenum rgbFormat;
+    struct radeon_framebuffer *rfb;
+
+    if (isPixmap)
       return GL_FALSE; /* not implemented */
-   }
-   else {
-      const GLboolean swDepth = GL_FALSE;
-      const GLboolean swAlpha = GL_FALSE;
-      const GLboolean swAccum = mesaVis->accumRedBits > 0;
-      const GLboolean swStencil = mesaVis->stencilBits > 0 &&
-         mesaVis->depthBits != 24;
-      struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis);
-
-      /* front color renderbuffer */
-      {
-         driRenderbuffer *frontRb
-            = driNewRenderbuffer(GL_RGBA,
-                                 driScrnPriv->pFB + screen->frontOffset,
-                                 screen->cpp,
-                                 screen->frontOffset, screen->frontPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(frontRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base);
-      }
 
-      /* back color renderbuffer */
-      if (mesaVis->doubleBufferMode) {
-         driRenderbuffer *backRb
-            = driNewRenderbuffer(GL_RGBA,
-                                 driScrnPriv->pFB + screen->backOffset,
-                                 screen->cpp,
-                                 screen->backOffset, screen->backPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(backRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base);
-      }
+    rfb = CALLOC_STRUCT(radeon_framebuffer);
+    if (!rfb)
+      return GL_FALSE;
 
-      /* depth renderbuffer */
-      if (mesaVis->depthBits == 16) {
-         driRenderbuffer *depthRb
-            = driNewRenderbuffer(GL_DEPTH_COMPONENT16,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(depthRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
-	 depthRb->depthHasSurface = screen->depthHasSurface;
-      }
-      else if (mesaVis->depthBits == 24) {
-         driRenderbuffer *depthRb
-            = driNewRenderbuffer(GL_DEPTH_COMPONENT24,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(depthRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
-	 depthRb->depthHasSurface = screen->depthHasSurface;
-      }
+    _mesa_initialize_framebuffer(&rfb->base, mesaVis);
+
+    if (mesaVis->redBits == 5)
+        rgbFormat = GL_RGB5;
+    else if (mesaVis->alphaBits == 0)
+        rgbFormat = GL_RGB8;
+    else
+        rgbFormat = GL_RGBA8;
+
+    /* front color renderbuffer */
+    rfb->color_rb[0] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
+    _mesa_add_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT, &rfb->color_rb[0]->base);
+    rfb->color_rb[0]->has_surface = 1;
+
+    /* back color renderbuffer */
+    if (mesaVis->doubleBufferMode) {
+      rfb->color_rb[1] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_BACK_LEFT, &rfb->color_rb[1]->base);
+	rfb->color_rb[1]->has_surface = 1;
+    }
 
-      /* stencil renderbuffer */
-      if (mesaVis->stencilBits > 0 && !swStencil) {
-         driRenderbuffer *stencilRb
-            = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(stencilRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base);
-	 stencilRb->depthHasSurface = screen->depthHasSurface;
+    if (mesaVis->depthBits == 24) {
+      if (mesaVis->stencilBits == 8) {
+	struct radeon_renderbuffer *depthStencilRb = radeon_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depthStencilRb->base);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_STENCIL, &depthStencilRb->base);
+	depthStencilRb->has_surface = screen->depthHasSurface;
+      } else {
+	/* depth renderbuffer */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT24, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
+	depth->has_surface = screen->depthHasSurface;
       }
+    } else if (mesaVis->depthBits == 16) {
+      /* just 16-bit depth buffer, no hw stencil */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT16, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
+	depth->has_surface = screen->depthHasSurface;
+    }
 
-      _mesa_add_soft_renderbuffers(fb,
-                                   GL_FALSE, /* color */
-                                   swDepth,
-                                   swStencil,
-                                   swAccum,
-                                   swAlpha,
-                                   GL_FALSE /* aux */);
-      driDrawPriv->driverPrivate = (void *) fb;
+    _mesa_add_soft_renderbuffers(&rfb->base,
+	    GL_FALSE, /* color */
+	    swDepth,
+	    swStencil,
+	    swAccum,
+	    swAlpha,
+	    GL_FALSE /* aux */);
+    driDrawPriv->driverPrivate = (void *) rfb;
 
-      return (driDrawPriv->driverPrivate != NULL);
-   }
+    return (driDrawPriv->driverPrivate != NULL);
 }
 
 
-static void
-radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
+static void radeon_cleanup_renderbuffers(struct radeon_framebuffer *rfb)
 {
-   _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL);
-}
+	struct radeon_renderbuffer *rb;
 
-#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
-/**
- * Choose the appropriate CreateContext function based on the chipset.
- * Eventually, all drivers will go through this process.
- */
-static GLboolean radeonCreateContext(const __GLcontextModes * glVisual,
-				     __DRIcontextPrivate * driContextPriv,
-				     void *sharedContextPriv)
-{
-	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
-	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
-
-	if (IS_R300_CLASS(screen))
-		return r300CreateContext(glVisual, driContextPriv, sharedContextPriv);
-        return GL_FALSE;
+	rb = rfb->color_rb[0];
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[1];
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
 }
 
-/**
- * Choose the appropriate DestroyContext function based on the chipset.
- */
-static void radeonDestroyContext(__DRIcontextPrivate * driContextPriv)
+void
+radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
 {
-	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
-
-	if (IS_R300_CLASS(radeon->radeonScreen))
-		return r300DestroyContext(driContextPriv);
+    struct radeon_framebuffer *rfb;
+    if (!driDrawPriv)
+	return;
+
+    rfb = (void*)driDrawPriv->driverPrivate;
+    if (!rfb)
+	return;
+    radeon_cleanup_renderbuffers(rfb);
+    _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL);
 }
 
 
-#endif
-
-
 /**
  * This is the driver specific part of the createNewScreen entry point.
  *
@@ -1191,6 +1604,11 @@ radeonInitScreen(__DRIscreenPrivate *psp)
    static const __DRIutilversion2 ddx_expected = { 4, 5, 0, 0 };
    static const __DRIversion dri_expected = { 4, 0, 0 };
    static const __DRIversion drm_expected = { 1, 24, 0 };
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   static const char *driver_name = "R600";
+   static const __DRIutilversion2 ddx_expected = { 4, 5, 0, 0 };
+   static const __DRIversion dri_expected = { 4, 0, 0 };
+   static const __DRIversion drm_expected = { 1, 24, 0 };
 #endif
    RADEONDRIPtr dri_priv = (RADEONDRIPtr) psp->pDevPriv;
 
@@ -1218,20 +1636,112 @@ radeonInitScreen(__DRIscreenPrivate *psp)
    driInitSingleExtension( NULL, NV_vp_extension );
    driInitSingleExtension( NULL, ATI_fs_extension );
    driInitExtensions( NULL, point_extensions, GL_FALSE );
-#elif defined(RADEON_COMMON_FOR_R300)
+#elif (defined(RADEON_COMMON_FOR_R300) || defined(RADEON_COMMON_FOR_R600))
    driInitSingleExtension( NULL, gl_20_extension );
 #endif
 
    if (!radeonInitDriver(psp))
        return NULL;
 
+   /* for now fill in all modes */
    return radeonFillInModes( psp,
 			     dri_priv->bpp,
 			     (dri_priv->bpp == 16) ? 16 : 24,
-			     (dri_priv->bpp == 16) ? 0  : 8,
-			     (dri_priv->backOffset != dri_priv->depthOffset) );
+			     (dri_priv->bpp == 16) ? 0  : 8, 1);
 }
+#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
 
+/**
+ * This is the driver specific part of the createNewScreen entry point.
+ * Called when using DRI2.
+ *
+ * \return the __GLcontextModes supported by this driver
+ */
+static const
+__DRIconfig **radeonInitScreen2(__DRIscreenPrivate *psp)
+{
+   GLenum fb_format[3];
+   GLenum fb_type[3];
+   /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+    * support pageflipping at all.
+    */
+   static const GLenum back_buffer_modes[] = {
+     GLX_NONE, GLX_SWAP_UNDEFINED_OML, /*, GLX_SWAP_COPY_OML*/
+   };
+   uint8_t depth_bits[4], stencil_bits[4], msaa_samples_array[1];
+   int color;
+   __DRIconfig **configs = NULL;
+
+   /* Calling driInitExtensions here, with a NULL context pointer,
+    * does not actually enable the extensions.  It just makes sure
+    * that all the dispatch offsets for all the extensions that
+    * *might* be enables are known.  This is needed because the
+    * dispatch offsets need to be known when _mesa_context_create
+    * is called, but we can't enable the extensions until we have a
+    * context pointer.
+    *
+    * Hello chicken.  Hello egg.  How are you two today?
+    */
+   driInitExtensions( NULL, card_extensions, GL_FALSE );
+   driInitExtensions( NULL, mm_extensions, GL_FALSE );
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+   driInitExtensions( NULL, blend_extensions, GL_FALSE );
+   driInitSingleExtension( NULL, ARB_vp_extension );
+   driInitSingleExtension( NULL, NV_vp_extension );
+   driInitSingleExtension( NULL, ATI_fs_extension );
+   driInitExtensions( NULL, point_extensions, GL_FALSE );
+#elif (defined(RADEON_COMMON_FOR_R300) || defined(RADEON_COMMON_FOR_R600))
+   driInitSingleExtension( NULL, gl_20_extension );
+#endif
+
+   if (!radeonInitDriver(psp)) {
+       return NULL;
+    }
+   depth_bits[0] = 0;
+   stencil_bits[0] = 0;
+   depth_bits[1] = 16;
+   stencil_bits[1] = 0;
+   depth_bits[2] = 24;
+   stencil_bits[2] = 0;
+   depth_bits[3] = 24;
+   stencil_bits[3] = 8;
+
+   msaa_samples_array[0] = 0;
+
+   fb_format[0] = GL_RGB;
+   fb_type[0] = GL_UNSIGNED_SHORT_5_6_5;
+
+   fb_format[1] = GL_BGR;
+   fb_type[1] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   fb_format[2] = GL_BGRA;
+   fb_type[2] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   for (color = 0; color < ARRAY_SIZE(fb_format); color++) {
+      __DRIconfig **new_configs;
+
+      new_configs = driCreateConfigs(fb_format[color], fb_type[color],
+				     depth_bits,
+				     stencil_bits,
+				     ARRAY_SIZE(depth_bits),
+				     back_buffer_modes,
+				     ARRAY_SIZE(back_buffer_modes),
+				     msaa_samples_array,
+				     ARRAY_SIZE(msaa_samples_array));
+      if (configs == NULL)
+	 configs = new_configs;
+      else
+	 configs = driConcatConfigs(configs, new_configs);
+   }
+
+   if (configs == NULL) {
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+              __LINE__);
+      return NULL;
+   }
+
+   return (const __DRIconfig **)configs;
+}
 
 /**
  * Get information about previous buffer swaps.
@@ -1239,36 +1749,42 @@ radeonInitScreen(__DRIscreenPrivate *psp)
 static int
 getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo )
 {
-#if !RADEON_COMMON || (RADEON_COMMON && defined(RADEON_COMMON_FOR_R300))
-   radeonContextPtr  rmesa;
-#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
-   r200ContextPtr  rmesa;
-#endif
+    struct radeon_framebuffer *rfb;
 
-   if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL)
-	|| (dPriv->driContextPriv->driverPrivate == NULL)
-	|| (sInfo == NULL) ) {
-      return -1;
+    if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL)
+	 || (dPriv->driContextPriv->driverPrivate == NULL)
+	 || (sInfo == NULL) ) {
+	return -1;
    }
 
-   rmesa = dPriv->driContextPriv->driverPrivate;
-   sInfo->swap_count = rmesa->swap_count;
-   sInfo->swap_ust = rmesa->swap_ust;
-   sInfo->swap_missed_count = rmesa->swap_missed_count;
+    rfb = dPriv->driverPrivate;
+    sInfo->swap_count = rfb->swap_count;
+    sInfo->swap_ust = rfb->swap_ust;
+    sInfo->swap_missed_count = rfb->swap_missed_count;
 
    sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0)
-       ? driCalculateSwapUsage( dPriv, 0, rmesa->swap_missed_ust )
+       ? driCalculateSwapUsage( dPriv, 0, rfb->swap_missed_ust )
        : 0.0;
 
    return 0;
 }
 
-#if !RADEON_COMMON || (RADEON_COMMON && defined(RADEON_COMMON_FOR_R300))
 const struct __DriverAPIRec driDriverAPI = {
    .InitScreen      = radeonInitScreen,
    .DestroyScreen   = radeonDestroyScreen,
-   .CreateContext   = radeonCreateContext,
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+   .CreateContext   = r200CreateContext,
+   .DestroyContext  = r200DestroyContext,
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   .CreateContext   = r600CreateContext,
+   .DestroyContext  = radeonDestroyContext,
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+   .CreateContext   = r300CreateContext,
    .DestroyContext  = radeonDestroyContext,
+#else
+   .CreateContext   = r100CreateContext,
+   .DestroyContext  = radeonDestroyContext,
+#endif
    .CreateBuffer    = radeonCreateBuffer,
    .DestroyBuffer   = radeonDestroyBuffer,
    .SwapBuffers     = radeonSwapBuffers,
@@ -1280,23 +1796,7 @@ const struct __DriverAPIRec driDriverAPI = {
    .WaitForSBC      = NULL,
    .SwapBuffersMSC  = NULL,
    .CopySubBuffer   = radeonCopySubBuffer,
+    /* DRI2 */
+   .InitScreen2     = radeonInitScreen2,
 };
-#else
-const struct __DriverAPIRec driDriverAPI = {
-   .InitScreen      = radeonInitScreen,
-   .DestroyScreen   = radeonDestroyScreen,
-   .CreateContext   = r200CreateContext,
-   .DestroyContext  = r200DestroyContext,
-   .CreateBuffer    = radeonCreateBuffer,
-   .DestroyBuffer   = radeonDestroyBuffer,
-   .SwapBuffers     = r200SwapBuffers,
-   .MakeCurrent     = r200MakeCurrent,
-   .UnbindContext   = r200UnbindContext,
-   .GetSwapInfo     = getSwapInfo,
-   .GetDrawableMSC  = driDrawableGetMSC32,
-   .WaitForMSC      = driWaitForMSC32,
-   .WaitForSBC      = NULL,
-   .SwapBuffersMSC  = NULL,
-   .CopySubBuffer   = r200CopySubBuffer,
-};
-#endif
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h
index b84c70bfae..15744e8828 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.h
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.h
@@ -54,11 +54,12 @@ typedef struct {
    drmAddress map;			/* Mapping of the DRM region */
 } radeonRegionRec, *radeonRegionPtr;
 
-typedef struct {
+typedef struct radeon_screen {
    int chip_family;
    int chip_flags;
    int cpp;
    int card_type;
+   int device_id; /* PCI ID */
    int AGPMode;
    unsigned int irq;			/* IRQ number (0 means none) */
 
@@ -98,14 +99,19 @@ typedef struct {
    GLboolean drmSupportsPointSprites;   /* need radeon kernel module >= 1.13 */
    GLboolean drmSupportsCubeMapsR100;   /* need radeon kernel module >= 1.15 */
    GLboolean drmSupportsVertexProgram;  /* need radeon kernel module >= 1.25 */
+   GLboolean drmSupportsOcclusionQueries; /* need radeon kernel module >= 1.30 */
    GLboolean depthHasSurface;
 
    /* Configuration cache with default values for all contexts */
    driOptionCache optionCache;
 
-   const __DRIextension *extensions[8];
+   const __DRIextension *extensions[16];
 
    int num_gb_pipes;
+   int num_z_pipes;
+   int kernel_mm;
+   drm_radeon_sarea_t *sarea;	/* Private SAREA data */
+   struct radeon_bo_manager *bom;
 } radeonScreenRec, *radeonScreenPtr;
 
 #define IS_R100_CLASS(screen) \
@@ -114,5 +120,8 @@ typedef struct {
 	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R200)
 #define IS_R300_CLASS(screen) \
 	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R300)
+#define IS_R600_CLASS(screen) \
+	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R600)
 
+extern void radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv);
 #endif /* __RADEON_SCREEN_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c
index 12051ff1c8..4e100d854e 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.c
+++ b/src/mesa/drivers/dri/radeon/radeon_span.c
@@ -43,46 +43,222 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/glheader.h"
 #include "swrast/swrast.h"
 
-#include "radeon_context.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "radeon_span.h"
-#include "radeon_tex.h"
-
-#include "drirenderbuffer.h"
 
 #define DBG 0
 
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb);
+
+
+/* r200 depth buffer is always tiled - this is the formula
+   according to the docs unless I typo'ed in it
+*/
+#if defined(RADEON_COMMON_FOR_R200)
+static GLubyte *r200_depth_2byte(const struct radeon_renderbuffer * rrb,
+				 GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    GLint offset;
+    if (rrb->has_surface) {
+	offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+	GLuint b;
+	offset = 0;
+	b = (((y  >> 4) * (rrb->pitch >> 8) + (x >> 6)));
+	offset += (b >> 1) << 12;
+	offset += (((rrb->pitch >> 8) & 0x1) ? (b & 0x1) : ((b & 0x1) ^ ((y >> 4) & 0x1))) << 11;
+	offset += ((y >> 2) & 0x3) << 9;
+	offset += ((x >> 3) & 0x1) << 8;
+	offset += ((x >> 4) & 0x3) << 6;
+	offset += ((x >> 2) & 0x1) << 5;
+	offset += ((y >> 1) & 0x1) << 4;
+	offset += ((x >> 1) & 0x1) << 3;
+	offset += (y & 0x1) << 2;
+	offset += (x & 0x1) << 1;
+    }
+    return &ptr[offset];
+}
+
+static GLubyte *r200_depth_4byte(const struct radeon_renderbuffer * rrb,
+				 GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    GLint offset;
+    if (rrb->has_surface) {
+	offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+	GLuint b;
+	offset = 0;
+	b = (((y & 0x7ff) >> 4) * (rrb->pitch >> 7) + (x >> 5));
+	offset += (b >> 1) << 12;
+	offset += (((rrb->pitch >> 7) & 0x1) ? (b & 0x1) : ((b & 0x1) ^ ((y >> 4) & 0x1))) << 11;
+	offset += ((y >> 2) & 0x3) << 9;
+	offset += ((x >> 2) & 0x1) << 8;
+	offset += ((x >> 3) & 0x3) << 6;
+	offset += ((y >> 1) & 0x1) << 5;
+	offset += ((x >> 1) & 0x1) << 4;
+	offset += (y & 0x1) << 3;
+	offset += (x & 0x1) << 2;
+    }
+    return &ptr[offset];
+}
+#endif
+
+/* radeon tiling on r300-r500 has 4 states,
+   macro-linear/micro-linear
+   macro-linear/micro-tiled
+   macro-tiled /micro-linear
+   macro-tiled /micro-tiled
+   1 byte surface 
+   2 byte surface - two types - we only provide 8x2 microtiling
+   4 byte surface
+   8/16 byte (unused)
+*/
+static GLubyte *radeon_ptr_4byte(const struct radeon_renderbuffer * rrb,
+			     GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
+    GLint offset;
+
+    if (rrb->has_surface || !(rrb->bo->flags & mask)) {
+        offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+        offset = 0;
+        if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) {
+		offset = ((y >> 4) * (rrb->pitch >> 7) + (x >> 5)) << 11;
+		offset += (((y >> 3) ^ (x >> 5)) & 0x1) << 10;
+		offset += (((y >> 4) ^ (x >> 4)) & 0x1) << 9;
+		offset += (((y >> 2) ^ (x >> 4)) & 0x1) << 8;
+		offset += (((y >> 3) ^ (x >> 3)) & 0x1) << 7;
+		offset += ((y >> 1) & 0x1) << 6;
+		offset += ((x >> 2) & 0x1) << 5;
+		offset += (y & 1) << 4;
+		offset += (x & 3) << 2;
+            } else {
+		offset = ((y >> 3) * (rrb->pitch >> 8) + (x >> 6)) << 11;
+		offset += (((y >> 2) ^ (x >> 6)) & 0x1) << 10;
+		offset += (((y >> 3) ^ (x >> 5)) & 0x1) << 9;
+		offset += (((y >> 1) ^ (x >> 5)) & 0x1) << 8;
+		offset += (((y >> 2) ^ (x >> 4)) & 0x1) << 7;
+		offset += (y & 1) << 6;
+		offset += (x & 15) << 2;
+            }
+        } else {
+	    offset = ((y >> 1) * (rrb->pitch >> 4) + (x >> 2)) << 5;
+	    offset += (y & 1) << 4;
+	    offset += (x & 3) << 2;
+        }
+    }
+    return &ptr[offset];
+}
+
+static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb,
+				     GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
+    GLint offset;
+
+    if (rrb->has_surface || !(rrb->bo->flags & mask)) {
+        offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+        offset = 0;
+        if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+            if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) {
+		offset = ((y >> 4) * (rrb->pitch >> 7) + (x >> 6)) << 11;
+		offset += (((y >> 3) ^ (x >> 6)) & 0x1) << 10;
+		offset += (((y >> 4) ^ (x >> 5)) & 0x1) << 9;
+		offset += (((y >> 2) ^ (x >> 5)) & 0x1) << 8;
+		offset += (((y >> 3) ^ (x >> 4)) & 0x1) << 7;
+		offset += ((y >> 1) & 0x1) << 6;
+		offset += ((x >> 3) & 0x1) << 5;
+		offset += (y & 1) << 4;
+		offset += (x & 3) << 2;
+            } else {
+		offset = ((y >> 3) * (rrb->pitch >> 8) + (x >> 7)) << 11;
+		offset += (((y >> 2) ^ (x >> 7)) & 0x1) << 10;
+		offset += (((y >> 3) ^ (x >> 6)) & 0x1) << 9;
+		offset += (((y >> 1) ^ (x >> 6)) & 0x1) << 8;
+		offset += (((y >> 2) ^ (x >> 5)) & 0x1) << 7;
+		offset += (y & 1) << 6;
+		offset += ((x >> 4) & 0x1) << 5;
+                offset += (x & 15) << 2;
+            }
+        } else {
+	    offset = ((y >> 1) * (rrb->pitch >> 4) + (x >> 3)) << 5;
+	    offset += (y & 0x1) << 4;
+	    offset += (x & 0x7) << 1;
+        }
+    }
+    return &ptr[offset];
+}
+
+#ifndef COMPILE_R300
+static uint32_t
+z24s8_to_s8z24(uint32_t val)
+{
+   return (val << 24) | (val >> 8);
+}
+
+static uint32_t
+s8z24_to_z24s8(uint32_t val)
+{
+   return (val >> 24) | (val << 8);
+}
+#endif
+
 /*
  * Note that all information needed to access pixels in a renderbuffer
  * should be obtained through the gl_renderbuffer parameter, not per-context
  * information.
  */
 #define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
+   struct radeon_renderbuffer *rrb = (void *) rb;		\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+   GLuint p;						\
+   (void)p;						\
+   radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
+   struct radeon_renderbuffer *rrb = (void *) rb;	\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+  radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
 
-#define Y_FLIP(Y) (bottom - (Y))
+#define Y_FLIP(_y) ((_y) * yScale + yBias)
 
 #define HW_LOCK()
 
 #define HW_UNLOCK()
 
+/* XXX FBO: this is identical to the macro in spantmp2.h except we get
+ * the cliprect info from the context, not the driDrawable.
+ * Move this into spantmp2.h someday.
+ */
+#define HW_CLIPLOOP()							\
+   do {									\
+      int _nc = num_cliprects;						\
+      while ( _nc-- ) {							\
+	 int minx = cliprects[_nc].x1 - x_off;				\
+	 int miny = cliprects[_nc].y1 - y_off;				\
+	 int maxx = cliprects[_nc].x2 - x_off;				\
+	 int maxy = cliprects[_nc].y2 - y_off;
+
 /* ================================================================
  * Color buffer
  */
@@ -94,7 +270,41 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define TAG(x)    radeon##x##_RGB565
 #define TAG2(x,y) radeon##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 16 bit, ARGB1555 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5_REV
+
+#define TAG(x)    radeon##x##_ARGB1555
+#define TAG2(x,y) radeon##x##_ARGB1555##y
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 16 bit, RGBA4 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4_REV
+
+#define TAG(x)    radeon##x##_ARGB4444
+#define TAG2(x,y) radeon##x##_ARGB4444##y
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 32 bit, xRGB8888 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    radeon##x##_xRGB8888
+#define TAG2(x,y) radeon##x##_xRGB8888##y
+#define GET_VALUE(_x, _y) ((*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) | 0xff000000))
+#define PUT_VALUE(_x, _y, d) { \
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
 #include "spantmp2.h"
 
 /* 32 bit, ARGB8888 color spanline and pixel functions
@@ -104,7 +314,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define TAG(x)    radeon##x##_ARGB8888
 #define TAG2(x,y) radeon##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
+#define GET_VALUE(_x, _y) (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)))
+#define PUT_VALUE(_x, _y, d) { \
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
 #include "spantmp2.h"
 
 /* ================================================================
@@ -121,106 +335,127 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * too...
  */
 
-static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 4 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0..1] = 0           */
-
-#ifdef COMPILE_R300
-		ba = (y / 8) * (pitch / 8) + (x / 8);
-#else
-		ba = (y / 16) * (pitch / 16) + (x / 16);
-#endif
-
-		address |= (x & 0x7) << 2;	/* a[2..4] = x[0..2]     */
-		address |= (y & 0x3) << 5;	/* a[5..6] = y[0..1]     */
-		address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5;	/* a[7]    = x[4] ^ y[2] */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7;	/* a[11]   = x[3] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-static INLINE GLuint
-radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 2 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0]    = 0           */
-
-		ba = (y / 16) * (pitch / 32) + (x / 32);
-
-		address |= (x & 0x7) << 1;	/* a[1..3] = x[0..2]     */
-		address |= (y & 0x7) << 4;	/* a[4..6] = y[0..2]     */
-		address |= (x & 0x8) << 4;	/* a[7]    = x[3]        */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= ((x & 0x10) ^ (y & 0x10)) << 7;	/* a[11]   = x[4] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
 /* 16-bit depth buffer functions
  */
 #define VALUE_TYPE GLushort
 
+#if defined(RADEON_COMMON_FOR_R200)
+#define WRITE_DEPTH( _x, _y, d )					\
+   *(GLushort *)r200_depth_2byte(rrb, _x + x_off, _y + y_off) = d
+#else
 #define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d;
+   *(GLushort *)radeon_ptr_2byte_8x2(rrb, _x + x_off, _y + y_off) = d
+#endif
 
+#if defined(RADEON_COMMON_FOR_R200)
 #define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo ));
+   d = *(GLushort *)r200_depth_2byte(rrb, _x + x_off, _y + y_off)
+#else
+#define READ_DEPTH( d, _x, _y )						\
+   d = *(GLushort *)radeon_ptr_2byte_8x2(rrb, _x + x_off, _y + y_off)
+#endif
 
 #define TAG(x) radeon##x##_z16
 #include "depthtmp.h"
 
-/* 24 bit depth, 8 bit stencil depthbuffer functions
+/* 24 bit depth
  *
  * Careful: It looks like the R300 uses ZZZS byte order while the R200
  * uses SZZZ for 24 bit depth, 8 bit stencil mode.
  */
 #define VALUE_TYPE GLuint
 
-#ifdef COMPILE_R300
+#if defined(COMPILE_R300)
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0x000000ff;							\
    tmp |= ((d << 8) & 0xffffff00);					\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
+} while (0)
+#elif defined(RADEON_COMMON_FOR_R200)
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
+   tmp &= 0xff000000;							\
+   tmp |= ((d) & 0x00ffffff);						\
+   *_ptr = tmp;					\
 } while (0)
 #else
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );	\
+   GLuint tmp = *_ptr;							\
    tmp &= 0xff000000;							\
    tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #endif
 
-#ifdef COMPILE_R300
+#if defined(COMPILE_R300)
 #define READ_DEPTH( d, _x, _y )						\
-  do { \
-    d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,		\
-					 _y + yo )) & 0xffffff00) >> 8; \
+  do {									\
+    d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0xffffff00) >> 8; \
+  }while(0)
+#elif defined(RADEON_COMMON_FOR_R200)
+#define READ_DEPTH( d, _x, _y )						\
+  do {									\
+    d = *(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off)) & 0x00ffffff; \
   }while(0)
 #else
+#define READ_DEPTH( d, _x, _y )	\
+  d = *(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off,	_y + y_off)) & 0x00ffffff;
+#endif
+
+#define TAG(x) radeon##x##_z24
+#include "depthtmp.h"
+
+/* 24 bit depth, 8 bit stencil depthbuffer functions
+ * EXT_depth_stencil
+ *
+ * Careful: It looks like the R300 uses ZZZS byte order while the R200
+ * uses SZZZ for 24 bit depth, 8 bit stencil mode.
+ */
+#define VALUE_TYPE GLuint
+
+#if defined(COMPILE_R300)
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
+#elif defined(RADEON_COMMON_FOR_R200)
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = z24s8_to_s8z24(d);					\
+   *_ptr = tmp;								\
+} while (0)
+#else
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );	\
+   GLuint tmp = z24s8_to_s8z24(d);					\
+   *_ptr = tmp;					\
+} while (0)
+#endif
+
+#if defined(COMPILE_R300)
+#define READ_DEPTH( d, _x, _y )						\
+  do { \
+    d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)));	\
+  }while(0)
+#elif defined(RADEON_COMMON_FOR_R200)
 #define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
+  do { \
+    d = s8z24_to_z24s8(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off)));	\
+  }while(0)
+#else
+#define READ_DEPTH( d, _x, _y )	do {					\
+    d = s8z24_to_z24s8(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off,	_y + y_off ))); \
+  } while (0)
 #endif
 
 #define TAG(x) radeon##x##_z24_s8
@@ -235,35 +470,51 @@ do {									\
 #ifdef COMPILE_R300
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off);		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0xffffff00;							\
    tmp |= (d) & 0xff;							\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
+} while (0)
+#elif defined(RADEON_COMMON_FOR_R200)
+#define WRITE_STENCIL( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)r200_depth_4byte(rrb, _x + x_off, _y + y_off);		\
+   GLuint tmp = *_ptr;				\
+   tmp &= 0x00ffffff;							\
+   tmp |= (((d) & 0xff) << 24);						\
+   *_ptr = tmp;					\
 } while (0)
 #else
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off);		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0x00ffffff;							\
    tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #endif
 
 #ifdef COMPILE_R300
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    d = tmp & 0x000000ff;						\
 } while (0)
+#elif defined(RADEON_COMMON_FOR_R200)
+#define READ_STENCIL( d, _x, _y )					\
+do {									\
+   GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
+   d = (tmp & 0xff000000) >> 24;					\
+} while (0)
 #else
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    d = (tmp & 0xff000000) >> 24;					\
 } while (0)
 #endif
@@ -271,29 +522,110 @@ do {									\
 #define TAG(x) radeon##x##_z24_s8
 #include "stenciltmp.h"
 
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
+
+static void map_unmap_rb(struct gl_renderbuffer *rb, int flag)
+{
+	struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+	int r;
+
+	if (rrb == NULL || !rrb->bo)
+		return;
+
+	if (flag) {
+		if (rrb->bo->bom->funcs->bo_wait)
+			radeon_bo_wait(rrb->bo);
+		r = radeon_bo_map(rrb->bo, 1);
+		if (r) {
+			fprintf(stderr, "(%s) error(%d) mapping buffer.\n",
+				__FUNCTION__, r);
+		}
+
+		radeonSetSpanFunctions(rrb);
+	} else {
+		radeon_bo_unmap(rrb->bo);
+		rb->GetRow = NULL;
+		rb->PutRow = NULL;
+	}
+}
+
+static void
+radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map)
+{
+	GLuint i, j;
+
+	/* color draw buffers */
+	for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++)
+		map_unmap_rb(ctx->DrawBuffer->_ColorDrawBuffers[j], map);
+
+	/* check for render to textures */
+	for (i = 0; i < BUFFER_COUNT; i++) {
+		struct gl_renderbuffer_attachment *att =
+			ctx->DrawBuffer->Attachment + i;
+		struct gl_texture_object *tex = att->Texture;
+		if (tex) {
+			/* Render to texture. Note that a mipmapped texture need not
+			 * be complete for render to texture, so we must restrict to
+			 * mapping only the attached image.
+			 */
+			radeon_texture_image *image = get_radeon_texture_image(tex->Image[att->CubeMapFace][att->TextureLevel]);
+			ASSERT(att->Renderbuffer);
+
+			if (map)
+				radeon_teximage_map(image, GL_TRUE);
+			else
+				radeon_teximage_unmap(image);
+		}
+	}
+
+	map_unmap_rb(ctx->ReadBuffer->_ColorReadBuffer, map);
+
+	/* depth buffer (Note wrapper!) */
+	if (ctx->DrawBuffer->_DepthBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_DepthBuffer->Wrapped, map);
+
+	if (ctx->DrawBuffer->_StencilBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_StencilBuffer->Wrapped, map);
+}
 
 static void radeonSpanRenderStart(GLcontext * ctx)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-#ifdef COMPILE_R300
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-	R300_FIREVERTICES(r300);
-#else
-	RADEON_FIREVERTICES(rmesa);
-#endif
-	LOCK_HARDWARE(rmesa);
-	radeonWaitForIdleLocked(rmesa);
+	int i;
+
+	radeon_firevertices(rmesa);
+
+	/* The locking and wait for idle should really only be needed in classic mode.
+	 * In a future memory manager based implementation, this should become
+	 * unnecessary due to the fact that mapping our buffers, textures, etc.
+	 * should implicitly wait for any previous rendering commands that must
+	 * be waited on. */
+	if (!rmesa->radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(rmesa);
+		radeonWaitForIdleLocked(rmesa);
+	}
+
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+		if (ctx->Texture.Unit[i]._ReallyEnabled)
+			ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current);
+	}
+
+	radeon_map_unmap_buffers(ctx, 1);
 }
 
 static void radeonSpanRenderFinish(GLcontext * ctx)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	int i;
 	_swrast_flush(ctx);
-	UNLOCK_HARDWARE(rmesa);
+	if (!rmesa->radeonScreen->driScreen->dri2.enabled) {
+		UNLOCK_HARDWARE(rmesa);
+	}
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+		if (ctx->Texture.Unit[i]._ReallyEnabled)
+			ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current);
+	}
+
+	radeon_map_unmap_buffers(ctx, 0);
 }
 
 void radeonInitSpanFuncs(GLcontext * ctx)
@@ -307,20 +639,27 @@ void radeonInitSpanFuncs(GLcontext * ctx)
 /**
  * Plug in the Get/Put routines for the given driRenderbuffer.
  */
-void radeonSetSpanFunctions(driRenderbuffer * drb, const GLvisual * vis)
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb)
 {
-	if (drb->Base.InternalFormat == GL_RGBA) {
-		if (vis->redBits == 5 && vis->greenBits == 6
-		    && vis->blueBits == 5) {
-			radeonInitPointers_RGB565(&drb->Base);
-		} else {
-			radeonInitPointers_ARGB8888(&drb->Base);
-		}
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-		radeonInitDepthPointers_z16(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-		radeonInitDepthPointers_z24_s8(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-		radeonInitStencilPointers_z24_s8(&drb->Base);
+	if (rrb->base._ActualFormat == GL_RGB5) {
+		radeonInitPointers_RGB565(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGB8) {
+		radeonInitPointers_xRGB8888(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGBA8) {
+		radeonInitPointers_ARGB8888(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGBA4) {
+		radeonInitPointers_ARGB4444(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGB5_A1) {
+		radeonInitPointers_ARGB1555(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT16) {
+		radeonInitDepthPointers_z16(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT24) {
+		radeonInitDepthPointers_z24(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH24_STENCIL8_EXT) {
+		radeonInitDepthPointers_z24_s8(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_STENCIL_INDEX8_EXT) {
+		radeonInitStencilPointers_z24_s8(&rrb->base);
+	} else {
+		fprintf(stderr, "radeonSetSpanFunctions: bad actual format: 0x%04X\n", rrb->base._ActualFormat);
 	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.h b/src/mesa/drivers/dri/radeon/radeon_span.h
index 9abe0864b1..ea6a2e7fb4 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.h
+++ b/src/mesa/drivers/dri/radeon/radeon_span.h
@@ -42,9 +42,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __RADEON_SPAN_H__
 #define __RADEON_SPAN_H__
 
-#include "drirenderbuffer.h"
-
 extern void radeonInitSpanFuncs(GLcontext * ctx);
-extern void radeonSetSpanFunctions(driRenderbuffer * rb, const GLvisual * vis);
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index b6561001e7..4d0d35ee0c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -40,6 +40,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/state.h"
 #include "main/context.h"
 #include "main/framebuffer.h"
+#include "main/simple_list.h"
 
 #include "vbo/vbo.h"
 #include "tnl/tnl.h"
@@ -47,6 +48,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast_setup/swrast_setup.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
 #include "radeon_tcl.h"
@@ -62,7 +64,7 @@ static void radeonUpdateSpecular( GLcontext *ctx );
 
 static void radeonAlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    int pp_misc = rmesa->hw.ctx.cmd[CTX_PP_MISC];
    GLubyte refByte;
 
@@ -106,7 +108,7 @@ static void radeonAlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 static void radeonBlendEquationSeparate( GLcontext *ctx,
 					 GLenum modeRGB, GLenum modeA )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] & ~RADEON_COMB_FCN_MASK;
    GLboolean fallback = GL_FALSE;
 
@@ -147,8 +149,8 @@ static void radeonBlendFuncSeparate( GLcontext *ctx,
 				     GLenum sfactorRGB, GLenum dfactorRGB,
 				     GLenum sfactorA, GLenum dfactorA )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] & 
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] &
       ~(RADEON_SRC_BLEND_MASK | RADEON_DST_BLEND_MASK);
    GLboolean fallback = GL_FALSE;
 
@@ -257,7 +259,7 @@ static void radeonBlendFuncSeparate( GLcontext *ctx,
 
 static void radeonDepthFunc( GLcontext *ctx, GLenum func )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, ctx );
    rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_Z_TEST_MASK;
@@ -293,7 +295,7 @@ static void radeonDepthFunc( GLcontext *ctx, GLenum func )
 
 static void radeonDepthMask( GLcontext *ctx, GLboolean flag )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    RADEON_STATECHANGE( rmesa, ctx );
 
    if ( ctx->Depth.Mask ) {
@@ -305,16 +307,16 @@ static void radeonDepthMask( GLcontext *ctx, GLboolean flag )
 
 static void radeonClearDepth( GLcontext *ctx, GLclampd d )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint format = (rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &
 		    RADEON_DEPTH_FORMAT_MASK);
 
    switch ( format ) {
    case RADEON_DEPTH_FORMAT_16BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x0000ffff;
+      rmesa->radeon.state.depth.clear = d * 0x0000ffff;
       break;
    case RADEON_DEPTH_FORMAT_24BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x00ffffff;
+      rmesa->radeon.state.depth.clear = d * 0x00ffffff;
       break;
    }
 }
@@ -327,7 +329,7 @@ static void radeonClearDepth( GLcontext *ctx, GLclampd d )
 
 static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    union { int i; float f; } c, d;
    GLchan col[4];
 
@@ -391,7 +393,7 @@ static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 	 rmesa->hw.fog.cmd[FOG_D] = d.i;
       }
       break;
-   case GL_FOG_COLOR: 
+   case GL_FOG_COLOR:
       RADEON_STATECHANGE( rmesa, ctx );
       UNCLAMPED_FLOAT_TO_RGB_CHAN( col, ctx->Fog.Color );
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~RADEON_FOG_COLOR_MASK;
@@ -406,109 +408,13 @@ static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
    }
 }
 
-
-/* =============================================================
- * Scissoring
- */
-
-
-static GLboolean intersect_rect( drm_clip_rect_t *out,
-				 drm_clip_rect_t *a,
-				 drm_clip_rect_t *b )
-{
-   *out = *a;
-   if ( b->x1 > out->x1 ) out->x1 = b->x1;
-   if ( b->y1 > out->y1 ) out->y1 = b->y1;
-   if ( b->x2 < out->x2 ) out->x2 = b->x2;
-   if ( b->y2 < out->y2 ) out->y2 = b->y2;
-   if ( out->x1 >= out->x2 ) return GL_FALSE;
-   if ( out->y1 >= out->y2 ) return GL_FALSE;
-   return GL_TRUE;
-}
-
-
-void radeonRecalcScissorRects( radeonContextPtr rmesa )
-{
-   drm_clip_rect_t *out;
-   int i;
-
-   /* Grow cliprect store?
-    */
-   if (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-      while (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-	 rmesa->state.scissor.numAllocedClipRects += 1;	/* zero case */
-	 rmesa->state.scissor.numAllocedClipRects *= 2;
-      }
-
-      if (rmesa->state.scissor.pClipRects)
-	 FREE(rmesa->state.scissor.pClipRects);
-
-      rmesa->state.scissor.pClipRects = 
-	 MALLOC( rmesa->state.scissor.numAllocedClipRects * 
-		 sizeof(drm_clip_rect_t) );
-
-      if ( rmesa->state.scissor.pClipRects == NULL ) {
-	 rmesa->state.scissor.numAllocedClipRects = 0;
-	 return;
-      }
-   }
-   
-   out = rmesa->state.scissor.pClipRects;
-   rmesa->state.scissor.numClipRects = 0;
-
-   for ( i = 0 ; i < rmesa->numClipRects ;  i++ ) {
-      if ( intersect_rect( out, 
-			   &rmesa->pClipRects[i], 
-			   &rmesa->state.scissor.rect ) ) {
-	 rmesa->state.scissor.numClipRects++;
-	 out++;
-      }
-   }
-}
-
-
-static void radeonUpdateScissor( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if ( rmesa->dri.drawable ) {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-
-      int x = ctx->Scissor.X;
-      int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
-      int w = ctx->Scissor.X + ctx->Scissor.Width - 1;
-      int h = dPriv->h - ctx->Scissor.Y - 1;
-
-      rmesa->state.scissor.rect.x1 = x + dPriv->x;
-      rmesa->state.scissor.rect.y1 = y + dPriv->y;
-      rmesa->state.scissor.rect.x2 = w + dPriv->x + 1;
-      rmesa->state.scissor.rect.y2 = h + dPriv->y + 1;
-
-      radeonRecalcScissorRects( rmesa );
-   }
-}
-
-
-static void radeonScissor( GLcontext *ctx,
-			   GLint x, GLint y, GLsizei w, GLsizei h )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if ( ctx->Scissor.Enabled ) {
-      RADEON_FIREVERTICES( rmesa );	/* don't pipeline cliprect changes */
-      radeonUpdateScissor( ctx );
-   }
-
-}
-
-
 /* =============================================================
  * Culling
  */
 
 static void radeonCullFace( GLcontext *ctx, GLenum unused )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint s = rmesa->hw.set.cmd[SET_SE_CNTL];
    GLuint t = rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL];
 
@@ -545,7 +451,7 @@ static void radeonCullFace( GLcontext *ctx, GLenum unused )
 
 static void radeonFrontFace( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, set );
    rmesa->hw.set.cmd[SET_SE_CNTL] &= ~RADEON_FFACE_CULL_DIR_MASK;
@@ -553,6 +459,10 @@ static void radeonFrontFace( GLcontext *ctx, GLenum mode )
    RADEON_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~RADEON_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= RADEON_FFACE_CULL_CW;
@@ -570,7 +480,7 @@ static void radeonFrontFace( GLcontext *ctx, GLenum mode )
  */
 static void radeonLineWidth( GLcontext *ctx, GLfloat widthf )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, lin );
    RADEON_STATECHANGE( rmesa, set );
@@ -587,10 +497,10 @@ static void radeonLineWidth( GLcontext *ctx, GLfloat widthf )
 
 static void radeonLineStipple( GLcontext *ctx, GLint factor, GLushort pattern )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, lin );
-   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] = 
+   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] =
       ((((GLuint)factor & 0xff) << 16) | ((GLuint)pattern));
 }
 
@@ -602,12 +512,19 @@ static void radeonColorMask( GLcontext *ctx,
 			     GLboolean r, GLboolean g,
 			     GLboolean b, GLboolean a )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint mask = radeonPackColor( rmesa->radeonScreen->cpp,
-				  ctx->Color.ColorMask[RCOMP],
-				  ctx->Color.ColorMask[GCOMP],
-				  ctx->Color.ColorMask[BCOMP],
-				  ctx->Color.ColorMask[ACOMP] );
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb;
+   GLuint mask;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+
+   mask = radeonPackColor( rrb->cpp,
+			   ctx->Color.ColorMask[RCOMP],
+			   ctx->Color.ColorMask[GCOMP],
+			   ctx->Color.ColorMask[BCOMP],
+			   ctx->Color.ColorMask[ACOMP] );
 
    if ( rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK] != mask ) {
       RADEON_STATECHANGE( rmesa, msk );
@@ -623,8 +540,9 @@ static void radeonColorMask( GLcontext *ctx,
 static void radeonPolygonOffset( GLcontext *ctx,
 				 GLfloat factor, GLfloat units )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->state.depth.scale };
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
    RADEON_STATECHANGE( rmesa, zbs );
@@ -632,41 +550,16 @@ static void radeonPolygonOffset( GLcontext *ctx,
    rmesa->hw.zbs.cmd[ZBS_SE_ZBIAS_CONSTANT] = constant.ui32;
 }
 
-static void radeonPolygonStipple( GLcontext *ctx, const GLubyte *mask )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint i;
-   drm_radeon_stipple_t stipple;
-
-   /* Must flip pattern upside down.
-    */
-   for ( i = 0 ; i < 32 ; i++ ) {
-      rmesa->state.stipple.mask[31 - i] = ((GLuint *) mask)[i];
-   }
-
-   /* TODO: push this into cmd mechanism
-    */
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* FIXME: Use window x,y offsets into stipple RAM.
-    */
-   stipple.mask = rmesa->state.stipple.mask;
-   drmCommandWrite( rmesa->dri.fd, DRM_RADEON_STIPPLE, 
-                    &stipple, sizeof(drm_radeon_stipple_t) );
-   UNLOCK_HARDWARE( rmesa );
-}
-
 static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean flag = (ctx->_TriangleCaps & DD_TRI_UNFILLED) != 0;
 
    /* Can't generally do unfilled via tcl, but some good special
-    * cases work. 
+    * cases work.
     */
    TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_UNFILLED, flag);
-   if (rmesa->TclFallback) {
+   if (rmesa->radeon.TclFallback) {
       radeonChooseRenderState( ctx );
       radeonChooseVertexState( ctx );
    }
@@ -686,7 +579,7 @@ static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
  */
 static void radeonUpdateSpecular( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    uint32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL];
    GLuint flag = 0;
 
@@ -711,7 +604,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_TCL_VTX_PK_DIFFUSE;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] |= RADEON_LIGHTING_ENABLE;
       p |=  RADEON_SPECULAR_ENABLE;
-      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &= 
+      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &=
 	 ~RADEON_DIFFUSE_SPECULAR_COMBINE;
    }
    else if (ctx->Light.Enabled) {
@@ -741,7 +634,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
 	    RADEON_TCL_COMPUTE_SPECULAR) != 0;
       }
    }
- 
+
    TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_FOGCOORDSPEC, flag);
 
    if (NEED_SECONDARY_COLOR(ctx)) {
@@ -757,7 +650,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
 
    /* Update vertex/render formats
     */
-   if (rmesa->TclFallback) { 
+   if (rmesa->radeon.TclFallback) {
       radeonChooseRenderState( ctx );
       radeonChooseVertexState( ctx );
    }
@@ -769,12 +662,12 @@ static void radeonUpdateSpecular( GLcontext *ctx )
  */
 
 
-/* Update on colormaterial, material emmissive/ambient, 
+/* Update on colormaterial, material emmissive/ambient,
  * lightmodel.globalambient
  */
 static void update_global_ambient( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    float *fcmd = (float *)RADEON_DB_STATE( glt );
 
    /* Need to do more if both emmissive & ambient are PREMULT:
@@ -782,23 +675,23 @@ static void update_global_ambient( GLcontext *ctx )
     */
    if ((rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &
        ((3 << RADEON_EMISSIVE_SOURCE_SHIFT) |
-	(3 << RADEON_AMBIENT_SOURCE_SHIFT))) == 0) 
+	(3 << RADEON_AMBIENT_SOURCE_SHIFT))) == 0)
    {
-      COPY_3V( &fcmd[GLT_RED], 
+      COPY_3V( &fcmd[GLT_RED],
 	       ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_EMISSION]);
       ACC_SCALE_3V( &fcmd[GLT_RED],
 		   ctx->Light.Model.Ambient,
 		   ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT]);
-   } 
+   }
    else
    {
       COPY_3V( &fcmd[GLT_RED], ctx->Light.Model.Ambient );
    }
-   
+
    RADEON_DB_STATECHANGE(rmesa, &rmesa->hw.glt);
 }
 
-/* Update on change to 
+/* Update on change to
  *    - light[p].colors
  *    - light[p].enabled
  */
@@ -809,13 +702,13 @@ static void update_light_colors( GLcontext *ctx, GLuint p )
 /*     fprintf(stderr, "%s\n", __FUNCTION__); */
 
    if (l->Enabled) {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+      r100ContextPtr rmesa = R100_CONTEXT(ctx);
       float *fcmd = (float *)RADEON_DB_STATE( lit[p] );
 
-      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );	 
+      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );
       COPY_4V( &fcmd[LIT_DIFFUSE_RED], l->Diffuse );
       COPY_4V( &fcmd[LIT_SPECULAR_RED], l->Specular );
-      
+
       RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.lit[p] );
    }
 }
@@ -829,7 +722,7 @@ static void check_twoside_fallback( GLcontext *ctx )
 
    if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
       if (ctx->Light.ColorMaterialEnabled &&
-	  (ctx->Light.ColorMaterialBitmask & BACK_MATERIAL_BITS) != 
+	  (ctx->Light.ColorMaterialBitmask & BACK_MATERIAL_BITS) !=
 	  ((ctx->Light.ColorMaterialBitmask & FRONT_MATERIAL_BITS)<<1))
 	 fallback = GL_TRUE;
       else {
@@ -837,7 +730,7 @@ static void check_twoside_fallback( GLcontext *ctx )
 	    if (memcmp( ctx->Light.Material.Attrib[i],
 			ctx->Light.Material.Attrib[i+1],
 			sizeof(GLfloat)*4) != 0) {
-	       fallback = GL_TRUE;  
+	       fallback = GL_TRUE;
 	       break;
 	    }
       }
@@ -849,14 +742,14 @@ static void check_twoside_fallback( GLcontext *ctx )
 
 static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+      r100ContextPtr rmesa = R100_CONTEXT(ctx);
       GLuint light_model_ctl1 = rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL];
 
       light_model_ctl1 &= ~((3 << RADEON_EMISSIVE_SOURCE_SHIFT) |
 			   (3 << RADEON_AMBIENT_SOURCE_SHIFT) |
 			   (3 << RADEON_DIFFUSE_SOURCE_SHIFT) |
-			   (3 << RADEON_SPECULAR_SOURCE_SHIFT)); 
-   
+			   (3 << RADEON_SPECULAR_SOURCE_SHIFT));
+
    if (ctx->Light.ColorMaterialEnabled) {
       GLuint mask = ctx->Light.ColorMaterialBitmask;
 
@@ -877,7 +770,7 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_STATE_MULT <<
 			     RADEON_AMBIENT_SOURCE_SHIFT);
       }
-	 
+
       if (mask & MAT_BIT_FRONT_DIFFUSE) {
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_VERTEX_DIFFUSE <<
 			     RADEON_DIFFUSE_SOURCE_SHIFT);
@@ -886,7 +779,7 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_STATE_MULT <<
 			     RADEON_DIFFUSE_SOURCE_SHIFT);
       }
-   
+
       if (mask & MAT_BIT_FRONT_SPECULAR) {
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_VERTEX_DIFFUSE <<
 			     RADEON_SPECULAR_SOURCE_SHIFT);
@@ -904,27 +797,27 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 		   (RADEON_LM_SOURCE_STATE_MULT << RADEON_DIFFUSE_SOURCE_SHIFT) |
 		   (RADEON_LM_SOURCE_STATE_MULT << RADEON_SPECULAR_SOURCE_SHIFT);
    }
-   
+
       if (light_model_ctl1 != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]) {
 	 RADEON_STATECHANGE( rmesa, tcl );
-	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = light_model_ctl1;      
+	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = light_model_ctl1;
    }
 }
 
 void radeonUpdateMaterial( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLfloat (*mat)[4] = ctx->Light.Material.Attrib;
    GLfloat *fcmd = (GLfloat *)RADEON_DB_STATE( mtl );
    GLuint mask = ~0;
-   
+
    if (ctx->Light.ColorMaterialEnabled)
       mask &= ~ctx->Light.ColorMaterialBitmask;
 
-   if (RADEON_DEBUG & DEBUG_STATE)
+   if (RADEON_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-      
+
    if (mask & MAT_BIT_FRONT_EMISSION) {
       fcmd[MTL_EMMISSIVE_RED]   = mat[MAT_ATTRIB_FRONT_EMISSION][0];
       fcmd[MTL_EMMISSIVE_GREEN] = mat[MAT_ATTRIB_FRONT_EMISSION][1];
@@ -974,11 +867,11 @@ void radeonUpdateMaterial( GLcontext *ctx )
  *
  * which are calculated in light.c and are correct for the current
  * lighting space (model or eye), hence dependencies on _NEW_MODELVIEW
- * and _MESA_NEW_NEED_EYE_COORDS.  
+ * and _MESA_NEW_NEED_EYE_COORDS.
  */
 static void update_light( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    /* Have to check these, or have an automatic shortcircuit mechanism
     * to remove noop statechanges. (Or just do a better job on the
@@ -991,12 +884,12 @@ static void update_light( GLcontext *ctx )
 	 tmp &= ~RADEON_LIGHT_IN_MODELSPACE;
       else
 	 tmp |= RADEON_LIGHT_IN_MODELSPACE;
-      
+
 
       /* Leave this test disabled: (unexplained q3 lockup) (even with
          new packets)
       */
-      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]) 
+      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL])
       {
 	 RADEON_STATECHANGE( rmesa, tcl );
 	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = tmp;
@@ -1020,10 +913,10 @@ static void update_light( GLcontext *ctx )
 	 if (ctx->Light.Light[p].Enabled) {
 	    struct gl_light *l = &ctx->Light.Light[p];
 	    GLfloat *fcmd = (GLfloat *)RADEON_DB_STATE( lit[p] );
-	    
+
 	    if (l->EyePosition[3] == 0.0) {
-	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm ); 
-	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm ); 
+	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm );
+	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm );
 	       fcmd[LIT_POSITION_W] = 0;
 	       fcmd[LIT_DIRECTION_W] = 0;
 	    } else {
@@ -1043,30 +936,30 @@ static void update_light( GLcontext *ctx )
 static void radeonLightfv( GLcontext *ctx, GLenum light,
 			   GLenum pname, const GLfloat *params )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLint p = light - GL_LIGHT0;
    struct gl_light *l = &ctx->Light.Light[p];
    GLfloat *fcmd = (GLfloat *)rmesa->hw.lit[p].cmd;
-   
+
 
    switch (pname) {
-   case GL_AMBIENT:		
+   case GL_AMBIENT:
    case GL_DIFFUSE:
    case GL_SPECULAR:
       update_light_colors( ctx, p );
       break;
 
-   case GL_SPOT_DIRECTION: 
-      /* picked up in update_light */	
+   case GL_SPOT_DIRECTION:
+      /* picked up in update_light */
       break;
 
    case GL_POSITION: {
-      /* positions picked up in update_light, but can do flag here */	
+      /* positions picked up in update_light, but can do flag here */
       GLuint flag;
       GLuint idx = TCL_PER_LIGHT_CTL_0 + p/2;
 
       /* FIXME: Set RANGE_ATTEN only when needed */
-      if (p&1) 
+      if (p&1)
 	 flag = RADEON_LIGHT_1_IS_LOCAL;
       else
 	 flag = RADEON_LIGHT_0_IS_LOCAL;
@@ -1158,16 +1051,16 @@ static void radeonLightfv( GLcontext *ctx, GLenum light,
    }
 }
 
-		  
+
 
 
 static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 				const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    switch (pname) {
-      case GL_LIGHT_MODEL_AMBIENT: 
+      case GL_LIGHT_MODEL_AMBIENT:
 	 update_global_ambient( ctx );
 	 break;
 
@@ -1188,7 +1081,7 @@ static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 
 	 check_twoside_fallback( ctx );
 
-	 if (rmesa->TclFallback) {
+	 if (rmesa->radeon.TclFallback) {
 	    radeonChooseRenderState( ctx );
 	    radeonChooseVertexState( ctx );
 	 }
@@ -1205,7 +1098,7 @@ static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 
 static void radeonShadeModel( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint s = rmesa->hw.set.cmd[SET_SE_CNTL];
 
    s &= ~(RADEON_DIFFUSE_SHADE_MASK |
@@ -1244,7 +1137,7 @@ static void radeonShadeModel( GLcontext *ctx, GLenum mode )
 static void radeonClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 {
    GLint p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p];
 
    RADEON_STATECHANGE( rmesa, ucp[p] );
@@ -1256,7 +1149,7 @@ static void radeonClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 
 static void radeonUpdateClipPlanes( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint p;
 
    for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
@@ -1281,7 +1174,7 @@ static void
 radeonStencilFuncSeparate( GLcontext *ctx, GLenum face, GLenum func,
                            GLint ref, GLuint mask )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint refmask = (((ctx->Stencil.Ref[0] & 0xff) << RADEON_STENCIL_REF_SHIFT) |
 		     ((ctx->Stencil.ValueMask[0] & 0xff) << RADEON_STENCIL_MASK_SHIFT));
 
@@ -1325,7 +1218,7 @@ radeonStencilFuncSeparate( GLcontext *ctx, GLenum face, GLenum func,
 static void
 radeonStencilMaskSeparate( GLcontext *ctx, GLenum face, GLuint mask )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, msk );
    rmesa->hw.msk.cmd[MSK_RB3D_STENCILREFMASK] &= ~RADEON_STENCIL_WRITE_MASK;
@@ -1336,20 +1229,20 @@ radeonStencilMaskSeparate( GLcontext *ctx, GLenum face, GLuint mask )
 static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
                                      GLenum zfail, GLenum zpass )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    /* radeon 7200 have stencil bug, DEC and INC_WRAP will actually both do DEC_WRAP,
       and DEC_WRAP (and INVERT) will do INVERT. No way to get correct INC_WRAP and DEC,
       but DEC_WRAP can be fixed by using DEC and INC_WRAP at least use INC. */
-   
+
    GLuint tempRADEON_STENCIL_FAIL_DEC_WRAP;
    GLuint tempRADEON_STENCIL_FAIL_INC_WRAP;
    GLuint tempRADEON_STENCIL_ZFAIL_DEC_WRAP;
    GLuint tempRADEON_STENCIL_ZFAIL_INC_WRAP;
    GLuint tempRADEON_STENCIL_ZPASS_DEC_WRAP;
    GLuint tempRADEON_STENCIL_ZPASS_INC_WRAP;
-   
-   if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_BROKEN_STENCIL) {
+
+   if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_BROKEN_STENCIL) {
       tempRADEON_STENCIL_FAIL_DEC_WRAP = RADEON_STENCIL_FAIL_DEC;
       tempRADEON_STENCIL_FAIL_INC_WRAP = RADEON_STENCIL_FAIL_INC;
       tempRADEON_STENCIL_ZFAIL_DEC_WRAP = RADEON_STENCIL_ZFAIL_DEC;
@@ -1365,7 +1258,7 @@ static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
       tempRADEON_STENCIL_ZPASS_DEC_WRAP = RADEON_STENCIL_ZPASS_DEC_WRAP;
       tempRADEON_STENCIL_ZPASS_INC_WRAP = RADEON_STENCIL_ZPASS_INC_WRAP;
    }
-   
+
    RADEON_STATECHANGE( rmesa, ctx );
    rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &= ~(RADEON_STENCIL_FAIL_MASK |
 					       RADEON_STENCIL_ZFAIL_MASK |
@@ -1455,9 +1348,9 @@ static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
 
 static void radeonClearStencil( GLcontext *ctx, GLint s )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   rmesa->state.stencil.clear = 
+   rmesa->radeon.state.stencil.clear =
       ((GLuint) (ctx->Stencil.Clear & 0xff) |
        (0xff << RADEON_STENCIL_MASK_SHIFT) |
        ((ctx->Stencil.WriteMask[0] & 0xff) << RADEON_STENCIL_WRITEMASK_SHIFT));
@@ -1481,20 +1374,30 @@ static void radeonClearStencil( GLcontext *ctx, GLint s )
  */
 void radeonUpdateWindow( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLfloat xoffset = (GLfloat)dPriv->x;
-   GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+   GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
-   RADEON_FIREVERTICES( rmesa );
    RADEON_STATECHANGE( rmesa, vpt );
 
    rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE]  = sx.ui32;
@@ -1514,6 +1417,8 @@ static void radeonViewport( GLcontext *ctx, GLint x, GLint y,
     * values, or keep the originals hanging around.
     */
    radeonUpdateWindow( ctx );
+
+   radeon_viewport(ctx, x, y, width, height);
 }
 
 static void radeonDepthRange( GLcontext *ctx, GLclampd nearval,
@@ -1524,8 +1429,8 @@ static void radeonDepthRange( GLcontext *ctx, GLclampd nearval,
 
 void radeonUpdateViewportOffset( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    GLfloat xoffset = (GLfloat)dPriv->x;
    GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
@@ -1555,8 +1460,8 @@ void radeonUpdateViewportOffset( GLcontext *ctx )
                 RADEON_STIPPLE_Y_OFFSET_MASK);
 
          /* add magic offsets, then invert */
-         stx = 31 - ((rmesa->dri.drawable->x - 1) & RADEON_STIPPLE_COORD_MASK);
-         sty = 31 - ((rmesa->dri.drawable->y + rmesa->dri.drawable->h - 1)
+         stx = 31 - ((dPriv->x - 1) & RADEON_STIPPLE_COORD_MASK);
+         sty = 31 - ((dPriv->y + dPriv->h - 1)
                      & RADEON_STIPPLE_COORD_MASK);
 
          m |= ((stx << RADEON_STIPPLE_X_OFFSET_SHIFT) |
@@ -1580,20 +1485,26 @@ void radeonUpdateViewportOffset( GLcontext *ctx )
 
 static void radeonClearColor( GLcontext *ctx, const GLfloat color[4] )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLubyte c[4];
+   struct radeon_renderbuffer *rrb;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+     
    CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
    CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]);
    CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]);
    CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]);
-   rmesa->state.color.clear = radeonPackColor( rmesa->radeonScreen->cpp,
+   rmesa->radeon.state.color.clear = radeonPackColor( rrb->cpp,
 					       c[0], c[1], c[2], c[3] );
 }
 
 
 static void radeonRenderMode( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    FALLBACK( rmesa, RADEON_FALLBACK_RENDER_MODE, (mode != GL_RENDER) );
 }
 
@@ -1619,7 +1530,7 @@ static GLuint radeon_rop_tab[] = {
 
 static void radeonLogicOpCode( GLcontext *ctx, GLenum opcode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint rop = (GLuint)opcode - GL_CLEAR;
 
    ASSERT( rop < 16 );
@@ -1628,108 +1539,16 @@ static void radeonLogicOpCode( GLcontext *ctx, GLenum opcode )
    rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = radeon_rop_tab[rop];
 }
 
-
-/**
- * Set up the cliprects for either front or back-buffer drawing.
- */
-void radeonSetCliprects( radeonContextPtr rmesa )
-{
-   __DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *const readable = rmesa->dri.readable;
-   GLframebuffer *const draw_fb = (GLframebuffer*) drawable->driverPrivate;
-   GLframebuffer *const read_fb = (GLframebuffer*) readable->driverPrivate;
-
-   if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* Can't ignore 2d windows if we are page flipping.
-       */
-      if ( drawable->numBackClipRects == 0 || rmesa->doPageFlip ) {
-	 rmesa->numClipRects = drawable->numClipRects;
-	 rmesa->pClipRects = drawable->pClipRects;
-      }
-      else {
-	 rmesa->numClipRects = drawable->numBackClipRects;
-	 rmesa->pClipRects = drawable->pBackClipRects;
-      }
-   }
-   else {
-      /* front buffer (or none, or multiple buffers */
-      rmesa->numClipRects = drawable->numClipRects;
-      rmesa->pClipRects = drawable->pClipRects;
-   }
-
-   if ((draw_fb->Width != drawable->w) || (draw_fb->Height != drawable->h)) {
-      _mesa_resize_framebuffer(rmesa->glCtx, draw_fb,
-			       drawable->w, drawable->h);
-      draw_fb->Initialized = GL_TRUE;
-   }
-
-   if (drawable != readable) {
-      if ((read_fb->Width != readable->w) || (read_fb->Height != readable->h)) {
-	 _mesa_resize_framebuffer(rmesa->glCtx, read_fb,
-				  readable->w, readable->h);
-	 read_fb->Initialized = GL_TRUE;
-      }
-   }
-
-   if (rmesa->state.scissor.enabled)
-      radeonRecalcScissorRects( rmesa );
-
-   rmesa->lastStamp = drawable->lastStamp;
-}
-
-
-/**
- * Called via glDrawBuffer.
- */
-static void radeonDrawBuffer( GLcontext *ctx, GLenum mode )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s %s\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   RADEON_FIREVERTICES(rmesa);	/* don't pipeline cliprect changes */
-
-   if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) {
-      /* 0 (GL_NONE) buffers or multiple color drawing buffers */
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   switch ( ctx->DrawBuffer->_ColorDrawBufferIndexes[0] ) {
-   case BUFFER_FRONT_LEFT:
-   case BUFFER_BACK_LEFT:
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   default:
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   radeonSetCliprects( rmesa );
-
-   /* We'll set the drawing engine's offset/pitch parameters later
-    * when we update other state.
-    */
-}
-
-static void radeonReadBuffer( GLcontext *ctx, GLenum mode )
-{
-   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
-}
-
-
 /* =============================================================
  * State enable/disable
  */
 
 static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint p, flag;
 
-   if ( RADEON_DEBUG & DEBUG_STATE )
+   if ( RADEON_DEBUG & RADEON_STATE )
       fprintf( stderr, "%s( %s = %s )\n", __FUNCTION__,
 	       _mesa_lookup_enum_by_nr( cap ),
 	       state ? "GL_TRUE" : "GL_FALSE" );
@@ -1787,7 +1606,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_CLIP_PLANE2:
    case GL_CLIP_PLANE3:
    case GL_CLIP_PLANE4:
-   case GL_CLIP_PLANE5: 
+   case GL_CLIP_PLANE5:
       p = cap-GL_CLIP_PLANE0;
       RADEON_STATECHANGE( rmesa, tcl );
       if (state) {
@@ -1821,10 +1640,10 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
       RADEON_STATECHANGE(rmesa, ctx );
       if ( state ) {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->radeon.state.color.roundEnable;
       } else {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->radeon.state.color.roundEnable;
       }
       break;
 
@@ -1852,13 +1671,13 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHT7:
       RADEON_STATECHANGE(rmesa, tcl);
       p = cap - GL_LIGHT0;
-      if (p&1) 
+      if (p&1)
 	 flag = (RADEON_LIGHT_1_ENABLE |
-		 RADEON_LIGHT_1_ENABLE_AMBIENT | 
+		 RADEON_LIGHT_1_ENABLE_AMBIENT |
 		 RADEON_LIGHT_1_ENABLE_SPECULAR);
       else
 	 flag = (RADEON_LIGHT_0_ENABLE |
-		 RADEON_LIGHT_0_ENABLE_AMBIENT | 
+		 RADEON_LIGHT_0_ENABLE_AMBIENT |
 		 RADEON_LIGHT_0_ENABLE_SPECULAR);
 
       if (state)
@@ -1866,7 +1685,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
       else
 	 rmesa->hw.tcl.cmd[p/2 + TCL_PER_LIGHT_CTL_0] &= ~flag;
 
-      /* 
+      /*
        */
       update_light_colors( ctx, p );
       break;
@@ -1904,7 +1723,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_ROP_ENABLE;
       }
       break;
-      
+
    case GL_NORMALIZE:
       RADEON_STATECHANGE( rmesa, tcl );
       if ( state ) {
@@ -1971,21 +1790,30 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    }
 
    case GL_SCISSOR_TEST:
-      RADEON_FIREVERTICES( rmesa );
-      rmesa->state.scissor.enabled = state;
+      radeon_firevertices(&rmesa->radeon);
+      rmesa->radeon.state.scissor.enabled = state;
       radeonUpdateScissor( ctx );
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->state.stencil.hwBuffer ) {
-	 RADEON_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    RADEON_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
       }
       break;
 
@@ -1995,7 +1823,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_TEXTURE_GEN_T:
       /* Picked up in radeonUpdateTextureState.
        */
-      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE; 
+      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE;
       break;
 
    case GL_COLOR_SUM_EXT:
@@ -2010,11 +1838,11 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 
 static void radeonLightingSpaceChange( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean tmp;
    RADEON_STATECHANGE( rmesa, tcl );
 
-   if (RADEON_DEBUG & DEBUG_STATE)
+   if (RADEON_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s %d BEFORE %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]);
 
@@ -2029,7 +1857,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &= ~RADEON_RESCALE_NORMALS;
    }
 
-   if (RADEON_DEBUG & DEBUG_STATE) 
+   if (RADEON_DEBUG & RADEON_STATE)
       fprintf(stderr, "%s %d AFTER %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]);
 }
@@ -2039,7 +1867,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx )
  */
 
 
-void radeonUploadTexMatrix( radeonContextPtr rmesa,
+void radeonUploadTexMatrix( r100ContextPtr rmesa,
 			    int unit, GLboolean swapcols )
 {
 /* Here's how this works: on r100, only 3 tex coords can be submitted, so the
@@ -2065,7 +1893,7 @@ void radeonUploadTexMatrix( radeonContextPtr rmesa,
    int idx = TEXMAT_0 + unit;
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] )) + MAT_ELT_0;
    int i;
-   struct gl_texture_unit tUnit = rmesa->glCtx->Texture.Unit[unit];
+   struct gl_texture_unit tUnit = rmesa->radeon.glCtx->Texture.Unit[unit];
    GLfloat *src = rmesa->tmpmat[unit].m;
 
    rmesa->TexMatColSwap &= ~(1 << unit);
@@ -2119,7 +1947,7 @@ void radeonUploadTexMatrix( radeonContextPtr rmesa,
 }
 
 
-static void upload_matrix( radeonContextPtr rmesa, GLfloat *src, int idx )
+static void upload_matrix( r100ContextPtr rmesa, GLfloat *src, int idx )
 {
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] ))+MAT_ELT_0;
    int i;
@@ -2135,7 +1963,7 @@ static void upload_matrix( radeonContextPtr rmesa, GLfloat *src, int idx )
    RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.mat[idx] );
 }
 
-static void upload_matrix_t( radeonContextPtr rmesa, GLfloat *src, int idx )
+static void upload_matrix_t( r100ContextPtr rmesa, GLfloat *src, int idx )
 {
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] ))+MAT_ELT_0;
    memcpy(dest, src, 16*sizeof(float));
@@ -2145,7 +1973,7 @@ static void upload_matrix_t( radeonContextPtr rmesa, GLfloat *src, int idx )
 
 static void update_texturematrix( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    GLuint tpc = rmesa->hw.tcl.cmd[TCL_TEXTURE_PROC_CTL];
    GLuint vs = rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXSEL];
    int unit;
@@ -2209,64 +2037,74 @@ static void update_texturematrix( GLcontext *ctx )
    }
 }
 
-
-/**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void
-radeonUpdateDrawBuffer(GLcontext *ctx)
+static GLboolean r100ValidateBuffers(GLcontext *ctx)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   driRenderbuffer *drb;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb;
+   int i, ret;
 
-   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-      /* draw to front */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
-   }
-   else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* draw to back */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
-   else {
-      /* drawing to multiple buffers, or none */
-      return;
+
+   /* depth buffer */
+   rrb = radeon_get_depthbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
-   assert(drb);
-   assert(drb->flippedPitch);
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+      radeonTexObj *t;
 
-   RADEON_STATECHANGE( rmesa, ctx );
+      if (!ctx->Texture.Unit[i]._ReallyEnabled)
+	 continue;
 
-   /* Note: we used the (possibly) page-flipped values */
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-     = ((drb->flippedOffset + rmesa->radeonScreen->fbLocation)
-	& RADEON_COLOROFFSET_MASK);
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= RADEON_COLOR_TILE_ENABLE;
+      t = rmesa->state.texture.unit[i].texobj;
+      if (t->image_override && t->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+      else if (t->mt->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
-}
 
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret)
+       return GL_FALSE;
+   return GL_TRUE;
+}
 
-void radeonValidateState( GLcontext *ctx )
+GLboolean radeonValidateState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint new_state = rmesa->NewGLState;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint new_state = rmesa->radeon.NewGLState;
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-     radeonUpdateDrawBuffer(ctx);
+   if (new_state & _NEW_BUFFERS) {
+     _mesa_update_framebuffer(ctx);
+     /* this updates the DrawBuffer's Width/Height if it's a FBO */
+     _mesa_update_draw_buffer_bounds(ctx);
+     RADEON_STATECHANGE(rmesa, ctx);
    }
 
    if (new_state & _NEW_TEXTURE) {
       radeonUpdateTextureState( ctx );
-      new_state |= rmesa->NewGLState; /* may add TEXTURE_MATRIX */
+      new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */
    }
 
+   /* we need to do a space check here */
+   if (!r100ValidateBuffers(ctx))
+     return GL_FALSE;
+
    /* Need an event driven matrix update?
     */
-   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION)) 
+   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       upload_matrix( rmesa, ctx->_ModelProjectMatrix.m, MODEL_PROJ );
 
    /* Need these for lighting (shouldn't upload otherwise)
@@ -2290,12 +2128,14 @@ void radeonValidateState( GLcontext *ctx )
    /* emit all active clip planes if projection matrix changes.
     */
    if (new_state & (_NEW_PROJECTION)) {
-      if (ctx->Transform.ClipPlanesEnabled) 
+      if (ctx->Transform.ClipPlanesEnabled)
 	 radeonUpdateClipPlanes( ctx );
    }
 
 
-   rmesa->NewGLState = 0;
+   rmesa->radeon.NewGLState = 0;
+
+   return GL_TRUE;
 }
 
 
@@ -2306,7 +2146,7 @@ static void radeonInvalidateState( GLcontext *ctx, GLuint new_state )
    _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
-   RADEON_CONTEXT(ctx)->NewGLState |= new_state;
+   R100_CONTEXT(ctx)->radeon.NewGLState |= new_state;
 }
 
 
@@ -2317,8 +2157,8 @@ static GLboolean check_material( GLcontext *ctx )
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLint i;
 
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT; 
-	i < _TNL_ATTRIB_MAT_BACK_INDEXES; 
+   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT;
+	i < _TNL_ATTRIB_MAT_BACK_INDEXES;
 	i++)
       if (tnl->vb.AttribPtr[i] &&
 	  tnl->vb.AttribPtr[i]->stride)
@@ -2326,20 +2166,21 @@ static GLboolean check_material( GLcontext *ctx )
 
    return GL_FALSE;
 }
-      
+
 
 static void radeonWrapRunPipeline( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean has_material;
 
    if (0)
-      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->NewGLState);
+      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->radeon.NewGLState);
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!radeonValidateState( ctx ))
+	 FALLBACK(rmesa, RADEON_FALLBACK_TEXTURE, GL_TRUE);
 
    has_material = (ctx->Light.Enabled && check_material( ctx ));
 
@@ -2348,7 +2189,7 @@ static void radeonWrapRunPipeline( GLcontext *ctx )
    }
 
    /* Run the pipeline.
-    */ 
+    */
    _tnl_run_pipeline( ctx );
 
    if (has_material) {
@@ -2356,12 +2197,28 @@ static void radeonWrapRunPipeline( GLcontext *ctx )
    }
 }
 
+static void radeonPolygonStipple( GLcontext *ctx, const GLubyte *mask )
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   GLint i;
+
+   radeon_firevertices(&r100->radeon);
+
+   RADEON_STATECHANGE(r100, stp);
+
+   /* Must flip pattern upside down.
+    */
+   for ( i = 31 ; i >= 0; i--) {
+     r100->hw.stp.cmd[3 + i] = ((GLuint *) mask)[i];
+   }
+}
+
 
 /* Initialize the driver's state functions.
  * Many of the ctx->Driver functions might have been initialized to
  * software defaults in the earlier _mesa_init_driver_functions() call.
  */
-void radeonInitStateFuncs( GLcontext *ctx )
+void radeonInitStateFuncs( GLcontext *ctx , GLboolean dri2 )
 {
    ctx->Driver.UpdateState		= radeonInvalidateState;
    ctx->Driver.LightingSpaceChange      = radeonLightingSpaceChange;
@@ -2394,7 +2251,10 @@ void radeonInitStateFuncs( GLcontext *ctx )
    ctx->Driver.LogicOpcode		= radeonLogicOpCode;
    ctx->Driver.PolygonMode		= radeonPolygonMode;
    ctx->Driver.PolygonOffset		= radeonPolygonOffset;
-   ctx->Driver.PolygonStipple		= radeonPolygonStipple;
+   if (dri2)
+      ctx->Driver.PolygonStipple		= radeonPolygonStipple;
+   else
+      ctx->Driver.PolygonStipple		= radeonPolygonStipplePreKMS;
    ctx->Driver.RenderMode		= radeonRenderMode;
    ctx->Driver.Scissor			= radeonScissor;
    ctx->Driver.ShadeModel		= radeonShadeModel;
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.h b/src/mesa/drivers/dri/radeon/radeon_state.h
index 2171879f75..c780cff0cf 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.h
+++ b/src/mesa/drivers/dri/radeon/radeon_state.h
@@ -39,30 +39,25 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "radeon_context.h"
 
-extern void radeonInitState( radeonContextPtr rmesa );
-extern void radeonInitStateFuncs( GLcontext *ctx );
+extern void radeonInitState( r100ContextPtr rmesa );
+extern void radeonInitStateFuncs( GLcontext *ctx , GLboolean dri2);
 
 extern void radeonUpdateMaterial( GLcontext *ctx );
 
-extern void radeonSetCliprects( radeonContextPtr rmesa );
-extern void radeonRecalcScissorRects( radeonContextPtr rmesa );
 extern void radeonUpdateViewportOffset( GLcontext *ctx );
 extern void radeonUpdateWindow( GLcontext *ctx );
 extern void radeonUpdateDrawBuffer( GLcontext *ctx );
-extern void radeonUploadTexMatrix( radeonContextPtr rmesa,
+extern void radeonUploadTexMatrix( r100ContextPtr rmesa,
 				   int unit, GLboolean swapcols );
 
-extern void radeonValidateState( GLcontext *ctx );
-
-extern void radeonPrintDirty( radeonContextPtr rmesa,
-			      const char *msg );
+extern GLboolean radeonValidateState( GLcontext *ctx );
 
 
 extern void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode );
 #define FALLBACK( rmesa, bit, mode ) do {				\
    if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",		\
 		     __FUNCTION__, bit, mode );				\
-   radeonFallback( rmesa->glCtx, bit, mode );				\
+   radeonFallback( rmesa->radeon.glCtx, bit, mode );				\
 } while (0)
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index 57dc380050..f3ad0dd17a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -38,39 +38,141 @@
 #include "swrast_setup/swrast_setup.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
+#include "radeon_queryobj.h"
+
+#include "../r200/r200_reg.h"
 
 #include "xmlpool.h"
 
+/* New (1.3) state mechanism.  3 commands (packet, scalar, vector) in
+ * 1.3 cmdbuffers allow all previous state to be updated as well as
+ * the tcl scalar and vector areas.
+ */
+static struct {
+	int start;
+	int len;
+	const char *name;
+} packet[RADEON_MAX_STATE_PACKETS] = {
+	{RADEON_PP_MISC, 7, "RADEON_PP_MISC"},
+	{RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"},
+	{RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"},
+	{RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"},
+	{RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"},
+	{RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"},
+	{RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"},
+	{RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"},
+	{RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"},
+	{RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"},
+	{RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"},
+	{RADEON_RE_MISC, 1, "RADEON_RE_MISC"},
+	{RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"},
+	{RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"},
+	{RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"},
+	{RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"},
+	{RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"},
+	{RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"},
+	{RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"},
+	{RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"},
+	{RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17,
+		    "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"},
+	{R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"},
+	{R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"},
+	{R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"},
+	{R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"},
+	{R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"},
+	{R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"},
+	{R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"},
+	{R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"},
+	{R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"},
+	{R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"},
+	{R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"},
+	{R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"},
+	{R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"},
+	{R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"},
+	{R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"},
+	{R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"},
+	{R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"},
+	{R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"},
+	{R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"},
+	{R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"},
+	{R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"},
+	{R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"},
+	{R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"},
+	{R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"},
+	{R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"},
+	{R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"},
+	{R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"},
+	{R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"},
+	{R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1,
+	 "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"},
+	{R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"},
+	{R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"},
+	{R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"},
+	{R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"},
+	{R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"},
+	{R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"},
+	{R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"},
+	{R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"},
+	{R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"},
+	{R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"},
+	{R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4,
+		    "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"},
+	{R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"},	/* 61 */
+	{R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */
+	{R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"},
+	{R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"},
+	{R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"},
+	{R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"},
+	{R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"},
+	{R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"},
+	{R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"},
+	{R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"},
+	{R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"},
+	{R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"},
+	{RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"},
+	{RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"},
+	{RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"},
+	{R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"},
+	{R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"},
+	{RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"},
+	{RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"},
+	{RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"},
+	{RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"},
+	{RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"},
+	{RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"},
+	{R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"},
+	{R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"},     /* 85 */
+	{R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"},
+	{R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"},
+	{R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"},
+	{R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"},
+	{R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"},
+	{R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"},
+	{R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"},
+	{R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"},
+	{R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"},
+};
+
 /* =============================================================
  * State initialization
  */
-
-void radeonPrintDirty( radeonContextPtr rmesa, const char *msg )
+static int cmdpkt( r100ContextPtr rmesa, int id ) 
 {
-   struct radeon_state_atom *l;
-
-   fprintf(stderr, msg);
-   fprintf(stderr, ": ");
+   drm_radeon_cmd_header_t h;
 
-   foreach(l, &rmesa->hw.atomlist) {
-      if (l->dirty || rmesa->hw.all_dirty)
-	 fprintf(stderr, "%s, ", l->name);
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     return CP_PACKET0(packet[id].start, packet[id].len - 1);
+   } else {
+     h.i = 0;
+     h.packet.cmd_type = RADEON_CMD_PACKET;
+     h.packet.packet_id = id;
    }
-
-   fprintf(stderr, "\n");
-}
-
-static int cmdpkt( int id ) 
-{
-   drm_radeon_cmd_header_t h;
-   h.i = 0;
-   h.packet.cmd_type = RADEON_CMD_PACKET;
-   h.packet.packet_id = id;
    return h.i;
 }
 
@@ -96,131 +198,523 @@ static int cmdscl( int offset, int stride, int count )
    return h.i;
 }
 
-#define CHECK( NM, FLAG )			\
-static GLboolean check_##NM( GLcontext *ctx )	\
-{						\
-   return FLAG;					\
+#define CHECK( NM, FLAG, ADD )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom )	\
+{							\
+   return FLAG ? atom->cmd_size + (ADD) : 0;			\
 }
 
-#define TCL_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx )		\
+#define TCL_CHECK( NM, FLAG, ADD )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom )	\
 {							\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);	\
-   return !rmesa->TclFallback && (FLAG);		\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);	\
+   return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size + (ADD) : 0;	\
 }
 
 
-CHECK( always, GL_TRUE )
-CHECK( never, GL_FALSE )
-CHECK( tex0, ctx->Texture.Unit[0]._ReallyEnabled )
-CHECK( tex1, ctx->Texture.Unit[1]._ReallyEnabled )
+CHECK( always, GL_TRUE, 0 )
+CHECK( always_add2, GL_TRUE, 2 )
+CHECK( always_add4, GL_TRUE, 4 )
+CHECK( never, GL_FALSE, 0 )
+CHECK( tex0_mm, ctx->Texture.Unit[0]._ReallyEnabled, 3 )
+CHECK( tex1_mm, ctx->Texture.Unit[1]._ReallyEnabled, 3 )
 /* need this for the cubic_map on disabled unit 2 bug, maybe r100 only? */
-CHECK( tex2, ctx->Texture._EnabledUnits )
-CHECK( cube0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT))
-CHECK( cube1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT))
-CHECK( cube2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT))
-CHECK( fog, ctx->Fog.Enabled )
-TCL_CHECK( tcl, GL_TRUE )
-TCL_CHECK( tcl_tex0, ctx->Texture.Unit[0]._ReallyEnabled )
-TCL_CHECK( tcl_tex1, ctx->Texture.Unit[1]._ReallyEnabled )
-TCL_CHECK( tcl_tex2, ctx->Texture.Unit[2]._ReallyEnabled )
-TCL_CHECK( tcl_lighting, ctx->Light.Enabled )
-TCL_CHECK( tcl_eyespace_or_lighting, ctx->_NeedEyeCoords || ctx->Light.Enabled )
-TCL_CHECK( tcl_lit0, ctx->Light.Enabled && ctx->Light.Light[0].Enabled )
-TCL_CHECK( tcl_lit1, ctx->Light.Enabled && ctx->Light.Light[1].Enabled )
-TCL_CHECK( tcl_lit2, ctx->Light.Enabled && ctx->Light.Light[2].Enabled )
-TCL_CHECK( tcl_lit3, ctx->Light.Enabled && ctx->Light.Light[3].Enabled )
-TCL_CHECK( tcl_lit4, ctx->Light.Enabled && ctx->Light.Light[4].Enabled )
-TCL_CHECK( tcl_lit5, ctx->Light.Enabled && ctx->Light.Light[5].Enabled )
-TCL_CHECK( tcl_lit6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled )
-TCL_CHECK( tcl_lit7, ctx->Light.Enabled && ctx->Light.Light[7].Enabled )
-TCL_CHECK( tcl_ucp0, (ctx->Transform.ClipPlanesEnabled & 0x1) )
-TCL_CHECK( tcl_ucp1, (ctx->Transform.ClipPlanesEnabled & 0x2) )
-TCL_CHECK( tcl_ucp2, (ctx->Transform.ClipPlanesEnabled & 0x4) )
-TCL_CHECK( tcl_ucp3, (ctx->Transform.ClipPlanesEnabled & 0x8) )
-TCL_CHECK( tcl_ucp4, (ctx->Transform.ClipPlanesEnabled & 0x10) )
-TCL_CHECK( tcl_ucp5, (ctx->Transform.ClipPlanesEnabled & 0x20) )
-TCL_CHECK( tcl_eyespace_or_fog, ctx->_NeedEyeCoords || ctx->Fog.Enabled ) 
-
-CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT))
-CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT))
-CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT))
+CHECK( tex2_mm, ctx->Texture._EnabledUnits, 3 )
+CHECK( tex0, ctx->Texture.Unit[0]._ReallyEnabled, 2 )
+CHECK( tex1, ctx->Texture.Unit[1]._ReallyEnabled, 2 )
+CHECK( tex2, ctx->Texture._EnabledUnits, 2 )
+CHECK( cube0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE )
+CHECK( cube1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE )
+CHECK( cube2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE )
+CHECK( cube0_mm, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE )
+CHECK( cube1_mm, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE )
+CHECK( cube2_mm, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE )
+CHECK( fog, ctx->Fog.Enabled, 0 )
+CHECK( fog_add4, ctx->Fog.Enabled, 4 )
+TCL_CHECK( tcl, GL_TRUE, 0 )
+TCL_CHECK( tcl_add4, GL_TRUE, 4 )
+TCL_CHECK( tcl_tex0, ctx->Texture.Unit[0]._ReallyEnabled, 0 )
+TCL_CHECK( tcl_tex1, ctx->Texture.Unit[1]._ReallyEnabled, 0 )
+TCL_CHECK( tcl_tex2, ctx->Texture.Unit[2]._ReallyEnabled, 0 )
+TCL_CHECK( tcl_tex0_add4, ctx->Texture.Unit[0]._ReallyEnabled, 4 )
+TCL_CHECK( tcl_tex1_add4, ctx->Texture.Unit[1]._ReallyEnabled, 4 )
+TCL_CHECK( tcl_tex2_add4, ctx->Texture.Unit[2]._ReallyEnabled, 4 )
+TCL_CHECK( tcl_lighting, ctx->Light.Enabled, 0 )
+TCL_CHECK( tcl_lighting_add4, ctx->Light.Enabled, 4 )
+TCL_CHECK( tcl_eyespace_or_lighting, ctx->_NeedEyeCoords || ctx->Light.Enabled, 0 )
+TCL_CHECK( tcl_eyespace_or_lighting_add4, ctx->_NeedEyeCoords || ctx->Light.Enabled, 4 )
+TCL_CHECK( tcl_lit0, ctx->Light.Enabled && ctx->Light.Light[0].Enabled, 0 )
+TCL_CHECK( tcl_lit1, ctx->Light.Enabled && ctx->Light.Light[1].Enabled, 0 )
+TCL_CHECK( tcl_lit2, ctx->Light.Enabled && ctx->Light.Light[2].Enabled, 0 )
+TCL_CHECK( tcl_lit3, ctx->Light.Enabled && ctx->Light.Light[3].Enabled, 0 )
+TCL_CHECK( tcl_lit4, ctx->Light.Enabled && ctx->Light.Light[4].Enabled, 0 )
+TCL_CHECK( tcl_lit5, ctx->Light.Enabled && ctx->Light.Light[5].Enabled, 0 )
+TCL_CHECK( tcl_lit6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled, 0 )
+TCL_CHECK( tcl_lit7, ctx->Light.Enabled && ctx->Light.Light[7].Enabled, 0 )
+TCL_CHECK( tcl_lit0_add6, ctx->Light.Enabled && ctx->Light.Light[0].Enabled, 6 )
+TCL_CHECK( tcl_lit1_add6, ctx->Light.Enabled && ctx->Light.Light[1].Enabled, 6 )
+TCL_CHECK( tcl_lit2_add6, ctx->Light.Enabled && ctx->Light.Light[2].Enabled, 6 )
+TCL_CHECK( tcl_lit3_add6, ctx->Light.Enabled && ctx->Light.Light[3].Enabled, 6 )
+TCL_CHECK( tcl_lit4_add6, ctx->Light.Enabled && ctx->Light.Light[4].Enabled, 6 )
+TCL_CHECK( tcl_lit5_add6, ctx->Light.Enabled && ctx->Light.Light[5].Enabled, 6 )
+TCL_CHECK( tcl_lit6_add6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled, 6 )
+TCL_CHECK( tcl_lit7_add6, ctx->Light.Enabled && ctx->Light.Light[7].Enabled, 6 )
+TCL_CHECK( tcl_ucp0, (ctx->Transform.ClipPlanesEnabled & 0x1), 0 )
+TCL_CHECK( tcl_ucp1, (ctx->Transform.ClipPlanesEnabled & 0x2), 0 )
+TCL_CHECK( tcl_ucp2, (ctx->Transform.ClipPlanesEnabled & 0x4), 0 )
+TCL_CHECK( tcl_ucp3, (ctx->Transform.ClipPlanesEnabled & 0x8), 0 )
+TCL_CHECK( tcl_ucp4, (ctx->Transform.ClipPlanesEnabled & 0x10), 0 )
+TCL_CHECK( tcl_ucp5, (ctx->Transform.ClipPlanesEnabled & 0x20), 0 )
+TCL_CHECK( tcl_ucp0_add4, (ctx->Transform.ClipPlanesEnabled & 0x1), 4 )
+TCL_CHECK( tcl_ucp1_add4, (ctx->Transform.ClipPlanesEnabled & 0x2), 4 )
+TCL_CHECK( tcl_ucp2_add4, (ctx->Transform.ClipPlanesEnabled & 0x4), 4 )
+TCL_CHECK( tcl_ucp3_add4, (ctx->Transform.ClipPlanesEnabled & 0x8), 4 )
+TCL_CHECK( tcl_ucp4_add4, (ctx->Transform.ClipPlanesEnabled & 0x10), 4 )
+TCL_CHECK( tcl_ucp5_add4, (ctx->Transform.ClipPlanesEnabled & 0x20), 4 )
+TCL_CHECK( tcl_eyespace_or_fog, ctx->_NeedEyeCoords || ctx->Fog.Enabled, 0 )
+TCL_CHECK( tcl_eyespace_or_fog_add4, ctx->_NeedEyeCoords || ctx->Fog.Enabled, 4 )
+
+CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT), 0 )
+CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT), 0 )
+CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT), 0 )
+
+#define OUT_VEC(hdr, data) do {			\
+    drm_radeon_cmd_header_t h;					\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1));	\
+    OUT_BATCH_TABLE((data), h.vectors.count);				\
+  } while(0)
+
+#define OUT_SCL(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+
+static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_SCL(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
 
 
+static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
 
-/* Initialize the context's hardware state.
- */
-void radeonInitState( radeonContextPtr rmesa )
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
+
+
+static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom)
 {
-   GLcontext *ctx = rmesa->glCtx;
-   GLuint color_fmt, depth_fmt, i;
-   GLint drawPitch, drawOffset;
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1);
+   OUT_SCL(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1);
+   END_BATCH();
+}
 
-   switch ( rmesa->radeonScreen->cpp ) {
-   case 2:
-      color_fmt = RADEON_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      color_fmt = RADEON_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" );
-      exit( -1 );
+static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   struct radeon_renderbuffer *rrb;
+   uint32_t cbpitch;
+   uint32_t zbpitch, depth_fmt;
+   uint32_t dwords = atom->check(ctx, atom);
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_BATCH_TABLE(atom->cmd, 5);
+
+   rrb = radeon_get_depthbuffer(&r100->radeon);
+   if (!rrb) {
+     OUT_BATCH(0);
+     OUT_BATCH(0);
+   } else {
+     zbpitch = (rrb->pitch / rrb->cpp);
+     if (r100->using_hyperz)
+       zbpitch |= RADEON_DEPTH_HYPERZ;
+
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+     OUT_BATCH(zbpitch);
+     if (rrb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+   }
+     
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(atom->cmd[CTX_CMD_1]);
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+
+   rrb = radeon_get_colorbuffer(&r100->radeon);
+   if (!rrb || !rrb->bo) {
+      OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+      OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]);
+   } else {
+      atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+      if (rrb->cpp == 4)
+         atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+      else
+         atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+
+      OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+      OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
    }
 
-   rmesa->state.color.clear = 0x00000000;
+   OUT_BATCH(atom->cmd[CTX_CMD_2]);
+
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]);
+   } else {
+     cbpitch = (rrb->pitch / rrb->cpp);
+     if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= RADEON_COLOR_TILE_ENABLE;
+     OUT_BATCH(cbpitch);
+   }
+
+   END_BATCH();
+}
+
+static int check_always_ctx( GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t dwords;
+
+   rrb = radeon_get_colorbuffer(&r100->radeon);
+   if (!rrb || !rrb->bo) {
+      return 0;
+   }
+
+   drb = radeon_get_depthbuffer(&r100->radeon);
+
+   dwords = 10;
+   if (drb)
+     dwords += 6;
+   if (rrb)
+     dwords += 8;
+
+   return dwords;
+}
+
+static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t cbpitch = 0;
+   uint32_t zbpitch = 0;
+   uint32_t dwords = atom->check(ctx, atom);
+   uint32_t depth_fmt;
+
+   rrb = radeon_get_colorbuffer(&r100->radeon);
+   if (!rrb || !rrb->bo) {
+      fprintf(stderr, "no rrb\n");
+      return;
+   }
+
+   atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+   if (rrb->cpp == 4)
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
+
+   cbpitch = (rrb->pitch / rrb->cpp);
+   if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+
+   drb = radeon_get_depthbuffer(&r100->radeon);
+   if (drb) {
+     zbpitch = (drb->pitch / drb->cpp);
+     if (drb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+     
+   }
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   /* In the CS case we need to split this up */
+   OUT_BATCH(CP_PACKET0(packet[0].start, 3));
+   OUT_BATCH_TABLE((atom->cmd + 1), 4);
+
+   if (drb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0));
+     OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0));
+     OUT_BATCH(zbpitch);
+   }
+
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0));
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1));
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+   OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+
+   if (rrb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0));
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0));
+     OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+   }
+
+   // if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) {
+   //   OUT_BATCH_TABLE((atom->cmd + 14), 4);
+   // }
+
+   END_BATCH();
+   BEGIN_BATCH_NO_AUTOSTATE(4);
+   OUT_BATCH(CP_PACKET0(RADEON_RE_TOP_LEFT, 0));
+   OUT_BATCH(0);
+   OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
+   if (rrb) {
+       OUT_BATCH(((rrb->base.Width - 1) << RADEON_RE_WIDTH_SHIFT) |
+                 ((rrb->base.Height - 1) << RADEON_RE_HEIGHT_SHIFT));
+   } else {
+       OUT_BATCH(0);
+   }
+   END_BATCH();
+}
+
+static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx, j;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (!(ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT))
+	return;
+
+   if (!t)
+	return;
+
+   if (!t->mt)
+	return;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_BATCH_TABLE(atom->cmd, 3);
+   lvl = &t->mt->levels[0];
+   for (j = 0; j < 5; j++) {
+	OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   }
+   END_BATCH();
+}
+
+static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->check(ctx, atom);
+   int i = atom->idx, j;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   uint32_t base_reg;
+
+   if (!(ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT))
+	return;
+
+   if (!t)
+	return;
+
+   if (!t->mt)
+	return;
+
+   switch(i) {
+	case 1: base_reg = RADEON_PP_CUBIC_OFFSET_T1_0; break;
+	case 2: base_reg = RADEON_PP_CUBIC_OFFSET_T2_0; break;
+	default:
+	case 0: base_reg = RADEON_PP_CUBIC_OFFSET_T0_0; break;
+   };
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_BATCH_TABLE(atom->cmd, 2);
+   lvl = &t->mt->levels[0];
+   for (j = 0; j < 5; j++) {
+	OUT_BATCH(CP_PACKET0(base_reg + (4 * j), 0));
+	OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   }
+   END_BATCH();
+}
+
+static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (t && t->mt && !t->image_override)
+     dwords += 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH_TABLE(atom->cmd, 3);
+   if (t && t->mt && !t->image_override) {
+     if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) {
+   	lvl = &t->mt->levels[0];
+	OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     } else {
+        OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   } else if (!t) {
+     /* workaround for old CS mechanism */
+     OUT_BATCH(r100->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]);
+     //     OUT_BATCH(r100->radeon.radeonScreen);
+   } else {
+     OUT_BATCH(t->override_offset);
+   }
+
+   OUT_BATCH_TABLE((atom->cmd+4), 5);
+   END_BATCH();
+}
+
+static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   int hastexture = 1;
+
+   if (!t)
+	hastexture = 0;
+   else {
+	if (!t->mt && !t->bo)
+		hastexture = 0;
+   }
+   dwords += 1;
+   if (hastexture)
+     dwords += 2;
+   else
+     dwords -= 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH(CP_PACKET0(RADEON_PP_TXFILTER_0 + (24 * i), 1));
+   OUT_BATCH_TABLE((atom->cmd + 1), 2);
+
+   if (hastexture) {
+     OUT_BATCH(CP_PACKET0(RADEON_PP_TXOFFSET_0 + (24 * i), 0));
+     if (t->mt && !t->image_override) {
+        if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) {
+            lvl = &t->mt->levels[0];
+	    OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+        } else {
+           OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+        }
+      } else {
+	if (t->bo)
+            OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+                            RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      }
+   }
+
+   OUT_BATCH(CP_PACKET0(RADEON_PP_TXCBLEND_0 + (i * 24), 1));
+   OUT_BATCH_TABLE((atom->cmd+4), 2);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_BORDER_COLOR_0 + (i * 4), 0));
+   OUT_BATCH((atom->cmd[TEX_PP_BORDER_COLOR]));
+   END_BATCH();
+}
+
+/* Initialize the context's hardware state.
+ */
+void radeonInitState( r100ContextPtr rmesa )
+{
+   GLcontext *ctx = rmesa->radeon.glCtx;
+   GLuint i;
+
+   rmesa->radeon.state.color.clear = 0x00000000;
 
    switch ( ctx->Visual.depthBits ) {
    case 16:
-      rmesa->state.depth.clear = 0x0000ffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff;
-      depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
-      rmesa->state.stencil.clear = 0x00000000;
+      rmesa->radeon.state.depth.clear = 0x0000ffff;
+      rmesa->radeon.state.stencil.clear = 0x00000000;
       break;
    case 24:
-      rmesa->state.depth.clear = 0x00ffffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff;
-      depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
-      rmesa->state.stencil.clear = 0xffff0000;
+      rmesa->radeon.state.depth.clear = 0x00ffffff;
+      rmesa->radeon.state.stencil.clear = 0xffff0000;
       break;
    default:
-      fprintf( stderr, "Error: Unsupported depth %d... exiting\n",
-	       ctx->Visual.depthBits );
-      exit( -1 );
+      break;
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
+   rmesa->radeon.Fallback = 0;
 
-   rmesa->Fallback = 0;
 
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      drawOffset = rmesa->radeonScreen->backOffset;
-      drawPitch  = rmesa->radeonScreen->backPitch;
-   } else {
-      drawOffset = rmesa->radeonScreen->frontOffset;
-      drawPitch  = rmesa->radeonScreen->frontPitch;
-   }
+   rmesa->radeon.hw.max_state_size = 0;
 
-   rmesa->hw.max_state_size = 0;
-
-#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG )				\
+#define ALLOC_STATE_IDX( ATOM, CHK, SZ, NM, FLAG, IDX )		\
    do {								\
       rmesa->hw.ATOM.cmd_size = SZ;				\
-      rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.name = NM;					\
+      rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int)); \
+      rmesa->hw.ATOM.name = NM;						\
       rmesa->hw.ATOM.is_tcl = FLAG;					\
       rmesa->hw.ATOM.check = check_##CHK;				\
-      rmesa->hw.ATOM.dirty = GL_TRUE;				\
-      rmesa->hw.max_state_size += SZ * sizeof(int);		\
+      rmesa->hw.ATOM.dirty = GL_TRUE;					\
+      rmesa->hw.ATOM.idx = IDX;					\
+      rmesa->radeon.hw.max_state_size += SZ * sizeof(int);		\
    } while (0)
-      
-      
+
+#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG )		\
+   ALLOC_STATE_IDX(ATOM, CHK, SZ, NM, FLAG, 0)
+
    /* Allocate state buffers:
     */
-   ALLOC_STATE( ctx, always, CTX_STATE_SIZE, "CTX/context", 0 );
+   ALLOC_STATE( ctx, always_add4, CTX_STATE_SIZE, "CTX/context", 0 );
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     rmesa->hw.ctx.emit = ctx_emit_cs;
+     rmesa->hw.ctx.check = check_always_ctx;
+   } else
+     rmesa->hw.ctx.emit = ctx_emit;
    ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 );
    ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 );
    ALLOC_STATE( vpt, always, VPT_STATE_SIZE, "VPT/viewport", 0 );
@@ -229,82 +723,133 @@ void radeonInitState( radeonContextPtr rmesa )
    ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 );
    ALLOC_STATE( tcl, always, TCL_STATE_SIZE, "TCL/tcl", 1 );
    ALLOC_STATE( mtl, tcl_lighting, MTL_STATE_SIZE, "MTL/material", 1 );
-   ALLOC_STATE( grd, always, GRD_STATE_SIZE, "GRD/guard-band", 1 );
-   ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 );
-   ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 );
-   ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 );
-   ALLOC_STATE( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0 );
-   ALLOC_STATE( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0 );
-   ALLOC_STATE( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0 );
-   if (rmesa->radeonScreen->drmSupportsCubeMapsR100)
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      ALLOC_STATE( grd, always_add2, GRD_STATE_SIZE, "GRD/guard-band", 1 );
+      ALLOC_STATE( fog, fog_add4, FOG_STATE_SIZE, "FOG/fog", 1 );
+      ALLOC_STATE( glt, tcl_lighting_add4, GLT_STATE_SIZE, "GLT/light-global", 1 );
+      ALLOC_STATE( eye, tcl_lighting_add4, EYE_STATE_SIZE, "EYE/eye-vector", 1 );
+      ALLOC_STATE_IDX( tex[0], tex0_mm, TEX_STATE_SIZE, "TEX/tex-0", 0, 0);
+      ALLOC_STATE_IDX( tex[1], tex1_mm, TEX_STATE_SIZE, "TEX/tex-1", 0, 1);
+      ALLOC_STATE_IDX( tex[2], tex2_mm, TEX_STATE_SIZE, "TEX/tex-2", 0, 2);
+      ALLOC_STATE( mat[0], tcl_add4, MAT_STATE_SIZE, "MAT/modelproject", 1 );
+      ALLOC_STATE( mat[1], tcl_eyespace_or_fog_add4, MAT_STATE_SIZE, "MAT/modelview", 1 );
+      ALLOC_STATE( mat[2], tcl_eyespace_or_lighting_add4, MAT_STATE_SIZE, "MAT/it-modelview", 1 );
+      ALLOC_STATE( mat[3], tcl_tex0_add4, MAT_STATE_SIZE, "MAT/texmat0", 1 );
+      ALLOC_STATE( mat[4], tcl_tex1_add4, MAT_STATE_SIZE, "MAT/texmat1", 1 );
+      ALLOC_STATE( mat[5], tcl_tex2_add4, MAT_STATE_SIZE, "MAT/texmat2", 1 );
+      ALLOC_STATE( lit[0], tcl_lit0_add6, LIT_STATE_SIZE, "LIT/light-0", 1 );
+      ALLOC_STATE( lit[1], tcl_lit1_add6, LIT_STATE_SIZE, "LIT/light-1", 1 );
+      ALLOC_STATE( lit[2], tcl_lit2_add6, LIT_STATE_SIZE, "LIT/light-2", 1 );
+      ALLOC_STATE( lit[3], tcl_lit3_add6, LIT_STATE_SIZE, "LIT/light-3", 1 );
+      ALLOC_STATE( lit[4], tcl_lit4_add6, LIT_STATE_SIZE, "LIT/light-4", 1 );
+      ALLOC_STATE( lit[5], tcl_lit5_add6, LIT_STATE_SIZE, "LIT/light-5", 1 );
+      ALLOC_STATE( lit[6], tcl_lit6_add6, LIT_STATE_SIZE, "LIT/light-6", 1 );
+      ALLOC_STATE( lit[7], tcl_lit7_add6, LIT_STATE_SIZE, "LIT/light-7", 1 );
+      ALLOC_STATE( ucp[0], tcl_ucp0_add4, UCP_STATE_SIZE, "UCP/userclip-0", 1 );
+      ALLOC_STATE( ucp[1], tcl_ucp1_add4, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
+      ALLOC_STATE( ucp[2], tcl_ucp2_add4, UCP_STATE_SIZE, "UCP/userclip-2", 1 );
+      ALLOC_STATE( ucp[3], tcl_ucp3_add4, UCP_STATE_SIZE, "UCP/userclip-3", 1 );
+      ALLOC_STATE( ucp[4], tcl_ucp4_add4, UCP_STATE_SIZE, "UCP/userclip-4", 1 );
+      ALLOC_STATE( ucp[5], tcl_ucp5_add4, UCP_STATE_SIZE, "UCP/userclip-5", 1 );
+   } else {
+      ALLOC_STATE( grd, always, GRD_STATE_SIZE, "GRD/guard-band", 1 );
+      ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 );
+      ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 );
+      ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 );
+      ALLOC_STATE_IDX( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0, 0);
+      ALLOC_STATE_IDX( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0, 1);
+      ALLOC_STATE_IDX( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0, 2);
+      ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 );
+      ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 );
+      ALLOC_STATE( mat[2], tcl_eyespace_or_lighting, MAT_STATE_SIZE, "MAT/it-modelview", 1 );
+      ALLOC_STATE( mat[3], tcl_tex0, MAT_STATE_SIZE, "MAT/texmat0", 1 );
+      ALLOC_STATE( mat[4], tcl_tex1, MAT_STATE_SIZE, "MAT/texmat1", 1 );
+      ALLOC_STATE( mat[5], tcl_tex2, MAT_STATE_SIZE, "MAT/texmat2", 1 );
+      ALLOC_STATE( lit[0], tcl_lit0, LIT_STATE_SIZE, "LIT/light-0", 1 );
+      ALLOC_STATE( lit[1], tcl_lit1, LIT_STATE_SIZE, "LIT/light-1", 1 );
+      ALLOC_STATE( lit[2], tcl_lit2, LIT_STATE_SIZE, "LIT/light-2", 1 );
+      ALLOC_STATE( lit[3], tcl_lit3, LIT_STATE_SIZE, "LIT/light-3", 1 );
+      ALLOC_STATE( lit[4], tcl_lit4, LIT_STATE_SIZE, "LIT/light-4", 1 );
+      ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 );
+      ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 );
+      ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 );
+      ALLOC_STATE( ucp[0], tcl_ucp0, UCP_STATE_SIZE, "UCP/userclip-0", 1 );
+      ALLOC_STATE( ucp[1], tcl_ucp1, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
+      ALLOC_STATE( ucp[2], tcl_ucp2, UCP_STATE_SIZE, "UCP/userclip-2", 1 );
+      ALLOC_STATE( ucp[3], tcl_ucp3, UCP_STATE_SIZE, "UCP/userclip-3", 1 );
+      ALLOC_STATE( ucp[4], tcl_ucp4, UCP_STATE_SIZE, "UCP/userclip-4", 1 );
+      ALLOC_STATE( ucp[5], tcl_ucp5, UCP_STATE_SIZE, "UCP/userclip-5", 1 );
+   }
+
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+       ALLOC_STATE( stp, always, STP_STATE_SIZE, "STP/stp", 0 );
+   }
+   
+   for (i = 0; i < 3; i++) {
+      if (rmesa->radeon.radeonScreen->kernel_mm)
+          rmesa->hw.tex[i].emit = tex_emit_cs;
+      else
+          rmesa->hw.tex[i].emit = tex_emit;
+   }
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100)
    {
-      ALLOC_STATE( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
-      ALLOC_STATE( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
-      ALLOC_STATE( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0 );
+      if (rmesa->radeon.radeonScreen->kernel_mm) {
+         ALLOC_STATE_IDX( cube[0], cube0_mm, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 );
+         ALLOC_STATE_IDX( cube[1], cube1_mm, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 );
+         ALLOC_STATE_IDX( cube[2], cube2_mm, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 );
+         for (i = 0; i < 3; i++)
+            rmesa->hw.cube[i].emit = cube_emit_cs;
+      } else {
+         ALLOC_STATE_IDX( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 );
+         ALLOC_STATE_IDX( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 );
+         ALLOC_STATE_IDX( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 );
+         for (i = 0; i < 3; i++)
+            rmesa->hw.cube[i].emit = cube_emit;
+      }
    }
    else
    {
-      ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
-      ALLOC_STATE( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
-      ALLOC_STATE( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0 );
-   }
-   ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 );
-   ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 );
-   ALLOC_STATE( mat[2], tcl_eyespace_or_lighting, MAT_STATE_SIZE, "MAT/it-modelview", 1 );
-   ALLOC_STATE( mat[3], tcl_tex0, MAT_STATE_SIZE, "MAT/texmat0", 1 );
-   ALLOC_STATE( mat[4], tcl_tex1, MAT_STATE_SIZE, "MAT/texmat1", 1 );
-   ALLOC_STATE( mat[5], tcl_tex2, MAT_STATE_SIZE, "MAT/texmat2", 1 );
-   ALLOC_STATE( ucp[0], tcl_ucp0, UCP_STATE_SIZE, "UCP/userclip-0", 1 );
-   ALLOC_STATE( ucp[1], tcl_ucp1, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
-   ALLOC_STATE( ucp[2], tcl_ucp2, UCP_STATE_SIZE, "UCP/userclip-2", 1 );
-   ALLOC_STATE( ucp[3], tcl_ucp3, UCP_STATE_SIZE, "UCP/userclip-3", 1 );
-   ALLOC_STATE( ucp[4], tcl_ucp4, UCP_STATE_SIZE, "UCP/userclip-4", 1 );
-   ALLOC_STATE( ucp[5], tcl_ucp5, UCP_STATE_SIZE, "UCP/userclip-5", 1 );
-   ALLOC_STATE( lit[0], tcl_lit0, LIT_STATE_SIZE, "LIT/light-0", 1 );
-   ALLOC_STATE( lit[1], tcl_lit1, LIT_STATE_SIZE, "LIT/light-1", 1 );
-   ALLOC_STATE( lit[2], tcl_lit2, LIT_STATE_SIZE, "LIT/light-2", 1 );
-   ALLOC_STATE( lit[3], tcl_lit3, LIT_STATE_SIZE, "LIT/light-3", 1 );
-   ALLOC_STATE( lit[4], tcl_lit4, LIT_STATE_SIZE, "LIT/light-4", 1 );
-   ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 );
-   ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 );
-   ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 );
-   ALLOC_STATE( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0 );
-   ALLOC_STATE( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0 );
-   ALLOC_STATE( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0 );
+      ALLOC_STATE_IDX( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 );
+      ALLOC_STATE_IDX( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 );
+      ALLOC_STATE_IDX( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 );
+   }
+   ALLOC_STATE_IDX( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0, 0 );
+   ALLOC_STATE_IDX( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0, 1 );
+   ALLOC_STATE_IDX( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0, 2 );
 
    radeonSetUpAtomList( rmesa );
 
    /* Fill in the packet headers:
     */
-   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC);
-   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL);
-   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH);
-   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN);
-   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH);
-   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK);
-   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE);
-   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL);
-   rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(RADEON_EMIT_SE_CNTL_STATUS);
-   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC);
-   rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_0);
-   rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_0);
-   rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_1);
-   rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_1);
-   rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_2);
-   rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_2);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_0);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T0);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_1);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T1);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_2);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T2);
-   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
-   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT);
+   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC);
+   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL);
+   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH);
+   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN);
+   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH);
+   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK);
+   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE);
+   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL);
+   rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL_STATUS);
+   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC);
+   rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_0);
+   rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_0);
+   rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_1);
+   rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_1);
+   rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_2);
+   rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_2);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T1);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_2);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T2);
+   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR);
+   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT);
    rmesa->hw.mtl.cmd[MTL_CMD_0] = 
-      cmdpkt(RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED);
-   rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_0);
-   rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_1);
-   rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_2);
+      cmdpkt(rmesa, RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED);
+   rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_0);
+   rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_1);
+   rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_2);
    rmesa->hw.grd.cmd[GRD_CMD_0] = 
       cmdscl( RADEON_SS_VERT_GUARD_CLIP_ADJ_ADDR, 1, 4 );
    rmesa->hw.fog.cmd[FOG_CMD_0] = 
@@ -331,6 +876,26 @@ void radeonInitState( radeonContextPtr rmesa )
 	 cmdvec( RADEON_VS_UCP_ADDR + i, 1, 4 );
    }
 
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      rmesa->hw.stp.cmd[STP_CMD_0] = CP_PACKET0(RADEON_RE_STIPPLE_ADDR, 0);
+      rmesa->hw.stp.cmd[STP_DATA_0] = 0;
+      rmesa->hw.stp.cmd[STP_CMD_1] = CP_PACKET0_ONE(RADEON_RE_STIPPLE_DATA, 31);
+
+      rmesa->hw.grd.emit = scl_emit;
+      rmesa->hw.fog.emit = vec_emit;
+      rmesa->hw.glt.emit = vec_emit;
+      rmesa->hw.eye.emit = vec_emit;
+      
+      for (i = 0; i < 6; i++)
+	 rmesa->hw.mat[i].emit = vec_emit;
+
+      for (i = 0; i < 8; i++)
+	 rmesa->hw.lit[i].emit = lit_emit;
+
+      for (i = 0; i < 6; i++)
+	 rmesa->hw.ucp[i].emit = vec_emit;
+   }
+
    rmesa->last_ReallyEnabled = -1;
 
    /* Initial Harware state:
@@ -352,19 +917,7 @@ void radeonInitState( radeonContextPtr rmesa )
 					    RADEON_SRC_BLEND_GL_ONE |
 					    RADEON_DST_BLEND_GL_ZERO );
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] =
-      rmesa->radeonScreen->depthOffset + rmesa->radeonScreen->fbLocation;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = 
-      ((rmesa->radeonScreen->depthPitch &
-	RADEON_DEPTHPITCH_MASK) |
-       RADEON_DEPTH_ENDIAN_NO_SWAP);
-       
-   if (rmesa->using_hyperz)
-       rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] |= RADEON_DEPTH_HYPERZ;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt |
-					       RADEON_Z_TEST_LESS |
+   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (RADEON_Z_TEST_LESS |
 					       RADEON_STENCIL_TEST_ALWAYS |
 					       RADEON_STENCIL_FAIL_KEEP |
 					       RADEON_STENCIL_ZPASS_KEEP |
@@ -374,7 +927,7 @@ void radeonInitState( radeonContextPtr rmesa )
    if (rmesa->using_hyperz) {
        rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_COMPRESSION_ENABLE |
 						   RADEON_Z_DECOMPRESSION_ENABLE;
-      if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
 	 /* works for q3, but slight rendering errors with glxgears ? */
 /*	 rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/
 	 /* need this otherwise get lots of lockups with q3 ??? */
@@ -386,10 +939,9 @@ void radeonInitState( radeonContextPtr rmesa )
 				     RADEON_ANTI_ALIAS_NONE);
 
    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = (RADEON_PLANE_MASK_ENABLE |
-				       color_fmt |
 				       RADEON_ZBLOCK16);
 
-   switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) {
+   switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) {
    case DRI_CONF_DITHER_XERRORDIFFRESET:
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_INIT;
       break;
@@ -397,30 +949,17 @@ void radeonInitState( radeonContextPtr rmesa )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_SCALE_DITHER_ENABLE;
       break;
    }
-   if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) ==
+   if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) ==
 	DRI_CONF_ROUND_ROUND )
-      rmesa->state.color.roundEnable = RADEON_ROUND_ENABLE;
+      rmesa->radeon.state.color.roundEnable = RADEON_ROUND_ENABLE;
    else
-      rmesa->state.color.roundEnable = 0;
-   if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) ==
+      rmesa->radeon.state.color.roundEnable = 0;
+   if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) ==
 	DRI_CONF_COLOR_REDUCTION_DITHER )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_ENABLE;
    else
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset +
-					       rmesa->radeonScreen->fbLocation)
-					      & RADEON_COLOROFFSET_MASK);
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable;
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch &
-					      RADEON_COLORPITCH_MASK) |
-					     RADEON_COLOR_ENDIAN_NO_SWAP);
-
-
-   /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= RADEON_COLOR_TILE_ENABLE;
-   }
 
    rmesa->hw.set.cmd[SET_SE_CNTL] = (RADEON_FFACE_CULL_CCW |
 				     RADEON_BFACE_SOLID |
@@ -444,7 +983,7 @@ void radeonInitState( radeonContextPtr rmesa )
   					    RADEON_VC_NO_SWAP;
 #endif
 
-   if (!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+   if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
      rmesa->hw.set.cmd[SET_SE_CNTL_STATUS] |= RADEON_TCL_BYPASS;
    }
 
@@ -491,8 +1030,8 @@ void radeonInitState( radeonContextPtr rmesa )
 	   (2 << RADEON_TXFORMAT_HEIGHT_SHIFT));
 
       /* Initialize the texture offset to the start of the card texture heap */
-      rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      //      rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] =
+      //	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 
       rmesa->hw.tex[i].cmd[TEX_PP_BORDER_COLOR] = 0;
       rmesa->hw.tex[i].cmd[TEX_PP_TXCBLEND] =  
@@ -513,15 +1052,15 @@ void radeonInitState( radeonContextPtr rmesa )
 
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_0] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_1] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_2] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_3] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_4] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
    }
 
    /* Can only add ST1 at the time of doing some multitex but can keep
@@ -612,6 +1151,14 @@ void radeonInitState( radeonContextPtr rmesa )
    rmesa->hw.eye.cmd[EYE_Y] = 0;
    rmesa->hw.eye.cmd[EYE_Z] = IEEE_ONE;
    rmesa->hw.eye.cmd[EYE_RESCALE_FACTOR] = IEEE_ONE;
-   
-   rmesa->hw.all_dirty = GL_TRUE;
+
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      radeon_init_query_stateobj(&rmesa->radeon, R100_QUERYOBJ_CMDSIZE);
+      rmesa->radeon.query.queryobj.cmd[R100_QUERYOBJ_CMD_0] = CP_PACKET0(RADEON_RB3D_ZPASS_DATA, 0);
+      rmesa->radeon.query.queryobj.cmd[R100_QUERYOBJ_DATA_0] = 0;
+   }
+     
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
+
+   rcommonInitCmdBuf(&rmesa->radeon);
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
index ebea1fecdc..e61f59eaea 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
@@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/enums.h"
 #include "main/imports.h"
 #include "main/macros.h"
+#include "main/simple_list.h"
 
 #include "swrast_setup/swrast_setup.h"
 #include "math/m_translate.h"
@@ -50,10 +51,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_state.h"
 #include "radeon_swtcl.h"
 #include "radeon_tcl.h"
+#include "radeon_debug.h"
 
 
-static void flush_last_swtcl_prim( radeonContextPtr rmesa  );
-
 /* R100: xyzw, c0, c1/fog, stq[0..2]  = 4+1+1+3*3 = 15  right? */
 /* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */
 #define RADEON_MAX_TNL_VERTEX_SIZE (15 * sizeof(GLfloat))	/* for mesa _tnl stage */
@@ -64,18 +64,18 @@ static void flush_last_swtcl_prim( radeonContextPtr rmesa  );
 
 #define EMIT_ATTR( ATTR, STYLE, F0 )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
    fmt_0 |= F0;								\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static GLuint radeon_cp_vc_frmts[3][2] =
@@ -87,7 +87,7 @@ static GLuint radeon_cp_vc_frmts[3][2] =
 
 static void radeonSetVertexFormat( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    DECLARE_RENDERINPUTS(index_bitset);
@@ -106,7 +106,7 @@ static void radeonSetVertexFormat( GLcontext *ctx )
    }
 
    assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-   rmesa->swtcl.vertex_attr_count = 0;
+   rmesa->radeon.swtcl.vertex_attr_count = 0;
 
    /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
     * build up a hardware vertex.
@@ -204,33 +204,52 @@ static void radeonSetVertexFormat( GLcontext *ctx )
       }
    }
 
-   if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) ||
+   if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) ||
 	fmt_0 != rmesa->swtcl.vertex_format) {
       RADEON_NEWPRIM(rmesa);
       rmesa->swtcl.vertex_format = fmt_0;
-      rmesa->swtcl.vertex_size =
+      rmesa->radeon.swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
-			      rmesa->swtcl.vertex_attrs, 
-			      rmesa->swtcl.vertex_attr_count,
+			      rmesa->radeon.swtcl.vertex_attrs, 
+			      rmesa->radeon.swtcl.vertex_attr_count,
 			      NULL, 0 );
-      rmesa->swtcl.vertex_size /= 4;
-      RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
-      if (RADEON_DEBUG & DEBUG_VERTS)
-	 fprintf( stderr, "%s: vertex_size= %d floats\n",
-		  __FUNCTION__, rmesa->swtcl.vertex_size);
+      rmesa->radeon.swtcl.vertex_size /= 4;
+      RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
+      radeon_print(RADEON_SWRENDER, RADEON_VERBOSE,
+	  "%s: vertex_size= %d floats\n",  __FUNCTION__, rmesa->radeon.swtcl.vertex_size);
    }
 }
 
+static void radeon_predict_emit_size( r100ContextPtr rmesa )
+{
+
+    if (!rmesa->radeon.swtcl.emit_prediction) {
+        const int state_size = radeonCountStateEmitSize( &rmesa->radeon );
+        const int scissor_size = 8;
+        const int prims_size = 8;
+        const int vertex_size = 7;
+
+        if (rcommonEnsureCmdBufSpace(&rmesa->radeon,
+                    state_size +
+                    (scissor_size + prims_size + vertex_size),
+                    __FUNCTION__))
+            rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize( &rmesa->radeon );
+        else
+            rmesa->radeon.swtcl.emit_prediction = state_size;
+        rmesa->radeon.swtcl.emit_prediction += scissor_size + prims_size + vertex_size
+            + rmesa->radeon.cmdbuf.cs->cdw;
+    }
+}
 
 static void radeonRenderStart( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+    r100ContextPtr rmesa = R100_CONTEXT( ctx );
 
-   radeonSetVertexFormat( ctx );
-   
-   if (rmesa->dma.flush != 0 && 
-       rmesa->dma.flush != flush_last_swtcl_prim)
-      rmesa->dma.flush( rmesa );
+    radeonSetVertexFormat( ctx );
+
+    if (rmesa->radeon.dma.flush != 0 &&
+            rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim)
+        rmesa->radeon.dma.flush( ctx );
 }
 
 
@@ -241,7 +260,7 @@ static void radeonRenderStart( GLcontext *ctx )
  */
 void radeonChooseVertexState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    TNLcontext *tnl = TNL_CONTEXT(ctx);
 
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
@@ -254,7 +273,7 @@ void radeonChooseVertexState( GLcontext *ctx )
     * rasterization fallback.  As this function will be called again when we
     * leave a rasterization fallback, we can just skip it for now.
     */
-   if (rmesa->Fallback != 0)
+   if (rmesa->radeon.Fallback != 0)
       return;
 
    /* HW perspective divide is a win, but tiny vertex formats are a
@@ -281,80 +300,33 @@ void radeonChooseVertexState( GLcontext *ctx )
    }
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( radeonContextPtr rmesa  )
+void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 {
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   rmesa->dma.flush = NULL;
 
-   if (rmesa->dma.current.buf) {
-      struct radeon_dma_region *current = &rmesa->dma.current;
-      GLuint current_offset = (rmesa->radeonScreen->gart_buffer_offset +
-			       current->buf->buf->idx * RADEON_BUFFER_SIZE + 
-			       current->start);
 
-      assert (!(rmesa->swtcl.hw_primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitVertexAOS( rmesa,
+			rmesa->radeon.swtcl.vertex_size,
+			first_elem(&rmesa->radeon.dma.reserved)->bo,
+			current_offset);
 
-      assert (current->start + 
-	      rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	      current->ptr);
+		      
+   radeonEmitVbufPrim( rmesa,
+		       rmesa->swtcl.vertex_format,
+		       rmesa->radeon.swtcl.hw_primitive,
+		       rmesa->radeon.swtcl.numverts);
+   if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw )
+     WARN_ONCE("Rendering was %d commands larger than predicted size."
+	 " We might overflow  command buffer.\n",
+	 rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction );
 
-      if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-	 radeonEnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ +
-			          rmesa->hw.max_state_size + VBUF_BUFSZ );
 
-	 radeonEmitVertexAOS( rmesa,
-			      rmesa->swtcl.vertex_size,
-			      current_offset);
+   rmesa->radeon.swtcl.emit_prediction = 0;
 
-	 radeonEmitVbufPrim( rmesa,
-			     rmesa->swtcl.vertex_format,
-			     rmesa->swtcl.hw_primitive,
-			     rmesa->swtcl.numverts);
-      }
-
-      rmesa->swtcl.numverts = 0;
-      current->start = current->ptr;
-   }
 }
 
-
-/* Alloc space in the current dma region.
- */
-static INLINE void *
-radeonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
-{
-   GLuint bytes = vsize * nverts;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      radeonRefillCurrentDmaRegion( rmesa );
-
-   if (!rmesa->dma.flush) {
-      rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-      rmesa->dma.flush = flush_last_swtcl_prim;
-   }
-
-   assert( vsize == rmesa->swtcl.vertex_size * 4 );
-   assert( rmesa->dma.flush == flush_last_swtcl_prim );
-   assert (rmesa->dma.current.start + 
-	   rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	   rmesa->dma.current.ptr);
-
-
-   {
-      GLubyte *head = (GLubyte *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->dma.current.ptr += bytes;
-      rmesa->swtcl.numverts += nverts;
-      return head;
-   }
-
-}
-
-
 /*
  * Render unclipped vertex buffers by emitting vertices directly to
  * dma buffers.  Use strip/fan hardware primitives where possible.
@@ -387,22 +359,31 @@ static const GLuint hw_prim[GL_POLYGON+1] = {
 };
 
 static INLINE void
-radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
+radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim )
 {
    RADEON_NEWPRIM( rmesa );
-   rmesa->swtcl.hw_primitive = hw_prim[prim];
-   assert(rmesa->dma.current.ptr == rmesa->dma.current.start);
+   rmesa->radeon.swtcl.hw_primitive = hw_prim[prim];
+   //   assert(rmesa->radeon.dma.current.ptr == rmesa->radeon.dma.current.start);
+}
+
+static void* radeon_alloc_verts( r100ContextPtr rmesa , GLuint nr, GLuint size )
+{
+   void *rv;
+   do {
+     radeon_predict_emit_size( rmesa );
+     rv = rcommonAllocDmaLowVerts( &rmesa->radeon, nr, size );
+   } while (!rv);
+   return rv;
 }
 
-#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
+#define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx)
 #define INIT( prim ) radeonDmaPrimitive( rmesa, prim )
 #define FLUSH()  RADEON_NEWPRIM( rmesa )
-#define GET_CURRENT_VB_MAX_VERTS() \
-  (((int)rmesa->dma.current.end - (int)rmesa->dma.current.ptr) / (rmesa->swtcl.vertex_size*4))
+#define GET_CURRENT_VB_MAX_VERTS()					10\
+//  (((int)rmesa->radeon.dma.current.end - (int)rmesa->radeon.dma.current.ptr) / (rmesa->radeon.swtcl.vertex_size*4))
 #define GET_SUBSEQUENT_VB_MAX_VERTS() \
-  ((RADEON_BUFFER_SIZE) / (rmesa->swtcl.vertex_size*4))
-#define ALLOC_VERTS( nr ) \
-  radeonAllocDmaLowVerts( rmesa, nr, rmesa->swtcl.vertex_size * 4 )
+  ((RADEON_BUFFER_SIZE) / (rmesa->radeon.swtcl.vertex_size*4))
+#define ALLOC_VERTS( nr ) radeon_alloc_verts( rmesa, nr, rmesa->radeon.swtcl.vertex_size * 4 )
 #define EMIT_VERTS( ctx, j, nr, buf ) \
   _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf)
 
@@ -418,16 +399,13 @@ radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
 static GLboolean radeon_run_render( GLcontext *ctx,
 				    struct tnl_pipeline_stage *stage )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    tnl_render_func *tab = TAG(render_tab_verts);
    GLuint i;
 
-   if (rmesa->swtcl.indexed_verts.buf) 
-      RELEASE_ELT_VERTS();
-   	
-   if (rmesa->swtcl.RenderIndex != 0 ||   
+   if (rmesa->radeon.swtcl.RenderIndex != 0 ||   
        !radeon_dma_validate_render( ctx, VB ))
       return GL_TRUE;		
 
@@ -442,8 +420,8 @@ static GLboolean radeon_run_render( GLcontext *ctx,
       if (!length)
 	 continue;
 
-      if (RADEON_DEBUG & DEBUG_PRIMS)
-	 fprintf(stderr, "radeon_render.c: prim %s %d..%d\n", 
+      radeon_print(RADEON_SWRENDER, RADEON_NORMAL,
+	  "radeon_render.c: prim %s %d..%d\n",
 		 _mesa_lookup_enum_by_nr(prim & PRIM_MODE_MASK), 
 		 start, start+length);
 
@@ -496,13 +474,13 @@ static void radeonResetLineStipple( GLcontext *ctx );
 
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
-#define CTX_ARG radeonContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) radeonAllocDmaLowVerts( rmesa, n, (size) * 4 )
+#define CTX_ARG r100ContextPtr rmesa
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) radeon_alloc_verts( rmesa, n, (size) * 4 )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);		\
-   const char *radeonverts = (char *)rmesa->swtcl.verts;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);		\
+   const char *radeonverts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (radeonVertex *)(radeonverts + ((x) * (vertsize) * sizeof(int)))
 #define VERTEX radeonVertex 
 #undef TAG
@@ -560,7 +538,7 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + ((e) * rmesa->swtcl.vertex_size * sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + ((e) * rmesa->radeon.swtcl.vertex_size * sizeof(int)))
 
 #define VERT_SET_RGBA( v, c )  					\
 do {								\
@@ -606,8 +584,8 @@ do {							\
 #undef INIT
 
 #define LOCAL_VARS(n)							\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   GLuint color[n], spec[n];						\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);			\
+   GLuint color[n] = {0}, spec[n] = {0};						\
    GLuint coloroffset = rmesa->swtcl.coloroffset;	\
    GLuint specoffset = rmesa->swtcl.specoffset;			\
    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
@@ -617,7 +595,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) radeonRasterPrimitive( ctx, reduced_hw_prim[x] )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -673,9 +651,9 @@ static void init_rast_tab( void )
 } while (0)
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *radeonverts = (char *)rmesa->swtcl.verts;		\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *radeonverts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -700,17 +678,17 @@ static void init_rast_tab( void )
 void radeonChooseRenderState( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint index = 0;
    GLuint flags = ctx->_TriangleCaps;
 
-   if (!rmesa->TclFallback || rmesa->Fallback) 
+   if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback) 
       return;
 
    if (flags & DD_TRI_LIGHT_TWOSIDE) index |= RADEON_TWOSIDE_BIT;
    if (flags & DD_TRI_UNFILLED)      index |= RADEON_UNFILLED_BIT;
 
-   if (index != rmesa->swtcl.RenderIndex) {
+   if (index != rmesa->radeon.swtcl.RenderIndex) {
       tnl->Driver.Render.Points = rast_tab[index].points;
       tnl->Driver.Render.Line = rast_tab[index].line;
       tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -727,7 +705,7 @@ void radeonChooseRenderState( GLcontext *ctx )
 	 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
       }
 
-      rmesa->swtcl.RenderIndex = index;
+      rmesa->radeon.swtcl.RenderIndex = index;
    }
 }
 
@@ -739,18 +717,18 @@ void radeonChooseRenderState( GLcontext *ctx )
 
 static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   if (rmesa->swtcl.hw_primitive != hwprim) {
+   if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       RADEON_NEWPRIM( rmesa );
-      rmesa->swtcl.hw_primitive = hwprim;
+      rmesa->radeon.swtcl.hw_primitive = hwprim;
    }
 }
 
 static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   rmesa->swtcl.render_primitive = prim;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   rmesa->radeon.swtcl.render_primitive = prim;
    if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) 
       radeonRasterPrimitive( ctx, reduced_hw_prim[prim] );
 }
@@ -761,7 +739,7 @@ static void radeonRenderFinish( GLcontext *ctx )
 
 static void radeonResetLineStipple( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    RADEON_STATECHANGE( rmesa, lin );
 }
 
@@ -795,25 +773,25 @@ static const char *getFallbackString(GLuint bit)
 
 void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->Fallback;
+   GLuint oldfallback = rmesa->radeon.Fallback;
 
    if (mode) {
-      rmesa->Fallback |= bit;
+      rmesa->radeon.Fallback |= bit;
       if (oldfallback == 0) {
-	 RADEON_FIREVERTICES( rmesa );
+	 radeon_firevertices(&rmesa->radeon);
 	 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE );
 	 _swsetup_Wakeup( ctx );
-	 rmesa->swtcl.RenderIndex = ~0;
-         if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+	 rmesa->radeon.swtcl.RenderIndex = ~0;
+         if (RADEON_DEBUG & RADEON_FALLBACKS) {
             fprintf(stderr, "Radeon begin rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
          }
       }
    }
    else {
-      rmesa->Fallback &= ~bit;
+      rmesa->radeon.Fallback &= ~bit;
       if (oldfallback == bit) {
 	 _swrast_flush( ctx );
 	 tnl->Driver.Render.Start = radeonRenderStart;
@@ -826,18 +804,18 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 	 tnl->Driver.Render.ResetLineStipple = radeonResetLineStipple;
 	 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_FALSE );
-	 if (rmesa->TclFallback) {
-	    /* These are already done if rmesa->TclFallback goes to
+	 if (rmesa->radeon.TclFallback) {
+	    /* These are already done if rmesa->radeon.TclFallback goes to
 	     * zero above. But not if it doesn't (RADEON_NO_TCL for
 	     * example?)
 	     */
 	    _tnl_invalidate_vertex_state( ctx, ~0 );
 	    _tnl_invalidate_vertices( ctx, ~0 );
-	    RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
+	    RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset );
 	    radeonChooseVertexState( ctx );
 	    radeonChooseRenderState( ctx );
 	 }
-         if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+         if (RADEON_DEBUG & RADEON_FALLBACKS) {
             fprintf(stderr, "Radeon end rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
          }
@@ -853,13 +831,14 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 void radeonInitSwtcl( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    static int firsttime = 1;
 
    if (firsttime) {
       init_rast_tab();
       firsttime = 0;
    }
+   rmesa->radeon.swtcl.emit_prediction = 0;
 
    tnl->Driver.Render.Start = radeonRenderStart;
    tnl->Driver.Render.Finish = radeonRenderFinish;
@@ -872,18 +851,9 @@ void radeonInitSwtcl( GLcontext *ctx )
    _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
 		       RADEON_MAX_TNL_VERTEX_SIZE);
    
-   rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->swtcl.render_primitive = GL_TRIANGLES;
-   rmesa->swtcl.hw_primitive = 0;
+   rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+   rmesa->radeon.swtcl.hw_primitive = 0;
 }
 
-
-void radeonDestroySwtcl( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (rmesa->swtcl.indexed_verts.buf) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
index e485052ad7..da89158eeb 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
@@ -40,7 +40,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 extern void radeonInitSwtcl( GLcontext *ctx );
-extern void radeonDestroySwtcl( GLcontext *ctx );
 
 extern void radeonChooseRenderState( GLcontext *ctx );
 extern void radeonChooseVertexState( GLcontext *ctx );
@@ -63,5 +62,5 @@ extern void radeon_translate_vertex( GLcontext *ctx,
 
 extern void radeon_print_vertex( GLcontext *ctx, const radeonVertex *v );
 
-
+extern void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c
index 779e9ae5df..b334ea05e5 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c
@@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
+#include "radeon_common.h"
 #include "radeon_context.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
@@ -49,6 +50,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tcl.h"
 #include "radeon_swtcl.h"
 #include "radeon_maos.h"
+#include "radeon_common_context.h"
 
 
 
@@ -104,7 +106,7 @@ static GLboolean discrete_prim[0x10] = {
 };
    
 
-#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
+#define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx)
 #define ELT_TYPE  GLushort
 
 #define ELT_INIT(prim, hw_prim) \
@@ -125,7 +127,7 @@ static GLboolean discrete_prim[0x10] = {
 
 #define RESET_STIPPLE() do {			\
    RADEON_STATECHANGE( rmesa, lin );		\
-   radeonEmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 #define AUTO_STIPPLE( mode )  do {		\
@@ -136,31 +138,26 @@ static GLboolean discrete_prim[0x10] = {
    else						\
       rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &=	\
 	 ~RADEON_LINE_PATTERN_AUTO_RESET;	\
-   radeonEmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);		\
 } while (0)
 
 
 
 #define ALLOC_ELTS(nr)	radeonAllocElts( rmesa, nr )
 
-static GLushort *radeonAllocElts( radeonContextPtr rmesa, GLuint nr ) 
+static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) 
 {
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+      if (rmesa->radeon.dma.flush)
+	 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
 
-   radeonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			   rmesa->hw.max_state_size + ELTS_BUFSZ(nr));
+      radeonEmitAOS( rmesa,
+		     rmesa->radeon.tcl.aos_count, 0 );
 
-   radeonEmitAOS( rmesa,
-		rmesa->tcl.aos_components,
-		rmesa->tcl.nr_aos_components, 0 );
-
-   return radeonAllocEltsOpenEnded( rmesa,
-				    rmesa->tcl.vertex_format, 
-				    rmesa->tcl.hw_primitive, nr );
+      return radeonAllocEltsOpenEnded( rmesa, rmesa->tcl.vertex_format,
+				       rmesa->tcl.hw_primitive, nr );
 }
 
-#define CLOSE_ELTS()  RADEON_NEWPRIM( rmesa )
+#define CLOSE_ELTS() if (0)  RADEON_NEWPRIM( rmesa )
 
 
 
@@ -174,15 +171,11 @@ static void radeonEmitPrim( GLcontext *ctx,
 		       GLuint start, 
 		       GLuint count)	
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    radeonTclPrimitive( ctx, prim, hwprim );
    
-   radeonEnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			    rmesa->hw.max_state_size + VBUF_BUFSZ );
-
    radeonEmitAOS( rmesa,
-		  rmesa->tcl.aos_components,
-		  rmesa->tcl.nr_aos_components,
+		  rmesa->radeon.tcl.aos_count,
 		  start );
    
    /* Why couldn't this packet have taken an offset param?
@@ -197,6 +190,8 @@ static void radeonEmitPrim( GLcontext *ctx,
    radeonEmitPrim( ctx, prim, hwprim, start, count );           \
    (void) rmesa; } while (0)
 
+#define MAX_CONVERSION_SIZE 40
+
 /* Try & join small primitives
  */
 #if 0
@@ -254,7 +249,7 @@ void radeonTclPrimitive( GLcontext *ctx,
 			 GLenum prim,
 			 int hw_prim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint se_cntl;
    GLuint newprim = hw_prim | RADEON_CP_VC_CNTL_TCL_ENABLE;
 
@@ -361,6 +356,73 @@ radeonComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord )
    }
 }
 
+/**
+ * Predict total emit size for next rendering operation so there is no flush in middle of rendering
+ * Prediction has to aim towards the best possible value that is worse than worst case scenario
+ */
+static GLuint radeonEnsureEmitSize( GLcontext * ctx , GLuint inputs )
+{
+  r100ContextPtr rmesa = R100_CONTEXT(ctx);
+  TNLcontext *tnl = TNL_CONTEXT(ctx);
+  struct vertex_buffer *VB = &tnl->vb;
+  GLuint space_required;
+  GLuint state_size;
+  GLuint nr_aos = 1; /* radeonEmitArrays does always emit one */
+  int i;
+  /* list of flags that are allocating aos object */
+  const GLuint flags_to_check[] = {
+    VERT_BIT_NORMAL,
+    VERT_BIT_COLOR0,
+    VERT_BIT_COLOR1,
+    VERT_BIT_FOG
+  };
+  /* predict number of aos to emit */
+  for (i=0; i < sizeof(flags_to_check)/sizeof(flags_to_check[0]); ++i)
+  {
+    if (inputs & flags_to_check[i])
+      ++nr_aos;
+  }
+  for (i = 0; i < ctx->Const.MaxTextureUnits; ++i)
+  {
+    if (inputs & VERT_BIT_TEX(i))
+      ++nr_aos;
+  }
+
+  {
+    /* count the prediction for state size */
+    space_required = 0;
+    state_size = radeonCountStateEmitSize( &rmesa->radeon );
+    /* tcl may be changed in radeonEmitArrays so account for it if not dirty */
+    if (!rmesa->hw.tcl.dirty)
+      state_size += rmesa->hw.tcl.check( rmesa->radeon.glCtx, &rmesa->hw.tcl );
+    /* predict size for elements */
+    for (i = 0; i < VB->PrimitiveCount; ++i)
+    {
+      if (!VB->Primitive[i].count)
+	continue;
+      /* If primitive.count is less than MAX_CONVERSION_SIZE
+	 rendering code may decide convert to elts.
+	 In that case we have to make pessimistic prediction.
+	 and use larger of 2 paths. */
+      const GLuint elts = ELTS_BUFSZ(nr_aos);
+      const GLuint index = INDEX_BUFSZ;
+      const GLuint vbuf = VBUF_BUFSZ;
+      if ( (!VB->Elts && VB->Primitive[i].count >= MAX_CONVERSION_SIZE)
+	  || vbuf > index + elts)
+	space_required += vbuf;
+      else
+	space_required += index + elts;
+      space_required += AOS_BUFSZ(nr_aos);
+    }
+    space_required += SCISSOR_BUFSZ;
+  }
+  /* flush the buffer in case we need more than is left. */
+  if (rcommonEnsureCmdBufSpace(&rmesa->radeon, space_required, __FUNCTION__))
+    return space_required + radeonCountStateEmitSize( &rmesa->radeon );
+  else
+    return space_required + state_size;
+}
+
 /**********************************************************************/
 /*                          Render pipeline stage                     */
 /**********************************************************************/
@@ -371,7 +433,7 @@ radeonComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord )
 static GLboolean radeon_run_tcl_render( GLcontext *ctx,
 					struct tnl_pipeline_stage *stage )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    GLuint inputs = VERT_BIT_POS | VERT_BIT_COLOR0;
@@ -379,7 +441,7 @@ static GLboolean radeon_run_tcl_render( GLcontext *ctx,
 
    /* TODO: separate this from the swtnl pipeline 
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       return GL_TRUE;	/* fallback to software t&l */
 
    if (VB->Count == 0)
@@ -411,6 +473,8 @@ static GLboolean radeon_run_tcl_render( GLcontext *ctx,
    }
 
    radeonReleaseArrays( ctx, ~0 );
+   GLuint emit_end = radeonEnsureEmitSize( ctx, inputs )
+     + rmesa->radeon.cmdbuf.cs->cdw;
    radeonEmitArrays( ctx, inputs );
 
    rmesa->tcl.Elts = VB->Elts;
@@ -430,6 +494,10 @@ static GLboolean radeon_run_tcl_render( GLcontext *ctx,
 	 radeonEmitPrimitive( ctx, start, start+length, prim );
    }
 
+   if (emit_end < rmesa->radeon.cmdbuf.cs->cdw)
+      WARN_ONCE("Rendering was %d commands larger than predicted size."
+	  " We might overflow  command buffer.\n", rmesa->radeon.cmdbuf.cs->cdw - emit_end);
+
    return GL_FALSE;		/* finished the pipe */
 }
 
@@ -461,7 +529,7 @@ const struct tnl_pipeline_stage _radeon_tcl_stage =
 
 static void transition_to_swtnl( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLuint se_cntl;
 
@@ -490,7 +558,7 @@ static void transition_to_swtnl( GLcontext *ctx )
 
 static void transition_to_hwtnl( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
 
@@ -509,17 +577,17 @@ static void transition_to_hwtnl( GLcontext *ctx )
 
    tnl->Driver.NotifyMaterialChange = radeonUpdateMaterial;
 
-   if ( rmesa->dma.flush )			
-      rmesa->dma.flush( rmesa );	
+   if ( rmesa->radeon.dma.flush )			
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	
 
-   rmesa->dma.flush = NULL;
+   rmesa->radeon.dma.flush = NULL;
    rmesa->swtcl.vertex_format = 0;
    
-   if (rmesa->swtcl.indexed_verts.buf) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
+   //   if (rmesa->swtcl.indexed_verts.buf) 
+   //      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
+   //			      __FUNCTION__ );
 
-   if (RADEON_DEBUG & DEBUG_FALLBACKS) 
+   if (RADEON_DEBUG & RADEON_FALLBACKS)
       fprintf(stderr, "Radeon end tcl fallback\n");
 }
 
@@ -550,22 +618,22 @@ static char *getFallbackString(GLuint bit)
 
 void radeonTclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->TclFallback;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint oldfallback = rmesa->radeon.TclFallback;
 
    if (mode) {
-      rmesa->TclFallback |= bit;
+      rmesa->radeon.TclFallback |= bit;
       if (oldfallback == 0) {
-	 if (RADEON_DEBUG & DEBUG_FALLBACKS) 
+	 if (RADEON_DEBUG & RADEON_FALLBACKS)
 	    fprintf(stderr, "Radeon begin tcl fallback %s\n",
 		    getFallbackString( bit ));
 	 transition_to_swtnl( ctx );
       }
    }
    else {
-      rmesa->TclFallback &= ~bit;
+      rmesa->radeon.TclFallback &= ~bit;
       if (oldfallback == bit) {
-	 if (RADEON_DEBUG & DEBUG_FALLBACKS) 
+	 if (RADEON_DEBUG & RADEON_FALLBACKS)
 	    fprintf(stderr, "Radeon end tcl fallback %s\n",
 		    getFallbackString( bit ));
 	 transition_to_hwtnl( ctx );
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c
index f2b6deb9c0..99865fff27 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tex.c
@@ -44,6 +44,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/texobj.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_swtcl.h"
@@ -170,10 +171,13 @@ static void radeonSetTexFilter( radeonTexObjPtr t, GLenum minf, GLenum magf )
 {
    GLuint anisotropy = (t->pp_txfilter & RADEON_MAX_ANISO_MASK);
 
+   /* Force revalidation to account for switches from/to mipmapping. */
+   t->validated = GL_FALSE;
+
    t->pp_txfilter &= ~(RADEON_MIN_FILTER_MASK | RADEON_MAG_FILTER_MASK);
 
    /* r100 chips can't handle mipmaps/aniso for cubemap/volume textures */
-   if ( t->base.tObj->Target == GL_TEXTURE_CUBE_MAP ) {
+   if ( t->base.Target == GL_TEXTURE_CUBE_MAP ) {
       switch ( minf ) {
       case GL_NEAREST:
       case GL_NEAREST_MIPMAP_NEAREST:
@@ -249,437 +253,17 @@ static void radeonSetTexBorderColor( radeonTexObjPtr t, const GLfloat color[4] )
    t->pp_border_color = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
 }
 
-
-/**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static radeonTexObjPtr radeonAllocTexObj( struct gl_texture_object *texObj )
-{
-   radeonTexObjPtr t;
-
-   t = CALLOC_STRUCT( radeon_tex_obj );
-   texObj->DriverData = t;
-   if ( t != NULL ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-	 fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)texObj, (void *)t );
-      }
-
-      /* Initialize non-image-dependent parts of the state:
-       */
-      t->base.tObj = texObj;
-      t->border_fallback = GL_FALSE;
-
-      t->pp_txfilter = RADEON_BORDER_MODE_OGL;
-      t->pp_txformat = (RADEON_TXFORMAT_ENDIAN_NO_SWAP |
-			RADEON_TXFORMAT_PERSPECTIVE_ENABLE);
-
-      make_empty_list( & t->base );
-
-      radeonSetTexWrap( t, texObj->WrapS, texObj->WrapT );
-      radeonSetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
-      radeonSetTexBorderColor( t, texObj->BorderColor );
-   }
-
-   return t;
-}
-
-
-static const struct gl_texture_format *
-radeonChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
-                           GLenum format, GLenum type )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   const GLboolean do32bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32 );
-   const GLboolean force16bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16 );
-   (void) format;
-
-   switch ( internalFormat ) {
-   case 4:
-   case GL_RGBA:
-   case GL_COMPRESSED_RGBA:
-      switch ( type ) {
-      case GL_UNSIGNED_INT_10_10_10_2:
-      case GL_UNSIGNED_INT_2_10_10_10_REV:
-	 return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb4444;
-      }
-
-   case 3:
-   case GL_RGB:
-   case GL_COMPRESSED_RGB:
-      switch ( type ) {
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_5_6_5:
-      case GL_UNSIGNED_SHORT_5_6_5_REV:
-	 return _dri_texformat_rgb565;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-      }
-
-   case GL_RGBA8:
-   case GL_RGB10_A2:
-   case GL_RGBA12:
-   case GL_RGBA16:
-      return !force16bpt ?
-	  _dri_texformat_argb8888 : _dri_texformat_argb4444;
-
-   case GL_RGBA4:
-   case GL_RGBA2:
-      return _dri_texformat_argb4444;
-
-   case GL_RGB5_A1:
-      return _dri_texformat_argb1555;
-
-   case GL_RGB8:
-   case GL_RGB10:
-   case GL_RGB12:
-   case GL_RGB16:
-      return !force16bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-
-   case GL_RGB5:
-   case GL_RGB4:
-   case GL_R3_G3_B2:
-      return _dri_texformat_rgb565;
-
-   case GL_ALPHA:
-   case GL_ALPHA4:
-   case GL_ALPHA8:
-   case GL_ALPHA12:
-   case GL_ALPHA16:
-   case GL_COMPRESSED_ALPHA:
-      return _dri_texformat_a8;
-
-   case 1:
-   case GL_LUMINANCE:
-   case GL_LUMINANCE4:
-   case GL_LUMINANCE8:
-   case GL_LUMINANCE12:
-   case GL_LUMINANCE16:
-   case GL_COMPRESSED_LUMINANCE:
-      return _dri_texformat_l8;
-
-   case 2:
-   case GL_LUMINANCE_ALPHA:
-   case GL_LUMINANCE4_ALPHA4:
-   case GL_LUMINANCE6_ALPHA2:
-   case GL_LUMINANCE8_ALPHA8:
-   case GL_LUMINANCE12_ALPHA4:
-   case GL_LUMINANCE12_ALPHA12:
-   case GL_LUMINANCE16_ALPHA16:
-   case GL_COMPRESSED_LUMINANCE_ALPHA:
-      return _dri_texformat_al88;
-
-   case GL_INTENSITY:
-   case GL_INTENSITY4:
-   case GL_INTENSITY8:
-   case GL_INTENSITY12:
-   case GL_INTENSITY16:
-   case GL_COMPRESSED_INTENSITY:
-      return _dri_texformat_i8;
-
-   case GL_YCBCR_MESA:
-      if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-          type == GL_UNSIGNED_BYTE)
-         return &_mesa_texformat_ycbcr;
-      else
-         return &_mesa_texformat_ycbcr_rev;
-
-   case GL_RGB_S3TC:
-   case GL_RGB4_S3TC:
-   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgb_dxt1;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgba_dxt1;
-
-   case GL_RGBA_S3TC:
-   case GL_RGBA4_S3TC:
-   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-      return &_mesa_texformat_rgba_dxt3;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-      return &_mesa_texformat_rgba_dxt5;
-
-   default:
-      _mesa_problem(ctx, "unexpected texture format in %s", __FUNCTION__);
-      return NULL;
-   }
-
-   return NULL; /* never get here */
-}
-
-
-static void radeonTexImage1D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage1d(ctx, target, level, internalFormat,
-                          width, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void radeonTexSubImage1D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset,
-                                 GLsizei width,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-			     format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void radeonTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage2d(ctx, target, level, internalFormat,
-                          width, height, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void radeonTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-			     height, format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-static void radeonCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLsizei imageSize, const GLvoid *data,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width,
-                                 height, border, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void radeonCompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format,
-                                 GLsizei imageSize, const GLvoid *data,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-                                 height, format, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
 #define SCALED_FLOAT_TO_BYTE( x, scale ) \
 		(((GLuint)((255.0F / scale) * (x))) / 2)
 
 static void radeonTexEnv( GLcontext *ctx, GLenum target,
 			  GLenum pname, const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint unit = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 
-   if ( RADEON_DEBUG & DEBUG_STATE ) {
+   if ( RADEON_DEBUG & RADEON_STATE ) {
       fprintf( stderr, "%s( %s )\n",
 	       __FUNCTION__, _mesa_lookup_enum_by_nr( pname ) );
    }
@@ -706,7 +290,7 @@ static void radeonTexEnv( GLcontext *ctx, GLenum target,
        * functions, one mapping [-1.0,0.0] to [-128,0] and one mapping
        * [0.0,4.0] to [0,127].
        */
-      min = driQueryOptionb (&rmesa->optionCache, "no_neg_lod_bias") ?
+      min = driQueryOptionb (&rmesa->radeon.optionCache, "no_neg_lod_bias") ?
 	  0.0 : -1.0;
       bias = CLAMP( *param, min, 4.0 );
       if ( bias == 0 ) {
@@ -739,12 +323,10 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target,
 				struct gl_texture_object *texObj,
 				GLenum pname, const GLfloat *params )
 {
-   radeonTexObjPtr t = (radeonTexObjPtr) texObj->DriverData;
+   radeonTexObj* t = radeon_tex_obj(texObj);
 
-   if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %s )\n", __FUNCTION__,
+   radeon_print(RADEON_TEXTURE, RADEON_VERBOSE, "%s( %s )\n", __FUNCTION__,
 	       _mesa_lookup_enum_by_nr( pname ) );
-   }
 
    switch ( pname ) {
    case GL_TEXTURE_MIN_FILTER:
@@ -767,57 +349,50 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target,
    case GL_TEXTURE_MAX_LEVEL:
    case GL_TEXTURE_MIN_LOD:
    case GL_TEXTURE_MAX_LOD:
+
       /* This isn't the most efficient solution but there doesn't appear to
        * be a nice alternative.  Since there's no LOD clamping,
        * we just have to rely on loading the right subset of mipmap levels
        * to simulate a clamped LOD.
        */
-      driSwapOutTextureObject( (driTextureObject *) t );
+      if (t->mt) {
+         radeon_miptree_unreference(t->mt);
+	 t->mt = 0;
+	 t->validated = GL_FALSE;
+      }
       break;
 
    default:
       return;
    }
-
-   /* Mark this texobj as dirty (one bit per tex unit)
-    */
-   t->dirty_state = TEX_ALL;
-}
-
-
-static void radeonBindTexture( GLcontext *ctx, GLenum target,
-			       struct gl_texture_object *texObj )
-{
-   if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, (void *)texObj,
-	       ctx->Texture.CurrentUnit );
-   }
-
-   assert( (target != GL_TEXTURE_1D && target != GL_TEXTURE_2D &&
-            target != GL_TEXTURE_RECTANGLE_NV && target != GL_TEXTURE_CUBE_MAP) ||
-           (texObj->DriverData != NULL) );
 }
 
-
 static void radeonDeleteTexture( GLcontext *ctx,
 				 struct gl_texture_object *texObj )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj* t = radeon_tex_obj(texObj);
+   int i;
 
-   if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj,
+   radeon_print(RADEON_TEXTURE, RADEON_NORMAL,
+	 "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj,
 	       _mesa_lookup_enum_by_nr( texObj->Target ) );
-   }
-
-   if ( t != NULL ) {
-      if ( rmesa ) {
-         RADEON_FIREVERTICES( rmesa );
-      }
 
-      driDestroyTextureObject( t );
+   if ( rmesa ) {
+     radeon_firevertices(&rmesa->radeon);
+     for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) {
+       if ( t == rmesa->state.texture.unit[i].texobj ) {
+	 rmesa->state.texture.unit[i].texobj = NULL;
+	 rmesa->hw.tex[i].dirty = GL_FALSE;
+	 rmesa->hw.cube[i].dirty = GL_FALSE;
+       }
+     }
    }
 
+   if (t->mt) {
+      radeon_miptree_unreference(t->mt);
+      t->mt = 0;
+   }
    /* Free mipmap images and the texture object itself */
    _mesa_delete_texture_object(ctx, texObj);
 }
@@ -837,7 +412,7 @@ static void radeonTexGen( GLcontext *ctx,
 			  GLenum pname,
 			  const GLfloat *params )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint unit = ctx->Texture.CurrentUnit;
    rmesa->recheck_texgen[unit] = GL_TRUE;
 }
@@ -851,29 +426,40 @@ static void radeonTexGen( GLcontext *ctx,
 static struct gl_texture_object *
 radeonNewTextureObject( GLcontext *ctx, GLuint name, GLenum target )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_object *obj;
-   obj = _mesa_new_texture_object(ctx, name, target);
-   if (!obj)
-      return NULL;
-   obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
-   radeonAllocTexObj( obj );
-   return obj;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+   _mesa_initialize_texture_object(&t->base, name, target);
+   t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+   t->border_fallback = GL_FALSE;
+
+   t->pp_txfilter = RADEON_BORDER_MODE_OGL;
+   t->pp_txformat = (RADEON_TXFORMAT_ENDIAN_NO_SWAP |
+		     RADEON_TXFORMAT_PERSPECTIVE_ENABLE);
+   
+   radeonSetTexWrap( t, t->base.WrapS, t->base.WrapT );
+   radeonSetTexMaxAnisotropy( t, t->base.MaxAnisotropy );
+   radeonSetTexFilter( t, t->base.MinFilter, t->base.MagFilter );
+   radeonSetTexBorderColor( t, t->base.BorderColor );
+   return &t->base;
 }
 
 
+
 void radeonInitTextureFuncs( struct dd_function_table *functions )
 {
-   functions->ChooseTextureFormat	= radeonChooseTextureFormat;
+   functions->ChooseTextureFormat	= radeonChooseTextureFormat_mesa;
    functions->TexImage1D		= radeonTexImage1D;
    functions->TexImage2D		= radeonTexImage2D;
    functions->TexSubImage1D		= radeonTexSubImage1D;
    functions->TexSubImage2D		= radeonTexSubImage2D;
+   functions->GetTexImage               = radeonGetTexImage;
+   functions->GetCompressedTexImage     = radeonGetCompressedTexImage;
 
    functions->NewTextureObject		= radeonNewTextureObject;
-   functions->BindTexture		= radeonBindTexture;
+   //   functions->BindTexture		= radeonBindTexture;
    functions->DeleteTexture		= radeonDeleteTexture;
-   functions->IsTextureResident		= driIsTextureResident;
 
    functions->TexEnv			= radeonTexEnv;
    functions->TexParameter		= radeonTexParameter;
@@ -882,5 +468,12 @@ void radeonInitTextureFuncs( struct dd_function_table *functions )
    functions->CompressedTexImage2D	= radeonCompressedTexImage2D;
    functions->CompressedTexSubImage2D	= radeonCompressedTexSubImage2D;
 
+   functions->GenerateMipmap = radeonGenerateMipmap;
+
+   functions->NewTextureImage = radeonNewTextureImage;
+   functions->FreeTexImageData = radeonFreeTexImageData;
+   functions->MapTexture = radeonMapTexture;
+   functions->UnmapTexture = radeonUnmapTexture;
+
    driInitTextureFormats();
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.h b/src/mesa/drivers/dri/radeon/radeon_tex.h
index 8000880828..a4aaddc74f 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex.h
+++ b/src/mesa/drivers/dri/radeon/radeon_tex.h
@@ -41,12 +41,16 @@ extern void radeonSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
                                unsigned long long offset, GLint depth,
                                GLuint pitch);
 
+extern void radeonSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv);
+extern void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			       __DRIdrawable *dPriv);
+
 extern void radeonUpdateTextureState( GLcontext *ctx );
 
-extern int radeonUploadTexImages( radeonContextPtr rmesa, radeonTexObjPtr t,
+extern int radeonUploadTexImages( r100ContextPtr rmesa, radeonTexObjPtr t,
 				  GLuint face );
 
-extern void radeonDestroyTexObj( radeonContextPtr rmesa, radeonTexObjPtr t );
+extern void radeonDestroyTexObj( r100ContextPtr rmesa, radeonTexObjPtr t );
 
 extern void radeonInitTextureFuncs( struct dd_function_table *functions );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_texmem.c b/src/mesa/drivers/dri/radeon/radeon_texmem.c
deleted file mode 100644
index 5f7bbe6a4c..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_texmem.c
+++ /dev/null
@@ -1,404 +0,0 @@
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *
- */
-#include <errno.h> 
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/macros.h"
-
-#include "radeon_context.h"
-#include "radeon_ioctl.h"
-#include "radeon_tex.h"
-
-#include <unistd.h>  /* for usleep() */
-
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void
-radeonDestroyTexObj( radeonContextPtr rmesa, radeonTexObjPtr t )
-{
-   if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)t, (void *)t->base.tObj );
-   }
-
-   if ( rmesa != NULL ) {
-      unsigned   i;
-
-
-      for ( i = 0 ; i < rmesa->glCtx->Const.MaxTextureUnits ; i++ ) {
-	 if ( t == rmesa->state.texture.unit[i].texobj ) {
-	    rmesa->state.texture.unit[i].texobj = NULL;
-	 }
-      }
-   }
-}
-
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-
-static void radeonUploadRectSubImage( radeonContextPtr rmesa,
-				      radeonTexObjPtr t, 
-				      struct gl_texture_image *texImage,
-				      GLint x, GLint y, 
-				      GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   int blit_format, dstPitch, done;
-
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = RADEON_GMC_DST_8BPP_CI;
-      break;
-   case 2:
-      blit_format = RADEON_GMC_DST_16BPP;
-      break;
-   case 4:
-      blit_format = RADEON_GMC_DST_32BPP;
-      break;
-   default:
-      fprintf( stderr, "radeonUploadRectSubImage: unknown blit_format (texelbytes=%d)\n", 
-      	       texFormat->TexelBytes);
-      return;
-   }
-
-   t->image[0][0].data = texImage->Data;
-
-   /* Currently don't need to cope with small pitches.
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-   dstPitch = t->pp_txpitch + 32;
-
-   {	/* FIXME: prefer GART-texturing if possible */
-      /* Data not in GART memory, or bad pitch.
-       */
-      for (done = 0; done < height ; ) {
-	 struct radeon_dma_region region;
-	 int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch );
-	 int src_pitch;
-	 char *tex;
-
-         src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-	 tex = (char *)texImage->Data + done * src_pitch;
-
-	 memset(&region, 0, sizeof(region));
-	 radeonAllocDmaRegion( rmesa, &region, lines * dstPitch, 1024 );
-
-	 /* Copy texdata to dma:
-	  */
-	 if (0)
-	    fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n",
-		    __FUNCTION__, src_pitch, dstPitch);
-
-	 if (src_pitch == dstPitch) {
-	    memcpy( region.address + region.start, tex, lines * src_pitch );
-	 } 
-	 else {
-	    char *buf = region.address + region.start;
-	    int i;
-	    for (i = 0 ; i < lines ; i++) {
-	       memcpy( buf, tex, src_pitch );
-	       buf += dstPitch;
-	       tex += src_pitch;
-	    }
-	 }
-
-	 radeonEmitWait( rmesa, RADEON_WAIT_3D );
-
-	 
-
-	 /* Blit to framebuffer
-	  */
-	 radeonEmitBlit( rmesa,
-		       blit_format,
-		       dstPitch, GET_START( &region ),
-		       dstPitch, t->bufAddr,
-		       0, 0,
-		       0, done,
-		       width, lines );
-	 
-	 radeonEmitWait( rmesa, RADEON_WAIT_2D );
-
-	 radeonReleaseDmaRegion( rmesa, &region, __FUNCTION__ );
-	 done += lines;
-      }
-   }
-}
-
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void uploadSubImage( radeonContextPtr rmesa, radeonTexObjPtr t, 
-			    GLint hwlevel,
-			    GLint x, GLint y, GLint width, GLint height,
-			    GLuint face )
-{
-   struct gl_texture_image *texImage = NULL;
-   GLuint offset;
-   GLint imageWidth, imageHeight;
-   GLint ret;
-   drm_radeon_texture_t tex;
-   drm_radeon_tex_image_t tmp;
-   const int level = hwlevel + t->base.firstLevel;
-
-   if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n", 
-	       __FUNCTION__, (void *)t, (void *)t->base.tObj, level, width, height, face );
-   }
-
-   ASSERT(face < 6);
-
-   /* Ensure we have a valid texture to upload */
-   if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) {
-      _mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-      return;
-   }
-
-   texImage = t->base.tObj->Image[face][level];
-
-   if ( !texImage ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level );
-      return;
-   }
-   if ( !texImage->Data ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ );
-      return;
-   }
-
-
-   if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-      assert(level == 0);
-      assert(hwlevel == 0);
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__);
-      radeonUploadRectSubImage( rmesa, t, texImage, x, y, width, height );
-      return;
-   }
-
-   imageWidth = texImage->Width;
-   imageHeight = texImage->Height;
-
-   offset = t->bufAddr + t->base.totalSize * face / 6;
-
-   if ( RADEON_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      GLint imageX = 0;
-      GLint imageY = 0;
-      GLint blitX = t->image[face][hwlevel].x;
-      GLint blitY = t->image[face][hwlevel].y;
-      GLint blitWidth = t->image[face][hwlevel].width;
-      GLint blitHeight = t->image[face][hwlevel].height;
-      fprintf( stderr, "   upload image: %d,%d at %d,%d\n",
-	       imageWidth, imageHeight, imageX, imageY );
-      fprintf( stderr, "   upload  blit: %d,%d at %d,%d\n",
-	       blitWidth, blitHeight, blitX, blitY );
-      fprintf( stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-	       (GLuint)offset, hwlevel, level );
-   }
-
-   t->image[face][hwlevel].data = texImage->Data;
-
-   /* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-    * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-    * We used to use 1, 2 and 4-byte texels and used to use the texture
-    * width to dictate the blit width - but that won't work for compressed
-    * textures. (Brian)
-    * NOTE: can't do that with texture tiling. (sroland)
-    */
-   tex.offset = offset;
-   tex.image = &tmp;
-   /* copy (x,y,width,height,data) */
-   memcpy( &tmp, &t->image[face][hwlevel], sizeof(drm_radeon_tex_image_t) );
-
-   if (texImage->TexFormat->TexelBytes) {
-      /* use multi-byte upload scheme */
-      tex.height = imageHeight;
-      tex.width = imageWidth;
-      tex.format = t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK;
-      tex.pitch = MAX2((texImage->Width * texImage->TexFormat->TexelBytes) / 64, 1);
-      tex.offset += tmp.x & ~1023;
-      tmp.x = tmp.x % 1024;
-      if (t->tile_bits & RADEON_TXO_MICRO_TILE_X2) {
-	 /* need something like "tiled coordinates" ? */
-	 tmp.y = tmp.x / (tex.pitch * 128) * 2;
-	 tmp.x = tmp.x % (tex.pitch * 128) / 2 / texImage->TexFormat->TexelBytes;
-	 tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-      }
-      else {
-	 tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-      }
-      if ((t->tile_bits & RADEON_TXO_MACRO_TILE) &&
-	 (texImage->Width * texImage->TexFormat->TexelBytes >= 256)) {
-	 /* radeon switches off macro tiling for small textures/mipmaps it seems */
-	 tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-      }
-   }
-   else {
-      /* In case of for instance 8x8 texture (2x2 dxt blocks), padding after the first two blocks is
-         needed (only with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-      /* set tex.height to 1/4 since 1 "macropixel" (dxt-block) has 4 real pixels. Needed
-         so the kernel module reads the right amount of data. */
-      tex.format = RADEON_TXFORMAT_I8; /* any 1-byte texel format */
-      tex.pitch = (BLIT_WIDTH_BYTES / 64);
-      tex.height = (imageHeight + 3) / 4;
-      tex.width = (imageWidth + 3) / 4;
-      switch (t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK) {
-      case RADEON_TXFORMAT_DXT1:
-         tex.width *= 8;
-         break;
-      case RADEON_TXFORMAT_DXT23:
-      case RADEON_TXFORMAT_DXT45:
-         tex.width *= 16;
-         break;
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
-   do {
-      ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE,
-                                 &tex, sizeof(drm_radeon_texture_t) );
-   } while ( ret == -EAGAIN );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret );
-      fprintf( stderr, "   offset=0x%08x\n",
-	       offset );
-      fprintf( stderr, "   image width=%d height=%d\n",
-	       imageWidth, imageHeight );
-      fprintf( stderr, "    blit width=%d height=%d data=%p\n",
-	       t->image[face][hwlevel].width, t->image[face][hwlevel].height,
-	       t->image[face][hwlevel].data );
-      exit( 1 );
-   }
-}
-
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- * 
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int radeonUploadTexImages( radeonContextPtr rmesa, radeonTexObjPtr t, GLuint face )
-{
-   int numLevels;
-
-   if ( !t || t->base.totalSize == 0 || t->image_override )
-      return 0;
-
-   if ( RADEON_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-	       (void *)rmesa->glCtx, (void *)t->base.tObj, t->base.totalSize,
-	       t->base.firstLevel, t->base.lastLevel );
-   }
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      radeonFinish( rmesa->glCtx );
-   }
-
-   LOCK_HARDWARE( rmesa );
-
-   if ( t->base.memBlock == NULL ) {
-      int heap;
-
-      heap = driAllocateTexture( rmesa->texture_heaps, rmesa->nr_heaps,
-				 (driTextureObject *) t );
-      if ( heap == -1 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 return -1;
-      }
-
-      /* Set the base offset of the texture image */
-      t->bufAddr = rmesa->radeonScreen->texOffset[heap] 
-	   + t->base.memBlock->ofs;
-      t->pp_txoffset = t->bufAddr;
-
-      if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-	 /* hope it's safe to add that here... */
-	 t->pp_txoffset |= t->tile_bits;
-      }
-
-      /* Mark this texobj as dirty on all units:
-       */
-      t->dirty_state = TEX_ALL;
-   }
-
-
-   /* Let the world know we've used this memory recently.
-    */
-   driUpdateTextureLRU( (driTextureObject *) t );
-   UNLOCK_HARDWARE( rmesa );
-
-
-   /* Upload any images that are new */
-   if (t->base.dirty_images[face]) {
-      int i;
-      for ( i = 0 ; i < numLevels ; i++ ) {
-         if ( (t->base.dirty_images[face] & (1 << (i+t->base.firstLevel))) != 0 ) {
-            uploadSubImage( rmesa, t, i, 0, 0, t->image[face][i].width,
-			    t->image[face][i].height, face );
-         }
-      }
-      t->base.dirty_images[face] = 0;
-   }
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      radeonFinish( rmesa->glCtx );
-   }
-
-   return 0;
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c
index b165205c09..9d252aa74c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texstate.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c
@@ -39,10 +39,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/macros.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "main/texobj.h"
 #include "main/enums.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_swtcl.h"
@@ -75,10 +77,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5) \
 			     && (tx_table[f].format != 0xffffffff) )
 
-static const struct {
+struct tx_table {
    GLuint format, filter;
-}
-tx_table[] =
+};
+
+static const struct tx_table tx_table[] =
 {
    _ALPHA(RGBA8888),
    _ALPHA_REV(RGBA8888),
@@ -111,252 +114,6 @@ tx_table[] =
 #undef _ALPHA
 #undef _INVALID
 
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c pp_txfilter, \c pp_txformat, etc. will be set here
- * too.
- * 
- * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
- */
-static void radeonSetTexImages( radeonContextPtr rmesa,
-				struct gl_texture_object *tObj )
-{
-   radeonTexObjPtr t = (radeonTexObjPtr)tObj->DriverData;
-   const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel];
-   GLint curOffset, blitWidth;
-   GLint i, texelBytes;
-   GLint numLevels;
-   GLint log2Width, log2Height, log2Depth;
-
-   /* Set the hardware texture format
-    */
-   if ( !t->image_override ) {
-      t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK |
-                          RADEON_TXFORMAT_ALPHA_IN_MAP);
-      t->pp_txfilter &= ~RADEON_YUV_TO_RGB;
-
-      if ( VALID_FORMAT( baseImage->TexFormat->MesaFormat ) ) {
-         t->pp_txformat |= tx_table[ baseImage->TexFormat->MesaFormat ].format;
-         t->pp_txfilter |= tx_table[ baseImage->TexFormat->MesaFormat ].filter;
-      }
-      else {
-         _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__);
-         return;
-      }
-   }
-
-   texelBytes = baseImage->TexFormat->TexelBytes;
-
-   /* Compute which mipmap levels we really want to send to the hardware.
-    */
-
-   if (tObj->Target != GL_TEXTURE_CUBE_MAP)
-      driCalculateTextureFirstLastLevel( (driTextureObject *) t );
-   else {
-      /* r100 can't handle mipmaps for cube/3d textures, so don't waste
-         memory for them */
-      t->base.firstLevel = t->base.lastLevel = tObj->BaseLevel;
-   }
-   log2Width  = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-   log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-   log2Depth  = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   assert(numLevels <= RADEON_MAX_TEXTURE_LEVELS);
-
-   /* Calculate mipmap offsets and dimensions for blitting (uploading)
-    * The idea is that we lay out the mipmap levels within a block of
-    * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-    */
-   curOffset = 0;
-   blitWidth = BLIT_WIDTH_BYTES;
-   t->tile_bits = 0;
-
-   /* figure out if this texture is suitable for tiling. */
-   if (texelBytes && (tObj->Target != GL_TEXTURE_RECTANGLE_NV)) {
-      if (rmesa->texmicrotile && (baseImage->Height > 1)) {
-	 /* allow 32 (bytes) x 1 mip (which will use two times the space
-	    the non-tiled version would use) max if base texture is large enough */
-	 if ((numLevels == 1) ||
-	   (((baseImage->Width * texelBytes / baseImage->Height) <= 32) &&
-	       (baseImage->Width * texelBytes > 64)) ||
-	    ((baseImage->Width * texelBytes / baseImage->Height) <= 16)) {
-	    /* R100 has two microtile bits (only the txoffset reg, not the blitter)
-	       weird: X2 + OPT: 32bit correct, 16bit completely hosed
-		      X2: 32bit correct, 16bit correct
-		      OPT: 32bit large mips correct, small mips hosed, 16bit completely hosed */
-	    t->tile_bits |= RADEON_TXO_MICRO_TILE_X2 /*| RADEON_TXO_MICRO_TILE_OPT*/;
-	 }
-      }
-      if ((baseImage->Width * texelBytes >= 256) && (baseImage->Height >= 16)) {
-	 /* R100 disables macro tiling only if mip width is smaller than 256 bytes, and not
-	    in the case if height is smaller than 16 (not 100% sure), as does the r200,
-	    so need to disable macro tiling in that case */
-	 if ((numLevels == 1) || ((baseImage->Width * texelBytes / baseImage->Height) <= 4)) {
-	    t->tile_bits |= RADEON_TXO_MACRO_TILE;
-	 }
-      }
-   }
-
-   for (i = 0; i < numLevels; i++) {
-      const struct gl_texture_image *texImage;
-      GLuint size;
-
-      texImage = tObj->Image[0][i + t->base.firstLevel];
-      if ( !texImage )
-	 break;
-
-      /* find image size in bytes */
-      if (texImage->IsCompressed) {
-      /* need to calculate the size AFTER padding even though the texture is
-         submitted without padding.
-         Only handle pot textures currently - don't know if npot is even possible,
-         size calculation would certainly need (trivial) adjustments.
-         Align (and later pad) to 32byte, not sure what that 64byte blit width is
-         good for? */
-         if ((t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK) == RADEON_TXFORMAT_DXT1) {
-            /* RGB_DXT1/RGBA_DXT1, 8 bytes per block */
-            if ((texImage->Width + 3) < 8) /* width one block */
-               size = texImage->CompressedSize * 4;
-            else if ((texImage->Width + 3) < 16)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-         }
-         else /* DXT3/5, 16 bytes per block */
-            if ((texImage->Width + 3) < 8)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-      }
-      else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-	 size = ((texImage->Width * texelBytes + 63) & ~63) * texImage->Height;
-      }
-      else if (t->tile_bits & RADEON_TXO_MICRO_TILE_X2) {
-	 /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-	    though the actual offset may be different (if texture is less than
-	    32 bytes width) to the untiled case */
-	 int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-	 size = (w * ((texImage->Height + 1) / 2)) * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      else {
-	 int w = (texImage->Width * texelBytes + 31) & ~31;
-	 size = w * texImage->Height * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      assert(size > 0);
-
-      /* Align to 32-byte offset.  It is faster to do this unconditionally
-       * (no branch penalty).
-       */
-
-      curOffset = (curOffset + 0x1f) & ~0x1f;
-
-      if (texelBytes) {
-	 t->image[0][i].x = curOffset; /* fix x and y coords up later together with offset */
-	 t->image[0][i].y = 0;
-	 t->image[0][i].width = MIN2(size / texelBytes, blitWidth);
-	 t->image[0][i].height = (size / texelBytes) / t->image[0][i].width;
-      }
-      else {
-         t->image[0][i].x = curOffset % BLIT_WIDTH_BYTES;
-         t->image[0][i].y = curOffset / BLIT_WIDTH_BYTES;
-         t->image[0][i].width  = MIN2(size, BLIT_WIDTH_BYTES);
-         t->image[0][i].height = size / t->image[0][i].width;     
-      }
-
-#if 0
-      /* for debugging only and only  applicable to non-rectangle targets */
-      assert(size % t->image[0][i].width == 0);
-      assert(t->image[0][i].x == 0
-             || (size < BLIT_WIDTH_BYTES && t->image[0][i].height == 1));
-#endif
-
-      if (0)
-         fprintf(stderr,
-                 "level %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-                 i, texImage->Width, texImage->Height,
-                 t->image[0][i].x, t->image[0][i].y,
-                 t->image[0][i].width, t->image[0][i].height, size, curOffset);
-
-      curOffset += size;
-
-   }
-
-   /* Align the total size of texture memory block.
-    */
-   t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-   /* Setup remaining cube face blits, if needed */
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      const GLuint faceSize = t->base.totalSize;
-      GLuint face;
-      /* reuse face 0 x/y/width/height - just update the offset when uploading */
-      for (face = 1; face < 6; face++) {
-         for (i = 0; i < numLevels; i++) {
-            t->image[face][i].x =  t->image[0][i].x;
-            t->image[face][i].y =  t->image[0][i].y;
-            t->image[face][i].width  = t->image[0][i].width;
-            t->image[face][i].height = t->image[0][i].height;
-         }
-      }
-      t->base.totalSize = 6 * faceSize; /* total texmem needed */
-   }
-
-   /* Hardware state:
-    */
-   t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK;
-   t->pp_txfilter |= (numLevels - 1) << RADEON_MAX_MIP_LEVEL_SHIFT;
-
-   t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK |
-		       RADEON_TXFORMAT_HEIGHT_MASK |
-                       RADEON_TXFORMAT_CUBIC_MAP_ENABLE |
-                       RADEON_TXFORMAT_F5_WIDTH_MASK |
-                       RADEON_TXFORMAT_F5_HEIGHT_MASK);
-   t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) |
-		      (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT));
-
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      assert(log2Width == log2Height);
-      t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) |
-                         (log2Height << RADEON_TXFORMAT_F5_HEIGHT_SHIFT) |
-                         (RADEON_TXFORMAT_CUBIC_MAP_ENABLE));
-      t->pp_cubic_faces = ((log2Width << RADEON_FACE_WIDTH_1_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_1_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_2_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_2_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_3_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_3_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_4_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_4_SHIFT));
-   }
-
-   t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) |
-                   ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16));
-
-   /* Only need to round to nearest 32 for textures, but the blitter
-    * requires 64-byte aligned pitches, and we may/may not need the
-    * blitter.   NPOT only!
-    */
-   if ( !t->image_override ) {
-      if (baseImage->IsCompressed)
-         t->pp_txpitch = (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-      else
-         t->pp_txpitch = ((tObj->Image[0][t->base.firstLevel]->Width * texelBytes) + 63) & ~(63);
-      t->pp_txpitch -= 32;
-   }
-
-   t->dirty_state = TEX_ALL;
-
-   /* FYI: radeonUploadTexImages( rmesa, t ); used to be called here */
-}
-
-
-
 /* ================================================================
  * Texture combine functions
  */
@@ -503,7 +260,7 @@ do {							\
 
 static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
    GLuint color_combine, alpha_combine;
    const GLuint color_combine0 = RADEON_COLOR_ARG_A_ZERO | RADEON_COLOR_ARG_B_ZERO
@@ -520,7 +277,7 @@ static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit )
    assert( (texUnit->_ReallyEnabled == 0)
 	   || (texUnit->_Current != NULL) );
 
-   if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
+   if ( RADEON_DEBUG & RADEON_TEXTURE ) {
       fprintf( stderr, "%s( %p, %d )\n", __FUNCTION__, (void *)ctx, unit );
    }
 
@@ -846,22 +603,21 @@ static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit )
 void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
                         unsigned long long offset, GLint depth, GLuint pitch)
 {
-	radeonContextPtr rmesa = pDRICtx->driverPrivate;
+	r100ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
-	    _mesa_lookup_texture(rmesa->glCtx, texname);
-	radeonTexObjPtr t;
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 
 	if (tObj == NULL)
 		return;
 
-	t = (radeonTexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
-
-	t->pp_txoffset = offset;
+	
+	t->bo = NULL;
+	t->override_offset = offset;
 	t->pp_txpitch = pitch - 32;
 
 	switch (depth) {
@@ -881,6 +637,125 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 	}
 }
 
+void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			 __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r100ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = tx_table[MESA_FORMAT_RGB888].format;
+		else
+			t->pp_txformat = tx_table[MESA_FORMAT_ARGB8888].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_ARGB8888].filter;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = tx_table[MESA_FORMAT_RGB888].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_RGB888].filter;
+		break;
+	case 2:
+		t->pp_txformat = tx_table[MESA_FORMAT_RGB565].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_RGB565].filter;
+		break;
+	}
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
+	t->pp_txpitch = pitch_val;
+        t->pp_txpitch -= 32;
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+
+void radeonSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        radeonSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
+
 #define TEXOBJ_TXFILTER_MASK (RADEON_MAX_MIP_LEVEL_MASK |	\
 			      RADEON_MIN_FILTER_MASK | 		\
 			      RADEON_MAG_FILTER_MASK |		\
@@ -901,12 +776,53 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
                               RADEON_TXFORMAT_NON_POWER2)
 
 
-static void import_tex_obj_state( radeonContextPtr rmesa,
+static void disable_tex_obj_state( r100ContextPtr rmesa, 
+				   int unit )
+{
+   RADEON_STATECHANGE( rmesa, tex[unit] );
+
+   RADEON_STATECHANGE( rmesa, tcl );
+   rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] &= ~(RADEON_ST_BIT(unit) |
+					     RADEON_Q_BIT(unit));
+   
+   if (rmesa->radeon.TclFallback & (RADEON_TCL_FALLBACK_TEXGEN_0<<unit)) {
+     TCL_FALLBACK( rmesa->radeon.glCtx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
+     rmesa->recheck_texgen[unit] = GL_TRUE;
+   }
+
+   if (rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) {
+     /* this seems to be a genuine (r100 only?) hw bug. Need to remove the
+	cubic_map bit on unit 2 when the unit is disabled, otherwise every
+	2nd (2d) mipmap on unit 0 will be broken (may not be needed for other
+	units, better be safe than sorry though).*/
+     RADEON_STATECHANGE( rmesa, tex[unit] );
+     rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] &= ~RADEON_TXFORMAT_CUBIC_MAP_ENABLE;
+   }
+
+   {
+      GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
+      GLuint tmp = rmesa->TexGenEnabled;
+
+      rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(RADEON_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_INPUT_MASK<<inputshift);
+      rmesa->TexGenNeedNormals[unit] = 0;
+      rmesa->TexGenEnabled |= 
+	(RADEON_TEXGEN_INPUT_TEXCOORD_0+unit) << inputshift;
+
+      if (tmp != rmesa->TexGenEnabled) {
+	rmesa->recheck_texgen[unit] = GL_TRUE;
+	rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
+      }
+   }
+}
+
+static void import_tex_obj_state( r100ContextPtr rmesa,
 				  int unit,
 				  radeonTexObjPtr texobj )
 {
 /* do not use RADEON_DB_STATE to avoid stale texture caches */
-   int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
+   uint32_t *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
 
    RADEON_STATECHANGE( rmesa, tex[unit] );
@@ -915,10 +831,9 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
    cmd[TEX_PP_TXFILTER] |= texobj->pp_txfilter & TEXOBJ_TXFILTER_MASK;
    cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
    cmd[TEX_PP_TXFORMAT] |= texobj->pp_txformat & TEXOBJ_TXFORMAT_MASK;
-   cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset;
    cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color;
 
-   if (texobj->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
+   if (texobj->base.Target == GL_TEXTURE_RECTANGLE_NV) {
       GLuint *txr_cmd = RADEON_DB_STATE( txr[unit] );
       txr_cmd[TXR_PP_TEX_SIZE] = texobj->pp_txsize; /* NPOT only! */
       txr_cmd[TXR_PP_TEX_PITCH] = texobj->pp_txpitch; /* NPOT only! */
@@ -928,22 +843,12 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
    else {
       se_coord_fmt &= ~(RADEON_VTX_ST0_NONPARAMETRIC << unit);
 
-      if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) {
-	 int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
-	 GLuint bytesPerFace = texobj->base.totalSize / 6;
-	 ASSERT(texobj->base.totalSize % 6 == 0);
+      if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) {
+	 uint32_t *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
 
 	 RADEON_STATECHANGE( rmesa, cube[unit] );
 	 cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
-	 /* dont know if this setup conforms to OpenGL.. 
-	  * at least it matches the behavior of mesa software renderer
-	  */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_0] = texobj->pp_txoffset; /* right */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_1] = texobj->pp_txoffset + 1 * bytesPerFace; /* left */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_2] = texobj->pp_txoffset + 2 * bytesPerFace; /* top */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_3] = texobj->pp_txoffset + 3 * bytesPerFace; /* bottom */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_4] = texobj->pp_txoffset + 4 * bytesPerFace; /* front */
-	 cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset + 5 * bytesPerFace; /* back */
+	 /* state filled out in the cube_emit */
       }
    }
 
@@ -952,13 +857,11 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
       rmesa->hw.set.cmd[SET_SE_COORDFMT] = se_coord_fmt;
    }
 
-   texobj->dirty_state &= ~(1<<unit);
+   rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
 }
 
 
-
-
-static void set_texgen_matrix( radeonContextPtr rmesa, 
+static void set_texgen_matrix( r100ContextPtr rmesa, 
 			       GLuint unit,
 			       const GLfloat *s_plane,
 			       const GLfloat *t_plane,
@@ -986,14 +889,14 @@ static void set_texgen_matrix( radeonContextPtr rmesa,
    rmesa->TexGenMatrix[unit].m[15] = q_plane[3];
 
    rmesa->TexGenEnabled |= RADEON_TEXMAT_0_ENABLE << unit;
-   rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+   rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
 }
 
 /* Returns GL_FALSE if fallback required.
  */
 static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
    GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
    GLuint tmp = rmesa->TexGenEnabled;
@@ -1030,7 +933,7 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
 	    (texUnit->GenS.Mode != texUnit->GenQ.Mode)) ) {
 	 /* Mixed modes, fallback:
 	  */
-	 if (RADEON_DEBUG & DEBUG_FALLBACKS)
+	 if (RADEON_DEBUG & RADEON_FALLBACKS)
 	    fprintf(stderr, "fallback mixed texgen\n");
 	 return GL_FALSE;
       }
@@ -1038,7 +941,7 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
    }
    else {
    /* some texgen mode not including both S and T bits */
-      if (RADEON_DEBUG & DEBUG_FALLBACKS)
+      if (RADEON_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback mixed texgen/nontexgen\n");
       return GL_FALSE;
    }
@@ -1088,289 +991,195 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
    default:
       /* Unsupported mode, fallback:
        */
-      if (RADEON_DEBUG & DEBUG_FALLBACKS) 
+      if (RADEON_DEBUG & RADEON_FALLBACKS)
 	 fprintf(stderr, "fallback GL_SPHERE_MAP\n");
       return GL_FALSE;
    }
 
    if (tmp != rmesa->TexGenEnabled) {
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
    return GL_TRUE;
 }
 
-
-static void disable_tex( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (RADEON_TEX_0_ENABLE<<unit)) {
-      /* Texture unit disabled */
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= ~(1UL << unit);
-	 rmesa->state.texture.unit[unit].texobj = NULL;
-      }
-
-      RADEON_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= 
-	  ~((RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit);
-
-      RADEON_STATECHANGE( rmesa, tcl );
-      rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] &= ~(RADEON_ST_BIT(unit) |
-						RADEON_Q_BIT(unit));
-
-      if (rmesa->TclFallback & (RADEON_TCL_FALLBACK_TEXGEN_0<<unit)) {
-	 TCL_FALLBACK( ctx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
-	 rmesa->recheck_texgen[unit] = GL_TRUE;
-      }
-
-      if (rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) {
-      /* this seems to be a genuine (r100 only?) hw bug. Need to remove the
-         cubic_map bit on unit 2 when the unit is disabled, otherwise every
-	 2nd (2d) mipmap on unit 0 will be broken (may not be needed for other
-	 units, better be safe than sorry though).*/
-	 RADEON_STATECHANGE( rmesa, tex[unit] );
-	 rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] &= ~RADEON_TXFORMAT_CUBIC_MAP_ENABLE;
-      }
-
-      {
-	 GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
-	 GLuint tmp = rmesa->TexGenEnabled;
-
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_INPUT_MASK<<inputshift);
-	 rmesa->TexGenNeedNormals[unit] = 0;
-	 rmesa->TexGenEnabled |= 
-	     (RADEON_TEXGEN_INPUT_TEXCOORD_0+unit) << inputshift;
-
-	 if (tmp != rmesa->TexGenEnabled) {
-	    rmesa->recheck_texgen[unit] = GL_TRUE;
-	    rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-	 }
-      }
-   }
-}
-
-static GLboolean enable_tex_2d( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & RADEON_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~RADEON_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
-   }
-
-   ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
-
-   if ( t->base.dirty_images[0] ) {
-      RADEON_FIREVERTICES( rmesa );
-      radeonSetTexImages( rmesa, tObj );
-      radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock && !t->image_override ) 
-	return GL_FALSE;
-   }
-
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_cube( GLcontext *ctx, int unit )
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int unit)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-   GLuint face;
+   const struct gl_texture_image *firstImage;
+   GLint log2Width, log2Height, log2Depth, texelBytes;
 
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & RADEON_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~RADEON_TXFORMAT_NON_POWER2;
-      for (face = 0; face < 6; face++)
-         t->base.dirty_images[face] = ~0;
+   if ( t->bo ) {
+	return GL_TRUE;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
+   firstImage = t->base.Image[0][t->mt->firstLevel];   
 
-   if ( t->base.dirty_images[0] || t->base.dirty_images[1] ||
-        t->base.dirty_images[2] || t->base.dirty_images[3] ||
-        t->base.dirty_images[4] || t->base.dirty_images[5] ) {
-      /* flush */
-      RADEON_FIREVERTICES( rmesa );
-      /* layout memory space, once for all faces */
-      radeonSetTexImages( rmesa, tObj );
+   if (firstImage->Border > 0) {
+      fprintf(stderr, "%s: border\n", __FUNCTION__);
+      return GL_FALSE;
    }
 
-   /* upload (per face) */
-   for (face = 0; face < 6; face++) {
-      if (t->base.dirty_images[face]) {
-         radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, face );
+   log2Width  = firstImage->WidthLog2;
+   log2Height = firstImage->HeightLog2;
+   log2Depth  = firstImage->DepthLog2;
+   texelBytes = firstImage->TexFormat->TexelBytes;
+
+   if (!t->image_override) {
+      if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+	const struct tx_table *table = tx_table;
+
+	 t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK |
+			     RADEON_TXFORMAT_ALPHA_IN_MAP);
+	 t->pp_txfilter &= ~RADEON_YUV_TO_RGB;	 
+	 
+	 t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format;
+	 t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter;
+      } else {
+	 _mesa_problem(NULL, "unexpected texture format in %s",
+		       __FUNCTION__);
+	 return GL_FALSE;
       }
    }
-      
-   if ( !t->base.memBlock ) {
-      /* texmem alloc failed, use s/w fallback */
-      return GL_FALSE;
+   
+   t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK;
+   t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << RADEON_MAX_MIP_LEVEL_SHIFT;
+	
+   t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK |
+		       RADEON_TXFORMAT_HEIGHT_MASK |
+		       RADEON_TXFORMAT_CUBIC_MAP_ENABLE |
+		       RADEON_TXFORMAT_F5_WIDTH_MASK |
+		       RADEON_TXFORMAT_F5_HEIGHT_MASK);
+   t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) |
+		      (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT));
+   
+   t->tile_bits = 0;
+   
+   if (t->base.Target == GL_TEXTURE_CUBE_MAP) {
+      ASSERT(log2Width == log2Height);
+      t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) |
+			 (log2Height << RADEON_TXFORMAT_F5_HEIGHT_SHIFT) |
+			 /* don't think we need this bit, if it exists at all - fglrx does not set it */
+			 (RADEON_TXFORMAT_CUBIC_MAP_ENABLE));
+      t->pp_cubic_faces = ((log2Width << RADEON_FACE_WIDTH_1_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_1_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_2_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_2_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_3_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_3_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_4_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_4_SHIFT));
    }
 
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_rect( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
+   t->pp_txsize = (((firstImage->Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((firstImage->Height - 1) << RADEON_TEX_VSIZE_SHIFT));
 
-   if (!(t->pp_txformat & RADEON_TXFORMAT_NON_POWER2)) {
-      t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
+   if ( !t->image_override ) {
+      if (firstImage->IsCompressed)
+         t->pp_txpitch = (firstImage->Width + 63) & ~(63);
+      else
+         t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63);
+      t->pp_txpitch -= 32;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-   if ( t->base.dirty_images[0] ) {
-      RADEON_FIREVERTICES( rmesa );
-      radeonSetTexImages( rmesa, tObj );
-      radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock &&
-           !t->image_override /* && !rmesa->prefer_gart_client_texturing  FIXME */ ) {
-	 fprintf(stderr, "%s: upload failed\n", __FUNCTION__);
-	 return GL_FALSE;
-      }
+   if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
+      t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
    }
 
    return GL_TRUE;
 }
 
-
-static GLboolean update_tex_common( GLcontext *ctx, int unit )
+static GLboolean radeon_validate_texture(GLcontext *ctx, struct gl_texture_object *texObj, int unit)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-   GLenum format;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj *t = radeon_tex_obj(texObj);
+   int ret;
 
-   /* Fallback if there's a texture border */
-   if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 ) {
-      fprintf(stderr, "%s: border\n", __FUNCTION__);
+   if (!radeon_validate_texture_miptree(ctx, texObj))
       return GL_FALSE;
-   }
+
+   ret = setup_hardware_state(rmesa, t, unit);
+   if (ret == GL_FALSE)
+     return GL_FALSE;
+
    /* yuv conversion only works in first unit */
    if (unit != 0 && (t->pp_txfilter & RADEON_YUV_TO_RGB))
       return GL_FALSE;
 
-   /* Update state if this is a different texture object to last
-    * time.
-    */
-   if ( rmesa->state.texture.unit[unit].texobj != t ) {
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= 
-	     ~(1UL << unit);
-      }
-
-      rmesa->state.texture.unit[unit].texobj = t;
-      t->base.bound |= (1UL << unit);
-      t->dirty_state |= 1<<unit;
-      driUpdateTextureLRU( (driTextureObject *) t ); /* XXX: should be locked! */
-   }
+   RADEON_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= 
+     (RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit;
 
+   RADEON_STATECHANGE( rmesa, tcl );
+   rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_ST_BIT(unit);
 
-   /* Newly enabled?
-    */
-   if ( !(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (RADEON_TEX_0_ENABLE<<unit))) {
-      RADEON_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= 
-	  (RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit;
+   rmesa->recheck_texgen[unit] = GL_TRUE;
 
-      RADEON_STATECHANGE( rmesa, tcl );
-
-      rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_ST_BIT(unit);
-
-      rmesa->recheck_texgen[unit] = GL_TRUE;
-   }
-
-   if (t->dirty_state & (1<<unit)) {
-      import_tex_obj_state( rmesa, unit, t );
-      /* may need to update texture matrix (for texrect adjustments) */
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-   }
+   import_tex_obj_state( rmesa, unit, t );
 
    if (rmesa->recheck_texgen[unit]) {
       GLboolean fallback = !radeon_validate_texgen( ctx, unit );
       TCL_FALLBACK( ctx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), fallback);
       rmesa->recheck_texgen[unit] = 0;
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
-   format = tObj->Image[0][tObj->BaseLevel]->_BaseFormat;
-   if ( rmesa->state.texture.unit[unit].format != format ||
-	rmesa->state.texture.unit[unit].envMode != texUnit->EnvMode ) {
-      rmesa->state.texture.unit[unit].format = format;
-      rmesa->state.texture.unit[unit].envMode = texUnit->EnvMode;
-      if ( ! radeonUpdateTextureEnv( ctx, unit ) ) {
-	 return GL_FALSE;
-      }
+   if ( ! radeonUpdateTextureEnv( ctx, unit ) ) {
+     return GL_FALSE;
    }
-
    FALLBACK( rmesa, RADEON_FALLBACK_BORDER_MODE, t->border_fallback );
+
+   t->validated = GL_TRUE;
    return !t->border_fallback;
 }
 
-
-
 static GLboolean radeonUpdateTextureUnit( GLcontext *ctx, int unit )
 {
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   if ( texUnit->_ReallyEnabled & (TEXTURE_RECT_BIT) ) {
-      return (enable_tex_rect( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( texUnit->_ReallyEnabled & (TEXTURE_1D_BIT | TEXTURE_2D_BIT) ) {
-      return (enable_tex_2d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( texUnit->_ReallyEnabled & (TEXTURE_CUBE_BIT) ) {
-      return (enable_tex_cube( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
+   if (ctx->Texture.Unit[unit]._ReallyEnabled & TEXTURE_3D_BIT) {
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_FALSE;
    }
-   else if ( texUnit->_ReallyEnabled ) {
-      return GL_FALSE;
+
+   if (!ctx->Texture.Unit[unit]._ReallyEnabled) {
+     /* disable the unit */
+     disable_tex_obj_state(rmesa, unit);
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_TRUE;
    }
-   else {
-      disable_tex( ctx, unit );
-      return GL_TRUE;
+
+   if (!radeon_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) {
+    _mesa_warning(ctx,
+		  "failed to validate texture for unit %d.\n",
+		  unit);
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_FALSE;
    }
+   rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+   return GL_TRUE;
 }
 
 void radeonUpdateTextureState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean ok;
 
+   /* set the ctx all textures off */
+   RADEON_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~((RADEON_TEX_ENABLE_MASK) | (RADEON_TEX_BLEND_ENABLE_MASK));
+
    ok = (radeonUpdateTextureUnit( ctx, 0 ) &&
 	 radeonUpdateTextureUnit( ctx, 1 ) &&
 	 radeonUpdateTextureUnit( ctx, 2 ));
 
    FALLBACK( rmesa, RADEON_FALLBACK_TEXTURE, !ok );
 
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       radeonChooseVertexState( ctx );
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c
new file mode 100644
index 0000000000..fad3d1ceda
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.c
@@ -0,0 +1,1052 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ * Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+ *
+ * The Weather Channel (TM) funded Tungsten Graphics to develop the
+ * initial release of the Radeon 8500 driver under the XFree86 license.
+ * This notice must be preserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/convolve.h"
+#include "main/mipmap.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/texgetimage.h"
+
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+
+#include "radeon_common.h"
+
+#include "radeon_mipmap_tree.h"
+
+
+static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride,
+	GLuint numrows, GLuint rowsize)
+{
+	assert(rowsize <= dststride);
+	assert(rowsize <= srcstride);
+
+	if (rowsize == srcstride && rowsize == dststride) {
+		memcpy(dst, src, numrows*rowsize);
+	} else {
+		GLuint i;
+		for(i = 0; i < numrows; ++i) {
+			memcpy(dst, src, rowsize);
+			dst += dststride;
+			src += srcstride;
+		}
+	}
+}
+
+/* textures */
+/**
+ * Allocate an empty texture image object.
+ */
+struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx)
+{
+	return CALLOC(sizeof(radeon_texture_image));
+}
+
+/**
+ * Free memory associated with this texture image.
+ */
+void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage)
+{
+	radeon_texture_image* image = get_radeon_texture_image(timage);
+
+	if (image->mt) {
+		radeon_miptree_unreference(image->mt);
+		image->mt = 0;
+		assert(!image->base.Data);
+	} else {
+		_mesa_free_texture_image_data(ctx, timage);
+	}
+	if (image->bo) {
+		radeon_bo_unref(image->bo);
+		image->bo = NULL;
+	}
+	if (timage->Data) {
+		_mesa_free_texmemory(timage->Data);
+		timage->Data = NULL;
+	}
+}
+
+/* Set Data pointer and additional data for mapped texture image */
+static void teximage_set_map_data(radeon_texture_image *image)
+{
+	radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+
+	image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset;
+	image->base.RowStride = lvl->rowstride / image->mt->bpp;
+}
+
+
+/**
+ * Map a single texture image for glTexImage and friends.
+ */
+void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable)
+{
+	if (image->mt) {
+		assert(!image->base.Data);
+
+		radeon_bo_map(image->mt->bo, write_enable);
+		teximage_set_map_data(image);
+	}
+}
+
+
+void radeon_teximage_unmap(radeon_texture_image *image)
+{
+	if (image->mt) {
+		assert(image->base.Data);
+
+		image->base.Data = 0;
+		radeon_bo_unmap(image->mt->bo);
+	}
+}
+
+static void map_override(GLcontext *ctx, radeonTexObj *t)
+{
+	radeon_texture_image *img = get_radeon_texture_image(t->base.Image[0][0]);
+
+	radeon_bo_map(t->bo, GL_FALSE);
+
+	img->base.Data = t->bo->ptr;
+	_mesa_set_fetch_functions(&img->base, 2);
+}
+
+static void unmap_override(GLcontext *ctx, radeonTexObj *t)
+{
+	radeon_texture_image *img = get_radeon_texture_image(t->base.Image[0][0]);
+
+	radeon_bo_unmap(t->bo);
+
+	img->base.Data = NULL;
+}
+
+/**
+ * Map a validated texture for reading during software rendering.
+ */
+void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	int face, level;
+
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+	  return;
+
+	/* for r100 3D sw fallbacks don't have mt */
+	if (t->image_override && t->bo)
+		map_override(ctx, t);
+
+	if (!t->mt)
+		return;
+
+	radeon_bo_map(t->mt->bo, GL_FALSE);
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level)
+			teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level]));
+	}
+}
+
+void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	int face, level;
+
+	if (t->image_override && t->bo)
+		unmap_override(ctx, t);
+	/* for r100 3D sw fallbacks don't have mt */
+	if (!t->mt)
+	  return;
+
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level)
+			texObj->Image[face][level]->Data = 0;
+	}
+	radeon_bo_unmap(t->mt->bo);
+}
+
+GLuint radeon_face_for_target(GLenum target)
+{
+	switch (target) {
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+		return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+	default:
+		return 0;
+	}
+}
+
+/**
+ * Wraps Mesa's implementation to ensure that the base level image is mapped.
+ *
+ * This relies on internal details of _mesa_generate_mipmap, in particular
+ * the fact that the memory for recreated texture images is always freed.
+ */
+static void radeon_generate_mipmap(GLcontext *ctx, GLenum target,
+				   struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	GLuint nr_faces = (t->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+	int i, face;
+
+
+	_mesa_generate_mipmap(ctx, target, texObj);
+
+	for (face = 0; face < nr_faces; face++) {
+		for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+			radeon_texture_image *image;
+
+			image = get_radeon_texture_image(texObj->Image[face][i]);
+
+			if (image == NULL)
+				break;
+
+			image->mtlevel = i;
+			image->mtface = face;
+
+			radeon_miptree_unreference(image->mt);
+			image->mt = NULL;
+		}
+	}
+	
+}
+
+void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj)
+{
+	GLuint face = radeon_face_for_target(target);
+	radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[face][texObj->BaseLevel]);
+
+	radeon_teximage_map(baseimage, GL_FALSE);
+	radeon_generate_mipmap(ctx, target, texObj);
+	radeon_teximage_unmap(baseimage);
+}
+
+
+/* try to find a format which will only need a memcopy */
+static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPtr rmesa,
+								 GLenum srcFormat,
+								 GLenum srcType, GLboolean fbo)
+{
+	const GLuint ui = 1;
+	const GLubyte littleEndian = *((const GLubyte *)&ui);
+
+	/* r100 can only do this */
+	if (IS_R100_CLASS(rmesa->radeonScreen) || fbo)
+	  return _dri_texformat_argb8888;
+
+	if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
+	    (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
+	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
+	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
+		return &_mesa_texformat_rgba8888;
+	} else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
+		   (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
+		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
+		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
+		return &_mesa_texformat_rgba8888_rev;
+	} else if (IS_R200_CLASS(rmesa->radeonScreen)) {
+		return _dri_texformat_argb8888;
+	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
+					    srcType == GL_UNSIGNED_INT_8_8_8_8)) {
+		return &_mesa_texformat_argb8888_rev;
+	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) ||
+					    srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) {
+		return &_mesa_texformat_argb8888;
+	} else
+		return _dri_texformat_argb8888;
+}
+
+const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type)
+{
+	return radeonChooseTextureFormat(ctx, internalFormat, format,
+					 type, 0);
+}
+
+const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type, GLboolean fbo)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	const GLboolean do32bpt =
+	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32);
+	const GLboolean force16bpt =
+	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16);
+	(void)format;
+
+#if 0
+	fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n",
+		_mesa_lookup_enum_by_nr(internalFormat), internalFormat,
+		_mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
+	fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt);
+#endif
+
+	switch (internalFormat) {
+	case 4:
+	case GL_RGBA:
+	case GL_COMPRESSED_RGBA:
+		switch (type) {
+		case GL_UNSIGNED_INT_10_10_10_2:
+		case GL_UNSIGNED_INT_2_10_10_10_REV:
+			return do32bpt ? _dri_texformat_argb8888 :
+			    _dri_texformat_argb1555;
+		case GL_UNSIGNED_SHORT_4_4_4_4:
+		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+			return _dri_texformat_argb4444;
+		case GL_UNSIGNED_SHORT_5_5_5_1:
+		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+			return _dri_texformat_argb1555;
+		default:
+			return do32bpt ? radeonChoose8888TexFormat(rmesa, format, type, fbo) :
+			    _dri_texformat_argb4444;
+		}
+
+	case 3:
+	case GL_RGB:
+	case GL_COMPRESSED_RGB:
+		switch (type) {
+		case GL_UNSIGNED_SHORT_4_4_4_4:
+		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+			return _dri_texformat_argb4444;
+		case GL_UNSIGNED_SHORT_5_5_5_1:
+		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+			return _dri_texformat_argb1555;
+		case GL_UNSIGNED_SHORT_5_6_5:
+		case GL_UNSIGNED_SHORT_5_6_5_REV:
+			return _dri_texformat_rgb565;
+		default:
+			return do32bpt ? _dri_texformat_argb8888 :
+			    _dri_texformat_rgb565;
+		}
+
+	case GL_RGBA8:
+	case GL_RGB10_A2:
+	case GL_RGBA12:
+	case GL_RGBA16:
+		return !force16bpt ?
+			radeonChoose8888TexFormat(rmesa, format, type, fbo) :
+			_dri_texformat_argb4444;
+
+	case GL_RGBA4:
+	case GL_RGBA2:
+		return _dri_texformat_argb4444;
+
+	case GL_RGB5_A1:
+		return _dri_texformat_argb1555;
+
+	case GL_RGB8:
+	case GL_RGB10:
+	case GL_RGB12:
+	case GL_RGB16:
+		return !force16bpt ? _dri_texformat_argb8888 :
+		    _dri_texformat_rgb565;
+
+	case GL_RGB5:
+	case GL_RGB4:
+	case GL_R3_G3_B2:
+		return _dri_texformat_rgb565;
+
+	case GL_ALPHA:
+	case GL_ALPHA4:
+	case GL_ALPHA8:
+	case GL_ALPHA12:
+	case GL_ALPHA16:
+	case GL_COMPRESSED_ALPHA:
+		/* r200: can't use a8 format since interpreting hw I8 as a8 would result
+		   in wrong rgb values (same as alpha value instead of 0). */
+		if (IS_R200_CLASS(rmesa->radeonScreen))
+			return _dri_texformat_al88;
+		else
+			return _dri_texformat_a8;
+	case 1:
+	case GL_LUMINANCE:
+	case GL_LUMINANCE4:
+	case GL_LUMINANCE8:
+	case GL_LUMINANCE12:
+	case GL_LUMINANCE16:
+	case GL_COMPRESSED_LUMINANCE:
+		return _dri_texformat_l8;
+
+	case 2:
+	case GL_LUMINANCE_ALPHA:
+	case GL_LUMINANCE4_ALPHA4:
+	case GL_LUMINANCE6_ALPHA2:
+	case GL_LUMINANCE8_ALPHA8:
+	case GL_LUMINANCE12_ALPHA4:
+	case GL_LUMINANCE12_ALPHA12:
+	case GL_LUMINANCE16_ALPHA16:
+	case GL_COMPRESSED_LUMINANCE_ALPHA:
+		return _dri_texformat_al88;
+
+	case GL_INTENSITY:
+	case GL_INTENSITY4:
+	case GL_INTENSITY8:
+	case GL_INTENSITY12:
+	case GL_INTENSITY16:
+	case GL_COMPRESSED_INTENSITY:
+		return _dri_texformat_i8;
+
+	case GL_YCBCR_MESA:
+		if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
+		    type == GL_UNSIGNED_BYTE)
+			return &_mesa_texformat_ycbcr;
+		else
+			return &_mesa_texformat_ycbcr_rev;
+
+	case GL_RGB_S3TC:
+	case GL_RGB4_S3TC:
+	case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+		return &_mesa_texformat_rgb_dxt1;
+
+	case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+		return &_mesa_texformat_rgba_dxt1;
+
+	case GL_RGBA_S3TC:
+	case GL_RGBA4_S3TC:
+	case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+		return &_mesa_texformat_rgba_dxt3;
+
+	case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+		return &_mesa_texformat_rgba_dxt5;
+
+	case GL_ALPHA16F_ARB:
+		return &_mesa_texformat_alpha_float16;
+	case GL_ALPHA32F_ARB:
+		return &_mesa_texformat_alpha_float32;
+	case GL_LUMINANCE16F_ARB:
+		return &_mesa_texformat_luminance_float16;
+	case GL_LUMINANCE32F_ARB:
+		return &_mesa_texformat_luminance_float32;
+	case GL_LUMINANCE_ALPHA16F_ARB:
+		return &_mesa_texformat_luminance_alpha_float16;
+	case GL_LUMINANCE_ALPHA32F_ARB:
+		return &_mesa_texformat_luminance_alpha_float32;
+	case GL_INTENSITY16F_ARB:
+		return &_mesa_texformat_intensity_float16;
+	case GL_INTENSITY32F_ARB:
+		return &_mesa_texformat_intensity_float32;
+	case GL_RGB16F_ARB:
+		return &_mesa_texformat_rgba_float16;
+	case GL_RGB32F_ARB:
+		return &_mesa_texformat_rgba_float32;
+	case GL_RGBA16F_ARB:
+		return &_mesa_texformat_rgba_float16;
+	case GL_RGBA32F_ARB:
+		return &_mesa_texformat_rgba_float32;
+
+	case GL_DEPTH_COMPONENT:
+	case GL_DEPTH_COMPONENT16:
+	case GL_DEPTH_COMPONENT24:
+	case GL_DEPTH_COMPONENT32:
+	case GL_DEPTH_STENCIL_EXT:
+	case GL_DEPTH24_STENCIL8_EXT:
+		return &_mesa_texformat_s8_z24;
+
+	/* EXT_texture_sRGB */
+	case GL_SRGB:
+	case GL_SRGB8:
+	case GL_SRGB_ALPHA:
+	case GL_SRGB8_ALPHA8:
+	case GL_COMPRESSED_SRGB:
+	case GL_COMPRESSED_SRGB_ALPHA:
+		return &_mesa_texformat_srgba8;
+
+	case GL_SLUMINANCE:
+	case GL_SLUMINANCE8:
+	case GL_COMPRESSED_SLUMINANCE:
+		return &_mesa_texformat_sl8;
+
+	case GL_SLUMINANCE_ALPHA:
+	case GL_SLUMINANCE8_ALPHA8:
+	case GL_COMPRESSED_SLUMINANCE_ALPHA:
+		return &_mesa_texformat_sla8;
+
+	default:
+		_mesa_problem(ctx,
+			      "unexpected internalFormat 0x%x in %s",
+			      (int)internalFormat, __func__);
+		return NULL;
+	}
+
+	return NULL;		/* never get here */
+}
+
+/**
+ * All glTexImage calls go through this function.
+ */
+static void radeon_teximage(
+	GLcontext *ctx, int dims,
+	GLenum target, GLint level,
+	GLint internalFormat,
+	GLint width, GLint height, GLint depth,
+	GLsizei imageSize,
+	GLenum format, GLenum type, const GLvoid * pixels,
+	const struct gl_pixelstore_attrib *packing,
+	struct gl_texture_object *texObj,
+	struct gl_texture_image *texImage,
+	int compressed)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	radeon_texture_image* image = get_radeon_texture_image(texImage);
+	GLuint dstRowStride;
+	GLint postConvWidth = width;
+	GLint postConvHeight = height;
+	GLuint texelBytes;
+	GLuint face = radeon_face_for_target(target);
+
+	radeon_firevertices(rmesa);
+
+	t->validated = GL_FALSE;
+
+	if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
+	       _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth,
+						  &postConvHeight);
+	}
+
+	/* Choose and fill in the texture format for this image */
+	texImage->TexFormat = radeonChooseTextureFormat(ctx, internalFormat, format, type, 0);
+	_mesa_set_fetch_functions(texImage, dims);
+
+	if (texImage->TexFormat->TexelBytes == 0) {
+		texelBytes = 0;
+		texImage->IsCompressed = GL_TRUE;
+		texImage->CompressedSize =
+			ctx->Driver.CompressedTextureSize(ctx, texImage->Width,
+					   texImage->Height, texImage->Depth,
+					   texImage->TexFormat->MesaFormat);
+	} else {
+		texImage->IsCompressed = GL_FALSE;
+		texImage->CompressedSize = 0;
+
+		texelBytes = texImage->TexFormat->TexelBytes;
+		/* Minimum pitch of 32 bytes */
+		if (postConvWidth * texelBytes < 32) {
+		  postConvWidth = 32 / texelBytes;
+		  texImage->RowStride = postConvWidth;
+		}
+		if (!image->mt) {      
+			assert(texImage->RowStride == postConvWidth);
+		}
+	}
+
+	/* Allocate memory for image */
+	radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */
+
+	if (t->mt &&
+	    t->mt->firstLevel == level &&
+	    t->mt->lastLevel == level &&
+	    t->mt->target != GL_TEXTURE_CUBE_MAP_ARB &&
+	    !radeon_miptree_matches_image(t->mt, texImage, face, level)) {
+	  radeon_miptree_unreference(t->mt);
+	  t->mt = NULL;
+	}
+
+	if (!t->mt)
+		radeon_try_alloc_miptree(rmesa, t, image, face, level);
+	if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) {
+		radeon_mipmap_level *lvl;
+		image->mt = t->mt;
+		image->mtlevel = level - t->mt->firstLevel;
+		image->mtface = face;
+		radeon_miptree_reference(t->mt);
+		lvl = &image->mt->levels[image->mtlevel];
+		dstRowStride = lvl->rowstride;
+	} else {
+		int size;
+		if (texImage->IsCompressed) {
+			size = texImage->CompressedSize;
+		} else {
+			size = texImage->Width * texImage->Height * texImage->Depth * texImage->TexFormat->TexelBytes;
+		}
+		texImage->Data = _mesa_alloc_texmemory(size);
+	}
+
+	/* Upload texture image; note that the spec allows pixels to be NULL */
+	if (compressed) {
+		pixels = _mesa_validate_pbo_compressed_teximage(
+			ctx, imageSize, pixels, packing, "glCompressedTexImage");
+	} else {
+		pixels = _mesa_validate_pbo_teximage(
+			ctx, dims, width, height, depth,
+			format, type, pixels, packing, "glTexImage");
+	}
+
+	if (pixels) {
+		radeon_teximage_map(image, GL_TRUE);
+		if (compressed) {
+			if (image->mt) {
+				uint32_t srcRowStride, bytesPerRow, rows;
+				srcRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+				bytesPerRow = srcRowStride;
+				rows = (height + 3) / 4;
+				copy_rows(texImage->Data, image->mt->levels[level].rowstride,
+					  pixels, srcRowStride, rows, bytesPerRow);
+			} else {
+				memcpy(texImage->Data, pixels, imageSize);
+			}
+		} else {
+			GLuint dstRowStride;
+			GLuint *dstImageOffsets;
+
+			if (image->mt) {
+				radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+				dstRowStride = lvl->rowstride;
+			} else {
+				dstRowStride = texImage->Width * texImage->TexFormat->TexelBytes;
+			}
+
+			if (dims == 3) {
+				int i;
+
+				dstImageOffsets = _mesa_malloc(depth * sizeof(GLuint)) ;
+				if (!dstImageOffsets)
+					_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+
+				for (i = 0; i < depth; ++i) {
+					dstImageOffsets[i] = dstRowStride/texImage->TexFormat->TexelBytes * height * i;
+				}
+			} else {
+				dstImageOffsets = texImage->ImageOffsets;
+			}
+
+			if (!texImage->TexFormat->StoreImage(ctx, dims,
+						texImage->_BaseFormat,
+						texImage->TexFormat,
+						texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */
+						dstRowStride,
+						dstImageOffsets,
+						width, height, depth,
+						format, type, pixels, packing))
+				_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+
+			if (dims == 3)
+				_mesa_free(dstImageOffsets);
+		}
+
+		/* SGIS_generate_mipmap */
+		if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+			radeon_generate_mipmap(ctx, target, texObj);
+		}
+	}
+
+	_mesa_unmap_teximage_pbo(ctx, packing);
+
+	if (pixels)
+	  radeon_teximage_unmap(image);
+
+
+}
+
+void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 1, target, level, internalFormat, width, 1, 1,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+			   GLint internalFormat,
+			   GLint width, GLint height, GLint border,
+			   GLenum format, GLenum type, const GLvoid * pixels,
+			   const struct gl_pixelstore_attrib *packing,
+			   struct gl_texture_object *texObj,
+			   struct gl_texture_image *texImage)
+
+{
+	radeon_teximage(ctx, 2, target, level, internalFormat, width, height, 1,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target,
+				     GLint level, GLint internalFormat,
+				     GLint width, GLint height, GLint border,
+				     GLsizei imageSize, const GLvoid * data,
+				     struct gl_texture_object *texObj,
+				     struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 2, target, level, internalFormat, width, height, 1,
+		imageSize, 0, 0, data, &ctx->Unpack, texObj, texImage, 1);
+}
+
+void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint depth,
+		      GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 3, target, level, internalFormat, width, height, depth,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+/**
+ * Update a subregion of the given texture image.
+ */
+static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int level,
+		GLint xoffset, GLint yoffset, GLint zoffset,
+		GLsizei width, GLsizei height, GLsizei depth,
+		GLsizei imageSize,
+		GLenum format, GLenum type,
+		const GLvoid * pixels,
+		const struct gl_pixelstore_attrib *packing,
+		struct gl_texture_object *texObj,
+		struct gl_texture_image *texImage,
+		int compressed)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	radeon_texture_image* image = get_radeon_texture_image(texImage);
+
+	radeon_firevertices(rmesa);
+
+	t->validated = GL_FALSE;
+	if (compressed) {
+		pixels = _mesa_validate_pbo_compressed_teximage(
+			ctx, imageSize, pixels, packing, "glCompressedTexImage");
+	} else {
+		pixels = _mesa_validate_pbo_teximage(ctx, dims,
+			width, height, depth, format, type, pixels, packing, "glTexSubImage1D");
+	}
+
+	if (pixels) {
+		GLint dstRowStride;
+		radeon_teximage_map(image, GL_TRUE);
+
+		if (image->mt) {
+			radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+			dstRowStride = lvl->rowstride;
+		} else {
+			dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes;
+		}
+
+		if (compressed) {
+			uint32_t srcRowStride, bytesPerRow, rows;
+			GLubyte *img_start;
+			if (!image->mt) {
+				dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, texImage->Width);
+				img_start = _mesa_compressed_image_address(xoffset, yoffset, 0,
+									   texImage->TexFormat->MesaFormat,
+									   texImage->Width, texImage->Data);
+			}
+			else {
+				uint32_t blocks_x = dstRowStride / (image->mt->bpp * 4);
+				img_start = texImage->Data + image->mt->bpp * 4 * (blocks_x * (yoffset / 4) + xoffset / 4);
+			}
+			srcRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+			bytesPerRow = srcRowStride;
+			rows = (height + 3) / 4;
+
+			copy_rows(img_start, dstRowStride,  pixels, srcRowStride, rows,  bytesPerRow);
+			
+		} else {
+			if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+							     texImage->TexFormat, texImage->Data,
+							     xoffset, yoffset, zoffset,
+							     dstRowStride,
+							     texImage->ImageOffsets,
+							     width, height, depth,
+							     format, type, pixels, packing))
+				_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+		}
+
+		/* GL_SGIS_generate_mipmap */
+		if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+			radeon_generate_mipmap(ctx, target, texObj);
+		}
+	}
+
+	radeon_teximage_unmap(image);
+
+	_mesa_unmap_teximage_pbo(ctx, packing);
+
+
+}
+
+void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset,
+			 GLsizei width,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 1, target, level, xoffset, 0, 0, width, 1, 1, 0,
+		format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset,
+			 GLsizei width, GLsizei height,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 2, target, level, xoffset, yoffset, 0, width, height, 1,
+			   0, format, type, pixels, packing, texObj, texImage,
+			   0);
+}
+
+void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target,
+				   GLint level, GLint xoffset,
+				   GLint yoffset, GLsizei width,
+				   GLsizei height, GLenum format,
+				   GLsizei imageSize, const GLvoid * data,
+				   struct gl_texture_object *texObj,
+				   struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 2, target, level, xoffset, yoffset, 0, width, height, 1,
+		imageSize, format, 0, data, &ctx->Unpack, texObj, texImage, 1);
+}
+
+
+void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset, GLint zoffset,
+			 GLsizei width, GLsizei height, GLsizei depth,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 3, target, level, xoffset, yoffset, zoffset, width, height, depth, 0,
+		format, type, pixels, packing, texObj, texImage, 0);
+}
+
+
+
+/**
+ * Ensure that the given image is stored in the given miptree from now on.
+ */
+static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level)
+{
+	radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel];
+	unsigned char *dest;
+
+	assert(image->mt != mt);
+	assert(dstlvl->width == image->base.Width);
+	assert(dstlvl->height == image->base.Height);
+	assert(dstlvl->depth == image->base.Depth);
+
+
+	radeon_bo_map(mt->bo, GL_TRUE);
+	dest = mt->bo->ptr + dstlvl->faces[face].offset;
+
+	if (image->mt) {
+		/* Format etc. should match, so we really just need a memcpy().
+		 * In fact, that memcpy() could be done by the hardware in many
+		 * cases, provided that we have a proper memory manager.
+		 */
+		radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel-image->mt->firstLevel];
+
+		assert(srclvl->size == dstlvl->size);
+		assert(srclvl->rowstride == dstlvl->rowstride);
+
+		radeon_bo_map(image->mt->bo, GL_FALSE);
+
+		memcpy(dest,
+			image->mt->bo->ptr + srclvl->faces[face].offset,
+			dstlvl->size);
+		radeon_bo_unmap(image->mt->bo);
+
+		radeon_miptree_unreference(image->mt);
+	} else {
+		uint32_t srcrowstride;
+		uint32_t height;
+		/* need to confirm this value is correct */
+		if (mt->compressed) {
+			height = (image->base.Height + 3) / 4;
+			srcrowstride = _mesa_compressed_row_stride(image->base.TexFormat->MesaFormat, image->base.Width);
+		} else {
+			height = image->base.Height * image->base.Depth;
+			srcrowstride = image->base.Width * image->base.TexFormat->TexelBytes;
+		}
+
+//		if (mt->tilebits)
+//			WARN_ONCE("%s: tiling not supported yet", __FUNCTION__);
+
+		copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride,
+			  height, srcrowstride);
+
+		_mesa_free_texmemory(image->base.Data);
+		image->base.Data = 0;
+	}
+
+	radeon_bo_unmap(mt->bo);
+
+	image->mt = mt;
+	image->mtface = face;
+	image->mtlevel = level;
+	radeon_miptree_reference(image->mt);
+}
+
+int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj *t = radeon_tex_obj(texObj);
+	radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]);
+	int face, level;
+
+	if (t->validated || t->image_override)
+		return GL_TRUE;
+
+	if (RADEON_DEBUG & RADEON_TEXTURE)
+		fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj);
+
+	if (baseimage->base.Border > 0)
+		return GL_FALSE;
+
+	/* Ensure a matching miptree exists.
+	 *
+	 * Differing mipmap trees can result when the app uses TexImage to
+	 * change texture dimensions.
+	 *
+	 * Prefer to use base image's miptree if it
+	 * exists, since that most likely contains more valid data (remember
+	 * that the base level is usually significantly larger than the rest
+	 * of the miptree, so cubemaps are the only possible exception).
+	 */
+	if (baseimage->mt &&
+	    baseimage->mt != t->mt &&
+	    radeon_miptree_matches_texture(baseimage->mt, &t->base)) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = baseimage->mt;
+		radeon_miptree_reference(t->mt);
+	} else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
+	}
+
+	if (!t->mt) {
+		if (RADEON_DEBUG & RADEON_TEXTURE)
+			fprintf(stderr, " Allocate new miptree\n");
+		radeon_try_alloc_miptree(rmesa, t, baseimage, 0, texObj->BaseLevel);
+		if (!t->mt) {
+			_mesa_problem(ctx, "radeon_validate_texture failed to alloc miptree");
+			return GL_FALSE;
+		}
+	}
+
+	/* Ensure all images are stored in the single main miptree */
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) {
+			radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]);
+			if (RADEON_DEBUG & RADEON_TEXTURE)
+				fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt);
+			if (t->mt == image->mt) {
+				if (RADEON_DEBUG & RADEON_TEXTURE)
+					fprintf(stderr, "OK\n");
+
+				continue;
+			}
+
+			if (RADEON_DEBUG & RADEON_TEXTURE)
+				fprintf(stderr, "migrating\n");
+			migrate_image_to_miptree(t->mt, image, face, level);
+		}
+	}
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Need to map texture image into memory before copying image data,
+ * then unmap it.
+ */
+static void
+radeon_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
+		     GLenum format, GLenum type, GLvoid * pixels,
+		     struct gl_texture_object *texObj,
+		     struct gl_texture_image *texImage, int compressed)
+{
+	radeon_texture_image *image = get_radeon_texture_image(texImage);
+
+	if (image->mt) {
+		/* Map the texture image read-only */
+		radeon_teximage_map(image, GL_FALSE);
+	} else {
+		/* Image hasn't been uploaded to a miptree yet */
+		assert(image->base.Data);
+	}
+
+	if (compressed) {
+		/* FIXME: this can't work for small textures (mips) which
+		         use different hw stride */
+		_mesa_get_compressed_teximage(ctx, target, level, pixels,
+					      texObj, texImage);
+	} else {
+		_mesa_get_teximage(ctx, target, level, format, type, pixels,
+				   texObj, texImage);
+	}
+     
+	if (image->mt) {
+		radeon_teximage_unmap(image);
+	}
+}
+
+void
+radeonGetTexImage(GLcontext * ctx, GLenum target, GLint level,
+		  GLenum format, GLenum type, GLvoid * pixels,
+		  struct gl_texture_object *texObj,
+		  struct gl_texture_image *texImage)
+{
+	radeon_get_tex_image(ctx, target, level, format, type, pixels,
+			     texObj, texImage, 0);
+}
+
+void
+radeonGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+			    GLvoid *pixels,
+			    struct gl_texture_object *texObj,
+			    struct gl_texture_image *texImage)
+{
+	radeon_get_tex_image(ctx, target, level, 0, 0, pixels,
+			     texObj, texImage, 1);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.h b/src/mesa/drivers/dri/radeon/radeon_texture.h
new file mode 100644
index 0000000000..888a55ba91
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.h
@@ -0,0 +1,122 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ * Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+ *
+ * The Weather Channel (TM) funded Tungsten Graphics to develop the
+ * initial release of the Radeon 8500 driver under the XFree86 license.
+ * This notice must be preserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef RADEON_TEXTURE_H
+#define RADEON_TEXTURE_H
+struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx);
+void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage);
+
+void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable);
+void radeon_teximage_unmap(radeon_texture_image *image);
+void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj);
+void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj);
+void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj);
+int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj);
+GLuint radeon_face_for_target(GLenum target);
+const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type);
+const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type, GLboolean fbo);
+
+void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target,
+				GLint level, GLint internalFormat,
+				GLint width, GLint height, GLint border,
+				GLsizei imageSize, const GLvoid * data,
+				struct gl_texture_object *texObj,
+				struct gl_texture_image *texImage);
+void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint depth,
+		      GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset,
+			 GLsizei width,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage);
+void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+				GLint xoffset, GLint yoffset,
+				GLsizei width, GLsizei height,
+				GLenum format, GLenum type,
+				const GLvoid * pixels,
+				const struct gl_pixelstore_attrib *packing,
+				struct gl_texture_object *texObj,
+				struct gl_texture_image *texImage);
+void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target,
+				   GLint level, GLint xoffset,
+				   GLint yoffset, GLsizei width,
+				   GLsizei height, GLenum format,
+				   GLsizei imageSize, const GLvoid * data,
+				   struct gl_texture_object *texObj,
+				   struct gl_texture_image *texImage);
+
+void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset, GLint zoffset,
+			 GLsizei width, GLsizei height, GLsizei depth,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage);
+
+void radeonGetTexImage(GLcontext * ctx, GLenum target, GLint level,
+		       GLenum format, GLenum type, GLvoid * pixels,
+		       struct gl_texture_object *texObj,
+		       struct gl_texture_image *texImage);
+void radeonGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+				 GLvoid *pixels,
+				 struct gl_texture_object *texObj,
+				 struct gl_texture_image *texImage);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/server/radeon_reg.h b/src/mesa/drivers/dri/radeon/server/radeon_reg.h
index ae2ccdf292..e81d7fdcd0 100644
--- a/src/mesa/drivers/dri/radeon/server/radeon_reg.h
+++ b/src/mesa/drivers/dri/radeon/server/radeon_reg.h
@@ -1601,6 +1601,8 @@
 #       define RADEON_STENCIL_VALUE_MASK      (0xff << 16)
 #       define RADEON_STENCIL_WRITEMASK_SHIFT 24
 #       define RADEON_STENCIL_WRITE_MASK      (0xff << 24)
+#define RADEON_RB3D_ZPASS_DATA              0x3290
+#define RADEON_RB3D_ZPASS_ADDR              0x3294
 #define RADEON_RB3D_ZSTENCILCNTL            0x1c2c
 #       define RADEON_DEPTH_FORMAT_MASK          (0xf << 0)
 #       define RADEON_DEPTH_FORMAT_16BIT_INT_Z   (0  <<  0)
@@ -1661,6 +1663,9 @@
 #       define RADEON_FORCE_Z_DIRTY              (1  << 29)
 #       define RADEON_Z_WRITE_ENABLE             (1  << 30)
 #       define RADEON_Z_DECOMPRESSION_ENABLE     (1  << 31)
+
+#define RADEON_RE_STIPPLE_ADDR              0x1cc8
+#define RADEON_RE_STIPPLE_DATA              0x1ccc
 #define RADEON_RE_LINE_PATTERN              0x1cd0
 #       define RADEON_LINE_PATTERN_MASK             0x0000ffff
 #       define RADEON_LINE_REPEAT_COUNT_SHIFT       16
@@ -2031,6 +2036,9 @@
 #define RADEON_CP_PACKET3_3D_DRAW_INDX              0xC0002A00
 #define RADEON_CP_PACKET3_LOAD_PALETTE              0xC0002C00
 #define RADEON_CP_PACKET3_3D_LOAD_VBPNTR            0xC0002F00
+#define R200_CP_CMD_3D_DRAW_VBUF_2      0xC0003400
+#define R200_CP_CMD_3D_DRAW_IMMD_2      0xC0003500
+#define R200_CP_CMD_3D_DRAW_INDX_2      0xC0003600
 #define RADEON_CP_PACKET3_CNTL_PAINT                0xC0009100
 #define RADEON_CP_PACKET3_CNTL_BITBLT               0xC0009200
 #define RADEON_CP_PACKET3_CNTL_SMALLTEXT            0xC0009300
diff --git a/src/mesa/drivers/dri/s3v/s3v_state.c b/src/mesa/drivers/dri/s3v/s3v_state.c
index c71c89a3e1..561f42c705 100644
--- a/src/mesa/drivers/dri/s3v/s3v_state.c
+++ b/src/mesa/drivers/dri/s3v/s3v_state.c
@@ -2,7 +2,6 @@
  * Author: Max Lingua <sunmax@libero.it>
  */
 
-#include <X11/Xlibint.h>
 #include "s3v_context.h"
 #include "s3v_macros.h"
 #include "s3v_dri.h"
@@ -24,7 +23,7 @@
 static void s3vUpdateAlphaMode( GLcontext *ctx )
 {
 	s3vContextPtr vmesa = S3V_CONTEXT(ctx);
-	CARD32 cmd = vmesa->CMD;
+	uint32_t cmd = vmesa->CMD;
 	cmd &= ~ALPHA_BLEND_MASK;
 
 	if ( ctx->Color.BlendEnabled ) {
@@ -173,7 +172,7 @@ static void s3vDDClear( GLcontext *ctx, GLbitfield mask )
 static void s3vUpdateZMode( GLcontext *ctx )
 {
 	s3vContextPtr vmesa = S3V_CONTEXT(ctx);
-	CARD32 cmd = vmesa->CMD;
+	uint32_t cmd = vmesa->CMD;
 
 	DEBUG(("Depth.Test = %i\n", ctx->Depth.Test));
 	DEBUG(("CMD was = 0x%x ", cmd));
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index a94f1c076c..931ceff0a8 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -23,7 +23,6 @@
  */
 
 
-#include <X11/Xlibint.h>
 #include <stdio.h>
 
 #include "main/context.h"
@@ -180,7 +179,7 @@ savageInitDriver(__DRIscreenPrivate *sPriv)
    }
 
    /* Allocate the private area */
-   savageScreen = (savageScreenPrivate *)Xmalloc(sizeof(savageScreenPrivate));
+   savageScreen = (savageScreenPrivate *)_mesa_malloc(sizeof(savageScreenPrivate));
    if (!savageScreen)
       return GL_FALSE;
 
@@ -227,7 +226,7 @@ savageInitDriver(__DRIscreenPrivate *sPriv)
 		  savageScreen->agpTextures.handle,
 		  savageScreen->agpTextures.size,
 		  (drmAddress *)&(savageScreen->agpTextures.map)) != 0) {
-	   Xfree(savageScreen);
+	   _mesa_free(savageScreen);
 	   sPriv->private = NULL;
 	   return GL_FALSE;
        }
@@ -247,7 +246,7 @@ savageInitDriver(__DRIscreenPrivate *sPriv)
 	      savageScreen->aperture.size, 
 	      (drmAddress *)&savageScreen->aperture.map) != 0) 
    {
-      Xfree(savageScreen);
+      _mesa_free(savageScreen);
       sPriv->private = NULL;
       return GL_FALSE;
    }
@@ -283,7 +282,7 @@ savageDestroyScreen(__DRIscreenPrivate *sPriv)
    /* free all option information */
    driDestroyOptionInfo (&savageScreen->optionCache);
 
-   Xfree(savageScreen);
+   _mesa_free(savageScreen);
    sPriv->private = NULL;
 }
 
@@ -301,7 +300,7 @@ savageCreateContext( const __GLcontextModes *mesaVis,
 						 savageScreen->sarea_priv_offset);
    int textureSize[SAVAGE_NR_TEX_HEAPS];
    int i;
-   imesa = (savageContextPtr)Xcalloc(sizeof(savageContext), 1);
+   imesa = (savageContextPtr)_mesa_calloc(sizeof(savageContext));
    if (!imesa) {
       return GL_FALSE;
    }
@@ -318,7 +317,7 @@ savageCreateContext( const __GLcontextModes *mesaVis,
       shareCtx = NULL;
    ctx = _mesa_create_context(mesaVis, shareCtx, &functions, imesa);
    if (!ctx) {
-      Xfree(imesa);
+      _mesa_free(imesa);
       return GL_FALSE;
    }
    driContextPriv->driverPrivate = imesa;
diff --git a/src/mesa/drivers/dri/savage/savagecontext.h b/src/mesa/drivers/dri/savage/savagecontext.h
index fd6399d6a6..53a37db1cb 100644
--- a/src/mesa/drivers/dri/savage/savagecontext.h
+++ b/src/mesa/drivers/dri/savage/savagecontext.h
@@ -31,7 +31,6 @@ typedef struct savage_context_t savageContext;
 typedef struct savage_context_t *savageContextPtr;
 typedef struct savage_texture_object_t *savageTextureObjectPtr;
 
-#include <X11/Xlibint.h>
 #include "dri_util.h"
 #include "main/mtypes.h"
 #include "xf86drm.h"
diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c
index fbfa49c99d..a858af30c1 100644
--- a/src/mesa/drivers/dri/swrast/swrast.c
+++ b/src/mesa/drivers/dri/swrast/swrast.c
@@ -63,7 +63,9 @@
 
 /* sw extensions not associated with some GL version */
 #define need_GL_ARB_shader_objects
+#define need_GL_ARB_vertex_array_object
 #define need_GL_ARB_vertex_program
+#define need_GL_ARB_sync
 #define need_GL_APPLE_vertex_array_object
 #define need_GL_ATI_fragment_shader
 #define need_GL_ATI_separate_stencil
@@ -94,7 +96,9 @@ const struct dri_extension card_extensions[] =
     { "GL_SGI_color_table",		GL_SGI_color_table_functions },
 
     { "GL_ARB_shader_objects",		GL_ARB_shader_objects_functions },
+    { "GL_ARB_vertex_array_object",	GL_ARB_vertex_array_object_functions },
     { "GL_ARB_vertex_program",		GL_ARB_vertex_program_functions },
+    { "GL_ARB_sync",			GL_ARB_sync_functions },
     { "GL_APPLE_vertex_array_object",	GL_APPLE_vertex_array_object_functions },
     { "GL_ATI_fragment_shader",		GL_ATI_fragment_shader_functions },
     { "GL_ATI_separate_stencil",	GL_ATI_separate_stencil_functions },
diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c
index 6ae5b6fd58..2c7be9f182 100644
--- a/src/mesa/drivers/x11/xm_api.c
+++ b/src/mesa/drivers/x11/xm_api.c
@@ -79,6 +79,7 @@
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
 #include "drivers/common/driverfuncs.h"
+#include "drivers/common/meta.h"
 
 /**
  * Global X driver lock
@@ -1317,6 +1318,7 @@ xmesa_convert_from_x_visual_type( int visualType )
 
 /* sw extensions not associated with some GL version */
 #define need_GL_ARB_shader_objects
+#define need_GL_ARB_sync
 #define need_GL_ARB_vertex_program
 #define need_GL_APPLE_vertex_array_object
 #define need_GL_ATI_fragment_shader
@@ -1346,6 +1348,7 @@ const struct dri_extension card_extensions[] =
    { "GL_SGI_color_table",		GL_SGI_color_table_functions },
 
    { "GL_ARB_shader_objects",		GL_ARB_shader_objects_functions },
+   { "GL_ARB_sync",			GL_ARB_sync_functions },
    { "GL_ARB_vertex_program",		GL_ARB_vertex_program_functions },
    { "GL_APPLE_vertex_array_object",	GL_APPLE_vertex_array_object_functions },
    { "GL_ATI_fragment_shader",		GL_ATI_fragment_shader_functions },
@@ -1641,6 +1644,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    xmesa_register_swrast_functions( mesaCtx );
    _swsetup_Wakeup(mesaCtx);
 
+   if (TEST_META_FUNCS)
+      _mesa_meta_init(mesaCtx);
+
    return c;
 }
 
@@ -1655,6 +1661,9 @@ void XMesaDestroyContext( XMesaContext c )
    FXdestroyContext( XMESA_BUFFER(mesaCtx->DrawBuffer) );
 #endif
 
+   if (TEST_META_FUNCS)
+      _mesa_meta_free( mesaCtx );
+
    _swsetup_DestroyContext( mesaCtx );
    _swrast_DestroyContext( mesaCtx );
    _tnl_DestroyContext( mesaCtx );
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index 305df548fa..465a40ac8d 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -51,6 +51,7 @@
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
+#include "drivers/common/meta.h"
 #include "xmesaP.h"
 
 
@@ -912,8 +913,9 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state )
    /*
     * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect
     * renderbuffer span/clear funcs.
+    * Check _NEW_COLOR to detect dither enable/disable.
     */
-   if (new_state & (_NEW_COLOR | _NEW_PIXEL | _NEW_BUFFERS)) {
+   if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) {
       XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer);
       struct xmesa_renderbuffer *front_xrb, *back_xrb;
 
@@ -1146,17 +1148,25 @@ xmesa_init_driver_functions( XMesaVisual xmvisual,
    driver->IndexMask = index_mask;
    driver->ColorMask = color_mask;
    driver->Enable = enable;
-   driver->Clear = clear_buffers;
    driver->Viewport = xmesa_viewport;
-#ifndef XFree86Server
-   driver->CopyPixels = xmesa_CopyPixels;
-   if (xmvisual->undithered_pf == PF_8R8G8B &&
-       xmvisual->dithered_pf == PF_8R8G8B &&
-       xmvisual->BitsPerPixel == 32) {
-      driver->DrawPixels = xmesa_DrawPixels_8R8G8B;
+   if (TEST_META_FUNCS) {
+      driver->Clear = _mesa_meta_clear;
+      driver->CopyPixels = _mesa_meta_copy_pixels;
+      driver->BlitFramebuffer = _mesa_meta_blit_framebuffer;
+      driver->DrawPixels = _mesa_meta_draw_pixels;
    }
-   else if (xmvisual->undithered_pf == PF_5R6G5B) {
-      driver->DrawPixels = xmesa_DrawPixels_5R6G5B;
+   else {
+      driver->Clear = clear_buffers;
+#ifndef XFree86Server
+      driver->CopyPixels = xmesa_CopyPixels;
+      if (xmvisual->undithered_pf == PF_8R8G8B &&
+          xmvisual->dithered_pf == PF_8R8G8B &&
+          xmvisual->BitsPerPixel == 32) {
+         driver->DrawPixels = xmesa_DrawPixels_8R8G8B;
+      }
+      else if (xmvisual->undithered_pf == PF_5R6G5B) {
+         driver->DrawPixels = xmesa_DrawPixels_5R6G5B;
+      }
    }
 #endif
    driver->TestProxyTexImage = test_proxy_teximage;
diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h
index 65e747d7b9..25db55862e 100644
--- a/src/mesa/drivers/x11/xmesaP.h
+++ b/src/mesa/drivers/x11/xmesaP.h
@@ -581,4 +581,8 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx );
 #define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */
 #endif
 
+
+#define TEST_META_FUNCS 0
+
+
 #endif
diff --git a/src/mesa/glapi/ARB_copy_buffer.xml b/src/mesa/glapi/ARB_copy_buffer.xml
new file mode 100644
index 0000000000..719816d817
--- /dev/null
+++ b/src/mesa/glapi/ARB_copy_buffer.xml
@@ -0,0 +1,24 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_copy_buffer" number="59">
+
+    <enum name="COPY_READ_BUFFER"   value="0x8F36"/>
+    <enum name="COPY_WRITE_BUFFER"  value="0x8F37"/>
+
+    <function name="CopyBufferSubData" offset="assign">
+        <param name="readTarget" type="GLenum"/>
+        <param name="writeTarget" type="GLenum"/>
+        <param name="readOffset" type="GLintptr"/>
+        <param name="writeOffset" type="GLintptr"/>
+        <param name="size" type="GLsizeiptr"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_map_buffer_range.xml b/src/mesa/glapi/ARB_map_buffer_range.xml
new file mode 100644
index 0000000000..afcb9b6ba6
--- /dev/null
+++ b/src/mesa/glapi/ARB_map_buffer_range.xml
@@ -0,0 +1,34 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_map_buffer_range" number="50">
+
+    <enum name="MAP_READ_BIT"                value="0x0001"/>
+    <enum name="MAP_WRITE_BIT"               value="0x0002"/>
+    <enum name="MAP_INVALIDATE_RANGE_BIT"    value="0x0004"/>
+    <enum name="MAP_INVALIDATE_BUFFER_BIT"   value="0x0008"/>
+    <enum name="MAP_FLUSH_EXPLICIT_BIT"      value="0x0010"/>
+    <enum name="MAP_UNSYNCHRONIZED_BIT"      value="0x0020"/>
+
+    <function name="MapBufferRange" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="offset" type="GLintptr"/>
+        <param name="length" type="GLsizeiptr"/>
+        <param name="access" type="GLbitfield"/>
+        <return type="GLvoid *"/>
+    </function>
+
+    <function name="FlushMappedBufferRange" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="offset" type="GLintptr"/>
+        <param name="length" type="GLsizeiptr"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_seamless_cube_map.xml b/src/mesa/glapi/ARB_seamless_cube_map.xml
new file mode 100644
index 0000000000..3cdc84d2b9
--- /dev/null
+++ b/src/mesa/glapi/ARB_seamless_cube_map.xml
@@ -0,0 +1,12 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<OpenGLAPI>
+
+<category name="GL_ARB_seamless_cubemap" number="65">
+    <enum name="TEXTURE_CUBE_MAP_SEAMLESS"     count="1"  value="0x88F4">
+        <size name="Get" mode="get"/>
+    </enum>
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_sync.xml b/src/mesa/glapi/ARB_sync.xml
new file mode 100644
index 0000000000..37f474980c
--- /dev/null
+++ b/src/mesa/glapi/ARB_sync.xml
@@ -0,0 +1,82 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<OpenGLAPI>
+<category name="GL_ARB_sync" number="61">
+    <type name="int64"   size="8"                  glx_name="CARD64"/>
+    <type name="uint64"  size="8"  unsigned="true" glx_name="CARD64"/>
+    <type name="sync"    size="8"  unsigned="true" glx_name="CARD64"/>
+
+    <enum name="MAX_SERVER_WAIT_TIMEOUT"       count="1"  value="0x9111">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="OBJECT_TYPE"                   count="1"  value="0x9112">
+        <size name="GetSynciv" mode="get"/>
+    </enum>
+    <enum name="SYNC_CONDITION"                count="1"  value="0x9113">
+        <size name="GetSynciv" mode="get"/>
+    </enum>
+    <enum name="SYNC_STATUS"                   count="1"  value="0x9114">
+        <size name="GetSynciv" mode="get"/>
+    </enum>
+    <enum name="SYNC_FLAGS"                    count="1"  value="0x9115">
+        <size name="GetSynciv" mode="get"/>
+    </enum>
+
+    <enum name="SYNC_FENCE"                               value="0x9116"/>
+    <enum name="SYNC_GPU_COMMANDS_COMPLETE"               value="0x9117"/>
+    <enum name="UNSIGNALED"                               value="0x9118"/>
+    <enum name="SIGNALED"                                 value="0x9119"/>
+    <enum name="ALREADY_SIGNALED"                         value="0x911A"/>
+    <enum name="TIMEOUT_EXPIRED"                          value="0x911B"/>
+    <enum name="CONDITION_SATISFIED"                      value="0x911C"/>
+    <enum name="WAIT_FAILED"                              value="0x911D"/>
+
+    <enum name="SYNC_FLUSH_COMMANDS_BIT"                  value="0x00000001"/>
+    <enum name="TIMEOUT_IGNORED"                  value="0xFFFFFFFFFFFFFFFF"/>
+
+
+
+    <function name="FenceSync" offset="assign">
+        <param name="condition" type="GLenum"/>
+        <param name="flags" type="GLbitfield"/>
+        <return type="GLsync"/>
+    </function>
+
+    <function name="IsSync" offset="assign">
+        <param name="sync" type="GLsync"/>
+	<return type="GLboolean"/>
+    </function>
+
+    <function name="DeleteSync" offset="assign">
+        <param name="sync" type="GLsync"/>
+    </function>
+
+    <function name="ClientWaitSync" offset="assign">
+        <param name="sync" type="GLsync"/>
+        <param name="flags" type="GLbitfield"/>
+	<param name="timeout" type="GLuint64"/>
+        <return type="GLenum"/>
+    </function>
+
+    <function name="WaitSync" offset="assign">
+        <param name="sync" type="GLsync"/>
+        <param name="flags" type="GLbitfield"/>
+	<param name="timeout" type="GLuint64"/>
+    </function>
+
+    <function name="GetInteger64v" offset="assign">
+        <param name="pname" type="GLenum"/>
+        <param name="params" type="GLint64 *" output="true" variable_param="pname"/>
+    </function>
+
+    <function name="GetSynciv" offset="assign">
+        <param name="sync" type="GLsync"/>
+        <param name="pname" type="GLenum"/>
+        <param name="bufSize" type="GLsizei"/>
+        <param name="length" type="GLsizei *" output="true"/>
+        <param name="values" type="GLint *" output="true" variable_param="pname"/>
+    </function>
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_vertex_array_object.xml b/src/mesa/glapi/ARB_vertex_array_object.xml
new file mode 100644
index 0000000000..3b4ab64f55
--- /dev/null
+++ b/src/mesa/glapi/ARB_vertex_array_object.xml
@@ -0,0 +1,34 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_vertex_array_object" number="54">
+
+    <enum name="VERTEX_ARRAY_BINDING" value="0x85B5"/>
+
+    <function name="BindVertexArray" offset="assign">
+        <param name="array" type="GLuint"/>
+    </function>
+
+    <function name="DeleteVertexArrays" alias="DeleteVertexArraysAPPLE">
+        <param name="n" type="GLsizei"/>
+        <param name="arrays" type="const GLuint *"/>
+    </function>
+
+    <function name="GenVertexArrays" offset="assign">
+        <param name="n" type="GLsizei"/>
+        <param name="arrays" type="GLuint *"/>
+    </function>
+
+    <function name="IsVertexArray" alias="IsVertexArrayAPPLE">
+        <param name="array" type="GLuint"/>
+        <return type="GLboolean"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/EXT_framebuffer_object.xml b/src/mesa/glapi/EXT_framebuffer_object.xml
index 1b0de2ad23..5559b48b11 100644
--- a/src/mesa/glapi/EXT_framebuffer_object.xml
+++ b/src/mesa/glapi/EXT_framebuffer_object.xml
@@ -192,42 +192,4 @@
     </function>
 </category>
 
-
-
-<category name="GL_EXT_texture_array" number="329">
-    <enum name="TEXTURE_1D_ARRAY_EXT"                      value="0x8C18"/>
-    <enum name="PROXY_TEXTURE_1D_ARRAY_EXT"                value="0x8C19"/>
-    <enum name="TEXTURE_2D_ARRAY_EXT"                      value="0x8C1A"/>
-    <enum name="PROXY_TEXTURE_2D_ARRAY_EXT"                value="0x8C1B"/>
-
-    <enum name="TEXTURE_BINDING_1D_ARRAY_EXT"   count="1"  value="0x8C1C">
-        <size name="Get" mode="get"/>
-    </enum>
-
-    <enum name="TEXTURE_BINDING_2D_ARRAY_EXT"   count="1"  value="0x8C1D">
-        <size name="Get" mode="get"/>
-    </enum>
-
-    <enum name="MAX_ARRAY_TEXTURE_LAYERS_EXT"   count="1"  value="0x88FF">
-        <size name="Get" mode="get"/>
-    </enum>
-
-    <enum name="COMPARE_REF_DEPTH_TO_TEXTURE_EXT" count="1" value="0x884E">
-        <size name="Get" mode="get"/>
-    </enum>
-
-    <enum name="FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT"   count="1" value="0x8CD4">
-        <size name="GetFramebufferAttachmentParameterivEXT" mode="get"/>
-    </enum>
-
-    <function name="FramebufferTextureLayerEXT" offset="assign">
-        <param name="target" type="GLenum"/>
-        <param name="attachment" type="GLenum"/>
-        <param name="texture" type="GLuint"/>
-        <param name="level" type="GLint"/>
-        <param name="layer" type="GLint"/>
-    </function>
-</category>
-
-
 </OpenGLAPI>
diff --git a/src/mesa/glapi/EXT_provoking_vertex.xml b/src/mesa/glapi/EXT_provoking_vertex.xml
new file mode 100644
index 0000000000..f528a2c7d3
--- /dev/null
+++ b/src/mesa/glapi/EXT_provoking_vertex.xml
@@ -0,0 +1,22 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_EXT_provoking_vertex" number="364">
+
+    <enum name="FIRST_VERTEX_CONVENTION_EXT"                  value="0x8E4D"/>
+    <enum name="LAST_VERTEX_CONVENTION_EXT"                   value="0x8E4E"/>
+    <enum name="PROVOKING_VERTEX_EXT"                         value="0x8E4F"/>
+    <enum name="QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT" value="0x8E4C"/>
+
+    <function name="ProvokingVertexEXT" offset="assign">
+        <param name="mode" type="GLenum"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/EXT_texture_array.xml b/src/mesa/glapi/EXT_texture_array.xml
new file mode 100644
index 0000000000..e5bd9f3c69
--- /dev/null
+++ b/src/mesa/glapi/EXT_texture_array.xml
@@ -0,0 +1,41 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<OpenGLAPI>
+
+<category name="GL_EXT_texture_array" number="329">
+    <enum name="TEXTURE_1D_ARRAY_EXT"                      value="0x8C18"/>
+    <enum name="PROXY_TEXTURE_1D_ARRAY_EXT"                value="0x8C19"/>
+    <enum name="TEXTURE_2D_ARRAY_EXT"                      value="0x8C1A"/>
+    <enum name="PROXY_TEXTURE_2D_ARRAY_EXT"                value="0x8C1B"/>
+
+    <enum name="TEXTURE_BINDING_1D_ARRAY_EXT"   count="1"  value="0x8C1C">
+        <size name="Get" mode="get"/>
+    </enum>
+
+    <enum name="TEXTURE_BINDING_2D_ARRAY_EXT"   count="1"  value="0x8C1D">
+        <size name="Get" mode="get"/>
+    </enum>
+
+    <enum name="MAX_ARRAY_TEXTURE_LAYERS_EXT"   count="1"  value="0x88FF">
+        <size name="Get" mode="get"/>
+    </enum>
+
+    <enum name="COMPARE_REF_DEPTH_TO_TEXTURE_EXT" count="1" value="0x884E">
+        <size name="Get" mode="get"/>
+    </enum>
+
+    <enum name="FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT"   count="1" value="0x8CD4">
+        <size name="GetFramebufferAttachmentParameterivEXT" mode="get"/>
+    </enum>
+
+    <function name="FramebufferTextureLayerEXT" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="texture" type="GLuint"/>
+        <param name="level" type="GLint"/>
+        <param name="layer" type="GLint"/>
+    </function>
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/Makefile b/src/mesa/glapi/Makefile
index f524167a47..65edab7cec 100644
--- a/src/mesa/glapi/Makefile
+++ b/src/mesa/glapi/Makefile
@@ -47,8 +47,14 @@ SERVER_OUTPUTS = \
 
 API_XML = gl_API.xml \
 	EXT_framebuffer_object.xml \
+	ARB_copy_buffer.xml \
 	ARB_framebuffer_object.xml \
-	APPLE_vertex_array_object.xml
+	ARB_map_buffer_range.xml \
+	ARB_seamless_cube_map.xml \
+	ARB_sync.xml \
+	ARB_vertex_array_object.xml \
+	APPLE_vertex_array_object.xml \
+	EXT_provoking_vertex.xml
 
 COMMON = gl_XML.py glX_XML.py license.py $(API_XML) typeexpr.py
 COMMON_GLX = $(COMMON) glX_API.xml glX_XML.py glX_proto_common.py
diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h
index 45b2fa077a..0d32045916 100644
--- a/src/mesa/glapi/dispatch.h
+++ b/src/mesa/glapi/dispatch.h
@@ -1746,6 +1746,42 @@
 #define CALL_RenderbufferStorageMultisample(disp, parameters) (*((disp)->RenderbufferStorageMultisample)) parameters
 #define GET_RenderbufferStorageMultisample(disp) ((disp)->RenderbufferStorageMultisample)
 #define SET_RenderbufferStorageMultisample(disp, fn) ((disp)->RenderbufferStorageMultisample = fn)
+#define CALL_FlushMappedBufferRange(disp, parameters) (*((disp)->FlushMappedBufferRange)) parameters
+#define GET_FlushMappedBufferRange(disp) ((disp)->FlushMappedBufferRange)
+#define SET_FlushMappedBufferRange(disp, fn) ((disp)->FlushMappedBufferRange = fn)
+#define CALL_MapBufferRange(disp, parameters) (*((disp)->MapBufferRange)) parameters
+#define GET_MapBufferRange(disp) ((disp)->MapBufferRange)
+#define SET_MapBufferRange(disp, fn) ((disp)->MapBufferRange = fn)
+#define CALL_BindVertexArray(disp, parameters) (*((disp)->BindVertexArray)) parameters
+#define GET_BindVertexArray(disp) ((disp)->BindVertexArray)
+#define SET_BindVertexArray(disp, fn) ((disp)->BindVertexArray = fn)
+#define CALL_GenVertexArrays(disp, parameters) (*((disp)->GenVertexArrays)) parameters
+#define GET_GenVertexArrays(disp) ((disp)->GenVertexArrays)
+#define SET_GenVertexArrays(disp, fn) ((disp)->GenVertexArrays = fn)
+#define CALL_CopyBufferSubData(disp, parameters) (*((disp)->CopyBufferSubData)) parameters
+#define GET_CopyBufferSubData(disp) ((disp)->CopyBufferSubData)
+#define SET_CopyBufferSubData(disp, fn) ((disp)->CopyBufferSubData = fn)
+#define CALL_ClientWaitSync(disp, parameters) (*((disp)->ClientWaitSync)) parameters
+#define GET_ClientWaitSync(disp) ((disp)->ClientWaitSync)
+#define SET_ClientWaitSync(disp, fn) ((disp)->ClientWaitSync = fn)
+#define CALL_DeleteSync(disp, parameters) (*((disp)->DeleteSync)) parameters
+#define GET_DeleteSync(disp) ((disp)->DeleteSync)
+#define SET_DeleteSync(disp, fn) ((disp)->DeleteSync = fn)
+#define CALL_FenceSync(disp, parameters) (*((disp)->FenceSync)) parameters
+#define GET_FenceSync(disp) ((disp)->FenceSync)
+#define SET_FenceSync(disp, fn) ((disp)->FenceSync = fn)
+#define CALL_GetInteger64v(disp, parameters) (*((disp)->GetInteger64v)) parameters
+#define GET_GetInteger64v(disp) ((disp)->GetInteger64v)
+#define SET_GetInteger64v(disp, fn) ((disp)->GetInteger64v = fn)
+#define CALL_GetSynciv(disp, parameters) (*((disp)->GetSynciv)) parameters
+#define GET_GetSynciv(disp) ((disp)->GetSynciv)
+#define SET_GetSynciv(disp, fn) ((disp)->GetSynciv = fn)
+#define CALL_IsSync(disp, parameters) (*((disp)->IsSync)) parameters
+#define GET_IsSync(disp) ((disp)->IsSync)
+#define SET_IsSync(disp, fn) ((disp)->IsSync = fn)
+#define CALL_WaitSync(disp, parameters) (*((disp)->WaitSync)) parameters
+#define GET_WaitSync(disp) ((disp)->WaitSync)
+#define SET_WaitSync(disp, fn) ((disp)->WaitSync = fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) (*((disp)->PolygonOffsetEXT)) parameters
 #define GET_PolygonOffsetEXT(disp) ((disp)->PolygonOffsetEXT)
 #define SET_PolygonOffsetEXT(disp, fn) ((disp)->PolygonOffsetEXT = fn)
@@ -2367,9 +2403,24 @@
 #define CALL_BlitFramebufferEXT(disp, parameters) (*((disp)->BlitFramebufferEXT)) parameters
 #define GET_BlitFramebufferEXT(disp) ((disp)->BlitFramebufferEXT)
 #define SET_BlitFramebufferEXT(disp, fn) ((disp)->BlitFramebufferEXT = fn)
+#define CALL_BufferParameteriAPPLE(disp, parameters) (*((disp)->BufferParameteriAPPLE)) parameters
+#define GET_BufferParameteriAPPLE(disp) ((disp)->BufferParameteriAPPLE)
+#define SET_BufferParameteriAPPLE(disp, fn) ((disp)->BufferParameteriAPPLE = fn)
+#define CALL_FlushMappedBufferRangeAPPLE(disp, parameters) (*((disp)->FlushMappedBufferRangeAPPLE)) parameters
+#define GET_FlushMappedBufferRangeAPPLE(disp) ((disp)->FlushMappedBufferRangeAPPLE)
+#define SET_FlushMappedBufferRangeAPPLE(disp, fn) ((disp)->FlushMappedBufferRangeAPPLE = fn)
 #define CALL_FramebufferTextureLayerEXT(disp, parameters) (*((disp)->FramebufferTextureLayerEXT)) parameters
 #define GET_FramebufferTextureLayerEXT(disp) ((disp)->FramebufferTextureLayerEXT)
 #define SET_FramebufferTextureLayerEXT(disp, fn) ((disp)->FramebufferTextureLayerEXT = fn)
+#define CALL_ProvokingVertexEXT(disp, parameters) (*((disp)->ProvokingVertexEXT)) parameters
+#define GET_ProvokingVertexEXT(disp) ((disp)->ProvokingVertexEXT)
+#define SET_ProvokingVertexEXT(disp, fn) ((disp)->ProvokingVertexEXT = fn)
+#define CALL_GetTexParameterPointervAPPLE(disp, parameters) (*((disp)->GetTexParameterPointervAPPLE)) parameters
+#define GET_GetTexParameterPointervAPPLE(disp) ((disp)->GetTexParameterPointervAPPLE)
+#define SET_GetTexParameterPointervAPPLE(disp, fn) ((disp)->GetTexParameterPointervAPPLE = fn)
+#define CALL_TextureRangeAPPLE(disp, parameters) (*((disp)->TextureRangeAPPLE)) parameters
+#define GET_TextureRangeAPPLE(disp) ((disp)->TextureRangeAPPLE)
+#define SET_TextureRangeAPPLE(disp, fn) ((disp)->TextureRangeAPPLE = fn)
 #define CALL_StencilFuncSeparateATI(disp, parameters) (*((disp)->StencilFuncSeparateATI)) parameters
 #define GET_StencilFuncSeparateATI(disp) ((disp)->StencilFuncSeparateATI)
 #define SET_StencilFuncSeparateATI(disp, fn) ((disp)->StencilFuncSeparateATI = fn)
@@ -2388,7 +2439,7 @@
 
 #else
 
-#define driDispatchRemapTable_size 367
+#define driDispatchRemapTable_size 384
 extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 
 #define AttachShader_remap_index 0
@@ -2545,219 +2596,236 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define GetAttribLocationARB_remap_index 151
 #define DrawBuffersARB_remap_index 152
 #define RenderbufferStorageMultisample_remap_index 153
-#define PolygonOffsetEXT_remap_index 154
-#define GetPixelTexGenParameterfvSGIS_remap_index 155
-#define GetPixelTexGenParameterivSGIS_remap_index 156
-#define PixelTexGenParameterfSGIS_remap_index 157
-#define PixelTexGenParameterfvSGIS_remap_index 158
-#define PixelTexGenParameteriSGIS_remap_index 159
-#define PixelTexGenParameterivSGIS_remap_index 160
-#define SampleMaskSGIS_remap_index 161
-#define SamplePatternSGIS_remap_index 162
-#define ColorPointerEXT_remap_index 163
-#define EdgeFlagPointerEXT_remap_index 164
-#define IndexPointerEXT_remap_index 165
-#define NormalPointerEXT_remap_index 166
-#define TexCoordPointerEXT_remap_index 167
-#define VertexPointerEXT_remap_index 168
-#define PointParameterfEXT_remap_index 169
-#define PointParameterfvEXT_remap_index 170
-#define LockArraysEXT_remap_index 171
-#define UnlockArraysEXT_remap_index 172
-#define CullParameterdvEXT_remap_index 173
-#define CullParameterfvEXT_remap_index 174
-#define SecondaryColor3bEXT_remap_index 175
-#define SecondaryColor3bvEXT_remap_index 176
-#define SecondaryColor3dEXT_remap_index 177
-#define SecondaryColor3dvEXT_remap_index 178
-#define SecondaryColor3fEXT_remap_index 179
-#define SecondaryColor3fvEXT_remap_index 180
-#define SecondaryColor3iEXT_remap_index 181
-#define SecondaryColor3ivEXT_remap_index 182
-#define SecondaryColor3sEXT_remap_index 183
-#define SecondaryColor3svEXT_remap_index 184
-#define SecondaryColor3ubEXT_remap_index 185
-#define SecondaryColor3ubvEXT_remap_index 186
-#define SecondaryColor3uiEXT_remap_index 187
-#define SecondaryColor3uivEXT_remap_index 188
-#define SecondaryColor3usEXT_remap_index 189
-#define SecondaryColor3usvEXT_remap_index 190
-#define SecondaryColorPointerEXT_remap_index 191
-#define MultiDrawArraysEXT_remap_index 192
-#define MultiDrawElementsEXT_remap_index 193
-#define FogCoordPointerEXT_remap_index 194
-#define FogCoorddEXT_remap_index 195
-#define FogCoorddvEXT_remap_index 196
-#define FogCoordfEXT_remap_index 197
-#define FogCoordfvEXT_remap_index 198
-#define PixelTexGenSGIX_remap_index 199
-#define BlendFuncSeparateEXT_remap_index 200
-#define FlushVertexArrayRangeNV_remap_index 201
-#define VertexArrayRangeNV_remap_index 202
-#define CombinerInputNV_remap_index 203
-#define CombinerOutputNV_remap_index 204
-#define CombinerParameterfNV_remap_index 205
-#define CombinerParameterfvNV_remap_index 206
-#define CombinerParameteriNV_remap_index 207
-#define CombinerParameterivNV_remap_index 208
-#define FinalCombinerInputNV_remap_index 209
-#define GetCombinerInputParameterfvNV_remap_index 210
-#define GetCombinerInputParameterivNV_remap_index 211
-#define GetCombinerOutputParameterfvNV_remap_index 212
-#define GetCombinerOutputParameterivNV_remap_index 213
-#define GetFinalCombinerInputParameterfvNV_remap_index 214
-#define GetFinalCombinerInputParameterivNV_remap_index 215
-#define ResizeBuffersMESA_remap_index 216
-#define WindowPos2dMESA_remap_index 217
-#define WindowPos2dvMESA_remap_index 218
-#define WindowPos2fMESA_remap_index 219
-#define WindowPos2fvMESA_remap_index 220
-#define WindowPos2iMESA_remap_index 221
-#define WindowPos2ivMESA_remap_index 222
-#define WindowPos2sMESA_remap_index 223
-#define WindowPos2svMESA_remap_index 224
-#define WindowPos3dMESA_remap_index 225
-#define WindowPos3dvMESA_remap_index 226
-#define WindowPos3fMESA_remap_index 227
-#define WindowPos3fvMESA_remap_index 228
-#define WindowPos3iMESA_remap_index 229
-#define WindowPos3ivMESA_remap_index 230
-#define WindowPos3sMESA_remap_index 231
-#define WindowPos3svMESA_remap_index 232
-#define WindowPos4dMESA_remap_index 233
-#define WindowPos4dvMESA_remap_index 234
-#define WindowPos4fMESA_remap_index 235
-#define WindowPos4fvMESA_remap_index 236
-#define WindowPos4iMESA_remap_index 237
-#define WindowPos4ivMESA_remap_index 238
-#define WindowPos4sMESA_remap_index 239
-#define WindowPos4svMESA_remap_index 240
-#define MultiModeDrawArraysIBM_remap_index 241
-#define MultiModeDrawElementsIBM_remap_index 242
-#define DeleteFencesNV_remap_index 243
-#define FinishFenceNV_remap_index 244
-#define GenFencesNV_remap_index 245
-#define GetFenceivNV_remap_index 246
-#define IsFenceNV_remap_index 247
-#define SetFenceNV_remap_index 248
-#define TestFenceNV_remap_index 249
-#define AreProgramsResidentNV_remap_index 250
-#define BindProgramNV_remap_index 251
-#define DeleteProgramsNV_remap_index 252
-#define ExecuteProgramNV_remap_index 253
-#define GenProgramsNV_remap_index 254
-#define GetProgramParameterdvNV_remap_index 255
-#define GetProgramParameterfvNV_remap_index 256
-#define GetProgramStringNV_remap_index 257
-#define GetProgramivNV_remap_index 258
-#define GetTrackMatrixivNV_remap_index 259
-#define GetVertexAttribPointervNV_remap_index 260
-#define GetVertexAttribdvNV_remap_index 261
-#define GetVertexAttribfvNV_remap_index 262
-#define GetVertexAttribivNV_remap_index 263
-#define IsProgramNV_remap_index 264
-#define LoadProgramNV_remap_index 265
-#define ProgramParameters4dvNV_remap_index 266
-#define ProgramParameters4fvNV_remap_index 267
-#define RequestResidentProgramsNV_remap_index 268
-#define TrackMatrixNV_remap_index 269
-#define VertexAttrib1dNV_remap_index 270
-#define VertexAttrib1dvNV_remap_index 271
-#define VertexAttrib1fNV_remap_index 272
-#define VertexAttrib1fvNV_remap_index 273
-#define VertexAttrib1sNV_remap_index 274
-#define VertexAttrib1svNV_remap_index 275
-#define VertexAttrib2dNV_remap_index 276
-#define VertexAttrib2dvNV_remap_index 277
-#define VertexAttrib2fNV_remap_index 278
-#define VertexAttrib2fvNV_remap_index 279
-#define VertexAttrib2sNV_remap_index 280
-#define VertexAttrib2svNV_remap_index 281
-#define VertexAttrib3dNV_remap_index 282
-#define VertexAttrib3dvNV_remap_index 283
-#define VertexAttrib3fNV_remap_index 284
-#define VertexAttrib3fvNV_remap_index 285
-#define VertexAttrib3sNV_remap_index 286
-#define VertexAttrib3svNV_remap_index 287
-#define VertexAttrib4dNV_remap_index 288
-#define VertexAttrib4dvNV_remap_index 289
-#define VertexAttrib4fNV_remap_index 290
-#define VertexAttrib4fvNV_remap_index 291
-#define VertexAttrib4sNV_remap_index 292
-#define VertexAttrib4svNV_remap_index 293
-#define VertexAttrib4ubNV_remap_index 294
-#define VertexAttrib4ubvNV_remap_index 295
-#define VertexAttribPointerNV_remap_index 296
-#define VertexAttribs1dvNV_remap_index 297
-#define VertexAttribs1fvNV_remap_index 298
-#define VertexAttribs1svNV_remap_index 299
-#define VertexAttribs2dvNV_remap_index 300
-#define VertexAttribs2fvNV_remap_index 301
-#define VertexAttribs2svNV_remap_index 302
-#define VertexAttribs3dvNV_remap_index 303
-#define VertexAttribs3fvNV_remap_index 304
-#define VertexAttribs3svNV_remap_index 305
-#define VertexAttribs4dvNV_remap_index 306
-#define VertexAttribs4fvNV_remap_index 307
-#define VertexAttribs4svNV_remap_index 308
-#define VertexAttribs4ubvNV_remap_index 309
-#define GetTexBumpParameterfvATI_remap_index 310
-#define GetTexBumpParameterivATI_remap_index 311
-#define TexBumpParameterfvATI_remap_index 312
-#define TexBumpParameterivATI_remap_index 313
-#define AlphaFragmentOp1ATI_remap_index 314
-#define AlphaFragmentOp2ATI_remap_index 315
-#define AlphaFragmentOp3ATI_remap_index 316
-#define BeginFragmentShaderATI_remap_index 317
-#define BindFragmentShaderATI_remap_index 318
-#define ColorFragmentOp1ATI_remap_index 319
-#define ColorFragmentOp2ATI_remap_index 320
-#define ColorFragmentOp3ATI_remap_index 321
-#define DeleteFragmentShaderATI_remap_index 322
-#define EndFragmentShaderATI_remap_index 323
-#define GenFragmentShadersATI_remap_index 324
-#define PassTexCoordATI_remap_index 325
-#define SampleMapATI_remap_index 326
-#define SetFragmentShaderConstantATI_remap_index 327
-#define PointParameteriNV_remap_index 328
-#define PointParameterivNV_remap_index 329
-#define ActiveStencilFaceEXT_remap_index 330
-#define BindVertexArrayAPPLE_remap_index 331
-#define DeleteVertexArraysAPPLE_remap_index 332
-#define GenVertexArraysAPPLE_remap_index 333
-#define IsVertexArrayAPPLE_remap_index 334
-#define GetProgramNamedParameterdvNV_remap_index 335
-#define GetProgramNamedParameterfvNV_remap_index 336
-#define ProgramNamedParameter4dNV_remap_index 337
-#define ProgramNamedParameter4dvNV_remap_index 338
-#define ProgramNamedParameter4fNV_remap_index 339
-#define ProgramNamedParameter4fvNV_remap_index 340
-#define DepthBoundsEXT_remap_index 341
-#define BlendEquationSeparateEXT_remap_index 342
-#define BindFramebufferEXT_remap_index 343
-#define BindRenderbufferEXT_remap_index 344
-#define CheckFramebufferStatusEXT_remap_index 345
-#define DeleteFramebuffersEXT_remap_index 346
-#define DeleteRenderbuffersEXT_remap_index 347
-#define FramebufferRenderbufferEXT_remap_index 348
-#define FramebufferTexture1DEXT_remap_index 349
-#define FramebufferTexture2DEXT_remap_index 350
-#define FramebufferTexture3DEXT_remap_index 351
-#define GenFramebuffersEXT_remap_index 352
-#define GenRenderbuffersEXT_remap_index 353
-#define GenerateMipmapEXT_remap_index 354
-#define GetFramebufferAttachmentParameterivEXT_remap_index 355
-#define GetRenderbufferParameterivEXT_remap_index 356
-#define IsFramebufferEXT_remap_index 357
-#define IsRenderbufferEXT_remap_index 358
-#define RenderbufferStorageEXT_remap_index 359
-#define BlitFramebufferEXT_remap_index 360
-#define FramebufferTextureLayerEXT_remap_index 361
-#define StencilFuncSeparateATI_remap_index 362
-#define ProgramEnvParameters4fvEXT_remap_index 363
-#define ProgramLocalParameters4fvEXT_remap_index 364
-#define GetQueryObjecti64vEXT_remap_index 365
-#define GetQueryObjectui64vEXT_remap_index 366
+#define FlushMappedBufferRange_remap_index 154
+#define MapBufferRange_remap_index 155
+#define BindVertexArray_remap_index 156
+#define GenVertexArrays_remap_index 157
+#define CopyBufferSubData_remap_index 158
+#define ClientWaitSync_remap_index 159
+#define DeleteSync_remap_index 160
+#define FenceSync_remap_index 161
+#define GetInteger64v_remap_index 162
+#define GetSynciv_remap_index 163
+#define IsSync_remap_index 164
+#define WaitSync_remap_index 165
+#define PolygonOffsetEXT_remap_index 166
+#define GetPixelTexGenParameterfvSGIS_remap_index 167
+#define GetPixelTexGenParameterivSGIS_remap_index 168
+#define PixelTexGenParameterfSGIS_remap_index 169
+#define PixelTexGenParameterfvSGIS_remap_index 170
+#define PixelTexGenParameteriSGIS_remap_index 171
+#define PixelTexGenParameterivSGIS_remap_index 172
+#define SampleMaskSGIS_remap_index 173
+#define SamplePatternSGIS_remap_index 174
+#define ColorPointerEXT_remap_index 175
+#define EdgeFlagPointerEXT_remap_index 176
+#define IndexPointerEXT_remap_index 177
+#define NormalPointerEXT_remap_index 178
+#define TexCoordPointerEXT_remap_index 179
+#define VertexPointerEXT_remap_index 180
+#define PointParameterfEXT_remap_index 181
+#define PointParameterfvEXT_remap_index 182
+#define LockArraysEXT_remap_index 183
+#define UnlockArraysEXT_remap_index 184
+#define CullParameterdvEXT_remap_index 185
+#define CullParameterfvEXT_remap_index 186
+#define SecondaryColor3bEXT_remap_index 187
+#define SecondaryColor3bvEXT_remap_index 188
+#define SecondaryColor3dEXT_remap_index 189
+#define SecondaryColor3dvEXT_remap_index 190
+#define SecondaryColor3fEXT_remap_index 191
+#define SecondaryColor3fvEXT_remap_index 192
+#define SecondaryColor3iEXT_remap_index 193
+#define SecondaryColor3ivEXT_remap_index 194
+#define SecondaryColor3sEXT_remap_index 195
+#define SecondaryColor3svEXT_remap_index 196
+#define SecondaryColor3ubEXT_remap_index 197
+#define SecondaryColor3ubvEXT_remap_index 198
+#define SecondaryColor3uiEXT_remap_index 199
+#define SecondaryColor3uivEXT_remap_index 200
+#define SecondaryColor3usEXT_remap_index 201
+#define SecondaryColor3usvEXT_remap_index 202
+#define SecondaryColorPointerEXT_remap_index 203
+#define MultiDrawArraysEXT_remap_index 204
+#define MultiDrawElementsEXT_remap_index 205
+#define FogCoordPointerEXT_remap_index 206
+#define FogCoorddEXT_remap_index 207
+#define FogCoorddvEXT_remap_index 208
+#define FogCoordfEXT_remap_index 209
+#define FogCoordfvEXT_remap_index 210
+#define PixelTexGenSGIX_remap_index 211
+#define BlendFuncSeparateEXT_remap_index 212
+#define FlushVertexArrayRangeNV_remap_index 213
+#define VertexArrayRangeNV_remap_index 214
+#define CombinerInputNV_remap_index 215
+#define CombinerOutputNV_remap_index 216
+#define CombinerParameterfNV_remap_index 217
+#define CombinerParameterfvNV_remap_index 218
+#define CombinerParameteriNV_remap_index 219
+#define CombinerParameterivNV_remap_index 220
+#define FinalCombinerInputNV_remap_index 221
+#define GetCombinerInputParameterfvNV_remap_index 222
+#define GetCombinerInputParameterivNV_remap_index 223
+#define GetCombinerOutputParameterfvNV_remap_index 224
+#define GetCombinerOutputParameterivNV_remap_index 225
+#define GetFinalCombinerInputParameterfvNV_remap_index 226
+#define GetFinalCombinerInputParameterivNV_remap_index 227
+#define ResizeBuffersMESA_remap_index 228
+#define WindowPos2dMESA_remap_index 229
+#define WindowPos2dvMESA_remap_index 230
+#define WindowPos2fMESA_remap_index 231
+#define WindowPos2fvMESA_remap_index 232
+#define WindowPos2iMESA_remap_index 233
+#define WindowPos2ivMESA_remap_index 234
+#define WindowPos2sMESA_remap_index 235
+#define WindowPos2svMESA_remap_index 236
+#define WindowPos3dMESA_remap_index 237
+#define WindowPos3dvMESA_remap_index 238
+#define WindowPos3fMESA_remap_index 239
+#define WindowPos3fvMESA_remap_index 240
+#define WindowPos3iMESA_remap_index 241
+#define WindowPos3ivMESA_remap_index 242
+#define WindowPos3sMESA_remap_index 243
+#define WindowPos3svMESA_remap_index 244
+#define WindowPos4dMESA_remap_index 245
+#define WindowPos4dvMESA_remap_index 246
+#define WindowPos4fMESA_remap_index 247
+#define WindowPos4fvMESA_remap_index 248
+#define WindowPos4iMESA_remap_index 249
+#define WindowPos4ivMESA_remap_index 250
+#define WindowPos4sMESA_remap_index 251
+#define WindowPos4svMESA_remap_index 252
+#define MultiModeDrawArraysIBM_remap_index 253
+#define MultiModeDrawElementsIBM_remap_index 254
+#define DeleteFencesNV_remap_index 255
+#define FinishFenceNV_remap_index 256
+#define GenFencesNV_remap_index 257
+#define GetFenceivNV_remap_index 258
+#define IsFenceNV_remap_index 259
+#define SetFenceNV_remap_index 260
+#define TestFenceNV_remap_index 261
+#define AreProgramsResidentNV_remap_index 262
+#define BindProgramNV_remap_index 263
+#define DeleteProgramsNV_remap_index 264
+#define ExecuteProgramNV_remap_index 265
+#define GenProgramsNV_remap_index 266
+#define GetProgramParameterdvNV_remap_index 267
+#define GetProgramParameterfvNV_remap_index 268
+#define GetProgramStringNV_remap_index 269
+#define GetProgramivNV_remap_index 270
+#define GetTrackMatrixivNV_remap_index 271
+#define GetVertexAttribPointervNV_remap_index 272
+#define GetVertexAttribdvNV_remap_index 273
+#define GetVertexAttribfvNV_remap_index 274
+#define GetVertexAttribivNV_remap_index 275
+#define IsProgramNV_remap_index 276
+#define LoadProgramNV_remap_index 277
+#define ProgramParameters4dvNV_remap_index 278
+#define ProgramParameters4fvNV_remap_index 279
+#define RequestResidentProgramsNV_remap_index 280
+#define TrackMatrixNV_remap_index 281
+#define VertexAttrib1dNV_remap_index 282
+#define VertexAttrib1dvNV_remap_index 283
+#define VertexAttrib1fNV_remap_index 284
+#define VertexAttrib1fvNV_remap_index 285
+#define VertexAttrib1sNV_remap_index 286
+#define VertexAttrib1svNV_remap_index 287
+#define VertexAttrib2dNV_remap_index 288
+#define VertexAttrib2dvNV_remap_index 289
+#define VertexAttrib2fNV_remap_index 290
+#define VertexAttrib2fvNV_remap_index 291
+#define VertexAttrib2sNV_remap_index 292
+#define VertexAttrib2svNV_remap_index 293
+#define VertexAttrib3dNV_remap_index 294
+#define VertexAttrib3dvNV_remap_index 295
+#define VertexAttrib3fNV_remap_index 296
+#define VertexAttrib3fvNV_remap_index 297
+#define VertexAttrib3sNV_remap_index 298
+#define VertexAttrib3svNV_remap_index 299
+#define VertexAttrib4dNV_remap_index 300
+#define VertexAttrib4dvNV_remap_index 301
+#define VertexAttrib4fNV_remap_index 302
+#define VertexAttrib4fvNV_remap_index 303
+#define VertexAttrib4sNV_remap_index 304
+#define VertexAttrib4svNV_remap_index 305
+#define VertexAttrib4ubNV_remap_index 306
+#define VertexAttrib4ubvNV_remap_index 307
+#define VertexAttribPointerNV_remap_index 308
+#define VertexAttribs1dvNV_remap_index 309
+#define VertexAttribs1fvNV_remap_index 310
+#define VertexAttribs1svNV_remap_index 311
+#define VertexAttribs2dvNV_remap_index 312
+#define VertexAttribs2fvNV_remap_index 313
+#define VertexAttribs2svNV_remap_index 314
+#define VertexAttribs3dvNV_remap_index 315
+#define VertexAttribs3fvNV_remap_index 316
+#define VertexAttribs3svNV_remap_index 317
+#define VertexAttribs4dvNV_remap_index 318
+#define VertexAttribs4fvNV_remap_index 319
+#define VertexAttribs4svNV_remap_index 320
+#define VertexAttribs4ubvNV_remap_index 321
+#define GetTexBumpParameterfvATI_remap_index 322
+#define GetTexBumpParameterivATI_remap_index 323
+#define TexBumpParameterfvATI_remap_index 324
+#define TexBumpParameterivATI_remap_index 325
+#define AlphaFragmentOp1ATI_remap_index 326
+#define AlphaFragmentOp2ATI_remap_index 327
+#define AlphaFragmentOp3ATI_remap_index 328
+#define BeginFragmentShaderATI_remap_index 329
+#define BindFragmentShaderATI_remap_index 330
+#define ColorFragmentOp1ATI_remap_index 331
+#define ColorFragmentOp2ATI_remap_index 332
+#define ColorFragmentOp3ATI_remap_index 333
+#define DeleteFragmentShaderATI_remap_index 334
+#define EndFragmentShaderATI_remap_index 335
+#define GenFragmentShadersATI_remap_index 336
+#define PassTexCoordATI_remap_index 337
+#define SampleMapATI_remap_index 338
+#define SetFragmentShaderConstantATI_remap_index 339
+#define PointParameteriNV_remap_index 340
+#define PointParameterivNV_remap_index 341
+#define ActiveStencilFaceEXT_remap_index 342
+#define BindVertexArrayAPPLE_remap_index 343
+#define DeleteVertexArraysAPPLE_remap_index 344
+#define GenVertexArraysAPPLE_remap_index 345
+#define IsVertexArrayAPPLE_remap_index 346
+#define GetProgramNamedParameterdvNV_remap_index 347
+#define GetProgramNamedParameterfvNV_remap_index 348
+#define ProgramNamedParameter4dNV_remap_index 349
+#define ProgramNamedParameter4dvNV_remap_index 350
+#define ProgramNamedParameter4fNV_remap_index 351
+#define ProgramNamedParameter4fvNV_remap_index 352
+#define DepthBoundsEXT_remap_index 353
+#define BlendEquationSeparateEXT_remap_index 354
+#define BindFramebufferEXT_remap_index 355
+#define BindRenderbufferEXT_remap_index 356
+#define CheckFramebufferStatusEXT_remap_index 357
+#define DeleteFramebuffersEXT_remap_index 358
+#define DeleteRenderbuffersEXT_remap_index 359
+#define FramebufferRenderbufferEXT_remap_index 360
+#define FramebufferTexture1DEXT_remap_index 361
+#define FramebufferTexture2DEXT_remap_index 362
+#define FramebufferTexture3DEXT_remap_index 363
+#define GenFramebuffersEXT_remap_index 364
+#define GenRenderbuffersEXT_remap_index 365
+#define GenerateMipmapEXT_remap_index 366
+#define GetFramebufferAttachmentParameterivEXT_remap_index 367
+#define GetRenderbufferParameterivEXT_remap_index 368
+#define IsFramebufferEXT_remap_index 369
+#define IsRenderbufferEXT_remap_index 370
+#define RenderbufferStorageEXT_remap_index 371
+#define BlitFramebufferEXT_remap_index 372
+#define BufferParameteriAPPLE_remap_index 373
+#define FlushMappedBufferRangeAPPLE_remap_index 374
+#define FramebufferTextureLayerEXT_remap_index 375
+#define ProvokingVertexEXT_remap_index 376
+#define GetTexParameterPointervAPPLE_remap_index 377
+#define TextureRangeAPPLE_remap_index 378
+#define StencilFuncSeparateATI_remap_index 379
+#define ProgramEnvParameters4fvEXT_remap_index 380
+#define ProgramLocalParameters4fvEXT_remap_index 381
+#define GetQueryObjecti64vEXT_remap_index 382
+#define GetQueryObjectui64vEXT_remap_index 383
 
 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters)
 #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index])
@@ -3221,6 +3289,42 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define CALL_RenderbufferStorageMultisample(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLsizei, GLenum, GLsizei, GLsizei)), driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], parameters)
 #define GET_RenderbufferStorageMultisample(disp) GET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index])
 #define SET_RenderbufferStorageMultisample(disp, fn) SET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], fn)
+#define CALL_FlushMappedBufferRange(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLintptr, GLsizeiptr)), driDispatchRemapTable[FlushMappedBufferRange_remap_index], parameters)
+#define GET_FlushMappedBufferRange(disp) GET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRange_remap_index])
+#define SET_FlushMappedBufferRange(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRange_remap_index], fn)
+#define CALL_MapBufferRange(disp, parameters) CALL_by_offset(disp, (GLvoid * (GLAPIENTRYP)(GLenum, GLintptr, GLsizeiptr, GLbitfield)), driDispatchRemapTable[MapBufferRange_remap_index], parameters)
+#define GET_MapBufferRange(disp) GET_by_offset(disp, driDispatchRemapTable[MapBufferRange_remap_index])
+#define SET_MapBufferRange(disp, fn) SET_by_offset(disp, driDispatchRemapTable[MapBufferRange_remap_index], fn)
+#define CALL_BindVertexArray(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint)), driDispatchRemapTable[BindVertexArray_remap_index], parameters)
+#define GET_BindVertexArray(disp) GET_by_offset(disp, driDispatchRemapTable[BindVertexArray_remap_index])
+#define SET_BindVertexArray(disp, fn) SET_by_offset(disp, driDispatchRemapTable[BindVertexArray_remap_index], fn)
+#define CALL_GenVertexArrays(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsizei, GLuint *)), driDispatchRemapTable[GenVertexArrays_remap_index], parameters)
+#define GET_GenVertexArrays(disp) GET_by_offset(disp, driDispatchRemapTable[GenVertexArrays_remap_index])
+#define SET_GenVertexArrays(disp, fn) SET_by_offset(disp, driDispatchRemapTable[GenVertexArrays_remap_index], fn)
+#define CALL_CopyBufferSubData(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLintptr, GLintptr, GLsizeiptr)), driDispatchRemapTable[CopyBufferSubData_remap_index], parameters)
+#define GET_CopyBufferSubData(disp) GET_by_offset(disp, driDispatchRemapTable[CopyBufferSubData_remap_index])
+#define SET_CopyBufferSubData(disp, fn) SET_by_offset(disp, driDispatchRemapTable[CopyBufferSubData_remap_index], fn)
+#define CALL_ClientWaitSync(disp, parameters) CALL_by_offset(disp, (GLenum (GLAPIENTRYP)(GLsync, GLbitfield, GLuint64)), driDispatchRemapTable[ClientWaitSync_remap_index], parameters)
+#define GET_ClientWaitSync(disp) GET_by_offset(disp, driDispatchRemapTable[ClientWaitSync_remap_index])
+#define SET_ClientWaitSync(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ClientWaitSync_remap_index], fn)
+#define CALL_DeleteSync(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsync)), driDispatchRemapTable[DeleteSync_remap_index], parameters)
+#define GET_DeleteSync(disp) GET_by_offset(disp, driDispatchRemapTable[DeleteSync_remap_index])
+#define SET_DeleteSync(disp, fn) SET_by_offset(disp, driDispatchRemapTable[DeleteSync_remap_index], fn)
+#define CALL_FenceSync(disp, parameters) CALL_by_offset(disp, (GLsync (GLAPIENTRYP)(GLenum, GLbitfield)), driDispatchRemapTable[FenceSync_remap_index], parameters)
+#define GET_FenceSync(disp) GET_by_offset(disp, driDispatchRemapTable[FenceSync_remap_index])
+#define SET_FenceSync(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FenceSync_remap_index], fn)
+#define CALL_GetInteger64v(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLint64 *)), driDispatchRemapTable[GetInteger64v_remap_index], parameters)
+#define GET_GetInteger64v(disp) GET_by_offset(disp, driDispatchRemapTable[GetInteger64v_remap_index])
+#define SET_GetInteger64v(disp, fn) SET_by_offset(disp, driDispatchRemapTable[GetInteger64v_remap_index], fn)
+#define CALL_GetSynciv(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsync, GLenum, GLsizei, GLsizei *, GLint *)), driDispatchRemapTable[GetSynciv_remap_index], parameters)
+#define GET_GetSynciv(disp) GET_by_offset(disp, driDispatchRemapTable[GetSynciv_remap_index])
+#define SET_GetSynciv(disp, fn) SET_by_offset(disp, driDispatchRemapTable[GetSynciv_remap_index], fn)
+#define CALL_IsSync(disp, parameters) CALL_by_offset(disp, (GLboolean (GLAPIENTRYP)(GLsync)), driDispatchRemapTable[IsSync_remap_index], parameters)
+#define GET_IsSync(disp) GET_by_offset(disp, driDispatchRemapTable[IsSync_remap_index])
+#define SET_IsSync(disp, fn) SET_by_offset(disp, driDispatchRemapTable[IsSync_remap_index], fn)
+#define CALL_WaitSync(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsync, GLbitfield, GLuint64)), driDispatchRemapTable[WaitSync_remap_index], parameters)
+#define GET_WaitSync(disp) GET_by_offset(disp, driDispatchRemapTable[WaitSync_remap_index])
+#define SET_WaitSync(disp, fn) SET_by_offset(disp, driDispatchRemapTable[WaitSync_remap_index], fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLfloat, GLfloat)), driDispatchRemapTable[PolygonOffsetEXT_remap_index], parameters)
 #define GET_PolygonOffsetEXT(disp) GET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index])
 #define SET_PolygonOffsetEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index], fn)
@@ -3842,9 +3946,24 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define CALL_BlitFramebufferEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLint, GLint, GLint, GLint, GLint, GLint, GLint, GLint, GLbitfield, GLenum)), driDispatchRemapTable[BlitFramebufferEXT_remap_index], parameters)
 #define GET_BlitFramebufferEXT(disp) GET_by_offset(disp, driDispatchRemapTable[BlitFramebufferEXT_remap_index])
 #define SET_BlitFramebufferEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[BlitFramebufferEXT_remap_index], fn)
+#define CALL_BufferParameteriAPPLE(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLint)), driDispatchRemapTable[BufferParameteriAPPLE_remap_index], parameters)
+#define GET_BufferParameteriAPPLE(disp) GET_by_offset(disp, driDispatchRemapTable[BufferParameteriAPPLE_remap_index])
+#define SET_BufferParameteriAPPLE(disp, fn) SET_by_offset(disp, driDispatchRemapTable[BufferParameteriAPPLE_remap_index], fn)
+#define CALL_FlushMappedBufferRangeAPPLE(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLintptr, GLsizeiptr)), driDispatchRemapTable[FlushMappedBufferRangeAPPLE_remap_index], parameters)
+#define GET_FlushMappedBufferRangeAPPLE(disp) GET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRangeAPPLE_remap_index])
+#define SET_FlushMappedBufferRangeAPPLE(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRangeAPPLE_remap_index], fn)
 #define CALL_FramebufferTextureLayerEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLuint, GLint, GLint)), driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], parameters)
 #define GET_FramebufferTextureLayerEXT(disp) GET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index])
 #define SET_FramebufferTextureLayerEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], fn)
+#define CALL_ProvokingVertexEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum)), driDispatchRemapTable[ProvokingVertexEXT_remap_index], parameters)
+#define GET_ProvokingVertexEXT(disp) GET_by_offset(disp, driDispatchRemapTable[ProvokingVertexEXT_remap_index])
+#define SET_ProvokingVertexEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProvokingVertexEXT_remap_index], fn)
+#define CALL_GetTexParameterPointervAPPLE(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLvoid **)), driDispatchRemapTable[GetTexParameterPointervAPPLE_remap_index], parameters)
+#define GET_GetTexParameterPointervAPPLE(disp) GET_by_offset(disp, driDispatchRemapTable[GetTexParameterPointervAPPLE_remap_index])
+#define SET_GetTexParameterPointervAPPLE(disp, fn) SET_by_offset(disp, driDispatchRemapTable[GetTexParameterPointervAPPLE_remap_index], fn)
+#define CALL_TextureRangeAPPLE(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLsizei, GLvoid *)), driDispatchRemapTable[TextureRangeAPPLE_remap_index], parameters)
+#define GET_TextureRangeAPPLE(disp) GET_by_offset(disp, driDispatchRemapTable[TextureRangeAPPLE_remap_index])
+#define SET_TextureRangeAPPLE(disp, fn) SET_by_offset(disp, driDispatchRemapTable[TextureRangeAPPLE_remap_index], fn)
 #define CALL_StencilFuncSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLint, GLuint)), driDispatchRemapTable[StencilFuncSeparateATI_remap_index], parameters)
 #define GET_StencilFuncSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index])
 #define SET_StencilFuncSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index], fn)
diff --git a/src/mesa/glapi/extension_helper.py b/src/mesa/glapi/extension_helper.py
index 64f64a2fd8..83471d89f5 100644
--- a/src/mesa/glapi/extension_helper.py
+++ b/src/mesa/glapi/extension_helper.py
@@ -170,7 +170,7 @@ class PrintGlExtensionGlue(gl_XML.gl_print_base):
 			condition = condition_for_function(f, abi, 0)
 			if len(condition):
 				print '#if %s' % (string.join(condition, " || "))
-				print 'static const char %s_names[] = ' % (f.name)
+				print 'static const char %s_names[] =' % (f.name)
 
 				parameter_signature = ''
 				for p in f.parameterIterator():
diff --git a/src/mesa/glapi/glX_proto_send.py b/src/mesa/glapi/glX_proto_send.py
index 501706acc7..daca1b767a 100644
--- a/src/mesa/glapi/glX_proto_send.py
+++ b/src/mesa/glapi/glX_proto_send.py
@@ -373,9 +373,13 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 				print '{'
 				print '    __GLXcontext * const gc = __glXGetCurrentContext();'
 				print ''
+				print '#ifdef GLX_DIRECT_RENDERING'
 				print '    if (gc->driContext) {'
 				print '    %sCALL_%s(GET_DISPATCH(), (%s));' % (ret_string, func.name, func.get_called_parameter_string())
-				print '    } else {'
+				print '    } else'
+				print '#endif'
+				print '    {'
+
 				footer = '}\n}\n'
 			else:
 				print '#define %s %d' % (func.opcode_name(), func.opcode_value())
diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml
index 4b66793e1c..1dfd92be08 100644
--- a/src/mesa/glapi/gl_API.xml
+++ b/src/mesa/glapi/gl_API.xml
@@ -7948,6 +7948,16 @@
 
 <xi:include href="ARB_framebuffer_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
 
+<xi:include href="ARB_copy_buffer.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_map_buffer_range.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_vertex_array_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_sync.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_seamless_cube_map.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
 
 <!-- Non-ARB extensions sorted by extension number. -->
 
@@ -12224,6 +12234,8 @@
 
 <xi:include href="EXT_framebuffer_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
 
+<xi:include href="EXT_provoking_vertex.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
 <category name="GL_EXT_framebuffer_blit" number="316">
     <enum name="READ_FRAMEBUFFER_EXT" value="0x8CA8"/>
     <enum name="DRAW_FRAMEBUFFER_EXT" value="0x8CA9"/>
@@ -12247,6 +12259,65 @@
     </function>
 </category>
 
+<category name="GL_APPLE_flush_buffer_range" number="321">
+    <enum name="BUFFER_SERIALIZED_MODIFY_APPLE" count="1" value="0x8A12">
+        <size name="GetBufferParameteriv" mode="get"/>
+    </enum>
+    <enum name="BUFFER_FLUSHING_UNMAP_APPLE" count="1" value="0x8A13">
+        <size name="GetBufferParameteriv" mode="get"/>
+    </enum>
+    <function name="BufferParameteriAPPLE" offset="assign" static_dispatch="false">
+        <param name="target" type="GLenum"/>
+        <param name="pname" type="GLenum"/>
+        <param name="param" type="GLint"/>
+    </function>
+    <function name="FlushMappedBufferRangeAPPLE" offset="assign" static_dispatch="false">
+        <param name="target" type="GLenum"/>
+        <param name="offset" type="GLintptr"/>
+        <param name="size" type="GLsizeiptr"/>
+    </function>
+</category>
+
+<xi:include href="EXT_texture_array.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<category name="GL_APPLE_texture_range" number="367">
+    <enum name="TEXTURE_STORAGE_HINT_APPLE" count="1" value="0x85BC">
+        <size name="TexParameteriv"/>
+        <size name="TexParameterfv"/>
+        <size name="GetTexParameteriv" mode="get"/>
+        <size name="GetTexParameterfv" mode="get"/>
+    </enum>
+    <enum name="STORAGE_PRIVATE_APPLE" count="1" value="0x85BD">
+        <size name="TexParameteriv"/>
+        <size name="TexParameterfv"/>
+    </enum>
+    <enum name="STORAGE_CACHED_APPLE" count="1" value="0x85BE">
+        <size name="TexParameteriv"/>
+        <size name="TexParameterfv"/>
+    </enum>
+    <enum name="STORAGE_SHARED_APPLE" count="1" value="0x85BF">
+        <size name="TexParameteriv"/>
+        <size name="TexParameterfv"/>
+    </enum>
+    <enum name="TEXTURE_RANGE_LENGTH_APPLE" count="1" value="0x85B7">
+        <size name="GetTexParameteriv" mode="get"/>
+        <size name="GetTexParameterfv" mode="get"/>
+    </enum>
+    <enum name="TEXTURE_RANGE_POINTER_APPLE" count="1" value="0x85B8">
+        <size name="GetTexParameterPointervAPPLE" mode="get"/>
+    </enum>
+    <function name="TextureRangeAPPLE" offset="assign" static_dispatch="false">
+        <param name="target" type="GLenum"/>
+        <param name="length" type="GLsizei"/>
+        <param name="pointer" type="GLvoid *"/>
+    </function>
+    <function name="GetTexParameterPointervAPPLE" offset="assign" static_dispatch="false">
+        <param name="target" type="GLenum"/>
+        <param name="pname" type="GLenum"/>
+        <param name="params" type="GLvoid **"/>
+    </function>
+</category>
+
 <!-- Unnumbered extensions sorted by name. -->
 
 <category name="GL_ATI_blend_equation_separate">
diff --git a/src/mesa/glapi/gl_x86-64_asm.py b/src/mesa/glapi/gl_x86-64_asm.py
index fa45406f47..f36ad3a5d8 100644
--- a/src/mesa/glapi/gl_x86-64_asm.py
+++ b/src/mesa/glapi/gl_x86-64_asm.py
@@ -138,7 +138,7 @@ class PrintGenericStubs(gl_XML.gl_print_base):
 		print '#  define GL_PREFIX(n) GLNAME(CONCAT(gl,n))'
 		print '# endif'
 		print ''
-		print '#if defined(PTHREADS) || defined(USE_XTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)'
+		print '#if defined(PTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)'
 		print '#  define THREADS'
 		print '#endif'
 		print ''
diff --git a/src/mesa/glapi/gl_x86_asm.py b/src/mesa/glapi/gl_x86_asm.py
index 0dbf3ebe0a..36f0e31fe2 100644
--- a/src/mesa/glapi/gl_x86_asm.py
+++ b/src/mesa/glapi/gl_x86_asm.py
@@ -79,7 +79,7 @@ class PrintGenericStubs(gl_XML.gl_print_base):
 		print '#define GLOBL_FN(x) GLOBL x'
 		print '#endif'
 		print ''
-		print '#if defined(PTHREADS) || defined(USE_XTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)'
+		print '#if defined(PTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)'
 		print '#  define THREADS'
 		print '#endif'
 		print ''
diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c
index 2b105d0f17..e36fccb354 100644
--- a/src/mesa/glapi/glapi.c
+++ b/src/mesa/glapi/glapi.c
@@ -198,6 +198,16 @@ PUBLIC const void *_glapi_Context = NULL;
 
 #if defined(THREADS)
 
+#ifdef WIN32_THREADS
+/* _glthread_DECLARE_STATIC_MUTEX is broken on windows.  There will be race! */
+#define CHECK_MULTITHREAD_LOCK()
+#define CHECK_MULTITHREAD_UNLOCK()
+#else
+_glthread_DECLARE_STATIC_MUTEX(ThreadCheckMutex);
+#define CHECK_MULTITHREAD_LOCK() _glthread_LOCK_MUTEX(ThreadCheckMutex)
+#define CHECK_MULTITHREAD_UNLOCK() _glthread_UNLOCK_MUTEX(ThreadCheckMutex)
+#endif
+
 static GLboolean ThreadSafe = GL_FALSE;  /**< In thread-safe mode? */
 _glthread_TSD _gl_DispatchTSD;           /**< Per-thread dispatch pointer */
 static _glthread_TSD ContextTSD;         /**< Per-thread context pointer */
@@ -231,23 +241,27 @@ void
 _glapi_check_multithread(void)
 {
 #if defined(THREADS) && !defined(GLX_USE_TLS)
-   if (!ThreadSafe) {
-      static unsigned long knownID;
-      static GLboolean firstCall = GL_TRUE;
-      if (firstCall) {
-         knownID = _glthread_GetID();
-         firstCall = GL_FALSE;
-      }
-      else if (knownID != _glthread_GetID()) {
-         ThreadSafe = GL_TRUE;
-         _glapi_set_dispatch(NULL);
-         _glapi_set_context(NULL);
-      }
+   static unsigned long knownID;
+   static GLboolean firstCall = GL_TRUE;
+
+   if (ThreadSafe)
+      return;
+
+   CHECK_MULTITHREAD_LOCK();
+   if (firstCall) {
+      /* initialize TSDs */
+      (void) _glthread_GetTSD(&ContextTSD);
+      (void) _glthread_GetTSD(&_gl_DispatchTSD);
+
+      knownID = _glthread_GetID();
+      firstCall = GL_FALSE;
    }
-   else if (!_glapi_get_dispatch()) {
-      /* make sure that this thread's dispatch pointer isn't null */
+   else if (knownID != _glthread_GetID()) {
+      ThreadSafe = GL_TRUE;
       _glapi_set_dispatch(NULL);
+      _glapi_set_context(NULL);
    }
+   CHECK_MULTITHREAD_UNLOCK();
 #endif
 }
 
diff --git a/src/mesa/glapi/glapi.h b/src/mesa/glapi/glapi.h
index 8f2cf66218..5fb5401229 100644
--- a/src/mesa/glapi/glapi.h
+++ b/src/mesa/glapi/glapi.h
@@ -80,8 +80,8 @@ typedef void (*_glapi_warning_func)(void *ctx, const char *str, ...);
  **/
 #if defined (GLX_USE_TLS)
 
-const extern void *_glapi_Context;
-const extern struct _glapi_table *_glapi_Dispatch;
+extern const void *_glapi_Context;
+extern const struct _glapi_table *_glapi_Dispatch;
 
 extern __thread void * _glapi_tls_Context
     __attribute__((tls_model("initial-exec")));
diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h
index 14f305f025..6d4ab09b07 100644
--- a/src/mesa/glapi/glapioffsets.h
+++ b/src/mesa/glapi/glapioffsets.h
@@ -594,220 +594,237 @@
 #define _gloffset_GetAttribLocationARB 559
 #define _gloffset_DrawBuffersARB 560
 #define _gloffset_RenderbufferStorageMultisample 561
-#define _gloffset_PolygonOffsetEXT 562
-#define _gloffset_GetPixelTexGenParameterfvSGIS 563
-#define _gloffset_GetPixelTexGenParameterivSGIS 564
-#define _gloffset_PixelTexGenParameterfSGIS 565
-#define _gloffset_PixelTexGenParameterfvSGIS 566
-#define _gloffset_PixelTexGenParameteriSGIS 567
-#define _gloffset_PixelTexGenParameterivSGIS 568
-#define _gloffset_SampleMaskSGIS 569
-#define _gloffset_SamplePatternSGIS 570
-#define _gloffset_ColorPointerEXT 571
-#define _gloffset_EdgeFlagPointerEXT 572
-#define _gloffset_IndexPointerEXT 573
-#define _gloffset_NormalPointerEXT 574
-#define _gloffset_TexCoordPointerEXT 575
-#define _gloffset_VertexPointerEXT 576
-#define _gloffset_PointParameterfEXT 577
-#define _gloffset_PointParameterfvEXT 578
-#define _gloffset_LockArraysEXT 579
-#define _gloffset_UnlockArraysEXT 580
-#define _gloffset_CullParameterdvEXT 581
-#define _gloffset_CullParameterfvEXT 582
-#define _gloffset_SecondaryColor3bEXT 583
-#define _gloffset_SecondaryColor3bvEXT 584
-#define _gloffset_SecondaryColor3dEXT 585
-#define _gloffset_SecondaryColor3dvEXT 586
-#define _gloffset_SecondaryColor3fEXT 587
-#define _gloffset_SecondaryColor3fvEXT 588
-#define _gloffset_SecondaryColor3iEXT 589
-#define _gloffset_SecondaryColor3ivEXT 590
-#define _gloffset_SecondaryColor3sEXT 591
-#define _gloffset_SecondaryColor3svEXT 592
-#define _gloffset_SecondaryColor3ubEXT 593
-#define _gloffset_SecondaryColor3ubvEXT 594
-#define _gloffset_SecondaryColor3uiEXT 595
-#define _gloffset_SecondaryColor3uivEXT 596
-#define _gloffset_SecondaryColor3usEXT 597
-#define _gloffset_SecondaryColor3usvEXT 598
-#define _gloffset_SecondaryColorPointerEXT 599
-#define _gloffset_MultiDrawArraysEXT 600
-#define _gloffset_MultiDrawElementsEXT 601
-#define _gloffset_FogCoordPointerEXT 602
-#define _gloffset_FogCoorddEXT 603
-#define _gloffset_FogCoorddvEXT 604
-#define _gloffset_FogCoordfEXT 605
-#define _gloffset_FogCoordfvEXT 606
-#define _gloffset_PixelTexGenSGIX 607
-#define _gloffset_BlendFuncSeparateEXT 608
-#define _gloffset_FlushVertexArrayRangeNV 609
-#define _gloffset_VertexArrayRangeNV 610
-#define _gloffset_CombinerInputNV 611
-#define _gloffset_CombinerOutputNV 612
-#define _gloffset_CombinerParameterfNV 613
-#define _gloffset_CombinerParameterfvNV 614
-#define _gloffset_CombinerParameteriNV 615
-#define _gloffset_CombinerParameterivNV 616
-#define _gloffset_FinalCombinerInputNV 617
-#define _gloffset_GetCombinerInputParameterfvNV 618
-#define _gloffset_GetCombinerInputParameterivNV 619
-#define _gloffset_GetCombinerOutputParameterfvNV 620
-#define _gloffset_GetCombinerOutputParameterivNV 621
-#define _gloffset_GetFinalCombinerInputParameterfvNV 622
-#define _gloffset_GetFinalCombinerInputParameterivNV 623
-#define _gloffset_ResizeBuffersMESA 624
-#define _gloffset_WindowPos2dMESA 625
-#define _gloffset_WindowPos2dvMESA 626
-#define _gloffset_WindowPos2fMESA 627
-#define _gloffset_WindowPos2fvMESA 628
-#define _gloffset_WindowPos2iMESA 629
-#define _gloffset_WindowPos2ivMESA 630
-#define _gloffset_WindowPos2sMESA 631
-#define _gloffset_WindowPos2svMESA 632
-#define _gloffset_WindowPos3dMESA 633
-#define _gloffset_WindowPos3dvMESA 634
-#define _gloffset_WindowPos3fMESA 635
-#define _gloffset_WindowPos3fvMESA 636
-#define _gloffset_WindowPos3iMESA 637
-#define _gloffset_WindowPos3ivMESA 638
-#define _gloffset_WindowPos3sMESA 639
-#define _gloffset_WindowPos3svMESA 640
-#define _gloffset_WindowPos4dMESA 641
-#define _gloffset_WindowPos4dvMESA 642
-#define _gloffset_WindowPos4fMESA 643
-#define _gloffset_WindowPos4fvMESA 644
-#define _gloffset_WindowPos4iMESA 645
-#define _gloffset_WindowPos4ivMESA 646
-#define _gloffset_WindowPos4sMESA 647
-#define _gloffset_WindowPos4svMESA 648
-#define _gloffset_MultiModeDrawArraysIBM 649
-#define _gloffset_MultiModeDrawElementsIBM 650
-#define _gloffset_DeleteFencesNV 651
-#define _gloffset_FinishFenceNV 652
-#define _gloffset_GenFencesNV 653
-#define _gloffset_GetFenceivNV 654
-#define _gloffset_IsFenceNV 655
-#define _gloffset_SetFenceNV 656
-#define _gloffset_TestFenceNV 657
-#define _gloffset_AreProgramsResidentNV 658
-#define _gloffset_BindProgramNV 659
-#define _gloffset_DeleteProgramsNV 660
-#define _gloffset_ExecuteProgramNV 661
-#define _gloffset_GenProgramsNV 662
-#define _gloffset_GetProgramParameterdvNV 663
-#define _gloffset_GetProgramParameterfvNV 664
-#define _gloffset_GetProgramStringNV 665
-#define _gloffset_GetProgramivNV 666
-#define _gloffset_GetTrackMatrixivNV 667
-#define _gloffset_GetVertexAttribPointervNV 668
-#define _gloffset_GetVertexAttribdvNV 669
-#define _gloffset_GetVertexAttribfvNV 670
-#define _gloffset_GetVertexAttribivNV 671
-#define _gloffset_IsProgramNV 672
-#define _gloffset_LoadProgramNV 673
-#define _gloffset_ProgramParameters4dvNV 674
-#define _gloffset_ProgramParameters4fvNV 675
-#define _gloffset_RequestResidentProgramsNV 676
-#define _gloffset_TrackMatrixNV 677
-#define _gloffset_VertexAttrib1dNV 678
-#define _gloffset_VertexAttrib1dvNV 679
-#define _gloffset_VertexAttrib1fNV 680
-#define _gloffset_VertexAttrib1fvNV 681
-#define _gloffset_VertexAttrib1sNV 682
-#define _gloffset_VertexAttrib1svNV 683
-#define _gloffset_VertexAttrib2dNV 684
-#define _gloffset_VertexAttrib2dvNV 685
-#define _gloffset_VertexAttrib2fNV 686
-#define _gloffset_VertexAttrib2fvNV 687
-#define _gloffset_VertexAttrib2sNV 688
-#define _gloffset_VertexAttrib2svNV 689
-#define _gloffset_VertexAttrib3dNV 690
-#define _gloffset_VertexAttrib3dvNV 691
-#define _gloffset_VertexAttrib3fNV 692
-#define _gloffset_VertexAttrib3fvNV 693
-#define _gloffset_VertexAttrib3sNV 694
-#define _gloffset_VertexAttrib3svNV 695
-#define _gloffset_VertexAttrib4dNV 696
-#define _gloffset_VertexAttrib4dvNV 697
-#define _gloffset_VertexAttrib4fNV 698
-#define _gloffset_VertexAttrib4fvNV 699
-#define _gloffset_VertexAttrib4sNV 700
-#define _gloffset_VertexAttrib4svNV 701
-#define _gloffset_VertexAttrib4ubNV 702
-#define _gloffset_VertexAttrib4ubvNV 703
-#define _gloffset_VertexAttribPointerNV 704
-#define _gloffset_VertexAttribs1dvNV 705
-#define _gloffset_VertexAttribs1fvNV 706
-#define _gloffset_VertexAttribs1svNV 707
-#define _gloffset_VertexAttribs2dvNV 708
-#define _gloffset_VertexAttribs2fvNV 709
-#define _gloffset_VertexAttribs2svNV 710
-#define _gloffset_VertexAttribs3dvNV 711
-#define _gloffset_VertexAttribs3fvNV 712
-#define _gloffset_VertexAttribs3svNV 713
-#define _gloffset_VertexAttribs4dvNV 714
-#define _gloffset_VertexAttribs4fvNV 715
-#define _gloffset_VertexAttribs4svNV 716
-#define _gloffset_VertexAttribs4ubvNV 717
-#define _gloffset_GetTexBumpParameterfvATI 718
-#define _gloffset_GetTexBumpParameterivATI 719
-#define _gloffset_TexBumpParameterfvATI 720
-#define _gloffset_TexBumpParameterivATI 721
-#define _gloffset_AlphaFragmentOp1ATI 722
-#define _gloffset_AlphaFragmentOp2ATI 723
-#define _gloffset_AlphaFragmentOp3ATI 724
-#define _gloffset_BeginFragmentShaderATI 725
-#define _gloffset_BindFragmentShaderATI 726
-#define _gloffset_ColorFragmentOp1ATI 727
-#define _gloffset_ColorFragmentOp2ATI 728
-#define _gloffset_ColorFragmentOp3ATI 729
-#define _gloffset_DeleteFragmentShaderATI 730
-#define _gloffset_EndFragmentShaderATI 731
-#define _gloffset_GenFragmentShadersATI 732
-#define _gloffset_PassTexCoordATI 733
-#define _gloffset_SampleMapATI 734
-#define _gloffset_SetFragmentShaderConstantATI 735
-#define _gloffset_PointParameteriNV 736
-#define _gloffset_PointParameterivNV 737
-#define _gloffset_ActiveStencilFaceEXT 738
-#define _gloffset_BindVertexArrayAPPLE 739
-#define _gloffset_DeleteVertexArraysAPPLE 740
-#define _gloffset_GenVertexArraysAPPLE 741
-#define _gloffset_IsVertexArrayAPPLE 742
-#define _gloffset_GetProgramNamedParameterdvNV 743
-#define _gloffset_GetProgramNamedParameterfvNV 744
-#define _gloffset_ProgramNamedParameter4dNV 745
-#define _gloffset_ProgramNamedParameter4dvNV 746
-#define _gloffset_ProgramNamedParameter4fNV 747
-#define _gloffset_ProgramNamedParameter4fvNV 748
-#define _gloffset_DepthBoundsEXT 749
-#define _gloffset_BlendEquationSeparateEXT 750
-#define _gloffset_BindFramebufferEXT 751
-#define _gloffset_BindRenderbufferEXT 752
-#define _gloffset_CheckFramebufferStatusEXT 753
-#define _gloffset_DeleteFramebuffersEXT 754
-#define _gloffset_DeleteRenderbuffersEXT 755
-#define _gloffset_FramebufferRenderbufferEXT 756
-#define _gloffset_FramebufferTexture1DEXT 757
-#define _gloffset_FramebufferTexture2DEXT 758
-#define _gloffset_FramebufferTexture3DEXT 759
-#define _gloffset_GenFramebuffersEXT 760
-#define _gloffset_GenRenderbuffersEXT 761
-#define _gloffset_GenerateMipmapEXT 762
-#define _gloffset_GetFramebufferAttachmentParameterivEXT 763
-#define _gloffset_GetRenderbufferParameterivEXT 764
-#define _gloffset_IsFramebufferEXT 765
-#define _gloffset_IsRenderbufferEXT 766
-#define _gloffset_RenderbufferStorageEXT 767
-#define _gloffset_BlitFramebufferEXT 768
-#define _gloffset_FramebufferTextureLayerEXT 769
-#define _gloffset_StencilFuncSeparateATI 770
-#define _gloffset_ProgramEnvParameters4fvEXT 771
-#define _gloffset_ProgramLocalParameters4fvEXT 772
-#define _gloffset_GetQueryObjecti64vEXT 773
-#define _gloffset_GetQueryObjectui64vEXT 774
-#define _gloffset_FIRST_DYNAMIC 775
+#define _gloffset_FlushMappedBufferRange 562
+#define _gloffset_MapBufferRange 563
+#define _gloffset_BindVertexArray 564
+#define _gloffset_GenVertexArrays 565
+#define _gloffset_CopyBufferSubData 566
+#define _gloffset_ClientWaitSync 567
+#define _gloffset_DeleteSync 568
+#define _gloffset_FenceSync 569
+#define _gloffset_GetInteger64v 570
+#define _gloffset_GetSynciv 571
+#define _gloffset_IsSync 572
+#define _gloffset_WaitSync 573
+#define _gloffset_PolygonOffsetEXT 574
+#define _gloffset_GetPixelTexGenParameterfvSGIS 575
+#define _gloffset_GetPixelTexGenParameterivSGIS 576
+#define _gloffset_PixelTexGenParameterfSGIS 577
+#define _gloffset_PixelTexGenParameterfvSGIS 578
+#define _gloffset_PixelTexGenParameteriSGIS 579
+#define _gloffset_PixelTexGenParameterivSGIS 580
+#define _gloffset_SampleMaskSGIS 581
+#define _gloffset_SamplePatternSGIS 582
+#define _gloffset_ColorPointerEXT 583
+#define _gloffset_EdgeFlagPointerEXT 584
+#define _gloffset_IndexPointerEXT 585
+#define _gloffset_NormalPointerEXT 586
+#define _gloffset_TexCoordPointerEXT 587
+#define _gloffset_VertexPointerEXT 588
+#define _gloffset_PointParameterfEXT 589
+#define _gloffset_PointParameterfvEXT 590
+#define _gloffset_LockArraysEXT 591
+#define _gloffset_UnlockArraysEXT 592
+#define _gloffset_CullParameterdvEXT 593
+#define _gloffset_CullParameterfvEXT 594
+#define _gloffset_SecondaryColor3bEXT 595
+#define _gloffset_SecondaryColor3bvEXT 596
+#define _gloffset_SecondaryColor3dEXT 597
+#define _gloffset_SecondaryColor3dvEXT 598
+#define _gloffset_SecondaryColor3fEXT 599
+#define _gloffset_SecondaryColor3fvEXT 600
+#define _gloffset_SecondaryColor3iEXT 601
+#define _gloffset_SecondaryColor3ivEXT 602
+#define _gloffset_SecondaryColor3sEXT 603
+#define _gloffset_SecondaryColor3svEXT 604
+#define _gloffset_SecondaryColor3ubEXT 605
+#define _gloffset_SecondaryColor3ubvEXT 606
+#define _gloffset_SecondaryColor3uiEXT 607
+#define _gloffset_SecondaryColor3uivEXT 608
+#define _gloffset_SecondaryColor3usEXT 609
+#define _gloffset_SecondaryColor3usvEXT 610
+#define _gloffset_SecondaryColorPointerEXT 611
+#define _gloffset_MultiDrawArraysEXT 612
+#define _gloffset_MultiDrawElementsEXT 613
+#define _gloffset_FogCoordPointerEXT 614
+#define _gloffset_FogCoorddEXT 615
+#define _gloffset_FogCoorddvEXT 616
+#define _gloffset_FogCoordfEXT 617
+#define _gloffset_FogCoordfvEXT 618
+#define _gloffset_PixelTexGenSGIX 619
+#define _gloffset_BlendFuncSeparateEXT 620
+#define _gloffset_FlushVertexArrayRangeNV 621
+#define _gloffset_VertexArrayRangeNV 622
+#define _gloffset_CombinerInputNV 623
+#define _gloffset_CombinerOutputNV 624
+#define _gloffset_CombinerParameterfNV 625
+#define _gloffset_CombinerParameterfvNV 626
+#define _gloffset_CombinerParameteriNV 627
+#define _gloffset_CombinerParameterivNV 628
+#define _gloffset_FinalCombinerInputNV 629
+#define _gloffset_GetCombinerInputParameterfvNV 630
+#define _gloffset_GetCombinerInputParameterivNV 631
+#define _gloffset_GetCombinerOutputParameterfvNV 632
+#define _gloffset_GetCombinerOutputParameterivNV 633
+#define _gloffset_GetFinalCombinerInputParameterfvNV 634
+#define _gloffset_GetFinalCombinerInputParameterivNV 635
+#define _gloffset_ResizeBuffersMESA 636
+#define _gloffset_WindowPos2dMESA 637
+#define _gloffset_WindowPos2dvMESA 638
+#define _gloffset_WindowPos2fMESA 639
+#define _gloffset_WindowPos2fvMESA 640
+#define _gloffset_WindowPos2iMESA 641
+#define _gloffset_WindowPos2ivMESA 642
+#define _gloffset_WindowPos2sMESA 643
+#define _gloffset_WindowPos2svMESA 644
+#define _gloffset_WindowPos3dMESA 645
+#define _gloffset_WindowPos3dvMESA 646
+#define _gloffset_WindowPos3fMESA 647
+#define _gloffset_WindowPos3fvMESA 648
+#define _gloffset_WindowPos3iMESA 649
+#define _gloffset_WindowPos3ivMESA 650
+#define _gloffset_WindowPos3sMESA 651
+#define _gloffset_WindowPos3svMESA 652
+#define _gloffset_WindowPos4dMESA 653
+#define _gloffset_WindowPos4dvMESA 654
+#define _gloffset_WindowPos4fMESA 655
+#define _gloffset_WindowPos4fvMESA 656
+#define _gloffset_WindowPos4iMESA 657
+#define _gloffset_WindowPos4ivMESA 658
+#define _gloffset_WindowPos4sMESA 659
+#define _gloffset_WindowPos4svMESA 660
+#define _gloffset_MultiModeDrawArraysIBM 661
+#define _gloffset_MultiModeDrawElementsIBM 662
+#define _gloffset_DeleteFencesNV 663
+#define _gloffset_FinishFenceNV 664
+#define _gloffset_GenFencesNV 665
+#define _gloffset_GetFenceivNV 666
+#define _gloffset_IsFenceNV 667
+#define _gloffset_SetFenceNV 668
+#define _gloffset_TestFenceNV 669
+#define _gloffset_AreProgramsResidentNV 670
+#define _gloffset_BindProgramNV 671
+#define _gloffset_DeleteProgramsNV 672
+#define _gloffset_ExecuteProgramNV 673
+#define _gloffset_GenProgramsNV 674
+#define _gloffset_GetProgramParameterdvNV 675
+#define _gloffset_GetProgramParameterfvNV 676
+#define _gloffset_GetProgramStringNV 677
+#define _gloffset_GetProgramivNV 678
+#define _gloffset_GetTrackMatrixivNV 679
+#define _gloffset_GetVertexAttribPointervNV 680
+#define _gloffset_GetVertexAttribdvNV 681
+#define _gloffset_GetVertexAttribfvNV 682
+#define _gloffset_GetVertexAttribivNV 683
+#define _gloffset_IsProgramNV 684
+#define _gloffset_LoadProgramNV 685
+#define _gloffset_ProgramParameters4dvNV 686
+#define _gloffset_ProgramParameters4fvNV 687
+#define _gloffset_RequestResidentProgramsNV 688
+#define _gloffset_TrackMatrixNV 689
+#define _gloffset_VertexAttrib1dNV 690
+#define _gloffset_VertexAttrib1dvNV 691
+#define _gloffset_VertexAttrib1fNV 692
+#define _gloffset_VertexAttrib1fvNV 693
+#define _gloffset_VertexAttrib1sNV 694
+#define _gloffset_VertexAttrib1svNV 695
+#define _gloffset_VertexAttrib2dNV 696
+#define _gloffset_VertexAttrib2dvNV 697
+#define _gloffset_VertexAttrib2fNV 698
+#define _gloffset_VertexAttrib2fvNV 699
+#define _gloffset_VertexAttrib2sNV 700
+#define _gloffset_VertexAttrib2svNV 701
+#define _gloffset_VertexAttrib3dNV 702
+#define _gloffset_VertexAttrib3dvNV 703
+#define _gloffset_VertexAttrib3fNV 704
+#define _gloffset_VertexAttrib3fvNV 705
+#define _gloffset_VertexAttrib3sNV 706
+#define _gloffset_VertexAttrib3svNV 707
+#define _gloffset_VertexAttrib4dNV 708
+#define _gloffset_VertexAttrib4dvNV 709
+#define _gloffset_VertexAttrib4fNV 710
+#define _gloffset_VertexAttrib4fvNV 711
+#define _gloffset_VertexAttrib4sNV 712
+#define _gloffset_VertexAttrib4svNV 713
+#define _gloffset_VertexAttrib4ubNV 714
+#define _gloffset_VertexAttrib4ubvNV 715
+#define _gloffset_VertexAttribPointerNV 716
+#define _gloffset_VertexAttribs1dvNV 717
+#define _gloffset_VertexAttribs1fvNV 718
+#define _gloffset_VertexAttribs1svNV 719
+#define _gloffset_VertexAttribs2dvNV 720
+#define _gloffset_VertexAttribs2fvNV 721
+#define _gloffset_VertexAttribs2svNV 722
+#define _gloffset_VertexAttribs3dvNV 723
+#define _gloffset_VertexAttribs3fvNV 724
+#define _gloffset_VertexAttribs3svNV 725
+#define _gloffset_VertexAttribs4dvNV 726
+#define _gloffset_VertexAttribs4fvNV 727
+#define _gloffset_VertexAttribs4svNV 728
+#define _gloffset_VertexAttribs4ubvNV 729
+#define _gloffset_GetTexBumpParameterfvATI 730
+#define _gloffset_GetTexBumpParameterivATI 731
+#define _gloffset_TexBumpParameterfvATI 732
+#define _gloffset_TexBumpParameterivATI 733
+#define _gloffset_AlphaFragmentOp1ATI 734
+#define _gloffset_AlphaFragmentOp2ATI 735
+#define _gloffset_AlphaFragmentOp3ATI 736
+#define _gloffset_BeginFragmentShaderATI 737
+#define _gloffset_BindFragmentShaderATI 738
+#define _gloffset_ColorFragmentOp1ATI 739
+#define _gloffset_ColorFragmentOp2ATI 740
+#define _gloffset_ColorFragmentOp3ATI 741
+#define _gloffset_DeleteFragmentShaderATI 742
+#define _gloffset_EndFragmentShaderATI 743
+#define _gloffset_GenFragmentShadersATI 744
+#define _gloffset_PassTexCoordATI 745
+#define _gloffset_SampleMapATI 746
+#define _gloffset_SetFragmentShaderConstantATI 747
+#define _gloffset_PointParameteriNV 748
+#define _gloffset_PointParameterivNV 749
+#define _gloffset_ActiveStencilFaceEXT 750
+#define _gloffset_BindVertexArrayAPPLE 751
+#define _gloffset_DeleteVertexArraysAPPLE 752
+#define _gloffset_GenVertexArraysAPPLE 753
+#define _gloffset_IsVertexArrayAPPLE 754
+#define _gloffset_GetProgramNamedParameterdvNV 755
+#define _gloffset_GetProgramNamedParameterfvNV 756
+#define _gloffset_ProgramNamedParameter4dNV 757
+#define _gloffset_ProgramNamedParameter4dvNV 758
+#define _gloffset_ProgramNamedParameter4fNV 759
+#define _gloffset_ProgramNamedParameter4fvNV 760
+#define _gloffset_DepthBoundsEXT 761
+#define _gloffset_BlendEquationSeparateEXT 762
+#define _gloffset_BindFramebufferEXT 763
+#define _gloffset_BindRenderbufferEXT 764
+#define _gloffset_CheckFramebufferStatusEXT 765
+#define _gloffset_DeleteFramebuffersEXT 766
+#define _gloffset_DeleteRenderbuffersEXT 767
+#define _gloffset_FramebufferRenderbufferEXT 768
+#define _gloffset_FramebufferTexture1DEXT 769
+#define _gloffset_FramebufferTexture2DEXT 770
+#define _gloffset_FramebufferTexture3DEXT 771
+#define _gloffset_GenFramebuffersEXT 772
+#define _gloffset_GenRenderbuffersEXT 773
+#define _gloffset_GenerateMipmapEXT 774
+#define _gloffset_GetFramebufferAttachmentParameterivEXT 775
+#define _gloffset_GetRenderbufferParameterivEXT 776
+#define _gloffset_IsFramebufferEXT 777
+#define _gloffset_IsRenderbufferEXT 778
+#define _gloffset_RenderbufferStorageEXT 779
+#define _gloffset_BlitFramebufferEXT 780
+#define _gloffset_BufferParameteriAPPLE 781
+#define _gloffset_FlushMappedBufferRangeAPPLE 782
+#define _gloffset_FramebufferTextureLayerEXT 783
+#define _gloffset_ProvokingVertexEXT 784
+#define _gloffset_GetTexParameterPointervAPPLE 785
+#define _gloffset_TextureRangeAPPLE 786
+#define _gloffset_StencilFuncSeparateATI 787
+#define _gloffset_ProgramEnvParameters4fvEXT 788
+#define _gloffset_ProgramLocalParameters4fvEXT 789
+#define _gloffset_GetQueryObjecti64vEXT 790
+#define _gloffset_GetQueryObjectui64vEXT 791
+#define _gloffset_FIRST_DYNAMIC 792
 
 #else
 
@@ -965,6 +982,18 @@
 #define _gloffset_GetAttribLocationARB driDispatchRemapTable[GetAttribLocationARB_remap_index]
 #define _gloffset_DrawBuffersARB driDispatchRemapTable[DrawBuffersARB_remap_index]
 #define _gloffset_RenderbufferStorageMultisample driDispatchRemapTable[RenderbufferStorageMultisample_remap_index]
+#define _gloffset_FlushMappedBufferRange driDispatchRemapTable[FlushMappedBufferRange_remap_index]
+#define _gloffset_MapBufferRange driDispatchRemapTable[MapBufferRange_remap_index]
+#define _gloffset_BindVertexArray driDispatchRemapTable[BindVertexArray_remap_index]
+#define _gloffset_GenVertexArrays driDispatchRemapTable[GenVertexArrays_remap_index]
+#define _gloffset_CopyBufferSubData driDispatchRemapTable[CopyBufferSubData_remap_index]
+#define _gloffset_ClientWaitSync driDispatchRemapTable[ClientWaitSync_remap_index]
+#define _gloffset_DeleteSync driDispatchRemapTable[DeleteSync_remap_index]
+#define _gloffset_FenceSync driDispatchRemapTable[FenceSync_remap_index]
+#define _gloffset_GetInteger64v driDispatchRemapTable[GetInteger64v_remap_index]
+#define _gloffset_GetSynciv driDispatchRemapTable[GetSynciv_remap_index]
+#define _gloffset_IsSync driDispatchRemapTable[IsSync_remap_index]
+#define _gloffset_WaitSync driDispatchRemapTable[WaitSync_remap_index]
 #define _gloffset_PolygonOffsetEXT driDispatchRemapTable[PolygonOffsetEXT_remap_index]
 #define _gloffset_GetPixelTexGenParameterfvSGIS driDispatchRemapTable[GetPixelTexGenParameterfvSGIS_remap_index]
 #define _gloffset_GetPixelTexGenParameterivSGIS driDispatchRemapTable[GetPixelTexGenParameterivSGIS_remap_index]
@@ -1172,7 +1201,12 @@
 #define _gloffset_IsRenderbufferEXT driDispatchRemapTable[IsRenderbufferEXT_remap_index]
 #define _gloffset_RenderbufferStorageEXT driDispatchRemapTable[RenderbufferStorageEXT_remap_index]
 #define _gloffset_BlitFramebufferEXT driDispatchRemapTable[BlitFramebufferEXT_remap_index]
+#define _gloffset_BufferParameteriAPPLE driDispatchRemapTable[BufferParameteriAPPLE_remap_index]
+#define _gloffset_FlushMappedBufferRangeAPPLE driDispatchRemapTable[FlushMappedBufferRangeAPPLE_remap_index]
 #define _gloffset_FramebufferTextureLayerEXT driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index]
+#define _gloffset_ProvokingVertexEXT driDispatchRemapTable[ProvokingVertexEXT_remap_index]
+#define _gloffset_GetTexParameterPointervAPPLE driDispatchRemapTable[GetTexParameterPointervAPPLE_remap_index]
+#define _gloffset_TextureRangeAPPLE driDispatchRemapTable[TextureRangeAPPLE_remap_index]
 #define _gloffset_StencilFuncSeparateATI driDispatchRemapTable[StencilFuncSeparateATI_remap_index]
 #define _gloffset_ProgramEnvParameters4fvEXT driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index]
 #define _gloffset_ProgramLocalParameters4fvEXT driDispatchRemapTable[ProgramLocalParameters4fvEXT_remap_index]
diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h
index 9ecb036461..855fcaa7fe 100644
--- a/src/mesa/glapi/glapitable.h
+++ b/src/mesa/glapi/glapitable.h
@@ -602,219 +602,236 @@ struct _glapi_table
    GLint (GLAPIENTRYP GetAttribLocationARB)(GLhandleARB program, const GLcharARB * name); /* 559 */
    void (GLAPIENTRYP DrawBuffersARB)(GLsizei n, const GLenum * bufs); /* 560 */
    void (GLAPIENTRYP RenderbufferStorageMultisample)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height); /* 561 */
-   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 562 */
-   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 563 */
-   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 564 */
-   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 565 */
-   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 566 */
-   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 567 */
-   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 568 */
-   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 569 */
-   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 570 */
-   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 571 */
-   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 572 */
-   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 573 */
-   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 574 */
-   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 575 */
-   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 576 */
-   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 577 */
-   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 578 */
-   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 579 */
-   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 580 */
-   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 581 */
-   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 582 */
-   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 583 */
-   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 584 */
-   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 585 */
-   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 586 */
-   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 587 */
-   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 588 */
-   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 589 */
-   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 590 */
-   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 591 */
-   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 592 */
-   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 593 */
-   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 594 */
-   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 595 */
-   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 596 */
-   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 597 */
-   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 598 */
-   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 599 */
-   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 600 */
-   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 601 */
-   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 602 */
-   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 603 */
-   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 604 */
-   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 605 */
-   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 606 */
-   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 607 */
-   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 608 */
-   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 609 */
-   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 610 */
-   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 611 */
-   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 612 */
-   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 613 */
-   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 614 */
-   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 615 */
-   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 616 */
-   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 617 */
-   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 618 */
-   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 619 */
-   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 620 */
-   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 621 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 622 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 623 */
-   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 624 */
-   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 625 */
-   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 626 */
-   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 627 */
-   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 628 */
-   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 629 */
-   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 630 */
-   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 631 */
-   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 632 */
-   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 633 */
-   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 634 */
-   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 635 */
-   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 636 */
-   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 637 */
-   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 638 */
-   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 639 */
-   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 640 */
-   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 641 */
-   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 642 */
-   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 643 */
-   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 644 */
-   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 645 */
-   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 646 */
-   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 647 */
-   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 648 */
-   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 649 */
-   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 650 */
-   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 651 */
-   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 652 */
-   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 653 */
-   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 654 */
-   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 655 */
-   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 656 */
-   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 657 */
-   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 658 */
-   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 659 */
-   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 660 */
-   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 661 */
-   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 662 */
-   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 663 */
-   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 664 */
-   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 665 */
-   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 666 */
-   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 667 */
-   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 668 */
-   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 669 */
-   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 670 */
-   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 671 */
-   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 672 */
-   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 673 */
-   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 674 */
-   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 675 */
-   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 676 */
-   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 677 */
-   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 678 */
-   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 679 */
-   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 680 */
-   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 681 */
-   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 682 */
-   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 683 */
-   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 684 */
-   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 685 */
-   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 686 */
-   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 687 */
-   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 688 */
-   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 689 */
-   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 690 */
-   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 691 */
-   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 692 */
-   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 693 */
-   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 694 */
-   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 695 */
-   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 696 */
-   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 697 */
-   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 698 */
-   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 699 */
-   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 700 */
-   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 701 */
-   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 702 */
-   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 703 */
-   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 704 */
-   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 705 */
-   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 706 */
-   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 707 */
-   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 708 */
-   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 709 */
-   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 710 */
-   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 711 */
-   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 712 */
-   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 713 */
-   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 714 */
-   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 715 */
-   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 716 */
-   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 717 */
-   void (GLAPIENTRYP GetTexBumpParameterfvATI)(GLenum pname, GLfloat * param); /* 718 */
-   void (GLAPIENTRYP GetTexBumpParameterivATI)(GLenum pname, GLint * param); /* 719 */
-   void (GLAPIENTRYP TexBumpParameterfvATI)(GLenum pname, const GLfloat * param); /* 720 */
-   void (GLAPIENTRYP TexBumpParameterivATI)(GLenum pname, const GLint * param); /* 721 */
-   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 722 */
-   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 723 */
-   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 724 */
-   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 725 */
-   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 726 */
-   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 727 */
-   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 728 */
-   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 729 */
-   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 730 */
-   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 731 */
-   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 732 */
-   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 733 */
-   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 734 */
-   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 735 */
-   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 736 */
-   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 737 */
-   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 738 */
-   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 739 */
-   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 740 */
-   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 741 */
-   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 742 */
-   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 743 */
-   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 744 */
-   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 745 */
-   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 746 */
-   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 747 */
-   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 748 */
-   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 749 */
-   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 750 */
-   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 751 */
-   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 752 */
-   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 753 */
-   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 754 */
-   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 755 */
-   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 756 */
-   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 757 */
-   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 758 */
-   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 759 */
-   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 760 */
-   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 761 */
-   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 762 */
-   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 763 */
-   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 764 */
-   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 765 */
-   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 766 */
-   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 767 */
-   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 768 */
-   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 769 */
-   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 770 */
-   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */
-   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 772 */
-   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 773 */
-   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 774 */
+   void (GLAPIENTRYP FlushMappedBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length); /* 562 */
+   GLvoid * (GLAPIENTRYP MapBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access); /* 563 */
+   void (GLAPIENTRYP BindVertexArray)(GLuint array); /* 564 */
+   void (GLAPIENTRYP GenVertexArrays)(GLsizei n, GLuint * arrays); /* 565 */
+   void (GLAPIENTRYP CopyBufferSubData)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size); /* 566 */
+   GLenum (GLAPIENTRYP ClientWaitSync)(GLsync sync, GLbitfield flags, GLuint64 timeout); /* 567 */
+   void (GLAPIENTRYP DeleteSync)(GLsync sync); /* 568 */
+   GLsync (GLAPIENTRYP FenceSync)(GLenum condition, GLbitfield flags); /* 569 */
+   void (GLAPIENTRYP GetInteger64v)(GLenum pname, GLint64 * params); /* 570 */
+   void (GLAPIENTRYP GetSynciv)(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei * length, GLint * values); /* 571 */
+   GLboolean (GLAPIENTRYP IsSync)(GLsync sync); /* 572 */
+   void (GLAPIENTRYP WaitSync)(GLsync sync, GLbitfield flags, GLuint64 timeout); /* 573 */
+   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 574 */
+   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 575 */
+   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 576 */
+   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 577 */
+   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 578 */
+   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 579 */
+   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 580 */
+   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 581 */
+   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 582 */
+   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 583 */
+   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 584 */
+   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 585 */
+   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 586 */
+   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 587 */
+   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 588 */
+   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 589 */
+   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 590 */
+   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 591 */
+   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 592 */
+   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 593 */
+   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 594 */
+   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 595 */
+   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 596 */
+   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 597 */
+   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 598 */
+   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 599 */
+   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 600 */
+   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 601 */
+   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 602 */
+   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 603 */
+   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 604 */
+   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 605 */
+   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 606 */
+   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 607 */
+   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 608 */
+   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 609 */
+   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 610 */
+   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 611 */
+   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 612 */
+   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 613 */
+   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 614 */
+   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 615 */
+   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 616 */
+   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 617 */
+   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 618 */
+   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 619 */
+   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 620 */
+   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 621 */
+   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 622 */
+   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 623 */
+   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 624 */
+   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 625 */
+   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 626 */
+   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 627 */
+   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 628 */
+   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 629 */
+   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 630 */
+   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 631 */
+   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 632 */
+   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 633 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 634 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 635 */
+   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 636 */
+   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 637 */
+   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 638 */
+   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 639 */
+   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 640 */
+   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 641 */
+   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 642 */
+   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 643 */
+   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 644 */
+   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 645 */
+   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 646 */
+   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 647 */
+   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 648 */
+   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 649 */
+   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 650 */
+   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 651 */
+   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 652 */
+   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 653 */
+   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 654 */
+   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 655 */
+   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 656 */
+   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 657 */
+   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 658 */
+   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 659 */
+   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 660 */
+   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 661 */
+   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 662 */
+   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 663 */
+   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 664 */
+   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 665 */
+   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 666 */
+   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 667 */
+   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 668 */
+   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 669 */
+   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 670 */
+   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 671 */
+   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 672 */
+   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 673 */
+   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 674 */
+   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 675 */
+   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 676 */
+   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 677 */
+   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 678 */
+   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 679 */
+   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 680 */
+   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 681 */
+   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 682 */
+   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 683 */
+   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 684 */
+   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 685 */
+   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 686 */
+   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 687 */
+   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 688 */
+   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 689 */
+   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 690 */
+   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 691 */
+   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 692 */
+   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 693 */
+   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 694 */
+   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 695 */
+   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 696 */
+   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 697 */
+   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 698 */
+   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 699 */
+   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 700 */
+   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 701 */
+   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 702 */
+   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 703 */
+   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 704 */
+   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 705 */
+   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 706 */
+   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 707 */
+   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 708 */
+   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 709 */
+   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 710 */
+   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 711 */
+   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 712 */
+   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 713 */
+   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 714 */
+   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 715 */
+   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 716 */
+   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 717 */
+   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 718 */
+   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 719 */
+   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 720 */
+   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 721 */
+   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 722 */
+   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 723 */
+   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 724 */
+   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 725 */
+   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 726 */
+   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 727 */
+   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 728 */
+   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 729 */
+   void (GLAPIENTRYP GetTexBumpParameterfvATI)(GLenum pname, GLfloat * param); /* 730 */
+   void (GLAPIENTRYP GetTexBumpParameterivATI)(GLenum pname, GLint * param); /* 731 */
+   void (GLAPIENTRYP TexBumpParameterfvATI)(GLenum pname, const GLfloat * param); /* 732 */
+   void (GLAPIENTRYP TexBumpParameterivATI)(GLenum pname, const GLint * param); /* 733 */
+   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 734 */
+   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 735 */
+   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 736 */
+   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 737 */
+   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 738 */
+   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 739 */
+   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 740 */
+   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 741 */
+   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 742 */
+   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 743 */
+   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 744 */
+   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 745 */
+   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 746 */
+   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 747 */
+   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 748 */
+   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 749 */
+   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 750 */
+   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 751 */
+   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 752 */
+   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 753 */
+   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 754 */
+   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 755 */
+   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 756 */
+   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 757 */
+   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 758 */
+   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 759 */
+   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 760 */
+   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 761 */
+   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 762 */
+   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 763 */
+   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 764 */
+   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 765 */
+   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 766 */
+   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 767 */
+   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 768 */
+   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 769 */
+   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 770 */
+   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 771 */
+   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 772 */
+   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 773 */
+   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 774 */
+   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 775 */
+   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 776 */
+   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 777 */
+   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 778 */
+   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 779 */
+   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 780 */
+   void (GLAPIENTRYP BufferParameteriAPPLE)(GLenum target, GLenum pname, GLint param); /* 781 */
+   void (GLAPIENTRYP FlushMappedBufferRangeAPPLE)(GLenum target, GLintptr offset, GLsizeiptr size); /* 782 */
+   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 783 */
+   void (GLAPIENTRYP ProvokingVertexEXT)(GLenum mode); /* 784 */
+   void (GLAPIENTRYP GetTexParameterPointervAPPLE)(GLenum target, GLenum pname, GLvoid ** params); /* 785 */
+   void (GLAPIENTRYP TextureRangeAPPLE)(GLenum target, GLsizei length, GLvoid * pointer); /* 786 */
+   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 787 */
+   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 788 */
+   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 789 */
+   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 790 */
+   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 791 */
 };
 
 #endif /* !defined( _GLAPI_TABLE_H_ ) */
diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h
index 7ccd9707c3..e08f9548f5 100644
--- a/src/mesa/glapi/glapitemp.h
+++ b/src/mesa/glapi/glapitemp.h
@@ -4011,63 +4011,123 @@ KEYWORD1 void KEYWORD2 NAME(RenderbufferStorageMultisample)(GLenum target, GLsiz
    DISPATCH(RenderbufferStorageMultisample, (target, samples, internalformat, width, height), (F, "glRenderbufferStorageMultisample(0x%x, %d, 0x%x, %d, %d);\n", target, samples, internalformat, width, height));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FlushMappedBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length)
+{
+   DISPATCH(FlushMappedBufferRange, (target, offset, length), (F, "glFlushMappedBufferRange(0x%x, %d, %d);\n", target, offset, length));
+}
+
+KEYWORD1 GLvoid * KEYWORD2 NAME(MapBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access)
+{
+   RETURN_DISPATCH(MapBufferRange, (target, offset, length, access), (F, "glMapBufferRange(0x%x, %d, %d, %d);\n", target, offset, length, access));
+}
+
+KEYWORD1 void KEYWORD2 NAME(BindVertexArray)(GLuint array)
+{
+   DISPATCH(BindVertexArray, (array), (F, "glBindVertexArray(%d);\n", array));
+}
+
+KEYWORD1 void KEYWORD2 NAME(GenVertexArrays)(GLsizei n, GLuint * arrays)
+{
+   DISPATCH(GenVertexArrays, (n, arrays), (F, "glGenVertexArrays(%d, %p);\n", n, (const void *) arrays));
+}
+
+KEYWORD1 void KEYWORD2 NAME(CopyBufferSubData)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size)
+{
+   DISPATCH(CopyBufferSubData, (readTarget, writeTarget, readOffset, writeOffset, size), (F, "glCopyBufferSubData(0x%x, 0x%x, %d, %d, %d);\n", readTarget, writeTarget, readOffset, writeOffset, size));
+}
+
+KEYWORD1 GLenum KEYWORD2 NAME(ClientWaitSync)(GLsync sync, GLbitfield flags, GLuint64 timeout)
+{
+   RETURN_DISPATCH(ClientWaitSync, (sync, flags, timeout), (F, "glClientWaitSync(%d, %d, %d);\n", sync, flags, timeout));
+}
+
+KEYWORD1 void KEYWORD2 NAME(DeleteSync)(GLsync sync)
+{
+   DISPATCH(DeleteSync, (sync), (F, "glDeleteSync(%d);\n", sync));
+}
+
+KEYWORD1 GLsync KEYWORD2 NAME(FenceSync)(GLenum condition, GLbitfield flags)
+{
+   RETURN_DISPATCH(FenceSync, (condition, flags), (F, "glFenceSync(0x%x, %d);\n", condition, flags));
+}
+
+KEYWORD1 void KEYWORD2 NAME(GetInteger64v)(GLenum pname, GLint64 * params)
+{
+   DISPATCH(GetInteger64v, (pname, params), (F, "glGetInteger64v(0x%x, %p);\n", pname, (const void *) params));
+}
+
+KEYWORD1 void KEYWORD2 NAME(GetSynciv)(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei * length, GLint * values)
+{
+   DISPATCH(GetSynciv, (sync, pname, bufSize, length, values), (F, "glGetSynciv(%d, 0x%x, %d, %p, %p);\n", sync, pname, bufSize, (const void *) length, (const void *) values));
+}
+
+KEYWORD1 GLboolean KEYWORD2 NAME(IsSync)(GLsync sync)
+{
+   RETURN_DISPATCH(IsSync, (sync), (F, "glIsSync(%d);\n", sync));
+}
+
+KEYWORD1 void KEYWORD2 NAME(WaitSync)(GLsync sync, GLbitfield flags, GLuint64 timeout)
+{
+   DISPATCH(WaitSync, (sync, flags, timeout), (F, "glWaitSync(%d, %d, %d);\n", sync, flags, timeout));
+}
+
 KEYWORD1 void KEYWORD2 NAME(PolygonOffsetEXT)(GLfloat factor, GLfloat bias)
 {
    DISPATCH(PolygonOffsetEXT, (factor, bias), (F, "glPolygonOffsetEXT(%f, %f);\n", factor, bias));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_575)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_575)(GLenum pname, GLfloat * params)
 {
    DISPATCH(GetPixelTexGenParameterfvSGIS, (pname, params), (F, "glGetPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_576)(GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_576)(GLenum pname, GLint * params)
 {
    DISPATCH(GetPixelTexGenParameterivSGIS, (pname, params), (F, "glGetPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param)
 {
    DISPATCH(PixelTexGenParameterfSGIS, (pname, param), (F, "glPixelTexGenParameterfSGIS(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PixelTexGenParameterfvSGIS, (pname, params), (F, "glPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_579)(GLenum pname, GLint param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_579)(GLenum pname, GLint param)
 {
    DISPATCH(PixelTexGenParameteriSGIS, (pname, param), (F, "glPixelTexGenParameteriSGIS(0x%x, %d);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_580)(GLenum pname, const GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_580)(GLenum pname, const GLint * params)
 {
    DISPATCH(PixelTexGenParameterivSGIS, (pname, params), (F, "glPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLclampf value, GLboolean invert);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLclampf value, GLboolean invert)
 {
    DISPATCH(SampleMaskSGIS, (value, invert), (F, "glSampleMaskSGIS(%f, %d);\n", value, invert));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pattern);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pattern)
 {
    DISPATCH(SamplePatternSGIS, (pattern), (F, "glSamplePatternSGIS(0x%x);\n", pattern));
 }
@@ -4117,9 +4177,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfEXT)(GLenum pname, GLfloat param)
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfEXT(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_589)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_589)(GLenum pname, GLfloat param)
 {
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfSGIS(0x%x, %f);\n", pname, param));
 }
@@ -4139,9 +4199,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfvEXT)(GLenum pname, const GLfloat * p
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_590)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_590)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4156,16 +4216,16 @@ KEYWORD1 void KEYWORD2 NAME(UnlockArraysEXT)(void)
    DISPATCH(UnlockArraysEXT, (), (F, "glUnlockArraysEXT();\n"));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_593)(GLenum pname, GLdouble * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_593)(GLenum pname, GLdouble * params)
 {
    DISPATCH(CullParameterdvEXT, (pname, params), (F, "glCullParameterdvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_594)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_594)(GLenum pname, GLfloat * params)
 {
    DISPATCH(CullParameterfvEXT, (pname, params), (F, "glCullParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4410,9 +4470,9 @@ KEYWORD1 void KEYWORD2 NAME(FogCoordfvEXT)(const GLfloat * coord)
    DISPATCH(FogCoordfvEXT, (coord), (F, "glFogCoordfvEXT(%p);\n", (const void *) coord));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_619)(GLenum mode);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_619)(GLenum mode)
 {
    DISPATCH(PixelTexGenSGIX, (mode), (F, "glPixelTexGenSGIX(0x%x);\n", mode));
 }
@@ -4427,9 +4487,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfac
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateEXT(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_620)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_620)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
 {
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateINGR(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
@@ -4794,65 +4854,65 @@ KEYWORD1 void KEYWORD2 NAME(WindowPos4svMESA)(const GLshort * v)
    DISPATCH(WindowPos4svMESA, (v), (F, "glWindowPos4svMESA(%p);\n", (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_661)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_661)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawArraysIBM, (mode, first, count, primcount, modestride), (F, "glMultiModeDrawArraysIBM(%p, %p, %p, %d, %d);\n", (const void *) mode, (const void *) first, (const void *) count, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_662)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_662)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawElementsIBM, (mode, count, type, indices, primcount, modestride), (F, "glMultiModeDrawElementsIBM(%p, %p, 0x%x, %p, %d, %d);\n", (const void *) mode, (const void *) count, type, (const void *) indices, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_663)(GLsizei n, const GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_663)(GLsizei n, const GLuint * fences)
 {
    DISPATCH(DeleteFencesNV, (n, fences), (F, "glDeleteFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_664)(GLuint fence);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_664)(GLuint fence)
 {
    DISPATCH(FinishFenceNV, (fence), (F, "glFinishFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_665)(GLsizei n, GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_665)(GLsizei n, GLuint * fences)
 {
    DISPATCH(GenFencesNV, (n, fences), (F, "glGenFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_666)(GLuint fence, GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_666)(GLuint fence, GLenum pname, GLint * params)
 {
    DISPATCH(GetFenceivNV, (fence, pname, params), (F, "glGetFenceivNV(%d, 0x%x, %p);\n", fence, pname, (const void *) params));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_667)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_667)(GLuint fence)
 {
    RETURN_DISPATCH(IsFenceNV, (fence), (F, "glIsFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_668)(GLuint fence, GLenum condition);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_668)(GLuint fence, GLenum condition)
 {
    DISPATCH(SetFenceNV, (fence, condition), (F, "glSetFenceNV(%d, 0x%x);\n", fence, condition));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_669)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_669)(GLuint fence)
 {
    RETURN_DISPATCH(TestFenceNV, (fence), (F, "glTestFenceNV(%d);\n", fence));
 }
@@ -5297,37 +5357,47 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterivNV)(GLenum pname, const GLint * para
    DISPATCH(PointParameterivNV, (pname, params), (F, "glPointParameterivNV(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLenum face);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum face);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLenum face)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum face)
 {
    DISPATCH(ActiveStencilFaceEXT, (face), (F, "glActiveStencilFaceEXT(0x%x);\n", face));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLuint array);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_751)(GLuint array);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLuint array)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_751)(GLuint array)
 {
    DISPATCH(BindVertexArrayAPPLE, (array), (F, "glBindVertexArrayAPPLE(%d);\n", array));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, const GLuint * arrays);
+KEYWORD1 void KEYWORD2 NAME(DeleteVertexArrays)(GLsizei n, const GLuint * arrays)
+{
+   DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArrays(%d, %p);\n", n, (const void *) arrays));
+}
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, const GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_752)(GLsizei n, const GLuint * arrays);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_752)(GLsizei n, const GLuint * arrays)
 {
    DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_741)(GLsizei n, GLuint * arrays);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_753)(GLsizei n, GLuint * arrays);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_741)(GLsizei n, GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_753)(GLsizei n, GLuint * arrays)
 {
    DISPATCH(GenVertexArraysAPPLE, (n, arrays), (F, "glGenVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_742)(GLuint array);
+KEYWORD1 GLboolean KEYWORD2 NAME(IsVertexArray)(GLuint array)
+{
+   RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArray(%d);\n", array));
+}
+
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_754)(GLuint array);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_742)(GLuint array)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_754)(GLuint array)
 {
    RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArrayAPPLE(%d);\n", array));
 }
@@ -5362,9 +5432,9 @@ KEYWORD1 void KEYWORD2 NAME(ProgramNamedParameter4fvNV)(GLuint id, GLsizei len,
    DISPATCH(ProgramNamedParameter4fvNV, (id, len, name, v), (F, "glProgramNamedParameter4fvNV(%d, %d, %p, %p);\n", id, len, (const void *) name, (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLclampd zmin, GLclampd zmax);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_761)(GLclampd zmin, GLclampd zmax);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLclampd zmin, GLclampd zmax)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_761)(GLclampd zmin, GLclampd zmax)
 {
    DISPATCH(DepthBoundsEXT, (zmin, zmax), (F, "glDepthBoundsEXT(%f, %f);\n", zmin, zmax));
 }
@@ -5374,9 +5444,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendEquationSeparate)(GLenum modeRGB, GLenum modeA)
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparate(0x%x, 0x%x);\n", modeRGB, modeA));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum modeRGB, GLenum modeA);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_762)(GLenum modeRGB, GLenum modeA);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum modeRGB, GLenum modeA)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_762)(GLenum modeRGB, GLenum modeA)
 {
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparateEXT(0x%x, 0x%x);\n", modeRGB, modeA));
 }
@@ -5556,13 +5626,27 @@ KEYWORD1 void KEYWORD2 NAME(BlitFramebuffer)(GLint srcX0, GLint srcY0, GLint src
    DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebuffer(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_780)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_780)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
 {
    DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebufferEXT(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
 }
 
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_781)(GLenum target, GLenum pname, GLint param);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_781)(GLenum target, GLenum pname, GLint param)
+{
+   DISPATCH(BufferParameteriAPPLE, (target, pname, param), (F, "glBufferParameteriAPPLE(0x%x, 0x%x, %d);\n", target, pname, param));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_782)(GLenum target, GLintptr offset, GLsizeiptr size);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_782)(GLenum target, GLintptr offset, GLsizeiptr size)
+{
+   DISPATCH(FlushMappedBufferRangeAPPLE, (target, offset, size), (F, "glFlushMappedBufferRangeAPPLE(0x%x, %d, %d);\n", target, offset, size));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayer)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer)
 {
    DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayer(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer));
@@ -5573,37 +5657,56 @@ KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayerEXT)(GLenum target, GLenum at
    DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayerEXT(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
+KEYWORD1 void KEYWORD2 NAME(ProvokingVertexEXT)(GLenum mode)
+{
+   DISPATCH(ProvokingVertexEXT, (mode), (F, "glProvokingVertexEXT(0x%x);\n", mode));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_785)(GLenum target, GLenum pname, GLvoid ** params);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_785)(GLenum target, GLenum pname, GLvoid ** params)
+{
+   DISPATCH(GetTexParameterPointervAPPLE, (target, pname, params), (F, "glGetTexParameterPointervAPPLE(0x%x, 0x%x, %p);\n", target, pname, (const void *) params));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_786)(GLenum target, GLsizei length, GLvoid * pointer);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_786)(GLenum target, GLsizei length, GLvoid * pointer)
+{
+   DISPATCH(TextureRangeAPPLE, (target, length, pointer), (F, "glTextureRangeAPPLE(0x%x, %d, %p);\n", target, length, (const void *) pointer));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_787)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_787)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
 {
    DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_788)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_788)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_789)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_789)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_790)(GLuint id, GLenum pname, GLint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_790)(GLuint id, GLenum pname, GLint64EXT * params)
 {
    DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_791)(GLuint id, GLenum pname, GLuint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_791)(GLuint id, GLenum pname, GLuint64EXT * params)
 {
    DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
@@ -6184,15 +6287,27 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(GetAttribLocationARB),
    TABLE_ENTRY(DrawBuffersARB),
    TABLE_ENTRY(RenderbufferStorageMultisample),
+   TABLE_ENTRY(FlushMappedBufferRange),
+   TABLE_ENTRY(MapBufferRange),
+   TABLE_ENTRY(BindVertexArray),
+   TABLE_ENTRY(GenVertexArrays),
+   TABLE_ENTRY(CopyBufferSubData),
+   TABLE_ENTRY(ClientWaitSync),
+   TABLE_ENTRY(DeleteSync),
+   TABLE_ENTRY(FenceSync),
+   TABLE_ENTRY(GetInteger64v),
+   TABLE_ENTRY(GetSynciv),
+   TABLE_ENTRY(IsSync),
+   TABLE_ENTRY(WaitSync),
    TABLE_ENTRY(PolygonOffsetEXT),
-   TABLE_ENTRY(_dispatch_stub_563),
-   TABLE_ENTRY(_dispatch_stub_564),
-   TABLE_ENTRY(_dispatch_stub_565),
-   TABLE_ENTRY(_dispatch_stub_566),
-   TABLE_ENTRY(_dispatch_stub_567),
-   TABLE_ENTRY(_dispatch_stub_568),
-   TABLE_ENTRY(_dispatch_stub_569),
-   TABLE_ENTRY(_dispatch_stub_570),
+   TABLE_ENTRY(_dispatch_stub_575),
+   TABLE_ENTRY(_dispatch_stub_576),
+   TABLE_ENTRY(_dispatch_stub_577),
+   TABLE_ENTRY(_dispatch_stub_578),
+   TABLE_ENTRY(_dispatch_stub_579),
+   TABLE_ENTRY(_dispatch_stub_580),
+   TABLE_ENTRY(_dispatch_stub_581),
+   TABLE_ENTRY(_dispatch_stub_582),
    TABLE_ENTRY(ColorPointerEXT),
    TABLE_ENTRY(EdgeFlagPointerEXT),
    TABLE_ENTRY(IndexPointerEXT),
@@ -6203,8 +6318,8 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(PointParameterfvEXT),
    TABLE_ENTRY(LockArraysEXT),
    TABLE_ENTRY(UnlockArraysEXT),
-   TABLE_ENTRY(_dispatch_stub_581),
-   TABLE_ENTRY(_dispatch_stub_582),
+   TABLE_ENTRY(_dispatch_stub_593),
+   TABLE_ENTRY(_dispatch_stub_594),
    TABLE_ENTRY(SecondaryColor3bEXT),
    TABLE_ENTRY(SecondaryColor3bvEXT),
    TABLE_ENTRY(SecondaryColor3dEXT),
@@ -6229,7 +6344,7 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(FogCoorddvEXT),
    TABLE_ENTRY(FogCoordfEXT),
    TABLE_ENTRY(FogCoordfvEXT),
-   TABLE_ENTRY(_dispatch_stub_607),
+   TABLE_ENTRY(_dispatch_stub_619),
    TABLE_ENTRY(BlendFuncSeparateEXT),
    TABLE_ENTRY(FlushVertexArrayRangeNV),
    TABLE_ENTRY(VertexArrayRangeNV),
@@ -6271,15 +6386,15 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(WindowPos4ivMESA),
    TABLE_ENTRY(WindowPos4sMESA),
    TABLE_ENTRY(WindowPos4svMESA),
-   TABLE_ENTRY(_dispatch_stub_649),
-   TABLE_ENTRY(_dispatch_stub_650),
-   TABLE_ENTRY(_dispatch_stub_651),
-   TABLE_ENTRY(_dispatch_stub_652),
-   TABLE_ENTRY(_dispatch_stub_653),
-   TABLE_ENTRY(_dispatch_stub_654),
-   TABLE_ENTRY(_dispatch_stub_655),
-   TABLE_ENTRY(_dispatch_stub_656),
-   TABLE_ENTRY(_dispatch_stub_657),
+   TABLE_ENTRY(_dispatch_stub_661),
+   TABLE_ENTRY(_dispatch_stub_662),
+   TABLE_ENTRY(_dispatch_stub_663),
+   TABLE_ENTRY(_dispatch_stub_664),
+   TABLE_ENTRY(_dispatch_stub_665),
+   TABLE_ENTRY(_dispatch_stub_666),
+   TABLE_ENTRY(_dispatch_stub_667),
+   TABLE_ENTRY(_dispatch_stub_668),
+   TABLE_ENTRY(_dispatch_stub_669),
    TABLE_ENTRY(AreProgramsResidentNV),
    TABLE_ENTRY(BindProgramNV),
    TABLE_ENTRY(DeleteProgramsNV),
@@ -6360,19 +6475,19 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(SetFragmentShaderConstantATI),
    TABLE_ENTRY(PointParameteriNV),
    TABLE_ENTRY(PointParameterivNV),
-   TABLE_ENTRY(_dispatch_stub_738),
-   TABLE_ENTRY(_dispatch_stub_739),
-   TABLE_ENTRY(_dispatch_stub_740),
-   TABLE_ENTRY(_dispatch_stub_741),
-   TABLE_ENTRY(_dispatch_stub_742),
+   TABLE_ENTRY(_dispatch_stub_750),
+   TABLE_ENTRY(_dispatch_stub_751),
+   TABLE_ENTRY(_dispatch_stub_752),
+   TABLE_ENTRY(_dispatch_stub_753),
+   TABLE_ENTRY(_dispatch_stub_754),
    TABLE_ENTRY(GetProgramNamedParameterdvNV),
    TABLE_ENTRY(GetProgramNamedParameterfvNV),
    TABLE_ENTRY(ProgramNamedParameter4dNV),
    TABLE_ENTRY(ProgramNamedParameter4dvNV),
    TABLE_ENTRY(ProgramNamedParameter4fNV),
    TABLE_ENTRY(ProgramNamedParameter4fvNV),
-   TABLE_ENTRY(_dispatch_stub_749),
-   TABLE_ENTRY(_dispatch_stub_750),
+   TABLE_ENTRY(_dispatch_stub_761),
+   TABLE_ENTRY(_dispatch_stub_762),
    TABLE_ENTRY(BindFramebufferEXT),
    TABLE_ENTRY(BindRenderbufferEXT),
    TABLE_ENTRY(CheckFramebufferStatusEXT),
@@ -6390,13 +6505,18 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(IsFramebufferEXT),
    TABLE_ENTRY(IsRenderbufferEXT),
    TABLE_ENTRY(RenderbufferStorageEXT),
-   TABLE_ENTRY(_dispatch_stub_768),
+   TABLE_ENTRY(_dispatch_stub_780),
+   TABLE_ENTRY(_dispatch_stub_781),
+   TABLE_ENTRY(_dispatch_stub_782),
    TABLE_ENTRY(FramebufferTextureLayerEXT),
-   TABLE_ENTRY(_dispatch_stub_770),
-   TABLE_ENTRY(_dispatch_stub_771),
-   TABLE_ENTRY(_dispatch_stub_772),
-   TABLE_ENTRY(_dispatch_stub_773),
-   TABLE_ENTRY(_dispatch_stub_774),
+   TABLE_ENTRY(ProvokingVertexEXT),
+   TABLE_ENTRY(_dispatch_stub_785),
+   TABLE_ENTRY(_dispatch_stub_786),
+   TABLE_ENTRY(_dispatch_stub_787),
+   TABLE_ENTRY(_dispatch_stub_788),
+   TABLE_ENTRY(_dispatch_stub_789),
+   TABLE_ENTRY(_dispatch_stub_790),
+   TABLE_ENTRY(_dispatch_stub_791),
    /* A whole bunch of no-op functions.  These might be called
     * when someone tries to call a dynamically-registered
     * extension function without a current rendering context.
@@ -6763,6 +6883,8 @@ static _glapi_proc UNUSED_TABLE_NAME[] = {
    TABLE_ENTRY(IsProgramARB),
    TABLE_ENTRY(PointParameteri),
    TABLE_ENTRY(PointParameteriv),
+   TABLE_ENTRY(DeleteVertexArrays),
+   TABLE_ENTRY(IsVertexArray),
    TABLE_ENTRY(BlendEquationSeparate),
    TABLE_ENTRY(BindFramebuffer),
    TABLE_ENTRY(BindRenderbuffer),
diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h
index 680893cfc4..ae6603262e 100644
--- a/src/mesa/glapi/glprocs.h
+++ b/src/mesa/glapi/glprocs.h
@@ -614,6 +614,18 @@ static const char gl_string_table[] =
     "glGetAttribLocationARB\0"
     "glDrawBuffersARB\0"
     "glRenderbufferStorageMultisample\0"
+    "glFlushMappedBufferRange\0"
+    "glMapBufferRange\0"
+    "glBindVertexArray\0"
+    "glGenVertexArrays\0"
+    "glCopyBufferSubData\0"
+    "glClientWaitSync\0"
+    "glDeleteSync\0"
+    "glFenceSync\0"
+    "glGetInteger64v\0"
+    "glGetSynciv\0"
+    "glIsSync\0"
+    "glWaitSync\0"
     "glPolygonOffsetEXT\0"
     "glGetPixelTexGenParameterfvSGIS\0"
     "glGetPixelTexGenParameterivSGIS\0"
@@ -821,7 +833,12 @@ static const char gl_string_table[] =
     "glIsRenderbufferEXT\0"
     "glRenderbufferStorageEXT\0"
     "glBlitFramebufferEXT\0"
+    "glBufferParameteriAPPLE\0"
+    "glFlushMappedBufferRangeAPPLE\0"
     "glFramebufferTextureLayerEXT\0"
+    "glProvokingVertexEXT\0"
+    "glGetTexParameterPointervAPPLE\0"
+    "glTextureRangeAPPLE\0"
     "glStencilFuncSeparateATI\0"
     "glProgramEnvParameters4fvEXT\0"
     "glProgramLocalParameters4fvEXT\0"
@@ -1104,6 +1121,8 @@ static const char gl_string_table[] =
     "glIsProgramARB\0"
     "glPointParameteri\0"
     "glPointParameteriv\0"
+    "glDeleteVertexArrays\0"
+    "glIsVertexArray\0"
     "glBlendEquationSeparate\0"
     "glBlendEquationSeparateATI\0"
     "glBindFramebuffer\0"
@@ -1142,39 +1161,43 @@ static const char gl_string_table[] =
 #define gl_dispatch_stub_364 mgl_dispatch_stub_364
 #define gl_dispatch_stub_365 mgl_dispatch_stub_365
 #define gl_dispatch_stub_366 mgl_dispatch_stub_366
-#define gl_dispatch_stub_563 mgl_dispatch_stub_563
-#define gl_dispatch_stub_564 mgl_dispatch_stub_564
-#define gl_dispatch_stub_565 mgl_dispatch_stub_565
-#define gl_dispatch_stub_566 mgl_dispatch_stub_566
-#define gl_dispatch_stub_567 mgl_dispatch_stub_567
-#define gl_dispatch_stub_568 mgl_dispatch_stub_568
-#define gl_dispatch_stub_569 mgl_dispatch_stub_569
-#define gl_dispatch_stub_570 mgl_dispatch_stub_570
+#define gl_dispatch_stub_575 mgl_dispatch_stub_575
+#define gl_dispatch_stub_576 mgl_dispatch_stub_576
+#define gl_dispatch_stub_577 mgl_dispatch_stub_577
+#define gl_dispatch_stub_578 mgl_dispatch_stub_578
+#define gl_dispatch_stub_579 mgl_dispatch_stub_579
+#define gl_dispatch_stub_580 mgl_dispatch_stub_580
 #define gl_dispatch_stub_581 mgl_dispatch_stub_581
 #define gl_dispatch_stub_582 mgl_dispatch_stub_582
-#define gl_dispatch_stub_607 mgl_dispatch_stub_607
-#define gl_dispatch_stub_649 mgl_dispatch_stub_649
-#define gl_dispatch_stub_650 mgl_dispatch_stub_650
-#define gl_dispatch_stub_651 mgl_dispatch_stub_651
-#define gl_dispatch_stub_652 mgl_dispatch_stub_652
-#define gl_dispatch_stub_653 mgl_dispatch_stub_653
-#define gl_dispatch_stub_654 mgl_dispatch_stub_654
-#define gl_dispatch_stub_655 mgl_dispatch_stub_655
-#define gl_dispatch_stub_656 mgl_dispatch_stub_656
-#define gl_dispatch_stub_657 mgl_dispatch_stub_657
-#define gl_dispatch_stub_738 mgl_dispatch_stub_738
-#define gl_dispatch_stub_739 mgl_dispatch_stub_739
-#define gl_dispatch_stub_740 mgl_dispatch_stub_740
-#define gl_dispatch_stub_741 mgl_dispatch_stub_741
-#define gl_dispatch_stub_742 mgl_dispatch_stub_742
-#define gl_dispatch_stub_749 mgl_dispatch_stub_749
+#define gl_dispatch_stub_593 mgl_dispatch_stub_593
+#define gl_dispatch_stub_594 mgl_dispatch_stub_594
+#define gl_dispatch_stub_619 mgl_dispatch_stub_619
+#define gl_dispatch_stub_661 mgl_dispatch_stub_661
+#define gl_dispatch_stub_662 mgl_dispatch_stub_662
+#define gl_dispatch_stub_663 mgl_dispatch_stub_663
+#define gl_dispatch_stub_664 mgl_dispatch_stub_664
+#define gl_dispatch_stub_665 mgl_dispatch_stub_665
+#define gl_dispatch_stub_666 mgl_dispatch_stub_666
+#define gl_dispatch_stub_667 mgl_dispatch_stub_667
+#define gl_dispatch_stub_668 mgl_dispatch_stub_668
+#define gl_dispatch_stub_669 mgl_dispatch_stub_669
 #define gl_dispatch_stub_750 mgl_dispatch_stub_750
-#define gl_dispatch_stub_768 mgl_dispatch_stub_768
-#define gl_dispatch_stub_770 mgl_dispatch_stub_770
-#define gl_dispatch_stub_771 mgl_dispatch_stub_771
-#define gl_dispatch_stub_772 mgl_dispatch_stub_772
-#define gl_dispatch_stub_773 mgl_dispatch_stub_773
-#define gl_dispatch_stub_774 mgl_dispatch_stub_774
+#define gl_dispatch_stub_751 mgl_dispatch_stub_751
+#define gl_dispatch_stub_752 mgl_dispatch_stub_752
+#define gl_dispatch_stub_753 mgl_dispatch_stub_753
+#define gl_dispatch_stub_754 mgl_dispatch_stub_754
+#define gl_dispatch_stub_761 mgl_dispatch_stub_761
+#define gl_dispatch_stub_762 mgl_dispatch_stub_762
+#define gl_dispatch_stub_780 mgl_dispatch_stub_780
+#define gl_dispatch_stub_781 mgl_dispatch_stub_781
+#define gl_dispatch_stub_782 mgl_dispatch_stub_782
+#define gl_dispatch_stub_785 mgl_dispatch_stub_785
+#define gl_dispatch_stub_786 mgl_dispatch_stub_786
+#define gl_dispatch_stub_787 mgl_dispatch_stub_787
+#define gl_dispatch_stub_788 mgl_dispatch_stub_788
+#define gl_dispatch_stub_789 mgl_dispatch_stub_789
+#define gl_dispatch_stub_790 mgl_dispatch_stub_790
+#define gl_dispatch_stub_791 mgl_dispatch_stub_791
 #endif /* USE_MGL_NAMESPACE */
 
 
@@ -1192,39 +1215,43 @@ void GLAPIENTRY gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params
 void GLAPIENTRY gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid * values);
 void GLAPIENTRY gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params);
 void GLAPIENTRY gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params);
-void GLAPIENTRY gl_dispatch_stub_563(GLenum pname, GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_564(GLenum pname, GLint * params);
-void GLAPIENTRY gl_dispatch_stub_565(GLenum pname, GLfloat param);
-void GLAPIENTRY gl_dispatch_stub_566(GLenum pname, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_567(GLenum pname, GLint param);
-void GLAPIENTRY gl_dispatch_stub_568(GLenum pname, const GLint * params);
-void GLAPIENTRY gl_dispatch_stub_569(GLclampf value, GLboolean invert);
-void GLAPIENTRY gl_dispatch_stub_570(GLenum pattern);
-void GLAPIENTRY gl_dispatch_stub_581(GLenum pname, GLdouble * params);
-void GLAPIENTRY gl_dispatch_stub_582(GLenum pname, GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_607(GLenum mode);
-void GLAPIENTRY gl_dispatch_stub_649(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
-void GLAPIENTRY gl_dispatch_stub_650(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
-void GLAPIENTRY gl_dispatch_stub_651(GLsizei n, const GLuint * fences);
-void GLAPIENTRY gl_dispatch_stub_652(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_653(GLsizei n, GLuint * fences);
-void GLAPIENTRY gl_dispatch_stub_654(GLuint fence, GLenum pname, GLint * params);
-GLboolean GLAPIENTRY gl_dispatch_stub_655(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_656(GLuint fence, GLenum condition);
-GLboolean GLAPIENTRY gl_dispatch_stub_657(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_738(GLenum face);
-void GLAPIENTRY gl_dispatch_stub_739(GLuint array);
-void GLAPIENTRY gl_dispatch_stub_740(GLsizei n, const GLuint * arrays);
-void GLAPIENTRY gl_dispatch_stub_741(GLsizei n, GLuint * arrays);
-GLboolean GLAPIENTRY gl_dispatch_stub_742(GLuint array);
-void GLAPIENTRY gl_dispatch_stub_749(GLclampd zmin, GLclampd zmax);
-void GLAPIENTRY gl_dispatch_stub_750(GLenum modeRGB, GLenum modeA);
-void GLAPIENTRY gl_dispatch_stub_768(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
-void GLAPIENTRY gl_dispatch_stub_770(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
-void GLAPIENTRY gl_dispatch_stub_771(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_772(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_773(GLuint id, GLenum pname, GLint64EXT * params);
-void GLAPIENTRY gl_dispatch_stub_774(GLuint id, GLenum pname, GLuint64EXT * params);
+void GLAPIENTRY gl_dispatch_stub_575(GLenum pname, GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_576(GLenum pname, GLint * params);
+void GLAPIENTRY gl_dispatch_stub_577(GLenum pname, GLfloat param);
+void GLAPIENTRY gl_dispatch_stub_578(GLenum pname, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_579(GLenum pname, GLint param);
+void GLAPIENTRY gl_dispatch_stub_580(GLenum pname, const GLint * params);
+void GLAPIENTRY gl_dispatch_stub_581(GLclampf value, GLboolean invert);
+void GLAPIENTRY gl_dispatch_stub_582(GLenum pattern);
+void GLAPIENTRY gl_dispatch_stub_593(GLenum pname, GLdouble * params);
+void GLAPIENTRY gl_dispatch_stub_594(GLenum pname, GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_619(GLenum mode);
+void GLAPIENTRY gl_dispatch_stub_661(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
+void GLAPIENTRY gl_dispatch_stub_662(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
+void GLAPIENTRY gl_dispatch_stub_663(GLsizei n, const GLuint * fences);
+void GLAPIENTRY gl_dispatch_stub_664(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_665(GLsizei n, GLuint * fences);
+void GLAPIENTRY gl_dispatch_stub_666(GLuint fence, GLenum pname, GLint * params);
+GLboolean GLAPIENTRY gl_dispatch_stub_667(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_668(GLuint fence, GLenum condition);
+GLboolean GLAPIENTRY gl_dispatch_stub_669(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_750(GLenum face);
+void GLAPIENTRY gl_dispatch_stub_751(GLuint array);
+void GLAPIENTRY gl_dispatch_stub_752(GLsizei n, const GLuint * arrays);
+void GLAPIENTRY gl_dispatch_stub_753(GLsizei n, GLuint * arrays);
+GLboolean GLAPIENTRY gl_dispatch_stub_754(GLuint array);
+void GLAPIENTRY gl_dispatch_stub_761(GLclampd zmin, GLclampd zmax);
+void GLAPIENTRY gl_dispatch_stub_762(GLenum modeRGB, GLenum modeA);
+void GLAPIENTRY gl_dispatch_stub_780(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+void GLAPIENTRY gl_dispatch_stub_781(GLenum target, GLenum pname, GLint param);
+void GLAPIENTRY gl_dispatch_stub_782(GLenum target, GLintptr offset, GLsizeiptr size);
+void GLAPIENTRY gl_dispatch_stub_785(GLenum target, GLenum pname, GLvoid ** params);
+void GLAPIENTRY gl_dispatch_stub_786(GLenum target, GLsizei length, GLvoid * pointer);
+void GLAPIENTRY gl_dispatch_stub_787(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
+void GLAPIENTRY gl_dispatch_stub_788(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_789(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_790(GLuint id, GLenum pname, GLint64EXT * params);
+void GLAPIENTRY gl_dispatch_stub_791(GLuint id, GLenum pname, GLuint64EXT * params);
 #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */
 
 static const glprocs_table_t static_functions[] = {
@@ -1790,517 +1817,536 @@ static const glprocs_table_t static_functions[] = {
     NAME_FUNC_OFFSET( 8911, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
     NAME_FUNC_OFFSET( 8934, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
     NAME_FUNC_OFFSET( 8951, glRenderbufferStorageMultisample, glRenderbufferStorageMultisample, NULL, _gloffset_RenderbufferStorageMultisample),
-    NAME_FUNC_OFFSET( 8984, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
-    NAME_FUNC_OFFSET( 9003, gl_dispatch_stub_563, gl_dispatch_stub_563, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9035, gl_dispatch_stub_564, gl_dispatch_stub_564, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9067, gl_dispatch_stub_565, gl_dispatch_stub_565, NULL, _gloffset_PixelTexGenParameterfSGIS),
-    NAME_FUNC_OFFSET( 9095, gl_dispatch_stub_566, gl_dispatch_stub_566, NULL, _gloffset_PixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9124, gl_dispatch_stub_567, gl_dispatch_stub_567, NULL, _gloffset_PixelTexGenParameteriSGIS),
-    NAME_FUNC_OFFSET( 9152, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_PixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9181, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET( 9198, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET( 9218, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
-    NAME_FUNC_OFFSET( 9236, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
-    NAME_FUNC_OFFSET( 9257, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
-    NAME_FUNC_OFFSET( 9275, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
-    NAME_FUNC_OFFSET( 9294, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9315, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
-    NAME_FUNC_OFFSET( 9334, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET( 9355, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET( 9377, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
-    NAME_FUNC_OFFSET( 9393, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
-    NAME_FUNC_OFFSET( 9411, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_CullParameterdvEXT),
-    NAME_FUNC_OFFSET( 9432, gl_dispatch_stub_582, gl_dispatch_stub_582, NULL, _gloffset_CullParameterfvEXT),
-    NAME_FUNC_OFFSET( 9453, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET( 9475, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET( 9498, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET( 9520, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET( 9543, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET( 9565, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET( 9588, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET( 9610, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET( 9633, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET( 9655, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET( 9678, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET( 9701, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET( 9725, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET( 9748, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET( 9772, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET( 9795, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET( 9819, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET( 9846, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET( 9867, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET( 9890, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9911, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET( 9926, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET( 9942, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET( 9957, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET( 9973, gl_dispatch_stub_607, gl_dispatch_stub_607, NULL, _gloffset_PixelTexGenSGIX),
-    NAME_FUNC_OFFSET( 9991, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(10014, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10040, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10061, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
-    NAME_FUNC_OFFSET(10079, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
-    NAME_FUNC_OFFSET(10098, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
-    NAME_FUNC_OFFSET(10121, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
-    NAME_FUNC_OFFSET(10145, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
-    NAME_FUNC_OFFSET(10168, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
-    NAME_FUNC_OFFSET(10192, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
-    NAME_FUNC_OFFSET(10215, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10247, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10279, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
-    NAME_FUNC_OFFSET(10312, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
-    NAME_FUNC_OFFSET(10345, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10382, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10419, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
-    NAME_FUNC_OFFSET(10439, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(10457, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(10476, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(10494, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(10513, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(10531, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(10550, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(10568, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(10587, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(10605, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(10624, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(10642, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(10661, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(10679, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(10698, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(10716, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(10735, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
-    NAME_FUNC_OFFSET(10753, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
-    NAME_FUNC_OFFSET(10772, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
-    NAME_FUNC_OFFSET(10790, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
-    NAME_FUNC_OFFSET(10809, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
-    NAME_FUNC_OFFSET(10827, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
-    NAME_FUNC_OFFSET(10846, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
-    NAME_FUNC_OFFSET(10864, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
-    NAME_FUNC_OFFSET(10883, gl_dispatch_stub_649, gl_dispatch_stub_649, NULL, _gloffset_MultiModeDrawArraysIBM),
-    NAME_FUNC_OFFSET(10908, gl_dispatch_stub_650, gl_dispatch_stub_650, NULL, _gloffset_MultiModeDrawElementsIBM),
-    NAME_FUNC_OFFSET(10935, gl_dispatch_stub_651, gl_dispatch_stub_651, NULL, _gloffset_DeleteFencesNV),
-    NAME_FUNC_OFFSET(10952, gl_dispatch_stub_652, gl_dispatch_stub_652, NULL, _gloffset_FinishFenceNV),
-    NAME_FUNC_OFFSET(10968, gl_dispatch_stub_653, gl_dispatch_stub_653, NULL, _gloffset_GenFencesNV),
-    NAME_FUNC_OFFSET(10982, gl_dispatch_stub_654, gl_dispatch_stub_654, NULL, _gloffset_GetFenceivNV),
-    NAME_FUNC_OFFSET(10997, gl_dispatch_stub_655, gl_dispatch_stub_655, NULL, _gloffset_IsFenceNV),
-    NAME_FUNC_OFFSET(11009, gl_dispatch_stub_656, gl_dispatch_stub_656, NULL, _gloffset_SetFenceNV),
-    NAME_FUNC_OFFSET(11022, gl_dispatch_stub_657, gl_dispatch_stub_657, NULL, _gloffset_TestFenceNV),
-    NAME_FUNC_OFFSET(11036, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
-    NAME_FUNC_OFFSET(11060, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(11076, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(11095, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
-    NAME_FUNC_OFFSET(11114, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(11130, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
-    NAME_FUNC_OFFSET(11156, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
-    NAME_FUNC_OFFSET(11182, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
-    NAME_FUNC_OFFSET(11203, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
-    NAME_FUNC_OFFSET(11220, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
-    NAME_FUNC_OFFSET(11241, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(11269, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
-    NAME_FUNC_OFFSET(11291, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
-    NAME_FUNC_OFFSET(11313, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
-    NAME_FUNC_OFFSET(11335, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(11349, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
-    NAME_FUNC_OFFSET(11365, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
-    NAME_FUNC_OFFSET(11390, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
-    NAME_FUNC_OFFSET(11415, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
-    NAME_FUNC_OFFSET(11443, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
-    NAME_FUNC_OFFSET(11459, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
-    NAME_FUNC_OFFSET(11478, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
-    NAME_FUNC_OFFSET(11498, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
-    NAME_FUNC_OFFSET(11517, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
-    NAME_FUNC_OFFSET(11537, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
-    NAME_FUNC_OFFSET(11556, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
-    NAME_FUNC_OFFSET(11576, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
-    NAME_FUNC_OFFSET(11595, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
-    NAME_FUNC_OFFSET(11615, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
-    NAME_FUNC_OFFSET(11634, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
-    NAME_FUNC_OFFSET(11654, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
-    NAME_FUNC_OFFSET(11673, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
-    NAME_FUNC_OFFSET(11693, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
-    NAME_FUNC_OFFSET(11712, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
-    NAME_FUNC_OFFSET(11732, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
-    NAME_FUNC_OFFSET(11751, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
-    NAME_FUNC_OFFSET(11771, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
-    NAME_FUNC_OFFSET(11790, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
-    NAME_FUNC_OFFSET(11810, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
-    NAME_FUNC_OFFSET(11829, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
-    NAME_FUNC_OFFSET(11849, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
-    NAME_FUNC_OFFSET(11868, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
-    NAME_FUNC_OFFSET(11888, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
-    NAME_FUNC_OFFSET(11907, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
-    NAME_FUNC_OFFSET(11927, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
-    NAME_FUNC_OFFSET(11947, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
-    NAME_FUNC_OFFSET(11968, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
-    NAME_FUNC_OFFSET(11992, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
-    NAME_FUNC_OFFSET(12013, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
-    NAME_FUNC_OFFSET(12034, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
-    NAME_FUNC_OFFSET(12055, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
-    NAME_FUNC_OFFSET(12076, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
-    NAME_FUNC_OFFSET(12097, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
-    NAME_FUNC_OFFSET(12118, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
-    NAME_FUNC_OFFSET(12139, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
-    NAME_FUNC_OFFSET(12160, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
-    NAME_FUNC_OFFSET(12181, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
-    NAME_FUNC_OFFSET(12202, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
-    NAME_FUNC_OFFSET(12223, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
-    NAME_FUNC_OFFSET(12244, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
-    NAME_FUNC_OFFSET(12266, glGetTexBumpParameterfvATI, glGetTexBumpParameterfvATI, NULL, _gloffset_GetTexBumpParameterfvATI),
-    NAME_FUNC_OFFSET(12293, glGetTexBumpParameterivATI, glGetTexBumpParameterivATI, NULL, _gloffset_GetTexBumpParameterivATI),
-    NAME_FUNC_OFFSET(12320, glTexBumpParameterfvATI, glTexBumpParameterfvATI, NULL, _gloffset_TexBumpParameterfvATI),
-    NAME_FUNC_OFFSET(12344, glTexBumpParameterivATI, glTexBumpParameterivATI, NULL, _gloffset_TexBumpParameterivATI),
-    NAME_FUNC_OFFSET(12368, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12390, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12412, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12434, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
-    NAME_FUNC_OFFSET(12459, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
-    NAME_FUNC_OFFSET(12483, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12505, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12527, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12549, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
-    NAME_FUNC_OFFSET(12575, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
-    NAME_FUNC_OFFSET(12598, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
-    NAME_FUNC_OFFSET(12622, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
-    NAME_FUNC_OFFSET(12640, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
-    NAME_FUNC_OFFSET(12655, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
-    NAME_FUNC_OFFSET(12686, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(12706, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(12727, gl_dispatch_stub_738, gl_dispatch_stub_738, NULL, _gloffset_ActiveStencilFaceEXT),
-    NAME_FUNC_OFFSET(12750, gl_dispatch_stub_739, gl_dispatch_stub_739, NULL, _gloffset_BindVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12773, gl_dispatch_stub_740, gl_dispatch_stub_740, NULL, _gloffset_DeleteVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12799, gl_dispatch_stub_741, gl_dispatch_stub_741, NULL, _gloffset_GenVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12822, gl_dispatch_stub_742, gl_dispatch_stub_742, NULL, _gloffset_IsVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12843, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
-    NAME_FUNC_OFFSET(12874, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
-    NAME_FUNC_OFFSET(12905, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
-    NAME_FUNC_OFFSET(12933, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
-    NAME_FUNC_OFFSET(12962, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
-    NAME_FUNC_OFFSET(12990, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
-    NAME_FUNC_OFFSET(13019, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_DepthBoundsEXT),
-    NAME_FUNC_OFFSET(13036, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(13063, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
-    NAME_FUNC_OFFSET(13084, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
-    NAME_FUNC_OFFSET(13106, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
-    NAME_FUNC_OFFSET(13134, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
-    NAME_FUNC_OFFSET(13158, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13183, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
-    NAME_FUNC_OFFSET(13212, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
-    NAME_FUNC_OFFSET(13238, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
-    NAME_FUNC_OFFSET(13264, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
-    NAME_FUNC_OFFSET(13290, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
-    NAME_FUNC_OFFSET(13311, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13333, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
-    NAME_FUNC_OFFSET(13353, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
-    NAME_FUNC_OFFSET(13394, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
-    NAME_FUNC_OFFSET(13426, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
-    NAME_FUNC_OFFSET(13445, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
-    NAME_FUNC_OFFSET(13465, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
-    NAME_FUNC_OFFSET(13490, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_BlitFramebufferEXT),
-    NAME_FUNC_OFFSET(13511, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
-    NAME_FUNC_OFFSET(13540, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_StencilFuncSeparateATI),
-    NAME_FUNC_OFFSET(13565, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramEnvParameters4fvEXT),
-    NAME_FUNC_OFFSET(13594, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_ProgramLocalParameters4fvEXT),
-    NAME_FUNC_OFFSET(13625, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjecti64vEXT),
-    NAME_FUNC_OFFSET(13649, gl_dispatch_stub_774, gl_dispatch_stub_774, NULL, _gloffset_GetQueryObjectui64vEXT),
-    NAME_FUNC_OFFSET(13674, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
-    NAME_FUNC_OFFSET(13692, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
-    NAME_FUNC_OFFSET(13709, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
-    NAME_FUNC_OFFSET(13725, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
-    NAME_FUNC_OFFSET(13750, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
-    NAME_FUNC_OFFSET(13770, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
-    NAME_FUNC_OFFSET(13790, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
-    NAME_FUNC_OFFSET(13813, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
-    NAME_FUNC_OFFSET(13836, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
-    NAME_FUNC_OFFSET(13856, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
-    NAME_FUNC_OFFSET(13873, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
-    NAME_FUNC_OFFSET(13890, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
-    NAME_FUNC_OFFSET(13905, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
-    NAME_FUNC_OFFSET(13929, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
-    NAME_FUNC_OFFSET(13948, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
-    NAME_FUNC_OFFSET(13967, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
-    NAME_FUNC_OFFSET(13983, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
-    NAME_FUNC_OFFSET(14002, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
-    NAME_FUNC_OFFSET(14025, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(14041, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(14057, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
-    NAME_FUNC_OFFSET(14084, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
-    NAME_FUNC_OFFSET(14111, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
-    NAME_FUNC_OFFSET(14131, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14150, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14169, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14199, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14229, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14259, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14289, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
-    NAME_FUNC_OFFSET(14308, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
-    NAME_FUNC_OFFSET(14331, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14356, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14381, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
-    NAME_FUNC_OFFSET(14408, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14436, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
-    NAME_FUNC_OFFSET(14463, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14491, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14520, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14549, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
-    NAME_FUNC_OFFSET(14575, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14606, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14637, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
-    NAME_FUNC_OFFSET(14661, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
-    NAME_FUNC_OFFSET(14684, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
-    NAME_FUNC_OFFSET(14702, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
-    NAME_FUNC_OFFSET(14731, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
-    NAME_FUNC_OFFSET(14760, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
-    NAME_FUNC_OFFSET(14775, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
-    NAME_FUNC_OFFSET(14801, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
-    NAME_FUNC_OFFSET(14827, glHistogram, glHistogram, NULL, _gloffset_Histogram),
-    NAME_FUNC_OFFSET(14842, glMinmax, glMinmax, NULL, _gloffset_Minmax),
-    NAME_FUNC_OFFSET(14854, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
-    NAME_FUNC_OFFSET(14874, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
-    NAME_FUNC_OFFSET(14891, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
-    NAME_FUNC_OFFSET(14907, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
-    NAME_FUNC_OFFSET(14926, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
-    NAME_FUNC_OFFSET(14949, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
-    NAME_FUNC_OFFSET(14965, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
-    NAME_FUNC_OFFSET(14987, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
-    NAME_FUNC_OFFSET(15005, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
-    NAME_FUNC_OFFSET(15024, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
-    NAME_FUNC_OFFSET(15042, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
-    NAME_FUNC_OFFSET(15061, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
-    NAME_FUNC_OFFSET(15079, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
-    NAME_FUNC_OFFSET(15098, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
-    NAME_FUNC_OFFSET(15116, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
-    NAME_FUNC_OFFSET(15135, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
-    NAME_FUNC_OFFSET(15153, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
-    NAME_FUNC_OFFSET(15172, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
-    NAME_FUNC_OFFSET(15190, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
-    NAME_FUNC_OFFSET(15209, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
-    NAME_FUNC_OFFSET(15227, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
-    NAME_FUNC_OFFSET(15246, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
-    NAME_FUNC_OFFSET(15264, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
-    NAME_FUNC_OFFSET(15283, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
-    NAME_FUNC_OFFSET(15301, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
-    NAME_FUNC_OFFSET(15320, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
-    NAME_FUNC_OFFSET(15338, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
-    NAME_FUNC_OFFSET(15357, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
-    NAME_FUNC_OFFSET(15375, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
-    NAME_FUNC_OFFSET(15394, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
-    NAME_FUNC_OFFSET(15412, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
-    NAME_FUNC_OFFSET(15431, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
-    NAME_FUNC_OFFSET(15449, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
-    NAME_FUNC_OFFSET(15468, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
-    NAME_FUNC_OFFSET(15486, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
-    NAME_FUNC_OFFSET(15505, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
-    NAME_FUNC_OFFSET(15523, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
-    NAME_FUNC_OFFSET(15542, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
-    NAME_FUNC_OFFSET(15560, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
-    NAME_FUNC_OFFSET(15579, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
-    NAME_FUNC_OFFSET(15602, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15625, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15648, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15671, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15694, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
-    NAME_FUNC_OFFSET(15711, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
-    NAME_FUNC_OFFSET(15734, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
-    NAME_FUNC_OFFSET(15757, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
-    NAME_FUNC_OFFSET(15780, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
-    NAME_FUNC_OFFSET(15806, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
-    NAME_FUNC_OFFSET(15832, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
-    NAME_FUNC_OFFSET(15858, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
-    NAME_FUNC_OFFSET(15882, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15909, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15935, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
-    NAME_FUNC_OFFSET(15955, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
-    NAME_FUNC_OFFSET(15975, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
-    NAME_FUNC_OFFSET(15995, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
-    NAME_FUNC_OFFSET(16018, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
-    NAME_FUNC_OFFSET(16042, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
-    NAME_FUNC_OFFSET(16065, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
-    NAME_FUNC_OFFSET(16089, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
-    NAME_FUNC_OFFSET(16106, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
-    NAME_FUNC_OFFSET(16124, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
-    NAME_FUNC_OFFSET(16141, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
-    NAME_FUNC_OFFSET(16159, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
-    NAME_FUNC_OFFSET(16176, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
-    NAME_FUNC_OFFSET(16194, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
-    NAME_FUNC_OFFSET(16211, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
-    NAME_FUNC_OFFSET(16229, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
-    NAME_FUNC_OFFSET(16246, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
-    NAME_FUNC_OFFSET(16264, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
-    NAME_FUNC_OFFSET(16281, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
-    NAME_FUNC_OFFSET(16299, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
-    NAME_FUNC_OFFSET(16316, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
-    NAME_FUNC_OFFSET(16334, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
-    NAME_FUNC_OFFSET(16351, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
-    NAME_FUNC_OFFSET(16369, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
-    NAME_FUNC_OFFSET(16386, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
-    NAME_FUNC_OFFSET(16404, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
-    NAME_FUNC_OFFSET(16423, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
-    NAME_FUNC_OFFSET(16442, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
-    NAME_FUNC_OFFSET(16461, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
-    NAME_FUNC_OFFSET(16480, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
-    NAME_FUNC_OFFSET(16500, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
-    NAME_FUNC_OFFSET(16520, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
-    NAME_FUNC_OFFSET(16540, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
-    NAME_FUNC_OFFSET(16558, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
-    NAME_FUNC_OFFSET(16575, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
-    NAME_FUNC_OFFSET(16593, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
-    NAME_FUNC_OFFSET(16610, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
-    NAME_FUNC_OFFSET(16628, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
-    NAME_FUNC_OFFSET(16646, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
-    NAME_FUNC_OFFSET(16663, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
-    NAME_FUNC_OFFSET(16681, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
-    NAME_FUNC_OFFSET(16700, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
-    NAME_FUNC_OFFSET(16719, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
-    NAME_FUNC_OFFSET(16738, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
-    NAME_FUNC_OFFSET(16760, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
-    NAME_FUNC_OFFSET(16773, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
-    NAME_FUNC_OFFSET(16786, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
-    NAME_FUNC_OFFSET(16802, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
-    NAME_FUNC_OFFSET(16818, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
-    NAME_FUNC_OFFSET(16831, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
-    NAME_FUNC_OFFSET(16854, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
-    NAME_FUNC_OFFSET(16874, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
-    NAME_FUNC_OFFSET(16893, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
-    NAME_FUNC_OFFSET(16904, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
-    NAME_FUNC_OFFSET(16916, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
-    NAME_FUNC_OFFSET(16930, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
-    NAME_FUNC_OFFSET(16943, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
-    NAME_FUNC_OFFSET(16959, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
-    NAME_FUNC_OFFSET(16970, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
-    NAME_FUNC_OFFSET(16983, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
-    NAME_FUNC_OFFSET(17002, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
-    NAME_FUNC_OFFSET(17022, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
-    NAME_FUNC_OFFSET(17035, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
-    NAME_FUNC_OFFSET(17045, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
-    NAME_FUNC_OFFSET(17061, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
-    NAME_FUNC_OFFSET(17080, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
-    NAME_FUNC_OFFSET(17098, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
-    NAME_FUNC_OFFSET(17119, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
-    NAME_FUNC_OFFSET(17134, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
-    NAME_FUNC_OFFSET(17149, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
-    NAME_FUNC_OFFSET(17163, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
-    NAME_FUNC_OFFSET(17178, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
-    NAME_FUNC_OFFSET(17190, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
-    NAME_FUNC_OFFSET(17203, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
-    NAME_FUNC_OFFSET(17215, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
-    NAME_FUNC_OFFSET(17228, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
-    NAME_FUNC_OFFSET(17240, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
-    NAME_FUNC_OFFSET(17253, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
-    NAME_FUNC_OFFSET(17265, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
-    NAME_FUNC_OFFSET(17278, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
-    NAME_FUNC_OFFSET(17290, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
-    NAME_FUNC_OFFSET(17303, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
-    NAME_FUNC_OFFSET(17315, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
-    NAME_FUNC_OFFSET(17328, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
-    NAME_FUNC_OFFSET(17340, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
-    NAME_FUNC_OFFSET(17353, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
-    NAME_FUNC_OFFSET(17365, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
-    NAME_FUNC_OFFSET(17378, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
-    NAME_FUNC_OFFSET(17397, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
-    NAME_FUNC_OFFSET(17416, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
-    NAME_FUNC_OFFSET(17435, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
-    NAME_FUNC_OFFSET(17448, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
-    NAME_FUNC_OFFSET(17466, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
-    NAME_FUNC_OFFSET(17487, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
-    NAME_FUNC_OFFSET(17505, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
-    NAME_FUNC_OFFSET(17525, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17539, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17556, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET(17572, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET(17591, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17609, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17630, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17652, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17671, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17693, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17716, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET(17735, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET(17755, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET(17774, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET(17794, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET(17813, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET(17833, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET(17852, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET(17872, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET(17891, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET(17911, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET(17931, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET(17952, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET(17972, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET(17993, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET(18013, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET(18034, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET(18058, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET(18076, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET(18096, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET(18114, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET(18126, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET(18139, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET(18151, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET(18164, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18184, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18208, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18222, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18239, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18254, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18272, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18286, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18303, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18318, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18336, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18350, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18367, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18382, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18400, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18414, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18431, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18446, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18464, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18478, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18495, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18510, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18528, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18542, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18559, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18574, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18592, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18606, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18623, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18638, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18656, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18670, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18687, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18702, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18720, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(18737, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(18757, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(18774, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18800, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18829, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(18844, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(18862, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(18881, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18905, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18932, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
-    NAME_FUNC_OFFSET(18950, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
-    NAME_FUNC_OFFSET(18969, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
-    NAME_FUNC_OFFSET(18994, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
-    NAME_FUNC_OFFSET(19015, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
-    NAME_FUNC_OFFSET(19037, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
-    NAME_FUNC_OFFSET(19063, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
-    NAME_FUNC_OFFSET(19086, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
-    NAME_FUNC_OFFSET(19109, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
-    NAME_FUNC_OFFSET(19132, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
-    NAME_FUNC_OFFSET(19150, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
-    NAME_FUNC_OFFSET(19169, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
-    NAME_FUNC_OFFSET(19186, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
-    NAME_FUNC_OFFSET(19224, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
-    NAME_FUNC_OFFSET(19253, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
-    NAME_FUNC_OFFSET(19269, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
-    NAME_FUNC_OFFSET(19286, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
-    NAME_FUNC_OFFSET(19308, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_BlitFramebufferEXT),
-    NAME_FUNC_OFFSET(19326, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
+    NAME_FUNC_OFFSET( 8984, glFlushMappedBufferRange, glFlushMappedBufferRange, NULL, _gloffset_FlushMappedBufferRange),
+    NAME_FUNC_OFFSET( 9009, glMapBufferRange, glMapBufferRange, NULL, _gloffset_MapBufferRange),
+    NAME_FUNC_OFFSET( 9026, glBindVertexArray, glBindVertexArray, NULL, _gloffset_BindVertexArray),
+    NAME_FUNC_OFFSET( 9044, glGenVertexArrays, glGenVertexArrays, NULL, _gloffset_GenVertexArrays),
+    NAME_FUNC_OFFSET( 9062, glCopyBufferSubData, glCopyBufferSubData, NULL, _gloffset_CopyBufferSubData),
+    NAME_FUNC_OFFSET( 9082, glClientWaitSync, glClientWaitSync, NULL, _gloffset_ClientWaitSync),
+    NAME_FUNC_OFFSET( 9099, glDeleteSync, glDeleteSync, NULL, _gloffset_DeleteSync),
+    NAME_FUNC_OFFSET( 9112, glFenceSync, glFenceSync, NULL, _gloffset_FenceSync),
+    NAME_FUNC_OFFSET( 9124, glGetInteger64v, glGetInteger64v, NULL, _gloffset_GetInteger64v),
+    NAME_FUNC_OFFSET( 9140, glGetSynciv, glGetSynciv, NULL, _gloffset_GetSynciv),
+    NAME_FUNC_OFFSET( 9152, glIsSync, glIsSync, NULL, _gloffset_IsSync),
+    NAME_FUNC_OFFSET( 9161, glWaitSync, glWaitSync, NULL, _gloffset_WaitSync),
+    NAME_FUNC_OFFSET( 9172, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
+    NAME_FUNC_OFFSET( 9191, gl_dispatch_stub_575, gl_dispatch_stub_575, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9223, gl_dispatch_stub_576, gl_dispatch_stub_576, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9255, gl_dispatch_stub_577, gl_dispatch_stub_577, NULL, _gloffset_PixelTexGenParameterfSGIS),
+    NAME_FUNC_OFFSET( 9283, gl_dispatch_stub_578, gl_dispatch_stub_578, NULL, _gloffset_PixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9312, gl_dispatch_stub_579, gl_dispatch_stub_579, NULL, _gloffset_PixelTexGenParameteriSGIS),
+    NAME_FUNC_OFFSET( 9340, gl_dispatch_stub_580, gl_dispatch_stub_580, NULL, _gloffset_PixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9369, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET( 9386, gl_dispatch_stub_582, gl_dispatch_stub_582, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET( 9406, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
+    NAME_FUNC_OFFSET( 9424, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
+    NAME_FUNC_OFFSET( 9445, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
+    NAME_FUNC_OFFSET( 9463, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
+    NAME_FUNC_OFFSET( 9482, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
+    NAME_FUNC_OFFSET( 9503, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
+    NAME_FUNC_OFFSET( 9522, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET( 9543, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET( 9565, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
+    NAME_FUNC_OFFSET( 9581, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
+    NAME_FUNC_OFFSET( 9599, gl_dispatch_stub_593, gl_dispatch_stub_593, NULL, _gloffset_CullParameterdvEXT),
+    NAME_FUNC_OFFSET( 9620, gl_dispatch_stub_594, gl_dispatch_stub_594, NULL, _gloffset_CullParameterfvEXT),
+    NAME_FUNC_OFFSET( 9641, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET( 9663, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET( 9686, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET( 9708, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET( 9731, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET( 9753, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET( 9776, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET( 9798, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET( 9821, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET( 9843, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET( 9866, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET( 9889, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET( 9913, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET( 9936, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET( 9960, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET( 9983, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET(10007, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET(10034, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET(10055, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET(10078, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET(10099, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET(10114, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET(10130, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET(10145, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET(10161, gl_dispatch_stub_619, gl_dispatch_stub_619, NULL, _gloffset_PixelTexGenSGIX),
+    NAME_FUNC_OFFSET(10179, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(10202, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10228, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10249, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
+    NAME_FUNC_OFFSET(10267, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
+    NAME_FUNC_OFFSET(10286, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
+    NAME_FUNC_OFFSET(10309, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
+    NAME_FUNC_OFFSET(10333, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
+    NAME_FUNC_OFFSET(10356, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
+    NAME_FUNC_OFFSET(10380, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
+    NAME_FUNC_OFFSET(10403, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10435, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10467, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
+    NAME_FUNC_OFFSET(10500, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
+    NAME_FUNC_OFFSET(10533, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10570, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10607, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
+    NAME_FUNC_OFFSET(10627, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(10645, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(10664, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(10682, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(10701, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(10719, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(10738, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(10756, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(10775, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(10793, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(10812, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(10830, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(10849, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(10867, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(10886, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(10904, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(10923, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
+    NAME_FUNC_OFFSET(10941, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
+    NAME_FUNC_OFFSET(10960, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
+    NAME_FUNC_OFFSET(10978, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
+    NAME_FUNC_OFFSET(10997, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
+    NAME_FUNC_OFFSET(11015, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
+    NAME_FUNC_OFFSET(11034, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
+    NAME_FUNC_OFFSET(11052, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
+    NAME_FUNC_OFFSET(11071, gl_dispatch_stub_661, gl_dispatch_stub_661, NULL, _gloffset_MultiModeDrawArraysIBM),
+    NAME_FUNC_OFFSET(11096, gl_dispatch_stub_662, gl_dispatch_stub_662, NULL, _gloffset_MultiModeDrawElementsIBM),
+    NAME_FUNC_OFFSET(11123, gl_dispatch_stub_663, gl_dispatch_stub_663, NULL, _gloffset_DeleteFencesNV),
+    NAME_FUNC_OFFSET(11140, gl_dispatch_stub_664, gl_dispatch_stub_664, NULL, _gloffset_FinishFenceNV),
+    NAME_FUNC_OFFSET(11156, gl_dispatch_stub_665, gl_dispatch_stub_665, NULL, _gloffset_GenFencesNV),
+    NAME_FUNC_OFFSET(11170, gl_dispatch_stub_666, gl_dispatch_stub_666, NULL, _gloffset_GetFenceivNV),
+    NAME_FUNC_OFFSET(11185, gl_dispatch_stub_667, gl_dispatch_stub_667, NULL, _gloffset_IsFenceNV),
+    NAME_FUNC_OFFSET(11197, gl_dispatch_stub_668, gl_dispatch_stub_668, NULL, _gloffset_SetFenceNV),
+    NAME_FUNC_OFFSET(11210, gl_dispatch_stub_669, gl_dispatch_stub_669, NULL, _gloffset_TestFenceNV),
+    NAME_FUNC_OFFSET(11224, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
+    NAME_FUNC_OFFSET(11248, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(11264, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(11283, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
+    NAME_FUNC_OFFSET(11302, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(11318, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
+    NAME_FUNC_OFFSET(11344, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
+    NAME_FUNC_OFFSET(11370, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
+    NAME_FUNC_OFFSET(11391, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
+    NAME_FUNC_OFFSET(11408, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
+    NAME_FUNC_OFFSET(11429, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(11457, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
+    NAME_FUNC_OFFSET(11479, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
+    NAME_FUNC_OFFSET(11501, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
+    NAME_FUNC_OFFSET(11523, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(11537, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
+    NAME_FUNC_OFFSET(11553, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
+    NAME_FUNC_OFFSET(11578, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
+    NAME_FUNC_OFFSET(11603, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
+    NAME_FUNC_OFFSET(11631, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
+    NAME_FUNC_OFFSET(11647, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
+    NAME_FUNC_OFFSET(11666, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
+    NAME_FUNC_OFFSET(11686, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
+    NAME_FUNC_OFFSET(11705, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
+    NAME_FUNC_OFFSET(11725, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
+    NAME_FUNC_OFFSET(11744, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
+    NAME_FUNC_OFFSET(11764, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
+    NAME_FUNC_OFFSET(11783, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
+    NAME_FUNC_OFFSET(11803, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
+    NAME_FUNC_OFFSET(11822, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
+    NAME_FUNC_OFFSET(11842, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
+    NAME_FUNC_OFFSET(11861, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
+    NAME_FUNC_OFFSET(11881, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
+    NAME_FUNC_OFFSET(11900, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
+    NAME_FUNC_OFFSET(11920, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
+    NAME_FUNC_OFFSET(11939, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
+    NAME_FUNC_OFFSET(11959, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
+    NAME_FUNC_OFFSET(11978, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
+    NAME_FUNC_OFFSET(11998, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
+    NAME_FUNC_OFFSET(12017, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
+    NAME_FUNC_OFFSET(12037, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
+    NAME_FUNC_OFFSET(12056, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
+    NAME_FUNC_OFFSET(12076, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
+    NAME_FUNC_OFFSET(12095, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
+    NAME_FUNC_OFFSET(12115, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
+    NAME_FUNC_OFFSET(12135, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
+    NAME_FUNC_OFFSET(12156, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
+    NAME_FUNC_OFFSET(12180, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
+    NAME_FUNC_OFFSET(12201, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
+    NAME_FUNC_OFFSET(12222, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
+    NAME_FUNC_OFFSET(12243, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
+    NAME_FUNC_OFFSET(12264, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
+    NAME_FUNC_OFFSET(12285, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
+    NAME_FUNC_OFFSET(12306, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
+    NAME_FUNC_OFFSET(12327, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
+    NAME_FUNC_OFFSET(12348, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
+    NAME_FUNC_OFFSET(12369, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
+    NAME_FUNC_OFFSET(12390, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
+    NAME_FUNC_OFFSET(12411, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
+    NAME_FUNC_OFFSET(12432, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
+    NAME_FUNC_OFFSET(12454, glGetTexBumpParameterfvATI, glGetTexBumpParameterfvATI, NULL, _gloffset_GetTexBumpParameterfvATI),
+    NAME_FUNC_OFFSET(12481, glGetTexBumpParameterivATI, glGetTexBumpParameterivATI, NULL, _gloffset_GetTexBumpParameterivATI),
+    NAME_FUNC_OFFSET(12508, glTexBumpParameterfvATI, glTexBumpParameterfvATI, NULL, _gloffset_TexBumpParameterfvATI),
+    NAME_FUNC_OFFSET(12532, glTexBumpParameterivATI, glTexBumpParameterivATI, NULL, _gloffset_TexBumpParameterivATI),
+    NAME_FUNC_OFFSET(12556, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12578, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12600, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12622, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
+    NAME_FUNC_OFFSET(12647, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
+    NAME_FUNC_OFFSET(12671, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12693, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12715, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12737, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
+    NAME_FUNC_OFFSET(12763, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
+    NAME_FUNC_OFFSET(12786, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
+    NAME_FUNC_OFFSET(12810, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
+    NAME_FUNC_OFFSET(12828, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
+    NAME_FUNC_OFFSET(12843, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
+    NAME_FUNC_OFFSET(12874, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(12894, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(12915, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_ActiveStencilFaceEXT),
+    NAME_FUNC_OFFSET(12938, gl_dispatch_stub_751, gl_dispatch_stub_751, NULL, _gloffset_BindVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(12961, gl_dispatch_stub_752, gl_dispatch_stub_752, NULL, _gloffset_DeleteVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(12987, gl_dispatch_stub_753, gl_dispatch_stub_753, NULL, _gloffset_GenVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(13010, gl_dispatch_stub_754, gl_dispatch_stub_754, NULL, _gloffset_IsVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(13031, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
+    NAME_FUNC_OFFSET(13062, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
+    NAME_FUNC_OFFSET(13093, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
+    NAME_FUNC_OFFSET(13121, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
+    NAME_FUNC_OFFSET(13150, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
+    NAME_FUNC_OFFSET(13178, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
+    NAME_FUNC_OFFSET(13207, gl_dispatch_stub_761, gl_dispatch_stub_761, NULL, _gloffset_DepthBoundsEXT),
+    NAME_FUNC_OFFSET(13224, gl_dispatch_stub_762, gl_dispatch_stub_762, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(13251, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(13272, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(13294, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(13322, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(13346, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13371, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(13400, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(13426, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(13452, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(13478, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(13499, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13521, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(13541, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(13582, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(13614, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(13633, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(13653, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(13678, gl_dispatch_stub_780, gl_dispatch_stub_780, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(13699, gl_dispatch_stub_781, gl_dispatch_stub_781, NULL, _gloffset_BufferParameteriAPPLE),
+    NAME_FUNC_OFFSET(13723, gl_dispatch_stub_782, gl_dispatch_stub_782, NULL, _gloffset_FlushMappedBufferRangeAPPLE),
+    NAME_FUNC_OFFSET(13753, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
+    NAME_FUNC_OFFSET(13782, glProvokingVertexEXT, glProvokingVertexEXT, NULL, _gloffset_ProvokingVertexEXT),
+    NAME_FUNC_OFFSET(13803, gl_dispatch_stub_785, gl_dispatch_stub_785, NULL, _gloffset_GetTexParameterPointervAPPLE),
+    NAME_FUNC_OFFSET(13834, gl_dispatch_stub_786, gl_dispatch_stub_786, NULL, _gloffset_TextureRangeAPPLE),
+    NAME_FUNC_OFFSET(13854, gl_dispatch_stub_787, gl_dispatch_stub_787, NULL, _gloffset_StencilFuncSeparateATI),
+    NAME_FUNC_OFFSET(13879, gl_dispatch_stub_788, gl_dispatch_stub_788, NULL, _gloffset_ProgramEnvParameters4fvEXT),
+    NAME_FUNC_OFFSET(13908, gl_dispatch_stub_789, gl_dispatch_stub_789, NULL, _gloffset_ProgramLocalParameters4fvEXT),
+    NAME_FUNC_OFFSET(13939, gl_dispatch_stub_790, gl_dispatch_stub_790, NULL, _gloffset_GetQueryObjecti64vEXT),
+    NAME_FUNC_OFFSET(13963, gl_dispatch_stub_791, gl_dispatch_stub_791, NULL, _gloffset_GetQueryObjectui64vEXT),
+    NAME_FUNC_OFFSET(13988, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
+    NAME_FUNC_OFFSET(14006, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
+    NAME_FUNC_OFFSET(14023, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
+    NAME_FUNC_OFFSET(14039, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
+    NAME_FUNC_OFFSET(14064, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
+    NAME_FUNC_OFFSET(14084, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
+    NAME_FUNC_OFFSET(14104, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
+    NAME_FUNC_OFFSET(14127, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
+    NAME_FUNC_OFFSET(14150, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
+    NAME_FUNC_OFFSET(14170, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
+    NAME_FUNC_OFFSET(14187, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
+    NAME_FUNC_OFFSET(14204, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
+    NAME_FUNC_OFFSET(14219, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
+    NAME_FUNC_OFFSET(14243, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
+    NAME_FUNC_OFFSET(14262, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
+    NAME_FUNC_OFFSET(14281, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
+    NAME_FUNC_OFFSET(14297, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
+    NAME_FUNC_OFFSET(14316, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
+    NAME_FUNC_OFFSET(14339, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(14355, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(14371, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
+    NAME_FUNC_OFFSET(14398, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
+    NAME_FUNC_OFFSET(14425, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
+    NAME_FUNC_OFFSET(14445, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14464, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14483, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14513, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14543, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14573, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14603, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
+    NAME_FUNC_OFFSET(14622, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
+    NAME_FUNC_OFFSET(14645, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14670, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14695, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
+    NAME_FUNC_OFFSET(14722, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14750, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
+    NAME_FUNC_OFFSET(14777, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14805, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14834, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14863, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
+    NAME_FUNC_OFFSET(14889, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14920, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14951, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
+    NAME_FUNC_OFFSET(14975, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
+    NAME_FUNC_OFFSET(14998, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
+    NAME_FUNC_OFFSET(15016, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
+    NAME_FUNC_OFFSET(15045, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
+    NAME_FUNC_OFFSET(15074, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
+    NAME_FUNC_OFFSET(15089, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
+    NAME_FUNC_OFFSET(15115, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
+    NAME_FUNC_OFFSET(15141, glHistogram, glHistogram, NULL, _gloffset_Histogram),
+    NAME_FUNC_OFFSET(15156, glMinmax, glMinmax, NULL, _gloffset_Minmax),
+    NAME_FUNC_OFFSET(15168, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
+    NAME_FUNC_OFFSET(15188, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
+    NAME_FUNC_OFFSET(15205, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
+    NAME_FUNC_OFFSET(15221, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
+    NAME_FUNC_OFFSET(15240, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
+    NAME_FUNC_OFFSET(15263, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
+    NAME_FUNC_OFFSET(15279, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
+    NAME_FUNC_OFFSET(15301, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
+    NAME_FUNC_OFFSET(15319, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
+    NAME_FUNC_OFFSET(15338, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
+    NAME_FUNC_OFFSET(15356, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
+    NAME_FUNC_OFFSET(15375, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
+    NAME_FUNC_OFFSET(15393, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
+    NAME_FUNC_OFFSET(15412, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
+    NAME_FUNC_OFFSET(15430, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
+    NAME_FUNC_OFFSET(15449, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
+    NAME_FUNC_OFFSET(15467, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
+    NAME_FUNC_OFFSET(15486, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
+    NAME_FUNC_OFFSET(15504, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
+    NAME_FUNC_OFFSET(15523, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
+    NAME_FUNC_OFFSET(15541, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
+    NAME_FUNC_OFFSET(15560, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
+    NAME_FUNC_OFFSET(15578, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
+    NAME_FUNC_OFFSET(15597, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
+    NAME_FUNC_OFFSET(15615, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
+    NAME_FUNC_OFFSET(15634, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
+    NAME_FUNC_OFFSET(15652, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
+    NAME_FUNC_OFFSET(15671, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
+    NAME_FUNC_OFFSET(15689, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
+    NAME_FUNC_OFFSET(15708, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
+    NAME_FUNC_OFFSET(15726, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
+    NAME_FUNC_OFFSET(15745, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
+    NAME_FUNC_OFFSET(15763, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
+    NAME_FUNC_OFFSET(15782, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
+    NAME_FUNC_OFFSET(15800, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
+    NAME_FUNC_OFFSET(15819, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
+    NAME_FUNC_OFFSET(15837, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
+    NAME_FUNC_OFFSET(15856, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
+    NAME_FUNC_OFFSET(15874, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
+    NAME_FUNC_OFFSET(15893, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
+    NAME_FUNC_OFFSET(15916, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15939, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(15962, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15985, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(16008, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
+    NAME_FUNC_OFFSET(16025, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
+    NAME_FUNC_OFFSET(16048, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
+    NAME_FUNC_OFFSET(16071, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
+    NAME_FUNC_OFFSET(16094, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
+    NAME_FUNC_OFFSET(16120, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
+    NAME_FUNC_OFFSET(16146, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
+    NAME_FUNC_OFFSET(16172, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
+    NAME_FUNC_OFFSET(16196, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(16223, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(16249, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
+    NAME_FUNC_OFFSET(16269, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
+    NAME_FUNC_OFFSET(16289, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
+    NAME_FUNC_OFFSET(16309, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
+    NAME_FUNC_OFFSET(16332, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
+    NAME_FUNC_OFFSET(16356, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
+    NAME_FUNC_OFFSET(16379, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
+    NAME_FUNC_OFFSET(16403, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
+    NAME_FUNC_OFFSET(16420, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
+    NAME_FUNC_OFFSET(16438, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
+    NAME_FUNC_OFFSET(16455, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
+    NAME_FUNC_OFFSET(16473, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
+    NAME_FUNC_OFFSET(16490, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
+    NAME_FUNC_OFFSET(16508, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
+    NAME_FUNC_OFFSET(16525, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
+    NAME_FUNC_OFFSET(16543, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
+    NAME_FUNC_OFFSET(16560, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
+    NAME_FUNC_OFFSET(16578, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
+    NAME_FUNC_OFFSET(16595, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
+    NAME_FUNC_OFFSET(16613, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
+    NAME_FUNC_OFFSET(16630, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
+    NAME_FUNC_OFFSET(16648, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
+    NAME_FUNC_OFFSET(16665, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
+    NAME_FUNC_OFFSET(16683, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
+    NAME_FUNC_OFFSET(16700, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
+    NAME_FUNC_OFFSET(16718, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
+    NAME_FUNC_OFFSET(16737, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
+    NAME_FUNC_OFFSET(16756, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
+    NAME_FUNC_OFFSET(16775, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
+    NAME_FUNC_OFFSET(16794, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
+    NAME_FUNC_OFFSET(16814, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
+    NAME_FUNC_OFFSET(16834, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
+    NAME_FUNC_OFFSET(16854, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
+    NAME_FUNC_OFFSET(16872, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
+    NAME_FUNC_OFFSET(16889, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
+    NAME_FUNC_OFFSET(16907, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
+    NAME_FUNC_OFFSET(16924, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
+    NAME_FUNC_OFFSET(16942, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
+    NAME_FUNC_OFFSET(16960, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
+    NAME_FUNC_OFFSET(16977, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
+    NAME_FUNC_OFFSET(16995, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
+    NAME_FUNC_OFFSET(17014, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
+    NAME_FUNC_OFFSET(17033, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
+    NAME_FUNC_OFFSET(17052, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
+    NAME_FUNC_OFFSET(17074, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
+    NAME_FUNC_OFFSET(17087, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
+    NAME_FUNC_OFFSET(17100, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
+    NAME_FUNC_OFFSET(17116, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
+    NAME_FUNC_OFFSET(17132, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
+    NAME_FUNC_OFFSET(17145, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
+    NAME_FUNC_OFFSET(17168, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
+    NAME_FUNC_OFFSET(17188, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
+    NAME_FUNC_OFFSET(17207, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
+    NAME_FUNC_OFFSET(17218, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
+    NAME_FUNC_OFFSET(17230, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
+    NAME_FUNC_OFFSET(17244, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
+    NAME_FUNC_OFFSET(17257, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
+    NAME_FUNC_OFFSET(17273, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
+    NAME_FUNC_OFFSET(17284, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
+    NAME_FUNC_OFFSET(17297, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
+    NAME_FUNC_OFFSET(17316, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
+    NAME_FUNC_OFFSET(17336, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
+    NAME_FUNC_OFFSET(17349, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
+    NAME_FUNC_OFFSET(17359, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
+    NAME_FUNC_OFFSET(17375, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
+    NAME_FUNC_OFFSET(17394, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
+    NAME_FUNC_OFFSET(17412, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
+    NAME_FUNC_OFFSET(17433, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
+    NAME_FUNC_OFFSET(17448, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
+    NAME_FUNC_OFFSET(17463, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
+    NAME_FUNC_OFFSET(17477, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
+    NAME_FUNC_OFFSET(17492, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
+    NAME_FUNC_OFFSET(17504, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
+    NAME_FUNC_OFFSET(17517, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
+    NAME_FUNC_OFFSET(17529, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
+    NAME_FUNC_OFFSET(17542, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
+    NAME_FUNC_OFFSET(17554, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
+    NAME_FUNC_OFFSET(17567, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
+    NAME_FUNC_OFFSET(17579, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
+    NAME_FUNC_OFFSET(17592, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
+    NAME_FUNC_OFFSET(17604, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
+    NAME_FUNC_OFFSET(17617, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
+    NAME_FUNC_OFFSET(17629, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
+    NAME_FUNC_OFFSET(17642, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
+    NAME_FUNC_OFFSET(17654, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
+    NAME_FUNC_OFFSET(17667, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
+    NAME_FUNC_OFFSET(17679, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
+    NAME_FUNC_OFFSET(17692, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
+    NAME_FUNC_OFFSET(17711, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
+    NAME_FUNC_OFFSET(17730, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
+    NAME_FUNC_OFFSET(17749, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
+    NAME_FUNC_OFFSET(17762, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
+    NAME_FUNC_OFFSET(17780, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
+    NAME_FUNC_OFFSET(17801, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
+    NAME_FUNC_OFFSET(17819, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
+    NAME_FUNC_OFFSET(17839, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17853, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17870, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET(17886, gl_dispatch_stub_582, gl_dispatch_stub_582, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET(17905, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17923, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17944, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17966, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17985, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(18007, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(18030, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET(18049, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET(18069, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET(18088, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET(18108, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET(18127, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET(18147, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET(18166, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET(18186, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET(18205, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET(18225, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET(18245, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET(18266, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET(18286, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET(18307, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET(18327, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET(18348, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET(18372, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET(18390, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET(18410, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET(18428, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET(18440, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET(18453, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET(18465, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET(18478, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18498, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18522, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18536, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18553, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18568, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18586, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18600, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18617, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18632, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18650, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18664, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18681, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18696, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18714, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18728, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18745, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18760, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18778, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18792, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18809, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18824, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18842, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18856, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18873, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18888, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18906, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18920, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18937, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18952, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18970, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(18984, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(19001, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(19016, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(19034, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(19051, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(19071, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(19088, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(19114, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(19143, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(19158, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(19176, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(19195, gl_dispatch_stub_752, gl_dispatch_stub_752, NULL, _gloffset_DeleteVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(19216, gl_dispatch_stub_754, gl_dispatch_stub_754, NULL, _gloffset_IsVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(19232, gl_dispatch_stub_762, gl_dispatch_stub_762, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(19256, gl_dispatch_stub_762, gl_dispatch_stub_762, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(19283, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(19301, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(19320, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(19345, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(19366, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(19388, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(19414, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(19437, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(19460, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(19483, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(19501, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(19520, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(19537, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(19575, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(19604, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(19620, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(19637, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(19659, gl_dispatch_stub_780, gl_dispatch_stub_780, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(19677, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
     NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0)
 };
 
diff --git a/src/mesa/glapi/glthread.c b/src/mesa/glapi/glthread.c
index e3abb0f4ae..737fd4d6a8 100644
--- a/src/mesa/glapi/glthread.c
+++ b/src/mesa/glapi/glthread.c
@@ -246,57 +246,6 @@ _glthread_SetTSD(_glthread_TSD *tsd, void *ptr)
 
 #endif /* WIN32_THREADS */
 
-
-
-/*
- * XFree86 has its own thread wrapper, Xthreads.h
- * We wrap it again for GL.
- */
-#ifdef USE_XTHREADS
-
-unsigned long
-_glthread_GetID(void)
-{
-   return (unsigned long) xthread_self();
-}
-
-
-void
-_glthread_InitTSD(_glthread_TSD *tsd)
-{
-   if (xthread_key_create(&tsd->key, NULL) != 0) {
-      perror(INIT_TSD_ERROR);
-      exit(-1);
-   }
-   tsd->initMagic = INIT_MAGIC;
-}
-
-
-void *
-_glthread_GetTSD(_glthread_TSD *tsd)
-{
-   void *ptr;
-   if (tsd->initMagic != INIT_MAGIC) {
-      _glthread_InitTSD(tsd);
-   }
-   xthread_get_specific(tsd->key, &ptr);
-   return ptr;
-}
-
-
-void
-_glthread_SetTSD(_glthread_TSD *tsd, void *ptr)
-{
-   if (tsd->initMagic != INIT_MAGIC) {
-      _glthread_InitTSD(tsd);
-   }
-   xthread_set_specific(tsd->key, ptr);
-}
-
-#endif /* XTHREAD */
-
-
-
 /*
  * BeOS threads
  */
diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h
index dfe09a9d59..8ec933a851 100644
--- a/src/mesa/glapi/glthread.h
+++ b/src/mesa/glapi/glthread.h
@@ -71,7 +71,7 @@
 
 
 #if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\
-     defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \
+     defined(WIN32_THREADS) || defined(BEOS_THREADS)) \
     && !defined(THREADS)
 # define THREADS
 #endif
@@ -218,48 +218,6 @@ typedef CRITICAL_SECTION _glthread_Mutex;
 #endif /* WIN32_THREADS */
 
 
-
-
-/*
- * XFree86 has its own thread wrapper, Xthreads.h
- * We wrap it again for GL.
- */
-#ifdef USE_XTHREADS
-#include <X11/Xthreads.h>
-
-typedef struct {
-   xthread_key_t key;
-   int initMagic;
-} _glthread_TSD;
-
-typedef xthread_t _glthread_Thread;
-
-typedef xmutex_rec _glthread_Mutex;
-
-#ifdef XMUTEX_INITIALIZER
-#define _glthread_DECLARE_STATIC_MUTEX(name) \
-   static _glthread_Mutex name = XMUTEX_INITIALIZER
-#else
-#define _glthread_DECLARE_STATIC_MUTEX(name) \
-   static _glthread_Mutex name
-#endif
-
-#define _glthread_INIT_MUTEX(name) \
-   xmutex_init(&(name))
-
-#define _glthread_DESTROY_MUTEX(name) \
-   xmutex_clear(&(name))
-
-#define _glthread_LOCK_MUTEX(name) \
-   (void) xmutex_lock(&(name))
-
-#define _glthread_UNLOCK_MUTEX(name) \
-   (void) xmutex_unlock(&(name))
-
-#endif /* USE_XTHREADS */
-
-
-
 /*
  * BeOS threads. R5.x required.
  */
diff --git a/src/mesa/main/api_arrayelt.c b/src/mesa/main/api_arrayelt.c
index d124c724c9..2462a1b003 100644
--- a/src/mesa/main/api_arrayelt.c
+++ b/src/mesa/main/api_arrayelt.c
@@ -28,6 +28,7 @@
 
 #include "glheader.h"
 #include "api_arrayelt.h"
+#include "bufferobj.h"
 #include "context.h"
 #include "imports.h"
 #include "macros.h"
@@ -1071,7 +1072,7 @@ void _ae_destroy_context( GLcontext *ctx )
 static void check_vbo( AEcontext *actx,
 		       struct gl_buffer_object *vbo )
 {
-   if (vbo->Name && !vbo->Pointer) {
+   if (_mesa_is_bufferobj(vbo) && !_mesa_bufferobj_mapped(vbo)) {
       GLuint i;
       for (i = 0; i < actx->nr_vbos; i++)
 	 if (actx->vbo[i] == vbo)
@@ -1094,48 +1095,49 @@ static void _ae_update_state( GLcontext *ctx )
    AEarray *aa = actx->arrays;
    AEattrib *at = actx->attribs;
    GLuint i;
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
 
    actx->nr_vbos = 0;
 
    /* conventional vertex arrays */
-  if (ctx->Array.ArrayObj->Index.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Index;
+   if (arrayObj->Index.Enabled) {
+      aa->array = &arrayObj->Index;
       aa->offset = IndexFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->EdgeFlag;
+   if (arrayObj->EdgeFlag.Enabled) {
+      aa->array = &arrayObj->EdgeFlag;
       aa->offset = _gloffset_EdgeFlagv;
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->Normal.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Normal;
+   if (arrayObj->Normal.Enabled) {
+      aa->array = &arrayObj->Normal;
       aa->offset = NormalFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->Color.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Color;
+   if (arrayObj->Color.Enabled) {
+      aa->array = &arrayObj->Color;
       aa->offset = ColorFuncs[aa->array->Size-3][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->SecondaryColor;
+   if (arrayObj->SecondaryColor.Enabled) {
+      aa->array = &arrayObj->SecondaryColor;
       aa->offset = SecondaryColorFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->FogCoord;
+   if (arrayObj->FogCoord.Enabled) {
+      aa->array = &arrayObj->FogCoord;
       aa->offset = FogCoordFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
    for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      struct gl_client_array *attribArray = &ctx->Array.ArrayObj->TexCoord[i];
+      struct gl_client_array *attribArray = &arrayObj->TexCoord[i];
       if (attribArray->Enabled) {
          /* NOTE: we use generic glVertexAttribNV functions here.
           * If we ever remove GL_NV_vertex_program this will have to change.
@@ -1152,8 +1154,8 @@ static void _ae_update_state( GLcontext *ctx )
    }
 
    /* generic vertex attribute arrays */   
-   for (i = 1; i < VERT_ATTRIB_MAX; i++) {  /* skip zero! */
-      struct gl_client_array *attribArray = &ctx->Array.ArrayObj->VertexAttrib[i];
+   for (i = 1; i < Elements(arrayObj->VertexAttrib); i++) {  /* skip zero! */
+      struct gl_client_array *attribArray = &arrayObj->VertexAttrib[i];
       if (attribArray->Enabled) {
          at->array = attribArray;
          /* Note: we can't grab the _glapi_Dispatch->VertexAttrib1fvNV
@@ -1179,18 +1181,18 @@ static void _ae_update_state( GLcontext *ctx )
    }
 
    /* finally, vertex position */
-   if (ctx->Array.ArrayObj->VertexAttrib[0].Enabled) {
+   if (arrayObj->VertexAttrib[0].Enabled) {
       /* Use glVertex(v) instead of glVertexAttrib(0, v) to be sure it's
        * issued as the last (provoking) attribute).
        */
-      aa->array = &ctx->Array.ArrayObj->VertexAttrib[0];
+      aa->array = &arrayObj->VertexAttrib[0];
       assert(aa->array->Size >= 2); /* XXX fix someday? */
       aa->offset = VertexFuncs[aa->array->Size-2][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   else if (ctx->Array.ArrayObj->Vertex.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Vertex;
+   else if (arrayObj->Vertex.Enabled) {
+      aa->array = &arrayObj->Vertex;
       aa->offset = VertexFuncs[aa->array->Size-2][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c
index 6f66ff47a0..02550ae108 100644
--- a/src/mesa/main/api_exec.c
+++ b/src/mesa/main/api_exec.c
@@ -107,6 +107,7 @@
 #include "state.h"
 #include "stencil.h"
 #include "texenv.h"
+#include "texgetimage.h"
 #include "teximage.h"
 #if FEATURE_texgen
 #include "texgen.h"
@@ -128,6 +129,9 @@
 #if FEATURE_ARB_shader_objects
 #include "shaders.h"
 #endif
+#if FEATURE_ARB_sync
+#include "syncobj.h"
+#endif
 #include "debug.h"
 #include "glapi/dispatch.h"
 
@@ -528,7 +532,6 @@ _mesa_init_exec_table(struct _glapi_table *exec)
    /* 148. GL_EXT_multi_draw_arrays */
 #if _HAVE_FULL_GL
    SET_MultiDrawArraysEXT(exec, _mesa_MultiDrawArraysEXT);
-   SET_MultiDrawElementsEXT(exec, _mesa_MultiDrawElementsEXT);
 #endif
 
    /* 173. GL_INGR_blend_func_separate */
@@ -641,6 +644,8 @@ _mesa_init_exec_table(struct _glapi_table *exec)
    /* ???. GL_EXT_depth_bounds_test */
    SET_DepthBoundsEXT(exec, _mesa_DepthBoundsEXT);
 
+   SET_ProvokingVertexEXT(exec, _mesa_ProvokingVertexEXT);
+
    /* ARB 1. GL_ARB_multitexture */
 #if _HAVE_FULL_GL
    SET_ActiveTextureARB(exec, _mesa_ActiveTextureARB);
@@ -821,6 +826,17 @@ _mesa_init_exec_table(struct _glapi_table *exec)
    SET_GetAttribLocationARB(exec, _mesa_GetAttribLocationARB);
 #endif    /* FEATURE_ARB_vertex_shader */
 
+   /* GL_ARB_sync */
+#if FEATURE_ARB_sync
+   SET_IsSync(exec, _mesa_IsSync);
+   SET_DeleteSync(exec, _mesa_DeleteSync);
+   SET_FenceSync(exec, _mesa_FenceSync);
+   SET_ClientWaitSync(exec, _mesa_ClientWaitSync);
+   SET_WaitSync(exec, _mesa_WaitSync);
+   SET_GetInteger64v(exec, _mesa_GetInteger64v);
+   SET_GetSynciv(exec, _mesa_GetSynciv);
+#endif
+
   /* GL_ATI_fragment_shader */
 #if FEATURE_ATI_fragment_shader
    SET_GenFragmentShadersATI(exec, _mesa_GenFragmentShadersATI);
@@ -894,5 +910,16 @@ _mesa_init_exec_table(struct _glapi_table *exec)
     */
    SET_RenderbufferStorageMultisample(exec, _mesa_RenderbufferStorageMultisample);
 #endif
-}
 
+#if FEATURE_ARB_map_buffer_range
+   SET_MapBufferRange(exec, _mesa_MapBufferRange);
+   SET_FlushMappedBufferRange(exec, _mesa_FlushMappedBufferRange);
+#endif
+
+   /* GL_ARB_copy_buffer */
+   SET_CopyBufferSubData(exec, _mesa_CopyBufferSubData);
+
+   /* GL_ARB_vertex_array_object */
+   SET_BindVertexArray(exec, _mesa_BindVertexArray);
+   SET_GenVertexArrays(exec, _mesa_GenVertexArrays);
+}
diff --git a/src/mesa/main/api_noop.c b/src/mesa/main/api_noop.c
index 66f9c4e6bd..09ba7e5062 100644
--- a/src/mesa/main/api_noop.c
+++ b/src/mesa/main/api_noop.c
@@ -772,6 +772,20 @@ _mesa_noop_DrawRangeElements(GLenum mode,
        CALL_DrawElements(GET_DISPATCH(), (mode, count, type, indices));
 }
 
+/* GL_EXT_multi_draw_arrays */
+void GLAPIENTRY
+_mesa_noop_MultiDrawElements(GLenum mode, const GLsizei *count, GLenum type,
+			     const GLvoid **indices, GLsizei primcount)
+{
+   GLsizei i;
+
+   for (i = 0; i < primcount; i++) {
+      if (count[i] > 0) {
+	 CALL_DrawElements(GET_DISPATCH(), (mode, count[i], type, indices[i]));
+      }
+   }
+}
+
 /*
  * Eval Mesh
  */
@@ -980,6 +994,7 @@ _mesa_noop_vtxfmt_init( GLvertexformat *vfmt )
    vfmt->DrawArrays = _mesa_noop_DrawArrays;
    vfmt->DrawElements = _mesa_noop_DrawElements;
    vfmt->DrawRangeElements = _mesa_noop_DrawRangeElements;
+   vfmt->MultiDrawElementsEXT = _mesa_noop_MultiDrawElements;
    vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
    vfmt->EvalMesh2 = _mesa_noop_EvalMesh2;
 }
diff --git a/src/mesa/main/api_noop.h b/src/mesa/main/api_noop.h
index 8bf4660800..a7956d00b3 100644
--- a/src/mesa/main/api_noop.h
+++ b/src/mesa/main/api_noop.h
@@ -40,6 +40,10 @@ _mesa_noop_EvalMesh2(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2);
 extern void GLAPIENTRY
 _mesa_noop_Materialfv(GLenum face, GLenum pname, const GLfloat *param);
 
+extern void GLAPIENTRY
+_mesa_noop_MultiDrawElements(GLenum mode, const GLsizei *count, GLenum type,
+			     const GLvoid **indices, GLsizei primcount);
+
 extern void
 _mesa_noop_vtxfmt_init(GLvertexformat *vfmt);
 
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 2c6f370df9..2df4f17389 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -24,12 +24,33 @@
 
 #include "glheader.h"
 #include "api_validate.h"
+#include "bufferobj.h"
 #include "context.h"
 #include "imports.h"
 #include "mtypes.h"
 #include "state.h"
 
 
+
+/**
+ * \return  number of bytes in array [count] of type.
+ */
+static GLsizei
+index_bytes(GLenum type, GLsizei count)
+{
+   if (type == GL_UNSIGNED_INT) {
+      return count * sizeof(GLuint);
+   }
+   else if (type == GL_UNSIGNED_BYTE) {
+      return count * sizeof(GLubyte);
+   }
+   else {
+      ASSERT(type == GL_UNSIGNED_SHORT);
+      return count * sizeof(GLushort);
+   }
+}
+
+
 /**
  * Find the max index in the given element/index buffer
  */
@@ -42,12 +63,10 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
    GLuint max = 0;
    GLuint i;
 
-   if (elementBuf->Name) {
+   if (_mesa_is_bufferobj(elementBuf)) {
       /* elements are in a user-defined buffer object.  need to map it */
-      map = ctx->Driver.MapBuffer(ctx,
-                                  GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                  GL_READ_ONLY,
-                                  elementBuf);
+      map = ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER,
+                                  GL_READ_ONLY, elementBuf);
       /* Actual address is the sum of pointers */
       indices = (const GLvoid *) ADD_POINTERS(map, (const GLubyte *) indices);
    }
@@ -70,20 +89,20 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
    }
 
    if (map) {
-      ctx->Driver.UnmapBuffer(ctx,
-                              GL_ELEMENT_ARRAY_BUFFER_ARB,
-                              ctx->Array.ElementArrayBufferObj);
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuf);
    }
 
    return max;
 }
 
+
+/**
+ * Check if OK to draw arrays/elements.
+ */
 static GLboolean
-check_valid_to_render(GLcontext *ctx, char *function)
+check_valid_to_render(GLcontext *ctx, const char *function)
 {
-   if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
-      _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
-                  "glDraw%s(incomplete framebuffer)", function);
+   if (!_mesa_valid_to_render(ctx, function)) {
       return GL_FALSE;
    }
 
@@ -105,6 +124,12 @@ check_valid_to_render(GLcontext *ctx, char *function)
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawElements(GLcontext *ctx,
 			    GLenum mode, GLsizei count, GLenum type,
@@ -134,33 +159,14 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   if (!check_valid_to_render(ctx, "Elements"))
+   if (!check_valid_to_render(ctx, "glDrawElements"))
       return GL_FALSE;
 
    /* Vertex buffer object tests */
-   if (ctx->Array.ElementArrayBufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
       /* use indices in the buffer object */
-      GLuint indexBytes;
-
-      if (!ctx->Array.ElementArrayBufferObj->Size) {
-         _mesa_warning(ctx,
-                       "glDrawElements called with empty array elements buffer");
-         return GL_FALSE;
-      }
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > (GLuint) ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -175,8 +181,10 @@ _mesa_validate_DrawElements(GLcontext *ctx,
       /* find max array index */
       GLuint max = max_buffer_index(ctx, count, type, indices,
                                     ctx->Array.ElementArrayBufferObj);
-      if (max >= ctx->Array._MaxElement) {
+      if (max >= ctx->Array.ArrayObj->_MaxElement) {
          /* the max element is out of bounds of one or more enabled arrays */
+         _mesa_warning(ctx, "glDrawElements() index=%u is "
+                       "out of bounds (max=%u)", max, ctx->Array.ArrayObj->_MaxElement);
          return GL_FALSE;
       }
    }
@@ -184,6 +192,12 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawRangeElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 				 GLuint start, GLuint end,
@@ -218,27 +232,14 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   if (!check_valid_to_render(ctx, "RangeElements"))
+   if (!check_valid_to_render(ctx, "glDrawRangeElements"))
       return GL_FALSE;
 
    /* Vertex buffer object tests */
-   if (ctx->Array.ElementArrayBufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
       /* use indices in the buffer object */
-      GLsizei indexBytes;
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawRangeElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -252,7 +253,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    if (ctx->Const.CheckArrayBounds) {
       GLuint max = max_buffer_index(ctx, count, type, indices,
                                     ctx->Array.ElementArrayBufferObj);
-      if (max >= ctx->Array._MaxElement) {
+      if (max >= ctx->Array.ArrayObj->_MaxElement) {
          /* the max element is out of bounds of one or more enabled arrays */
          return GL_FALSE;
       }
@@ -265,6 +266,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 /**
  * Called from the tnl module to error check the function parameters and
  * verify that we really can draw something.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
  */
 GLboolean
 _mesa_validate_DrawArrays(GLcontext *ctx,
@@ -286,11 +288,11 @@ _mesa_validate_DrawArrays(GLcontext *ctx,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   if (!check_valid_to_render(ctx, "Arrays"))
+   if (!check_valid_to_render(ctx, "glDrawArrays"))
       return GL_FALSE;
 
    if (ctx->Const.CheckArrayBounds) {
-      if (start + count > (GLint) ctx->Array._MaxElement)
+      if (start + count > (GLint) ctx->Array.ArrayObj->_MaxElement)
          return GL_FALSE;
    }
 
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index 0fa5f0de55..fd35d4e38c 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -1,9 +1,10 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  * (C) Copyright IBM Corporation 2006
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -46,6 +47,7 @@
 #include "bufferobj.h"
 #endif
 #include "arrayobj.h"
+#include "macros.h"
 #include "glapi/dispatch.h"
 
 
@@ -61,10 +63,11 @@
 static INLINE struct gl_array_object *
 lookup_arrayobj(GLcontext *ctx, GLuint id)
 {
-   return (id == 0) 
-     ? NULL 
-     : (struct gl_array_object *) _mesa_HashLookup(ctx->Shared->ArrayObjects,
-						   id);
+   if (id == 0)
+      return NULL;
+   else
+      return (struct gl_array_object *)
+         _mesa_HashLookup(ctx->Array.Objects, id);
 }
 
 
@@ -79,6 +82,7 @@ unbind_array_object_vbos(GLcontext *ctx, struct gl_array_object *obj)
    GLuint i;
 
    _mesa_reference_buffer_object(ctx, &obj->Vertex.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->Weight.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->Normal.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->Color.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->SecondaryColor.BufferObj, NULL);
@@ -86,10 +90,10 @@ unbind_array_object_vbos(GLcontext *ctx, struct gl_array_object *obj)
    _mesa_reference_buffer_object(ctx, &obj->Index.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->EdgeFlag.BufferObj, NULL);
 
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+   for (i = 0; i < Elements(obj->TexCoord); i++)
       _mesa_reference_buffer_object(ctx, &obj->TexCoord[i].BufferObj, NULL);
 
-   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+   for (i = 0; i < Elements(obj->VertexAttrib); i++)
       _mesa_reference_buffer_object(ctx, &obj->VertexAttrib[i].BufferObj,NULL);
 }
 
@@ -198,7 +202,8 @@ init_array(GLcontext *ctx,
    array->Normalized = GL_FALSE;
 #if FEATURE_ARB_vertex_buffer_object
    /* Vertex array buffers */
-   array->BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &array->BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
 
@@ -220,16 +225,17 @@ _mesa_initialize_array_object( GLcontext *ctx,
 
    /* Init the individual arrays */
    init_array(ctx, &obj->Vertex, 4, GL_FLOAT);
+   init_array(ctx, &obj->Weight, 1, GL_FLOAT);
    init_array(ctx, &obj->Normal, 3, GL_FLOAT);
    init_array(ctx, &obj->Color, 4, GL_FLOAT);
    init_array(ctx, &obj->SecondaryColor, 4, GL_FLOAT);
    init_array(ctx, &obj->FogCoord, 1, GL_FLOAT);
    init_array(ctx, &obj->Index, 1, GL_FLOAT);
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+   for (i = 0; i < Elements(obj->TexCoord); i++) {
       init_array(ctx, &obj->TexCoord[i], 4, GL_FLOAT);
    }
    init_array(ctx, &obj->EdgeFlag, 1, GL_BOOL);
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(obj->VertexAttrib); i++) {
       init_array(ctx, &obj->VertexAttrib[i], 4, GL_FLOAT);
    }
 
@@ -242,12 +248,12 @@ _mesa_initialize_array_object( GLcontext *ctx,
 /**
  * Add the given array object to the array object pool.
  */
-void
-_mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+save_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* insert into hash table */
-      _mesa_HashInsert(ctx->Shared->ArrayObjects, obj->Name, obj);
+      _mesa_HashInsert(ctx->Array.Objects, obj->Name, obj);
    }
 }
 
@@ -256,13 +262,90 @@ _mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
  * Remove the given array object from the array object pool.
  * Do not deallocate the array object though.
  */
-void
-_mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* remove from hash table */
-      _mesa_HashRemove(ctx->Shared->ArrayObjects, obj->Name);
+      _mesa_HashRemove(ctx->Array.Objects, obj->Name);
+   }
+}
+
+
+
+/**
+ * Compute the index of the last array element that can be safely accessed
+ * in a vertex array.  We can really only do this when the array lives in
+ * a VBO.
+ * The array->_MaxElement field will be updated.
+ * Later in glDrawArrays/Elements/etc we can do some bounds checking.
+ */
+static void
+compute_max_element(struct gl_client_array *array)
+{
+   if (array->BufferObj->Name) {
+      /* Compute the max element we can access in the VBO without going
+       * out of bounds.
+       */
+      array->_MaxElement = ((GLsizeiptrARB) array->BufferObj->Size
+                            - (GLsizeiptrARB) array->Ptr + array->StrideB
+                            - array->_ElementSize) / array->StrideB;
+      if (0)
+         _mesa_printf("%s Object %u  Size %u  MaxElement %u\n",
+                      __FUNCTION__,
+                      array->BufferObj->Name,
+                      (GLuint) array->BufferObj->Size,
+                      array->_MaxElement);
    }
+   else {
+      /* user-space array, no idea how big it is */
+      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
+   }
+}
+
+
+/**
+ * Helper for update_arrays().
+ * \return  min(current min, array->_MaxElement).
+ */
+static GLuint
+update_min(GLuint min, struct gl_client_array *array)
+{
+   compute_max_element(array);
+   if (array->Enabled)
+      return MIN2(min, array->_MaxElement);
+   else
+      return min;
+}
+
+
+/**
+ * Examine vertex arrays to update the gl_array_object::_MaxElement field.
+ */
+void
+_mesa_update_array_object_max_element(GLcontext *ctx,
+                                      struct gl_array_object *arrayObj)
+{
+   GLuint i, min = ~0;
+
+   min = update_min(min, &arrayObj->Vertex);
+   min = update_min(min, &arrayObj->Weight);
+   min = update_min(min, &arrayObj->Normal);
+   min = update_min(min, &arrayObj->Color);
+   min = update_min(min, &arrayObj->SecondaryColor);
+   min = update_min(min, &arrayObj->FogCoord);
+   min = update_min(min, &arrayObj->Index);
+   min = update_min(min, &arrayObj->EdgeFlag);
+#if FEATURE_point_size_array
+   min = update_min(min, &arrayObj->PointSize);
+#endif
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
+      min = update_min(min, &arrayObj->TexCoord[i]);
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      min = update_min(min, &arrayObj->VertexAttrib[i]);
+
+   /* _MaxElement is one past the last legal array element */
+   arrayObj->_MaxElement = min;
 }
 
 
@@ -270,18 +353,15 @@ _mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 /* API Functions                                                      */
 /**********************************************************************/
 
+
 /**
- * Bind a new array.
- *
- * \todo
- * The binding could be done more efficiently by comparing the non-NULL
- * pointers in the old and new objects.  The only arrays that are "dirty" are
- * the ones that are non-NULL in either object.
+ * Helper for _mesa_BindVertexArray() and _mesa_BindVertexArrayAPPLE().
+ * \param genRequired  specifies behavour when id was not generated with
+ *                     glGenVertexArrays().
  */
-void GLAPIENTRY
-_mesa_BindVertexArrayAPPLE( GLuint id )
+static void
+bind_vertex_array(GLcontext *ctx, GLuint id, GLboolean genRequired)
 {
-   GET_CURRENT_CONTEXT(ctx);
    struct gl_array_object * const oldObj = ctx->Array.ArrayObj;
    struct gl_array_object *newObj = NULL;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -304,14 +384,18 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
       /* non-default array object */
       newObj = lookup_arrayobj(ctx, id);
       if (!newObj) {
-         /* If this is a new array object id, allocate an array object now.
-	  */
+         if (genRequired) {
+            _mesa_error(ctx, GL_INVALID_OPERATION, "glBindVertexArray(id)");
+            return;
+         }
+
+         /* For APPLE version, generate a new array object now */
 	 newObj = (*ctx->Driver.NewArrayObject)(ctx, id);
          if (!newObj) {
             _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindVertexArrayAPPLE");
             return;
          }
-         _mesa_save_array_object(ctx, newObj);
+         save_array_object(ctx, newObj);
       }
    }
 
@@ -321,7 +405,37 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
 
    /* Pass BindVertexArray call to device driver */
    if (ctx->Driver.BindArrayObject && newObj)
-      (*ctx->Driver.BindArrayObject)( ctx, newObj );
+      ctx->Driver.BindArrayObject(ctx, newObj);
+}
+
+
+/**
+ * ARB version of glBindVertexArray()
+ * This function behaves differently from glBindVertexArrayAPPLE() in
+ * that this function requires all ids to have been previously generated
+ * by glGenVertexArrays[APPLE]().
+ */
+void GLAPIENTRY
+_mesa_BindVertexArray( GLuint id )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   bind_vertex_array(ctx, id, GL_TRUE);
+}
+
+
+/**
+ * Bind a new array.
+ *
+ * \todo
+ * The binding could be done more efficiently by comparing the non-NULL
+ * pointers in the old and new objects.  The only arrays that are "dirty" are
+ * the ones that are non-NULL in either object.
+ */
+void GLAPIENTRY
+_mesa_BindVertexArrayAPPLE( GLuint id )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   bind_vertex_array(ctx, id, GL_FALSE);
 }
 
 
@@ -343,8 +457,6 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
       return;
    }
 
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
-
    for (i = 0; i < n; i++) {
       struct gl_array_object *obj = lookup_arrayobj(ctx, ids[i]);
 
@@ -360,7 +472,7 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
 	 }
 
 	 /* The ID is immediately freed for re-use */
-	 _mesa_remove_array_object(ctx, obj);
+	 remove_array_object(ctx, obj);
 
          /* Unreference the array object. 
           * If refcount hits zero, the object will be deleted.
@@ -368,21 +480,19 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
          _mesa_reference_array_object(ctx, &obj, NULL);
       }
    }
-
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
 }
 
 
 /**
  * Generate a set of unique array object IDs and store them in \c arrays.
- * 
+ * Helper for _mesa_GenVertexArrays[APPLE]() functions below.
  * \param n       Number of IDs to generate.
  * \param arrays  Array of \c n locations to store the IDs.
+ * \param vboOnly Will arrays have to reside in VBOs?
  */
-void GLAPIENTRY
-_mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
+static void 
+gen_vertex_arrays(GLcontext *ctx, GLsizei n, GLuint *arrays, GLboolean vboOnly)
 {
-   GET_CURRENT_CONTEXT(ctx);
    GLuint first;
    GLint i;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -396,12 +506,7 @@ _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
       return;
    }
 
-   /*
-    * This must be atomic (generation and allocation of array object IDs)
-    */
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
-
-   first = _mesa_HashFindFreeKeyBlock(ctx->Shared->ArrayObjects, n);
+   first = _mesa_HashFindFreeKeyBlock(ctx->Array.Objects, n);
 
    /* Allocate new, empty array objects and return identifiers */
    for (i = 0; i < n; i++) {
@@ -410,15 +515,37 @@ _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
 
       obj = (*ctx->Driver.NewArrayObject)( ctx, name );
       if (!obj) {
-         _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenVertexArraysAPPLE");
          return;
       }
-      _mesa_save_array_object(ctx, obj);
+      obj->VBOonly = vboOnly;
+      save_array_object(ctx, obj);
       arrays[i] = first + i;
    }
+}
+
+
+/**
+ * ARB version of glGenVertexArrays()
+ * All arrays will be required to live in VBOs.
+ */
+void GLAPIENTRY
+_mesa_GenVertexArrays(GLsizei n, GLuint *arrays)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   gen_vertex_arrays(ctx, n, arrays, GL_TRUE);
+}
+
 
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+/**
+ * APPLE version of glGenVertexArraysAPPLE()
+ * Arrays may live in VBOs or ordinary memory.
+ */
+void GLAPIENTRY
+_mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   gen_vertex_arrays(ctx, n, arrays, GL_FALSE);
 }
 
 
@@ -439,9 +566,7 @@ _mesa_IsVertexArrayAPPLE( GLuint id )
    if (id == 0)
       return GL_FALSE;
 
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
    obj = lookup_arrayobj(ctx, id);
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
 
    return (obj != NULL) ? GL_TRUE : GL_FALSE;
 }
diff --git a/src/mesa/main/arrayobj.h b/src/mesa/main/arrayobj.h
index 9c4036af5a..8999edc724 100644
--- a/src/mesa/main/arrayobj.h
+++ b/src/mesa/main/arrayobj.h
@@ -1,9 +1,10 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
  * (C) Copyright IBM Corporation 2006
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -56,22 +57,25 @@ extern void
 _mesa_initialize_array_object( GLcontext *ctx,
                                struct gl_array_object *obj, GLuint name );
 
-extern void
-_mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj );
 
 extern void
-_mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj );
-
+_mesa_update_array_object_max_element(GLcontext *ctx,
+                                      struct gl_array_object *arrayObj);
 
 
 /*
  * API functions
  */
 
+
+void GLAPIENTRY _mesa_BindVertexArray( GLuint id );
+
 void GLAPIENTRY _mesa_BindVertexArrayAPPLE( GLuint id );
 
 void GLAPIENTRY _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids);
 
+void GLAPIENTRY _mesa_GenVertexArrays(GLsizei n, GLuint *arrays);
+
 void GLAPIENTRY _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *buffer);
 
 GLboolean GLAPIENTRY _mesa_IsVertexArrayAPPLE( GLuint id );
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 4d85d54bb9..ab99ca1c64 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -174,24 +174,30 @@ struct texture_state
 
 
 /**
- * Allocate a new attribute state node.  These nodes have a
- * "kind" value and a pointer to a struct of state data.
+ * Allocate new attribute node of given type/kind.  Attach payload data.
+ * Insert it into the linked list named by 'head'.
  */
-static struct gl_attrib_node *
-new_attrib_node( GLbitfield kind )
+static void
+save_attrib_data(struct gl_attrib_node **head,
+                 GLbitfield kind, void *payload)
 {
-   struct gl_attrib_node *an = MALLOC_STRUCT(gl_attrib_node);
-   if (an) {
-      an->kind = kind;
+   struct gl_attrib_node *n = MALLOC_STRUCT(gl_attrib_node);
+   if (n) {
+      n->kind = kind;
+      n->data = payload;
+      /* insert at head */
+      n->next = *head;
+      *head = n;
+   }
+   else {
+      /* out of memory! */
    }
-   return an;
 }
 
 
 void GLAPIENTRY
 _mesa_PushAttrib(GLbitfield mask)
 {
-   struct gl_attrib_node *newnode;
    struct gl_attrib_node *head;
 
    GET_CURRENT_CONTEXT(ctx);
@@ -213,10 +219,7 @@ _mesa_PushAttrib(GLbitfield mask)
       struct gl_accum_attrib *attr;
       attr = MALLOC_STRUCT( gl_accum_attrib );
       MEMCPY( attr, &ctx->Accum, sizeof(struct gl_accum_attrib) );
-      newnode = new_attrib_node( GL_ACCUM_BUFFER_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_ACCUM_BUFFER_BIT, attr);
    }
 
    if (mask & GL_COLOR_BUFFER_BIT) {
@@ -227,10 +230,7 @@ _mesa_PushAttrib(GLbitfield mask)
       /* push the Draw FBO's DrawBuffer[] state, not ctx->Color.DrawBuffer[] */
       for (i = 0; i < ctx->Const.MaxDrawBuffers; i ++)
          attr->DrawBuffer[i] = ctx->DrawBuffer->ColorDrawBuffer[i];
-      newnode = new_attrib_node( GL_COLOR_BUFFER_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_COLOR_BUFFER_BIT, attr);
    }
 
    if (mask & GL_CURRENT_BIT) {
@@ -238,20 +238,14 @@ _mesa_PushAttrib(GLbitfield mask)
       FLUSH_CURRENT( ctx, 0 );
       attr = MALLOC_STRUCT( gl_current_attrib );
       MEMCPY( attr, &ctx->Current, sizeof(struct gl_current_attrib) );
-      newnode = new_attrib_node( GL_CURRENT_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_CURRENT_BIT, attr);
    }
 
    if (mask & GL_DEPTH_BUFFER_BIT) {
       struct gl_depthbuffer_attrib *attr;
       attr = MALLOC_STRUCT( gl_depthbuffer_attrib );
       MEMCPY( attr, &ctx->Depth, sizeof(struct gl_depthbuffer_attrib) );
-      newnode = new_attrib_node( GL_DEPTH_BUFFER_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_DEPTH_BUFFER_BIT, attr);
    }
 
    if (mask & GL_ENABLE_BIT) {
@@ -331,40 +325,28 @@ _mesa_PushAttrib(GLbitfield mask)
       attr->VertexProgram = ctx->VertexProgram.Enabled;
       attr->VertexProgramPointSize = ctx->VertexProgram.PointSizeEnabled;
       attr->VertexProgramTwoSide = ctx->VertexProgram.TwoSideEnabled;
-      newnode = new_attrib_node( GL_ENABLE_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_ENABLE_BIT, attr);
    }
 
    if (mask & GL_EVAL_BIT) {
       struct gl_eval_attrib *attr;
       attr = MALLOC_STRUCT( gl_eval_attrib );
       MEMCPY( attr, &ctx->Eval, sizeof(struct gl_eval_attrib) );
-      newnode = new_attrib_node( GL_EVAL_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_EVAL_BIT, attr);
    }
 
    if (mask & GL_FOG_BIT) {
       struct gl_fog_attrib *attr;
       attr = MALLOC_STRUCT( gl_fog_attrib );
       MEMCPY( attr, &ctx->Fog, sizeof(struct gl_fog_attrib) );
-      newnode = new_attrib_node( GL_FOG_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_FOG_BIT, attr);
    }
 
    if (mask & GL_HINT_BIT) {
       struct gl_hint_attrib *attr;
       attr = MALLOC_STRUCT( gl_hint_attrib );
       MEMCPY( attr, &ctx->Hint, sizeof(struct gl_hint_attrib) );
-      newnode = new_attrib_node( GL_HINT_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_HINT_BIT, attr);
    }
 
    if (mask & GL_LIGHTING_BIT) {
@@ -372,30 +354,21 @@ _mesa_PushAttrib(GLbitfield mask)
       FLUSH_CURRENT(ctx, 0);	/* flush material changes */
       attr = MALLOC_STRUCT( gl_light_attrib );
       MEMCPY( attr, &ctx->Light, sizeof(struct gl_light_attrib) );
-      newnode = new_attrib_node( GL_LIGHTING_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_LIGHTING_BIT, attr);
    }
 
    if (mask & GL_LINE_BIT) {
       struct gl_line_attrib *attr;
       attr = MALLOC_STRUCT( gl_line_attrib );
       MEMCPY( attr, &ctx->Line, sizeof(struct gl_line_attrib) );
-      newnode = new_attrib_node( GL_LINE_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_LINE_BIT, attr);
    }
 
    if (mask & GL_LIST_BIT) {
       struct gl_list_attrib *attr;
       attr = MALLOC_STRUCT( gl_list_attrib );
       MEMCPY( attr, &ctx->List, sizeof(struct gl_list_attrib) );
-      newnode = new_attrib_node( GL_LIST_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_LIST_BIT, attr);
    }
 
    if (mask & GL_PIXEL_MODE_BIT) {
@@ -404,60 +377,42 @@ _mesa_PushAttrib(GLbitfield mask)
       MEMCPY( attr, &ctx->Pixel, sizeof(struct gl_pixel_attrib) );
       /* push the Read FBO's ReadBuffer state, not ctx->Pixel.ReadBuffer */
       attr->ReadBuffer = ctx->ReadBuffer->ColorReadBuffer;
-      newnode = new_attrib_node( GL_PIXEL_MODE_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_PIXEL_MODE_BIT, attr);
    }
 
    if (mask & GL_POINT_BIT) {
       struct gl_point_attrib *attr;
       attr = MALLOC_STRUCT( gl_point_attrib );
       MEMCPY( attr, &ctx->Point, sizeof(struct gl_point_attrib) );
-      newnode = new_attrib_node( GL_POINT_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_POINT_BIT, attr);
    }
 
    if (mask & GL_POLYGON_BIT) {
       struct gl_polygon_attrib *attr;
       attr = MALLOC_STRUCT( gl_polygon_attrib );
       MEMCPY( attr, &ctx->Polygon, sizeof(struct gl_polygon_attrib) );
-      newnode = new_attrib_node( GL_POLYGON_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_POLYGON_BIT, attr);
    }
 
    if (mask & GL_POLYGON_STIPPLE_BIT) {
       GLuint *stipple;
       stipple = (GLuint *) MALLOC( 32*sizeof(GLuint) );
       MEMCPY( stipple, ctx->PolygonStipple, 32*sizeof(GLuint) );
-      newnode = new_attrib_node( GL_POLYGON_STIPPLE_BIT );
-      newnode->data = stipple;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_POLYGON_STIPPLE_BIT, stipple);
    }
 
    if (mask & GL_SCISSOR_BIT) {
       struct gl_scissor_attrib *attr;
       attr = MALLOC_STRUCT( gl_scissor_attrib );
       MEMCPY( attr, &ctx->Scissor, sizeof(struct gl_scissor_attrib) );
-      newnode = new_attrib_node( GL_SCISSOR_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_SCISSOR_BIT, attr);
    }
 
    if (mask & GL_STENCIL_BUFFER_BIT) {
       struct gl_stencil_attrib *attr;
       attr = MALLOC_STRUCT( gl_stencil_attrib );
       MEMCPY( attr, &ctx->Stencil, sizeof(struct gl_stencil_attrib) );
-      newnode = new_attrib_node( GL_STENCIL_BUFFER_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_STENCIL_BUFFER_BIT, attr);
    }
 
    if (mask & GL_TEXTURE_BIT) {
@@ -494,30 +449,21 @@ _mesa_PushAttrib(GLbitfield mask)
 
       _mesa_unlock_context_textures(ctx);
 
-      newnode = new_attrib_node( GL_TEXTURE_BIT );
-      newnode->data = texstate;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_TEXTURE_BIT, texstate);
    }
 
    if (mask & GL_TRANSFORM_BIT) {
       struct gl_transform_attrib *attr;
       attr = MALLOC_STRUCT( gl_transform_attrib );
       MEMCPY( attr, &ctx->Transform, sizeof(struct gl_transform_attrib) );
-      newnode = new_attrib_node( GL_TRANSFORM_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_TRANSFORM_BIT, attr);
    }
 
    if (mask & GL_VIEWPORT_BIT) {
       struct gl_viewport_attrib *attr;
       attr = MALLOC_STRUCT( gl_viewport_attrib );
       MEMCPY( attr, &ctx->Viewport, sizeof(struct gl_viewport_attrib) );
-      newnode = new_attrib_node( GL_VIEWPORT_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_VIEWPORT_BIT, attr);
    }
 
    /* GL_ARB_multisample */
@@ -525,10 +471,7 @@ _mesa_PushAttrib(GLbitfield mask)
       struct gl_multisample_attrib *attr;
       attr = MALLOC_STRUCT( gl_multisample_attrib );
       MEMCPY( attr, &ctx->Multisample, sizeof(struct gl_multisample_attrib) );
-      newnode = new_attrib_node( GL_MULTISAMPLE_BIT_ARB );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_MULTISAMPLE_BIT_ARB, attr);
    }
 
 end:
@@ -1324,20 +1267,22 @@ _mesa_PopAttrib(void)
  * counts when pushing/popping the GL_CLIENT_VERTEX_ARRAY_BIT attribute group.
  */
 static void
-adjust_buffer_object_ref_counts(struct gl_array_attrib *array, GLint step)
+adjust_buffer_object_ref_counts(struct gl_array_object *arrayObj, GLint step)
 {
    GLuint i;
-   array->ArrayObj->Vertex.BufferObj->RefCount += step;
-   array->ArrayObj->Normal.BufferObj->RefCount += step;
-   array->ArrayObj->Color.BufferObj->RefCount += step;
-   array->ArrayObj->SecondaryColor.BufferObj->RefCount += step;
-   array->ArrayObj->FogCoord.BufferObj->RefCount += step;
-   array->ArrayObj->Index.BufferObj->RefCount += step;
-   array->ArrayObj->EdgeFlag.BufferObj->RefCount += step;
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
-      array->ArrayObj->TexCoord[i].BufferObj->RefCount += step;
-   for (i = 0; i < VERT_ATTRIB_MAX; i++)
-      array->ArrayObj->VertexAttrib[i].BufferObj->RefCount += step;
+
+   arrayObj->Vertex.BufferObj->RefCount += step;
+   arrayObj->Weight.BufferObj->RefCount += step;
+   arrayObj->Normal.BufferObj->RefCount += step;
+   arrayObj->Color.BufferObj->RefCount += step;
+   arrayObj->SecondaryColor.BufferObj->RefCount += step;
+   arrayObj->FogCoord.BufferObj->RefCount += step;
+   arrayObj->Index.BufferObj->RefCount += step;
+   arrayObj->EdgeFlag.BufferObj->RefCount += step;
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
+      arrayObj->TexCoord[i].BufferObj->RefCount += step;
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      arrayObj->VertexAttrib[i].BufferObj->RefCount += step;
 }
 
 
@@ -1371,7 +1316,6 @@ copy_pixelstore(GLcontext *ctx,
 void GLAPIENTRY
 _mesa_PushClientAttrib(GLbitfield mask)
 {
-   struct gl_attrib_node *newnode;
    struct gl_attrib_node *head;
 
    GET_CURRENT_CONTEXT(ctx);
@@ -1392,17 +1336,11 @@ _mesa_PushClientAttrib(GLbitfield mask)
       /* packing attribs */
       attr = CALLOC_STRUCT( gl_pixelstore_attrib );
       copy_pixelstore(ctx, attr, &ctx->Pack);
-      newnode = new_attrib_node( GL_CLIENT_PACK_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_CLIENT_PACK_BIT, attr);
       /* unpacking attribs */
       attr = CALLOC_STRUCT( gl_pixelstore_attrib );
       copy_pixelstore(ctx, attr, &ctx->Unpack);
-      newnode = new_attrib_node( GL_CLIENT_UNPACK_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_CLIENT_UNPACK_BIT, attr);
    }
 
    if (mask & GL_CLIENT_VERTEX_ARRAY_BIT) {
@@ -1423,12 +1361,10 @@ _mesa_PushClientAttrib(GLbitfield mask)
 
       attr->ArrayObj = obj;
 
-      newnode = new_attrib_node( GL_CLIENT_VERTEX_ARRAY_BIT );
-      newnode->data = attr;
-      newnode->next = head;
-      head = newnode;
+      save_attrib_data(&head, GL_CLIENT_VERTEX_ARRAY_BIT, attr);
+
       /* bump reference counts on buffer objects */
-      adjust_buffer_object_ref_counts(&ctx->Array, 1);
+      adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, 1);
    }
 
    ctx->ClientAttribStack[ctx->ClientAttribStackDepth] = head;
@@ -1478,7 +1414,7 @@ _mesa_PopClientAttrib(void)
 	    struct gl_array_attrib * data =
 	      (struct gl_array_attrib *) node->data;
 
-            adjust_buffer_object_ref_counts(&ctx->Array, -1);
+            adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, -1);
 	 
             ctx->Array.ActiveTexture = data->ActiveTexture;
 	    if (data->LockCount != 0)
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index c8d160baa9..b95e00af5b 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -1,9 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
- * Copyright (C) 1999-2009  VMware, Inc.  All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -39,10 +39,15 @@
 #include "bufferobj.h"
 
 
+/* Debug flags */
+/*#define VBO_DEBUG*/
+/*#define BOUNDS_CHECK*/
+
+
 #ifdef FEATURE_OES_mapbuffer
-#define DEFAULT_ACCESS GL_WRITE_ONLY;
+#define DEFAULT_ACCESS GL_MAP_WRITE_BIT
 #else
-#define DEFAULT_ACCESS GL_READ_WRITE;
+#define DEFAULT_ACCESS (GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)
 #endif
 
 
@@ -73,6 +78,16 @@ get_buffer(GLcontext *ctx, GLenum target)
       case GL_PIXEL_UNPACK_BUFFER_EXT:
          bufObj = ctx->Unpack.BufferObj;
          break;
+      case GL_COPY_READ_BUFFER:
+         if (ctx->Extensions.ARB_copy_buffer) {
+            bufObj = ctx->CopyReadBuffer;
+         }
+         break;
+      case GL_COPY_WRITE_BUFFER:
+         if (ctx->Extensions.ARB_copy_buffer) {
+            bufObj = ctx->CopyWriteBuffer;
+         }
+         break;
       default:
          /* error must be recorded by caller */
          return NULL;
@@ -86,6 +101,24 @@ get_buffer(GLcontext *ctx, GLenum target)
 
 
 /**
+ * Convert a GLbitfield describing the mapped buffer access flags
+ * into one of GL_READ_WRITE, GL_READ_ONLY, or GL_WRITE_ONLY.
+ */
+static GLenum
+simplified_access_mode(GLbitfield access)
+{
+   const GLbitfield rwFlags = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT;
+   if ((access & rwFlags) == rwFlags)
+      return GL_READ_WRITE;
+   if ((access & GL_MAP_READ_BIT) == GL_MAP_READ_BIT)
+      return GL_READ_ONLY;
+   if ((access & GL_MAP_WRITE_BIT) == GL_MAP_WRITE_BIT)
+      return GL_WRITE_ONLY;
+   return GL_READ_WRITE; /* this should never happen, but no big deal */
+}
+
+
+/**
  * Tests the subdata range parameters and sets the GL error code for
  * \c glBufferSubDataARB and \c glGetBufferSubDataARB.
  *
@@ -123,7 +156,7 @@ buffer_object_subdata_range_good( GLcontext * ctx, GLenum target,
       _mesa_error(ctx, GL_INVALID_ENUM, "%s(target)", caller);
       return NULL;
    }
-   if (bufObj->Name == 0) {
+   if (!_mesa_is_bufferobj(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "%s", caller);
       return NULL;
    }
@@ -132,7 +165,7 @@ buffer_object_subdata_range_good( GLcontext * ctx, GLenum target,
 		  "%s(size + offset > buffer size)", caller);
       return NULL;
    }
-   if (bufObj->Pointer) {
+   if (_mesa_bufferobj_mapped(bufObj)) {
       /* Buffer is currently mapped */
       _mesa_error(ctx, GL_INVALID_OPERATION, "%s", caller);
       return NULL;
@@ -147,7 +180,7 @@ buffer_object_subdata_range_good( GLcontext * ctx, GLenum target,
  * 
  * Default callback for the \c dd_function_table::NewBufferObject() hook.
  */
-struct gl_buffer_object *
+static struct gl_buffer_object *
 _mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target )
 {
    struct gl_buffer_object *obj;
@@ -165,7 +198,7 @@ _mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target )
  * 
  * Default callback for the \c dd_function_table::DeleteBuffer() hook.
  */
-void
+static void
 _mesa_delete_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
 {
    (void) ctx;
@@ -194,7 +227,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
       return;
 
    if (*ptr) {
-      /* Unreference the old texture */
+      /* Unreference the old buffer */
       GLboolean deleteFlag = GL_FALSE;
       struct gl_buffer_object *oldObj = *ptr;
 
@@ -227,7 +260,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
    ASSERT(!*ptr);
 
    if (bufObj) {
-      /* reference new texture */
+      /* reference new buffer */
       /*_glthread_LOCK_MUTEX(tex->Mutex);*/
       if (bufObj->RefCount == 0) {
          /* this buffer's being deleted (look just above) */
@@ -261,7 +294,7 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
    obj->RefCount = 1;
    obj->Name = name;
    obj->Usage = GL_STATIC_DRAW_ARB;
-   obj->Access = DEFAULT_ACCESS;
+   obj->AccessFlags = DEFAULT_ACCESS;
 }
 
 
@@ -281,9 +314,10 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
  * \param usage   Hints about how the data will be used.
  * \param bufObj  Object to be used.
  *
+ * \return GL_TRUE for success, GL_FALSE for failure
  * \sa glBufferDataARB, dd_function_table::BufferData.
  */
-void
+static GLboolean
 _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
 		   const GLvoid * data, GLenum usage,
 		   struct gl_buffer_object * bufObj )
@@ -301,6 +335,11 @@ _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
       if (data) {
 	 _mesa_memcpy( bufObj->Data, data, size );
       }
+
+      return GL_TRUE;
+   }
+   else {
+      return GL_FALSE;
    }
 }
 
@@ -322,7 +361,7 @@ _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
  *
  * \sa glBufferSubDataARB, dd_function_table::BufferSubData.
  */
-void
+static void
 _mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
 		      GLsizeiptrARB size, const GLvoid * data,
 		      struct gl_buffer_object * bufObj )
@@ -355,7 +394,7 @@ _mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
  *
  * \sa glBufferGetSubDataARB, dd_function_table::GetBufferSubData.
  */
-void
+static void
 _mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
 			  GLsizeiptrARB size, GLvoid * data,
 			  struct gl_buffer_object * bufObj )
@@ -382,60 +421,132 @@ _mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
  *
  * \sa glMapBufferARB, dd_function_table::MapBuffer
  */
-void *
+static void *
 _mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
 		  struct gl_buffer_object *bufObj )
 {
    (void) ctx;
    (void) target;
    (void) access;
-   ASSERT(!bufObj->OnCard);
    /* Just return a direct pointer to the data */
-   if (bufObj->Pointer) {
+   if (_mesa_bufferobj_mapped(bufObj)) {
       /* already mapped! */
       return NULL;
    }
    bufObj->Pointer = bufObj->Data;
+   bufObj->Length = bufObj->Size;
+   bufObj->Offset = 0;
+   return bufObj->Pointer;
+}
+
+
+/**
+ * Default fallback for \c dd_function_table::MapBufferRange().
+ * Called via glMapBufferRange().
+ */
+static void *
+_mesa_buffer_map_range( GLcontext *ctx, GLenum target, GLintptr offset,
+                        GLsizeiptr length, GLbitfield access,
+                        struct gl_buffer_object *bufObj )
+{
+   (void) ctx;
+   (void) target;
+   assert(!_mesa_bufferobj_mapped(bufObj));
+   /* Just return a direct pointer to the data */
+   bufObj->Pointer = bufObj->Data + offset;
+   bufObj->Length = length;
+   bufObj->Offset = offset;
+   bufObj->AccessFlags = access;
    return bufObj->Pointer;
 }
 
 
 /**
+ * Default fallback for \c dd_function_table::FlushMappedBufferRange().
+ * Called via glFlushMappedBufferRange().
+ */
+static void
+_mesa_buffer_flush_mapped_range( GLcontext *ctx, GLenum target, 
+                                 GLintptr offset, GLsizeiptr length,
+                                 struct gl_buffer_object *obj )
+{
+   (void) ctx;
+   (void) target;
+   (void) offset;
+   (void) length;
+   (void) obj;
+   /* no-op */
+}
+
+
+/**
  * Default callback for \c dd_function_table::MapBuffer().
  *
  * The input parameters will have been already tested for errors.
  *
  * \sa glUnmapBufferARB, dd_function_table::UnmapBuffer
  */
-GLboolean
+static GLboolean
 _mesa_buffer_unmap( GLcontext *ctx, GLenum target,
                     struct gl_buffer_object *bufObj )
 {
    (void) ctx;
    (void) target;
-   ASSERT(!bufObj->OnCard);
    /* XXX we might assert here that bufObj->Pointer is non-null */
    bufObj->Pointer = NULL;
+   bufObj->Length = 0;
+   bufObj->Offset = 0;
+   bufObj->AccessFlags = 0x0;
    return GL_TRUE;
 }
 
 
 /**
+ * Default fallback for \c dd_function_table::CopyBufferSubData().
+ * Called via glCopyBuffserSubData().
+ */
+static void
+_mesa_copy_buffer_subdata(GLcontext *ctx,
+                          struct gl_buffer_object *src,
+                          struct gl_buffer_object *dst,
+                          GLintptr readOffset, GLintptr writeOffset,
+                          GLsizeiptr size)
+{
+   GLubyte *srcPtr, *dstPtr;
+
+   /* buffer should not already be mapped */
+   assert(!_mesa_bufferobj_mapped(src));
+   assert(!_mesa_bufferobj_mapped(dst));
+
+   srcPtr = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_COPY_READ_BUFFER,
+                                              GL_READ_ONLY, src);
+   dstPtr = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_COPY_WRITE_BUFFER,
+                                              GL_WRITE_ONLY, dst);
+
+   if (srcPtr && dstPtr)
+      _mesa_memcpy(dstPtr + writeOffset, srcPtr + readOffset, size);
+
+   ctx->Driver.UnmapBuffer(ctx, GL_COPY_READ_BUFFER, src);
+   ctx->Driver.UnmapBuffer(ctx, GL_COPY_WRITE_BUFFER, dst);
+}
+
+
+
+/**
  * Initialize the state associated with buffer objects
  */
 void
 _mesa_init_buffer_objects( GLcontext *ctx )
 {
-   /* Allocate the default buffer object and set refcount so high that
-    * it never gets deleted.
-    * XXX with recent/improved refcounting this may not longer be needed.
-    */
-   ctx->Array.NullBufferObj = _mesa_new_buffer_object(ctx, 0, 0);
-   if (ctx->Array.NullBufferObj)
-      ctx->Array.NullBufferObj->RefCount = 1000;
-
-   ctx->Array.ArrayBufferObj = ctx->Array.NullBufferObj;
-   ctx->Array.ElementArrayBufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj,
+                                 ctx->Shared->NullBufferObj);
+   _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj,
+                                 ctx->Shared->NullBufferObj);
+
+   _mesa_reference_buffer_object(ctx, &ctx->CopyReadBuffer,
+                                 ctx->Shared->NullBufferObj);
+   _mesa_reference_buffer_object(ctx, &ctx->CopyWriteBuffer,
+                                 ctx->Shared->NullBufferObj);
 }
 
 
@@ -462,8 +573,22 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
    case GL_PIXEL_UNPACK_BUFFER_EXT:
       bindTarget = &ctx->Unpack.BufferObj;
       break;
+   case GL_COPY_READ_BUFFER:
+      if (ctx->Extensions.ARB_copy_buffer) {
+         bindTarget = &ctx->CopyReadBuffer;
+      }
+      break;
+   case GL_COPY_WRITE_BUFFER:
+      if (ctx->Extensions.ARB_copy_buffer) {
+         bindTarget = &ctx->CopyWriteBuffer;
+      }
+      break;
    default:
-      _mesa_error(ctx, GL_INVALID_ENUM, "glBindBufferARB(target)");
+      ; /* no-op / we'll hit the follow error test next */
+   }
+
+   if (!bindTarget) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glBindBufferARB(target 0x%x)");
       return;
    }
 
@@ -479,7 +604,7 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
       /* The spec says there's not a buffer object named 0, but we use
        * one internally because it simplifies things.
        */
-      newBufObj = ctx->Array.NullBufferObj;
+      newBufObj = ctx->Shared->NullBufferObj;
    }
    else {
       /* non-default buffer object */
@@ -500,7 +625,7 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
    _mesa_reference_buffer_object(ctx, bindTarget, newBufObj);
 
    /* Pass BindBuffer call to device driver */
-   if (ctx->Driver.BindBuffer && newBufObj)
+   if (ctx->Driver.BindBuffer)
       ctx->Driver.BindBuffer( ctx, target, newBufObj );
 }
 
@@ -533,6 +658,8 @@ _mesa_update_default_objects_buffer_objects(GLcontext *ctx)
  * currently mapped.  Whoever calls this function should check for that.
  * Remember, we can't use a PBO when it's mapped!
  *
+ * If we're not using a PBO, this is a no-op.
+ *
  * \param width  width of image to read/write
  * \param height  height of image to read/write
  * \param depth  depth of image to read/write
@@ -551,7 +678,8 @@ _mesa_validate_pbo_access(GLuint dimensions,
    GLvoid *start, *end;
    const GLubyte *sizeAddr; /* buffer size, cast to a pointer */
 
-   ASSERT(pack->BufferObj->Name != 0);
+   if (!_mesa_is_bufferobj(pack->BufferObj))
+      return GL_TRUE;  /* no PBO, OK */
 
    if (pack->BufferObj->Size == 0)
       /* no buffer! */
@@ -583,21 +711,22 @@ _mesa_validate_pbo_access(GLuint dimensions,
 
 
 /**
- * If the source of glBitmap data is a PBO, check that we won't read out
- * of buffer bounds, then map the buffer.
- * If not sourcing from a PBO, just return the bitmap pointer.
- * This is a helper function for (some) drivers.
- * Return NULL if error.
- * If non-null return, must call _mesa_unmap_bitmap_pbo() when done.
+ * For commands that read from a PBO (glDrawPixels, glTexImage,
+ * glPolygonStipple, etc), if we're reading from a PBO, map it read-only
+ * and return the pointer into the PBO.  If we're not reading from a
+ * PBO, return \p src as-is.
+ * If non-null return, must call _mesa_unmap_pbo_source() when done.
+ *
+ * \return NULL if error, else pointer to start of data
  */
-const GLubyte *
-_mesa_map_bitmap_pbo(GLcontext *ctx,
+const GLvoid *
+_mesa_map_pbo_source(GLcontext *ctx,
                      const struct gl_pixelstore_attrib *unpack,
-                     const GLubyte *bitmap)
+                     const GLvoid *src)
 {
    const GLubyte *buf;
 
-   if (unpack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       /* unpack from PBO */
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
@@ -605,11 +734,11 @@ _mesa_map_bitmap_pbo(GLcontext *ctx,
       if (!buf)
          return NULL;
 
-      buf = ADD_POINTERS(buf, bitmap);
+      buf = ADD_POINTERS(buf, src);
    }
    else {
       /* unpack from normal memory */
-      buf = bitmap;
+      buf = src;
    }
 
    return buf;
@@ -617,57 +746,55 @@ _mesa_map_bitmap_pbo(GLcontext *ctx,
 
 
 /**
- * Counterpart to _mesa_map_bitmap_pbo()
- * This is a helper function for (some) drivers.
- */
-void
-_mesa_unmap_bitmap_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack)
-{
-   if (unpack->BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              unpack->BufferObj);
-   }
-}
-
-
-/**
- * \sa _mesa_map_bitmap_pbo
+ * Combine PBO-read validation and mapping.
+ * If any GL errors are detected, they'll be recorded and NULL returned.
+ * \sa _mesa_validate_pbo_access
+ * \sa _mesa_map_pbo_source
+ * A call to this function should have a matching call to
+ * _mesa_unmap_pbo_source().
  */
 const GLvoid *
-_mesa_map_drawpix_pbo(GLcontext *ctx,
-                      const struct gl_pixelstore_attrib *unpack,
-                      const GLvoid *pixels)
+_mesa_map_validate_pbo_source(GLcontext *ctx,
+                              GLuint dimensions,
+                              const struct gl_pixelstore_attrib *unpack,
+                              GLsizei width, GLsizei height, GLsizei depth,
+                              GLenum format, GLenum type, const GLvoid *ptr,
+                              const char *where)
 {
-   const GLvoid *buf;
+   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
 
-   if (unpack->BufferObj->Name) {
-      /* unpack from PBO */
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              unpack->BufferObj);
-      if (!buf)
-         return NULL;
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* non-PBO access: no validation to be done */
+      return ptr;
+   }
 
-      buf = ADD_POINTERS(buf, pixels);
+   if (!_mesa_validate_pbo_access(dimensions, unpack,
+                                  width, height, depth, format, type, ptr)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "%s(out of bounds PBO access)", where);
+      return NULL;
    }
-   else {
-      /* unpack from normal memory */
-      buf = pixels;
+
+   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
+      /* buffer is already mapped - that's an error */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
+      return NULL;
    }
 
-   return buf;
+   ptr = _mesa_map_pbo_source(ctx, unpack, ptr);
+   return ptr;
 }
 
 
 /**
- * \sa _mesa_unmap_bitmap_pbo
+ * Counterpart to _mesa_map_pbo_source()
  */
 void
-_mesa_unmap_drawpix_pbo(GLcontext *ctx,
-                        const struct gl_pixelstore_attrib *unpack)
+_mesa_unmap_pbo_source(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack)
 {
-   if (unpack->BufferObj->Name) {
+   ASSERT(unpack != &ctx->Pack); /* catch pack/unpack mismatch */
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                               unpack->BufferObj);
    }
@@ -675,18 +802,21 @@ _mesa_unmap_drawpix_pbo(GLcontext *ctx,
 
 
 /**
- * If PBO is bound, map the buffer, return dest pointer in mapped buffer.
- * Call _mesa_unmap_readpix_pbo() when finished
- * \return NULL if error
+ * For commands that write to a PBO (glReadPixels, glGetColorTable, etc),
+ * if we're writing to a PBO, map it write-only and return the pointer
+ * into the PBO.  If we're not writing to a PBO, return \p dst as-is.
+ * If non-null return, must call _mesa_unmap_pbo_dest() when done.
+ *
+ * \return NULL if error, else pointer to start of data
  */
 void *
-_mesa_map_readpix_pbo(GLcontext *ctx,
-                      const struct gl_pixelstore_attrib *pack,
-                      GLvoid *dest)
+_mesa_map_pbo_dest(GLcontext *ctx,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest)
 {
    void *buf;
 
-   if (pack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(pack->BufferObj)) {
       /* pack into PBO */
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
@@ -706,13 +836,55 @@ _mesa_map_readpix_pbo(GLcontext *ctx,
 
 
 /**
- * Counterpart to _mesa_map_readpix_pbo()
+ * Combine PBO-write validation and mapping.
+ * If any GL errors are detected, they'll be recorded and NULL returned.
+ * \sa _mesa_validate_pbo_access
+ * \sa _mesa_map_pbo_dest
+ * A call to this function should have a matching call to
+ * _mesa_unmap_pbo_dest().
+ */
+GLvoid *
+_mesa_map_validate_pbo_dest(GLcontext *ctx,
+                            GLuint dimensions,
+                            const struct gl_pixelstore_attrib *unpack,
+                            GLsizei width, GLsizei height, GLsizei depth,
+                            GLenum format, GLenum type, GLvoid *ptr,
+                            const char *where)
+{
+   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
+
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* non-PBO access: no validation to be done */
+      return ptr;
+   }
+
+   if (!_mesa_validate_pbo_access(dimensions, unpack,
+                                  width, height, depth, format, type, ptr)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "%s(out of bounds PBO access)", where);
+      return NULL;
+   }
+
+   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
+      /* buffer is already mapped - that's an error */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
+      return NULL;
+   }
+
+   ptr = _mesa_map_pbo_dest(ctx, unpack, ptr);
+   return ptr;
+}
+
+
+/**
+ * Counterpart to _mesa_map_pbo_dest()
  */
 void
-_mesa_unmap_readpix_pbo(GLcontext *ctx,
-                        const struct gl_pixelstore_attrib *pack)
+_mesa_unmap_pbo_dest(GLcontext *ctx,
+                     const struct gl_pixelstore_attrib *pack)
 {
-   if (pack->BufferObj->Name) {
+   ASSERT(pack != &ctx->Unpack); /* catch pack/unpack mismatch */
+   if (_mesa_is_bufferobj(pack->BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, pack->BufferObj);
    }
 }
@@ -746,11 +918,37 @@ unbind(GLcontext *ctx,
        struct gl_buffer_object *obj)
 {
    if (*ptr == obj) {
-      _mesa_reference_buffer_object(ctx, ptr, ctx->Array.NullBufferObj);
+      _mesa_reference_buffer_object(ctx, ptr, ctx->Shared->NullBufferObj);
    }
 }
 
 
+/**
+ * Plug default/fallback buffer object functions into the device
+ * driver hooks.
+ */
+void
+_mesa_init_buffer_object_functions(struct dd_function_table *driver)
+{
+   /* GL_ARB_vertex/pixel_buffer_object */
+   driver->NewBufferObject = _mesa_new_buffer_object;
+   driver->DeleteBuffer = _mesa_delete_buffer_object;
+   driver->BindBuffer = NULL;
+   driver->BufferData = _mesa_buffer_data;
+   driver->BufferSubData = _mesa_buffer_subdata;
+   driver->GetBufferSubData = _mesa_buffer_get_subdata;
+   driver->MapBuffer = _mesa_buffer_map;
+   driver->UnmapBuffer = _mesa_buffer_unmap;
+
+   /* GL_ARB_map_buffer_range */
+   driver->MapBufferRange = _mesa_buffer_map_range;
+   driver->FlushMappedBufferRange = _mesa_buffer_flush_mapped_range;
+
+   /* GL_ARB_copy_buffer */
+   driver->CopyBufferSubData = _mesa_copy_buffer_subdata;
+}
+
+
 
 /**********************************************************************/
 /* API Functions                                                      */
@@ -789,30 +987,32 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
    for (i = 0; i < n; i++) {
       struct gl_buffer_object *bufObj = _mesa_lookup_bufferobj(ctx, ids[i]);
       if (bufObj) {
+         struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
          GLuint j;
 
          ASSERT(bufObj->Name == ids[i]);
 
-         if (bufObj->Pointer) {
+         if (_mesa_bufferobj_mapped(bufObj)) {
             /* if mapped, unmap it now */
             ctx->Driver.UnmapBuffer(ctx, 0, bufObj);
-            bufObj->Access = DEFAULT_ACCESS;
+            bufObj->AccessFlags = DEFAULT_ACCESS;
             bufObj->Pointer = NULL;
          }
 
          /* unbind any vertex pointers bound to this buffer */
-         unbind(ctx, &ctx->Array.ArrayObj->Vertex.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Normal.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Color.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->SecondaryColor.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->FogCoord.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Index.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->EdgeFlag.BufferObj, bufObj);
-         for (j = 0; j < MAX_TEXTURE_COORD_UNITS; j++) {
-            unbind(ctx, &ctx->Array.ArrayObj->TexCoord[j].BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Vertex.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Weight.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Normal.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Color.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->SecondaryColor.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->FogCoord.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Index.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->EdgeFlag.BufferObj, bufObj);
+         for (j = 0; j < Elements(arrayObj->TexCoord); j++) {
+            unbind(ctx, &arrayObj->TexCoord[j].BufferObj, bufObj);
          }
-         for (j = 0; j < VERT_ATTRIB_MAX; j++) {
-            unbind(ctx, &ctx->Array.ArrayObj->VertexAttrib[j].BufferObj, bufObj);
+         for (j = 0; j < Elements(arrayObj->VertexAttrib); j++) {
+            unbind(ctx, &arrayObj->VertexAttrib[j].BufferObj, bufObj);
          }
 
          if (ctx->Array.ArrayBufferObj == bufObj) {
@@ -946,22 +1146,35 @@ _mesa_BufferDataARB(GLenum target, GLsizeiptrARB size,
       _mesa_error(ctx, GL_INVALID_ENUM, "glBufferDataARB(target)" );
       return;
    }
-   if (bufObj->Name == 0) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glBufferDataARB" );
+   if (!_mesa_is_bufferobj(bufObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glBufferDataARB(buffer 0)" );
       return;
    }
    
-   if (bufObj->Pointer) {
+   if (_mesa_bufferobj_mapped(bufObj)) {
       /* Unmap the existing buffer.  We'll replace it now.  Not an error. */
       ctx->Driver.UnmapBuffer(ctx, target, bufObj);
-      bufObj->Access = DEFAULT_ACCESS;
-      bufObj->Pointer = NULL;
+      bufObj->AccessFlags = DEFAULT_ACCESS;
+      ASSERT(bufObj->Pointer == NULL);
    }  
 
-   ASSERT(ctx->Driver.BufferData);
+   FLUSH_VERTICES(ctx, _NEW_BUFFER_OBJECT);
 
-   /* Give the buffer object to the driver!  <data> may be null! */
-   ctx->Driver.BufferData( ctx, target, size, data, usage, bufObj );
+   bufObj->Written = GL_TRUE;
+
+#ifdef VBO_DEBUG
+   _mesa_printf("glBufferDataARB(%u, sz %ld, from %p, usage 0x%x)\n",
+                bufObj->Name, size, data, usage);
+#endif
+
+#ifdef BOUNDS_CHECK
+   size += 100;
+#endif
+
+   ASSERT(ctx->Driver.BufferData);
+   if (!ctx->Driver.BufferData( ctx, target, size, data, usage, bufObj )) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBufferDataARB(access)");
+   }
 }
 
 
@@ -980,6 +1193,8 @@ _mesa_BufferSubDataARB(GLenum target, GLintptrARB offset,
       return;
    }
 
+   bufObj->Written = GL_TRUE;
+
    ASSERT(ctx->Driver.BufferSubData);
    ctx->Driver.BufferSubData( ctx, target, offset, size, data, bufObj );
 }
@@ -1010,13 +1225,20 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
 {
    GET_CURRENT_CONTEXT(ctx);
    struct gl_buffer_object * bufObj;
+   GLbitfield accessFlags;
+   void *map;
+
    ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, NULL);
 
    switch (access) {
       case GL_READ_ONLY_ARB:
+         accessFlags = GL_MAP_READ_BIT;
+         break;
       case GL_WRITE_ONLY_ARB:
+         accessFlags = GL_MAP_WRITE_BIT;
+         break;
       case GL_READ_WRITE_ARB:
-         /* OK */
+         accessFlags = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glMapBufferARB(access)");
@@ -1028,22 +1250,56 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
       _mesa_error(ctx, GL_INVALID_ENUM, "glMapBufferARB(target)" );
       return NULL;
    }
-   if (bufObj->Name == 0) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glMapBufferARB" );
+   if (!_mesa_is_bufferobj(bufObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glMapBufferARB(buffer 0)" );
       return NULL;
    }
-   if (bufObj->Pointer) {
+   if (_mesa_bufferobj_mapped(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glMapBufferARB(already mapped)");
       return NULL;
    }
 
    ASSERT(ctx->Driver.MapBuffer);
-   bufObj->Pointer = ctx->Driver.MapBuffer( ctx, target, access, bufObj );
-   if (!bufObj->Pointer) {
+   map = ctx->Driver.MapBuffer( ctx, target, access, bufObj );
+   if (!map) {
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMapBufferARB(access)");
+      return NULL;
+   }
+   else {
+      /* The driver callback should have set these fields.
+       * This is important because other modules (like VBO) might call
+       * the driver function directly.
+       */
+      ASSERT(bufObj->Pointer == map);
+      ASSERT(bufObj->Length == bufObj->Size);
+      ASSERT(bufObj->Offset == 0);
+      bufObj->AccessFlags = accessFlags;
+   }
+
+   if (access == GL_WRITE_ONLY_ARB || access == GL_READ_WRITE_ARB)
+      bufObj->Written = GL_TRUE;
+
+#ifdef VBO_DEBUG
+   _mesa_printf("glMapBufferARB(%u, sz %ld, access 0x%x)\n",
+                bufObj->Name, bufObj->Size, access);
+   if (access == GL_WRITE_ONLY_ARB) {
+      GLuint i;
+      GLubyte *b = (GLubyte *) bufObj->Pointer;
+      for (i = 0; i < bufObj->Size; i++)
+         b[i] = i & 0xff;
    }
+#endif
 
-   bufObj->Access = access;
+#ifdef BOUNDS_CHECK
+   {
+      GLubyte *buf = (GLubyte *) bufObj->Pointer;
+      GLuint i;
+      /* buffer is 100 bytes larger than requested, fill with magic value */
+      for (i = 0; i < 100; i++) {
+         buf[bufObj->Size - i - 1] = 123;
+      }
+   }
+#endif
 
    return bufObj->Pointer;
 }
@@ -1062,18 +1318,56 @@ _mesa_UnmapBufferARB(GLenum target)
       _mesa_error(ctx, GL_INVALID_ENUM, "glUnmapBufferARB(target)" );
       return GL_FALSE;
    }
-   if (bufObj->Name == 0) {
+   if (!_mesa_is_bufferobj(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glUnmapBufferARB" );
       return GL_FALSE;
    }
-   if (!bufObj->Pointer) {
+   if (!_mesa_bufferobj_mapped(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glUnmapBufferARB");
       return GL_FALSE;
    }
 
+#ifdef BOUNDS_CHECK
+   if (bufObj->Access != GL_READ_ONLY_ARB) {
+      GLubyte *buf = (GLubyte *) bufObj->Pointer;
+      GLuint i;
+      /* check that last 100 bytes are still = magic value */
+      for (i = 0; i < 100; i++) {
+         GLuint pos = bufObj->Size - i - 1;
+         if (buf[pos] != 123) {
+            _mesa_warning(ctx, "Out of bounds buffer object write detected"
+                          " at position %d (value = %u)\n",
+                          pos, buf[pos]);
+         }
+      }
+   }
+#endif
+
+#ifdef VBO_DEBUG
+   if (bufObj->AccessFlags & GL_MAP_WRITE_BIT) {
+      GLuint i, unchanged = 0;
+      GLubyte *b = (GLubyte *) bufObj->Pointer;
+      GLint pos = -1;
+      /* check which bytes changed */
+      for (i = 0; i < bufObj->Size - 1; i++) {
+         if (b[i] == (i & 0xff) && b[i+1] == ((i+1) & 0xff)) {
+            unchanged++;
+            if (pos == -1)
+               pos = i;
+         }
+      }
+      if (unchanged) {
+         _mesa_printf("glUnmapBufferARB(%u): %u of %ld unchanged, starting at %d\n",
+                      bufObj->Name, unchanged, bufObj->Size, pos);
+      }
+   }
+#endif
+
    status = ctx->Driver.UnmapBuffer( ctx, target, bufObj );
-   bufObj->Access = DEFAULT_ACCESS;
-   bufObj->Pointer = NULL;
+   bufObj->AccessFlags = DEFAULT_ACCESS;
+   ASSERT(bufObj->Pointer == NULL);
+   ASSERT(bufObj->Offset == 0);
+   ASSERT(bufObj->Length == 0);
 
    return status;
 }
@@ -1091,7 +1385,7 @@ _mesa_GetBufferParameterivARB(GLenum target, GLenum pname, GLint *params)
       _mesa_error(ctx, GL_INVALID_ENUM, "GetBufferParameterivARB(target)" );
       return;
    }
-   if (bufObj->Name == 0) {
+   if (!_mesa_is_bufferobj(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "GetBufferParameterivARB" );
       return;
    }
@@ -1104,10 +1398,10 @@ _mesa_GetBufferParameterivARB(GLenum target, GLenum pname, GLint *params)
          *params = bufObj->Usage;
          break;
       case GL_BUFFER_ACCESS_ARB:
-         *params = bufObj->Access;
+         *params = simplified_access_mode(bufObj->AccessFlags);
          break;
       case GL_BUFFER_MAPPED_ARB:
-         *params = (bufObj->Pointer != NULL);
+         *params = _mesa_bufferobj_mapped(bufObj);
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetBufferParameterivARB(pname)");
@@ -1133,10 +1427,252 @@ _mesa_GetBufferPointervARB(GLenum target, GLenum pname, GLvoid **params)
       _mesa_error(ctx, GL_INVALID_ENUM, "glGetBufferPointervARB(target)" );
       return;
    }
-   if (bufObj->Name == 0) {
+   if (!_mesa_is_bufferobj(bufObj)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glGetBufferPointervARB" );
       return;
    }
 
    *params = bufObj->Pointer;
 }
+
+
+void GLAPIENTRY
+_mesa_CopyBufferSubData(GLenum readTarget, GLenum writeTarget,
+                        GLintptr readOffset, GLintptr writeOffset,
+                        GLsizeiptr size)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *src, *dst;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   src = get_buffer(ctx, readTarget);
+   if (!src || !_mesa_is_bufferobj(src)) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glCopyBuffserSubData(readTarget = 0x%x)", readTarget);
+      return;
+   }
+
+   dst = get_buffer(ctx, writeTarget);
+   if (!dst || !_mesa_is_bufferobj(dst)) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glCopyBuffserSubData(writeTarget = 0x%x)", writeTarget);
+      return;
+   }
+
+   if (_mesa_bufferobj_mapped(src)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glCopyBuffserSubData(readBuffer is mapped)");
+      return;
+   }
+
+   if (_mesa_bufferobj_mapped(dst)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glCopyBuffserSubData(writeBuffer is mapped)");
+      return;
+   }
+
+   if (readOffset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(readOffset = %d)", readOffset);
+      return;
+   }
+
+   if (writeOffset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(writeOffset = %d)", writeOffset);
+      return;
+   }
+
+   if (readOffset + size > src->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(readOffset + size = %d)",
+                  readOffset, size);
+      return;
+   }
+
+   if (writeOffset + size > dst->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(writeOffset + size = %d)",
+                  writeOffset, size);
+      return;
+   }
+
+   if (src == dst) {
+      if (readOffset + size <= writeOffset) {
+         /* OK */
+      }
+      else if (writeOffset + size <= readOffset) {
+         /* OK */
+      }
+      else {
+         /* overlapping src/dst is illegal */
+         _mesa_error(ctx, GL_INVALID_VALUE,
+                     "glCopyBuffserSubData(overlapping src/dst)");
+         return;
+      }
+   }
+
+   ctx->Driver.CopyBufferSubData(ctx, src, dst, readOffset, writeOffset, size);
+}
+
+
+/**
+ * See GL_ARB_map_buffer_range spec
+ */
+void * GLAPIENTRY
+_mesa_MapBufferRange(GLenum target, GLintptr offset, GLsizeiptr length,
+                     GLbitfield access)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *bufObj;
+   void *map;
+
+   ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, NULL);
+
+   if (!ctx->Extensions.ARB_map_buffer_range) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(extension not supported)");
+      return NULL;
+   }
+
+   if (offset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset = %ld)", offset);
+      return NULL;
+   }
+
+   if (length < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(length = %ld)", length);
+      return NULL;
+   }
+
+   if ((access & (GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)) == 0) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(access indicates neither read or write)");
+      return NULL;
+   }
+
+   if (access & GL_MAP_READ_BIT) {
+      if ((access & GL_MAP_INVALIDATE_RANGE_BIT) ||
+          (access & GL_MAP_INVALIDATE_BUFFER_BIT) ||
+          (access & GL_MAP_UNSYNCHRONIZED_BIT)) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glMapBufferRange(invalid access flags)");
+         return NULL;
+      }
+   }
+
+   if ((access & GL_MAP_FLUSH_EXPLICIT_BIT) &&
+       ((access & GL_MAP_WRITE_BIT) == 0)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(invalid access flags)");
+      return NULL;
+   }
+
+   bufObj = get_buffer(ctx, target);
+   if (!bufObj || !_mesa_is_bufferobj(bufObj)) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glMapBufferRange(target = 0x%x)", target);
+      return NULL;
+   }
+
+   if (offset + length > bufObj->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset + length > size)");
+      return NULL;
+   }
+
+   if (_mesa_bufferobj_mapped(bufObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(buffer already mapped)");
+      return NULL;
+   }
+      
+   ASSERT(ctx->Driver.MapBufferRange);
+   map = ctx->Driver.MapBufferRange(ctx, target, offset, length,
+                                    access, bufObj);
+   if (!map) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMapBufferARB(access)");
+   }
+   else {
+      /* The driver callback should have set all these fields.
+       * This is important because other modules (like VBO) might call
+       * the driver function directly.
+       */
+      ASSERT(bufObj->Pointer == map);
+      ASSERT(bufObj->Length == length);
+      ASSERT(bufObj->Offset == offset);
+      ASSERT(bufObj->AccessFlags == access);
+   }
+
+   return map;
+}
+
+
+/**
+ * See GL_ARB_map_buffer_range spec
+ */
+void GLAPIENTRY
+_mesa_FlushMappedBufferRange(GLenum target, GLintptr offset, GLsizeiptr length)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *bufObj;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!ctx->Extensions.ARB_map_buffer_range) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(extension not supported)");
+      return;
+   }
+
+   if (offset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset = %ld)", offset);
+      return;
+   }
+
+   if (length < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(length = %ld)", length);
+      return;
+   }
+
+   bufObj = get_buffer(ctx, target);
+   if (!bufObj) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glMapBufferRange(target = 0x%x)", target);
+      return;
+   }
+
+   if (!_mesa_is_bufferobj(bufObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(current buffer is 0)");
+      return;
+   }
+
+   if (!_mesa_bufferobj_mapped(bufObj)) {
+      /* buffer is not mapped */
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(buffer is not mapped)");
+      return;
+   }
+
+   if ((bufObj->AccessFlags & GL_MAP_FLUSH_EXPLICIT_BIT) == 0) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(GL_MAP_FLUSH_EXPLICIT_BIT not set)");
+      return;
+   }
+
+   if (offset + length > bufObj->Length) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+             "glMapBufferRange(offset %ld + length %ld > mapped length %ld)",
+             offset, length, bufObj->Length);
+      return;
+   }
+
+   ASSERT(bufObj->AccessFlags & GL_MAP_WRITE_BIT);
+
+   if (ctx->Driver.FlushMappedBufferRange)
+      ctx->Driver.FlushMappedBufferRange(ctx, target, offset, length, bufObj);
+}
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 3c08f0083c..9f732ec0c0 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -35,17 +36,32 @@
  * Internal functions
  */
 
+
+/** Is the given buffer object currently mapped? */
+static INLINE GLboolean
+_mesa_bufferobj_mapped(const struct gl_buffer_object *obj)
+{
+   return obj->Pointer != NULL;
+}
+
+/**
+ * Is the given buffer object a user-created buffer object?
+ * Mesa uses default buffer objects in several places.  Default buffers
+ * always have Name==0.  User created buffers have Name!=0.
+ */
+static INLINE GLboolean
+_mesa_is_bufferobj(const struct gl_buffer_object *obj)
+{
+   return obj->Name != 0;
+}
+
+
 extern void
 _mesa_init_buffer_objects( GLcontext *ctx );
 
 extern void
 _mesa_update_default_objects_buffer_objects(GLcontext *ctx);
 
-extern struct gl_buffer_object *
-_mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target );
-
-extern void
-_mesa_delete_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj );
 
 extern struct gl_buffer_object *
 _mesa_lookup_bufferobj(GLcontext *ctx, GLuint buffer);
@@ -59,63 +75,50 @@ _mesa_reference_buffer_object(GLcontext *ctx,
                               struct gl_buffer_object **ptr,
                               struct gl_buffer_object *bufObj);
 
-extern void
-_mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
-		   const GLvoid * data, GLenum usage,
-		   struct gl_buffer_object * bufObj );
-
-extern void
-_mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
-		      GLsizeiptrARB size, const GLvoid * data,
-		      struct gl_buffer_object * bufObj );
-
-extern void
-_mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
-			  GLsizeiptrARB size, GLvoid * data,
-			  struct gl_buffer_object * bufObj );
-
-extern void *
-_mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
-		  struct gl_buffer_object * bufObj );
-
-extern GLboolean
-_mesa_buffer_unmap( GLcontext *ctx, GLenum target,
-                    struct gl_buffer_object * bufObj );
-
 extern GLboolean
 _mesa_validate_pbo_access(GLuint dimensions,
                           const struct gl_pixelstore_attrib *pack,
                           GLsizei width, GLsizei height, GLsizei depth,
                           GLenum format, GLenum type, const GLvoid *ptr);
 
-extern const GLubyte *
-_mesa_map_bitmap_pbo(GLcontext *ctx,
+extern const GLvoid *
+_mesa_map_pbo_source(GLcontext *ctx,
                      const struct gl_pixelstore_attrib *unpack,
-                     const GLubyte *bitmap);
-
-extern void
-_mesa_unmap_bitmap_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack);
+                     const GLvoid *src);
 
 extern const GLvoid *
-_mesa_map_drawpix_pbo(GLcontext *ctx,
-                      const struct gl_pixelstore_attrib *unpack,
-                      const GLvoid *pixels);
+_mesa_map_validate_pbo_source(GLcontext *ctx,
+                              GLuint dimensions,
+                              const struct gl_pixelstore_attrib *unpack,
+                              GLsizei width, GLsizei height, GLsizei depth,
+                              GLenum format, GLenum type, const GLvoid *ptr,
+                              const char *where);
 
 extern void
-_mesa_unmap_drawpix_pbo(GLcontext *ctx,
-                        const struct gl_pixelstore_attrib *unpack);
+_mesa_unmap_pbo_source(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack);
 
 extern void *
-_mesa_map_readpix_pbo(GLcontext *ctx,
-                      const struct gl_pixelstore_attrib *pack,
-                      GLvoid *dest);
+_mesa_map_pbo_dest(GLcontext *ctx,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest);
+
+extern GLvoid *
+_mesa_map_validate_pbo_dest(GLcontext *ctx,
+                            GLuint dimensions,
+                            const struct gl_pixelstore_attrib *unpack,
+                            GLsizei width, GLsizei height, GLsizei depth,
+                            GLenum format, GLenum type, GLvoid *ptr,
+                            const char *where);
 
 extern void
-_mesa_unmap_readpix_pbo(GLcontext *ctx,
-                        const struct gl_pixelstore_attrib *pack);
+_mesa_unmap_pbo_dest(GLcontext *ctx,
+                     const struct gl_pixelstore_attrib *pack);
 
 
+extern void
+_mesa_init_buffer_object_functions(struct dd_function_table *driver);
+
 
 /*
  * API functions
@@ -154,4 +157,16 @@ _mesa_GetBufferParameterivARB(GLenum target, GLenum pname, GLint *params);
 extern void GLAPIENTRY
 _mesa_GetBufferPointervARB(GLenum target, GLenum pname, GLvoid **params);
 
+extern void GLAPIENTRY
+_mesa_CopyBufferSubData(GLenum readTarget, GLenum writeTarget,
+                        GLintptr readOffset, GLintptr writeOffset,
+                        GLsizeiptr size);
+
+extern void * GLAPIENTRY
+_mesa_MapBufferRange(GLenum target, GLintptr offset, GLsizeiptr length,
+                     GLbitfield access);
+
+extern void GLAPIENTRY
+_mesa_FlushMappedBufferRange(GLenum target, GLintptr offset, GLsizeiptr length);
+
 #endif
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index c5f13345f0..d8b5f3b1f4 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -443,7 +443,7 @@ _mesa_readbuffer(GLcontext *ctx, GLenum buffer, GLint bufferIndex)
    fb->ColorReadBuffer = buffer;
    fb->_ColorReadBufferIndex = bufferIndex;
 
-   ctx->NewState |= _NEW_PIXEL;
+   ctx->NewState |= _NEW_BUFFERS;
 }
 
 
diff --git a/src/mesa/main/colortab.c b/src/mesa/main/colortab.c
index bd9cf438b4..5a7de5f209 100644
--- a/src/mesa/main/colortab.c
+++ b/src/mesa/main/colortab.c
@@ -179,26 +179,12 @@ store_colortable_entries(GLcontext *ctx, struct gl_color_table *table,
 			 GLfloat bScale, GLfloat bBias,
 			 GLfloat aScale, GLfloat aBias)
 {
-   if (ctx->Unpack.BufferObj->Name) {
-      /* Get/unpack the color table data from a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Unpack, count, 1, 1,
-                                     format, type, data)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glColor[Sub]Table(bad PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glColor[Sub]Table(PBO mapped)");
-         return;
-      }
-      data = ADD_POINTERS(buf, data);
-   }
-
+   data = _mesa_map_validate_pbo_source(ctx, 
+                                        1, &ctx->Unpack, count, 1, 1,
+                                        format, type, data,
+                                        "glColor[Sub]Table");
+   if (!data)
+      return;
 
    {
       /* convert user-provided data to GLfloat values */
@@ -279,10 +265,7 @@ store_colortable_entries(GLcontext *ctx, struct gl_color_table *table,
       }
    }
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 }
 
 
@@ -696,34 +679,17 @@ _mesa_GetColorTable( GLenum target, GLenum format,
       return;
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack color table into PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Pack, table->Size, 1, 1,
-                                     format, type, data)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetColorTable(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetColorTable(PBO is mapped)");
-         return;
-      }
-      data = ADD_POINTERS(buf, data);
-   }
+   data = _mesa_map_validate_pbo_dest(ctx, 
+                                      1, &ctx->Pack, table->Size, 1, 1,
+                                      format, type, data,
+                                      "glGetColorTable");
+   if (!data)
+      return;
 
    _mesa_pack_rgba_span_float(ctx, table->Size, rgba,
                               format, type, data, &ctx->Pack, 0x0);
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index 114119006a..8a09efdb53 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -170,10 +170,48 @@
 /** For GL_EXT_texture_lod_bias (typically MAX_TEXTURE_LEVELS - 1) */
 #define MAX_TEXTURE_LOD_BIAS 12.0
 
+/** For any program target/extension */
+/*@{*/
+#define MAX_PROGRAM_INSTRUCTIONS       (16 * 1024)
+
+/**
+ * Per-program constants (power of two)
+ *
+ * \c MAX_PROGRAM_LOCAL_PARAMS and \c MAX_UNIFORMS are just the assmebly shader
+ * and GLSL shader names for the same thing.  They should \b always have the
+ * same value.  Each refers to the number of vec4 values supplied as
+ * per-program parameters.
+ */
+/*@{*/
+#define MAX_PROGRAM_LOCAL_PARAMS       1024
+#define MAX_UNIFORMS                   1024
+/*@}*/
+
+/**
+ * Per-context constants (power of two)
+ *
+ * \note
+ * This value should always be less than or equal to \c MAX_PROGRAM_LOCAL_PARAMS
+ * and \c MAX_VERTEX_PROGRAM_PARAMS.  Otherwise some applications will make
+ * incorrect assumptions.
+ */
+#define MAX_PROGRAM_ENV_PARAMS         256
+
+#define MAX_PROGRAM_MATRICES           8
+#define MAX_PROGRAM_MATRIX_STACK_DEPTH 4
+#define MAX_PROGRAM_CALL_DEPTH         8
+#define MAX_PROGRAM_TEMPS              256
+#define MAX_PROGRAM_ADDRESS_REGS       2
+#define MAX_VARYING                    16    /**< number of float[4] vectors */
+#define MAX_SAMPLERS                   MAX_TEXTURE_IMAGE_UNITS
+#define MAX_PROGRAM_INPUTS             32
+#define MAX_PROGRAM_OUTPUTS            32
+/*@}*/
+
 /** For GL_ARB_vertex_program */
 /*@{*/
 #define MAX_VERTEX_PROGRAM_ADDRESS_REGS 1
-#define MAX_VERTEX_PROGRAM_ATTRIBS     16
+#define MAX_VERTEX_PROGRAM_PARAMS       MAX_UNIFORMS
 /*@}*/
 
 /** For GL_ARB_fragment_program */
@@ -181,23 +219,6 @@
 #define MAX_FRAGMENT_PROGRAM_ADDRESS_REGS 0
 /*@}*/
 
-/** For any program target/extension */
-/*@{*/
-#define MAX_PROGRAM_INSTRUCTIONS  (16 * 1024)
-#define MAX_PROGRAM_LOCAL_PARAMS 256 /**< per-program constants (power of two) */
-#define MAX_PROGRAM_ENV_PARAMS 256   /**< per-context constants (power of two) */
-#define MAX_PROGRAM_MATRICES 8
-#define MAX_PROGRAM_MATRIX_STACK_DEPTH 4
-#define MAX_PROGRAM_CALL_DEPTH 8
-#define MAX_PROGRAM_TEMPS 256
-#define MAX_PROGRAM_ADDRESS_REGS 2
-#define MAX_UNIFORMS 1024  /**< number of vec4 uniforms */
-#define MAX_VARYING 16     /**< number of float[4] vectors */
-#define MAX_SAMPLERS MAX_TEXTURE_IMAGE_UNITS
-#define MAX_PROGRAM_INPUTS 32
-#define MAX_PROGRAM_OUTPUTS 32
-/*@}*/
-
 /** For GL_NV_vertex_program */
 /*@{*/
 #define MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS 128
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 0a4c9acdfe..f6d4ac4595 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -129,6 +129,9 @@
 #if FEATURE_ARB_occlusion_query
 #include "queryobj.h"
 #endif
+#if FEATURE_ARB_sync
+#include "syncobj.h"
+#endif
 #if FEATURE_drawpix
 #include "rastpos.h"
 #endif
@@ -150,6 +153,7 @@
 #include "glapi/glapioffsets.h"
 #include "glapi/glapitable.h"
 #include "shader/program.h"
+#include "shader/prog_print.h"
 #include "shader/shader_api.h"
 #if FEATURE_ATI_fragment_shader
 #include "shader/atifragshader.h"
@@ -455,7 +459,7 @@ _mesa_init_current(GLcontext *ctx)
    GLuint i;
 
    /* Init all to (0,0,0,1) */
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(ctx->Current.Attrib); i++) {
       ASSIGN_4V( ctx->Current.Attrib[i], 0.0, 0.0, 0.0, 1.0 );
    }
 
@@ -486,7 +490,7 @@ init_program_limits(GLenum type, struct gl_program_constants *prog)
    prog->MaxUniformComponents = 4 * MAX_UNIFORMS;
 
    if (type == GL_VERTEX_PROGRAM_ARB) {
-      prog->MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
+      prog->MaxParameters = MAX_VERTEX_PROGRAM_PARAMS;
       prog->MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
       prog->MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
    }
@@ -496,15 +500,17 @@ init_program_limits(GLenum type, struct gl_program_constants *prog)
       prog->MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
    }
 
-   /* copy the above limits to init native limits */
-   prog->MaxNativeInstructions = prog->MaxInstructions;
-   prog->MaxNativeAluInstructions = prog->MaxAluInstructions;
-   prog->MaxNativeTexInstructions = prog->MaxTexInstructions;
-   prog->MaxNativeTexIndirections = prog->MaxTexIndirections;
-   prog->MaxNativeAttribs = prog->MaxAttribs;
-   prog->MaxNativeTemps = prog->MaxTemps;
-   prog->MaxNativeAddressRegs = prog->MaxAddressRegs;
-   prog->MaxNativeParameters = prog->MaxParameters;
+   /* Set the native limits to zero.  This implies that there is no native
+    * support for shaders.  Let the drivers fill in the actual values.
+    */
+   prog->MaxNativeInstructions = 0;
+   prog->MaxNativeAluInstructions = 0;
+   prog->MaxNativeTexInstructions = 0;
+   prog->MaxNativeTexIndirections = 0;
+   prog->MaxNativeAttribs = 0;
+   prog->MaxNativeTemps = 0;
+   prog->MaxNativeAddressRegs = 0;
+   prog->MaxNativeParameters = 0;
 }
 
 
@@ -589,9 +595,15 @@ _mesa_init_constants(GLcontext *ctx)
    /* GL_ARB_framebuffer_object */
    ctx->Const.MaxSamples = 0;
 
+   /* GL_ARB_sync */
+   ctx->Const.MaxServerWaitTimeout = (GLuint64) ~0;
+
    /* GL_ATI_envmap_bumpmap */
    ctx->Const.SupportedBumpUnits = SUPPORTED_ATI_BUMP_UNITS;
 
+   /* GL_EXT_provoking_vertex */
+   ctx->Const.QuadsFollowProvokingVertexConvention = GL_TRUE;
+
    /* sanity checks */
    ASSERT(ctx->Const.MaxTextureUnits == MIN2(ctx->Const.MaxTextureImageUnits,
                                              ctx->Const.MaxTextureCoordUnits));
@@ -602,6 +614,10 @@ _mesa_init_constants(GLcontext *ctx)
    ASSERT(MAX_NV_VERTEX_PROGRAM_TEMPS <= MAX_PROGRAM_TEMPS);
    ASSERT(MAX_NV_VERTEX_PROGRAM_INPUTS <= VERT_ATTRIB_MAX);
    ASSERT(MAX_NV_VERTEX_PROGRAM_OUTPUTS <= VERT_RESULT_MAX);
+
+   /* check that we don't exceed various 32-bit bitfields */
+   ASSERT(VERT_RESULT_MAX <= 32);
+   ASSERT(FRAG_ATTRIB_MAX <= 32);
 }
 
 
@@ -705,6 +721,9 @@ init_attrib_groups(GLcontext *ctx)
 #if FEATURE_ARB_occlusion_query
    _mesa_init_query( ctx );
 #endif
+#if FEATURE_ARB_sync
+   _mesa_init_sync( ctx );
+#endif
 #if FEATURE_drawpix
    _mesa_init_rastpos( ctx );
 #endif
@@ -1004,6 +1023,10 @@ _mesa_free_context_data( GLcontext *ctx )
 #if FEATURE_ARB_occlusion_query
    _mesa_free_query_data(ctx);
 #endif
+#if FEATURE_ARB_sync
+   _mesa_free_sync_data(ctx);
+#endif
+   _mesa_free_varray_data(ctx);
 
    _mesa_delete_array_object(ctx, ctx->Array.DefaultArrayObj);
 
@@ -1017,10 +1040,6 @@ _mesa_free_context_data( GLcontext *ctx )
    _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, NULL);
 #endif
 
-#if FEATURE_ARB_vertex_buffer_object
-   _mesa_delete_buffer_object(ctx, ctx->Array.NullBufferObj);
-#endif
-
    /* free dispatch tables */
    _mesa_free(ctx->Exec);
    _mesa_free(ctx->Save);
@@ -1572,4 +1591,82 @@ _mesa_set_mvp_with_dp4( GLcontext *ctx,
 }
 
 
+
+/**
+ * Prior to drawing anything with glBegin, glDrawArrays, etc. this function
+ * is called to see if it's valid to render.  This involves checking that
+ * the current shader is valid and the framebuffer is complete.
+ * If an error is detected it'll be recorded here.
+ * \return GL_TRUE if OK to render, GL_FALSE if not
+ */
+GLboolean
+_mesa_valid_to_render(GLcontext *ctx, const char *where)
+{
+   if (ctx->Shader.CurrentProgram) {
+      /* using shaders */
+      if (!ctx->Shader.CurrentProgram->LinkStatus) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "%s(shader not linked), where");
+         return GL_FALSE;
+      }
+#if 0 /* not normally enabled */
+      {
+         char errMsg[100];
+         if (!_mesa_validate_shader_program(ctx, ctx->Shader.CurrentProgram,
+                                            errMsg)) {
+            _mesa_warning(ctx, "Shader program %u is invalid: %s",
+                          ctx->Shader.CurrentProgram->Name, errMsg);
+         }
+      }
+#endif
+   }
+   else {
+      if (ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "%s(vertex program not valid)", where);
+         return GL_FALSE;
+      }
+      if (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "%s(fragment program not valid)", where);
+         return GL_FALSE;
+      }
+   }
+
+   if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+      _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
+                  "%s(incomplete framebuffer)", where);
+      return GL_FALSE;
+   }
+
+#ifdef DEBUG
+   if (ctx->Shader.Flags & GLSL_LOG) {
+      struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+      if (shProg) {
+         if (!shProg->_Used) {
+            /* This is the first time this shader is being used.
+             * Append shader's constants/uniforms to log file.
+             */
+            GLuint i;
+            for (i = 0; i < shProg->NumShaders; i++) {
+               struct gl_shader *sh = shProg->Shaders[i];
+               if (sh->Type == GL_VERTEX_SHADER) {
+                  _mesa_append_uniforms_to_file(sh,
+                                                &shProg->VertexProgram->Base);
+               }
+               else if (sh->Type == GL_FRAGMENT_SHADER) {
+                  _mesa_append_uniforms_to_file(sh,
+                                                &shProg->FragmentProgram->Base);
+               }
+            }
+            shProg->_Used = GL_TRUE;
+         }
+      }
+   }
+#endif
+
+   return GL_TRUE;
+}
+
+
 /*@}*/
diff --git a/src/mesa/main/context.h b/src/mesa/main/context.h
index 0531ae8ee8..5587695fa0 100644
--- a/src/mesa/main/context.h
+++ b/src/mesa/main/context.h
@@ -159,6 +159,11 @@ _mesa_set_mvp_with_dp4( GLcontext *ctx,
                         GLboolean flag );
 
 
+extern GLboolean
+_mesa_valid_to_render(GLcontext *ctx, const char *where);
+
+
+
 /** \name Miscellaneous */
 /*@{*/
 
@@ -174,7 +179,6 @@ _mesa_Flush( void );
 /*@}*/
 
 
-
 /**
  * \name Macros for flushing buffered rendering commands before state changes,
  * checking if inside glBegin/glEnd, etc.
diff --git a/src/mesa/main/convolve.c b/src/mesa/main/convolve.c
index 814c6a0a5a..70951112a1 100644
--- a/src/mesa/main/convolve.c
+++ b/src/mesa/main/convolve.c
@@ -144,39 +144,19 @@ _mesa_ConvolutionFilter1D(GLenum target, GLenum internalFormat, GLsizei width, G
    ctx->Convolution1D.Width = width;
    ctx->Convolution1D.Height = 1;
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack filter from PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Unpack, width, 1, 1,
-                                     format, type, image)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glConvolutionFilter1D(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glConvolutionFilter1D(PBO is mapped)");
-         return;
-      }
-      image = ADD_POINTERS(buf, image);
-   }
-   else if (!image) {
+   image = _mesa_map_validate_pbo_source(ctx, 
+                                        1, &ctx->Unpack, width, 1, 1,
+                                        format, type, image,
+                                        "glConvolutionFilter1D");
+   if (!image)
       return;
-   }
 
    _mesa_unpack_color_span_float(ctx, width, GL_RGBA,
                                  ctx->Convolution1D.Filter,
                                  format, type, image, &ctx->Unpack,
                                  0); /* transferOps */
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 
    _mesa_scale_and_bias_rgba(width,
                              (GLfloat (*)[4]) ctx->Convolution1D.Filter,
@@ -242,29 +222,12 @@ _mesa_ConvolutionFilter2D(GLenum target, GLenum internalFormat, GLsizei width, G
    ctx->Convolution2D.Width = width;
    ctx->Convolution2D.Height = height;
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack filter from PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1,
-                                     format, type, image)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glConvolutionFilter2D(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glConvolutionFilter2D(PBO is mapped)");
-         return;
-      }
-      image = ADD_POINTERS(buf, image);
-   }
-   else if (!image) {
+   image = _mesa_map_validate_pbo_source(ctx, 
+                                         2, &ctx->Unpack, width, height, 1,
+                                         format, type, image,
+                                         "glConvolutionFilter2D");
+   if (!image)
       return;
-   }
 
    /* Unpack filter image.  We always store filters in RGBA format. */
    for (i = 0; i < height; i++) {
@@ -276,10 +239,7 @@ _mesa_ConvolutionFilter2D(GLenum target, GLenum internalFormat, GLsizei width, G
                                     0); /* transferOps */
    }
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 
    _mesa_scale_and_bias_rgba(width * height,
                              (GLfloat (*)[4]) ctx->Convolution2D.Filter,
@@ -598,27 +558,12 @@ _mesa_GetConvolutionFilter(GLenum target, GLenum format, GLenum type,
          return;
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* Pack the filter into a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, &ctx->Pack,
-                                     filter->Width, filter->Height,
-                                     1, format, type, image)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetConvolutionFilter(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetConvolutionFilter(PBO is mapped)");
-         return;
-      }
-      image = ADD_POINTERS(image, buf);
-   }
+   image = _mesa_map_validate_pbo_dest(ctx, 2, &ctx->Pack,
+                                       filter->Width, filter->Height, 1,
+                                       format, type, image,
+                                       "glGetConvolutionFilter");
+   if (!image)
+      return;
 
    for (row = 0; row < filter->Height; row++) {
       GLvoid *dst = _mesa_image_address2d(&ctx->Pack, image, filter->Width,
@@ -629,10 +574,7 @@ _mesa_GetConvolutionFilter(GLenum target, GLenum format, GLenum type,
                                  format, type, dst, &ctx->Pack, 0x0);
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
@@ -802,59 +744,35 @@ _mesa_GetSeparableFilter(GLenum target, GLenum format, GLenum type,
 
    filter = &ctx->Separable2D;
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* Pack filter into PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Pack, filter->Width, 1, 1,
-                                     format, type, row)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetSeparableFilter(invalid PBO access, width)");
-         return;
-      }
-      if (!_mesa_validate_pbo_access(1, &ctx->Pack, filter->Height, 1, 1,
-                                     format, type, column)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetSeparableFilter(invalid PBO access, height)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetSeparableFilter(PBO is mapped)");
-         return;
-      }
-      row = ADD_POINTERS(buf, row);
-      column = ADD_POINTERS(buf, column);
-   }
-
-   /* Row filter */
+   /* Get row filter */
+   row = _mesa_map_validate_pbo_dest(ctx, 1, &ctx->Pack,
+                                     filter->Width, 1, 1,
+                                     format, type, row,
+                                     "glGetConvolutionFilter");
    if (row) {
       GLvoid *dst = _mesa_image_address1d(&ctx->Pack, row, filter->Width,
                                           format, type, 0);
       _mesa_pack_rgba_span_float(ctx, filter->Width,
                                  (GLfloat (*)[4]) filter->Filter,
                                  format, type, dst, &ctx->Pack, 0x0);
+      _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
    }
 
-   /* Column filter */
+   /* get column filter */
+   column = _mesa_map_validate_pbo_dest(ctx, 1, &ctx->Pack,
+                                        filter->Height, 1, 1,
+                                        format, type, column,
+                                        "glGetConvolutionFilter");
    if (column) {
       GLvoid *dst = _mesa_image_address1d(&ctx->Pack, column, filter->Height,
                                           format, type, 0);
       GLfloat (*src)[4] = (GLfloat (*)[4]) (filter->Filter + colStart);
       _mesa_pack_rgba_span_float(ctx, filter->Height, src,
                                  format, type, dst, &ctx->Pack, 0x0);
+      _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
    }
 
    (void) span;  /* unused at this time */
-
-   if (ctx->Pack.BufferObj->Name) {
-      /* Pack filter into PBO */
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
 }
 
 
@@ -905,41 +823,16 @@ _mesa_SeparableFilter2D(GLenum target, GLenum internalFormat, GLsizei width, GLs
    ctx->Separable2D.Width = width;
    ctx->Separable2D.Height = height;
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack filter from PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Unpack, width, 1, 1,
-                                     format, type, row)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glSeparableFilter2D(invalid PBO access, width)");
-         return;
-      }
-      if (!_mesa_validate_pbo_access(1, &ctx->Unpack, height, 1, 1,
-                                     format, type, column)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glSeparableFilter2D(invalid PBO access, height)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glSeparableFilter2D(PBO is mapped)");
-         return;
-      }
-      row = ADD_POINTERS(buf, row);
-      column = ADD_POINTERS(buf, column);
-   }
-
    /* unpack row filter */
+   row = _mesa_map_validate_pbo_source(ctx, 1, &ctx->Unpack,
+                                       width, 1, 1,
+                                       format, type, row,
+                                       "glSeparableFilter2D");
    if (row) {
       _mesa_unpack_color_span_float(ctx, width, GL_RGBA,
                                     ctx->Separable2D.Filter,
                                     format, type, row, &ctx->Unpack,
-                                    0);  /* transferOps */
-
+                                    0x0);  /* transferOps */
       _mesa_scale_and_bias_rgba(width,
                              (GLfloat (*)[4]) ctx->Separable2D.Filter,
                              ctx->Pixel.ConvolutionFilterScale[2][0],
@@ -950,9 +843,14 @@ _mesa_SeparableFilter2D(GLenum target, GLenum internalFormat, GLsizei width, GLs
                              ctx->Pixel.ConvolutionFilterBias[2][1],
                              ctx->Pixel.ConvolutionFilterBias[2][2],
                              ctx->Pixel.ConvolutionFilterBias[2][3]);
+      _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
    }
 
    /* unpack column filter */
+   column = _mesa_map_validate_pbo_source(ctx, 1, &ctx->Unpack,
+                                          height, 1, 1,
+                                          format, type, column,
+                                          "glSeparableFilter2D");
    if (column) {
       _mesa_unpack_color_span_float(ctx, height, GL_RGBA,
                                     &ctx->Separable2D.Filter[colStart],
@@ -969,9 +867,10 @@ _mesa_SeparableFilter2D(GLenum target, GLenum internalFormat, GLsizei width, GLs
                        ctx->Pixel.ConvolutionFilterBias[2][1],
                        ctx->Pixel.ConvolutionFilterBias[2][2],
                        ctx->Pixel.ConvolutionFilterBias[2][3]);
+      _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
    }
 
-   if (ctx->Unpack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Unpack.BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                               ctx->Unpack.BufferObj);
    }
diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h
index 32b1d4e9fa..4a700b5cb4 100644
--- a/src/mesa/main/dd.h
+++ b/src/mesa/main/dd.h
@@ -766,6 +766,8 @@ struct dd_function_table {
    /** Return the value or values of a selected parameter */
    GLboolean (*GetIntegerv)(GLcontext *ctx, GLenum pname, GLint *result);
    /** Return the value or values of a selected parameter */
+   GLboolean (*GetInteger64v)(GLcontext *ctx, GLenum pname, GLint64 *result);
+   /** Return the value or values of a selected parameter */
    GLboolean (*GetPointerv)(GLcontext *ctx, GLenum pname, GLvoid **result);
    /*@}*/
    
@@ -783,9 +785,9 @@ struct dd_function_table {
    
    void (*DeleteBuffer)( GLcontext *ctx, struct gl_buffer_object *obj );
 
-   void (*BufferData)( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
-		       const GLvoid *data, GLenum usage,
-		       struct gl_buffer_object *obj );
+   GLboolean (*BufferData)( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
+                            const GLvoid *data, GLenum usage,
+                            struct gl_buffer_object *obj );
 
    void (*BufferSubData)( GLcontext *ctx, GLenum target, GLintptrARB offset,
 			  GLsizeiptrARB size, const GLvoid *data,
@@ -798,6 +800,12 @@ struct dd_function_table {
    void * (*MapBuffer)( GLcontext *ctx, GLenum target, GLenum access,
 			struct gl_buffer_object *obj );
 
+   void (*CopyBufferSubData)( GLcontext *ctx,
+                              struct gl_buffer_object *src,
+                              struct gl_buffer_object *dst,
+                              GLintptr readOffset, GLintptr writeOffset,
+                              GLsizeiptr size );
+
    /* May return NULL if MESA_MAP_NOWAIT_BIT is set in access:
     */
    void * (*MapBufferRange)( GLcontext *ctx, GLenum target,
@@ -1036,6 +1044,22 @@ struct dd_function_table {
     */
    void (*EndCallList)( GLcontext *ctx );
 
+
+#if FEATURE_ARB_sync
+   /**
+    * \name GL_ARB_sync interfaces
+    */
+   /*@{*/
+   struct gl_sync_object * (*NewSyncObject)(GLcontext *, GLenum);
+   void (*FenceSync)(GLcontext *, struct gl_sync_object *, GLenum, GLbitfield);
+   void (*DeleteSyncObject)(GLcontext *, struct gl_sync_object *);
+   void (*CheckSync)(GLcontext *, struct gl_sync_object *);
+   void (*ClientWaitSync)(GLcontext *, struct gl_sync_object *,
+			  GLbitfield, GLuint64);
+   void (*ServerWaitSync)(GLcontext *, struct gl_sync_object *,
+			  GLbitfield, GLuint64);
+   /*@}*/
+#endif
 };
 
 
@@ -1144,7 +1168,11 @@ typedef struct {
    void (GLAPIENTRYP DrawRangeElements)( GLenum mode, GLuint start,
 			      GLuint end, GLsizei count,
 			      GLenum type, const GLvoid *indices );
-   /*@}*/
+   void (GLAPIENTRYP MultiDrawElementsEXT)( GLenum mode, const GLsizei *count,
+					    GLenum type,
+					    const GLvoid **indices,
+					    GLsizei primcount);
+  /*@}*/
 
    /**
     * \name Eval
diff --git a/src/mesa/main/debug.c b/src/mesa/main/debug.c
index b54e47919d..8492c8561d 100644
--- a/src/mesa/main/debug.c
+++ b/src/mesa/main/debug.c
@@ -24,12 +24,16 @@
  */
 
 #include "mtypes.h"
+#include "attrib.h"
 #include "colormac.h"
 #include "context.h"
 #include "hash.h"
 #include "imports.h"
 #include "debug.h"
 #include "get.h"
+#include "pixelstore.h"
+#include "readpix.h"
+#include "texgetimage.h"
 #include "texobj.h"
 #include "texformat.h"
 
@@ -231,11 +235,11 @@ _mesa_init_debug( GLcontext *ctx )
  */
 static void
 write_ppm(const char *filename, const GLubyte *buffer, int width, int height,
-          int comps, int rcomp, int gcomp, int bcomp)
+          int comps, int rcomp, int gcomp, int bcomp, GLboolean invert)
 {
    FILE *f = fopen( filename, "w" );
    if (f) {
-      int i, x, y;
+      int x, y;
       const GLubyte *ptr = buffer;
       fprintf(f,"P6\n");
       fprintf(f,"# ppm-file created by osdemo.c\n");
@@ -243,10 +247,11 @@ write_ppm(const char *filename, const GLubyte *buffer, int width, int height,
       fprintf(f,"255\n");
       fclose(f);
       f = fopen( filename, "ab" );  /* reopen in binary append mode */
-      for (y=height-1; y>=0; y--) {
-         for (x=0; x<width; x++) {
-            i = (y*width + x) * comps;
-            fputc(ptr[i+rcomp], f);   /* write red */
+      for (y=0; y < height; y++) {
+         for (x = 0; x < width; x++) {
+            int yy = invert ? (height - 1 - y) : y;
+            int i = (yy * width + x) * comps;
+            fputc(ptr[i+rcomp], f); /* write red */
             fputc(ptr[i+gcomp], f); /* write green */
             fputc(ptr[i+bcomp], f); /* write blue */
          }
@@ -260,47 +265,34 @@ write_ppm(const char *filename, const GLubyte *buffer, int width, int height,
  * Write level[0] image to a ppm file.
  */
 static void
-write_texture_image(struct gl_texture_object *texObj)
+write_texture_image(struct gl_texture_object *texObj, GLuint face, GLuint level)
 {
-   const struct gl_texture_image *img = texObj->Image[0][0];
-   if (img && img->Data) {
+   struct gl_texture_image *img = texObj->Image[face][level];
+   if (img) {
+      GET_CURRENT_CONTEXT(ctx);
+      struct gl_pixelstore_attrib store;
+      GLubyte *buffer;
       char s[100];
 
+      buffer = (GLubyte *) _mesa_malloc(img->Width * img->Height
+                                        * img->Depth * 4);
+
+      store = ctx->Pack; /* save */
+      ctx->Pack = ctx->DefaultPacking;
+
+      ctx->Driver.GetTexImage(ctx, texObj->Target, level,
+                              GL_RGBA, GL_UNSIGNED_BYTE,
+                              buffer, texObj, img);
+
       /* make filename */
-      sprintf(s, "/tmp/teximage%u.ppm", texObj->Name);
-
-      switch (img->TexFormat->MesaFormat) {
-      case MESA_FORMAT_RGBA8888:
-         write_ppm(s, img->Data, img->Width, img->Height, 4, 3, 2, 1);
-         break;
-      case MESA_FORMAT_ARGB8888:
-         write_ppm(s, img->Data, img->Width, img->Height, 4, 2, 1, 0);
-         break;
-      case MESA_FORMAT_RGB888:
-         write_ppm(s, img->Data, img->Width, img->Height, 3, 2, 1, 0);
-         break;
-      case MESA_FORMAT_RGB565:
-         {
-            GLubyte *buf2 = (GLubyte *) _mesa_malloc(img->Width * img->Height * 3);
-            GLuint i;
-            for (i = 0; i < img->Width * img->Height; i++) {
-               GLint r, g, b;
-               GLushort s = ((GLushort *) img->Data)[i];
-               r = UBYTE_TO_CHAN( ((s >> 8) & 0xf8) | ((s >> 13) & 0x7) );
-               g = UBYTE_TO_CHAN( ((s >> 3) & 0xfc) | ((s >>  9) & 0x3) );
-               b = UBYTE_TO_CHAN( ((s << 3) & 0xf8) | ((s >>  2) & 0x7) );
-               buf2[i*3+1] = r;
-               buf2[i*3+2] = g;
-               buf2[i*3+3] = b;
-            }
-            write_ppm(s, buf2, img->Width, img->Height, 3, 2, 1, 0);
-            _mesa_free(buf2);
-         }
-         break;
-      default:
-         printf("XXXX unsupported mesa tex format %d in %s\n",
-                img->TexFormat->MesaFormat, __FUNCTION__);
-      }
+      _mesa_sprintf(s, "/tmp/teximage%u.ppm", texObj->Name);
+
+      _mesa_printf("  Writing image level %u to %s\n", level, s);
+      write_ppm(s, buffer, img->Width, img->Height, 4, 0, 1, 2, GL_FALSE);
+
+      ctx->Pack = store; /* restore */
+
+      _mesa_free(buffer);
    }
 }
 
@@ -313,17 +305,21 @@ dump_texture_cb(GLuint id, void *data, void *userData)
 {
    struct gl_texture_object *texObj = (struct gl_texture_object *) data;
    int i;
+   GLboolean written = GL_FALSE;
    (void) userData;
 
-   printf("Texture %u\n", texObj->Name);
-   printf("  Target 0x%x\n", texObj->Target);
+   _mesa_printf("Texture %u\n", texObj->Name);
+   _mesa_printf("  Target 0x%x\n", texObj->Target);
    for (i = 0; i < MAX_TEXTURE_LEVELS; i++) {
       struct gl_texture_image *texImg = texObj->Image[0][i];
       if (texImg) {
-         printf("  Image %u: %d x %d x %d at %p\n", i,
-                texImg->Width, texImg->Height, texImg->Depth, texImg->Data);
-         if (DumpImages && i == 0) {
-            write_texture_image(texObj);
+         _mesa_printf("  Image %u: %d x %d x %d, format %u at %p\n", i,
+                      texImg->Width, texImg->Height, texImg->Depth,
+                      texImg->TexFormat->MesaFormat, texImg->Data);
+         if (DumpImages && !written) {
+            GLuint face = 0;
+            write_texture_image(texObj, face, i);
+            written = GL_TRUE;
          }
       }
    }
@@ -341,3 +337,104 @@ _mesa_dump_textures(GLboolean dumpImages)
    DumpImages = dumpImages;
    _mesa_HashWalk(ctx->Shared->TexObjects, dump_texture_cb, ctx);
 }
+
+
+void
+_mesa_dump_color_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLubyte *buf;
+
+   buf = (GLubyte *) _mesa_malloc(w * h * 4);
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, buf);
+
+   _mesa_printf("ReadBuffer %p 0x%x  DrawBuffer %p 0x%x\n",
+                ctx->ReadBuffer->_ColorReadBuffer,
+                ctx->ReadBuffer->ColorReadBuffer,
+                ctx->DrawBuffer->_ColorDrawBuffers[0],
+                ctx->DrawBuffer->ColorDrawBuffer[0]);
+   _mesa_printf("Writing %d x %d color buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf, w, h, 4, 0, 1, 2, GL_TRUE);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+}
+
+
+void
+_mesa_dump_depth_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLuint *buf;
+   GLubyte *buf2;
+   GLuint i;
+
+   buf = (GLuint *) _mesa_malloc(w * h * 4);  /* 4 bpp */
+   buf2 = (GLubyte *) _mesa_malloc(w * h * 3); /* 3 bpp */
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, buf);
+
+   /* spread 24 bits of Z across R, G, B */
+   for (i = 0; i < w * h; i++) {
+      buf2[i*3+0] = (buf[i] >> 24) & 0xff;
+      buf2[i*3+1] = (buf[i] >> 16) & 0xff;
+      buf2[i*3+2] = (buf[i] >>  8) & 0xff;
+   }
+
+   _mesa_printf("Writing %d x %d depth buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf2, w, h, 3, 0, 1, 2, GL_TRUE);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+   _mesa_free(buf2);
+}
+
+
+void
+_mesa_dump_stencil_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLubyte *buf;
+   GLubyte *buf2;
+   GLuint i;
+
+   buf = (GLubyte *) _mesa_malloc(w * h);  /* 1 bpp */
+   buf2 = (GLubyte *) _mesa_malloc(w * h * 3); /* 3 bpp */
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_STENCIL_INDEX, GL_UNSIGNED_BYTE, buf);
+
+   for (i = 0; i < w * h; i++) {
+      buf2[i*3+0] = buf[i];
+      buf2[i*3+1] = (buf[i] & 127) * 2;
+      buf2[i*3+2] = (buf[i] - 128) * 2;
+   }
+
+   _mesa_printf("Writing %d x %d stencil buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf2, w, h, 3, 0, 1, 2, GL_TRUE);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+   _mesa_free(buf2);
+}
diff --git a/src/mesa/main/debug.h b/src/mesa/main/debug.h
index 1862ec75b7..bb384c4324 100644
--- a/src/mesa/main/debug.h
+++ b/src/mesa/main/debug.h
@@ -60,4 +60,13 @@ extern void _mesa_init_debug( GLcontext *ctx );
 extern void
 _mesa_dump_textures(GLboolean dumpImages);
 
+extern void
+_mesa_dump_color_buffer(const char *filename);
+
+extern void
+_mesa_dump_depth_buffer(const char *filename);
+
+extern void
+_mesa_dump_stencil_buffer(const char *filename);
+
 #endif
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index 74537d79e9..8cff9ea64a 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -352,6 +352,9 @@ typedef enum
    OPCODE_EVAL_P1,
    OPCODE_EVAL_P2,
 
+   /* GL_EXT_provoking_vertex */
+   OPCODE_PROVOKING_VERTEX,
+
    /* The following three are meta instructions */
    OPCODE_ERROR,                /* raise compiled-in error */
    OPCODE_CONTINUE,
@@ -688,7 +691,7 @@ unpack_image(GLcontext *ctx, GLuint dimensions,
              GLenum format, GLenum type, const GLvoid * pixels,
              const struct gl_pixelstore_attrib *unpack)
 {
-   if (unpack->BufferObj->Name == 0) {
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
       /* no PBO */
       GLvoid *image = _mesa_unpack_image(dimensions, width, height, depth,
                                          format, type, pixels, unpack);
@@ -4937,7 +4940,7 @@ save_Attr1fNV(GLenum attr, GLfloat x)
       n[2].f = x;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 1;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, 0, 0, 1);
 
@@ -4959,7 +4962,7 @@ save_Attr2fNV(GLenum attr, GLfloat x, GLfloat y)
       n[3].f = y;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 2;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, 0, 1);
 
@@ -4982,7 +4985,7 @@ save_Attr3fNV(GLenum attr, GLfloat x, GLfloat y, GLfloat z)
       n[4].f = z;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 3;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, z, 1);
 
@@ -5006,7 +5009,7 @@ save_Attr4fNV(GLenum attr, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
       n[5].f = w;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 4;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, z, w);
 
@@ -5783,6 +5786,25 @@ save_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
 #endif
 
 
+/** GL_EXT_provoking_vertex */
+static void GLAPIENTRY
+save_ProvokingVertexEXT(GLenum mode)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   Node *n;
+   ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx);
+   n = ALLOC_INSTRUCTION(ctx, OPCODE_PROVOKING_VERTEX, 1);
+   if (n) {
+      n[1].e = mode;
+   }
+   if (ctx->ExecuteFlag) {
+      /*CALL_ProvokingVertexEXT(ctx->Exec, (mode));*/
+      _mesa_ProvokingVertexEXT(mode);
+   }
+}
+
+
+
 /**
  * Save an error-generating command into display list.
  *
@@ -6347,6 +6369,9 @@ execute_list(GLcontext *ctx, GLuint list)
          case OPCODE_SHADE_MODEL:
             CALL_ShadeModel(ctx->Exec, (n[1].e));
             break;
+         case OPCODE_PROVOKING_VERTEX:
+            CALL_ProvokingVertexEXT(ctx->Exec, (n[1].e));
+            break;
          case OPCODE_STENCIL_FUNC:
             CALL_StencilFunc(ctx->Exec, (n[1].e, n[2].i, n[3].ui));
             break;
@@ -7726,18 +7751,6 @@ exec_MultiDrawArraysEXT(GLenum mode, GLint * first,
    CALL_MultiDrawArraysEXT(ctx->Exec, (mode, first, count, primcount));
 }
 
-/* GL_EXT_multi_draw_arrays */
-static void GLAPIENTRY
-exec_MultiDrawElementsEXT(GLenum mode, const GLsizei * count,
-                          GLenum type, const GLvoid ** indices,
-                          GLsizei primcount)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FLUSH_VERTICES(ctx, 0);
-   CALL_MultiDrawElementsEXT(ctx->Exec,
-                             (mode, count, type, indices, primcount));
-}
-
 /* GL_IBM_multimode_draw_arrays */
 static void GLAPIENTRY
 exec_MultiModeDrawArraysIBM(const GLenum * mode, const GLint * first,
@@ -8083,7 +8096,6 @@ _mesa_init_dlist_table(struct _glapi_table *table)
 
    /* 148. GL_EXT_multi_draw_arrays */
    SET_MultiDrawArraysEXT(table, exec_MultiDrawArraysEXT);
-   SET_MultiDrawElementsEXT(table, exec_MultiDrawElementsEXT);
 
    /* 149. GL_EXT_fog_coord */
    SET_FogCoordPointerEXT(table, exec_FogCoordPointerEXT);
@@ -8310,6 +8322,18 @@ _mesa_init_dlist_table(struct _glapi_table *table)
    SET_ProgramEnvParameters4fvEXT(table, save_ProgramEnvParameters4fvEXT);
    SET_ProgramLocalParameters4fvEXT(table, save_ProgramLocalParameters4fvEXT);
 #endif
+
+   /* ARB 50. GL_ARB_map_buffer_range */
+#if FEATURE_ARB_map_buffer_range
+   SET_MapBufferRange(table, _mesa_MapBufferRange); /* no dlist save */
+   SET_FlushMappedBufferRange(table, _mesa_FlushMappedBufferRange); /* no dl */
+#endif
+
+   /* ARB 59. GL_ARB_copy_buffer */
+   SET_CopyBufferSubData(table, _mesa_CopyBufferSubData); /* no dlist save */
+
+   /* 364. GL_EXT_provoking_vertex */
+   SET_ProvokingVertexEXT(table, save_ProvokingVertexEXT);
 }
 
 
@@ -8546,6 +8570,11 @@ print_list(GLcontext *ctx, GLuint list)
             _mesa_printf("EVAL_P2 %d %d\n", n[1].i, n[2].i);
             break;
 
+         case OPCODE_PROVOKING_VERTEX:
+            _mesa_printf("ProvokingVertex %s\n",
+                         _mesa_lookup_enum_by_nr(n[1].ui));
+            break;
+
             /*
              * meta opcodes/commands
              */
@@ -8681,6 +8710,7 @@ _mesa_save_vtxfmt_init(GLvertexformat * vfmt)
    vfmt->DrawArrays = 0;
    vfmt->DrawElements = 0;
    vfmt->DrawRangeElements = 0;
+   vfmt->MultiDrawElemementsEXT = 0;
 #endif
 }
 
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index 6682b5e725..aef6585641 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -27,6 +27,7 @@
 #include "bufferobj.h"
 #include "context.h"
 #include "drawpix.h"
+#include "enums.h"
 #include "feedback.h"
 #include "framebuffer.h"
 #include "image.h"
@@ -34,6 +35,18 @@
 #include "state.h"
 
 
+
+/**
+ * If a fragment program is enabled, check that it's valid.
+ * \return GL_TRUE if valid, GL_FALSE otherwise
+ */
+static GLboolean
+valid_fragment_program(GLcontext *ctx)
+{
+   return !(ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled);
+}
+
+
 #if _HAVE_FULL_GL
 
 /*
@@ -51,29 +64,34 @@ _mesa_DrawPixels( GLsizei width, GLsizei height,
       return;
    }
 
+   /* We're not using the current vertex program, and the driver may install
+    * it's own.
+    */
+   _mesa_set_vp_override(ctx, GL_TRUE);
+
    if (ctx->NewState) {
       _mesa_update_state(ctx);
    }
 
-   if (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled) {
+   if (!valid_fragment_program(ctx)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glDrawPixels (invalid fragment program)");
-      return;
+      goto end;
    }
 
    if (_mesa_error_check_format_type(ctx, format, type, GL_TRUE)) {
-      /* found an error */
-      return;
+      /* the error was already recorded */
+      goto end;
    }
 
    if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
       _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
                   "glDrawPixels(incomplete framebuffer)" );
-      return;
+      goto end;
    }
 
    if (!ctx->Current.RasterPosValid) {
-      return;
+      goto end; /* no-op, not an error */
    }
 
    if (ctx->RenderMode == GL_RENDER) {
@@ -88,13 +106,13 @@ _mesa_DrawPixels( GLsizei width, GLsizei height,
                                            format, type, pixels)) {
                _mesa_error(ctx, GL_INVALID_OPERATION,
                            "glDrawPixels(invalid PBO access)");
-               return;
+               goto end;
             }
-            if (ctx->Unpack.BufferObj->Pointer) {
+            if (_mesa_bufferobj_mapped(ctx->Unpack.BufferObj)) {
                /* buffer is mapped - that's an error */
                _mesa_error(ctx, GL_INVALID_OPERATION,
                            "glDrawPixels(PBO is mapped)");
-               return;
+               goto end;
             }
          }
 
@@ -116,6 +134,9 @@ _mesa_DrawPixels( GLsizei width, GLsizei height,
       ASSERT(ctx->RenderMode == GL_SELECT);
       /* Do nothing.  See OpenGL Spec, Appendix B, Corollary 6. */
    }
+
+end:
+   _mesa_set_vp_override(ctx, GL_FALSE);
 }
 
 
@@ -126,37 +147,55 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
+   if (width < 0 || height < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE, "glCopyPixels(width or height < 0)");
+      return;
+   }
+
+   /* Note: more detailed 'type' checking is done by the
+    * _mesa_source/dest_buffer_exists() calls below.  That's where we
+    * check if the stencil buffer exists, etc.
+    */
+   if (type != GL_COLOR &&
+       type != GL_DEPTH &&
+       type != GL_STENCIL &&
+       type != GL_DEPTH_STENCIL) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glCopyPixels(type=%s)",
+                  _mesa_lookup_enum_by_nr(type));
+      return;
+   }
+
+   /* We're not using the current vertex program, and the driver may install
+    * it's own.
+    */
+   _mesa_set_vp_override(ctx, GL_TRUE);
+
    if (ctx->NewState) {
       _mesa_update_state(ctx);
    }
 
-   if (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled) {
+   if (!valid_fragment_program(ctx)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glCopyPixels (invalid fragment program)");
-      return;
-   }
-
-   if (width < 0 || height < 0) {
-      _mesa_error(ctx, GL_INVALID_VALUE, "glCopyPixels(width or height < 0)");
-      return;
+      goto end;
    }
 
    if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT ||
        ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
       _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
                   "glCopyPixels(incomplete framebuffer)" );
-      return;
+      goto end;
    }
 
    if (!_mesa_source_buffer_exists(ctx, type) ||
        !_mesa_dest_buffer_exists(ctx, type)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glCopyPixels(missing source or dest buffer)");
-      return;
+      goto end;
    }
 
    if (!ctx->Current.RasterPosValid || width ==0 || height == 0) {
-      return;
+      goto end; /* no-op, not an error */
    }
 
    if (ctx->RenderMode == GL_RENDER) {
@@ -181,6 +220,9 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height,
       ASSERT(ctx->RenderMode == GL_SELECT);
       /* Do nothing.  See OpenGL Spec, Appendix B, Corollary 6. */
    }
+
+end:
+   _mesa_set_vp_override(ctx, GL_FALSE);
 }
 
 #endif /* _HAVE_FULL_GL */
@@ -208,7 +250,7 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
       _mesa_update_state(ctx);
    }
 
-   if (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled) {
+   if (!valid_fragment_program(ctx)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glBitmap (invalid fragment program)");
       return;
@@ -222,27 +264,30 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
 
    if (ctx->RenderMode == GL_RENDER) {
       /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
-      const GLfloat epsilon = 0.0001F;
-      GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig);
-      GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig);
-
-      if (ctx->Unpack.BufferObj->Name) {
-         /* unpack from PBO */
-         if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1,
-                                        GL_COLOR_INDEX, GL_BITMAP,
-                                        (GLvoid *) bitmap)) {
-            _mesa_error(ctx, GL_INVALID_OPERATION,
-                        "glBitmap(invalid PBO access)");
-            return;
-         }
-         if (ctx->Unpack.BufferObj->Pointer) {
-            /* buffer is mapped - that's an error */
-            _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)");
-            return;
+      if (width > 0 && height > 0) {
+         const GLfloat epsilon = 0.0001F;
+         GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig);
+         GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig);
+
+         if (ctx->Unpack.BufferObj->Name) {
+            /* unpack from PBO */
+            if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1,
+                                           GL_COLOR_INDEX, GL_BITMAP,
+                                           (GLvoid *) bitmap)) {
+               _mesa_error(ctx, GL_INVALID_OPERATION,
+                           "glBitmap(invalid PBO access)");
+               return;
+            }
+            if (_mesa_bufferobj_mapped(ctx->Unpack.BufferObj)) {
+               /* buffer is mapped - that's an error */
+               _mesa_error(ctx, GL_INVALID_OPERATION,
+                           "glBitmap(PBO is mapped)");
+               return;
+            }
          }
-      }
 
-      ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
+         ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
+      }
    }
 #if _HAVE_FULL_GL
    else if (ctx->RenderMode == GL_FEEDBACK) {
@@ -264,68 +309,3 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
    ctx->Current.RasterPos[0] += xmove;
    ctx->Current.RasterPos[1] += ymove;
 }
-
-
-
-#if 0  /* experimental */
-/*
- * Execute glDrawDepthPixelsMESA().  This function accepts both a color
- * image and depth (Z) image.  Rasterization produces fragments with
- * color and Z taken from these images.  This function is intended for
- * Z-compositing.  Normally, this operation requires two glDrawPixels
- * calls with stencil testing.
- */
-void GLAPIENTRY
-_mesa_DrawDepthPixelsMESA( GLsizei width, GLsizei height,
-                           GLenum colorFormat, GLenum colorType,
-                           const GLvoid *colors,
-                           GLenum depthType, const GLvoid *depths )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
-
-   if (width < 0 || height < 0) {
-      _mesa_error( ctx, GL_INVALID_VALUE,
-                   "glDrawDepthPixelsMESA(width or height < 0" );
-      return;
-   }
-
-   if (!ctx->Current.RasterPosValid) {
-      return;
-   }
-
-   if (ctx->NewState) {
-      _mesa_update_state(ctx);
-   }
-
-   if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
-      _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
-                  "glDrawDepthPixelsMESA(incomplete framebuffer)");
-      return;
-   }
-
-   if (ctx->RenderMode == GL_RENDER) {
-      /* Round, to satisfy conformance tests (matches SGI's OpenGL) */
-      GLint x = IROUND(ctx->Current.RasterPos[0]);
-      GLint y = IROUND(ctx->Current.RasterPos[1]);
-      ctx->Driver.DrawDepthPixelsMESA(ctx, x, y, width, height,
-                                      colorFormat, colorType, colors,
-                                      depthType, depths, &ctx->Unpack);
-   }
-   else if (ctx->RenderMode == GL_FEEDBACK) {
-      /* Feedback the current raster pos info */
-      FLUSH_CURRENT( ctx, 0 );
-      _mesa_feedback_token( ctx, (GLfloat) (GLint) GL_DRAW_PIXEL_TOKEN );
-      _mesa_feedback_vertex( ctx,
-                             ctx->Current.RasterPos,
-                             ctx->Current.RasterColor,
-                             ctx->Current.RasterIndex,
-                             ctx->Current.RasterTexCoords[0] );
-   }
-   else {
-      ASSERT(ctx->RenderMode == GL_SELECT);
-      /* Do nothing.  See OpenGL Spec, Appendix B, Corollary 6. */
-   }
-}
-
-#endif
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index f432be183c..4bc54771e9 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -119,6 +119,7 @@ client_state(GLcontext *ctx, GLenum cap, GLboolean state)
          CHECK_EXTENSION(NV_vertex_program, cap);
          {
             GLint n = (GLint) cap - GL_VERTEX_ATTRIB_ARRAY0_NV;
+            ASSERT(n < Elements(ctx->Array.ArrayObj->VertexAttrib));
             var = &ctx->Array.ArrayObj->VertexAttrib[n].Enabled;
             flag = _NEW_ARRAY_ATTRIB(n);
          }
@@ -222,14 +223,16 @@ get_texcoord_unit(GLcontext *ctx)
 
 /**
  * Helper function to enable or disable a texture target.
+ * \param bit  one of the TEXTURE_x_BIT values
+ * \return GL_TRUE if state is changing or GL_FALSE if no change
  */
 static GLboolean
-enable_texture(GLcontext *ctx, GLboolean state, GLbitfield bit)
+enable_texture(GLcontext *ctx, GLboolean state, GLbitfield texBit)
 {
    const GLuint curr = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[curr];
-   const GLuint newenabled = (!state)
-       ? (texUnit->Enabled & ~bit) :  (texUnit->Enabled | bit);
+   const GLbitfield newenabled = state
+      ? (texUnit->Enabled | texBit) : (texUnit->Enabled & ~texBit);
 
    if (!ctx->DrawBuffer->Visual.rgbMode || texUnit->Enabled == newenabled)
        return GL_FALSE;
@@ -969,6 +972,11 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          }
          break;
 
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+	 CHECK_EXTENSION(ARB_seamless_cube_map, cap);
+	 ctx->Texture.CubeMapSeamless = state;
+	 break;
+
       default:
          _mesa_error(ctx, GL_INVALID_ENUM,
                      "%s(0x%x)", state ? "glEnable" : "glDisable", cap);
@@ -1314,6 +1322,7 @@ _mesa_IsEnabled( GLenum cap )
          CHECK_EXTENSION(NV_vertex_program);
          {
             GLint n = (GLint) cap - GL_VERTEX_ATTRIB_ARRAY0_NV;
+            ASSERT(n < Elements(ctx->Array.ArrayObj->VertexAttrib));
             return (ctx->Array.ArrayObj->VertexAttrib[n].Enabled != 0);
          }
       case GL_MAP1_VERTEX_ATTRIB0_4_NV:
@@ -1391,6 +1400,11 @@ _mesa_IsEnabled( GLenum cap )
 	 CHECK_EXTENSION(ATI_fragment_shader);
 	 return ctx->ATIFragmentShader.Enabled;
 #endif /* FEATURE_ATI_fragment_shader */
+
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+	 CHECK_EXTENSION(ARB_seamless_cube_map);
+	 return ctx->Texture.CubeMapSeamless;
+
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glIsEnabled(0x%x)", (int) cap);
 	 return GL_FALSE;
diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c
index dff5544086..9f650dadd3 100644
--- a/src/mesa/main/enums.c
+++ b/src/mesa/main/enums.c
@@ -82,6 +82,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_ALPHA_TEST\0"
    "GL_ALPHA_TEST_FUNC\0"
    "GL_ALPHA_TEST_REF\0"
+   "GL_ALREADY_SIGNALED\0"
    "GL_ALWAYS\0"
    "GL_AMBIENT\0"
    "GL_AMBIENT_AND_DIFFUSE\0"
@@ -139,10 +140,12 @@ LONGSTRING static const char enum_string_table[] =
    "GL_BOOL_VEC4_ARB\0"
    "GL_BUFFER_ACCESS\0"
    "GL_BUFFER_ACCESS_ARB\0"
+   "GL_BUFFER_FLUSHING_UNMAP_APPLE\0"
    "GL_BUFFER_MAPPED\0"
    "GL_BUFFER_MAPPED_ARB\0"
    "GL_BUFFER_MAP_POINTER\0"
    "GL_BUFFER_MAP_POINTER_ARB\0"
+   "GL_BUFFER_SERIALIZED_MODIFY_APPLE\0"
    "GL_BUFFER_SIZE\0"
    "GL_BUFFER_SIZE_ARB\0"
    "GL_BUFFER_USAGE\0"
@@ -303,6 +306,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_COMPRESSED_SRGB\0"
    "GL_COMPRESSED_SRGB_ALPHA\0"
    "GL_COMPRESSED_TEXTURE_FORMATS\0"
+   "GL_CONDITION_SATISFIED\0"
    "GL_CONSTANT\0"
    "GL_CONSTANT_ALPHA\0"
    "GL_CONSTANT_ALPHA_EXT\0"
@@ -334,6 +338,8 @@ LONGSTRING static const char enum_string_table[] =
    "GL_COPY\0"
    "GL_COPY_INVERTED\0"
    "GL_COPY_PIXEL_TOKEN\0"
+   "GL_COPY_READ_BUFFER\0"
+   "GL_COPY_WRITE_BUFFER\0"
    "GL_CULL_FACE\0"
    "GL_CULL_FACE_MODE\0"
    "GL_CULL_VERTEX_EXT\0"
@@ -512,6 +518,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_FEEDBACK_BUFFER_SIZE\0"
    "GL_FEEDBACK_BUFFER_TYPE\0"
    "GL_FILL\0"
+   "GL_FIRST_VERTEX_CONVENTION_EXT\0"
    "GL_FLAT\0"
    "GL_FLOAT\0"
    "GL_FLOAT_MAT2\0"
@@ -694,6 +701,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_INVERSE_TRANSPOSE_NV\0"
    "GL_INVERT\0"
    "GL_KEEP\0"
+   "GL_LAST_VERTEX_CONVENTION_EXT\0"
    "GL_LEFT\0"
    "GL_LEQUAL\0"
    "GL_LESS\0"
@@ -819,7 +827,13 @@ LONGSTRING static const char enum_string_table[] =
    "GL_MAP2_VERTEX_ATTRIB8_4_NV\0"
    "GL_MAP2_VERTEX_ATTRIB9_4_NV\0"
    "GL_MAP_COLOR\0"
+   "GL_MAP_FLUSH_EXPLICIT_BIT\0"
+   "GL_MAP_INVALIDATE_BUFFER_BIT\0"
+   "GL_MAP_INVALIDATE_RANGE_BIT\0"
+   "GL_MAP_READ_BIT\0"
    "GL_MAP_STENCIL\0"
+   "GL_MAP_UNSYNCHRONIZED_BIT\0"
+   "GL_MAP_WRITE_BIT\0"
    "GL_MATRIX0_ARB\0"
    "GL_MATRIX0_NV\0"
    "GL_MATRIX10_ARB\0"
@@ -932,6 +946,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_MAX_RECTANGLE_TEXTURE_SIZE_NV\0"
    "GL_MAX_RENDERBUFFER_SIZE_EXT\0"
    "GL_MAX_SAMPLES\0"
+   "GL_MAX_SERVER_WAIT_TIMEOUT\0"
    "GL_MAX_SHININESS_NV\0"
    "GL_MAX_SPOT_EXPONENT_NV\0"
    "GL_MAX_TEXTURE_COORDS\0"
@@ -1060,6 +1075,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_OBJECT_PLANE\0"
    "GL_OBJECT_SHADER_SOURCE_LENGTH_ARB\0"
    "GL_OBJECT_SUBTYPE_ARB\0"
+   "GL_OBJECT_TYPE\0"
    "GL_OBJECT_TYPE_ARB\0"
    "GL_OBJECT_VALIDATE_STATUS_ARB\0"
    "GL_OCCLUSION_TEST_HP\0"
@@ -1271,6 +1287,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_PROJECTION\0"
    "GL_PROJECTION_MATRIX\0"
    "GL_PROJECTION_STACK_DEPTH\0"
+   "GL_PROVOKING_VERTEX_EXT\0"
    "GL_PROXY_COLOR_TABLE\0"
    "GL_PROXY_HISTOGRAM\0"
    "GL_PROXY_HISTOGRAM_EXT\0"
@@ -1291,6 +1308,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_Q\0"
    "GL_QUADRATIC_ATTENUATION\0"
    "GL_QUADS\0"
+   "GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT\0"
    "GL_QUAD_MESH_SUN\0"
    "GL_QUAD_STRIP\0"
    "GL_QUERY_COUNTER_BITS\0"
@@ -1438,6 +1456,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_SHARED_TEXTURE_PALETTE_EXT\0"
    "GL_SHININESS\0"
    "GL_SHORT\0"
+   "GL_SIGNALED\0"
    "GL_SIGNED_NORMALIZED\0"
    "GL_SINGLE_COLOR\0"
    "GL_SINGLE_COLOR_EXT\0"
@@ -1530,6 +1549,9 @@ LONGSTRING static const char enum_string_table[] =
    "GL_STENCIL_VALUE_MASK\0"
    "GL_STENCIL_WRITEMASK\0"
    "GL_STEREO\0"
+   "GL_STORAGE_CACHED_APPLE\0"
+   "GL_STORAGE_PRIVATE_APPLE\0"
+   "GL_STORAGE_SHARED_APPLE\0"
    "GL_STREAM_COPY\0"
    "GL_STREAM_COPY_ARB\0"
    "GL_STREAM_DRAW\0"
@@ -1539,6 +1561,12 @@ LONGSTRING static const char enum_string_table[] =
    "GL_SUBPIXEL_BITS\0"
    "GL_SUBTRACT\0"
    "GL_SUBTRACT_ARB\0"
+   "GL_SYNC_CONDITION\0"
+   "GL_SYNC_FENCE\0"
+   "GL_SYNC_FLAGS\0"
+   "GL_SYNC_FLUSH_COMMANDS_BIT\0"
+   "GL_SYNC_GPU_COMMANDS_COMPLETE\0"
+   "GL_SYNC_STATUS\0"
    "GL_T\0"
    "GL_T2F_C3F_V3F\0"
    "GL_T2F_C4F_N3F_V3F\0"
@@ -1679,6 +1707,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB\0"
    "GL_TEXTURE_CUBE_MAP_POSITIVE_Z\0"
    "GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB\0"
+   "GL_TEXTURE_CUBE_MAP_SEAMLESS\0"
    "GL_TEXTURE_DEPTH\0"
    "GL_TEXTURE_DEPTH_SIZE\0"
    "GL_TEXTURE_DEPTH_SIZE_ARB\0"
@@ -1718,6 +1747,8 @@ LONGSTRING static const char enum_string_table[] =
    "GL_TEXTURE_MIN_FILTER\0"
    "GL_TEXTURE_MIN_LOD\0"
    "GL_TEXTURE_PRIORITY\0"
+   "GL_TEXTURE_RANGE_LENGTH_APPLE\0"
+   "GL_TEXTURE_RANGE_POINTER_APPLE\0"
    "GL_TEXTURE_RECTANGLE_ARB\0"
    "GL_TEXTURE_RECTANGLE_NV\0"
    "GL_TEXTURE_RED_SIZE\0"
@@ -1725,12 +1756,15 @@ LONGSTRING static const char enum_string_table[] =
    "GL_TEXTURE_RESIDENT\0"
    "GL_TEXTURE_STACK_DEPTH\0"
    "GL_TEXTURE_STENCIL_SIZE\0"
+   "GL_TEXTURE_STORAGE_HINT_APPLE\0"
    "GL_TEXTURE_TOO_LARGE_EXT\0"
    "GL_TEXTURE_UNSIGNED_REMAP_MODE_NV\0"
    "GL_TEXTURE_WIDTH\0"
    "GL_TEXTURE_WRAP_R\0"
    "GL_TEXTURE_WRAP_S\0"
    "GL_TEXTURE_WRAP_T\0"
+   "GL_TIMEOUT_EXPIRED\0"
+   "GL_TIMEOUT_IGNORED\0"
    "GL_TIME_ELAPSED_EXT\0"
    "GL_TRACK_MATRIX_NV\0"
    "GL_TRACK_MATRIX_TRANSFORM_NV\0"
@@ -1758,6 +1792,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_UNPACK_SKIP_PIXELS\0"
    "GL_UNPACK_SKIP_ROWS\0"
    "GL_UNPACK_SWAP_BYTES\0"
+   "GL_UNSIGNALED\0"
    "GL_UNSIGNED_BYTE\0"
    "GL_UNSIGNED_BYTE_2_3_3_REV\0"
    "GL_UNSIGNED_BYTE_3_3_2\0"
@@ -1787,6 +1822,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_VENDOR\0"
    "GL_VERSION\0"
    "GL_VERTEX_ARRAY\0"
+   "GL_VERTEX_ARRAY_BINDING\0"
    "GL_VERTEX_ARRAY_BINDING_APPLE\0"
    "GL_VERTEX_ARRAY_BUFFER_BINDING\0"
    "GL_VERTEX_ARRAY_BUFFER_BINDING_ARB\0"
@@ -1839,6 +1875,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_VERTEX_STATE_PROGRAM_NV\0"
    "GL_VIEWPORT\0"
    "GL_VIEWPORT_BIT\0"
+   "GL_WAIT_FAILED\0"
    "GL_WEIGHT_ARRAY_ARB\0"
    "GL_WEIGHT_ARRAY_BUFFER_BINDING\0"
    "GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB\0"
@@ -1858,7 +1895,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_ZOOM_Y\0"
    ;
 
-static const enum_elt all_enums[1820] =
+static const enum_elt all_enums[1857] =
 {
    {     0, 0x00000600 }, /* GL_2D */
    {     6, 0x00001407 }, /* GL_2_BYTES */
@@ -1906,3103 +1943,3168 @@ static const enum_elt all_enums[1820] =
    {   728, 0x00000BC0 }, /* GL_ALPHA_TEST */
    {   742, 0x00000BC1 }, /* GL_ALPHA_TEST_FUNC */
    {   761, 0x00000BC2 }, /* GL_ALPHA_TEST_REF */
-   {   779, 0x00000207 }, /* GL_ALWAYS */
-   {   789, 0x00001200 }, /* GL_AMBIENT */
-   {   800, 0x00001602 }, /* GL_AMBIENT_AND_DIFFUSE */
-   {   823, 0x00001501 }, /* GL_AND */
-   {   830, 0x00001504 }, /* GL_AND_INVERTED */
-   {   846, 0x00001502 }, /* GL_AND_REVERSE */
-   {   861, 0x00008892 }, /* GL_ARRAY_BUFFER */
-   {   877, 0x00008894 }, /* GL_ARRAY_BUFFER_BINDING */
-   {   901, 0x00008894 }, /* GL_ARRAY_BUFFER_BINDING_ARB */
-   {   929, 0x00008B85 }, /* GL_ATTACHED_SHADERS */
-   {   949, 0x00008645 }, /* GL_ATTRIB_ARRAY_POINTER_NV */
-   {   976, 0x00008623 }, /* GL_ATTRIB_ARRAY_SIZE_NV */
-   {  1000, 0x00008624 }, /* GL_ATTRIB_ARRAY_STRIDE_NV */
-   {  1026, 0x00008625 }, /* GL_ATTRIB_ARRAY_TYPE_NV */
-   {  1050, 0x00000BB0 }, /* GL_ATTRIB_STACK_DEPTH */
-   {  1072, 0x00000D80 }, /* GL_AUTO_NORMAL */
-   {  1087, 0x00000409 }, /* GL_AUX0 */
-   {  1095, 0x0000040A }, /* GL_AUX1 */
-   {  1103, 0x0000040B }, /* GL_AUX2 */
-   {  1111, 0x0000040C }, /* GL_AUX3 */
-   {  1119, 0x00000C00 }, /* GL_AUX_BUFFERS */
-   {  1134, 0x00000405 }, /* GL_BACK */
-   {  1142, 0x00000402 }, /* GL_BACK_LEFT */
-   {  1155, 0x00000403 }, /* GL_BACK_RIGHT */
-   {  1169, 0x000080E0 }, /* GL_BGR */
-   {  1176, 0x000080E1 }, /* GL_BGRA */
-   {  1184, 0x00001A00 }, /* GL_BITMAP */
-   {  1194, 0x00000704 }, /* GL_BITMAP_TOKEN */
-   {  1210, 0x00000BE2 }, /* GL_BLEND */
-   {  1219, 0x00008005 }, /* GL_BLEND_COLOR */
-   {  1234, 0x00008005 }, /* GL_BLEND_COLOR_EXT */
-   {  1253, 0x00000BE0 }, /* GL_BLEND_DST */
-   {  1266, 0x000080CA }, /* GL_BLEND_DST_ALPHA */
-   {  1285, 0x000080C8 }, /* GL_BLEND_DST_RGB */
-   {  1302, 0x00008009 }, /* GL_BLEND_EQUATION */
-   {  1320, 0x0000883D }, /* GL_BLEND_EQUATION_ALPHA */
-   {  1344, 0x0000883D }, /* GL_BLEND_EQUATION_ALPHA_EXT */
-   {  1372, 0x00008009 }, /* GL_BLEND_EQUATION_EXT */
-   {  1394, 0x00008009 }, /* GL_BLEND_EQUATION_RGB */
-   {  1416, 0x00008009 }, /* GL_BLEND_EQUATION_RGB_EXT */
-   {  1442, 0x00000BE1 }, /* GL_BLEND_SRC */
-   {  1455, 0x000080CB }, /* GL_BLEND_SRC_ALPHA */
-   {  1474, 0x000080C9 }, /* GL_BLEND_SRC_RGB */
-   {  1491, 0x00001905 }, /* GL_BLUE */
-   {  1499, 0x00000D1B }, /* GL_BLUE_BIAS */
-   {  1512, 0x00000D54 }, /* GL_BLUE_BITS */
-   {  1525, 0x00000D1A }, /* GL_BLUE_SCALE */
-   {  1539, 0x00008B56 }, /* GL_BOOL */
-   {  1547, 0x00008B56 }, /* GL_BOOL_ARB */
-   {  1559, 0x00008B57 }, /* GL_BOOL_VEC2 */
-   {  1572, 0x00008B57 }, /* GL_BOOL_VEC2_ARB */
-   {  1589, 0x00008B58 }, /* GL_BOOL_VEC3 */
-   {  1602, 0x00008B58 }, /* GL_BOOL_VEC3_ARB */
-   {  1619, 0x00008B59 }, /* GL_BOOL_VEC4 */
-   {  1632, 0x00008B59 }, /* GL_BOOL_VEC4_ARB */
-   {  1649, 0x000088BB }, /* GL_BUFFER_ACCESS */
-   {  1666, 0x000088BB }, /* GL_BUFFER_ACCESS_ARB */
-   {  1687, 0x000088BC }, /* GL_BUFFER_MAPPED */
-   {  1704, 0x000088BC }, /* GL_BUFFER_MAPPED_ARB */
-   {  1725, 0x000088BD }, /* GL_BUFFER_MAP_POINTER */
-   {  1747, 0x000088BD }, /* GL_BUFFER_MAP_POINTER_ARB */
-   {  1773, 0x00008764 }, /* GL_BUFFER_SIZE */
-   {  1788, 0x00008764 }, /* GL_BUFFER_SIZE_ARB */
-   {  1807, 0x00008765 }, /* GL_BUFFER_USAGE */
-   {  1823, 0x00008765 }, /* GL_BUFFER_USAGE_ARB */
-   {  1843, 0x0000877B }, /* GL_BUMP_ENVMAP_ATI */
-   {  1862, 0x00008777 }, /* GL_BUMP_NUM_TEX_UNITS_ATI */
-   {  1888, 0x00008775 }, /* GL_BUMP_ROT_MATRIX_ATI */
-   {  1911, 0x00008776 }, /* GL_BUMP_ROT_MATRIX_SIZE_ATI */
-   {  1939, 0x0000877C }, /* GL_BUMP_TARGET_ATI */
-   {  1958, 0x00008778 }, /* GL_BUMP_TEX_UNITS_ATI */
-   {  1980, 0x00001400 }, /* GL_BYTE */
-   {  1988, 0x00002A24 }, /* GL_C3F_V3F */
-   {  1999, 0x00002A26 }, /* GL_C4F_N3F_V3F */
-   {  2014, 0x00002A22 }, /* GL_C4UB_V2F */
-   {  2026, 0x00002A23 }, /* GL_C4UB_V3F */
-   {  2038, 0x00000901 }, /* GL_CCW */
-   {  2045, 0x00002900 }, /* GL_CLAMP */
-   {  2054, 0x0000812D }, /* GL_CLAMP_TO_BORDER */
-   {  2073, 0x0000812D }, /* GL_CLAMP_TO_BORDER_ARB */
-   {  2096, 0x0000812D }, /* GL_CLAMP_TO_BORDER_SGIS */
-   {  2120, 0x0000812F }, /* GL_CLAMP_TO_EDGE */
-   {  2137, 0x0000812F }, /* GL_CLAMP_TO_EDGE_SGIS */
-   {  2159, 0x00001500 }, /* GL_CLEAR */
-   {  2168, 0x000084E1 }, /* GL_CLIENT_ACTIVE_TEXTURE */
-   {  2193, 0x000084E1 }, /* GL_CLIENT_ACTIVE_TEXTURE_ARB */
-   {  2222, 0xFFFFFFFF }, /* GL_CLIENT_ALL_ATTRIB_BITS */
-   {  2248, 0x00000BB1 }, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
-   {  2277, 0x00000001 }, /* GL_CLIENT_PIXEL_STORE_BIT */
-   {  2303, 0x00000002 }, /* GL_CLIENT_VERTEX_ARRAY_BIT */
-   {  2330, 0x00003000 }, /* GL_CLIP_PLANE0 */
-   {  2345, 0x00003001 }, /* GL_CLIP_PLANE1 */
-   {  2360, 0x00003002 }, /* GL_CLIP_PLANE2 */
-   {  2375, 0x00003003 }, /* GL_CLIP_PLANE3 */
-   {  2390, 0x00003004 }, /* GL_CLIP_PLANE4 */
-   {  2405, 0x00003005 }, /* GL_CLIP_PLANE5 */
-   {  2420, 0x000080F0 }, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
-   {  2453, 0x00000A00 }, /* GL_COEFF */
-   {  2462, 0x00001800 }, /* GL_COLOR */
-   {  2471, 0x00008076 }, /* GL_COLOR_ARRAY */
-   {  2486, 0x00008898 }, /* GL_COLOR_ARRAY_BUFFER_BINDING */
-   {  2516, 0x00008898 }, /* GL_COLOR_ARRAY_BUFFER_BINDING_ARB */
-   {  2550, 0x00008090 }, /* GL_COLOR_ARRAY_POINTER */
-   {  2573, 0x00008081 }, /* GL_COLOR_ARRAY_SIZE */
-   {  2593, 0x00008083 }, /* GL_COLOR_ARRAY_STRIDE */
-   {  2615, 0x00008082 }, /* GL_COLOR_ARRAY_TYPE */
-   {  2635, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0 */
-   {  2656, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0_EXT */
-   {  2681, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1 */
-   {  2702, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10 */
-   {  2724, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10_EXT */
-   {  2750, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11 */
-   {  2772, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11_EXT */
-   {  2798, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12 */
-   {  2820, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12_EXT */
-   {  2846, 0x00008CED }, /* GL_COLOR_ATTACHMENT13 */
-   {  2868, 0x00008CED }, /* GL_COLOR_ATTACHMENT13_EXT */
-   {  2894, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14 */
-   {  2916, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14_EXT */
-   {  2942, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15 */
-   {  2964, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15_EXT */
-   {  2990, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1_EXT */
-   {  3015, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2 */
-   {  3036, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2_EXT */
-   {  3061, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3 */
-   {  3082, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3_EXT */
-   {  3107, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4 */
-   {  3128, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4_EXT */
-   {  3153, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5 */
-   {  3174, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5_EXT */
-   {  3199, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6 */
-   {  3220, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6_EXT */
-   {  3245, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7 */
-   {  3266, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7_EXT */
-   {  3291, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8 */
-   {  3312, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8_EXT */
-   {  3337, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9 */
-   {  3358, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9_EXT */
-   {  3383, 0x00004000 }, /* GL_COLOR_BUFFER_BIT */
-   {  3403, 0x00000C22 }, /* GL_COLOR_CLEAR_VALUE */
-   {  3424, 0x00001900 }, /* GL_COLOR_INDEX */
-   {  3439, 0x00001603 }, /* GL_COLOR_INDEXES */
-   {  3456, 0x00000BF2 }, /* GL_COLOR_LOGIC_OP */
-   {  3474, 0x00000B57 }, /* GL_COLOR_MATERIAL */
-   {  3492, 0x00000B55 }, /* GL_COLOR_MATERIAL_FACE */
-   {  3515, 0x00000B56 }, /* GL_COLOR_MATERIAL_PARAMETER */
-   {  3543, 0x000080B1 }, /* GL_COLOR_MATRIX */
-   {  3559, 0x000080B1 }, /* GL_COLOR_MATRIX_SGI */
-   {  3579, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH */
-   {  3607, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH_SGI */
-   {  3639, 0x00008458 }, /* GL_COLOR_SUM */
-   {  3652, 0x00008458 }, /* GL_COLOR_SUM_ARB */
-   {  3669, 0x000080D0 }, /* GL_COLOR_TABLE */
-   {  3684, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE */
-   {  3710, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_EXT */
-   {  3740, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_SGI */
-   {  3770, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS */
-   {  3790, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS_SGI */
-   {  3814, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE */
-   {  3839, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_EXT */
-   {  3868, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_SGI */
-   {  3897, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT */
-   {  3919, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_EXT */
-   {  3945, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_SGI */
-   {  3971, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE */
-   {  3997, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_EXT */
-   {  4027, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_SGI */
-   {  4057, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE */
-   {  4087, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_EXT */
-   {  4121, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_SGI */
-   {  4155, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
-   {  4185, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_EXT */
-   {  4219, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_SGI */
-   {  4253, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE */
-   {  4277, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_EXT */
-   {  4305, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_SGI */
-   {  4333, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE */
-   {  4354, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE_SGI */
-   {  4379, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH */
-   {  4400, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_EXT */
-   {  4425, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_SGI */
-   {  4450, 0x00000C23 }, /* GL_COLOR_WRITEMASK */
-   {  4469, 0x00008570 }, /* GL_COMBINE */
-   {  4480, 0x00008503 }, /* GL_COMBINE4 */
-   {  4492, 0x00008572 }, /* GL_COMBINE_ALPHA */
-   {  4509, 0x00008572 }, /* GL_COMBINE_ALPHA_ARB */
-   {  4530, 0x00008572 }, /* GL_COMBINE_ALPHA_EXT */
-   {  4551, 0x00008570 }, /* GL_COMBINE_ARB */
-   {  4566, 0x00008570 }, /* GL_COMBINE_EXT */
-   {  4581, 0x00008571 }, /* GL_COMBINE_RGB */
-   {  4596, 0x00008571 }, /* GL_COMBINE_RGB_ARB */
-   {  4615, 0x00008571 }, /* GL_COMBINE_RGB_EXT */
-   {  4634, 0x0000884E }, /* GL_COMPARE_REF_DEPTH_TO_TEXTURE_EXT */
-   {  4670, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE */
-   {  4694, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE_ARB */
-   {  4722, 0x00001300 }, /* GL_COMPILE */
-   {  4733, 0x00001301 }, /* GL_COMPILE_AND_EXECUTE */
-   {  4756, 0x00008B81 }, /* GL_COMPILE_STATUS */
-   {  4774, 0x000084E9 }, /* GL_COMPRESSED_ALPHA */
-   {  4794, 0x000084E9 }, /* GL_COMPRESSED_ALPHA_ARB */
-   {  4818, 0x000084EC }, /* GL_COMPRESSED_INTENSITY */
-   {  4842, 0x000084EC }, /* GL_COMPRESSED_INTENSITY_ARB */
-   {  4870, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE */
-   {  4894, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA */
-   {  4924, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA_ARB */
-   {  4958, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE_ARB */
-   {  4986, 0x000084ED }, /* GL_COMPRESSED_RGB */
-   {  5004, 0x000084EE }, /* GL_COMPRESSED_RGBA */
-   {  5023, 0x000084EE }, /* GL_COMPRESSED_RGBA_ARB */
-   {  5046, 0x000086B1 }, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
-   {  5075, 0x000083F1 }, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
-   {  5108, 0x000083F2 }, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
-   {  5141, 0x000083F3 }, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
-   {  5174, 0x000084ED }, /* GL_COMPRESSED_RGB_ARB */
-   {  5196, 0x000086B0 }, /* GL_COMPRESSED_RGB_FXT1_3DFX */
-   {  5224, 0x000083F0 }, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
-   {  5256, 0x00008C4A }, /* GL_COMPRESSED_SLUMINANCE */
-   {  5281, 0x00008C4B }, /* GL_COMPRESSED_SLUMINANCE_ALPHA */
-   {  5312, 0x00008C48 }, /* GL_COMPRESSED_SRGB */
-   {  5331, 0x00008C49 }, /* GL_COMPRESSED_SRGB_ALPHA */
-   {  5356, 0x000086A3 }, /* GL_COMPRESSED_TEXTURE_FORMATS */
-   {  5386, 0x00008576 }, /* GL_CONSTANT */
-   {  5398, 0x00008003 }, /* GL_CONSTANT_ALPHA */
-   {  5416, 0x00008003 }, /* GL_CONSTANT_ALPHA_EXT */
-   {  5438, 0x00008576 }, /* GL_CONSTANT_ARB */
-   {  5454, 0x00001207 }, /* GL_CONSTANT_ATTENUATION */
-   {  5478, 0x00008151 }, /* GL_CONSTANT_BORDER_HP */
-   {  5500, 0x00008001 }, /* GL_CONSTANT_COLOR */
-   {  5518, 0x00008001 }, /* GL_CONSTANT_COLOR_EXT */
-   {  5540, 0x00008576 }, /* GL_CONSTANT_EXT */
-   {  5556, 0x00008010 }, /* GL_CONVOLUTION_1D */
-   {  5574, 0x00008011 }, /* GL_CONVOLUTION_2D */
-   {  5592, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR */
-   {  5620, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR_HP */
-   {  5651, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE */
-   {  5678, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE_EXT */
-   {  5709, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS */
-   {  5736, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS_EXT */
-   {  5767, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE */
-   {  5795, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE_EXT */
-   {  5827, 0x00008017 }, /* GL_CONVOLUTION_FORMAT */
-   {  5849, 0x00008017 }, /* GL_CONVOLUTION_FORMAT_EXT */
-   {  5875, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT */
-   {  5897, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT_EXT */
-   {  5923, 0x00008018 }, /* GL_CONVOLUTION_WIDTH */
-   {  5944, 0x00008018 }, /* GL_CONVOLUTION_WIDTH_EXT */
-   {  5969, 0x00008862 }, /* GL_COORD_REPLACE */
-   {  5986, 0x00008862 }, /* GL_COORD_REPLACE_ARB */
-   {  6007, 0x00008862 }, /* GL_COORD_REPLACE_NV */
-   {  6027, 0x00001503 }, /* GL_COPY */
-   {  6035, 0x0000150C }, /* GL_COPY_INVERTED */
-   {  6052, 0x00000706 }, /* GL_COPY_PIXEL_TOKEN */
-   {  6072, 0x00000B44 }, /* GL_CULL_FACE */
-   {  6085, 0x00000B45 }, /* GL_CULL_FACE_MODE */
-   {  6103, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
-   {  6122, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-   {  6154, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-   {  6189, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
-   {  6210, 0x00000001 }, /* GL_CURRENT_BIT */
-   {  6225, 0x00000B00 }, /* GL_CURRENT_COLOR */
-   {  6242, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
-   {  6263, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
-   {  6289, 0x00000B01 }, /* GL_CURRENT_INDEX */
-   {  6306, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
-   {  6328, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
-   {  6356, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
-   {  6377, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-   {  6411, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
-   {  6444, 0x00000B02 }, /* GL_CURRENT_NORMAL */
-   {  6462, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-   {  6492, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
-   {  6511, 0x00008865 }, /* GL_CURRENT_QUERY */
-   {  6528, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
-   {  6549, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
-   {  6573, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
-   {  6600, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
-   {  6624, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
-   {  6651, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
-   {  6684, 0x0000845F }, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
-   {  6718, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-   {  6751, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
-   {  6778, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
-   {  6804, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
-   {  6829, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
-   {  6858, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
-   {  6880, 0x00000900 }, /* GL_CW */
-   {  6886, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
-   {  6907, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
-   {  6928, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
-   {  6948, 0x00002101 }, /* GL_DECAL */
-   {  6957, 0x00001E03 }, /* GL_DECR */
-   {  6965, 0x00008508 }, /* GL_DECR_WRAP */
-   {  6978, 0x00008508 }, /* GL_DECR_WRAP_EXT */
-   {  6995, 0x00008B80 }, /* GL_DELETE_STATUS */
-   {  7012, 0x00001801 }, /* GL_DEPTH */
-   {  7021, 0x000088F0 }, /* GL_DEPTH24_STENCIL8 */
-   {  7041, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT */
-   {  7061, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
-   {  7085, 0x00000D1F }, /* GL_DEPTH_BIAS */
-   {  7099, 0x00000D56 }, /* GL_DEPTH_BITS */
-   {  7113, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
-   {  7133, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
-   {  7158, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
-   {  7178, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
-   {  7196, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
-   {  7217, 0x00001902 }, /* GL_DEPTH_COMPONENT */
-   {  7236, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
-   {  7257, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
-   {  7282, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
-   {  7308, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
-   {  7329, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
-   {  7354, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
-   {  7380, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
-   {  7401, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
-   {  7426, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
-   {  7452, 0x00000B74 }, /* GL_DEPTH_FUNC */
-   {  7466, 0x00000B70 }, /* GL_DEPTH_RANGE */
-   {  7481, 0x00000D1E }, /* GL_DEPTH_SCALE */
-   {  7496, 0x000084F9 }, /* GL_DEPTH_STENCIL */
-   {  7513, 0x0000821A }, /* GL_DEPTH_STENCIL_ATTACHMENT */
-   {  7541, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
-   {  7561, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-   {  7589, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-   {  7617, 0x00000B71 }, /* GL_DEPTH_TEST */
-   {  7631, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
-   {  7653, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
-   {  7679, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
-   {  7698, 0x00001201 }, /* GL_DIFFUSE */
-   {  7709, 0x00000BD0 }, /* GL_DITHER */
-   {  7719, 0x00000A02 }, /* GL_DOMAIN */
-   {  7729, 0x00001100 }, /* GL_DONT_CARE */
-   {  7742, 0x000086AE }, /* GL_DOT3_RGB */
-   {  7754, 0x000086AF }, /* GL_DOT3_RGBA */
-   {  7767, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
-   {  7784, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
-   {  7801, 0x000086AE }, /* GL_DOT3_RGB_ARB */
-   {  7817, 0x00008740 }, /* GL_DOT3_RGB_EXT */
-   {  7833, 0x0000140A }, /* GL_DOUBLE */
-   {  7843, 0x00000C32 }, /* GL_DOUBLEBUFFER */
-   {  7859, 0x00000C01 }, /* GL_DRAW_BUFFER */
-   {  7874, 0x00008825 }, /* GL_DRAW_BUFFER0 */
-   {  7890, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
-   {  7910, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
-   {  7930, 0x00008826 }, /* GL_DRAW_BUFFER1 */
-   {  7946, 0x0000882F }, /* GL_DRAW_BUFFER10 */
-   {  7963, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
-   {  7984, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
-   {  8005, 0x00008830 }, /* GL_DRAW_BUFFER11 */
-   {  8022, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
-   {  8043, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
-   {  8064, 0x00008831 }, /* GL_DRAW_BUFFER12 */
-   {  8081, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
-   {  8102, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
-   {  8123, 0x00008832 }, /* GL_DRAW_BUFFER13 */
-   {  8140, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
-   {  8161, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
-   {  8182, 0x00008833 }, /* GL_DRAW_BUFFER14 */
-   {  8199, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
-   {  8220, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
-   {  8241, 0x00008834 }, /* GL_DRAW_BUFFER15 */
-   {  8258, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
-   {  8279, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
-   {  8300, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
-   {  8320, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
-   {  8340, 0x00008827 }, /* GL_DRAW_BUFFER2 */
-   {  8356, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
-   {  8376, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
-   {  8396, 0x00008828 }, /* GL_DRAW_BUFFER3 */
-   {  8412, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
-   {  8432, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
-   {  8452, 0x00008829 }, /* GL_DRAW_BUFFER4 */
-   {  8468, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
-   {  8488, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
-   {  8508, 0x0000882A }, /* GL_DRAW_BUFFER5 */
-   {  8524, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
-   {  8544, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
-   {  8564, 0x0000882B }, /* GL_DRAW_BUFFER6 */
-   {  8580, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
-   {  8600, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
-   {  8620, 0x0000882C }, /* GL_DRAW_BUFFER7 */
-   {  8636, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
-   {  8656, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
-   {  8676, 0x0000882D }, /* GL_DRAW_BUFFER8 */
-   {  8692, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
-   {  8712, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
-   {  8732, 0x0000882E }, /* GL_DRAW_BUFFER9 */
-   {  8748, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
-   {  8768, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
-   {  8788, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER */
-   {  8808, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-   {  8840, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
-   {  8864, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
-   {  8884, 0x00000304 }, /* GL_DST_ALPHA */
-   {  8897, 0x00000306 }, /* GL_DST_COLOR */
-   {  8910, 0x0000877A }, /* GL_DU8DV8_ATI */
-   {  8924, 0x00008779 }, /* GL_DUDV_ATI */
-   {  8936, 0x000088EA }, /* GL_DYNAMIC_COPY */
-   {  8952, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
-   {  8972, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
-   {  8988, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
-   {  9008, 0x000088E9 }, /* GL_DYNAMIC_READ */
-   {  9024, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
-   {  9044, 0x00000B43 }, /* GL_EDGE_FLAG */
-   {  9057, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
-   {  9076, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-   {  9110, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
-   {  9148, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
-   {  9175, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-   {  9201, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
-   {  9225, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-   {  9257, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
-   {  9293, 0x00001600 }, /* GL_EMISSION */
-   {  9305, 0x00002000 }, /* GL_ENABLE_BIT */
-   {  9319, 0x00000202 }, /* GL_EQUAL */
-   {  9328, 0x00001509 }, /* GL_EQUIV */
-   {  9337, 0x00010000 }, /* GL_EVAL_BIT */
-   {  9349, 0x00000800 }, /* GL_EXP */
-   {  9356, 0x00000801 }, /* GL_EXP2 */
-   {  9364, 0x00001F03 }, /* GL_EXTENSIONS */
-   {  9378, 0x00002400 }, /* GL_EYE_LINEAR */
-   {  9392, 0x00002502 }, /* GL_EYE_PLANE */
-   {  9405, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
-   {  9430, 0x0000855B }, /* GL_EYE_RADIAL_NV */
-   {  9447, 0x00000000 }, /* GL_FALSE */
-   {  9456, 0x00001101 }, /* GL_FASTEST */
-   {  9467, 0x00001C01 }, /* GL_FEEDBACK */
-   {  9479, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
-   {  9506, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
-   {  9530, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
-   {  9554, 0x00001B02 }, /* GL_FILL */
-   {  9562, 0x00001D00 }, /* GL_FLAT */
-   {  9570, 0x00001406 }, /* GL_FLOAT */
-   {  9579, 0x00008B5A }, /* GL_FLOAT_MAT2 */
-   {  9593, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
-   {  9611, 0x00008B65 }, /* GL_FLOAT_MAT2x3 */
-   {  9627, 0x00008B66 }, /* GL_FLOAT_MAT2x4 */
-   {  9643, 0x00008B5B }, /* GL_FLOAT_MAT3 */
-   {  9657, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
-   {  9675, 0x00008B67 }, /* GL_FLOAT_MAT3x2 */
-   {  9691, 0x00008B68 }, /* GL_FLOAT_MAT3x4 */
-   {  9707, 0x00008B5C }, /* GL_FLOAT_MAT4 */
-   {  9721, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
-   {  9739, 0x00008B69 }, /* GL_FLOAT_MAT4x2 */
-   {  9755, 0x00008B6A }, /* GL_FLOAT_MAT4x3 */
-   {  9771, 0x00008B50 }, /* GL_FLOAT_VEC2 */
-   {  9785, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
-   {  9803, 0x00008B51 }, /* GL_FLOAT_VEC3 */
-   {  9817, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
-   {  9835, 0x00008B52 }, /* GL_FLOAT_VEC4 */
-   {  9849, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
-   {  9867, 0x00000B60 }, /* GL_FOG */
-   {  9874, 0x00000080 }, /* GL_FOG_BIT */
-   {  9885, 0x00000B66 }, /* GL_FOG_COLOR */
-   {  9898, 0x00008451 }, /* GL_FOG_COORD */
-   {  9911, 0x00008451 }, /* GL_FOG_COORDINATE */
-   {  9929, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
-   {  9953, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-   {  9992, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
-   { 10035, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-   { 10067, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-   { 10098, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-   { 10127, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
-   { 10152, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
-   { 10171, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
-   { 10205, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
-   { 10232, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
-   { 10258, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
-   { 10282, 0x00008450 }, /* GL_FOG_COORD_SRC */
-   { 10299, 0x00000B62 }, /* GL_FOG_DENSITY */
-   { 10314, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
-   { 10338, 0x00000B64 }, /* GL_FOG_END */
-   { 10349, 0x00000C54 }, /* GL_FOG_HINT */
-   { 10361, 0x00000B61 }, /* GL_FOG_INDEX */
-   { 10374, 0x00000B65 }, /* GL_FOG_MODE */
-   { 10386, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
-   { 10405, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
-   { 10430, 0x00000B63 }, /* GL_FOG_START */
-   { 10443, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
-   { 10461, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
-   { 10485, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
-   { 10504, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
-   { 10527, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-   { 10562, 0x00008D40 }, /* GL_FRAMEBUFFER */
-   { 10577, 0x00008215 }, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
-   { 10614, 0x00008214 }, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
-   { 10650, 0x00008210 }, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
-   { 10691, 0x00008211 }, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
-   { 10732, 0x00008216 }, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
-   { 10769, 0x00008213 }, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
-   { 10806, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
-   { 10844, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
-   { 10886, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
-   { 10924, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
-   { 10966, 0x00008212 }, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
-   { 11001, 0x00008217 }, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
-   { 11040, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
-   { 11089, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
-   { 11137, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
-   { 11189, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
-   { 11229, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
-   { 11273, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
-   { 11313, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
-   { 11357, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
-   { 11384, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE */
-   { 11408, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
-   { 11436, 0x00008218 }, /* GL_FRAMEBUFFER_DEFAULT */
-   { 11459, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
-   { 11478, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
-   { 11515, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
-   { 11556, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-   { 11597, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-   { 11639, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-   { 11690, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-   { 11728, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
-   { 11773, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
-   { 11822, 0x00008D56 }, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
-   { 11860, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-   { 11902, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-   { 11934, 0x00008219 }, /* GL_FRAMEBUFFER_UNDEFINED */
-   { 11959, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED */
-   { 11986, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
-   { 12017, 0x00000404 }, /* GL_FRONT */
-   { 12026, 0x00000408 }, /* GL_FRONT_AND_BACK */
-   { 12044, 0x00000B46 }, /* GL_FRONT_FACE */
-   { 12058, 0x00000400 }, /* GL_FRONT_LEFT */
-   { 12072, 0x00000401 }, /* GL_FRONT_RIGHT */
-   { 12087, 0x00008006 }, /* GL_FUNC_ADD */
-   { 12099, 0x00008006 }, /* GL_FUNC_ADD_EXT */
-   { 12115, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
-   { 12140, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
-   { 12169, 0x0000800A }, /* GL_FUNC_SUBTRACT */
-   { 12186, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
-   { 12207, 0x00008191 }, /* GL_GENERATE_MIPMAP */
-   { 12226, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
-   { 12250, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
-   { 12279, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
-   { 12303, 0x00000206 }, /* GL_GEQUAL */
-   { 12313, 0x00000204 }, /* GL_GREATER */
-   { 12324, 0x00001904 }, /* GL_GREEN */
-   { 12333, 0x00000D19 }, /* GL_GREEN_BIAS */
-   { 12347, 0x00000D53 }, /* GL_GREEN_BITS */
-   { 12361, 0x00000D18 }, /* GL_GREEN_SCALE */
-   { 12376, 0x00008000 }, /* GL_HINT_BIT */
-   { 12388, 0x00008024 }, /* GL_HISTOGRAM */
-   { 12401, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
-   { 12425, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
-   { 12453, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
-   { 12476, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
-   { 12503, 0x00008024 }, /* GL_HISTOGRAM_EXT */
-   { 12520, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
-   { 12540, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
-   { 12564, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
-   { 12588, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
-   { 12616, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-   { 12644, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
-   { 12676, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
-   { 12698, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
-   { 12724, 0x0000802D }, /* GL_HISTOGRAM_SINK */
-   { 12742, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
-   { 12764, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
-   { 12783, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
-   { 12806, 0x0000862A }, /* GL_IDENTITY_NV */
-   { 12821, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
-   { 12841, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-   { 12881, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-   { 12919, 0x00001E02 }, /* GL_INCR */
-   { 12927, 0x00008507 }, /* GL_INCR_WRAP */
-   { 12940, 0x00008507 }, /* GL_INCR_WRAP_EXT */
-   { 12957, 0x00008222 }, /* GL_INDEX */
-   { 12966, 0x00008077 }, /* GL_INDEX_ARRAY */
-   { 12981, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-   { 13011, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
-   { 13045, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
-   { 13068, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
-   { 13090, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
-   { 13110, 0x00000D51 }, /* GL_INDEX_BITS */
-   { 13124, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
-   { 13145, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
-   { 13163, 0x00000C30 }, /* GL_INDEX_MODE */
-   { 13177, 0x00000D13 }, /* GL_INDEX_OFFSET */
-   { 13193, 0x00000D12 }, /* GL_INDEX_SHIFT */
-   { 13208, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
-   { 13227, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
-   { 13246, 0x00001404 }, /* GL_INT */
-   { 13253, 0x00008049 }, /* GL_INTENSITY */
-   { 13266, 0x0000804C }, /* GL_INTENSITY12 */
-   { 13281, 0x0000804C }, /* GL_INTENSITY12_EXT */
-   { 13300, 0x0000804D }, /* GL_INTENSITY16 */
-   { 13315, 0x0000804D }, /* GL_INTENSITY16_EXT */
-   { 13334, 0x0000804A }, /* GL_INTENSITY4 */
-   { 13348, 0x0000804A }, /* GL_INTENSITY4_EXT */
-   { 13366, 0x0000804B }, /* GL_INTENSITY8 */
-   { 13380, 0x0000804B }, /* GL_INTENSITY8_EXT */
-   { 13398, 0x00008049 }, /* GL_INTENSITY_EXT */
-   { 13415, 0x00008575 }, /* GL_INTERPOLATE */
-   { 13430, 0x00008575 }, /* GL_INTERPOLATE_ARB */
-   { 13449, 0x00008575 }, /* GL_INTERPOLATE_EXT */
-   { 13468, 0x00008B53 }, /* GL_INT_VEC2 */
-   { 13480, 0x00008B53 }, /* GL_INT_VEC2_ARB */
-   { 13496, 0x00008B54 }, /* GL_INT_VEC3 */
-   { 13508, 0x00008B54 }, /* GL_INT_VEC3_ARB */
-   { 13524, 0x00008B55 }, /* GL_INT_VEC4 */
-   { 13536, 0x00008B55 }, /* GL_INT_VEC4_ARB */
-   { 13552, 0x00000500 }, /* GL_INVALID_ENUM */
-   { 13568, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION */
-   { 13601, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
-   { 13638, 0x00000502 }, /* GL_INVALID_OPERATION */
-   { 13659, 0x00000501 }, /* GL_INVALID_VALUE */
-   { 13676, 0x0000862B }, /* GL_INVERSE_NV */
-   { 13690, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
-   { 13714, 0x0000150A }, /* GL_INVERT */
-   { 13724, 0x00001E00 }, /* GL_KEEP */
-   { 13732, 0x00000406 }, /* GL_LEFT */
-   { 13740, 0x00000203 }, /* GL_LEQUAL */
-   { 13750, 0x00000201 }, /* GL_LESS */
-   { 13758, 0x00004000 }, /* GL_LIGHT0 */
-   { 13768, 0x00004001 }, /* GL_LIGHT1 */
-   { 13778, 0x00004002 }, /* GL_LIGHT2 */
-   { 13788, 0x00004003 }, /* GL_LIGHT3 */
-   { 13798, 0x00004004 }, /* GL_LIGHT4 */
-   { 13808, 0x00004005 }, /* GL_LIGHT5 */
-   { 13818, 0x00004006 }, /* GL_LIGHT6 */
-   { 13828, 0x00004007 }, /* GL_LIGHT7 */
-   { 13838, 0x00000B50 }, /* GL_LIGHTING */
-   { 13850, 0x00000040 }, /* GL_LIGHTING_BIT */
-   { 13866, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
-   { 13889, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-   { 13918, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
-   { 13951, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-   { 13979, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
-   { 14003, 0x00001B01 }, /* GL_LINE */
-   { 14011, 0x00002601 }, /* GL_LINEAR */
-   { 14021, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
-   { 14043, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-   { 14073, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-   { 14104, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
-   { 14128, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
-   { 14153, 0x00000001 }, /* GL_LINES */
-   { 14162, 0x00000004 }, /* GL_LINE_BIT */
-   { 14174, 0x00000002 }, /* GL_LINE_LOOP */
-   { 14187, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
-   { 14207, 0x00000B20 }, /* GL_LINE_SMOOTH */
-   { 14222, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
-   { 14242, 0x00000B24 }, /* GL_LINE_STIPPLE */
-   { 14258, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
-   { 14282, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
-   { 14305, 0x00000003 }, /* GL_LINE_STRIP */
-   { 14319, 0x00000702 }, /* GL_LINE_TOKEN */
-   { 14333, 0x00000B21 }, /* GL_LINE_WIDTH */
-   { 14347, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
-   { 14373, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
-   { 14393, 0x00008B82 }, /* GL_LINK_STATUS */
-   { 14408, 0x00000B32 }, /* GL_LIST_BASE */
-   { 14421, 0x00020000 }, /* GL_LIST_BIT */
-   { 14433, 0x00000B33 }, /* GL_LIST_INDEX */
-   { 14447, 0x00000B30 }, /* GL_LIST_MODE */
-   { 14460, 0x00000101 }, /* GL_LOAD */
-   { 14468, 0x00000BF1 }, /* GL_LOGIC_OP */
-   { 14480, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
-   { 14497, 0x00008CA1 }, /* GL_LOWER_LEFT */
-   { 14511, 0x00001909 }, /* GL_LUMINANCE */
-   { 14524, 0x00008041 }, /* GL_LUMINANCE12 */
-   { 14539, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
-   { 14562, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
-   { 14589, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
-   { 14611, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
-   { 14637, 0x00008041 }, /* GL_LUMINANCE12_EXT */
-   { 14656, 0x00008042 }, /* GL_LUMINANCE16 */
-   { 14671, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
-   { 14694, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
-   { 14721, 0x00008042 }, /* GL_LUMINANCE16_EXT */
-   { 14740, 0x0000803F }, /* GL_LUMINANCE4 */
-   { 14754, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
-   { 14775, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
-   { 14800, 0x0000803F }, /* GL_LUMINANCE4_EXT */
-   { 14818, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
-   { 14839, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
-   { 14864, 0x00008040 }, /* GL_LUMINANCE8 */
-   { 14878, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
-   { 14899, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
-   { 14924, 0x00008040 }, /* GL_LUMINANCE8_EXT */
-   { 14942, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
-   { 14961, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
-   { 14977, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
-   { 14997, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
-   { 15019, 0x00000D91 }, /* GL_MAP1_INDEX */
-   { 15033, 0x00000D92 }, /* GL_MAP1_NORMAL */
-   { 15048, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
-   { 15072, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
-   { 15096, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
-   { 15120, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
-   { 15144, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
-   { 15161, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
-   { 15178, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-   { 15206, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-   { 15235, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-   { 15264, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-   { 15293, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-   { 15322, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-   { 15351, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-   { 15380, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-   { 15408, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-   { 15436, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-   { 15464, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-   { 15492, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-   { 15520, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-   { 15548, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-   { 15576, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-   { 15604, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-   { 15632, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
-   { 15648, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
-   { 15668, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
-   { 15690, 0x00000DB1 }, /* GL_MAP2_INDEX */
-   { 15704, 0x00000DB2 }, /* GL_MAP2_NORMAL */
-   { 15719, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
-   { 15743, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
-   { 15767, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
-   { 15791, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
-   { 15815, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
-   { 15832, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
-   { 15849, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-   { 15877, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-   { 15906, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-   { 15935, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-   { 15964, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-   { 15993, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-   { 16022, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-   { 16051, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-   { 16079, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-   { 16107, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-   { 16135, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-   { 16163, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-   { 16191, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-   { 16219, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
-   { 16247, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-   { 16275, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-   { 16303, 0x00000D10 }, /* GL_MAP_COLOR */
-   { 16316, 0x00000D11 }, /* GL_MAP_STENCIL */
-   { 16331, 0x000088C0 }, /* GL_MATRIX0_ARB */
-   { 16346, 0x00008630 }, /* GL_MATRIX0_NV */
-   { 16360, 0x000088CA }, /* GL_MATRIX10_ARB */
-   { 16376, 0x000088CB }, /* GL_MATRIX11_ARB */
-   { 16392, 0x000088CC }, /* GL_MATRIX12_ARB */
-   { 16408, 0x000088CD }, /* GL_MATRIX13_ARB */
-   { 16424, 0x000088CE }, /* GL_MATRIX14_ARB */
-   { 16440, 0x000088CF }, /* GL_MATRIX15_ARB */
-   { 16456, 0x000088D0 }, /* GL_MATRIX16_ARB */
-   { 16472, 0x000088D1 }, /* GL_MATRIX17_ARB */
-   { 16488, 0x000088D2 }, /* GL_MATRIX18_ARB */
-   { 16504, 0x000088D3 }, /* GL_MATRIX19_ARB */
-   { 16520, 0x000088C1 }, /* GL_MATRIX1_ARB */
-   { 16535, 0x00008631 }, /* GL_MATRIX1_NV */
-   { 16549, 0x000088D4 }, /* GL_MATRIX20_ARB */
-   { 16565, 0x000088D5 }, /* GL_MATRIX21_ARB */
-   { 16581, 0x000088D6 }, /* GL_MATRIX22_ARB */
-   { 16597, 0x000088D7 }, /* GL_MATRIX23_ARB */
-   { 16613, 0x000088D8 }, /* GL_MATRIX24_ARB */
-   { 16629, 0x000088D9 }, /* GL_MATRIX25_ARB */
-   { 16645, 0x000088DA }, /* GL_MATRIX26_ARB */
-   { 16661, 0x000088DB }, /* GL_MATRIX27_ARB */
-   { 16677, 0x000088DC }, /* GL_MATRIX28_ARB */
-   { 16693, 0x000088DD }, /* GL_MATRIX29_ARB */
-   { 16709, 0x000088C2 }, /* GL_MATRIX2_ARB */
-   { 16724, 0x00008632 }, /* GL_MATRIX2_NV */
-   { 16738, 0x000088DE }, /* GL_MATRIX30_ARB */
-   { 16754, 0x000088DF }, /* GL_MATRIX31_ARB */
-   { 16770, 0x000088C3 }, /* GL_MATRIX3_ARB */
-   { 16785, 0x00008633 }, /* GL_MATRIX3_NV */
-   { 16799, 0x000088C4 }, /* GL_MATRIX4_ARB */
-   { 16814, 0x00008634 }, /* GL_MATRIX4_NV */
-   { 16828, 0x000088C5 }, /* GL_MATRIX5_ARB */
-   { 16843, 0x00008635 }, /* GL_MATRIX5_NV */
-   { 16857, 0x000088C6 }, /* GL_MATRIX6_ARB */
-   { 16872, 0x00008636 }, /* GL_MATRIX6_NV */
-   { 16886, 0x000088C7 }, /* GL_MATRIX7_ARB */
-   { 16901, 0x00008637 }, /* GL_MATRIX7_NV */
-   { 16915, 0x000088C8 }, /* GL_MATRIX8_ARB */
-   { 16930, 0x000088C9 }, /* GL_MATRIX9_ARB */
-   { 16945, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
-   { 16971, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-   { 17005, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-   { 17036, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-   { 17069, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-   { 17100, 0x00000BA0 }, /* GL_MATRIX_MODE */
-   { 17115, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
-   { 17137, 0x00008008 }, /* GL_MAX */
-   { 17144, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
-   { 17167, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-   { 17199, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
-   { 17225, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-   { 17258, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-   { 17284, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 17318, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
-   { 17337, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
-   { 17366, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-   { 17398, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
-   { 17434, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-   { 17470, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
-   { 17510, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
-   { 17536, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
-   { 17566, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
-   { 17591, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
-   { 17620, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-   { 17649, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
-   { 17682, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
-   { 17702, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
-   { 17726, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
-   { 17750, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
-   { 17774, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
-   { 17799, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
-   { 17817, 0x00008008 }, /* GL_MAX_EXT */
-   { 17828, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-   { 17863, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
-   { 17902, 0x00000D31 }, /* GL_MAX_LIGHTS */
-   { 17916, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
-   { 17936, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-   { 17974, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-   { 18003, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
-   { 18027, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
-   { 18055, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
-   { 18078, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 18115, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 18151, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-   { 18178, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-   { 18207, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-   { 18241, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-   { 18277, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-   { 18304, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-   { 18336, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-   { 18372, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-   { 18401, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-   { 18430, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
-   { 18458, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-   { 18496, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 18540, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 18583, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 18617, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 18656, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 18693, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 18731, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 18774, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 18817, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-   { 18847, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-   { 18878, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 18914, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 18950, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
-   { 18980, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-   { 19014, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
-   { 19047, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
-   { 19076, 0x00008D57 }, /* GL_MAX_SAMPLES */
-   { 19091, 0x00008504 }, /* GL_MAX_SHININESS_NV */
-   { 19111, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
-   { 19135, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
-   { 19157, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
-   { 19183, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-   { 19210, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
-   { 19241, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
-   { 19265, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 19299, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
-   { 19319, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
-   { 19346, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
-   { 19367, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
-   { 19392, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
-   { 19417, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
-   { 19452, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
-   { 19474, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
-   { 19500, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
-   { 19522, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
-   { 19548, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-   { 19582, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
-   { 19620, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-   { 19653, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
-   { 19690, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
-   { 19714, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
-   { 19735, 0x00008007 }, /* GL_MIN */
-   { 19742, 0x0000802E }, /* GL_MINMAX */
-   { 19752, 0x0000802E }, /* GL_MINMAX_EXT */
-   { 19766, 0x0000802F }, /* GL_MINMAX_FORMAT */
-   { 19783, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
-   { 19804, 0x00008030 }, /* GL_MINMAX_SINK */
-   { 19819, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
-   { 19838, 0x00008007 }, /* GL_MIN_EXT */
-   { 19849, 0x00008370 }, /* GL_MIRRORED_REPEAT */
-   { 19868, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
-   { 19891, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
-   { 19914, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
-   { 19934, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
-   { 19954, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-   { 19984, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
-   { 20012, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-   { 20040, 0x00001700 }, /* GL_MODELVIEW */
-   { 20053, 0x00001700 }, /* GL_MODELVIEW0_ARB */
-   { 20071, 0x0000872A }, /* GL_MODELVIEW10_ARB */
-   { 20090, 0x0000872B }, /* GL_MODELVIEW11_ARB */
-   { 20109, 0x0000872C }, /* GL_MODELVIEW12_ARB */
-   { 20128, 0x0000872D }, /* GL_MODELVIEW13_ARB */
-   { 20147, 0x0000872E }, /* GL_MODELVIEW14_ARB */
-   { 20166, 0x0000872F }, /* GL_MODELVIEW15_ARB */
-   { 20185, 0x00008730 }, /* GL_MODELVIEW16_ARB */
-   { 20204, 0x00008731 }, /* GL_MODELVIEW17_ARB */
-   { 20223, 0x00008732 }, /* GL_MODELVIEW18_ARB */
-   { 20242, 0x00008733 }, /* GL_MODELVIEW19_ARB */
-   { 20261, 0x0000850A }, /* GL_MODELVIEW1_ARB */
-   { 20279, 0x00008734 }, /* GL_MODELVIEW20_ARB */
-   { 20298, 0x00008735 }, /* GL_MODELVIEW21_ARB */
-   { 20317, 0x00008736 }, /* GL_MODELVIEW22_ARB */
-   { 20336, 0x00008737 }, /* GL_MODELVIEW23_ARB */
-   { 20355, 0x00008738 }, /* GL_MODELVIEW24_ARB */
-   { 20374, 0x00008739 }, /* GL_MODELVIEW25_ARB */
-   { 20393, 0x0000873A }, /* GL_MODELVIEW26_ARB */
-   { 20412, 0x0000873B }, /* GL_MODELVIEW27_ARB */
-   { 20431, 0x0000873C }, /* GL_MODELVIEW28_ARB */
-   { 20450, 0x0000873D }, /* GL_MODELVIEW29_ARB */
-   { 20469, 0x00008722 }, /* GL_MODELVIEW2_ARB */
-   { 20487, 0x0000873E }, /* GL_MODELVIEW30_ARB */
-   { 20506, 0x0000873F }, /* GL_MODELVIEW31_ARB */
-   { 20525, 0x00008723 }, /* GL_MODELVIEW3_ARB */
-   { 20543, 0x00008724 }, /* GL_MODELVIEW4_ARB */
-   { 20561, 0x00008725 }, /* GL_MODELVIEW5_ARB */
-   { 20579, 0x00008726 }, /* GL_MODELVIEW6_ARB */
-   { 20597, 0x00008727 }, /* GL_MODELVIEW7_ARB */
-   { 20615, 0x00008728 }, /* GL_MODELVIEW8_ARB */
-   { 20633, 0x00008729 }, /* GL_MODELVIEW9_ARB */
-   { 20651, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
-   { 20671, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
-   { 20698, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
-   { 20723, 0x00002100 }, /* GL_MODULATE */
-   { 20735, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
-   { 20755, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
-   { 20782, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
-   { 20807, 0x00000103 }, /* GL_MULT */
-   { 20815, 0x0000809D }, /* GL_MULTISAMPLE */
-   { 20830, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
-   { 20850, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
-   { 20869, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
-   { 20888, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
-   { 20912, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
-   { 20935, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-   { 20965, 0x00002A25 }, /* GL_N3F_V3F */
-   { 20976, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
-   { 20996, 0x0000150E }, /* GL_NAND */
-   { 21004, 0x00002600 }, /* GL_NEAREST */
-   { 21015, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-   { 21046, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-   { 21078, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
-   { 21103, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
-   { 21129, 0x00000200 }, /* GL_NEVER */
-   { 21138, 0x00001102 }, /* GL_NICEST */
-   { 21148, 0x00000000 }, /* GL_NONE */
-   { 21156, 0x00001505 }, /* GL_NOOP */
-   { 21164, 0x00001508 }, /* GL_NOR */
-   { 21171, 0x00000BA1 }, /* GL_NORMALIZE */
-   { 21184, 0x00008075 }, /* GL_NORMAL_ARRAY */
-   { 21200, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
-   { 21231, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
-   { 21266, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
-   { 21290, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
-   { 21313, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
-   { 21334, 0x00008511 }, /* GL_NORMAL_MAP */
-   { 21348, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
-   { 21366, 0x00008511 }, /* GL_NORMAL_MAP_NV */
-   { 21383, 0x00000205 }, /* GL_NOTEQUAL */
-   { 21395, 0x00000000 }, /* GL_NO_ERROR */
-   { 21407, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
-   { 21441, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
-   { 21479, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
-   { 21511, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
-   { 21553, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
-   { 21583, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
-   { 21623, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
-   { 21654, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
-   { 21683, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
-   { 21711, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
-   { 21741, 0x00002401 }, /* GL_OBJECT_LINEAR */
-   { 21758, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
-   { 21784, 0x00002501 }, /* GL_OBJECT_PLANE */
-   { 21800, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
-   { 21835, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
-   { 21857, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
-   { 21876, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
-   { 21906, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
-   { 21927, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
-   { 21955, 0x00000001 }, /* GL_ONE */
-   { 21962, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
-   { 21990, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
-   { 22022, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
-   { 22050, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
-   { 22082, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
-   { 22105, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
-   { 22128, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
-   { 22151, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
-   { 22174, 0x00008598 }, /* GL_OPERAND0_ALPHA */
-   { 22192, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
-   { 22214, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
-   { 22236, 0x00008590 }, /* GL_OPERAND0_RGB */
-   { 22252, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
-   { 22272, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
-   { 22292, 0x00008599 }, /* GL_OPERAND1_ALPHA */
-   { 22310, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
-   { 22332, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
-   { 22354, 0x00008591 }, /* GL_OPERAND1_RGB */
-   { 22370, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
-   { 22390, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
-   { 22410, 0x0000859A }, /* GL_OPERAND2_ALPHA */
-   { 22428, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
-   { 22450, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
-   { 22472, 0x00008592 }, /* GL_OPERAND2_RGB */
-   { 22488, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
-   { 22508, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
-   { 22528, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
-   { 22549, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
-   { 22568, 0x00001507 }, /* GL_OR */
-   { 22574, 0x00000A01 }, /* GL_ORDER */
-   { 22583, 0x0000150D }, /* GL_OR_INVERTED */
-   { 22598, 0x0000150B }, /* GL_OR_REVERSE */
-   { 22612, 0x00000505 }, /* GL_OUT_OF_MEMORY */
-   { 22629, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
-   { 22647, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
-   { 22668, 0x00008758 }, /* GL_PACK_INVERT_MESA */
-   { 22688, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
-   { 22706, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
-   { 22725, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
-   { 22745, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
-   { 22765, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
-   { 22783, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
-   { 22802, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
-   { 22827, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
-   { 22851, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
-   { 22872, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
-   { 22894, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
-   { 22916, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
-   { 22941, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
-   { 22965, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
-   { 22986, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
-   { 23008, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
-   { 23030, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
-   { 23052, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
-   { 23083, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
-   { 23103, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-   { 23128, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
-   { 23148, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-   { 23173, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
-   { 23193, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-   { 23218, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
-   { 23238, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-   { 23263, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
-   { 23283, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-   { 23308, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
-   { 23328, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-   { 23353, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
-   { 23373, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-   { 23398, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
-   { 23418, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-   { 23443, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
-   { 23463, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-   { 23488, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
-   { 23508, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-   { 23533, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
-   { 23551, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER */
-   { 23572, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING */
-   { 23601, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
-   { 23634, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
-   { 23659, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER */
-   { 23682, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
-   { 23713, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
-   { 23748, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
-   { 23775, 0x00001B00 }, /* GL_POINT */
-   { 23784, 0x00000000 }, /* GL_POINTS */
-   { 23794, 0x00000002 }, /* GL_POINT_BIT */
-   { 23807, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
-   { 23837, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
-   { 23871, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
-   { 23905, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
-   { 23940, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
-   { 23969, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
-   { 24002, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
-   { 24035, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
-   { 24069, 0x00000B11 }, /* GL_POINT_SIZE */
-   { 24083, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
-   { 24109, 0x00008127 }, /* GL_POINT_SIZE_MAX */
-   { 24127, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
-   { 24149, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
-   { 24171, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
-   { 24194, 0x00008126 }, /* GL_POINT_SIZE_MIN */
-   { 24212, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
-   { 24234, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
-   { 24256, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
-   { 24279, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
-   { 24299, 0x00000B10 }, /* GL_POINT_SMOOTH */
-   { 24315, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
-   { 24336, 0x00008861 }, /* GL_POINT_SPRITE */
-   { 24352, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
-   { 24372, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
-   { 24401, 0x00008861 }, /* GL_POINT_SPRITE_NV */
-   { 24420, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
-   { 24446, 0x00000701 }, /* GL_POINT_TOKEN */
-   { 24461, 0x00000009 }, /* GL_POLYGON */
-   { 24472, 0x00000008 }, /* GL_POLYGON_BIT */
-   { 24487, 0x00000B40 }, /* GL_POLYGON_MODE */
-   { 24503, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
-   { 24526, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
-   { 24551, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
-   { 24574, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
-   { 24597, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
-   { 24621, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
-   { 24645, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
-   { 24663, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
-   { 24686, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
-   { 24705, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
-   { 24728, 0x00000703 }, /* GL_POLYGON_TOKEN */
-   { 24745, 0x00001203 }, /* GL_POSITION */
-   { 24757, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-   { 24789, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
-   { 24825, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-   { 24858, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
-   { 24895, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-   { 24926, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
-   { 24961, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-   { 24993, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
-   { 25029, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 25062, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-   { 25094, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
-   { 25130, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-   { 25163, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
-   { 25200, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-   { 25230, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
-   { 25264, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-   { 25295, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
-   { 25330, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-   { 25361, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
-   { 25396, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-   { 25428, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
-   { 25464, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-   { 25494, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
-   { 25528, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-   { 25559, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
-   { 25594, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-   { 25626, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-   { 25657, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
-   { 25692, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-   { 25724, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
-   { 25760, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
-   { 25789, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
-   { 25822, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
-   { 25852, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
-   { 25886, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-   { 25925, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-   { 25958, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-   { 25998, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-   { 26032, 0x00008578 }, /* GL_PREVIOUS */
-   { 26044, 0x00008578 }, /* GL_PREVIOUS_ARB */
-   { 26060, 0x00008578 }, /* GL_PREVIOUS_EXT */
-   { 26076, 0x00008577 }, /* GL_PRIMARY_COLOR */
-   { 26093, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
-   { 26114, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
-   { 26135, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 26168, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 26200, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
-   { 26223, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
-   { 26246, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
-   { 26276, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
-   { 26305, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
-   { 26333, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
-   { 26355, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-   { 26383, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-   { 26411, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
-   { 26433, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
-   { 26454, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 26494, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 26533, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 26563, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 26598, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 26631, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 26665, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 26704, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 26743, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
-   { 26765, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
-   { 26791, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
-   { 26815, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
-   { 26838, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
-   { 26860, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
-   { 26881, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
-   { 26902, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
-   { 26929, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 26961, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 26993, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-   { 27028, 0x00001701 }, /* GL_PROJECTION */
-   { 27042, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
-   { 27063, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
-   { 27089, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
-   { 27110, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
-   { 27129, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
-   { 27152, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 27191, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-   { 27229, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
-   { 27249, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-   { 27279, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
-   { 27303, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
-   { 27323, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-   { 27353, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
-   { 27377, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
-   { 27397, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-   { 27430, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
-   { 27456, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
-   { 27486, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-   { 27517, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
-   { 27547, 0x00002003 }, /* GL_Q */
-   { 27552, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
-   { 27577, 0x00000007 }, /* GL_QUADS */
-   { 27586, 0x00008614 }, /* GL_QUAD_MESH_SUN */
-   { 27603, 0x00000008 }, /* GL_QUAD_STRIP */
-   { 27617, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
-   { 27639, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
-   { 27665, 0x00008866 }, /* GL_QUERY_RESULT */
-   { 27681, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
-   { 27701, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
-   { 27727, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
-   { 27757, 0x00002002 }, /* GL_R */
-   { 27762, 0x00002A10 }, /* GL_R3_G3_B2 */
-   { 27774, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-   { 27807, 0x00000C02 }, /* GL_READ_BUFFER */
-   { 27822, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER */
-   { 27842, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-   { 27874, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
-   { 27898, 0x000088B8 }, /* GL_READ_ONLY */
-   { 27911, 0x000088B8 }, /* GL_READ_ONLY_ARB */
-   { 27928, 0x000088BA }, /* GL_READ_WRITE */
-   { 27942, 0x000088BA }, /* GL_READ_WRITE_ARB */
-   { 27960, 0x00001903 }, /* GL_RED */
-   { 27967, 0x00008016 }, /* GL_REDUCE */
-   { 27977, 0x00008016 }, /* GL_REDUCE_EXT */
-   { 27991, 0x00000D15 }, /* GL_RED_BIAS */
-   { 28003, 0x00000D52 }, /* GL_RED_BITS */
-   { 28015, 0x00000D14 }, /* GL_RED_SCALE */
-   { 28028, 0x00008512 }, /* GL_REFLECTION_MAP */
-   { 28046, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
-   { 28068, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
-   { 28089, 0x00001C00 }, /* GL_RENDER */
-   { 28099, 0x00008D41 }, /* GL_RENDERBUFFER */
-   { 28115, 0x00008D53 }, /* GL_RENDERBUFFER_ALPHA_SIZE */
-   { 28142, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
-   { 28170, 0x00008D52 }, /* GL_RENDERBUFFER_BLUE_SIZE */
-   { 28196, 0x00008D54 }, /* GL_RENDERBUFFER_DEPTH_SIZE */
-   { 28223, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
-   { 28243, 0x00008D51 }, /* GL_RENDERBUFFER_GREEN_SIZE */
-   { 28270, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT */
-   { 28293, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
-   { 28320, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
-   { 28352, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
-   { 28388, 0x00008D50 }, /* GL_RENDERBUFFER_RED_SIZE */
-   { 28413, 0x00008CAB }, /* GL_RENDERBUFFER_SAMPLES */
-   { 28437, 0x00008D55 }, /* GL_RENDERBUFFER_STENCIL_SIZE */
-   { 28466, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH */
-   { 28488, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
-   { 28514, 0x00001F01 }, /* GL_RENDERER */
-   { 28526, 0x00000C40 }, /* GL_RENDER_MODE */
-   { 28541, 0x00002901 }, /* GL_REPEAT */
-   { 28551, 0x00001E01 }, /* GL_REPLACE */
-   { 28562, 0x00008062 }, /* GL_REPLACE_EXT */
-   { 28577, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
-   { 28600, 0x0000803A }, /* GL_RESCALE_NORMAL */
-   { 28618, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
-   { 28640, 0x00000102 }, /* GL_RETURN */
-   { 28650, 0x00001907 }, /* GL_RGB */
-   { 28657, 0x00008052 }, /* GL_RGB10 */
-   { 28666, 0x00008059 }, /* GL_RGB10_A2 */
-   { 28678, 0x00008059 }, /* GL_RGB10_A2_EXT */
-   { 28694, 0x00008052 }, /* GL_RGB10_EXT */
-   { 28707, 0x00008053 }, /* GL_RGB12 */
-   { 28716, 0x00008053 }, /* GL_RGB12_EXT */
-   { 28729, 0x00008054 }, /* GL_RGB16 */
-   { 28738, 0x00008054 }, /* GL_RGB16_EXT */
-   { 28751, 0x0000804E }, /* GL_RGB2_EXT */
-   { 28763, 0x0000804F }, /* GL_RGB4 */
-   { 28771, 0x0000804F }, /* GL_RGB4_EXT */
-   { 28783, 0x000083A1 }, /* GL_RGB4_S3TC */
-   { 28796, 0x00008050 }, /* GL_RGB5 */
-   { 28804, 0x00008057 }, /* GL_RGB5_A1 */
-   { 28815, 0x00008057 }, /* GL_RGB5_A1_EXT */
-   { 28830, 0x00008050 }, /* GL_RGB5_EXT */
-   { 28842, 0x00008051 }, /* GL_RGB8 */
-   { 28850, 0x00008051 }, /* GL_RGB8_EXT */
-   { 28862, 0x00001908 }, /* GL_RGBA */
-   { 28870, 0x0000805A }, /* GL_RGBA12 */
-   { 28880, 0x0000805A }, /* GL_RGBA12_EXT */
-   { 28894, 0x0000805B }, /* GL_RGBA16 */
-   { 28904, 0x0000805B }, /* GL_RGBA16_EXT */
-   { 28918, 0x00008055 }, /* GL_RGBA2 */
-   { 28927, 0x00008055 }, /* GL_RGBA2_EXT */
-   { 28940, 0x00008056 }, /* GL_RGBA4 */
-   { 28949, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
-   { 28968, 0x00008056 }, /* GL_RGBA4_EXT */
-   { 28981, 0x000083A3 }, /* GL_RGBA4_S3TC */
-   { 28995, 0x00008058 }, /* GL_RGBA8 */
-   { 29004, 0x00008058 }, /* GL_RGBA8_EXT */
-   { 29017, 0x00008F97 }, /* GL_RGBA8_SNORM */
-   { 29032, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
-   { 29050, 0x00000C31 }, /* GL_RGBA_MODE */
-   { 29063, 0x000083A2 }, /* GL_RGBA_S3TC */
-   { 29076, 0x00008F93 }, /* GL_RGBA_SNORM */
-   { 29090, 0x000083A0 }, /* GL_RGB_S3TC */
-   { 29102, 0x00008573 }, /* GL_RGB_SCALE */
-   { 29115, 0x00008573 }, /* GL_RGB_SCALE_ARB */
-   { 29132, 0x00008573 }, /* GL_RGB_SCALE_EXT */
-   { 29149, 0x00000407 }, /* GL_RIGHT */
-   { 29158, 0x00002000 }, /* GL_S */
-   { 29163, 0x00008B5D }, /* GL_SAMPLER_1D */
-   { 29177, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
-   { 29198, 0x00008B5E }, /* GL_SAMPLER_2D */
-   { 29212, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
-   { 29233, 0x00008B5F }, /* GL_SAMPLER_3D */
-   { 29247, 0x00008B60 }, /* GL_SAMPLER_CUBE */
-   { 29263, 0x000080A9 }, /* GL_SAMPLES */
-   { 29274, 0x000086B4 }, /* GL_SAMPLES_3DFX */
-   { 29290, 0x000080A9 }, /* GL_SAMPLES_ARB */
-   { 29305, 0x00008914 }, /* GL_SAMPLES_PASSED */
-   { 29323, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
-   { 29345, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-   { 29373, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
-   { 29405, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
-   { 29428, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
-   { 29455, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
-   { 29473, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
-   { 29496, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
-   { 29518, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
-   { 29537, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
-   { 29560, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
-   { 29586, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
-   { 29616, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
-   { 29641, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
-   { 29670, 0x00080000 }, /* GL_SCISSOR_BIT */
-   { 29685, 0x00000C10 }, /* GL_SCISSOR_BOX */
-   { 29700, 0x00000C11 }, /* GL_SCISSOR_TEST */
-   { 29716, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
-   { 29741, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-   { 29781, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
-   { 29825, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-   { 29858, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-   { 29888, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-   { 29920, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-   { 29950, 0x00001C02 }, /* GL_SELECT */
-   { 29960, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
-   { 29988, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
-   { 30013, 0x00008012 }, /* GL_SEPARABLE_2D */
-   { 30029, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
-   { 30056, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
-   { 30087, 0x0000150F }, /* GL_SET */
-   { 30094, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
-   { 30115, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
-   { 30139, 0x00008B4F }, /* GL_SHADER_TYPE */
-   { 30154, 0x00000B54 }, /* GL_SHADE_MODEL */
-   { 30169, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
-   { 30197, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
-   { 30220, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-   { 30250, 0x00001601 }, /* GL_SHININESS */
-   { 30263, 0x00001402 }, /* GL_SHORT */
-   { 30272, 0x00008F9C }, /* GL_SIGNED_NORMALIZED */
-   { 30293, 0x000081F9 }, /* GL_SINGLE_COLOR */
-   { 30309, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
-   { 30329, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
-   { 30348, 0x00008C46 }, /* GL_SLUMINANCE */
-   { 30362, 0x00008C47 }, /* GL_SLUMINANCE8 */
-   { 30377, 0x00008C45 }, /* GL_SLUMINANCE8_ALPHA8 */
-   { 30399, 0x00008C44 }, /* GL_SLUMINANCE_ALPHA */
-   { 30419, 0x00001D01 }, /* GL_SMOOTH */
-   { 30429, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
-   { 30462, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
-   { 30489, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
-   { 30522, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
-   { 30549, 0x00008588 }, /* GL_SOURCE0_ALPHA */
-   { 30566, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
-   { 30587, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
-   { 30608, 0x00008580 }, /* GL_SOURCE0_RGB */
-   { 30623, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
-   { 30642, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
-   { 30661, 0x00008589 }, /* GL_SOURCE1_ALPHA */
-   { 30678, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
-   { 30699, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
-   { 30720, 0x00008581 }, /* GL_SOURCE1_RGB */
-   { 30735, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
-   { 30754, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
-   { 30773, 0x0000858A }, /* GL_SOURCE2_ALPHA */
-   { 30790, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
-   { 30811, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
-   { 30832, 0x00008582 }, /* GL_SOURCE2_RGB */
-   { 30847, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
-   { 30866, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
-   { 30885, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
-   { 30905, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
-   { 30923, 0x00001202 }, /* GL_SPECULAR */
-   { 30935, 0x00002402 }, /* GL_SPHERE_MAP */
-   { 30949, 0x00001206 }, /* GL_SPOT_CUTOFF */
-   { 30964, 0x00001204 }, /* GL_SPOT_DIRECTION */
-   { 30982, 0x00001205 }, /* GL_SPOT_EXPONENT */
-   { 30999, 0x00008588 }, /* GL_SRC0_ALPHA */
-   { 31013, 0x00008580 }, /* GL_SRC0_RGB */
-   { 31025, 0x00008589 }, /* GL_SRC1_ALPHA */
-   { 31039, 0x00008581 }, /* GL_SRC1_RGB */
-   { 31051, 0x0000858A }, /* GL_SRC2_ALPHA */
-   { 31065, 0x00008582 }, /* GL_SRC2_RGB */
-   { 31077, 0x00000302 }, /* GL_SRC_ALPHA */
-   { 31090, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
-   { 31112, 0x00000300 }, /* GL_SRC_COLOR */
-   { 31125, 0x00008C40 }, /* GL_SRGB */
-   { 31133, 0x00008C41 }, /* GL_SRGB8 */
-   { 31142, 0x00008C43 }, /* GL_SRGB8_ALPHA8 */
-   { 31158, 0x00008C42 }, /* GL_SRGB_ALPHA */
-   { 31172, 0x00000503 }, /* GL_STACK_OVERFLOW */
-   { 31190, 0x00000504 }, /* GL_STACK_UNDERFLOW */
-   { 31209, 0x000088E6 }, /* GL_STATIC_COPY */
-   { 31224, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
-   { 31243, 0x000088E4 }, /* GL_STATIC_DRAW */
-   { 31258, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
-   { 31277, 0x000088E5 }, /* GL_STATIC_READ */
-   { 31292, 0x000088E5 }, /* GL_STATIC_READ_ARB */
-   { 31311, 0x00001802 }, /* GL_STENCIL */
-   { 31322, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT */
-   { 31344, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
-   { 31370, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
-   { 31391, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
-   { 31416, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
-   { 31437, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
-   { 31462, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-   { 31494, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
-   { 31530, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-   { 31562, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
-   { 31598, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
-   { 31618, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
-   { 31645, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
-   { 31671, 0x00000D57 }, /* GL_STENCIL_BITS */
-   { 31687, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
-   { 31709, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
-   { 31732, 0x00000B94 }, /* GL_STENCIL_FAIL */
-   { 31748, 0x00000B92 }, /* GL_STENCIL_FUNC */
-   { 31764, 0x00001901 }, /* GL_STENCIL_INDEX */
-   { 31781, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
-   { 31804, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
-   { 31826, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
-   { 31848, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
-   { 31870, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
-   { 31891, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
-   { 31918, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
-   { 31945, 0x00000B97 }, /* GL_STENCIL_REF */
-   { 31960, 0x00000B90 }, /* GL_STENCIL_TEST */
-   { 31976, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
-   { 32005, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
-   { 32027, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
-   { 32048, 0x00000C33 }, /* GL_STEREO */
-   { 32058, 0x000088E2 }, /* GL_STREAM_COPY */
-   { 32073, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
-   { 32092, 0x000088E0 }, /* GL_STREAM_DRAW */
-   { 32107, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
-   { 32126, 0x000088E1 }, /* GL_STREAM_READ */
-   { 32141, 0x000088E1 }, /* GL_STREAM_READ_ARB */
-   { 32160, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
-   { 32177, 0x000084E7 }, /* GL_SUBTRACT */
-   { 32189, 0x000084E7 }, /* GL_SUBTRACT_ARB */
-   { 32205, 0x00002001 }, /* GL_T */
-   { 32210, 0x00002A2A }, /* GL_T2F_C3F_V3F */
-   { 32225, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
-   { 32244, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
-   { 32260, 0x00002A2B }, /* GL_T2F_N3F_V3F */
-   { 32275, 0x00002A27 }, /* GL_T2F_V3F */
-   { 32286, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
-   { 32305, 0x00002A28 }, /* GL_T4F_V4F */
-   { 32316, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
-   { 32339, 0x00001702 }, /* GL_TEXTURE */
-   { 32350, 0x000084C0 }, /* GL_TEXTURE0 */
-   { 32362, 0x000084C0 }, /* GL_TEXTURE0_ARB */
-   { 32378, 0x000084C1 }, /* GL_TEXTURE1 */
-   { 32390, 0x000084CA }, /* GL_TEXTURE10 */
-   { 32403, 0x000084CA }, /* GL_TEXTURE10_ARB */
-   { 32420, 0x000084CB }, /* GL_TEXTURE11 */
-   { 32433, 0x000084CB }, /* GL_TEXTURE11_ARB */
-   { 32450, 0x000084CC }, /* GL_TEXTURE12 */
-   { 32463, 0x000084CC }, /* GL_TEXTURE12_ARB */
-   { 32480, 0x000084CD }, /* GL_TEXTURE13 */
-   { 32493, 0x000084CD }, /* GL_TEXTURE13_ARB */
-   { 32510, 0x000084CE }, /* GL_TEXTURE14 */
-   { 32523, 0x000084CE }, /* GL_TEXTURE14_ARB */
-   { 32540, 0x000084CF }, /* GL_TEXTURE15 */
-   { 32553, 0x000084CF }, /* GL_TEXTURE15_ARB */
-   { 32570, 0x000084D0 }, /* GL_TEXTURE16 */
-   { 32583, 0x000084D0 }, /* GL_TEXTURE16_ARB */
-   { 32600, 0x000084D1 }, /* GL_TEXTURE17 */
-   { 32613, 0x000084D1 }, /* GL_TEXTURE17_ARB */
-   { 32630, 0x000084D2 }, /* GL_TEXTURE18 */
-   { 32643, 0x000084D2 }, /* GL_TEXTURE18_ARB */
-   { 32660, 0x000084D3 }, /* GL_TEXTURE19 */
-   { 32673, 0x000084D3 }, /* GL_TEXTURE19_ARB */
-   { 32690, 0x000084C1 }, /* GL_TEXTURE1_ARB */
-   { 32706, 0x000084C2 }, /* GL_TEXTURE2 */
-   { 32718, 0x000084D4 }, /* GL_TEXTURE20 */
-   { 32731, 0x000084D4 }, /* GL_TEXTURE20_ARB */
-   { 32748, 0x000084D5 }, /* GL_TEXTURE21 */
-   { 32761, 0x000084D5 }, /* GL_TEXTURE21_ARB */
-   { 32778, 0x000084D6 }, /* GL_TEXTURE22 */
-   { 32791, 0x000084D6 }, /* GL_TEXTURE22_ARB */
-   { 32808, 0x000084D7 }, /* GL_TEXTURE23 */
-   { 32821, 0x000084D7 }, /* GL_TEXTURE23_ARB */
-   { 32838, 0x000084D8 }, /* GL_TEXTURE24 */
-   { 32851, 0x000084D8 }, /* GL_TEXTURE24_ARB */
-   { 32868, 0x000084D9 }, /* GL_TEXTURE25 */
-   { 32881, 0x000084D9 }, /* GL_TEXTURE25_ARB */
-   { 32898, 0x000084DA }, /* GL_TEXTURE26 */
-   { 32911, 0x000084DA }, /* GL_TEXTURE26_ARB */
-   { 32928, 0x000084DB }, /* GL_TEXTURE27 */
-   { 32941, 0x000084DB }, /* GL_TEXTURE27_ARB */
-   { 32958, 0x000084DC }, /* GL_TEXTURE28 */
-   { 32971, 0x000084DC }, /* GL_TEXTURE28_ARB */
-   { 32988, 0x000084DD }, /* GL_TEXTURE29 */
-   { 33001, 0x000084DD }, /* GL_TEXTURE29_ARB */
-   { 33018, 0x000084C2 }, /* GL_TEXTURE2_ARB */
-   { 33034, 0x000084C3 }, /* GL_TEXTURE3 */
-   { 33046, 0x000084DE }, /* GL_TEXTURE30 */
-   { 33059, 0x000084DE }, /* GL_TEXTURE30_ARB */
-   { 33076, 0x000084DF }, /* GL_TEXTURE31 */
-   { 33089, 0x000084DF }, /* GL_TEXTURE31_ARB */
-   { 33106, 0x000084C3 }, /* GL_TEXTURE3_ARB */
-   { 33122, 0x000084C4 }, /* GL_TEXTURE4 */
-   { 33134, 0x000084C4 }, /* GL_TEXTURE4_ARB */
-   { 33150, 0x000084C5 }, /* GL_TEXTURE5 */
-   { 33162, 0x000084C5 }, /* GL_TEXTURE5_ARB */
-   { 33178, 0x000084C6 }, /* GL_TEXTURE6 */
-   { 33190, 0x000084C6 }, /* GL_TEXTURE6_ARB */
-   { 33206, 0x000084C7 }, /* GL_TEXTURE7 */
-   { 33218, 0x000084C7 }, /* GL_TEXTURE7_ARB */
-   { 33234, 0x000084C8 }, /* GL_TEXTURE8 */
-   { 33246, 0x000084C8 }, /* GL_TEXTURE8_ARB */
-   { 33262, 0x000084C9 }, /* GL_TEXTURE9 */
-   { 33274, 0x000084C9 }, /* GL_TEXTURE9_ARB */
-   { 33290, 0x00000DE0 }, /* GL_TEXTURE_1D */
-   { 33304, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
-   { 33328, 0x00000DE1 }, /* GL_TEXTURE_2D */
-   { 33342, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
-   { 33366, 0x0000806F }, /* GL_TEXTURE_3D */
-   { 33380, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
-   { 33402, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
-   { 33428, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
-   { 33450, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
-   { 33472, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-   { 33504, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
-   { 33526, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-   { 33558, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
-   { 33580, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
-   { 33608, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
-   { 33640, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-   { 33673, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
-   { 33705, 0x00040000 }, /* GL_TEXTURE_BIT */
-   { 33720, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
-   { 33741, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
-   { 33766, 0x00001005 }, /* GL_TEXTURE_BORDER */
-   { 33784, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
-   { 33808, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-   { 33839, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-   { 33869, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-   { 33899, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-   { 33934, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-   { 33965, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 34003, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
-   { 34030, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-   { 34062, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-   { 34096, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
-   { 34120, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
-   { 34148, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
-   { 34172, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
-   { 34200, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-   { 34233, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
-   { 34257, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
-   { 34279, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
-   { 34301, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
-   { 34327, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
-   { 34361, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-   { 34394, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
-   { 34431, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
-   { 34459, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
-   { 34491, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
-   { 34514, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-   { 34552, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
-   { 34594, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-   { 34625, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-   { 34653, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-   { 34683, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-   { 34711, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
-   { 34731, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
-   { 34755, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-   { 34786, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
-   { 34821, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-   { 34852, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
-   { 34887, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-   { 34918, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
-   { 34953, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-   { 34984, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
-   { 35019, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-   { 35050, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
-   { 35085, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-   { 35116, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
-   { 35151, 0x00008071 }, /* GL_TEXTURE_DEPTH */
-   { 35168, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
-   { 35190, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
-   { 35216, 0x00002300 }, /* GL_TEXTURE_ENV */
-   { 35231, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
-   { 35252, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
-   { 35272, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
-   { 35298, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
-   { 35318, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
-   { 35335, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
-   { 35352, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
-   { 35369, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
-   { 35386, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
-   { 35411, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
-   { 35433, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
-   { 35459, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
-   { 35477, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
-   { 35503, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
-   { 35529, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
-   { 35559, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
-   { 35586, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
-   { 35611, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
-   { 35631, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
-   { 35655, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-   { 35682, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-   { 35709, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-   { 35736, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
-   { 35762, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
-   { 35792, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
-   { 35814, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
-   { 35832, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 35862, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-   { 35890, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-   { 35918, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-   { 35946, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
-   { 35967, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
-   { 35986, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
-   { 36008, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
-   { 36027, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
-   { 36047, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
-   { 36072, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
-   { 36096, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
-   { 36116, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
-   { 36140, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
-   { 36160, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
-   { 36183, 0x000088F1 }, /* GL_TEXTURE_STENCIL_SIZE */
-   { 36207, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
-   { 36232, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-   { 36266, 0x00001000 }, /* GL_TEXTURE_WIDTH */
-   { 36283, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
-   { 36301, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
-   { 36319, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
-   { 36337, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
-   { 36357, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
-   { 36376, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-   { 36405, 0x00001000 }, /* GL_TRANSFORM_BIT */
-   { 36422, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
-   { 36448, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
-   { 36478, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-   { 36510, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-   { 36540, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
-   { 36574, 0x0000862C }, /* GL_TRANSPOSE_NV */
-   { 36590, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-   { 36621, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
-   { 36656, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-   { 36684, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
-   { 36716, 0x00000004 }, /* GL_TRIANGLES */
-   { 36729, 0x00000006 }, /* GL_TRIANGLE_FAN */
-   { 36745, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
-   { 36766, 0x00000005 }, /* GL_TRIANGLE_STRIP */
-   { 36784, 0x00000001 }, /* GL_TRUE */
-   { 36792, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
-   { 36812, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
-   { 36835, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
-   { 36855, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
-   { 36876, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
-   { 36898, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
-   { 36920, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
-   { 36940, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
-   { 36961, 0x00001401 }, /* GL_UNSIGNED_BYTE */
-   { 36978, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-   { 37005, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
-   { 37028, 0x00001405 }, /* GL_UNSIGNED_INT */
-   { 37044, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
-   { 37071, 0x000084FA }, /* GL_UNSIGNED_INT_24_8 */
-   { 37092, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
-   { 37116, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-   { 37147, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
-   { 37171, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-   { 37199, 0x00008C17 }, /* GL_UNSIGNED_NORMALIZED */
-   { 37222, 0x00001403 }, /* GL_UNSIGNED_SHORT */
-   { 37240, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-   { 37270, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-   { 37296, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-   { 37326, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-   { 37352, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
-   { 37376, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-   { 37404, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-   { 37432, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
-   { 37459, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-   { 37491, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
-   { 37522, 0x00008CA2 }, /* GL_UPPER_LEFT */
-   { 37536, 0x00002A20 }, /* GL_V2F */
-   { 37543, 0x00002A21 }, /* GL_V3F */
-   { 37550, 0x00008B83 }, /* GL_VALIDATE_STATUS */
-   { 37569, 0x00001F00 }, /* GL_VENDOR */
-   { 37579, 0x00001F02 }, /* GL_VERSION */
-   { 37590, 0x00008074 }, /* GL_VERTEX_ARRAY */
-   { 37606, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-   { 37636, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-   { 37667, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
-   { 37702, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
-   { 37726, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
-   { 37747, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
-   { 37770, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
-   { 37791, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-   { 37818, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-   { 37846, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-   { 37874, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-   { 37902, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-   { 37930, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-   { 37958, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-   { 37986, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-   { 38013, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-   { 38040, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-   { 38067, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-   { 38094, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-   { 38121, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-   { 38148, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-   { 38175, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-   { 38202, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-   { 38229, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-   { 38267, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
-   { 38309, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-   { 38340, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
-   { 38375, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-   { 38409, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
-   { 38447, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-   { 38478, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
-   { 38513, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-   { 38541, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
-   { 38573, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-   { 38603, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
-   { 38637, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-   { 38665, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
-   { 38697, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
-   { 38717, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
-   { 38739, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
-   { 38768, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
-   { 38789, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-   { 38818, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
-   { 38851, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
-   { 38883, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-   { 38910, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
-   { 38941, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
-   { 38971, 0x00008B31 }, /* GL_VERTEX_SHADER */
-   { 38988, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
-   { 39009, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
-   { 39036, 0x00000BA2 }, /* GL_VIEWPORT */
-   { 39048, 0x00000800 }, /* GL_VIEWPORT_BIT */
-   { 39064, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
-   { 39084, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-   { 39115, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
-   { 39150, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-   { 39178, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-   { 39203, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-   { 39230, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-   { 39255, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
-   { 39279, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
-   { 39298, 0x000088B9 }, /* GL_WRITE_ONLY */
-   { 39312, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
-   { 39330, 0x00001506 }, /* GL_XOR */
-   { 39337, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
-   { 39356, 0x00008757 }, /* GL_YCBCR_MESA */
-   { 39370, 0x00000000 }, /* GL_ZERO */
-   { 39378, 0x00000D16 }, /* GL_ZOOM_X */
-   { 39388, 0x00000D17 }, /* GL_ZOOM_Y */
+   {   779, 0x0000911A }, /* GL_ALREADY_SIGNALED */
+   {   799, 0x00000207 }, /* GL_ALWAYS */
+   {   809, 0x00001200 }, /* GL_AMBIENT */
+   {   820, 0x00001602 }, /* GL_AMBIENT_AND_DIFFUSE */
+   {   843, 0x00001501 }, /* GL_AND */
+   {   850, 0x00001504 }, /* GL_AND_INVERTED */
+   {   866, 0x00001502 }, /* GL_AND_REVERSE */
+   {   881, 0x00008892 }, /* GL_ARRAY_BUFFER */
+   {   897, 0x00008894 }, /* GL_ARRAY_BUFFER_BINDING */
+   {   921, 0x00008894 }, /* GL_ARRAY_BUFFER_BINDING_ARB */
+   {   949, 0x00008B85 }, /* GL_ATTACHED_SHADERS */
+   {   969, 0x00008645 }, /* GL_ATTRIB_ARRAY_POINTER_NV */
+   {   996, 0x00008623 }, /* GL_ATTRIB_ARRAY_SIZE_NV */
+   {  1020, 0x00008624 }, /* GL_ATTRIB_ARRAY_STRIDE_NV */
+   {  1046, 0x00008625 }, /* GL_ATTRIB_ARRAY_TYPE_NV */
+   {  1070, 0x00000BB0 }, /* GL_ATTRIB_STACK_DEPTH */
+   {  1092, 0x00000D80 }, /* GL_AUTO_NORMAL */
+   {  1107, 0x00000409 }, /* GL_AUX0 */
+   {  1115, 0x0000040A }, /* GL_AUX1 */
+   {  1123, 0x0000040B }, /* GL_AUX2 */
+   {  1131, 0x0000040C }, /* GL_AUX3 */
+   {  1139, 0x00000C00 }, /* GL_AUX_BUFFERS */
+   {  1154, 0x00000405 }, /* GL_BACK */
+   {  1162, 0x00000402 }, /* GL_BACK_LEFT */
+   {  1175, 0x00000403 }, /* GL_BACK_RIGHT */
+   {  1189, 0x000080E0 }, /* GL_BGR */
+   {  1196, 0x000080E1 }, /* GL_BGRA */
+   {  1204, 0x00001A00 }, /* GL_BITMAP */
+   {  1214, 0x00000704 }, /* GL_BITMAP_TOKEN */
+   {  1230, 0x00000BE2 }, /* GL_BLEND */
+   {  1239, 0x00008005 }, /* GL_BLEND_COLOR */
+   {  1254, 0x00008005 }, /* GL_BLEND_COLOR_EXT */
+   {  1273, 0x00000BE0 }, /* GL_BLEND_DST */
+   {  1286, 0x000080CA }, /* GL_BLEND_DST_ALPHA */
+   {  1305, 0x000080C8 }, /* GL_BLEND_DST_RGB */
+   {  1322, 0x00008009 }, /* GL_BLEND_EQUATION */
+   {  1340, 0x0000883D }, /* GL_BLEND_EQUATION_ALPHA */
+   {  1364, 0x0000883D }, /* GL_BLEND_EQUATION_ALPHA_EXT */
+   {  1392, 0x00008009 }, /* GL_BLEND_EQUATION_EXT */
+   {  1414, 0x00008009 }, /* GL_BLEND_EQUATION_RGB */
+   {  1436, 0x00008009 }, /* GL_BLEND_EQUATION_RGB_EXT */
+   {  1462, 0x00000BE1 }, /* GL_BLEND_SRC */
+   {  1475, 0x000080CB }, /* GL_BLEND_SRC_ALPHA */
+   {  1494, 0x000080C9 }, /* GL_BLEND_SRC_RGB */
+   {  1511, 0x00001905 }, /* GL_BLUE */
+   {  1519, 0x00000D1B }, /* GL_BLUE_BIAS */
+   {  1532, 0x00000D54 }, /* GL_BLUE_BITS */
+   {  1545, 0x00000D1A }, /* GL_BLUE_SCALE */
+   {  1559, 0x00008B56 }, /* GL_BOOL */
+   {  1567, 0x00008B56 }, /* GL_BOOL_ARB */
+   {  1579, 0x00008B57 }, /* GL_BOOL_VEC2 */
+   {  1592, 0x00008B57 }, /* GL_BOOL_VEC2_ARB */
+   {  1609, 0x00008B58 }, /* GL_BOOL_VEC3 */
+   {  1622, 0x00008B58 }, /* GL_BOOL_VEC3_ARB */
+   {  1639, 0x00008B59 }, /* GL_BOOL_VEC4 */
+   {  1652, 0x00008B59 }, /* GL_BOOL_VEC4_ARB */
+   {  1669, 0x000088BB }, /* GL_BUFFER_ACCESS */
+   {  1686, 0x000088BB }, /* GL_BUFFER_ACCESS_ARB */
+   {  1707, 0x00008A13 }, /* GL_BUFFER_FLUSHING_UNMAP_APPLE */
+   {  1738, 0x000088BC }, /* GL_BUFFER_MAPPED */
+   {  1755, 0x000088BC }, /* GL_BUFFER_MAPPED_ARB */
+   {  1776, 0x000088BD }, /* GL_BUFFER_MAP_POINTER */
+   {  1798, 0x000088BD }, /* GL_BUFFER_MAP_POINTER_ARB */
+   {  1824, 0x00008A12 }, /* GL_BUFFER_SERIALIZED_MODIFY_APPLE */
+   {  1858, 0x00008764 }, /* GL_BUFFER_SIZE */
+   {  1873, 0x00008764 }, /* GL_BUFFER_SIZE_ARB */
+   {  1892, 0x00008765 }, /* GL_BUFFER_USAGE */
+   {  1908, 0x00008765 }, /* GL_BUFFER_USAGE_ARB */
+   {  1928, 0x0000877B }, /* GL_BUMP_ENVMAP_ATI */
+   {  1947, 0x00008777 }, /* GL_BUMP_NUM_TEX_UNITS_ATI */
+   {  1973, 0x00008775 }, /* GL_BUMP_ROT_MATRIX_ATI */
+   {  1996, 0x00008776 }, /* GL_BUMP_ROT_MATRIX_SIZE_ATI */
+   {  2024, 0x0000877C }, /* GL_BUMP_TARGET_ATI */
+   {  2043, 0x00008778 }, /* GL_BUMP_TEX_UNITS_ATI */
+   {  2065, 0x00001400 }, /* GL_BYTE */
+   {  2073, 0x00002A24 }, /* GL_C3F_V3F */
+   {  2084, 0x00002A26 }, /* GL_C4F_N3F_V3F */
+   {  2099, 0x00002A22 }, /* GL_C4UB_V2F */
+   {  2111, 0x00002A23 }, /* GL_C4UB_V3F */
+   {  2123, 0x00000901 }, /* GL_CCW */
+   {  2130, 0x00002900 }, /* GL_CLAMP */
+   {  2139, 0x0000812D }, /* GL_CLAMP_TO_BORDER */
+   {  2158, 0x0000812D }, /* GL_CLAMP_TO_BORDER_ARB */
+   {  2181, 0x0000812D }, /* GL_CLAMP_TO_BORDER_SGIS */
+   {  2205, 0x0000812F }, /* GL_CLAMP_TO_EDGE */
+   {  2222, 0x0000812F }, /* GL_CLAMP_TO_EDGE_SGIS */
+   {  2244, 0x00001500 }, /* GL_CLEAR */
+   {  2253, 0x000084E1 }, /* GL_CLIENT_ACTIVE_TEXTURE */
+   {  2278, 0x000084E1 }, /* GL_CLIENT_ACTIVE_TEXTURE_ARB */
+   {  2307, 0xFFFFFFFF }, /* GL_CLIENT_ALL_ATTRIB_BITS */
+   {  2333, 0x00000BB1 }, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
+   {  2362, 0x00000001 }, /* GL_CLIENT_PIXEL_STORE_BIT */
+   {  2388, 0x00000002 }, /* GL_CLIENT_VERTEX_ARRAY_BIT */
+   {  2415, 0x00003000 }, /* GL_CLIP_PLANE0 */
+   {  2430, 0x00003001 }, /* GL_CLIP_PLANE1 */
+   {  2445, 0x00003002 }, /* GL_CLIP_PLANE2 */
+   {  2460, 0x00003003 }, /* GL_CLIP_PLANE3 */
+   {  2475, 0x00003004 }, /* GL_CLIP_PLANE4 */
+   {  2490, 0x00003005 }, /* GL_CLIP_PLANE5 */
+   {  2505, 0x000080F0 }, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
+   {  2538, 0x00000A00 }, /* GL_COEFF */
+   {  2547, 0x00001800 }, /* GL_COLOR */
+   {  2556, 0x00008076 }, /* GL_COLOR_ARRAY */
+   {  2571, 0x00008898 }, /* GL_COLOR_ARRAY_BUFFER_BINDING */
+   {  2601, 0x00008898 }, /* GL_COLOR_ARRAY_BUFFER_BINDING_ARB */
+   {  2635, 0x00008090 }, /* GL_COLOR_ARRAY_POINTER */
+   {  2658, 0x00008081 }, /* GL_COLOR_ARRAY_SIZE */
+   {  2678, 0x00008083 }, /* GL_COLOR_ARRAY_STRIDE */
+   {  2700, 0x00008082 }, /* GL_COLOR_ARRAY_TYPE */
+   {  2720, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0 */
+   {  2741, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0_EXT */
+   {  2766, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1 */
+   {  2787, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10 */
+   {  2809, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10_EXT */
+   {  2835, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11 */
+   {  2857, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11_EXT */
+   {  2883, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12 */
+   {  2905, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12_EXT */
+   {  2931, 0x00008CED }, /* GL_COLOR_ATTACHMENT13 */
+   {  2953, 0x00008CED }, /* GL_COLOR_ATTACHMENT13_EXT */
+   {  2979, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14 */
+   {  3001, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14_EXT */
+   {  3027, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15 */
+   {  3049, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15_EXT */
+   {  3075, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1_EXT */
+   {  3100, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2 */
+   {  3121, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2_EXT */
+   {  3146, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3 */
+   {  3167, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3_EXT */
+   {  3192, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4 */
+   {  3213, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4_EXT */
+   {  3238, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5 */
+   {  3259, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5_EXT */
+   {  3284, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6 */
+   {  3305, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6_EXT */
+   {  3330, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7 */
+   {  3351, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7_EXT */
+   {  3376, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8 */
+   {  3397, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8_EXT */
+   {  3422, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9 */
+   {  3443, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9_EXT */
+   {  3468, 0x00004000 }, /* GL_COLOR_BUFFER_BIT */
+   {  3488, 0x00000C22 }, /* GL_COLOR_CLEAR_VALUE */
+   {  3509, 0x00001900 }, /* GL_COLOR_INDEX */
+   {  3524, 0x00001603 }, /* GL_COLOR_INDEXES */
+   {  3541, 0x00000BF2 }, /* GL_COLOR_LOGIC_OP */
+   {  3559, 0x00000B57 }, /* GL_COLOR_MATERIAL */
+   {  3577, 0x00000B55 }, /* GL_COLOR_MATERIAL_FACE */
+   {  3600, 0x00000B56 }, /* GL_COLOR_MATERIAL_PARAMETER */
+   {  3628, 0x000080B1 }, /* GL_COLOR_MATRIX */
+   {  3644, 0x000080B1 }, /* GL_COLOR_MATRIX_SGI */
+   {  3664, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH */
+   {  3692, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH_SGI */
+   {  3724, 0x00008458 }, /* GL_COLOR_SUM */
+   {  3737, 0x00008458 }, /* GL_COLOR_SUM_ARB */
+   {  3754, 0x000080D0 }, /* GL_COLOR_TABLE */
+   {  3769, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE */
+   {  3795, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_EXT */
+   {  3825, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_SGI */
+   {  3855, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS */
+   {  3875, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS_SGI */
+   {  3899, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE */
+   {  3924, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_EXT */
+   {  3953, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_SGI */
+   {  3982, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT */
+   {  4004, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_EXT */
+   {  4030, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_SGI */
+   {  4056, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE */
+   {  4082, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_EXT */
+   {  4112, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_SGI */
+   {  4142, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE */
+   {  4172, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_EXT */
+   {  4206, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_SGI */
+   {  4240, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
+   {  4270, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_EXT */
+   {  4304, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_SGI */
+   {  4338, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE */
+   {  4362, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_EXT */
+   {  4390, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_SGI */
+   {  4418, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE */
+   {  4439, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE_SGI */
+   {  4464, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH */
+   {  4485, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_EXT */
+   {  4510, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_SGI */
+   {  4535, 0x00000C23 }, /* GL_COLOR_WRITEMASK */
+   {  4554, 0x00008570 }, /* GL_COMBINE */
+   {  4565, 0x00008503 }, /* GL_COMBINE4 */
+   {  4577, 0x00008572 }, /* GL_COMBINE_ALPHA */
+   {  4594, 0x00008572 }, /* GL_COMBINE_ALPHA_ARB */
+   {  4615, 0x00008572 }, /* GL_COMBINE_ALPHA_EXT */
+   {  4636, 0x00008570 }, /* GL_COMBINE_ARB */
+   {  4651, 0x00008570 }, /* GL_COMBINE_EXT */
+   {  4666, 0x00008571 }, /* GL_COMBINE_RGB */
+   {  4681, 0x00008571 }, /* GL_COMBINE_RGB_ARB */
+   {  4700, 0x00008571 }, /* GL_COMBINE_RGB_EXT */
+   {  4719, 0x0000884E }, /* GL_COMPARE_REF_DEPTH_TO_TEXTURE_EXT */
+   {  4755, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE */
+   {  4779, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE_ARB */
+   {  4807, 0x00001300 }, /* GL_COMPILE */
+   {  4818, 0x00001301 }, /* GL_COMPILE_AND_EXECUTE */
+   {  4841, 0x00008B81 }, /* GL_COMPILE_STATUS */
+   {  4859, 0x000084E9 }, /* GL_COMPRESSED_ALPHA */
+   {  4879, 0x000084E9 }, /* GL_COMPRESSED_ALPHA_ARB */
+   {  4903, 0x000084EC }, /* GL_COMPRESSED_INTENSITY */
+   {  4927, 0x000084EC }, /* GL_COMPRESSED_INTENSITY_ARB */
+   {  4955, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE */
+   {  4979, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA */
+   {  5009, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA_ARB */
+   {  5043, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE_ARB */
+   {  5071, 0x000084ED }, /* GL_COMPRESSED_RGB */
+   {  5089, 0x000084EE }, /* GL_COMPRESSED_RGBA */
+   {  5108, 0x000084EE }, /* GL_COMPRESSED_RGBA_ARB */
+   {  5131, 0x000086B1 }, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
+   {  5160, 0x000083F1 }, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
+   {  5193, 0x000083F2 }, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
+   {  5226, 0x000083F3 }, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
+   {  5259, 0x000084ED }, /* GL_COMPRESSED_RGB_ARB */
+   {  5281, 0x000086B0 }, /* GL_COMPRESSED_RGB_FXT1_3DFX */
+   {  5309, 0x000083F0 }, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
+   {  5341, 0x00008C4A }, /* GL_COMPRESSED_SLUMINANCE */
+   {  5366, 0x00008C4B }, /* GL_COMPRESSED_SLUMINANCE_ALPHA */
+   {  5397, 0x00008C48 }, /* GL_COMPRESSED_SRGB */
+   {  5416, 0x00008C49 }, /* GL_COMPRESSED_SRGB_ALPHA */
+   {  5441, 0x000086A3 }, /* GL_COMPRESSED_TEXTURE_FORMATS */
+   {  5471, 0x0000911C }, /* GL_CONDITION_SATISFIED */
+   {  5494, 0x00008576 }, /* GL_CONSTANT */
+   {  5506, 0x00008003 }, /* GL_CONSTANT_ALPHA */
+   {  5524, 0x00008003 }, /* GL_CONSTANT_ALPHA_EXT */
+   {  5546, 0x00008576 }, /* GL_CONSTANT_ARB */
+   {  5562, 0x00001207 }, /* GL_CONSTANT_ATTENUATION */
+   {  5586, 0x00008151 }, /* GL_CONSTANT_BORDER_HP */
+   {  5608, 0x00008001 }, /* GL_CONSTANT_COLOR */
+   {  5626, 0x00008001 }, /* GL_CONSTANT_COLOR_EXT */
+   {  5648, 0x00008576 }, /* GL_CONSTANT_EXT */
+   {  5664, 0x00008010 }, /* GL_CONVOLUTION_1D */
+   {  5682, 0x00008011 }, /* GL_CONVOLUTION_2D */
+   {  5700, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR */
+   {  5728, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR_HP */
+   {  5759, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE */
+   {  5786, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE_EXT */
+   {  5817, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS */
+   {  5844, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS_EXT */
+   {  5875, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE */
+   {  5903, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE_EXT */
+   {  5935, 0x00008017 }, /* GL_CONVOLUTION_FORMAT */
+   {  5957, 0x00008017 }, /* GL_CONVOLUTION_FORMAT_EXT */
+   {  5983, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT */
+   {  6005, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT_EXT */
+   {  6031, 0x00008018 }, /* GL_CONVOLUTION_WIDTH */
+   {  6052, 0x00008018 }, /* GL_CONVOLUTION_WIDTH_EXT */
+   {  6077, 0x00008862 }, /* GL_COORD_REPLACE */
+   {  6094, 0x00008862 }, /* GL_COORD_REPLACE_ARB */
+   {  6115, 0x00008862 }, /* GL_COORD_REPLACE_NV */
+   {  6135, 0x00001503 }, /* GL_COPY */
+   {  6143, 0x0000150C }, /* GL_COPY_INVERTED */
+   {  6160, 0x00000706 }, /* GL_COPY_PIXEL_TOKEN */
+   {  6180, 0x00008F36 }, /* GL_COPY_READ_BUFFER */
+   {  6200, 0x00008F37 }, /* GL_COPY_WRITE_BUFFER */
+   {  6221, 0x00000B44 }, /* GL_CULL_FACE */
+   {  6234, 0x00000B45 }, /* GL_CULL_FACE_MODE */
+   {  6252, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
+   {  6271, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+   {  6303, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+   {  6338, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
+   {  6359, 0x00000001 }, /* GL_CURRENT_BIT */
+   {  6374, 0x00000B00 }, /* GL_CURRENT_COLOR */
+   {  6391, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
+   {  6412, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
+   {  6438, 0x00000B01 }, /* GL_CURRENT_INDEX */
+   {  6455, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
+   {  6477, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
+   {  6505, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
+   {  6526, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+   {  6560, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
+   {  6593, 0x00000B02 }, /* GL_CURRENT_NORMAL */
+   {  6611, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+   {  6641, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
+   {  6660, 0x00008865 }, /* GL_CURRENT_QUERY */
+   {  6677, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
+   {  6698, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
+   {  6722, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
+   {  6749, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
+   {  6773, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
+   {  6800, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
+   {  6833, 0x0000845F }, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
+   {  6867, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+   {  6900, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
+   {  6927, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
+   {  6953, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
+   {  6978, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
+   {  7007, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
+   {  7029, 0x00000900 }, /* GL_CW */
+   {  7035, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
+   {  7056, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
+   {  7077, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
+   {  7097, 0x00002101 }, /* GL_DECAL */
+   {  7106, 0x00001E03 }, /* GL_DECR */
+   {  7114, 0x00008508 }, /* GL_DECR_WRAP */
+   {  7127, 0x00008508 }, /* GL_DECR_WRAP_EXT */
+   {  7144, 0x00008B80 }, /* GL_DELETE_STATUS */
+   {  7161, 0x00001801 }, /* GL_DEPTH */
+   {  7170, 0x000088F0 }, /* GL_DEPTH24_STENCIL8 */
+   {  7190, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT */
+   {  7210, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
+   {  7234, 0x00000D1F }, /* GL_DEPTH_BIAS */
+   {  7248, 0x00000D56 }, /* GL_DEPTH_BITS */
+   {  7262, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
+   {  7282, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
+   {  7307, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
+   {  7327, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
+   {  7345, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
+   {  7366, 0x00001902 }, /* GL_DEPTH_COMPONENT */
+   {  7385, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
+   {  7406, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
+   {  7431, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
+   {  7457, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
+   {  7478, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
+   {  7503, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
+   {  7529, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
+   {  7550, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
+   {  7575, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
+   {  7601, 0x00000B74 }, /* GL_DEPTH_FUNC */
+   {  7615, 0x00000B70 }, /* GL_DEPTH_RANGE */
+   {  7630, 0x00000D1E }, /* GL_DEPTH_SCALE */
+   {  7645, 0x000084F9 }, /* GL_DEPTH_STENCIL */
+   {  7662, 0x0000821A }, /* GL_DEPTH_STENCIL_ATTACHMENT */
+   {  7690, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
+   {  7710, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+   {  7738, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+   {  7766, 0x00000B71 }, /* GL_DEPTH_TEST */
+   {  7780, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
+   {  7802, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
+   {  7828, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
+   {  7847, 0x00001201 }, /* GL_DIFFUSE */
+   {  7858, 0x00000BD0 }, /* GL_DITHER */
+   {  7868, 0x00000A02 }, /* GL_DOMAIN */
+   {  7878, 0x00001100 }, /* GL_DONT_CARE */
+   {  7891, 0x000086AE }, /* GL_DOT3_RGB */
+   {  7903, 0x000086AF }, /* GL_DOT3_RGBA */
+   {  7916, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
+   {  7933, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
+   {  7950, 0x000086AE }, /* GL_DOT3_RGB_ARB */
+   {  7966, 0x00008740 }, /* GL_DOT3_RGB_EXT */
+   {  7982, 0x0000140A }, /* GL_DOUBLE */
+   {  7992, 0x00000C32 }, /* GL_DOUBLEBUFFER */
+   {  8008, 0x00000C01 }, /* GL_DRAW_BUFFER */
+   {  8023, 0x00008825 }, /* GL_DRAW_BUFFER0 */
+   {  8039, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
+   {  8059, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
+   {  8079, 0x00008826 }, /* GL_DRAW_BUFFER1 */
+   {  8095, 0x0000882F }, /* GL_DRAW_BUFFER10 */
+   {  8112, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
+   {  8133, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
+   {  8154, 0x00008830 }, /* GL_DRAW_BUFFER11 */
+   {  8171, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
+   {  8192, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
+   {  8213, 0x00008831 }, /* GL_DRAW_BUFFER12 */
+   {  8230, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
+   {  8251, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
+   {  8272, 0x00008832 }, /* GL_DRAW_BUFFER13 */
+   {  8289, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
+   {  8310, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
+   {  8331, 0x00008833 }, /* GL_DRAW_BUFFER14 */
+   {  8348, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
+   {  8369, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
+   {  8390, 0x00008834 }, /* GL_DRAW_BUFFER15 */
+   {  8407, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
+   {  8428, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
+   {  8449, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
+   {  8469, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
+   {  8489, 0x00008827 }, /* GL_DRAW_BUFFER2 */
+   {  8505, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
+   {  8525, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
+   {  8545, 0x00008828 }, /* GL_DRAW_BUFFER3 */
+   {  8561, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
+   {  8581, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
+   {  8601, 0x00008829 }, /* GL_DRAW_BUFFER4 */
+   {  8617, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
+   {  8637, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
+   {  8657, 0x0000882A }, /* GL_DRAW_BUFFER5 */
+   {  8673, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
+   {  8693, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
+   {  8713, 0x0000882B }, /* GL_DRAW_BUFFER6 */
+   {  8729, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
+   {  8749, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
+   {  8769, 0x0000882C }, /* GL_DRAW_BUFFER7 */
+   {  8785, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
+   {  8805, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
+   {  8825, 0x0000882D }, /* GL_DRAW_BUFFER8 */
+   {  8841, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
+   {  8861, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
+   {  8881, 0x0000882E }, /* GL_DRAW_BUFFER9 */
+   {  8897, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
+   {  8917, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
+   {  8937, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER */
+   {  8957, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+   {  8989, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
+   {  9013, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
+   {  9033, 0x00000304 }, /* GL_DST_ALPHA */
+   {  9046, 0x00000306 }, /* GL_DST_COLOR */
+   {  9059, 0x0000877A }, /* GL_DU8DV8_ATI */
+   {  9073, 0x00008779 }, /* GL_DUDV_ATI */
+   {  9085, 0x000088EA }, /* GL_DYNAMIC_COPY */
+   {  9101, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
+   {  9121, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
+   {  9137, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
+   {  9157, 0x000088E9 }, /* GL_DYNAMIC_READ */
+   {  9173, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
+   {  9193, 0x00000B43 }, /* GL_EDGE_FLAG */
+   {  9206, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
+   {  9225, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+   {  9259, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
+   {  9297, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
+   {  9324, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+   {  9350, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
+   {  9374, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+   {  9406, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
+   {  9442, 0x00001600 }, /* GL_EMISSION */
+   {  9454, 0x00002000 }, /* GL_ENABLE_BIT */
+   {  9468, 0x00000202 }, /* GL_EQUAL */
+   {  9477, 0x00001509 }, /* GL_EQUIV */
+   {  9486, 0x00010000 }, /* GL_EVAL_BIT */
+   {  9498, 0x00000800 }, /* GL_EXP */
+   {  9505, 0x00000801 }, /* GL_EXP2 */
+   {  9513, 0x00001F03 }, /* GL_EXTENSIONS */
+   {  9527, 0x00002400 }, /* GL_EYE_LINEAR */
+   {  9541, 0x00002502 }, /* GL_EYE_PLANE */
+   {  9554, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
+   {  9579, 0x0000855B }, /* GL_EYE_RADIAL_NV */
+   {  9596, 0x00000000 }, /* GL_FALSE */
+   {  9605, 0x00001101 }, /* GL_FASTEST */
+   {  9616, 0x00001C01 }, /* GL_FEEDBACK */
+   {  9628, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
+   {  9655, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
+   {  9679, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
+   {  9703, 0x00001B02 }, /* GL_FILL */
+   {  9711, 0x00008E4D }, /* GL_FIRST_VERTEX_CONVENTION_EXT */
+   {  9742, 0x00001D00 }, /* GL_FLAT */
+   {  9750, 0x00001406 }, /* GL_FLOAT */
+   {  9759, 0x00008B5A }, /* GL_FLOAT_MAT2 */
+   {  9773, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
+   {  9791, 0x00008B65 }, /* GL_FLOAT_MAT2x3 */
+   {  9807, 0x00008B66 }, /* GL_FLOAT_MAT2x4 */
+   {  9823, 0x00008B5B }, /* GL_FLOAT_MAT3 */
+   {  9837, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
+   {  9855, 0x00008B67 }, /* GL_FLOAT_MAT3x2 */
+   {  9871, 0x00008B68 }, /* GL_FLOAT_MAT3x4 */
+   {  9887, 0x00008B5C }, /* GL_FLOAT_MAT4 */
+   {  9901, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
+   {  9919, 0x00008B69 }, /* GL_FLOAT_MAT4x2 */
+   {  9935, 0x00008B6A }, /* GL_FLOAT_MAT4x3 */
+   {  9951, 0x00008B50 }, /* GL_FLOAT_VEC2 */
+   {  9965, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
+   {  9983, 0x00008B51 }, /* GL_FLOAT_VEC3 */
+   {  9997, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
+   { 10015, 0x00008B52 }, /* GL_FLOAT_VEC4 */
+   { 10029, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
+   { 10047, 0x00000B60 }, /* GL_FOG */
+   { 10054, 0x00000080 }, /* GL_FOG_BIT */
+   { 10065, 0x00000B66 }, /* GL_FOG_COLOR */
+   { 10078, 0x00008451 }, /* GL_FOG_COORD */
+   { 10091, 0x00008451 }, /* GL_FOG_COORDINATE */
+   { 10109, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
+   { 10133, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+   { 10172, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
+   { 10215, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+   { 10247, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+   { 10278, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+   { 10307, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
+   { 10332, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
+   { 10351, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
+   { 10385, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
+   { 10412, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
+   { 10438, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
+   { 10462, 0x00008450 }, /* GL_FOG_COORD_SRC */
+   { 10479, 0x00000B62 }, /* GL_FOG_DENSITY */
+   { 10494, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
+   { 10518, 0x00000B64 }, /* GL_FOG_END */
+   { 10529, 0x00000C54 }, /* GL_FOG_HINT */
+   { 10541, 0x00000B61 }, /* GL_FOG_INDEX */
+   { 10554, 0x00000B65 }, /* GL_FOG_MODE */
+   { 10566, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
+   { 10585, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
+   { 10610, 0x00000B63 }, /* GL_FOG_START */
+   { 10623, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
+   { 10641, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
+   { 10665, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
+   { 10684, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
+   { 10707, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+   { 10742, 0x00008D40 }, /* GL_FRAMEBUFFER */
+   { 10757, 0x00008215 }, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+   { 10794, 0x00008214 }, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+   { 10830, 0x00008210 }, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+   { 10871, 0x00008211 }, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+   { 10912, 0x00008216 }, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+   { 10949, 0x00008213 }, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+   { 10986, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+   { 11024, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
+   { 11066, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+   { 11104, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
+   { 11146, 0x00008212 }, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+   { 11181, 0x00008217 }, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+   { 11220, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
+   { 11269, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+   { 11317, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
+   { 11369, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+   { 11409, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
+   { 11453, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+   { 11493, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
+   { 11537, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
+   { 11564, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE */
+   { 11588, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
+   { 11616, 0x00008218 }, /* GL_FRAMEBUFFER_DEFAULT */
+   { 11639, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
+   { 11658, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+   { 11695, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
+   { 11736, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+   { 11777, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+   { 11819, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+   { 11870, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+   { 11908, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+   { 11953, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
+   { 12002, 0x00008D56 }, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+   { 12040, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+   { 12082, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+   { 12114, 0x00008219 }, /* GL_FRAMEBUFFER_UNDEFINED */
+   { 12139, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED */
+   { 12166, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
+   { 12197, 0x00000404 }, /* GL_FRONT */
+   { 12206, 0x00000408 }, /* GL_FRONT_AND_BACK */
+   { 12224, 0x00000B46 }, /* GL_FRONT_FACE */
+   { 12238, 0x00000400 }, /* GL_FRONT_LEFT */
+   { 12252, 0x00000401 }, /* GL_FRONT_RIGHT */
+   { 12267, 0x00008006 }, /* GL_FUNC_ADD */
+   { 12279, 0x00008006 }, /* GL_FUNC_ADD_EXT */
+   { 12295, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
+   { 12320, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
+   { 12349, 0x0000800A }, /* GL_FUNC_SUBTRACT */
+   { 12366, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
+   { 12387, 0x00008191 }, /* GL_GENERATE_MIPMAP */
+   { 12406, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
+   { 12430, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
+   { 12459, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
+   { 12483, 0x00000206 }, /* GL_GEQUAL */
+   { 12493, 0x00000204 }, /* GL_GREATER */
+   { 12504, 0x00001904 }, /* GL_GREEN */
+   { 12513, 0x00000D19 }, /* GL_GREEN_BIAS */
+   { 12527, 0x00000D53 }, /* GL_GREEN_BITS */
+   { 12541, 0x00000D18 }, /* GL_GREEN_SCALE */
+   { 12556, 0x00008000 }, /* GL_HINT_BIT */
+   { 12568, 0x00008024 }, /* GL_HISTOGRAM */
+   { 12581, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
+   { 12605, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
+   { 12633, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
+   { 12656, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
+   { 12683, 0x00008024 }, /* GL_HISTOGRAM_EXT */
+   { 12700, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
+   { 12720, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
+   { 12744, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
+   { 12768, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
+   { 12796, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+   { 12824, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
+   { 12856, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
+   { 12878, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
+   { 12904, 0x0000802D }, /* GL_HISTOGRAM_SINK */
+   { 12922, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
+   { 12944, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
+   { 12963, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
+   { 12986, 0x0000862A }, /* GL_IDENTITY_NV */
+   { 13001, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
+   { 13021, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+   { 13061, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+   { 13099, 0x00001E02 }, /* GL_INCR */
+   { 13107, 0x00008507 }, /* GL_INCR_WRAP */
+   { 13120, 0x00008507 }, /* GL_INCR_WRAP_EXT */
+   { 13137, 0x00008222 }, /* GL_INDEX */
+   { 13146, 0x00008077 }, /* GL_INDEX_ARRAY */
+   { 13161, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+   { 13191, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
+   { 13225, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
+   { 13248, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
+   { 13270, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
+   { 13290, 0x00000D51 }, /* GL_INDEX_BITS */
+   { 13304, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
+   { 13325, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
+   { 13343, 0x00000C30 }, /* GL_INDEX_MODE */
+   { 13357, 0x00000D13 }, /* GL_INDEX_OFFSET */
+   { 13373, 0x00000D12 }, /* GL_INDEX_SHIFT */
+   { 13388, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
+   { 13407, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
+   { 13426, 0x00001404 }, /* GL_INT */
+   { 13433, 0x00008049 }, /* GL_INTENSITY */
+   { 13446, 0x0000804C }, /* GL_INTENSITY12 */
+   { 13461, 0x0000804C }, /* GL_INTENSITY12_EXT */
+   { 13480, 0x0000804D }, /* GL_INTENSITY16 */
+   { 13495, 0x0000804D }, /* GL_INTENSITY16_EXT */
+   { 13514, 0x0000804A }, /* GL_INTENSITY4 */
+   { 13528, 0x0000804A }, /* GL_INTENSITY4_EXT */
+   { 13546, 0x0000804B }, /* GL_INTENSITY8 */
+   { 13560, 0x0000804B }, /* GL_INTENSITY8_EXT */
+   { 13578, 0x00008049 }, /* GL_INTENSITY_EXT */
+   { 13595, 0x00008575 }, /* GL_INTERPOLATE */
+   { 13610, 0x00008575 }, /* GL_INTERPOLATE_ARB */
+   { 13629, 0x00008575 }, /* GL_INTERPOLATE_EXT */
+   { 13648, 0x00008B53 }, /* GL_INT_VEC2 */
+   { 13660, 0x00008B53 }, /* GL_INT_VEC2_ARB */
+   { 13676, 0x00008B54 }, /* GL_INT_VEC3 */
+   { 13688, 0x00008B54 }, /* GL_INT_VEC3_ARB */
+   { 13704, 0x00008B55 }, /* GL_INT_VEC4 */
+   { 13716, 0x00008B55 }, /* GL_INT_VEC4_ARB */
+   { 13732, 0x00000500 }, /* GL_INVALID_ENUM */
+   { 13748, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION */
+   { 13781, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
+   { 13818, 0x00000502 }, /* GL_INVALID_OPERATION */
+   { 13839, 0x00000501 }, /* GL_INVALID_VALUE */
+   { 13856, 0x0000862B }, /* GL_INVERSE_NV */
+   { 13870, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
+   { 13894, 0x0000150A }, /* GL_INVERT */
+   { 13904, 0x00001E00 }, /* GL_KEEP */
+   { 13912, 0x00008E4E }, /* GL_LAST_VERTEX_CONVENTION_EXT */
+   { 13942, 0x00000406 }, /* GL_LEFT */
+   { 13950, 0x00000203 }, /* GL_LEQUAL */
+   { 13960, 0x00000201 }, /* GL_LESS */
+   { 13968, 0x00004000 }, /* GL_LIGHT0 */
+   { 13978, 0x00004001 }, /* GL_LIGHT1 */
+   { 13988, 0x00004002 }, /* GL_LIGHT2 */
+   { 13998, 0x00004003 }, /* GL_LIGHT3 */
+   { 14008, 0x00004004 }, /* GL_LIGHT4 */
+   { 14018, 0x00004005 }, /* GL_LIGHT5 */
+   { 14028, 0x00004006 }, /* GL_LIGHT6 */
+   { 14038, 0x00004007 }, /* GL_LIGHT7 */
+   { 14048, 0x00000B50 }, /* GL_LIGHTING */
+   { 14060, 0x00000040 }, /* GL_LIGHTING_BIT */
+   { 14076, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
+   { 14099, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+   { 14128, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
+   { 14161, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+   { 14189, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
+   { 14213, 0x00001B01 }, /* GL_LINE */
+   { 14221, 0x00002601 }, /* GL_LINEAR */
+   { 14231, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
+   { 14253, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+   { 14283, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+   { 14314, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
+   { 14338, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
+   { 14363, 0x00000001 }, /* GL_LINES */
+   { 14372, 0x00000004 }, /* GL_LINE_BIT */
+   { 14384, 0x00000002 }, /* GL_LINE_LOOP */
+   { 14397, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
+   { 14417, 0x00000B20 }, /* GL_LINE_SMOOTH */
+   { 14432, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
+   { 14452, 0x00000B24 }, /* GL_LINE_STIPPLE */
+   { 14468, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
+   { 14492, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
+   { 14515, 0x00000003 }, /* GL_LINE_STRIP */
+   { 14529, 0x00000702 }, /* GL_LINE_TOKEN */
+   { 14543, 0x00000B21 }, /* GL_LINE_WIDTH */
+   { 14557, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
+   { 14583, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
+   { 14603, 0x00008B82 }, /* GL_LINK_STATUS */
+   { 14618, 0x00000B32 }, /* GL_LIST_BASE */
+   { 14631, 0x00020000 }, /* GL_LIST_BIT */
+   { 14643, 0x00000B33 }, /* GL_LIST_INDEX */
+   { 14657, 0x00000B30 }, /* GL_LIST_MODE */
+   { 14670, 0x00000101 }, /* GL_LOAD */
+   { 14678, 0x00000BF1 }, /* GL_LOGIC_OP */
+   { 14690, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
+   { 14707, 0x00008CA1 }, /* GL_LOWER_LEFT */
+   { 14721, 0x00001909 }, /* GL_LUMINANCE */
+   { 14734, 0x00008041 }, /* GL_LUMINANCE12 */
+   { 14749, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
+   { 14772, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
+   { 14799, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
+   { 14821, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
+   { 14847, 0x00008041 }, /* GL_LUMINANCE12_EXT */
+   { 14866, 0x00008042 }, /* GL_LUMINANCE16 */
+   { 14881, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
+   { 14904, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
+   { 14931, 0x00008042 }, /* GL_LUMINANCE16_EXT */
+   { 14950, 0x0000803F }, /* GL_LUMINANCE4 */
+   { 14964, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
+   { 14985, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
+   { 15010, 0x0000803F }, /* GL_LUMINANCE4_EXT */
+   { 15028, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
+   { 15049, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
+   { 15074, 0x00008040 }, /* GL_LUMINANCE8 */
+   { 15088, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
+   { 15109, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
+   { 15134, 0x00008040 }, /* GL_LUMINANCE8_EXT */
+   { 15152, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
+   { 15171, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
+   { 15187, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
+   { 15207, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
+   { 15229, 0x00000D91 }, /* GL_MAP1_INDEX */
+   { 15243, 0x00000D92 }, /* GL_MAP1_NORMAL */
+   { 15258, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
+   { 15282, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
+   { 15306, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
+   { 15330, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
+   { 15354, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
+   { 15371, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
+   { 15388, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+   { 15416, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+   { 15445, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+   { 15474, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+   { 15503, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+   { 15532, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+   { 15561, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+   { 15590, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+   { 15618, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+   { 15646, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+   { 15674, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+   { 15702, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+   { 15730, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+   { 15758, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+   { 15786, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+   { 15814, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+   { 15842, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
+   { 15858, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
+   { 15878, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
+   { 15900, 0x00000DB1 }, /* GL_MAP2_INDEX */
+   { 15914, 0x00000DB2 }, /* GL_MAP2_NORMAL */
+   { 15929, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
+   { 15953, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
+   { 15977, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
+   { 16001, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
+   { 16025, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
+   { 16042, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
+   { 16059, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+   { 16087, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+   { 16116, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+   { 16145, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+   { 16174, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+   { 16203, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+   { 16232, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+   { 16261, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+   { 16289, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+   { 16317, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+   { 16345, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+   { 16373, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+   { 16401, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+   { 16429, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
+   { 16457, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+   { 16485, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+   { 16513, 0x00000D10 }, /* GL_MAP_COLOR */
+   { 16526, 0x00000010 }, /* GL_MAP_FLUSH_EXPLICIT_BIT */
+   { 16552, 0x00000008 }, /* GL_MAP_INVALIDATE_BUFFER_BIT */
+   { 16581, 0x00000004 }, /* GL_MAP_INVALIDATE_RANGE_BIT */
+   { 16609, 0x00000001 }, /* GL_MAP_READ_BIT */
+   { 16625, 0x00000D11 }, /* GL_MAP_STENCIL */
+   { 16640, 0x00000020 }, /* GL_MAP_UNSYNCHRONIZED_BIT */
+   { 16666, 0x00000002 }, /* GL_MAP_WRITE_BIT */
+   { 16683, 0x000088C0 }, /* GL_MATRIX0_ARB */
+   { 16698, 0x00008630 }, /* GL_MATRIX0_NV */
+   { 16712, 0x000088CA }, /* GL_MATRIX10_ARB */
+   { 16728, 0x000088CB }, /* GL_MATRIX11_ARB */
+   { 16744, 0x000088CC }, /* GL_MATRIX12_ARB */
+   { 16760, 0x000088CD }, /* GL_MATRIX13_ARB */
+   { 16776, 0x000088CE }, /* GL_MATRIX14_ARB */
+   { 16792, 0x000088CF }, /* GL_MATRIX15_ARB */
+   { 16808, 0x000088D0 }, /* GL_MATRIX16_ARB */
+   { 16824, 0x000088D1 }, /* GL_MATRIX17_ARB */
+   { 16840, 0x000088D2 }, /* GL_MATRIX18_ARB */
+   { 16856, 0x000088D3 }, /* GL_MATRIX19_ARB */
+   { 16872, 0x000088C1 }, /* GL_MATRIX1_ARB */
+   { 16887, 0x00008631 }, /* GL_MATRIX1_NV */
+   { 16901, 0x000088D4 }, /* GL_MATRIX20_ARB */
+   { 16917, 0x000088D5 }, /* GL_MATRIX21_ARB */
+   { 16933, 0x000088D6 }, /* GL_MATRIX22_ARB */
+   { 16949, 0x000088D7 }, /* GL_MATRIX23_ARB */
+   { 16965, 0x000088D8 }, /* GL_MATRIX24_ARB */
+   { 16981, 0x000088D9 }, /* GL_MATRIX25_ARB */
+   { 16997, 0x000088DA }, /* GL_MATRIX26_ARB */
+   { 17013, 0x000088DB }, /* GL_MATRIX27_ARB */
+   { 17029, 0x000088DC }, /* GL_MATRIX28_ARB */
+   { 17045, 0x000088DD }, /* GL_MATRIX29_ARB */
+   { 17061, 0x000088C2 }, /* GL_MATRIX2_ARB */
+   { 17076, 0x00008632 }, /* GL_MATRIX2_NV */
+   { 17090, 0x000088DE }, /* GL_MATRIX30_ARB */
+   { 17106, 0x000088DF }, /* GL_MATRIX31_ARB */
+   { 17122, 0x000088C3 }, /* GL_MATRIX3_ARB */
+   { 17137, 0x00008633 }, /* GL_MATRIX3_NV */
+   { 17151, 0x000088C4 }, /* GL_MATRIX4_ARB */
+   { 17166, 0x00008634 }, /* GL_MATRIX4_NV */
+   { 17180, 0x000088C5 }, /* GL_MATRIX5_ARB */
+   { 17195, 0x00008635 }, /* GL_MATRIX5_NV */
+   { 17209, 0x000088C6 }, /* GL_MATRIX6_ARB */
+   { 17224, 0x00008636 }, /* GL_MATRIX6_NV */
+   { 17238, 0x000088C7 }, /* GL_MATRIX7_ARB */
+   { 17253, 0x00008637 }, /* GL_MATRIX7_NV */
+   { 17267, 0x000088C8 }, /* GL_MATRIX8_ARB */
+   { 17282, 0x000088C9 }, /* GL_MATRIX9_ARB */
+   { 17297, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
+   { 17323, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+   { 17357, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+   { 17388, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+   { 17421, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+   { 17452, 0x00000BA0 }, /* GL_MATRIX_MODE */
+   { 17467, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
+   { 17489, 0x00008008 }, /* GL_MAX */
+   { 17496, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
+   { 17519, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+   { 17551, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
+   { 17577, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+   { 17610, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+   { 17636, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 17670, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
+   { 17689, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+   { 17718, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+   { 17750, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
+   { 17786, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+   { 17822, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
+   { 17862, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
+   { 17888, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
+   { 17918, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
+   { 17943, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
+   { 17972, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+   { 18001, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
+   { 18034, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
+   { 18054, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
+   { 18078, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
+   { 18102, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
+   { 18126, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
+   { 18151, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
+   { 18169, 0x00008008 }, /* GL_MAX_EXT */
+   { 18180, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+   { 18215, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
+   { 18254, 0x00000D31 }, /* GL_MAX_LIGHTS */
+   { 18268, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
+   { 18288, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+   { 18326, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+   { 18355, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
+   { 18379, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
+   { 18407, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
+   { 18430, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 18467, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 18503, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+   { 18530, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+   { 18559, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+   { 18593, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
+   { 18629, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+   { 18656, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+   { 18688, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+   { 18724, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+   { 18753, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+   { 18782, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
+   { 18810, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+   { 18848, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 18892, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 18935, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 18969, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 19008, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 19045, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 19083, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 19126, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 19169, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+   { 19199, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+   { 19230, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 19266, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 19302, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
+   { 19332, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+   { 19366, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
+   { 19399, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+   { 19428, 0x00008D57 }, /* GL_MAX_SAMPLES */
+   { 19443, 0x00009111 }, /* GL_MAX_SERVER_WAIT_TIMEOUT */
+   { 19470, 0x00008504 }, /* GL_MAX_SHININESS_NV */
+   { 19490, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
+   { 19514, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
+   { 19536, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
+   { 19562, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+   { 19589, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19620, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
+   { 19644, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 19678, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
+   { 19698, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
+   { 19725, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
+   { 19746, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
+   { 19771, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
+   { 19796, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
+   { 19831, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
+   { 19853, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
+   { 19879, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
+   { 19901, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
+   { 19927, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+   { 19961, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19999, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+   { 20032, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
+   { 20069, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
+   { 20093, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
+   { 20114, 0x00008007 }, /* GL_MIN */
+   { 20121, 0x0000802E }, /* GL_MINMAX */
+   { 20131, 0x0000802E }, /* GL_MINMAX_EXT */
+   { 20145, 0x0000802F }, /* GL_MINMAX_FORMAT */
+   { 20162, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
+   { 20183, 0x00008030 }, /* GL_MINMAX_SINK */
+   { 20198, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
+   { 20217, 0x00008007 }, /* GL_MIN_EXT */
+   { 20228, 0x00008370 }, /* GL_MIRRORED_REPEAT */
+   { 20247, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
+   { 20270, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
+   { 20293, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
+   { 20313, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
+   { 20333, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+   { 20363, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
+   { 20391, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+   { 20419, 0x00001700 }, /* GL_MODELVIEW */
+   { 20432, 0x00001700 }, /* GL_MODELVIEW0_ARB */
+   { 20450, 0x0000872A }, /* GL_MODELVIEW10_ARB */
+   { 20469, 0x0000872B }, /* GL_MODELVIEW11_ARB */
+   { 20488, 0x0000872C }, /* GL_MODELVIEW12_ARB */
+   { 20507, 0x0000872D }, /* GL_MODELVIEW13_ARB */
+   { 20526, 0x0000872E }, /* GL_MODELVIEW14_ARB */
+   { 20545, 0x0000872F }, /* GL_MODELVIEW15_ARB */
+   { 20564, 0x00008730 }, /* GL_MODELVIEW16_ARB */
+   { 20583, 0x00008731 }, /* GL_MODELVIEW17_ARB */
+   { 20602, 0x00008732 }, /* GL_MODELVIEW18_ARB */
+   { 20621, 0x00008733 }, /* GL_MODELVIEW19_ARB */
+   { 20640, 0x0000850A }, /* GL_MODELVIEW1_ARB */
+   { 20658, 0x00008734 }, /* GL_MODELVIEW20_ARB */
+   { 20677, 0x00008735 }, /* GL_MODELVIEW21_ARB */
+   { 20696, 0x00008736 }, /* GL_MODELVIEW22_ARB */
+   { 20715, 0x00008737 }, /* GL_MODELVIEW23_ARB */
+   { 20734, 0x00008738 }, /* GL_MODELVIEW24_ARB */
+   { 20753, 0x00008739 }, /* GL_MODELVIEW25_ARB */
+   { 20772, 0x0000873A }, /* GL_MODELVIEW26_ARB */
+   { 20791, 0x0000873B }, /* GL_MODELVIEW27_ARB */
+   { 20810, 0x0000873C }, /* GL_MODELVIEW28_ARB */
+   { 20829, 0x0000873D }, /* GL_MODELVIEW29_ARB */
+   { 20848, 0x00008722 }, /* GL_MODELVIEW2_ARB */
+   { 20866, 0x0000873E }, /* GL_MODELVIEW30_ARB */
+   { 20885, 0x0000873F }, /* GL_MODELVIEW31_ARB */
+   { 20904, 0x00008723 }, /* GL_MODELVIEW3_ARB */
+   { 20922, 0x00008724 }, /* GL_MODELVIEW4_ARB */
+   { 20940, 0x00008725 }, /* GL_MODELVIEW5_ARB */
+   { 20958, 0x00008726 }, /* GL_MODELVIEW6_ARB */
+   { 20976, 0x00008727 }, /* GL_MODELVIEW7_ARB */
+   { 20994, 0x00008728 }, /* GL_MODELVIEW8_ARB */
+   { 21012, 0x00008729 }, /* GL_MODELVIEW9_ARB */
+   { 21030, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
+   { 21050, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
+   { 21077, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
+   { 21102, 0x00002100 }, /* GL_MODULATE */
+   { 21114, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
+   { 21134, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
+   { 21161, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
+   { 21186, 0x00000103 }, /* GL_MULT */
+   { 21194, 0x0000809D }, /* GL_MULTISAMPLE */
+   { 21209, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
+   { 21229, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
+   { 21248, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
+   { 21267, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
+   { 21291, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
+   { 21314, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+   { 21344, 0x00002A25 }, /* GL_N3F_V3F */
+   { 21355, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
+   { 21375, 0x0000150E }, /* GL_NAND */
+   { 21383, 0x00002600 }, /* GL_NEAREST */
+   { 21394, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+   { 21425, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+   { 21457, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
+   { 21482, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
+   { 21508, 0x00000200 }, /* GL_NEVER */
+   { 21517, 0x00001102 }, /* GL_NICEST */
+   { 21527, 0x00000000 }, /* GL_NONE */
+   { 21535, 0x00001505 }, /* GL_NOOP */
+   { 21543, 0x00001508 }, /* GL_NOR */
+   { 21550, 0x00000BA1 }, /* GL_NORMALIZE */
+   { 21563, 0x00008075 }, /* GL_NORMAL_ARRAY */
+   { 21579, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+   { 21610, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
+   { 21645, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
+   { 21669, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
+   { 21692, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
+   { 21713, 0x00008511 }, /* GL_NORMAL_MAP */
+   { 21727, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
+   { 21745, 0x00008511 }, /* GL_NORMAL_MAP_NV */
+   { 21762, 0x00000205 }, /* GL_NOTEQUAL */
+   { 21774, 0x00000000 }, /* GL_NO_ERROR */
+   { 21786, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+   { 21820, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
+   { 21858, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
+   { 21890, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
+   { 21932, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
+   { 21962, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
+   { 22002, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
+   { 22033, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
+   { 22062, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
+   { 22090, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
+   { 22120, 0x00002401 }, /* GL_OBJECT_LINEAR */
+   { 22137, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
+   { 22163, 0x00002501 }, /* GL_OBJECT_PLANE */
+   { 22179, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
+   { 22214, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
+   { 22236, 0x00009112 }, /* GL_OBJECT_TYPE */
+   { 22251, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
+   { 22270, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
+   { 22300, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
+   { 22321, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
+   { 22349, 0x00000001 }, /* GL_ONE */
+   { 22356, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+   { 22384, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
+   { 22416, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
+   { 22444, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
+   { 22476, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
+   { 22499, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
+   { 22522, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
+   { 22545, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
+   { 22568, 0x00008598 }, /* GL_OPERAND0_ALPHA */
+   { 22586, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
+   { 22608, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
+   { 22630, 0x00008590 }, /* GL_OPERAND0_RGB */
+   { 22646, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
+   { 22666, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
+   { 22686, 0x00008599 }, /* GL_OPERAND1_ALPHA */
+   { 22704, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
+   { 22726, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
+   { 22748, 0x00008591 }, /* GL_OPERAND1_RGB */
+   { 22764, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
+   { 22784, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
+   { 22804, 0x0000859A }, /* GL_OPERAND2_ALPHA */
+   { 22822, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
+   { 22844, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
+   { 22866, 0x00008592 }, /* GL_OPERAND2_RGB */
+   { 22882, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
+   { 22902, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
+   { 22922, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
+   { 22943, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
+   { 22962, 0x00001507 }, /* GL_OR */
+   { 22968, 0x00000A01 }, /* GL_ORDER */
+   { 22977, 0x0000150D }, /* GL_OR_INVERTED */
+   { 22992, 0x0000150B }, /* GL_OR_REVERSE */
+   { 23006, 0x00000505 }, /* GL_OUT_OF_MEMORY */
+   { 23023, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
+   { 23041, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
+   { 23062, 0x00008758 }, /* GL_PACK_INVERT_MESA */
+   { 23082, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
+   { 23100, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
+   { 23119, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
+   { 23139, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
+   { 23159, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
+   { 23177, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
+   { 23196, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
+   { 23221, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
+   { 23245, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
+   { 23266, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
+   { 23288, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
+   { 23310, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
+   { 23335, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
+   { 23359, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
+   { 23380, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
+   { 23402, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
+   { 23424, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
+   { 23446, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
+   { 23477, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
+   { 23497, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+   { 23522, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
+   { 23542, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+   { 23567, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
+   { 23587, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+   { 23612, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
+   { 23632, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+   { 23657, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
+   { 23677, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+   { 23702, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
+   { 23722, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+   { 23747, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
+   { 23767, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+   { 23792, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
+   { 23812, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+   { 23837, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
+   { 23857, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+   { 23882, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
+   { 23902, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+   { 23927, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
+   { 23945, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER */
+   { 23966, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING */
+   { 23995, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
+   { 24028, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
+   { 24053, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER */
+   { 24076, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
+   { 24107, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
+   { 24142, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
+   { 24169, 0x00001B00 }, /* GL_POINT */
+   { 24178, 0x00000000 }, /* GL_POINTS */
+   { 24188, 0x00000002 }, /* GL_POINT_BIT */
+   { 24201, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
+   { 24231, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
+   { 24265, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
+   { 24299, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
+   { 24334, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
+   { 24363, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
+   { 24396, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
+   { 24429, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
+   { 24463, 0x00000B11 }, /* GL_POINT_SIZE */
+   { 24477, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
+   { 24503, 0x00008127 }, /* GL_POINT_SIZE_MAX */
+   { 24521, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
+   { 24543, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
+   { 24565, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
+   { 24588, 0x00008126 }, /* GL_POINT_SIZE_MIN */
+   { 24606, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
+   { 24628, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
+   { 24650, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
+   { 24673, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
+   { 24693, 0x00000B10 }, /* GL_POINT_SMOOTH */
+   { 24709, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
+   { 24730, 0x00008861 }, /* GL_POINT_SPRITE */
+   { 24746, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
+   { 24766, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
+   { 24795, 0x00008861 }, /* GL_POINT_SPRITE_NV */
+   { 24814, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
+   { 24840, 0x00000701 }, /* GL_POINT_TOKEN */
+   { 24855, 0x00000009 }, /* GL_POLYGON */
+   { 24866, 0x00000008 }, /* GL_POLYGON_BIT */
+   { 24881, 0x00000B40 }, /* GL_POLYGON_MODE */
+   { 24897, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
+   { 24920, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
+   { 24945, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
+   { 24968, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
+   { 24991, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
+   { 25015, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
+   { 25039, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
+   { 25057, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
+   { 25080, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
+   { 25099, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
+   { 25122, 0x00000703 }, /* GL_POLYGON_TOKEN */
+   { 25139, 0x00001203 }, /* GL_POSITION */
+   { 25151, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+   { 25183, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
+   { 25219, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+   { 25252, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
+   { 25289, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+   { 25320, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
+   { 25355, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+   { 25387, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
+   { 25423, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 25456, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+   { 25488, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
+   { 25524, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+   { 25557, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
+   { 25594, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+   { 25624, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
+   { 25658, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+   { 25689, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
+   { 25724, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+   { 25755, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
+   { 25790, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+   { 25822, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
+   { 25858, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+   { 25888, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
+   { 25922, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+   { 25953, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
+   { 25988, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+   { 26020, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+   { 26051, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
+   { 26086, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+   { 26118, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
+   { 26154, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
+   { 26183, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
+   { 26216, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
+   { 26246, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
+   { 26280, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+   { 26319, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+   { 26352, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+   { 26392, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+   { 26426, 0x00008578 }, /* GL_PREVIOUS */
+   { 26438, 0x00008578 }, /* GL_PREVIOUS_ARB */
+   { 26454, 0x00008578 }, /* GL_PREVIOUS_EXT */
+   { 26470, 0x00008577 }, /* GL_PRIMARY_COLOR */
+   { 26487, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
+   { 26508, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
+   { 26529, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 26562, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 26594, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
+   { 26617, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
+   { 26640, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
+   { 26670, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
+   { 26699, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
+   { 26727, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
+   { 26749, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+   { 26777, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+   { 26805, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
+   { 26827, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
+   { 26848, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 26888, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 26927, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 26957, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 26992, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 27025, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 27059, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 27098, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 27137, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
+   { 27159, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
+   { 27185, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
+   { 27209, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
+   { 27232, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
+   { 27254, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
+   { 27275, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
+   { 27296, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
+   { 27323, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 27355, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 27387, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+   { 27422, 0x00001701 }, /* GL_PROJECTION */
+   { 27436, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
+   { 27457, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
+   { 27483, 0x00008E4F }, /* GL_PROVOKING_VERTEX_EXT */
+   { 27507, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
+   { 27528, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
+   { 27547, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
+   { 27570, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 27609, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+   { 27647, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
+   { 27667, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+   { 27697, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
+   { 27721, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
+   { 27741, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+   { 27771, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
+   { 27795, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
+   { 27815, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+   { 27848, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
+   { 27874, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
+   { 27904, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+   { 27935, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
+   { 27965, 0x00002003 }, /* GL_Q */
+   { 27970, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
+   { 27995, 0x00000007 }, /* GL_QUADS */
+   { 28004, 0x00008E4C }, /* GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT */
+   { 28052, 0x00008614 }, /* GL_QUAD_MESH_SUN */
+   { 28069, 0x00000008 }, /* GL_QUAD_STRIP */
+   { 28083, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
+   { 28105, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
+   { 28131, 0x00008866 }, /* GL_QUERY_RESULT */
+   { 28147, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
+   { 28167, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
+   { 28193, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
+   { 28223, 0x00002002 }, /* GL_R */
+   { 28228, 0x00002A10 }, /* GL_R3_G3_B2 */
+   { 28240, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+   { 28273, 0x00000C02 }, /* GL_READ_BUFFER */
+   { 28288, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER */
+   { 28308, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+   { 28340, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
+   { 28364, 0x000088B8 }, /* GL_READ_ONLY */
+   { 28377, 0x000088B8 }, /* GL_READ_ONLY_ARB */
+   { 28394, 0x000088BA }, /* GL_READ_WRITE */
+   { 28408, 0x000088BA }, /* GL_READ_WRITE_ARB */
+   { 28426, 0x00001903 }, /* GL_RED */
+   { 28433, 0x00008016 }, /* GL_REDUCE */
+   { 28443, 0x00008016 }, /* GL_REDUCE_EXT */
+   { 28457, 0x00000D15 }, /* GL_RED_BIAS */
+   { 28469, 0x00000D52 }, /* GL_RED_BITS */
+   { 28481, 0x00000D14 }, /* GL_RED_SCALE */
+   { 28494, 0x00008512 }, /* GL_REFLECTION_MAP */
+   { 28512, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
+   { 28534, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
+   { 28555, 0x00001C00 }, /* GL_RENDER */
+   { 28565, 0x00008D41 }, /* GL_RENDERBUFFER */
+   { 28581, 0x00008D53 }, /* GL_RENDERBUFFER_ALPHA_SIZE */
+   { 28608, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
+   { 28636, 0x00008D52 }, /* GL_RENDERBUFFER_BLUE_SIZE */
+   { 28662, 0x00008D54 }, /* GL_RENDERBUFFER_DEPTH_SIZE */
+   { 28689, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
+   { 28709, 0x00008D51 }, /* GL_RENDERBUFFER_GREEN_SIZE */
+   { 28736, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT */
+   { 28759, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
+   { 28786, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+   { 28818, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
+   { 28854, 0x00008D50 }, /* GL_RENDERBUFFER_RED_SIZE */
+   { 28879, 0x00008CAB }, /* GL_RENDERBUFFER_SAMPLES */
+   { 28903, 0x00008D55 }, /* GL_RENDERBUFFER_STENCIL_SIZE */
+   { 28932, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH */
+   { 28954, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
+   { 28980, 0x00001F01 }, /* GL_RENDERER */
+   { 28992, 0x00000C40 }, /* GL_RENDER_MODE */
+   { 29007, 0x00002901 }, /* GL_REPEAT */
+   { 29017, 0x00001E01 }, /* GL_REPLACE */
+   { 29028, 0x00008062 }, /* GL_REPLACE_EXT */
+   { 29043, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
+   { 29066, 0x0000803A }, /* GL_RESCALE_NORMAL */
+   { 29084, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
+   { 29106, 0x00000102 }, /* GL_RETURN */
+   { 29116, 0x00001907 }, /* GL_RGB */
+   { 29123, 0x00008052 }, /* GL_RGB10 */
+   { 29132, 0x00008059 }, /* GL_RGB10_A2 */
+   { 29144, 0x00008059 }, /* GL_RGB10_A2_EXT */
+   { 29160, 0x00008052 }, /* GL_RGB10_EXT */
+   { 29173, 0x00008053 }, /* GL_RGB12 */
+   { 29182, 0x00008053 }, /* GL_RGB12_EXT */
+   { 29195, 0x00008054 }, /* GL_RGB16 */
+   { 29204, 0x00008054 }, /* GL_RGB16_EXT */
+   { 29217, 0x0000804E }, /* GL_RGB2_EXT */
+   { 29229, 0x0000804F }, /* GL_RGB4 */
+   { 29237, 0x0000804F }, /* GL_RGB4_EXT */
+   { 29249, 0x000083A1 }, /* GL_RGB4_S3TC */
+   { 29262, 0x00008050 }, /* GL_RGB5 */
+   { 29270, 0x00008057 }, /* GL_RGB5_A1 */
+   { 29281, 0x00008057 }, /* GL_RGB5_A1_EXT */
+   { 29296, 0x00008050 }, /* GL_RGB5_EXT */
+   { 29308, 0x00008051 }, /* GL_RGB8 */
+   { 29316, 0x00008051 }, /* GL_RGB8_EXT */
+   { 29328, 0x00001908 }, /* GL_RGBA */
+   { 29336, 0x0000805A }, /* GL_RGBA12 */
+   { 29346, 0x0000805A }, /* GL_RGBA12_EXT */
+   { 29360, 0x0000805B }, /* GL_RGBA16 */
+   { 29370, 0x0000805B }, /* GL_RGBA16_EXT */
+   { 29384, 0x00008055 }, /* GL_RGBA2 */
+   { 29393, 0x00008055 }, /* GL_RGBA2_EXT */
+   { 29406, 0x00008056 }, /* GL_RGBA4 */
+   { 29415, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
+   { 29434, 0x00008056 }, /* GL_RGBA4_EXT */
+   { 29447, 0x000083A3 }, /* GL_RGBA4_S3TC */
+   { 29461, 0x00008058 }, /* GL_RGBA8 */
+   { 29470, 0x00008058 }, /* GL_RGBA8_EXT */
+   { 29483, 0x00008F97 }, /* GL_RGBA8_SNORM */
+   { 29498, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
+   { 29516, 0x00000C31 }, /* GL_RGBA_MODE */
+   { 29529, 0x000083A2 }, /* GL_RGBA_S3TC */
+   { 29542, 0x00008F93 }, /* GL_RGBA_SNORM */
+   { 29556, 0x000083A0 }, /* GL_RGB_S3TC */
+   { 29568, 0x00008573 }, /* GL_RGB_SCALE */
+   { 29581, 0x00008573 }, /* GL_RGB_SCALE_ARB */
+   { 29598, 0x00008573 }, /* GL_RGB_SCALE_EXT */
+   { 29615, 0x00000407 }, /* GL_RIGHT */
+   { 29624, 0x00002000 }, /* GL_S */
+   { 29629, 0x00008B5D }, /* GL_SAMPLER_1D */
+   { 29643, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
+   { 29664, 0x00008B5E }, /* GL_SAMPLER_2D */
+   { 29678, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
+   { 29699, 0x00008B5F }, /* GL_SAMPLER_3D */
+   { 29713, 0x00008B60 }, /* GL_SAMPLER_CUBE */
+   { 29729, 0x000080A9 }, /* GL_SAMPLES */
+   { 29740, 0x000086B4 }, /* GL_SAMPLES_3DFX */
+   { 29756, 0x000080A9 }, /* GL_SAMPLES_ARB */
+   { 29771, 0x00008914 }, /* GL_SAMPLES_PASSED */
+   { 29789, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
+   { 29811, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+   { 29839, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
+   { 29871, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
+   { 29894, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
+   { 29921, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
+   { 29939, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
+   { 29962, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
+   { 29984, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
+   { 30003, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
+   { 30026, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
+   { 30052, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
+   { 30082, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
+   { 30107, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
+   { 30136, 0x00080000 }, /* GL_SCISSOR_BIT */
+   { 30151, 0x00000C10 }, /* GL_SCISSOR_BOX */
+   { 30166, 0x00000C11 }, /* GL_SCISSOR_TEST */
+   { 30182, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
+   { 30207, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+   { 30247, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
+   { 30291, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+   { 30324, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+   { 30354, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+   { 30386, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+   { 30416, 0x00001C02 }, /* GL_SELECT */
+   { 30426, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
+   { 30454, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
+   { 30479, 0x00008012 }, /* GL_SEPARABLE_2D */
+   { 30495, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
+   { 30522, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
+   { 30553, 0x0000150F }, /* GL_SET */
+   { 30560, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
+   { 30581, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
+   { 30605, 0x00008B4F }, /* GL_SHADER_TYPE */
+   { 30620, 0x00000B54 }, /* GL_SHADE_MODEL */
+   { 30635, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
+   { 30663, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
+   { 30686, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+   { 30716, 0x00001601 }, /* GL_SHININESS */
+   { 30729, 0x00001402 }, /* GL_SHORT */
+   { 30738, 0x00009119 }, /* GL_SIGNALED */
+   { 30750, 0x00008F9C }, /* GL_SIGNED_NORMALIZED */
+   { 30771, 0x000081F9 }, /* GL_SINGLE_COLOR */
+   { 30787, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
+   { 30807, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
+   { 30826, 0x00008C46 }, /* GL_SLUMINANCE */
+   { 30840, 0x00008C47 }, /* GL_SLUMINANCE8 */
+   { 30855, 0x00008C45 }, /* GL_SLUMINANCE8_ALPHA8 */
+   { 30877, 0x00008C44 }, /* GL_SLUMINANCE_ALPHA */
+   { 30897, 0x00001D01 }, /* GL_SMOOTH */
+   { 30907, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
+   { 30940, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
+   { 30967, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
+   { 31000, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
+   { 31027, 0x00008588 }, /* GL_SOURCE0_ALPHA */
+   { 31044, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
+   { 31065, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
+   { 31086, 0x00008580 }, /* GL_SOURCE0_RGB */
+   { 31101, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
+   { 31120, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
+   { 31139, 0x00008589 }, /* GL_SOURCE1_ALPHA */
+   { 31156, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
+   { 31177, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
+   { 31198, 0x00008581 }, /* GL_SOURCE1_RGB */
+   { 31213, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
+   { 31232, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
+   { 31251, 0x0000858A }, /* GL_SOURCE2_ALPHA */
+   { 31268, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
+   { 31289, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
+   { 31310, 0x00008582 }, /* GL_SOURCE2_RGB */
+   { 31325, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
+   { 31344, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
+   { 31363, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
+   { 31383, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
+   { 31401, 0x00001202 }, /* GL_SPECULAR */
+   { 31413, 0x00002402 }, /* GL_SPHERE_MAP */
+   { 31427, 0x00001206 }, /* GL_SPOT_CUTOFF */
+   { 31442, 0x00001204 }, /* GL_SPOT_DIRECTION */
+   { 31460, 0x00001205 }, /* GL_SPOT_EXPONENT */
+   { 31477, 0x00008588 }, /* GL_SRC0_ALPHA */
+   { 31491, 0x00008580 }, /* GL_SRC0_RGB */
+   { 31503, 0x00008589 }, /* GL_SRC1_ALPHA */
+   { 31517, 0x00008581 }, /* GL_SRC1_RGB */
+   { 31529, 0x0000858A }, /* GL_SRC2_ALPHA */
+   { 31543, 0x00008582 }, /* GL_SRC2_RGB */
+   { 31555, 0x00000302 }, /* GL_SRC_ALPHA */
+   { 31568, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
+   { 31590, 0x00000300 }, /* GL_SRC_COLOR */
+   { 31603, 0x00008C40 }, /* GL_SRGB */
+   { 31611, 0x00008C41 }, /* GL_SRGB8 */
+   { 31620, 0x00008C43 }, /* GL_SRGB8_ALPHA8 */
+   { 31636, 0x00008C42 }, /* GL_SRGB_ALPHA */
+   { 31650, 0x00000503 }, /* GL_STACK_OVERFLOW */
+   { 31668, 0x00000504 }, /* GL_STACK_UNDERFLOW */
+   { 31687, 0x000088E6 }, /* GL_STATIC_COPY */
+   { 31702, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
+   { 31721, 0x000088E4 }, /* GL_STATIC_DRAW */
+   { 31736, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
+   { 31755, 0x000088E5 }, /* GL_STATIC_READ */
+   { 31770, 0x000088E5 }, /* GL_STATIC_READ_ARB */
+   { 31789, 0x00001802 }, /* GL_STENCIL */
+   { 31800, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT */
+   { 31822, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
+   { 31848, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
+   { 31869, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
+   { 31894, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
+   { 31915, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
+   { 31940, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+   { 31972, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
+   { 32008, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+   { 32040, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
+   { 32076, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
+   { 32096, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
+   { 32123, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
+   { 32149, 0x00000D57 }, /* GL_STENCIL_BITS */
+   { 32165, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
+   { 32187, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
+   { 32210, 0x00000B94 }, /* GL_STENCIL_FAIL */
+   { 32226, 0x00000B92 }, /* GL_STENCIL_FUNC */
+   { 32242, 0x00001901 }, /* GL_STENCIL_INDEX */
+   { 32259, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
+   { 32282, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
+   { 32304, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
+   { 32326, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
+   { 32348, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
+   { 32369, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
+   { 32396, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
+   { 32423, 0x00000B97 }, /* GL_STENCIL_REF */
+   { 32438, 0x00000B90 }, /* GL_STENCIL_TEST */
+   { 32454, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+   { 32483, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
+   { 32505, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
+   { 32526, 0x00000C33 }, /* GL_STEREO */
+   { 32536, 0x000085BE }, /* GL_STORAGE_CACHED_APPLE */
+   { 32560, 0x000085BD }, /* GL_STORAGE_PRIVATE_APPLE */
+   { 32585, 0x000085BF }, /* GL_STORAGE_SHARED_APPLE */
+   { 32609, 0x000088E2 }, /* GL_STREAM_COPY */
+   { 32624, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
+   { 32643, 0x000088E0 }, /* GL_STREAM_DRAW */
+   { 32658, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
+   { 32677, 0x000088E1 }, /* GL_STREAM_READ */
+   { 32692, 0x000088E1 }, /* GL_STREAM_READ_ARB */
+   { 32711, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
+   { 32728, 0x000084E7 }, /* GL_SUBTRACT */
+   { 32740, 0x000084E7 }, /* GL_SUBTRACT_ARB */
+   { 32756, 0x00009113 }, /* GL_SYNC_CONDITION */
+   { 32774, 0x00009116 }, /* GL_SYNC_FENCE */
+   { 32788, 0x00009115 }, /* GL_SYNC_FLAGS */
+   { 32802, 0x00000001 }, /* GL_SYNC_FLUSH_COMMANDS_BIT */
+   { 32829, 0x00009117 }, /* GL_SYNC_GPU_COMMANDS_COMPLETE */
+   { 32859, 0x00009114 }, /* GL_SYNC_STATUS */
+   { 32874, 0x00002001 }, /* GL_T */
+   { 32879, 0x00002A2A }, /* GL_T2F_C3F_V3F */
+   { 32894, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
+   { 32913, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
+   { 32929, 0x00002A2B }, /* GL_T2F_N3F_V3F */
+   { 32944, 0x00002A27 }, /* GL_T2F_V3F */
+   { 32955, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
+   { 32974, 0x00002A28 }, /* GL_T4F_V4F */
+   { 32985, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
+   { 33008, 0x00001702 }, /* GL_TEXTURE */
+   { 33019, 0x000084C0 }, /* GL_TEXTURE0 */
+   { 33031, 0x000084C0 }, /* GL_TEXTURE0_ARB */
+   { 33047, 0x000084C1 }, /* GL_TEXTURE1 */
+   { 33059, 0x000084CA }, /* GL_TEXTURE10 */
+   { 33072, 0x000084CA }, /* GL_TEXTURE10_ARB */
+   { 33089, 0x000084CB }, /* GL_TEXTURE11 */
+   { 33102, 0x000084CB }, /* GL_TEXTURE11_ARB */
+   { 33119, 0x000084CC }, /* GL_TEXTURE12 */
+   { 33132, 0x000084CC }, /* GL_TEXTURE12_ARB */
+   { 33149, 0x000084CD }, /* GL_TEXTURE13 */
+   { 33162, 0x000084CD }, /* GL_TEXTURE13_ARB */
+   { 33179, 0x000084CE }, /* GL_TEXTURE14 */
+   { 33192, 0x000084CE }, /* GL_TEXTURE14_ARB */
+   { 33209, 0x000084CF }, /* GL_TEXTURE15 */
+   { 33222, 0x000084CF }, /* GL_TEXTURE15_ARB */
+   { 33239, 0x000084D0 }, /* GL_TEXTURE16 */
+   { 33252, 0x000084D0 }, /* GL_TEXTURE16_ARB */
+   { 33269, 0x000084D1 }, /* GL_TEXTURE17 */
+   { 33282, 0x000084D1 }, /* GL_TEXTURE17_ARB */
+   { 33299, 0x000084D2 }, /* GL_TEXTURE18 */
+   { 33312, 0x000084D2 }, /* GL_TEXTURE18_ARB */
+   { 33329, 0x000084D3 }, /* GL_TEXTURE19 */
+   { 33342, 0x000084D3 }, /* GL_TEXTURE19_ARB */
+   { 33359, 0x000084C1 }, /* GL_TEXTURE1_ARB */
+   { 33375, 0x000084C2 }, /* GL_TEXTURE2 */
+   { 33387, 0x000084D4 }, /* GL_TEXTURE20 */
+   { 33400, 0x000084D4 }, /* GL_TEXTURE20_ARB */
+   { 33417, 0x000084D5 }, /* GL_TEXTURE21 */
+   { 33430, 0x000084D5 }, /* GL_TEXTURE21_ARB */
+   { 33447, 0x000084D6 }, /* GL_TEXTURE22 */
+   { 33460, 0x000084D6 }, /* GL_TEXTURE22_ARB */
+   { 33477, 0x000084D7 }, /* GL_TEXTURE23 */
+   { 33490, 0x000084D7 }, /* GL_TEXTURE23_ARB */
+   { 33507, 0x000084D8 }, /* GL_TEXTURE24 */
+   { 33520, 0x000084D8 }, /* GL_TEXTURE24_ARB */
+   { 33537, 0x000084D9 }, /* GL_TEXTURE25 */
+   { 33550, 0x000084D9 }, /* GL_TEXTURE25_ARB */
+   { 33567, 0x000084DA }, /* GL_TEXTURE26 */
+   { 33580, 0x000084DA }, /* GL_TEXTURE26_ARB */
+   { 33597, 0x000084DB }, /* GL_TEXTURE27 */
+   { 33610, 0x000084DB }, /* GL_TEXTURE27_ARB */
+   { 33627, 0x000084DC }, /* GL_TEXTURE28 */
+   { 33640, 0x000084DC }, /* GL_TEXTURE28_ARB */
+   { 33657, 0x000084DD }, /* GL_TEXTURE29 */
+   { 33670, 0x000084DD }, /* GL_TEXTURE29_ARB */
+   { 33687, 0x000084C2 }, /* GL_TEXTURE2_ARB */
+   { 33703, 0x000084C3 }, /* GL_TEXTURE3 */
+   { 33715, 0x000084DE }, /* GL_TEXTURE30 */
+   { 33728, 0x000084DE }, /* GL_TEXTURE30_ARB */
+   { 33745, 0x000084DF }, /* GL_TEXTURE31 */
+   { 33758, 0x000084DF }, /* GL_TEXTURE31_ARB */
+   { 33775, 0x000084C3 }, /* GL_TEXTURE3_ARB */
+   { 33791, 0x000084C4 }, /* GL_TEXTURE4 */
+   { 33803, 0x000084C4 }, /* GL_TEXTURE4_ARB */
+   { 33819, 0x000084C5 }, /* GL_TEXTURE5 */
+   { 33831, 0x000084C5 }, /* GL_TEXTURE5_ARB */
+   { 33847, 0x000084C6 }, /* GL_TEXTURE6 */
+   { 33859, 0x000084C6 }, /* GL_TEXTURE6_ARB */
+   { 33875, 0x000084C7 }, /* GL_TEXTURE7 */
+   { 33887, 0x000084C7 }, /* GL_TEXTURE7_ARB */
+   { 33903, 0x000084C8 }, /* GL_TEXTURE8 */
+   { 33915, 0x000084C8 }, /* GL_TEXTURE8_ARB */
+   { 33931, 0x000084C9 }, /* GL_TEXTURE9 */
+   { 33943, 0x000084C9 }, /* GL_TEXTURE9_ARB */
+   { 33959, 0x00000DE0 }, /* GL_TEXTURE_1D */
+   { 33973, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
+   { 33997, 0x00000DE1 }, /* GL_TEXTURE_2D */
+   { 34011, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
+   { 34035, 0x0000806F }, /* GL_TEXTURE_3D */
+   { 34049, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
+   { 34071, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
+   { 34097, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
+   { 34119, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
+   { 34141, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+   { 34173, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
+   { 34195, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+   { 34227, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
+   { 34249, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
+   { 34277, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
+   { 34309, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+   { 34342, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
+   { 34374, 0x00040000 }, /* GL_TEXTURE_BIT */
+   { 34389, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
+   { 34410, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
+   { 34435, 0x00001005 }, /* GL_TEXTURE_BORDER */
+   { 34453, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
+   { 34477, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+   { 34508, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+   { 34538, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+   { 34568, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+   { 34603, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+   { 34634, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 34672, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
+   { 34699, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+   { 34731, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+   { 34765, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
+   { 34789, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
+   { 34817, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
+   { 34841, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
+   { 34869, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+   { 34902, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
+   { 34926, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
+   { 34948, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
+   { 34970, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
+   { 34996, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
+   { 35030, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+   { 35063, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
+   { 35100, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
+   { 35128, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
+   { 35160, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
+   { 35183, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+   { 35221, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
+   { 35263, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+   { 35294, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+   { 35322, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+   { 35352, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+   { 35380, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
+   { 35400, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
+   { 35424, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+   { 35455, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
+   { 35490, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+   { 35521, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
+   { 35556, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+   { 35587, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
+   { 35622, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+   { 35653, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
+   { 35688, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+   { 35719, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
+   { 35754, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+   { 35785, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
+   { 35820, 0x000088F4 }, /* GL_TEXTURE_CUBE_MAP_SEAMLESS */
+   { 35849, 0x00008071 }, /* GL_TEXTURE_DEPTH */
+   { 35866, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
+   { 35888, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
+   { 35914, 0x00002300 }, /* GL_TEXTURE_ENV */
+   { 35929, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
+   { 35950, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
+   { 35970, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
+   { 35996, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
+   { 36016, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
+   { 36033, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
+   { 36050, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
+   { 36067, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
+   { 36084, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
+   { 36109, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
+   { 36131, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
+   { 36157, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
+   { 36175, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
+   { 36201, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
+   { 36227, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
+   { 36257, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
+   { 36284, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
+   { 36309, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
+   { 36329, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
+   { 36353, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+   { 36380, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+   { 36407, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+   { 36434, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
+   { 36460, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
+   { 36490, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
+   { 36512, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
+   { 36530, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 36560, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+   { 36588, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+   { 36616, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+   { 36644, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
+   { 36665, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
+   { 36684, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
+   { 36706, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
+   { 36725, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
+   { 36745, 0x000085B7 }, /* GL_TEXTURE_RANGE_LENGTH_APPLE */
+   { 36775, 0x000085B8 }, /* GL_TEXTURE_RANGE_POINTER_APPLE */
+   { 36806, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
+   { 36831, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
+   { 36855, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
+   { 36875, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
+   { 36899, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
+   { 36919, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
+   { 36942, 0x000088F1 }, /* GL_TEXTURE_STENCIL_SIZE */
+   { 36966, 0x000085BC }, /* GL_TEXTURE_STORAGE_HINT_APPLE */
+   { 36996, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
+   { 37021, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+   { 37055, 0x00001000 }, /* GL_TEXTURE_WIDTH */
+   { 37072, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
+   { 37090, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
+   { 37108, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
+   { 37126, 0x0000911B }, /* GL_TIMEOUT_EXPIRED */
+   { 37145, 0xFFFFFFFFFFFFFFFF }, /* GL_TIMEOUT_IGNORED */
+   { 37164, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
+   { 37184, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
+   { 37203, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+   { 37232, 0x00001000 }, /* GL_TRANSFORM_BIT */
+   { 37249, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
+   { 37275, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
+   { 37305, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+   { 37337, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+   { 37367, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
+   { 37401, 0x0000862C }, /* GL_TRANSPOSE_NV */
+   { 37417, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+   { 37448, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
+   { 37483, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+   { 37511, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
+   { 37543, 0x00000004 }, /* GL_TRIANGLES */
+   { 37556, 0x00000006 }, /* GL_TRIANGLE_FAN */
+   { 37572, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
+   { 37593, 0x00000005 }, /* GL_TRIANGLE_STRIP */
+   { 37611, 0x00000001 }, /* GL_TRUE */
+   { 37619, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
+   { 37639, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
+   { 37662, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
+   { 37682, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
+   { 37703, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
+   { 37725, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
+   { 37747, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
+   { 37767, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
+   { 37788, 0x00009118 }, /* GL_UNSIGNALED */
+   { 37802, 0x00001401 }, /* GL_UNSIGNED_BYTE */
+   { 37819, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+   { 37846, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
+   { 37869, 0x00001405 }, /* GL_UNSIGNED_INT */
+   { 37885, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
+   { 37912, 0x000084FA }, /* GL_UNSIGNED_INT_24_8 */
+   { 37933, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
+   { 37957, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+   { 37988, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
+   { 38012, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+   { 38040, 0x00008C17 }, /* GL_UNSIGNED_NORMALIZED */
+   { 38063, 0x00001403 }, /* GL_UNSIGNED_SHORT */
+   { 38081, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+   { 38111, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+   { 38137, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+   { 38167, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+   { 38193, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
+   { 38217, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+   { 38245, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+   { 38273, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
+   { 38300, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+   { 38332, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
+   { 38363, 0x00008CA2 }, /* GL_UPPER_LEFT */
+   { 38377, 0x00002A20 }, /* GL_V2F */
+   { 38384, 0x00002A21 }, /* GL_V3F */
+   { 38391, 0x00008B83 }, /* GL_VALIDATE_STATUS */
+   { 38410, 0x00001F00 }, /* GL_VENDOR */
+   { 38420, 0x00001F02 }, /* GL_VERSION */
+   { 38431, 0x00008074 }, /* GL_VERTEX_ARRAY */
+   { 38447, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING */
+   { 38471, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
+   { 38501, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+   { 38532, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
+   { 38567, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
+   { 38591, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
+   { 38612, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
+   { 38635, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
+   { 38656, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+   { 38683, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+   { 38711, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+   { 38739, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+   { 38767, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+   { 38795, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+   { 38823, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+   { 38851, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+   { 38878, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+   { 38905, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+   { 38932, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+   { 38959, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+   { 38986, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+   { 39013, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+   { 39040, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+   { 39067, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+   { 39094, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+   { 39132, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
+   { 39174, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+   { 39205, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
+   { 39240, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+   { 39274, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
+   { 39312, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+   { 39343, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
+   { 39378, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+   { 39406, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
+   { 39438, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+   { 39468, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
+   { 39502, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+   { 39530, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
+   { 39562, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
+   { 39582, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
+   { 39604, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
+   { 39633, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
+   { 39654, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+   { 39683, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
+   { 39716, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
+   { 39748, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+   { 39775, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
+   { 39806, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
+   { 39836, 0x00008B31 }, /* GL_VERTEX_SHADER */
+   { 39853, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
+   { 39874, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
+   { 39901, 0x00000BA2 }, /* GL_VIEWPORT */
+   { 39913, 0x00000800 }, /* GL_VIEWPORT_BIT */
+   { 39929, 0x0000911D }, /* GL_WAIT_FAILED */
+   { 39944, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
+   { 39964, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+   { 39995, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
+   { 40030, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+   { 40058, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+   { 40083, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+   { 40110, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+   { 40135, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
+   { 40159, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
+   { 40178, 0x000088B9 }, /* GL_WRITE_ONLY */
+   { 40192, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
+   { 40210, 0x00001506 }, /* GL_XOR */
+   { 40217, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
+   { 40236, 0x00008757 }, /* GL_YCBCR_MESA */
+   { 40250, 0x00000000 }, /* GL_ZERO */
+   { 40258, 0x00000D16 }, /* GL_ZOOM_X */
+   { 40268, 0x00000D17 }, /* GL_ZOOM_Y */
 };
 
-static const unsigned reduced_enums[1319] =
+static const unsigned reduced_enums[1347] =
 {
-       469, /* GL_FALSE */
-       683, /* GL_LINES */
-       685, /* GL_LINE_LOOP */
-       692, /* GL_LINE_STRIP */
-      1709, /* GL_TRIANGLES */
-      1712, /* GL_TRIANGLE_STRIP */
-      1710, /* GL_TRIANGLE_FAN */
-      1254, /* GL_QUADS */
-      1256, /* GL_QUAD_STRIP */
-      1142, /* GL_POLYGON */
-      1154, /* GL_POLYGON_STIPPLE_BIT */
-      1103, /* GL_PIXEL_MODE_BIT */
-       670, /* GL_LIGHTING_BIT */
-       497, /* GL_FOG_BIT */
+       475, /* GL_FALSE */
+       691, /* GL_LINES */
+       693, /* GL_LINE_LOOP */
+       700, /* GL_LINE_STRIP */
+      1743, /* GL_TRIANGLES */
+      1746, /* GL_TRIANGLE_STRIP */
+      1744, /* GL_TRIANGLE_FAN */
+      1271, /* GL_QUADS */
+      1274, /* GL_QUAD_STRIP */
+      1158, /* GL_POLYGON */
+      1170, /* GL_POLYGON_STIPPLE_BIT */
+      1119, /* GL_PIXEL_MODE_BIT */
+       678, /* GL_LIGHTING_BIT */
+       504, /* GL_FOG_BIT */
          8, /* GL_ACCUM */
-       702, /* GL_LOAD */
-      1308, /* GL_RETURN */
-       976, /* GL_MULT */
+       710, /* GL_LOAD */
+      1326, /* GL_RETURN */
+       991, /* GL_MULT */
         23, /* GL_ADD */
-       992, /* GL_NEVER */
-       660, /* GL_LESS */
-       459, /* GL_EQUAL */
-       659, /* GL_LEQUAL */
-       583, /* GL_GREATER */
-      1007, /* GL_NOTEQUAL */
-       582, /* GL_GEQUAL */
-        46, /* GL_ALWAYS */
-      1448, /* GL_SRC_COLOR */
-      1036, /* GL_ONE_MINUS_SRC_COLOR */
-      1446, /* GL_SRC_ALPHA */
-      1035, /* GL_ONE_MINUS_SRC_ALPHA */
-       438, /* GL_DST_ALPHA */
-      1033, /* GL_ONE_MINUS_DST_ALPHA */
-       439, /* GL_DST_COLOR */
-      1034, /* GL_ONE_MINUS_DST_COLOR */
-      1447, /* GL_SRC_ALPHA_SATURATE */
-       570, /* GL_FRONT_LEFT */
-       571, /* GL_FRONT_RIGHT */
-        68, /* GL_BACK_LEFT */
-        69, /* GL_BACK_RIGHT */
-       567, /* GL_FRONT */
-        67, /* GL_BACK */
-       658, /* GL_LEFT */
-      1350, /* GL_RIGHT */
-       568, /* GL_FRONT_AND_BACK */
-        62, /* GL_AUX0 */
-        63, /* GL_AUX1 */
-        64, /* GL_AUX2 */
-        65, /* GL_AUX3 */
-       649, /* GL_INVALID_ENUM */
-       653, /* GL_INVALID_VALUE */
-       652, /* GL_INVALID_OPERATION */
-      1453, /* GL_STACK_OVERFLOW */
-      1454, /* GL_STACK_UNDERFLOW */
-      1061, /* GL_OUT_OF_MEMORY */
-       650, /* GL_INVALID_FRAMEBUFFER_OPERATION */
+      1007, /* GL_NEVER */
+       668, /* GL_LESS */
+       465, /* GL_EQUAL */
+       667, /* GL_LEQUAL */
+       590, /* GL_GREATER */
+      1022, /* GL_NOTEQUAL */
+       589, /* GL_GEQUAL */
+        47, /* GL_ALWAYS */
+      1467, /* GL_SRC_COLOR */
+      1052, /* GL_ONE_MINUS_SRC_COLOR */
+      1465, /* GL_SRC_ALPHA */
+      1051, /* GL_ONE_MINUS_SRC_ALPHA */
+       444, /* GL_DST_ALPHA */
+      1049, /* GL_ONE_MINUS_DST_ALPHA */
+       445, /* GL_DST_COLOR */
+      1050, /* GL_ONE_MINUS_DST_COLOR */
+      1466, /* GL_SRC_ALPHA_SATURATE */
+       577, /* GL_FRONT_LEFT */
+       578, /* GL_FRONT_RIGHT */
+        69, /* GL_BACK_LEFT */
+        70, /* GL_BACK_RIGHT */
+       574, /* GL_FRONT */
+        68, /* GL_BACK */
+       666, /* GL_LEFT */
+      1368, /* GL_RIGHT */
+       575, /* GL_FRONT_AND_BACK */
+        63, /* GL_AUX0 */
+        64, /* GL_AUX1 */
+        65, /* GL_AUX2 */
+        66, /* GL_AUX3 */
+       656, /* GL_INVALID_ENUM */
+       660, /* GL_INVALID_VALUE */
+       659, /* GL_INVALID_OPERATION */
+      1472, /* GL_STACK_OVERFLOW */
+      1473, /* GL_STACK_UNDERFLOW */
+      1077, /* GL_OUT_OF_MEMORY */
+       657, /* GL_INVALID_FRAMEBUFFER_OPERATION */
          0, /* GL_2D */
          2, /* GL_3D */
          3, /* GL_3D_COLOR */
          4, /* GL_3D_COLOR_TEXTURE */
          6, /* GL_4D_COLOR_TEXTURE */
-      1081, /* GL_PASS_THROUGH_TOKEN */
-      1141, /* GL_POINT_TOKEN */
-       693, /* GL_LINE_TOKEN */
-      1155, /* GL_POLYGON_TOKEN */
-        73, /* GL_BITMAP_TOKEN */
-       437, /* GL_DRAW_PIXEL_TOKEN */
-       297, /* GL_COPY_PIXEL_TOKEN */
-       686, /* GL_LINE_RESET_TOKEN */
-       462, /* GL_EXP */
-       463, /* GL_EXP2 */
-       331, /* GL_CW */
-       122, /* GL_CCW */
-       143, /* GL_COEFF */
-      1058, /* GL_ORDER */
-       375, /* GL_DOMAIN */
-       305, /* GL_CURRENT_COLOR */
-       308, /* GL_CURRENT_INDEX */
-       314, /* GL_CURRENT_NORMAL */
-       327, /* GL_CURRENT_TEXTURE_COORDS */
-       319, /* GL_CURRENT_RASTER_COLOR */
-       321, /* GL_CURRENT_RASTER_INDEX */
-       325, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-       322, /* GL_CURRENT_RASTER_POSITION */
-       323, /* GL_CURRENT_RASTER_POSITION_VALID */
-       320, /* GL_CURRENT_RASTER_DISTANCE */
-      1134, /* GL_POINT_SMOOTH */
-      1123, /* GL_POINT_SIZE */
-      1133, /* GL_POINT_SIZE_RANGE */
-      1124, /* GL_POINT_SIZE_GRANULARITY */
-       687, /* GL_LINE_SMOOTH */
-       694, /* GL_LINE_WIDTH */
-       696, /* GL_LINE_WIDTH_RANGE */
-       695, /* GL_LINE_WIDTH_GRANULARITY */
-       689, /* GL_LINE_STIPPLE */
-       690, /* GL_LINE_STIPPLE_PATTERN */
-       691, /* GL_LINE_STIPPLE_REPEAT */
-       701, /* GL_LIST_MODE */
-       860, /* GL_MAX_LIST_NESTING */
-       698, /* GL_LIST_BASE */
-       700, /* GL_LIST_INDEX */
-      1144, /* GL_POLYGON_MODE */
-      1151, /* GL_POLYGON_SMOOTH */
-      1153, /* GL_POLYGON_STIPPLE */
-       448, /* GL_EDGE_FLAG */
-       298, /* GL_CULL_FACE */
-       299, /* GL_CULL_FACE_MODE */
-       569, /* GL_FRONT_FACE */
-       669, /* GL_LIGHTING */
-       674, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-       675, /* GL_LIGHT_MODEL_TWO_SIDE */
-       671, /* GL_LIGHT_MODEL_AMBIENT */
-      1396, /* GL_SHADE_MODEL */
-       190, /* GL_COLOR_MATERIAL_FACE */
-       191, /* GL_COLOR_MATERIAL_PARAMETER */
-       189, /* GL_COLOR_MATERIAL */
-       496, /* GL_FOG */
-       518, /* GL_FOG_INDEX */
-       514, /* GL_FOG_DENSITY */
-       522, /* GL_FOG_START */
-       516, /* GL_FOG_END */
-       519, /* GL_FOG_MODE */
-       498, /* GL_FOG_COLOR */
-       362, /* GL_DEPTH_RANGE */
-       369, /* GL_DEPTH_TEST */
-       372, /* GL_DEPTH_WRITEMASK */
-       350, /* GL_DEPTH_CLEAR_VALUE */
-       361, /* GL_DEPTH_FUNC */
+      1097, /* GL_PASS_THROUGH_TOKEN */
+      1157, /* GL_POINT_TOKEN */
+       701, /* GL_LINE_TOKEN */
+      1171, /* GL_POLYGON_TOKEN */
+        74, /* GL_BITMAP_TOKEN */
+       443, /* GL_DRAW_PIXEL_TOKEN */
+       301, /* GL_COPY_PIXEL_TOKEN */
+       694, /* GL_LINE_RESET_TOKEN */
+       468, /* GL_EXP */
+       469, /* GL_EXP2 */
+       337, /* GL_CW */
+       125, /* GL_CCW */
+       146, /* GL_COEFF */
+      1074, /* GL_ORDER */
+       381, /* GL_DOMAIN */
+       311, /* GL_CURRENT_COLOR */
+       314, /* GL_CURRENT_INDEX */
+       320, /* GL_CURRENT_NORMAL */
+       333, /* GL_CURRENT_TEXTURE_COORDS */
+       325, /* GL_CURRENT_RASTER_COLOR */
+       327, /* GL_CURRENT_RASTER_INDEX */
+       331, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+       328, /* GL_CURRENT_RASTER_POSITION */
+       329, /* GL_CURRENT_RASTER_POSITION_VALID */
+       326, /* GL_CURRENT_RASTER_DISTANCE */
+      1150, /* GL_POINT_SMOOTH */
+      1139, /* GL_POINT_SIZE */
+      1149, /* GL_POINT_SIZE_RANGE */
+      1140, /* GL_POINT_SIZE_GRANULARITY */
+       695, /* GL_LINE_SMOOTH */
+       702, /* GL_LINE_WIDTH */
+       704, /* GL_LINE_WIDTH_RANGE */
+       703, /* GL_LINE_WIDTH_GRANULARITY */
+       697, /* GL_LINE_STIPPLE */
+       698, /* GL_LINE_STIPPLE_PATTERN */
+       699, /* GL_LINE_STIPPLE_REPEAT */
+       709, /* GL_LIST_MODE */
+       874, /* GL_MAX_LIST_NESTING */
+       706, /* GL_LIST_BASE */
+       708, /* GL_LIST_INDEX */
+      1160, /* GL_POLYGON_MODE */
+      1167, /* GL_POLYGON_SMOOTH */
+      1169, /* GL_POLYGON_STIPPLE */
+       454, /* GL_EDGE_FLAG */
+       304, /* GL_CULL_FACE */
+       305, /* GL_CULL_FACE_MODE */
+       576, /* GL_FRONT_FACE */
+       677, /* GL_LIGHTING */
+       682, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+       683, /* GL_LIGHT_MODEL_TWO_SIDE */
+       679, /* GL_LIGHT_MODEL_AMBIENT */
+      1414, /* GL_SHADE_MODEL */
+       193, /* GL_COLOR_MATERIAL_FACE */
+       194, /* GL_COLOR_MATERIAL_PARAMETER */
+       192, /* GL_COLOR_MATERIAL */
+       503, /* GL_FOG */
+       525, /* GL_FOG_INDEX */
+       521, /* GL_FOG_DENSITY */
+       529, /* GL_FOG_START */
+       523, /* GL_FOG_END */
+       526, /* GL_FOG_MODE */
+       505, /* GL_FOG_COLOR */
+       368, /* GL_DEPTH_RANGE */
+       375, /* GL_DEPTH_TEST */
+       378, /* GL_DEPTH_WRITEMASK */
+       356, /* GL_DEPTH_CLEAR_VALUE */
+       367, /* GL_DEPTH_FUNC */
         12, /* GL_ACCUM_CLEAR_VALUE */
-      1489, /* GL_STENCIL_TEST */
-      1477, /* GL_STENCIL_CLEAR_VALUE */
-      1479, /* GL_STENCIL_FUNC */
-      1491, /* GL_STENCIL_VALUE_MASK */
-      1478, /* GL_STENCIL_FAIL */
-      1486, /* GL_STENCIL_PASS_DEPTH_FAIL */
-      1487, /* GL_STENCIL_PASS_DEPTH_PASS */
-      1488, /* GL_STENCIL_REF */
-      1492, /* GL_STENCIL_WRITEMASK */
-       829, /* GL_MATRIX_MODE */
-       997, /* GL_NORMALIZE */
-      1801, /* GL_VIEWPORT */
-       971, /* GL_MODELVIEW_STACK_DEPTH */
-      1234, /* GL_PROJECTION_STACK_DEPTH */
-      1687, /* GL_TEXTURE_STACK_DEPTH */
-       969, /* GL_MODELVIEW_MATRIX */
-      1233, /* GL_PROJECTION_MATRIX */
-      1672, /* GL_TEXTURE_MATRIX */
-        60, /* GL_ATTRIB_STACK_DEPTH */
-       133, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
+      1508, /* GL_STENCIL_TEST */
+      1496, /* GL_STENCIL_CLEAR_VALUE */
+      1498, /* GL_STENCIL_FUNC */
+      1510, /* GL_STENCIL_VALUE_MASK */
+      1497, /* GL_STENCIL_FAIL */
+      1505, /* GL_STENCIL_PASS_DEPTH_FAIL */
+      1506, /* GL_STENCIL_PASS_DEPTH_PASS */
+      1507, /* GL_STENCIL_REF */
+      1511, /* GL_STENCIL_WRITEMASK */
+       843, /* GL_MATRIX_MODE */
+      1012, /* GL_NORMALIZE */
+      1837, /* GL_VIEWPORT */
+       986, /* GL_MODELVIEW_STACK_DEPTH */
+      1250, /* GL_PROJECTION_STACK_DEPTH */
+      1718, /* GL_TEXTURE_STACK_DEPTH */
+       984, /* GL_MODELVIEW_MATRIX */
+      1249, /* GL_PROJECTION_MATRIX */
+      1701, /* GL_TEXTURE_MATRIX */
+        61, /* GL_ATTRIB_STACK_DEPTH */
+       136, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
         43, /* GL_ALPHA_TEST */
         44, /* GL_ALPHA_TEST_FUNC */
         45, /* GL_ALPHA_TEST_REF */
-       374, /* GL_DITHER */
-        77, /* GL_BLEND_DST */
-        86, /* GL_BLEND_SRC */
-        74, /* GL_BLEND */
-       704, /* GL_LOGIC_OP_MODE */
-       623, /* GL_INDEX_LOGIC_OP */
-       188, /* GL_COLOR_LOGIC_OP */
-        66, /* GL_AUX_BUFFERS */
-       385, /* GL_DRAW_BUFFER */
-      1266, /* GL_READ_BUFFER */
-      1377, /* GL_SCISSOR_BOX */
-      1378, /* GL_SCISSOR_TEST */
-       622, /* GL_INDEX_CLEAR_VALUE */
-       627, /* GL_INDEX_WRITEMASK */
-       185, /* GL_COLOR_CLEAR_VALUE */
-       227, /* GL_COLOR_WRITEMASK */
-       624, /* GL_INDEX_MODE */
-      1343, /* GL_RGBA_MODE */
-       384, /* GL_DOUBLEBUFFER */
-      1493, /* GL_STEREO */
-      1301, /* GL_RENDER_MODE */
-      1082, /* GL_PERSPECTIVE_CORRECTION_HINT */
-      1135, /* GL_POINT_SMOOTH_HINT */
-       688, /* GL_LINE_SMOOTH_HINT */
-      1152, /* GL_POLYGON_SMOOTH_HINT */
-       517, /* GL_FOG_HINT */
-      1653, /* GL_TEXTURE_GEN_S */
-      1654, /* GL_TEXTURE_GEN_T */
-      1652, /* GL_TEXTURE_GEN_R */
-      1651, /* GL_TEXTURE_GEN_Q */
-      1095, /* GL_PIXEL_MAP_I_TO_I */
-      1101, /* GL_PIXEL_MAP_S_TO_S */
-      1097, /* GL_PIXEL_MAP_I_TO_R */
-      1093, /* GL_PIXEL_MAP_I_TO_G */
-      1091, /* GL_PIXEL_MAP_I_TO_B */
-      1089, /* GL_PIXEL_MAP_I_TO_A */
-      1099, /* GL_PIXEL_MAP_R_TO_R */
-      1087, /* GL_PIXEL_MAP_G_TO_G */
-      1085, /* GL_PIXEL_MAP_B_TO_B */
-      1083, /* GL_PIXEL_MAP_A_TO_A */
-      1096, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-      1102, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-      1098, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-      1094, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-      1092, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-      1090, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-      1100, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-      1088, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-      1086, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-      1084, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-      1721, /* GL_UNPACK_SWAP_BYTES */
-      1716, /* GL_UNPACK_LSB_FIRST */
-      1717, /* GL_UNPACK_ROW_LENGTH */
-      1720, /* GL_UNPACK_SKIP_ROWS */
-      1719, /* GL_UNPACK_SKIP_PIXELS */
-      1714, /* GL_UNPACK_ALIGNMENT */
-      1070, /* GL_PACK_SWAP_BYTES */
-      1065, /* GL_PACK_LSB_FIRST */
-      1066, /* GL_PACK_ROW_LENGTH */
-      1069, /* GL_PACK_SKIP_ROWS */
-      1068, /* GL_PACK_SKIP_PIXELS */
-      1062, /* GL_PACK_ALIGNMENT */
-       782, /* GL_MAP_COLOR */
-       783, /* GL_MAP_STENCIL */
-       626, /* GL_INDEX_SHIFT */
-       625, /* GL_INDEX_OFFSET */
-      1279, /* GL_RED_SCALE */
-      1277, /* GL_RED_BIAS */
-      1818, /* GL_ZOOM_X */
-      1819, /* GL_ZOOM_Y */
-       587, /* GL_GREEN_SCALE */
-       585, /* GL_GREEN_BIAS */
-        92, /* GL_BLUE_SCALE */
-        90, /* GL_BLUE_BIAS */
+       380, /* GL_DITHER */
+        78, /* GL_BLEND_DST */
+        87, /* GL_BLEND_SRC */
+        75, /* GL_BLEND */
+       712, /* GL_LOGIC_OP_MODE */
+       630, /* GL_INDEX_LOGIC_OP */
+       191, /* GL_COLOR_LOGIC_OP */
+        67, /* GL_AUX_BUFFERS */
+       391, /* GL_DRAW_BUFFER */
+      1284, /* GL_READ_BUFFER */
+      1395, /* GL_SCISSOR_BOX */
+      1396, /* GL_SCISSOR_TEST */
+       629, /* GL_INDEX_CLEAR_VALUE */
+       634, /* GL_INDEX_WRITEMASK */
+       188, /* GL_COLOR_CLEAR_VALUE */
+       230, /* GL_COLOR_WRITEMASK */
+       631, /* GL_INDEX_MODE */
+      1361, /* GL_RGBA_MODE */
+       390, /* GL_DOUBLEBUFFER */
+      1512, /* GL_STEREO */
+      1319, /* GL_RENDER_MODE */
+      1098, /* GL_PERSPECTIVE_CORRECTION_HINT */
+      1151, /* GL_POINT_SMOOTH_HINT */
+       696, /* GL_LINE_SMOOTH_HINT */
+      1168, /* GL_POLYGON_SMOOTH_HINT */
+       524, /* GL_FOG_HINT */
+      1682, /* GL_TEXTURE_GEN_S */
+      1683, /* GL_TEXTURE_GEN_T */
+      1681, /* GL_TEXTURE_GEN_R */
+      1680, /* GL_TEXTURE_GEN_Q */
+      1111, /* GL_PIXEL_MAP_I_TO_I */
+      1117, /* GL_PIXEL_MAP_S_TO_S */
+      1113, /* GL_PIXEL_MAP_I_TO_R */
+      1109, /* GL_PIXEL_MAP_I_TO_G */
+      1107, /* GL_PIXEL_MAP_I_TO_B */
+      1105, /* GL_PIXEL_MAP_I_TO_A */
+      1115, /* GL_PIXEL_MAP_R_TO_R */
+      1103, /* GL_PIXEL_MAP_G_TO_G */
+      1101, /* GL_PIXEL_MAP_B_TO_B */
+      1099, /* GL_PIXEL_MAP_A_TO_A */
+      1112, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+      1118, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+      1114, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+      1110, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+      1108, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+      1106, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+      1116, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+      1104, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+      1102, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+      1100, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+      1755, /* GL_UNPACK_SWAP_BYTES */
+      1750, /* GL_UNPACK_LSB_FIRST */
+      1751, /* GL_UNPACK_ROW_LENGTH */
+      1754, /* GL_UNPACK_SKIP_ROWS */
+      1753, /* GL_UNPACK_SKIP_PIXELS */
+      1748, /* GL_UNPACK_ALIGNMENT */
+      1086, /* GL_PACK_SWAP_BYTES */
+      1081, /* GL_PACK_LSB_FIRST */
+      1082, /* GL_PACK_ROW_LENGTH */
+      1085, /* GL_PACK_SKIP_ROWS */
+      1084, /* GL_PACK_SKIP_PIXELS */
+      1078, /* GL_PACK_ALIGNMENT */
+       790, /* GL_MAP_COLOR */
+       795, /* GL_MAP_STENCIL */
+       633, /* GL_INDEX_SHIFT */
+       632, /* GL_INDEX_OFFSET */
+      1297, /* GL_RED_SCALE */
+      1295, /* GL_RED_BIAS */
+      1855, /* GL_ZOOM_X */
+      1856, /* GL_ZOOM_Y */
+       594, /* GL_GREEN_SCALE */
+       592, /* GL_GREEN_BIAS */
+        93, /* GL_BLUE_SCALE */
+        91, /* GL_BLUE_BIAS */
         42, /* GL_ALPHA_SCALE */
         40, /* GL_ALPHA_BIAS */
-       363, /* GL_DEPTH_SCALE */
-       344, /* GL_DEPTH_BIAS */
-       855, /* GL_MAX_EVAL_ORDER */
-       859, /* GL_MAX_LIGHTS */
-       838, /* GL_MAX_CLIP_PLANES */
-       904, /* GL_MAX_TEXTURE_SIZE */
-       865, /* GL_MAX_PIXEL_MAP_TABLE */
-       834, /* GL_MAX_ATTRIB_STACK_DEPTH */
-       862, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-       863, /* GL_MAX_NAME_STACK_DEPTH */
-       891, /* GL_MAX_PROJECTION_STACK_DEPTH */
-       905, /* GL_MAX_TEXTURE_STACK_DEPTH */
-       919, /* GL_MAX_VIEWPORT_DIMS */
-       835, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-      1500, /* GL_SUBPIXEL_BITS */
-       621, /* GL_INDEX_BITS */
-      1278, /* GL_RED_BITS */
-       586, /* GL_GREEN_BITS */
-        91, /* GL_BLUE_BITS */
+       369, /* GL_DEPTH_SCALE */
+       350, /* GL_DEPTH_BIAS */
+       869, /* GL_MAX_EVAL_ORDER */
+       873, /* GL_MAX_LIGHTS */
+       852, /* GL_MAX_CLIP_PLANES */
+       919, /* GL_MAX_TEXTURE_SIZE */
+       879, /* GL_MAX_PIXEL_MAP_TABLE */
+       848, /* GL_MAX_ATTRIB_STACK_DEPTH */
+       876, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+       877, /* GL_MAX_NAME_STACK_DEPTH */
+       905, /* GL_MAX_PROJECTION_STACK_DEPTH */
+       920, /* GL_MAX_TEXTURE_STACK_DEPTH */
+       934, /* GL_MAX_VIEWPORT_DIMS */
+       849, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+      1522, /* GL_SUBPIXEL_BITS */
+       628, /* GL_INDEX_BITS */
+      1296, /* GL_RED_BITS */
+       593, /* GL_GREEN_BITS */
+        92, /* GL_BLUE_BITS */
         41, /* GL_ALPHA_BITS */
-       345, /* GL_DEPTH_BITS */
-      1475, /* GL_STENCIL_BITS */
+       351, /* GL_DEPTH_BITS */
+      1494, /* GL_STENCIL_BITS */
         14, /* GL_ACCUM_RED_BITS */
         13, /* GL_ACCUM_GREEN_BITS */
         10, /* GL_ACCUM_BLUE_BITS */
          9, /* GL_ACCUM_ALPHA_BITS */
-       985, /* GL_NAME_STACK_DEPTH */
-        61, /* GL_AUTO_NORMAL */
-       728, /* GL_MAP1_COLOR_4 */
-       731, /* GL_MAP1_INDEX */
-       732, /* GL_MAP1_NORMAL */
-       733, /* GL_MAP1_TEXTURE_COORD_1 */
-       734, /* GL_MAP1_TEXTURE_COORD_2 */
-       735, /* GL_MAP1_TEXTURE_COORD_3 */
-       736, /* GL_MAP1_TEXTURE_COORD_4 */
-       737, /* GL_MAP1_VERTEX_3 */
-       738, /* GL_MAP1_VERTEX_4 */
-       755, /* GL_MAP2_COLOR_4 */
-       758, /* GL_MAP2_INDEX */
-       759, /* GL_MAP2_NORMAL */
-       760, /* GL_MAP2_TEXTURE_COORD_1 */
-       761, /* GL_MAP2_TEXTURE_COORD_2 */
-       762, /* GL_MAP2_TEXTURE_COORD_3 */
-       763, /* GL_MAP2_TEXTURE_COORD_4 */
-       764, /* GL_MAP2_VERTEX_3 */
-       765, /* GL_MAP2_VERTEX_4 */
-       729, /* GL_MAP1_GRID_DOMAIN */
-       730, /* GL_MAP1_GRID_SEGMENTS */
-       756, /* GL_MAP2_GRID_DOMAIN */
-       757, /* GL_MAP2_GRID_SEGMENTS */
-      1577, /* GL_TEXTURE_1D */
-      1579, /* GL_TEXTURE_2D */
-       472, /* GL_FEEDBACK_BUFFER_POINTER */
-       473, /* GL_FEEDBACK_BUFFER_SIZE */
-       474, /* GL_FEEDBACK_BUFFER_TYPE */
-      1387, /* GL_SELECTION_BUFFER_POINTER */
-      1388, /* GL_SELECTION_BUFFER_SIZE */
-      1691, /* GL_TEXTURE_WIDTH */
-      1658, /* GL_TEXTURE_HEIGHT */
-      1614, /* GL_TEXTURE_COMPONENTS */
-      1598, /* GL_TEXTURE_BORDER_COLOR */
-      1597, /* GL_TEXTURE_BORDER */
-       376, /* GL_DONT_CARE */
-       470, /* GL_FASTEST */
-       993, /* GL_NICEST */
-        47, /* GL_AMBIENT */
-       373, /* GL_DIFFUSE */
-      1435, /* GL_SPECULAR */
-      1156, /* GL_POSITION */
-      1438, /* GL_SPOT_DIRECTION */
-      1439, /* GL_SPOT_EXPONENT */
-      1437, /* GL_SPOT_CUTOFF */
-       271, /* GL_CONSTANT_ATTENUATION */
-       678, /* GL_LINEAR_ATTENUATION */
-      1253, /* GL_QUADRATIC_ATTENUATION */
-       241, /* GL_COMPILE */
-       242, /* GL_COMPILE_AND_EXECUTE */
-       117, /* GL_BYTE */
-      1722, /* GL_UNSIGNED_BYTE */
-      1401, /* GL_SHORT */
-      1733, /* GL_UNSIGNED_SHORT */
-       629, /* GL_INT */
-      1725, /* GL_UNSIGNED_INT */
-       477, /* GL_FLOAT */
+      1000, /* GL_NAME_STACK_DEPTH */
+        62, /* GL_AUTO_NORMAL */
+       736, /* GL_MAP1_COLOR_4 */
+       739, /* GL_MAP1_INDEX */
+       740, /* GL_MAP1_NORMAL */
+       741, /* GL_MAP1_TEXTURE_COORD_1 */
+       742, /* GL_MAP1_TEXTURE_COORD_2 */
+       743, /* GL_MAP1_TEXTURE_COORD_3 */
+       744, /* GL_MAP1_TEXTURE_COORD_4 */
+       745, /* GL_MAP1_VERTEX_3 */
+       746, /* GL_MAP1_VERTEX_4 */
+       763, /* GL_MAP2_COLOR_4 */
+       766, /* GL_MAP2_INDEX */
+       767, /* GL_MAP2_NORMAL */
+       768, /* GL_MAP2_TEXTURE_COORD_1 */
+       769, /* GL_MAP2_TEXTURE_COORD_2 */
+       770, /* GL_MAP2_TEXTURE_COORD_3 */
+       771, /* GL_MAP2_TEXTURE_COORD_4 */
+       772, /* GL_MAP2_VERTEX_3 */
+       773, /* GL_MAP2_VERTEX_4 */
+       737, /* GL_MAP1_GRID_DOMAIN */
+       738, /* GL_MAP1_GRID_SEGMENTS */
+       764, /* GL_MAP2_GRID_DOMAIN */
+       765, /* GL_MAP2_GRID_SEGMENTS */
+      1605, /* GL_TEXTURE_1D */
+      1607, /* GL_TEXTURE_2D */
+       478, /* GL_FEEDBACK_BUFFER_POINTER */
+       479, /* GL_FEEDBACK_BUFFER_SIZE */
+       480, /* GL_FEEDBACK_BUFFER_TYPE */
+      1405, /* GL_SELECTION_BUFFER_POINTER */
+      1406, /* GL_SELECTION_BUFFER_SIZE */
+      1723, /* GL_TEXTURE_WIDTH */
+      1687, /* GL_TEXTURE_HEIGHT */
+      1642, /* GL_TEXTURE_COMPONENTS */
+      1626, /* GL_TEXTURE_BORDER_COLOR */
+      1625, /* GL_TEXTURE_BORDER */
+       382, /* GL_DONT_CARE */
+       476, /* GL_FASTEST */
+      1008, /* GL_NICEST */
+        48, /* GL_AMBIENT */
+       379, /* GL_DIFFUSE */
+      1454, /* GL_SPECULAR */
+      1172, /* GL_POSITION */
+      1457, /* GL_SPOT_DIRECTION */
+      1458, /* GL_SPOT_EXPONENT */
+      1456, /* GL_SPOT_CUTOFF */
+       275, /* GL_CONSTANT_ATTENUATION */
+       686, /* GL_LINEAR_ATTENUATION */
+      1270, /* GL_QUADRATIC_ATTENUATION */
+       244, /* GL_COMPILE */
+       245, /* GL_COMPILE_AND_EXECUTE */
+       120, /* GL_BYTE */
+      1757, /* GL_UNSIGNED_BYTE */
+      1419, /* GL_SHORT */
+      1768, /* GL_UNSIGNED_SHORT */
+       636, /* GL_INT */
+      1760, /* GL_UNSIGNED_INT */
+       484, /* GL_FLOAT */
          1, /* GL_2_BYTES */
          5, /* GL_3_BYTES */
          7, /* GL_4_BYTES */
-       383, /* GL_DOUBLE */
-       129, /* GL_CLEAR */
-        49, /* GL_AND */
-        51, /* GL_AND_REVERSE */
-       295, /* GL_COPY */
-        50, /* GL_AND_INVERTED */
-       995, /* GL_NOOP */
-      1814, /* GL_XOR */
-      1057, /* GL_OR */
-       996, /* GL_NOR */
-       460, /* GL_EQUIV */
-       656, /* GL_INVERT */
-      1060, /* GL_OR_REVERSE */
-       296, /* GL_COPY_INVERTED */
-      1059, /* GL_OR_INVERTED */
-       986, /* GL_NAND */
-      1392, /* GL_SET */
-       457, /* GL_EMISSION */
-      1400, /* GL_SHININESS */
-        48, /* GL_AMBIENT_AND_DIFFUSE */
-       187, /* GL_COLOR_INDEXES */
-       936, /* GL_MODELVIEW */
-      1232, /* GL_PROJECTION */
-      1512, /* GL_TEXTURE */
-       144, /* GL_COLOR */
-       340, /* GL_DEPTH */
-      1461, /* GL_STENCIL */
-       186, /* GL_COLOR_INDEX */
-      1480, /* GL_STENCIL_INDEX */
-       351, /* GL_DEPTH_COMPONENT */
-      1274, /* GL_RED */
-       584, /* GL_GREEN */
-        89, /* GL_BLUE */
+       389, /* GL_DOUBLE */
+       132, /* GL_CLEAR */
+        50, /* GL_AND */
+        52, /* GL_AND_REVERSE */
+       299, /* GL_COPY */
+        51, /* GL_AND_INVERTED */
+      1010, /* GL_NOOP */
+      1851, /* GL_XOR */
+      1073, /* GL_OR */
+      1011, /* GL_NOR */
+       466, /* GL_EQUIV */
+       663, /* GL_INVERT */
+      1076, /* GL_OR_REVERSE */
+       300, /* GL_COPY_INVERTED */
+      1075, /* GL_OR_INVERTED */
+      1001, /* GL_NAND */
+      1410, /* GL_SET */
+       463, /* GL_EMISSION */
+      1418, /* GL_SHININESS */
+        49, /* GL_AMBIENT_AND_DIFFUSE */
+       190, /* GL_COLOR_INDEXES */
+       951, /* GL_MODELVIEW */
+      1248, /* GL_PROJECTION */
+      1540, /* GL_TEXTURE */
+       147, /* GL_COLOR */
+       346, /* GL_DEPTH */
+      1480, /* GL_STENCIL */
+       189, /* GL_COLOR_INDEX */
+      1499, /* GL_STENCIL_INDEX */
+       357, /* GL_DEPTH_COMPONENT */
+      1292, /* GL_RED */
+       591, /* GL_GREEN */
+        90, /* GL_BLUE */
         31, /* GL_ALPHA */
-      1309, /* GL_RGB */
-      1328, /* GL_RGBA */
-       706, /* GL_LUMINANCE */
-       727, /* GL_LUMINANCE_ALPHA */
-        72, /* GL_BITMAP */
-      1112, /* GL_POINT */
-       676, /* GL_LINE */
-       475, /* GL_FILL */
-      1283, /* GL_RENDER */
-       471, /* GL_FEEDBACK */
-      1386, /* GL_SELECT */
-       476, /* GL_FLAT */
-      1410, /* GL_SMOOTH */
-       657, /* GL_KEEP */
-      1303, /* GL_REPLACE */
-       611, /* GL_INCR */
-       336, /* GL_DECR */
-      1748, /* GL_VENDOR */
-      1300, /* GL_RENDERER */
-      1749, /* GL_VERSION */
-       464, /* GL_EXTENSIONS */
-      1351, /* GL_S */
-      1503, /* GL_T */
-      1263, /* GL_R */
-      1252, /* GL_Q */
-       972, /* GL_MODULATE */
-       335, /* GL_DECAL */
-      1648, /* GL_TEXTURE_ENV_MODE */
-      1647, /* GL_TEXTURE_ENV_COLOR */
-      1646, /* GL_TEXTURE_ENV */
-       465, /* GL_EYE_LINEAR */
-      1019, /* GL_OBJECT_LINEAR */
-      1436, /* GL_SPHERE_MAP */
-      1650, /* GL_TEXTURE_GEN_MODE */
-      1021, /* GL_OBJECT_PLANE */
-       466, /* GL_EYE_PLANE */
-       987, /* GL_NEAREST */
-       677, /* GL_LINEAR */
-       991, /* GL_NEAREST_MIPMAP_NEAREST */
-       682, /* GL_LINEAR_MIPMAP_NEAREST */
-       990, /* GL_NEAREST_MIPMAP_LINEAR */
-       681, /* GL_LINEAR_MIPMAP_LINEAR */
-      1671, /* GL_TEXTURE_MAG_FILTER */
-      1679, /* GL_TEXTURE_MIN_FILTER */
-      1693, /* GL_TEXTURE_WRAP_S */
-      1694, /* GL_TEXTURE_WRAP_T */
-       123, /* GL_CLAMP */
-      1302, /* GL_REPEAT */
-      1150, /* GL_POLYGON_OFFSET_UNITS */
-      1149, /* GL_POLYGON_OFFSET_POINT */
-      1148, /* GL_POLYGON_OFFSET_LINE */
-      1264, /* GL_R3_G3_B2 */
-      1745, /* GL_V2F */
-      1746, /* GL_V3F */
-       120, /* GL_C4UB_V2F */
-       121, /* GL_C4UB_V3F */
-       118, /* GL_C3F_V3F */
-       984, /* GL_N3F_V3F */
-       119, /* GL_C4F_N3F_V3F */
-      1508, /* GL_T2F_V3F */
-      1510, /* GL_T4F_V4F */
-      1506, /* GL_T2F_C4UB_V3F */
-      1504, /* GL_T2F_C3F_V3F */
-      1507, /* GL_T2F_N3F_V3F */
-      1505, /* GL_T2F_C4F_N3F_V3F */
-      1509, /* GL_T4F_C4F_N3F_V4F */
-       136, /* GL_CLIP_PLANE0 */
-       137, /* GL_CLIP_PLANE1 */
-       138, /* GL_CLIP_PLANE2 */
-       139, /* GL_CLIP_PLANE3 */
-       140, /* GL_CLIP_PLANE4 */
-       141, /* GL_CLIP_PLANE5 */
-       661, /* GL_LIGHT0 */
-       662, /* GL_LIGHT1 */
-       663, /* GL_LIGHT2 */
-       664, /* GL_LIGHT3 */
-       665, /* GL_LIGHT4 */
-       666, /* GL_LIGHT5 */
-       667, /* GL_LIGHT6 */
-       668, /* GL_LIGHT7 */
-       588, /* GL_HINT_BIT */
-       273, /* GL_CONSTANT_COLOR */
-      1031, /* GL_ONE_MINUS_CONSTANT_COLOR */
-       268, /* GL_CONSTANT_ALPHA */
-      1029, /* GL_ONE_MINUS_CONSTANT_ALPHA */
-        75, /* GL_BLEND_COLOR */
-       572, /* GL_FUNC_ADD */
-       920, /* GL_MIN */
-       831, /* GL_MAX */
-        80, /* GL_BLEND_EQUATION */
-       576, /* GL_FUNC_SUBTRACT */
-       574, /* GL_FUNC_REVERSE_SUBTRACT */
-       276, /* GL_CONVOLUTION_1D */
-       277, /* GL_CONVOLUTION_2D */
-      1389, /* GL_SEPARABLE_2D */
-       280, /* GL_CONVOLUTION_BORDER_MODE */
-       284, /* GL_CONVOLUTION_FILTER_SCALE */
-       282, /* GL_CONVOLUTION_FILTER_BIAS */
-      1275, /* GL_REDUCE */
-       286, /* GL_CONVOLUTION_FORMAT */
-       290, /* GL_CONVOLUTION_WIDTH */
-       288, /* GL_CONVOLUTION_HEIGHT */
-       846, /* GL_MAX_CONVOLUTION_WIDTH */
-       844, /* GL_MAX_CONVOLUTION_HEIGHT */
-      1189, /* GL_POST_CONVOLUTION_RED_SCALE */
-      1185, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-      1180, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-      1176, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-      1187, /* GL_POST_CONVOLUTION_RED_BIAS */
-      1183, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-      1178, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-      1174, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-       589, /* GL_HISTOGRAM */
-      1236, /* GL_PROXY_HISTOGRAM */
-       605, /* GL_HISTOGRAM_WIDTH */
-       595, /* GL_HISTOGRAM_FORMAT */
-       601, /* GL_HISTOGRAM_RED_SIZE */
-       597, /* GL_HISTOGRAM_GREEN_SIZE */
-       592, /* GL_HISTOGRAM_BLUE_SIZE */
-       590, /* GL_HISTOGRAM_ALPHA_SIZE */
-       599, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-       603, /* GL_HISTOGRAM_SINK */
-       921, /* GL_MINMAX */
-       923, /* GL_MINMAX_FORMAT */
-       925, /* GL_MINMAX_SINK */
-      1511, /* GL_TABLE_TOO_LARGE_EXT */
-      1724, /* GL_UNSIGNED_BYTE_3_3_2 */
-      1735, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-      1737, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-      1730, /* GL_UNSIGNED_INT_8_8_8_8 */
-      1726, /* GL_UNSIGNED_INT_10_10_10_2 */
-      1147, /* GL_POLYGON_OFFSET_FILL */
-      1146, /* GL_POLYGON_OFFSET_FACTOR */
-      1145, /* GL_POLYGON_OFFSET_BIAS */
-      1306, /* GL_RESCALE_NORMAL */
+      1327, /* GL_RGB */
+      1346, /* GL_RGBA */
+       714, /* GL_LUMINANCE */
+       735, /* GL_LUMINANCE_ALPHA */
+        73, /* GL_BITMAP */
+      1128, /* GL_POINT */
+       684, /* GL_LINE */
+       481, /* GL_FILL */
+      1301, /* GL_RENDER */
+       477, /* GL_FEEDBACK */
+      1404, /* GL_SELECT */
+       483, /* GL_FLAT */
+      1429, /* GL_SMOOTH */
+       664, /* GL_KEEP */
+      1321, /* GL_REPLACE */
+       618, /* GL_INCR */
+       342, /* GL_DECR */
+      1783, /* GL_VENDOR */
+      1318, /* GL_RENDERER */
+      1784, /* GL_VERSION */
+       470, /* GL_EXTENSIONS */
+      1369, /* GL_S */
+      1531, /* GL_T */
+      1281, /* GL_R */
+      1269, /* GL_Q */
+       987, /* GL_MODULATE */
+       341, /* GL_DECAL */
+      1677, /* GL_TEXTURE_ENV_MODE */
+      1676, /* GL_TEXTURE_ENV_COLOR */
+      1675, /* GL_TEXTURE_ENV */
+       471, /* GL_EYE_LINEAR */
+      1034, /* GL_OBJECT_LINEAR */
+      1455, /* GL_SPHERE_MAP */
+      1679, /* GL_TEXTURE_GEN_MODE */
+      1036, /* GL_OBJECT_PLANE */
+       472, /* GL_EYE_PLANE */
+      1002, /* GL_NEAREST */
+       685, /* GL_LINEAR */
+      1006, /* GL_NEAREST_MIPMAP_NEAREST */
+       690, /* GL_LINEAR_MIPMAP_NEAREST */
+      1005, /* GL_NEAREST_MIPMAP_LINEAR */
+       689, /* GL_LINEAR_MIPMAP_LINEAR */
+      1700, /* GL_TEXTURE_MAG_FILTER */
+      1708, /* GL_TEXTURE_MIN_FILTER */
+      1725, /* GL_TEXTURE_WRAP_S */
+      1726, /* GL_TEXTURE_WRAP_T */
+       126, /* GL_CLAMP */
+      1320, /* GL_REPEAT */
+      1166, /* GL_POLYGON_OFFSET_UNITS */
+      1165, /* GL_POLYGON_OFFSET_POINT */
+      1164, /* GL_POLYGON_OFFSET_LINE */
+      1282, /* GL_R3_G3_B2 */
+      1780, /* GL_V2F */
+      1781, /* GL_V3F */
+       123, /* GL_C4UB_V2F */
+       124, /* GL_C4UB_V3F */
+       121, /* GL_C3F_V3F */
+       999, /* GL_N3F_V3F */
+       122, /* GL_C4F_N3F_V3F */
+      1536, /* GL_T2F_V3F */
+      1538, /* GL_T4F_V4F */
+      1534, /* GL_T2F_C4UB_V3F */
+      1532, /* GL_T2F_C3F_V3F */
+      1535, /* GL_T2F_N3F_V3F */
+      1533, /* GL_T2F_C4F_N3F_V3F */
+      1537, /* GL_T4F_C4F_N3F_V4F */
+       139, /* GL_CLIP_PLANE0 */
+       140, /* GL_CLIP_PLANE1 */
+       141, /* GL_CLIP_PLANE2 */
+       142, /* GL_CLIP_PLANE3 */
+       143, /* GL_CLIP_PLANE4 */
+       144, /* GL_CLIP_PLANE5 */
+       669, /* GL_LIGHT0 */
+       670, /* GL_LIGHT1 */
+       671, /* GL_LIGHT2 */
+       672, /* GL_LIGHT3 */
+       673, /* GL_LIGHT4 */
+       674, /* GL_LIGHT5 */
+       675, /* GL_LIGHT6 */
+       676, /* GL_LIGHT7 */
+       595, /* GL_HINT_BIT */
+       277, /* GL_CONSTANT_COLOR */
+      1047, /* GL_ONE_MINUS_CONSTANT_COLOR */
+       272, /* GL_CONSTANT_ALPHA */
+      1045, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+        76, /* GL_BLEND_COLOR */
+       579, /* GL_FUNC_ADD */
+       935, /* GL_MIN */
+       845, /* GL_MAX */
+        81, /* GL_BLEND_EQUATION */
+       583, /* GL_FUNC_SUBTRACT */
+       581, /* GL_FUNC_REVERSE_SUBTRACT */
+       280, /* GL_CONVOLUTION_1D */
+       281, /* GL_CONVOLUTION_2D */
+      1407, /* GL_SEPARABLE_2D */
+       284, /* GL_CONVOLUTION_BORDER_MODE */
+       288, /* GL_CONVOLUTION_FILTER_SCALE */
+       286, /* GL_CONVOLUTION_FILTER_BIAS */
+      1293, /* GL_REDUCE */
+       290, /* GL_CONVOLUTION_FORMAT */
+       294, /* GL_CONVOLUTION_WIDTH */
+       292, /* GL_CONVOLUTION_HEIGHT */
+       860, /* GL_MAX_CONVOLUTION_WIDTH */
+       858, /* GL_MAX_CONVOLUTION_HEIGHT */
+      1205, /* GL_POST_CONVOLUTION_RED_SCALE */
+      1201, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+      1196, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+      1192, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+      1203, /* GL_POST_CONVOLUTION_RED_BIAS */
+      1199, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+      1194, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+      1190, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+       596, /* GL_HISTOGRAM */
+      1253, /* GL_PROXY_HISTOGRAM */
+       612, /* GL_HISTOGRAM_WIDTH */
+       602, /* GL_HISTOGRAM_FORMAT */
+       608, /* GL_HISTOGRAM_RED_SIZE */
+       604, /* GL_HISTOGRAM_GREEN_SIZE */
+       599, /* GL_HISTOGRAM_BLUE_SIZE */
+       597, /* GL_HISTOGRAM_ALPHA_SIZE */
+       606, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+       610, /* GL_HISTOGRAM_SINK */
+       936, /* GL_MINMAX */
+       938, /* GL_MINMAX_FORMAT */
+       940, /* GL_MINMAX_SINK */
+      1539, /* GL_TABLE_TOO_LARGE_EXT */
+      1759, /* GL_UNSIGNED_BYTE_3_3_2 */
+      1770, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+      1772, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+      1765, /* GL_UNSIGNED_INT_8_8_8_8 */
+      1761, /* GL_UNSIGNED_INT_10_10_10_2 */
+      1163, /* GL_POLYGON_OFFSET_FILL */
+      1162, /* GL_POLYGON_OFFSET_FACTOR */
+      1161, /* GL_POLYGON_OFFSET_BIAS */
+      1324, /* GL_RESCALE_NORMAL */
         36, /* GL_ALPHA4 */
         38, /* GL_ALPHA8 */
         32, /* GL_ALPHA12 */
         34, /* GL_ALPHA16 */
-       717, /* GL_LUMINANCE4 */
-       723, /* GL_LUMINANCE8 */
-       707, /* GL_LUMINANCE12 */
-       713, /* GL_LUMINANCE16 */
-       718, /* GL_LUMINANCE4_ALPHA4 */
-       721, /* GL_LUMINANCE6_ALPHA2 */
-       724, /* GL_LUMINANCE8_ALPHA8 */
-       710, /* GL_LUMINANCE12_ALPHA4 */
-       708, /* GL_LUMINANCE12_ALPHA12 */
-       714, /* GL_LUMINANCE16_ALPHA16 */
-       630, /* GL_INTENSITY */
-       635, /* GL_INTENSITY4 */
-       637, /* GL_INTENSITY8 */
-       631, /* GL_INTENSITY12 */
-       633, /* GL_INTENSITY16 */
-      1318, /* GL_RGB2_EXT */
-      1319, /* GL_RGB4 */
-      1322, /* GL_RGB5 */
-      1326, /* GL_RGB8 */
-      1310, /* GL_RGB10 */
-      1314, /* GL_RGB12 */
-      1316, /* GL_RGB16 */
-      1333, /* GL_RGBA2 */
-      1335, /* GL_RGBA4 */
-      1323, /* GL_RGB5_A1 */
-      1339, /* GL_RGBA8 */
-      1311, /* GL_RGB10_A2 */
-      1329, /* GL_RGBA12 */
-      1331, /* GL_RGBA16 */
-      1684, /* GL_TEXTURE_RED_SIZE */
-      1656, /* GL_TEXTURE_GREEN_SIZE */
-      1595, /* GL_TEXTURE_BLUE_SIZE */
-      1582, /* GL_TEXTURE_ALPHA_SIZE */
-      1669, /* GL_TEXTURE_LUMINANCE_SIZE */
-      1660, /* GL_TEXTURE_INTENSITY_SIZE */
-      1304, /* GL_REPLACE_EXT */
-      1240, /* GL_PROXY_TEXTURE_1D */
-      1243, /* GL_PROXY_TEXTURE_2D */
-      1689, /* GL_TEXTURE_TOO_LARGE_EXT */
-      1681, /* GL_TEXTURE_PRIORITY */
-      1686, /* GL_TEXTURE_RESIDENT */
-      1585, /* GL_TEXTURE_BINDING_1D */
-      1587, /* GL_TEXTURE_BINDING_2D */
-      1589, /* GL_TEXTURE_BINDING_3D */
-      1067, /* GL_PACK_SKIP_IMAGES */
-      1063, /* GL_PACK_IMAGE_HEIGHT */
-      1718, /* GL_UNPACK_SKIP_IMAGES */
-      1715, /* GL_UNPACK_IMAGE_HEIGHT */
-      1581, /* GL_TEXTURE_3D */
-      1246, /* GL_PROXY_TEXTURE_3D */
-      1643, /* GL_TEXTURE_DEPTH */
-      1692, /* GL_TEXTURE_WRAP_R */
-       832, /* GL_MAX_3D_TEXTURE_SIZE */
-      1750, /* GL_VERTEX_ARRAY */
-       998, /* GL_NORMAL_ARRAY */
-       145, /* GL_COLOR_ARRAY */
-       615, /* GL_INDEX_ARRAY */
-      1622, /* GL_TEXTURE_COORD_ARRAY */
-       449, /* GL_EDGE_FLAG_ARRAY */
-      1755, /* GL_VERTEX_ARRAY_SIZE */
-      1757, /* GL_VERTEX_ARRAY_TYPE */
-      1756, /* GL_VERTEX_ARRAY_STRIDE */
-      1003, /* GL_NORMAL_ARRAY_TYPE */
-      1002, /* GL_NORMAL_ARRAY_STRIDE */
-       149, /* GL_COLOR_ARRAY_SIZE */
-       151, /* GL_COLOR_ARRAY_TYPE */
-       150, /* GL_COLOR_ARRAY_STRIDE */
-       620, /* GL_INDEX_ARRAY_TYPE */
-       619, /* GL_INDEX_ARRAY_STRIDE */
-      1626, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-      1628, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-      1627, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-       453, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-      1754, /* GL_VERTEX_ARRAY_POINTER */
-      1001, /* GL_NORMAL_ARRAY_POINTER */
-       148, /* GL_COLOR_ARRAY_POINTER */
-       618, /* GL_INDEX_ARRAY_POINTER */
-      1625, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-       452, /* GL_EDGE_FLAG_ARRAY_POINTER */
-       977, /* GL_MULTISAMPLE */
-      1363, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-      1365, /* GL_SAMPLE_ALPHA_TO_ONE */
-      1370, /* GL_SAMPLE_COVERAGE */
-      1367, /* GL_SAMPLE_BUFFERS */
-      1358, /* GL_SAMPLES */
-      1374, /* GL_SAMPLE_COVERAGE_VALUE */
-      1372, /* GL_SAMPLE_COVERAGE_INVERT */
-       192, /* GL_COLOR_MATRIX */
-       194, /* GL_COLOR_MATRIX_STACK_DEPTH */
-       840, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-      1172, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-      1168, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-      1163, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-      1159, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-      1170, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-      1166, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-      1161, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-      1157, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-      1605, /* GL_TEXTURE_COLOR_TABLE_SGI */
-      1247, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-      1607, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-        79, /* GL_BLEND_DST_RGB */
-        88, /* GL_BLEND_SRC_RGB */
-        78, /* GL_BLEND_DST_ALPHA */
-        87, /* GL_BLEND_SRC_ALPHA */
-       198, /* GL_COLOR_TABLE */
-      1182, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-      1165, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-      1235, /* GL_PROXY_COLOR_TABLE */
-      1239, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-      1238, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
-       222, /* GL_COLOR_TABLE_SCALE */
-       202, /* GL_COLOR_TABLE_BIAS */
-       207, /* GL_COLOR_TABLE_FORMAT */
-       224, /* GL_COLOR_TABLE_WIDTH */
-       219, /* GL_COLOR_TABLE_RED_SIZE */
-       210, /* GL_COLOR_TABLE_GREEN_SIZE */
-       204, /* GL_COLOR_TABLE_BLUE_SIZE */
-       199, /* GL_COLOR_TABLE_ALPHA_SIZE */
-       216, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
-       213, /* GL_COLOR_TABLE_INTENSITY_SIZE */
-        70, /* GL_BGR */
-        71, /* GL_BGRA */
-       854, /* GL_MAX_ELEMENTS_VERTICES */
-       853, /* GL_MAX_ELEMENTS_INDICES */
-      1659, /* GL_TEXTURE_INDEX_SIZE_EXT */
-       142, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
-      1129, /* GL_POINT_SIZE_MIN */
-      1125, /* GL_POINT_SIZE_MAX */
-      1119, /* GL_POINT_FADE_THRESHOLD_SIZE */
-      1115, /* GL_POINT_DISTANCE_ATTENUATION */
-       124, /* GL_CLAMP_TO_BORDER */
-       127, /* GL_CLAMP_TO_EDGE */
-      1680, /* GL_TEXTURE_MIN_LOD */
-      1678, /* GL_TEXTURE_MAX_LOD */
-      1584, /* GL_TEXTURE_BASE_LEVEL */
-      1677, /* GL_TEXTURE_MAX_LEVEL */
-       608, /* GL_IGNORE_BORDER_HP */
-       272, /* GL_CONSTANT_BORDER_HP */
-      1305, /* GL_REPLICATE_BORDER_HP */
-       278, /* GL_CONVOLUTION_BORDER_COLOR */
-      1026, /* GL_OCCLUSION_TEST_HP */
-      1027, /* GL_OCCLUSION_TEST_RESULT_HP */
-       679, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-      1599, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-      1601, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-      1603, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-      1604, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1602, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-      1600, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-       836, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-       837, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1192, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-      1194, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-      1191, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-      1193, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-      1667, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-      1668, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-      1666, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-       578, /* GL_GENERATE_MIPMAP */
-       579, /* GL_GENERATE_MIPMAP_HINT */
-       520, /* GL_FOG_OFFSET_SGIX */
-       521, /* GL_FOG_OFFSET_VALUE_SGIX */
-      1613, /* GL_TEXTURE_COMPARE_SGIX */
-      1612, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-      1663, /* GL_TEXTURE_LEQUAL_R_SGIX */
-      1655, /* GL_TEXTURE_GEQUAL_R_SGIX */
-       352, /* GL_DEPTH_COMPONENT16 */
-       355, /* GL_DEPTH_COMPONENT24 */
-       358, /* GL_DEPTH_COMPONENT32 */
-       300, /* GL_CULL_VERTEX_EXT */
-       302, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-       301, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-      1811, /* GL_WRAP_BORDER_SUN */
-      1606, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-       672, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-      1403, /* GL_SINGLE_COLOR */
-      1390, /* GL_SEPARATE_SPECULAR_COLOR */
-      1399, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-       531, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
-       532, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
-       539, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
-       534, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
-       530, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
-       529, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
-       533, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
-       540, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
-       551, /* GL_FRAMEBUFFER_DEFAULT */
-       564, /* GL_FRAMEBUFFER_UNDEFINED */
-       365, /* GL_DEPTH_STENCIL_ATTACHMENT */
-       614, /* GL_INDEX */
-      1723, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-      1738, /* GL_UNSIGNED_SHORT_5_6_5 */
-      1739, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-      1736, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-      1734, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-      1731, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-      1729, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-      1675, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-      1676, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-      1674, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-       928, /* GL_MIRRORED_REPEAT */
-      1346, /* GL_RGB_S3TC */
-      1321, /* GL_RGB4_S3TC */
-      1344, /* GL_RGBA_S3TC */
-      1338, /* GL_RGBA4_S3TC */
-      1342, /* GL_RGBA_DXT5_S3TC */
-      1336, /* GL_RGBA4_DXT5_S3TC */
-       261, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
-       256, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
-       257, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
-       258, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
-       989, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-       988, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-       680, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-       507, /* GL_FOG_COORDINATE_SOURCE */
-       499, /* GL_FOG_COORD */
-       523, /* GL_FRAGMENT_DEPTH */
-       306, /* GL_CURRENT_FOG_COORD */
-       506, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-       505, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-       504, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-       501, /* GL_FOG_COORDINATE_ARRAY */
-       196, /* GL_COLOR_SUM */
-       326, /* GL_CURRENT_SECONDARY_COLOR */
-      1383, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-      1385, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-      1384, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-      1382, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-      1379, /* GL_SECONDARY_COLOR_ARRAY */
-       324, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
+       725, /* GL_LUMINANCE4 */
+       731, /* GL_LUMINANCE8 */
+       715, /* GL_LUMINANCE12 */
+       721, /* GL_LUMINANCE16 */
+       726, /* GL_LUMINANCE4_ALPHA4 */
+       729, /* GL_LUMINANCE6_ALPHA2 */
+       732, /* GL_LUMINANCE8_ALPHA8 */
+       718, /* GL_LUMINANCE12_ALPHA4 */
+       716, /* GL_LUMINANCE12_ALPHA12 */
+       722, /* GL_LUMINANCE16_ALPHA16 */
+       637, /* GL_INTENSITY */
+       642, /* GL_INTENSITY4 */
+       644, /* GL_INTENSITY8 */
+       638, /* GL_INTENSITY12 */
+       640, /* GL_INTENSITY16 */
+      1336, /* GL_RGB2_EXT */
+      1337, /* GL_RGB4 */
+      1340, /* GL_RGB5 */
+      1344, /* GL_RGB8 */
+      1328, /* GL_RGB10 */
+      1332, /* GL_RGB12 */
+      1334, /* GL_RGB16 */
+      1351, /* GL_RGBA2 */
+      1353, /* GL_RGBA4 */
+      1341, /* GL_RGB5_A1 */
+      1357, /* GL_RGBA8 */
+      1329, /* GL_RGB10_A2 */
+      1347, /* GL_RGBA12 */
+      1349, /* GL_RGBA16 */
+      1715, /* GL_TEXTURE_RED_SIZE */
+      1685, /* GL_TEXTURE_GREEN_SIZE */
+      1623, /* GL_TEXTURE_BLUE_SIZE */
+      1610, /* GL_TEXTURE_ALPHA_SIZE */
+      1698, /* GL_TEXTURE_LUMINANCE_SIZE */
+      1689, /* GL_TEXTURE_INTENSITY_SIZE */
+      1322, /* GL_REPLACE_EXT */
+      1257, /* GL_PROXY_TEXTURE_1D */
+      1260, /* GL_PROXY_TEXTURE_2D */
+      1721, /* GL_TEXTURE_TOO_LARGE_EXT */
+      1710, /* GL_TEXTURE_PRIORITY */
+      1717, /* GL_TEXTURE_RESIDENT */
+      1613, /* GL_TEXTURE_BINDING_1D */
+      1615, /* GL_TEXTURE_BINDING_2D */
+      1617, /* GL_TEXTURE_BINDING_3D */
+      1083, /* GL_PACK_SKIP_IMAGES */
+      1079, /* GL_PACK_IMAGE_HEIGHT */
+      1752, /* GL_UNPACK_SKIP_IMAGES */
+      1749, /* GL_UNPACK_IMAGE_HEIGHT */
+      1609, /* GL_TEXTURE_3D */
+      1263, /* GL_PROXY_TEXTURE_3D */
+      1672, /* GL_TEXTURE_DEPTH */
+      1724, /* GL_TEXTURE_WRAP_R */
+       846, /* GL_MAX_3D_TEXTURE_SIZE */
+      1785, /* GL_VERTEX_ARRAY */
+      1013, /* GL_NORMAL_ARRAY */
+       148, /* GL_COLOR_ARRAY */
+       622, /* GL_INDEX_ARRAY */
+      1650, /* GL_TEXTURE_COORD_ARRAY */
+       455, /* GL_EDGE_FLAG_ARRAY */
+      1791, /* GL_VERTEX_ARRAY_SIZE */
+      1793, /* GL_VERTEX_ARRAY_TYPE */
+      1792, /* GL_VERTEX_ARRAY_STRIDE */
+      1018, /* GL_NORMAL_ARRAY_TYPE */
+      1017, /* GL_NORMAL_ARRAY_STRIDE */
+       152, /* GL_COLOR_ARRAY_SIZE */
+       154, /* GL_COLOR_ARRAY_TYPE */
+       153, /* GL_COLOR_ARRAY_STRIDE */
+       627, /* GL_INDEX_ARRAY_TYPE */
+       626, /* GL_INDEX_ARRAY_STRIDE */
+      1654, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+      1656, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+      1655, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+       459, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+      1790, /* GL_VERTEX_ARRAY_POINTER */
+      1016, /* GL_NORMAL_ARRAY_POINTER */
+       151, /* GL_COLOR_ARRAY_POINTER */
+       625, /* GL_INDEX_ARRAY_POINTER */
+      1653, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+       458, /* GL_EDGE_FLAG_ARRAY_POINTER */
+       992, /* GL_MULTISAMPLE */
+      1381, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+      1383, /* GL_SAMPLE_ALPHA_TO_ONE */
+      1388, /* GL_SAMPLE_COVERAGE */
+      1385, /* GL_SAMPLE_BUFFERS */
+      1376, /* GL_SAMPLES */
+      1392, /* GL_SAMPLE_COVERAGE_VALUE */
+      1390, /* GL_SAMPLE_COVERAGE_INVERT */
+       195, /* GL_COLOR_MATRIX */
+       197, /* GL_COLOR_MATRIX_STACK_DEPTH */
+       854, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+      1188, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+      1184, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+      1179, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+      1175, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+      1186, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+      1182, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+      1177, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+      1173, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+      1633, /* GL_TEXTURE_COLOR_TABLE_SGI */
+      1264, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+      1635, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+        80, /* GL_BLEND_DST_RGB */
+        89, /* GL_BLEND_SRC_RGB */
+        79, /* GL_BLEND_DST_ALPHA */
+        88, /* GL_BLEND_SRC_ALPHA */
+       201, /* GL_COLOR_TABLE */
+      1198, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+      1181, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+      1252, /* GL_PROXY_COLOR_TABLE */
+      1256, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+      1255, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+       225, /* GL_COLOR_TABLE_SCALE */
+       205, /* GL_COLOR_TABLE_BIAS */
+       210, /* GL_COLOR_TABLE_FORMAT */
+       227, /* GL_COLOR_TABLE_WIDTH */
+       222, /* GL_COLOR_TABLE_RED_SIZE */
+       213, /* GL_COLOR_TABLE_GREEN_SIZE */
+       207, /* GL_COLOR_TABLE_BLUE_SIZE */
+       202, /* GL_COLOR_TABLE_ALPHA_SIZE */
+       219, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
+       216, /* GL_COLOR_TABLE_INTENSITY_SIZE */
+        71, /* GL_BGR */
+        72, /* GL_BGRA */
+       868, /* GL_MAX_ELEMENTS_VERTICES */
+       867, /* GL_MAX_ELEMENTS_INDICES */
+      1688, /* GL_TEXTURE_INDEX_SIZE_EXT */
+       145, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
+      1145, /* GL_POINT_SIZE_MIN */
+      1141, /* GL_POINT_SIZE_MAX */
+      1135, /* GL_POINT_FADE_THRESHOLD_SIZE */
+      1131, /* GL_POINT_DISTANCE_ATTENUATION */
+       127, /* GL_CLAMP_TO_BORDER */
+       130, /* GL_CLAMP_TO_EDGE */
+      1709, /* GL_TEXTURE_MIN_LOD */
+      1707, /* GL_TEXTURE_MAX_LOD */
+      1612, /* GL_TEXTURE_BASE_LEVEL */
+      1706, /* GL_TEXTURE_MAX_LEVEL */
+       615, /* GL_IGNORE_BORDER_HP */
+       276, /* GL_CONSTANT_BORDER_HP */
+      1323, /* GL_REPLICATE_BORDER_HP */
+       282, /* GL_CONVOLUTION_BORDER_COLOR */
+      1042, /* GL_OCCLUSION_TEST_HP */
+      1043, /* GL_OCCLUSION_TEST_RESULT_HP */
+       687, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+      1627, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+      1629, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+      1631, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+      1632, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1630, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+      1628, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+       850, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+       851, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1208, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+      1210, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+      1207, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+      1209, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+      1696, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+      1697, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+      1695, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+       585, /* GL_GENERATE_MIPMAP */
+       586, /* GL_GENERATE_MIPMAP_HINT */
+       527, /* GL_FOG_OFFSET_SGIX */
+       528, /* GL_FOG_OFFSET_VALUE_SGIX */
+      1641, /* GL_TEXTURE_COMPARE_SGIX */
+      1640, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+      1692, /* GL_TEXTURE_LEQUAL_R_SGIX */
+      1684, /* GL_TEXTURE_GEQUAL_R_SGIX */
+       358, /* GL_DEPTH_COMPONENT16 */
+       361, /* GL_DEPTH_COMPONENT24 */
+       364, /* GL_DEPTH_COMPONENT32 */
+       306, /* GL_CULL_VERTEX_EXT */
+       308, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+       307, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+      1848, /* GL_WRAP_BORDER_SUN */
+      1634, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+       680, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+      1422, /* GL_SINGLE_COLOR */
+      1408, /* GL_SEPARATE_SPECULAR_COLOR */
+      1417, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+       538, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+       539, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+       546, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+       541, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+       537, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+       536, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+       540, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+       547, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+       558, /* GL_FRAMEBUFFER_DEFAULT */
+       571, /* GL_FRAMEBUFFER_UNDEFINED */
+       371, /* GL_DEPTH_STENCIL_ATTACHMENT */
+       621, /* GL_INDEX */
+      1758, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+      1773, /* GL_UNSIGNED_SHORT_5_6_5 */
+      1774, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+      1771, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+      1769, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+      1766, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+      1764, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+      1704, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+      1705, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+      1703, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+       943, /* GL_MIRRORED_REPEAT */
+      1364, /* GL_RGB_S3TC */
+      1339, /* GL_RGB4_S3TC */
+      1362, /* GL_RGBA_S3TC */
+      1356, /* GL_RGBA4_S3TC */
+      1360, /* GL_RGBA_DXT5_S3TC */
+      1354, /* GL_RGBA4_DXT5_S3TC */
+       264, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
+       259, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
+       260, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
+       261, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
+      1004, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+      1003, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+       688, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+       514, /* GL_FOG_COORDINATE_SOURCE */
+       506, /* GL_FOG_COORD */
+       530, /* GL_FRAGMENT_DEPTH */
+       312, /* GL_CURRENT_FOG_COORD */
+       513, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+       512, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+       511, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+       508, /* GL_FOG_COORDINATE_ARRAY */
+       199, /* GL_COLOR_SUM */
+       332, /* GL_CURRENT_SECONDARY_COLOR */
+      1401, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+      1403, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+      1402, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+      1400, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+      1397, /* GL_SECONDARY_COLOR_ARRAY */
+       330, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
         28, /* GL_ALIASED_POINT_SIZE_RANGE */
         27, /* GL_ALIASED_LINE_WIDTH_RANGE */
-      1513, /* GL_TEXTURE0 */
-      1515, /* GL_TEXTURE1 */
-      1537, /* GL_TEXTURE2 */
-      1559, /* GL_TEXTURE3 */
-      1565, /* GL_TEXTURE4 */
-      1567, /* GL_TEXTURE5 */
-      1569, /* GL_TEXTURE6 */
-      1571, /* GL_TEXTURE7 */
-      1573, /* GL_TEXTURE8 */
-      1575, /* GL_TEXTURE9 */
-      1516, /* GL_TEXTURE10 */
-      1518, /* GL_TEXTURE11 */
-      1520, /* GL_TEXTURE12 */
-      1522, /* GL_TEXTURE13 */
-      1524, /* GL_TEXTURE14 */
-      1526, /* GL_TEXTURE15 */
-      1528, /* GL_TEXTURE16 */
-      1530, /* GL_TEXTURE17 */
-      1532, /* GL_TEXTURE18 */
-      1534, /* GL_TEXTURE19 */
-      1538, /* GL_TEXTURE20 */
-      1540, /* GL_TEXTURE21 */
-      1542, /* GL_TEXTURE22 */
-      1544, /* GL_TEXTURE23 */
-      1546, /* GL_TEXTURE24 */
-      1548, /* GL_TEXTURE25 */
-      1550, /* GL_TEXTURE26 */
-      1552, /* GL_TEXTURE27 */
-      1554, /* GL_TEXTURE28 */
-      1556, /* GL_TEXTURE29 */
-      1560, /* GL_TEXTURE30 */
-      1562, /* GL_TEXTURE31 */
+      1541, /* GL_TEXTURE0 */
+      1543, /* GL_TEXTURE1 */
+      1565, /* GL_TEXTURE2 */
+      1587, /* GL_TEXTURE3 */
+      1593, /* GL_TEXTURE4 */
+      1595, /* GL_TEXTURE5 */
+      1597, /* GL_TEXTURE6 */
+      1599, /* GL_TEXTURE7 */
+      1601, /* GL_TEXTURE8 */
+      1603, /* GL_TEXTURE9 */
+      1544, /* GL_TEXTURE10 */
+      1546, /* GL_TEXTURE11 */
+      1548, /* GL_TEXTURE12 */
+      1550, /* GL_TEXTURE13 */
+      1552, /* GL_TEXTURE14 */
+      1554, /* GL_TEXTURE15 */
+      1556, /* GL_TEXTURE16 */
+      1558, /* GL_TEXTURE17 */
+      1560, /* GL_TEXTURE18 */
+      1562, /* GL_TEXTURE19 */
+      1566, /* GL_TEXTURE20 */
+      1568, /* GL_TEXTURE21 */
+      1570, /* GL_TEXTURE22 */
+      1572, /* GL_TEXTURE23 */
+      1574, /* GL_TEXTURE24 */
+      1576, /* GL_TEXTURE25 */
+      1578, /* GL_TEXTURE26 */
+      1580, /* GL_TEXTURE27 */
+      1582, /* GL_TEXTURE28 */
+      1584, /* GL_TEXTURE29 */
+      1588, /* GL_TEXTURE30 */
+      1590, /* GL_TEXTURE31 */
         18, /* GL_ACTIVE_TEXTURE */
-       130, /* GL_CLIENT_ACTIVE_TEXTURE */
-       906, /* GL_MAX_TEXTURE_UNITS */
-      1702, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-      1705, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-      1707, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-      1699, /* GL_TRANSPOSE_COLOR_MATRIX */
-      1501, /* GL_SUBTRACT */
-       894, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
-       244, /* GL_COMPRESSED_ALPHA */
-       248, /* GL_COMPRESSED_LUMINANCE */
-       249, /* GL_COMPRESSED_LUMINANCE_ALPHA */
-       246, /* GL_COMPRESSED_INTENSITY */
-       252, /* GL_COMPRESSED_RGB */
-       253, /* GL_COMPRESSED_RGBA */
-      1620, /* GL_TEXTURE_COMPRESSION_HINT */
-      1682, /* GL_TEXTURE_RECTANGLE_ARB */
-      1592, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-      1250, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-       892, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-       364, /* GL_DEPTH_STENCIL */
-      1727, /* GL_UNSIGNED_INT_24_8 */
-       902, /* GL_MAX_TEXTURE_LOD_BIAS */
-      1673, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-       903, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-      1649, /* GL_TEXTURE_FILTER_CONTROL */
-      1664, /* GL_TEXTURE_LOD_BIAS */
-       229, /* GL_COMBINE4 */
-       896, /* GL_MAX_SHININESS_NV */
-       897, /* GL_MAX_SPOT_EXPONENT_NV */
-       612, /* GL_INCR_WRAP */
-       337, /* GL_DECR_WRAP */
-       948, /* GL_MODELVIEW1_ARB */
-      1004, /* GL_NORMAL_MAP */
-      1280, /* GL_REFLECTION_MAP */
-      1629, /* GL_TEXTURE_CUBE_MAP */
-      1590, /* GL_TEXTURE_BINDING_CUBE_MAP */
-      1637, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-      1631, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-      1639, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-      1633, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-      1641, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-      1635, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-      1248, /* GL_PROXY_TEXTURE_CUBE_MAP */
-       848, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-       983, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-       515, /* GL_FOG_DISTANCE_MODE_NV */
-       468, /* GL_EYE_RADIAL_NV */
-       467, /* GL_EYE_PLANE_ABSOLUTE_NV */
-       228, /* GL_COMBINE */
-       235, /* GL_COMBINE_RGB */
-       230, /* GL_COMBINE_ALPHA */
-      1347, /* GL_RGB_SCALE */
+       133, /* GL_CLIENT_ACTIVE_TEXTURE */
+       921, /* GL_MAX_TEXTURE_UNITS */
+      1736, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+      1739, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+      1741, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+      1733, /* GL_TRANSPOSE_COLOR_MATRIX */
+      1523, /* GL_SUBTRACT */
+       908, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+       247, /* GL_COMPRESSED_ALPHA */
+       251, /* GL_COMPRESSED_LUMINANCE */
+       252, /* GL_COMPRESSED_LUMINANCE_ALPHA */
+       249, /* GL_COMPRESSED_INTENSITY */
+       255, /* GL_COMPRESSED_RGB */
+       256, /* GL_COMPRESSED_RGBA */
+      1648, /* GL_TEXTURE_COMPRESSION_HINT */
+      1713, /* GL_TEXTURE_RECTANGLE_ARB */
+      1620, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+      1267, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+       906, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+       370, /* GL_DEPTH_STENCIL */
+      1762, /* GL_UNSIGNED_INT_24_8 */
+       917, /* GL_MAX_TEXTURE_LOD_BIAS */
+      1702, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+       918, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+      1678, /* GL_TEXTURE_FILTER_CONTROL */
+      1693, /* GL_TEXTURE_LOD_BIAS */
+       232, /* GL_COMBINE4 */
+       911, /* GL_MAX_SHININESS_NV */
+       912, /* GL_MAX_SPOT_EXPONENT_NV */
+       619, /* GL_INCR_WRAP */
+       343, /* GL_DECR_WRAP */
+       963, /* GL_MODELVIEW1_ARB */
+      1019, /* GL_NORMAL_MAP */
+      1298, /* GL_REFLECTION_MAP */
+      1657, /* GL_TEXTURE_CUBE_MAP */
+      1618, /* GL_TEXTURE_BINDING_CUBE_MAP */
+      1665, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+      1659, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+      1667, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+      1661, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+      1669, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+      1663, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+      1265, /* GL_PROXY_TEXTURE_CUBE_MAP */
+       862, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+       998, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+       522, /* GL_FOG_DISTANCE_MODE_NV */
+       474, /* GL_EYE_RADIAL_NV */
+       473, /* GL_EYE_PLANE_ABSOLUTE_NV */
+       231, /* GL_COMBINE */
+       238, /* GL_COMBINE_RGB */
+       233, /* GL_COMBINE_ALPHA */
+      1365, /* GL_RGB_SCALE */
         24, /* GL_ADD_SIGNED */
-       640, /* GL_INTERPOLATE */
-       267, /* GL_CONSTANT */
-      1198, /* GL_PRIMARY_COLOR */
-      1195, /* GL_PREVIOUS */
-      1418, /* GL_SOURCE0_RGB */
-      1424, /* GL_SOURCE1_RGB */
-      1430, /* GL_SOURCE2_RGB */
-      1434, /* GL_SOURCE3_RGB_NV */
-      1415, /* GL_SOURCE0_ALPHA */
-      1421, /* GL_SOURCE1_ALPHA */
-      1427, /* GL_SOURCE2_ALPHA */
-      1433, /* GL_SOURCE3_ALPHA_NV */
-      1040, /* GL_OPERAND0_RGB */
-      1046, /* GL_OPERAND1_RGB */
-      1052, /* GL_OPERAND2_RGB */
-      1056, /* GL_OPERAND3_RGB_NV */
-      1037, /* GL_OPERAND0_ALPHA */
-      1043, /* GL_OPERAND1_ALPHA */
-      1049, /* GL_OPERAND2_ALPHA */
-      1055, /* GL_OPERAND3_ALPHA_NV */
-      1751, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-      1815, /* GL_YCBCR_422_APPLE */
-      1740, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-      1742, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-      1405, /* GL_SLICE_ACCUM_SUN */
-      1255, /* GL_QUAD_MESH_SUN */
-      1711, /* GL_TRIANGLE_MESH_SUN */
-      1789, /* GL_VERTEX_PROGRAM_ARB */
-      1800, /* GL_VERTEX_STATE_PROGRAM_NV */
-      1776, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-      1782, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-      1784, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-      1786, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-       328, /* GL_CURRENT_VERTEX_ATTRIB */
-      1211, /* GL_PROGRAM_LENGTH_ARB */
-      1225, /* GL_PROGRAM_STRING_ARB */
-       970, /* GL_MODELVIEW_PROJECTION_NV */
-       607, /* GL_IDENTITY_NV */
-       654, /* GL_INVERSE_NV */
-      1704, /* GL_TRANSPOSE_NV */
-       655, /* GL_INVERSE_TRANSPOSE_NV */
-       878, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-       877, /* GL_MAX_PROGRAM_MATRICES_ARB */
-       785, /* GL_MATRIX0_NV */
-       797, /* GL_MATRIX1_NV */
-       809, /* GL_MATRIX2_NV */
-       813, /* GL_MATRIX3_NV */
-       815, /* GL_MATRIX4_NV */
-       817, /* GL_MATRIX5_NV */
-       819, /* GL_MATRIX6_NV */
-       821, /* GL_MATRIX7_NV */
-       312, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-       309, /* GL_CURRENT_MATRIX_ARB */
-      1792, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-      1795, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-      1223, /* GL_PROGRAM_PARAMETER_NV */
-      1780, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-      1227, /* GL_PROGRAM_TARGET_NV */
-      1224, /* GL_PROGRAM_RESIDENT_NV */
-      1696, /* GL_TRACK_MATRIX_NV */
-      1697, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-      1790, /* GL_VERTEX_PROGRAM_BINDING_NV */
-      1205, /* GL_PROGRAM_ERROR_POSITION_ARB */
-       349, /* GL_DEPTH_CLAMP_NV */
-      1758, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-      1765, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-      1766, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-      1767, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-      1768, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-      1769, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-      1770, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-      1771, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-      1772, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-      1773, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-      1759, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-      1760, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-      1761, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-      1762, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-      1763, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-      1764, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-       739, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-       746, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-       747, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-       748, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-       749, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-       750, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-       751, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-       752, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-       753, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-       754, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-       740, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-       741, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-       742, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-       743, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-       744, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-       745, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-       766, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-       773, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-       774, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-       775, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-       776, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-       777, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-       778, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-      1204, /* GL_PROGRAM_BINDING_ARB */
-       780, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-       781, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-       767, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-       768, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-       769, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-       770, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-       771, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-       772, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-      1618, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-      1615, /* GL_TEXTURE_COMPRESSED */
-      1009, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
-       266, /* GL_COMPRESSED_TEXTURE_FORMATS */
-       918, /* GL_MAX_VERTEX_UNITS_ARB */
+       647, /* GL_INTERPOLATE */
+       271, /* GL_CONSTANT */
+      1214, /* GL_PRIMARY_COLOR */
+      1211, /* GL_PREVIOUS */
+      1437, /* GL_SOURCE0_RGB */
+      1443, /* GL_SOURCE1_RGB */
+      1449, /* GL_SOURCE2_RGB */
+      1453, /* GL_SOURCE3_RGB_NV */
+      1434, /* GL_SOURCE0_ALPHA */
+      1440, /* GL_SOURCE1_ALPHA */
+      1446, /* GL_SOURCE2_ALPHA */
+      1452, /* GL_SOURCE3_ALPHA_NV */
+      1056, /* GL_OPERAND0_RGB */
+      1062, /* GL_OPERAND1_RGB */
+      1068, /* GL_OPERAND2_RGB */
+      1072, /* GL_OPERAND3_RGB_NV */
+      1053, /* GL_OPERAND0_ALPHA */
+      1059, /* GL_OPERAND1_ALPHA */
+      1065, /* GL_OPERAND2_ALPHA */
+      1071, /* GL_OPERAND3_ALPHA_NV */
+      1786, /* GL_VERTEX_ARRAY_BINDING */
+      1711, /* GL_TEXTURE_RANGE_LENGTH_APPLE */
+      1712, /* GL_TEXTURE_RANGE_POINTER_APPLE */
+      1852, /* GL_YCBCR_422_APPLE */
+      1775, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+      1777, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+      1720, /* GL_TEXTURE_STORAGE_HINT_APPLE */
+      1514, /* GL_STORAGE_PRIVATE_APPLE */
+      1513, /* GL_STORAGE_CACHED_APPLE */
+      1515, /* GL_STORAGE_SHARED_APPLE */
+      1424, /* GL_SLICE_ACCUM_SUN */
+      1273, /* GL_QUAD_MESH_SUN */
+      1745, /* GL_TRIANGLE_MESH_SUN */
+      1825, /* GL_VERTEX_PROGRAM_ARB */
+      1836, /* GL_VERTEX_STATE_PROGRAM_NV */
+      1812, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+      1818, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+      1820, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+      1822, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+       334, /* GL_CURRENT_VERTEX_ATTRIB */
+      1227, /* GL_PROGRAM_LENGTH_ARB */
+      1241, /* GL_PROGRAM_STRING_ARB */
+       985, /* GL_MODELVIEW_PROJECTION_NV */
+       614, /* GL_IDENTITY_NV */
+       661, /* GL_INVERSE_NV */
+      1738, /* GL_TRANSPOSE_NV */
+       662, /* GL_INVERSE_TRANSPOSE_NV */
+       892, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+       891, /* GL_MAX_PROGRAM_MATRICES_ARB */
+       799, /* GL_MATRIX0_NV */
+       811, /* GL_MATRIX1_NV */
+       823, /* GL_MATRIX2_NV */
+       827, /* GL_MATRIX3_NV */
+       829, /* GL_MATRIX4_NV */
+       831, /* GL_MATRIX5_NV */
+       833, /* GL_MATRIX6_NV */
+       835, /* GL_MATRIX7_NV */
+       318, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+       315, /* GL_CURRENT_MATRIX_ARB */
+      1828, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+      1831, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+      1239, /* GL_PROGRAM_PARAMETER_NV */
+      1816, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+      1243, /* GL_PROGRAM_TARGET_NV */
+      1240, /* GL_PROGRAM_RESIDENT_NV */
+      1730, /* GL_TRACK_MATRIX_NV */
+      1731, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+      1826, /* GL_VERTEX_PROGRAM_BINDING_NV */
+      1221, /* GL_PROGRAM_ERROR_POSITION_ARB */
+       355, /* GL_DEPTH_CLAMP_NV */
+      1794, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+      1801, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+      1802, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+      1803, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+      1804, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+      1805, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+      1806, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+      1807, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+      1808, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+      1809, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+      1795, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+      1796, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+      1797, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+      1798, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+      1799, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+      1800, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+       747, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+       754, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+       755, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+       756, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+       757, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+       758, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+       759, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+       760, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+       761, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+       762, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+       748, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+       749, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+       750, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+       751, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+       752, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+       753, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+       774, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+       781, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+       782, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+       783, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+       784, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+       785, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+       786, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+      1220, /* GL_PROGRAM_BINDING_ARB */
+       788, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+       789, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+       775, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+       776, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+       777, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+       778, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+       779, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+       780, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+      1646, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+      1643, /* GL_TEXTURE_COMPRESSED */
+      1024, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+       269, /* GL_COMPRESSED_TEXTURE_FORMATS */
+       933, /* GL_MAX_VERTEX_UNITS_ARB */
         22, /* GL_ACTIVE_VERTEX_UNITS_ARB */
-      1810, /* GL_WEIGHT_SUM_UNITY_ARB */
-      1788, /* GL_VERTEX_BLEND_ARB */
-       330, /* GL_CURRENT_WEIGHT_ARB */
-      1809, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-      1808, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-      1807, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-      1806, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-      1803, /* GL_WEIGHT_ARRAY_ARB */
-       377, /* GL_DOT3_RGB */
-       378, /* GL_DOT3_RGBA */
-       260, /* GL_COMPRESSED_RGB_FXT1_3DFX */
-       255, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
-       978, /* GL_MULTISAMPLE_3DFX */
-      1368, /* GL_SAMPLE_BUFFERS_3DFX */
-      1359, /* GL_SAMPLES_3DFX */
-       959, /* GL_MODELVIEW2_ARB */
-       962, /* GL_MODELVIEW3_ARB */
-       963, /* GL_MODELVIEW4_ARB */
-       964, /* GL_MODELVIEW5_ARB */
-       965, /* GL_MODELVIEW6_ARB */
-       966, /* GL_MODELVIEW7_ARB */
-       967, /* GL_MODELVIEW8_ARB */
-       968, /* GL_MODELVIEW9_ARB */
-       938, /* GL_MODELVIEW10_ARB */
-       939, /* GL_MODELVIEW11_ARB */
-       940, /* GL_MODELVIEW12_ARB */
-       941, /* GL_MODELVIEW13_ARB */
-       942, /* GL_MODELVIEW14_ARB */
-       943, /* GL_MODELVIEW15_ARB */
-       944, /* GL_MODELVIEW16_ARB */
-       945, /* GL_MODELVIEW17_ARB */
-       946, /* GL_MODELVIEW18_ARB */
-       947, /* GL_MODELVIEW19_ARB */
-       949, /* GL_MODELVIEW20_ARB */
-       950, /* GL_MODELVIEW21_ARB */
-       951, /* GL_MODELVIEW22_ARB */
-       952, /* GL_MODELVIEW23_ARB */
-       953, /* GL_MODELVIEW24_ARB */
-       954, /* GL_MODELVIEW25_ARB */
-       955, /* GL_MODELVIEW26_ARB */
-       956, /* GL_MODELVIEW27_ARB */
-       957, /* GL_MODELVIEW28_ARB */
-       958, /* GL_MODELVIEW29_ARB */
-       960, /* GL_MODELVIEW30_ARB */
-       961, /* GL_MODELVIEW31_ARB */
-       382, /* GL_DOT3_RGB_EXT */
-       380, /* GL_DOT3_RGBA_EXT */
-       932, /* GL_MIRROR_CLAMP_EXT */
-       935, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-       973, /* GL_MODULATE_ADD_ATI */
-       974, /* GL_MODULATE_SIGNED_ADD_ATI */
-       975, /* GL_MODULATE_SUBTRACT_ATI */
-      1816, /* GL_YCBCR_MESA */
-      1064, /* GL_PACK_INVERT_MESA */
-       333, /* GL_DEBUG_OBJECT_MESA */
-       334, /* GL_DEBUG_PRINT_MESA */
-       332, /* GL_DEBUG_ASSERT_MESA */
-       107, /* GL_BUFFER_SIZE */
-       109, /* GL_BUFFER_USAGE */
-       113, /* GL_BUMP_ROT_MATRIX_ATI */
-       114, /* GL_BUMP_ROT_MATRIX_SIZE_ATI */
-       112, /* GL_BUMP_NUM_TEX_UNITS_ATI */
-       116, /* GL_BUMP_TEX_UNITS_ATI */
-       441, /* GL_DUDV_ATI */
-       440, /* GL_DU8DV8_ATI */
-       111, /* GL_BUMP_ENVMAP_ATI */
-       115, /* GL_BUMP_TARGET_ATI */
-      1466, /* GL_STENCIL_BACK_FUNC */
-      1464, /* GL_STENCIL_BACK_FAIL */
-      1468, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-      1470, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-       524, /* GL_FRAGMENT_PROGRAM_ARB */
-      1202, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-      1230, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-      1229, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-      1214, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-      1220, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-      1219, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       867, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-       890, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-       889, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-       880, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-       886, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-       885, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       850, /* GL_MAX_DRAW_BUFFERS */
-       386, /* GL_DRAW_BUFFER0 */
-       389, /* GL_DRAW_BUFFER1 */
-       410, /* GL_DRAW_BUFFER2 */
-       413, /* GL_DRAW_BUFFER3 */
-       416, /* GL_DRAW_BUFFER4 */
-       419, /* GL_DRAW_BUFFER5 */
-       422, /* GL_DRAW_BUFFER6 */
-       425, /* GL_DRAW_BUFFER7 */
-       428, /* GL_DRAW_BUFFER8 */
-       431, /* GL_DRAW_BUFFER9 */
-       390, /* GL_DRAW_BUFFER10 */
-       393, /* GL_DRAW_BUFFER11 */
-       396, /* GL_DRAW_BUFFER12 */
-       399, /* GL_DRAW_BUFFER13 */
-       402, /* GL_DRAW_BUFFER14 */
-       405, /* GL_DRAW_BUFFER15 */
-        81, /* GL_BLEND_EQUATION_ALPHA */
-       830, /* GL_MATRIX_PALETTE_ARB */
-       861, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-       864, /* GL_MAX_PALETTE_MATRICES_ARB */
-       315, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-       824, /* GL_MATRIX_INDEX_ARRAY_ARB */
-       310, /* GL_CURRENT_MATRIX_INDEX_ARB */
-       826, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-       828, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-       827, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-       825, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-      1644, /* GL_TEXTURE_DEPTH_SIZE */
-       370, /* GL_DEPTH_TEXTURE_MODE */
-      1610, /* GL_TEXTURE_COMPARE_MODE */
-      1608, /* GL_TEXTURE_COMPARE_FUNC */
-       239, /* GL_COMPARE_R_TO_TEXTURE */
-      1136, /* GL_POINT_SPRITE */
-       292, /* GL_COORD_REPLACE */
-      1140, /* GL_POINT_SPRITE_R_MODE_NV */
-      1257, /* GL_QUERY_COUNTER_BITS */
-       317, /* GL_CURRENT_QUERY */
-      1259, /* GL_QUERY_RESULT */
-      1261, /* GL_QUERY_RESULT_AVAILABLE */
-       912, /* GL_MAX_VERTEX_ATTRIBS */
-      1778, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-       368, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-       367, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-       898, /* GL_MAX_TEXTURE_COORDS */
-       900, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-      1207, /* GL_PROGRAM_ERROR_STRING_ARB */
-      1209, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-      1208, /* GL_PROGRAM_FORMAT_ARB */
-      1690, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-       347, /* GL_DEPTH_BOUNDS_TEST_EXT */
-       346, /* GL_DEPTH_BOUNDS_EXT */
-        52, /* GL_ARRAY_BUFFER */
-       454, /* GL_ELEMENT_ARRAY_BUFFER */
-        53, /* GL_ARRAY_BUFFER_BINDING */
-       455, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-      1752, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-       999, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
-       146, /* GL_COLOR_ARRAY_BUFFER_BINDING */
-       616, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-      1623, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-       450, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-      1380, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-       502, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-      1804, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-      1774, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-      1210, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-       873, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-      1216, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-       882, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-      1228, /* GL_PROGRAM_TEMPORARIES_ARB */
-       888, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-      1218, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-       884, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-      1222, /* GL_PROGRAM_PARAMETERS_ARB */
-       887, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-      1217, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-       883, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-      1203, /* GL_PROGRAM_ATTRIBS_ARB */
-       868, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-      1215, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-       881, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-      1201, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-       866, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-      1213, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       879, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       874, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-       870, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-      1231, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-      1701, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-      1270, /* GL_READ_ONLY */
-      1812, /* GL_WRITE_ONLY */
-      1272, /* GL_READ_WRITE */
-       101, /* GL_BUFFER_ACCESS */
-       103, /* GL_BUFFER_MAPPED */
-       105, /* GL_BUFFER_MAP_POINTER */
-      1695, /* GL_TIME_ELAPSED_EXT */
-       784, /* GL_MATRIX0_ARB */
-       796, /* GL_MATRIX1_ARB */
-       808, /* GL_MATRIX2_ARB */
-       812, /* GL_MATRIX3_ARB */
-       814, /* GL_MATRIX4_ARB */
-       816, /* GL_MATRIX5_ARB */
-       818, /* GL_MATRIX6_ARB */
-       820, /* GL_MATRIX7_ARB */
-       822, /* GL_MATRIX8_ARB */
-       823, /* GL_MATRIX9_ARB */
-       786, /* GL_MATRIX10_ARB */
-       787, /* GL_MATRIX11_ARB */
-       788, /* GL_MATRIX12_ARB */
-       789, /* GL_MATRIX13_ARB */
-       790, /* GL_MATRIX14_ARB */
-       791, /* GL_MATRIX15_ARB */
-       792, /* GL_MATRIX16_ARB */
-       793, /* GL_MATRIX17_ARB */
-       794, /* GL_MATRIX18_ARB */
-       795, /* GL_MATRIX19_ARB */
-       798, /* GL_MATRIX20_ARB */
-       799, /* GL_MATRIX21_ARB */
-       800, /* GL_MATRIX22_ARB */
-       801, /* GL_MATRIX23_ARB */
-       802, /* GL_MATRIX24_ARB */
-       803, /* GL_MATRIX25_ARB */
-       804, /* GL_MATRIX26_ARB */
-       805, /* GL_MATRIX27_ARB */
-       806, /* GL_MATRIX28_ARB */
-       807, /* GL_MATRIX29_ARB */
-       810, /* GL_MATRIX30_ARB */
-       811, /* GL_MATRIX31_ARB */
-      1496, /* GL_STREAM_DRAW */
-      1498, /* GL_STREAM_READ */
-      1494, /* GL_STREAM_COPY */
-      1457, /* GL_STATIC_DRAW */
-      1459, /* GL_STATIC_READ */
-      1455, /* GL_STATIC_COPY */
-       444, /* GL_DYNAMIC_DRAW */
-       446, /* GL_DYNAMIC_READ */
-       442, /* GL_DYNAMIC_COPY */
-      1104, /* GL_PIXEL_PACK_BUFFER */
-      1108, /* GL_PIXEL_UNPACK_BUFFER */
-      1105, /* GL_PIXEL_PACK_BUFFER_BINDING */
-      1109, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
-       341, /* GL_DEPTH24_STENCIL8 */
-      1688, /* GL_TEXTURE_STENCIL_SIZE */
-       871, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-       869, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-       872, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-       876, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-       875, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-       833, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-      1490, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+      1847, /* GL_WEIGHT_SUM_UNITY_ARB */
+      1824, /* GL_VERTEX_BLEND_ARB */
+       336, /* GL_CURRENT_WEIGHT_ARB */
+      1846, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+      1845, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+      1844, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+      1843, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+      1840, /* GL_WEIGHT_ARRAY_ARB */
+       383, /* GL_DOT3_RGB */
+       384, /* GL_DOT3_RGBA */
+       263, /* GL_COMPRESSED_RGB_FXT1_3DFX */
+       258, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
+       993, /* GL_MULTISAMPLE_3DFX */
+      1386, /* GL_SAMPLE_BUFFERS_3DFX */
+      1377, /* GL_SAMPLES_3DFX */
+       974, /* GL_MODELVIEW2_ARB */
+       977, /* GL_MODELVIEW3_ARB */
+       978, /* GL_MODELVIEW4_ARB */
+       979, /* GL_MODELVIEW5_ARB */
+       980, /* GL_MODELVIEW6_ARB */
+       981, /* GL_MODELVIEW7_ARB */
+       982, /* GL_MODELVIEW8_ARB */
+       983, /* GL_MODELVIEW9_ARB */
+       953, /* GL_MODELVIEW10_ARB */
+       954, /* GL_MODELVIEW11_ARB */
+       955, /* GL_MODELVIEW12_ARB */
+       956, /* GL_MODELVIEW13_ARB */
+       957, /* GL_MODELVIEW14_ARB */
+       958, /* GL_MODELVIEW15_ARB */
+       959, /* GL_MODELVIEW16_ARB */
+       960, /* GL_MODELVIEW17_ARB */
+       961, /* GL_MODELVIEW18_ARB */
+       962, /* GL_MODELVIEW19_ARB */
+       964, /* GL_MODELVIEW20_ARB */
+       965, /* GL_MODELVIEW21_ARB */
+       966, /* GL_MODELVIEW22_ARB */
+       967, /* GL_MODELVIEW23_ARB */
+       968, /* GL_MODELVIEW24_ARB */
+       969, /* GL_MODELVIEW25_ARB */
+       970, /* GL_MODELVIEW26_ARB */
+       971, /* GL_MODELVIEW27_ARB */
+       972, /* GL_MODELVIEW28_ARB */
+       973, /* GL_MODELVIEW29_ARB */
+       975, /* GL_MODELVIEW30_ARB */
+       976, /* GL_MODELVIEW31_ARB */
+       388, /* GL_DOT3_RGB_EXT */
+       386, /* GL_DOT3_RGBA_EXT */
+       947, /* GL_MIRROR_CLAMP_EXT */
+       950, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+       988, /* GL_MODULATE_ADD_ATI */
+       989, /* GL_MODULATE_SIGNED_ADD_ATI */
+       990, /* GL_MODULATE_SUBTRACT_ATI */
+      1853, /* GL_YCBCR_MESA */
+      1080, /* GL_PACK_INVERT_MESA */
+       339, /* GL_DEBUG_OBJECT_MESA */
+       340, /* GL_DEBUG_PRINT_MESA */
+       338, /* GL_DEBUG_ASSERT_MESA */
+       110, /* GL_BUFFER_SIZE */
+       112, /* GL_BUFFER_USAGE */
+       116, /* GL_BUMP_ROT_MATRIX_ATI */
+       117, /* GL_BUMP_ROT_MATRIX_SIZE_ATI */
+       115, /* GL_BUMP_NUM_TEX_UNITS_ATI */
+       119, /* GL_BUMP_TEX_UNITS_ATI */
+       447, /* GL_DUDV_ATI */
+       446, /* GL_DU8DV8_ATI */
+       114, /* GL_BUMP_ENVMAP_ATI */
+       118, /* GL_BUMP_TARGET_ATI */
+      1485, /* GL_STENCIL_BACK_FUNC */
+      1483, /* GL_STENCIL_BACK_FAIL */
+      1487, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+      1489, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+       531, /* GL_FRAGMENT_PROGRAM_ARB */
+      1218, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+      1246, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+      1245, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+      1230, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+      1236, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+      1235, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       881, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+       904, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+       903, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+       894, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+       900, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+       899, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       864, /* GL_MAX_DRAW_BUFFERS */
+       392, /* GL_DRAW_BUFFER0 */
+       395, /* GL_DRAW_BUFFER1 */
+       416, /* GL_DRAW_BUFFER2 */
+       419, /* GL_DRAW_BUFFER3 */
+       422, /* GL_DRAW_BUFFER4 */
+       425, /* GL_DRAW_BUFFER5 */
+       428, /* GL_DRAW_BUFFER6 */
+       431, /* GL_DRAW_BUFFER7 */
+       434, /* GL_DRAW_BUFFER8 */
+       437, /* GL_DRAW_BUFFER9 */
+       396, /* GL_DRAW_BUFFER10 */
+       399, /* GL_DRAW_BUFFER11 */
+       402, /* GL_DRAW_BUFFER12 */
+       405, /* GL_DRAW_BUFFER13 */
+       408, /* GL_DRAW_BUFFER14 */
+       411, /* GL_DRAW_BUFFER15 */
+        82, /* GL_BLEND_EQUATION_ALPHA */
+       844, /* GL_MATRIX_PALETTE_ARB */
+       875, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+       878, /* GL_MAX_PALETTE_MATRICES_ARB */
+       321, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+       838, /* GL_MATRIX_INDEX_ARRAY_ARB */
+       316, /* GL_CURRENT_MATRIX_INDEX_ARB */
+       840, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+       842, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+       841, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+       839, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+      1673, /* GL_TEXTURE_DEPTH_SIZE */
+       376, /* GL_DEPTH_TEXTURE_MODE */
+      1638, /* GL_TEXTURE_COMPARE_MODE */
+      1636, /* GL_TEXTURE_COMPARE_FUNC */
+       242, /* GL_COMPARE_R_TO_TEXTURE */
+      1152, /* GL_POINT_SPRITE */
+       296, /* GL_COORD_REPLACE */
+      1156, /* GL_POINT_SPRITE_R_MODE_NV */
+      1275, /* GL_QUERY_COUNTER_BITS */
+       323, /* GL_CURRENT_QUERY */
+      1277, /* GL_QUERY_RESULT */
+      1279, /* GL_QUERY_RESULT_AVAILABLE */
+       927, /* GL_MAX_VERTEX_ATTRIBS */
+      1814, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+       374, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+       373, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+       913, /* GL_MAX_TEXTURE_COORDS */
+       915, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+      1223, /* GL_PROGRAM_ERROR_STRING_ARB */
+      1225, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+      1224, /* GL_PROGRAM_FORMAT_ARB */
+      1722, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+       353, /* GL_DEPTH_BOUNDS_TEST_EXT */
+       352, /* GL_DEPTH_BOUNDS_EXT */
+        53, /* GL_ARRAY_BUFFER */
+       460, /* GL_ELEMENT_ARRAY_BUFFER */
+        54, /* GL_ARRAY_BUFFER_BINDING */
+       461, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+      1788, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+      1014, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+       149, /* GL_COLOR_ARRAY_BUFFER_BINDING */
+       623, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+      1651, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+       456, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+      1398, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+       509, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+      1841, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+      1810, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+      1226, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+       887, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+      1232, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+       896, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+      1244, /* GL_PROGRAM_TEMPORARIES_ARB */
+       902, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+      1234, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+       898, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+      1238, /* GL_PROGRAM_PARAMETERS_ARB */
+       901, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+      1233, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+       897, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+      1219, /* GL_PROGRAM_ATTRIBS_ARB */
+       882, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+      1231, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+       895, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+      1217, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+       880, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+      1229, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       893, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       888, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+       884, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+      1247, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+      1735, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+      1288, /* GL_READ_ONLY */
+      1849, /* GL_WRITE_ONLY */
+      1290, /* GL_READ_WRITE */
+       102, /* GL_BUFFER_ACCESS */
+       105, /* GL_BUFFER_MAPPED */
+       107, /* GL_BUFFER_MAP_POINTER */
+      1729, /* GL_TIME_ELAPSED_EXT */
+       798, /* GL_MATRIX0_ARB */
+       810, /* GL_MATRIX1_ARB */
+       822, /* GL_MATRIX2_ARB */
+       826, /* GL_MATRIX3_ARB */
+       828, /* GL_MATRIX4_ARB */
+       830, /* GL_MATRIX5_ARB */
+       832, /* GL_MATRIX6_ARB */
+       834, /* GL_MATRIX7_ARB */
+       836, /* GL_MATRIX8_ARB */
+       837, /* GL_MATRIX9_ARB */
+       800, /* GL_MATRIX10_ARB */
+       801, /* GL_MATRIX11_ARB */
+       802, /* GL_MATRIX12_ARB */
+       803, /* GL_MATRIX13_ARB */
+       804, /* GL_MATRIX14_ARB */
+       805, /* GL_MATRIX15_ARB */
+       806, /* GL_MATRIX16_ARB */
+       807, /* GL_MATRIX17_ARB */
+       808, /* GL_MATRIX18_ARB */
+       809, /* GL_MATRIX19_ARB */
+       812, /* GL_MATRIX20_ARB */
+       813, /* GL_MATRIX21_ARB */
+       814, /* GL_MATRIX22_ARB */
+       815, /* GL_MATRIX23_ARB */
+       816, /* GL_MATRIX24_ARB */
+       817, /* GL_MATRIX25_ARB */
+       818, /* GL_MATRIX26_ARB */
+       819, /* GL_MATRIX27_ARB */
+       820, /* GL_MATRIX28_ARB */
+       821, /* GL_MATRIX29_ARB */
+       824, /* GL_MATRIX30_ARB */
+       825, /* GL_MATRIX31_ARB */
+      1518, /* GL_STREAM_DRAW */
+      1520, /* GL_STREAM_READ */
+      1516, /* GL_STREAM_COPY */
+      1476, /* GL_STATIC_DRAW */
+      1478, /* GL_STATIC_READ */
+      1474, /* GL_STATIC_COPY */
+       450, /* GL_DYNAMIC_DRAW */
+       452, /* GL_DYNAMIC_READ */
+       448, /* GL_DYNAMIC_COPY */
+      1120, /* GL_PIXEL_PACK_BUFFER */
+      1124, /* GL_PIXEL_UNPACK_BUFFER */
+      1121, /* GL_PIXEL_PACK_BUFFER_BINDING */
+      1125, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
+       347, /* GL_DEPTH24_STENCIL8 */
+      1719, /* GL_TEXTURE_STENCIL_SIZE */
+      1671, /* GL_TEXTURE_CUBE_MAP_SEAMLESS */
+       883, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+       886, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+       890, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+       889, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+       847, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+      1509, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
         17, /* GL_ACTIVE_STENCIL_FACE_EXT */
-       933, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-      1361, /* GL_SAMPLES_PASSED */
-       525, /* GL_FRAGMENT_SHADER */
-      1798, /* GL_VERTEX_SHADER */
-      1221, /* GL_PROGRAM_OBJECT_ARB */
-      1393, /* GL_SHADER_OBJECT_ARB */
-       857, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-       916, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-       910, /* GL_MAX_VARYING_FLOATS */
-       914, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-       842, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-      1024, /* GL_OBJECT_TYPE_ARB */
-      1395, /* GL_SHADER_TYPE */
-       490, /* GL_FLOAT_VEC2 */
-       492, /* GL_FLOAT_VEC3 */
-       494, /* GL_FLOAT_VEC4 */
-       643, /* GL_INT_VEC2 */
-       645, /* GL_INT_VEC3 */
-       647, /* GL_INT_VEC4 */
-        93, /* GL_BOOL */
-        95, /* GL_BOOL_VEC2 */
-        97, /* GL_BOOL_VEC3 */
-        99, /* GL_BOOL_VEC4 */
-       478, /* GL_FLOAT_MAT2 */
-       482, /* GL_FLOAT_MAT3 */
-       486, /* GL_FLOAT_MAT4 */
-      1352, /* GL_SAMPLER_1D */
-      1354, /* GL_SAMPLER_2D */
-      1356, /* GL_SAMPLER_3D */
-      1357, /* GL_SAMPLER_CUBE */
-      1353, /* GL_SAMPLER_1D_SHADOW */
-      1355, /* GL_SAMPLER_2D_SHADOW */
-       480, /* GL_FLOAT_MAT2x3 */
-       481, /* GL_FLOAT_MAT2x4 */
-       484, /* GL_FLOAT_MAT3x2 */
-       485, /* GL_FLOAT_MAT3x4 */
-       488, /* GL_FLOAT_MAT4x2 */
-       489, /* GL_FLOAT_MAT4x3 */
-       339, /* GL_DELETE_STATUS */
-       243, /* GL_COMPILE_STATUS */
-       697, /* GL_LINK_STATUS */
-      1747, /* GL_VALIDATE_STATUS */
-       628, /* GL_INFO_LOG_LENGTH */
-        55, /* GL_ATTACHED_SHADERS */
+       948, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+      1379, /* GL_SAMPLES_PASSED */
+       109, /* GL_BUFFER_SERIALIZED_MODIFY_APPLE */
+       104, /* GL_BUFFER_FLUSHING_UNMAP_APPLE */
+       532, /* GL_FRAGMENT_SHADER */
+      1834, /* GL_VERTEX_SHADER */
+      1237, /* GL_PROGRAM_OBJECT_ARB */
+      1411, /* GL_SHADER_OBJECT_ARB */
+       871, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+       931, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+       925, /* GL_MAX_VARYING_FLOATS */
+       929, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+       856, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+      1040, /* GL_OBJECT_TYPE_ARB */
+      1413, /* GL_SHADER_TYPE */
+       497, /* GL_FLOAT_VEC2 */
+       499, /* GL_FLOAT_VEC3 */
+       501, /* GL_FLOAT_VEC4 */
+       650, /* GL_INT_VEC2 */
+       652, /* GL_INT_VEC3 */
+       654, /* GL_INT_VEC4 */
+        94, /* GL_BOOL */
+        96, /* GL_BOOL_VEC2 */
+        98, /* GL_BOOL_VEC3 */
+       100, /* GL_BOOL_VEC4 */
+       485, /* GL_FLOAT_MAT2 */
+       489, /* GL_FLOAT_MAT3 */
+       493, /* GL_FLOAT_MAT4 */
+      1370, /* GL_SAMPLER_1D */
+      1372, /* GL_SAMPLER_2D */
+      1374, /* GL_SAMPLER_3D */
+      1375, /* GL_SAMPLER_CUBE */
+      1371, /* GL_SAMPLER_1D_SHADOW */
+      1373, /* GL_SAMPLER_2D_SHADOW */
+       487, /* GL_FLOAT_MAT2x3 */
+       488, /* GL_FLOAT_MAT2x4 */
+       491, /* GL_FLOAT_MAT3x2 */
+       492, /* GL_FLOAT_MAT3x4 */
+       495, /* GL_FLOAT_MAT4x2 */
+       496, /* GL_FLOAT_MAT4x3 */
+       345, /* GL_DELETE_STATUS */
+       246, /* GL_COMPILE_STATUS */
+       705, /* GL_LINK_STATUS */
+      1782, /* GL_VALIDATE_STATUS */
+       635, /* GL_INFO_LOG_LENGTH */
+        56, /* GL_ATTACHED_SHADERS */
         20, /* GL_ACTIVE_UNIFORMS */
         21, /* GL_ACTIVE_UNIFORM_MAX_LENGTH */
-      1394, /* GL_SHADER_SOURCE_LENGTH */
+      1412, /* GL_SHADER_SOURCE_LENGTH */
         15, /* GL_ACTIVE_ATTRIBUTES */
         16, /* GL_ACTIVE_ATTRIBUTE_MAX_LENGTH */
-       527, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-      1397, /* GL_SHADING_LANGUAGE_VERSION */
-       316, /* GL_CURRENT_PROGRAM */
-      1073, /* GL_PALETTE4_RGB8_OES */
-      1075, /* GL_PALETTE4_RGBA8_OES */
-      1071, /* GL_PALETTE4_R5_G6_B5_OES */
-      1074, /* GL_PALETTE4_RGBA4_OES */
-      1072, /* GL_PALETTE4_RGB5_A1_OES */
-      1078, /* GL_PALETTE8_RGB8_OES */
-      1080, /* GL_PALETTE8_RGBA8_OES */
-      1076, /* GL_PALETTE8_R5_G6_B5_OES */
-      1079, /* GL_PALETTE8_RGBA4_OES */
-      1077, /* GL_PALETTE8_RGB5_A1_OES */
-       610, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-       609, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-      1732, /* GL_UNSIGNED_NORMALIZED */
-      1578, /* GL_TEXTURE_1D_ARRAY_EXT */
-      1241, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-      1580, /* GL_TEXTURE_2D_ARRAY_EXT */
-      1244, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-      1586, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-      1588, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-      1449, /* GL_SRGB */
-      1450, /* GL_SRGB8 */
-      1452, /* GL_SRGB_ALPHA */
-      1451, /* GL_SRGB8_ALPHA8 */
-      1409, /* GL_SLUMINANCE_ALPHA */
-      1408, /* GL_SLUMINANCE8_ALPHA8 */
-      1406, /* GL_SLUMINANCE */
-      1407, /* GL_SLUMINANCE8 */
-       264, /* GL_COMPRESSED_SRGB */
-       265, /* GL_COMPRESSED_SRGB_ALPHA */
-       262, /* GL_COMPRESSED_SLUMINANCE */
-       263, /* GL_COMPRESSED_SLUMINANCE_ALPHA */
-      1138, /* GL_POINT_SPRITE_COORD_ORIGIN */
-       705, /* GL_LOWER_LEFT */
-      1744, /* GL_UPPER_LEFT */
-      1472, /* GL_STENCIL_BACK_REF */
-      1473, /* GL_STENCIL_BACK_VALUE_MASK */
-      1474, /* GL_STENCIL_BACK_WRITEMASK */
-       435, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-      1286, /* GL_RENDERBUFFER_BINDING_EXT */
-      1267, /* GL_READ_FRAMEBUFFER */
-       434, /* GL_DRAW_FRAMEBUFFER */
-      1268, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-      1296, /* GL_RENDERBUFFER_SAMPLES */
-       537, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
-       535, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
-       546, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
-       542, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
-       544, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
-       549, /* GL_FRAMEBUFFER_COMPLETE */
-       553, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
-       559, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
-       557, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-       555, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-       558, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-       556, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-       562, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-       565, /* GL_FRAMEBUFFER_UNSUPPORTED */
-       563, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-       839, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
-       152, /* GL_COLOR_ATTACHMENT0 */
-       154, /* GL_COLOR_ATTACHMENT1 */
-       168, /* GL_COLOR_ATTACHMENT2 */
-       170, /* GL_COLOR_ATTACHMENT3 */
-       172, /* GL_COLOR_ATTACHMENT4 */
-       174, /* GL_COLOR_ATTACHMENT5 */
-       176, /* GL_COLOR_ATTACHMENT6 */
-       178, /* GL_COLOR_ATTACHMENT7 */
-       180, /* GL_COLOR_ATTACHMENT8 */
-       182, /* GL_COLOR_ATTACHMENT9 */
-       155, /* GL_COLOR_ATTACHMENT10 */
-       157, /* GL_COLOR_ATTACHMENT11 */
-       159, /* GL_COLOR_ATTACHMENT12 */
-       161, /* GL_COLOR_ATTACHMENT13 */
-       163, /* GL_COLOR_ATTACHMENT14 */
-       165, /* GL_COLOR_ATTACHMENT15 */
-       342, /* GL_DEPTH_ATTACHMENT */
-      1462, /* GL_STENCIL_ATTACHMENT */
-       528, /* GL_FRAMEBUFFER */
-      1284, /* GL_RENDERBUFFER */
-      1298, /* GL_RENDERBUFFER_WIDTH */
-      1291, /* GL_RENDERBUFFER_HEIGHT */
-      1293, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
-      1485, /* GL_STENCIL_INDEX_EXT */
-      1482, /* GL_STENCIL_INDEX1_EXT */
-      1483, /* GL_STENCIL_INDEX4_EXT */
-      1484, /* GL_STENCIL_INDEX8_EXT */
-      1481, /* GL_STENCIL_INDEX16_EXT */
-      1295, /* GL_RENDERBUFFER_RED_SIZE */
-      1290, /* GL_RENDERBUFFER_GREEN_SIZE */
-      1287, /* GL_RENDERBUFFER_BLUE_SIZE */
-      1285, /* GL_RENDERBUFFER_ALPHA_SIZE */
-      1288, /* GL_RENDERBUFFER_DEPTH_SIZE */
-      1297, /* GL_RENDERBUFFER_STENCIL_SIZE */
-       561, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
-       895, /* GL_MAX_SAMPLES */
-      1345, /* GL_RGBA_SNORM */
-      1341, /* GL_RGBA8_SNORM */
-      1402, /* GL_SIGNED_NORMALIZED */
-       461, /* GL_EVAL_BIT */
-      1265, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-       699, /* GL_LIST_BIT */
-      1594, /* GL_TEXTURE_BIT */
-      1376, /* GL_SCISSOR_BIT */
+       534, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+      1415, /* GL_SHADING_LANGUAGE_VERSION */
+       322, /* GL_CURRENT_PROGRAM */
+      1089, /* GL_PALETTE4_RGB8_OES */
+      1091, /* GL_PALETTE4_RGBA8_OES */
+      1087, /* GL_PALETTE4_R5_G6_B5_OES */
+      1090, /* GL_PALETTE4_RGBA4_OES */
+      1088, /* GL_PALETTE4_RGB5_A1_OES */
+      1094, /* GL_PALETTE8_RGB8_OES */
+      1096, /* GL_PALETTE8_RGBA8_OES */
+      1092, /* GL_PALETTE8_R5_G6_B5_OES */
+      1095, /* GL_PALETTE8_RGBA4_OES */
+      1093, /* GL_PALETTE8_RGB5_A1_OES */
+       617, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+       616, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+      1767, /* GL_UNSIGNED_NORMALIZED */
+      1606, /* GL_TEXTURE_1D_ARRAY_EXT */
+      1258, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+      1608, /* GL_TEXTURE_2D_ARRAY_EXT */
+      1261, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+      1614, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+      1616, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+      1468, /* GL_SRGB */
+      1469, /* GL_SRGB8 */
+      1471, /* GL_SRGB_ALPHA */
+      1470, /* GL_SRGB8_ALPHA8 */
+      1428, /* GL_SLUMINANCE_ALPHA */
+      1427, /* GL_SLUMINANCE8_ALPHA8 */
+      1425, /* GL_SLUMINANCE */
+      1426, /* GL_SLUMINANCE8 */
+       267, /* GL_COMPRESSED_SRGB */
+       268, /* GL_COMPRESSED_SRGB_ALPHA */
+       265, /* GL_COMPRESSED_SLUMINANCE */
+       266, /* GL_COMPRESSED_SLUMINANCE_ALPHA */
+      1154, /* GL_POINT_SPRITE_COORD_ORIGIN */
+       713, /* GL_LOWER_LEFT */
+      1779, /* GL_UPPER_LEFT */
+      1491, /* GL_STENCIL_BACK_REF */
+      1492, /* GL_STENCIL_BACK_VALUE_MASK */
+      1493, /* GL_STENCIL_BACK_WRITEMASK */
+       441, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+      1304, /* GL_RENDERBUFFER_BINDING_EXT */
+      1285, /* GL_READ_FRAMEBUFFER */
+       440, /* GL_DRAW_FRAMEBUFFER */
+      1286, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+      1314, /* GL_RENDERBUFFER_SAMPLES */
+       544, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+       542, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+       553, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+       549, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+       551, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+       556, /* GL_FRAMEBUFFER_COMPLETE */
+       560, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+       566, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+       564, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+       562, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+       565, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+       563, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+       569, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+       572, /* GL_FRAMEBUFFER_UNSUPPORTED */
+       570, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+       853, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+       155, /* GL_COLOR_ATTACHMENT0 */
+       157, /* GL_COLOR_ATTACHMENT1 */
+       171, /* GL_COLOR_ATTACHMENT2 */
+       173, /* GL_COLOR_ATTACHMENT3 */
+       175, /* GL_COLOR_ATTACHMENT4 */
+       177, /* GL_COLOR_ATTACHMENT5 */
+       179, /* GL_COLOR_ATTACHMENT6 */
+       181, /* GL_COLOR_ATTACHMENT7 */
+       183, /* GL_COLOR_ATTACHMENT8 */
+       185, /* GL_COLOR_ATTACHMENT9 */
+       158, /* GL_COLOR_ATTACHMENT10 */
+       160, /* GL_COLOR_ATTACHMENT11 */
+       162, /* GL_COLOR_ATTACHMENT12 */
+       164, /* GL_COLOR_ATTACHMENT13 */
+       166, /* GL_COLOR_ATTACHMENT14 */
+       168, /* GL_COLOR_ATTACHMENT15 */
+       348, /* GL_DEPTH_ATTACHMENT */
+      1481, /* GL_STENCIL_ATTACHMENT */
+       535, /* GL_FRAMEBUFFER */
+      1302, /* GL_RENDERBUFFER */
+      1316, /* GL_RENDERBUFFER_WIDTH */
+      1309, /* GL_RENDERBUFFER_HEIGHT */
+      1311, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+      1504, /* GL_STENCIL_INDEX_EXT */
+      1501, /* GL_STENCIL_INDEX1_EXT */
+      1502, /* GL_STENCIL_INDEX4_EXT */
+      1503, /* GL_STENCIL_INDEX8_EXT */
+      1500, /* GL_STENCIL_INDEX16_EXT */
+      1313, /* GL_RENDERBUFFER_RED_SIZE */
+      1308, /* GL_RENDERBUFFER_GREEN_SIZE */
+      1305, /* GL_RENDERBUFFER_BLUE_SIZE */
+      1303, /* GL_RENDERBUFFER_ALPHA_SIZE */
+      1306, /* GL_RENDERBUFFER_DEPTH_SIZE */
+      1315, /* GL_RENDERBUFFER_STENCIL_SIZE */
+       568, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+       909, /* GL_MAX_SAMPLES */
+      1272, /* GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT */
+       482, /* GL_FIRST_VERTEX_CONVENTION_EXT */
+       665, /* GL_LAST_VERTEX_CONVENTION_EXT */
+      1251, /* GL_PROVOKING_VERTEX_EXT */
+       302, /* GL_COPY_READ_BUFFER */
+       303, /* GL_COPY_WRITE_BUFFER */
+      1363, /* GL_RGBA_SNORM */
+      1359, /* GL_RGBA8_SNORM */
+      1421, /* GL_SIGNED_NORMALIZED */
+       910, /* GL_MAX_SERVER_WAIT_TIMEOUT */
+      1039, /* GL_OBJECT_TYPE */
+      1525, /* GL_SYNC_CONDITION */
+      1530, /* GL_SYNC_STATUS */
+      1527, /* GL_SYNC_FLAGS */
+      1526, /* GL_SYNC_FENCE */
+      1529, /* GL_SYNC_GPU_COMMANDS_COMPLETE */
+      1756, /* GL_UNSIGNALED */
+      1420, /* GL_SIGNALED */
+        46, /* GL_ALREADY_SIGNALED */
+      1727, /* GL_TIMEOUT_EXPIRED */
+       270, /* GL_CONDITION_SATISFIED */
+      1839, /* GL_WAIT_FAILED */
+       467, /* GL_EVAL_BIT */
+      1283, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+       707, /* GL_LIST_BIT */
+      1622, /* GL_TEXTURE_BIT */
+      1394, /* GL_SCISSOR_BIT */
         29, /* GL_ALL_ATTRIB_BITS */
-       980, /* GL_MULTISAMPLE_BIT */
+       995, /* GL_MULTISAMPLE_BIT */
         30, /* GL_ALL_CLIENT_ATTRIB_BITS */
+      1728, /* GL_TIMEOUT_IGNORED */
 };
 
 typedef int (*cfunc)(const void *, const void *);
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index ebbe28acf4..e3070b1547 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.3
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -44,6 +45,7 @@ static const struct {
    const char *name;
    int flag_offset;
 } default_extensions[] = {
+   { OFF, "GL_ARB_copy_buffer",                F(ARB_copy_buffer) },
    { OFF, "GL_ARB_depth_texture",              F(ARB_depth_texture) },
    { ON,  "GL_ARB_draw_buffers",               F(ARB_draw_buffers) },
    { OFF, "GL_ARB_fragment_program",           F(ARB_fragment_program) },
@@ -52,17 +54,20 @@ static const struct {
    { OFF, "GL_ARB_framebuffer_object",         F(ARB_framebuffer_object) },
    { OFF, "GL_ARB_half_float_pixel",           F(ARB_half_float_pixel) },
    { OFF, "GL_ARB_imaging",                    F(ARB_imaging) },
+   { OFF, "GL_ARB_map_buffer_range",           F(ARB_map_buffer_range) },
    { ON,  "GL_ARB_multisample",                F(ARB_multisample) },
    { OFF, "GL_ARB_multitexture",               F(ARB_multitexture) },
    { OFF, "GL_ARB_occlusion_query",            F(ARB_occlusion_query) },
    { OFF, "GL_ARB_pixel_buffer_object",        F(EXT_pixel_buffer_object) },
    { OFF, "GL_ARB_point_parameters",           F(EXT_point_parameters) },
    { OFF, "GL_ARB_point_sprite",               F(ARB_point_sprite) },
+   { OFF, "GL_ARB_seamless_cube_map",          F(ARB_seamless_cube_map) },
    { OFF, "GL_ARB_shader_objects",             F(ARB_shader_objects) },
    { OFF, "GL_ARB_shading_language_100",       F(ARB_shading_language_100) },
    { OFF, "GL_ARB_shading_language_120",       F(ARB_shading_language_120) },
    { OFF, "GL_ARB_shadow",                     F(ARB_shadow) },
    { OFF, "GL_ARB_shadow_ambient",             F(ARB_shadow_ambient) },
+   { OFF, "GL_ARB_sync",                       F(ARB_sync) },
    { OFF, "GL_ARB_texture_border_clamp",       F(ARB_texture_border_clamp) },
    { ON,  "GL_ARB_texture_compression",        F(ARB_texture_compression) },
    { OFF, "GL_ARB_texture_cube_map",           F(ARB_texture_cube_map) },
@@ -75,6 +80,8 @@ static const struct {
    { OFF, "GL_ARB_texture_non_power_of_two",   F(ARB_texture_non_power_of_two)},
    { OFF, "GL_ARB_texture_rectangle",          F(NV_texture_rectangle) },
    { ON,  "GL_ARB_transpose_matrix",           F(ARB_transpose_matrix) },
+   { OFF, "GL_ARB_vertex_array_bgra",          F(EXT_vertex_array_bgra) },
+   { OFF, "GL_ARB_vertex_array_object",        F(ARB_vertex_array_object) },
    { ON,  "GL_ARB_vertex_buffer_object",       F(ARB_vertex_buffer_object) },
    { OFF, "GL_ARB_vertex_program",             F(ARB_vertex_program) },
    { OFF, "GL_ARB_vertex_shader",              F(ARB_vertex_shader) },
@@ -106,6 +113,7 @@ static const struct {
    { OFF, "GL_EXT_pixel_buffer_object",        F(EXT_pixel_buffer_object) },
    { OFF, "GL_EXT_point_parameters",           F(EXT_point_parameters) },
    { ON,  "GL_EXT_polygon_offset",             F(EXT_polygon_offset) },
+   { OFF, "GL_EXT_provoking_vertex",           F(EXT_provoking_vertex) },
    { ON,  "GL_EXT_rescale_normal",             F(EXT_rescale_normal) },
    { OFF, "GL_EXT_secondary_color",            F(EXT_secondary_color) },
    { ON,  "GL_EXT_separate_specular_color",    F(EXT_separate_specular_color) },
@@ -183,6 +191,7 @@ static const struct {
 void
 _mesa_enable_sw_extensions(GLcontext *ctx)
 {
+   ctx->Extensions.ARB_copy_buffer = GL_TRUE;
    ctx->Extensions.ARB_depth_texture = GL_TRUE;
    /*ctx->Extensions.ARB_draw_buffers = GL_TRUE;*/
 #if FEATURE_ARB_fragment_program
@@ -197,6 +206,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #endif
    ctx->Extensions.ARB_half_float_pixel = GL_TRUE;
    ctx->Extensions.ARB_imaging = GL_TRUE;
+   ctx->Extensions.ARB_map_buffer_range = GL_TRUE;
    ctx->Extensions.ARB_multitexture = GL_TRUE;
 #if FEATURE_ARB_occlusion_query
    ctx->Extensions.ARB_occlusion_query = GL_TRUE;
@@ -221,6 +231,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    /*ctx->Extensions.ARB_texture_float = GL_TRUE;*/
    ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
    ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE;
+   ctx->Extensions.ARB_vertex_array_object = GL_TRUE;
 #if FEATURE_ARB_vertex_program
    ctx->Extensions.ARB_vertex_program = GL_TRUE;
 #endif
@@ -230,6 +241,9 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #if FEATURE_ARB_vertex_buffer_object
    /*ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;*/
 #endif
+#if FEATURE_ARB_sync
+   ctx->Extensions.ARB_sync = GL_TRUE;
+#endif
    ctx->Extensions.APPLE_vertex_array_object = GL_TRUE;
    ctx->Extensions.ATI_envmap_bumpmap = GL_TRUE;
 #if FEATURE_ATI_fragment_shader
@@ -261,6 +275,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
 #endif
    ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_provoking_vertex = GL_TRUE;
    ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_shared_texture_palette = GL_TRUE;
@@ -441,8 +456,9 @@ _mesa_enable_2_1_extensions(GLcontext *ctx)
 
 /**
  * Either enable or disable the named extension.
+ * \return GL_TRUE for success, GL_FALSE if invalid extension name
  */
-static void
+static GLboolean
 set_extension( GLcontext *ctx, const char *name, GLboolean state )
 {
    GLboolean *base = (GLboolean *) &ctx->Extensions;
@@ -451,7 +467,7 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
    if (ctx->Extensions.String) {
       /* The string was already queried - can't change it now! */
       _mesa_problem(ctx, "Trying to enable/disable extension after glGetString(GL_EXTENSIONS): %s", name);
-      return;
+      return GL_FALSE;
    }
 
    for (i = 0 ; i < Elements(default_extensions) ; i++) {
@@ -460,10 +476,10 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
             GLboolean *enabled = base + default_extensions[i].flag_offset;
             *enabled = state;
          }
-         return;
+         return GL_TRUE;
       }
    }
-   _mesa_problem(ctx, "Trying to enable unknown extension: %s", name);
+   return GL_FALSE;
 }
 
 
@@ -474,7 +490,8 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
 void
 _mesa_enable_extension( GLcontext *ctx, const char *name )
 {
-   set_extension(ctx, name, GL_TRUE);
+   if (!set_extension(ctx, name, GL_TRUE))
+      _mesa_problem(ctx, "Trying to enable unknown extension: %s", name);
 }
 
 
@@ -485,7 +502,8 @@ _mesa_enable_extension( GLcontext *ctx, const char *name )
 void
 _mesa_disable_extension( GLcontext *ctx, const char *name )
 {
-   set_extension(ctx, name, GL_FALSE);
+   if (!set_extension(ctx, name, GL_FALSE))
+      _mesa_problem(ctx, "Trying to disable unknown extension: %s", name);
 }
 
 
@@ -510,6 +528,80 @@ _mesa_extension_is_enabled( GLcontext *ctx, const char *name )
 
 
 /**
+ * Append string 'b' onto string 'a'.  Free 'a' and return new string.
+ */
+static char *
+append(const char *a, const char *b)
+{
+   const GLuint aLen = a ? _mesa_strlen(a) : 0;
+   const GLuint bLen = b ? _mesa_strlen(b) : 0;
+   char *s = _mesa_calloc(aLen + bLen + 1);
+   if (s) {
+      if (a)
+         _mesa_memcpy(s, a, aLen);
+      if (b)
+         _mesa_memcpy(s + aLen, b, bLen);
+      s[aLen + bLen] = '\0';
+   }
+   if (a)
+      _mesa_free((void *) a);
+   return s;
+}
+
+
+/**
+ * Check the MESA_EXTENSION_OVERRIDE env var.
+ * For extension names that are recognized, turn them on.  For extension
+ * names that are recognized and prefixed with '-', turn them off.
+ * Return a string of the unknown/leftover names.
+ */
+static const char *
+get_extension_override( GLcontext *ctx )
+{
+   const char *envExt = _mesa_getenv("MESA_EXTENSION_OVERRIDE");
+   char *extraExt = NULL;
+   char ext[1000];
+   GLuint extLen = 0;
+   GLuint i;
+   GLboolean disableExt = GL_FALSE;
+
+   if (!envExt)
+      return NULL;
+
+   for (i = 0; ; i++) {
+      if (envExt[i] == '\0' || envExt[i] == ' ') {
+         /* terminate/process 'ext' if extLen > 0 */
+         if (extLen > 0) {
+            assert(extLen < sizeof(ext));
+            /* enable extension named by 'ext' */
+            ext[extLen] = 0;
+            if (!set_extension(ctx, ext, !disableExt)) {
+               /* unknown extension name, append it to extraExt */
+               if (extraExt) {
+                  extraExt = append(extraExt, " ");
+               }
+               extraExt = append(extraExt, ext);
+            }
+            extLen = 0;
+            disableExt = GL_FALSE;
+         }
+         if (envExt[i] == '\0')
+            break;
+      }
+      else if (envExt[i] == '-') {
+         disableExt = GL_TRUE;
+      }
+      else {
+         /* accumulate this non-space character */
+         ext[extLen++] = envExt[i];
+      }
+   }
+
+   return extraExt;
+}
+
+
+/**
  * Run through the default_extensions array above and set the
  * ctx->Extensions.ARB/EXT_* flags accordingly.
  * To be called during context initialization.
@@ -537,8 +629,9 @@ GLubyte *
 _mesa_make_extension_string( GLcontext *ctx )
 {
    const GLboolean *base = (const GLboolean *) &ctx->Extensions;
+   const char *extraExt = get_extension_override(ctx);
    GLuint extStrLen = 0;
-   GLubyte *s;
+   char *s;
    GLuint i;
 
    /* first, compute length of the extension string */
@@ -548,7 +641,14 @@ _mesa_make_extension_string( GLcontext *ctx )
          extStrLen += (GLuint)_mesa_strlen(default_extensions[i].name) + 1;
       }
    }
-   s = (GLubyte *) _mesa_malloc(extStrLen);
+
+   if (extraExt)
+      extStrLen += _mesa_strlen(extraExt) + 1; /* +1 for space */
+
+   /* allocate the extension string */
+   s = (char *) _mesa_malloc(extStrLen);
+   if (!s)
+      return NULL;
 
    /* second, build the extension string */
    extStrLen = 0;
@@ -558,13 +658,18 @@ _mesa_make_extension_string( GLcontext *ctx )
          GLuint len = (GLuint)_mesa_strlen(default_extensions[i].name);
          _mesa_memcpy(s + extStrLen, default_extensions[i].name, len);
          extStrLen += len;
-         s[extStrLen] = (GLubyte) ' ';
+         s[extStrLen] = ' ';
          extStrLen++;
       }
    }
    ASSERT(extStrLen > 0);
 
-   s[extStrLen - 1] = 0;
+   s[extStrLen - 1] = 0; /* -1 to overwrite trailing the ' ' */
 
-   return s;
+   if (extraExt) {
+      s = append(s, " ");
+      s = append(s, extraExt);
+   }
+
+   return (GLubyte *) s;
 }
diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
index 8e21a27f89..356476e35a 100644
--- a/src/mesa/main/ffvertex_prog.c
+++ b/src/mesa/main/ffvertex_prog.c
@@ -46,6 +46,9 @@
 #include "shader/prog_statevars.h"
 
 
+/** Max of number of lights and texture coord units */
+#define NUM_UNITS MAX2(MAX_TEXTURE_COORD_UNITS, MAX_LIGHTS)
+
 struct state_key {
    unsigned light_color_material_mask:12;
    unsigned light_global_enabled:1;
@@ -77,7 +80,7 @@ struct state_key {
       unsigned texgen_mode1:4;
       unsigned texgen_mode2:4;
       unsigned texgen_mode3:4;
-   } unit[8];
+   } unit[NUM_UNITS];
 };
 
 
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index 5a13c88a7a..dc79b8ca61 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -817,7 +817,7 @@ _mesa_update_framebuffer(GLcontext *ctx)
 
 /**
  * Check if the renderbuffer for a read operation (glReadPixels, glCopyPixels,
- * glCopyTex[Sub]Image, etc. exists.
+ * glCopyTex[Sub]Image, etc) exists.
  * \param format  a basic image format such as GL_RGB, GL_RGBA, GL_ALPHA,
  *                GL_DEPTH_COMPONENT, etc. or GL_COLOR, GL_DEPTH, GL_STENCIL.
  * \return GL_TRUE if buffer exists, GL_FALSE otherwise
@@ -825,8 +825,12 @@ _mesa_update_framebuffer(GLcontext *ctx)
 GLboolean
 _mesa_source_buffer_exists(GLcontext *ctx, GLenum format)
 {
-   const struct gl_renderbuffer_attachment *att
-      = ctx->ReadBuffer->Attachment;
+   const struct gl_renderbuffer_attachment *att = ctx->ReadBuffer->Attachment;
+
+   /* If we don't know the framebuffer status, update it now */
+   if (ctx->ReadBuffer->_Status == 0) {
+      _mesa_test_framebuffer_completeness(ctx, ctx->ReadBuffer);
+   }
 
    if (ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
       return GL_FALSE;
@@ -850,10 +854,8 @@ _mesa_source_buffer_exists(GLcontext *ctx, GLenum format)
       if (ctx->ReadBuffer->_ColorReadBuffer == NULL) {
          return GL_FALSE;
       }
-      /* XXX enable this post 6.5 release:
       ASSERT(ctx->ReadBuffer->_ColorReadBuffer->RedBits > 0 ||
              ctx->ReadBuffer->_ColorReadBuffer->IndexBits > 0);
-      */
       break;
    case GL_DEPTH:
    case GL_DEPTH_COMPONENT:
@@ -891,13 +893,17 @@ _mesa_source_buffer_exists(GLcontext *ctx, GLenum format)
 
 /**
  * As above, but for drawing operations.
- * XXX code do some code merging w/ above function.
+ * XXX could do some code merging w/ above function.
  */
 GLboolean
 _mesa_dest_buffer_exists(GLcontext *ctx, GLenum format)
 {
-   const struct gl_renderbuffer_attachment *att
-      = ctx->ReadBuffer->Attachment;
+   const struct gl_renderbuffer_attachment *att = ctx->DrawBuffer->Attachment;
+
+   /* If we don't know the framebuffer status, update it now */
+   if (ctx->DrawBuffer->_Status == 0) {
+      _mesa_test_framebuffer_completeness(ctx, ctx->DrawBuffer);
+   }
 
    if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
       return GL_FALSE;
@@ -918,7 +924,7 @@ _mesa_dest_buffer_exists(GLcontext *ctx, GLenum format)
    case GL_BGRA:
    case GL_ABGR_EXT:
    case GL_COLOR_INDEX:
-      /* nothing special */
+      /* Nothing special since GL_DRAW_BUFFER could be GL_NONE. */
       /* Could assert that colorbuffer has RedBits > 0 */
       break;
    case GL_DEPTH:
@@ -945,7 +951,7 @@ _mesa_dest_buffer_exists(GLcontext *ctx, GLenum format)
       break;
    default:
       _mesa_problem(ctx,
-                    "Unexpected format 0x%x in _mesa_source_buffer_exists",
+                    "Unexpected format 0x%x in _mesa_dest_buffer_exists",
                     format);
       return GL_FALSE;
    }
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index 1ed6fc3383..477ed01030 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -19,9 +19,15 @@
 
 #define INT_TO_BOOLEAN(I)     ( (I) ? GL_TRUE : GL_FALSE )
 
+#define INT64_TO_BOOLEAN(I)   ( (I) ? GL_TRUE : GL_FALSE )
+#define INT64_TO_INT(I)       ( (GLint)((I > INT_MAX) ? INT_MAX : ((I < INT_MIN) ? INT_MIN : (I))) )
+
 #define BOOLEAN_TO_INT(B)     ( (GLint) (B) )
+#define BOOLEAN_TO_INT64(B)   ( (GLint64) (B) )
 #define BOOLEAN_TO_FLOAT(B)   ( (B) ? 1.0F : 0.0F )
 
+#define ENUM_TO_INT64(E)      ( (GLint64) (E) )
+
 
 /*
  * Check if named extension is enabled, if not generate error and return.
@@ -276,6 +282,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
       case GL_CURRENT_TEXTURE_COORDS:
          {
          const GLuint texUnit = ctx->Texture.CurrentUnit;
+         FLUSH_CURRENT(ctx, 0);
          params[0] = FLOAT_TO_BOOLEAN(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][0]);
          params[1] = FLOAT_TO_BOOLEAN(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][1]);
          params[2] = FLOAT_TO_BOOLEAN(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][2]);
@@ -1835,6 +1842,14 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->ReadBuffer->Name);
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetBooleanv");
+         params[0] = ctx->Light.ProvokingVertex;
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetBooleanv");
+         params[0] = ctx->Const.QuadsFollowProvokingVertexConvention;
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -1871,6 +1886,14 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          CHECK_EXT1(APPLE_vertex_array_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->Name);
          break;
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+         CHECK_EXT1(ARB_seamless_cube_map, "GetBooleanv");
+         params[0] = ctx->Texture.CubeMapSeamless;
+         break;
+      case GL_MAX_SERVER_WAIT_TIMEOUT:
+         CHECK_EXT1(ARB_sync, "GetBooleanv");
+         params[0] = INT64_TO_BOOLEAN(ctx->Const.MaxServerWaitTimeout);
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetBooleanv(pname=0x%x)", pname);
    }
@@ -2090,6 +2113,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
       case GL_CURRENT_TEXTURE_COORDS:
          {
          const GLuint texUnit = ctx->Texture.CurrentUnit;
+         FLUSH_CURRENT(ctx, 0);
          params[0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][0];
          params[1] = ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][1];
          params[2] = ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][2];
@@ -3649,6 +3673,14 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetFloatv");
          params[0] = (GLfloat)(ctx->ReadBuffer->Name);
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetFloatv");
+         params[0] = BOOLEAN_TO_FLOAT(ctx->Light.ProvokingVertex);
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetFloatv");
+         params[0] = BOOLEAN_TO_FLOAT(ctx->Const.QuadsFollowProvokingVertexConvention);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetFloatv");
          params[0] = (GLfloat)(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -3685,6 +3717,14 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          CHECK_EXT1(APPLE_vertex_array_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->Name);
          break;
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+         CHECK_EXT1(ARB_seamless_cube_map, "GetFloatv");
+         params[0] = BOOLEAN_TO_FLOAT(ctx->Texture.CubeMapSeamless);
+         break;
+      case GL_MAX_SERVER_WAIT_TIMEOUT:
+         CHECK_EXT1(ARB_sync, "GetFloatv");
+         params[0] = (GLfloat)(ctx->Const.MaxServerWaitTimeout);
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetFloatv(pname=0x%x)", pname);
    }
@@ -3904,6 +3944,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
       case GL_CURRENT_TEXTURE_COORDS:
          {
          const GLuint texUnit = ctx->Texture.CurrentUnit;
+         FLUSH_CURRENT(ctx, 0);
          params[0] = IROUND(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][0]);
          params[1] = IROUND(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][1]);
          params[2] = IROUND(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][2]);
@@ -5463,6 +5504,14 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetIntegerv");
          params[0] = ctx->ReadBuffer->Name;
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetIntegerv");
+         params[0] = BOOLEAN_TO_INT(ctx->Light.ProvokingVertex);
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetIntegerv");
+         params[0] = BOOLEAN_TO_INT(ctx->Const.QuadsFollowProvokingVertexConvention);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetIntegerv");
          params[0] = ctx->Const.FragmentProgram.MaxUniformComponents;
@@ -5499,11 +5548,1852 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          CHECK_EXT1(APPLE_vertex_array_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->Name;
          break;
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+         CHECK_EXT1(ARB_seamless_cube_map, "GetIntegerv");
+         params[0] = BOOLEAN_TO_INT(ctx->Texture.CubeMapSeamless);
+         break;
+      case GL_MAX_SERVER_WAIT_TIMEOUT:
+         CHECK_EXT1(ARB_sync, "GetIntegerv");
+         params[0] = INT64_TO_INT(ctx->Const.MaxServerWaitTimeout);
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetIntegerv(pname=0x%x)", pname);
    }
 }
 
+#if FEATURE_ARB_sync
+void GLAPIENTRY
+_mesa_GetInteger64v( GLenum pname, GLint64 *params )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!params)
+      return;
+
+   if (ctx->NewState)
+      _mesa_update_state(ctx);
+
+   if (ctx->Driver.GetInteger64v &&
+       ctx->Driver.GetInteger64v(ctx, pname, params))
+      return;
+
+   switch (pname) {
+      case GL_ACCUM_RED_BITS:
+         params[0] = ctx->DrawBuffer->Visual.accumRedBits;
+         break;
+      case GL_ACCUM_GREEN_BITS:
+         params[0] = ctx->DrawBuffer->Visual.accumGreenBits;
+         break;
+      case GL_ACCUM_BLUE_BITS:
+         params[0] = ctx->DrawBuffer->Visual.accumBlueBits;
+         break;
+      case GL_ACCUM_ALPHA_BITS:
+         params[0] = ctx->DrawBuffer->Visual.accumAlphaBits;
+         break;
+      case GL_ACCUM_CLEAR_VALUE:
+         params[0] = FLOAT_TO_INT64(ctx->Accum.ClearColor[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Accum.ClearColor[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Accum.ClearColor[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Accum.ClearColor[3]);
+         break;
+      case GL_ALPHA_BIAS:
+         params[0] = IROUND64(ctx->Pixel.AlphaBias);
+         break;
+      case GL_ALPHA_BITS:
+         params[0] = ctx->DrawBuffer->Visual.alphaBits;
+         break;
+      case GL_ALPHA_SCALE:
+         params[0] = IROUND64(ctx->Pixel.AlphaScale);
+         break;
+      case GL_ALPHA_TEST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Color.AlphaEnabled);
+         break;
+      case GL_ALPHA_TEST_FUNC:
+         params[0] = ENUM_TO_INT64(ctx->Color.AlphaFunc);
+         break;
+      case GL_ALPHA_TEST_REF:
+         params[0] = FLOAT_TO_INT64(ctx->Color.AlphaRef);
+         break;
+      case GL_ATTRIB_STACK_DEPTH:
+         params[0] = ctx->AttribStackDepth;
+         break;
+      case GL_AUTO_NORMAL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.AutoNormal);
+         break;
+      case GL_AUX_BUFFERS:
+         params[0] = ctx->DrawBuffer->Visual.numAuxBuffers;
+         break;
+      case GL_BLEND:
+         params[0] = BOOLEAN_TO_INT64(ctx->Color.BlendEnabled);
+         break;
+      case GL_BLEND_DST:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendDstRGB);
+         break;
+      case GL_BLEND_SRC:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendSrcRGB);
+         break;
+      case GL_BLEND_SRC_RGB_EXT:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendSrcRGB);
+         break;
+      case GL_BLEND_DST_RGB_EXT:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendDstRGB);
+         break;
+      case GL_BLEND_SRC_ALPHA_EXT:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendSrcA);
+         break;
+      case GL_BLEND_DST_ALPHA_EXT:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendDstA);
+         break;
+      case GL_BLEND_EQUATION:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendEquationRGB );
+         break;
+      case GL_BLEND_EQUATION_ALPHA_EXT:
+         params[0] = ENUM_TO_INT64(ctx->Color.BlendEquationA );
+         break;
+      case GL_BLEND_COLOR_EXT:
+         params[0] = FLOAT_TO_INT64(ctx->Color.BlendColor[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Color.BlendColor[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Color.BlendColor[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Color.BlendColor[3]);
+         break;
+      case GL_BLUE_BIAS:
+         params[0] = IROUND64(ctx->Pixel.BlueBias);
+         break;
+      case GL_BLUE_BITS:
+         params[0] = ctx->DrawBuffer->Visual.blueBits;
+         break;
+      case GL_BLUE_SCALE:
+         params[0] = IROUND64(ctx->Pixel.BlueScale);
+         break;
+      case GL_CLIENT_ATTRIB_STACK_DEPTH:
+         params[0] = ctx->ClientAttribStackDepth;
+         break;
+      case GL_CLIP_PLANE0:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 0) & 1);
+         break;
+      case GL_CLIP_PLANE1:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 1) & 1);
+         break;
+      case GL_CLIP_PLANE2:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 2) & 1);
+         break;
+      case GL_CLIP_PLANE3:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 3) & 1);
+         break;
+      case GL_CLIP_PLANE4:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 4) & 1);
+         break;
+      case GL_CLIP_PLANE5:
+         params[0] = BOOLEAN_TO_INT64((ctx->Transform.ClipPlanesEnabled >> 5) & 1);
+         break;
+      case GL_COLOR_CLEAR_VALUE:
+         params[0] = FLOAT_TO_INT64(ctx->Color.ClearColor[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Color.ClearColor[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Color.ClearColor[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Color.ClearColor[3]);
+         break;
+      case GL_COLOR_MATERIAL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.ColorMaterialEnabled);
+         break;
+      case GL_COLOR_MATERIAL_FACE:
+         params[0] = ENUM_TO_INT64(ctx->Light.ColorMaterialFace);
+         break;
+      case GL_COLOR_MATERIAL_PARAMETER:
+         params[0] = ENUM_TO_INT64(ctx->Light.ColorMaterialMode);
+         break;
+      case GL_COLOR_WRITEMASK:
+         params[0] = ctx->Color.ColorMask[RCOMP] ? 1 : 0;
+         params[1] = ctx->Color.ColorMask[GCOMP] ? 1 : 0;
+         params[2] = ctx->Color.ColorMask[BCOMP] ? 1 : 0;
+         params[3] = ctx->Color.ColorMask[ACOMP] ? 1 : 0;
+         break;
+      case GL_CULL_FACE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.CullFlag);
+         break;
+      case GL_CULL_FACE_MODE:
+         params[0] = ENUM_TO_INT64(ctx->Polygon.CullFaceMode);
+         break;
+      case GL_CURRENT_COLOR:
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0]);
+         params[1] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1]);
+         params[2] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2]);
+         params[3] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3]);
+         }
+         break;
+      case GL_CURRENT_INDEX:
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0]);
+         }
+         break;
+      case GL_CURRENT_NORMAL:
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0]);
+         params[1] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1]);
+         params[2] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2]);
+         }
+         break;
+      case GL_CURRENT_RASTER_COLOR:
+         params[0] = FLOAT_TO_INT64(ctx->Current.RasterColor[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Current.RasterColor[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Current.RasterColor[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Current.RasterColor[3]);
+         break;
+      case GL_CURRENT_RASTER_DISTANCE:
+         params[0] = IROUND64(ctx->Current.RasterDistance);
+         break;
+      case GL_CURRENT_RASTER_INDEX:
+         params[0] = IROUND64(ctx->Current.RasterIndex);
+         break;
+      case GL_CURRENT_RASTER_POSITION:
+         params[0] = IROUND64(ctx->Current.RasterPos[0]);
+         params[1] = IROUND64(ctx->Current.RasterPos[1]);
+         params[2] = IROUND64(ctx->Current.RasterPos[2]);
+         params[3] = IROUND64(ctx->Current.RasterPos[3]);
+         break;
+      case GL_CURRENT_RASTER_SECONDARY_COLOR:
+         params[0] = FLOAT_TO_INT64(ctx->Current.RasterSecondaryColor[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Current.RasterSecondaryColor[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Current.RasterSecondaryColor[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Current.RasterSecondaryColor[3]);
+         break;
+      case GL_CURRENT_RASTER_TEXTURE_COORDS:
+         {
+         const GLuint texUnit = ctx->Texture.CurrentUnit;
+         params[0] = IROUND64(ctx->Current.RasterTexCoords[texUnit][0]);
+         params[1] = IROUND64(ctx->Current.RasterTexCoords[texUnit][1]);
+         params[2] = IROUND64(ctx->Current.RasterTexCoords[texUnit][2]);
+         params[3] = IROUND64(ctx->Current.RasterTexCoords[texUnit][3]);
+         }
+         break;
+      case GL_CURRENT_RASTER_POSITION_VALID:
+         params[0] = BOOLEAN_TO_INT64(ctx->Current.RasterPosValid);
+         break;
+      case GL_CURRENT_TEXTURE_COORDS:
+         {
+         const GLuint texUnit = ctx->Texture.CurrentUnit;
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][0]);
+         params[1] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][1]);
+         params[2] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][2]);
+         params[3] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][3]);
+         }
+         break;
+      case GL_DEPTH_BIAS:
+         params[0] = IROUND64(ctx->Pixel.DepthBias);
+         break;
+      case GL_DEPTH_BITS:
+         params[0] = ctx->DrawBuffer->Visual.depthBits;
+         break;
+      case GL_DEPTH_CLEAR_VALUE:
+         params[0] = FLOAT_TO_INT64(((GLfloat) ctx->Depth.Clear));
+         break;
+      case GL_DEPTH_FUNC:
+         params[0] = ENUM_TO_INT64(ctx->Depth.Func);
+         break;
+      case GL_DEPTH_RANGE:
+         params[0] = FLOAT_TO_INT64(ctx->Viewport.Near);
+         params[1] = FLOAT_TO_INT64(ctx->Viewport.Far);
+         break;
+      case GL_DEPTH_SCALE:
+         params[0] = IROUND64(ctx->Pixel.DepthScale);
+         break;
+      case GL_DEPTH_TEST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Depth.Test);
+         break;
+      case GL_DEPTH_WRITEMASK:
+         params[0] = BOOLEAN_TO_INT64(ctx->Depth.Mask);
+         break;
+      case GL_DITHER:
+         params[0] = BOOLEAN_TO_INT64(ctx->Color.DitherFlag);
+         break;
+      case GL_DOUBLEBUFFER:
+         params[0] = BOOLEAN_TO_INT64(ctx->DrawBuffer->Visual.doubleBufferMode);
+         break;
+      case GL_DRAW_BUFFER:
+         params[0] = ENUM_TO_INT64(ctx->DrawBuffer->ColorDrawBuffer[0]);
+         break;
+      case GL_EDGE_FLAG:
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = BOOLEAN_TO_INT64((ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0));
+         }
+         break;
+      case GL_FEEDBACK_BUFFER_SIZE:
+         params[0] = ctx->Feedback.BufferSize;
+         break;
+      case GL_FEEDBACK_BUFFER_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Feedback.Type);
+         break;
+      case GL_FOG:
+         params[0] = BOOLEAN_TO_INT64(ctx->Fog.Enabled);
+         break;
+      case GL_FOG_COLOR:
+         params[0] = FLOAT_TO_INT64(ctx->Fog.Color[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Fog.Color[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Fog.Color[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Fog.Color[3]);
+         break;
+      case GL_FOG_DENSITY:
+         params[0] = IROUND64(ctx->Fog.Density);
+         break;
+      case GL_FOG_END:
+         params[0] = IROUND64(ctx->Fog.End);
+         break;
+      case GL_FOG_HINT:
+         params[0] = ENUM_TO_INT64(ctx->Hint.Fog);
+         break;
+      case GL_FOG_INDEX:
+         params[0] = IROUND64(ctx->Fog.Index);
+         break;
+      case GL_FOG_MODE:
+         params[0] = ENUM_TO_INT64(ctx->Fog.Mode);
+         break;
+      case GL_FOG_START:
+         params[0] = IROUND64(ctx->Fog.Start);
+         break;
+      case GL_FRONT_FACE:
+         params[0] = ENUM_TO_INT64(ctx->Polygon.FrontFace);
+         break;
+      case GL_GREEN_BIAS:
+         params[0] = IROUND64(ctx->Pixel.GreenBias);
+         break;
+      case GL_GREEN_BITS:
+         params[0] = ctx->DrawBuffer->Visual.greenBits;
+         break;
+      case GL_GREEN_SCALE:
+         params[0] = IROUND64(ctx->Pixel.GreenScale);
+         break;
+      case GL_INDEX_BITS:
+         params[0] = ctx->DrawBuffer->Visual.indexBits;
+         break;
+      case GL_INDEX_CLEAR_VALUE:
+         params[0] = ctx->Color.ClearIndex;
+         break;
+      case GL_INDEX_MODE:
+         params[0] = BOOLEAN_TO_INT64(!ctx->DrawBuffer->Visual.rgbMode);
+         break;
+      case GL_INDEX_OFFSET:
+         params[0] = ctx->Pixel.IndexOffset;
+         break;
+      case GL_INDEX_SHIFT:
+         params[0] = ctx->Pixel.IndexShift;
+         break;
+      case GL_INDEX_WRITEMASK:
+         params[0] = ctx->Color.IndexMask;
+         break;
+      case GL_LIGHT0:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[0].Enabled);
+         break;
+      case GL_LIGHT1:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[1].Enabled);
+         break;
+      case GL_LIGHT2:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[2].Enabled);
+         break;
+      case GL_LIGHT3:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[3].Enabled);
+         break;
+      case GL_LIGHT4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[4].Enabled);
+         break;
+      case GL_LIGHT5:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[5].Enabled);
+         break;
+      case GL_LIGHT6:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[6].Enabled);
+         break;
+      case GL_LIGHT7:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Light[7].Enabled);
+         break;
+      case GL_LIGHTING:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Enabled);
+         break;
+      case GL_LIGHT_MODEL_AMBIENT:
+         params[0] = FLOAT_TO_INT64(ctx->Light.Model.Ambient[0]);
+         params[1] = FLOAT_TO_INT64(ctx->Light.Model.Ambient[1]);
+         params[2] = FLOAT_TO_INT64(ctx->Light.Model.Ambient[2]);
+         params[3] = FLOAT_TO_INT64(ctx->Light.Model.Ambient[3]);
+         break;
+      case GL_LIGHT_MODEL_COLOR_CONTROL:
+         params[0] = ENUM_TO_INT64(ctx->Light.Model.ColorControl);
+         break;
+      case GL_LIGHT_MODEL_LOCAL_VIEWER:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Model.LocalViewer);
+         break;
+      case GL_LIGHT_MODEL_TWO_SIDE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.Model.TwoSide);
+         break;
+      case GL_LINE_SMOOTH:
+         params[0] = BOOLEAN_TO_INT64(ctx->Line.SmoothFlag);
+         break;
+      case GL_LINE_SMOOTH_HINT:
+         params[0] = ENUM_TO_INT64(ctx->Hint.LineSmooth);
+         break;
+      case GL_LINE_STIPPLE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Line.StippleFlag);
+         break;
+      case GL_LINE_STIPPLE_PATTERN:
+         params[0] = ctx->Line.StipplePattern;
+         break;
+      case GL_LINE_STIPPLE_REPEAT:
+         params[0] = ctx->Line.StippleFactor;
+         break;
+      case GL_LINE_WIDTH:
+         params[0] = IROUND64(ctx->Line.Width);
+         break;
+      case GL_LINE_WIDTH_GRANULARITY:
+         params[0] = IROUND64(ctx->Const.LineWidthGranularity);
+         break;
+      case GL_LINE_WIDTH_RANGE:
+         params[0] = IROUND64(ctx->Const.MinLineWidthAA);
+         params[1] = IROUND64(ctx->Const.MaxLineWidthAA);
+         break;
+      case GL_ALIASED_LINE_WIDTH_RANGE:
+         params[0] = IROUND64(ctx->Const.MinLineWidth);
+         params[1] = IROUND64(ctx->Const.MaxLineWidth);
+         break;
+      case GL_LIST_BASE:
+         params[0] = ctx->List.ListBase;
+         break;
+      case GL_LIST_INDEX:
+         params[0] = (ctx->ListState.CurrentList ? ctx->ListState.CurrentList->Name : 0);
+         break;
+      case GL_LIST_MODE:
+         {
+         GLenum mode;
+         if (!ctx->CompileFlag)
+            mode = 0;
+         else if (ctx->ExecuteFlag)
+            mode = GL_COMPILE_AND_EXECUTE;
+         else
+            mode = GL_COMPILE;
+         params[0] = ENUM_TO_INT64(mode);
+         }
+         break;
+      case GL_INDEX_LOGIC_OP:
+         params[0] = BOOLEAN_TO_INT64(ctx->Color.IndexLogicOpEnabled);
+         break;
+      case GL_COLOR_LOGIC_OP:
+         params[0] = BOOLEAN_TO_INT64(ctx->Color.ColorLogicOpEnabled);
+         break;
+      case GL_LOGIC_OP_MODE:
+         params[0] = ENUM_TO_INT64(ctx->Color.LogicOp);
+         break;
+      case GL_MAP1_COLOR_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Color4);
+         break;
+      case GL_MAP1_GRID_DOMAIN:
+         params[0] = IROUND64(ctx->Eval.MapGrid1u1);
+         params[1] = IROUND64(ctx->Eval.MapGrid1u2);
+         break;
+      case GL_MAP1_GRID_SEGMENTS:
+         params[0] = ctx->Eval.MapGrid1un;
+         break;
+      case GL_MAP1_INDEX:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Index);
+         break;
+      case GL_MAP1_NORMAL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Normal);
+         break;
+      case GL_MAP1_TEXTURE_COORD_1:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1TextureCoord1);
+         break;
+      case GL_MAP1_TEXTURE_COORD_2:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1TextureCoord2);
+         break;
+      case GL_MAP1_TEXTURE_COORD_3:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1TextureCoord3);
+         break;
+      case GL_MAP1_TEXTURE_COORD_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1TextureCoord4);
+         break;
+      case GL_MAP1_VERTEX_3:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Vertex3);
+         break;
+      case GL_MAP1_VERTEX_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Vertex4);
+         break;
+      case GL_MAP2_COLOR_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2Color4);
+         break;
+      case GL_MAP2_GRID_DOMAIN:
+         params[0] = IROUND64(ctx->Eval.MapGrid2u1);
+         params[1] = IROUND64(ctx->Eval.MapGrid2u2);
+         params[2] = IROUND64(ctx->Eval.MapGrid2v1);
+         params[3] = IROUND64(ctx->Eval.MapGrid2v2);
+         break;
+      case GL_MAP2_GRID_SEGMENTS:
+         params[0] = ctx->Eval.MapGrid2un;
+         params[1] = ctx->Eval.MapGrid2vn;
+         break;
+      case GL_MAP2_INDEX:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2Index);
+         break;
+      case GL_MAP2_NORMAL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2Normal);
+         break;
+      case GL_MAP2_TEXTURE_COORD_1:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2TextureCoord1);
+         break;
+      case GL_MAP2_TEXTURE_COORD_2:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2TextureCoord2);
+         break;
+      case GL_MAP2_TEXTURE_COORD_3:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2TextureCoord3);
+         break;
+      case GL_MAP2_TEXTURE_COORD_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2TextureCoord4);
+         break;
+      case GL_MAP2_VERTEX_3:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2Vertex3);
+         break;
+      case GL_MAP2_VERTEX_4:
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map2Vertex4);
+         break;
+      case GL_MAP_COLOR:
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.MapColorFlag);
+         break;
+      case GL_MAP_STENCIL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.MapStencilFlag);
+         break;
+      case GL_MATRIX_MODE:
+         params[0] = ENUM_TO_INT64(ctx->Transform.MatrixMode);
+         break;
+      case GL_MAX_ATTRIB_STACK_DEPTH:
+         params[0] = MAX_ATTRIB_STACK_DEPTH;
+         break;
+      case GL_MAX_CLIENT_ATTRIB_STACK_DEPTH:
+         params[0] = MAX_CLIENT_ATTRIB_STACK_DEPTH;
+         break;
+      case GL_MAX_CLIP_PLANES:
+         params[0] = ctx->Const.MaxClipPlanes;
+         break;
+      case GL_MAX_ELEMENTS_VERTICES:
+         params[0] = ctx->Const.MaxArrayLockSize;
+         break;
+      case GL_MAX_ELEMENTS_INDICES:
+         params[0] = ctx->Const.MaxArrayLockSize;
+         break;
+      case GL_MAX_EVAL_ORDER:
+         params[0] = MAX_EVAL_ORDER;
+         break;
+      case GL_MAX_LIGHTS:
+         params[0] = ctx->Const.MaxLights;
+         break;
+      case GL_MAX_LIST_NESTING:
+         params[0] = MAX_LIST_NESTING;
+         break;
+      case GL_MAX_MODELVIEW_STACK_DEPTH:
+         params[0] = MAX_MODELVIEW_STACK_DEPTH;
+         break;
+      case GL_MAX_NAME_STACK_DEPTH:
+         params[0] = MAX_NAME_STACK_DEPTH;
+         break;
+      case GL_MAX_PIXEL_MAP_TABLE:
+         params[0] = MAX_PIXEL_MAP_TABLE;
+         break;
+      case GL_MAX_PROJECTION_STACK_DEPTH:
+         params[0] = MAX_PROJECTION_STACK_DEPTH;
+         break;
+      case GL_MAX_TEXTURE_SIZE:
+         params[0] = 1 << (ctx->Const.MaxTextureLevels - 1);
+         break;
+      case GL_MAX_3D_TEXTURE_SIZE:
+         params[0] = 1 << (ctx->Const.Max3DTextureLevels - 1);
+         break;
+      case GL_MAX_TEXTURE_STACK_DEPTH:
+         params[0] = MAX_TEXTURE_STACK_DEPTH;
+         break;
+      case GL_MAX_VIEWPORT_DIMS:
+         params[0] = ctx->Const.MaxViewportWidth;
+         params[1] = ctx->Const.MaxViewportHeight;
+         break;
+      case GL_MODELVIEW_MATRIX:
+         {
+         const GLfloat *matrix = ctx->ModelviewMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[1]);
+         params[2] = IROUND64(matrix[2]);
+         params[3] = IROUND64(matrix[3]);
+         params[4] = IROUND64(matrix[4]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[6]);
+         params[7] = IROUND64(matrix[7]);
+         params[8] = IROUND64(matrix[8]);
+         params[9] = IROUND64(matrix[9]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[11]);
+         params[12] = IROUND64(matrix[12]);
+         params[13] = IROUND64(matrix[13]);
+         params[14] = IROUND64(matrix[14]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_MODELVIEW_STACK_DEPTH:
+         params[0] = ctx->ModelviewMatrixStack.Depth + 1;
+         break;
+      case GL_NAME_STACK_DEPTH:
+         params[0] = ctx->Select.NameStackDepth;
+         break;
+      case GL_NORMALIZE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Transform.Normalize);
+         break;
+      case GL_PACK_ALIGNMENT:
+         params[0] = ctx->Pack.Alignment;
+         break;
+      case GL_PACK_LSB_FIRST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Pack.LsbFirst);
+         break;
+      case GL_PACK_ROW_LENGTH:
+         params[0] = ctx->Pack.RowLength;
+         break;
+      case GL_PACK_SKIP_PIXELS:
+         params[0] = ctx->Pack.SkipPixels;
+         break;
+      case GL_PACK_SKIP_ROWS:
+         params[0] = ctx->Pack.SkipRows;
+         break;
+      case GL_PACK_SWAP_BYTES:
+         params[0] = BOOLEAN_TO_INT64(ctx->Pack.SwapBytes);
+         break;
+      case GL_PACK_SKIP_IMAGES_EXT:
+         params[0] = ctx->Pack.SkipImages;
+         break;
+      case GL_PACK_IMAGE_HEIGHT_EXT:
+         params[0] = ctx->Pack.ImageHeight;
+         break;
+      case GL_PACK_INVERT_MESA:
+         params[0] = BOOLEAN_TO_INT64(ctx->Pack.Invert);
+         break;
+      case GL_PERSPECTIVE_CORRECTION_HINT:
+         params[0] = ENUM_TO_INT64(ctx->Hint.PerspectiveCorrection);
+         break;
+      case GL_PIXEL_MAP_A_TO_A_SIZE:
+         params[0] = ctx->PixelMaps.AtoA.Size;
+         break;
+      case GL_PIXEL_MAP_B_TO_B_SIZE:
+         params[0] = ctx->PixelMaps.BtoB.Size;
+         break;
+      case GL_PIXEL_MAP_G_TO_G_SIZE:
+         params[0] = ctx->PixelMaps.GtoG.Size;
+         break;
+      case GL_PIXEL_MAP_I_TO_A_SIZE:
+         params[0] = ctx->PixelMaps.ItoA.Size;
+         break;
+      case GL_PIXEL_MAP_I_TO_B_SIZE:
+         params[0] = ctx->PixelMaps.ItoB.Size;
+         break;
+      case GL_PIXEL_MAP_I_TO_G_SIZE:
+         params[0] = ctx->PixelMaps.ItoG.Size;
+         break;
+      case GL_PIXEL_MAP_I_TO_I_SIZE:
+         params[0] = ctx->PixelMaps.ItoI.Size;
+         break;
+      case GL_PIXEL_MAP_I_TO_R_SIZE:
+         params[0] = ctx->PixelMaps.ItoR.Size;
+         break;
+      case GL_PIXEL_MAP_R_TO_R_SIZE:
+         params[0] = ctx->PixelMaps.RtoR.Size;
+         break;
+      case GL_PIXEL_MAP_S_TO_S_SIZE:
+         params[0] = ctx->PixelMaps.StoS.Size;
+         break;
+      case GL_POINT_SIZE:
+         params[0] = IROUND64(ctx->Point.Size);
+         break;
+      case GL_POINT_SIZE_GRANULARITY:
+         params[0] = IROUND64(ctx->Const.PointSizeGranularity);
+         break;
+      case GL_POINT_SIZE_RANGE:
+         params[0] = IROUND64(ctx->Const.MinPointSizeAA);
+         params[1] = IROUND64(ctx->Const.MaxPointSizeAA);
+         break;
+      case GL_ALIASED_POINT_SIZE_RANGE:
+         params[0] = IROUND64(ctx->Const.MinPointSize);
+         params[1] = IROUND64(ctx->Const.MaxPointSize);
+         break;
+      case GL_POINT_SMOOTH:
+         params[0] = BOOLEAN_TO_INT64(ctx->Point.SmoothFlag);
+         break;
+      case GL_POINT_SMOOTH_HINT:
+         params[0] = ENUM_TO_INT64(ctx->Hint.PointSmooth);
+         break;
+      case GL_POINT_SIZE_MIN_EXT:
+         params[0] = IROUND64(ctx->Point.MinSize);
+         break;
+      case GL_POINT_SIZE_MAX_EXT:
+         params[0] = IROUND64(ctx->Point.MaxSize);
+         break;
+      case GL_POINT_FADE_THRESHOLD_SIZE_EXT:
+         params[0] = IROUND64(ctx->Point.Threshold);
+         break;
+      case GL_DISTANCE_ATTENUATION_EXT:
+         params[0] = IROUND64(ctx->Point.Params[0]);
+         params[1] = IROUND64(ctx->Point.Params[1]);
+         params[2] = IROUND64(ctx->Point.Params[2]);
+         break;
+      case GL_POLYGON_MODE:
+         params[0] = ENUM_TO_INT64(ctx->Polygon.FrontMode);
+         params[1] = ENUM_TO_INT64(ctx->Polygon.BackMode);
+         break;
+      case GL_POLYGON_OFFSET_BIAS_EXT:
+         params[0] = IROUND64(ctx->Polygon.OffsetUnits);
+         break;
+      case GL_POLYGON_OFFSET_FACTOR:
+         params[0] = IROUND64(ctx->Polygon.OffsetFactor );
+         break;
+      case GL_POLYGON_OFFSET_UNITS:
+         params[0] = IROUND64(ctx->Polygon.OffsetUnits );
+         break;
+      case GL_POLYGON_OFFSET_POINT:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.OffsetPoint);
+         break;
+      case GL_POLYGON_OFFSET_LINE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.OffsetLine);
+         break;
+      case GL_POLYGON_OFFSET_FILL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.OffsetFill);
+         break;
+      case GL_POLYGON_SMOOTH:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.SmoothFlag);
+         break;
+      case GL_POLYGON_SMOOTH_HINT:
+         params[0] = ENUM_TO_INT64(ctx->Hint.PolygonSmooth);
+         break;
+      case GL_POLYGON_STIPPLE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Polygon.StippleFlag);
+         break;
+      case GL_PROJECTION_MATRIX:
+         {
+         const GLfloat *matrix = ctx->ProjectionMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[1]);
+         params[2] = IROUND64(matrix[2]);
+         params[3] = IROUND64(matrix[3]);
+         params[4] = IROUND64(matrix[4]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[6]);
+         params[7] = IROUND64(matrix[7]);
+         params[8] = IROUND64(matrix[8]);
+         params[9] = IROUND64(matrix[9]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[11]);
+         params[12] = IROUND64(matrix[12]);
+         params[13] = IROUND64(matrix[13]);
+         params[14] = IROUND64(matrix[14]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_PROJECTION_STACK_DEPTH:
+         params[0] = ctx->ProjectionMatrixStack.Depth + 1;
+         break;
+      case GL_READ_BUFFER:
+         params[0] = ENUM_TO_INT64(ctx->ReadBuffer->ColorReadBuffer);
+         break;
+      case GL_RED_BIAS:
+         params[0] = IROUND64(ctx->Pixel.RedBias);
+         break;
+      case GL_RED_BITS:
+         params[0] = ctx->DrawBuffer->Visual.redBits;
+         break;
+      case GL_RED_SCALE:
+         params[0] = IROUND64(ctx->Pixel.RedScale);
+         break;
+      case GL_RENDER_MODE:
+         params[0] = ENUM_TO_INT64(ctx->RenderMode);
+         break;
+      case GL_RESCALE_NORMAL:
+         params[0] = BOOLEAN_TO_INT64(ctx->Transform.RescaleNormals);
+         break;
+      case GL_RGBA_MODE:
+         params[0] = BOOLEAN_TO_INT64(ctx->DrawBuffer->Visual.rgbMode);
+         break;
+      case GL_SCISSOR_BOX:
+         params[0] = ctx->Scissor.X;
+         params[1] = ctx->Scissor.Y;
+         params[2] = ctx->Scissor.Width;
+         params[3] = ctx->Scissor.Height;
+         break;
+      case GL_SCISSOR_TEST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Scissor.Enabled);
+         break;
+      case GL_SELECTION_BUFFER_SIZE:
+         params[0] = ctx->Select.BufferSize;
+         break;
+      case GL_SHADE_MODEL:
+         params[0] = ENUM_TO_INT64(ctx->Light.ShadeModel);
+         break;
+      case GL_SHARED_TEXTURE_PALETTE_EXT:
+         params[0] = BOOLEAN_TO_INT64(ctx->Texture.SharedPalette);
+         break;
+      case GL_STENCIL_BITS:
+         params[0] = ctx->DrawBuffer->Visual.stencilBits;
+         break;
+      case GL_STENCIL_CLEAR_VALUE:
+         params[0] = ctx->Stencil.Clear;
+         break;
+      case GL_STENCIL_FAIL:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.FailFunc[ctx->Stencil.ActiveFace]);
+         break;
+      case GL_STENCIL_FUNC:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.Function[ctx->Stencil.ActiveFace]);
+         break;
+      case GL_STENCIL_PASS_DEPTH_FAIL:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.ZFailFunc[ctx->Stencil.ActiveFace]);
+         break;
+      case GL_STENCIL_PASS_DEPTH_PASS:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.ZPassFunc[ctx->Stencil.ActiveFace]);
+         break;
+      case GL_STENCIL_REF:
+         params[0] = ctx->Stencil.Ref[ctx->Stencil.ActiveFace];
+         break;
+      case GL_STENCIL_TEST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Stencil.Enabled);
+         break;
+      case GL_STENCIL_VALUE_MASK:
+         params[0] = ctx->Stencil.ValueMask[ctx->Stencil.ActiveFace];
+         break;
+      case GL_STENCIL_WRITEMASK:
+         params[0] = ctx->Stencil.WriteMask[ctx->Stencil.ActiveFace];
+         break;
+      case GL_STEREO:
+         params[0] = BOOLEAN_TO_INT64(ctx->DrawBuffer->Visual.stereoMode);
+         break;
+      case GL_SUBPIXEL_BITS:
+         params[0] = ctx->Const.SubPixelBits;
+         break;
+      case GL_TEXTURE_1D:
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_1D));
+         break;
+      case GL_TEXTURE_2D:
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_2D));
+         break;
+      case GL_TEXTURE_3D:
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_3D));
+         break;
+      case GL_TEXTURE_1D_ARRAY_EXT:
+         CHECK_EXT1(MESA_texture_array, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_1D_ARRAY_EXT));
+         break;
+      case GL_TEXTURE_2D_ARRAY_EXT:
+         CHECK_EXT1(MESA_texture_array, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_2D_ARRAY_EXT));
+         break;
+      case GL_TEXTURE_BINDING_1D:
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_1D_INDEX]->Name;
+         break;
+      case GL_TEXTURE_BINDING_2D:
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_2D_INDEX]->Name;
+         break;
+      case GL_TEXTURE_BINDING_3D:
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_3D_INDEX]->Name;
+         break;
+      case GL_TEXTURE_BINDING_1D_ARRAY_EXT:
+         CHECK_EXT1(MESA_texture_array, "GetInteger64v");
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_1D_ARRAY_INDEX]->Name;
+         break;
+      case GL_TEXTURE_BINDING_2D_ARRAY_EXT:
+         CHECK_EXT1(MESA_texture_array, "GetInteger64v");
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_2D_ARRAY_INDEX]->Name;
+         break;
+      case GL_TEXTURE_GEN_S:
+         params[0] = BOOLEAN_TO_INT64(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & S_BIT) ? 1 : 0));
+         break;
+      case GL_TEXTURE_GEN_T:
+         params[0] = BOOLEAN_TO_INT64(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & T_BIT) ? 1 : 0));
+         break;
+      case GL_TEXTURE_GEN_R:
+         params[0] = BOOLEAN_TO_INT64(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & R_BIT) ? 1 : 0));
+         break;
+      case GL_TEXTURE_GEN_Q:
+         params[0] = BOOLEAN_TO_INT64(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & Q_BIT) ? 1 : 0));
+         break;
+      case GL_TEXTURE_MATRIX:
+         {
+         const GLfloat *matrix = ctx->TextureMatrixStack[ctx->Texture.CurrentUnit].Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[1]);
+         params[2] = IROUND64(matrix[2]);
+         params[3] = IROUND64(matrix[3]);
+         params[4] = IROUND64(matrix[4]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[6]);
+         params[7] = IROUND64(matrix[7]);
+         params[8] = IROUND64(matrix[8]);
+         params[9] = IROUND64(matrix[9]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[11]);
+         params[12] = IROUND64(matrix[12]);
+         params[13] = IROUND64(matrix[13]);
+         params[14] = IROUND64(matrix[14]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_TEXTURE_STACK_DEPTH:
+         params[0] = ctx->TextureMatrixStack[ctx->Texture.CurrentUnit].Depth + 1;
+         break;
+      case GL_UNPACK_ALIGNMENT:
+         params[0] = ctx->Unpack.Alignment;
+         break;
+      case GL_UNPACK_LSB_FIRST:
+         params[0] = BOOLEAN_TO_INT64(ctx->Unpack.LsbFirst);
+         break;
+      case GL_UNPACK_ROW_LENGTH:
+         params[0] = ctx->Unpack.RowLength;
+         break;
+      case GL_UNPACK_SKIP_PIXELS:
+         params[0] = ctx->Unpack.SkipPixels;
+         break;
+      case GL_UNPACK_SKIP_ROWS:
+         params[0] = ctx->Unpack.SkipRows;
+         break;
+      case GL_UNPACK_SWAP_BYTES:
+         params[0] = BOOLEAN_TO_INT64(ctx->Unpack.SwapBytes);
+         break;
+      case GL_UNPACK_SKIP_IMAGES_EXT:
+         params[0] = ctx->Unpack.SkipImages;
+         break;
+      case GL_UNPACK_IMAGE_HEIGHT_EXT:
+         params[0] = ctx->Unpack.ImageHeight;
+         break;
+      case GL_UNPACK_CLIENT_STORAGE_APPLE:
+         params[0] = BOOLEAN_TO_INT64(ctx->Unpack.ClientStorage);
+         break;
+      case GL_VIEWPORT:
+         params[0] = ctx->Viewport.X;
+         params[1] = ctx->Viewport.Y;
+         params[2] = ctx->Viewport.Width;
+         params[3] = ctx->Viewport.Height;
+         break;
+      case GL_ZOOM_X:
+         params[0] = IROUND64(ctx->Pixel.ZoomX);
+         break;
+      case GL_ZOOM_Y:
+         params[0] = IROUND64(ctx->Pixel.ZoomY);
+         break;
+      case GL_VERTEX_ARRAY:
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->Vertex.Enabled);
+         break;
+      case GL_VERTEX_ARRAY_SIZE:
+         params[0] = ctx->Array.ArrayObj->Vertex.Size;
+         break;
+      case GL_VERTEX_ARRAY_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->Vertex.Type);
+         break;
+      case GL_VERTEX_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->Vertex.Stride;
+         break;
+      case GL_VERTEX_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_NORMAL_ARRAY:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->Normal.Enabled);
+         break;
+      case GL_NORMAL_ARRAY_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->Normal.Type);
+         break;
+      case GL_NORMAL_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->Normal.Stride;
+         break;
+      case GL_NORMAL_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_COLOR_ARRAY:
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->Color.Enabled);
+         break;
+      case GL_COLOR_ARRAY_SIZE:
+         params[0] = ctx->Array.ArrayObj->Color.Size;
+         break;
+      case GL_COLOR_ARRAY_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->Color.Type);
+         break;
+      case GL_COLOR_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->Color.Stride;
+         break;
+      case GL_COLOR_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_INDEX_ARRAY:
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->Index.Enabled);
+         break;
+      case GL_INDEX_ARRAY_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->Index.Type);
+         break;
+      case GL_INDEX_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->Index.Stride;
+         break;
+      case GL_INDEX_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_TEXTURE_COORD_ARRAY:
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].Enabled);
+         break;
+      case GL_TEXTURE_COORD_ARRAY_SIZE:
+         params[0] = ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].Size;
+         break;
+      case GL_TEXTURE_COORD_ARRAY_TYPE:
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].Type);
+         break;
+      case GL_TEXTURE_COORD_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].Stride;
+         break;
+      case GL_TEXTURE_COORD_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_EDGE_FLAG_ARRAY:
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->EdgeFlag.Enabled);
+         break;
+      case GL_EDGE_FLAG_ARRAY_STRIDE:
+         params[0] = ctx->Array.ArrayObj->EdgeFlag.Stride;
+         break;
+      case GL_EDGE_FLAG_ARRAY_COUNT_EXT:
+         params[0] = 0;
+         break;
+      case GL_MAX_TEXTURE_UNITS_ARB:
+         CHECK_EXT1(ARB_multitexture, "GetInteger64v");
+         params[0] = ctx->Const.MaxTextureUnits;
+         break;
+      case GL_ACTIVE_TEXTURE_ARB:
+         CHECK_EXT1(ARB_multitexture, "GetInteger64v");
+         params[0] = GL_TEXTURE0_ARB + ctx->Texture.CurrentUnit;
+         break;
+      case GL_CLIENT_ACTIVE_TEXTURE_ARB:
+         CHECK_EXT1(ARB_multitexture, "GetInteger64v");
+         params[0] = GL_TEXTURE0_ARB + ctx->Array.ActiveTexture;
+         break;
+      case GL_TEXTURE_CUBE_MAP_ARB:
+         CHECK_EXT1(ARB_texture_cube_map, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_CUBE_MAP_ARB));
+         break;
+      case GL_TEXTURE_BINDING_CUBE_MAP_ARB:
+         CHECK_EXT1(ARB_texture_cube_map, "GetInteger64v");
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_CUBE_INDEX]->Name;
+         break;
+      case GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB:
+         CHECK_EXT1(ARB_texture_cube_map, "GetInteger64v");
+         params[0] = (1 << (ctx->Const.MaxCubeTextureLevels - 1));
+         break;
+      case GL_TEXTURE_COMPRESSION_HINT_ARB:
+         params[0] = ctx->Hint.TextureCompression;
+         break;
+      case GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB:
+         params[0] = _mesa_get_compressed_formats(ctx, NULL, GL_FALSE);
+         break;
+      case GL_COMPRESSED_TEXTURE_FORMATS_ARB:
+         {
+         GLint formats[100];
+         GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE);
+         ASSERT(n <= 100);
+         for (i = 0; i < n; i++)
+            params[i] = ENUM_TO_INT64(formats[i]);
+         }
+         break;
+      case GL_ARRAY_ELEMENT_LOCK_FIRST_EXT:
+         CHECK_EXT1(EXT_compiled_vertex_array, "GetInteger64v");
+         params[0] = ctx->Array.LockFirst;
+         break;
+      case GL_ARRAY_ELEMENT_LOCK_COUNT_EXT:
+         CHECK_EXT1(EXT_compiled_vertex_array, "GetInteger64v");
+         params[0] = ctx->Array.LockCount;
+         break;
+      case GL_TRANSPOSE_COLOR_MATRIX_ARB:
+         {
+         const GLfloat *matrix = ctx->ColorMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[4]);
+         params[2] = IROUND64(matrix[8]);
+         params[3] = IROUND64(matrix[12]);
+         params[4] = IROUND64(matrix[1]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[9]);
+         params[7] = IROUND64(matrix[13]);
+         params[8] = IROUND64(matrix[2]);
+         params[9] = IROUND64(matrix[6]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[14]);
+         params[12] = IROUND64(matrix[3]);
+         params[13] = IROUND64(matrix[7]);
+         params[14] = IROUND64(matrix[11]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_TRANSPOSE_MODELVIEW_MATRIX_ARB:
+         {
+         const GLfloat *matrix = ctx->ModelviewMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[4]);
+         params[2] = IROUND64(matrix[8]);
+         params[3] = IROUND64(matrix[12]);
+         params[4] = IROUND64(matrix[1]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[9]);
+         params[7] = IROUND64(matrix[13]);
+         params[8] = IROUND64(matrix[2]);
+         params[9] = IROUND64(matrix[6]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[14]);
+         params[12] = IROUND64(matrix[3]);
+         params[13] = IROUND64(matrix[7]);
+         params[14] = IROUND64(matrix[11]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_TRANSPOSE_PROJECTION_MATRIX_ARB:
+         {
+         const GLfloat *matrix = ctx->ProjectionMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[4]);
+         params[2] = IROUND64(matrix[8]);
+         params[3] = IROUND64(matrix[12]);
+         params[4] = IROUND64(matrix[1]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[9]);
+         params[7] = IROUND64(matrix[13]);
+         params[8] = IROUND64(matrix[2]);
+         params[9] = IROUND64(matrix[6]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[14]);
+         params[12] = IROUND64(matrix[3]);
+         params[13] = IROUND64(matrix[7]);
+         params[14] = IROUND64(matrix[11]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_TRANSPOSE_TEXTURE_MATRIX_ARB:
+         {
+         const GLfloat *matrix = ctx->TextureMatrixStack[ctx->Texture.CurrentUnit].Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[4]);
+         params[2] = IROUND64(matrix[8]);
+         params[3] = IROUND64(matrix[12]);
+         params[4] = IROUND64(matrix[1]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[9]);
+         params[7] = IROUND64(matrix[13]);
+         params[8] = IROUND64(matrix[2]);
+         params[9] = IROUND64(matrix[6]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[14]);
+         params[12] = IROUND64(matrix[3]);
+         params[13] = IROUND64(matrix[7]);
+         params[14] = IROUND64(matrix[11]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_COLOR_MATRIX_SGI:
+         {
+         const GLfloat *matrix = ctx->ColorMatrixStack.Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[1]);
+         params[2] = IROUND64(matrix[2]);
+         params[3] = IROUND64(matrix[3]);
+         params[4] = IROUND64(matrix[4]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[6]);
+         params[7] = IROUND64(matrix[7]);
+         params[8] = IROUND64(matrix[8]);
+         params[9] = IROUND64(matrix[9]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[11]);
+         params[12] = IROUND64(matrix[12]);
+         params[13] = IROUND64(matrix[13]);
+         params[14] = IROUND64(matrix[14]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_COLOR_MATRIX_STACK_DEPTH_SGI:
+         params[0] = ctx->ColorMatrixStack.Depth + 1;
+         break;
+      case GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI:
+         params[0] = MAX_COLOR_STACK_DEPTH;
+         break;
+      case GL_POST_COLOR_MATRIX_RED_SCALE_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixScale[0]);
+         break;
+      case GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixScale[1]);
+         break;
+      case GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixScale[2]);
+         break;
+      case GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixScale[3]);
+         break;
+      case GL_POST_COLOR_MATRIX_RED_BIAS_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixBias[0]);
+         break;
+      case GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixBias[1]);
+         break;
+      case GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixBias[2]);
+         break;
+      case GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI:
+         params[0] = IROUND64(ctx->Pixel.PostColorMatrixBias[3]);
+         break;
+      case GL_CONVOLUTION_1D_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.Convolution1DEnabled);
+         break;
+      case GL_CONVOLUTION_2D_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.Convolution2DEnabled);
+         break;
+      case GL_SEPARABLE_2D_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.Separable2DEnabled);
+         break;
+      case GL_POST_CONVOLUTION_RED_SCALE_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionScale[0]);
+         break;
+      case GL_POST_CONVOLUTION_GREEN_SCALE_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionScale[1]);
+         break;
+      case GL_POST_CONVOLUTION_BLUE_SCALE_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionScale[2]);
+         break;
+      case GL_POST_CONVOLUTION_ALPHA_SCALE_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionScale[3]);
+         break;
+      case GL_POST_CONVOLUTION_RED_BIAS_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionBias[0]);
+         break;
+      case GL_POST_CONVOLUTION_GREEN_BIAS_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionBias[1]);
+         break;
+      case GL_POST_CONVOLUTION_BLUE_BIAS_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionBias[2]);
+         break;
+      case GL_POST_CONVOLUTION_ALPHA_BIAS_EXT:
+         CHECK_EXT1(EXT_convolution, "GetInteger64v");
+         params[0] = IROUND64(ctx->Pixel.PostConvolutionBias[3]);
+         break;
+      case GL_HISTOGRAM:
+         CHECK_EXT1(EXT_histogram, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.HistogramEnabled);
+         break;
+      case GL_MINMAX:
+         CHECK_EXT1(EXT_histogram, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.MinMaxEnabled);
+         break;
+      case GL_COLOR_TABLE_SGI:
+         CHECK_EXT1(SGI_color_table, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.ColorTableEnabled[COLORTABLE_PRECONVOLUTION]);
+         break;
+      case GL_POST_CONVOLUTION_COLOR_TABLE_SGI:
+         CHECK_EXT1(SGI_color_table, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.ColorTableEnabled[COLORTABLE_POSTCONVOLUTION]);
+         break;
+      case GL_POST_COLOR_MATRIX_COLOR_TABLE_SGI:
+         CHECK_EXT1(SGI_color_table, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Pixel.ColorTableEnabled[COLORTABLE_POSTCOLORMATRIX]);
+         break;
+      case GL_TEXTURE_COLOR_TABLE_SGI:
+         CHECK_EXT1(SGI_texture_color_table, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Texture.Unit[ctx->Texture.CurrentUnit].ColorTableEnabled);
+         break;
+      case GL_COLOR_SUM_EXT:
+         CHECK_EXT2(EXT_secondary_color, ARB_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Fog.ColorSumEnabled);
+         break;
+      case GL_CURRENT_SECONDARY_COLOR_EXT:
+         CHECK_EXT1(EXT_secondary_color, "GetInteger64v");
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0]);
+         params[1] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1]);
+         params[2] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2]);
+         params[3] = FLOAT_TO_INT64(ctx->Current.Attrib[VERT_ATTRIB_COLOR1][3]);
+         }
+         break;
+      case GL_SECONDARY_COLOR_ARRAY_EXT:
+         CHECK_EXT1(EXT_secondary_color, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->SecondaryColor.Enabled);
+         break;
+      case GL_SECONDARY_COLOR_ARRAY_TYPE_EXT:
+         CHECK_EXT1(EXT_secondary_color, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->SecondaryColor.Type);
+         break;
+      case GL_SECONDARY_COLOR_ARRAY_STRIDE_EXT:
+         CHECK_EXT1(EXT_secondary_color, "GetInteger64v");
+         params[0] = ctx->Array.ArrayObj->SecondaryColor.Stride;
+         break;
+      case GL_SECONDARY_COLOR_ARRAY_SIZE_EXT:
+         CHECK_EXT1(EXT_secondary_color, "GetInteger64v");
+         params[0] = ctx->Array.ArrayObj->SecondaryColor.Size;
+         break;
+      case GL_CURRENT_FOG_COORDINATE_EXT:
+         CHECK_EXT1(EXT_fog_coord, "GetInteger64v");
+         {
+         FLUSH_CURRENT(ctx, 0);
+         params[0] = IROUND64(ctx->Current.Attrib[VERT_ATTRIB_FOG][0]);
+         }
+         break;
+      case GL_FOG_COORDINATE_ARRAY_EXT:
+         CHECK_EXT1(EXT_fog_coord, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->FogCoord.Enabled);
+         break;
+      case GL_FOG_COORDINATE_ARRAY_TYPE_EXT:
+         CHECK_EXT1(EXT_fog_coord, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Array.ArrayObj->FogCoord.Type);
+         break;
+      case GL_FOG_COORDINATE_ARRAY_STRIDE_EXT:
+         CHECK_EXT1(EXT_fog_coord, "GetInteger64v");
+         params[0] = ctx->Array.ArrayObj->FogCoord.Stride;
+         break;
+      case GL_FOG_COORDINATE_SOURCE_EXT:
+         CHECK_EXT1(EXT_fog_coord, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Fog.FogCoordinateSource);
+         break;
+      case GL_MAX_TEXTURE_LOD_BIAS_EXT:
+         CHECK_EXT1(EXT_texture_lod_bias, "GetInteger64v");
+         params[0] = IROUND64(ctx->Const.MaxTextureLodBias);
+         break;
+      case GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT:
+         CHECK_EXT1(EXT_texture_filter_anisotropic, "GetInteger64v");
+         params[0] = IROUND64(ctx->Const.MaxTextureMaxAnisotropy);
+         break;
+      case GL_MULTISAMPLE_ARB:
+         params[0] = BOOLEAN_TO_INT64(ctx->Multisample.Enabled);
+         break;
+      case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
+         params[0] = BOOLEAN_TO_INT64(ctx->Multisample.SampleAlphaToCoverage);
+         break;
+      case GL_SAMPLE_ALPHA_TO_ONE_ARB:
+         params[0] = BOOLEAN_TO_INT64(ctx->Multisample.SampleAlphaToOne);
+         break;
+      case GL_SAMPLE_COVERAGE_ARB:
+         params[0] = BOOLEAN_TO_INT64(ctx->Multisample.SampleCoverage);
+         break;
+      case GL_SAMPLE_COVERAGE_VALUE_ARB:
+         params[0] = IROUND64(ctx->Multisample.SampleCoverageValue);
+         break;
+      case GL_SAMPLE_COVERAGE_INVERT_ARB:
+         params[0] = BOOLEAN_TO_INT64(ctx->Multisample.SampleCoverageInvert);
+         break;
+      case GL_SAMPLE_BUFFERS_ARB:
+         params[0] = ctx->DrawBuffer->Visual.sampleBuffers;
+         break;
+      case GL_SAMPLES_ARB:
+         params[0] = ctx->DrawBuffer->Visual.samples;
+         break;
+      case GL_RASTER_POSITION_UNCLIPPED_IBM:
+         CHECK_EXT1(IBM_rasterpos_clip, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Transform.RasterPositionUnclipped);
+         break;
+      case GL_POINT_SPRITE_NV:
+         CHECK_EXT2(NV_point_sprite, ARB_point_sprite, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Point.PointSprite);
+         break;
+      case GL_POINT_SPRITE_R_MODE_NV:
+         CHECK_EXT1(NV_point_sprite, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Point.SpriteRMode);
+         break;
+      case GL_POINT_SPRITE_COORD_ORIGIN:
+         CHECK_EXT2(NV_point_sprite, ARB_point_sprite, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Point.SpriteOrigin);
+         break;
+      case GL_GENERATE_MIPMAP_HINT_SGIS:
+         CHECK_EXT1(SGIS_generate_mipmap, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Hint.GenerateMipmap);
+         break;
+      case GL_VERTEX_PROGRAM_BINDING_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = (ctx->VertexProgram.Current ? ctx->VertexProgram.Current->Base.Id : 0);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY0_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[0].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY1_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[1].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY2_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[2].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY3_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[3].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[4].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY5_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[5].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY6_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[6].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY7_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[7].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY8_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[8].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY9_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[9].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY10_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[10].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY11_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[11].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY12_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[12].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY13_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[13].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY14_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[14].Enabled);
+         break;
+      case GL_VERTEX_ATTRIB_ARRAY15_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Array.ArrayObj->VertexAttrib[15].Enabled);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB0_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[0]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB1_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[1]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB2_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[2]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB3_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[3]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB4_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[4]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB5_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[5]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB6_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[6]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB7_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[7]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB8_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[8]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB9_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[9]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB10_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[10]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB11_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[11]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB12_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[12]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB13_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[13]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB14_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[14]);
+         break;
+      case GL_MAP1_VERTEX_ATTRIB15_4_NV:
+         CHECK_EXT1(NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Eval.Map1Attrib[15]);
+         break;
+      case GL_FRAGMENT_PROGRAM_NV:
+         CHECK_EXT1(NV_fragment_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->FragmentProgram.Enabled);
+         break;
+      case GL_FRAGMENT_PROGRAM_BINDING_NV:
+         CHECK_EXT1(NV_fragment_program, "GetInteger64v");
+         params[0] = ctx->FragmentProgram.Current ? ctx->FragmentProgram.Current->Base.Id : 0;
+         break;
+      case GL_MAX_FRAGMENT_PROGRAM_LOCAL_PARAMETERS_NV:
+         CHECK_EXT1(NV_fragment_program, "GetInteger64v");
+         params[0] = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
+         break;
+      case GL_TEXTURE_RECTANGLE_NV:
+         CHECK_EXT1(NV_texture_rectangle, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(_mesa_IsEnabled(GL_TEXTURE_RECTANGLE_NV));
+         break;
+      case GL_TEXTURE_BINDING_RECTANGLE_NV:
+         CHECK_EXT1(NV_texture_rectangle, "GetInteger64v");
+         params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].CurrentTex[TEXTURE_RECT_INDEX]->Name;
+         break;
+      case GL_MAX_RECTANGLE_TEXTURE_SIZE_NV:
+         CHECK_EXT1(NV_texture_rectangle, "GetInteger64v");
+         params[0] = ctx->Const.MaxTextureRectSize;
+         break;
+      case GL_STENCIL_TEST_TWO_SIDE_EXT:
+         CHECK_EXT1(EXT_stencil_two_side, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Stencil.TestTwoSide);
+         break;
+      case GL_ACTIVE_STENCIL_FACE_EXT:
+         CHECK_EXT1(EXT_stencil_two_side, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Stencil.ActiveFace ? GL_BACK : GL_FRONT);
+         break;
+      case GL_MAX_SHININESS_NV:
+         CHECK_EXT1(NV_light_max_exponent, "GetInteger64v");
+         params[0] = IROUND64(ctx->Const.MaxShininess);
+         break;
+      case GL_MAX_SPOT_EXPONENT_NV:
+         CHECK_EXT1(NV_light_max_exponent, "GetInteger64v");
+         params[0] = IROUND64(ctx->Const.MaxSpotExponent);
+         break;
+      case GL_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayBufferObj->Name;
+         break;
+      case GL_VERTEX_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->Vertex.BufferObj->Name;
+         break;
+      case GL_NORMAL_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->Normal.BufferObj->Name;
+         break;
+      case GL_COLOR_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->Color.BufferObj->Name;
+         break;
+      case GL_INDEX_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->Index.BufferObj->Name;
+         break;
+      case GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].BufferObj->Name;
+         break;
+      case GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name;
+         break;
+      case GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->SecondaryColor.BufferObj->Name;
+         break;
+      case GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ArrayObj->FogCoord.BufferObj->Name;
+         break;
+      case GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB:
+         params[0] = ctx->Array.ElementArrayBufferObj->Name;
+         break;
+      case GL_PIXEL_PACK_BUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_pixel_buffer_object, "GetInteger64v");
+         params[0] = ctx->Pack.BufferObj->Name;
+         break;
+      case GL_PIXEL_UNPACK_BUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_pixel_buffer_object, "GetInteger64v");
+         params[0] = ctx->Unpack.BufferObj->Name;
+         break;
+      case GL_VERTEX_PROGRAM_ARB:
+         CHECK_EXT2(ARB_vertex_program, NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->VertexProgram.Enabled);
+         break;
+      case GL_VERTEX_PROGRAM_POINT_SIZE_ARB:
+         CHECK_EXT2(ARB_vertex_program, NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->VertexProgram.PointSizeEnabled);
+         break;
+      case GL_VERTEX_PROGRAM_TWO_SIDE_ARB:
+         CHECK_EXT2(ARB_vertex_program, NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->VertexProgram.TwoSideEnabled);
+         break;
+      case GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB:
+         CHECK_EXT3(ARB_vertex_program, ARB_fragment_program, NV_vertex_program, "GetInteger64v");
+         params[0] = ctx->Const.MaxProgramMatrixStackDepth;
+         break;
+      case GL_MAX_PROGRAM_MATRICES_ARB:
+         CHECK_EXT3(ARB_vertex_program, ARB_fragment_program, NV_vertex_program, "GetInteger64v");
+         params[0] = ctx->Const.MaxProgramMatrices;
+         break;
+      case GL_CURRENT_MATRIX_STACK_DEPTH_ARB:
+         CHECK_EXT3(ARB_vertex_program, ARB_fragment_program, NV_vertex_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->CurrentStack->Depth + 1);
+         break;
+      case GL_CURRENT_MATRIX_ARB:
+         CHECK_EXT3(ARB_vertex_program, ARB_fragment_program, NV_fragment_program, "GetInteger64v");
+         {
+         const GLfloat *matrix = ctx->CurrentStack->Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[1]);
+         params[2] = IROUND64(matrix[2]);
+         params[3] = IROUND64(matrix[3]);
+         params[4] = IROUND64(matrix[4]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[6]);
+         params[7] = IROUND64(matrix[7]);
+         params[8] = IROUND64(matrix[8]);
+         params[9] = IROUND64(matrix[9]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[11]);
+         params[12] = IROUND64(matrix[12]);
+         params[13] = IROUND64(matrix[13]);
+         params[14] = IROUND64(matrix[14]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_TRANSPOSE_CURRENT_MATRIX_ARB:
+         CHECK_EXT2(ARB_vertex_program, ARB_fragment_program, "GetInteger64v");
+         {
+         const GLfloat *matrix = ctx->CurrentStack->Top->m;
+         params[0] = IROUND64(matrix[0]);
+         params[1] = IROUND64(matrix[4]);
+         params[2] = IROUND64(matrix[8]);
+         params[3] = IROUND64(matrix[12]);
+         params[4] = IROUND64(matrix[1]);
+         params[5] = IROUND64(matrix[5]);
+         params[6] = IROUND64(matrix[9]);
+         params[7] = IROUND64(matrix[13]);
+         params[8] = IROUND64(matrix[2]);
+         params[9] = IROUND64(matrix[6]);
+         params[10] = IROUND64(matrix[10]);
+         params[11] = IROUND64(matrix[14]);
+         params[12] = IROUND64(matrix[3]);
+         params[13] = IROUND64(matrix[7]);
+         params[14] = IROUND64(matrix[11]);
+         params[15] = IROUND64(matrix[15]);
+         }
+         break;
+      case GL_MAX_VERTEX_ATTRIBS_ARB:
+         CHECK_EXT1(ARB_vertex_program, "GetInteger64v");
+         params[0] = ctx->Const.VertexProgram.MaxAttribs;
+         break;
+      case GL_PROGRAM_ERROR_POSITION_ARB:
+         CHECK_EXT4(NV_vertex_program, ARB_vertex_program, NV_fragment_program, ARB_fragment_program, "GetInteger64v");
+         params[0] = ctx->Program.ErrorPos;
+         break;
+      case GL_FRAGMENT_PROGRAM_ARB:
+         CHECK_EXT1(ARB_fragment_program, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->FragmentProgram.Enabled);
+         break;
+      case GL_MAX_TEXTURE_COORDS_ARB:
+         CHECK_EXT2(ARB_fragment_program, NV_fragment_program, "GetInteger64v");
+         params[0] = ctx->Const.MaxTextureCoordUnits;
+         break;
+      case GL_MAX_TEXTURE_IMAGE_UNITS_ARB:
+         CHECK_EXT2(ARB_fragment_program, NV_fragment_program, "GetInteger64v");
+         params[0] = ctx->Const.MaxTextureImageUnits;
+         break;
+      case GL_DEPTH_BOUNDS_TEST_EXT:
+         CHECK_EXT1(EXT_depth_bounds_test, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Depth.BoundsTest);
+         break;
+      case GL_DEPTH_BOUNDS_EXT:
+         CHECK_EXT1(EXT_depth_bounds_test, "GetInteger64v");
+         params[0] = IROUND64(ctx->Depth.BoundsMin);
+         params[1] = IROUND64(ctx->Depth.BoundsMax);
+         break;
+      case GL_MAX_DRAW_BUFFERS_ARB:
+         params[0] = ctx->Const.MaxDrawBuffers;
+         break;
+      case GL_DRAW_BUFFER0_ARB:
+         params[0] = ENUM_TO_INT64(ctx->DrawBuffer->ColorDrawBuffer[0]);
+         break;
+      case GL_DRAW_BUFFER1_ARB:
+         {
+         GLenum buffer;
+         if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
+            _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
+            return;
+         }
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[1];
+         params[0] = ENUM_TO_INT64(buffer);
+         }
+         break;
+      case GL_DRAW_BUFFER2_ARB:
+         {
+         GLenum buffer;
+         if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
+            _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
+            return;
+         }
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[2];
+         params[0] = ENUM_TO_INT64(buffer);
+         }
+         break;
+      case GL_DRAW_BUFFER3_ARB:
+         {
+         GLenum buffer;
+         if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
+            _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
+            return;
+         }
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[3];
+         params[0] = ENUM_TO_INT64(buffer);
+         }
+         break;
+      case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES:
+         CHECK_EXT1(OES_read_format, "GetInteger64v");
+         params[0] = ctx->Const.ColorReadType;
+         break;
+      case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES:
+         CHECK_EXT1(OES_read_format, "GetInteger64v");
+         params[0] = ctx->Const.ColorReadFormat;
+         break;
+      case GL_NUM_FRAGMENT_REGISTERS_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 6;
+         break;
+      case GL_NUM_FRAGMENT_CONSTANTS_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 8;
+         break;
+      case GL_NUM_PASSES_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 2;
+         break;
+      case GL_NUM_INSTRUCTIONS_PER_PASS_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 8;
+         break;
+      case GL_NUM_INSTRUCTIONS_TOTAL_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 16;
+         break;
+      case GL_COLOR_ALPHA_PAIRING_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(GL_TRUE);
+         break;
+      case GL_NUM_LOOPBACK_COMPONENTS_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 3;
+         break;
+      case GL_NUM_INPUT_INTERPOLATOR_COMPONENTS_ATI:
+         CHECK_EXT1(ATI_fragment_shader, "GetInteger64v");
+         params[0] = 3;
+         break;
+      case GL_STENCIL_BACK_FUNC:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.Function[1]);
+         break;
+      case GL_STENCIL_BACK_VALUE_MASK:
+         params[0] = ctx->Stencil.ValueMask[1];
+         break;
+      case GL_STENCIL_BACK_WRITEMASK:
+         params[0] = ctx->Stencil.WriteMask[1];
+         break;
+      case GL_STENCIL_BACK_REF:
+         params[0] = ctx->Stencil.Ref[1];
+         break;
+      case GL_STENCIL_BACK_FAIL:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.FailFunc[1]);
+         break;
+      case GL_STENCIL_BACK_PASS_DEPTH_FAIL:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.ZFailFunc[1]);
+         break;
+      case GL_STENCIL_BACK_PASS_DEPTH_PASS:
+         params[0] = ENUM_TO_INT64(ctx->Stencil.ZPassFunc[1]);
+         break;
+      case GL_FRAMEBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_object, "GetInteger64v");
+         params[0] = ctx->DrawBuffer->Name;
+         break;
+      case GL_RENDERBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_object, "GetInteger64v");
+         params[0] = ctx->CurrentRenderbuffer ? ctx->CurrentRenderbuffer->Name : 0;
+         break;
+      case GL_MAX_COLOR_ATTACHMENTS_EXT:
+         CHECK_EXT1(EXT_framebuffer_object, "GetInteger64v");
+         params[0] = ctx->Const.MaxColorAttachments;
+         break;
+      case GL_MAX_RENDERBUFFER_SIZE_EXT:
+         CHECK_EXT1(EXT_framebuffer_object, "GetInteger64v");
+         params[0] = ctx->Const.MaxRenderbufferSize;
+         break;
+      case GL_READ_FRAMEBUFFER_BINDING_EXT:
+         CHECK_EXT1(EXT_framebuffer_blit, "GetInteger64v");
+         params[0] = ctx->ReadBuffer->Name;
+         break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Light.ProvokingVertex);
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Const.QuadsFollowProvokingVertexConvention);
+         break;
+      case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
+         CHECK_EXT1(ARB_fragment_shader, "GetInteger64v");
+         params[0] = ctx->Const.FragmentProgram.MaxUniformComponents;
+         break;
+      case GL_FRAGMENT_SHADER_DERIVATIVE_HINT_ARB:
+         CHECK_EXT1(ARB_fragment_shader, "GetInteger64v");
+         params[0] = ENUM_TO_INT64(ctx->Hint.FragmentShaderDerivative);
+         break;
+      case GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB:
+         CHECK_EXT1(ARB_vertex_shader, "GetInteger64v");
+         params[0] = ctx->Const.VertexProgram.MaxUniformComponents;
+         break;
+      case GL_MAX_VARYING_FLOATS_ARB:
+         CHECK_EXT1(ARB_vertex_shader, "GetInteger64v");
+         params[0] = ctx->Const.MaxVarying * 4;
+         break;
+      case GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB:
+         CHECK_EXT1(ARB_vertex_shader, "GetInteger64v");
+         params[0] = ctx->Const.MaxVertexTextureImageUnits;
+         break;
+      case GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB:
+         CHECK_EXT1(ARB_vertex_shader, "GetInteger64v");
+         params[0] = MAX_COMBINED_TEXTURE_IMAGE_UNITS;
+         break;
+      case GL_CURRENT_PROGRAM:
+         CHECK_EXT1(ARB_shader_objects, "GetInteger64v");
+         params[0] = ctx->Shader.CurrentProgram ? ctx->Shader.CurrentProgram->Name : 0;
+         break;
+      case GL_MAX_SAMPLES:
+         CHECK_EXT1(ARB_framebuffer_object, "GetInteger64v");
+         params[0] = ctx->Const.MaxSamples;
+         break;
+      case GL_VERTEX_ARRAY_BINDING_APPLE:
+         CHECK_EXT1(APPLE_vertex_array_object, "GetInteger64v");
+         params[0] = ctx->Array.ArrayObj->Name;
+         break;
+      case GL_TEXTURE_CUBE_MAP_SEAMLESS:
+         CHECK_EXT1(ARB_seamless_cube_map, "GetInteger64v");
+         params[0] = BOOLEAN_TO_INT64(ctx->Texture.CubeMapSeamless);
+         break;
+      case GL_MAX_SERVER_WAIT_TIMEOUT:
+         CHECK_EXT1(ARB_sync, "GetInteger64v");
+         params[0] = ctx->Const.MaxServerWaitTimeout;
+         break;
+      default:
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetInteger64v(pname=0x%x)", pname);
+   }
+}
+#endif /* FEATURE_ARB_sync */
+
 
 void GLAPIENTRY
 _mesa_GetDoublev( GLenum pname, GLdouble *params )
diff --git a/src/mesa/main/get.h b/src/mesa/main/get.h
index 9b47b230a9..77a9a7d04b 100644
--- a/src/mesa/main/get.h
+++ b/src/mesa/main/get.h
@@ -48,6 +48,9 @@ extern void GLAPIENTRY
 _mesa_GetIntegerv( GLenum pname, GLint *params );
 
 extern void GLAPIENTRY
+_mesa_GetInteger64v( GLenum pname, GLint64 *params );
+
+extern void GLAPIENTRY
 _mesa_GetPointerv( GLenum pname, GLvoid **params );
 
 extern const GLubyte * GLAPIENTRY
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 43ee5fff10..2878c1b552 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -35,6 +35,7 @@ GLfloat = 3
 GLdouble = 4
 GLboolean = 5
 GLfloatN = 6    # A normalized value, such as a color or depth range
+GLint64 = 7
 
 
 TypeStrings = {
@@ -42,7 +43,8 @@ TypeStrings = {
 	GLenum : "GLenum",
 	GLfloat : "GLfloat",
 	GLdouble : "GLdouble",
-	GLboolean : "GLboolean"
+	GLboolean : "GLboolean",
+	GLint64 : "GLint64"
 }
 
 
@@ -176,7 +178,8 @@ StateVars = [
 	   "ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][1]",
 	   "ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][2]",
 	   "ctx->Current.Attrib[VERT_ATTRIB_TEX0 + texUnit][3]"],
-	  "const GLuint texUnit = ctx->Texture.CurrentUnit;", None ),
+	  """const GLuint texUnit = ctx->Texture.CurrentUnit;
+         FLUSH_CURRENT(ctx, 0);""", None ),
 	( "GL_DEPTH_BIAS", GLfloat, ["ctx->Pixel.DepthBias"], "", None ),
 	( "GL_DEPTH_BITS", GLint, ["ctx->DrawBuffer->Visual.depthBits"],
 	  "", None ),
@@ -976,6 +979,13 @@ StateVars = [
 	( "GL_READ_FRAMEBUFFER_BINDING_EXT", GLint, ["ctx->ReadBuffer->Name"], "",
 	  ["EXT_framebuffer_blit"] ),
 
+	# GL_EXT_provoking_vertex
+	( "GL_PROVOKING_VERTEX_EXT", GLboolean,
+	  ["ctx->Light.ProvokingVertex"], "", ["EXT_provoking_vertex"] ),
+	( "GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT", GLboolean,
+	  ["ctx->Const.QuadsFollowProvokingVertexConvention"], "",
+	  ["EXT_provoking_vertex"] ),
+
 	# GL_ARB_fragment_shader
 	( "GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB", GLint,
 	  ["ctx->Const.FragmentProgram.MaxUniformComponents"], "",
@@ -1008,6 +1018,14 @@ StateVars = [
 	# GL_APPLE_vertex_array_object
 	( "GL_VERTEX_ARRAY_BINDING_APPLE", GLint, ["ctx->Array.ArrayObj->Name"], "",
 	  ["APPLE_vertex_array_object"] ),
+
+	# GL_ARB_seamless_cube_map
+	( "GL_TEXTURE_CUBE_MAP_SEAMLESS", GLboolean, ["ctx->Texture.CubeMapSeamless"], "",
+	  ["ARB_seamless_cube_map"] ),
+
+	# GL_ARB_sync
+	( "GL_MAX_SERVER_WAIT_TIMEOUT", GLint64, ["ctx->Const.MaxServerWaitTimeout"], "",
+	  ["ARB_sync"] ),
 ]
 
 
@@ -1017,10 +1035,16 @@ def ConversionFunc(fromType, toType):
 		return ""
 	elif fromType == GLfloat and toType == GLint:
 		return "IROUND"
+	elif fromType == GLfloat and toType == GLint64:
+		return "IROUND64"
 	elif fromType == GLfloatN and toType == GLfloat:
 		return ""
 	elif fromType == GLint and toType == GLfloat: # but not GLfloatN!
 		return "(GLfloat)"
+	elif fromType == GLint and toType == GLint64:
+		return ""
+	elif fromType == GLint64 and toType == GLfloat: # but not GLfloatN!
+		return "(GLfloat)"
 	else:
 		if fromType == GLfloatN:
 			fromType = GLfloat
@@ -1035,6 +1059,7 @@ def EmitGetFunction(stateVars, returnType):
 	"""Emit the code to implement glGetBooleanv, glGetIntegerv or glGetFloatv."""
 	assert (returnType == GLboolean or
 			returnType == GLint or
+			returnType == GLint64 or
 			returnType == GLfloat)
 
 	strType = TypeStrings[returnType]
@@ -1045,9 +1070,14 @@ def EmitGetFunction(stateVars, returnType):
 		function = "GetBooleanv"
 	elif returnType == GLfloat:
 		function = "GetFloatv"
+	elif returnType == GLint64:
+		function = "GetInteger64v"
 	else:
 		abort()
 
+	if returnType == GLint64:
+		print "#if FEATURE_ARB_sync"
+
 	print "void GLAPIENTRY"
 	print "_mesa_%s( GLenum pname, %s *params )" % (function, strType)
 	print "{"
@@ -1101,6 +1131,8 @@ def EmitGetFunction(stateVars, returnType):
 	print '         _mesa_error(ctx, GL_INVALID_ENUM, "gl%s(pname=0x%%x)", pname);' % function
 	print "   }"
 	print "}"
+	if returnType == GLint64:
+		print "#endif /* FEATURE_ARB_sync */"
 	print ""
 	return
 
@@ -1129,9 +1161,15 @@ def EmitHeader():
 
 #define INT_TO_BOOLEAN(I)     ( (I) ? GL_TRUE : GL_FALSE )
 
+#define INT64_TO_BOOLEAN(I)   ( (I) ? GL_TRUE : GL_FALSE )
+#define INT64_TO_INT(I)       ( (GLint)((I > INT_MAX) ? INT_MAX : ((I < INT_MIN) ? INT_MIN : (I))) )
+
 #define BOOLEAN_TO_INT(B)     ( (GLint) (B) )
+#define BOOLEAN_TO_INT64(B)   ( (GLint64) (B) )
 #define BOOLEAN_TO_FLOAT(B)   ( (B) ? 1.0F : 0.0F )
 
+#define ENUM_TO_INT64(E)      ( (GLint64) (E) )
+
 
 /*
  * Check if named extension is enabled, if not generate error and return.
@@ -1209,5 +1247,6 @@ EmitHeader()
 EmitGetFunction(StateVars, GLboolean)
 EmitGetFunction(StateVars, GLfloat)
 EmitGetFunction(StateVars, GLint)
+EmitGetFunction(StateVars, GLint64)
 EmitGetDoublev()
 
diff --git a/src/mesa/main/getstring.c b/src/mesa/main/getstring.c
index 41fd786d7d..6599ed9698 100644
--- a/src/mesa/main/getstring.c
+++ b/src/mesa/main/getstring.c
@@ -266,5 +266,6 @@ _mesa_GetError( void )
       _mesa_debug(ctx, "glGetError <-- %s\n", _mesa_lookup_enum_by_nr(e));
 
    ctx->ErrorValue = (GLenum) GL_NO_ERROR;
+   ctx->ErrorDebugCount = 0;
    return e;
 }
diff --git a/src/mesa/main/histogram.c b/src/mesa/main/histogram.c
index 5fee4fd0e3..ceb0d5a6a8 100644
--- a/src/mesa/main/histogram.c
+++ b/src/mesa/main/histogram.c
@@ -649,29 +649,11 @@ _mesa_GetMinmax(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvo
       return;
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack min/max values into a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Pack, 2, 1, 1,
-                                     format, type, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetMinMax(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION,"glGetMinMax(PBO is mapped)");
-         return;
-      }
-      values = ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
-      /* not an error */
+
+   values = _mesa_map_validate_pbo_dest(ctx, 1, &ctx->Pack, 2, 1, 1,
+                                        format, type, values, "glGetMinmax");
+   if (!values)
       return;
-   }
 
    {
       GLfloat minmax[2][4];
@@ -687,10 +669,7 @@ _mesa_GetMinmax(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvo
                                  format, type, values, &ctx->Pack, 0x0);
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 
    if (reset) {
       _mesa_ResetMinmax(GL_MINMAX);
@@ -733,38 +712,18 @@ _mesa_GetHistogram(GLenum target, GLboolean reset, GLenum format, GLenum type, G
       return;
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack min/max values into a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(1, &ctx->Pack, ctx->Histogram.Width, 1, 1,
-                                     format, type, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetHistogram(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx,GL_INVALID_OPERATION,"glGetHistogram(PBO is mapped)");
-         return;
-      }
-      values = ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
-      /* not an error */
+   values = _mesa_map_validate_pbo_dest(ctx, 1, &ctx->Pack,
+                                        ctx->Histogram.Width, 1, 1,
+                                        format, type, values,
+                                        "glGetHistogram");
+   if (!values)
       return;
-   }
 
    pack_histogram(ctx, ctx->Histogram.Width,
                   (CONST GLuint (*)[4]) ctx->Histogram.Count,
                   format, type, values, &ctx->Pack);
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 
    if (reset) {
       GLuint i;
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index 1a4d22256f..baecbab0a3 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -182,6 +182,8 @@ _mesa_sizeof_type( GLenum type )
 	 return sizeof(GLint);
       case GL_FLOAT:
 	 return sizeof(GLfloat);
+      case GL_DOUBLE:
+	 return sizeof(GLdouble);
       case GL_HALF_FLOAT_ARB:
 	 return sizeof(GLhalfARB);
       default:
@@ -529,6 +531,226 @@ _mesa_is_legal_format_and_type( GLcontext *ctx, GLenum format, GLenum type )
 
 
 /**
+ * Test if the given image format is a color/RGBA format (i.e., not color
+ * index, depth, stencil, etc).
+ * \param format  the image format value (may by an internal texture format)
+ * \return GL_TRUE if its a color/RGBA format, GL_FALSE otherwise.
+ */
+GLboolean
+_mesa_is_color_format(GLenum format)
+{
+   switch (format) {
+      case GL_RED:
+      case GL_GREEN:
+      case GL_BLUE:
+      case GL_ALPHA:
+      case GL_ALPHA4:
+      case GL_ALPHA8:
+      case GL_ALPHA12:
+      case GL_ALPHA16:
+      case 1:
+      case GL_LUMINANCE:
+      case GL_LUMINANCE4:
+      case GL_LUMINANCE8:
+      case GL_LUMINANCE12:
+      case GL_LUMINANCE16:
+      case 2:
+      case GL_LUMINANCE_ALPHA:
+      case GL_LUMINANCE4_ALPHA4:
+      case GL_LUMINANCE6_ALPHA2:
+      case GL_LUMINANCE8_ALPHA8:
+      case GL_LUMINANCE12_ALPHA4:
+      case GL_LUMINANCE12_ALPHA12:
+      case GL_LUMINANCE16_ALPHA16:
+      case GL_INTENSITY:
+      case GL_INTENSITY4:
+      case GL_INTENSITY8:
+      case GL_INTENSITY12:
+      case GL_INTENSITY16:
+      case 3:
+      case GL_RGB:
+      case GL_BGR:
+      case GL_R3_G3_B2:
+      case GL_RGB4:
+      case GL_RGB5:
+      case GL_RGB8:
+      case GL_RGB10:
+      case GL_RGB12:
+      case GL_RGB16:
+      case 4:
+      case GL_ABGR_EXT:
+      case GL_RGBA:
+      case GL_BGRA:
+      case GL_RGBA2:
+      case GL_RGBA4:
+      case GL_RGB5_A1:
+      case GL_RGBA8:
+      case GL_RGB10_A2:
+      case GL_RGBA12:
+      case GL_RGBA16:
+      /* float texture formats */
+      case GL_ALPHA16F_ARB:
+      case GL_ALPHA32F_ARB:
+      case GL_LUMINANCE16F_ARB:
+      case GL_LUMINANCE32F_ARB:
+      case GL_LUMINANCE_ALPHA16F_ARB:
+      case GL_LUMINANCE_ALPHA32F_ARB:
+      case GL_INTENSITY16F_ARB:
+      case GL_INTENSITY32F_ARB:
+      case GL_RGB16F_ARB:
+      case GL_RGB32F_ARB:
+      case GL_RGBA16F_ARB:
+      case GL_RGBA32F_ARB:
+      /* compressed formats */
+      case GL_COMPRESSED_ALPHA:
+      case GL_COMPRESSED_LUMINANCE:
+      case GL_COMPRESSED_LUMINANCE_ALPHA:
+      case GL_COMPRESSED_INTENSITY:
+      case GL_COMPRESSED_RGB:
+      case GL_COMPRESSED_RGBA:
+      case GL_RGB_S3TC:
+      case GL_RGB4_S3TC:
+      case GL_RGBA_S3TC:
+      case GL_RGBA4_S3TC:
+      case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+      case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+      case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+      case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+      case GL_COMPRESSED_RGB_FXT1_3DFX:
+      case GL_COMPRESSED_RGBA_FXT1_3DFX:
+#if FEATURE_EXT_texture_sRGB
+      case GL_SRGB_EXT:
+      case GL_SRGB8_EXT:
+      case GL_SRGB_ALPHA_EXT:
+      case GL_SRGB8_ALPHA8_EXT:
+      case GL_SLUMINANCE_ALPHA_EXT:
+      case GL_SLUMINANCE8_ALPHA8_EXT:
+      case GL_SLUMINANCE_EXT:
+      case GL_SLUMINANCE8_EXT:
+      case GL_COMPRESSED_SRGB_EXT:
+      case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+      case GL_COMPRESSED_SRGB_ALPHA_EXT:
+      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+      case GL_COMPRESSED_SLUMINANCE_EXT:
+      case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
+#endif /* FEATURE_EXT_texture_sRGB */
+         return GL_TRUE;
+      /* signed texture formats */
+      case GL_RGBA_SNORM:
+      case GL_RGBA8_SNORM:
+         return GL_TRUE;
+      case GL_YCBCR_MESA:  /* not considered to be RGB */
+         /* fall-through */
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
+ * Test if the given image format is a color index format.
+ */
+GLboolean
+_mesa_is_index_format(GLenum format)
+{
+   switch (format) {
+      case GL_COLOR_INDEX:
+      case GL_COLOR_INDEX1_EXT:
+      case GL_COLOR_INDEX2_EXT:
+      case GL_COLOR_INDEX4_EXT:
+      case GL_COLOR_INDEX8_EXT:
+      case GL_COLOR_INDEX12_EXT:
+      case GL_COLOR_INDEX16_EXT:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
+ * Test if the given image format is a depth component format.
+ */
+GLboolean
+_mesa_is_depth_format(GLenum format)
+{
+   switch (format) {
+      case GL_DEPTH_COMPONENT:
+      case GL_DEPTH_COMPONENT16:
+      case GL_DEPTH_COMPONENT24:
+      case GL_DEPTH_COMPONENT32:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
+ * Test if the given image format is a stencil format.
+ */
+GLboolean
+_mesa_is_stencil_format(GLenum format)
+{
+   switch (format) {
+      case GL_STENCIL_INDEX:
+      case GL_DEPTH_STENCIL:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
+ * Test if the given image format is a YCbCr format.
+ */
+GLboolean
+_mesa_is_ycbcr_format(GLenum format)
+{
+   switch (format) {
+      case GL_YCBCR_MESA:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
+ * Test if the given image format is a depth+stencil format.
+ */
+GLboolean
+_mesa_is_depthstencil_format(GLenum format)
+{
+   switch (format) {
+      case GL_DEPTH24_STENCIL8_EXT:
+      case GL_DEPTH_STENCIL_EXT:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+/**
+ * Test if the given image format is a dudv format.
+ */
+GLboolean
+_mesa_is_dudv_format(GLenum format)
+{
+   switch (format) {
+      case GL_DUDV_ATI:
+      case GL_DU8DV8_ATI:
+         return GL_TRUE;
+      default:
+         return GL_FALSE;
+   }
+}
+
+
+/**
  * Return the address of a specific pixel in an image (1D, 2D or 3D).
  *
  * Pixel unpacking/packing parameters are observed according to \p packing.
@@ -1037,6 +1259,91 @@ _mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source,
 }
 
 
+/**
+ * "Expand" a bitmap from 1-bit per pixel to 8-bits per pixel.
+ * This is typically used to convert a bitmap into a GLubyte/pixel texture.
+ * "On" bits will set texels to \p onValue.
+ * "Off" bits will not modify texels.
+ * \param width  src bitmap width in pixels
+ * \param height  src bitmap height in pixels
+ * \param unpack  bitmap unpacking state
+ * \param bitmap  the src bitmap data
+ * \param destBuffer  start of dest buffer
+ * \param destStride  row stride in dest buffer
+ * \param onValue  if bit is 1, set destBuffer pixel to this value
+ */
+void
+_mesa_expand_bitmap(GLsizei width, GLsizei height,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLubyte *bitmap,
+                    GLubyte *destBuffer, GLint destStride,
+                    GLubyte onValue)
+{
+   const GLubyte *srcRow = (const GLubyte *)
+      _mesa_image_address2d(unpack, bitmap, width, height,
+                            GL_COLOR_INDEX, GL_BITMAP, 0, 0);
+   const GLint srcStride = _mesa_image_row_stride(unpack, width,
+                                                  GL_COLOR_INDEX, GL_BITMAP);
+   GLint row, col;
+
+#define SET_PIXEL(COL, ROW) \
+   destBuffer[(ROW) * destStride + (COL)] = onValue;
+
+   for (row = 0; row < height; row++) {
+      const GLubyte *src = srcRow;
+
+      if (unpack->LsbFirst) {
+         /* Lsb first */
+         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 128U) {
+               src++;
+               mask = 1U;
+            }
+            else {
+               mask = mask << 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 1)
+            src++;
+      }
+      else {
+         /* Msb first */
+         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 1U) {
+               src++;
+               mask = 128U;
+            }
+            else {
+               mask = mask >> 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 128)
+            src++;
+      }
+
+      srcRow += srcStride;
+   } /* row */
+
+#undef SET_PIXEL
+}
+
+
 /**********************************************************************/
 /*****                  Pixel processing functions               ******/
 /**********************************************************************/
diff --git a/src/mesa/main/image.h b/src/mesa/main/image.h
index ee17accb80..72717d6787 100644
--- a/src/mesa/main/image.h
+++ b/src/mesa/main/image.h
@@ -54,6 +54,27 @@ _mesa_bytes_per_pixel( GLenum format, GLenum type );
 extern GLboolean
 _mesa_is_legal_format_and_type( GLcontext *ctx, GLenum format, GLenum type );
 
+extern GLboolean
+_mesa_is_color_format(GLenum format);
+
+extern GLboolean
+_mesa_is_index_format(GLenum format);
+
+extern GLboolean
+_mesa_is_depth_format(GLenum format);
+
+extern GLboolean
+_mesa_is_stencil_format(GLenum format);
+
+extern GLboolean
+_mesa_is_ycbcr_format(GLenum format);
+
+extern GLboolean
+_mesa_is_depthstencil_format(GLenum format);
+
+extern GLboolean
+_mesa_is_dudv_format(GLenum format);
+
 
 extern GLvoid *
 _mesa_image_address( GLuint dimensions,
@@ -113,6 +134,13 @@ extern void
 _mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source,
                    GLubyte *dest, const struct gl_pixelstore_attrib *packing );
 
+extern void
+_mesa_expand_bitmap(GLsizei width, GLsizei height,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLubyte *bitmap,
+                    GLubyte *destBuffer, GLint destStride,
+                    GLubyte onValue);
+
 
 /** \name Pixel processing functions */
 /*@{*/
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 1722579e82..6ffaddcde9 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -911,6 +911,20 @@ _mesa_strtod( const char *s, char **end )
    return strtod(s, end);
 }
 
+/** Compute simple checksum/hash for a string */
+unsigned int
+_mesa_str_checksum(const char *str)
+{
+   /* This could probably be much better */
+   unsigned int sum, i;
+   const char *c;
+   sum = i = 1;
+   for (c = str; *c; c++)
+      sum += *c * (i % 100);
+   return sum;
+}
+
+
 /*@}*/
 
 
@@ -1021,23 +1035,59 @@ output_if_debug(const char *prefixString, const char *outputString,
    }
 }
 
-static const char *error_string( GLenum error );
 
-static void flush_delayed_errors( GLcontext *ctx )
+/**
+ * Return string version of GL error code.
+ */
+static const char *
+error_string( GLenum error )
+{
+   switch (error) {
+   case GL_NO_ERROR:
+      return "GL_NO_ERROR";
+   case GL_INVALID_VALUE:
+      return "GL_INVALID_VALUE";
+   case GL_INVALID_ENUM:
+      return "GL_INVALID_ENUM";
+   case GL_INVALID_OPERATION:
+      return "GL_INVALID_OPERATION";
+   case GL_STACK_OVERFLOW:
+      return "GL_STACK_OVERFLOW";
+   case GL_STACK_UNDERFLOW:
+      return "GL_STACK_UNDERFLOW";
+   case GL_OUT_OF_MEMORY:
+      return "GL_OUT_OF_MEMORY";
+   case GL_TABLE_TOO_LARGE:
+      return "GL_TABLE_TOO_LARGE";
+   case GL_INVALID_FRAMEBUFFER_OPERATION_EXT:
+      return "GL_INVALID_FRAMEBUFFER_OPERATION";
+   default:
+      return "unknown";
+   }
+}
+
+
+/**
+ * When a new type of error is recorded, print a message describing
+ * previous errors which were accumulated.
+ */
+static void
+flush_delayed_errors( GLcontext *ctx )
 {
-   char s2[MAXSTRING];
+   char s[MAXSTRING];
 
    if (ctx->ErrorDebugCount) {
-      _mesa_snprintf(s2, MAXSTRING, "%d similar %s errors", 
+      _mesa_snprintf(s, MAXSTRING, "%d similar %s errors", 
                      ctx->ErrorDebugCount,
                      error_string(ctx->ErrorValue));
 
-      output_if_debug("Mesa: ", s2, GL_TRUE);
+      output_if_debug("Mesa", s, GL_TRUE);
 
       ctx->ErrorDebugCount = 0;
    }
 }
 
+
 /**
  * Report a warning (a recoverable error condition) to stderr if
  * either DEBUG is defined or the MESA_DEBUG env var is set.
@@ -1083,31 +1133,6 @@ _mesa_problem( const GLcontext *ctx, const char *fmtString, ... )
    fprintf(stderr, "Please report at bugzilla.freedesktop.org\n");
 }
 
-static const char *error_string( GLenum error )
-{
-   switch (error) {
-   case GL_NO_ERROR:
-      return "GL_NO_ERROR";
-   case GL_INVALID_VALUE:
-      return "GL_INVALID_VALUE";
-   case GL_INVALID_ENUM:
-      return "GL_INVALID_ENUM";
-   case GL_INVALID_OPERATION:
-      return "GL_INVALID_OPERATION";
-   case GL_STACK_OVERFLOW:
-      return "GL_STACK_OVERFLOW";
-   case GL_STACK_UNDERFLOW:
-      return "GL_STACK_UNDERFLOW";
-   case GL_OUT_OF_MEMORY:
-      return "GL_OUT_OF_MEMORY";
-   case GL_TABLE_TOO_LARGE:
-      return "GL_TABLE_TOO_LARGE";
-   case GL_INVALID_FRAMEBUFFER_OPERATION_EXT:
-      return "GL_INVALID_FRAMEBUFFER_OPERATION";
-   default:
-      return "unknown";
-   }
-}
 
 /**
  * Record an OpenGL state error.  These usually occur when the user
diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h
index 7b61e22e93..7d4012a856 100644
--- a/src/mesa/main/imports.h
+++ b/src/mesa/main/imports.h
@@ -291,6 +291,7 @@ long iround(float f);
 #define IROUND(f)  ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 #endif
 
+#define IROUND64(f)  ((GLint64) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F)))
 
 /***
  *** IROUND_POS: return (as an integer) positive float rounded to nearest int
@@ -586,6 +587,9 @@ _mesa_atoi( const char *s );
 extern double
 _mesa_strtod( const char *s, char **end );
 
+extern unsigned int
+_mesa_str_checksum(const char *str);
+
 extern int
 _mesa_sprintf( char *str, const char *fmt, ... );
 
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index 7479fe8fc4..10c89f4368 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.0
+ * Version:  7.5
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -64,6 +65,37 @@ _mesa_ShadeModel( GLenum mode )
 
 
 /**
+ * Set the provoking vertex (the vertex which specifies the prim's
+ * color when flat shading) to either the first or last vertex of the
+ * triangle or line.
+ */
+void GLAPIENTRY
+_mesa_ProvokingVertexEXT(GLenum mode)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (MESA_VERBOSE&VERBOSE_API)
+      _mesa_debug(ctx, "glProvokingVertexEXT 0x%x\n", mode);
+
+   switch (mode) {
+   case GL_FIRST_VERTEX_CONVENTION_EXT:
+   case GL_LAST_VERTEX_CONVENTION_EXT:
+      break;
+   default:
+      _mesa_error(ctx, GL_INVALID_ENUM, "glProvokingVertexEXT(0x%x)", mode);
+      return;
+   }
+
+   if (ctx->Light.ProvokingVertex == mode)
+      return;
+
+   FLUSH_VERTICES(ctx, _NEW_LIGHT);
+   ctx->Light.ProvokingVertex = mode;
+}
+
+
+/**
  * Helper function called by _mesa_Lightfv and _mesa_PopAttrib to set
  * per-light state.
  * For GL_POSITION and GL_SPOT_DIRECTION the params position/direction
@@ -1348,6 +1380,7 @@ _mesa_init_lighting( GLcontext *ctx )
    init_lightmodel( &ctx->Light.Model );
    init_material( &ctx->Light.Material );
    ctx->Light.ShadeModel = GL_SMOOTH;
+   ctx->Light.ProvokingVertex = GL_LAST_VERTEX_CONVENTION_EXT;
    ctx->Light.Enabled = GL_FALSE;
    ctx->Light.ColorMaterialFace = GL_FRONT_AND_BACK;
    ctx->Light.ColorMaterialMode = GL_AMBIENT_AND_DIFFUSE;
diff --git a/src/mesa/main/light.h b/src/mesa/main/light.h
index b97e17b5be..9c1a5eefad 100644
--- a/src/mesa/main/light.h
+++ b/src/mesa/main/light.h
@@ -1,13 +1,9 @@
-/**
- * \file light.h
- * Lighting.
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  3.5
+ * Version:  7.5
  *
- * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -37,6 +33,10 @@
 extern void GLAPIENTRY
 _mesa_ShadeModel( GLenum mode );
 
+extern void GLAPIENTRY
+_mesa_ProvokingVertexEXT(GLenum mode);
+
+
 #if _HAVE_FULL_GL
 extern void GLAPIENTRY
 _mesa_ColorMaterial( GLenum face, GLenum mode );
diff --git a/src/mesa/main/macros.h b/src/mesa/main/macros.h
index 4ca7957ffa..3d9a1aba98 100644
--- a/src/mesa/main/macros.h
+++ b/src/mesa/main/macros.h
@@ -99,6 +99,9 @@ extern GLfloat _mesa_ubyte_to_float_color_tab[256];
 /* a close approximation: */
 #define FLOAT_TO_INT(X)     ( (GLint) (2147483647.0 * (X)) )
 
+/** Convert GLfloat in [-1.0,1.0] to GLint64 in [-(1<<63),(1 << 63) -1] */
+#define FLOAT_TO_INT64(X)     ( (GLint64) (9223372036854775807.0 * (double)(X)) )
+
 
 /** Convert GLint in [-2147483648,2147483647] to GLfloat in [-1.0,1.0], texture/fb data */
 #define INT_TO_FLOAT_TEX(I)    ((I) == -2147483648 ? -1.0F : (I) * (1.0F/2147483647.0))
diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h
index 62c3ead3e1..e23cdb1f42 100644
--- a/src/mesa/main/mfeatures.h
+++ b/src/mesa/main/mfeatures.h
@@ -61,6 +61,7 @@
 #define FEATURE_ARB_occlusion_query  _HAVE_FULL_GL
 #define FEATURE_ARB_fragment_program  _HAVE_FULL_GL
 #define FEATURE_ARB_framebuffer_object  _HAVE_FULL_GL
+#define FEATURE_ARB_map_buffer_range  _HAVE_FULL_GL
 #define FEATURE_ARB_pixel_buffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_buffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_program  _HAVE_FULL_GL
@@ -69,6 +70,7 @@
 #define FEATURE_ARB_shader_objects (FEATURE_ARB_vertex_shader || FEATURE_ARB_fragment_shader)
 #define FEATURE_ARB_shading_language_100 FEATURE_ARB_shader_objects
 #define FEATURE_ARB_shading_language_120 FEATURE_ARB_shader_objects
+#define FEATURE_ARB_sync _HAVE_FULL_GL
 
 #define FEATURE_EXT_framebuffer_blit _HAVE_FULL_GL
 #define FEATURE_EXT_framebuffer_object _HAVE_FULL_GL
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index b306700484..3dca09d9f2 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -1481,9 +1481,12 @@ next_mipmap_level_size(GLenum target, GLint border,
 
 
 /**
- * For GL_SGIX_generate_mipmap:
- * Generate a complete set of mipmaps from texObj's base-level image.
+ * Automatic mipmap generation.
+ * This is the fallback/default function for ctx->Driver.GenerateMipmap().
+ * Generate a complete set of mipmaps from texObj's BaseLevel image.
  * Stop at texObj's MaxLevel or when we get to the 1x1 texture.
+ * For cube maps, target will be one of
+ * GL_TEXTURE_CUBE_MAP_POSITIVE/NEGATIVE_X/Y/Z; never GL_TEXTURE_CUBE_MAP.
  */
 void
 _mesa_generate_mipmap(GLcontext *ctx, GLenum target,
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 3b0ebcb7ae..6b64bf8139 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -40,6 +40,7 @@
 #include "main/mfeatures.h"
 #include "glapi/glapi.h"
 #include "math/m_matrix.h"	/* GLmatrix */
+#include "main/simple_list.h"	/* struct simple_node */
 
 
 /**
@@ -83,6 +84,7 @@
 /*@{*/
 struct _mesa_HashTable;
 struct gl_attrib_node;
+struct gl_meta_state;
 struct gl_pixelstore_attrib;
 struct gl_program_cache;
 struct gl_texture_format;
@@ -227,7 +229,9 @@ typedef enum
    FRAG_ATTRIB_TEX5 = 9,
    FRAG_ATTRIB_TEX6 = 10,
    FRAG_ATTRIB_TEX7 = 11,
-   FRAG_ATTRIB_VAR0 = 12,  /**< shader varying */
+   FRAG_ATTRIB_FACE = 12,  /**< front/back face */
+   FRAG_ATTRIB_PNTC = 13,  /**< sprite/point coord */
+   FRAG_ATTRIB_VAR0 = 14,  /**< shader varying */
    FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING)
 } gl_frag_attrib;
 
@@ -239,6 +243,8 @@ typedef enum
 #define FRAG_BIT_COL0  (1 << FRAG_ATTRIB_COL0)
 #define FRAG_BIT_COL1  (1 << FRAG_ATTRIB_COL1)
 #define FRAG_BIT_FOGC  (1 << FRAG_ATTRIB_FOGC)
+#define FRAG_BIT_FACE  (1 << FRAG_ATTRIB_FACE)
+#define FRAG_BIT_PNTC  (1 << FRAG_ATTRIB_PNTC)
 #define FRAG_BIT_TEX0  (1 << FRAG_ATTRIB_TEX0)
 #define FRAG_BIT_TEX1  (1 << FRAG_ATTRIB_TEX1)
 #define FRAG_BIT_TEX2  (1 << FRAG_ATTRIB_TEX2)
@@ -770,6 +776,7 @@ struct gl_light_attrib
 
    GLboolean Enabled;			/**< Lighting enabled flag */
    GLenum ShadeModel;			/**< GL_FLAT or GL_SMOOTH */
+   GLenum ProvokingVertex;              /**< GL_EXT_provoking_vertex */
    GLenum ColorMaterialFace;		/**< GL_FRONT, BACK or FRONT_AND_BACK */
    GLenum ColorMaterialMode;		/**< GL_AMBIENT, GL_DIFFUSE, etc */
    GLbitfield ColorMaterialBitmask;	/**< bitmask formed from Face and Mode */
@@ -1028,7 +1035,7 @@ struct gl_stencil_attrib
 
 /**
  * An index for each type of texture object.  These correspond to the GL
- * target target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
+ * texture target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
  * Note: the order is from highest priority to lowest priority.
  */
 typedef enum
@@ -1439,6 +1446,9 @@ struct gl_texture_attrib
 
    struct gl_texture_object *ProxyTex[NUM_TEXTURE_TARGETS];
 
+   /** GL_ARB_seamless_cubemap */
+   GLboolean CubeMapSeamless;
+
    /** GL_EXT_shared_texture_palette */
    GLboolean SharedPalette;
    struct gl_color_table Palette;
@@ -1498,14 +1508,17 @@ struct gl_buffer_object
 {
    GLint RefCount;
    GLuint Name;
-   GLenum Usage;
-   GLenum Access;
-   GLvoid *Pointer;          /**< Only valid while buffer is mapped */
-   GLintptr Offset;          /**< mapped offset */
-   GLsizeiptr Length;        /**< mapped length */
-   GLsizeiptrARB Size;       /**< Size of storage in bytes */
-   GLubyte *Data;            /**< Location of storage either in RAM or VRAM. */
-   GLboolean OnCard;         /**< Is buffer in VRAM? (hardware drivers) */
+   GLenum Usage;        /**< GL_STREAM_DRAW_ARB, GL_STREAM_READ_ARB, etc. */
+   GLsizeiptrARB Size;  /**< Size of buffer storage in bytes */
+   GLubyte *Data;       /**< Location of storage either in RAM or VRAM. */
+   /** Fields describing a mapped buffer */
+   /*@{*/
+   GLbitfield AccessFlags; /**< Mask of GL_MAP_x_BIT flags */
+   GLvoid *Pointer;     /**< User-space address of mapping */
+   GLintptr Offset;     /**< Mapped offset */
+   GLsizeiptr Length;   /**< Mapped length */
+   /*@}*/
+   GLboolean Written;   /**< Ever written to? (for debugging) */
 };
 
 
@@ -1541,10 +1554,10 @@ struct gl_client_array
    const GLubyte *Ptr;          /**< Points to array data */
    GLboolean Enabled;		/**< Enabled flag is a boolean */
    GLboolean Normalized;        /**< GL_ARB_vertex_program */
+   GLuint _ElementSize;         /**< size of each element in bytes */
 
-   /**< GL_ARB_vertex_buffer_object */
-   struct gl_buffer_object *BufferObj;
-   GLuint _MaxElement;
+   struct gl_buffer_object *BufferObj;/**< GL_ARB_vertex_buffer_object */
+   GLuint _MaxElement;          /**< max element index into array buffer + 1 */
 };
 
 
@@ -1559,10 +1572,12 @@ struct gl_array_object
 
    GLint RefCount;
    _glthread_Mutex Mutex;
+   GLboolean VBOonly;  /**< require all arrays to live in VBOs? */
 
    /** Conventional vertex arrays */
    /*@{*/
    struct gl_client_array Vertex;
+   struct gl_client_array Weight;
    struct gl_client_array Normal;
    struct gl_client_array Color;
    struct gl_client_array SecondaryColor;
@@ -1573,11 +1588,22 @@ struct gl_array_object
    struct gl_client_array PointSize;
    /*@}*/
 
-   /** Generic arrays for vertex programs/shaders */
-   struct gl_client_array VertexAttrib[VERT_ATTRIB_MAX];
+   /**
+    * Generic arrays for vertex programs/shaders.
+    * For NV vertex programs, these attributes alias and take priority
+    * over the conventional attribs above.  For ARB vertex programs and
+    * GLSL vertex shaders, these attributes are separate.
+    */
+   struct gl_client_array VertexAttrib[MAX_VERTEX_GENERIC_ATTRIBS];
 
    /** Mask of _NEW_ARRAY_* values indicating which arrays are enabled */
    GLbitfield _Enabled;
+
+   /**
+    * Min of all enabled arrays' _MaxElement.  When arrays reside inside VBOs
+    * we can determine the max legal (in bounds) glDrawElements array index.
+    */
+   GLuint _MaxElement;
 };
 
 
@@ -1586,9 +1612,15 @@ struct gl_array_object
  */
 struct gl_array_attrib
 {
+   /** Currently bound array object. See _mesa_BindVertexArrayAPPLE() */
    struct gl_array_object *ArrayObj;
+
+   /** The default vertex array object */
    struct gl_array_object *DefaultArrayObj;
 
+   /** Array objects (GL_ARB/APPLE_vertex_array_object) */
+   struct _mesa_HashTable *Objects;
+
    GLint ActiveTexture;		/**< Client Active Texture */
    GLuint LockFirst;            /**< GL_EXT_compiled_vertex_array */
    GLuint LockCount;            /**< GL_EXT_compiled_vertex_array */
@@ -1596,11 +1628,9 @@ struct gl_array_attrib
    GLbitfield NewState;		/**< mask of _NEW_ARRAY_* values */
 
 #if FEATURE_ARB_vertex_buffer_object
-   struct gl_buffer_object *NullBufferObj;
    struct gl_buffer_object *ArrayBufferObj;
    struct gl_buffer_object *ElementArrayBufferObj;
 #endif
-   GLuint _MaxElement;          /* Min of all enabled array's maxes */
 };
 
 
@@ -1770,7 +1800,7 @@ struct gl_program
    /** Map from sampler unit to texture unit (set by glUniform1i()) */
    GLubyte SamplerUnits[MAX_SAMPLERS];
    /** Which texture target is being sampled (TEXTURE_1D/2D/3D/etc_INDEX) */
-   GLubyte SamplerTargets[MAX_SAMPLERS];
+   gl_texture_index SamplerTargets[MAX_SAMPLERS];
 
    /** Logical counts */
    /*@{*/
@@ -1813,9 +1843,6 @@ struct gl_fragment_program
    struct gl_program Base;   /**< base class */
    GLenum FogOption;
    GLboolean UsesKill;          /**< shader uses KIL instruction */
-   GLboolean UsesPointCoord;    /**< shader uses gl_PointCoord */
-   GLboolean UsesFrontFacing;   /**< shader used gl_FrontFacing */
-   GLboolean UsesFogFragCoord;  /**< shader used gl_FogFragCoord */
 };
 
 
@@ -1960,6 +1987,21 @@ struct gl_query_state
 };
 
 
+/** Sync object state */
+struct gl_sync_object {
+   struct simple_node link;
+   GLenum Type;               /**< GL_SYNC_FENCE */
+   GLuint Name;               /**< Fence name */
+   GLint RefCount;            /**< Reference count */
+   GLboolean DeletePending;   /**< Object was deleted while there were still
+			       * live references (e.g., sync not yet finished)
+			       */
+   GLenum SyncCondition;
+   GLbitfield Flags;          /**< Flags passed to glFenceSync */
+   GLuint StatusFlag:1;       /**< Has the sync object been signaled? */
+};
+
+
 /** Set by #pragma directives */
 struct gl_sl_pragmas
 {
@@ -1983,6 +2025,7 @@ struct gl_shader
    GLboolean Main;  /**< shader defines main() */
    GLboolean UnresolvedRefs;
    const GLchar *Source;  /**< Source code string */
+   GLuint SourceChecksum;       /**< for debug/logging purposes */
    struct gl_program *Program;  /**< Post-compile assembly code */
    GLchar *InfoLog;
    struct gl_sl_pragmas Pragmas;
@@ -2013,6 +2056,7 @@ struct gl_shader_program
    struct gl_program_parameter_list *Varying;
    GLboolean LinkStatus;   /**< GL_LINK_STATUS */
    GLboolean Validated;
+   GLboolean _Used;        /**< Ever used for drawing? */
    GLchar *InfoLog;
 };   
 
@@ -2068,6 +2112,8 @@ struct gl_shared_state
    GLuint TextureStateStamp;	        /**< state notification for shared tex */
    /*@}*/
 
+   /** Default buffer object for vertex arrays that aren't in VBOs */
+   struct gl_buffer_object *NullBufferObj;
 
    /**
     * \name Vertex/fragment programs
@@ -2101,8 +2147,9 @@ struct gl_shared_state
    struct _mesa_HashTable *FrameBuffers;
 #endif
 
-   /** Objects associated with the GL_APPLE_vertex_array_object extension. */
-   struct _mesa_HashTable *ArrayObjects;
+#if FEATURE_ARB_sync
+   struct simple_node SyncObjects;
+#endif
 
    void *DriverData;  /**< Device driver shared state */
 };
@@ -2407,6 +2454,15 @@ struct gl_constants
    GLuint MaxVarying;  /**< Number of float[4] varying parameters */
 
    GLbitfield SupportedBumpUnits; /**> units supporting GL_ATI_envmap_bumpmap as targets */
+
+   /**
+    * Maximum amount of time, measured in nanseconds, that the server can wait.
+    */
+   GLuint64 MaxServerWaitTimeout;
+
+
+   /**< GL_EXT_provoking_vertex */
+   GLboolean QuadsFollowProvokingVertexConvention;
 };
 
 
@@ -2417,6 +2473,7 @@ struct gl_constants
 struct gl_extensions
 {
    GLboolean dummy;  /* don't remove this! */
+   GLboolean ARB_copy_buffer;
    GLboolean ARB_depth_texture;
    GLboolean ARB_draw_buffers;
    GLboolean ARB_fragment_program;
@@ -2425,15 +2482,18 @@ struct gl_extensions
    GLboolean ARB_framebuffer_object;
    GLboolean ARB_half_float_pixel;
    GLboolean ARB_imaging;
+   GLboolean ARB_map_buffer_range;
    GLboolean ARB_multisample;
    GLboolean ARB_multitexture;
    GLboolean ARB_occlusion_query;
    GLboolean ARB_point_sprite;
+   GLboolean ARB_seamless_cube_map;
    GLboolean ARB_shader_objects;
    GLboolean ARB_shading_language_100;
    GLboolean ARB_shading_language_120;
    GLboolean ARB_shadow;
    GLboolean ARB_shadow_ambient; /* or GL_ARB_shadow_ambient */
+   GLboolean ARB_sync;
    GLboolean ARB_texture_border_clamp;
    GLboolean ARB_texture_compression;
    GLboolean ARB_texture_cube_map;
@@ -2444,6 +2504,7 @@ struct gl_extensions
    GLboolean ARB_texture_mirrored_repeat;
    GLboolean ARB_texture_non_power_of_two;
    GLboolean ARB_transpose_matrix;
+   GLboolean ARB_vertex_array_object;
    GLboolean ARB_vertex_buffer_object;
    GLboolean ARB_vertex_program;
    GLboolean ARB_vertex_shader;
@@ -2475,6 +2536,7 @@ struct gl_extensions
    GLboolean EXT_pixel_buffer_object;
    GLboolean EXT_point_parameters;
    GLboolean EXT_polygon_offset;
+   GLboolean EXT_provoking_vertex;
    GLboolean EXT_rescale_normal;
    GLboolean EXT_shadow_funcs;
    GLboolean EXT_secondary_color;
@@ -2623,6 +2685,7 @@ struct gl_matrix_stack
 #define _NEW_PROGRAM            0x8000000  /**< __GLcontextRec::VertexProgram */
 #define _NEW_CURRENT_ATTRIB     0x10000000  /**< __GLcontextRec::Current */
 #define _NEW_PROGRAM_CONSTANTS  0x20000000
+#define _NEW_BUFFER_OBJECT      0x40000000
 #define _NEW_ALL ~0
 /*@}*/
 
@@ -2948,8 +3011,13 @@ struct __GLcontextRec
    struct gl_shader_state Shader; /**< GLSL shader object state */
 
    struct gl_query_state Query;  /**< occlusion, timer queries */
+
+   struct gl_buffer_object *CopyReadBuffer; /**< GL_ARB_copy_buffer */
+   struct gl_buffer_object *CopyWriteBuffer; /**< GL_ARB_copy_buffer */
    /*@}*/
 
+   struct gl_meta_state *Meta;  /**< for "meta" operations */
+
 #if FEATURE_EXT_framebuffer_object
    struct gl_renderbuffer *CurrentRenderbuffer;
 #endif
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 57ae9c721a..fcef6dfd42 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -136,6 +136,33 @@ store_pixelmap(GLcontext *ctx, GLenum map, GLsizei mapsize,
 }
 
 
+/**
+ * Convenience wrapper for _mesa_validate_pbo_access() for gl[Get]PixelMap().
+ */
+static GLboolean
+validate_pbo_access(GLcontext *ctx, struct gl_pixelstore_attrib *pack,
+                    GLsizei mapsize, GLenum format, GLenum type,
+                    const GLvoid *ptr)
+{
+   GLboolean ok;
+
+   /* Note, need to use DefaultPacking and Unpack's buffer object */
+   ctx->DefaultPacking.BufferObj = pack->BufferObj;
+
+   ok = _mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
+                                  format, type, ptr);
+
+   /* restore */
+   ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
+
+   if (!ok) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glPixelMap(invalid PBO access)");
+   }
+   return ok;
+}
+
+
 void GLAPIENTRY
 _mesa_PixelMapfv( GLenum map, GLsizei mapsize, const GLfloat *values )
 {
@@ -158,40 +185,23 @@ _mesa_PixelMapfv( GLenum map, GLsizei mapsize, const GLfloat *values )
 
    FLUSH_VERTICES(ctx, _NEW_PIXEL);
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack pixelmap from PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Unpack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Unpack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_FLOAT, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glPixelMapfv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Unpack, mapsize,
+                            GL_INTENSITY, GL_FLOAT, values)) {
+      return;
+   }
+
+   values = (const GLfloat *) _mesa_map_pbo_source(ctx, &ctx->Unpack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Unpack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glPixelMapfv(PBO is mapped)");
-         return;
       }
-      values = (const GLfloat *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
    store_pixelmap(ctx, map, mapsize, values);
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 }
 
 
@@ -217,31 +227,17 @@ _mesa_PixelMapuiv(GLenum map, GLsizei mapsize, const GLuint *values )
 
    FLUSH_VERTICES(ctx, _NEW_PIXEL);
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack pixelmap from PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Unpack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Unpack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_UNSIGNED_INT, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glPixelMapuiv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Unpack, mapsize,
+                            GL_INTENSITY, GL_UNSIGNED_INT, values)) {
+      return;
+   }
+
+   values = (const GLuint *) _mesa_map_pbo_source(ctx, &ctx->Unpack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Unpack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glPixelMapuiv(PBO is mapped)");
-         return;
       }
-      values = (const GLuint *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
@@ -259,10 +255,7 @@ _mesa_PixelMapuiv(GLenum map, GLsizei mapsize, const GLuint *values )
       }
    }
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 
    store_pixelmap(ctx, map, mapsize, fvalues);
 }
@@ -290,32 +283,17 @@ _mesa_PixelMapusv(GLenum map, GLsizei mapsize, const GLushort *values )
 
    FLUSH_VERTICES(ctx, _NEW_PIXEL);
 
-   if (ctx->Unpack.BufferObj->Name) {
-      /* unpack pixelmap from PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Unpack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Unpack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_UNSIGNED_SHORT,
-                                     values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glPixelMapusv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Unpack, mapsize,
+                            GL_INTENSITY, GL_UNSIGNED_SHORT, values)) {
+      return;
+   }
+
+   values = (const GLushort *) _mesa_map_pbo_source(ctx, &ctx->Unpack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Unpack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glPixelMapusv(PBO is mapped)");
-         return;
       }
-      values = (const GLushort *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
@@ -333,10 +311,7 @@ _mesa_PixelMapusv(GLenum map, GLsizei mapsize, const GLushort *values )
       }
    }
 
-   if (ctx->Unpack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 
    store_pixelmap(ctx, map, mapsize, fvalues);
 }
@@ -359,31 +334,17 @@ _mesa_GetPixelMapfv( GLenum map, GLfloat *values )
 
    mapsize = pm->Size;
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack pixelmap into PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Pack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Pack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_FLOAT, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetPixelMapfv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Pack, mapsize,
+                            GL_INTENSITY, GL_FLOAT, values)) {
+      return;
+   }
+
+   values = (GLfloat *) _mesa_map_pbo_dest(ctx, &ctx->Pack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glGetPixelMapfv(PBO is mapped)");
-         return;
       }
-      values = (GLfloat *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
@@ -397,10 +358,7 @@ _mesa_GetPixelMapfv( GLenum map, GLfloat *values )
       MEMCPY(values, pm->Map, mapsize * sizeof(GLfloat));
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
@@ -420,31 +378,17 @@ _mesa_GetPixelMapuiv( GLenum map, GLuint *values )
    }
    mapsize = pm->Size;
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack pixelmap into PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Pack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Pack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_UNSIGNED_INT, values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetPixelMapuiv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Pack, mapsize,
+                            GL_INTENSITY, GL_UNSIGNED_INT, values)) {
+      return;
+   }
+
+   values = (GLuint *) _mesa_map_pbo_dest(ctx, &ctx->Pack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glGetPixelMapuiv(PBO is mapped)");
-         return;
       }
-      values = (GLuint *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
@@ -458,10 +402,7 @@ _mesa_GetPixelMapuiv( GLenum map, GLuint *values )
       }
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
@@ -481,32 +422,17 @@ _mesa_GetPixelMapusv( GLenum map, GLushort *values )
    }
    mapsize = pm ? pm->Size : 0;
 
-   if (ctx->Pack.BufferObj->Name) {
-      /* pack pixelmap into PBO */
-      GLubyte *buf;
-      /* Note, need to use DefaultPacking and Pack's buffer object */
-      ctx->DefaultPacking.BufferObj = ctx->Pack.BufferObj;
-      if (!_mesa_validate_pbo_access(1, &ctx->DefaultPacking, mapsize, 1, 1,
-                                     GL_INTENSITY, GL_UNSIGNED_SHORT,
-                                     values)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetPixelMapusv(invalid PBO access)");
-         return;
-      }
-      /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
+   if (!validate_pbo_access(ctx, &ctx->Pack, mapsize,
+                            GL_INTENSITY, GL_UNSIGNED_SHORT, values)) {
+      return;
+   }
+
+   values = (GLushort *) _mesa_map_pbo_dest(ctx, &ctx->Pack, values);
+   if (!values) {
+      if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glGetPixelMapusv(PBO is mapped)");
-         return;
       }
-      values = (GLushort *) ADD_POINTERS(buf, values);
-   }
-   else if (!values) {
       return;
    }
 
@@ -528,10 +454,7 @@ _mesa_GetPixelMapusv( GLenum map, GLushort *values )
       }
    }
 
-   if (ctx->Pack.BufferObj->Name) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c
index ff1a6344cc..6a641f83f2 100644
--- a/src/mesa/main/pixelstore.c
+++ b/src/mesa/main/pixelstore.c
@@ -245,7 +245,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Pack.ClientStorage = GL_FALSE;
    ctx->Pack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Pack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Pack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
    ctx->Unpack.Alignment = 4;
    ctx->Unpack.RowLength = 0;
@@ -258,7 +259,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Unpack.ClientStorage = GL_FALSE;
    ctx->Unpack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Unpack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 
    /*
@@ -278,6 +280,7 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->DefaultPacking.ClientStorage = GL_FALSE;
    ctx->DefaultPacking.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->DefaultPacking.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
diff --git a/src/mesa/main/polygon.c b/src/mesa/main/polygon.c
index 564250b881..376a87a396 100644
--- a/src/mesa/main/polygon.c
+++ b/src/mesa/main/polygon.c
@@ -193,32 +193,16 @@ _mesa_PolygonMode( GLenum face, GLenum mode )
 void
 _mesa_polygon_stipple(GLcontext *ctx, const GLubyte *pattern)
 {
-   if (ctx->Unpack.BufferObj->Name) {
-      /* Get/unpack the stipple pattern from a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, &ctx->Unpack, 32, 32, 1,
-                                     GL_COLOR_INDEX, GL_BITMAP, pattern)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glPolygonStipple(bad PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              ctx->Unpack.BufferObj);
-      if (!buf) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glPolygonStipple(PBO mapped)");
-         return;
-      }
-      buf = ADD_POINTERS(buf, pattern);
-      _mesa_unpack_polygon_stipple(buf, ctx->PolygonStipple, &ctx->Unpack);
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              ctx->Unpack.BufferObj);
-   }
-   else {
-      /* Get/unpack the stipple pattern from user memory */
-      _mesa_unpack_polygon_stipple(pattern, ctx->PolygonStipple, &ctx->Unpack);
-   }
+   pattern = _mesa_map_validate_pbo_source(ctx, 2,
+                                           &ctx->Unpack, 32, 32, 1,
+                                           GL_COLOR_INDEX, GL_BITMAP, pattern,
+                                           "glPolygonStipple");
+   if (!pattern)
+      return;
+
+   _mesa_unpack_polygon_stipple(pattern, ctx->PolygonStipple, &ctx->Unpack);
+
+   _mesa_unmap_pbo_source(ctx, &ctx->Unpack);
 }
 
 
@@ -255,35 +239,16 @@ _mesa_GetPolygonStipple( GLubyte *dest )
    if (MESA_VERBOSE&VERBOSE_API)
       _mesa_debug(ctx, "glGetPolygonStipple\n");
 
-   /* XXX someday we may put this code into a separate function and call
-    * it with ctx->Driver.GetPolygonStipple().
-    */
-   if (ctx->Pack.BufferObj->Name) {
-      /* Put/pack the stipple pattern into a PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, &ctx->Pack, 32, 32, 1,
-                                     GL_COLOR_INDEX, GL_BITMAP, dest)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetPolygonStipple(bad PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              ctx->Pack.BufferObj);
-      if (!buf) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glGetPolygonStipple(PBO mapped)");
-         return;
-      }
-      buf = ADD_POINTERS(buf, dest);
-      _mesa_pack_polygon_stipple(ctx->PolygonStipple, buf, &ctx->Pack);
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              ctx->Pack.BufferObj);
-   }
-   else {
-      /* Put/pack the stipple pattern into user memory */
-      _mesa_pack_polygon_stipple(ctx->PolygonStipple, dest, &ctx->Pack);
-   }
+   dest = _mesa_map_validate_pbo_dest(ctx, 2,
+                                      &ctx->Pack, 32, 32, 1,
+                                      GL_COLOR_INDEX, GL_BITMAP, dest,
+                                      "glGetPolygonStipple");
+   if (!dest)
+      return;
+
+   _mesa_pack_polygon_stipple(ctx->PolygonStipple, dest, &ctx->Pack);
+
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 }
 
 
diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c
index c25b31af02..a73c6e0508 100644
--- a/src/mesa/main/queryobj.c
+++ b/src/mesa/main/queryobj.c
@@ -38,7 +38,7 @@
  * \param id - the new object's ID
  * \return pointer to new query_object object or NULL if out of memory.
  */
-struct gl_query_object *
+static struct gl_query_object *
 _mesa_new_query_object(GLcontext *ctx, GLuint id)
 {
    struct gl_query_object *q = MALLOC_STRUCT(gl_query_object);
@@ -57,7 +57,7 @@ _mesa_new_query_object(GLcontext *ctx, GLuint id)
  * Begin a query.  Software driver fallback.
  * Called via ctx->Driver.BeginQuery().
  */
-void
+static void
 _mesa_begin_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* no-op */
@@ -68,7 +68,7 @@ _mesa_begin_query(GLcontext *ctx, struct gl_query_object *q)
  * End a query.  Software driver fallback.
  * Called via ctx->Driver.EndQuery().
  */
-void
+static void
 _mesa_end_query(GLcontext *ctx, struct gl_query_object *q)
 {
    q->Ready = GL_TRUE;
@@ -79,7 +79,7 @@ _mesa_end_query(GLcontext *ctx, struct gl_query_object *q)
  * Wait for query to complete.  Software driver fallback.
  * Called via ctx->Driver.WaitQuery().
  */
-void
+static void
 _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* For software drivers, _mesa_end_query() should have completed the query.
@@ -94,7 +94,7 @@ _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q)
  * Check if a query results are ready.  Software driver fallback.
  * Called via ctx->Driver.CheckQuery().
  */
-void
+static void
 _mesa_check_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* No-op for sw rendering.
@@ -107,7 +107,7 @@ _mesa_check_query(GLcontext *ctx, struct gl_query_object *q)
  * Delete a query object.  Called via ctx->Driver.DeleteQuery().
  * Not removed from hash table here.
  */
-void
+static void
 _mesa_delete_query(GLcontext *ctx, struct gl_query_object *q)
 {
    _mesa_free(q);
@@ -123,6 +123,18 @@ lookup_query_object(GLcontext *ctx, GLuint id)
 
 
 
+void
+_mesa_init_query_object_functions(struct dd_function_table *driver)
+{
+   driver->NewQueryObject = _mesa_new_query_object;
+   driver->DeleteQuery = _mesa_delete_query;
+   driver->BeginQuery = _mesa_begin_query;
+   driver->EndQuery = _mesa_end_query;
+   driver->WaitQuery = _mesa_wait_query;
+   driver->CheckQuery = _mesa_check_query;
+}
+
+
 void GLAPIENTRY
 _mesa_GenQueriesARB(GLsizei n, GLuint *ids)
 {
diff --git a/src/mesa/main/queryobj.h b/src/mesa/main/queryobj.h
index bc02b65b54..ee775ef959 100644
--- a/src/mesa/main/queryobj.h
+++ b/src/mesa/main/queryobj.h
@@ -27,9 +27,6 @@
 #define OCCLUDE_H
 
 
-extern struct gl_query_object *
-_mesa_new_query_object(GLcontext *ctx, GLuint id);
-
 extern void
 _mesa_init_query(GLcontext *ctx);
 
@@ -37,19 +34,7 @@ extern void
 _mesa_free_query_data(GLcontext *ctx);
 
 extern void
-_mesa_delete_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_begin_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_end_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_wait_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_check_query(GLcontext *ctx, struct gl_query_object *q);
+_mesa_init_query_object_functions(struct dd_function_table *driver);
 
 
 extern void GLAPIENTRY
diff --git a/src/mesa/main/readpix.c b/src/mesa/main/readpix.c
index 2326776ecb..feea1d375f 100644
--- a/src/mesa/main/readpix.c
+++ b/src/mesa/main/readpix.c
@@ -44,6 +44,10 @@ _mesa_error_check_format_type(GLcontext *ctx, GLenum format, GLenum type,
                               GLboolean drawing)
 {
    const char *readDraw = drawing ? "Draw" : "Read";
+   const GLboolean reading = !drawing;
+
+   /* state validation should have already been done */
+   ASSERT(ctx->NewState == 0x0);
 
    if (ctx->Extensions.EXT_packed_depth_stencil
        && type == GL_UNSIGNED_INT_24_8_EXT
@@ -73,32 +77,45 @@ _mesa_error_check_format_type(GLcontext *ctx, GLenum format, GLenum type,
    case GL_RGBA:
    case GL_BGRA:
    case GL_ABGR_EXT:
-      if (drawing && !ctx->Visual.rgbMode) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
+      if (drawing) {
+         if (!ctx->DrawBuffer->Visual.rgbMode) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
                    "glDrawPixels(drawing RGB pixels into color index buffer)");
-         return GL_TRUE;
+            return GL_TRUE;
+         }
       }
-      if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glReadPixels(no color buffer)");
-         return GL_TRUE;
+      else {
+         /* reading */
+         if (!_mesa_source_buffer_exists(ctx, GL_COLOR)) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glReadPixels(no color buffer)");
+            return GL_TRUE;
+         }
       }
       break;
    case GL_COLOR_INDEX:
-      if (!drawing && ctx->Visual.rgbMode) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                    "glReadPixels(reading color index format from RGB buffer)");
-         return GL_TRUE;
+      if (drawing) {
+         if (ctx->DrawBuffer->Visual.rgbMode &&
+             (ctx->PixelMaps.ItoR.Size == 0 ||
+              ctx->PixelMaps.ItoG.Size == 0 ||
+              ctx->PixelMaps.ItoB.Size == 0)) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                   "glDrawPixels(drawing color index pixels into RGB buffer)");
+            return GL_TRUE;
+         }
       }
-      if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glReadPixels(no color buffer)");
-         return GL_TRUE;
+      else {
+         /* reading */
+         if (!_mesa_source_buffer_exists(ctx, GL_COLOR)) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glReadPixels(no color buffer)");
+            return GL_TRUE;
+         }
       }
       break;
    case GL_STENCIL_INDEX:
       if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) ||
-          (!drawing && !_mesa_source_buffer_exists(ctx, format))) {
+          (reading && !_mesa_source_buffer_exists(ctx, format))) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "gl%sPixels(no stencil buffer)", readDraw);
          return GL_TRUE;
@@ -118,7 +135,7 @@ _mesa_error_check_format_type(GLcontext *ctx, GLenum format, GLenum type,
          return GL_TRUE;
       }
       if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) ||
-          (!drawing && !_mesa_source_buffer_exists(ctx, format))) {
+          (reading && !_mesa_source_buffer_exists(ctx, format))) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "gl%sPixels(no depth or stencil buffer)", readDraw);
          return GL_TRUE;
@@ -173,7 +190,7 @@ _mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height,
    if (width == 0 || height == 0)
       return; /* nothing to do */
 
-   if (ctx->Pack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
       if (!_mesa_validate_pbo_access(2, &ctx->Pack, width, height, 1,
                                      format, type, pixels)) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
@@ -181,7 +198,7 @@ _mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height,
          return;
       }
 
-      if (ctx->Pack.BufferObj->Pointer) {
+      if (_mesa_bufferobj_mapped(ctx->Pack.BufferObj)) {
          /* buffer is mapped - that's an error */
          _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)");
          return;
diff --git a/src/mesa/main/shared.c b/src/mesa/main/shared.c
index ae8dd2a836..643ad3354e 100644
--- a/src/mesa/main/shared.c
+++ b/src/mesa/main/shared.c
@@ -33,6 +33,7 @@
 #include "mtypes.h"
 #include "hash.h"
 #include "arrayobj.h"
+#include "bufferobj.h"
 #include "shared.h"
 #include "shader/program.h"
 #include "shader/shader_api.h"
@@ -42,7 +43,9 @@
 #if FEATURE_ATI_fragment_shader
 #include "shader/atifragshader.h"
 #endif
-
+#if FEATURE_ARB_sync
+#include "syncobj.h"
+#endif
 
 /**
  * Allocate and initialize a shared context state structure.
@@ -92,7 +95,12 @@ _mesa_alloc_shared_state(GLcontext *ctx)
    shared->BufferObjects = _mesa_NewHashTable();
 #endif
 
-   shared->ArrayObjects = _mesa_NewHashTable();
+   /* Allocate the default buffer object and set refcount so high that
+    * it never gets deleted.
+    * XXX with recent/improved refcounting this may not longer be needed.
+    */
+   shared->NullBufferObj = ctx->Driver.NewBufferObject(ctx, 0, 0);
+   shared->NullBufferObj->RefCount = 1000 * 1000 * 1000;
 
    /* Create default texture objects */
    for (i = 0; i < NUM_TEXTURE_TARGETS; i++) {
@@ -121,6 +129,10 @@ _mesa_alloc_shared_state(GLcontext *ctx)
    shared->RenderBuffers = _mesa_NewHashTable();
 #endif
 
+#if FEATURE_ARB_sync
+   make_empty_list(& shared->SyncObjects);
+#endif
+
    return shared;
 }
 
@@ -190,7 +202,7 @@ delete_bufferobj_cb(GLuint id, void *data, void *userData)
 {
    struct gl_buffer_object *bufObj = (struct gl_buffer_object *) data;
    GLcontext *ctx = (GLcontext *) userData;
-   if (bufObj->Pointer) {
+   if (_mesa_bufferobj_mapped(bufObj)) {
       ctx->Driver.UnmapBuffer(ctx, 0, bufObj);
       bufObj->Pointer = NULL;
    }
@@ -199,18 +211,6 @@ delete_bufferobj_cb(GLuint id, void *data, void *userData)
 
 
 /**
- * Callback for deleting an array object.  Called by _mesa_HashDeleteAll().
- */
-static void
-delete_arrayobj_cb(GLuint id, void *data, void *userData)
-{
-   struct gl_array_object *arrayObj = (struct gl_array_object *) data;
-   GLcontext *ctx = (GLcontext *) userData;
-   _mesa_delete_array_object(ctx, arrayObj);
-}
-
-
-/**
  * Callback for freeing shader program data. Call it before delete_shader_cb
  * to avoid memory access error.
  */
@@ -312,9 +312,6 @@ _mesa_free_shared_state(GLcontext *ctx, struct gl_shared_state *shared)
    _mesa_HashDeleteAll(shared->Programs, delete_program_cb, ctx);
    _mesa_DeleteHashTable(shared->Programs);
 
-   _mesa_HashDeleteAll(shared->ArrayObjects, delete_arrayobj_cb, ctx);
-   _mesa_DeleteHashTable(shared->ArrayObjects);
-
 #if FEATURE_ARB_vertex_program
    _mesa_reference_vertprog(ctx, &shared->DefaultVertexProgram, NULL);
 #endif
@@ -341,6 +338,21 @@ _mesa_free_shared_state(GLcontext *ctx, struct gl_shared_state *shared)
    _mesa_DeleteHashTable(shared->RenderBuffers);
 #endif
 
+#if FEATURE_ARB_vertex_buffer_object
+   ctx->Driver.DeleteBuffer(ctx, shared->NullBufferObj);
+#endif
+
+#if FEATURE_ARB_sync
+   {
+      struct simple_node *node;
+      struct simple_node *temp;
+
+      foreach_s(node, temp, & shared->SyncObjects) {
+	 _mesa_unref_sync_object(ctx, (struct gl_sync_object *) node);
+      }
+   }
+#endif
+
    /*
     * Free texture objects (after FBOs since some textures might have
     * been bound to FBOs).
diff --git a/src/mesa/main/simple_list.h b/src/mesa/main/simple_list.h
index 63475f6f74..ff7f888238 100644
--- a/src/mesa/main/simple_list.h
+++ b/src/mesa/main/simple_list.h
@@ -37,6 +37,11 @@
 #ifndef _SIMPLE_LIST_H
 #define _SIMPLE_LIST_H
 
+struct simple_node {
+   struct simple_node *next;
+   struct simple_node *prev;
+};
+
 /**
  * Remove an element from list.
  *
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 51726461b6..140a998df2 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -64,114 +64,159 @@ update_separate_specular(GLcontext *ctx)
 
 
 /**
- * Update state dependent on vertex arrays.
+ * Compute the index of the last array element that can be safely accessed
+ * in a vertex array.  We can really only do this when the array lives in
+ * a VBO.
+ * The array->_MaxElement field will be updated.
+ * Later in glDrawArrays/Elements/etc we can do some bounds checking.
+ */
+static void
+compute_max_element(struct gl_client_array *array)
+{
+   assert(array->Enabled);
+   if (array->BufferObj->Name) {
+      GLsizeiptrARB offset = (GLsizeiptrARB) array->Ptr;
+      GLsizeiptrARB obj_size = (GLsizeiptrARB) array->BufferObj->Size;
+
+      if (offset < obj_size) {
+	 array->_MaxElement = (obj_size - offset +
+			       array->StrideB -
+			       array->_ElementSize) / array->StrideB;
+      } else {
+	 array->_MaxElement = 0;
+      }
+      /* Compute the max element we can access in the VBO without going
+       * out of bounds.
+       */
+      array->_MaxElement = ((GLsizeiptrARB) array->BufferObj->Size
+                            - (GLsizeiptrARB) array->Ptr + array->StrideB
+                            - array->_ElementSize) / array->StrideB;
+   }
+   else {
+      /* user-space array, no idea how big it is */
+      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
+   }
+}
+
+
+/**
+ * Helper for update_arrays().
+ * \return  min(current min, array->_MaxElement).
+ */
+static GLuint
+update_min(GLuint min, struct gl_client_array *array)
+{
+   compute_max_element(array);
+   return MIN2(min, array->_MaxElement);
+}
+
+
+/**
+ * Update ctx->Array._MaxElement (the max legal index into all enabled arrays).
+ * Need to do this upon new array state or new buffer object state.
  */
 static void
 update_arrays( GLcontext *ctx )
 {
-   GLuint i, min;
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   GLuint i, min = ~0;
 
    /* find min of _MaxElement values for all enabled arrays */
 
    /* 0 */
    if (ctx->VertexProgram._Current
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
-      min = ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS]._MaxElement;
+       && arrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_POS]);
    }
-   else if (ctx->Array.ArrayObj->Vertex.Enabled) {
-      min = ctx->Array.ArrayObj->Vertex._MaxElement;
-   }
-   else {
-      /* can't draw anything without vertex positions! */
-      min = 0;
+   else if (arrayObj->Vertex.Enabled) {
+      min = update_min(min, &arrayObj->Vertex);
    }
 
    /* 1 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]);
    }
    /* no conventional vertex weight array */
 
    /* 2 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]);
    }
-   else if (ctx->Array.ArrayObj->Normal.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Normal._MaxElement);
+   else if (arrayObj->Normal.Enabled) {
+      min = update_min(min, &arrayObj->Normal);
    }
 
    /* 3 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]);
    }
-   else if (ctx->Array.ArrayObj->Color.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Color._MaxElement);
+   else if (arrayObj->Color.Enabled) {
+      min = update_min(min, &arrayObj->Color);
    }
 
    /* 4 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]);
    }
-   else if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->SecondaryColor._MaxElement);
+   else if (arrayObj->SecondaryColor.Enabled) {
+      min = update_min(min, &arrayObj->SecondaryColor);
    }
 
    /* 5 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_FOG]);
    }
-   else if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->FogCoord._MaxElement);
+   else if (arrayObj->FogCoord.Enabled) {
+      min = update_min(min, &arrayObj->FogCoord);
    }
 
    /* 6 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]);
    }
-   else if (ctx->Array.ArrayObj->Index.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Index._MaxElement);
+   else if (arrayObj->Index.Enabled) {
+      min = update_min(min, &arrayObj->Index);
    }
 
-
    /* 7 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]);
    }
 
    /* 8..15 */
    for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++) {
       if (ctx->VertexProgram._Enabled
-          && ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+          && arrayObj->VertexAttrib[i].Enabled) {
+         min = update_min(min, &arrayObj->VertexAttrib[i]);
       }
       else if (i - VERT_ATTRIB_TEX0 < ctx->Const.MaxTextureCoordUnits
-               && ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0]._MaxElement);
+               && arrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
+         min = update_min(min, &arrayObj->TexCoord[i - VERT_ATTRIB_TEX0]);
       }
    }
 
    /* 16..31 */
    if (ctx->VertexProgram._Current) {
-      for (i = 0; i < Elements(ctx->Array.ArrayObj->VertexAttrib); i++) {
-         if (ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-            min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+      for (i = 0; i < Elements(arrayObj->VertexAttrib); i++) {
+         if (arrayObj->VertexAttrib[i].Enabled) {
+            min = update_min(min, &arrayObj->VertexAttrib[i]);
          }
       }
    }
 
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->EdgeFlag._MaxElement);
+   if (arrayObj->EdgeFlag.Enabled) {
+      min = update_min(min, &arrayObj->EdgeFlag);
    }
 
    /* _MaxElement is one past the last legal array element */
-   ctx->Array._MaxElement = min;
+   arrayObj->_MaxElement = min;
 }
 
 
@@ -501,7 +546,7 @@ _mesa_update_state_locked( GLcontext *ctx )
 
    /* Determine which state flags effect vertex/fragment program state */
    if (ctx->FragmentProgram._MaintainTexEnvProgram) {
-      prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR |
+      prog_flags |= (_NEW_TEXTURE | _NEW_FOG |
 		     _NEW_ARRAY | _NEW_LIGHT | _NEW_POINT | _NEW_RENDERMODE |
 		     _NEW_PROGRAM);
    }
@@ -548,7 +593,7 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _DD_NEW_SEPARATE_SPECULAR)
       update_separate_specular( ctx );
 
-   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM))
+   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM | _NEW_BUFFER_OBJECT))
       update_arrays( ctx );
 
    if (new_state & (_NEW_BUFFERS | _NEW_VIEWPORT))
@@ -668,6 +713,6 @@ _mesa_set_vp_override(GLcontext *ctx, GLboolean flag)
       /* Set one of the bits which will trigger fragment program
        * regeneration:
        */
-      ctx->NewState |= _NEW_ARRAY; 
+      ctx->NewState |= _NEW_PROGRAM;
    }
 }
diff --git a/src/mesa/main/syncobj.c b/src/mesa/main/syncobj.c
new file mode 100644
index 0000000000..64f923ff91
--- /dev/null
+++ b/src/mesa/main/syncobj.c
@@ -0,0 +1,409 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file syncobj.c
+ * Sync object management.
+ *
+ * Unlike textures and other objects that are shared between contexts, sync
+ * objects are not bound to the context.  As a result, the reference counting
+ * and delete behavior of sync objects is slightly different.  References to
+ * sync objects are added:
+ *
+ *    - By \c glFencSynce.  This sets the initial reference count to 1.
+ *    - At the start of \c glClientWaitSync.  The reference is held for the
+ *      duration of the wait call.
+ *
+ * References are removed:
+ *
+ *    - By \c glDeleteSync.
+ *    - At the end of \c glClientWaitSync.
+ *
+ * Additionally, drivers may call \c _mesa_ref_sync_object and
+ * \c _mesa_unref_sync_object as needed to implement \c ServerWaitSync.
+ *
+ * As with shader objects, sync object names become invalid as soon as
+ * \c glDeleteSync is called.  For this reason \c glDeleteSync sets the
+ * \c DeletePending flag.  All functions validate object handles by testing
+ * this flag.
+ *
+ * \note
+ * Only \c GL_ARB_sync objects are shared between contexts.  If support is ever
+ * added for either \c GL_NV_fence or \c GL_APPLE_fence different semantics
+ * will need to be implemented.
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "context.h"
+#include "macros.h"
+
+#if FEATURE_ARB_sync
+#include "syncobj.h"
+
+static struct gl_sync_object *
+_mesa_new_sync_object(GLcontext *ctx, GLenum type)
+{
+   struct gl_sync_object *s = MALLOC_STRUCT(gl_sync_object);
+   (void) ctx;
+   (void) type;
+
+   return s;
+}
+
+
+static void
+_mesa_delete_sync_object(GLcontext *ctx, struct gl_sync_object *syncObj)
+{
+   (void) ctx;
+   _mesa_free(syncObj);
+}
+
+
+static void
+_mesa_fence_sync(GLcontext *ctx, struct gl_sync_object *syncObj,
+		 GLenum condition, GLbitfield flags)
+{
+   (void) ctx;
+   (void) condition;
+   (void) flags;
+
+   syncObj->StatusFlag = 1;
+}
+
+
+static void
+_mesa_check_sync(GLcontext *ctx, struct gl_sync_object *syncObj)
+{
+   (void) ctx;
+   (void) syncObj;
+
+   /* No-op for software rendering.  Hardware drivers will need to determine
+    * whether the state of the sync object has changed.
+    */
+}
+
+
+static void
+_mesa_wait_sync(GLcontext *ctx, struct gl_sync_object *syncObj,
+		GLbitfield flags, GLuint64 timeout)
+{
+   (void) ctx;
+   (void) syncObj;
+   (void) flags;
+   (void) timeout;
+
+   /* No-op for software rendering.  Hardware drivers will need to wait until
+    * the state of the sync object changes or the timeout expires.
+    */
+}
+
+
+void
+_mesa_init_sync_object_functions(struct dd_function_table *driver)
+{
+   driver->NewSyncObject = _mesa_new_sync_object;
+   driver->FenceSync = _mesa_fence_sync;
+   driver->DeleteSyncObject = _mesa_delete_sync_object;
+   driver->CheckSync = _mesa_check_sync;
+
+   /* Use the same no-op wait function for both.
+    */
+   driver->ClientWaitSync = _mesa_wait_sync;
+   driver->ServerWaitSync = _mesa_wait_sync;
+}
+
+
+/**
+ * Allocate/init the context state related to sync objects.
+ */
+void
+_mesa_init_sync(GLcontext *ctx)
+{
+   (void) ctx;
+}
+
+
+/**
+ * Free the context state related to sync objects.
+ */
+void
+_mesa_free_sync_data(GLcontext *ctx)
+{
+   (void) ctx;
+}
+
+
+static int
+_mesa_validate_sync(struct gl_sync_object *syncObj)
+{
+   return (syncObj != NULL)
+      && (syncObj->Type == GL_SYNC_FENCE)
+      && !syncObj->DeletePending;
+}
+
+
+void
+_mesa_ref_sync_object(GLcontext *ctx, struct gl_sync_object *syncObj)
+{
+   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
+   syncObj->RefCount++;
+   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+}
+
+
+void
+_mesa_unref_sync_object(GLcontext *ctx, struct gl_sync_object *syncObj)
+{
+   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
+   syncObj->RefCount--;
+   if (syncObj->RefCount == 0) {
+      remove_from_list(& syncObj->link);
+      _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+
+      ctx->Driver.DeleteSyncObject(ctx, syncObj);
+   } else {
+      _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+   }
+}
+
+
+GLboolean
+_mesa_IsSync(GLsync sync)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *const syncObj = (struct gl_sync_object *) sync;
+   ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, GL_FALSE);
+
+   return _mesa_validate_sync(syncObj) ? GL_TRUE : GL_FALSE;
+}
+
+
+void
+_mesa_DeleteSync(GLsync sync)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *const syncObj = (struct gl_sync_object *) sync;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   /* From the GL_ARB_sync spec:
+    *
+    *    DeleteSync will silently ignore a <sync> value of zero. An
+    *    INVALID_VALUE error is generated if <sync> is neither zero nor the
+    *    name of a sync object.
+    */
+   if (sync == 0) {
+      return;
+   }
+
+   if (!_mesa_validate_sync(syncObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glDeleteSync");
+      return;
+   }
+
+   /* If there are no client-waits or server-waits pending on this sync, delete
+    * the underlying object.
+    */
+   syncObj->DeletePending = GL_TRUE;
+   _mesa_unref_sync_object(ctx, syncObj);
+}
+
+
+GLsync
+_mesa_FenceSync(GLenum condition, GLbitfield flags)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *syncObj;
+   ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, 0);
+
+   if (condition != GL_SYNC_GPU_COMMANDS_COMPLETE) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glFenceSync(condition=0x%x)",
+		  condition);
+      return 0;
+   }
+
+   if (flags != 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE, "glFenceSync(flags=0x%x)",
+		  condition);
+      return 0;
+   }
+
+   syncObj = ctx->Driver.NewSyncObject(ctx, GL_SYNC_FENCE);
+   if (syncObj != NULL) {
+      syncObj->Type = GL_SYNC_FENCE;
+      /* The name is not currently used, and it is never visible to
+       * applications.  If sync support is extended to provide support for
+       * NV_fence, this field will be used.  We'll also need to add an object
+       * ID hashtable.
+       */
+      syncObj->Name = 1;
+      syncObj->RefCount = 1;
+      syncObj->DeletePending = GL_FALSE;
+      syncObj->SyncCondition = condition;
+      syncObj->Flags = flags;
+      syncObj->StatusFlag = 0;
+
+      ctx->Driver.FenceSync(ctx, syncObj, condition, flags);
+
+      _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
+      insert_at_tail(& ctx->Shared->SyncObjects, & syncObj->link);
+      _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+
+      return (GLsync) syncObj;
+   }
+
+   return NULL;
+}
+
+
+GLenum
+_mesa_ClientWaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *const syncObj = (struct gl_sync_object *) sync;
+   GLenum ret;
+   ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, GL_WAIT_FAILED);
+
+   if (!_mesa_validate_sync(syncObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glClientWaitSync");
+      return GL_WAIT_FAILED;
+   }
+
+   if ((flags & ~GL_SYNC_FLUSH_COMMANDS_BIT) != 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glClientWaitSync(flags=0x%x)", flags);
+      return GL_WAIT_FAILED;
+   }
+
+   _mesa_ref_sync_object(ctx, syncObj);
+
+   /* From the GL_ARB_sync spec:
+    *
+    *    ClientWaitSync returns one of four status values. A return value of
+    *    ALREADY_SIGNALED indicates that <sync> was signaled at the time
+    *    ClientWaitSync was called. ALREADY_SIGNALED will always be returned
+    *    if <sync> was signaled, even if the value of <timeout> is zero.
+    */
+   ctx->Driver.CheckSync(ctx, syncObj);
+   if (syncObj->StatusFlag) {
+      ret = GL_ALREADY_SIGNALED;
+   } else {
+      ctx->Driver.ClientWaitSync(ctx, syncObj, flags, timeout);
+
+      ret = syncObj->StatusFlag ? GL_CONDITION_SATISFIED : GL_TIMEOUT_EXPIRED;
+   }
+
+   _mesa_unref_sync_object(ctx, syncObj);
+   return ret;
+}
+
+
+void
+_mesa_WaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *const syncObj = (struct gl_sync_object *) sync;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!_mesa_validate_sync(syncObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glWaitSync");
+      return;
+   }
+
+   if (flags != 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glWaitSync(flags=0x%x)", flags);
+      return;
+   }
+
+   /* From the GL_ARB_sync spec:
+    *
+    *     If the value of <timeout> is zero, then WaitSync does nothing.
+    */
+   if (timeout == 0) {
+      return;
+   }
+
+   ctx->Driver.ServerWaitSync(ctx, syncObj, flags, timeout);
+}
+
+
+void
+_mesa_GetSynciv(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei *length,
+		GLint *values)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_sync_object *const syncObj = (struct gl_sync_object *) sync;
+   GLsizei size = 0;
+   GLint v[1];
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!_mesa_validate_sync(syncObj)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetSynciv");
+      return;
+   }
+
+   switch (pname) {
+   case GL_OBJECT_TYPE:
+      v[0] = syncObj->Type;
+      size = 1;
+      break;
+
+   case GL_SYNC_CONDITION:
+      v[0] = syncObj->SyncCondition;
+      size = 1;
+      break;
+
+   case GL_SYNC_STATUS:
+      /* Update the state of the sync by dipping into the driver.  Note that
+       * this call won't block.  It just updates state in the common object
+       * data from the current driver state.
+       */
+      ctx->Driver.CheckSync(ctx, syncObj);
+
+      v[0] = (syncObj->StatusFlag) ? GL_SIGNALED : GL_UNSIGNALED;
+      size = 1;
+      break;
+
+   case GL_SYNC_FLAGS:
+      v[0] = syncObj->Flags;
+      size = 1;
+      break;
+
+   default:
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetSynciv(pname=0x%x)\n", pname);
+      return;
+   }
+
+   if (size > 0) {
+      const GLsizei copy_count = MIN2(size, bufSize);
+
+      _mesa_memcpy(values, v, sizeof(GLint) * copy_count);
+   }
+
+   if (length != NULL) {
+      *length = size;
+   }
+}
+
+#endif /* FEATURE_ARB_sync */
diff --git a/src/mesa/main/syncobj.h b/src/mesa/main/syncobj.h
new file mode 100644
index 0000000000..fc160af289
--- /dev/null
+++ b/src/mesa/main/syncobj.h
@@ -0,0 +1,70 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file syncobj.h
+ * Sync object management.
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+#ifndef SYNCOBJ_H
+#define SYNCOBJ_H
+
+#include "context.h"
+
+extern void
+_mesa_init_sync_object_functions(struct dd_function_table *driver);
+
+extern void
+_mesa_init_sync(GLcontext *);
+
+extern void
+_mesa_free_sync_data(GLcontext *);
+
+extern void
+_mesa_ref_sync_object(GLcontext *ctx, struct gl_sync_object *syncObj);
+
+extern void
+_mesa_unref_sync_object(GLcontext *ctx, struct gl_sync_object *syncObj);
+
+extern GLboolean
+_mesa_IsSync(GLsync sync);
+
+extern void
+_mesa_DeleteSync(GLsync sync);
+
+extern GLsync
+_mesa_FenceSync(GLenum condition, GLbitfield flags);
+
+extern GLenum
+_mesa_ClientWaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout);
+
+extern void
+_mesa_WaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout);
+
+extern void
+_mesa_GetSynciv(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei *length,
+		GLint *values);
+
+#endif /* SYNCOBJ_H */
diff --git a/src/mesa/main/texenv.c b/src/mesa/main/texenv.c
index 4c04a7ed37..6d86a4275c 100644
--- a/src/mesa/main/texenv.c
+++ b/src/mesa/main/texenv.c
@@ -35,6 +35,7 @@
 #include "main/enums.h"
 #include "main/macros.h"
 #include "main/texenv.h"
+#include "main/texstate.h"
 
 
 #define TE_ERROR(errCode, msg, value)				\
@@ -466,7 +467,7 @@ _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param )
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (target == GL_TEXTURE_ENV) {
       switch (pname) {
@@ -793,7 +794,7 @@ _mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params )
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (target == GL_TEXTURE_ENV) {
       if (pname == GL_TEXTURE_ENV_COLOR) {
@@ -857,7 +858,7 @@ _mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params )
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (target == GL_TEXTURE_ENV) {
       if (pname == GL_TEXTURE_ENV_COLOR) {
@@ -908,12 +909,26 @@ _mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params )
    }
 }
 
-/* why does ATI_envmap_bumpmap require new entrypoints? Should just
-   reuse TexEnv ones... */
+
+/**
+ * Why does ATI_envmap_bumpmap require new entrypoints? Should just
+ * reuse TexEnv ones...
+ */
 void GLAPIENTRY
 _mesa_TexBumpParameterivATI( GLenum pname, const GLint *param )
 {
    GLfloat p[4];
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!ctx->Extensions.ATI_envmap_bumpmap) {
+      /* This isn't an "official" error case, but let's tell the user
+       * that something's wrong.
+       */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glTexBumpParameterivATI");
+      return;
+   }
+
    if (pname == GL_BUMP_ROT_MATRIX_ATI) {
       /* hope that conversion is correct here */
       p[0] = INT_TO_FLOAT( param[0] );
@@ -923,11 +938,12 @@ _mesa_TexBumpParameterivATI( GLenum pname, const GLint *param )
    }
    else {
       p[0] = (GLfloat) param[0];
-      p[1] = p[2] = p[3] = 0;  /* init to zero, just to be safe */
+      p[1] = p[2] = p[3] = 0.0F;  /* init to zero, just to be safe */
    }
    _mesa_TexBumpParameterfvATI( pname, p );
 }
 
+
 void GLAPIENTRY
 _mesa_TexBumpParameterfvATI( GLenum pname, const GLfloat *param )
 {
@@ -935,8 +951,12 @@ _mesa_TexBumpParameterfvATI( GLenum pname, const GLfloat *param )
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   /* should return error if extension not supported? */
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   if (!ctx->Extensions.ATI_envmap_bumpmap) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glTexBumpParameterfvATI");
+      return;
+   }
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (pname == GL_BUMP_ROT_MATRIX_ATI) {
       if (TEST_EQ_4V(param, texUnit->RotMatrix))
@@ -955,17 +975,21 @@ _mesa_TexBumpParameterfvATI( GLenum pname, const GLfloat *param )
    }
 }
 
+
 void GLAPIENTRY
 _mesa_GetTexBumpParameterivATI( GLenum pname, GLint *param )
 {
    const struct gl_texture_unit *texUnit;
    GLuint i;
-   GLint temp = 0;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   /* should return error if extension not supported? */
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   if (!ctx->Extensions.ATI_envmap_bumpmap) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexBumpParameterivATI");
+      return;
+   }
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (pname == GL_BUMP_ROT_MATRIX_SIZE_ATI) {
       /* spec leaves open to support larger matrices.
@@ -982,12 +1006,13 @@ _mesa_GetTexBumpParameterivATI( GLenum pname, GLint *param )
       param[3] = FLOAT_TO_INT(texUnit->RotMatrix[3]);
    }
    else if (pname == GL_BUMP_NUM_TEX_UNITS_ATI) {
+      GLint count = 0;
       for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
          if (ctx->Const.SupportedBumpUnits & (1 << i)) {
-            temp++;
+            count++;
          }
       }
-      *param = temp;
+      *param = count;
    }
    else if (pname == GL_BUMP_TEX_UNITS_ATI) {
       for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
@@ -1002,23 +1027,27 @@ _mesa_GetTexBumpParameterivATI( GLenum pname, GLint *param )
    }
 }
 
+
 void GLAPIENTRY
 _mesa_GetTexBumpParameterfvATI( GLenum pname, GLfloat *param )
 {
    const struct gl_texture_unit *texUnit;
    GLuint i;
-   GLint temp = 0;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   /* should return error if extension not supported? */
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   if (!ctx->Extensions.ATI_envmap_bumpmap) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexBumpParameterfvATI");
+      return;
+   }
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    if (pname == GL_BUMP_ROT_MATRIX_SIZE_ATI) {
       /* spec leaves open to support larger matrices.
          Don't think anyone would ever want to use it
          (and apps might not understand it) so hardcode this. */
-      *param = (GLfloat) 4;
+      *param = 4.0F;
    }
    else if (pname == GL_BUMP_ROT_MATRIX_ATI) {
       param[0] = texUnit->RotMatrix[0];
@@ -1027,12 +1056,13 @@ _mesa_GetTexBumpParameterfvATI( GLenum pname, GLfloat *param )
       param[3] = texUnit->RotMatrix[3];
    }
    else if (pname == GL_BUMP_NUM_TEX_UNITS_ATI) {
+      GLint count = 0;
       for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
          if (ctx->Const.SupportedBumpUnits & (1 << i)) {
-            temp++;
+            count++;
          }
       }
-      *param = (GLfloat) temp;
+      *param = (GLfloat) count;
    }
    else if (pname == GL_BUMP_TEX_UNITS_ATI) {
       for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
@@ -1046,4 +1076,3 @@ _mesa_GetTexBumpParameterfvATI( GLenum pname, GLfloat *param )
       return;
    }
 }
-
diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c
index 454d97506e..2f3e47e69e 100644
--- a/src/mesa/main/texenvprogram.c
+++ b/src/mesa/main/texenvprogram.c
@@ -27,8 +27,7 @@
  **************************************************************************/
 
 #include "glheader.h"
-#include "macros.h"
-#include "enums.h"
+#include "imports.h"
 #include "shader/program.h"
 #include "shader/prog_parameter.h"
 #include "shader/prog_cache.h"
@@ -62,6 +61,18 @@ struct texenvprog_cache_item
    struct texenvprog_cache_item *next;
 };
 
+static GLboolean
+texenv_doing_secondary_color(GLcontext *ctx)
+{
+   if (ctx->Light.Enabled &&
+       (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR))
+      return GL_TRUE;
+
+   if (ctx->Fog.ColorSumEnabled)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
 
 /**
  * Up to nine instructions per tex unit, plus fog, specular color.
@@ -71,8 +82,13 @@ struct texenvprog_cache_item
 #define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
 
 struct mode_opt {
-   GLuint Source:4;
-   GLuint Operand:3;
+#ifdef __GNUC__
+   __extension__ GLubyte Source:4;  /**< SRC_x */
+   __extension__ GLubyte Operand:3; /**< OPR_x */
+#else
+   GLubyte Source;  /**< SRC_x */
+   GLubyte Operand; /**< OPR_x */
+#endif
 };
 
 struct state_key {
@@ -80,24 +96,26 @@ struct state_key {
    GLuint enabled_units:8;
    GLuint separate_specular:1;
    GLuint fog_enabled:1;
-   GLuint fog_mode:2;
+   GLuint fog_mode:2;          /**< FOG_x */
    GLuint inputs_available:12;
 
+   /* NOTE: This array of structs must be last! (see "keySize" below) */
    struct {
       GLuint enabled:1;
-      GLuint source_index:3;   /* one of TEXTURE_1D/2D/3D/CUBE/RECT_INDEX */
+      GLuint source_index:3;   /**< TEXTURE_x_INDEX */
       GLuint shadow:1;
       GLuint ScaleShiftRGB:2;
       GLuint ScaleShiftA:2;
 
-      GLuint NumArgsRGB:3;
-      GLuint ModeRGB:5;
-      struct mode_opt OptRGB[MAX_COMBINER_TERMS];
+      GLuint NumArgsRGB:3;  /**< up to MAX_COMBINER_TERMS */
+      GLuint ModeRGB:5;     /**< MODE_x */
+
+      GLuint NumArgsA:3;  /**< up to MAX_COMBINER_TERMS */
+      GLuint ModeA:5;     /**< MODE_x */
 
-      GLuint NumArgsA:3;
-      GLuint ModeA:5;
+      struct mode_opt OptRGB[MAX_COMBINER_TERMS];
       struct mode_opt OptA[MAX_COMBINER_TERMS];
-   } unit[8];
+   } unit[MAX_TEXTURE_UNITS];
 };
 
 #define FOG_LINEAR  0
@@ -228,25 +246,51 @@ static GLuint translate_mode( GLenum envMode, GLenum mode )
    }
 }
 
-#define TEXTURE_UNKNOWN_INDEX 7
-static GLuint translate_tex_src_bit( GLbitfield bit )
+
+/**
+ * Do we need to clamp the results of the given texture env/combine mode?
+ * If the inputs to the mode are in [0,1] we don't always have to clamp
+ * the results.
+ */
+static GLboolean
+need_saturate( GLuint mode )
 {
-   /* make sure number of switch cases is correct */
-   assert(NUM_TEXTURE_TARGETS == 7);
-   switch (bit) {
-   case TEXTURE_1D_BIT:   return TEXTURE_1D_INDEX;
-   case TEXTURE_2D_BIT:   return TEXTURE_2D_INDEX;
-   case TEXTURE_RECT_BIT: return TEXTURE_RECT_INDEX;
-   case TEXTURE_3D_BIT:   return TEXTURE_3D_INDEX;
-   case TEXTURE_CUBE_BIT: return TEXTURE_CUBE_INDEX;
-   case TEXTURE_1D_ARRAY_BIT: return TEXTURE_1D_ARRAY_INDEX;
-   case TEXTURE_2D_ARRAY_BIT: return TEXTURE_2D_ARRAY_INDEX;
+   switch (mode) {
+   case MODE_REPLACE:
+   case MODE_MODULATE:
+   case MODE_INTERPOLATE:
+      return GL_FALSE;
+   case MODE_ADD:
+   case MODE_ADD_SIGNED:
+   case MODE_SUBTRACT:
+   case MODE_DOT3_RGB:
+   case MODE_DOT3_RGB_EXT:
+   case MODE_DOT3_RGBA:
+   case MODE_DOT3_RGBA_EXT:
+   case MODE_MODULATE_ADD_ATI:
+   case MODE_MODULATE_SIGNED_ADD_ATI:
+   case MODE_MODULATE_SUBTRACT_ATI:
+   case MODE_ADD_PRODUCTS:
+   case MODE_ADD_PRODUCTS_SIGNED:
+   case MODE_BUMP_ENVMAP_ATI:
+      return GL_TRUE;
    default:
       assert(0);
-      return TEXTURE_UNKNOWN_INDEX;
    }
 }
 
+
+
+/**
+ * Translate TEXTURE_x_BIT to TEXTURE_x_INDEX.
+ */
+static GLuint translate_tex_src_bit( GLbitfield bit )
+{
+   ASSERT(bit);
+   return _mesa_ffs(bit) - 1;
+}
+
+
 #define VERT_BIT_TEX_ANY    (0xff << VERT_ATTRIB_TEX0)
 #define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
 
@@ -299,7 +343,7 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx )
       if (ctx->Light.Enabled) {
          fp_inputs |= FRAG_BIT_COL0;
 
-         if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+         if (texenv_doing_secondary_color(ctx))
             fp_inputs |= FRAG_BIT_COL1;
       }
 
@@ -310,8 +354,10 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx )
       /* Then look at what might be varying as a result of enabled
        * arrays, etc:
        */
-      if (varying_inputs & VERT_BIT_COLOR0) fp_inputs |= FRAG_BIT_COL0;
-      if (varying_inputs & VERT_BIT_COLOR1) fp_inputs |= FRAG_BIT_COL1;
+      if (varying_inputs & VERT_BIT_COLOR0)
+         fp_inputs |= FRAG_BIT_COL0;
+      if (varying_inputs & VERT_BIT_COLOR1)
+         fp_inputs |= FRAG_BIT_COL1;
 
       fp_inputs |= (((varying_inputs & VERT_BIT_TEX_ANY) >> VERT_ATTRIB_TEX0) 
                     << FRAG_ATTRIB_TEX0);
@@ -340,8 +386,10 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx )
       if (ctx->Point.PointSprite)
          vp_outputs |= FRAG_BITS_TEX_ANY;
 
-      if (vp_outputs & (1 << VERT_RESULT_COL0)) fp_inputs |= FRAG_BIT_COL0;
-      if (vp_outputs & (1 << VERT_RESULT_COL1)) fp_inputs |= FRAG_BIT_COL1;
+      if (vp_outputs & (1 << VERT_RESULT_COL0))
+         fp_inputs |= FRAG_BIT_COL0;
+      if (vp_outputs & (1 << VERT_RESULT_COL1))
+         fp_inputs |= FRAG_BIT_COL1;
 
       fp_inputs |= (((vp_outputs & VERT_RESULT_TEX_ANY) >> VERT_RESULT_TEX0) 
                     << FRAG_ATTRIB_TEX0);
@@ -355,55 +403,55 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx )
  * Examine current texture environment state and generate a unique
  * key to identify it.
  */
-static void make_state_key( GLcontext *ctx,  struct state_key *key )
+static GLuint make_state_key( GLcontext *ctx,  struct state_key *key )
 {
    GLuint i, j;
    GLbitfield inputs_referenced = FRAG_BIT_COL0;
-   GLbitfield inputs_available = get_fp_input_mask( ctx );
+   const GLbitfield inputs_available = get_fp_input_mask( ctx );
+   GLuint keySize;
 
    memset(key, 0, sizeof(*key));
 
    /* _NEW_TEXTURE */
    for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
       const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+      const struct gl_texture_object *texObj = texUnit->_Current;
+      const struct gl_tex_env_combine_state *comb = texUnit->_CurrentCombine;
       GLenum format;
 
       if (!texUnit->_ReallyEnabled || !texUnit->Enabled)
          continue;
 
-      format = texUnit->_Current->Image[0][texUnit->_Current->BaseLevel]->_BaseFormat;
+      format = texObj->Image[0][texObj->BaseLevel]->_BaseFormat;
 
       key->unit[i].enabled = 1;
       key->enabled_units |= (1<<i);
-      key->nr_enabled_units = i+1;
+      key->nr_enabled_units = i + 1;
       inputs_referenced |= FRAG_BIT_TEX(i);
 
-      key->unit[i].source_index = 
-	 translate_tex_src_bit(texUnit->_ReallyEnabled);		
-      key->unit[i].shadow = ((texUnit->_Current->CompareMode == GL_COMPARE_R_TO_TEXTURE) && 
+      key->unit[i].source_index =
+         translate_tex_src_bit(texUnit->_ReallyEnabled);
+
+      key->unit[i].shadow = ((texObj->CompareMode == GL_COMPARE_R_TO_TEXTURE) &&
                              ((format == GL_DEPTH_COMPONENT) || 
                               (format == GL_DEPTH_STENCIL_EXT)));
 
-      key->unit[i].NumArgsRGB = texUnit->_CurrentCombine->_NumArgsRGB;
-      key->unit[i].NumArgsA = texUnit->_CurrentCombine->_NumArgsA;
+      key->unit[i].NumArgsRGB = comb->_NumArgsRGB;
+      key->unit[i].NumArgsA = comb->_NumArgsA;
 
       key->unit[i].ModeRGB =
-	 translate_mode(texUnit->EnvMode, texUnit->_CurrentCombine->ModeRGB);
+	 translate_mode(texUnit->EnvMode, comb->ModeRGB);
       key->unit[i].ModeA =
-	 translate_mode(texUnit->EnvMode, texUnit->_CurrentCombine->ModeA);
+	 translate_mode(texUnit->EnvMode, comb->ModeA);
 
-      key->unit[i].ScaleShiftRGB = texUnit->_CurrentCombine->ScaleShiftRGB;
-      key->unit[i].ScaleShiftA = texUnit->_CurrentCombine->ScaleShiftA;
+      key->unit[i].ScaleShiftRGB = comb->ScaleShiftRGB;
+      key->unit[i].ScaleShiftA = comb->ScaleShiftA;
 
       for (j = 0; j < MAX_COMBINER_TERMS; j++) {
-         key->unit[i].OptRGB[j].Operand =
-	    translate_operand(texUnit->_CurrentCombine->OperandRGB[j]);
-         key->unit[i].OptA[j].Operand =
-	    translate_operand(texUnit->_CurrentCombine->OperandA[j]);
-         key->unit[i].OptRGB[j].Source =
-	    translate_source(texUnit->_CurrentCombine->SourceRGB[j]);
-         key->unit[i].OptA[j].Source =
-	    translate_source(texUnit->_CurrentCombine->SourceA[j]);
+         key->unit[i].OptRGB[j].Operand = translate_operand(comb->OperandRGB[j]);
+         key->unit[i].OptA[j].Operand = translate_operand(comb->OperandA[j]);
+         key->unit[i].OptRGB[j].Source = translate_source(comb->SourceRGB[j]);
+         key->unit[i].OptA[j].Source = translate_source(comb->SourceA[j]);
       }
 
       if (key->unit[i].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
@@ -417,8 +465,8 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
        }
    }
 
-   /* _DD_NEW_SEPARATE_SPECULAR */
-   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
+   /* _NEW_LIGHT | _NEW_FOG */
+   if (texenv_doing_secondary_color(ctx)) {
       key->separate_specular = 1;
       inputs_referenced |= FRAG_BIT_COL1;
    }
@@ -431,8 +479,15 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
    }
 
    key->inputs_available = (inputs_available & inputs_referenced);
+
+   /* compute size of state key, ignoring unused texture units */
+   keySize = sizeof(*key) - sizeof(key->unit)
+      + key->nr_enabled_units * sizeof(key->unit[0]);
+
+   return keySize;
 }
 
+
 /**
  * Use uregs to represent registers internally, translate to Mesa's
  * expected formats on emit.  
@@ -450,10 +505,8 @@ struct ureg {
    GLuint file:4;
    GLuint idx:8;
    GLuint negatebase:1;
-   GLuint abs:1;
-   GLuint negateabs:1;
    GLuint swz:12;
-   GLuint pad:5;
+   GLuint pad:7;
 };
 
 static const struct ureg undef = { 
@@ -461,8 +514,6 @@ static const struct ureg undef = {
    ~0,
    0,
    0,
-   0,
-   0,
    0
 };
 
@@ -471,7 +522,6 @@ static const struct ureg undef = {
  */
 struct texenv_fragment_program {
    struct gl_fragment_program *program;
-   GLcontext *ctx;
    struct state_key *state;
 
    GLbitfield alu_temps;	/**< Track texture indirections, see spec. */
@@ -508,8 +558,6 @@ static struct ureg make_ureg(GLuint file, GLuint idx)
    reg.file = file;
    reg.idx = idx;
    reg.negatebase = 0;
-   reg.abs = 0;
-   reg.negateabs = 0;
    reg.swz = SWIZZLE_NOOP;
    reg.pad = 0;
    return reg;
@@ -675,7 +723,7 @@ static void emit_arg( struct prog_src_register *reg,
    reg->Index = ureg.idx;
    reg->Swizzle = ureg.swz;
    reg->Negate = ureg.negatebase ? NEGATE_XYZW : NEGATE_NONE;
-   reg->Abs = ureg.abs;
+   reg->Abs = GL_FALSE;
 }
 
 static void emit_dst( struct prog_dst_register *dst,
@@ -698,7 +746,7 @@ emit_op(struct texenv_fragment_program *p,
 	struct ureg src1,
 	struct ureg src2 )
 {
-   GLuint nr = p->program->Base.NumInstructions++;
+   const GLuint nr = p->program->Base.NumInstructions++;
    struct prog_instruction *inst = &p->program->Base.Instructions[nr];
 
    assert(nr < MAX_INSTRUCTIONS);
@@ -936,17 +984,22 @@ static struct ureg emit_combine_source( struct texenv_fragment_program *p,
    }
 }
 
+/**
+ * Check if the RGB and Alpha sources and operands match for the given
+ * texture unit's combinder state.  When the RGB and A sources and
+ * operands match, we can emit fewer instructions.
+ */
 static GLboolean args_match( const struct state_key *key, GLuint unit )
 {
-   GLuint i, nr = key->unit[unit].NumArgsRGB;
+   GLuint i, numArgs = key->unit[unit].NumArgsRGB;
 
-   for (i = 0 ; i < nr ; i++) {
+   for (i = 0; i < numArgs; i++) {
       if (key->unit[unit].OptA[i].Source != key->unit[unit].OptRGB[i].Source) 
 	 return GL_FALSE;
 
-      switch(key->unit[unit].OptA[i].Operand) {
+      switch (key->unit[unit].OptA[i].Operand) {
       case OPR_SRC_ALPHA: 
-	 switch(key->unit[unit].OptRGB[i].Operand) {
+	 switch (key->unit[unit].OptRGB[i].Operand) {
 	 case OPR_SRC_COLOR: 
 	 case OPR_SRC_ALPHA: 
 	    break;
@@ -955,7 +1008,7 @@ static GLboolean args_match( const struct state_key *key, GLuint unit )
 	 }
 	 break;
       case OPR_ONE_MINUS_SRC_ALPHA: 
-	 switch(key->unit[unit].OptRGB[i].Operand) {
+	 switch (key->unit[unit].OptRGB[i].Operand) {
 	 case OPR_ONE_MINUS_SRC_COLOR: 
 	 case OPR_ONE_MINUS_SRC_ALPHA: 
 	    break;
@@ -1097,7 +1150,7 @@ static struct ureg
 emit_texenv(struct texenv_fragment_program *p, GLuint unit)
 {
    const struct state_key *key = p->state;
-   GLboolean saturate;
+   GLboolean rgb_saturate, alpha_saturate;
    GLuint rgb_shift, alpha_shift;
    struct ureg out, dest;
 
@@ -1127,7 +1180,19 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
    /* If we'll do rgb/alpha shifting don't saturate in emit_combine().
     * We don't want to clamp twice.
     */
-   saturate = !(rgb_shift || alpha_shift);
+   if (rgb_shift)
+      rgb_saturate = GL_FALSE;  /* saturate after rgb shift */
+   else if (need_saturate(key->unit[unit].ModeRGB))
+      rgb_saturate = GL_TRUE;
+   else
+      rgb_saturate = GL_FALSE;
+
+   if (alpha_shift)
+      alpha_saturate = GL_FALSE;  /* saturate after alpha shift */
+   else if (need_saturate(key->unit[unit].ModeA))
+      alpha_saturate = GL_TRUE;
+   else
+      alpha_saturate = GL_FALSE;
 
    /* If this is the very last calculation, emit direct to output reg:
     */
@@ -1143,7 +1208,7 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
     */
    if (key->unit[unit].ModeRGB == key->unit[unit].ModeA &&
        args_match(key, unit)) {
-      out = emit_combine( p, dest, WRITEMASK_XYZW, saturate,
+      out = emit_combine( p, dest, WRITEMASK_XYZW, rgb_saturate,
 			  unit,
 			  key->unit[unit].NumArgsRGB,
 			  key->unit[unit].ModeRGB,
@@ -1151,7 +1216,7 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
    }
    else if (key->unit[unit].ModeRGB == MODE_DOT3_RGBA_EXT ||
 	    key->unit[unit].ModeRGB == MODE_DOT3_RGBA) {
-      out = emit_combine( p, dest, WRITEMASK_XYZW, saturate,
+      out = emit_combine( p, dest, WRITEMASK_XYZW, rgb_saturate,
 			  unit,
 			  key->unit[unit].NumArgsRGB,
 			  key->unit[unit].ModeRGB,
@@ -1161,12 +1226,12 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
       /* Need to do something to stop from re-emitting identical
        * argument calculations here:
        */
-      out = emit_combine( p, dest, WRITEMASK_XYZ, saturate,
+      out = emit_combine( p, dest, WRITEMASK_XYZ, rgb_saturate,
 			  unit,
 			  key->unit[unit].NumArgsRGB,
 			  key->unit[unit].ModeRGB,
 			  key->unit[unit].OptRGB);
-      out = emit_combine( p, dest, WRITEMASK_W, saturate,
+      out = emit_combine( p, dest, WRITEMASK_W, alpha_saturate,
 			  unit,
 			  key->unit[unit].NumArgsA,
 			  key->unit[unit].ModeA,
@@ -1177,8 +1242,7 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
     */
    if (alpha_shift || rgb_shift) {
       struct ureg shift;
-
-      saturate = GL_TRUE;  /* always saturate at this point */
+      GLboolean saturate = GL_TRUE;  /* always saturate at this point */
 
       if (rgb_shift == alpha_shift) {
 	 shift = register_scalar_const(p, (GLfloat)(1<<rgb_shift));
@@ -1204,7 +1268,7 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
 static void load_texture( struct texenv_fragment_program *p, GLuint unit )
 {
    if (is_undef(p->src_texture[unit])) {
-      GLuint texTarget = p->state->unit[unit].source_index;
+      const GLuint texTarget = p->state->unit[unit].source_index;
       struct ureg texcoord;
       struct ureg tmp = get_tex_temp( p );
 
@@ -1216,9 +1280,6 @@ static void load_texture( struct texenv_fragment_program *p, GLuint unit )
          texcoord = p->texcoord_tex[unit];
       }
 
-      if (texTarget == TEXTURE_UNKNOWN_INDEX)
-         program_error(p, "TexSrcBit");
-			  
       /* TODO: Use D0_MASK_XY where possible.
        */
       if (p->state->unit[unit].enabled) {
@@ -1277,7 +1338,7 @@ static GLboolean load_texenv_source( struct texenv_fragment_program *p,
  * Generate instructions for loading all texture source terms.
  */
 static GLboolean
-load_texunit_sources( struct texenv_fragment_program *p, int unit )
+load_texunit_sources( struct texenv_fragment_program *p, GLuint unit )
 {
    const struct state_key *key = p->state;
    GLuint i;
@@ -1297,7 +1358,7 @@ load_texunit_sources( struct texenv_fragment_program *p, int unit )
  * Generate instructions for loading bump map textures.
  */
 static GLboolean
-load_texunit_bumpmap( struct texenv_fragment_program *p, int unit )
+load_texunit_bumpmap( struct texenv_fragment_program *p, GLuint unit )
 {
    const struct state_key *key = p->state;
    GLuint bumpedUnitNr = key->unit[unit].OptRGB[1].Source - SRC_TEXTURE0;
@@ -1313,17 +1374,20 @@ load_texunit_bumpmap( struct texenv_fragment_program *p, int unit )
    texcDst = get_tex_temp( p );
    p->texcoord_tex[bumpedUnitNr] = texcDst;
 
-   /* apply rot matrix and add coords to be available in next phase */
-   /* dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1) */
-   /* note only 2 coords are affected the rest are left unchanged (mul by 0) */
+   /* Apply rot matrix and add coords to be available in next phase.
+    * dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1)
+    * note only 2 coords are affected the rest are left unchanged (mul by 0)
+    */
    emit_arith( p, OPCODE_MAD, texcDst, WRITEMASK_XYZW, 0,
                swizzle1(bumpMapRes, SWIZZLE_X), rotMat0, texcSrc );
    emit_arith( p, OPCODE_MAD, texcDst, WRITEMASK_XYZW, 0,
                swizzle1(bumpMapRes, SWIZZLE_Y), rotMat1, texcDst );
 
-   /* move 0,0,0,1 into bumpmap src if someone (crossbar) is foolish
-      enough to access this later, should optimize away */
-   emit_arith( p, OPCODE_MOV, bumpMapRes, WRITEMASK_XYZW, 0, constdudvcolor, undef, undef );
+   /* Move 0,0,0,1 into bumpmap src if someone (crossbar) is foolish
+    * enough to access this later, should optimize away.
+    */
+   emit_arith( p, OPCODE_MOV, bumpMapRes, WRITEMASK_XYZW, 0,
+               constdudvcolor, undef, undef );
 
    return GL_TRUE;
 }
@@ -1342,7 +1406,6 @@ create_new_program(GLcontext *ctx, struct state_key *key,
    struct ureg cf, out;
 
    _mesa_memset(&p, 0, sizeof(p));
-   p.ctx = ctx;
    p.state = key;
    p.program = program;
 
@@ -1351,17 +1414,17 @@ create_new_program(GLcontext *ctx, struct state_key *key,
     */
    p.program->Base.Instructions = instBuffer;
    p.program->Base.Target = GL_FRAGMENT_PROGRAM_ARB;
-   p.program->Base.NumTexIndirections = 1;
+   p.program->Base.String = NULL;
+   p.program->Base.NumTexIndirections = 1; /* is this right? */
    p.program->Base.NumTexInstructions = 0;
    p.program->Base.NumAluInstructions = 0;
-   p.program->Base.String = NULL;
-   p.program->Base.NumInstructions =
-   p.program->Base.NumTemporaries =
-   p.program->Base.NumParameters =
-   p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
+   p.program->Base.NumInstructions = 0;
+   p.program->Base.NumTemporaries = 0;
+   p.program->Base.NumParameters = 0;
+   p.program->Base.NumAttributes = 0;
+   p.program->Base.NumAddressRegs = 0;
    p.program->Base.Parameters = _mesa_new_parameter_list();
-
-   p.program->Base.InputsRead = 0;
+   p.program->Base.InputsRead = 0x0;
    p.program->Base.OutputsWritten = 1 << FRAG_RESULT_COLOR;
 
    for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
@@ -1378,10 +1441,12 @@ create_new_program(GLcontext *ctx, struct state_key *key,
    release_temps(ctx, &p);
 
    if (key->enabled_units) {
-       GLboolean needbumpstage = GL_FALSE;
+      GLboolean needbumpstage = GL_FALSE;
+
       /* Zeroth pass - bump map textures first */
-      for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++)
-	 if (key->unit[unit].enabled && key->unit[unit].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
+      for (unit = 0; unit < key->nr_enabled_units; unit++)
+	 if (key->unit[unit].enabled &&
+             key->unit[unit].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
 	    needbumpstage = GL_TRUE;
 	    load_texunit_bumpmap( &p, unit );
 	 }
@@ -1392,7 +1457,7 @@ create_new_program(GLcontext *ctx, struct state_key *key,
        * all referenced texture sources and emit texld instructions
        * for each:
        */
-      for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++)
+      for (unit = 0; unit < key->nr_enabled_units; unit++)
 	 if (key->unit[unit].enabled) {
 	    load_texunit_sources( &p, unit );
 	    p.last_tex_stage = unit;
@@ -1400,8 +1465,8 @@ create_new_program(GLcontext *ctx, struct state_key *key,
 
       /* Second pass - emit combine instructions to build final color:
        */
-      for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++)
-	 if (key->enabled_units & (1<<unit)) {
+      for (unit = 0; unit < key->nr_enabled_units; unit++)
+	 if (key->unit[unit].enabled) {
 	    p.src_previous = emit_texenv( &p, unit );
             reserve_temp(&p, p.src_previous); /* don't re-use this temp reg */
 	    release_temps(ctx, &p);	/* release all temps */
@@ -1434,9 +1499,11 @@ create_new_program(GLcontext *ctx, struct state_key *key,
        * a reduced value and not what is expected in FogOption
        */
       p.program->FogOption = ctx->Fog.Mode;
-      p.program->Base.InputsRead |= FRAG_BIT_FOGC; /* XXX new */
-   } else
+      p.program->Base.InputsRead |= FRAG_BIT_FOGC;
+   }
+   else {
       p.program->FogOption = GL_NONE;
+   }
 
    if (p.program->Base.NumTexIndirections > ctx->Const.FragmentProgram.MaxTexIndirections) 
       program_error(&p, "Exceeded max nr indirect texture lookups");
@@ -1489,12 +1556,13 @@ _mesa_get_fixed_func_fragment_program(GLcontext *ctx)
 {
    struct gl_fragment_program *prog;
    struct state_key key;
+   GLuint keySize;
 	
-   make_state_key(ctx, &key);
+   keySize = make_state_key(ctx, &key);
       
    prog = (struct gl_fragment_program *)
       _mesa_search_program_cache(ctx->FragmentProgram.Cache,
-                                 &key, sizeof(key));
+                                 &key, keySize);
 
    if (!prog) {
       prog = (struct gl_fragment_program *) 
@@ -1503,7 +1571,7 @@ _mesa_get_fixed_func_fragment_program(GLcontext *ctx)
       create_new_program(ctx, &key, prog);
 
       _mesa_program_cache_insert(ctx, ctx->FragmentProgram.Cache,
-                                 &key, sizeof(key), &prog->Base);
+                                 &key, keySize, &prog->Base);
    }
 
    return prog;
diff --git a/src/mesa/main/texgetimage.c b/src/mesa/main/texgetimage.c
index 02409d8009..14d6fc7659 100644
--- a/src/mesa/main/texgetimage.c
+++ b/src/mesa/main/texgetimage.c
@@ -30,11 +30,14 @@
 
 
 #include "glheader.h"
+#include "bufferobj.h"
 #include "context.h"
 #include "image.h"
 #include "texcompress.h"
 #include "texformat.h"
 #include "texgetimage.h"
+#include "teximage.h"
+#include "texstate.h"
 
 
 
@@ -116,7 +119,7 @@ _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
 {
    const GLuint dimensions = (target == GL_TEXTURE_3D) ? 3 : 2;
 
-   if (ctx->Pack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
       /* Packing texture image into a PBO.
        * Map the (potentially) VRAM-based buffer into our process space so
        * we can write into it with the code below.
@@ -296,7 +299,7 @@ _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
       } /* img */
    }
 
-   if (ctx->Pack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                               ctx->Pack.BufferObj);
    }
@@ -316,7 +319,7 @@ _mesa_get_compressed_teximage(GLcontext *ctx, GLenum target, GLint level,
 {
    GLuint size;
 
-   if (ctx->Pack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
       /* pack texture image into a PBO */
       GLubyte *buf;
       if ((const GLubyte *) img + texImage->CompressedSize >
@@ -349,8 +352,230 @@ _mesa_get_compressed_teximage(GLcontext *ctx, GLenum target, GLint level,
    /* just memcpy, no pixelstore or pixel transfer */
    _mesa_memcpy(img, texImage->Data, size);
 
-   if (ctx->Pack.BufferObj->Name) {
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                               ctx->Pack.BufferObj);
    }
 }
+
+
+
+/**
+ * Do error checking for a glGetTexImage() call.
+ * \return GL_TRUE if any error, GL_FALSE if no errors.
+ */
+static GLboolean
+getteximage_error_check(GLcontext *ctx, GLenum target, GLint level,
+                        GLenum format, GLenum type, GLvoid *pixels )
+{
+   const struct gl_texture_unit *texUnit;
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   const GLuint maxLevels = _mesa_max_texture_levels(ctx, target);
+
+   if (maxLevels == 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(target=0x%x)", target);
+      return GL_TRUE;
+   }
+
+   if (level < 0 || level >= maxLevels) {
+      _mesa_error( ctx, GL_INVALID_VALUE, "glGetTexImage(level)" );
+      return GL_TRUE;
+   }
+
+   if (_mesa_sizeof_packed_type(type) <= 0) {
+      _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexImage(type)" );
+      return GL_TRUE;
+   }
+
+   if (_mesa_components_in_format(format) <= 0 ||
+       format == GL_STENCIL_INDEX) {
+      _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexImage(format)" );
+      return GL_TRUE;
+   }
+
+   if (!ctx->Extensions.EXT_paletted_texture && _mesa_is_index_format(format)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
+      return GL_TRUE;
+   }
+
+   if (!ctx->Extensions.ARB_depth_texture && _mesa_is_depth_format(format)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
+      return GL_TRUE;
+   }
+
+   if (!ctx->Extensions.MESA_ycbcr_texture && _mesa_is_ycbcr_format(format)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
+      return GL_TRUE;
+   }
+
+   if (!ctx->Extensions.EXT_packed_depth_stencil
+       && _mesa_is_depthstencil_format(format)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
+      return GL_TRUE;
+   }
+
+   if (!ctx->Extensions.ATI_envmap_bumpmap
+       && _mesa_is_dudv_format(format)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
+      return GL_TRUE;
+   }
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+
+   if (!texObj || _mesa_is_proxy_texture(target)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(target)");
+      return GL_TRUE;
+   }
+
+   texImage = _mesa_select_tex_image(ctx, texObj, target, level);
+   if (!texImage) {
+      /* out of memory */
+      return GL_TRUE;
+   }
+      
+   /* Make sure the requested image format is compatible with the
+    * texture's format.  Note that a color index texture can be converted
+    * to RGBA so that combo is allowed.
+    */
+   if (_mesa_is_color_format(format)
+       && !_mesa_is_color_format(texImage->TexFormat->BaseFormat)
+       && !_mesa_is_index_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+   else if (_mesa_is_index_format(format)
+            && !_mesa_is_index_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+   else if (_mesa_is_depth_format(format)
+            && !_mesa_is_depth_format(texImage->TexFormat->BaseFormat)
+            && !_mesa_is_depthstencil_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+   else if (_mesa_is_ycbcr_format(format)
+            && !_mesa_is_ycbcr_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+   else if (_mesa_is_depthstencil_format(format)
+            && !_mesa_is_depthstencil_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+   else if (_mesa_is_dudv_format(format)
+            && !_mesa_is_dudv_format(texImage->TexFormat->BaseFormat)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
+      return GL_TRUE;
+   }
+
+   if (_mesa_is_bufferobj(ctx->Pack.BufferObj)) {
+      /* packing texture image into a PBO */
+      const GLuint dimensions = (target == GL_TEXTURE_3D) ? 3 : 2;
+      if (!_mesa_validate_pbo_access(dimensions, &ctx->Pack, texImage->Width,
+                                     texImage->Height, texImage->Depth,
+                                     format, type, pixels)) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glGetTexImage(invalid PBO access)");
+         return GL_TRUE;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+
+
+/**
+ * Get texture image.  Called by glGetTexImage.
+ *
+ * \param target texture target.
+ * \param level image level.
+ * \param format pixel data format for returned image.
+ * \param type pixel data type for returned image.
+ * \param pixels returned pixel data.
+ */
+void GLAPIENTRY
+_mesa_GetTexImage( GLenum target, GLint level, GLenum format,
+                   GLenum type, GLvoid *pixels )
+{
+   const struct gl_texture_unit *texUnit;
+   struct gl_texture_object *texObj;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
+
+   if (getteximage_error_check(ctx, target, level, format, type, pixels)) {
+      return;
+   }
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+
+   _mesa_lock_texture(ctx, texObj);
+   {
+      struct gl_texture_image *texImage =
+         _mesa_select_tex_image(ctx, texObj, target, level);
+
+      /* typically, this will call _mesa_get_teximage() */
+      ctx->Driver.GetTexImage(ctx, target, level, format, type, pixels,
+                              texObj, texImage);
+   }
+   _mesa_unlock_texture(ctx, texObj);
+}
+
+
+void GLAPIENTRY
+_mesa_GetCompressedTexImageARB(GLenum target, GLint level, GLvoid *img)
+{
+   const struct gl_texture_unit *texUnit;
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   GLint maxLevels;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
+
+   texUnit = _mesa_get_current_tex_unit(ctx);
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   if (!texObj) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetCompressedTexImageARB");
+      return;
+   }
+
+   maxLevels = _mesa_max_texture_levels(ctx, target);
+   ASSERT(maxLevels > 0); /* 0 indicates bad target, caught above */
+
+   if (level < 0 || level >= maxLevels) {
+      _mesa_error(ctx, GL_INVALID_VALUE, "glGetCompressedTexImageARB(level)");
+      return;
+   }
+
+   if (_mesa_is_proxy_texture(target)) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetCompressedTexImageARB(target)");
+      return;
+   }
+
+   _mesa_lock_texture(ctx, texObj);
+   {
+      texImage = _mesa_select_tex_image(ctx, texObj, target, level);
+      if (texImage) {
+         if (texImage->IsCompressed) {
+            /* this typically calls _mesa_get_compressed_teximage() */
+            ctx->Driver.GetCompressedTexImage(ctx, target, level, img,
+                                              texObj, texImage);
+         }
+         else {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glGetCompressedTexImageARB");
+         }
+      }
+      else {
+         /* probably invalid mipmap level */
+         _mesa_error(ctx, GL_INVALID_VALUE,
+                     "glGetCompressedTexImageARB(level)");
+      }
+   }
+   _mesa_unlock_texture(ctx, texObj);
+}
diff --git a/src/mesa/main/texgetimage.h b/src/mesa/main/texgetimage.h
index 01f486e8f0..088d27c7e1 100644
--- a/src/mesa/main/texgetimage.h
+++ b/src/mesa/main/texgetimage.h
@@ -43,4 +43,13 @@ _mesa_get_compressed_teximage(GLcontext *ctx, GLenum target, GLint level,
 
 
 
+extern void GLAPIENTRY
+_mesa_GetTexImage( GLenum target, GLint level,
+                   GLenum format, GLenum type, GLvoid *pixels );
+
+
+extern void GLAPIENTRY
+_mesa_GetCompressedTexImageARB(GLenum target, GLint lod, GLvoid *img);
+
+
 #endif /* TEXGETIMAGE_H */
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 6e21066537..8228303040 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -1,6 +1,6 @@
 /*
  * mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
@@ -182,6 +182,8 @@ logbase2( int n )
  *
  * This is the format which is used during texture application (i.e. the
  * texture format and env mode determine the arithmetic used.
+ *
+ * XXX this could be static
  */
 GLint
 _mesa_base_tex_format( GLcontext *ctx, GLint internalFormat )
@@ -415,211 +417,6 @@ _mesa_base_tex_format( GLcontext *ctx, GLint internalFormat )
 
 
 /**
- * Test if the given image format is a color/RGBA format (i.e., not color
- * index, depth, stencil, etc).
- * \param format  the image format value (may by an internal texture format)
- * \return GL_TRUE if its a color/RGBA format, GL_FALSE otherwise.
- * XXX maybe move this func to image.c
- */
-GLboolean
-_mesa_is_color_format(GLenum format)
-{
-   switch (format) {
-      case GL_RED:
-      case GL_GREEN:
-      case GL_BLUE:
-      case GL_ALPHA:
-      case GL_ALPHA4:
-      case GL_ALPHA8:
-      case GL_ALPHA12:
-      case GL_ALPHA16:
-      case 1:
-      case GL_LUMINANCE:
-      case GL_LUMINANCE4:
-      case GL_LUMINANCE8:
-      case GL_LUMINANCE12:
-      case GL_LUMINANCE16:
-      case 2:
-      case GL_LUMINANCE_ALPHA:
-      case GL_LUMINANCE4_ALPHA4:
-      case GL_LUMINANCE6_ALPHA2:
-      case GL_LUMINANCE8_ALPHA8:
-      case GL_LUMINANCE12_ALPHA4:
-      case GL_LUMINANCE12_ALPHA12:
-      case GL_LUMINANCE16_ALPHA16:
-      case GL_INTENSITY:
-      case GL_INTENSITY4:
-      case GL_INTENSITY8:
-      case GL_INTENSITY12:
-      case GL_INTENSITY16:
-      case 3:
-      case GL_RGB:
-      case GL_BGR:
-      case GL_R3_G3_B2:
-      case GL_RGB4:
-      case GL_RGB5:
-      case GL_RGB8:
-      case GL_RGB10:
-      case GL_RGB12:
-      case GL_RGB16:
-      case 4:
-      case GL_ABGR_EXT:
-      case GL_RGBA:
-      case GL_BGRA:
-      case GL_RGBA2:
-      case GL_RGBA4:
-      case GL_RGB5_A1:
-      case GL_RGBA8:
-      case GL_RGB10_A2:
-      case GL_RGBA12:
-      case GL_RGBA16:
-      /* float texture formats */
-      case GL_ALPHA16F_ARB:
-      case GL_ALPHA32F_ARB:
-      case GL_LUMINANCE16F_ARB:
-      case GL_LUMINANCE32F_ARB:
-      case GL_LUMINANCE_ALPHA16F_ARB:
-      case GL_LUMINANCE_ALPHA32F_ARB:
-      case GL_INTENSITY16F_ARB:
-      case GL_INTENSITY32F_ARB:
-      case GL_RGB16F_ARB:
-      case GL_RGB32F_ARB:
-      case GL_RGBA16F_ARB:
-      case GL_RGBA32F_ARB:
-      /* compressed formats */
-      case GL_COMPRESSED_ALPHA:
-      case GL_COMPRESSED_LUMINANCE:
-      case GL_COMPRESSED_LUMINANCE_ALPHA:
-      case GL_COMPRESSED_INTENSITY:
-      case GL_COMPRESSED_RGB:
-      case GL_COMPRESSED_RGBA:
-      case GL_RGB_S3TC:
-      case GL_RGB4_S3TC:
-      case GL_RGBA_S3TC:
-      case GL_RGBA4_S3TC:
-      case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-      case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-      case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-      case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-      case GL_COMPRESSED_RGB_FXT1_3DFX:
-      case GL_COMPRESSED_RGBA_FXT1_3DFX:
-#if FEATURE_EXT_texture_sRGB
-      case GL_SRGB_EXT:
-      case GL_SRGB8_EXT:
-      case GL_SRGB_ALPHA_EXT:
-      case GL_SRGB8_ALPHA8_EXT:
-      case GL_SLUMINANCE_ALPHA_EXT:
-      case GL_SLUMINANCE8_ALPHA8_EXT:
-      case GL_SLUMINANCE_EXT:
-      case GL_SLUMINANCE8_EXT:
-      case GL_COMPRESSED_SRGB_EXT:
-      case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
-      case GL_COMPRESSED_SRGB_ALPHA_EXT:
-      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
-      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
-      case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
-      case GL_COMPRESSED_SLUMINANCE_EXT:
-      case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
-#endif /* FEATURE_EXT_texture_sRGB */
-         return GL_TRUE;
-      /* signed texture formats */
-      case GL_RGBA_SNORM:
-      case GL_RGBA8_SNORM:
-         return GL_TRUE;
-      case GL_YCBCR_MESA:  /* not considered to be RGB */
-         /* fall-through */
-      default:
-         return GL_FALSE;
-   }
-}
-
-
-/**
- * Test if the given image format is a color index format.
- */
-static GLboolean
-is_index_format(GLenum format)
-{
-   switch (format) {
-      case GL_COLOR_INDEX:
-      case GL_COLOR_INDEX1_EXT:
-      case GL_COLOR_INDEX2_EXT:
-      case GL_COLOR_INDEX4_EXT:
-      case GL_COLOR_INDEX8_EXT:
-      case GL_COLOR_INDEX12_EXT:
-      case GL_COLOR_INDEX16_EXT:
-         return GL_TRUE;
-      default:
-         return GL_FALSE;
-   }
-}
-
-
-/**
- * Test if the given image format is a depth component format.
- */
-static GLboolean
-is_depth_format(GLenum format)
-{
-   switch (format) {
-      case GL_DEPTH_COMPONENT16:
-      case GL_DEPTH_COMPONENT24:
-      case GL_DEPTH_COMPONENT32:
-      case GL_DEPTH_COMPONENT:
-         return GL_TRUE;
-      default:
-         return GL_FALSE;
-   }
-}
-
-
-/**
- * Test if the given image format is a YCbCr format.
- */
-static GLboolean
-is_ycbcr_format(GLenum format)
-{
-   switch (format) {
-      case GL_YCBCR_MESA:
-         return GL_TRUE;
-      default:
-         return GL_FALSE;
-   }
-}
-
-
-/**
- * Test if the given image format is a Depth/Stencil format.
- */
-static GLboolean
-is_depthstencil_format(GLenum format)
-{
-   switch (format) {
-      case GL_DEPTH24_STENCIL8_EXT:
-      case GL_DEPTH_STENCIL_EXT:
-         return GL_TRUE;
-      default:
-         return GL_FALSE;
-   }
-}
-
-/**
- * Test if the given image format is a dudv format.
- */
-static GLboolean
-is_dudv_format(GLenum format)
-{
-   switch (format) {
-      case GL_DUDV_ATI:
-      case GL_DU8DV8_ATI:
-         return GL_TRUE;
-      default:
-         return GL_FALSE;
-   }
-}
-
-
-/**
  * Test if it is a supported compressed format.
  * 
  * \param internalFormat the internal format token provided by the user.
@@ -679,37 +476,14 @@ _mesa_set_tex_image(struct gl_texture_object *tObj,
                     GLenum target, GLint level,
                     struct gl_texture_image *texImage)
 {
+   const GLuint face = _mesa_tex_target_to_face(target);
+
    ASSERT(tObj);
    ASSERT(texImage);
-   /* XXX simplify this with _mesa_tex_target_to_face() */
-   switch (target) {
-      case GL_TEXTURE_1D:
-      case GL_TEXTURE_2D:
-      case GL_TEXTURE_3D:
-      case GL_TEXTURE_1D_ARRAY_EXT:
-      case GL_TEXTURE_2D_ARRAY_EXT:
-         tObj->Image[0][level] = texImage;
-         break;
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB:
-         {
-            GLuint face = ((GLuint) target - 
-                           (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X);
-            tObj->Image[face][level] = texImage;
-         }
-         break;
-      case GL_TEXTURE_RECTANGLE_NV:
-         ASSERT(level == 0);
-         tObj->Image[0][level] = texImage;
-         break;
-      default:
-         _mesa_problem(NULL, "bad target in _mesa_set_tex_image()");
-         return;
-   }
+   ASSERT(target != GL_TEXTURE_RECTANGLE_NV || level == 0);
+
+   tObj->Image[face][level] = texImage;
+
    /* Set the 'back' pointer */
    texImage->TexObject = tObj;
 }
@@ -788,6 +562,9 @@ _mesa_delete_texture_image( GLcontext *ctx, struct gl_texture_image *texImage )
 GLboolean
 _mesa_is_proxy_texture(GLenum target)
 {
+   /* NUM_TEXTURE_TARGETS should match number of terms below */
+   assert(NUM_TEXTURE_TARGETS == 7);
+
    return (target == GL_PROXY_TEXTURE_1D ||
            target == GL_PROXY_TEXTURE_2D ||
            target == GL_PROXY_TEXTURE_3D ||
@@ -864,74 +641,28 @@ _mesa_select_tex_object(GLcontext *ctx, const struct gl_texture_unit *texUnit,
 
 
 /**
- * Get the texture image struct which corresponds to target and level
- * of the given texture unit.
+ * Get a texture image pointer from a texture object, given a texture
+ * target and mipmap level.  The target and level parameters should
+ * have already been error-checked.
  *
  * \param ctx GL context.
  * \param texObj texture unit.
  * \param target texture target.
  * \param level image level.
  *
- * \return pointer to the texture image structure on success, or NULL on failure.
- *
- * \sa gl_texture_unit.
+ * \return pointer to the texture image structure, or NULL on failure.
  */
 struct gl_texture_image *
 _mesa_select_tex_image(GLcontext *ctx, const struct gl_texture_object *texObj,
 		       GLenum target, GLint level)
 {
-   ASSERT(texObj);
-
-   if (level < 0 || level >= MAX_TEXTURE_LEVELS) 
-      return NULL;
-
-   /* XXX simplify this with _mesa_tex_target_to_face() */
-   switch (target) {
-      case GL_TEXTURE_1D:
-      case GL_PROXY_TEXTURE_1D:
-      case GL_TEXTURE_2D:
-      case GL_PROXY_TEXTURE_2D:
-      case GL_TEXTURE_3D:
-      case GL_PROXY_TEXTURE_3D:
-         return texObj->Image[0][level];
-
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: 
-         if (ctx->Extensions.ARB_texture_cube_map) {
-	    GLuint face = ((GLuint) target - 
-			   (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X);
-            return texObj->Image[face][level];
-	 }
-         else
-            return NULL;
-
-      case GL_PROXY_TEXTURE_CUBE_MAP_ARB:
-         if (ctx->Extensions.ARB_texture_cube_map)
-            return texObj->Image[0][level];
-         else
-            return NULL;
-
-      case GL_TEXTURE_RECTANGLE_NV:
-      case GL_PROXY_TEXTURE_RECTANGLE_NV:
-         if (ctx->Extensions.NV_texture_rectangle && level == 0) 
-            return texObj->Image[0][level];
-         else 
-            return NULL;
+   const GLuint face = _mesa_tex_target_to_face(target);
 
-      case GL_TEXTURE_1D_ARRAY_EXT:
-      case GL_PROXY_TEXTURE_1D_ARRAY_EXT:
-      case GL_TEXTURE_2D_ARRAY_EXT:
-      case GL_PROXY_TEXTURE_2D_ARRAY_EXT:
-         return (ctx->Extensions.MESA_texture_array)
-            ? texObj->Image[0][level] : NULL;
+   ASSERT(texObj);
+   ASSERT(level >= 0);
+   ASSERT(level < MAX_TEXTURE_LEVELS);
 
-      default:
-         return NULL;
-   }
+   return texObj->Image[face][level];
 }
 
 
@@ -1053,10 +784,6 @@ _mesa_max_texture_levels(GLcontext *ctx, GLenum target)
    case GL_PROXY_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_PROXY_TEXTURE_2D:
-   case GL_TEXTURE_1D_ARRAY_EXT:
-   case GL_PROXY_TEXTURE_1D_ARRAY_EXT:
-   case GL_TEXTURE_2D_ARRAY_EXT:
-   case GL_PROXY_TEXTURE_2D_ARRAY_EXT:
       return ctx->Const.MaxTextureLevels;
    case GL_TEXTURE_3D:
    case GL_PROXY_TEXTURE_3D:
@@ -1069,10 +796,17 @@ _mesa_max_texture_levels(GLcontext *ctx, GLenum target)
    case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB:
    case GL_TEXTURE_CUBE_MAP_ARB:
    case GL_PROXY_TEXTURE_CUBE_MAP_ARB:
-      return ctx->Const.MaxCubeTextureLevels;
+      return ctx->Extensions.ARB_texture_cube_map
+         ? ctx->Const.MaxCubeTextureLevels : 0;
    case GL_TEXTURE_RECTANGLE_NV:
    case GL_PROXY_TEXTURE_RECTANGLE_NV:
-      return 1;
+      return ctx->Extensions.NV_texture_rectangle ? 1 : 0;
+   case GL_TEXTURE_1D_ARRAY_EXT:
+   case GL_PROXY_TEXTURE_1D_ARRAY_EXT:
+   case GL_TEXTURE_2D_ARRAY_EXT:
+   case GL_PROXY_TEXTURE_2D_ARRAY_EXT:
+      return ctx->Extensions.MESA_texture_array
+         ? ctx->Const.MaxTextureLevels : 0;
    default:
       return 0; /* bad target */
    }
@@ -1273,6 +1007,23 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target,
 
 
 /**
+ * Free and clear fields of the gl_texture_image struct.
+ *
+ * \param ctx GL context.
+ * \param texImage texture image structure to be cleared.
+ *
+ * After the call, \p texImage will have no data associated with it.  Its
+ * fields are cleared so that its parent object will test incomplete.
+ */
+void
+_mesa_clear_texture_image(GLcontext *ctx, struct gl_texture_image *texImage)
+{
+   ctx->Driver.FreeTexImageData(ctx, texImage);
+   clear_teximage_fields(texImage);
+}
+
+
+/**
  * This is the fallback for Driver.TestProxyTexImage().  Test the texture
  * level, width, height and depth against the ctx->Const limits for textures.
  *
@@ -1586,13 +1337,13 @@ texture_error_check( GLcontext *ctx, GLenum target,
 
    /* make sure internal format and format basically agree */
    colorFormat = _mesa_is_color_format(format);
-   indexFormat = is_index_format(format);
+   indexFormat = _mesa_is_index_format(format);
    if ((_mesa_is_color_format(internalFormat) && !colorFormat && !indexFormat) ||
-       (is_index_format(internalFormat) && !indexFormat) ||
-       (is_depth_format(internalFormat) != is_depth_format(format)) ||
-       (is_ycbcr_format(internalFormat) != is_ycbcr_format(format)) ||
-       (is_depthstencil_format(internalFormat) != is_depthstencil_format(format)) ||
-       (is_dudv_format(internalFormat) != is_dudv_format(format))) {
+       (_mesa_is_index_format(internalFormat) && !indexFormat) ||
+       (_mesa_is_depth_format(internalFormat) != _mesa_is_depth_format(format)) ||
+       (_mesa_is_ycbcr_format(internalFormat) != _mesa_is_ycbcr_format(format)) ||
+       (_mesa_is_depthstencil_format(internalFormat) != _mesa_is_depthstencil_format(format)) ||
+       (_mesa_is_dudv_format(internalFormat) != _mesa_is_dudv_format(format))) {
       if (!isProxy)
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glTexImage%dD(incompatible internalFormat 0x%x, format 0x%x)",
@@ -1926,14 +1677,14 @@ copytexture_error_check( GLcontext *ctx, GLuint dimensions,
       return GL_TRUE;
    }
 
-   /* NOTE: the format and type aren't really significant for
-    * TestProxyTexImage().  Only the internalformat really matters.
    if (!_mesa_source_buffer_exists(ctx, format)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glCopyTexImage%dD(missing readbuffer)", dimensions);
       return GL_TRUE;
    }
 
+   /* NOTE: the format and type aren't really significant for
+    * TestProxyTexImage().  Only the internalformat really matters.
     */
    type = GL_FLOAT;
 
@@ -2027,7 +1778,7 @@ copytexture_error_check( GLcontext *ctx, GLuint dimensions,
          return GL_TRUE;
       }
    }
-   else if (is_depth_format(internalFormat)) {
+   else if (_mesa_is_depth_format(internalFormat)) {
       /* make sure we have depth/stencil buffers */
       if (!ctx->ReadBuffer->_DepthBuffer) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
@@ -2035,7 +1786,7 @@ copytexture_error_check( GLcontext *ctx, GLuint dimensions,
          return GL_TRUE;
       }
    }
-   else if (is_depthstencil_format(internalFormat)) {
+   else if (_mesa_is_depthstencil_format(internalFormat)) {
       /* make sure we have depth/stencil buffers */
       if (!ctx->ReadBuffer->_DepthBuffer || !ctx->ReadBuffer->_StencilBuffer) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
@@ -2260,147 +2011,6 @@ copytexsubimage_error_check2( GLcontext *ctx, GLuint dimensions,
 }
 
 
-/**
- * Get texture image.  Called by glGetTexImage.
- *
- * \param target texture target.
- * \param level image level.
- * \param format pixel data format for returned image.
- * \param type pixel data type for returned image.
- * \param pixels returned pixel data.
- */
-void GLAPIENTRY
-_mesa_GetTexImage( GLenum target, GLint level, GLenum format,
-                   GLenum type, GLvoid *pixels )
-{
-   const struct gl_texture_unit *texUnit;
-   struct gl_texture_object *texObj;
-   struct gl_texture_image *texImage;
-   GLint maxLevels = 0;
-   GET_CURRENT_CONTEXT(ctx);
-   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
-
-   texUnit = &(ctx->Texture.Unit[ctx->Texture.CurrentUnit]);
-   texObj = _mesa_select_tex_object(ctx, texUnit, target);
-   if (!texObj || _mesa_is_proxy_texture(target)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(target)");
-      return;
-   }
-
-   maxLevels = _mesa_max_texture_levels(ctx, target);
-   ASSERT(maxLevels > 0);  /* 0 indicates bad target, caught above */
-
-   if (level < 0 || level >= maxLevels) {
-      _mesa_error( ctx, GL_INVALID_VALUE, "glGetTexImage(level)" );
-      return;
-   }
-
-   if (_mesa_sizeof_packed_type(type) <= 0) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexImage(type)" );
-      return;
-   }
-
-   if (_mesa_components_in_format(format) <= 0 ||
-       format == GL_STENCIL_INDEX) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexImage(format)" );
-      return;
-   }
-
-   if (!ctx->Extensions.EXT_paletted_texture && is_index_format(format)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
-      return;
-   }
-
-   if (!ctx->Extensions.ARB_depth_texture && is_depth_format(format)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
-      return;
-   }
-
-   if (!ctx->Extensions.MESA_ycbcr_texture && is_ycbcr_format(format)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
-      return;
-   }
-
-   if (!ctx->Extensions.EXT_packed_depth_stencil
-       && is_depthstencil_format(format)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
-      return;
-   }
-
-   if (!ctx->Extensions.ATI_envmap_bumpmap
-       && is_dudv_format(format)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
-      return;
-   }
-
-   _mesa_lock_texture(ctx, texObj);
-   {
-      texImage = _mesa_select_tex_image(ctx, texObj, target, level);
-      if (!texImage) {
-	 /* invalid mipmap level, not an error */
-	 goto out;
-      }
-
-
-      /* Make sure the requested image format is compatible with the
-       * texture's format.  Note that a color index texture can be converted
-       * to RGBA so that combo is allowed.
-       */
-      if (_mesa_is_color_format(format)
-	  && !_mesa_is_color_format(texImage->TexFormat->BaseFormat)
-	  && !is_index_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-      else if (is_index_format(format)
-	       && !is_index_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-      else if (is_depth_format(format)
-	       && !is_depth_format(texImage->TexFormat->BaseFormat)
-	       && !is_depthstencil_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-      else if (is_ycbcr_format(format)
-	       && !is_ycbcr_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-      else if (is_depthstencil_format(format)
-	       && !is_depthstencil_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-      else if (is_dudv_format(format)
-	       && !is_dudv_format(texImage->TexFormat->BaseFormat)) {
-	 _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexImage(format mismatch)");
-	 goto out;
-      }
-
-      if (ctx->Pack.BufferObj->Name) {
-	 /* packing texture image into a PBO */
-	 const GLuint dimensions = (target == GL_TEXTURE_3D) ? 3 : 2;
-	 if (!_mesa_validate_pbo_access(dimensions, &ctx->Pack, texImage->Width,
-					texImage->Height, texImage->Depth,
-					format, type, pixels)) {
-	    _mesa_error(ctx, GL_INVALID_OPERATION,
-			"glGetTexImage(invalid PBO access)");
-	    goto out;
-	 }
-      }
-
-      /* typically, this will call _mesa_get_teximage() */
-      ctx->Driver.GetTexImage(ctx, target, level, format, type, pixels,
-			      texObj, texImage);
-
-   }
- out:
-   _mesa_unlock_texture(ctx, texObj);
-}
-
-
 /** Callback info for walking over FBO hash table */
 struct cb_info
 {
@@ -2544,7 +2154,7 @@ _mesa_TexImage1D( GLenum target, GLint level, GLint internalFormat,
       if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
       _mesa_lock_texture(ctx, texObj);
       {
@@ -2652,7 +2262,7 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat,
       if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
       _mesa_lock_texture(ctx, texObj);
       {
@@ -2755,7 +2365,7 @@ _mesa_TexImage3D( GLenum target, GLint level, GLint internalFormat,
       if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
       _mesa_lock_texture(ctx, texObj);
       {
@@ -2861,7 +2471,7 @@ _mesa_TexSubImage1D( GLenum target, GLint level,
    }
 
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    assert(texObj);
 
@@ -2921,7 +2531,7 @@ _mesa_TexSubImage2D( GLenum target, GLint level,
       return;   /* error was detected */
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    _mesa_lock_texture(ctx, texObj);
    {
@@ -2973,7 +2583,7 @@ _mesa_TexSubImage3D( GLenum target, GLint level,
       return;   /* error was detected */
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
    _mesa_lock_texture(ctx, texObj);
@@ -3034,7 +2644,7 @@ _mesa_CopyTexImage1D( GLenum target, GLint level,
                                postConvWidth, 1, border))
       return;
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    _mesa_lock_texture(ctx, texObj);
    {
@@ -3100,7 +2710,7 @@ _mesa_CopyTexImage2D( GLenum target, GLint level, GLenum internalFormat,
                                postConvWidth, postConvHeight, border))
       return;
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
    _mesa_lock_texture(ctx, texObj);
@@ -3160,7 +2770,7 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level,
    if (copytexsubimage_error_check1(ctx, 1, target, level))
       return;
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
    _mesa_lock_texture(ctx, texObj);
@@ -3215,7 +2825,7 @@ _mesa_CopyTexSubImage2D( GLenum target, GLint level,
    if (copytexsubimage_error_check1(ctx, 2, target, level))
       return;
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
    _mesa_lock_texture(ctx, texObj);
@@ -3270,7 +2880,7 @@ _mesa_CopyTexSubImage3D( GLenum target, GLint level,
    if (copytexsubimage_error_check1(ctx, 3, target, level))
       return;
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
    _mesa_lock_texture(ctx, texObj);
@@ -3524,7 +3134,7 @@ _mesa_CompressedTexImage1DARB(GLenum target, GLint level,
          return;
       }
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
       _mesa_lock_texture(ctx, texObj);
@@ -3578,7 +3188,7 @@ _mesa_CompressedTexImage1DARB(GLenum target, GLint level,
          struct gl_texture_unit *texUnit;
          struct gl_texture_object *texObj;
          struct gl_texture_image *texImage;
-         texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+         texUnit = _mesa_get_current_tex_unit(ctx);
 	 texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
 	 _mesa_lock_texture(ctx, texObj);
@@ -3621,7 +3231,7 @@ _mesa_CompressedTexImage2DARB(GLenum target, GLint level,
          return;
       }
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
       _mesa_lock_texture(ctx, texObj);
@@ -3677,7 +3287,7 @@ _mesa_CompressedTexImage2DARB(GLenum target, GLint level,
          struct gl_texture_unit *texUnit;
          struct gl_texture_object *texObj;
          struct gl_texture_image *texImage;
-         texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+         texUnit = _mesa_get_current_tex_unit(ctx);
 	 texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
 	 _mesa_lock_texture(ctx, texObj);
@@ -3717,7 +3327,7 @@ _mesa_CompressedTexImage3DARB(GLenum target, GLint level,
          return;
       }
 
-      texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+      texUnit = _mesa_get_current_tex_unit(ctx);
       texObj = _mesa_select_tex_object(ctx, texUnit, target);
       _mesa_lock_texture(ctx, texObj);
       {
@@ -3771,7 +3381,7 @@ _mesa_CompressedTexImage3DARB(GLenum target, GLint level,
          struct gl_texture_unit *texUnit;
          struct gl_texture_object *texObj;
          struct gl_texture_image *texImage;
-         texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+         texUnit = _mesa_get_current_tex_unit(ctx);
 	 texObj = _mesa_select_tex_object(ctx, texUnit, target);
 	 _mesa_lock_texture(ctx, texObj);
 	 {
@@ -3810,7 +3420,7 @@ _mesa_CompressedTexSubImage1DARB(GLenum target, GLint level, GLint xoffset,
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    _mesa_lock_texture(ctx, texObj);
    {
@@ -3867,7 +3477,7 @@ _mesa_CompressedTexSubImage2DARB(GLenum target, GLint level, GLint xoffset,
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    _mesa_lock_texture(ctx, texObj);
    {
@@ -3924,7 +3534,7 @@ _mesa_CompressedTexSubImage3DARB(GLenum target, GLint level, GLint xoffset,
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
    _mesa_lock_texture(ctx, texObj);
    {
@@ -3961,55 +3571,3 @@ _mesa_CompressedTexSubImage3DARB(GLenum target, GLint level, GLint xoffset,
 }
 
 
-void GLAPIENTRY
-_mesa_GetCompressedTexImageARB(GLenum target, GLint level, GLvoid *img)
-{
-   const struct gl_texture_unit *texUnit;
-   struct gl_texture_object *texObj;
-   struct gl_texture_image *texImage;
-   GLint maxLevels;
-   GET_CURRENT_CONTEXT(ctx);
-   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
-
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-   texObj = _mesa_select_tex_object(ctx, texUnit, target);
-   if (!texObj) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetCompressedTexImageARB");
-      return;
-   }
-
-   maxLevels = _mesa_max_texture_levels(ctx, target);
-   ASSERT(maxLevels > 0); /* 0 indicates bad target, caught above */
-
-   if (level < 0 || level >= maxLevels) {
-      _mesa_error(ctx, GL_INVALID_VALUE, "glGetCompressedTexImageARB(level)");
-      return;
-   }
-
-   if (_mesa_is_proxy_texture(target)) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetCompressedTexImageARB(target)");
-      return;
-   }
-
-   _mesa_lock_texture(ctx, texObj);
-   {
-      texImage = _mesa_select_tex_image(ctx, texObj, target, level);
-      if (texImage) {
-         if (texImage->IsCompressed) {
-            /* this typically calls _mesa_get_compressed_teximage() */
-            ctx->Driver.GetCompressedTexImage(ctx, target, level, img,
-                                              texObj, texImage);
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_OPERATION,
-                        "glGetCompressedTexImageARB");
-         }
-      }
-      else {
-         /* probably invalid mipmap level */
-         _mesa_error(ctx, GL_INVALID_VALUE,
-                     "glGetCompressedTexImageARB(level)");
-      }
-   }
-   _mesa_unlock_texture(ctx, texObj);
-}
diff --git a/src/mesa/main/teximage.h b/src/mesa/main/teximage.h
index eb60a1fa8f..094177da79 100644
--- a/src/mesa/main/teximage.h
+++ b/src/mesa/main/teximage.h
@@ -73,6 +73,10 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target,
 
 
 extern void
+_mesa_clear_texture_image(GLcontext *ctx, struct gl_texture_image *texImage);
+
+
+extern void
 _mesa_set_tex_image(struct gl_texture_object *tObj,
                     GLenum target, GLint level,
                     struct gl_texture_image *texImage);
@@ -111,10 +115,6 @@ extern GLuint
 _mesa_tex_target_to_face(GLenum target);
 
 
-extern GLboolean
-_mesa_is_color_format(GLenum format);
-
-
 /**
  * Lock a texture for updating.  See also _mesa_lock_context_textures().
  */
@@ -164,11 +164,6 @@ _mesa_TexImage3DEXT( GLenum target, GLint level, GLenum internalformat,
 
 
 extern void GLAPIENTRY
-_mesa_GetTexImage( GLenum target, GLint level,
-                   GLenum format, GLenum type, GLvoid *pixels );
-
-
-extern void GLAPIENTRY
 _mesa_TexSubImage1D( GLenum target, GLint level, GLint xoffset,
                      GLsizei width,
                      GLenum format, GLenum type,
@@ -260,9 +255,6 @@ _mesa_CompressedTexSubImage3DARB(GLenum target, GLint level, GLint xoffset,
                                  GLsizei height, GLsizei depth, GLenum format,
                                  GLsizei imageSize, const GLvoid *data);
 
-extern void GLAPIENTRY
-_mesa_GetCompressedTexImageARB(GLenum target, GLint lod, GLvoid *img);
-
 /*@}*/
 
 #endif
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index 89ee69770e..d09c439250 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -261,6 +261,32 @@ _mesa_copy_texture_object( struct gl_texture_object *dest,
 
 
 /**
+ * Clear all texture images of the given texture object.
+ *
+ * \param ctx GL context.
+ * \param t texture object.
+ *
+ * \sa _mesa_clear_texture_image().
+ */
+void
+_mesa_clear_texture_object(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+   GLuint i, j;
+
+   if (texObj->Target == 0)
+      return;
+
+   for (i = 0; i < MAX_FACES; i++) {
+      for (j = 0; j < MAX_TEXTURE_LEVELS; j++) {
+         struct gl_texture_image *texImage = texObj->Image[i][j];
+         if (texImage)
+            _mesa_clear_texture_image(ctx, texImage);
+      }
+   }
+}
+
+
+/**
  * Check if the given texture object is valid by examining its Target field.
  * For debugging only.
  */
@@ -281,7 +307,8 @@ valid_texture_object(const struct gl_texture_object *tex)
       _mesa_problem(NULL, "invalid reference to a deleted texture object");
       return GL_FALSE;
    default:
-      _mesa_problem(NULL, "invalid texture object Target value");
+      _mesa_problem(NULL, "invalid texture object Target 0x%x, Id = %u",
+                    tex->Target, tex->Name);
       return GL_FALSE;
    }
 }
@@ -307,7 +334,7 @@ _mesa_reference_texobj(struct gl_texture_object **ptr,
       GLboolean deleteFlag = GL_FALSE;
       struct gl_texture_object *oldTex = *ptr;
 
-      assert(valid_texture_object(oldTex));
+      ASSERT(valid_texture_object(oldTex));
 
       _glthread_LOCK_MUTEX(oldTex->Mutex);
       ASSERT(oldTex->RefCount > 0);
@@ -330,7 +357,7 @@ _mesa_reference_texobj(struct gl_texture_object **ptr,
 
    if (tex) {
       /* reference new texture */
-      assert(valid_texture_object(tex));
+      ASSERT(valid_texture_object(tex));
       _glthread_LOCK_MUTEX(tex->Mutex);
       if (tex->RefCount == 0) {
          /* this texture's being deleted (look just above) */
@@ -664,6 +691,24 @@ _mesa_test_texobj_completeness( const GLcontext *ctx,
 
 
 /**
+ * Mark a texture object dirty.  It forces the object to be incomplete
+ * and optionally forces the context to re-validate its state.
+ *
+ * \param ctx GL context.
+ * \param texObj texture object.
+ * \param invalidate_state also invalidate context state.
+ */
+void
+_mesa_dirty_texobj(GLcontext *ctx, struct gl_texture_object *texObj,
+                   GLboolean invalidate_state)
+{
+   texObj->_Complete = GL_FALSE;
+   if (invalidate_state)
+      ctx->NewState |= _NEW_TEXTURE;
+}
+
+
+/**
  * Return pointer to a default/fallback texture.
  * The texture is a 2D 8x8 RGBA texture with all texels = (0,0,0,1).
  * That's the value a sampler should get when sampling from an
@@ -714,7 +759,6 @@ _mesa_get_fallback_texture(GLcontext *ctx)
 }
 
 
-
 /*@}*/
 
 
@@ -1159,10 +1203,9 @@ _mesa_IsTexture( GLuint texture )
 
 
 /**
- * Simplest implementation of texture locking: Grab the a new mutex in
- * the shared context.  Examine the shared context state timestamp and
- * if there has been a change, set the appropriate bits in
- * ctx->NewState.
+ * Simplest implementation of texture locking: grab the shared tex
+ * mutex.  Examine the shared context state timestamp and if there has
+ * been a change, set the appropriate bits in ctx->NewState.
  *
  * This is used to deal with synchronizing things when a texture object
  * is used/modified by different contexts (or threads) which are sharing
diff --git a/src/mesa/main/texobj.h b/src/mesa/main/texobj.h
index 2599c0816a..9bfebd45c8 100644
--- a/src/mesa/main/texobj.h
+++ b/src/mesa/main/texobj.h
@@ -58,6 +58,9 @@ _mesa_copy_texture_object( struct gl_texture_object *dest,
                            const struct gl_texture_object *src );
 
 extern void
+_mesa_clear_texture_object(GLcontext *ctx, struct gl_texture_object *obj);
+
+extern void
 _mesa_reference_texobj(struct gl_texture_object **ptr,
                        struct gl_texture_object *tex);
 
@@ -65,6 +68,10 @@ extern void
 _mesa_test_texobj_completeness( const GLcontext *ctx,
                                 struct gl_texture_object *obj );
 
+extern void
+_mesa_dirty_texobj(GLcontext *ctx, struct gl_texture_object *texObj,
+                   GLboolean invalidate_state);
+
 extern struct gl_texture_object *
 _mesa_get_fallback_texture(GLcontext *ctx);
 
@@ -76,7 +83,6 @@ _mesa_lock_context_textures( GLcontext *ctx );
 
 /*@}*/
 
-
 /**
  * \name API functions
  */
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index 35c3c920e7..b2fbe2205b 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -38,6 +38,7 @@
 #include "main/texcompress.h"
 #include "main/texparam.h"
 #include "main/teximage.h"
+#include "main/texstate.h"
 #include "shader/prog_instruction.h"
 
 
@@ -88,7 +89,7 @@ get_texobj(GLcontext *ctx, GLenum target)
       return NULL;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    switch (target) {
    case GL_TEXTURE_1D:
@@ -469,8 +470,10 @@ set_tex_parameterf(GLcontext *ctx,
          return GL_TRUE;
       }
       else {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
+         static GLuint count = 0;
+         if (count++ < 10)
+            _mesa_error(ctx, GL_INVALID_ENUM,
+                        "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
       }
       return GL_FALSE;
 
@@ -714,44 +717,6 @@ _mesa_GetTexLevelParameterfv( GLenum target, GLint level,
 }
 
 
-static GLuint
-tex_image_dimensions(GLcontext *ctx, GLenum target)
-{
-   switch (target) {
-      case GL_TEXTURE_1D:
-      case GL_PROXY_TEXTURE_1D:
-         return 1;
-      case GL_TEXTURE_2D:
-      case GL_PROXY_TEXTURE_2D:
-         return 2;
-      case GL_TEXTURE_3D:
-      case GL_PROXY_TEXTURE_3D:
-         return 3;
-      case GL_TEXTURE_CUBE_MAP:
-      case GL_PROXY_TEXTURE_CUBE_MAP:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-      case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-      case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-         return ctx->Extensions.ARB_texture_cube_map ? 2 : 0;
-      case GL_TEXTURE_RECTANGLE_NV:
-      case GL_PROXY_TEXTURE_RECTANGLE_NV:
-         return ctx->Extensions.NV_texture_rectangle ? 2 : 0;
-      case GL_TEXTURE_1D_ARRAY_EXT:
-      case GL_PROXY_TEXTURE_1D_ARRAY_EXT:
-         return ctx->Extensions.MESA_texture_array ? 2 : 0;
-      case GL_TEXTURE_2D_ARRAY_EXT:
-      case GL_PROXY_TEXTURE_2D_ARRAY_EXT:
-         return ctx->Extensions.MESA_texture_array ? 3 : 0;
-      default:
-         _mesa_problem(ctx, "bad target in _mesa_tex_target_dimensions()");
-         return 0;
-   }
-}
-
-
 void GLAPIENTRY
 _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
                               GLenum pname, GLint *params )
@@ -759,7 +724,6 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
    const struct gl_texture_unit *texUnit;
    struct gl_texture_object *texObj;
    const struct gl_texture_image *img = NULL;
-   GLuint dimensions;
    GLboolean isProxy;
    GLint maxLevels;
    GET_CURRENT_CONTEXT(ctx);
@@ -771,19 +735,13 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    /* this will catch bad target values */
-   dimensions = tex_image_dimensions(ctx, target);  /* 1, 2 or 3 */
-   if (dimensions == 0) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexLevelParameter[if]v(target)");
-      return;
-   }
-
    maxLevels = _mesa_max_texture_levels(ctx, target);
    if (maxLevels == 0) {
-      /* should not happen since <target> was just checked above */
-      _mesa_problem(ctx, "maxLevels=0 in _mesa_GetTexLevelParameter");
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glGetTexLevelParameter[if]v(target=0x%x)", target);
       return;
    }
 
@@ -880,7 +838,8 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
                         "glGetTexLevelParameter[if]v(pname)");
          break;
       case GL_TEXTURE_STENCIL_SIZE_EXT:
-         if (ctx->Extensions.EXT_packed_depth_stencil) {
+         if (ctx->Extensions.EXT_packed_depth_stencil ||
+             ctx->Extensions.ARB_framebuffer_object) {
             *params = img->TexFormat->StencilBits;
          }
          else {
@@ -999,7 +958,7 @@ _mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params )
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    obj = _mesa_select_tex_object(ctx, texUnit, target);
    if (!obj) {
@@ -1166,7 +1125,7 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
       return;
    }
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   texUnit = _mesa_get_current_tex_unit(ctx);
 
    obj = _mesa_select_tex_object(ctx, texUnit, target);
    if (!obj) {
diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c
index 6e0c0c688a..861c5f37c4 100644
--- a/src/mesa/main/texstate.c
+++ b/src/mesa/main/texstate.c
@@ -729,14 +729,7 @@ init_texture_unit( GLcontext *ctx, GLuint unit )
    ASSIGN_4V( texUnit->GenT.EyePlane, 0.0, 1.0, 0.0, 0.0 );
    ASSIGN_4V( texUnit->GenR.EyePlane, 0.0, 0.0, 0.0, 0.0 );
    ASSIGN_4V( texUnit->GenQ.EyePlane, 0.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenS.ObjectPlane, 1.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenT.ObjectPlane, 0.0, 1.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenR.ObjectPlane, 0.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenQ.ObjectPlane, 0.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenS.EyePlane, 1.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenT.EyePlane, 0.0, 1.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenR.EyePlane, 0.0, 0.0, 0.0, 0.0 );
-   ASSIGN_4V( texUnit->GenQ.EyePlane, 0.0, 0.0, 0.0, 0.0 );
+
    /* no mention of this in spec, but maybe id matrix expected? */
    ASSIGN_4V( texUnit->RotMatrix, 1.0, 0.0, 0.0, 1.0 );
 
diff --git a/src/mesa/main/texstate.h b/src/mesa/main/texstate.h
index a7d7088c62..17ac68000c 100644
--- a/src/mesa/main/texstate.h
+++ b/src/mesa/main/texstate.h
@@ -35,6 +35,18 @@
 #include "mtypes.h"
 
 
+/**
+ * Return pointer to current texture unit.
+ * This the texture unit set by glActiveTexture(), not glClientActiveTexture().
+ */
+static INLINE struct gl_texture_unit *
+_mesa_get_current_tex_unit(GLcontext *ctx)
+{
+   ASSERT(ctx->Texture.CurrentUnit < Elements(ctx->Texture.Unit));
+   return &(ctx->Texture.Unit[ctx->Texture.CurrentUnit]);
+}
+
+
 extern void
 _mesa_copy_texture_state( const GLcontext *src, GLcontext *dst );
 
@@ -48,16 +60,14 @@ _mesa_print_texunit_state( GLcontext *ctx, GLuint unit );
  */
 /*@{*/
 
-
-/*
- * GL_ARB_multitexture
- */
 extern void GLAPIENTRY
 _mesa_ActiveTextureARB( GLenum target );
 
 extern void GLAPIENTRY
 _mesa_ClientActiveTextureARB( GLenum target );
 
+/*@}*/
+
 
 /**
  * \name Initialization, state maintenance
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index bfced1b3f4..a22db628d3 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -3138,7 +3138,7 @@ _mesa_validate_pbo_teximage(GLcontext *ctx, GLuint dimensions,
 {
    GLubyte *buf;
 
-   if (unpack->BufferObj->Name == 0) {
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
       /* no PBO */
       return pixels;
    }
@@ -3174,7 +3174,7 @@ _mesa_validate_pbo_compressed_teximage(GLcontext *ctx,
 {
    GLubyte *buf;
 
-   if (packing->BufferObj->Name == 0) {
+   if (!_mesa_is_bufferobj(packing->BufferObj)) {
       /* not using a PBO - return pointer unchanged */
       return pixels;
    }
@@ -3204,7 +3204,7 @@ void
 _mesa_unmap_teximage_pbo(GLcontext *ctx,
                          const struct gl_pixelstore_attrib *unpack)
 {
-   if (unpack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                               unpack->BufferObj);
    }
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 8f6621ee45..6cd2a2f4f6 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,6 +30,7 @@
 #include "context.h"
 #include "enable.h"
 #include "enums.h"
+#include "hash.h"
 #include "mtypes.h"
 #include "varray.h"
 #include "arrayobj.h"
@@ -36,26 +38,39 @@
 
 
 /**
- * Update the fields of a vertex array object.
- * We need to do a few special things for arrays that live in
- * vertex buffer objects.
+ * Set the fields of a vertex array.
+ * Also do an error check for GL_ARB_vertex_array_object: check that
+ * all arrays reside in VBOs when using a vertex array object.
  *
  * \param array  the array to update
  * \param dirtyBit  which bit to set in ctx->Array.NewState for this array
  * \param elementSize  size of each array element, in bytes
  * \param size  components per element (1, 2, 3 or 4)
  * \param type  datatype of each component (GL_FLOAT, GL_INT, etc)
+ * \param format  either GL_RGBA or GL_BGRA
  * \param stride  stride between elements, in elements
  * \param normalized  are integer types converted to floats in [-1, 1]?
  * \param ptr  the address (or offset inside VBO) of the array data
+ * \return GL_TRUE if no error, GL_FALSE if error
  */
-static void
+static GLboolean
 update_array(GLcontext *ctx, struct gl_client_array *array,
              GLbitfield dirtyBit, GLsizei elementSize,
              GLint size, GLenum type, GLenum format,
              GLsizei stride, GLboolean normalized, const GLvoid *ptr)
 {
    ASSERT(format == GL_RGBA || format == GL_BGRA);
+
+   if (ctx->Array.ArrayObj->VBOonly &&
+       ctx->Array.ArrayBufferObj->Name == 0) {
+      /* GL_ARB_vertex_array_object requires that all arrays reside in VBOs.
+       * Generate GL_INVALID_OPERATION if that's not true.
+       */
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glVertex/Normal/EtcPointer(non-VBO array)");
+      return GL_FALSE;
+   }
+
    array->Size = size;
    array->Type = type;
    array->Format = format;
@@ -63,24 +78,15 @@ update_array(GLcontext *ctx, struct gl_client_array *array,
    array->StrideB = stride ? stride : elementSize;
    array->Normalized = normalized;
    array->Ptr = (const GLubyte *) ptr;
-#if FEATURE_ARB_vertex_buffer_object
+   array->_ElementSize = elementSize;
+
    _mesa_reference_buffer_object(ctx, &array->BufferObj,
                                  ctx->Array.ArrayBufferObj);
 
-   /* Compute the index of the last array element that's inside the buffer.
-    * Later in glDrawArrays we'll check if start + count > _MaxElement to
-    * be sure we won't go out of bounds.
-    */
-   if (ctx->Array.ArrayBufferObj->Name)
-      array->_MaxElement = ((GLsizeiptrARB) ctx->Array.ArrayBufferObj->Size
-                            - (GLsizeiptrARB) array->Ptr + array->StrideB
-                            - elementSize) / array->StrideB;
-   else
-#endif
-      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
-
    ctx->NewState |= _NEW_ARRAY;
    ctx->Array.NewState |= dirtyBit;
+
+   return GL_TRUE;
 }
 
 
@@ -129,12 +135,14 @@ _mesa_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
          break;
 #endif
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glVertexPointer(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glVertexPointer(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Vertex, _NEW_ARRAY_VERTEX,
-                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Vertex, _NEW_ARRAY_VERTEX,
+                     elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.VertexPointer)
       ctx->Driver.VertexPointer( ctx, size, type, stride, ptr );
@@ -179,12 +187,14 @@ _mesa_NormalPointer(GLenum type, GLsizei stride, const GLvoid *ptr )
          break;
 #endif
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glNormalPointer(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glNormalPointer(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Normal, _NEW_ARRAY_NORMAL,
-                elementSize, 3, type, GL_RGBA, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Normal, _NEW_ARRAY_NORMAL,
+                     elementSize, 3, type, GL_RGBA, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.NormalPointer)
       ctx->Driver.NormalPointer( ctx, type, stride, ptr );
@@ -257,12 +267,14 @@ _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
          break;
 #endif
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glColorPointer(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glColorPointer(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Color, _NEW_ARRAY_COLOR0,
-                elementSize, size, type, format, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Color, _NEW_ARRAY_COLOR0,
+                     elementSize, size, type, format, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.ColorPointer)
       ctx->Driver.ColorPointer( ctx, size, type, stride, ptr );
@@ -293,8 +305,9 @@ _mesa_FogCoordPointerEXT(GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->FogCoord, _NEW_ARRAY_FOGCOORD,
-                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->FogCoord, _NEW_ARRAY_FOGCOORD,
+                     elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.FogCoordPointer)
       ctx->Driver.FogCoordPointer( ctx, type, stride, ptr );
@@ -334,8 +347,9 @@ _mesa_IndexPointer(GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Index, _NEW_ARRAY_INDEX,
-                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Index, _NEW_ARRAY_INDEX,
+                     elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.IndexPointer)
       ctx->Driver.IndexPointer( ctx, type, stride, ptr );
@@ -404,12 +418,15 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type,
          elementSize = size * sizeof(GLdouble);
          break;
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glSecondaryColorPointer(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glSecondaryColorPointer(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->SecondaryColor, _NEW_ARRAY_COLOR1,
-                elementSize, size, type, format, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->SecondaryColor,
+                     _NEW_ARRAY_COLOR1, elementSize, size, type,
+                     format, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.SecondaryColorPointer)
       ctx->Driver.SecondaryColorPointer( ctx, size, type, stride, ptr );
@@ -463,13 +480,15 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride,
          break;
 #endif
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glTexCoordPointer(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glTexCoordPointer(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->TexCoord[unit],
-                _NEW_ARRAY_TEXCOORD(unit),
-                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->TexCoord[unit],
+                     _NEW_ARRAY_TEXCOORD(unit),
+                     elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.TexCoordPointer)
       ctx->Driver.TexCoordPointer( ctx, size, type, stride, ptr );
@@ -487,9 +506,10 @@ _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr)
       return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
-                sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, GL_RGBA,
-                stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
+                     sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, GL_RGBA,
+                     stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.EdgeFlagPointer)
       ctx->Driver.EdgeFlagPointer( ctx, stride, ptr );
@@ -528,6 +548,12 @@ _mesa_PointSizePointer(GLenum type, GLsizei stride, const GLvoid *ptr)
 
 
 #if FEATURE_NV_vertex_program
+/**
+ * Set a vertex attribute array.
+ * Note that these arrays DO alias the conventional GL vertex arrays
+ * (position, normal, color, fog, texcoord, etc).
+ * The generic attribute slots at #16 and above are not touched.
+ */
 void GLAPIENTRY
 _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
                             GLsizei stride, const GLvoid *ptr)
@@ -589,13 +615,15 @@ _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
          elementSize = size * sizeof(GLdouble);
          break;
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glVertexAttribPointerNV(type)" );
+         _mesa_error( ctx, GL_INVALID_ENUM, "glVertexAttribPointerNV(type=%s)",
+                      _mesa_lookup_enum_by_nr(type));
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
-                _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, format, stride, normalized, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
+                     _NEW_ARRAY_ATTRIB(index),
+                     elementSize, size, type, format, stride, normalized, ptr))
+      return;
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer( ctx, index, size, type, stride, ptr );
@@ -604,6 +632,11 @@ _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
 
 
 #if FEATURE_ARB_vertex_program
+/**
+ * Set a generic vertex attribute array.
+ * Note that these arrays DO NOT alias the conventional GL vertex arrays
+ * (position, normal, color, fog, texcoord, etc).
+ */
 void GLAPIENTRY
 _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
                              GLboolean normalized,
@@ -687,9 +720,10 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
-                _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, format, stride, normalized, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
+                     _NEW_ARRAY_ATTRIB(index),
+                     elementSize, size, type, format, stride, normalized, ptr))
+      return;
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer(ctx, index, size, type, stride, ptr);
@@ -1004,24 +1038,6 @@ _mesa_MultiDrawArraysEXT( GLenum mode, GLint *first,
 }
 
 
-/* GL_EXT_multi_draw_arrays */
-void GLAPIENTRY
-_mesa_MultiDrawElementsEXT( GLenum mode, const GLsizei *count, GLenum type,
-                            const GLvoid **indices, GLsizei primcount )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLint i;
-
-   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
-
-   for (i = 0; i < primcount; i++) {
-      if (count[i] > 0) {
-         CALL_DrawElements(ctx->Exec, (mode, count[i], type, indices[i]));
-      }
-   }
-}
-
-
 /* GL_IBM_multimode_draw_arrays */
 void GLAPIENTRY
 _mesa_MultiModeDrawArraysIBM( const GLenum * mode, const GLint * first,
@@ -1065,6 +1081,75 @@ _mesa_MultiModeDrawElementsIBM( const GLenum * mode, const GLsizei * count,
 
 
 /**
+ * Copy one client vertex array to another.
+ */
+void
+_mesa_copy_client_array(GLcontext *ctx,
+                        struct gl_client_array *dst,
+                        struct gl_client_array *src)
+{
+   dst->Size = src->Size;
+   dst->Type = src->Type;
+   dst->Format = src->Format;
+   dst->Stride = src->Stride;
+   dst->StrideB = src->StrideB;
+   dst->Ptr = src->Ptr;
+   dst->Enabled = src->Enabled;
+   dst->Normalized = src->Normalized;
+   dst->_ElementSize = src->_ElementSize;
+   _mesa_reference_buffer_object(ctx, &dst->BufferObj, src->BufferObj);
+   dst->_MaxElement = src->_MaxElement;
+}
+
+
+
+/**
+ * Print vertex array's fields.
+ */
+static void
+print_array(const char *name, GLint index, const struct gl_client_array *array)
+{
+   if (index >= 0)
+      _mesa_printf("  %s[%d]: ", name, index);
+   else
+      _mesa_printf("  %s: ", name);
+   _mesa_printf("Ptr=%p, Type=0x%x, Size=%d, ElemSize=%u, Stride=%d, Buffer=%u(Size %u), MaxElem=%u\n",
+                array->Ptr, array->Type, array->Size,
+                array->_ElementSize, array->StrideB,
+                array->BufferObj->Name, array->BufferObj->Size,
+                array->_MaxElement);
+}
+
+
+/**
+ * Print current vertex object/array info.  For debug.
+ */
+void
+_mesa_print_arrays(GLcontext *ctx)
+{
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   GLuint i;
+
+   _mesa_update_array_object_max_element(ctx, arrayObj);
+
+   _mesa_printf("Array Object %u\n", arrayObj->Name);
+   if (arrayObj->Vertex.Enabled)
+      print_array("Vertex", -1, &arrayObj->Vertex);
+   if (arrayObj->Normal.Enabled)
+      print_array("Normal", -1, &arrayObj->Normal);
+   if (arrayObj->Color.Enabled)
+      print_array("Color", -1, &arrayObj->Color);
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
+      if (arrayObj->TexCoord[i].Enabled)
+         print_array("TexCoord", i, &arrayObj->TexCoord[i]);
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      if (arrayObj->VertexAttrib[i].Enabled)
+         print_array("Attrib", i, &arrayObj->VertexAttrib[i]);
+   _mesa_printf("  _MaxElement = %u\n", arrayObj->_MaxElement);
+}
+
+
+/**
  * Initialize vertex array state for given context.
  */
 void 
@@ -1074,4 +1159,29 @@ _mesa_init_varray(GLcontext *ctx)
    _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj,
                                 ctx->Array.DefaultArrayObj);
    ctx->Array.ActiveTexture = 0;   /* GL_ARB_multitexture */
+
+   ctx->Array.Objects = _mesa_NewHashTable();
+}
+
+
+/**
+ * Callback for deleting an array object.  Called by _mesa_HashDeleteAll().
+ */
+static void
+delete_arrayobj_cb(GLuint id, void *data, void *userData)
+{
+   struct gl_array_object *arrayObj = (struct gl_array_object *) data;
+   GLcontext *ctx = (GLcontext *) userData;
+   _mesa_delete_array_object(ctx, arrayObj);
+}
+
+
+/**
+ * Free vertex array state for given context.
+ */
+void 
+_mesa_free_varray_data(GLcontext *ctx)
+{
+   _mesa_HashDeleteAll(ctx->Array.Objects, delete_arrayobj_cb, ctx);
+   _mesa_DeleteHashTable(ctx->Array.Objects);
 }
diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h
index 97d5c8219d..becc67c29d 100644
--- a/src/mesa/main/varray.h
+++ b/src/mesa/main/varray.h
@@ -1,18 +1,9 @@
-/**
- * \file varray.h
- * Vertex arrays.
- *
- * \if subset
- * (No-op)
- *
- * \endif
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  4.1
+ * Version:  7.6
  *
- * Copyright (C) 1999-2002  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -169,14 +160,26 @@ _mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
                         GLenum type, const GLvoid *indices);
 
 
+extern void
+_mesa_copy_client_array(GLcontext *ctx,
+                        struct gl_client_array *dst,
+                        struct gl_client_array *src);
+
+
+extern void
+_mesa_print_arrays(GLcontext *ctx);
 
 extern void
 _mesa_init_varray( GLcontext * ctx );
 
+extern void 
+_mesa_free_varray_data(GLcontext *ctx);
+
 #else
 
 /** No-op */
 #define _mesa_init_varray( c )  ((void)0)
+#define _mesa_free_varray_data( c )  ((void)0)
 
 #endif
 
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index 995e46d318..d4d3dd1a94 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5.2
+ * Version:  7.6
  *
- * Copyright (C) 1999-2009  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,9 +30,9 @@
 
 /* Mesa version */
 #define MESA_MAJOR 7
-#define MESA_MINOR 5
-#define MESA_PATCH 2
-#define MESA_VERSION_STRING "7.5.2-devel"
+#define MESA_MINOR 6
+#define MESA_PATCH 0
+#define MESA_VERSION_STRING "7.6-devel"
 
 /* To make version comparison easy */
 #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
diff --git a/src/mesa/main/viewport.c b/src/mesa/main/viewport.c
index 50e0402d27..309308c983 100644
--- a/src/mesa/main/viewport.c
+++ b/src/mesa/main/viewport.c
@@ -120,6 +120,10 @@ _mesa_DepthRange(GLclampd nearval, GLclampd farval)
    if (MESA_VERBOSE&VERBOSE_API)
       _mesa_debug(ctx, "glDepthRange %f %f\n", nearval, farval);
 
+   if (ctx->Viewport.Near == nearval &&
+       ctx->Viewport.Far == farval)
+      return;
+
    ctx->Viewport.Near = (GLfloat) CLAMP(nearval, 0.0, 1.0);
    ctx->Viewport.Far = (GLfloat) CLAMP(farval, 0.0, 1.0);
    ctx->NewState |= _NEW_VIEWPORT;
diff --git a/src/mesa/main/vtxfmt.c b/src/mesa/main/vtxfmt.c
index 1f807dc3dc..8d6f560a80 100644
--- a/src/mesa/main/vtxfmt.c
+++ b/src/mesa/main/vtxfmt.c
@@ -133,6 +133,7 @@ install_vtxfmt( struct _glapi_table *tab, const GLvertexformat *vfmt )
    SET_DrawArrays(tab, vfmt->DrawArrays);
    SET_DrawElements(tab, vfmt->DrawElements);
    SET_DrawRangeElements(tab, vfmt->DrawRangeElements);
+   SET_MultiDrawElementsEXT(tab, vfmt->MultiDrawElementsEXT);
    SET_EvalMesh1(tab, vfmt->EvalMesh1);
    SET_EvalMesh2(tab, vfmt->EvalMesh2);
    ASSERT(tab->EvalMesh2);
diff --git a/src/mesa/main/vtxfmt_tmp.h b/src/mesa/main/vtxfmt_tmp.h
index 6f5d01e40f..1308d0aa46 100644
--- a/src/mesa/main/vtxfmt_tmp.h
+++ b/src/mesa/main/vtxfmt_tmp.h
@@ -335,6 +335,17 @@ static void GLAPIENTRY TAG(DrawElements)( GLenum mode, GLsizei count, GLenum typ
    CALL_DrawElements(GET_DISPATCH(), ( mode, count, type, indices ));
 }
 
+static void GLAPIENTRY TAG(MultiDrawElementsEXT)( GLenum mode,
+						  const GLsizei *count,
+						  GLenum type,
+						  const GLvoid **indices,
+						  GLsizei primcount)
+{
+   PRE_LOOPBACK( MultiDrawElementsEXT );
+   CALL_MultiDrawElementsEXT(GET_DISPATCH(), ( mode, count, type, indices,
+					       primcount ));
+}
+
 static void GLAPIENTRY TAG(DrawRangeElements)( GLenum mode, GLuint start,
 				    GLuint end, GLsizei count,
 				    GLenum type, const GLvoid *indices )
@@ -522,6 +533,7 @@ static GLvertexformat TAG(vtxfmt) = {
    TAG(DrawArrays),
    TAG(DrawElements),
    TAG(DrawRangeElements),
+   TAG(MultiDrawElementsEXT),
    TAG(EvalMesh1),
    TAG(EvalMesh2)
 };
diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c
index c5e2fd1de1..4cbab11a35 100644
--- a/src/mesa/math/m_vector.c
+++ b/src/mesa/math/m_vector.c
@@ -1,4 +1,3 @@
-
 /*
  * Mesa 3-D graphics library
  * Version:  3.5
@@ -37,11 +36,12 @@
 
 
 
-/*
+/**
  * Given a vector [count][4] of floats, set all the [][elt] values
  * to 0 (if elt = 0, 1, 2) or 1.0 (if elt = 3).
  */
-void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
+void
+_mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
 {
    static const GLubyte elem_bits[4] = {
       VEC_DIRTY_0,
@@ -54,12 +54,13 @@ void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
    GLfloat (*data)[4] = (GLfloat (*)[4])vec->start;
    GLuint i;
 
-   for (i = 0 ; i < count ; i++)
+   for (i = 0; i < count; i++)
       data[i][elt] = v;
 
    vec->flags &= ~elem_bits[elt];
 }
 
+
 static const GLubyte size_bits[5] = {
    0,
    VEC_SIZE_1,
@@ -69,61 +70,53 @@ static const GLubyte size_bits[5] = {
 };
 
 
-
-/*
+/**
  * Initialize GLvector objects.
- * Input: v - the vector object to initialize.
- *        flags - bitwise-OR of VEC_* flags
- *        storage - pointer to storage for the vector's data
+ * \param v  the vector object to initialize.
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param storage  pointer to storage for the vector's data
  */
-
-
-void _mesa_vector4f_init( GLvector4f *v, GLuint flags, GLfloat (*storage)[4] )
+void
+_mesa_vector4f_init( GLvector4f *v, GLbitfield flags, GLfloat (*storage)[4] )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;   /* may change: 2-4 for vertices and 1-4 for texcoords */
    v->data = storage;
    v->start = (GLfloat *) storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags ;
+   v->flags = size_bits[4] | flags;
 }
 
 
-
-
-/*
+/**
  * Initialize GLvector objects and allocate storage.
- * Input: v - the vector object
- *        sz - unused????
- *        flags - bitwise-OR of VEC_* flags
- *        count - number of elements to allocate in vector
- *        alignment - desired memory alignment for the data (in bytes)
+ * \param v  the vector object
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param count  number of elements to allocate in vector
+ * \param alignment  desired memory alignment for the data (in bytes)
  */
-
-
-void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags, GLuint count,
-			GLuint alignment )
+void
+_mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags, GLuint count,
+                      GLuint alignment )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;
    v->storage = ALIGN_MALLOC( count * 4 * sizeof(GLfloat), alignment );
+   v->storage_count = count;
    v->start = (GLfloat *) v->storage;
    v->data = (GLfloat (*)[4]) v->storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags | VEC_MALLOC ;
+   v->flags = size_bits[4] | flags | VEC_MALLOC;
 }
 
 
-
-
-/*
+/**
  * Vector deallocation.  Free whatever memory is pointed to by the
  * vector's storage field if the VEC_MALLOC flag is set.
  * DO NOT free the GLvector object itself, though.
  */
-
-
-void _mesa_vector4f_free( GLvector4f *v )
+void
+_mesa_vector4f_free( GLvector4f *v )
 {
    if (v->flags & VEC_MALLOC) {
       ALIGN_FREE( v->storage );
@@ -135,13 +128,15 @@ void _mesa_vector4f_free( GLvector4f *v )
 }
 
 
-/*
+/**
  * For debugging
  */
-void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
+void
+_mesa_vector4f_print( const GLvector4f *v, const GLubyte *cullmask,
+                      GLboolean culling )
 {
-   GLfloat c[4] = { 0, 0, 0, 1 };
-   const char *templates[5] = {
+   static const GLfloat c[4] = { 0, 0, 0, 1 };
+   static const char *templates[5] = {
       "%d:\t0, 0, 0, 1\n",
       "%d:\t%f, 0, 0, 1\n",
       "%d:\t%f, %f, 0, 1\n",
@@ -154,30 +149,32 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
    GLuint j, i = 0, count;
 
    _mesa_printf("data-start\n");
-   for ( ; d != v->start ; STRIDE_F(d, v->stride), i++)
+   for (; d != v->start; STRIDE_F(d, v->stride), i++)
       _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
 
    _mesa_printf("start-count(%u)\n", v->count);
    count = i + v->count;
 
    if (culling) {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 if (cullmask[i])
 	    _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
    else {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
 
-   for (j = v->size ; j < 4; j++) {
+   for (j = v->size; j < 4; j++) {
       if ((v->flags & (1<<j)) == 0) {
 
 	 _mesa_printf("checking col %u is clean as advertised ", j);
 
-	 for (i = 0, d = (GLfloat *) v->data ;
-	      i < count && d[j] == c[j] ;
-	      i++, STRIDE_F(d, v->stride)) {};
+	 for (i = 0, d = (GLfloat *) v->data;
+	      i < count && d[j] == c[j];
+	      i++, STRIDE_F(d, v->stride)) {
+            /* no-op */
+         }
 
 	 if (i == count)
 	    _mesa_printf(" --> ok\n");
@@ -186,5 +183,3 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
       }
    }
 }
-
-
diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h
index 647388ac7d..71281d5758 100644
--- a/src/mesa/math/m_vector.h
+++ b/src/mesa/math/m_vector.h
@@ -31,7 +31,6 @@
 #define _M_VECTOR_H_
 
 #include "main/glheader.h"
-#include "main/mtypes.h"		/* hack for GLchan */
 
 
 #define VEC_DIRTY_0        0x1
@@ -50,7 +49,8 @@
 
 
 
-/* Wrap all the information about vectors up in a struct.  Has
+/**
+ * Wrap all the information about vectors up in a struct.  Has
  * additional fields compared to the other vectors to help us track of
  * different vertex sizes, and whether we need to clean columns out
  * because they contain non-(0,0,0,1) values.
@@ -60,29 +60,27 @@
  * the transformation routines.
  */
 typedef struct {
-   GLfloat (*data)[4];	/* may be malloc'd or point to client data */
-   GLfloat *start;	/* points somewhere inside of <data> */
-   GLuint count;	/* size of the vector (in elements) */
-   GLuint stride;	/* stride from one element to the next (in bytes) */
-   GLuint size;		/* 2-4 for vertices and 1-4 for texcoords */
-   GLuint flags;	/* which columns are dirty */
-   void *storage;	/* self-allocated storage */
+   GLfloat (*data)[4];	/**< may be malloc'd or point to client data */
+   GLfloat *start;	/**< points somewhere inside of <data> */
+   GLuint count;	/**< size of the vector (in elements) */
+   GLuint stride;	/**< stride from one element to the next (in bytes) */
+   GLuint size;		/**< 2-4 for vertices and 1-4 for texcoords */
+   GLbitfield flags;	/**< bitmask of VEC_x flags */
+   void *storage;	/**< self-allocated storage */
+   GLuint storage_count; /**< storage size in elements */
 } GLvector4f;
 
 
-extern void _mesa_vector4f_init( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_init( GLvector4f *v, GLbitfield flags,
 			      GLfloat (*storage)[4] );
-extern void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags,
 			       GLuint count, GLuint alignment );
 extern void _mesa_vector4f_free( GLvector4f *v );
-extern void _mesa_vector4f_print( GLvector4f *v, GLubyte *, GLboolean );
+extern void _mesa_vector4f_print( const GLvector4f *v, const GLubyte *, GLboolean );
 extern void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint nr, GLuint elt );
 
 
-
-
-
-/*
+/**
  * Given vector <v>, return a pointer (cast to <type *> to the <i>-th element.
  *
  * End up doing a lot of slow imuls if not careful.
diff --git a/src/mesa/shader/.gitignore b/src/mesa/shader/.gitignore
new file mode 100644
index 0000000000..086fd9a705
--- /dev/null
+++ b/src/mesa/shader/.gitignore
@@ -0,0 +1 @@
+program_parse.output
diff --git a/src/mesa/shader/Makefile b/src/mesa/shader/Makefile
new file mode 100644
index 0000000000..400a543bda
--- /dev/null
+++ b/src/mesa/shader/Makefile
@@ -0,0 +1,7 @@
+all: program_parse.tab.c lex.yy.c
+
+program_parse.tab.c program_parse.tab.h: program_parse.y
+	bison -v -d $<
+
+lex.yy.c: program_lexer.l
+	flex --never-interactive $<
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 8607940dc1..05ee4f563e 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -63,3817 +63,7 @@ having three separate program parameter arrays.
 #include "prog_parameter.h"
 #include "prog_statevars.h"
 #include "prog_instruction.h"
-
-/**
- * This is basically a union of the vertex_program and fragment_program
- * structs that we can use to parse the program into
- *
- * XXX we can probably get rid of this entirely someday.
- */
-struct arb_program
-{
-   struct gl_program Base;
-
-   GLuint Position;       /* Just used for error reporting while parsing */
-   GLuint MajorVersion;
-   GLuint MinorVersion;
-
-   /* ARB_vertex_progmra options */
-   GLboolean HintPositionInvariant;
-
-   /* ARB_fragment_progmra options */
-   GLenum PrecisionOption; /* GL_DONT_CARE, GL_NICEST or GL_FASTEST */
-   GLenum FogOption;       /* GL_NONE, GL_LINEAR, GL_EXP or GL_EXP2 */
-
-   /* ARB_fragment_program specifics */
-   GLbitfield TexturesUsed[MAX_TEXTURE_IMAGE_UNITS]; 
-   GLbitfield ShadowSamplers;
-   GLuint NumAluInstructions; 
-   GLuint NumTexInstructions;
-   GLuint NumTexIndirections;
-
-   GLboolean UsesKill;
-};
-
-
-
-/* TODO:
- *    Fragment Program Stuff:
- *    -----------------------------------------------------
- *
- *    - things from Michal's email
- *       + overflow on atoi
- *       + not-overflowing floats (don't use parse_integer..)
- *       + can remove range checking in arbparse.c
- *
- *    - check all limits of number of various variables
- *      + parameters
- *
- *    - test! test! test!
- *
- *    Vertex Program Stuff:
- *    -----------------------------------------------------
- *    - Optimize param array usage and count limits correctly, see spec,
- *         section 2.14.3.7
- *       + Record if an array is reference absolutly or relatively (or both)
- *       + For absolute arrays, store a bitmap of accesses
- *       + For single parameters, store an access flag
- *       + After parsing, make a parameter cleanup and merging pass, where
- *           relative arrays are layed out first, followed by abs arrays, and
- *           finally single state.
- *       + Remap offsets for param src and dst registers
- *       + Now we can properly count parameter usage
- *
- *    - Multiple state binding errors in param arrays (see spec, just before
- *         section 2.14.3.3)
- *    - grep for XXX
- *
- *    Mesa Stuff
- *    -----------------------------------------------------
- *    - User clipping planes vs. PositionInvariant
- *    - Is it sufficient to just multiply by the mvp to transform in the
- *        PositionInvariant case? Or do we need something more involved?
- *
- *    - vp_src swizzle is GLubyte, fp_src swizzle is GLuint
- *    - fetch state listed in program_parameters list
- *       + WTF should this go???
- *       + currently in nvvertexec.c and s_nvfragprog.c
- *
- *    - allow for multiple address registers (and fetch address regs properly)
- *
- *    Cosmetic Stuff
- *    -----------------------------------------------------
- * 	- remove any leftover unused grammer.c stuff (dict_ ?)
- * 	- fix grammer.c error handling so its not static
- * 	- #ifdef around stuff pertaining to extentions
- *
- *    Outstanding Questions:
- *    -----------------------------------------------------
- *    - ARB_matrix_palette / ARB_vertex_blend -- not supported
- *      what gets hacked off because of this:
- *       + VERTEX_ATTRIB_MATRIXINDEX
- *       + VERTEX_ATTRIB_WEIGHT
- *       + MATRIX_MODELVIEW
- *       + MATRIX_PALETTE
- *
- *    - When can we fetch env/local params from their own register files, and
- *      when to we have to fetch them into the main state register file?
- *      (think arrays)
- *
- *    Grammar Changes:
- *    -----------------------------------------------------
- */
-
-/* Changes since moving the file to shader directory
-
-2004-III-4 ------------------------------------------------------------
-- added #include "grammar_mesa.h"
-- removed grammar specific code part (it resides now in grammar.c)
-- added GL_ARB_fragment_program_shadow tokens
-- modified #include "arbparse_syn.h"
-- major changes inside _mesa_parse_arb_program()
-- check the program string for '\0' characters
-- copy the program string to a one-byte-longer location to have
-  it null-terminated
-- position invariance test (not writing to result.position) moved
-  to syntax part
-*/
-
-typedef GLubyte *production;
-
-
-/**
- * This is the text describing the rules to parse the grammar
- */
-LONGSTRING static char arb_grammar_text[] =
-#include "arbprogram_syn.h"
-;
-
-/**
- * These should match up with the values defined in arbprogram.syn
- */
-
-/*
-    Changes:
-    - changed and merged V_* and F_* opcode values to OP_*.
-    - added GL_ARB_fragment_program_shadow specific tokens (michal)
-*/
-#define  REVISION                                   0x0a
-
-/* program type */
-#define  FRAGMENT_PROGRAM                           0x01
-#define  VERTEX_PROGRAM                             0x02
-
-/* program section */
-#define  OPTION                                     0x01
-#define  INSTRUCTION                                0x02
-#define  DECLARATION                                0x03
-#define  END                                        0x04
-
-/* GL_ARB_fragment_program option */
-#define  ARB_PRECISION_HINT_FASTEST                 0x00
-#define  ARB_PRECISION_HINT_NICEST                  0x01
-#define  ARB_FOG_EXP                                0x02
-#define  ARB_FOG_EXP2                               0x03
-#define  ARB_FOG_LINEAR                             0x04
-
-/* GL_ARB_vertex_program option */
-#define  ARB_POSITION_INVARIANT                     0x05
-
-/* GL_ARB_fragment_program_shadow option */
-#define  ARB_FRAGMENT_PROGRAM_SHADOW                0x06
-
-/* GL_ARB_draw_buffers option */
-#define  ARB_DRAW_BUFFERS                           0x07
-
-/* GL_MESA_texture_array option */
-#define  MESA_TEXTURE_ARRAY                        0x08
-
-/* GL_ARB_fragment_program instruction class */
-#define  OP_ALU_INST                                0x00
-#define  OP_TEX_INST                                0x01
-
-/* GL_ARB_vertex_program instruction class */
-/*       OP_ALU_INST */
-
-/* GL_ARB_fragment_program instruction type */
-#define  OP_ALU_VECTOR                               0x00
-#define  OP_ALU_SCALAR                               0x01
-#define  OP_ALU_BINSC                                0x02
-#define  OP_ALU_BIN                                  0x03
-#define  OP_ALU_TRI                                  0x04
-#define  OP_ALU_SWZ                                  0x05
-#define  OP_TEX_SAMPLE                               0x06
-#define  OP_TEX_KIL                                  0x07
-
-/* GL_ARB_vertex_program instruction type */
-#define  OP_ALU_ARL                                  0x08
-/*       OP_ALU_VECTOR */
-/*       OP_ALU_SCALAR */
-/*       OP_ALU_BINSC */
-/*       OP_ALU_BIN */
-/*       OP_ALU_TRI */
-/*       OP_ALU_SWZ */
-
-/* GL_ARB_fragment_program instruction code */
-#define  OP_ABS                                     0x00
-#define  OP_ABS_SAT                                 0x1B
-#define  OP_FLR                                     0x09
-#define  OP_FLR_SAT                                 0x26
-#define  OP_FRC                                     0x0A
-#define  OP_FRC_SAT                                 0x27
-#define  OP_LIT                                     0x0C
-#define  OP_LIT_SAT                                 0x2A
-#define  OP_MOV                                     0x11
-#define  OP_MOV_SAT                                 0x30
-#define  OP_COS                                     0x1F
-#define  OP_COS_SAT                                 0x20
-#define  OP_EX2                                     0x07
-#define  OP_EX2_SAT                                 0x25
-#define  OP_LG2                                     0x0B
-#define  OP_LG2_SAT                                 0x29
-#define  OP_RCP                                     0x14
-#define  OP_RCP_SAT                                 0x33
-#define  OP_RSQ                                     0x15
-#define  OP_RSQ_SAT                                 0x34
-#define  OP_SIN                                     0x38
-#define  OP_SIN_SAT                                 0x39
-#define  OP_SCS                                     0x35
-#define  OP_SCS_SAT                                 0x36
-#define  OP_POW                                     0x13
-#define  OP_POW_SAT                                 0x32
-#define  OP_ADD                                     0x01
-#define  OP_ADD_SAT                                 0x1C
-#define  OP_DP3                                     0x03
-#define  OP_DP3_SAT                                 0x21
-#define  OP_DP4                                     0x04
-#define  OP_DP4_SAT                                 0x22
-#define  OP_DPH                                     0x05
-#define  OP_DPH_SAT                                 0x23
-#define  OP_DST                                     0x06
-#define  OP_DST_SAT                                 0x24
-#define  OP_MAX                                     0x0F
-#define  OP_MAX_SAT                                 0x2E
-#define  OP_MIN                                     0x10
-#define  OP_MIN_SAT                                 0x2F
-#define  OP_MUL                                     0x12
-#define  OP_MUL_SAT                                 0x31
-#define  OP_SGE                                     0x16
-#define  OP_SGE_SAT                                 0x37
-#define  OP_SLT                                     0x17
-#define  OP_SLT_SAT                                 0x3A
-#define  OP_SUB                                     0x18
-#define  OP_SUB_SAT                                 0x3B
-#define  OP_XPD                                     0x1A
-#define  OP_XPD_SAT                                 0x43
-#define  OP_CMP                                     0x1D
-#define  OP_CMP_SAT                                 0x1E
-#define  OP_LRP                                     0x2B
-#define  OP_LRP_SAT                                 0x2C
-#define  OP_MAD                                     0x0E
-#define  OP_MAD_SAT                                 0x2D
-#define  OP_SWZ                                     0x19
-#define  OP_SWZ_SAT                                 0x3C
-#define  OP_TEX                                     0x3D
-#define  OP_TEX_SAT                                 0x3E
-#define  OP_TXB                                     0x3F
-#define  OP_TXB_SAT                                 0x40
-#define  OP_TXP                                     0x41
-#define  OP_TXP_SAT                                 0x42
-#define  OP_KIL                                     0x28
-
-/* GL_ARB_vertex_program instruction code */
-#define  OP_ARL                                     0x02
-/*       OP_ABS */
-/*       OP_FLR */
-/*       OP_FRC */
-/*       OP_LIT */
-/*       OP_MOV */
-/*       OP_EX2 */
-#define  OP_EXP                                     0x08
-/*       OP_LG2 */
-#define  OP_LOG                                     0x0D
-/*       OP_RCP */
-/*       OP_RSQ */
-/*       OP_POW */
-/*       OP_ADD */
-/*       OP_DP3 */
-/*       OP_DP4 */
-/*       OP_DPH */
-/*       OP_DST */
-/*       OP_MAX */
-/*       OP_MIN */
-/*       OP_MUL */
-/*       OP_SGE */
-/*       OP_SLT */
-/*       OP_SUB */
-/*       OP_XPD */
-/*       OP_MAD */
-/*       OP_SWZ */
-
-/* fragment attribute binding */
-#define  FRAGMENT_ATTRIB_COLOR                      0x01
-#define  FRAGMENT_ATTRIB_TEXCOORD                   0x02
-#define  FRAGMENT_ATTRIB_FOGCOORD                   0x03
-#define  FRAGMENT_ATTRIB_POSITION                   0x04
-
-/* vertex attribute binding */
-#define  VERTEX_ATTRIB_POSITION                     0x01
-#define  VERTEX_ATTRIB_WEIGHT                       0x02
-#define  VERTEX_ATTRIB_NORMAL                       0x03
-#define  VERTEX_ATTRIB_COLOR                        0x04
-#define  VERTEX_ATTRIB_FOGCOORD                     0x05
-#define  VERTEX_ATTRIB_TEXCOORD                     0x06
-#define  VERTEX_ATTRIB_MATRIXINDEX                  0x07
-#define  VERTEX_ATTRIB_GENERIC                      0x08
-
-/* fragment result binding */
-#define  FRAGMENT_RESULT_COLOR                      0x01
-#define  FRAGMENT_RESULT_DEPTH                      0x02
-
-/* vertex result binding */
-#define  VERTEX_RESULT_POSITION                     0x01
-#define  VERTEX_RESULT_COLOR                        0x02
-#define  VERTEX_RESULT_FOGCOORD                     0x03
-#define  VERTEX_RESULT_POINTSIZE                    0x04
-#define  VERTEX_RESULT_TEXCOORD                     0x05
-
-/* texture target */
-#define  TEXTARGET_1D                               0x01
-#define  TEXTARGET_2D                               0x02
-#define  TEXTARGET_3D                               0x03
-#define  TEXTARGET_RECT                             0x04
-#define  TEXTARGET_CUBE                             0x05
-/* GL_ARB_fragment_program_shadow */
-#define  TEXTARGET_SHADOW1D                         0x06
-#define  TEXTARGET_SHADOW2D                         0x07
-#define  TEXTARGET_SHADOWRECT                       0x08
-/* GL_MESA_texture_array */
-#define  TEXTARGET_1D_ARRAY                         0x09
-#define  TEXTARGET_2D_ARRAY                         0x0a
-#define  TEXTARGET_SHADOW1D_ARRAY                   0x0b
-#define  TEXTARGET_SHADOW2D_ARRAY                   0x0c
-
-/* face type */
-#define  FACE_FRONT                                 0x00
-#define  FACE_BACK                                  0x01
-
-/* color type */
-#define  COLOR_PRIMARY                              0x00
-#define  COLOR_SECONDARY                            0x01
-
-/* component */
-#define  COMPONENT_X                                0x00
-#define  COMPONENT_Y                                0x01
-#define  COMPONENT_Z                                0x02
-#define  COMPONENT_W                                0x03
-#define  COMPONENT_0                                0x04
-#define  COMPONENT_1                                0x05
-
-/* array index type */
-#define  ARRAY_INDEX_ABSOLUTE                       0x00
-#define  ARRAY_INDEX_RELATIVE                       0x01
-
-/* matrix name */
-#define  MATRIX_MODELVIEW                           0x01
-#define  MATRIX_PROJECTION                          0x02
-#define  MATRIX_MVP                                 0x03
-#define  MATRIX_TEXTURE                             0x04
-#define  MATRIX_PALETTE                             0x05
-#define  MATRIX_PROGRAM                             0x06
-
-/* matrix modifier */
-#define  MATRIX_MODIFIER_IDENTITY                   0x00
-#define  MATRIX_MODIFIER_INVERSE                    0x01
-#define  MATRIX_MODIFIER_TRANSPOSE                  0x02
-#define  MATRIX_MODIFIER_INVTRANS                   0x03
-
-/* constant type */
-#define  CONSTANT_SCALAR                            0x01
-#define  CONSTANT_VECTOR                            0x02
-
-/* program param type */
-#define  PROGRAM_PARAM_ENV                          0x01
-#define  PROGRAM_PARAM_LOCAL                        0x02
-
-/* register type */
-#define  REGISTER_ATTRIB                            0x01
-#define  REGISTER_PARAM                             0x02
-#define  REGISTER_RESULT                            0x03
-#define  REGISTER_ESTABLISHED_NAME                  0x04
-
-/* param binding */
-#define  PARAM_NULL                                 0x00
-#define  PARAM_ARRAY_ELEMENT                        0x01
-#define  PARAM_STATE_ELEMENT                        0x02
-#define  PARAM_PROGRAM_ELEMENT                      0x03
-#define  PARAM_PROGRAM_ELEMENTS                     0x04
-#define  PARAM_CONSTANT                             0x05
-
-/* param state property */
-#define  STATE_MATERIAL_PARSER                      0x01
-#define  STATE_LIGHT_PARSER                         0x02
-#define  STATE_LIGHT_MODEL                          0x03
-#define  STATE_LIGHT_PROD                           0x04
-#define  STATE_FOG                                  0x05
-#define  STATE_MATRIX_ROWS                          0x06
-/* GL_ARB_fragment_program */
-#define  STATE_TEX_ENV                              0x07
-#define  STATE_DEPTH                                0x08
-/* GL_ARB_vertex_program */
-#define  STATE_TEX_GEN                              0x09
-#define  STATE_CLIP_PLANE                           0x0A
-#define  STATE_POINT                                0x0B
-
-/* state material property */
-#define  MATERIAL_AMBIENT                           0x01
-#define  MATERIAL_DIFFUSE                           0x02
-#define  MATERIAL_SPECULAR                          0x03
-#define  MATERIAL_EMISSION                          0x04
-#define  MATERIAL_SHININESS                         0x05
-
-/* state light property */
-#define  LIGHT_AMBIENT                              0x01
-#define  LIGHT_DIFFUSE                              0x02
-#define  LIGHT_SPECULAR                             0x03
-#define  LIGHT_POSITION                             0x04
-#define  LIGHT_ATTENUATION                          0x05
-#define  LIGHT_HALF                                 0x06
-#define  LIGHT_SPOT_DIRECTION                       0x07
-
-/* state light model property */
-#define  LIGHT_MODEL_AMBIENT                        0x01
-#define  LIGHT_MODEL_SCENECOLOR                     0x02
-
-/* state light product property */
-#define  LIGHT_PROD_AMBIENT                         0x01
-#define  LIGHT_PROD_DIFFUSE                         0x02
-#define  LIGHT_PROD_SPECULAR                        0x03
-
-/* state texture environment property */
-#define  TEX_ENV_COLOR                              0x01
-
-/* state texture generation coord property */
-#define  TEX_GEN_EYE                                0x01
-#define  TEX_GEN_OBJECT                             0x02
-
-/* state fog property */
-#define  FOG_COLOR                                  0x01
-#define  FOG_PARAMS                                 0x02
-
-/* state depth property */
-#define  DEPTH_RANGE                                0x01
-
-/* state point parameters property */
-#define  POINT_SIZE                                 0x01
-#define  POINT_ATTENUATION                          0x02
-
-/* declaration */
-#define  ATTRIB                                     0x01
-#define  PARAM                                      0x02
-#define  TEMP                                       0x03
-#define  OUTPUT                                     0x04
-#define  ALIAS                                      0x05
-/* GL_ARB_vertex_program */
-#define  ADDRESS                                    0x06
-
-/*-----------------------------------------------------------------------
- * From here on down is the semantic checking portion
- *
- */
-
-/**
- * Variable Table Handling functions
- */
-typedef enum
-{
-   vt_none,
-   vt_address,
-   vt_attrib,
-   vt_param,
-   vt_temp,
-   vt_output,
-   vt_alias
-} var_type;
-
-
-/**
- * Setting an explicit field for each of the binding properties is a bit
- * wasteful of space, but it should be much more clear when reading later on..
- */
-struct var_cache
-{
-   const GLubyte *name;         /* don't free() - no need */
-   var_type type;
-   GLuint address_binding;      /* The index of the address register we should
-                                 * be using                                        */
-   GLuint attrib_binding;       /* For type vt_attrib, see nvfragprog.h for values */
-   GLuint attrib_is_generic;    /* If the attrib was specified through a generic
-                                 * vertex attrib                                   */
-   GLuint temp_binding;         /* The index of the temp register we are to use    */
-   GLuint output_binding;       /* Output/result register number */
-   struct var_cache *alias_binding;     /* For type vt_alias, points to the var_cache entry
-                                         * that this is aliased to                         */
-   GLuint param_binding_type;   /* {PROGRAM_STATE_VAR, PROGRAM_LOCAL_PARAM,
-                                 *    PROGRAM_ENV_PARAM}                           */
-   GLuint param_binding_begin;  /* This is the offset into the program_parameter_list where
-                                 * the tokens representing our bound state (or constants)
-                                 * start */
-   GLuint param_binding_length; /* This is how many entries in the the program_parameter_list
-                                 * we take up with our state tokens or constants. Note that
-                                 * this is _not_ the same as the number of param registers
-                                 * we eventually use */
-   struct var_cache *next;
-};
-
-static GLvoid
-var_cache_create (struct var_cache **va)
-{
-   *va = (struct var_cache *) _mesa_malloc (sizeof (struct var_cache));
-   if (*va) {
-      (**va).name = NULL;
-      (**va).type = vt_none;
-      (**va).attrib_binding = ~0;
-      (**va).attrib_is_generic = 0;
-      (**va).temp_binding = ~0;
-      (**va).output_binding = ~0;
-      (**va).param_binding_type = ~0;
-      (**va).param_binding_begin = ~0;
-      (**va).param_binding_length = ~0;
-      (**va).alias_binding = NULL;
-      (**va).next = NULL;
-   }
-}
-
-static GLvoid
-var_cache_destroy (struct var_cache **va)
-{
-   if (*va) {
-      var_cache_destroy (&(**va).next);
-      _mesa_free (*va);
-      *va = NULL;
-   }
-}
-
-static GLvoid
-var_cache_append (struct var_cache **va, struct var_cache *nv)
-{
-   if (*va)
-      var_cache_append (&(**va).next, nv);
-   else
-      *va = nv;
-}
-
-static struct var_cache *
-var_cache_find (struct var_cache *va, const GLubyte * name)
-{
-   /*struct var_cache *first = va;*/
-
-   while (va) {
-      if (!_mesa_strcmp ( (const char*) name, (const char*) va->name)) {
-         if (va->type == vt_alias)
-            return va->alias_binding;
-         return va;
-      }
-
-      va = va->next;
-   }
-
-   return NULL;
-}
-
-
-
-/**
- * Called when an error is detected while parsing/compiling a program.
- * Sets the ctx->Program.ErrorString field to descript and records a
- * GL_INVALID_OPERATION error.
- * \param position  position of error in program string
- * \param descrip  verbose error description
- */
-static void
-program_error(GLcontext *ctx, GLint position, const char *descrip)
-{
-   if (descrip) {
-      const char *prefix = "glProgramString(", *suffix = ")";
-      char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) +
-                                        _mesa_strlen(prefix) +
-                                        _mesa_strlen(suffix) + 1);
-      if (str) {
-         _mesa_sprintf(str, "%s%s%s", prefix, descrip, suffix);
-         _mesa_error(ctx, GL_INVALID_OPERATION, str);
-         _mesa_free(str);
-      }
-   }
-   _mesa_set_program_error(ctx, position, descrip);
-}
-
-
-/**
- * As above, but with an extra string parameter for more info.
- */
-static void
-program_error2(GLcontext *ctx, GLint position, const char *descrip,
-               const char *var)
-{
-   if (descrip) {
-      const char *prefix = "glProgramString(", *suffix = ")";
-      char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) +
-                                        _mesa_strlen(": ") +
-                                        _mesa_strlen(var) +
-                                        _mesa_strlen(prefix) +
-                                        _mesa_strlen(suffix) + 1);
-      if (str) {
-         _mesa_sprintf(str, "%s%s: %s%s", prefix, descrip, var, suffix);
-         _mesa_error(ctx, GL_INVALID_OPERATION, str);
-         _mesa_free(str);
-      }
-   }
-   {
-      char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) +
-                                        _mesa_strlen(": ") +
-                                        _mesa_strlen(var) + 1);
-      if (str) {
-         _mesa_sprintf(str, "%s: %s", descrip, var);
-      }
-      _mesa_set_program_error(ctx, position, str);
-      if (str) {
-         _mesa_free(str);
-      }
-   }
-}
-
-
-
-/**
- * constructs an integer from 4 GLubytes in LE format
- */
-static GLuint
-parse_position (const GLubyte ** inst)
-{
-   GLuint value;
-
-   value =  (GLuint) (*(*inst)++);
-   value += (GLuint) (*(*inst)++) * 0x100;
-   value += (GLuint) (*(*inst)++) * 0x10000;
-   value += (GLuint) (*(*inst)++) * 0x1000000;
-
-   return value;
-}
-
-/**
- * This will, given a string, lookup the string as a variable name in the
- * var cache. If the name is found, the var cache node corresponding to the
- * var name is returned. If it is not found, a new entry is allocated
- *
- * \param  I     Points into the binary array where the string identifier begins
- * \param  found 1 if the string was found in the var_cache, 0 if it was allocated
- * \return       The location on the var_cache corresponding the the string starting at I
- */
-static struct var_cache *
-parse_string (const GLubyte ** inst, struct var_cache **vc_head,
-              struct arb_program *Program, GLuint * found)
-{
-   const GLubyte *i = *inst;
-   struct var_cache *va = NULL;
-   (void) Program;
-
-   *inst += _mesa_strlen ((char *) i) + 1;
-
-   va = var_cache_find (*vc_head, i);
-
-   if (va) {
-      *found = 1;
-      return va;
-   }
-
-   *found = 0;
-   var_cache_create (&va);
-   va->name = (const GLubyte *) i;
-
-   var_cache_append (vc_head, va);
-
-   return va;
-}
-
-static char *
-parse_string_without_adding (const GLubyte ** inst, struct arb_program *Program)
-{
-   const GLubyte *i = *inst;
-   (void) Program;
-   
-   *inst += _mesa_strlen ((char *) i) + 1;
-
-   return (char *) i;
-}
-
-/**
- * \return -1 if we parse '-', return 1 otherwise
- */
-static GLint
-parse_sign (const GLubyte ** inst)
-{
-   /*return *(*inst)++ != '+'; */
-
-   if (**inst == '-') {
-      (*inst)++;
-      return -1;
-   }
-   else if (**inst == '+') {
-      (*inst)++;
-      return 1;
-   }
-
-   return 1;
-}
-
-/**
- * parses and returns signed integer
- */
-static GLint
-parse_integer (const GLubyte ** inst, struct arb_program *Program)
-{
-   GLint sign;
-   GLint value;
-
-   /* check if *inst points to '+' or '-'
-    * if yes, grab the sign and increment *inst
-    */
-   sign = parse_sign (inst);
-
-   /* now check if *inst points to 0
-    * if yes, increment the *inst and return the default value
-    */
-   if (**inst == 0) {
-      (*inst)++;
-      return 0;
-   }
-
-   /* parse the integer as you normally would do it */
-   value = _mesa_atoi (parse_string_without_adding (inst, Program));
-
-   /* now, after terminating 0 there is a position
-    * to parse it - parse_position()
-    */
-   Program->Position = parse_position (inst);
-
-   return value * sign;
-}
-
-/**
-  Accumulate this string of digits, and return them as 
-  a large integer represented in floating point (for range).
-  If scale is not NULL, also accumulates a power-of-ten
-  integer scale factor that represents the number of digits 
-  in the string.
-*/
-static GLdouble
-parse_float_string(const GLubyte ** inst, struct arb_program *Program, GLdouble *scale)
-{
-   GLdouble value = 0.0;
-   GLdouble oscale = 1.0;
-
-   if (**inst == 0) { /* this string of digits is empty-- do nothing */
-      (*inst)++;
-   }
-   else { /* nonempty string-- parse out the digits */
-      while (**inst >= '0' && **inst <= '9') {
-         GLubyte digit = *((*inst)++);
-         value = value * 10.0 + (GLint) (digit - '0');
-         oscale *= 10.0;
-      }
-      assert(**inst == 0); /* integer string should end with 0 */
-      (*inst)++; /* skip over terminating 0 */
-      Program->Position = parse_position(inst); /* skip position (from integer) */
-   }
-   if (scale)
-      *scale = oscale;
-   return value;
-}
-
-/**
-  Parse an unsigned floating-point number from this stream of tokenized
-  characters.  Example floating-point formats supported:
-     12.34
-     12
-     0.34
-     .34
-     12.34e-4
- */
-static GLfloat
-parse_float (const GLubyte ** inst, struct arb_program *Program)
-{
-   GLint exponent;
-   GLdouble whole, fraction, fracScale = 1.0;
-
-   whole = parse_float_string(inst, Program, 0);
-   fraction = parse_float_string(inst, Program, &fracScale);
-   
-   /* Parse signed exponent */
-   exponent = parse_integer(inst, Program);   /* This is the exponent */
-
-   /* Assemble parts of floating-point number: */
-   return (GLfloat) ((whole + fraction / fracScale) *
-                     _mesa_pow(10.0, (GLfloat) exponent));
-}
-
-
-/**
- */
-static GLfloat
-parse_signed_float (const GLubyte ** inst, struct arb_program *Program)
-{
-   GLint sign = parse_sign (inst);
-   GLfloat value = parse_float (inst, Program);
-   return value * sign;
-}
-
-/**
- * This picks out a constant value from the parsed array. The constant vector is r
- * returned in the *values array, which should be of length 4.
- *
- * \param values - The 4 component vector with the constant value in it
- */
-static GLvoid
-parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Program,
-                GLboolean use)
-{
-   GLuint components, i;
-
-
-   switch (*(*inst)++) {
-      case CONSTANT_SCALAR:
-         if (use == GL_TRUE) {
-            values[0] =
-               values[1] =
-               values[2] = values[3] = parse_float (inst, Program);
-         }
-         else {
-            values[0] =
-               values[1] =
-               values[2] = values[3] = parse_signed_float (inst, Program);
-         }
-
-         break;
-      case CONSTANT_VECTOR:
-         values[0] = values[1] = values[2] = 0;
-         values[3] = 1;
-         components = *(*inst)++;
-         for (i = 0; i < components; i++) {
-            values[i] = parse_signed_float (inst, Program);
-         }
-         break;
-   }
-}
-
-/**
- * \param offset The offset from the address register that we should
- *                address
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_relative_offset(GLcontext *ctx, const GLubyte **inst,
-                      struct arb_program *Program, GLint *offset)
-{
-   (void) ctx;
-   *offset = parse_integer(inst, Program);
-   return 0;
-}
-
-/**
- * \param  color 0 if color type is primary, 1 if color type is secondary
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_color_type (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program,
-                  GLint * color)
-{
-   (void) ctx; (void) Program;
-   *color = *(*inst)++ != COLOR_PRIMARY;
-   return 0;
-}
-
-/**
- * Get an integer corresponding to a generic vertex attribute.
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_generic_attrib_num(GLcontext *ctx, const GLubyte ** inst,
-                       struct arb_program *Program, GLuint *attrib)
-{
-   GLint i = parse_integer(inst, Program);
-
-   if ((i < 0) || (i >= MAX_VERTEX_PROGRAM_ATTRIBS))
-   {
-      program_error(ctx, Program->Position,
-                    "Invalid generic vertex attribute index");
-      return 1;
-   }
-
-   *attrib = (GLuint) i;
-
-   return 0;
-}
-
-
-/**
- * \param color The index of the color buffer to write into
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_output_color_num (GLcontext * ctx, const GLubyte ** inst,
-                    struct arb_program *Program, GLuint * color)
-{
-   GLint i = parse_integer (inst, Program);
-
-   if ((i < 0) || (i >= (int)ctx->Const.MaxDrawBuffers)) {
-      *color = 0;
-      program_error(ctx, Program->Position, "Invalid draw buffer index");
-      return 1;
-   }
-
-   *color = (GLuint) i;
-   return 0;
-}
-
-
-/**
- * Validate the index of a texture coordinate
- *
- * \param coord The texture unit index
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_texcoord_num (GLcontext * ctx, const GLubyte ** inst,
-                    struct arb_program *Program, GLuint * coord)
-{
-   GLint i = parse_integer (inst, Program);
-
-   if ((i < 0) || (i >= (int)ctx->Const.MaxTextureCoordUnits)) {
-      program_error(ctx, Program->Position, "Invalid texture coordinate index");
-      return 1;
-   }
-
-   *coord = (GLuint) i;
-   return 0;
-}
-
-
-/**
- * Validate the index of a texture image unit
- *
- * \param coord The texture unit index
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_teximage_num (GLcontext * ctx, const GLubyte ** inst,
-                    struct arb_program *Program, GLuint * coord)
-{
-   GLint i = parse_integer (inst, Program);
-
-   if ((i < 0) || (i >= (int)ctx->Const.MaxTextureImageUnits)) {
-      char s[100];
-      _mesa_snprintf(s, sizeof(s), "Invalid texture image index %d (%u is max)",
-                     i, ctx->Const.MaxTextureImageUnits);
-      program_error(ctx, Program->Position, s);
-      return 1;
-   }
-
-   *coord = (GLuint) i;
-   return 0;
-}
-
-
-/**
- * \param coord The weight index
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_weight_num (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program,
-                  GLint * coord)
-{
-   *coord = parse_integer (inst, Program);
-
-   if ((*coord < 0) || (*coord >= 1)) {
-      program_error(ctx, Program->Position, "Invalid weight index");
-      return 1;
-   }
-
-   return 0;
-}
-
-/**
- * \param coord The clip plane index
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_clipplane_num (GLcontext * ctx, const GLubyte ** inst,
-                     struct arb_program *Program, GLint * coord)
-{
-   *coord = parse_integer (inst, Program);
-
-   if ((*coord < 0) || (*coord >= (GLint) ctx->Const.MaxClipPlanes)) {
-      program_error(ctx, Program->Position, "Invalid clip plane index");
-      return 1;
-   }
-
-   return 0;
-}
-
-
-/**
- * \return 0 on front face, 1 on back face
- */
-static GLuint
-parse_face_type (const GLubyte ** inst)
-{
-   switch (*(*inst)++) {
-      case FACE_FRONT:
-         return 0;
-
-      case FACE_BACK:
-         return 1;
-   }
-   return 0;
-}
-
-
-/**
- * Given a matrix and a modifier token on the binary array, return tokens
- * that _mesa_fetch_state() [program.c] can understand.
- *
- * \param matrix - the matrix we are talking about
- * \param matrix_idx - the index of the matrix we have (for texture & program matricies)
- * \param matrix_modifier - the matrix modifier (trans, inv, etc)
- * \return 0 on sucess, 1 on failure
- */
-static GLuint
-parse_matrix (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program,
-              GLint * matrix, GLint * matrix_idx, GLint * matrix_modifier)
-{
-   GLubyte mat = *(*inst)++;
-
-   *matrix_idx = 0;
-
-   switch (mat) {
-      case MATRIX_MODELVIEW:
-         *matrix = STATE_MODELVIEW_MATRIX;
-         *matrix_idx = parse_integer (inst, Program);
-         if (*matrix_idx > 0) {
-            program_error(ctx, Program->Position,
-                          "ARB_vertex_blend not supported");
-            return 1;
-         }
-         break;
-
-      case MATRIX_PROJECTION:
-         *matrix = STATE_PROJECTION_MATRIX;
-         break;
-
-      case MATRIX_MVP:
-         *matrix = STATE_MVP_MATRIX;
-         break;
-
-      case MATRIX_TEXTURE:
-         *matrix = STATE_TEXTURE_MATRIX;
-         *matrix_idx = parse_integer (inst, Program);
-         if (*matrix_idx >= (GLint) ctx->Const.MaxTextureUnits) {
-            program_error(ctx, Program->Position, "Invalid Texture Unit");
-            /* bad *matrix_id */
-            return 1;
-         }
-         break;
-
-         /* This is not currently supported (ARB_matrix_palette) */
-      case MATRIX_PALETTE:
-         *matrix_idx = parse_integer (inst, Program);
-         program_error(ctx, Program->Position,
-                       "ARB_matrix_palette not supported");
-         return 1;
-         break;
-
-      case MATRIX_PROGRAM:
-         *matrix = STATE_PROGRAM_MATRIX;
-         *matrix_idx = parse_integer (inst, Program);
-         if (*matrix_idx >= (GLint) ctx->Const.MaxProgramMatrices) {
-            program_error(ctx, Program->Position, "Invalid Program Matrix");
-            /* bad *matrix_idx */
-            return 1;
-         }
-         break;
-   }
-
-   switch (*(*inst)++) {
-      case MATRIX_MODIFIER_IDENTITY:
-         *matrix_modifier = 0;
-         break;
-      case MATRIX_MODIFIER_INVERSE:
-         *matrix_modifier = STATE_MATRIX_INVERSE;
-         break;
-      case MATRIX_MODIFIER_TRANSPOSE:
-         *matrix_modifier = STATE_MATRIX_TRANSPOSE;
-         break;
-      case MATRIX_MODIFIER_INVTRANS:
-         *matrix_modifier = STATE_MATRIX_INVTRANS;
-         break;
-   }
-
-   return 0;
-}
-
-
-/**
- * This parses a state string (rather, the binary version of it) into
- * a 6-token sequence as described in _mesa_fetch_state() [program.c]
- *
- * \param inst         - the start in the binary arry to start working from
- * \param state_tokens - the storage for the 6-token state description
- * \return             - 0 on sucess, 1 on error
- */
-static GLuint
-parse_state_single_item (GLcontext * ctx, const GLubyte ** inst,
-                         struct arb_program *Program,
-                         gl_state_index state_tokens[STATE_LENGTH])
-{
-   GLubyte token = *(*inst)++;
-
-   switch (token) {
-      case STATE_MATERIAL_PARSER:
-         state_tokens[0] = STATE_MATERIAL;
-         state_tokens[1] = parse_face_type (inst);
-         switch (*(*inst)++) {
-            case MATERIAL_AMBIENT:
-               state_tokens[2] = STATE_AMBIENT;
-               break;
-            case MATERIAL_DIFFUSE:
-               state_tokens[2] = STATE_DIFFUSE;
-               break;
-            case MATERIAL_SPECULAR:
-               state_tokens[2] = STATE_SPECULAR;
-               break;
-            case MATERIAL_EMISSION:
-               state_tokens[2] = STATE_EMISSION;
-	       break;
-            case MATERIAL_SHININESS:
-               state_tokens[2] = STATE_SHININESS;
-               break;
-         }
-         break;
-
-      case STATE_LIGHT_PARSER:
-         state_tokens[0] = STATE_LIGHT;
-         state_tokens[1] = parse_integer (inst, Program);
-
-         /* Check the value of state_tokens[1] against the # of lights */
-         if (state_tokens[1] >= (GLint) ctx->Const.MaxLights) {
-            program_error(ctx, Program->Position, "Invalid Light Number");
-            /* bad state_tokens[1] */
-            return 1;
-         }
-
-         switch (*(*inst)++) {
-            case LIGHT_AMBIENT:
-               state_tokens[2] = STATE_AMBIENT;
-               break;
-            case LIGHT_DIFFUSE:
-               state_tokens[2] = STATE_DIFFUSE;
-               break;
-            case LIGHT_SPECULAR:
-               state_tokens[2] = STATE_SPECULAR;
-               break;
-            case LIGHT_POSITION:
-               state_tokens[2] = STATE_POSITION;
-               break;
-            case LIGHT_ATTENUATION:
-               state_tokens[2] = STATE_ATTENUATION;
-               break;
-            case LIGHT_HALF:
-               state_tokens[2] = STATE_HALF_VECTOR;
-               break;
-            case LIGHT_SPOT_DIRECTION:
-               state_tokens[2] = STATE_SPOT_DIRECTION;
-               break;
-         }
-         break;
-
-      case STATE_LIGHT_MODEL:
-         switch (*(*inst)++) {
-            case LIGHT_MODEL_AMBIENT:
-               state_tokens[0] = STATE_LIGHTMODEL_AMBIENT;
-               break;
-            case LIGHT_MODEL_SCENECOLOR:
-               state_tokens[0] = STATE_LIGHTMODEL_SCENECOLOR;
-               state_tokens[1] = parse_face_type (inst);
-               break;
-         }
-         break;
-
-      case STATE_LIGHT_PROD:
-         state_tokens[0] = STATE_LIGHTPROD;
-         state_tokens[1] = parse_integer (inst, Program);
-
-         /* Check the value of state_tokens[1] against the # of lights */
-         if (state_tokens[1] >= (GLint) ctx->Const.MaxLights) {
-            program_error(ctx, Program->Position, "Invalid Light Number");
-            /* bad state_tokens[1] */
-            return 1;
-         }
-
-         state_tokens[2] = parse_face_type (inst);
-         switch (*(*inst)++) {
-            case LIGHT_PROD_AMBIENT:
-               state_tokens[3] = STATE_AMBIENT;
-               break;
-            case LIGHT_PROD_DIFFUSE:
-               state_tokens[3] = STATE_DIFFUSE;
-               break;
-            case LIGHT_PROD_SPECULAR:
-               state_tokens[3] = STATE_SPECULAR;
-               break;
-         }
-         break;
-
-
-      case STATE_FOG:
-         switch (*(*inst)++) {
-            case FOG_COLOR:
-               state_tokens[0] = STATE_FOG_COLOR;
-               break;
-            case FOG_PARAMS:
-               state_tokens[0] = STATE_FOG_PARAMS;
-               break;
-         }
-         break;
-
-      case STATE_TEX_ENV:
-         state_tokens[1] = parse_integer (inst, Program);
-         switch (*(*inst)++) {
-            case TEX_ENV_COLOR:
-               state_tokens[0] = STATE_TEXENV_COLOR;
-               break;
-         }
-         break;
-
-      case STATE_TEX_GEN:
-         {
-            GLuint type, coord;
-
-            state_tokens[0] = STATE_TEXGEN;
-            /*state_tokens[1] = parse_integer (inst, Program);*/    /* Texture Unit */
-
-            if (parse_texcoord_num (ctx, inst, Program, &coord))
-               return 1;
-	    state_tokens[1] = coord;
-
-            /* EYE or OBJECT */
-            type = *(*inst)++;
-
-            /* 0 - s, 1 - t, 2 - r, 3 - q */
-            coord = *(*inst)++;
-
-            if (type == TEX_GEN_EYE) {
-               switch (coord) {
-                  case COMPONENT_X:
-                     state_tokens[2] = STATE_TEXGEN_EYE_S;
-                     break;
-                  case COMPONENT_Y:
-                     state_tokens[2] = STATE_TEXGEN_EYE_T;
-                     break;
-                  case COMPONENT_Z:
-                     state_tokens[2] = STATE_TEXGEN_EYE_R;
-                     break;
-                  case COMPONENT_W:
-                     state_tokens[2] = STATE_TEXGEN_EYE_Q;
-                     break;
-                  default:
-                     _mesa_problem(ctx, "bad texgen component in "
-                                   "parse_state_single_item()");
-               }
-            }
-            else {
-               switch (coord) {
-                  case COMPONENT_X:
-                     state_tokens[2] = STATE_TEXGEN_OBJECT_S;
-                     break;
-                  case COMPONENT_Y:
-                     state_tokens[2] = STATE_TEXGEN_OBJECT_T;
-                     break;
-                  case COMPONENT_Z:
-                     state_tokens[2] = STATE_TEXGEN_OBJECT_R;
-                     break;
-                  case COMPONENT_W:
-                     state_tokens[2] = STATE_TEXGEN_OBJECT_Q;
-                     break;
-                  default:
-                     _mesa_problem(ctx, "bad texgen component in "
-                                   "parse_state_single_item()");
-               }
-            }
-         }
-         break;
-
-      case STATE_DEPTH:
-         switch (*(*inst)++) {
-            case DEPTH_RANGE:
-               state_tokens[0] = STATE_DEPTH_RANGE;
-               break;
-         }
-         break;
-
-      case STATE_CLIP_PLANE:
-         state_tokens[0] = STATE_CLIPPLANE;
-         if (parse_clipplane_num (ctx, inst, Program,
-                                  (GLint *) &state_tokens[1]))
-            return 1;
-         break;
-
-      case STATE_POINT:
-         switch (*(*inst)++) {
-            case POINT_SIZE:
-               state_tokens[0] = STATE_POINT_SIZE;
-               break;
-
-            case POINT_ATTENUATION:
-               state_tokens[0] = STATE_POINT_ATTENUATION;
-               break;
-         }
-         break;
-
-         /* XXX: I think this is the correct format for a matrix row */
-      case STATE_MATRIX_ROWS:
-         if (parse_matrix(ctx, inst, Program,
-                          (GLint *) &state_tokens[0],
-                          (GLint *) &state_tokens[1],
-                          (GLint *) &state_tokens[4]))
-            return 1;
-
-         state_tokens[2] = parse_integer (inst, Program);       /* The first row to grab */
-
-         if ((**inst) != 0) {                                   /* Either the last row, 0 */
-            state_tokens[3] = parse_integer (inst, Program);
-            if (state_tokens[3] < state_tokens[2]) {
-               program_error(ctx, Program->Position,
-                             "Second matrix index less than the first");
-               /* state_tokens[4] vs. state_tokens[3] */
-               return 1;
-            }
-         }
-         else {
-            state_tokens[3] = state_tokens[2];
-            (*inst)++;
-         }
-         break;
-   }
-
-   return 0;
-}
-
-/**
- * This parses a state string (rather, the binary version of it) into
- * a 6-token similar for the state fetching code in program.c
- *
- * One might ask, why fetch these parameters into just like  you fetch
- * state when they are already stored in other places?
- *
- * Because of array offsets -> We can stick env/local parameters in the
- * middle of a parameter array and then index someplace into the array
- * when we execute.
- *
- * One optimization might be to only do this for the cases where the
- * env/local parameters end up inside of an array, and leave the
- * single parameters (or arrays of pure env/local pareameters) in their
- * respective register files.
- *
- * For ENV parameters, the format is:
- *    state_tokens[0] = STATE_FRAGMENT_PROGRAM / STATE_VERTEX_PROGRAM
- *    state_tokens[1] = STATE_ENV
- *    state_tokens[2] = the parameter index
- *
- * for LOCAL parameters, the format is:
- *    state_tokens[0] = STATE_FRAGMENT_PROGRAM / STATE_VERTEX_PROGRAM
- *    state_tokens[1] = STATE_LOCAL
- *    state_tokens[2] = the parameter index
- *
- * \param inst         - the start in the binary arry to start working from
- * \param state_tokens - the storage for the 6-token state description
- * \return             - 0 on sucess, 1 on failure
- */
-static GLuint
-parse_program_single_item (GLcontext * ctx, const GLubyte ** inst,
-                           struct arb_program *Program,
-                           gl_state_index state_tokens[STATE_LENGTH])
-{
-   if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB)
-      state_tokens[0] = STATE_FRAGMENT_PROGRAM;
-   else
-      state_tokens[0] = STATE_VERTEX_PROGRAM;
-
-
-   switch (*(*inst)++) {
-      case PROGRAM_PARAM_ENV:
-         state_tokens[1] = STATE_ENV;
-         state_tokens[2] = parse_integer (inst, Program);
-
-         /* Check state_tokens[2] against the number of ENV parameters available */
-         if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) &&
-              (state_tokens[2] >= (GLint) ctx->Const.FragmentProgram.MaxEnvParams))
-             ||
-             ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) &&
-              (state_tokens[2] >= (GLint) ctx->Const.VertexProgram.MaxEnvParams))) {
-            program_error(ctx, Program->Position,
-                          "Invalid Program Env Parameter");
-            /* bad state_tokens[2] */
-            return 1;
-         }
-
-         break;
-
-      case PROGRAM_PARAM_LOCAL:
-         state_tokens[1] = STATE_LOCAL;
-         state_tokens[2] = parse_integer (inst, Program);
-
-         /* Check state_tokens[2] against the number of LOCAL parameters available */
-         if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) &&
-              (state_tokens[2] >= (GLint) ctx->Const.FragmentProgram.MaxLocalParams))
-             ||
-             ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) &&
-              (state_tokens[2] >= (GLint) ctx->Const.VertexProgram.MaxLocalParams))) {
-            program_error(ctx, Program->Position,
-                          "Invalid Program Local Parameter");
-            /* bad state_tokens[2] */
-            return 1;
-         }
-         break;
-   }
-
-   return 0;
-}
-
-/**
- * For ARB_vertex_program, programs are not allowed to use both an explicit
- * vertex attribute and a generic vertex attribute corresponding to the same
- * state. See section 2.14.3.1 of the GL_ARB_vertex_program spec.
- *
- * This will walk our var_cache and make sure that nobody does anything fishy.
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-generic_attrib_check(struct var_cache *vc_head)
-{
-   int a;
-   struct var_cache *curr;
-   GLboolean explicitAttrib[MAX_VERTEX_PROGRAM_ATTRIBS],
-      genericAttrib[MAX_VERTEX_PROGRAM_ATTRIBS];
-
-   for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) {
-      explicitAttrib[a] = GL_FALSE;
-      genericAttrib[a] = GL_FALSE;
-   }
-
-   curr = vc_head;
-   while (curr) {
-      if (curr->type == vt_attrib) {
-         if (curr->attrib_is_generic) {
-            GLuint attr = (curr->attrib_binding == 0)
-               ? 0 : (curr->attrib_binding - VERT_ATTRIB_GENERIC0);
-            assert(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
-            genericAttrib[attr] = GL_TRUE;
-         }
-         else {
-            assert(curr->attrib_binding < MAX_VERTEX_PROGRAM_ATTRIBS);
-            explicitAttrib[ curr->attrib_binding ] = GL_TRUE;
-         }
-      }
-
-      curr = curr->next;
-   }
-
-   for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) {
-      if ((explicitAttrib[a]) && (genericAttrib[a]))
-         return 1;
-   }
-
-   return 0;
-}
-
-/**
- * This will handle the binding side of an ATTRIB var declaration
- *
- * \param inputReg  returns the input register index, one of the
- *                  VERT_ATTRIB_* or FRAG_ATTRIB_* values.
- * \return returns 0 on success, 1 on error
- */
-static GLuint
-parse_attrib_binding(GLcontext * ctx, const GLubyte ** inst,
-                     struct arb_program *Program,
-                     GLuint *inputReg, GLuint *is_generic)
-{
-   GLint err = 0;
-
-   *is_generic = 0;
-
-   if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-      switch (*(*inst)++) {
-         case FRAGMENT_ATTRIB_COLOR:
-            {
-               GLint coord;
-               err = parse_color_type (ctx, inst, Program, &coord);
-               *inputReg = FRAG_ATTRIB_COL0 + coord;
-            }
-            break;
-         case FRAGMENT_ATTRIB_TEXCOORD:
-            {
-               GLuint texcoord = 0;
-               err = parse_texcoord_num (ctx, inst, Program, &texcoord);
-               *inputReg = FRAG_ATTRIB_TEX0 + texcoord;
-            }
-            break;
-         case FRAGMENT_ATTRIB_FOGCOORD:
-            *inputReg = FRAG_ATTRIB_FOGC;
-            break;
-         case FRAGMENT_ATTRIB_POSITION:
-            *inputReg = FRAG_ATTRIB_WPOS;
-            break;
-         default:
-            err = 1;
-            break;
-      }
-   }
-   else {
-      switch (*(*inst)++) {
-         case VERTEX_ATTRIB_POSITION:
-            *inputReg = VERT_ATTRIB_POS;
-            break;
-
-         case VERTEX_ATTRIB_WEIGHT:
-            {
-               GLint weight;
-               err = parse_weight_num (ctx, inst, Program, &weight);
-               *inputReg = VERT_ATTRIB_WEIGHT;
-#if 1
-               /* hack for Warcraft (see bug 8060) */
-               _mesa_warning(ctx, "Application error: vertex program uses 'vertex.weight' but GL_ARB_vertex_blend not supported.");
-               break;
-#else
-               program_error(ctx, Program->Position,
-                             "ARB_vertex_blend not supported");
-               return 1;
-#endif
-            }
-
-         case VERTEX_ATTRIB_NORMAL:
-            *inputReg = VERT_ATTRIB_NORMAL;
-            break;
-
-         case VERTEX_ATTRIB_COLOR:
-            {
-               GLint color;
-               err = parse_color_type (ctx, inst, Program, &color);
-               if (color) {
-                  *inputReg = VERT_ATTRIB_COLOR1;
-               }
-               else {
-                  *inputReg = VERT_ATTRIB_COLOR0;
-               }
-            }
-            break;
-
-         case VERTEX_ATTRIB_FOGCOORD:
-            *inputReg = VERT_ATTRIB_FOG;
-            break;
-
-         case VERTEX_ATTRIB_TEXCOORD:
-            {
-               GLuint unit = 0;
-               err = parse_texcoord_num (ctx, inst, Program, &unit);
-               *inputReg = VERT_ATTRIB_TEX0 + unit;
-            }
-            break;
-
-         case VERTEX_ATTRIB_MATRIXINDEX:
-            /* Not supported at this time */
-            {
-               const char *msg = "ARB_palette_matrix not supported";
-               parse_integer (inst, Program);
-               program_error(ctx, Program->Position, msg);
-            }
-            return 1;
-
-         case VERTEX_ATTRIB_GENERIC:
-            {
-               GLuint attrib;
-               err = parse_generic_attrib_num(ctx, inst, Program, &attrib);
-               if (!err) {
-                  *is_generic = 1;
-                  /* Add VERT_ATTRIB_GENERIC0 here because ARB_vertex_program's
-                   * attributes do not alias the conventional vertex
-                   * attributes.
-                   */
-                  if (attrib > 0)
-                     *inputReg = attrib + VERT_ATTRIB_GENERIC0;
-                  else
-                     *inputReg = 0;
-               }
-            }
-            break;
-
-         default:
-            err = 1;
-            break;
-      }
-   }
-
-   if (err) {
-      program_error(ctx, Program->Position, "Bad attribute binding");
-   }
-
-   return err;
-}
-
-
-/**
- * This translates between a binary token for an output variable type
- * and the mesa token for the same thing.
- *
- * \param inst       The parsed tokens
- * \param outputReg  Returned index/number of the output register,
- *                   one of the VERT_RESULT_* or FRAG_RESULT_* values.
- */
-static GLuint
-parse_result_binding(GLcontext *ctx, const GLubyte **inst,
-                     GLuint *outputReg, struct arb_program *Program)
-{
-   const GLubyte token = *(*inst)++;
-
-   switch (token) {
-      case FRAGMENT_RESULT_COLOR:
-         if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-            GLuint out_color;
-
-            /* This gets result of the color buffer we're supposed to 
-             * draw into.  This pertains to GL_ARB_draw_buffers.
-             */
-            parse_output_color_num(ctx, inst, Program, &out_color);
-            ASSERT(out_color < MAX_DRAW_BUFFERS);
-            *outputReg = FRAG_RESULT_COLOR;
-         }
-         else {
-            /* for vtx programs, this is VERTEX_RESULT_POSITION */
-            *outputReg = VERT_RESULT_HPOS;
-         }
-         break;
-
-      case FRAGMENT_RESULT_DEPTH:
-         if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-            /* for frag programs, this is FRAGMENT_RESULT_DEPTH */
-            *outputReg = FRAG_RESULT_DEPTH;
-         }
-         else {
-            /* for vtx programs, this is VERTEX_RESULT_COLOR */
-            GLint color_type;
-            GLuint face_type = parse_face_type(inst);
-	    GLint err = parse_color_type(ctx, inst, Program, &color_type);
-            if (err)
-               return 1;
-
-            if (face_type) {
-               /* back face */
-               if (color_type) {
-                  *outputReg = VERT_RESULT_BFC1; /* secondary color */
-               }
-               else {
-                  *outputReg = VERT_RESULT_BFC0; /* primary color */
-               }
-            }
-            else {
-               /* front face */
-               if (color_type) {
-                  *outputReg = VERT_RESULT_COL1; /* secondary color */
-               }
-               /* primary color */
-               else {
-                  *outputReg = VERT_RESULT_COL0; /* primary color */
-               }
-            }
-         }
-         break;
-
-      case VERTEX_RESULT_FOGCOORD:
-         *outputReg = VERT_RESULT_FOGC;
-         break;
-
-      case VERTEX_RESULT_POINTSIZE:
-         *outputReg = VERT_RESULT_PSIZ;
-         break;
-
-      case VERTEX_RESULT_TEXCOORD:
-         {
-            GLuint unit;
-            if (parse_texcoord_num (ctx, inst, Program, &unit))
-               return 1;
-            *outputReg = VERT_RESULT_TEX0 + unit;
-         }
-         break;
-   }
-
-   Program->Base.OutputsWritten |= (1 << *outputReg);
-
-   return 0;
-}
-
-
-/**
- * This handles the declaration of ATTRIB variables
- *
- * XXX: Still needs
- *      parse_vert_attrib_binding(), or something like that
- *
- * \return 0 on sucess, 1 on error
- */
-static GLint
-parse_attrib (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-              struct arb_program *Program)
-{
-   GLuint found;
-   struct var_cache *attrib_var;
-
-   attrib_var = parse_string (inst, vc_head, Program, &found);
-   Program->Position = parse_position (inst);
-   if (found) {
-      program_error2(ctx, Program->Position,
-                     "Duplicate variable declaration",
-                     (char *) attrib_var->name);
-      return 1;
-   }
-
-   attrib_var->type = vt_attrib;
-
-   if (parse_attrib_binding(ctx, inst, Program, &attrib_var->attrib_binding,
-                            &attrib_var->attrib_is_generic))
-      return 1;
-
-   if (generic_attrib_check(*vc_head)) {
-      program_error(ctx, Program->Position,
-                    "Cannot use both a generic vertex attribute "
-                    "and a specific attribute of the same type");
-      return 1;
-   }
-
-   Program->Base.NumAttributes++;
-   return 0;
-}
-
-/**
- * \param use -- TRUE if we're called when declaring implicit parameters,
- *               FALSE if we're declaraing variables. This has to do with
- *               if we get a signed or unsigned float for scalar constants
- */
-static GLuint
-parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
-                      struct var_cache *param_var,
-                      struct arb_program *Program, GLboolean use)
-{
-   GLint idx;
-   GLuint err = 0;
-   gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0};
-   GLfloat const_values[4];
-
-   GLubyte token = *(*inst)++;
-
-   switch (token) {
-      case PARAM_STATE_ELEMENT:
-         if (parse_state_single_item (ctx, inst, Program, state_tokens))
-            return 1;
-
-         /* If we adding STATE_MATRIX that has multiple rows, we need to
-          * unroll it and call _mesa_add_state_reference() for each row
-          */
-         if ((state_tokens[0] == STATE_MODELVIEW_MATRIX ||
-              state_tokens[0] == STATE_PROJECTION_MATRIX ||
-              state_tokens[0] == STATE_MVP_MATRIX ||
-              state_tokens[0] == STATE_TEXTURE_MATRIX ||
-              state_tokens[0] == STATE_PROGRAM_MATRIX)
-             && (state_tokens[2] != state_tokens[3])) {
-            GLint row;
-            const GLint first_row = state_tokens[2];
-            const GLint last_row = state_tokens[3];
-
-            for (row = first_row; row <= last_row; row++) {
-               state_tokens[2] = state_tokens[3] = row;
-
-               idx = _mesa_add_state_reference(Program->Base.Parameters,
-                                               state_tokens);
-               if (param_var->param_binding_begin == ~0U)
-                  param_var->param_binding_begin = idx;
-               param_var->param_binding_length++;
-            }
-         }
-         else {
-            idx = _mesa_add_state_reference(Program->Base.Parameters,
-                                            state_tokens);
-            if (param_var->param_binding_begin == ~0U)
-               param_var->param_binding_begin = idx;
-            param_var->param_binding_length++;
-         }
-         break;
-
-      case PARAM_PROGRAM_ELEMENT:
-         if (parse_program_single_item (ctx, inst, Program, state_tokens))
-            return 1;
-         idx = _mesa_add_state_reference (Program->Base.Parameters, state_tokens);
-         if (param_var->param_binding_begin == ~0U)
-            param_var->param_binding_begin = idx;
-         param_var->param_binding_length++;
-
-         /* Check if there is more: 0 -> we're done, else its an integer */
-         if (**inst) {
-            GLuint out_of_range, new_idx;
-            GLuint start_idx = state_tokens[2] + 1;
-            GLuint end_idx = parse_integer (inst, Program);
-
-            out_of_range = 0;
-            if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-               if (((state_tokens[1] == STATE_ENV)
-                    && (end_idx >= ctx->Const.FragmentProgram.MaxEnvParams))
-                   || ((state_tokens[1] == STATE_LOCAL)
-                       && (end_idx >=
-                           ctx->Const.FragmentProgram.MaxLocalParams)))
-                  out_of_range = 1;
-            }
-            else {
-               if (((state_tokens[1] == STATE_ENV)
-                    && (end_idx >= ctx->Const.VertexProgram.MaxEnvParams))
-                   || ((state_tokens[1] == STATE_LOCAL)
-                       && (end_idx >=
-                           ctx->Const.VertexProgram.MaxLocalParams)))
-                  out_of_range = 1;
-            }
-            if (out_of_range) {
-               program_error(ctx, Program->Position,
-                             "Invalid Program Parameter"); /*end_idx*/
-               return 1;
-            }
-
-            for (new_idx = start_idx; new_idx <= end_idx; new_idx++) {
-               state_tokens[2] = new_idx;
-               idx = _mesa_add_state_reference(Program->Base.Parameters,
-                                               state_tokens);
-               param_var->param_binding_length++;
-            }
-         }
-         else {
-            (*inst)++;
-         }
-         break;
-
-      case PARAM_CONSTANT:
-         /* parsing something like {1.0, 2.0, 3.0, 4.0} */
-         parse_constant (inst, const_values, Program, use);
-         idx = _mesa_add_named_constant(Program->Base.Parameters,
-                                        (char *) param_var->name,
-                                        const_values, 4);
-         if (param_var->param_binding_begin == ~0U)
-            param_var->param_binding_begin = idx;
-         param_var->param_binding_type = PROGRAM_STATE_VAR;
-         /* Note: when we reference this parameter in an instruction later,
-          * we'll check if it's really a constant/immediate and set the
-          * instruction register type appropriately.
-          */
-         param_var->param_binding_length++;
-         break;
-
-      default:
-         program_error(ctx, Program->Position,
-                       "Unexpected token (in parse_param_elements())");
-         return 1;
-   }
-
-   Program->Base.NumParameters = Program->Base.Parameters->NumParameters;
-
-   /* Make sure we haven't blown past our parameter limits */
-   if (((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) &&
-        (Program->Base.NumParameters >
-         ctx->Const.VertexProgram.MaxLocalParams))
-       || ((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB)
-           && (Program->Base.NumParameters >
-               ctx->Const.FragmentProgram.MaxLocalParams))) {
-      program_error(ctx, Program->Position, "Too many parameter variables");
-      return 1;
-   }
-
-   return err;
-}
-
-
-/**
- * This picks out PARAM program parameter bindings.
- *
- * XXX: This needs to be stressed & tested
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_param (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-             struct arb_program *Program)
-{
-   GLuint found, err;
-   GLint specified_length;
-   struct var_cache *param_var;
-
-   err = 0;
-   param_var = parse_string (inst, vc_head, Program, &found);
-   Program->Position = parse_position (inst);
-
-   if (found) {
-      program_error2(ctx, Program->Position,
-                     "Duplicate variable declaration",
-                     (char *) param_var->name);
-      return 1;
-   }
-
-   specified_length = parse_integer (inst, Program);
-
-   if (specified_length < 0) {
-      program_error(ctx, Program->Position, "Negative parameter array length");
-      return 1;
-   }
-
-   param_var->type = vt_param;
-   param_var->param_binding_length = 0;
-
-   /* Right now, everything is shoved into the main state register file.
-    *
-    * In the future, it would be nice to leave things ENV/LOCAL params
-    * in their respective register files, if possible
-    */
-   param_var->param_binding_type = PROGRAM_STATE_VAR;
-
-   /* Remember to:
-    * *   - add each guy to the parameter list
-    * *   - increment the param_var->param_binding_len
-    * *   - store the param_var->param_binding_begin for the first one
-    * *   - compare the actual len to the specified len at the end
-    */
-   while (**inst != PARAM_NULL) {
-      if (parse_param_elements (ctx, inst, param_var, Program, GL_FALSE))
-         return 1;
-   }
-
-   /* Test array length here! */
-   if (specified_length) {
-      if (specified_length != (int)param_var->param_binding_length) {
-         program_error(ctx, Program->Position,
-              "Declared parameter array length does not match parameter list");
-         return 1;
-      }
-   }
-
-   (*inst)++;
-
-   return 0;
-}
-
-/**
- *
- */
-static GLuint
-parse_param_use (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-                 struct arb_program *Program, struct var_cache **new_var)
-{
-   struct var_cache *param_var;
-
-   /* First, insert a dummy entry into the var_cache */
-   var_cache_create (&param_var);
-   param_var->name = (const GLubyte *) " ";
-   param_var->type = vt_param;
-
-   param_var->param_binding_length = 0;
-   /* Don't fill in binding_begin; We use the default value of -1
-    * to tell if its already initialized, elsewhere.
-    *
-    * param_var->param_binding_begin  = 0;
-    */
-   param_var->param_binding_type = PROGRAM_STATE_VAR;
-
-   var_cache_append (vc_head, param_var);
-
-   /* Then fill it with juicy parameter goodness */
-   if (parse_param_elements (ctx, inst, param_var, Program, GL_TRUE))
-      return 1;
-
-   *new_var = param_var;
-
-   return 0;
-}
-
-
-/**
- * This handles the declaration of TEMP variables
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_temp (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-            struct arb_program *Program)
-{
-   GLuint found;
-   struct var_cache *temp_var;
-
-   while (**inst != 0) {
-      temp_var = parse_string (inst, vc_head, Program, &found);
-      Program->Position = parse_position (inst);
-      if (found) {
-         program_error2(ctx, Program->Position,
-                        "Duplicate variable declaration",
-                        (char *) temp_var->name);
-         return 1;
-      }
-
-      temp_var->type = vt_temp;
-
-      if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) &&
-           (Program->Base.NumTemporaries >=
-            ctx->Const.FragmentProgram.MaxTemps))
-          || ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB)
-              && (Program->Base.NumTemporaries >=
-                  ctx->Const.VertexProgram.MaxTemps))) {
-         program_error(ctx, Program->Position,
-                       "Too many TEMP variables declared");
-         return 1;
-      }
-
-      temp_var->temp_binding = Program->Base.NumTemporaries;
-      Program->Base.NumTemporaries++;
-   }
-   (*inst)++;
-
-   return 0;
-}
-
-/**
- * This handles variables of the OUTPUT variety
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_output (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-              struct arb_program *Program)
-{
-   GLuint found;
-   struct var_cache *output_var;
-   GLuint err;
-
-   output_var = parse_string (inst, vc_head, Program, &found);
-   Program->Position = parse_position (inst);
-   if (found) {
-      program_error2(ctx, Program->Position,
-                     "Duplicate variable declaration",
-                     (char *) output_var->name);
-      return 1;
-   }
-
-   output_var->type = vt_output;
-
-   err = parse_result_binding(ctx, inst, &output_var->output_binding, Program);
-   return err;
-}
-
-/**
- * This handles variables of the ALIAS kind
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_alias (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-             struct arb_program *Program)
-{
-   GLuint found;
-   struct var_cache *temp_var;
-
-   temp_var = parse_string (inst, vc_head, Program, &found);
-   Program->Position = parse_position (inst);
-
-   if (found) {
-      program_error2(ctx, Program->Position,
-                    "Duplicate variable declaration",
-                     (char *) temp_var->name);
-      return 1;
-   }
-
-   temp_var->type = vt_alias;
-   temp_var->alias_binding =  parse_string (inst, vc_head, Program, &found);
-   Program->Position = parse_position (inst);
-
-   if (!found)
-   {
-      program_error2(ctx, Program->Position,
-                     "Undefined alias value",
-                     (char *) temp_var->alias_binding->name);
-      return 1;
-   }
-
-   return 0;
-}
-
-/**
- * This handles variables of the ADDRESS kind
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_address (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-               struct arb_program *Program)
-{
-   GLuint found;
-   struct var_cache *temp_var;
-
-   while (**inst != 0) {
-      temp_var = parse_string (inst, vc_head, Program, &found);
-      Program->Position = parse_position (inst);
-      if (found) {
-         program_error2(ctx, Program->Position,
-                        "Duplicate variable declaration",
-                        (char *) temp_var->name);
-         return 1;
-      }
-
-      temp_var->type = vt_address;
-
-      if (Program->Base.NumAddressRegs >=
-          ctx->Const.VertexProgram.MaxAddressRegs) {
-         const char *msg = "Too many ADDRESS variables declared";
-         program_error(ctx, Program->Position, msg);
-         return 1;
-      }
-
-      temp_var->address_binding = Program->Base.NumAddressRegs;
-      Program->Base.NumAddressRegs++;
-   }
-   (*inst)++;
-
-   return 0;
-}
-
-/**
- * Parse a program declaration
- *
- * \return 0 on sucess, 1 on error
- */
-static GLint
-parse_declaration (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head,
-                   struct arb_program *Program)
-{
-   GLint err = 0;
-
-   switch (*(*inst)++) {
-      case ADDRESS:
-         err = parse_address (ctx, inst, vc_head, Program);
-         break;
-
-      case ALIAS:
-         err = parse_alias (ctx, inst, vc_head, Program);
-         break;
-
-      case ATTRIB:
-         err = parse_attrib (ctx, inst, vc_head, Program);
-         break;
-
-      case OUTPUT:
-         err = parse_output (ctx, inst, vc_head, Program);
-         break;
-
-      case PARAM:
-         err = parse_param (ctx, inst, vc_head, Program);
-         break;
-
-      case TEMP:
-         err = parse_temp (ctx, inst, vc_head, Program);
-         break;
-   }
-
-   return err;
-}
-
-/**
- * Handle the parsing out of a masked destination register, either for a
- * vertex or fragment program.
- *
- * If we are a vertex program, make sure we don't write to
- * result.position if we have specified that the program is
- * position invariant
- *
- * \param File      - The register file we write to
- * \param Index     - The register index we write to
- * \param WriteMask - The mask controlling which components we write (1->write)
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_masked_dst_reg (GLcontext * ctx, const GLubyte ** inst,
-                      struct var_cache **vc_head, struct arb_program *Program,
-                      gl_register_file *File, GLuint *Index, GLint *WriteMask)
-{
-   GLuint tmp, result;
-   struct var_cache *dst;
-
-   /* We either have a result register specified, or a
-    * variable that may or may not be writable
-    */
-   switch (*(*inst)++) {
-      case REGISTER_RESULT:
-         if (parse_result_binding(ctx, inst, Index, Program))
-            return 1;
-         *File = PROGRAM_OUTPUT;
-         break;
-
-      case REGISTER_ESTABLISHED_NAME:
-         dst = parse_string (inst, vc_head, Program, &result);
-         Program->Position = parse_position (inst);
-
-         /* If the name has never been added to our symbol table, we're hosed */
-         if (!result) {
-            program_error(ctx, Program->Position, "0: Undefined variable");
-            return 1;
-         }
-
-         switch (dst->type) {
-            case vt_output:
-               *File = PROGRAM_OUTPUT;
-               *Index = dst->output_binding;
-               break;
-
-            case vt_temp:
-               *File = PROGRAM_TEMPORARY;
-               *Index = dst->temp_binding;
-               break;
-
-               /* If the var type is not vt_output or vt_temp, no go */
-            default:
-               program_error(ctx, Program->Position,
-                             "Destination register is read only");
-               return 1;
-         }
-         break;
-
-      default:
-         program_error(ctx, Program->Position,
-                       "Unexpected opcode in parse_masked_dst_reg()");
-         return 1;
-   }
-
-
-   /* Position invariance test */
-   /* This test is done now in syntax portion - when position invariance OPTION
-      is specified, "result.position" rule is disabled so there is no way
-      to write the position
-   */
-   /*if ((Program->HintPositionInvariant) && (*File == PROGRAM_OUTPUT) &&
-      (*Index == 0))   {
-      program_error(ctx, Program->Position,
-                  "Vertex program specified position invariance and wrote vertex position");
-   }*/
-
-   /* And then the mask.
-    *  w,a -> bit 0
-    *  z,b -> bit 1
-    *  y,g -> bit 2
-    *  x,r -> bit 3
-    *
-    * ==> Need to reverse the order of bits for this!
-    */
-   tmp =  (GLint) *(*inst)++;
-   *WriteMask = (((tmp>>3) & 0x1) |
-		 ((tmp>>1) & 0x2) |
-		 ((tmp<<1) & 0x4) |
-		 ((tmp<<3) & 0x8));
-
-   return 0;
-}
-
-
-/**
- * Handle the parsing of a address register
- *
- * \param Index     - The register index we write to
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_address_reg (GLcontext * ctx, const GLubyte ** inst,
-                          struct var_cache **vc_head,
-                          struct arb_program *Program, GLint * Index)
-{
-   struct var_cache *dst;
-   GLuint result;
-
-   *Index = 0; /* XXX */
-
-   dst = parse_string (inst, vc_head, Program, &result);
-   Program->Position = parse_position (inst);
-
-   /* If the name has never been added to our symbol table, we're hosed */
-   if (!result) {
-      program_error(ctx, Program->Position, "Undefined variable");
-      return 1;
-   }
-
-   if (dst->type != vt_address) {
-      program_error(ctx, Program->Position, "Variable is not of type ADDRESS");
-      return 1;
-   }
-
-   return 0;
-}
-
-#if 0 /* unused */
-/**
- * Handle the parsing out of a masked address register
- *
- * \param Index     - The register index we write to
- * \param WriteMask - The mask controlling which components we write (1->write)
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_masked_address_reg (GLcontext * ctx, const GLubyte ** inst,
-                          struct var_cache **vc_head,
-                          struct arb_program *Program, GLint * Index,
-                          GLboolean * WriteMask)
-{
-   if (parse_address_reg (ctx, inst, vc_head, Program, Index))
-      return 1;
-
-   /* This should be 0x8 */
-   (*inst)++;
-
-   /* Writemask of .x is implied */
-   WriteMask[0] = 1;
-   WriteMask[1] = WriteMask[2] = WriteMask[3] = 0;
-
-   return 0;
-}
-#endif
-
-/**
- * Parse out a swizzle mask.
- *
- * Basically convert COMPONENT_X/Y/Z/W to SWIZZLE_X/Y/Z/W
- *
- * The len parameter allows us to grab 4 components for a vector
- * swizzle, or just 1 component for a scalar src register selection
- */
-static void
-parse_swizzle_mask(const GLubyte ** inst, GLubyte *swizzle, GLint len)
-{
-   GLint i;
-
-   for (i = 0; i < 4; i++)
-      swizzle[i] = i;
-
-   for (i = 0; i < len; i++) {
-      switch (*(*inst)++) {
-         case COMPONENT_X:
-            swizzle[i] = SWIZZLE_X;
-            break;
-         case COMPONENT_Y:
-            swizzle[i] = SWIZZLE_Y;
-            break;
-         case COMPONENT_Z:
-            swizzle[i] = SWIZZLE_Z;
-            break;
-         case COMPONENT_W:
-            swizzle[i] = SWIZZLE_W;
-            break;
-         default:
-            _mesa_problem(NULL, "bad component in parse_swizzle_mask()");
-            return;
-      }
-   }
-}
-
-
-/**
- * Parse an extended swizzle mask which is a sequence of
- * four x/y/z/w/0/1 tokens.
- * \return swizzle  four swizzle values
- * \return negateMask  four element bitfield
- */
-static void
-parse_extended_swizzle_mask(const GLubyte **inst, GLubyte swizzle[4],
-                            GLubyte *negateMask)
-{
-   GLint i;
-
-   *negateMask = 0x0;
-   for (i = 0; i < 4; i++) {
-      GLubyte swz;
-      if (parse_sign(inst) == -1)
-         *negateMask |= (1 << i);
-
-      swz = *(*inst)++;
-
-      switch (swz) {
-         case COMPONENT_0:
-            swizzle[i] = SWIZZLE_ZERO;
-            break;
-         case COMPONENT_1:
-            swizzle[i] = SWIZZLE_ONE;
-            break;
-         case COMPONENT_X:
-            swizzle[i] = SWIZZLE_X;
-            break;
-         case COMPONENT_Y:
-            swizzle[i] = SWIZZLE_Y;
-            break;
-         case COMPONENT_Z:
-            swizzle[i] = SWIZZLE_Z;
-            break;
-         case COMPONENT_W:
-            swizzle[i] = SWIZZLE_W;
-            break;
-         default:
-            _mesa_problem(NULL, "bad case in parse_extended_swizzle_mask()");
-            return;
-      }
-   }
-}
-
-
-static GLuint
-parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
-               struct var_cache **vc_head,
-               struct arb_program *Program,
-               gl_register_file * File, GLint * Index,
-               GLboolean *IsRelOffset )
-{
-   struct var_cache *src;
-   GLuint binding, is_generic, found;
-   GLint offset;
-
-   *IsRelOffset = 0;
-
-   /* And the binding for the src */
-   switch (*(*inst)++) {
-      case REGISTER_ATTRIB:
-         if (parse_attrib_binding
-             (ctx, inst, Program, &binding, &is_generic))
-            return 1;
-         *File = PROGRAM_INPUT;
-         *Index = binding;
-
-         /* We need to insert a dummy variable into the var_cache so we can
-          * catch generic vertex attrib aliasing errors
-          */
-         var_cache_create(&src);
-         src->type = vt_attrib;
-         src->name = (const GLubyte *) "Dummy Attrib Variable";
-         src->attrib_binding = binding;
-         src->attrib_is_generic = is_generic;
-         var_cache_append(vc_head, src);
-         if (generic_attrib_check(*vc_head)) {
-            program_error(ctx, Program->Position,
-                          "Cannot use both a generic vertex attribute "
-                          "and a specific attribute of the same type");
-            return 1;
-         }
-         break;
-
-      case REGISTER_PARAM:
-         switch (**inst) {
-            case PARAM_ARRAY_ELEMENT:
-               (*inst)++;
-               src = parse_string (inst, vc_head, Program, &found);
-               Program->Position = parse_position (inst);
-
-               if (!found) {
-                  program_error2(ctx, Program->Position,
-                                 "Undefined variable",
-                                 (char *) src->name);
-                  return 1;
-               }
-
-               *File = (gl_register_file) src->param_binding_type;
-
-               switch (*(*inst)++) {
-                  case ARRAY_INDEX_ABSOLUTE:
-                     offset = parse_integer (inst, Program);
-
-                     if ((offset < 0)
-                         || (offset >= (int)src->param_binding_length)) {
-                        program_error(ctx, Program->Position,
-                                      "Index out of range");
-                        /* offset, src->name */
-                        return 1;
-                     }
-
-                     *Index = src->param_binding_begin + offset;
-                     break;
-
-                  case ARRAY_INDEX_RELATIVE:
-                     {
-                        GLint addr_reg_idx, rel_off;
-
-                        /* First, grab the address regiseter */
-                        if (parse_address_reg (ctx, inst, vc_head, Program, &addr_reg_idx))
-                           return 1;
-
-                        /* And the .x */
-                        ((*inst)++);
-                        ((*inst)++);
-                        ((*inst)++);
-                        ((*inst)++);
-
-                        /* Then the relative offset */
-                        if (parse_relative_offset(ctx, inst, Program, &rel_off)) return 1;
-
-                        /* And store it properly */
-                        *Index = src->param_binding_begin + rel_off;
-                        *IsRelOffset = 1;
-                     }
-                     break;
-               }
-               break;
-
-            default:
-               if (parse_param_use (ctx, inst, vc_head, Program, &src))
-                  return 1;
-
-               *File = (gl_register_file) src->param_binding_type;
-               *Index = src->param_binding_begin;
-               break;
-         }
-         break;
-
-      case REGISTER_ESTABLISHED_NAME:
-         src = parse_string (inst, vc_head, Program, &found);
-         Program->Position = parse_position (inst);
-
-         /* If the name has never been added to our symbol table, we're hosed */
-         if (!found) {
-            program_error(ctx, Program->Position,
-                          "3: Undefined variable"); /* src->name */
-            return 1;
-         }
-
-         switch (src->type) {
-            case vt_attrib:
-               *File = PROGRAM_INPUT;
-               *Index = src->attrib_binding;
-               break;
-
-               /* XXX: We have to handle offsets someplace in here!  -- or are those above? */
-            case vt_param:
-               *File = (gl_register_file) src->param_binding_type;
-               *Index = src->param_binding_begin;
-               break;
-
-            case vt_temp:
-               *File = PROGRAM_TEMPORARY;
-               *Index = src->temp_binding;
-               break;
-
-               /* If the var type is vt_output no go */
-            default:
-               program_error(ctx, Program->Position,
-                             "destination register is read only");
-               /* bad src->name */
-               return 1;
-         }
-         break;
-
-      default:
-         program_error(ctx, Program->Position,
-                       "Unknown token in parse_src_reg");
-         return 1;
-   }
-
-   if (*File == PROGRAM_STATE_VAR) {
-      gl_register_file file;
-
-      /* If we're referencing the Program->Parameters[] array, check if the
-       * parameter is really a constant/literal.  If so, set File to CONSTANT.
-       */
-      assert(*Index < (GLint) Program->Base.Parameters->NumParameters);
-      file = Program->Base.Parameters->Parameters[*Index].Type;
-      if (file == PROGRAM_CONSTANT)
-         *File = PROGRAM_CONSTANT;
-   }
-
-   /* Add attributes to InputsRead only if they are used the program.
-    * This avoids the handling of unused ATTRIB declarations in the drivers. */
-   if (*File == PROGRAM_INPUT)
-      Program->Base.InputsRead |= (1 << *Index);
-
-   return 0;
-}
-
-
-/**
- * Parse vertex/fragment program vector source register.
- */
-static GLuint
-parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst,
-                     struct var_cache **vc_head,
-                     struct arb_program *program,
-                     struct prog_src_register *reg)
-{
-   gl_register_file file;
-   GLint index;
-   GLubyte negateMask;
-   GLubyte swizzle[4];
-   GLboolean isRelOffset;
-
-   /* Grab the sign */
-   negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
-
-   /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
-      return 1;
-
-   /* finally, the swizzle */
-   parse_swizzle_mask(inst, swizzle, 4);
-
-   reg->File = file;
-   reg->Index = index;
-   reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]);
-   reg->Negate = negateMask;
-   reg->RelAddr = isRelOffset;
-   return 0;
-}
-
-
-/**
- * Parse vertex/fragment program scalar source register.
- */
-static GLuint
-parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst,
-                     struct var_cache **vc_head,
-                     struct arb_program *program,
-                     struct prog_src_register *reg)
-{
-   gl_register_file file;
-   GLint index;
-   GLubyte negateMask;
-   GLubyte swizzle[4];
-   GLboolean isRelOffset;
-
-   /* Grab the sign */
-   negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
-
-   /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
-      return 1;
-
-   /* finally, the swizzle */
-   parse_swizzle_mask(inst, swizzle, 1);
-
-   reg->File = file;
-   reg->Index = index;
-   reg->Swizzle = (swizzle[0] << 0);
-   reg->Negate = negateMask;
-   reg->RelAddr = isRelOffset;
-   return 0;
-}
-
-
-/**
- * Parse vertex/fragment program destination register.
- * \return 1 if error, 0 if no error.
- */
-static GLuint 
-parse_dst_reg(GLcontext * ctx, const GLubyte ** inst,
-              struct var_cache **vc_head, struct arb_program *program,
-              struct prog_dst_register *reg )
-{
-   GLint mask;
-   GLuint idx;
-   gl_register_file file;
-
-   if (parse_masked_dst_reg (ctx, inst, vc_head, program, &file, &idx, &mask))
-      return 1;
-
-   reg->File = file;
-   reg->Index = idx;
-   reg->WriteMask = mask;
-   return 0;
-}
-
-
-/**
- * This is a big mother that handles getting opcodes into the instruction
- * and handling the src & dst registers for fragment program instructions
- * \return 1 if error, 0 if no error
- */
-static GLuint
-parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst,
-                      struct var_cache **vc_head, struct arb_program *Program,
-                      struct prog_instruction *fp)
-{
-   GLint a;
-   GLuint texcoord;
-   GLubyte instClass, type, code;
-   GLboolean rel;
-   GLuint shadow_tex = 0;
-
-   _mesa_init_instructions(fp, 1);
-
-   /* OP_ALU_INST or OP_TEX_INST */
-   instClass = *(*inst)++;
-
-   /* OP_ALU_{VECTOR, SCALAR, BINSC, BIN, TRI, SWZ},
-    * OP_TEX_{SAMPLE, KIL}
-    */
-   type = *(*inst)++;
-
-   /* The actual opcode name */
-   code = *(*inst)++;
-
-   /* Increment the correct count */
-   switch (instClass) {
-      case OP_ALU_INST:
-         Program->NumAluInstructions++;
-         break;
-      case OP_TEX_INST:
-         Program->NumTexInstructions++;
-         break;
-   }
-
-   switch (type) {
-      case OP_ALU_VECTOR:
-         switch (code) {
-            case OP_ABS_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_ABS:
-               fp->Opcode = OPCODE_ABS;
-               break;
-
-            case OP_FLR_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_FLR:
-               fp->Opcode = OPCODE_FLR;
-               break;
-
-            case OP_FRC_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_FRC:
-               fp->Opcode = OPCODE_FRC;
-               break;
-
-            case OP_LIT_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_LIT:
-               fp->Opcode = OPCODE_LIT;
-               break;
-
-            case OP_MOV_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_MOV:
-               fp->Opcode = OPCODE_MOV;
-               break;
-         }
-
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-         if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0]))
-            return 1;
-         break;
-
-      case OP_ALU_SCALAR:
-         switch (code) {
-            case OP_COS_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_COS:
-               fp->Opcode = OPCODE_COS;
-               break;
-
-            case OP_EX2_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_EX2:
-               fp->Opcode = OPCODE_EX2;
-               break;
-
-            case OP_LG2_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_LG2:
-               fp->Opcode = OPCODE_LG2;
-               break;
-
-            case OP_RCP_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_RCP:
-               fp->Opcode = OPCODE_RCP;
-               break;
-
-            case OP_RSQ_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_RSQ:
-               fp->Opcode = OPCODE_RSQ;
-               break;
-
-            case OP_SIN_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SIN:
-               fp->Opcode = OPCODE_SIN;
-               break;
-
-            case OP_SCS_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SCS:
-
-               fp->Opcode = OPCODE_SCS;
-               break;
-         }
-
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-         if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0]))
-            return 1;
-         break;
-
-      case OP_ALU_BINSC:
-         switch (code) {
-            case OP_POW_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_POW:
-               fp->Opcode = OPCODE_POW;
-               break;
-         }
-
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-         for (a = 0; a < 2; a++) {
-	    if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a]))
-               return 1;
-         }
-         break;
-
-
-      case OP_ALU_BIN:
-         switch (code) {
-            case OP_ADD_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_ADD:
-               fp->Opcode = OPCODE_ADD;
-               break;
-
-            case OP_DP3_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_DP3:
-               fp->Opcode = OPCODE_DP3;
-               break;
-
-            case OP_DP4_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_DP4:
-               fp->Opcode = OPCODE_DP4;
-               break;
-
-            case OP_DPH_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_DPH:
-               fp->Opcode = OPCODE_DPH;
-               break;
-
-            case OP_DST_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_DST:
-               fp->Opcode = OPCODE_DST;
-               break;
-
-            case OP_MAX_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_MAX:
-               fp->Opcode = OPCODE_MAX;
-               break;
-
-            case OP_MIN_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_MIN:
-               fp->Opcode = OPCODE_MIN;
-               break;
-
-            case OP_MUL_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_MUL:
-               fp->Opcode = OPCODE_MUL;
-               break;
-
-            case OP_SGE_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SGE:
-               fp->Opcode = OPCODE_SGE;
-               break;
-
-            case OP_SLT_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SLT:
-               fp->Opcode = OPCODE_SLT;
-               break;
-
-            case OP_SUB_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SUB:
-               fp->Opcode = OPCODE_SUB;
-               break;
-
-            case OP_XPD_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_XPD:
-               fp->Opcode = OPCODE_XPD;
-               break;
-         }
-
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-         for (a = 0; a < 2; a++) {
-	    if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a]))
-	       return 1;
-         }
-         break;
-
-      case OP_ALU_TRI:
-         switch (code) {
-            case OP_CMP_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_CMP:
-               fp->Opcode = OPCODE_CMP;
-               break;
-
-            case OP_LRP_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_LRP:
-               fp->Opcode = OPCODE_LRP;
-               break;
-
-            case OP_MAD_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_MAD:
-               fp->Opcode = OPCODE_MAD;
-               break;
-         }
-
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-         for (a = 0; a < 3; a++) {
-	    if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a]))
-	       return 1;
-         }
-         break;
-
-      case OP_ALU_SWZ:
-         switch (code) {
-            case OP_SWZ_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_SWZ:
-               fp->Opcode = OPCODE_SWZ;
-               break;
-         }
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-	 {
-	    GLubyte swizzle[4];
-	    GLubyte negateMask;
-            gl_register_file file;
-	    GLint index;
-
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &rel))
-	       return 1;
-	    parse_extended_swizzle_mask(inst, swizzle, &negateMask);
-	    fp->SrcReg[0].File = file;
-	    fp->SrcReg[0].Index = index;
-	    fp->SrcReg[0].Negate = negateMask;
-	    fp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0],
-                                                  swizzle[1],
-                                                  swizzle[2],
-                                                  swizzle[3]);
-	 }
-         break;
-
-      case OP_TEX_SAMPLE:
-         switch (code) {
-            case OP_TEX_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_TEX:
-               fp->Opcode = OPCODE_TEX;
-               break;
-
-            case OP_TXP_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_TXP:
-               fp->Opcode = OPCODE_TXP;
-               break;
-
-            case OP_TXB_SAT:
-               fp->SaturateMode = SATURATE_ZERO_ONE;
-            case OP_TXB:
-               fp->Opcode = OPCODE_TXB;
-               break;
-         }
-
-         if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg))
-            return 1;
-
-	 if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0]))
-            return 1;
-
-         /* texImageUnit */
-         if (parse_teximage_num (ctx, inst, Program, &texcoord))
-            return 1;
-         fp->TexSrcUnit = texcoord;
-
-         /* texTarget */
-         switch (*(*inst)++) {
-            case TEXTARGET_SHADOW1D:
-               shadow_tex = 1 << texcoord;
-               /* FALLTHROUGH */
-            case TEXTARGET_1D:
-               fp->TexSrcTarget = TEXTURE_1D_INDEX;
-               break;
-            case TEXTARGET_SHADOW2D:
-               shadow_tex = 1 << texcoord;
-               /* FALLTHROUGH */
-            case TEXTARGET_2D:
-               fp->TexSrcTarget = TEXTURE_2D_INDEX;
-               break;
-            case TEXTARGET_3D:
-               fp->TexSrcTarget = TEXTURE_3D_INDEX;
-               break;
-            case TEXTARGET_SHADOWRECT:
-               shadow_tex = 1 << texcoord;
-               /* FALLTHROUGH */
-            case TEXTARGET_RECT:
-               fp->TexSrcTarget = TEXTURE_RECT_INDEX;
-               break;
-            case TEXTARGET_CUBE:
-               fp->TexSrcTarget = TEXTURE_CUBE_INDEX;
-               break;
-            case TEXTARGET_SHADOW1D_ARRAY:
-               shadow_tex = 1 << texcoord;
-               /* FALLTHROUGH */
-            case TEXTARGET_1D_ARRAY:
-               fp->TexSrcTarget = TEXTURE_1D_ARRAY_INDEX;
-               break;
-            case TEXTARGET_SHADOW2D_ARRAY:
-               shadow_tex = 1 << texcoord;
-               /* FALLTHROUGH */
-            case TEXTARGET_2D_ARRAY:
-               fp->TexSrcTarget = TEXTURE_2D_ARRAY_INDEX;
-               break;
-         }
-
-         if (shadow_tex)
-            fp->TexShadow = 1;
-
-         /* Don't test the first time a particular sampler is seen.  Each time
-          * after that, make sure the shadow state is the same.
-          */
-         if ((_mesa_bitcount(Program->TexturesUsed[texcoord]) > 0)
-             && ((Program->ShadowSamplers & (1 << texcoord)) != shadow_tex)) {
-            program_error(ctx, Program->Position,
-                          "texture image unit used for shadow sampling and non-shadow sampling");
-            return 1;
-         }
-
-         Program->TexturesUsed[texcoord] |= (1 << fp->TexSrcTarget);
-         /* Check that both "2D" and "CUBE" (for example) aren't both used */
-         if (_mesa_bitcount(Program->TexturesUsed[texcoord]) > 1) {
-            program_error(ctx, Program->Position,
-                          "multiple targets used on one texture image unit");
-            return 1;
-         }
-      
-
-         Program->ShadowSamplers |= shadow_tex;
-         break;
-
-      case OP_TEX_KIL:
-         Program->UsesKill = 1;
-	 if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0]))
-            return 1;
-         fp->Opcode = OPCODE_KIL;
-         break;
-      default:
-         _mesa_problem(ctx, "bad type 0x%x in parse_fp_instruction()", type);
-         return 1;
-   }
-
-   return 0;
-}
-
-
-/**
- * Handle the parsing out of a masked address register
- *
- * \param Index     - The register index we write to
- * \param WriteMask - The mask controlling which components we write (1->write)
- *
- * \return 0 on sucess, 1 on error
- */
-static GLuint
-parse_vp_address_reg (GLcontext * ctx, const GLubyte ** inst,
-		      struct var_cache **vc_head,
-		      struct arb_program *Program,
-		      struct prog_dst_register *reg)
-{
-   GLint idx;
-
-   if (parse_address_reg (ctx, inst, vc_head, Program, &idx))
-      return 1;
-
-   /* This should be 0x8 */
-   (*inst)++;
-
-   reg->File = PROGRAM_ADDRESS;
-   reg->Index = idx;
-
-   /* Writemask of .x is implied */
-   reg->WriteMask = 0x1;
-   return 0;
-}
-
-
-/**
- * This is a big mother that handles getting opcodes into the instruction
- * and handling the src & dst registers for vertex program instructions
- */
-static GLuint
-parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst,
-                      struct var_cache **vc_head, struct arb_program *Program,
-                      struct prog_instruction *vp)
-{
-   GLint a;
-   GLubyte type, code;
-
-   /* OP_ALU_{ARL, VECTOR, SCALAR, BINSC, BIN, TRI, SWZ} */
-   type = *(*inst)++;
-
-   /* The actual opcode name */
-   code = *(*inst)++;
-
-   _mesa_init_instructions(vp, 1);
-
-   switch (type) {
-         /* XXX: */
-      case OP_ALU_ARL:
-         vp->Opcode = OPCODE_ARL;
-
-         /* Remember to set SrcReg.RelAddr; */
-
-         /* Get the masked address register [dst] */
-         if (parse_vp_address_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-         vp->DstReg.File = PROGRAM_ADDRESS;
-
-         /* Get a scalar src register */
-	 if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0]))
-            return 1;
-
-         break;
-
-      case OP_ALU_VECTOR:
-         switch (code) {
-            case OP_ABS:
-               vp->Opcode = OPCODE_ABS;
-               break;
-            case OP_FLR:
-               vp->Opcode = OPCODE_FLR;
-               break;
-            case OP_FRC:
-               vp->Opcode = OPCODE_FRC;
-               break;
-            case OP_LIT:
-               vp->Opcode = OPCODE_LIT;
-               break;
-            case OP_MOV:
-               vp->Opcode = OPCODE_MOV;
-               break;
-         }
-
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-         if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0]))
-            return 1;
-         break;
-
-      case OP_ALU_SCALAR:
-         switch (code) {
-            case OP_EX2:
-               vp->Opcode = OPCODE_EX2;
-               break;
-            case OP_EXP:
-               vp->Opcode = OPCODE_EXP;
-               break;
-            case OP_LG2:
-               vp->Opcode = OPCODE_LG2;
-               break;
-            case OP_LOG:
-               vp->Opcode = OPCODE_LOG;
-               break;
-            case OP_RCP:
-               vp->Opcode = OPCODE_RCP;
-               break;
-            case OP_RSQ:
-               vp->Opcode = OPCODE_RSQ;
-               break;
-         }
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-	 if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0]))
-            return 1;
-         break;
-
-      case OP_ALU_BINSC:
-         switch (code) {
-            case OP_POW:
-               vp->Opcode = OPCODE_POW;
-               break;
-         }
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-         for (a = 0; a < 2; a++) {
-	    if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a]))
-               return 1;
-         }
-         break;
-
-      case OP_ALU_BIN:
-         switch (code) {
-            case OP_ADD:
-               vp->Opcode = OPCODE_ADD;
-               break;
-            case OP_DP3:
-               vp->Opcode = OPCODE_DP3;
-               break;
-            case OP_DP4:
-               vp->Opcode = OPCODE_DP4;
-               break;
-            case OP_DPH:
-               vp->Opcode = OPCODE_DPH;
-               break;
-            case OP_DST:
-               vp->Opcode = OPCODE_DST;
-               break;
-            case OP_MAX:
-               vp->Opcode = OPCODE_MAX;
-               break;
-            case OP_MIN:
-               vp->Opcode = OPCODE_MIN;
-               break;
-            case OP_MUL:
-               vp->Opcode = OPCODE_MUL;
-               break;
-            case OP_SGE:
-               vp->Opcode = OPCODE_SGE;
-               break;
-            case OP_SLT:
-               vp->Opcode = OPCODE_SLT;
-               break;
-            case OP_SUB:
-               vp->Opcode = OPCODE_SUB;
-               break;
-            case OP_XPD:
-               vp->Opcode = OPCODE_XPD;
-               break;
-         }
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-         for (a = 0; a < 2; a++) {
-	    if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a]))
-               return 1;
-         }
-         break;
-
-      case OP_ALU_TRI:
-         switch (code) {
-            case OP_MAD:
-               vp->Opcode = OPCODE_MAD;
-               break;
-         }
-
-         if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-            return 1;
-
-         for (a = 0; a < 3; a++) {
-	    if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a]))
-               return 1;
-         }
-         break;
-
-      case OP_ALU_SWZ:
-         switch (code) {
-            case OP_SWZ:
-               vp->Opcode = OPCODE_SWZ;
-               break;
-         }
-	 {
-	    GLubyte swizzle[4]; 
-	    GLubyte negateMask;
-	    GLboolean relAddr;
-            gl_register_file file;
-	    GLint index;
-
-	    if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
-	       return 1;
-
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &relAddr))
-	       return 1;
-	    parse_extended_swizzle_mask (inst, swizzle, &negateMask);
-	    vp->SrcReg[0].File = file;
-	    vp->SrcReg[0].Index = index;
-	    vp->SrcReg[0].Negate = negateMask;
-	    vp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0],
-                                                  swizzle[1],
-                                                  swizzle[2],
-                                                  swizzle[3]);
-	    vp->SrcReg[0].RelAddr = relAddr;
-	 }
-         break;
-   }
-   return 0;
-}
-
-#if DEBUG_PARSING
-
-static GLvoid
-debug_variables (GLcontext * ctx, struct var_cache *vc_head,
-                 struct arb_program *Program)
-{
-   struct var_cache *vc;
-   GLint a, b;
-
-   fprintf (stderr, "debug_variables, vc_head: %p\n", (void*) vc_head);
-
-   /* First of all, print out the contents of the var_cache */
-   vc = vc_head;
-   while (vc) {
-      fprintf (stderr, "[%p]\n", (void*) vc);
-      switch (vc->type) {
-         case vt_none:
-            fprintf (stderr, "UNDEFINED %s\n", vc->name);
-            break;
-         case vt_attrib:
-            fprintf (stderr, "ATTRIB    %s\n", vc->name);
-            fprintf (stderr, "          binding: 0x%x\n", vc->attrib_binding);
-            break;
-         case vt_param:
-            fprintf (stderr, "PARAM     %s  begin: %d len: %d\n", vc->name,
-                     vc->param_binding_begin, vc->param_binding_length);
-            b = vc->param_binding_begin;
-            for (a = 0; a < vc->param_binding_length; a++) {
-               fprintf (stderr, "%s\n",
-                        Program->Base.Parameters->Parameters[a + b].Name);
-               if (Program->Base.Parameters->Parameters[a + b].Type == PROGRAM_STATE_VAR) {
-                  char *s;
-                  s = _mesa_program_state_string(Program->Base.Parameters->Parameters
-                                                 [a + b].StateIndexes);
-                  fprintf(stderr, "%s\n", s);
-                  _mesa_free(s);
-               }
-               else
-                  fprintf (stderr, "%f %f %f %f\n",
-                           Program->Base.Parameters->ParameterValues[a + b][0],
-                           Program->Base.Parameters->ParameterValues[a + b][1],
-                           Program->Base.Parameters->ParameterValues[a + b][2],
-                           Program->Base.Parameters->ParameterValues[a + b][3]);
-            }
-            break;
-         case vt_temp:
-            fprintf (stderr, "TEMP      %s\n", vc->name);
-            fprintf (stderr, "          binding: 0x%x\n", vc->temp_binding);
-            break;
-         case vt_output:
-            fprintf (stderr, "OUTPUT    %s\n", vc->name);
-            fprintf (stderr, "          binding: 0x%x\n", vc->output_binding);
-            break;
-         case vt_alias:
-            fprintf (stderr, "ALIAS     %s\n", vc->name);
-            fprintf (stderr, "          binding: 0x%p (%s)\n",
-                     (void*) vc->alias_binding, vc->alias_binding->name);
-            break;
-         default:
-            /* nothing */
-            ;
-      }
-      vc = vc->next;
-   }
-}
-
-#endif /* DEBUG_PARSING */
-
-
-/**
- * The main loop for parsing a fragment or vertex program
- *
- * \return 1 on error, 0 on success
- */
-static GLint
-parse_instructions(GLcontext * ctx, const GLubyte * inst,
-                   struct var_cache **vc_head, struct arb_program *Program)
-{
-   const GLuint maxInst = (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB)
-      ? ctx->Const.FragmentProgram.MaxInstructions
-      : ctx->Const.VertexProgram.MaxInstructions;
-   GLint err = 0;
-
-   ASSERT(MAX_PROGRAM_INSTRUCTIONS >= maxInst);
-
-   Program->MajorVersion = (GLuint) * inst++;
-   Program->MinorVersion = (GLuint) * inst++;
-
-   while (*inst != END) {
-      switch (*inst++) {
-
-         case OPTION:
-            switch (*inst++) {
-               case ARB_PRECISION_HINT_FASTEST:
-                  Program->PrecisionOption = GL_FASTEST;
-                  break;
-
-               case ARB_PRECISION_HINT_NICEST:
-                  Program->PrecisionOption = GL_NICEST;
-                  break;
-
-               case ARB_FOG_EXP:
-                  Program->FogOption = GL_EXP;
-                  break;
-
-               case ARB_FOG_EXP2:
-                  Program->FogOption = GL_EXP2;
-                  break;
-
-               case ARB_FOG_LINEAR:
-                  Program->FogOption = GL_LINEAR;
-                  break;
-
-               case ARB_POSITION_INVARIANT:
-                  if (Program->Base.Target == GL_VERTEX_PROGRAM_ARB)
-                     Program->HintPositionInvariant = GL_TRUE;
-                  break;
-
-               case ARB_FRAGMENT_PROGRAM_SHADOW:
-	          if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-	             /* TODO ARB_fragment_program_shadow code */
-		  }
-		  break;
-
-               case ARB_DRAW_BUFFERS:
-	          if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-                     /* do nothing for now */
-                  }
-                  break;
-
-               case MESA_TEXTURE_ARRAY:
-		  /* do nothing for now */
-                  break;
-            }
-            break;
-
-         case INSTRUCTION:
-            /* check length */
-            if (Program->Base.NumInstructions + 1 >= maxInst) {
-               program_error(ctx, Program->Position,
-                             "Max instruction count exceeded");
-               return 1;
-            }
-            Program->Position = parse_position (&inst);
-            /* parse the current instruction */
-            if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-               err = parse_fp_instruction (ctx, &inst, vc_head, Program,
-                      &Program->Base.Instructions[Program->Base.NumInstructions]);
-            }
-            else {
-               err = parse_vp_instruction (ctx, &inst, vc_head, Program,
-                      &Program->Base.Instructions[Program->Base.NumInstructions]);
-            }
-
-            /* increment instuction count */
-            Program->Base.NumInstructions++;
-            break;
-
-         case DECLARATION:
-            err = parse_declaration (ctx, &inst, vc_head, Program);
-            break;
-
-         default:
-            break;
-      }
-
-      if (err)
-         break;
-   }
-
-   /* Finally, tag on an OPCODE_END instruction */
-   {
-      const GLuint numInst = Program->Base.NumInstructions;
-      _mesa_init_instructions(Program->Base.Instructions + numInst, 1);
-      Program->Base.Instructions[numInst].Opcode = OPCODE_END;
-   }
-   Program->Base.NumInstructions++;
-
-   /*
-    * Initialize native counts to logical counts.  The device driver may
-    * change them if program is translated into a hardware program.
-    */
-   Program->Base.NumNativeInstructions = Program->Base.NumInstructions;
-   Program->Base.NumNativeTemporaries = Program->Base.NumTemporaries;
-   Program->Base.NumNativeParameters = Program->Base.NumParameters;
-   Program->Base.NumNativeAttributes = Program->Base.NumAttributes;
-   Program->Base.NumNativeAddressRegs = Program->Base.NumAddressRegs;
-
-   return err;
-}
-
-
-/* XXX temporary */
-LONGSTRING static char core_grammar_text[] =
-#include "shader/grammar/grammar_syn.h"
-;
-
-
-/**
- * Set a grammar parameter.
- * \param name the grammar parameter
- * \param value the new parameter value
- * \return 0 if OK, 1 if error
- */
-static int
-set_reg8 (GLcontext *ctx, grammar id, const char *name, GLubyte value)
-{
-   char error_msg[300];
-   GLint error_pos;
-
-   if (grammar_set_reg8 (id, (const byte *) name, value))
-      return 0;
-
-   grammar_get_last_error ((byte *) error_msg, 300, &error_pos);
-   _mesa_set_program_error (ctx, error_pos, error_msg);
-   _mesa_error (ctx, GL_INVALID_OPERATION, "Grammar Register Error");
-   return 1;
-}
-
-
-/**
- * Enable support for the given language option in the parser.
- * \return 1 if OK, 0 if error
- */
-static int
-enable_ext(GLcontext *ctx, grammar id, const char *name)
-{
-   return !set_reg8(ctx, id, name, 1);
-}
-
-
-/**
- * Enable parser extensions based on which OpenGL extensions are supported
- * by this rendering context.
- *
- * \return GL_TRUE if OK, GL_FALSE if error.
- */
-static GLboolean
-enable_parser_extensions(GLcontext *ctx, grammar id)
-{
-#if 0
-   /* These are not supported at this time */
-   if ((ctx->Extensions.ARB_vertex_blend ||
-        ctx->Extensions.EXT_vertex_weighting)
-       && !enable_ext(ctx, id, "vertex_blend"))
-      return GL_FALSE;
-   if (ctx->Extensions.ARB_matrix_palette
-       && !enable_ext(ctx, id, "matrix_palette"))
-      return GL_FALSE;
-#endif
-   if (ctx->Extensions.ARB_fragment_program_shadow
-       && !enable_ext(ctx, id, "fragment_program_shadow"))
-      return GL_FALSE;
-   if (ctx->Extensions.EXT_point_parameters
-       && !enable_ext(ctx, id, "point_parameters"))
-      return GL_FALSE;
-   if (ctx->Extensions.EXT_secondary_color
-       && !enable_ext(ctx, id, "secondary_color"))
-      return GL_FALSE;
-   if (ctx->Extensions.EXT_fog_coord
-       && !enable_ext(ctx, id, "fog_coord"))
-      return GL_FALSE;
-   if (ctx->Extensions.NV_texture_rectangle
-       && !enable_ext(ctx, id, "texture_rectangle"))
-      return GL_FALSE;
-   if (!enable_ext(ctx, id, "draw_buffers"))
-      return GL_FALSE;
-   if (ctx->Extensions.MESA_texture_array
-       && !enable_ext(ctx, id, "texture_array"))
-      return GL_FALSE;
-#if 1
-   /* hack for Warcraft (see bug 8060) */
-   enable_ext(ctx, id, "vertex_blend");
-#endif
-
-   return GL_TRUE;
-}
-
-
-/**
- * This kicks everything off.
- *
- * \param ctx - The GL Context
- * \param str - The program string
- * \param len - The program string length
- * \param program - The arb_program struct to return all the parsed info in
- * \return GL_TRUE on sucess, GL_FALSE on error
- */
-static GLboolean
-_mesa_parse_arb_program(GLcontext *ctx, GLenum target,
-                        const GLubyte *str, GLsizei len,
-                        struct arb_program *program)
-{
-   GLint a, err, error_pos;
-   char error_msg[300];
-   GLuint parsed_len;
-   struct var_cache *vc_head;
-   grammar arbprogram_syn_id;
-   GLubyte *parsed, *inst;
-   GLubyte *strz = NULL;
-   static int arbprogram_syn_is_ok = 0;		/* XXX temporary */
-
-   /* set the program target before parsing */
-   program->Base.Target = target;
-
-   /* Reset error state */
-   _mesa_set_program_error(ctx, -1, NULL);
-
-   /* check if arb_grammar_text (arbprogram.syn) is syntactically correct */
-   if (!arbprogram_syn_is_ok) {
-      /* One-time initialization of parsing system */
-      grammar grammar_syn_id;
-      GLuint parsed_len;
-
-      grammar_syn_id = grammar_load_from_text ((byte *) core_grammar_text);
-      if (grammar_syn_id == 0) {
-         grammar_get_last_error ((byte *) error_msg, 300, &error_pos);
-         /* XXX this is not a GL error - it's an implementation bug! - FIX */
-         _mesa_set_program_error (ctx, error_pos, error_msg);
-         _mesa_error (ctx, GL_INVALID_OPERATION,
-                      "glProgramStringARB(Error loading grammar rule set)");
-         return GL_FALSE;
-      }
-
-      err = !grammar_check(grammar_syn_id, (byte *) arb_grammar_text,
-                           &parsed, &parsed_len);
-
-      /* 'parsed' is unused here */
-      _mesa_free (parsed);
-      parsed = NULL;
-
-      /* NOTE: we can't destroy grammar_syn_id right here because
-       * grammar_destroy() can reset the last error
-       */
-      if (err) {
-         /* XXX this is not a GL error - it's an implementation bug! - FIX */
-         grammar_get_last_error ((byte *) error_msg, 300, &error_pos);
-         _mesa_set_program_error (ctx, error_pos, error_msg);
-         _mesa_error (ctx, GL_INVALID_OPERATION,
-                      "glProgramString(Error loading grammar rule set");
-         grammar_destroy (grammar_syn_id);
-         return GL_FALSE;
-      }
-
-      grammar_destroy (grammar_syn_id);
-
-      arbprogram_syn_is_ok = 1;
-   }
-
-   /* create the grammar object */
-   arbprogram_syn_id = grammar_load_from_text ((byte *) arb_grammar_text);
-   if (arbprogram_syn_id == 0) {
-      /* XXX this is not a GL error - it's an implementation bug! - FIX */
-      grammar_get_last_error ((GLubyte *) error_msg, 300, &error_pos);
-      _mesa_set_program_error (ctx, error_pos, error_msg);
-      _mesa_error (ctx, GL_INVALID_OPERATION,
-                   "glProgramString(Error loading grammer rule set)");
-      return GL_FALSE;
-   }
-
-   /* Set program_target register value */
-   if (set_reg8 (ctx, arbprogram_syn_id, "program_target",
-      program->Base.Target == GL_FRAGMENT_PROGRAM_ARB ? 0x10 : 0x20)) {
-      grammar_destroy (arbprogram_syn_id);
-      return GL_FALSE;
-   }
-
-   if (!enable_parser_extensions(ctx, arbprogram_syn_id)) {
-      grammar_destroy(arbprogram_syn_id);
-      return GL_FALSE;
-   }
-
-   /* check for NULL character occurences */
-   {
-      GLint i;
-      for (i = 0; i < len; i++) {
-         if (str[i] == '\0') {
-            program_error(ctx, i, "illegal character");
-            grammar_destroy (arbprogram_syn_id);
-            return GL_FALSE;
-         }
-      }
-   }
-
-   /* copy the program string to a null-terminated string */
-   strz = (GLubyte *) _mesa_malloc (len + 1);
-   if (!strz) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glProgramStringARB");
-      grammar_destroy (arbprogram_syn_id);
-      return GL_FALSE;
-   }
-   _mesa_memcpy (strz, str, len);
-   strz[len] = '\0';
-
-   /* do a fast check on program string - initial production buffer is 4K */
-   err = !grammar_fast_check(arbprogram_syn_id, strz,
-                             &parsed, &parsed_len, 0x1000);
-
-   /* Syntax parse error */
-   if (err) {
-      grammar_get_last_error((GLubyte *) error_msg, 300, &error_pos);
-      program_error(ctx, error_pos, error_msg);
-
-#if DEBUG_PARSING
-      /* useful for debugging */
-      do {
-         int line, col;
-         char *s;
-         fprintf(stderr, "program: %s\n", (char *) strz);
-         fprintf(stderr, "Error Pos: %d\n", ctx->Program.ErrorPos);
-         s = (char *) _mesa_find_line_column(strz, strz+ctx->Program.ErrorPos,
-                                             &line, &col);
-         fprintf(stderr, "line %d col %d: %s\n", line, col, s);
-      } while (0);
-#endif
-
-      _mesa_free(strz);
-      _mesa_free(parsed);
-
-      grammar_destroy (arbprogram_syn_id);
-      return GL_FALSE;
-   }
-
-   grammar_destroy (arbprogram_syn_id);
-
-   /*
-    * Program string is syntactically correct at this point
-    * Parse the tokenized version of the program now, generating
-    * vertex/fragment program instructions.
-    */
-
-   /* Initialize the arb_program struct */
-   program->Base.String = strz;
-   program->Base.Instructions = _mesa_alloc_instructions(MAX_PROGRAM_INSTRUCTIONS);
-   program->Base.NumInstructions =
-   program->Base.NumTemporaries =
-   program->Base.NumParameters =
-   program->Base.NumAttributes = program->Base.NumAddressRegs = 0;
-   program->Base.Parameters = _mesa_new_parameter_list ();
-   program->Base.InputsRead = 0x0;
-   program->Base.OutputsWritten = 0x0;
-   program->Position = 0;
-   program->MajorVersion = program->MinorVersion = 0;
-   program->PrecisionOption = GL_DONT_CARE;
-   program->FogOption = GL_NONE;
-   program->HintPositionInvariant = GL_FALSE;
-   for (a = 0; a < MAX_TEXTURE_IMAGE_UNITS; a++)
-      program->TexturesUsed[a] = 0x0;
-   program->ShadowSamplers = 0x0;
-   program->NumAluInstructions =
-   program->NumTexInstructions =
-   program->NumTexIndirections = 0;
-   program->UsesKill = 0;
-
-   vc_head = NULL;
-   err = GL_FALSE;
-
-   /* Start examining the tokens in the array */
-   inst = parsed;
-
-   /* Check the grammer rev */
-   if (*inst++ != REVISION) {
-      program_error (ctx, 0, "Grammar version mismatch");
-      err = GL_TRUE;
-   }
-   else {
-      /* ignore program target */
-      inst++;
-      err = parse_instructions(ctx, inst, &vc_head, program);
-   }
-
-   /*debug_variables(ctx, vc_head, program); */
-
-   /* We're done with the parsed binary array */
-   var_cache_destroy (&vc_head);
-
-   _mesa_free (parsed);
-
-   /* Reallocate the instruction array from size [MAX_PROGRAM_INSTRUCTIONS]
-    * to size [ap.Base.NumInstructions].
-    */
-   program->Base.Instructions
-      = _mesa_realloc_instructions(program->Base.Instructions,
-                                   MAX_PROGRAM_INSTRUCTIONS,
-                                   program->Base.NumInstructions);
-
-   return !err;
-}
-
+#include "program_parser.h"
 
 
 void
@@ -3881,11 +71,18 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
                                  const GLvoid *str, GLsizei len,
                                  struct gl_fragment_program *program)
 {
-   struct arb_program ap;
+   struct gl_program prog;
+   struct asm_parser_state state;
    GLuint i;
 
    ASSERT(target == GL_FRAGMENT_PROGRAM_ARB);
-   if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, &ap)) {
+
+   memset(&prog, 0, sizeof(prog));
+   memset(&state, 0, sizeof(state));
+   state.prog = &prog;
+
+   if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len,
+				&state)) {
       /* Error in the program. Just return. */
       return;
    }
@@ -3893,51 +90,50 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
    /* Copy the relevant contents of the arb_program struct into the
     * fragment_program struct.
     */
-   program->Base.String          = ap.Base.String;
-   program->Base.NumInstructions = ap.Base.NumInstructions;
-   program->Base.NumTemporaries  = ap.Base.NumTemporaries;
-   program->Base.NumParameters   = ap.Base.NumParameters;
-   program->Base.NumAttributes   = ap.Base.NumAttributes;
-   program->Base.NumAddressRegs  = ap.Base.NumAddressRegs;
-   program->Base.NumNativeInstructions = ap.Base.NumNativeInstructions;
-   program->Base.NumNativeTemporaries = ap.Base.NumNativeTemporaries;
-   program->Base.NumNativeParameters = ap.Base.NumNativeParameters;
-   program->Base.NumNativeAttributes = ap.Base.NumNativeAttributes;
-   program->Base.NumNativeAddressRegs = ap.Base.NumNativeAddressRegs;
-   program->Base.NumAluInstructions   = ap.Base.NumAluInstructions;
-   program->Base.NumTexInstructions   = ap.Base.NumTexInstructions;
-   program->Base.NumTexIndirections   = ap.Base.NumTexIndirections;
-   program->Base.NumNativeAluInstructions = ap.Base.NumAluInstructions;
-   program->Base.NumNativeTexInstructions = ap.Base.NumTexInstructions;
-   program->Base.NumNativeTexIndirections = ap.Base.NumTexIndirections;
-   program->Base.InputsRead      = ap.Base.InputsRead;
-   program->Base.OutputsWritten  = ap.Base.OutputsWritten;
+   program->Base.String          = prog.String;
+   program->Base.NumInstructions = prog.NumInstructions;
+   program->Base.NumTemporaries  = prog.NumTemporaries;
+   program->Base.NumParameters   = prog.NumParameters;
+   program->Base.NumAttributes   = prog.NumAttributes;
+   program->Base.NumAddressRegs  = prog.NumAddressRegs;
+   program->Base.NumNativeInstructions = prog.NumNativeInstructions;
+   program->Base.NumNativeTemporaries = prog.NumNativeTemporaries;
+   program->Base.NumNativeParameters = prog.NumNativeParameters;
+   program->Base.NumNativeAttributes = prog.NumNativeAttributes;
+   program->Base.NumNativeAddressRegs = prog.NumNativeAddressRegs;
+   program->Base.NumAluInstructions   = prog.NumAluInstructions;
+   program->Base.NumTexInstructions   = prog.NumTexInstructions;
+   program->Base.NumTexIndirections   = prog.NumTexIndirections;
+   program->Base.NumNativeAluInstructions = prog.NumAluInstructions;
+   program->Base.NumNativeTexInstructions = prog.NumTexInstructions;
+   program->Base.NumNativeTexIndirections = prog.NumTexIndirections;
+   program->Base.InputsRead      = prog.InputsRead;
+   program->Base.OutputsWritten  = prog.OutputsWritten;
    for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) {
-      program->Base.TexturesUsed[i] = ap.TexturesUsed[i];
-      if (ap.TexturesUsed[i])
+      program->Base.TexturesUsed[i] = prog.TexturesUsed[i];
+      if (prog.TexturesUsed[i])
          program->Base.SamplersUsed |= (1 << i);
    }
-   program->Base.ShadowSamplers = ap.ShadowSamplers;
-   program->FogOption          = ap.FogOption;
-   program->UsesKill          = ap.UsesKill;
+   program->Base.ShadowSamplers = prog.ShadowSamplers;
+   switch (state.option.Fog) {
+   case OPTION_FOG_EXP:    program->FogOption = GL_EXP;    break;
+   case OPTION_FOG_EXP2:   program->FogOption = GL_EXP2;   break;
+   case OPTION_FOG_LINEAR: program->FogOption = GL_LINEAR; break;
+   default:                program->FogOption = GL_NONE;   break;
+   }
+
+   program->UsesKill            = state.fragment.UsesKill;
 
    if (program->FogOption)
       program->Base.InputsRead |= FRAG_BIT_FOGC;
 
-   /* XXX: assume that ARB fragment programs don't have access to the
-    * FrontFacing and PointCoord values stuffed into the fog
-    * coordinate in GLSL shaders.
-    */
-   if (program->Base.InputsRead & FRAG_BIT_FOGC)
-      program->UsesFogFragCoord = GL_TRUE;
-      
    if (program->Base.Instructions)
       _mesa_free(program->Base.Instructions);
-   program->Base.Instructions = ap.Base.Instructions;
+   program->Base.Instructions = prog.Instructions;
 
    if (program->Base.Parameters)
       _mesa_free_parameter_list(program->Base.Parameters);
-   program->Base.Parameters    = ap.Base.Parameters;
+   program->Base.Parameters    = prog.Parameters;
 
    /* Append fog instructions now if the program has "OPTION ARB_fog_exp"
     * or similar.  We used to leave this up to drivers, but it appears
@@ -3967,11 +163,17 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target,
 			       const GLvoid *str, GLsizei len,
 			       struct gl_vertex_program *program)
 {
-   struct arb_program ap;
+   struct gl_program prog;
+   struct asm_parser_state state;
 
    ASSERT(target == GL_VERTEX_PROGRAM_ARB);
 
-   if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, &ap)) {
+   memset(&prog, 0, sizeof(prog));
+   memset(&state, 0, sizeof(state));
+   state.prog = &prog;
+
+   if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len,
+				&state)) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glProgramString(bad program)");
       return;
    }
@@ -3979,28 +181,29 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target,
    /* Copy the relevant contents of the arb_program struct into the 
     * vertex_program struct.
     */
-   program->Base.String          = ap.Base.String;
-   program->Base.NumInstructions = ap.Base.NumInstructions;
-   program->Base.NumTemporaries  = ap.Base.NumTemporaries;
-   program->Base.NumParameters   = ap.Base.NumParameters;
-   program->Base.NumAttributes   = ap.Base.NumAttributes;
-   program->Base.NumAddressRegs  = ap.Base.NumAddressRegs;
-   program->Base.NumNativeInstructions = ap.Base.NumNativeInstructions;
-   program->Base.NumNativeTemporaries = ap.Base.NumNativeTemporaries;
-   program->Base.NumNativeParameters = ap.Base.NumNativeParameters;
-   program->Base.NumNativeAttributes = ap.Base.NumNativeAttributes;
-   program->Base.NumNativeAddressRegs = ap.Base.NumNativeAddressRegs;
-   program->Base.InputsRead     = ap.Base.InputsRead;
-   program->Base.OutputsWritten = ap.Base.OutputsWritten;
-   program->IsPositionInvariant = ap.HintPositionInvariant;
+   program->Base.String          = prog.String;
+   program->Base.NumInstructions = prog.NumInstructions;
+   program->Base.NumTemporaries  = prog.NumTemporaries;
+   program->Base.NumParameters   = prog.NumParameters;
+   program->Base.NumAttributes   = prog.NumAttributes;
+   program->Base.NumAddressRegs  = prog.NumAddressRegs;
+   program->Base.NumNativeInstructions = prog.NumNativeInstructions;
+   program->Base.NumNativeTemporaries = prog.NumNativeTemporaries;
+   program->Base.NumNativeParameters = prog.NumNativeParameters;
+   program->Base.NumNativeAttributes = prog.NumNativeAttributes;
+   program->Base.NumNativeAddressRegs = prog.NumNativeAddressRegs;
+   program->Base.InputsRead     = prog.InputsRead;
+   program->Base.OutputsWritten = prog.OutputsWritten;
+   program->IsPositionInvariant = (state.option.PositionInvariant)
+      ? GL_TRUE : GL_FALSE;
 
    if (program->Base.Instructions)
       _mesa_free(program->Base.Instructions);
-   program->Base.Instructions = ap.Base.Instructions;
+   program->Base.Instructions = prog.Instructions;
 
    if (program->Base.Parameters)
       _mesa_free_parameter_list(program->Base.Parameters);
-   program->Base.Parameters = ap.Base.Parameters; 
+   program->Base.Parameters = prog.Parameters; 
 
 #if DEBUG_VP
    _mesa_printf("____________Vertex program %u __________\n", program->Base.Id);
diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c
index a0331054bb..4d8cff0700 100644
--- a/src/mesa/shader/arbprogram.c
+++ b/src/mesa/shader/arbprogram.c
@@ -74,8 +74,6 @@ _mesa_BindProgram(GLenum target, GLuint id)
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-
    /* Error-check target and get curProg */
    if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */
         (ctx->Extensions.NV_vertex_program ||
@@ -132,6 +130,9 @@ _mesa_BindProgram(GLenum target, GLuint id)
       return;
    }
 
+   /* signal new program (and its new constants) */
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+
    /* bind newProg */
    if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */
       _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
@@ -253,6 +254,8 @@ _mesa_EnableVertexAttribArrayARB(GLuint index)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    FLUSH_VERTICES(ctx, _NEW_ARRAY);
    ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_TRUE;
    ctx->Array.ArrayObj->_Enabled |= _NEW_ARRAY_ATTRIB(index);
@@ -272,6 +275,8 @@ _mesa_DisableVertexAttribArrayARB(GLuint index)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    FLUSH_VERTICES(ctx, _NEW_ARRAY);
    ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_FALSE;
    ctx->Array.ArrayObj->_Enabled &= ~_NEW_ARRAY_ATTRIB(index);
@@ -298,32 +303,41 @@ _mesa_GetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble *params)
 }
 
 
+/**
+ * Return info for a generic vertex attribute array (no alias with
+ * legacy vertex attributes (pos, normal, color, etc)).
+ */
 void GLAPIENTRY
 _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS) {
+   if (index >= MAX_VERTEX_GENERIC_ATTRIBS) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glGetVertexAttribfvARB(index)");
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Enabled;
+         params[0] = (GLfloat) array->Enabled;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = (GLfloat) array->Size;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = (GLfloat) array->Stride;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = (GLfloat) array->Type;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Normalized;
+         params[0] = array->Normalized;
          break;
       case GL_CURRENT_VERTEX_ATTRIB_ARB:
          if (index == 0) {
@@ -335,7 +349,7 @@ _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params)
          COPY_4V(params, ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + index]);
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
+         params[0] = (GLfloat) array->BufferObj->Name;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribfvARB(pname)");
@@ -379,6 +393,8 @@ _mesa_GetVertexAttribPointervARB(GLuint index, GLenum pname, GLvoid **pointer)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    *pointer = (GLvoid *) ctx->Array.ArrayObj->VertexAttrib[index].Ptr;
 }
 
@@ -489,7 +505,7 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (target == GL_FRAGMENT_PROGRAM_ARB
        && ctx->Extensions.ARB_fragment_program) {
@@ -537,7 +553,7 @@ _mesa_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLfloat * dest;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameters4fv(count)");
@@ -631,7 +647,7 @@ _mesa_ProgramLocalParameter4fARB(GLenum target, GLuint index,
    struct gl_program *prog;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if ((target == GL_FRAGMENT_PROGRAM_NV
         && ctx->Extensions.NV_fragment_program) ||
@@ -685,7 +701,7 @@ _mesa_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLint i;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramLocalParameters4fv(count)");
diff --git a/src/mesa/shader/arbprogram.syn b/src/mesa/shader/arbprogram.syn
deleted file mode 100644
index b12c6a0eda..0000000000
--- a/src/mesa/shader/arbprogram.syn
+++ /dev/null
@@ -1,2824 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.2
- *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
- /**
- * \file arbprogram.syn
- * ARB_fragment/vertex_program syntax
- * \author Michal Krol
- */
-
-.syntax program;
-
-/*
-   This value must be incremented every time emit code values or structure of the production
-   array changes. This value is placed at the beginning of the production array. The loader
-   compares the value with its REVISION value. If they do not match, the loader is not up
-   to date.
-*/
-.emtcode REVISION                                   0x0a
-
-/* program type */
-.emtcode FRAGMENT_PROGRAM                           0x01
-.emtcode VERTEX_PROGRAM                             0x02
-
-/* program section */
-.emtcode OPTION                                     0x01
-.emtcode INSTRUCTION                                0x02
-.emtcode DECLARATION                                0x03
-.emtcode END                                        0x04
-
-/* GL_ARB_fragment_program option */
-.emtcode ARB_PRECISION_HINT_FASTEST                 0x00
-.emtcode ARB_PRECISION_HINT_NICEST                  0x01
-.emtcode ARB_FOG_EXP                                0x02
-.emtcode ARB_FOG_EXP2                               0x03
-.emtcode ARB_FOG_LINEAR                             0x04
-
-/* GL_ARB_vertex_program option */
-.emtcode ARB_POSITION_INVARIANT                     0x05
-
-/* GL_ARB_fragment_program_shadow option */
-.emtcode ARB_FRAGMENT_PROGRAM_SHADOW                0x06
-
-/* GL_ARB_draw_buffers option */
-.emtcode ARB_DRAW_BUFFERS                           0x07
-
-/* GL_MESA_texture_array option */
-.emtcode MESA_TEXTURE_ARRAY                         0x08
-
-/* GL_ARB_fragment_program instruction class */
-.emtcode OP_ALU_INST                                0x00
-.emtcode OP_TEX_INST                                0x01
-
-/* GL_ARB_vertex_program instruction class */
-/*       OP_ALU_INST */
-
-/* GL_ARB_fragment_program instruction type */
-.emtcode OP_ALU_VECTOR                               0x00
-.emtcode OP_ALU_SCALAR                               0x01
-.emtcode OP_ALU_BINSC                                0x02
-.emtcode OP_ALU_BIN                                  0x03
-.emtcode OP_ALU_TRI                                  0x04
-.emtcode OP_ALU_SWZ                                  0x05
-.emtcode OP_TEX_SAMPLE                               0x06
-.emtcode OP_TEX_KIL                                  0x07
-
-/* GL_ARB_vertex_program instruction type */
-.emtcode OP_ALU_ARL                                  0x08
-/*       OP_ALU_VECTOR */
-/*       OP_ALU_SCALAR */
-/*       OP_ALU_BINSC */
-/*       OP_ALU_BIN */
-/*       OP_ALU_TRI */
-/*       OP_ALU_SWZ */
-
-/* GL_ARB_fragment_program instruction code */
-.emtcode OP_ABS                                     0x00
-.emtcode OP_ABS_SAT                                 0x1B
-.emtcode OP_FLR                                     0x09
-.emtcode OP_FLR_SAT                                 0x26
-.emtcode OP_FRC                                     0x0A
-.emtcode OP_FRC_SAT                                 0x27
-.emtcode OP_LIT                                     0x0C
-.emtcode OP_LIT_SAT                                 0x2A
-.emtcode OP_MOV                                     0x11
-.emtcode OP_MOV_SAT                                 0x30
-.emtcode OP_COS                                     0x1F
-.emtcode OP_COS_SAT                                 0x20
-.emtcode OP_EX2                                     0x07
-.emtcode OP_EX2_SAT                                 0x25
-.emtcode OP_LG2                                     0x0B
-.emtcode OP_LG2_SAT                                 0x29
-.emtcode OP_RCP                                     0x14
-.emtcode OP_RCP_SAT                                 0x33
-.emtcode OP_RSQ                                     0x15
-.emtcode OP_RSQ_SAT                                 0x34
-.emtcode OP_SIN                                     0x38
-.emtcode OP_SIN_SAT                                 0x39
-.emtcode OP_SCS                                     0x35
-.emtcode OP_SCS_SAT                                 0x36
-.emtcode OP_POW                                     0x13
-.emtcode OP_POW_SAT                                 0x32
-.emtcode OP_ADD                                     0x01
-.emtcode OP_ADD_SAT                                 0x1C
-.emtcode OP_DP3                                     0x03
-.emtcode OP_DP3_SAT                                 0x21
-.emtcode OP_DP4                                     0x04
-.emtcode OP_DP4_SAT                                 0x22
-.emtcode OP_DPH                                     0x05
-.emtcode OP_DPH_SAT                                 0x23
-.emtcode OP_DST                                     0x06
-.emtcode OP_DST_SAT                                 0x24
-.emtcode OP_MAX                                     0x0F
-.emtcode OP_MAX_SAT                                 0x2E
-.emtcode OP_MIN                                     0x10
-.emtcode OP_MIN_SAT                                 0x2F
-.emtcode OP_MUL                                     0x12
-.emtcode OP_MUL_SAT                                 0x31
-.emtcode OP_SGE                                     0x16
-.emtcode OP_SGE_SAT                                 0x37
-.emtcode OP_SLT                                     0x17
-.emtcode OP_SLT_SAT                                 0x3A
-.emtcode OP_SUB                                     0x18
-.emtcode OP_SUB_SAT                                 0x3B
-.emtcode OP_XPD                                     0x1A
-.emtcode OP_XPD_SAT                                 0x43
-.emtcode OP_CMP                                     0x1D
-.emtcode OP_CMP_SAT                                 0x1E
-.emtcode OP_LRP                                     0x2B
-.emtcode OP_LRP_SAT                                 0x2C
-.emtcode OP_MAD                                     0x0E
-.emtcode OP_MAD_SAT                                 0x2D
-.emtcode OP_SWZ                                     0x19
-.emtcode OP_SWZ_SAT                                 0x3C
-.emtcode OP_TEX                                     0x3D
-.emtcode OP_TEX_SAT                                 0x3E
-.emtcode OP_TXB                                     0x3F
-.emtcode OP_TXB_SAT                                 0x40
-.emtcode OP_TXP                                     0x41
-.emtcode OP_TXP_SAT                                 0x42
-.emtcode OP_KIL                                     0x28
-
-/* GL_ARB_vertex_program instruction code */
-.emtcode OP_ARL                                     0x02
-/*       OP_ABS */
-/*       OP_FLR */
-/*       OP_FRC */
-/*       OP_LIT */
-/*       OP_MOV */
-/*       OP_EX2 */
-.emtcode OP_EXP                                     0x08
-/*       OP_LG2 */
-.emtcode OP_LOG                                     0x0D
-/*       OP_RCP */
-/*       OP_RSQ */
-/*       OP_POW */
-/*       OP_ADD */
-/*       OP_DP3 */
-/*       OP_DP4 */
-/*       OP_DPH */
-/*       OP_DST */
-/*       OP_MAX */
-/*       OP_MIN */
-/*       OP_MUL */
-/*       OP_SGE */
-/*       OP_SLT */
-/*       OP_SUB */
-/*       OP_XPD */
-/*       OP_MAD */
-/*       OP_SWZ */
-
-/* fragment attribute binding */
-.emtcode FRAGMENT_ATTRIB_COLOR                      0x01
-.emtcode FRAGMENT_ATTRIB_TEXCOORD                   0x02
-.emtcode FRAGMENT_ATTRIB_FOGCOORD                   0x03
-.emtcode FRAGMENT_ATTRIB_POSITION                   0x04
-
-/* vertex attribute binding */
-.emtcode VERTEX_ATTRIB_POSITION                     0x01
-.emtcode VERTEX_ATTRIB_WEIGHT                       0x02
-.emtcode VERTEX_ATTRIB_NORMAL                       0x03
-.emtcode VERTEX_ATTRIB_COLOR                        0x04
-.emtcode VERTEX_ATTRIB_FOGCOORD                     0x05
-.emtcode VERTEX_ATTRIB_TEXCOORD                     0x06
-.emtcode VERTEX_ATTRIB_MATRIXINDEX                  0x07
-.emtcode VERTEX_ATTRIB_GENERIC                      0x08
-
-/* fragment result binding */
-.emtcode FRAGMENT_RESULT_COLOR                      0x01
-.emtcode FRAGMENT_RESULT_DEPTH                      0x02
-
-/* vertex result binding */
-.emtcode VERTEX_RESULT_POSITION                     0x01
-.emtcode VERTEX_RESULT_COLOR                        0x02
-.emtcode VERTEX_RESULT_FOGCOORD                     0x03
-.emtcode VERTEX_RESULT_POINTSIZE                    0x04
-.emtcode VERTEX_RESULT_TEXCOORD                     0x05
-
-/* texture target */
-.emtcode TEXTARGET_1D                               0x01
-.emtcode TEXTARGET_2D                               0x02
-.emtcode TEXTARGET_3D                               0x03
-.emtcode TEXTARGET_RECT                             0x04
-.emtcode TEXTARGET_CUBE                             0x05
-/* GL_ARB_fragment_program_shadow */
-.emtcode TEXTARGET_SHADOW1D                         0x06
-.emtcode TEXTARGET_SHADOW2D                         0x07
-.emtcode TEXTARGET_SHADOWRECT                       0x08
-/* GL_MESA_texture_array */
-.emtcode TEXTARGET_1D_ARRAY                         0x09
-.emtcode TEXTARGET_2D_ARRAY                         0x0a
-.emtcode TEXTARGET_SHADOW1D_ARRAY                   0x0b
-.emtcode TEXTARGET_SHADOW2D_ARRAY                   0x0c
-
-/* face type */
-.emtcode FACE_FRONT                                 0x00
-.emtcode FACE_BACK                                  0x01
-
-/* color type */
-.emtcode COLOR_PRIMARY                              0x00
-.emtcode COLOR_SECONDARY                            0x01
-
-/* component */
-.emtcode COMPONENT_X                                0x00
-.emtcode COMPONENT_Y                                0x01
-.emtcode COMPONENT_Z                                0x02
-.emtcode COMPONENT_W                                0x03
-.emtcode COMPONENT_0                                0x04
-.emtcode COMPONENT_1                                0x05
-
-/* array index type */
-.emtcode ARRAY_INDEX_ABSOLUTE                       0x00
-.emtcode ARRAY_INDEX_RELATIVE                       0x01
-
-/* matrix name */
-.emtcode MATRIX_MODELVIEW                           0x01
-.emtcode MATRIX_PROJECTION                          0x02
-.emtcode MATRIX_MVP                                 0x03
-.emtcode MATRIX_TEXTURE                             0x04
-.emtcode MATRIX_PALETTE                             0x05
-.emtcode MATRIX_PROGRAM                             0x06
-
-/* matrix modifier */
-.emtcode MATRIX_MODIFIER_IDENTITY                   0x00
-.emtcode MATRIX_MODIFIER_INVERSE                    0x01
-.emtcode MATRIX_MODIFIER_TRANSPOSE                  0x02
-.emtcode MATRIX_MODIFIER_INVTRANS                   0x03
-
-/* constant type */
-.emtcode CONSTANT_SCALAR                            0x01
-.emtcode CONSTANT_VECTOR                            0x02
-
-/* program param type */
-.emtcode PROGRAM_PARAM_ENV                          0x01
-.emtcode PROGRAM_PARAM_LOCAL                        0x02
-
-/* register type */
-.emtcode REGISTER_ATTRIB                            0x01
-.emtcode REGISTER_PARAM                             0x02
-.emtcode REGISTER_RESULT                            0x03
-.emtcode REGISTER_ESTABLISHED_NAME                  0x04
-
-/* param binding */
-.emtcode PARAM_NULL                                 0x00
-.emtcode PARAM_ARRAY_ELEMENT                        0x01
-.emtcode PARAM_STATE_ELEMENT                        0x02
-.emtcode PARAM_PROGRAM_ELEMENT                      0x03
-.emtcode PARAM_PROGRAM_ELEMENTS                     0x04
-.emtcode PARAM_CONSTANT                             0x05
-
-/* param state property */
-.emtcode STATE_MATERIAL                             0x01
-.emtcode STATE_LIGHT                                0x02
-.emtcode STATE_LIGHT_MODEL                          0x03
-.emtcode STATE_LIGHT_PROD                           0x04
-.emtcode STATE_FOG                                  0x05
-.emtcode STATE_MATRIX_ROWS                          0x06
-/* GL_ARB_fragment_program */
-.emtcode STATE_TEX_ENV                              0x07
-.emtcode STATE_DEPTH                                0x08
-/* GL_ARB_vertex_program */
-.emtcode STATE_TEX_GEN                              0x09
-.emtcode STATE_CLIP_PLANE                           0x0A
-.emtcode STATE_POINT                                0x0B
-
-/* state material property */
-.emtcode MATERIAL_AMBIENT                           0x01
-.emtcode MATERIAL_DIFFUSE                           0x02
-.emtcode MATERIAL_SPECULAR                          0x03
-.emtcode MATERIAL_EMISSION                          0x04
-.emtcode MATERIAL_SHININESS                         0x05
-
-/* state light property */
-.emtcode LIGHT_AMBIENT                              0x01
-.emtcode LIGHT_DIFFUSE                              0x02
-.emtcode LIGHT_SPECULAR                             0x03
-.emtcode LIGHT_POSITION                             0x04
-.emtcode LIGHT_ATTENUATION                          0x05
-.emtcode LIGHT_HALF                                 0x06
-.emtcode LIGHT_SPOT_DIRECTION                       0x07
-
-/* state light model property */
-.emtcode LIGHT_MODEL_AMBIENT                        0x01
-.emtcode LIGHT_MODEL_SCENECOLOR                     0x02
-
-/* state light product property */
-.emtcode LIGHT_PROD_AMBIENT                         0x01
-.emtcode LIGHT_PROD_DIFFUSE                         0x02
-.emtcode LIGHT_PROD_SPECULAR                        0x03
-
-/* state texture environment property */
-.emtcode TEX_ENV_COLOR                              0x01
-
-/* state texture generation coord property */
-.emtcode TEX_GEN_EYE                                0x01
-.emtcode TEX_GEN_OBJECT                             0x02
-
-/* state fog property */
-.emtcode FOG_COLOR                                  0x01
-.emtcode FOG_PARAMS                                 0x02
-
-/* state depth property */
-.emtcode DEPTH_RANGE                                0x01
-
-/* state point parameters property */
-.emtcode POINT_SIZE                                 0x01
-.emtcode POINT_ATTENUATION                          0x02
-
-/* declaration */
-.emtcode ATTRIB                                     0x01
-.emtcode PARAM                                      0x02
-.emtcode TEMP                                       0x03
-.emtcode OUTPUT                                     0x04
-.emtcode ALIAS                                      0x05
-/* GL_ARB_vertex_program */
-.emtcode ADDRESS                                    0x06
-
-/* error messages */
-.errtext UNKNOWN_PROGRAM_SIGNATURE                  "1001: '$e_signature$': unknown program signature"
-.errtext MISSING_END_OR_INVALID_STATEMENT           "1002: '$e_statement$': invalid statement"
-.errtext CODE_AFTER_END                             "1003: '$e_statement$': code after 'END' keyword"
-.errtext INVALID_PROGRAM_OPTION                     "1004: '$e_identifier$': invalid program option"
-.errtext EXT_SWIZ_COMP_EXPECTED                     "1005: extended swizzle component expected but '$e_token$' found"
-.errtext TEX_TARGET_EXPECTED                        "1006: texture target expected but '$e_token$' found"
-.errtext TEXTURE_EXPECTED                           "1007: 'texture' expected but '$e_identifier$' found"
-.errtext SOURCE_REGISTER_EXPECTED                   "1008: source register expected but '$e_token$' found"
-.errtext DESTINATION_REGISTER_EXPECTED              "1009: destination register expected but '$e_token$' found"
-.errtext INVALID_ADDRESS_COMPONENT                  "1010: '$e_identifier$': invalid address component"
-.errtext INVALID_ADDRESS_WRITEMASK                  "1011: '$e_identifier$': invalid address writemask"
-.errtext INVALID_COMPONENT                          "1012: '$e_charordigit$': invalid component"
-.errtext INVALID_SUFFIX                             "1013: '$e_identifier$': invalid suffix"
-.errtext INVALID_WRITEMASK                          "1014: '$e_identifier$': invalid writemask"
-.errtext FRAGMENT_EXPECTED                          "1015: 'fragment' expected but '$e_identifier$' found"
-.errtext VERTEX_EXPECTED                            "1016: 'vertex' expected but '$e_identifier$' found"
-.errtext INVALID_FRAGMENT_PROPERTY                  "1017: '$e_identifier$': invalid fragment property"
-.errtext INVALID_VERTEX_PROPERTY                    "1018: '$e_identifier$': invalid vertex property"
-.errtext INVALID_STATE_PROPERTY                     "1019: '$e_identifier$': invalid state property"
-.errtext INVALID_MATERIAL_PROPERTY                  "1020: '$e_identifier$': invalid material property"
-.errtext INVALID_LIGHT_PROPERTY                     "1021: '$e_identifier$': invalid light property"
-.errtext INVALID_SPOT_PROPERTY                      "1022: '$e_identifier$': invalid spot property"
-.errtext INVALID_LIGHTMODEL_PROPERTY                "1023: '$e_identifier$': invalid light model property"
-.errtext INVALID_LIGHTPROD_PROPERTY                 "1024: '$e_identifier$': invalid light product property"
-.errtext INVALID_TEXENV_PROPERTY                    "1025: '$e_identifier$': invalid texture environment property"
-.errtext INVALID_TEXGEN_PROPERTY                    "1026: '$e_identifier$': invalid texture generating property"
-.errtext INVALID_TEXGEN_COORD                       "1027: '$e_identifier$': invalid texture generating coord"
-.errtext INVALID_FOG_PROPERTY                       "1028: '$e_identifier$': invalid fog property"
-.errtext INVALID_DEPTH_PROPERTY                     "1029: '$e_identifier$': invalid depth property"
-.errtext INVALID_CLIPPLANE_PROPERTY                 "1030: '$e_identifier$': invalid clip plane property"
-.errtext INVALID_POINT_PROPERTY                     "1031: '$e_identifier$': invalid point property"
-.errtext MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED   "1032: matrix row selector or modifier expected but '$e_token$' found"
-.errtext INVALID_MATRIX_NAME                        "1033: '$e_identifier$': invalid matrix name"
-.errtext INVALID_PROGRAM_PROPERTY                   "1034: '$e_identifier$': invalid program property"
-.errtext RESULT_EXPECTED                            "1035: 'result' expected but '$e_token$' found"
-.errtext INVALID_RESULT_PROPERTY                    "1036: '$e_identifier$': invalid result property"
-.errtext INVALID_FACE_PROPERTY                      "1037: '$e_identifier$': invalid face property"
-.errtext INVALID_COLOR_PROPERTY                     "1038: '$e_identifier$': invalid color property"
-.errtext IDENTIFIER_EXPECTED                        "1039: identifier expected but '$e_token$' found"
-.errtext RESERVED_KEYWORD                           "1040: use of reserved keyword as an identifier"
-.errtext INTEGER_EXPECTED                           "1041: integer value expected but '$e_token$' found"
-.errtext MISSING_SEMICOLON                          "1042: ';' expected but '$e_token$' found"
-.errtext MISSING_COMMA                              "1043: ',' expected but '$e_token$' found"
-.errtext MISSING_LBRACKET                           "1044: '[' expected but '$e_token$' found"
-.errtext MISSING_RBRACKET                           "1045: ']' expected but '$e_token$' found"
-.errtext MISSING_DOT                                "1046: '.' expected but '$e_token$' found"
-.errtext MISSING_EQUAL                              "1047: '=' expected but '$e_token$' found"
-.errtext MISSING_LBRACE                             "1048: '{' expected but '$e_token$' found"
-.errtext MISSING_RBRACE                             "1049: '}' expected but '$e_token$' found"
-.errtext MISSING_DOTDOT                             "1050: '..' expected but '$e_token$' found"
-.errtext MISSING_FRACTION_OR_EXPONENT               "1051: missing fraction part or exponent"
-.errtext MISSING_DOT_OR_EXPONENT                    "1052: missing '.' or exponent"
-.errtext EXPONENT_VALUE_EXPECTED                    "1053: exponent value expected"
-.errtext INTEGER_OUT_OF_RANGE                       "1054: integer value out of range"
-.errtext OPERATION_NEEDS_DESTINATION_VARIABLE       "1055: operation needs destination variable"
-.errtext OPERATION_NEEDS_SOURCE_VARIABLE            "1056: operation needs source variable"
-.errtext ADDRESS_REGISTER_EXPECTED                  "1057: address register expected but '$e_token$' found"
-.errtext ADDRESS_REGISTER_OR_INTEGER_EXPECTED       "1058: address register or integer literal expected but '$e_token$' found"
-
-/* extension presence condition registers */
-
-/* GL_ARB_vertex_blend */
-/* GL_EXT_vertex_weighting */
-.regbyte vertex_blend                               0x00
-
-/* GL_ARB_matrix_palette */
-.regbyte matrix_palette                             0x00
-
-/* GL_ARB_point_parameters */
-/* GL_EXT_point_parameters */
-.regbyte point_parameters                           0x00
-
-/* GL_EXT_secondary_color */
-.regbyte secondary_color                            0x00
-
-/* GL_EXT_fog_coord */
-.regbyte fog_coord                                  0x00
-
-/* GL_EXT_texture_rectangle */
-/* GL_NV_texture_rectangle */
-.regbyte texture_rectangle                          0x00
-
-/* GL_ARB_fragment_program_shadow */
-.regbyte fragment_program_shadow                    0x00
-
-/* GL_ARB_draw_buffers */
-.regbyte draw_buffers                               0x00
-
-/* GL_MESA_texture_array */
-.regbyte texture_array                              0x00
-
-/* option presence condition registers */
-/* they are all initially set to zero - when a particular OPTION is encountered, the appropriate */
-/* register is set to 1 to indicate that the OPTION was specified. */
-
-/* GL_ARB_fragment_program */
-.regbyte ARB_precision_hint_fastest                 0x00
-.regbyte ARB_precision_hint_nicest                  0x00
-.regbyte ARB_fog_exp                                0x00
-.regbyte ARB_fog_exp2                               0x00
-.regbyte ARB_fog_linear                             0x00
-
-/* GL_ARB_vertex_program */
-.regbyte ARB_position_invariant                     0x00
-
-/* GL_ARB_fragment_program_shadow */
-.regbyte ARB_fragment_program_shadow                0x00
-
-/* GL_ARB_draw_buffers */
-.regbyte ARB_draw_buffers                           0x00
-
-/* GL_MESA_texture_array */
-.regbyte MESA_texture_array                         0x00
-
-/* program target condition register */
-/* this syntax script deals with two program targets - VERTEX_PROGRAM and FRAGMENT_PROGRAM. */
-/* to distinguish between them we need a register that will store for us the current target. */
-/* the client will typically set the register to apropriate value before parsing a particular */
-/* program. the mapping between program targets and their values is listed below. */
-/* */
-/* program target               register value    */
-/* ---------------------------------------------- */
-/* FRAGMENT_PROGRAM             0x10              */
-/* VERTEX_PROGRAM               0x20              */
-/* */
-/* the initial value of the register is 0 to catch potential errors with not setting the register */
-/* with the proper value. */
-.regbyte program_target                             0x00
-
-/*
-    <program>              ::= <optionSequence> <statementSequence> "END"
-*/
-program
-    programs .error UNKNOWN_PROGRAM_SIGNATURE .emit REVISION;
-programs
-    .if (program_target == 0x10) frag_program_1_0 .emit FRAGMENT_PROGRAM .emit 0x01 .emit 0x00 .or
-    .if (program_target == 0x20) vert_program_1_0 .emit VERTEX_PROGRAM .emit 0x01 .emit 0x00;
-frag_program_1_0
-    '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'f' .and 'p' .and '1' .and '.' .and '0' .and
-    optional_space .and fp_optionSequence .and fp_statementSequence .and
-    "END" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and
-    '\0' .error CODE_AFTER_END;
-vert_program_1_0
-    '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'v' .and 'p' .and '1' .and '.' .and '0' .and
-    optional_space .and vp_optionSequence .and vp_statementSequence .and
-    "END" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and
-    '\0' .error CODE_AFTER_END;
-
-/*
-    <optionSequence>       ::= <optionSequence> <option>
-                             | ""
-*/
-fp_optionSequence
-    .loop fp_option;
-vp_optionSequence
-    .loop vp_option;
-
-/*
-    <option>               ::= "OPTION" <identifier> ";"
-
-NOTE: options ARB_precision_hint_nicest and ARB_precision_hint_fastest are exclusive. When one of
-      these options is encountered, the other one is automatically disabled.
-      the same applies to options ARB_fog_exp, ARB_fog_exp2 and ARB_fog_linear.
-*/
-fp_option
-    "OPTION" .emit OPTION .and space .error IDENTIFIER_EXPECTED .and
-    fp_optionString .error INVALID_PROGRAM_OPTION .and semicolon;
-vp_option
-    "OPTION" .emit OPTION .and space .error IDENTIFIER_EXPECTED .and
-    vp_optionString .error INVALID_PROGRAM_OPTION .and semicolon;
-fp_optionString
-    .if (ARB_precision_hint_nicest == 0x00) "ARB_precision_hint_fastest"
-        .emit ARB_PRECISION_HINT_FASTEST .load ARB_precision_hint_fastest 0x01 .or
-    .if (ARB_precision_hint_fastest == 0x00) "ARB_precision_hint_nicest"
-        .emit ARB_PRECISION_HINT_NICEST .load ARB_precision_hint_nicest 0x01 .or
-    fp_ARB_fog_exp .emit ARB_FOG_EXP .load ARB_fog_exp 0x01 .or
-    fp_ARB_fog_exp2 .emit ARB_FOG_EXP2 .load ARB_fog_exp2 0x01 .or
-    fp_ARB_fog_linear .emit ARB_FOG_LINEAR .load ARB_fog_linear 0x01 .or
-    .if (fragment_program_shadow != 0x00) "ARB_fragment_program_shadow"
-        .emit ARB_FRAGMENT_PROGRAM_SHADOW .load ARB_fragment_program_shadow 0x01 .or
-    .if (draw_buffers != 0x00) "ARB_draw_buffers" .emit ARB_DRAW_BUFFERS
-        .load ARB_draw_buffers 0x01 .or
-    .if (texture_array != 0x00) "MESA_texture_array" .emit MESA_TEXTURE_ARRAY
-        .load MESA_texture_array 0x01;
-vp_optionString
-    "ARB_position_invariant" .emit ARB_POSITION_INVARIANT .load ARB_position_invariant 0x01;
-fp_ARB_fog_exp
-    .if (ARB_fog_exp2 == 0x00) .true .and .if (ARB_fog_linear == 0x00) "ARB_fog_exp";
-fp_ARB_fog_exp2
-    .if (ARB_fog_exp == 0x00) .true .and .if (ARB_fog_linear == 0x00) "ARB_fog_exp2";
-fp_ARB_fog_linear
-    .if (ARB_fog_exp == 0x00) .true .and .if (ARB_fog_exp2 == 0x00) "ARB_fog_linear";
-
-/*
-    <statementSequence>    ::= <statementSequence> <statement>
-                             | ""
-*/
-fp_statementSequence
-    .loop fp_statement;
-vp_statementSequence
-    .loop vp_statement;
-
-/*
-    <statement>            ::= <instruction> ";"
-                             | <namingStatement> ";"
-
-NOTE: ".emit $" in the definitions below means that we output instruction position (offset of
-      the first character of instruction) for program debugging purposes.
-*/
-fp_statement
-    fp_statement_1 .or fp_statement_2;
-vp_statement
-    vp_statement_1 .or vp_statement_2;
-fp_statement_1
-    fp_instruction .emit INSTRUCTION .emit $ .and semicolon;
-fp_statement_2
-    fp_namingStatement .emit DECLARATION .and semicolon;
-vp_statement_1
-    vp_instruction .emit INSTRUCTION .emit $ .and semicolon;
-vp_statement_2
-    vp_namingStatement .emit DECLARATION .and semicolon;
-
-/*
-fragment program
-    <instruction>          ::= <ALUInstruction>
-                             | <TexInstruction>
-
-vertex program
-    <instruction>          ::= <ARL_instruction>
-                             | <VECTORop_instruction>
-                             | <SCALARop_instruction>
-                             | <BINSCop_instruction>
-                             | <BINop_instruction>
-                             | <TRIop_instruction>
-                             | <SWZ_instruction>
-*/
-fp_instruction
-    ALUInstruction .emit OP_ALU_INST .or
-    TexInstruction .emit OP_TEX_INST;
-vp_instruction
-    ARL_instruction .emit OP_ALU_ARL .or
-    vp_VECTORop_instruction .emit OP_ALU_VECTOR .or
-    vp_SCALARop_instruction .emit OP_ALU_SCALAR .or
-    vp_BINSCop_instruction .emit OP_ALU_BINSC .or
-    vp_BINop_instruction .emit OP_ALU_BIN .or
-    vp_TRIop_instruction .emit OP_ALU_TRI .or
-    vp_SWZ_instruction .emit OP_ALU_SWZ;
-
-/*
-fragment program
-    <ALUInstruction>       ::= <VECTORop_instruction>
-                             | <SCALARop_instruction>
-                             | <BINSCop_instruction>
-                             | <BINop_instruction>
-                             | <TRIop_instruction>
-                             | <SWZ_instruction>
-*/
-ALUInstruction
-    fp_VECTORop_instruction .emit OP_ALU_VECTOR .or
-    fp_SCALARop_instruction .emit OP_ALU_SCALAR .or
-    fp_BINSCop_instruction .emit OP_ALU_BINSC .or
-    fp_BINop_instruction .emit OP_ALU_BIN .or
-    fp_TRIop_instruction .emit OP_ALU_TRI .or
-    fp_SWZ_instruction .emit OP_ALU_SWZ;
-
-/*
-fragment program
-    <TexInstruction>       ::= <SAMPLE_instruction>
-                             | <KIL_instruction>
-*/
-TexInstruction
-    SAMPLE_instruction .emit OP_TEX_SAMPLE .or
-    KIL_instruction .emit OP_TEX_KIL;
-
-/*
-vertex program
-    <ARL_instruction>      ::= "ARL" <maskedAddrReg> "," <scalarSrcReg>
-*/
-ARL_instruction
-    "ARL" .emit OP_ARL .and space_dst .and maskedAddrReg .and comma .and vp_scalarSrcReg;
-
-/*
-fragment program
-    <VECTORop_instruction> ::= <VECTORop> <maskedDstReg> "," 
-                               <vectorSrcReg>
-
-vertex program
-    <VECTORop_instruction> ::= <VECTORop> <maskedDstReg> "," <swizzleSrcReg>
-*/
-fp_VECTORop_instruction
-    fp_VECTORop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg;
-vp_VECTORop_instruction
-    vp_VECTORop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg;
-
-/*
-fragment program
-    <VECTORop>             ::= "ABS" | "ABS_SAT"
-                             | "FLR" | "FLR_SAT"
-                             | "FRC" | "FRC_SAT"
-                             | "LIT" | "LIT_SAT"
-                             | "MOV" | "MOV_SAT"
-
-vertex program
-    <VECTORop>             ::= "ABS"
-                             | "FLR"
-                             | "FRC"
-                             | "LIT"
-                             | "MOV"
-*/
-fp_VECTORop
-    "ABS" .emit OP_ABS .or "ABS_SAT" .emit OP_ABS_SAT .or
-    "FLR" .emit OP_FLR .or "FLR_SAT" .emit OP_FLR_SAT .or
-    "FRC" .emit OP_FRC .or "FRC_SAT" .emit OP_FRC_SAT .or
-    "LIT" .emit OP_LIT .or "LIT_SAT" .emit OP_LIT_SAT .or
-    "MOV" .emit OP_MOV .or "MOV_SAT" .emit OP_MOV_SAT;
-vp_VECTORop
-    "ABS" .emit OP_ABS .or
-    "FLR" .emit OP_FLR .or
-    "FRC" .emit OP_FRC .or
-    "LIT" .emit OP_LIT .or
-    "MOV" .emit OP_MOV;
-
-/*
-    <SCALARop_instruction> ::= <SCALARop> <maskedDstReg> "," <scalarSrcReg>
-*/
-fp_SCALARop_instruction
-    fp_SCALARop .and space_dst .and fp_maskedDstReg .and comma .and fp_scalarSrcReg;
-vp_SCALARop_instruction
-    vp_SCALARop .and space_dst .and vp_maskedDstReg .and comma .and vp_scalarSrcReg;
-
-/*
-fragment program
-    <SCALARop>             ::= "COS" | "COS_SAT"
-                             | "EX2" | "EX2_SAT"
-                             | "LG2" | "LG2_SAT"
-                             | "RCP" | "RCP_SAT"
-                             | "RSQ" | "RSQ_SAT"
-                             | "SIN" | "SIN_SAT"
-                             | "SCS" | "SCS_SAT"
-
-vertex program
-    <SCALARop>             ::= "EX2"
-                             | "EXP"
-                             | "LG2"
-                             | "LOG"
-                             | "RCP"
-                             | "RSQ"
-*/
-fp_SCALARop
-    "COS" .emit OP_COS .or "COS_SAT" .emit OP_COS_SAT .or
-    "EX2" .emit OP_EX2 .or "EX2_SAT" .emit OP_EX2_SAT .or
-    "LG2" .emit OP_LG2 .or "LG2_SAT" .emit OP_LG2_SAT .or
-    "RCP" .emit OP_RCP .or "RCP_SAT" .emit OP_RCP_SAT .or
-    "RSQ" .emit OP_RSQ .or "RSQ_SAT" .emit OP_RSQ_SAT .or
-    "SIN" .emit OP_SIN .or "SIN_SAT" .emit OP_SIN_SAT .or
-    "SCS" .emit OP_SCS .or "SCS_SAT" .emit OP_SCS_SAT;
-vp_SCALARop
-    "EX2" .emit OP_EX2 .or
-    "EXP" .emit OP_EXP .or
-    "LG2" .emit OP_LG2 .or
-    "LOG" .emit OP_LOG .or
-    "RCP" .emit OP_RCP .or
-    "RSQ" .emit OP_RSQ;
-
-/*
-    <BINSCop_instruction>  ::= <BINSCop> <maskedDstReg> "," <scalarSrcReg> ","
-                               <scalarSrcReg>
-*/
-fp_BINSCop_instruction
-    fp_BINSCop .and space_dst .and fp_maskedDstReg .and comma .and fp_scalarSrcReg .and comma .and
-    fp_scalarSrcReg;
-vp_BINSCop_instruction
-    vp_BINSCop .and space_dst .and vp_maskedDstReg .and comma .and vp_scalarSrcReg .and comma .and
-    vp_scalarSrcReg;
-
-/*
-fragment program
-    <BINSCop>              ::= "POW" | "POW_SAT"
-
-vertex program
-    <BINSCop>              ::= "POW"
-*/
-fp_BINSCop
-    "POW" .emit OP_POW .or "POW_SAT" .emit OP_POW_SAT;
-vp_BINSCop
-    "POW" .emit OP_POW;
-
-/*
-fragment program
-    <BINop_instruction>    ::= <BINop> <maskedDstReg> ","
-                               <vectorSrcReg> "," <vectorSrcReg>
-
-vertex program
-    <BINop_instruction>    ::= <BINop> <maskedDstReg> ","
-                               <swizzleSrcReg> "," <swizzleSrcReg>
-*/
-fp_BINop_instruction
-    fp_BINop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and
-    vectorSrcReg;
-vp_BINop_instruction
-    vp_BINop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg .and comma .and
-    swizzleSrcReg;
-
-/*
-fragment program
-    <BINop>                ::= "ADD" | "ADD_SAT"
-                             | "DP3" | "DP3_SAT"
-                             | "DP4" | "DP4_SAT"
-                             | "DPH" | "DPH_SAT"
-                             | "DST" | "DST_SAT"
-                             | "MAX" | "MAX_SAT"
-                             | "MIN" | "MIN_SAT"
-                             | "MUL" | "MUL_SAT"
-                             | "SGE" | "SGE_SAT"
-                             | "SLT" | "SLT_SAT"
-                             | "SUB" | "SUB_SAT"
-                             | "XPD" | "XPD_SAT"
-
-vertex program
-    <BINop>                ::= "ADD"
-                             | "DP3"
-                             | "DP4"
-                             | "DPH"
-                             | "DST"
-                             | "MAX"
-                             | "MIN"
-                             | "MUL"
-                             | "SGE"
-                             | "SLT"
-                             | "SUB"
-                             | "XPD"
-*/
-fp_BINop
-    "ADD" .emit OP_ADD .or "ADD_SAT" .emit OP_ADD_SAT .or
-    "DP3" .emit OP_DP3 .or "DP3_SAT" .emit OP_DP3_SAT .or
-    "DP4" .emit OP_DP4 .or "DP4_SAT" .emit OP_DP4_SAT .or
-    "DPH" .emit OP_DPH .or "DPH_SAT" .emit OP_DPH_SAT .or
-    "DST" .emit OP_DST .or "DST_SAT" .emit OP_DST_SAT .or
-    "MAX" .emit OP_MAX .or "MAX_SAT" .emit OP_MAX_SAT .or
-    "MIN" .emit OP_MIN .or "MIN_SAT" .emit OP_MIN_SAT .or
-    "MUL" .emit OP_MUL .or "MUL_SAT" .emit OP_MUL_SAT .or
-    "SGE" .emit OP_SGE .or "SGE_SAT" .emit OP_SGE_SAT .or
-    "SLT" .emit OP_SLT .or "SLT_SAT" .emit OP_SLT_SAT .or
-    "SUB" .emit OP_SUB .or "SUB_SAT" .emit OP_SUB_SAT .or
-    "XPD" .emit OP_XPD .or "XPD_SAT" .emit OP_XPD_SAT;
-vp_BINop
-    "ADD" .emit OP_ADD .or
-    "DP3" .emit OP_DP3 .or
-    "DP4" .emit OP_DP4 .or
-    "DPH" .emit OP_DPH .or
-    "DST" .emit OP_DST .or
-    "MAX" .emit OP_MAX .or
-    "MIN" .emit OP_MIN .or
-    "MUL" .emit OP_MUL .or
-    "SGE" .emit OP_SGE .or
-    "SLT" .emit OP_SLT .or
-    "SUB" .emit OP_SUB .or
-    "XPD" .emit OP_XPD;
-
-/*
-fragment program
-    <TRIop_instruction>    ::= <TRIop> <maskedDstReg> ","
-                               <vectorSrcReg> "," <vectorSrcReg> ","
-                               <vectorSrcReg>
-
-vertex program
-    <TRIop_instruction>    ::= <TRIop> <maskedDstReg> ","
-                               <swizzleSrcReg> "," <swizzleSrcReg> ","
-                               <swizzleSrcReg>
-*/
-fp_TRIop_instruction
-    fp_TRIop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and
-    vectorSrcReg .and comma .and vectorSrcReg;
-vp_TRIop_instruction
-    vp_TRIop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg .and comma .and
-    swizzleSrcReg .and comma .and swizzleSrcReg;
-
-/*
-fragment program
-    <TRIop>                ::= "CMP" | "CMP_SAT"
-                             | "LRP" | "LRP_SAT"
-                             | "MAD" | "MAD_SAT"
-
-vertex program
-    <TRIop>                ::= "MAD"
-*/
-fp_TRIop
-    "CMP" .emit OP_CMP .or "CMP_SAT" .emit OP_CMP_SAT .or
-    "LRP" .emit OP_LRP .or "LRP_SAT" .emit OP_LRP_SAT .or
-    "MAD" .emit OP_MAD .or "MAD_SAT" .emit OP_MAD_SAT;
-vp_TRIop
-    "MAD" .emit OP_MAD;
-
-/*
-fragment program
-    <SWZ_instruction>      ::= <SWZop> <maskedDstReg> "," 
-                               <srcReg> "," <extendedSwizzle>
-
-vertex program
-    <SWZ_instruction>      ::= "SWZ" <maskedDstReg> "," <srcReg> "," 
-                               <extendedSwizzle>
-*/
-fp_SWZ_instruction
-    SWZop .and space_dst .and fp_maskedDstReg .and comma .and fp_srcReg .and comma .and
-    fp_extendedSwizzle .error EXT_SWIZ_COMP_EXPECTED;
-vp_SWZ_instruction
-    "SWZ" .emit OP_SWZ .and space_dst .and vp_maskedDstReg .and comma .and vp_srcReg .and comma .and
-    vp_extendedSwizzle .error EXT_SWIZ_COMP_EXPECTED;
-
-/*
-fragment program
-    <SWZop>                ::= "SWZ" | "SWZ_SAT"
-*/
-SWZop
-    "SWZ" .emit OP_SWZ .or "SWZ_SAT" .emit OP_SWZ_SAT;
-
-/*
-fragment program
-    <SAMPLE_instruction>   ::= <SAMPLEop> <maskedDstReg> ","
-                               <vectorSrcReg> "," <texImageUnit> "," 
-                               <texTarget>
-*/
-SAMPLE_instruction
-    SAMPLEop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and
-    texImageUnit .and comma .and texTarget .error TEX_TARGET_EXPECTED;
-
-/*
-fragment program
-    <SAMPLEop>             ::= "TEX" | "TEX_SAT"
-                             | "TXP" | "TXP_SAT"
-                             | "TXB" | "TXB_SAT"
-*/
-SAMPLEop
-    "TEX" .emit OP_TEX .or "TEX_SAT" .emit OP_TEX_SAT .or
-    "TXB" .emit OP_TXB .or "TXB_SAT" .emit OP_TXB_SAT .or
-    "TXP" .emit OP_TXP .or "TXP_SAT" .emit OP_TXP_SAT;
-
-/*
-fragment program
-    <KIL_instruction>      ::= "KIL" <vectorSrcReg>
-*/
-KIL_instruction
-    "KIL" .emit OP_KIL .and space_src .and vectorSrcReg;
-
-/*
-fragment program
-    <texImageUnit>         ::= "texture" <optTexImageUnitNum>
-*/
-texImageUnit
-    "texture" .error TEXTURE_EXPECTED .and optTexImageUnitNum;
-
-/*
-fragment program
-    <texTarget>            ::= "1D"
-                             | "2D"
-                             | "3D"
-                             | "CUBE"
-                             | "RECT"
-                             | <shadowTarget> (if option ARB_fragment_program_shadow present)
-                             | <arrayTarget> (if option MESA_texture_array present)
-*/
-texTarget
-    "1D" .emit TEXTARGET_1D .or
-    "2D" .emit TEXTARGET_2D .or
-    "3D" .emit TEXTARGET_3D .or
-    .if (texture_rectangle != 0x00) "RECT" .emit TEXTARGET_RECT .or
-    "CUBE" .emit TEXTARGET_CUBE .or
-    .if (ARB_fragment_program_shadow != 0x00) shadowTarget .or
-    .if (MESA_texture_array != 0x00) arrayTarget;
-
-/*
-GL_ARB_fragment_program_shadow
-    <shadowTarget>         ::= "SHADOW1D"
-                             | "SHADOW2D"
-                             | "SHADOWRECT"
-                             | <shadowArrayTarget> (if option MESA_texture_array present)
-*/
-shadowTarget
-    "SHADOW1D" .emit TEXTARGET_SHADOW1D .or
-    "SHADOW2D" .emit TEXTARGET_SHADOW2D .or
-    .if (texture_rectangle != 0x00) "SHADOWRECT" .emit TEXTARGET_SHADOWRECT .or
-    .if (MESA_texture_array != 0x00) shadowArrayTarget;
-
-/*
-GL_MESA_texture_array
-
-    <arrayTarget>         ::= "ARRAY1D"
-                             | "ARRAY2D"
-
-    <shadowArrayTarget>   ::= "SHADOWARRAY1D"
-                             | "SHADOWARRAY2D"
-*/
-
-arrayTarget
-    "ARRAY1D" .emit TEXTARGET_1D_ARRAY .or
-    "ARRAY2D" .emit TEXTARGET_2D_ARRAY;
-
-shadowArrayTarget
-    "SHADOWARRAY1D" .emit TEXTARGET_SHADOW1D_ARRAY .or
-    "SHADOWARRAY2D" .emit TEXTARGET_SHADOW2D_ARRAY;
-
-/*
-fragment program
-    <optTexImageUnitNum>   ::= ""
-                             | "[" <texImageUnitNum> "]"
-*/
-optTexImageUnitNum
-    optTexImageUnitNum_1 .or .true .emit 0x00;
-optTexImageUnitNum_1
-    lbracket_ne .and texImageUnitNum .and rbracket;
-
-/*
-fragment program
-    <texImageUnitNum>      ::= <integer> from 0 to 
-                               MAX_TEXTURE_IMAGE_UNITS_ARB-1
-*/
-texImageUnitNum
-    integer;
-
-/*
-    <scalarSrcReg>         ::= <optionalSign> <srcReg> <scalarSuffix>
-*/
-fp_scalarSrcReg
-    optionalSign .and fp_srcReg .and fp_scalarSuffix;
-vp_scalarSrcReg
-    optionalSign .and vp_srcReg .and vp_scalarSuffix;
-
-/*
-vertex program
-    <swizzleSrcReg>        ::= <optionalSign> <srcReg> <swizzleSuffix>
-*/
-swizzleSrcReg
-    optionalSign .and vp_srcReg .and swizzleSuffix;
-
-/*
-fragment program
-    <vectorSrcReg>         ::= <optionalSign> <srcReg> <optionalSuffix> 
-*/
-vectorSrcReg
-    optionalSign .and fp_srcReg .and optionalSuffix;
-
-/*
-    <maskedDstReg>         ::= <dstReg> <optionalMask>
-*/
-fp_maskedDstReg
-    fp_dstReg .and fp_optionalMask;
-vp_maskedDstReg
-    vp_dstReg .and vp_optionalMask;
-
-/*
-vertex program
-    <maskedAddrReg>        ::= <addrReg> <addrWriteMask>
-*/
-maskedAddrReg
-    addrReg .error ADDRESS_REGISTER_EXPECTED .and addrWriteMask;
-
-/*
-fragment program
-    <extendedSwizzle>      ::= <xyzwExtendedSwizzle>
-                             | <rgbaExtendedSwizzle>
-
-vertex program
-    <extendedSwizzle>      ::= <extSwizComp> "," <extSwizComp> "," 
-                                 <extSwizComp> "," <extSwizComp>
-
-NOTE: do NOT change the order of <rgbaExtendedSwizzle> and <xyzwExtendedSwizzle> rulez
-*/
-fp_extendedSwizzle
-    rgbaExtendedSwizzle .or xyzwExtendedSwizzle;
-vp_extendedSwizzle
-    extSwizComp .and comma .and
-    extSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    extSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    extSwizComp .error EXT_SWIZ_COMP_EXPECTED;
-
-/*
-fragment program
-    <xyzwExtendedSwizzle>  ::= <xyzwExtSwizComp> "," <xyzwExtSwizComp> "," 
-                               <xyzwExtSwizComp> "," <xyzwExtSwizComp>
-*/
-xyzwExtendedSwizzle
-    xyzwExtSwizComp .and comma .and
-    xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;
-
-/*
-fragment program
-    <rgbaExtendedSwizzle>  ::= <rgbaExtSwizComp> "," <rgbaExtSwizComp> "," 
-                               <rgbaExtSwizComp> "," <rgbaExtSwizComp>
-*/
-rgbaExtendedSwizzle
-    rgbaExtendedSwizzle_1 .or rgbaExtendedSwizzle_2 .or rgbaExtendedSwizzle_3 .or
-    rgbaExtendedSwizzle_4;
-rgbaExtendedSwizzle_1
-    rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_digit .and comma .and
-    rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp;
-rgbaExtendedSwizzle_2
-    rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_digit .and comma .and
-    rgbaExtSwizComp_alpha .and comma .and rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;
-rgbaExtendedSwizzle_3
-    rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_alpha .and comma .and
-    rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;
-rgbaExtendedSwizzle_4
-    rgbaExtSwizComp_alpha .and comma .and 
-    rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and
-    rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;
-
-/*
-fragment program
-    <xyzwExtSwizComp>      ::= <optionalSign> <xyzwExtSwizSel>
-*/
-xyzwExtSwizComp
-    optionalSign .and xyzwExtSwizSel;
-
-/*
-fragment program
-    <rgbaExtSwizComp>      ::= <optionalSign> <rgbaExtSwizSel>
-*/
-rgbaExtSwizComp
-    optionalSign .and rgbaExtSwizSel;
-rgbaExtSwizComp_digit
-    optionalSign .and rgbaExtSwizSel_digit;
-rgbaExtSwizComp_alpha
-    optionalSign .and rgbaExtSwizSel_alpha;
-
-/*
-vertex program
-    <extSwizComp>          ::= <optionalSign> <extSwizSel>
-*/
-extSwizComp
-    optionalSign .and extSwizSel;
-
-/*
-fragment program
-    <xyzwExtSwizSel>       ::= "0" 
-                             | "1" 
-                             | <xyzwComponent>
-*/
-xyzwExtSwizSel
-    "0" .emit COMPONENT_0 .or "1" .emit COMPONENT_1 .or xyzwComponent_single;
-
-/*
-fragment program
-    <rgbaExtSwizSel>       ::= "0" 
-                             | "1" 
-                             | <rgbaComponent>
-*/
-rgbaExtSwizSel
-    rgbaExtSwizSel_digit .or rgbaExtSwizSel_alpha;
-rgbaExtSwizSel_digit
-    "0" .emit COMPONENT_0 .or "1" .emit COMPONENT_1;
-rgbaExtSwizSel_alpha
-    rgbaComponent_single;
-
-/*
-vertex program
-    <extSwizSel>           ::= "0" 
-                             | "1" 
-                             | <component>
-*/
-extSwizSel
-    "0" .emit COMPONENT_0 .or "1" .emit COMPONENT_1 .or vp_component_single;
-
-/*
-fragment program
-    <srcReg>               ::= <fragmentAttribReg>
-                             | <temporaryReg>
-                             | <progParamReg>
-
-vertex program
-    <srcReg>               ::= <vertexAttribReg>
-                             | <temporaryReg>
-                             | <progParamReg>
-*/
-fp_srcReg
-    fp_srcReg_1 .error SOURCE_REGISTER_EXPECTED;
-vp_srcReg
-    vp_srcReg_1 .error SOURCE_REGISTER_EXPECTED;
-fp_srcReg_1
-    fragmentAttribReg .emit REGISTER_ATTRIB .or
-    fp_progParamReg .emit REGISTER_PARAM .or
-    fp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;
-vp_srcReg_1
-    vertexAttribReg .emit REGISTER_ATTRIB .or
-    vp_progParamReg .emit REGISTER_PARAM .or
-    vp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;
-
-/*
-fragment program
-    <dstReg>               ::= <temporaryReg>
-                             | <fragmentResultReg>
-
-vertex program
-    <dstReg>               ::= <temporaryReg>
-                             | <vertexResultReg>
-*/
-fp_dstReg
-    fp_dstReg_1 .error DESTINATION_REGISTER_EXPECTED;
-vp_dstReg
-    vp_dstReg_1 .error DESTINATION_REGISTER_EXPECTED;
-fp_dstReg_1
-    fragmentResultReg .emit REGISTER_RESULT .or
-    fp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;
-vp_dstReg_1
-    vertexResultReg .emit REGISTER_RESULT .or
-    vp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;
-
-/*
-fragment program
-    <fragmentAttribReg>    ::= <establishedName>
-                             | <fragAttribBinding>
-
-NOTE: <establishedName> is driven by <temporaryReg> rule at the end of <srcReg>
-*/
-fragmentAttribReg
-    /*fp_establishedName .or */fragAttribBinding;
-
-/*
-vertex program
-    <vertexAttribReg>      ::= <establishedName>
-                             | <vtxAttribBinding>
-
-NOTE: <establishedName> is driven by <temporaryReg> rule at the end of <srcReg>
-*/
-vertexAttribReg
-    vtxAttribBinding;
-
-/*
-    <temporaryReg>         ::= <establishedName>
-*/
-fp_temporaryReg
-    fp_establishedName_no_error_on_identifier;
-vp_temporaryReg
-    vp_establishedName_no_error_on_identifier;
-
-/*
-fragment program
-    <progParamReg>         ::= <progParamSingle>
-                             | <progParamArray> "[" <progParamArrayAbs> "]"
-                             | <paramSingleItemUse>
-
-vertex program
-    <progParamReg>         ::= <progParamSingle>
-                             | <progParamArray> "[" <progParamArrayMem> "]"
-                             | <paramSingleItemUse>
-*/
-fp_progParamReg
-    fp_paramSingleItemUse .or fp_progParamReg_1 .or fp_progParamSingle;
-vp_progParamReg
-    vp_paramSingleItemUse .or vp_progParamReg_1 .or vp_progParamSingle;
-fp_progParamReg_1
-    fp_progParamArray .emit PARAM_ARRAY_ELEMENT .and lbracket_ne .and progParamArrayAbs .and
-    rbracket;
-vp_progParamReg_1
-    vp_progParamArray .emit PARAM_ARRAY_ELEMENT .and lbracket_ne .and progParamArrayMem .and
-    rbracket;
-
-/*
-    <progParamSingle>      ::= <establishedName>
-
-NOTE: <establishedName> is driven by <temporaryReg> rule at the end of <srcReg>
-*/
-fp_progParamSingle
-    .false;
-vp_progParamSingle
-    .false;
-
-/*
-    <progParamArray>       ::= <establishedName>
-*/
-fp_progParamArray
-    fp_establishedName_no_error_on_identifier;
-vp_progParamArray
-    vp_establishedName_no_error_on_identifier;
-
-/*
-vertex program
-    <progParamArrayMem>    ::= <progParamArrayAbs>
-                             | <progParamArrayRel>
-*/
-progParamArrayMem
-    progParamArrayAbs .or progParamArrayRel;
-
-/*
-    <progParamArrayAbs>    ::= <integer>
-*/
-progParamArrayAbs
-    integer_ne .emit ARRAY_INDEX_ABSOLUTE;
-
-/*
-vertex program
-    <progParamArrayRel>    ::= <addrReg> <addrComponent> <addrRegRelOffset>
-*/
-progParamArrayRel
-    addrReg .error ADDRESS_REGISTER_OR_INTEGER_EXPECTED .emit ARRAY_INDEX_RELATIVE .and
-    addrComponent .and addrRegRelOffset;
-
-/*
-vertex program
-    <addrRegRelOffset>     ::= ""
-                             | "+" <addrRegPosOffset>
-                             | "-" <addrRegNegOffset>
-*/
-addrRegRelOffset
-    addrRegRelOffset_1 .or addrRegRelOffset_2 .or .true .emit 0x00;
-addrRegRelOffset_1
-    plus_ne .and addrRegPosOffset;
-addrRegRelOffset_2
-    minus_ne .and addrRegNegOffset;
-
-/*
-vertex program
-    <addrRegPosOffset>     ::= <integer> from 0 to 63
-*/
-addrRegPosOffset
-    integer_0_63;
-
-/*
-vertex program
-    <addrRegNegOffset>     ::= <integer> from 0 to 64
-*/
-addrRegNegOffset
-    integer_0_64;
-
-/*
-fragment program
-    <fragmentResultReg>    ::= <establishedName>
-                             | <resultBinding>
-
-NOTE: <establishedName> is driven by <temporaryReg> rule at the end of <dstReg>
-*/
-fragmentResultReg
-    fp_resultBinding;
-
-/*
-vertex program
-    <vertexResultReg>      ::= <establishedName>
-                             | <resultBinding>
-
-NOTE: <establishedName> is driven by <temporaryReg> rule at the end of <dstReg>
-*/
-vertexResultReg
-    vp_resultBinding;
-
-/*
-vertex program
-    <addrReg>              ::= <establishedName>
-*/
-addrReg
-    vp_establishedName_no_error_on_identifier;
-
-/*
-vertex program
-    <addrComponent>        ::= "." "x"
-*/
-addrComponent
-    dot .and "x" .error INVALID_ADDRESS_COMPONENT .emit COMPONENT_X .emit COMPONENT_X
-    .emit COMPONENT_X .emit COMPONENT_X;
-
-/*
-vertex program
-    <addrWriteMask>        ::= "." "x"
-*/
-addrWriteMask
-    dot .and "x" .error INVALID_ADDRESS_WRITEMASK .emit 0x08;
-
-/*
-    <scalarSuffix>         ::= "." <component>
-*/
-fp_scalarSuffix
-    dot .and fp_component_single .error INVALID_COMPONENT;
-vp_scalarSuffix
-    dot .and vp_component_single .error INVALID_COMPONENT;
-
-/*
-vertex program
-    <swizzleSuffix>        ::= ""
-                             | "." <component>
-                             | "." <component> <component>
-                                   <component> <component>
-*/
-swizzleSuffix
-    swizzleSuffix_1 .or
-    .true .emit COMPONENT_X .emit COMPONENT_Y .emit COMPONENT_Z .emit COMPONENT_W;
-swizzleSuffix_1
-    dot_ne .and swizzleSuffix_2 .error INVALID_SUFFIX;
-swizzleSuffix_2
-    swizzleSuffix_3 .or swizzleSuffix_4;
-swizzleSuffix_3
-    vp_component_multi .and vp_component_multi .and vp_component_multi .error INVALID_COMPONENT .and
-    vp_component_multi .error INVALID_COMPONENT;
-swizzleSuffix_4
-    "x" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or
-    "y" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or
-    "z" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or
-    "w" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W;
-
-/*
-fragment program
-    <optionalSuffix>       ::= "" 
-                             | "." <component> 
-                             | "." <xyzwComponent> <xyzwComponent>
-                                   <xyzwComponent> <xyzwComponent>
-                             | "." <rgbaComponent> <rgbaComponent>
-                                   <rgbaComponent> <rgbaComponent>
-*/
-optionalSuffix
-    optionalSuffix_1 .or
-    .true .emit COMPONENT_X .emit COMPONENT_Y .emit COMPONENT_Z .emit COMPONENT_W;
-optionalSuffix_1
-    dot_ne .and optionalSuffix_2 .error INVALID_SUFFIX;
-optionalSuffix_2
-    optionalSuffix_3 .or optionalSuffix_4 .or optionalSuffix_5;
-optionalSuffix_3
-    xyzwComponent_multi .and xyzwComponent_multi .and
-    xyzwComponent_multi .error INVALID_COMPONENT .and xyzwComponent_multi .error INVALID_COMPONENT;
-optionalSuffix_4
-    rgbaComponent_multi .and rgbaComponent_multi .and
-    rgbaComponent_multi .error INVALID_COMPONENT .and rgbaComponent_multi .error INVALID_COMPONENT;
-optionalSuffix_5
-    "x" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or
-    "y" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or
-    "z" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or
-    "w" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .or
-    "r" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or
-    "g" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or
-    "b" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or
-    "a" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W;
-
-/*
-fragment program
-    <component>            ::= <xyzwComponent>
-                             | <rgbaComponent>
-
-vertex program
-    <component>            ::= "x"
-                             | "y"
-                             | "z"
-                             | "w"
-*/
-fp_component_single
-    xyzwComponent_single .or rgbaComponent_single;
-vp_component_multi
-    'x' .emit COMPONENT_X .or 'y' .emit COMPONENT_Y .or 'z' .emit COMPONENT_Z .or
-    'w' .emit COMPONENT_W;
-vp_component_single
-    "x" .emit COMPONENT_X .or "y" .emit COMPONENT_Y .or "z" .emit COMPONENT_Z .or
-    "w" .emit COMPONENT_W;
-
-/*
-fragment program
-    <xyzwComponent>        ::= "x" | "y" | "z" | "w"
-*/
-xyzwComponent_multi
-    'x' .emit COMPONENT_X .or 'y' .emit COMPONENT_Y .or 'z' .emit COMPONENT_Z .or
-    'w' .emit COMPONENT_W;
-xyzwComponent_single
-    "x" .emit COMPONENT_X .or "y" .emit COMPONENT_Y .or "z" .emit COMPONENT_Z .or
-    "w" .emit COMPONENT_W;
-
-/*
-fragment program
-    <rgbaComponent>        ::= "r" | "g" | "b" | "a"
-*/
-rgbaComponent_multi
-    'r' .emit COMPONENT_X .or 'g' .emit COMPONENT_Y .or 'b' .emit COMPONENT_Z .or
-    'a' .emit COMPONENT_W;
-rgbaComponent_single
-    "r" .emit COMPONENT_X .or "g" .emit COMPONENT_Y .or "b" .emit COMPONENT_Z .or
-    "a" .emit COMPONENT_W;
-
-/*
-fragment program
-    <optionalMask>         ::= ""
-                             | <xyzwMask>
-                             | <rgbaMask>
-
-vertex program
-    <optionalMask>         ::= ""
-                             | "." "x"
-                             | "." "y"
-                             | "." "xy"
-                             | "." "z"
-                             | "." "xz"
-                             | "." "yz"
-                             | "." "xyz"
-                             | "." "w"
-                             | "." "xw"
-                             | "." "yw"
-                             | "." "xyw"
-                             | "." "zw"
-                             | "." "xzw"
-                             | "." "yzw"
-                             | "." "xyzw"
-
-NOTE: do NOT change the order of <rgbaMask> and <xyzwMask> rulez
-*/
-fp_optionalMask
-    rgbaMask .or xyzwMask .or .true .emit 0x0F;
-vp_optionalMask
-    xyzwMask .or .true .emit 0x0F;
-
-/*
-fragment program
-    <xyzwMask>             ::= "." "x"
-                             | "." "y"
-                             | "." "xy"
-                             | "." "z"
-                             | "." "xz"
-                             | "." "yz"
-                             | "." "xyz"
-                             | "." "w"
-                             | "." "xw"
-                             | "." "yw"
-                             | "." "xyw"
-                             | "." "zw"
-                             | "." "xzw"
-                             | "." "yzw"
-                             | "." "xyzw"
-
-NOTE: <xyzwMask> is also referenced by the vertex program symbol <optionalMask>.
-*/
-xyzwMask
-    dot_ne .and xyzwMask_1 .error INVALID_WRITEMASK;
-xyzwMask_1
-    "xyzw" .emit 0x0F .or "xyz" .emit 0x0E .or "xyw" .emit 0x0D .or "xy" .emit 0x0C .or
-    "xzw" .emit 0x0B .or "xz" .emit 0x0A .or "xw" .emit 0x09 .or "x" .emit 0x08 .or
-    "yzw" .emit 0x07 .or "yz" .emit 0x06 .or "yw" .emit 0x05 .or "y" .emit 0x04 .or
-    "zw" .emit 0x03 .or "z" .emit 0x02 .or "w" .emit 0x01;
-
-/*
-fragment program
-    <rgbaMask>             ::= "." "r"
-                             | "." "g"
-                             | "." "rg"
-                             | "." "b"
-                             | "." "rb"
-                             | "." "gb"
-                             | "." "rgb"
-                             | "." "a"
-                             | "." "ra"
-                             | "." "ga"
-                             | "." "rga"
-                             | "." "ba"
-                             | "." "rba"
-                             | "." "gba"
-                             | "." "rgba"
-*/
-rgbaMask
-    dot_ne .and rgbaMask_1;
-rgbaMask_1
-    "rgba" .emit 0x0F .or "rgb" .emit 0x0E .or "rga" .emit 0x0D .or "rg" .emit 0x0C .or
-    "rba" .emit 0x0B .or "rb" .emit 0x0A .or "ra" .emit 0x09 .or "r" .emit 0x08 .or
-    "gba" .emit 0x07 .or "gb" .emit 0x06 .or "ga" .emit 0x05 .or "g" .emit 0x04 .or
-    "ba" .emit 0x03 .or "b" .emit 0x02 .or "a" .emit 0x01;
-
-/*
-fragment program
-    <namingStatement>      ::= <ATTRIB_statement>
-                             | <PARAM_statement>
-                             | <TEMP_statement>
-                             | <OUTPUT_statement>
-                             | <ALIAS_statement>
-
-vertex program
-    <namingStatement>      ::= <ATTRIB_statement>
-                             | <PARAM_statement>
-                             | <TEMP_statement>
-                             | <ADDRESS_statement>
-                             | <OUTPUT_statement>
-                             | <ALIAS_statement>
-*/
-fp_namingStatement
-    fp_ATTRIB_statement .emit ATTRIB .or
-    fp_PARAM_statement .emit PARAM .or
-    fp_TEMP_statement .emit TEMP .or
-    fp_OUTPUT_statement .emit OUTPUT .or
-    fp_ALIAS_statement .emit ALIAS;
-vp_namingStatement
-    vp_ATTRIB_statement .emit ATTRIB .or
-    vp_PARAM_statement .emit PARAM .or
-    vp_TEMP_statement .emit TEMP .or
-    ADDRESS_statement .emit ADDRESS .or
-    vp_OUTPUT_statement .emit OUTPUT .or
-    vp_ALIAS_statement .emit ALIAS;
-
-/*
-fragment program
-    <ATTRIB_statement>     ::= "ATTRIB" <establishName> "="
-                                 <fragAttribBinding>
-
-vertex program
-    <ATTRIB_statement>     ::= "ATTRIB" <establishName> "="
-                                 <vtxAttribBinding>
-*/
-fp_ATTRIB_statement
-    "ATTRIB" .and space .and fp_establishName .and equal .and
-    fragAttribBinding .error FRAGMENT_EXPECTED;
-vp_ATTRIB_statement
-    "ATTRIB" .and space .and vp_establishName .and equal .and
-    vtxAttribBinding .error VERTEX_EXPECTED;
-
-/*
-fragment program
-    <fragAttribBinding>    ::= "fragment" "." <fragAttribItem>
-*/
-fragAttribBinding
-    "fragment" .and dot .and fragAttribItem .error INVALID_FRAGMENT_PROPERTY;
-
-/*
-vertex program
-    <vtxAttribBinding>     ::= "vertex" "." <vtxAttribItem>
-*/
-vtxAttribBinding
-    "vertex" .and dot .and vtxAttribItem .error INVALID_VERTEX_PROPERTY;
-
-/*
-fragment program
-    <fragAttribItem>       ::= "color" <optColorType>
-                             | "texcoord" <optTexCoordNum>
-                             | "fogcoord"
-                             | "position"
-*/
-fragAttribItem
-    fragAttribItem_1 .emit FRAGMENT_ATTRIB_COLOR .or
-    fragAttribItem_2 .emit FRAGMENT_ATTRIB_TEXCOORD .or
-    .if (fog_coord != 0x00) "fogcoord" .emit FRAGMENT_ATTRIB_FOGCOORD .or
-    "position" .emit FRAGMENT_ATTRIB_POSITION;
-fragAttribItem_1
-    "color" .and optColorType;
-fragAttribItem_2
-    "texcoord" .and optTexCoordNum;
-
-/*
-vertex program
-    <vtxAttribItem>        ::= "position"
-                             | "weight" <vtxOptWeightNum>
-                             | "normal"
-                             | "color" <optColorType>
-                             | "fogcoord"
-                             | "texcoord" <optTexCoordNum>
-                             | "matrixindex" "[" <vtxWeightNum> "]"
-                             | "attrib" "[" <vtxAttribNum> "]"
-*/
-vtxAttribItem
-    "position" .emit VERTEX_ATTRIB_POSITION .or
-    .if (vertex_blend != 0x00) vtxAttribItem_1 .emit VERTEX_ATTRIB_WEIGHT .or
-    "normal" .emit VERTEX_ATTRIB_NORMAL .or
-    vtxAttribItem_2 .emit VERTEX_ATTRIB_COLOR .or
-    "fogcoord" .emit VERTEX_ATTRIB_FOGCOORD .or
-    vtxAttribItem_3 .emit VERTEX_ATTRIB_TEXCOORD .or
-    .if (matrix_palette != 0x00) vtxAttribItem_4 .emit VERTEX_ATTRIB_MATRIXINDEX .or
-    vtxAttribItem_5 .emit VERTEX_ATTRIB_GENERIC;
-vtxAttribItem_1
-    "weight" .and vtxOptWeightNum;
-vtxAttribItem_2
-    "color" .and optColorType;
-vtxAttribItem_3
-    "texcoord" .and optTexCoordNum;
-vtxAttribItem_4
-    "matrixindex" .and lbracket .and vtxWeightNum .and rbracket;
-vtxAttribItem_5
-    "attrib" .and lbracket .and vtxAttribNum .and rbracket;
-
-/*
-vertex program
-    <vtxAttribNum>         ::= <integer> from 0 to MAX_VERTEX_ATTRIBS_ARB-1
-*/
-vtxAttribNum
-    integer;
-
-/*
-vertex program
-    <vtxOptWeightNum>      ::= ""
-                             | "[" <vtxWeightNum> "]"
-*/
-vtxOptWeightNum
-    vtxOptWeightNum_1 .or .true .emit 0x00;
-vtxOptWeightNum_1
-    lbracket_ne .and vtxWeightNum .and rbracket;
-
-/*
-vertex program
-    <vtxWeightNum>         ::= <integer> from 0 to MAX_VERTEX_UNITS_ARB-1,
-                                 must be divisible by four
-*/
-vtxWeightNum
-    integer;
-
-/*
-    <PARAM_statement>      ::= <PARAM_singleStmt>
-                             | <PARAM_multipleStmt>
-*/
-fp_PARAM_statement
-    fp_PARAM_multipleStmt .or fp_PARAM_singleStmt;
-vp_PARAM_statement
-    vp_PARAM_multipleStmt .or vp_PARAM_singleStmt;
-
-/*
-    <PARAM_singleStmt>     ::= "PARAM" <establishName> <paramSingleInit>
-*/
-fp_PARAM_singleStmt
-    "PARAM" .and space .and fp_establishName .and .true .emit 0x00 .and fp_paramSingleInit .and
-    .true .emit PARAM_NULL;
-vp_PARAM_singleStmt
-    "PARAM" .and space .and vp_establishName .and .true .emit 0x00 .and vp_paramSingleInit .and
-    .true .emit PARAM_NULL;
-
-/*
-    <PARAM_multipleStmt>   ::= "PARAM" <establishName> "[" <optArraySize> "]"
-                                   <paramMultipleInit>
-*/
-fp_PARAM_multipleStmt
-    "PARAM" .and space .and fp_establishName .and lbracket_ne .and optArraySize .and rbracket .and
-    fp_paramMultipleInit .and .true .emit PARAM_NULL;
-vp_PARAM_multipleStmt
-    "PARAM" .and space .and vp_establishName .and lbracket_ne .and optArraySize .and rbracket .and
-    vp_paramMultipleInit .and .true .emit PARAM_NULL;
-
-/*
-    <optArraySize>         ::= ""
-                             | <integer> from 1 to MAX_PROGRAM_PARAMETERS_ARB
-                                 (maximum number of allowed program 
-                                  parameter bindings)
-*/
-optArraySize
-    optional_integer;
-
-/*
-    <paramSingleInit>      ::= "=" <paramSingleItemDecl>
-*/
-fp_paramSingleInit
-    equal .and fp_paramSingleItemDecl;
-vp_paramSingleInit
-    equal .and vp_paramSingleItemDecl;
-
-/*
-    <paramMultipleInit>    ::= "=" "{" <paramMultInitList> "}"
-*/
-fp_paramMultipleInit
-    equal .and lbrace .and fp_paramMultInitList .and rbrace;
-vp_paramMultipleInit
-    equal .and lbrace .and vp_paramMultInitList .and rbrace;
-
-/*
-    <paramMultInitList>    ::= <paramMultipleItem>
-                             | <paramMultipleItem> "," <paramMultiInitList>
-*/
-fp_paramMultInitList
-    fp_paramMultInitList_1 .or fp_paramMultipleItem;
-vp_paramMultInitList
-    vp_paramMultInitList_1 .or vp_paramMultipleItem;
-fp_paramMultInitList_1
-    fp_paramMultipleItem .and comma_ne .and fp_paramMultInitList;
-vp_paramMultInitList_1
-    vp_paramMultipleItem .and comma_ne .and vp_paramMultInitList;
-
-/*
-    <paramSingleItemDecl>  ::= <stateSingleItem>
-                             | <programSingleItem>
-                             | <paramConstDecl>
-*/
-fp_paramSingleItemDecl
-    fp_stateSingleItem .emit PARAM_STATE_ELEMENT .or
-    programSingleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstDecl .emit PARAM_CONSTANT;
-vp_paramSingleItemDecl
-    vp_stateSingleItem .emit PARAM_STATE_ELEMENT .or
-    programSingleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstDecl .emit PARAM_CONSTANT;
-
-/*
-    <paramSingleItemUse>   ::= <stateSingleItem>
-                             | <programSingleItem>
-                             | <paramConstUse>
-*/
-fp_paramSingleItemUse
-    fp_stateSingleItem .emit PARAM_STATE_ELEMENT .or
-    programSingleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstUse .emit PARAM_CONSTANT;
-vp_paramSingleItemUse
-    vp_stateSingleItem .emit PARAM_STATE_ELEMENT .or
-    programSingleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstUse .emit PARAM_CONSTANT;
-
-/*
-    <paramMultipleItem>    ::= <stateMultipleItem>
-                             | <programMultipleItem>
-                             | <paramConstDecl>
-*/
-fp_paramMultipleItem
-    fp_stateMultipleItem .emit PARAM_STATE_ELEMENT .or
-    programMultipleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstDecl .emit PARAM_CONSTANT;
-vp_paramMultipleItem
-    vp_stateMultipleItem .emit PARAM_STATE_ELEMENT .or
-    programMultipleItem .emit PARAM_PROGRAM_ELEMENT .or
-    paramConstDecl .emit PARAM_CONSTANT;
-
-/*
-    <stateMultipleItem>    ::= <stateSingleItem>
-                             | "state" "." <stateMatrixRows>
-*/
-fp_stateMultipleItem
-    stateMultipleItem_1 .or fp_stateSingleItem;
-vp_stateMultipleItem
-    stateMultipleItem_1 .or vp_stateSingleItem;
-stateMultipleItem_1
-    "state" .and dot .and stateMatrixRows .emit STATE_MATRIX_ROWS;
-
-/*
-fragment program
-    <stateSingleItem>      ::= "state" "." <stateMaterialItem>
-                             | "state" "." <stateLightItem>
-                             | "state" "." <stateLightModelItem>
-                             | "state" "." <stateLightProdItem>
-                             | "state" "." <stateTexEnvItem>
-                             | "state" "." <stateFogItem>
-                             | "state" "." <stateDepthItem>
-                             | "state" "." <stateMatrixRow>
-
-vertex program
-    <stateSingleItem>      ::= "state" "." <stateMaterialItem>
-                             | "state" "." <stateLightItem>
-                             | "state" "." <stateLightModelItem>
-                             | "state" "." <stateLightProdItem>
-                             | "state" "." <stateTexGenItem>
-                             | "state" "." <stateFogItem>
-                             | "state" "." <stateClipPlaneItem>
-                             | "state" "." <statePointItem>
-                             | "state" "." <stateMatrixRow>
-*/
-fp_stateSingleItem
-    "state" .and dot .and fp_stateSingleItem_1 .error INVALID_STATE_PROPERTY;
-vp_stateSingleItem
-    "state" .and dot .and vp_stateSingleItem_1 .error INVALID_STATE_PROPERTY;
-fp_stateSingleItem_1
-    stateSingleItem_1 .or stateSingleItem_2 .or stateSingleItem_3 .or stateSingleItem_4 .or
-    stateSingleItem_5 .or stateSingleItem_7 .or stateSingleItem_8 .or stateSingleItem_11;
-vp_stateSingleItem_1
-    stateSingleItem_1 .or stateSingleItem_2 .or stateSingleItem_3 .or stateSingleItem_4 .or
-    stateSingleItem_6 .or stateSingleItem_7 .or stateSingleItem_9 .or stateSingleItem_10 .or
-    stateSingleItem_11;
-stateSingleItem_1
-    stateMaterialItem .emit STATE_MATERIAL;
-stateSingleItem_2
-    stateLightItem .emit STATE_LIGHT;
-stateSingleItem_3
-    stateLightModelItem .emit STATE_LIGHT_MODEL;
-stateSingleItem_4
-    stateLightProdItem .emit STATE_LIGHT_PROD;
-stateSingleItem_5
-    stateTexEnvItem .emit STATE_TEX_ENV;
-stateSingleItem_6
-    stateTexGenItem .emit STATE_TEX_GEN;
-stateSingleItem_7
-    stateFogItem .emit STATE_FOG;
-stateSingleItem_8
-    stateDepthItem .emit STATE_DEPTH;
-stateSingleItem_9
-    stateClipPlaneItem .emit STATE_CLIP_PLANE;
-stateSingleItem_10
-    statePointItem .emit STATE_POINT;
-stateSingleItem_11
-    stateMatrixRow .emit STATE_MATRIX_ROWS;
-
-/*
-    <stateMaterialItem>    ::= "material" <optFaceType> "." <stateMatProperty>
-*/
-stateMaterialItem
-    "material" .and optFaceType .and dot .and stateMatProperty .error INVALID_MATERIAL_PROPERTY;
-
-/*
-    <stateMatProperty>     ::= "ambient"
-                             | "diffuse"
-                             | "specular"
-                             | "emission"
-                             | "shininess"
-*/
-stateMatProperty
-    "ambient" .emit MATERIAL_AMBIENT .or
-    "diffuse" .emit MATERIAL_DIFFUSE .or
-    "specular" .emit MATERIAL_SPECULAR .or
-    "emission" .emit MATERIAL_EMISSION .or
-    "shininess" .emit MATERIAL_SHININESS;
-
-/*
-    <stateLightItem>       ::= "light" "[" <stateLightNumber> "]" "." 
-                                 <stateLightProperty>
-*/
-stateLightItem
-    "light" .and lbracket .and stateLightNumber .and rbracket .and dot .and
-    stateLightProperty .error INVALID_LIGHT_PROPERTY;
-
-/*
-    <stateLightProperty>   ::= "ambient"
-                             | "diffuse" 
-                             | "specular"
-                             | "position"
-                             | "attenuation"
-                             | "spot" "." <stateSpotProperty>
-                             | "half"
-*/
-stateLightProperty
-    "ambient" .emit LIGHT_AMBIENT .or
-    "diffuse" .emit LIGHT_DIFFUSE .or
-    "specular" .emit LIGHT_SPECULAR .or
-    "position" .emit LIGHT_POSITION .or
-    "attenuation" .emit LIGHT_ATTENUATION .or
-    stateLightProperty_1 .emit LIGHT_SPOT_DIRECTION .or
-    "half" .emit LIGHT_HALF;
-stateLightProperty_1
-    "spot" .and dot .and stateSpotProperty .error INVALID_SPOT_PROPERTY;
-
-/*
-    <stateSpotProperty>    ::= "direction" 
-*/
-stateSpotProperty
-    "direction";
-
-/*
-    <stateLightModelItem>  ::= "lightmodel" <stateLModProperty>
-*/
-stateLightModelItem
-    "lightmodel" .and stateLModProperty .error INVALID_LIGHTMODEL_PROPERTY;
-
-/*
-    <stateLModProperty>    ::= "." "ambient"
-                             | <optFaceType> "." "scenecolor"
-*/
-stateLModProperty
-    stateLModProperty_1 .or stateLModProperty_2;
-stateLModProperty_1
-    dot .and "ambient" .emit LIGHT_MODEL_AMBIENT;
-stateLModProperty_2
-    stateLModProperty_3 .emit LIGHT_MODEL_SCENECOLOR;
-stateLModProperty_3
-    optFaceType .and dot .and "scenecolor";
-
-/*
-    <stateLightProdItem>   ::= "lightprod" "[" <stateLightNumber> "]"
-                                 <optFaceType> "." <stateLProdProperty>
-*/
-stateLightProdItem
-    "lightprod" .and lbracket .and stateLightNumber .and rbracket .and optFaceType .and dot .and
-    stateLProdProperty .error INVALID_LIGHTPROD_PROPERTY;
-
-/*
-    <stateLProdProperty>   ::= "ambient"
-                             | "diffuse"
-                             | "specular"
-*/
-stateLProdProperty
-    "ambient" .emit LIGHT_PROD_AMBIENT .or
-    "diffuse" .emit LIGHT_PROD_DIFFUSE .or
-    "specular" .emit LIGHT_PROD_SPECULAR;
-
-/*
-    <stateLightNumber>     ::= <integer> from 0 to MAX_LIGHTS-1
-*/
-stateLightNumber
-    integer;
-
-/*
-fragment program
-    <stateTexEnvItem>      ::= "texenv" <optLegacyTexUnitNum> "." 
-                                 <stateTexEnvProperty>
-*/
-stateTexEnvItem
-    "texenv" .and optLegacyTexUnitNum .and dot .and
-    stateTexEnvProperty .error INVALID_TEXENV_PROPERTY;
-
-/*
-fragment program
-    <stateTexEnvProperty>  ::= "color"
-*/
-stateTexEnvProperty
-    "color" .emit TEX_ENV_COLOR;
-
-/*
-fragment program
-    <optLegacyTexUnitNum>  ::= ""
-                             | "[" <legacyTexUnitNum> "]"
-*/
-optLegacyTexUnitNum
-    optLegacyTexUnitNum_1 .or .true .emit 0x00;
-optLegacyTexUnitNum_1
-    lbracket_ne .and legacyTexUnitNum .and rbracket;
-
-/*
-fragment program
-    <legacyTexUnitNum>     ::= <integer> from 0 to MAX_TEXTURE_UNITS-1
-*/
-legacyTexUnitNum
-    integer;
-
-/*
-vertex program
-    <stateTexGenItem>      ::= "texgen" <optTexCoordNum> "."
-                                 <stateTexGenType> "." <stateTexGenCoord>
-*/
-stateTexGenItem
-    "texgen" .and optTexCoordNum .and dot .and stateTexGenType .error INVALID_TEXGEN_PROPERTY .and
-    dot .and stateTexGenCoord .error INVALID_TEXGEN_COORD;
-
-/*
-vertex program
-    <stateTexGenType>      ::= "eye"
-                             | "object"
-*/
-stateTexGenType
-    "eye" .emit TEX_GEN_EYE .or
-    "object" .emit TEX_GEN_OBJECT;
-
-/*
-vertex program
-    <stateTexGenCoord>     ::= "s"
-                             | "t"
-                             | "r"
-                             | "q"
-*/
-stateTexGenCoord
-    "s" .emit COMPONENT_X .or
-    "t" .emit COMPONENT_Y .or
-    "r" .emit COMPONENT_Z .or
-    "q" .emit COMPONENT_W;
-
-/*
-    <stateFogItem>         ::= "fog" "." <stateFogProperty>
-*/
-stateFogItem
-    "fog" .and dot .and stateFogProperty .error INVALID_FOG_PROPERTY;
-
-/*
-    <stateFogProperty>     ::= "color"
-                             | "params"
-*/
-stateFogProperty
-    "color" .emit FOG_COLOR .or
-    "params" .emit FOG_PARAMS;
-
-/*
-fragment program
-    <stateDepthItem>       ::= "depth" "." <stateDepthProperty>
-*/
-stateDepthItem
-    "depth" .and dot .and stateDepthProperty .error INVALID_DEPTH_PROPERTY;
-
-/*
-fragment program
-    <stateDepthProperty>   ::= "range"
-*/
-stateDepthProperty
-    "range" .emit DEPTH_RANGE;
-
-/*
-vertex program
-    <stateClipPlaneItem>   ::= "clip" "[" <stateClipPlaneNum> "]" "." "plane"
-*/
-stateClipPlaneItem
-    "clip" .and lbracket .and stateClipPlaneNum .and rbracket .and dot .and
-    "plane" .error INVALID_CLIPPLANE_PROPERTY;
-
-/*
-vertex program
-    <stateClipPlaneNum>    ::= <integer> from 0 to MAX_CLIP_PLANES-1
-*/
-stateClipPlaneNum
-    integer;
-
-/*
-vertex program
-    <statePointItem>       ::= "point" "." <statePointProperty>
-*/
-statePointItem
-    "point" .and dot .and statePointProperty .error INVALID_POINT_PROPERTY;
-
-/*
-vertex program
-    <statePointProperty>   ::= "size"
-                             | "attenuation"
-*/
-statePointProperty
-    "size" .emit POINT_SIZE .or
-    .if (point_parameters != 0x00) "attenuation" .emit POINT_ATTENUATION;
-
-/*
-    <stateMatrixRow>       ::= <stateMatrixItem> "." "row" "[" 
-                                  <stateMatrixRowNum> "]"
-*/
-stateMatrixRow
-    stateMatrixItem .and dot .and "row" .error MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED .and
-    lbracket .and stateMatrixRowNum .and rbracket .emit 0x0;
-
-/*
-    <stateMatrixRows>      ::= <stateMatrixItem> <optMatrixRows>
-*/
-stateMatrixRows
-    stateMatrixItem .and optMatrixRows;
-
-/*
-    <optMatrixRows>        ::= ""
-                             | "." "row" "[" <stateMatrixRowNum> ".." 
-                                  <stateMatrixRowNum> "]"
-*/
-optMatrixRows
-    optMatrixRows_1 .or .true .emit 0x0 .emit '3' .emit 0x0 .emit $;
-optMatrixRows_1
-    dot_ne .and "row" .error MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED .and lbracket .and
-    stateMatrixRowNum .and dotdot .and stateMatrixRowNum .and rbracket;
-
-/*
-    <stateMatrixItem>      ::= "matrix" "." <stateMatrixName> 
-                               <stateOptMatModifier>
-*/
-stateMatrixItem
-    "matrix" .and dot .and stateMatrixName .error INVALID_MATRIX_NAME .and stateOptMatModifier;
-
-/*
-    <stateOptMatModifier>  ::= ""
-                             | "." <stateMatModifier>
-*/
-stateOptMatModifier
-    stateOptMatModifier_1 .or .true .emit MATRIX_MODIFIER_IDENTITY;
-stateOptMatModifier_1
-    dot_ne .and stateMatModifier;
-
-/*
-    <stateMatModifier>     ::= "inverse" 
-                             | "transpose" 
-                             | "invtrans"
-*/
-stateMatModifier
-    "inverse" .emit MATRIX_MODIFIER_INVERSE .or
-    "transpose" .emit MATRIX_MODIFIER_TRANSPOSE .or
-    "invtrans" .emit MATRIX_MODIFIER_INVTRANS;
-
-/*
-    <stateMatrixRowNum>    ::= <integer> from 0 to 3
-*/
-stateMatrixRowNum
-    integer_0_3;
-
-/*
-    <stateMatrixName>      ::= "modelview" <stateOptModMatNum>
-                             | "projection"
-                             | "mvp"
-                             | "texture" <optTexCoordNum>
-                             | "palette" "[" <statePaletteMatNum> "]"
-                             | "program" "[" <stateProgramMatNum> "]"
-*/
-stateMatrixName
-    stateMatrixName_1_1 .emit MATRIX_MODELVIEW .or
-    "projection" .emit MATRIX_PROJECTION .or
-    "mvp" .emit MATRIX_MVP .or
-    stateMatrixName_1_2 .emit MATRIX_TEXTURE .or
-    .if (matrix_palette != 0x00) stateMatrixName_1_3 .emit MATRIX_PALETTE .or
-    stateMatrixName_1_4 .emit MATRIX_PROGRAM;
-stateMatrixName_1_1
-    "modelview" .and stateOptModMatNum;
-stateMatrixName_1_2
-    "texture" .and optTexCoordNum;
-stateMatrixName_1_3
-    "palette" .and lbracket .and statePaletteMatNum .and rbracket;
-stateMatrixName_1_4
-    "program" .and lbracket .and stateProgramMatNum .and rbracket;
-
-/*
-    <stateOptModMatNum>    ::= ""
-                             | "[" <stateModMatNum> "]"
-*/
-stateOptModMatNum
-    .if (vertex_blend != 0x00) stateOptModMatNum_1 .or
-    .true .emit 0x00;
-stateOptModMatNum_1
-    lbracket_ne .and stateModMatNum .and rbracket;
-
-/*
-    <stateModMatNum>       ::= <integer> from 0 to MAX_VERTEX_UNITS_ARB-1
-*/
-stateModMatNum
-    integer;
-
-/*
-    <optTexCoordNum>       ::= ""
-                             | "[" <texCoordNum> "]"
-*/
-optTexCoordNum
-    optTexCoordNum_1 .or .true .emit 0x00;
-optTexCoordNum_1
-    lbracket_ne .and texCoordNum .and rbracket;
-
-/*
-    <texCoordNum>          ::= <integer> from 0 to MAX_TEXTURE_COORDS_ARB-1
-*/
-texCoordNum
-    integer;
-
-/*
-    <statePaletteMatNum>   ::= <integer> from 0 to MAX_PALETTE_MATRICES_ARB-1
-*/
-statePaletteMatNum
-    integer;
-
-/*
-    <stateProgramMatNum>   ::= <integer> from 0 to MAX_PROGRAM_MATRICES_ARB-1
-*/
-stateProgramMatNum
-    integer;
-
-/*
-    <programSingleItem>    ::= <progEnvParam>
-                             | <progLocalParam>
-
-NOTE: <programSingleItem> has been modified for correct error handling. If program property
-      is neither "env" nor "local" INVALID_PROGRAM_PROPERTY is generated.
-*/
-programSingleItem
-    "program" .and dot .and programSingleItem_1 .error INVALID_PROGRAM_PROPERTY;
-programSingleItem_1
-    progEnvParam .or progLocalParam;
-
-/*
-    <programMultipleItem>  ::= <progEnvParams>
-                             | <progLocalParams>
-
-NOTE: <programMultipleItem> has been modified for correct error handling. If program property
-      is neither "env" nor "local" INVALID_PROGRAM_PROPERTY is generated.
-*/
-programMultipleItem
-    "program" .and dot .and programMultipleItem_1 .error INVALID_PROGRAM_PROPERTY;
-programMultipleItem_1
-    progEnvParams .or progLocalParams;
-
-/*
-    <progEnvParams>        ::= "program" "." "env" 
-                                 "[" <progEnvParamNums> "]"
-
-NOTE: "program" "." has been moved to <programMultipleItem>.
-*/
-progEnvParams
-    "env" .emit PROGRAM_PARAM_ENV .and lbracket .and progEnvParamNums .and rbracket;
-
-/*
-    <progEnvParamNums>     ::= <progEnvParamNum>
-                             | <progEnvParamNum> ".." <progEnvParamNum>
-*/
-progEnvParamNums
-    progEnvParamNums_1 .or progEnvParamNums_2;
-progEnvParamNums_1
-    progEnvParamNum .and dotdot_ne .and progEnvParamNum;
-progEnvParamNums_2
-    progEnvParamNum .and .true .emit 0x00;
-
-/*
-    <progEnvParam>         ::= "program" "." "env" 
-                                 "[" <progEnvParamNum> "]"
-
-NOTE: "program" "." has been moved to <programSingleItem>.
-*/
-progEnvParam
-    "env" .emit PROGRAM_PARAM_ENV .and lbracket .and progEnvParamNum .and rbracket .emit 0x00;
-
-/*
-    <progLocalParams>      ::= "program" "." "local" 
-                                 "[" <progLocalParamNums> "]"
-
-NOTE: "program" "." has been moved to <programMultipleItem>.
-*/
-progLocalParams
-    "local" .emit PROGRAM_PARAM_LOCAL .and lbracket .and progLocalParamNums .and rbracket;
-
-/*
-    <progLocalParamNums>   ::= <progLocalParamNum>
-                             | <progLocalParamNum> ".." <progLocalParamNum>
-*/
-progLocalParamNums
-    progLocalParamNums_1 .or progLocalParamNums_2;
-progLocalParamNums_1
-    progLocalParamNum .and dotdot_ne .and progLocalParamNum;
-progLocalParamNums_2
-    progLocalParamNum .and .true .emit 0x00;
-
-/*
-    <progLocalParam>       ::= "program" "." "local" 
-                                 "[" <progLocalParamNum> "]"
-
-NOTE: "program" "." has been moved to <programSingleItem>.
-*/
-progLocalParam
-    "local" .emit PROGRAM_PARAM_LOCAL .and lbracket .and progLocalParamNum .and rbracket .emit 0x00;
-
-/*
-    <progEnvParamNum>      ::= <integer> from 0 to
-                               MAX_PROGRAM_ENV_PARAMETERS_ARB - 1
-*/
-progEnvParamNum
-    integer;
-
-/*
-    <progLocalParamNum>    ::= <integer> from 0 to
-                               MAX_PROGRAM_LOCAL_PARAMETERS_ARB - 1
-*/
-progLocalParamNum
-    integer;
-
-/*
-    <paramConstDecl>       ::= <paramConstScalarDecl>
-                             | <paramConstVector>
-*/
-paramConstDecl
-    paramConstScalarDecl .emit CONSTANT_SCALAR .or paramConstVector .emit CONSTANT_VECTOR;
-
-/*
-    <paramConstUse>        ::= <paramConstScalarUse>
-                             | <paramConstVector>
-*/
-paramConstUse
-    paramConstScalarUse .emit CONSTANT_SCALAR .or paramConstVector .emit CONSTANT_VECTOR;
-
-/*
-    <paramConstScalarDecl> ::= <signedFloatConstant>
-*/
-paramConstScalarDecl
-    signedFloatConstant;
-
-/*
-    <paramConstScalarUse>  ::= <floatConstant>
-*/
-paramConstScalarUse
-    floatConstant;
-
-/*
-    <paramConstVector>     ::= "{" <signedFloatConstant> "}"
-                             | "{" <signedFloatConstant> "," 
-                                   <signedFloatConstant> "}"
-                             | "{" <signedFloatConstant> "," 
-                                   <signedFloatConstant> ","
-                                   <signedFloatConstant> "}"
-                             | "{" <signedFloatConstant> "," 
-                                   <signedFloatConstant> ","
-                                   <signedFloatConstant> "," 
-                                   <signedFloatConstant> "}"
-*/
-paramConstVector
-    paramConstVector_4 .emit 0x04 .or paramConstVector_3 .emit 0x03 .or
-    paramConstVector_2 .emit 0x02 .or paramConstVector_1 .emit 0x01;
-paramConstVector_1
-    lbrace_ne .and signedFloatConstant .and rbrace;
-paramConstVector_2
-    lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and rbrace;
-paramConstVector_3
-    lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and comma_ne .and
-    signedFloatConstant .and rbrace;
-paramConstVector_4
-    lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and comma_ne .and
-    signedFloatConstant .and comma_ne .and signedFloatConstant .and rbrace;
-
-/*
-    <signedFloatConstant>  ::= <optionalSign> <floatConstant>
-*/
-signedFloatConstant
-    optionalSign .and floatConstant;
-
-/*
-    <floatConstant>        ::= see text
-    The <floatConstant> rule matches a floating-point constant consisting
-    of an integer part, a decimal point, a fraction part, an "e" or
-    "E", and an optionally signed integer exponent.  The integer and
-    fraction parts both consist of a sequence of one or more digits ("0"
-    through "9").  Either the integer part or the fraction parts (not
-    both) may be missing; either the decimal point or the "e" (or "E")
-    and the exponent (not both) may be missing.
-*/
-floatConstant
-    float;
-
-/*
-    <optionalSign>         ::= ""
-                             | "-"
-                             | "+"
-*/
-optionalSign
-    optional_sign_ne;
-
-/*
-    <TEMP_statement>       ::= "TEMP" <varNameList>
-*/
-fp_TEMP_statement
-    "TEMP" .and space .and fp_varNameList .and .true .emit 0x00;
-vp_TEMP_statement
-    "TEMP" .and space .and vp_varNameList .and .true .emit 0x00;
-
-/*
-vertex program
-    <ADDRESS_statement>    ::= "ADDRESS" <varNameList>
-*/
-ADDRESS_statement
-    "ADDRESS" .and space .and vp_varNameList .and .true .emit 0x00;
-
-/*
-    <varNameList>          ::= <establishName>
-                             | <establishName> "," <varNameList>
-*/
-fp_varNameList
-    fp_varNameList_1 .or fp_establishName;
-vp_varNameList
-    vp_varNameList_1 .or vp_establishName;
-fp_varNameList_1
-    fp_establishName .and comma_ne .and fp_varNameList;
-vp_varNameList_1
-    vp_establishName .and comma_ne .and vp_varNameList;
-
-/*
-    <OUTPUT_statement>     ::= "OUTPUT" <establishName> "="
-                                 <resultBinding>
-*/
-fp_OUTPUT_statement
-    "OUTPUT" .and space .and fp_establishName .and equal .and
-    fp_resultBinding .error RESULT_EXPECTED;
-vp_OUTPUT_statement
-    "OUTPUT" .and space .and vp_establishName .and equal .and
-    vp_resultBinding .error RESULT_EXPECTED;
-
-/*
-fragment program
-    <resultBinding>        ::= "result" "." "color"
-                             | "result" "." "color" <optOutputColorNum> (if option ARB_draw_buffers present)
-                             | "result" "." "depth"
-
-vertex program
-    <resultBinding>        ::= "result" "." "position"
-                             | "result" "." <resultColBinding>
-                             | "result" "." "fogcoord"
-                             | "result" "." "pointsize"
-                             | "result" "." "texcoord" <optTexCoordNum>
-*/
-fp_resultBinding
-    "result" .and dot .and fp_resultBinding_1 .error INVALID_RESULT_PROPERTY;
-vp_resultBinding
-    "result" .and dot .and vp_resultBinding_1 .error INVALID_RESULT_PROPERTY;
-fp_resultBinding_1
-    fp_resultBinding_2 .emit FRAGMENT_RESULT_COLOR .or
-    "depth" .emit FRAGMENT_RESULT_DEPTH;
-fp_resultBinding_2
-    "color" .and optOutputColorNum;
-vp_resultBinding_1
-    .if (ARB_position_invariant == 0x00) "position" .emit VERTEX_RESULT_POSITION .or
-    resultColBinding .emit VERTEX_RESULT_COLOR .or
-    "fogcoord" .emit VERTEX_RESULT_FOGCOORD .or
-    "pointsize" .emit VERTEX_RESULT_POINTSIZE .or
-    vp_resultBinding_2 .emit VERTEX_RESULT_TEXCOORD;
-vp_resultBinding_2
-    "texcoord" .and optTexCoordNum;
-
-/*
-GL_ARB_draw_buffers
-    <optOutputColorNum>    ::= ""
-                             | "[" <outputColorNum> "]"
-*/
-optOutputColorNum
-    .if (ARB_draw_buffers != 0x00) optOutputColorNum_1 .or .true .emit 0x00;
-optOutputColorNum_1
-    lbracket_ne .and outputColorNum .and rbracket;
-
-/*
-GL_ARB_draw_buffers
-    <outputColorNum>       ::= <integer> from 0 to MAX_DRAW_BUFFERS_ARB-1
-*/
-outputColorNum
-    integer;
-
-/*
-vertex program
-    <resultColBinding>     ::= "color" <optFaceType> <optColorType>
-*/
-resultColBinding
-    "color" .and optFaceType .and optColorType;
-
-/*
-    <optFaceType>          ::= ""
-                             | "." "front"
-                             | "." "back"
-*/
-optFaceType
-    FaceType .or .true .emit FACE_FRONT;
-FaceType
-    dot_ne .and FaceProperty;
-FaceProperty
-    "front" .emit FACE_FRONT .or "back" .emit FACE_BACK;
-
-/*
-    <optColorType>         ::= ""
-                             | "." "primary"
-                             | "." "secondary"
-*/
-optColorType
-    ColorType .or .true .emit COLOR_PRIMARY;
-ColorType
-    dot_ne .and ColorProperty;
-ColorProperty
-    "primary" .emit COLOR_PRIMARY .or
-    .if (secondary_color != 0x00) "secondary" .emit COLOR_SECONDARY;
-
-/*
-    <ALIAS_statement>      ::= "ALIAS" <establishName> "="
-                                 <establishedName>
-*/
-fp_ALIAS_statement
-    "ALIAS" .and fp_ALIAS_statement_1 .error IDENTIFIER_EXPECTED .and equal .and fp_establishedName;
-vp_ALIAS_statement
-    "ALIAS" .and vp_ALIAS_statement_1 .error IDENTIFIER_EXPECTED .and equal .and vp_establishedName;
-fp_ALIAS_statement_1
-    space .and fp_establishName;
-vp_ALIAS_statement_1
-    space .and vp_establishName;
-
-/*
-    <establishName>        ::= <identifier>
-*/
-fp_establishName
-    fp_identifier;
-vp_establishName
-    vp_identifier;
-
-/*
-    <establishedName>      ::= <identifier>
-*/
-fp_establishedName
-    fp_identifier;
-vp_establishedName
-    vp_identifier;
-fp_establishedName_no_error_on_identifier
-    fp_identifier_ne;
-vp_establishedName_no_error_on_identifier
-    vp_identifier_ne;
-
-/*
-fragment program
-    <identifier>           ::= see text
-    The <identifier> rule matches a sequence of one or more letters ("A"
-    through "Z", "a" through "z"), digits ("0" through "9), underscores 
-    ("_"), or dollar signs ("$"); the first character must not be a 
-    number.  Upper and lower case letters are considered different 
-    (names are case-sensitive).  The following strings are reserved 
-    keywords and may not be used as identifiers:
-
-        ABS, ABS_SAT, ADD, ADD_SAT, ALIAS, ATTRIB, CMP, CMP_SAT, COS,
-        COS_SAT, DP3, DP3_SAT, DP4, DP4_SAT, DPH, DPH_SAT, DST, DST_SAT, 
-        END, EX2, EX2_SAT, FLR, FLR_SAT, FRC, FRC_SAT, KIL, LG2, 
-        LG2_SAT, LIT, LIT_SAT, LRP, LRP_SAT, MAD, MAD_SAT, MAX, MAX_SAT, 
-        MIN, MIN_SAT, MOV, MOV_SAT, MUL, MUL_SAT, OPTION, OUTPUT, PARAM, 
-        POW, POW_SAT, RCP, RCP_SAT, RSQ, RSQ_SAT, SIN, SIN_SAT, SCS, 
-        SCS_SAT, SGE, SGE_SAT, SLT, SLT_SAT, SUB, SUB_SAT, SWZ, SWZ_SAT, 
-        TEMP, TEX, TEX_SAT, TXB, TXB_SAT, TXP, TXP_SAT, XPD, XPD_SAT, 
-        fragment, program, result, state, and texture.
-
-vertex program
-    <identifier>           ::= see text
-    The <identifier> rule matches a sequence of one or more letters ("A"
-    through "Z", "a" through "z"), digits ("0" through "9), underscores ("_"),
-    or dollar signs ("$"); the first character must not be a number.  Upper
-    and lower case letters are considered different (names are
-    case-sensitive).  The following strings are reserved keywords and may not
-    be used as identifiers:
-
-        ABS, ADD, ADDRESS, ALIAS, ARL, ATTRIB, DP3, DP4, DPH, DST, END, EX2,
-        EXP, FLR, FRC, LG2, LIT, LOG, MAD, MAX, MIN, MOV, MUL, OPTION, OUTPUT,
-        PARAM, POW, RCP, RSQ, SGE, SLT, SUB, SWZ, TEMP, XPD, program, result,
-        state, and vertex.
-*/
-fp_identifier
-    fp_identifier_ne .error IDENTIFIER_EXPECTED;
-vp_identifier
-    vp_identifier_ne .error IDENTIFIER_EXPECTED;
-fp_identifier_ne
-    fp_not_reserved_identifier .and identifier_ne;
-vp_identifier_ne
-    vp_not_reserved_identifier .and identifier_ne;
-
-fp_not_reserved_identifier
-    fp_not_reserved_identifier_1 .or .true;
-fp_not_reserved_identifier_1
-    fp_reserved_identifier .and .false .error RESERVED_KEYWORD;
-vp_not_reserved_identifier
-    vp_not_reserved_identifier_1 .or .true;
-vp_not_reserved_identifier_1
-    vp_reserved_identifier .and .false .error RESERVED_KEYWORD;
-
-fp_reserved_identifier
-    "ABS" .or "ABS_SAT" .or "ADD" .or "ADD_SAT" .or "ALIAS" .or "ATTRIB" .or "CMP" .or "CMP_SAT" .or
-    "COS" .or "COS_SAT" .or "DP3" .or "DP3_SAT" .or "DP4" .or "DP4_SAT" .or "DPH" .or "DPH_SAT" .or
-    "DST" .or "DST_SAT" .or "END" .or "EX2" .or "EX2_SAT" .or "FLR" .or "FLR_SAT" .or "FRC" .or
-    "FRC_SAT" .or "KIL" .or "LG2" .or "LG2_SAT" .or "LIT" .or "LIT_SAT" .or "LRP" .or "LRP_SAT" .or
-    "MAD" .or "MAD_SAT" .or "MAX" .or "MAX_SAT" .or "MIN" .or "MIN_SAT" .or "MOV" .or "MOV_SAT" .or
-    "MUL" .or "MUL_SAT" .or "OPTION" .or "OUTPUT" .or "PARAM" .or "POW" .or "POW_SAT" .or "RCP" .or
-    "RCP_SAT" .or "RSQ" .or "RSQ_SAT" .or "SIN" .or "SIN_SAT" .or "SCS" .or "SCS_SAT" .or "SGE" .or
-    "SGE_SAT" .or "SLT" .or "SLT_SAT" .or "SUB" .or "SUB_SAT" .or "SWZ" .or "SWZ_SAT" .or "TEMP" .or
-    "TEX" .or "TEX_SAT" .or "TXB" .or "TXB_SAT" .or "TXP" .or "TXP_SAT" .or "XPD" .or "XPD_SAT" .or
-    "fragment" .or "program" .or "result" .or "state" .or "texture";
-vp_reserved_identifier
-    "ABS" .or "ADD" .or "ADDRESS" .or "ALIAS" .or "ARL" .or "ATTRIB" .or "DP3" .or "DP4" .or
-    "DPH" .or "DST" .or "END" .or "EX2" .or "EXP" .or "FLR" .or "FRC" .or "LG2" .or "LIT" .or
-    "LOG" .or "MAD" .or "MAX" .or "MIN" .or "MOV" .or "MUL" .or "OPTION" .or "OUTPUT" .or
-    "PARAM" .or "POW" .or "RCP" .or "RSQ" .or "SGE" .or "SLT" .or "SUB" .or "SWZ" .or "TEMP" .or
-    "XPD" .or "program" .or "result" .or "state" .or "vertex";
-
-/*
-    The <integer> rule matches an integer constant.  The integer consists
-    of a sequence of one or more digits ("0" through "9").
-*/
-integer
-    integer_ne .error INTEGER_EXPECTED;
-
-zero
-    '0';
-
-leading_zeroes
-    .loop zero;
-
-no_digit
-    no_digit_1 .or .true;
-no_digit_1
-    digit10 .and .false .error INTEGER_OUT_OF_RANGE;
-
-all_zeroes
-    all_zeroes_1 .or no_digit_1;
-all_zeroes_1
-    '0' .and .loop zero .and no_digit;
-
-integer_0_3
-    integer_0_3_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;
-integer_0_3_1
-    integer_0_3_2 .or all_zeroes .emit '0';
-integer_0_3_2       /* [1, 3] */
-    leading_zeroes .and '1'-'3' .emit * .and no_digit;
-
-integer_0_63
-    integer_0_63_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;
-integer_0_63_1
-    integer_0_63_2 .or integer_0_63_3 .or integer_0_63_4 .or integer_0_63_5 .or
-    all_zeroes .emit '0';
-integer_0_63_2      /* [7, 9] */
-    leading_zeroes .and '7'-'9' .emit * .and no_digit;
-integer_0_63_3      /* [10, 59] */
-    leading_zeroes .and '1'-'5' .emit * .and '0'-'9' .emit * .and no_digit;
-integer_0_63_4      /* [60, 63] */
-    leading_zeroes .and '6' .emit * .and '0'-'3' .emit * .and no_digit;
-integer_0_63_5      /* [1, 6] */
-    leading_zeroes .and '1'-'6' .emit * .and no_digit;
-
-integer_0_64
-    integer_0_64_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;
-integer_0_64_1
-    integer_0_64_2 .or integer_0_64_3 .or integer_0_64_4 .or integer_0_64_5 .or
-    all_zeroes .emit '0';
-integer_0_64_2      /* [7, 9] */
-    leading_zeroes .and '7'-'9' .emit * .and no_digit;
-integer_0_64_3      /* [10, 59] */
-    leading_zeroes .and '1'-'5' .emit * .and '0'-'9' .emit * .and no_digit;
-integer_0_64_4      /* [60, 64] */
-    leading_zeroes .and '6' .emit * .and '0'-'4' .emit * .and no_digit;
-integer_0_64_5      /* [1, 6] */
-    leading_zeroes .and '1'-'6' .emit * .and no_digit;
-
-optional_space
-    space .or .true;
-
-space_dst
-    space .error OPERATION_NEEDS_DESTINATION_VARIABLE;
-
-space_src
-    space .error OPERATION_NEEDS_SOURCE_VARIABLE;
-
-space
-    single_space .and .loop single_space;
-
-single_space
-    white_char .or comment_block;
-
-white_char
-    ' ' .or '\t' .or '\n' .or '\r';
-
-comment_block
-    '#' .and .loop comment_char .and optional_new_line;
-
-/* All ASCII characters except '\r', '\n' and '\0' */
-comment_char
-    '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C';
-
-optional_new_line
-    '\n' .or crlf .or .true;
-
-crlf
-    '\r' .and '\n';
-
-semicolon
-    optional_space .and ';' .error MISSING_SEMICOLON .and optional_space;
-
-comma
-    optional_space .and ',' .error MISSING_COMMA .and optional_space;
-
-comma_ne
-    optional_space .and ',' .and optional_space;
-
-lbracket
-    optional_space .and '[' .error MISSING_LBRACKET .and optional_space;
-
-lbracket_ne
-    optional_space .and '[' .and optional_space;
-
-rbracket
-    optional_space .and ']' .error MISSING_RBRACKET .and optional_space;
-
-dot
-    optional_space .and '.' .error MISSING_DOT .and optional_space;
-
-dot_ne
-    optional_space .and '.' .and optional_space;
-
-equal
-    optional_space .and '=' .error MISSING_EQUAL .and optional_space;
-
-lbrace
-    optional_space .and '{' .error MISSING_LBRACE .and optional_space;
-
-lbrace_ne
-    optional_space .and '{' .and optional_space;
-
-rbrace
-    optional_space .and '}' .error MISSING_RBRACE .and optional_space;
-
-dotdot
-    optional_space .and '.' .and '.' .error MISSING_DOTDOT .and optional_space;
-
-dotdot_ne
-    optional_space .and '.' .and '.' .and optional_space;
-
-/*
-    The definition below accepts the following floating point number formats:
-    .99 .99e99 99. 99.99 99.99e99 99.e99 99e99
-    Also 99 format was considered and accepted because of a large number of existing program
-    strings with such a format.
-*/
-float
-    float_1 .or float_2 .or float_legacy;
-float_1
-    '.' .emit 0x00 .and integer_ne .error MISSING_FRACTION_OR_EXPONENT .and optional_exponent;
-float_2
-    integer_ne .and float_3;
-float_3
-    float_4 .or float_5;
-float_4
-    '.' .and optional_integer .and optional_exponent;
-float_5
-    exponent .emit 0x00;
-float_legacy
-    integer_ne .and .true .emit 0x00 .emit 0x00;
-
-/*
-    Below is a correct version of <float> definiton.
-*/
-/*
-float
-    float_1 .or float_2;
-float_1
-    '.' .emit 0x00 .and integer_ne .error MISSING_FRACTION_OR_EXPONENT .and optional_exponent;
-float_2
-    integer_ne .and float_3 .error MISSING_DOT_OR_EXPONENT;
-float_3
-    float_4 .or float_5;
-float_4
-    '.' .and optional_integer .and optional_exponent;
-float_5
-    exponent .emit 0x00;
-*/
-
-integer_ne
-    integer_ne_1 .and .true .emit 0x00 .emit $;
-integer_ne_1
-    digit10 .emit * .and .loop digit10 .emit *;
-
-optional_integer
-    integer_ne .or .true .emit 0x00;
-
-/*
-NOTE: If exponent part is omited we treat it as if it was "E+1".
-*/
-optional_exponent
-    exponent .or .true .emit 0x00;
-
-exponent
-    exponent_1 .and optional_sign_ne .and integer_ne .error EXPONENT_VALUE_EXPECTED;
-exponent_1
-    'e' .or 'E';
-
-optional_sign_ne
-    minus_ne .or plus_ne .or .true;
-
-plus_ne
-    optional_space .and '+' .and optional_space;
-
-minus_ne
-    optional_space .and '-' .emit '-' .and optional_space;
-
-identifier_ne
-    first_idchar .emit * .and .loop follow_idchar .emit * .and .true .emit 0x00 .emit $;
-
-follow_idchar
-    first_idchar .or digit10;
-
-first_idchar
-    'a'-'z' .or 'A'-'Z' .or '_' .or '$';
-
-digit10
-    '0'-'9';
-
-/*
-    string filtering - if a string is encountered in grammar ("blabla"), the symbol below is
-    executed to create the string. The symbol must not throw any errors and emit bytes - it should
-    stop if it encounters invalid character. After this the resulting string (from starting
-    position up to the invalid character (but without it) is compared with the grammar string.
-*/
-.string __string_filter;
-
-__string_filter
-    .loop __identifier_char;
-
-__identifier_char
-    'a'-'z' .or 'A'-'Z' .or '_' .or '$' .or '0'-'9';
-
-/*
-    error token filtering
-*/
-e_signature
-    e_signature_char .and .loop e_signature_char;
-e_signature_char
-    '!' .or '.' .or 'A'-'Z' .or 'a'-'z' .or '0'-'9';
-
-e_statement
-    .loop e_statement_not_term;
-/* All ASCII characters to one of '\r', '\n', '\0' and ';' */
-e_statement_not_term
-    '\x3C'-'\xFF' .or '\x0E'-'\x3A' .or '\x01'-'\x09' .or '\x0B'-'\x0C';
-
-e_identifier
-    e_identifier_first .and .loop e_identifier_next;
-e_identifier_first
-    'a'-'z' .or 'A'-'Z' .or '_' .or '$';
-e_identifier_next
-    e_identifier_first .or '0'-'9';
-
-e_token
-    e_identifier .or e_token_number .or '[' .or ']' .or '.' .or '{' .or '}' .or '=' .or '+' .or
-    '-' .or ',' .or ';';
-e_token_number
-    e_token_digit .and .loop e_token_digit;
-e_token_digit
-    '0'-'9';
-
-e_charordigit
-    'A'-'Z' .or 'a'-'z' .or '0'-'9';
-
diff --git a/src/mesa/shader/arbprogram_syn.h b/src/mesa/shader/arbprogram_syn.h
deleted file mode 100644
index d95a5dede4..0000000000
--- a/src/mesa/shader/arbprogram_syn.h
+++ /dev/null
@@ -1,1350 +0,0 @@
-
-/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */
-
-" \n"
-".syntax program;\n"
-".emtcode REVISION 0x0a\n"
-".emtcode FRAGMENT_PROGRAM 0x01\n"
-".emtcode VERTEX_PROGRAM 0x02\n"
-".emtcode OPTION 0x01\n"
-".emtcode INSTRUCTION 0x02\n"
-".emtcode DECLARATION 0x03\n"
-".emtcode END 0x04\n"
-".emtcode ARB_PRECISION_HINT_FASTEST 0x00\n"
-".emtcode ARB_PRECISION_HINT_NICEST 0x01\n"
-".emtcode ARB_FOG_EXP 0x02\n"
-".emtcode ARB_FOG_EXP2 0x03\n"
-".emtcode ARB_FOG_LINEAR 0x04\n"
-".emtcode ARB_POSITION_INVARIANT 0x05\n"
-".emtcode ARB_FRAGMENT_PROGRAM_SHADOW 0x06\n"
-".emtcode ARB_DRAW_BUFFERS 0x07\n"
-".emtcode MESA_TEXTURE_ARRAY 0x08\n"
-".emtcode OP_ALU_INST 0x00\n"
-".emtcode OP_TEX_INST 0x01\n"
-".emtcode OP_ALU_VECTOR 0x00\n"
-".emtcode OP_ALU_SCALAR 0x01\n"
-".emtcode OP_ALU_BINSC 0x02\n"
-".emtcode OP_ALU_BIN 0x03\n"
-".emtcode OP_ALU_TRI 0x04\n"
-".emtcode OP_ALU_SWZ 0x05\n"
-".emtcode OP_TEX_SAMPLE 0x06\n"
-".emtcode OP_TEX_KIL 0x07\n"
-".emtcode OP_ALU_ARL 0x08\n"
-".emtcode OP_ABS 0x00\n"
-".emtcode OP_ABS_SAT 0x1B\n"
-".emtcode OP_FLR 0x09\n"
-".emtcode OP_FLR_SAT 0x26\n"
-".emtcode OP_FRC 0x0A\n"
-".emtcode OP_FRC_SAT 0x27\n"
-".emtcode OP_LIT 0x0C\n"
-".emtcode OP_LIT_SAT 0x2A\n"
-".emtcode OP_MOV 0x11\n"
-".emtcode OP_MOV_SAT 0x30\n"
-".emtcode OP_COS 0x1F\n"
-".emtcode OP_COS_SAT 0x20\n"
-".emtcode OP_EX2 0x07\n"
-".emtcode OP_EX2_SAT 0x25\n"
-".emtcode OP_LG2 0x0B\n"
-".emtcode OP_LG2_SAT 0x29\n"
-".emtcode OP_RCP 0x14\n"
-".emtcode OP_RCP_SAT 0x33\n"
-".emtcode OP_RSQ 0x15\n"
-".emtcode OP_RSQ_SAT 0x34\n"
-".emtcode OP_SIN 0x38\n"
-".emtcode OP_SIN_SAT 0x39\n"
-".emtcode OP_SCS 0x35\n"
-".emtcode OP_SCS_SAT 0x36\n"
-".emtcode OP_POW 0x13\n"
-".emtcode OP_POW_SAT 0x32\n"
-".emtcode OP_ADD 0x01\n"
-".emtcode OP_ADD_SAT 0x1C\n"
-".emtcode OP_DP3 0x03\n"
-".emtcode OP_DP3_SAT 0x21\n"
-".emtcode OP_DP4 0x04\n"
-".emtcode OP_DP4_SAT 0x22\n"
-".emtcode OP_DPH 0x05\n"
-".emtcode OP_DPH_SAT 0x23\n"
-".emtcode OP_DST 0x06\n"
-".emtcode OP_DST_SAT 0x24\n"
-".emtcode OP_MAX 0x0F\n"
-".emtcode OP_MAX_SAT 0x2E\n"
-".emtcode OP_MIN 0x10\n"
-".emtcode OP_MIN_SAT 0x2F\n"
-".emtcode OP_MUL 0x12\n"
-".emtcode OP_MUL_SAT 0x31\n"
-".emtcode OP_SGE 0x16\n"
-".emtcode OP_SGE_SAT 0x37\n"
-".emtcode OP_SLT 0x17\n"
-".emtcode OP_SLT_SAT 0x3A\n"
-".emtcode OP_SUB 0x18\n"
-".emtcode OP_SUB_SAT 0x3B\n"
-".emtcode OP_XPD 0x1A\n"
-".emtcode OP_XPD_SAT 0x43\n"
-".emtcode OP_CMP 0x1D\n"
-".emtcode OP_CMP_SAT 0x1E\n"
-".emtcode OP_LRP 0x2B\n"
-".emtcode OP_LRP_SAT 0x2C\n"
-".emtcode OP_MAD 0x0E\n"
-".emtcode OP_MAD_SAT 0x2D\n"
-".emtcode OP_SWZ 0x19\n"
-".emtcode OP_SWZ_SAT 0x3C\n"
-".emtcode OP_TEX 0x3D\n"
-".emtcode OP_TEX_SAT 0x3E\n"
-".emtcode OP_TXB 0x3F\n"
-".emtcode OP_TXB_SAT 0x40\n"
-".emtcode OP_TXP 0x41\n"
-".emtcode OP_TXP_SAT 0x42\n"
-".emtcode OP_KIL 0x28\n"
-".emtcode OP_ARL 0x02\n"
-".emtcode OP_EXP 0x08\n"
-".emtcode OP_LOG 0x0D\n"
-".emtcode FRAGMENT_ATTRIB_COLOR 0x01\n"
-".emtcode FRAGMENT_ATTRIB_TEXCOORD 0x02\n"
-".emtcode FRAGMENT_ATTRIB_FOGCOORD 0x03\n"
-".emtcode FRAGMENT_ATTRIB_POSITION 0x04\n"
-".emtcode VERTEX_ATTRIB_POSITION 0x01\n"
-".emtcode VERTEX_ATTRIB_WEIGHT 0x02\n"
-".emtcode VERTEX_ATTRIB_NORMAL 0x03\n"
-".emtcode VERTEX_ATTRIB_COLOR 0x04\n"
-".emtcode VERTEX_ATTRIB_FOGCOORD 0x05\n"
-".emtcode VERTEX_ATTRIB_TEXCOORD 0x06\n"
-".emtcode VERTEX_ATTRIB_MATRIXINDEX 0x07\n"
-".emtcode VERTEX_ATTRIB_GENERIC 0x08\n"
-".emtcode FRAGMENT_RESULT_COLOR 0x01\n"
-".emtcode FRAGMENT_RESULT_DEPTH 0x02\n"
-".emtcode VERTEX_RESULT_POSITION 0x01\n"
-".emtcode VERTEX_RESULT_COLOR 0x02\n"
-".emtcode VERTEX_RESULT_FOGCOORD 0x03\n"
-".emtcode VERTEX_RESULT_POINTSIZE 0x04\n"
-".emtcode VERTEX_RESULT_TEXCOORD 0x05\n"
-".emtcode TEXTARGET_1D 0x01\n"
-".emtcode TEXTARGET_2D 0x02\n"
-".emtcode TEXTARGET_3D 0x03\n"
-".emtcode TEXTARGET_RECT 0x04\n"
-".emtcode TEXTARGET_CUBE 0x05\n"
-".emtcode TEXTARGET_SHADOW1D 0x06\n"
-".emtcode TEXTARGET_SHADOW2D 0x07\n"
-".emtcode TEXTARGET_SHADOWRECT 0x08\n"
-".emtcode TEXTARGET_1D_ARRAY 0x09\n"
-".emtcode TEXTARGET_2D_ARRAY 0x0a\n"
-".emtcode TEXTARGET_SHADOW1D_ARRAY 0x0b\n"
-".emtcode TEXTARGET_SHADOW2D_ARRAY 0x0c\n"
-".emtcode FACE_FRONT 0x00\n"
-".emtcode FACE_BACK 0x01\n"
-".emtcode COLOR_PRIMARY 0x00\n"
-".emtcode COLOR_SECONDARY 0x01\n"
-".emtcode COMPONENT_X 0x00\n"
-".emtcode COMPONENT_Y 0x01\n"
-".emtcode COMPONENT_Z 0x02\n"
-".emtcode COMPONENT_W 0x03\n"
-".emtcode COMPONENT_0 0x04\n"
-".emtcode COMPONENT_1 0x05\n"
-".emtcode ARRAY_INDEX_ABSOLUTE 0x00\n"
-".emtcode ARRAY_INDEX_RELATIVE 0x01\n"
-".emtcode MATRIX_MODELVIEW 0x01\n"
-".emtcode MATRIX_PROJECTION 0x02\n"
-".emtcode MATRIX_MVP 0x03\n"
-".emtcode MATRIX_TEXTURE 0x04\n"
-".emtcode MATRIX_PALETTE 0x05\n"
-".emtcode MATRIX_PROGRAM 0x06\n"
-".emtcode MATRIX_MODIFIER_IDENTITY 0x00\n"
-".emtcode MATRIX_MODIFIER_INVERSE 0x01\n"
-".emtcode MATRIX_MODIFIER_TRANSPOSE 0x02\n"
-".emtcode MATRIX_MODIFIER_INVTRANS 0x03\n"
-".emtcode CONSTANT_SCALAR 0x01\n"
-".emtcode CONSTANT_VECTOR 0x02\n"
-".emtcode PROGRAM_PARAM_ENV 0x01\n"
-".emtcode PROGRAM_PARAM_LOCAL 0x02\n"
-".emtcode REGISTER_ATTRIB 0x01\n"
-".emtcode REGISTER_PARAM 0x02\n"
-".emtcode REGISTER_RESULT 0x03\n"
-".emtcode REGISTER_ESTABLISHED_NAME 0x04\n"
-".emtcode PARAM_NULL 0x00\n"
-".emtcode PARAM_ARRAY_ELEMENT 0x01\n"
-".emtcode PARAM_STATE_ELEMENT 0x02\n"
-".emtcode PARAM_PROGRAM_ELEMENT 0x03\n"
-".emtcode PARAM_PROGRAM_ELEMENTS 0x04\n"
-".emtcode PARAM_CONSTANT 0x05\n"
-".emtcode STATE_MATERIAL 0x01\n"
-".emtcode STATE_LIGHT 0x02\n"
-".emtcode STATE_LIGHT_MODEL 0x03\n"
-".emtcode STATE_LIGHT_PROD 0x04\n"
-".emtcode STATE_FOG 0x05\n"
-".emtcode STATE_MATRIX_ROWS 0x06\n"
-".emtcode STATE_TEX_ENV 0x07\n"
-".emtcode STATE_DEPTH 0x08\n"
-".emtcode STATE_TEX_GEN 0x09\n"
-".emtcode STATE_CLIP_PLANE 0x0A\n"
-".emtcode STATE_POINT 0x0B\n"
-".emtcode MATERIAL_AMBIENT 0x01\n"
-".emtcode MATERIAL_DIFFUSE 0x02\n"
-".emtcode MATERIAL_SPECULAR 0x03\n"
-".emtcode MATERIAL_EMISSION 0x04\n"
-".emtcode MATERIAL_SHININESS 0x05\n"
-".emtcode LIGHT_AMBIENT 0x01\n"
-".emtcode LIGHT_DIFFUSE 0x02\n"
-".emtcode LIGHT_SPECULAR 0x03\n"
-".emtcode LIGHT_POSITION 0x04\n"
-".emtcode LIGHT_ATTENUATION 0x05\n"
-".emtcode LIGHT_HALF 0x06\n"
-".emtcode LIGHT_SPOT_DIRECTION 0x07\n"
-".emtcode LIGHT_MODEL_AMBIENT 0x01\n"
-".emtcode LIGHT_MODEL_SCENECOLOR 0x02\n"
-".emtcode LIGHT_PROD_AMBIENT 0x01\n"
-".emtcode LIGHT_PROD_DIFFUSE 0x02\n"
-".emtcode LIGHT_PROD_SPECULAR 0x03\n"
-".emtcode TEX_ENV_COLOR 0x01\n"
-".emtcode TEX_GEN_EYE 0x01\n"
-".emtcode TEX_GEN_OBJECT 0x02\n"
-".emtcode FOG_COLOR 0x01\n"
-".emtcode FOG_PARAMS 0x02\n"
-".emtcode DEPTH_RANGE 0x01\n"
-".emtcode POINT_SIZE 0x01\n"
-".emtcode POINT_ATTENUATION 0x02\n"
-".emtcode ATTRIB 0x01\n"
-".emtcode PARAM 0x02\n"
-".emtcode TEMP 0x03\n"
-".emtcode OUTPUT 0x04\n"
-".emtcode ALIAS 0x05\n"
-".emtcode ADDRESS 0x06\n"
-".errtext UNKNOWN_PROGRAM_SIGNATURE \"1001: '$e_signature$': unknown program signature\"\n"
-".errtext MISSING_END_OR_INVALID_STATEMENT \"1002: '$e_statement$': invalid statement\"\n"
-".errtext CODE_AFTER_END \"1003: '$e_statement$': code after 'END' keyword\"\n"
-".errtext INVALID_PROGRAM_OPTION \"1004: '$e_identifier$': invalid program option\"\n"
-".errtext EXT_SWIZ_COMP_EXPECTED \"1005: extended swizzle component expected but '$e_token$' found\"\n"
-".errtext TEX_TARGET_EXPECTED \"1006: texture target expected but '$e_token$' found\"\n"
-".errtext TEXTURE_EXPECTED \"1007: 'texture' expected but '$e_identifier$' found\"\n"
-".errtext SOURCE_REGISTER_EXPECTED \"1008: source register expected but '$e_token$' found\"\n"
-".errtext DESTINATION_REGISTER_EXPECTED \"1009: destination register expected but '$e_token$' found\"\n"
-".errtext INVALID_ADDRESS_COMPONENT \"1010: '$e_identifier$': invalid address component\"\n"
-".errtext INVALID_ADDRESS_WRITEMASK \"1011: '$e_identifier$': invalid address writemask\"\n"
-".errtext INVALID_COMPONENT \"1012: '$e_charordigit$': invalid component\"\n"
-".errtext INVALID_SUFFIX \"1013: '$e_identifier$': invalid suffix\"\n"
-".errtext INVALID_WRITEMASK \"1014: '$e_identifier$': invalid writemask\"\n"
-".errtext FRAGMENT_EXPECTED \"1015: 'fragment' expected but '$e_identifier$' found\"\n"
-".errtext VERTEX_EXPECTED \"1016: 'vertex' expected but '$e_identifier$' found\"\n"
-".errtext INVALID_FRAGMENT_PROPERTY \"1017: '$e_identifier$': invalid fragment property\"\n"
-".errtext INVALID_VERTEX_PROPERTY \"1018: '$e_identifier$': invalid vertex property\"\n"
-".errtext INVALID_STATE_PROPERTY \"1019: '$e_identifier$': invalid state property\"\n"
-".errtext INVALID_MATERIAL_PROPERTY \"1020: '$e_identifier$': invalid material property\"\n"
-".errtext INVALID_LIGHT_PROPERTY \"1021: '$e_identifier$': invalid light property\"\n"
-".errtext INVALID_SPOT_PROPERTY \"1022: '$e_identifier$': invalid spot property\"\n"
-".errtext INVALID_LIGHTMODEL_PROPERTY \"1023: '$e_identifier$': invalid light model property\"\n"
-".errtext INVALID_LIGHTPROD_PROPERTY \"1024: '$e_identifier$': invalid light product property\"\n"
-".errtext INVALID_TEXENV_PROPERTY \"1025: '$e_identifier$': invalid texture environment property\"\n"
-".errtext INVALID_TEXGEN_PROPERTY \"1026: '$e_identifier$': invalid texture generating property\"\n"
-".errtext INVALID_TEXGEN_COORD \"1027: '$e_identifier$': invalid texture generating coord\"\n"
-".errtext INVALID_FOG_PROPERTY \"1028: '$e_identifier$': invalid fog property\"\n"
-".errtext INVALID_DEPTH_PROPERTY \"1029: '$e_identifier$': invalid depth property\"\n"
-".errtext INVALID_CLIPPLANE_PROPERTY \"1030: '$e_identifier$': invalid clip plane property\"\n"
-".errtext INVALID_POINT_PROPERTY \"1031: '$e_identifier$': invalid point property\"\n"
-".errtext MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED \"1032: matrix row selector or modifier expected but '$e_token$' found\"\n"
-".errtext INVALID_MATRIX_NAME \"1033: '$e_identifier$': invalid matrix name\"\n"
-".errtext INVALID_PROGRAM_PROPERTY \"1034: '$e_identifier$': invalid program property\"\n"
-".errtext RESULT_EXPECTED \"1035: 'result' expected but '$e_token$' found\"\n"
-".errtext INVALID_RESULT_PROPERTY \"1036: '$e_identifier$': invalid result property\"\n"
-".errtext INVALID_FACE_PROPERTY \"1037: '$e_identifier$': invalid face property\"\n"
-".errtext INVALID_COLOR_PROPERTY \"1038: '$e_identifier$': invalid color property\"\n"
-".errtext IDENTIFIER_EXPECTED \"1039: identifier expected but '$e_token$' found\"\n"
-".errtext RESERVED_KEYWORD \"1040: use of reserved keyword as an identifier\"\n"
-".errtext INTEGER_EXPECTED \"1041: integer value expected but '$e_token$' found\"\n"
-".errtext MISSING_SEMICOLON \"1042: ';' expected but '$e_token$' found\"\n"
-".errtext MISSING_COMMA \"1043: ',' expected but '$e_token$' found\"\n"
-".errtext MISSING_LBRACKET \"1044: '[' expected but '$e_token$' found\"\n"
-".errtext MISSING_RBRACKET \"1045: ']' expected but '$e_token$' found\"\n"
-".errtext MISSING_DOT \"1046: '.' expected but '$e_token$' found\"\n"
-".errtext MISSING_EQUAL \"1047: '=' expected but '$e_token$' found\"\n"
-".errtext MISSING_LBRACE \"1048: '{' expected but '$e_token$' found\"\n"
-".errtext MISSING_RBRACE \"1049: '}' expected but '$e_token$' found\"\n"
-".errtext MISSING_DOTDOT \"1050: '..' expected but '$e_token$' found\"\n"
-".errtext MISSING_FRACTION_OR_EXPONENT \"1051: missing fraction part or exponent\"\n"
-".errtext MISSING_DOT_OR_EXPONENT \"1052: missing '.' or exponent\"\n"
-".errtext EXPONENT_VALUE_EXPECTED \"1053: exponent value expected\"\n"
-".errtext INTEGER_OUT_OF_RANGE \"1054: integer value out of range\"\n"
-".errtext OPERATION_NEEDS_DESTINATION_VARIABLE \"1055: operation needs destination variable\"\n"
-".errtext OPERATION_NEEDS_SOURCE_VARIABLE \"1056: operation needs source variable\"\n"
-".errtext ADDRESS_REGISTER_EXPECTED \"1057: address register expected but '$e_token$' found\"\n"
-".errtext ADDRESS_REGISTER_OR_INTEGER_EXPECTED \"1058: address register or integer literal expected but '$e_token$' found\"\n"
-".regbyte vertex_blend 0x00\n"
-".regbyte matrix_palette 0x00\n"
-".regbyte point_parameters 0x00\n"
-".regbyte secondary_color 0x00\n"
-".regbyte fog_coord 0x00\n"
-".regbyte texture_rectangle 0x00\n"
-".regbyte fragment_program_shadow 0x00\n"
-".regbyte draw_buffers 0x00\n"
-".regbyte texture_array 0x00\n"
-".regbyte ARB_precision_hint_fastest 0x00\n"
-".regbyte ARB_precision_hint_nicest 0x00\n"
-".regbyte ARB_fog_exp 0x00\n"
-".regbyte ARB_fog_exp2 0x00\n"
-".regbyte ARB_fog_linear 0x00\n"
-".regbyte ARB_position_invariant 0x00\n"
-".regbyte ARB_fragment_program_shadow 0x00\n"
-".regbyte ARB_draw_buffers 0x00\n"
-".regbyte MESA_texture_array 0x00\n"
-".regbyte program_target 0x00\n"
-"program\n"
-" programs .error UNKNOWN_PROGRAM_SIGNATURE .emit REVISION;\n"
-"programs\n"
-" .if (program_target == 0x10) frag_program_1_0 .emit FRAGMENT_PROGRAM .emit 0x01 .emit 0x00 .or\n"
-" .if (program_target == 0x20) vert_program_1_0 .emit VERTEX_PROGRAM .emit 0x01 .emit 0x00;\n"
-"frag_program_1_0\n"
-" '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'f' .and 'p' .and '1' .and '.' .and '0' .and\n"
-" optional_space .and fp_optionSequence .and fp_statementSequence .and\n"
-" \"END\" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and\n"
-" '\\0' .error CODE_AFTER_END;\n"
-"vert_program_1_0\n"
-" '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'v' .and 'p' .and '1' .and '.' .and '0' .and\n"
-" optional_space .and vp_optionSequence .and vp_statementSequence .and\n"
-" \"END\" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and\n"
-" '\\0' .error CODE_AFTER_END;\n"
-"fp_optionSequence\n"
-" .loop fp_option;\n"
-"vp_optionSequence\n"
-" .loop vp_option;\n"
-"fp_option\n"
-" \"OPTION\" .emit OPTION .and space .error IDENTIFIER_EXPECTED .and\n"
-" fp_optionString .error INVALID_PROGRAM_OPTION .and semicolon;\n"
-"vp_option\n"
-" \"OPTION\" .emit OPTION .and space .error IDENTIFIER_EXPECTED .and\n"
-" vp_optionString .error INVALID_PROGRAM_OPTION .and semicolon;\n"
-"fp_optionString\n"
-" .if (ARB_precision_hint_nicest == 0x00) \"ARB_precision_hint_fastest\"\n"
-" .emit ARB_PRECISION_HINT_FASTEST .load ARB_precision_hint_fastest 0x01 .or\n"
-" .if (ARB_precision_hint_fastest == 0x00) \"ARB_precision_hint_nicest\"\n"
-" .emit ARB_PRECISION_HINT_NICEST .load ARB_precision_hint_nicest 0x01 .or\n"
-" fp_ARB_fog_exp .emit ARB_FOG_EXP .load ARB_fog_exp 0x01 .or\n"
-" fp_ARB_fog_exp2 .emit ARB_FOG_EXP2 .load ARB_fog_exp2 0x01 .or\n"
-" fp_ARB_fog_linear .emit ARB_FOG_LINEAR .load ARB_fog_linear 0x01 .or\n"
-" .if (fragment_program_shadow != 0x00) \"ARB_fragment_program_shadow\"\n"
-" .emit ARB_FRAGMENT_PROGRAM_SHADOW .load ARB_fragment_program_shadow 0x01 .or\n"
-" .if (draw_buffers != 0x00) \"ARB_draw_buffers\" .emit ARB_DRAW_BUFFERS\n"
-" .load ARB_draw_buffers 0x01 .or\n"
-" .if (texture_array != 0x00) \"MESA_texture_array\" .emit MESA_TEXTURE_ARRAY\n"
-" .load MESA_texture_array 0x01;\n"
-"vp_optionString\n"
-" \"ARB_position_invariant\" .emit ARB_POSITION_INVARIANT .load ARB_position_invariant 0x01;\n"
-"fp_ARB_fog_exp\n"
-" .if (ARB_fog_exp2 == 0x00) .true .and .if (ARB_fog_linear == 0x00) \"ARB_fog_exp\";\n"
-"fp_ARB_fog_exp2\n"
-" .if (ARB_fog_exp == 0x00) .true .and .if (ARB_fog_linear == 0x00) \"ARB_fog_exp2\";\n"
-"fp_ARB_fog_linear\n"
-" .if (ARB_fog_exp == 0x00) .true .and .if (ARB_fog_exp2 == 0x00) \"ARB_fog_linear\";\n"
-"fp_statementSequence\n"
-" .loop fp_statement;\n"
-"vp_statementSequence\n"
-" .loop vp_statement;\n"
-"fp_statement\n"
-" fp_statement_1 .or fp_statement_2;\n"
-"vp_statement\n"
-" vp_statement_1 .or vp_statement_2;\n"
-"fp_statement_1\n"
-" fp_instruction .emit INSTRUCTION .emit $ .and semicolon;\n"
-"fp_statement_2\n"
-" fp_namingStatement .emit DECLARATION .and semicolon;\n"
-"vp_statement_1\n"
-" vp_instruction .emit INSTRUCTION .emit $ .and semicolon;\n"
-"vp_statement_2\n"
-" vp_namingStatement .emit DECLARATION .and semicolon;\n"
-"fp_instruction\n"
-" ALUInstruction .emit OP_ALU_INST .or\n"
-" TexInstruction .emit OP_TEX_INST;\n"
-"vp_instruction\n"
-" ARL_instruction .emit OP_ALU_ARL .or\n"
-" vp_VECTORop_instruction .emit OP_ALU_VECTOR .or\n"
-" vp_SCALARop_instruction .emit OP_ALU_SCALAR .or\n"
-" vp_BINSCop_instruction .emit OP_ALU_BINSC .or\n"
-" vp_BINop_instruction .emit OP_ALU_BIN .or\n"
-" vp_TRIop_instruction .emit OP_ALU_TRI .or\n"
-" vp_SWZ_instruction .emit OP_ALU_SWZ;\n"
-"ALUInstruction\n"
-" fp_VECTORop_instruction .emit OP_ALU_VECTOR .or\n"
-" fp_SCALARop_instruction .emit OP_ALU_SCALAR .or\n"
-" fp_BINSCop_instruction .emit OP_ALU_BINSC .or\n"
-" fp_BINop_instruction .emit OP_ALU_BIN .or\n"
-" fp_TRIop_instruction .emit OP_ALU_TRI .or\n"
-" fp_SWZ_instruction .emit OP_ALU_SWZ;\n"
-"TexInstruction\n"
-" SAMPLE_instruction .emit OP_TEX_SAMPLE .or\n"
-" KIL_instruction .emit OP_TEX_KIL;\n"
-"ARL_instruction\n"
-" \"ARL\" .emit OP_ARL .and space_dst .and maskedAddrReg .and comma .and vp_scalarSrcReg;\n"
-"fp_VECTORop_instruction\n"
-" fp_VECTORop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg;\n"
-"vp_VECTORop_instruction\n"
-" vp_VECTORop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg;\n"
-"fp_VECTORop\n"
-" \"ABS\" .emit OP_ABS .or \"ABS_SAT\" .emit OP_ABS_SAT .or\n"
-" \"FLR\" .emit OP_FLR .or \"FLR_SAT\" .emit OP_FLR_SAT .or\n"
-" \"FRC\" .emit OP_FRC .or \"FRC_SAT\" .emit OP_FRC_SAT .or\n"
-" \"LIT\" .emit OP_LIT .or \"LIT_SAT\" .emit OP_LIT_SAT .or\n"
-" \"MOV\" .emit OP_MOV .or \"MOV_SAT\" .emit OP_MOV_SAT;\n"
-"vp_VECTORop\n"
-" \"ABS\" .emit OP_ABS .or\n"
-" \"FLR\" .emit OP_FLR .or\n"
-" \"FRC\" .emit OP_FRC .or\n"
-" \"LIT\" .emit OP_LIT .or\n"
-" \"MOV\" .emit OP_MOV;\n"
-"fp_SCALARop_instruction\n"
-" fp_SCALARop .and space_dst .and fp_maskedDstReg .and comma .and fp_scalarSrcReg;\n"
-"vp_SCALARop_instruction\n"
-" vp_SCALARop .and space_dst .and vp_maskedDstReg .and comma .and vp_scalarSrcReg;\n"
-"fp_SCALARop\n"
-" \"COS\" .emit OP_COS .or \"COS_SAT\" .emit OP_COS_SAT .or\n"
-" \"EX2\" .emit OP_EX2 .or \"EX2_SAT\" .emit OP_EX2_SAT .or\n"
-" \"LG2\" .emit OP_LG2 .or \"LG2_SAT\" .emit OP_LG2_SAT .or\n"
-" \"RCP\" .emit OP_RCP .or \"RCP_SAT\" .emit OP_RCP_SAT .or\n"
-" \"RSQ\" .emit OP_RSQ .or \"RSQ_SAT\" .emit OP_RSQ_SAT .or\n"
-" \"SIN\" .emit OP_SIN .or \"SIN_SAT\" .emit OP_SIN_SAT .or\n"
-" \"SCS\" .emit OP_SCS .or \"SCS_SAT\" .emit OP_SCS_SAT;\n"
-"vp_SCALARop\n"
-" \"EX2\" .emit OP_EX2 .or\n"
-" \"EXP\" .emit OP_EXP .or\n"
-" \"LG2\" .emit OP_LG2 .or\n"
-" \"LOG\" .emit OP_LOG .or\n"
-" \"RCP\" .emit OP_RCP .or\n"
-" \"RSQ\" .emit OP_RSQ;\n"
-"fp_BINSCop_instruction\n"
-" fp_BINSCop .and space_dst .and fp_maskedDstReg .and comma .and fp_scalarSrcReg .and comma .and\n"
-" fp_scalarSrcReg;\n"
-"vp_BINSCop_instruction\n"
-" vp_BINSCop .and space_dst .and vp_maskedDstReg .and comma .and vp_scalarSrcReg .and comma .and\n"
-" vp_scalarSrcReg;\n"
-"fp_BINSCop\n"
-" \"POW\" .emit OP_POW .or \"POW_SAT\" .emit OP_POW_SAT;\n"
-"vp_BINSCop\n"
-" \"POW\" .emit OP_POW;\n"
-"fp_BINop_instruction\n"
-" fp_BINop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and\n"
-" vectorSrcReg;\n"
-"vp_BINop_instruction\n"
-" vp_BINop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg .and comma .and\n"
-" swizzleSrcReg;\n"
-"fp_BINop\n"
-" \"ADD\" .emit OP_ADD .or \"ADD_SAT\" .emit OP_ADD_SAT .or\n"
-" \"DP3\" .emit OP_DP3 .or \"DP3_SAT\" .emit OP_DP3_SAT .or\n"
-" \"DP4\" .emit OP_DP4 .or \"DP4_SAT\" .emit OP_DP4_SAT .or\n"
-" \"DPH\" .emit OP_DPH .or \"DPH_SAT\" .emit OP_DPH_SAT .or\n"
-" \"DST\" .emit OP_DST .or \"DST_SAT\" .emit OP_DST_SAT .or\n"
-" \"MAX\" .emit OP_MAX .or \"MAX_SAT\" .emit OP_MAX_SAT .or\n"
-" \"MIN\" .emit OP_MIN .or \"MIN_SAT\" .emit OP_MIN_SAT .or\n"
-" \"MUL\" .emit OP_MUL .or \"MUL_SAT\" .emit OP_MUL_SAT .or\n"
-" \"SGE\" .emit OP_SGE .or \"SGE_SAT\" .emit OP_SGE_SAT .or\n"
-" \"SLT\" .emit OP_SLT .or \"SLT_SAT\" .emit OP_SLT_SAT .or\n"
-" \"SUB\" .emit OP_SUB .or \"SUB_SAT\" .emit OP_SUB_SAT .or\n"
-" \"XPD\" .emit OP_XPD .or \"XPD_SAT\" .emit OP_XPD_SAT;\n"
-"vp_BINop\n"
-" \"ADD\" .emit OP_ADD .or\n"
-" \"DP3\" .emit OP_DP3 .or\n"
-" \"DP4\" .emit OP_DP4 .or\n"
-" \"DPH\" .emit OP_DPH .or\n"
-" \"DST\" .emit OP_DST .or\n"
-" \"MAX\" .emit OP_MAX .or\n"
-" \"MIN\" .emit OP_MIN .or\n"
-" \"MUL\" .emit OP_MUL .or\n"
-" \"SGE\" .emit OP_SGE .or\n"
-" \"SLT\" .emit OP_SLT .or\n"
-" \"SUB\" .emit OP_SUB .or\n"
-" \"XPD\" .emit OP_XPD;\n"
-"fp_TRIop_instruction\n"
-" fp_TRIop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and\n"
-" vectorSrcReg .and comma .and vectorSrcReg;\n"
-"vp_TRIop_instruction\n"
-" vp_TRIop .and space_dst .and vp_maskedDstReg .and comma .and swizzleSrcReg .and comma .and\n"
-" swizzleSrcReg .and comma .and swizzleSrcReg;\n"
-"fp_TRIop\n"
-" \"CMP\" .emit OP_CMP .or \"CMP_SAT\" .emit OP_CMP_SAT .or\n"
-" \"LRP\" .emit OP_LRP .or \"LRP_SAT\" .emit OP_LRP_SAT .or\n"
-" \"MAD\" .emit OP_MAD .or \"MAD_SAT\" .emit OP_MAD_SAT;\n"
-"vp_TRIop\n"
-" \"MAD\" .emit OP_MAD;\n"
-"fp_SWZ_instruction\n"
-" SWZop .and space_dst .and fp_maskedDstReg .and comma .and fp_srcReg .and comma .and\n"
-" fp_extendedSwizzle .error EXT_SWIZ_COMP_EXPECTED;\n"
-"vp_SWZ_instruction\n"
-" \"SWZ\" .emit OP_SWZ .and space_dst .and vp_maskedDstReg .and comma .and vp_srcReg .and comma .and\n"
-" vp_extendedSwizzle .error EXT_SWIZ_COMP_EXPECTED;\n"
-"SWZop\n"
-" \"SWZ\" .emit OP_SWZ .or \"SWZ_SAT\" .emit OP_SWZ_SAT;\n"
-"SAMPLE_instruction\n"
-" SAMPLEop .and space_dst .and fp_maskedDstReg .and comma .and vectorSrcReg .and comma .and\n"
-" texImageUnit .and comma .and texTarget .error TEX_TARGET_EXPECTED;\n"
-"SAMPLEop\n"
-" \"TEX\" .emit OP_TEX .or \"TEX_SAT\" .emit OP_TEX_SAT .or\n"
-" \"TXB\" .emit OP_TXB .or \"TXB_SAT\" .emit OP_TXB_SAT .or\n"
-" \"TXP\" .emit OP_TXP .or \"TXP_SAT\" .emit OP_TXP_SAT;\n"
-"KIL_instruction\n"
-" \"KIL\" .emit OP_KIL .and space_src .and vectorSrcReg;\n"
-"texImageUnit\n"
-" \"texture\" .error TEXTURE_EXPECTED .and optTexImageUnitNum;\n"
-"texTarget\n"
-" \"1D\" .emit TEXTARGET_1D .or\n"
-" \"2D\" .emit TEXTARGET_2D .or\n"
-" \"3D\" .emit TEXTARGET_3D .or\n"
-" .if (texture_rectangle != 0x00) \"RECT\" .emit TEXTARGET_RECT .or\n"
-" \"CUBE\" .emit TEXTARGET_CUBE .or\n"
-" .if (ARB_fragment_program_shadow != 0x00) shadowTarget .or\n"
-" .if (MESA_texture_array != 0x00) arrayTarget;\n"
-"shadowTarget\n"
-" \"SHADOW1D\" .emit TEXTARGET_SHADOW1D .or\n"
-" \"SHADOW2D\" .emit TEXTARGET_SHADOW2D .or\n"
-" .if (texture_rectangle != 0x00) \"SHADOWRECT\" .emit TEXTARGET_SHADOWRECT .or\n"
-" .if (MESA_texture_array != 0x00) shadowArrayTarget;\n"
-"arrayTarget\n"
-" \"ARRAY1D\" .emit TEXTARGET_1D_ARRAY .or\n"
-" \"ARRAY2D\" .emit TEXTARGET_2D_ARRAY;\n"
-"shadowArrayTarget\n"
-" \"SHADOWARRAY1D\" .emit TEXTARGET_SHADOW1D_ARRAY .or\n"
-" \"SHADOWARRAY2D\" .emit TEXTARGET_SHADOW2D_ARRAY;\n"
-"optTexImageUnitNum\n"
-" optTexImageUnitNum_1 .or .true .emit 0x00;\n"
-"optTexImageUnitNum_1\n"
-" lbracket_ne .and texImageUnitNum .and rbracket;\n"
-"texImageUnitNum\n"
-" integer;\n"
-"fp_scalarSrcReg\n"
-" optionalSign .and fp_srcReg .and fp_scalarSuffix;\n"
-"vp_scalarSrcReg\n"
-" optionalSign .and vp_srcReg .and vp_scalarSuffix;\n"
-"swizzleSrcReg\n"
-" optionalSign .and vp_srcReg .and swizzleSuffix;\n"
-"vectorSrcReg\n"
-" optionalSign .and fp_srcReg .and optionalSuffix;\n"
-"fp_maskedDstReg\n"
-" fp_dstReg .and fp_optionalMask;\n"
-"vp_maskedDstReg\n"
-" vp_dstReg .and vp_optionalMask;\n"
-"maskedAddrReg\n"
-" addrReg .error ADDRESS_REGISTER_EXPECTED .and addrWriteMask;\n"
-"fp_extendedSwizzle\n"
-" rgbaExtendedSwizzle .or xyzwExtendedSwizzle;\n"
-"vp_extendedSwizzle\n"
-" extSwizComp .and comma .and\n"
-" extSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" extSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" extSwizComp .error EXT_SWIZ_COMP_EXPECTED;\n"
-"xyzwExtendedSwizzle\n"
-" xyzwExtSwizComp .and comma .and\n"
-" xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" xyzwExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;\n"
-"rgbaExtendedSwizzle\n"
-" rgbaExtendedSwizzle_1 .or rgbaExtendedSwizzle_2 .or rgbaExtendedSwizzle_3 .or\n"
-" rgbaExtendedSwizzle_4;\n"
-"rgbaExtendedSwizzle_1\n"
-" rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_digit .and comma .and\n"
-" rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp;\n"
-"rgbaExtendedSwizzle_2\n"
-" rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_digit .and comma .and\n"
-" rgbaExtSwizComp_alpha .and comma .and rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;\n"
-"rgbaExtendedSwizzle_3\n"
-" rgbaExtSwizComp_digit .and comma .and rgbaExtSwizComp_alpha .and comma .and\n"
-" rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;\n"
-"rgbaExtendedSwizzle_4\n"
-" rgbaExtSwizComp_alpha .and comma .and \n"
-"rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED .and comma .and\n"
-" rgbaExtSwizComp .error EXT_SWIZ_COMP_EXPECTED;\n"
-"xyzwExtSwizComp\n"
-" optionalSign .and xyzwExtSwizSel;\n"
-"rgbaExtSwizComp\n"
-" optionalSign .and rgbaExtSwizSel;\n"
-"rgbaExtSwizComp_digit\n"
-" optionalSign .and rgbaExtSwizSel_digit;\n"
-"rgbaExtSwizComp_alpha\n"
-" optionalSign .and rgbaExtSwizSel_alpha;\n"
-"extSwizComp\n"
-" optionalSign .and extSwizSel;\n"
-"xyzwExtSwizSel\n"
-" \"0\" .emit COMPONENT_0 .or \"1\" .emit COMPONENT_1 .or xyzwComponent_single;\n"
-"rgbaExtSwizSel\n"
-" rgbaExtSwizSel_digit .or rgbaExtSwizSel_alpha;\n"
-"rgbaExtSwizSel_digit\n"
-" \"0\" .emit COMPONENT_0 .or \"1\" .emit COMPONENT_1;\n"
-"rgbaExtSwizSel_alpha\n"
-" rgbaComponent_single;\n"
-"extSwizSel\n"
-" \"0\" .emit COMPONENT_0 .or \"1\" .emit COMPONENT_1 .or vp_component_single;\n"
-"fp_srcReg\n"
-" fp_srcReg_1 .error SOURCE_REGISTER_EXPECTED;\n"
-"vp_srcReg\n"
-" vp_srcReg_1 .error SOURCE_REGISTER_EXPECTED;\n"
-"fp_srcReg_1\n"
-" fragmentAttribReg .emit REGISTER_ATTRIB .or\n"
-" fp_progParamReg .emit REGISTER_PARAM .or\n"
-" fp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;\n"
-"vp_srcReg_1\n"
-" vertexAttribReg .emit REGISTER_ATTRIB .or\n"
-" vp_progParamReg .emit REGISTER_PARAM .or\n"
-" vp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;\n"
-"fp_dstReg\n"
-" fp_dstReg_1 .error DESTINATION_REGISTER_EXPECTED;\n"
-"vp_dstReg\n"
-" vp_dstReg_1 .error DESTINATION_REGISTER_EXPECTED;\n"
-"fp_dstReg_1\n"
-" fragmentResultReg .emit REGISTER_RESULT .or\n"
-" fp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;\n"
-"vp_dstReg_1\n"
-" vertexResultReg .emit REGISTER_RESULT .or\n"
-" vp_temporaryReg .emit REGISTER_ESTABLISHED_NAME;\n"
-"fragmentAttribReg\n"
-" fragAttribBinding;\n"
-"vertexAttribReg\n"
-" vtxAttribBinding;\n"
-"fp_temporaryReg\n"
-" fp_establishedName_no_error_on_identifier;\n"
-"vp_temporaryReg\n"
-" vp_establishedName_no_error_on_identifier;\n"
-"fp_progParamReg\n"
-" fp_paramSingleItemUse .or fp_progParamReg_1 .or fp_progParamSingle;\n"
-"vp_progParamReg\n"
-" vp_paramSingleItemUse .or vp_progParamReg_1 .or vp_progParamSingle;\n"
-"fp_progParamReg_1\n"
-" fp_progParamArray .emit PARAM_ARRAY_ELEMENT .and lbracket_ne .and progParamArrayAbs .and\n"
-" rbracket;\n"
-"vp_progParamReg_1\n"
-" vp_progParamArray .emit PARAM_ARRAY_ELEMENT .and lbracket_ne .and progParamArrayMem .and\n"
-" rbracket;\n"
-"fp_progParamSingle\n"
-" .false;\n"
-"vp_progParamSingle\n"
-" .false;\n"
-"fp_progParamArray\n"
-" fp_establishedName_no_error_on_identifier;\n"
-"vp_progParamArray\n"
-" vp_establishedName_no_error_on_identifier;\n"
-"progParamArrayMem\n"
-" progParamArrayAbs .or progParamArrayRel;\n"
-"progParamArrayAbs\n"
-" integer_ne .emit ARRAY_INDEX_ABSOLUTE;\n"
-"progParamArrayRel\n"
-" addrReg .error ADDRESS_REGISTER_OR_INTEGER_EXPECTED .emit ARRAY_INDEX_RELATIVE .and\n"
-" addrComponent .and addrRegRelOffset;\n"
-"addrRegRelOffset\n"
-" addrRegRelOffset_1 .or addrRegRelOffset_2 .or .true .emit 0x00;\n"
-"addrRegRelOffset_1\n"
-" plus_ne .and addrRegPosOffset;\n"
-"addrRegRelOffset_2\n"
-" minus_ne .and addrRegNegOffset;\n"
-"addrRegPosOffset\n"
-" integer_0_63;\n"
-"addrRegNegOffset\n"
-" integer_0_64;\n"
-"fragmentResultReg\n"
-" fp_resultBinding;\n"
-"vertexResultReg\n"
-" vp_resultBinding;\n"
-"addrReg\n"
-" vp_establishedName_no_error_on_identifier;\n"
-"addrComponent\n"
-" dot .and \"x\" .error INVALID_ADDRESS_COMPONENT .emit COMPONENT_X .emit COMPONENT_X\n"
-" .emit COMPONENT_X .emit COMPONENT_X;\n"
-"addrWriteMask\n"
-" dot .and \"x\" .error INVALID_ADDRESS_WRITEMASK .emit 0x08;\n"
-"fp_scalarSuffix\n"
-" dot .and fp_component_single .error INVALID_COMPONENT;\n"
-"vp_scalarSuffix\n"
-" dot .and vp_component_single .error INVALID_COMPONENT;\n"
-"swizzleSuffix\n"
-" swizzleSuffix_1 .or\n"
-" .true .emit COMPONENT_X .emit COMPONENT_Y .emit COMPONENT_Z .emit COMPONENT_W;\n"
-"swizzleSuffix_1\n"
-" dot_ne .and swizzleSuffix_2 .error INVALID_SUFFIX;\n"
-"swizzleSuffix_2\n"
-" swizzleSuffix_3 .or swizzleSuffix_4;\n"
-"swizzleSuffix_3\n"
-" vp_component_multi .and vp_component_multi .and vp_component_multi .error INVALID_COMPONENT .and\n"
-" vp_component_multi .error INVALID_COMPONENT;\n"
-"swizzleSuffix_4\n"
-" \"x\" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or\n"
-" \"y\" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or\n"
-" \"z\" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or\n"
-" \"w\" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W;\n"
-"optionalSuffix\n"
-" optionalSuffix_1 .or\n"
-" .true .emit COMPONENT_X .emit COMPONENT_Y .emit COMPONENT_Z .emit COMPONENT_W;\n"
-"optionalSuffix_1\n"
-" dot_ne .and optionalSuffix_2 .error INVALID_SUFFIX;\n"
-"optionalSuffix_2\n"
-" optionalSuffix_3 .or optionalSuffix_4 .or optionalSuffix_5;\n"
-"optionalSuffix_3\n"
-" xyzwComponent_multi .and xyzwComponent_multi .and\n"
-" xyzwComponent_multi .error INVALID_COMPONENT .and xyzwComponent_multi .error INVALID_COMPONENT;\n"
-"optionalSuffix_4\n"
-" rgbaComponent_multi .and rgbaComponent_multi .and\n"
-" rgbaComponent_multi .error INVALID_COMPONENT .and rgbaComponent_multi .error INVALID_COMPONENT;\n"
-"optionalSuffix_5\n"
-" \"x\" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or\n"
-" \"y\" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or\n"
-" \"z\" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or\n"
-" \"w\" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .or\n"
-" \"r\" .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .emit COMPONENT_X .or\n"
-" \"g\" .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .emit COMPONENT_Y .or\n"
-" \"b\" .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .emit COMPONENT_Z .or\n"
-" \"a\" .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W .emit COMPONENT_W;\n"
-"fp_component_single\n"
-" xyzwComponent_single .or rgbaComponent_single;\n"
-"vp_component_multi\n"
-" 'x' .emit COMPONENT_X .or 'y' .emit COMPONENT_Y .or 'z' .emit COMPONENT_Z .or\n"
-" 'w' .emit COMPONENT_W;\n"
-"vp_component_single\n"
-" \"x\" .emit COMPONENT_X .or \"y\" .emit COMPONENT_Y .or \"z\" .emit COMPONENT_Z .or\n"
-" \"w\" .emit COMPONENT_W;\n"
-"xyzwComponent_multi\n"
-" 'x' .emit COMPONENT_X .or 'y' .emit COMPONENT_Y .or 'z' .emit COMPONENT_Z .or\n"
-" 'w' .emit COMPONENT_W;\n"
-"xyzwComponent_single\n"
-" \"x\" .emit COMPONENT_X .or \"y\" .emit COMPONENT_Y .or \"z\" .emit COMPONENT_Z .or\n"
-" \"w\" .emit COMPONENT_W;\n"
-"rgbaComponent_multi\n"
-" 'r' .emit COMPONENT_X .or 'g' .emit COMPONENT_Y .or 'b' .emit COMPONENT_Z .or\n"
-" 'a' .emit COMPONENT_W;\n"
-"rgbaComponent_single\n"
-" \"r\" .emit COMPONENT_X .or \"g\" .emit COMPONENT_Y .or \"b\" .emit COMPONENT_Z .or\n"
-" \"a\" .emit COMPONENT_W;\n"
-"fp_optionalMask\n"
-" rgbaMask .or xyzwMask .or .true .emit 0x0F;\n"
-"vp_optionalMask\n"
-" xyzwMask .or .true .emit 0x0F;\n"
-"xyzwMask\n"
-" dot_ne .and xyzwMask_1 .error INVALID_WRITEMASK;\n"
-"xyzwMask_1\n"
-" \"xyzw\" .emit 0x0F .or \"xyz\" .emit 0x0E .or \"xyw\" .emit 0x0D .or \"xy\" .emit 0x0C .or\n"
-" \"xzw\" .emit 0x0B .or \"xz\" .emit 0x0A .or \"xw\" .emit 0x09 .or \"x\" .emit 0x08 .or\n"
-" \"yzw\" .emit 0x07 .or \"yz\" .emit 0x06 .or \"yw\" .emit 0x05 .or \"y\" .emit 0x04 .or\n"
-" \"zw\" .emit 0x03 .or \"z\" .emit 0x02 .or \"w\" .emit 0x01;\n"
-"rgbaMask\n"
-" dot_ne .and rgbaMask_1;\n"
-"rgbaMask_1\n"
-" \"rgba\" .emit 0x0F .or \"rgb\" .emit 0x0E .or \"rga\" .emit 0x0D .or \"rg\" .emit 0x0C .or\n"
-" \"rba\" .emit 0x0B .or \"rb\" .emit 0x0A .or \"ra\" .emit 0x09 .or \"r\" .emit 0x08 .or\n"
-" \"gba\" .emit 0x07 .or \"gb\" .emit 0x06 .or \"ga\" .emit 0x05 .or \"g\" .emit 0x04 .or\n"
-" \"ba\" .emit 0x03 .or \"b\" .emit 0x02 .or \"a\" .emit 0x01;\n"
-"fp_namingStatement\n"
-" fp_ATTRIB_statement .emit ATTRIB .or\n"
-" fp_PARAM_statement .emit PARAM .or\n"
-" fp_TEMP_statement .emit TEMP .or\n"
-" fp_OUTPUT_statement .emit OUTPUT .or\n"
-" fp_ALIAS_statement .emit ALIAS;\n"
-"vp_namingStatement\n"
-" vp_ATTRIB_statement .emit ATTRIB .or\n"
-" vp_PARAM_statement .emit PARAM .or\n"
-" vp_TEMP_statement .emit TEMP .or\n"
-" ADDRESS_statement .emit ADDRESS .or\n"
-" vp_OUTPUT_statement .emit OUTPUT .or\n"
-" vp_ALIAS_statement .emit ALIAS;\n"
-"fp_ATTRIB_statement\n"
-" \"ATTRIB\" .and space .and fp_establishName .and equal .and\n"
-" fragAttribBinding .error FRAGMENT_EXPECTED;\n"
-"vp_ATTRIB_statement\n"
-" \"ATTRIB\" .and space .and vp_establishName .and equal .and\n"
-" vtxAttribBinding .error VERTEX_EXPECTED;\n"
-"fragAttribBinding\n"
-" \"fragment\" .and dot .and fragAttribItem .error INVALID_FRAGMENT_PROPERTY;\n"
-"vtxAttribBinding\n"
-" \"vertex\" .and dot .and vtxAttribItem .error INVALID_VERTEX_PROPERTY;\n"
-"fragAttribItem\n"
-" fragAttribItem_1 .emit FRAGMENT_ATTRIB_COLOR .or\n"
-" fragAttribItem_2 .emit FRAGMENT_ATTRIB_TEXCOORD .or\n"
-" .if (fog_coord != 0x00) \"fogcoord\" .emit FRAGMENT_ATTRIB_FOGCOORD .or\n"
-" \"position\" .emit FRAGMENT_ATTRIB_POSITION;\n"
-"fragAttribItem_1\n"
-" \"color\" .and optColorType;\n"
-"fragAttribItem_2\n"
-" \"texcoord\" .and optTexCoordNum;\n"
-"vtxAttribItem\n"
-" \"position\" .emit VERTEX_ATTRIB_POSITION .or\n"
-" .if (vertex_blend != 0x00) vtxAttribItem_1 .emit VERTEX_ATTRIB_WEIGHT .or\n"
-" \"normal\" .emit VERTEX_ATTRIB_NORMAL .or\n"
-" vtxAttribItem_2 .emit VERTEX_ATTRIB_COLOR .or\n"
-" \"fogcoord\" .emit VERTEX_ATTRIB_FOGCOORD .or\n"
-" vtxAttribItem_3 .emit VERTEX_ATTRIB_TEXCOORD .or\n"
-" .if (matrix_palette != 0x00) vtxAttribItem_4 .emit VERTEX_ATTRIB_MATRIXINDEX .or\n"
-" vtxAttribItem_5 .emit VERTEX_ATTRIB_GENERIC;\n"
-"vtxAttribItem_1\n"
-" \"weight\" .and vtxOptWeightNum;\n"
-"vtxAttribItem_2\n"
-" \"color\" .and optColorType;\n"
-"vtxAttribItem_3\n"
-" \"texcoord\" .and optTexCoordNum;\n"
-"vtxAttribItem_4\n"
-" \"matrixindex\" .and lbracket .and vtxWeightNum .and rbracket;\n"
-"vtxAttribItem_5\n"
-" \"attrib\" .and lbracket .and vtxAttribNum .and rbracket;\n"
-"vtxAttribNum\n"
-" integer;\n"
-"vtxOptWeightNum\n"
-" vtxOptWeightNum_1 .or .true .emit 0x00;\n"
-"vtxOptWeightNum_1\n"
-" lbracket_ne .and vtxWeightNum .and rbracket;\n"
-"vtxWeightNum\n"
-" integer;\n"
-"fp_PARAM_statement\n"
-" fp_PARAM_multipleStmt .or fp_PARAM_singleStmt;\n"
-"vp_PARAM_statement\n"
-" vp_PARAM_multipleStmt .or vp_PARAM_singleStmt;\n"
-"fp_PARAM_singleStmt\n"
-" \"PARAM\" .and space .and fp_establishName .and .true .emit 0x00 .and fp_paramSingleInit .and\n"
-" .true .emit PARAM_NULL;\n"
-"vp_PARAM_singleStmt\n"
-" \"PARAM\" .and space .and vp_establishName .and .true .emit 0x00 .and vp_paramSingleInit .and\n"
-" .true .emit PARAM_NULL;\n"
-"fp_PARAM_multipleStmt\n"
-" \"PARAM\" .and space .and fp_establishName .and lbracket_ne .and optArraySize .and rbracket .and\n"
-" fp_paramMultipleInit .and .true .emit PARAM_NULL;\n"
-"vp_PARAM_multipleStmt\n"
-" \"PARAM\" .and space .and vp_establishName .and lbracket_ne .and optArraySize .and rbracket .and\n"
-" vp_paramMultipleInit .and .true .emit PARAM_NULL;\n"
-"optArraySize\n"
-" optional_integer;\n"
-"fp_paramSingleInit\n"
-" equal .and fp_paramSingleItemDecl;\n"
-"vp_paramSingleInit\n"
-" equal .and vp_paramSingleItemDecl;\n"
-"fp_paramMultipleInit\n"
-" equal .and lbrace .and fp_paramMultInitList .and rbrace;\n"
-"vp_paramMultipleInit\n"
-" equal .and lbrace .and vp_paramMultInitList .and rbrace;\n"
-"fp_paramMultInitList\n"
-" fp_paramMultInitList_1 .or fp_paramMultipleItem;\n"
-"vp_paramMultInitList\n"
-" vp_paramMultInitList_1 .or vp_paramMultipleItem;\n"
-"fp_paramMultInitList_1\n"
-" fp_paramMultipleItem .and comma_ne .and fp_paramMultInitList;\n"
-"vp_paramMultInitList_1\n"
-" vp_paramMultipleItem .and comma_ne .and vp_paramMultInitList;\n"
-"fp_paramSingleItemDecl\n"
-" fp_stateSingleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programSingleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstDecl .emit PARAM_CONSTANT;\n"
-"vp_paramSingleItemDecl\n"
-" vp_stateSingleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programSingleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstDecl .emit PARAM_CONSTANT;\n"
-"fp_paramSingleItemUse\n"
-" fp_stateSingleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programSingleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstUse .emit PARAM_CONSTANT;\n"
-"vp_paramSingleItemUse\n"
-" vp_stateSingleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programSingleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstUse .emit PARAM_CONSTANT;\n"
-"fp_paramMultipleItem\n"
-" fp_stateMultipleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programMultipleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstDecl .emit PARAM_CONSTANT;\n"
-"vp_paramMultipleItem\n"
-" vp_stateMultipleItem .emit PARAM_STATE_ELEMENT .or\n"
-" programMultipleItem .emit PARAM_PROGRAM_ELEMENT .or\n"
-" paramConstDecl .emit PARAM_CONSTANT;\n"
-"fp_stateMultipleItem\n"
-" stateMultipleItem_1 .or fp_stateSingleItem;\n"
-"vp_stateMultipleItem\n"
-" stateMultipleItem_1 .or vp_stateSingleItem;\n"
-"stateMultipleItem_1\n"
-" \"state\" .and dot .and stateMatrixRows .emit STATE_MATRIX_ROWS;\n"
-"fp_stateSingleItem\n"
-" \"state\" .and dot .and fp_stateSingleItem_1 .error INVALID_STATE_PROPERTY;\n"
-"vp_stateSingleItem\n"
-" \"state\" .and dot .and vp_stateSingleItem_1 .error INVALID_STATE_PROPERTY;\n"
-"fp_stateSingleItem_1\n"
-" stateSingleItem_1 .or stateSingleItem_2 .or stateSingleItem_3 .or stateSingleItem_4 .or\n"
-" stateSingleItem_5 .or stateSingleItem_7 .or stateSingleItem_8 .or stateSingleItem_11;\n"
-"vp_stateSingleItem_1\n"
-" stateSingleItem_1 .or stateSingleItem_2 .or stateSingleItem_3 .or stateSingleItem_4 .or\n"
-" stateSingleItem_6 .or stateSingleItem_7 .or stateSingleItem_9 .or stateSingleItem_10 .or\n"
-" stateSingleItem_11;\n"
-"stateSingleItem_1\n"
-" stateMaterialItem .emit STATE_MATERIAL;\n"
-"stateSingleItem_2\n"
-" stateLightItem .emit STATE_LIGHT;\n"
-"stateSingleItem_3\n"
-" stateLightModelItem .emit STATE_LIGHT_MODEL;\n"
-"stateSingleItem_4\n"
-" stateLightProdItem .emit STATE_LIGHT_PROD;\n"
-"stateSingleItem_5\n"
-" stateTexEnvItem .emit STATE_TEX_ENV;\n"
-"stateSingleItem_6\n"
-" stateTexGenItem .emit STATE_TEX_GEN;\n"
-"stateSingleItem_7\n"
-" stateFogItem .emit STATE_FOG;\n"
-"stateSingleItem_8\n"
-" stateDepthItem .emit STATE_DEPTH;\n"
-"stateSingleItem_9\n"
-" stateClipPlaneItem .emit STATE_CLIP_PLANE;\n"
-"stateSingleItem_10\n"
-" statePointItem .emit STATE_POINT;\n"
-"stateSingleItem_11\n"
-" stateMatrixRow .emit STATE_MATRIX_ROWS;\n"
-"stateMaterialItem\n"
-" \"material\" .and optFaceType .and dot .and stateMatProperty .error INVALID_MATERIAL_PROPERTY;\n"
-"stateMatProperty\n"
-" \"ambient\" .emit MATERIAL_AMBIENT .or\n"
-" \"diffuse\" .emit MATERIAL_DIFFUSE .or\n"
-" \"specular\" .emit MATERIAL_SPECULAR .or\n"
-" \"emission\" .emit MATERIAL_EMISSION .or\n"
-" \"shininess\" .emit MATERIAL_SHININESS;\n"
-"stateLightItem\n"
-" \"light\" .and lbracket .and stateLightNumber .and rbracket .and dot .and\n"
-" stateLightProperty .error INVALID_LIGHT_PROPERTY;\n"
-"stateLightProperty\n"
-" \"ambient\" .emit LIGHT_AMBIENT .or\n"
-" \"diffuse\" .emit LIGHT_DIFFUSE .or\n"
-" \"specular\" .emit LIGHT_SPECULAR .or\n"
-" \"position\" .emit LIGHT_POSITION .or\n"
-" \"attenuation\" .emit LIGHT_ATTENUATION .or\n"
-" stateLightProperty_1 .emit LIGHT_SPOT_DIRECTION .or\n"
-" \"half\" .emit LIGHT_HALF;\n"
-"stateLightProperty_1\n"
-" \"spot\" .and dot .and stateSpotProperty .error INVALID_SPOT_PROPERTY;\n"
-"stateSpotProperty\n"
-" \"direction\";\n"
-"stateLightModelItem\n"
-" \"lightmodel\" .and stateLModProperty .error INVALID_LIGHTMODEL_PROPERTY;\n"
-"stateLModProperty\n"
-" stateLModProperty_1 .or stateLModProperty_2;\n"
-"stateLModProperty_1\n"
-" dot .and \"ambient\" .emit LIGHT_MODEL_AMBIENT;\n"
-"stateLModProperty_2\n"
-" stateLModProperty_3 .emit LIGHT_MODEL_SCENECOLOR;\n"
-"stateLModProperty_3\n"
-" optFaceType .and dot .and \"scenecolor\";\n"
-"stateLightProdItem\n"
-" \"lightprod\" .and lbracket .and stateLightNumber .and rbracket .and optFaceType .and dot .and\n"
-" stateLProdProperty .error INVALID_LIGHTPROD_PROPERTY;\n"
-"stateLProdProperty\n"
-" \"ambient\" .emit LIGHT_PROD_AMBIENT .or\n"
-" \"diffuse\" .emit LIGHT_PROD_DIFFUSE .or\n"
-" \"specular\" .emit LIGHT_PROD_SPECULAR;\n"
-"stateLightNumber\n"
-" integer;\n"
-"stateTexEnvItem\n"
-" \"texenv\" .and optLegacyTexUnitNum .and dot .and\n"
-" stateTexEnvProperty .error INVALID_TEXENV_PROPERTY;\n"
-"stateTexEnvProperty\n"
-" \"color\" .emit TEX_ENV_COLOR;\n"
-"optLegacyTexUnitNum\n"
-" optLegacyTexUnitNum_1 .or .true .emit 0x00;\n"
-"optLegacyTexUnitNum_1\n"
-" lbracket_ne .and legacyTexUnitNum .and rbracket;\n"
-"legacyTexUnitNum\n"
-" integer;\n"
-"stateTexGenItem\n"
-" \"texgen\" .and optTexCoordNum .and dot .and stateTexGenType .error INVALID_TEXGEN_PROPERTY .and\n"
-" dot .and stateTexGenCoord .error INVALID_TEXGEN_COORD;\n"
-"stateTexGenType\n"
-" \"eye\" .emit TEX_GEN_EYE .or\n"
-" \"object\" .emit TEX_GEN_OBJECT;\n"
-"stateTexGenCoord\n"
-" \"s\" .emit COMPONENT_X .or\n"
-" \"t\" .emit COMPONENT_Y .or\n"
-" \"r\" .emit COMPONENT_Z .or\n"
-" \"q\" .emit COMPONENT_W;\n"
-"stateFogItem\n"
-" \"fog\" .and dot .and stateFogProperty .error INVALID_FOG_PROPERTY;\n"
-"stateFogProperty\n"
-" \"color\" .emit FOG_COLOR .or\n"
-" \"params\" .emit FOG_PARAMS;\n"
-"stateDepthItem\n"
-" \"depth\" .and dot .and stateDepthProperty .error INVALID_DEPTH_PROPERTY;\n"
-"stateDepthProperty\n"
-" \"range\" .emit DEPTH_RANGE;\n"
-"stateClipPlaneItem\n"
-" \"clip\" .and lbracket .and stateClipPlaneNum .and rbracket .and dot .and\n"
-" \"plane\" .error INVALID_CLIPPLANE_PROPERTY;\n"
-"stateClipPlaneNum\n"
-" integer;\n"
-"statePointItem\n"
-" \"point\" .and dot .and statePointProperty .error INVALID_POINT_PROPERTY;\n"
-"statePointProperty\n"
-" \"size\" .emit POINT_SIZE .or\n"
-" .if (point_parameters != 0x00) \"attenuation\" .emit POINT_ATTENUATION;\n"
-"stateMatrixRow\n"
-" stateMatrixItem .and dot .and \"row\" .error MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED .and\n"
-" lbracket .and stateMatrixRowNum .and rbracket .emit 0x0;\n"
-"stateMatrixRows\n"
-" stateMatrixItem .and optMatrixRows;\n"
-"optMatrixRows\n"
-" optMatrixRows_1 .or .true .emit 0x0 .emit '3' .emit 0x0 .emit $;\n"
-"optMatrixRows_1\n"
-" dot_ne .and \"row\" .error MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED .and lbracket .and\n"
-" stateMatrixRowNum .and dotdot .and stateMatrixRowNum .and rbracket;\n"
-"stateMatrixItem\n"
-" \"matrix\" .and dot .and stateMatrixName .error INVALID_MATRIX_NAME .and stateOptMatModifier;\n"
-"stateOptMatModifier\n"
-" stateOptMatModifier_1 .or .true .emit MATRIX_MODIFIER_IDENTITY;\n"
-"stateOptMatModifier_1\n"
-" dot_ne .and stateMatModifier;\n"
-"stateMatModifier\n"
-" \"inverse\" .emit MATRIX_MODIFIER_INVERSE .or\n"
-" \"transpose\" .emit MATRIX_MODIFIER_TRANSPOSE .or\n"
-" \"invtrans\" .emit MATRIX_MODIFIER_INVTRANS;\n"
-"stateMatrixRowNum\n"
-" integer_0_3;\n"
-"stateMatrixName\n"
-" stateMatrixName_1_1 .emit MATRIX_MODELVIEW .or\n"
-" \"projection\" .emit MATRIX_PROJECTION .or\n"
-" \"mvp\" .emit MATRIX_MVP .or\n"
-" stateMatrixName_1_2 .emit MATRIX_TEXTURE .or\n"
-" .if (matrix_palette != 0x00) stateMatrixName_1_3 .emit MATRIX_PALETTE .or\n"
-" stateMatrixName_1_4 .emit MATRIX_PROGRAM;\n"
-"stateMatrixName_1_1\n"
-" \"modelview\" .and stateOptModMatNum;\n"
-"stateMatrixName_1_2\n"
-" \"texture\" .and optTexCoordNum;\n"
-"stateMatrixName_1_3\n"
-" \"palette\" .and lbracket .and statePaletteMatNum .and rbracket;\n"
-"stateMatrixName_1_4\n"
-" \"program\" .and lbracket .and stateProgramMatNum .and rbracket;\n"
-"stateOptModMatNum\n"
-" .if (vertex_blend != 0x00) stateOptModMatNum_1 .or\n"
-" .true .emit 0x00;\n"
-"stateOptModMatNum_1\n"
-" lbracket_ne .and stateModMatNum .and rbracket;\n"
-"stateModMatNum\n"
-" integer;\n"
-"optTexCoordNum\n"
-" optTexCoordNum_1 .or .true .emit 0x00;\n"
-"optTexCoordNum_1\n"
-" lbracket_ne .and texCoordNum .and rbracket;\n"
-"texCoordNum\n"
-" integer;\n"
-"statePaletteMatNum\n"
-" integer;\n"
-"stateProgramMatNum\n"
-" integer;\n"
-"programSingleItem\n"
-" \"program\" .and dot .and programSingleItem_1 .error INVALID_PROGRAM_PROPERTY;\n"
-"programSingleItem_1\n"
-" progEnvParam .or progLocalParam;\n"
-"programMultipleItem\n"
-" \"program\" .and dot .and programMultipleItem_1 .error INVALID_PROGRAM_PROPERTY;\n"
-"programMultipleItem_1\n"
-" progEnvParams .or progLocalParams;\n"
-"progEnvParams\n"
-" \"env\" .emit PROGRAM_PARAM_ENV .and lbracket .and progEnvParamNums .and rbracket;\n"
-"progEnvParamNums\n"
-" progEnvParamNums_1 .or progEnvParamNums_2;\n"
-"progEnvParamNums_1\n"
-" progEnvParamNum .and dotdot_ne .and progEnvParamNum;\n"
-"progEnvParamNums_2\n"
-" progEnvParamNum .and .true .emit 0x00;\n"
-"progEnvParam\n"
-" \"env\" .emit PROGRAM_PARAM_ENV .and lbracket .and progEnvParamNum .and rbracket .emit 0x00;\n"
-"progLocalParams\n"
-" \"local\" .emit PROGRAM_PARAM_LOCAL .and lbracket .and progLocalParamNums .and rbracket;\n"
-"progLocalParamNums\n"
-" progLocalParamNums_1 .or progLocalParamNums_2;\n"
-"progLocalParamNums_1\n"
-" progLocalParamNum .and dotdot_ne .and progLocalParamNum;\n"
-"progLocalParamNums_2\n"
-" progLocalParamNum .and .true .emit 0x00;\n"
-"progLocalParam\n"
-" \"local\" .emit PROGRAM_PARAM_LOCAL .and lbracket .and progLocalParamNum .and rbracket .emit 0x00;\n"
-"progEnvParamNum\n"
-" integer;\n"
-"progLocalParamNum\n"
-" integer;\n"
-"paramConstDecl\n"
-" paramConstScalarDecl .emit CONSTANT_SCALAR .or paramConstVector .emit CONSTANT_VECTOR;\n"
-"paramConstUse\n"
-" paramConstScalarUse .emit CONSTANT_SCALAR .or paramConstVector .emit CONSTANT_VECTOR;\n"
-"paramConstScalarDecl\n"
-" signedFloatConstant;\n"
-"paramConstScalarUse\n"
-" floatConstant;\n"
-"paramConstVector\n"
-" paramConstVector_4 .emit 0x04 .or paramConstVector_3 .emit 0x03 .or\n"
-" paramConstVector_2 .emit 0x02 .or paramConstVector_1 .emit 0x01;\n"
-"paramConstVector_1\n"
-" lbrace_ne .and signedFloatConstant .and rbrace;\n"
-"paramConstVector_2\n"
-" lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and rbrace;\n"
-"paramConstVector_3\n"
-" lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and comma_ne .and\n"
-" signedFloatConstant .and rbrace;\n"
-"paramConstVector_4\n"
-" lbrace_ne .and signedFloatConstant .and comma_ne .and signedFloatConstant .and comma_ne .and\n"
-" signedFloatConstant .and comma_ne .and signedFloatConstant .and rbrace;\n"
-"signedFloatConstant\n"
-" optionalSign .and floatConstant;\n"
-"floatConstant\n"
-" float;\n"
-"optionalSign\n"
-" optional_sign_ne;\n"
-"fp_TEMP_statement\n"
-" \"TEMP\" .and space .and fp_varNameList .and .true .emit 0x00;\n"
-"vp_TEMP_statement\n"
-" \"TEMP\" .and space .and vp_varNameList .and .true .emit 0x00;\n"
-"ADDRESS_statement\n"
-" \"ADDRESS\" .and space .and vp_varNameList .and .true .emit 0x00;\n"
-"fp_varNameList\n"
-" fp_varNameList_1 .or fp_establishName;\n"
-"vp_varNameList\n"
-" vp_varNameList_1 .or vp_establishName;\n"
-"fp_varNameList_1\n"
-" fp_establishName .and comma_ne .and fp_varNameList;\n"
-"vp_varNameList_1\n"
-" vp_establishName .and comma_ne .and vp_varNameList;\n"
-"fp_OUTPUT_statement\n"
-" \"OUTPUT\" .and space .and fp_establishName .and equal .and\n"
-" fp_resultBinding .error RESULT_EXPECTED;\n"
-"vp_OUTPUT_statement\n"
-" \"OUTPUT\" .and space .and vp_establishName .and equal .and\n"
-" vp_resultBinding .error RESULT_EXPECTED;\n"
-"fp_resultBinding\n"
-" \"result\" .and dot .and fp_resultBinding_1 .error INVALID_RESULT_PROPERTY;\n"
-"vp_resultBinding\n"
-" \"result\" .and dot .and vp_resultBinding_1 .error INVALID_RESULT_PROPERTY;\n"
-"fp_resultBinding_1\n"
-" fp_resultBinding_2 .emit FRAGMENT_RESULT_COLOR .or\n"
-" \"depth\" .emit FRAGMENT_RESULT_DEPTH;\n"
-"fp_resultBinding_2\n"
-" \"color\" .and optOutputColorNum;\n"
-"vp_resultBinding_1\n"
-" .if (ARB_position_invariant == 0x00) \"position\" .emit VERTEX_RESULT_POSITION .or\n"
-" resultColBinding .emit VERTEX_RESULT_COLOR .or\n"
-" \"fogcoord\" .emit VERTEX_RESULT_FOGCOORD .or\n"
-" \"pointsize\" .emit VERTEX_RESULT_POINTSIZE .or\n"
-" vp_resultBinding_2 .emit VERTEX_RESULT_TEXCOORD;\n"
-"vp_resultBinding_2\n"
-" \"texcoord\" .and optTexCoordNum;\n"
-"optOutputColorNum\n"
-" .if (ARB_draw_buffers != 0x00) optOutputColorNum_1 .or .true .emit 0x00;\n"
-"optOutputColorNum_1\n"
-" lbracket_ne .and outputColorNum .and rbracket;\n"
-"outputColorNum\n"
-" integer;\n"
-"resultColBinding\n"
-" \"color\" .and optFaceType .and optColorType;\n"
-"optFaceType\n"
-" FaceType .or .true .emit FACE_FRONT;\n"
-"FaceType\n"
-" dot_ne .and FaceProperty;\n"
-"FaceProperty\n"
-" \"front\" .emit FACE_FRONT .or \"back\" .emit FACE_BACK;\n"
-"optColorType\n"
-" ColorType .or .true .emit COLOR_PRIMARY;\n"
-"ColorType\n"
-" dot_ne .and ColorProperty;\n"
-"ColorProperty\n"
-" \"primary\" .emit COLOR_PRIMARY .or\n"
-" .if (secondary_color != 0x00) \"secondary\" .emit COLOR_SECONDARY;\n"
-"fp_ALIAS_statement\n"
-" \"ALIAS\" .and fp_ALIAS_statement_1 .error IDENTIFIER_EXPECTED .and equal .and fp_establishedName;\n"
-"vp_ALIAS_statement\n"
-" \"ALIAS\" .and vp_ALIAS_statement_1 .error IDENTIFIER_EXPECTED .and equal .and vp_establishedName;\n"
-"fp_ALIAS_statement_1\n"
-" space .and fp_establishName;\n"
-"vp_ALIAS_statement_1\n"
-" space .and vp_establishName;\n"
-"fp_establishName\n"
-" fp_identifier;\n"
-"vp_establishName\n"
-" vp_identifier;\n"
-"fp_establishedName\n"
-" fp_identifier;\n"
-"vp_establishedName\n"
-" vp_identifier;\n"
-"fp_establishedName_no_error_on_identifier\n"
-" fp_identifier_ne;\n"
-"vp_establishedName_no_error_on_identifier\n"
-" vp_identifier_ne;\n"
-"fp_identifier\n"
-" fp_identifier_ne .error IDENTIFIER_EXPECTED;\n"
-"vp_identifier\n"
-" vp_identifier_ne .error IDENTIFIER_EXPECTED;\n"
-"fp_identifier_ne\n"
-" fp_not_reserved_identifier .and identifier_ne;\n"
-"vp_identifier_ne\n"
-" vp_not_reserved_identifier .and identifier_ne;\n"
-"fp_not_reserved_identifier\n"
-" fp_not_reserved_identifier_1 .or .true;\n"
-"fp_not_reserved_identifier_1\n"
-" fp_reserved_identifier .and .false .error RESERVED_KEYWORD;\n"
-"vp_not_reserved_identifier\n"
-" vp_not_reserved_identifier_1 .or .true;\n"
-"vp_not_reserved_identifier_1\n"
-" vp_reserved_identifier .and .false .error RESERVED_KEYWORD;\n"
-"fp_reserved_identifier\n"
-" \"ABS\" .or \"ABS_SAT\" .or \"ADD\" .or \"ADD_SAT\" .or \"ALIAS\" .or \"ATTRIB\" .or \"CMP\" .or \"CMP_SAT\" .or\n"
-" \"COS\" .or \"COS_SAT\" .or \"DP3\" .or \"DP3_SAT\" .or \"DP4\" .or \"DP4_SAT\" .or \"DPH\" .or \"DPH_SAT\" .or\n"
-" \"DST\" .or \"DST_SAT\" .or \"END\" .or \"EX2\" .or \"EX2_SAT\" .or \"FLR\" .or \"FLR_SAT\" .or \"FRC\" .or\n"
-" \"FRC_SAT\" .or \"KIL\" .or \"LG2\" .or \"LG2_SAT\" .or \"LIT\" .or \"LIT_SAT\" .or \"LRP\" .or \"LRP_SAT\" .or\n"
-" \"MAD\" .or \"MAD_SAT\" .or \"MAX\" .or \"MAX_SAT\" .or \"MIN\" .or \"MIN_SAT\" .or \"MOV\" .or \"MOV_SAT\" .or\n"
-" \"MUL\" .or \"MUL_SAT\" .or \"OPTION\" .or \"OUTPUT\" .or \"PARAM\" .or \"POW\" .or \"POW_SAT\" .or \"RCP\" .or\n"
-" \"RCP_SAT\" .or \"RSQ\" .or \"RSQ_SAT\" .or \"SIN\" .or \"SIN_SAT\" .or \"SCS\" .or \"SCS_SAT\" .or \"SGE\" .or\n"
-" \"SGE_SAT\" .or \"SLT\" .or \"SLT_SAT\" .or \"SUB\" .or \"SUB_SAT\" .or \"SWZ\" .or \"SWZ_SAT\" .or \"TEMP\" .or\n"
-" \"TEX\" .or \"TEX_SAT\" .or \"TXB\" .or \"TXB_SAT\" .or \"TXP\" .or \"TXP_SAT\" .or \"XPD\" .or \"XPD_SAT\" .or\n"
-" \"fragment\" .or \"program\" .or \"result\" .or \"state\" .or \"texture\";\n"
-"vp_reserved_identifier\n"
-" \"ABS\" .or \"ADD\" .or \"ADDRESS\" .or \"ALIAS\" .or \"ARL\" .or \"ATTRIB\" .or \"DP3\" .or \"DP4\" .or\n"
-" \"DPH\" .or \"DST\" .or \"END\" .or \"EX2\" .or \"EXP\" .or \"FLR\" .or \"FRC\" .or \"LG2\" .or \"LIT\" .or\n"
-" \"LOG\" .or \"MAD\" .or \"MAX\" .or \"MIN\" .or \"MOV\" .or \"MUL\" .or \"OPTION\" .or \"OUTPUT\" .or\n"
-" \"PARAM\" .or \"POW\" .or \"RCP\" .or \"RSQ\" .or \"SGE\" .or \"SLT\" .or \"SUB\" .or \"SWZ\" .or \"TEMP\" .or\n"
-" \"XPD\" .or \"program\" .or \"result\" .or \"state\" .or \"vertex\";\n"
-"integer\n"
-" integer_ne .error INTEGER_EXPECTED;\n"
-"zero\n"
-" '0';\n"
-"leading_zeroes\n"
-" .loop zero;\n"
-"no_digit\n"
-" no_digit_1 .or .true;\n"
-"no_digit_1\n"
-" digit10 .and .false .error INTEGER_OUT_OF_RANGE;\n"
-"all_zeroes\n"
-" all_zeroes_1 .or no_digit_1;\n"
-"all_zeroes_1\n"
-" '0' .and .loop zero .and no_digit;\n"
-"integer_0_3\n"
-" integer_0_3_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;\n"
-"integer_0_3_1\n"
-" integer_0_3_2 .or all_zeroes .emit '0';\n"
-"integer_0_3_2 \n"
-" leading_zeroes .and '1'-'3' .emit * .and no_digit;\n"
-"integer_0_63\n"
-" integer_0_63_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;\n"
-"integer_0_63_1\n"
-" integer_0_63_2 .or integer_0_63_3 .or integer_0_63_4 .or integer_0_63_5 .or\n"
-" all_zeroes .emit '0';\n"
-"integer_0_63_2 \n"
-" leading_zeroes .and '7'-'9' .emit * .and no_digit;\n"
-"integer_0_63_3 \n"
-" leading_zeroes .and '1'-'5' .emit * .and '0'-'9' .emit * .and no_digit;\n"
-"integer_0_63_4 \n"
-" leading_zeroes .and '6' .emit * .and '0'-'3' .emit * .and no_digit;\n"
-"integer_0_63_5 \n"
-" leading_zeroes .and '1'-'6' .emit * .and no_digit;\n"
-"integer_0_64\n"
-" integer_0_64_1 .error INTEGER_EXPECTED .and .true .emit 0x00 .emit $;\n"
-"integer_0_64_1\n"
-" integer_0_64_2 .or integer_0_64_3 .or integer_0_64_4 .or integer_0_64_5 .or\n"
-" all_zeroes .emit '0';\n"
-"integer_0_64_2 \n"
-" leading_zeroes .and '7'-'9' .emit * .and no_digit;\n"
-"integer_0_64_3 \n"
-" leading_zeroes .and '1'-'5' .emit * .and '0'-'9' .emit * .and no_digit;\n"
-"integer_0_64_4 \n"
-" leading_zeroes .and '6' .emit * .and '0'-'4' .emit * .and no_digit;\n"
-"integer_0_64_5 \n"
-" leading_zeroes .and '1'-'6' .emit * .and no_digit;\n"
-"optional_space\n"
-" space .or .true;\n"
-"space_dst\n"
-" space .error OPERATION_NEEDS_DESTINATION_VARIABLE;\n"
-"space_src\n"
-" space .error OPERATION_NEEDS_SOURCE_VARIABLE;\n"
-"space\n"
-" single_space .and .loop single_space;\n"
-"single_space\n"
-" white_char .or comment_block;\n"
-"white_char\n"
-" ' ' .or '\\t' .or '\\n' .or '\\r';\n"
-"comment_block\n"
-" '#' .and .loop comment_char .and optional_new_line;\n"
-"comment_char\n"
-" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n"
-"optional_new_line\n"
-" '\\n' .or crlf .or .true;\n"
-"crlf\n"
-" '\\r' .and '\\n';\n"
-"semicolon\n"
-" optional_space .and ';' .error MISSING_SEMICOLON .and optional_space;\n"
-"comma\n"
-" optional_space .and ',' .error MISSING_COMMA .and optional_space;\n"
-"comma_ne\n"
-" optional_space .and ',' .and optional_space;\n"
-"lbracket\n"
-" optional_space .and '[' .error MISSING_LBRACKET .and optional_space;\n"
-"lbracket_ne\n"
-" optional_space .and '[' .and optional_space;\n"
-"rbracket\n"
-" optional_space .and ']' .error MISSING_RBRACKET .and optional_space;\n"
-"dot\n"
-" optional_space .and '.' .error MISSING_DOT .and optional_space;\n"
-"dot_ne\n"
-" optional_space .and '.' .and optional_space;\n"
-"equal\n"
-" optional_space .and '=' .error MISSING_EQUAL .and optional_space;\n"
-"lbrace\n"
-" optional_space .and '{' .error MISSING_LBRACE .and optional_space;\n"
-"lbrace_ne\n"
-" optional_space .and '{' .and optional_space;\n"
-"rbrace\n"
-" optional_space .and '}' .error MISSING_RBRACE .and optional_space;\n"
-"dotdot\n"
-" optional_space .and '.' .and '.' .error MISSING_DOTDOT .and optional_space;\n"
-"dotdot_ne\n"
-" optional_space .and '.' .and '.' .and optional_space;\n"
-"float\n"
-" float_1 .or float_2 .or float_legacy;\n"
-"float_1\n"
-" '.' .emit 0x00 .and integer_ne .error MISSING_FRACTION_OR_EXPONENT .and optional_exponent;\n"
-"float_2\n"
-" integer_ne .and float_3;\n"
-"float_3\n"
-" float_4 .or float_5;\n"
-"float_4\n"
-" '.' .and optional_integer .and optional_exponent;\n"
-"float_5\n"
-" exponent .emit 0x00;\n"
-"float_legacy\n"
-" integer_ne .and .true .emit 0x00 .emit 0x00;\n"
-"integer_ne\n"
-" integer_ne_1 .and .true .emit 0x00 .emit $;\n"
-"integer_ne_1\n"
-" digit10 .emit * .and .loop digit10 .emit *;\n"
-"optional_integer\n"
-" integer_ne .or .true .emit 0x00;\n"
-"optional_exponent\n"
-" exponent .or .true .emit 0x00;\n"
-"exponent\n"
-" exponent_1 .and optional_sign_ne .and integer_ne .error EXPONENT_VALUE_EXPECTED;\n"
-"exponent_1\n"
-" 'e' .or 'E';\n"
-"optional_sign_ne\n"
-" minus_ne .or plus_ne .or .true;\n"
-"plus_ne\n"
-" optional_space .and '+' .and optional_space;\n"
-"minus_ne\n"
-" optional_space .and '-' .emit '-' .and optional_space;\n"
-"identifier_ne\n"
-" first_idchar .emit * .and .loop follow_idchar .emit * .and .true .emit 0x00 .emit $;\n"
-"follow_idchar\n"
-" first_idchar .or digit10;\n"
-"first_idchar\n"
-" 'a'-'z' .or 'A'-'Z' .or '_' .or '$';\n"
-"digit10\n"
-" '0'-'9';\n"
-".string __string_filter;\n"
-"__string_filter\n"
-" .loop __identifier_char;\n"
-"__identifier_char\n"
-" 'a'-'z' .or 'A'-'Z' .or '_' .or '$' .or '0'-'9';\n"
-"e_signature\n"
-" e_signature_char .and .loop e_signature_char;\n"
-"e_signature_char\n"
-" '!' .or '.' .or 'A'-'Z' .or 'a'-'z' .or '0'-'9';\n"
-"e_statement\n"
-" .loop e_statement_not_term;\n"
-"e_statement_not_term\n"
-" '\\x3C'-'\\xFF' .or '\\x0E'-'\\x3A' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n"
-"e_identifier\n"
-" e_identifier_first .and .loop e_identifier_next;\n"
-"e_identifier_first\n"
-" 'a'-'z' .or 'A'-'Z' .or '_' .or '$';\n"
-"e_identifier_next\n"
-" e_identifier_first .or '0'-'9';\n"
-"e_token\n"
-" e_identifier .or e_token_number .or '[' .or ']' .or '.' .or '{' .or '}' .or '=' .or '+' .or\n"
-" '-' .or ',' .or ';';\n"
-"e_token_number\n"
-" e_token_digit .and .loop e_token_digit;\n"
-"e_token_digit\n"
-" '0'-'9';\n"
-"e_charordigit\n"
-" 'A'-'Z' .or 'a'-'z' .or '0'-'9';\n"
-""
diff --git a/src/mesa/shader/hash_table.c b/src/mesa/shader/hash_table.c
new file mode 100644
index 0000000000..881179f9d8
--- /dev/null
+++ b/src/mesa/shader/hash_table.c
@@ -0,0 +1,163 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file hash_table.c
+ * \brief Implementation of a generic, opaque hash table data type.
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+#include <stdlib.h>
+#include <string.h>
+
+#include <assert.h>
+
+#include "main/imports.h"
+#include "main/simple_list.h"
+#include "hash_table.h"
+
+struct node {
+   struct node *next;
+   struct node *prev;
+};
+
+struct hash_table {
+    hash_func_t    hash;
+    hash_compare_func_t  compare;
+
+    unsigned num_buckets;
+    struct node buckets[1];
+};
+
+
+struct hash_node {
+    struct node link;
+    const void *key;
+    void *data;
+};
+
+
+struct hash_table *
+hash_table_ctor(unsigned num_buckets, hash_func_t hash,
+                hash_compare_func_t compare)
+{
+    struct hash_table *ht;
+    unsigned i;
+
+
+    if (num_buckets < 16) {
+        num_buckets = 16;
+    }
+
+    ht = _mesa_malloc(sizeof(*ht) + ((num_buckets - 1) 
+				     * sizeof(ht->buckets[0])));
+    if (ht != NULL) {
+        ht->hash = hash;
+        ht->compare = compare;
+        ht->num_buckets = num_buckets;
+
+        for (i = 0; i < num_buckets; i++) {
+            make_empty_list(& ht->buckets[i]);
+        }
+    }
+
+    return ht;
+}
+
+
+void
+hash_table_dtor(struct hash_table *ht)
+{
+   hash_table_clear(ht);
+   _mesa_free(ht);
+}
+
+
+void
+hash_table_clear(struct hash_table *ht)
+{
+   struct node *node;
+   struct node *temp;
+   unsigned i;
+
+
+   for (i = 0; i < ht->num_buckets; i++) {
+      foreach_s(node, temp, & ht->buckets[i]) {
+	 remove_from_list(node);
+	 _mesa_free(node);
+      }
+
+      assert(is_empty_list(& ht->buckets[i]));
+   }
+}
+
+
+void *
+hash_table_find(struct hash_table *ht, const void *key)
+{
+    const unsigned hash_value = (*ht->hash)(key);
+    const unsigned bucket = hash_value % ht->num_buckets;
+    struct node *node;
+
+    foreach(node, & ht->buckets[bucket]) {
+       struct hash_node *hn = (struct hash_node *) node;
+
+       if ((*ht->compare)(hn->key, key) == 0) {
+	  return hn->data;
+       }
+    }
+
+    return NULL;
+}
+
+
+void
+hash_table_insert(struct hash_table *ht, void *data, const void *key)
+{
+    const unsigned hash_value = (*ht->hash)(key);
+    const unsigned bucket = hash_value % ht->num_buckets;
+    struct hash_node *node;
+
+    node = _mesa_calloc(sizeof(*node));
+
+    node->data = data;
+    node->key = key;
+
+    insert_at_head(& ht->buckets[bucket], & node->link);
+}
+
+
+unsigned
+hash_table_string_hash(const void *key)
+{
+    const char *str = (const char *) key;
+    unsigned hash = 5381;
+
+
+    while (*str != '\0') {
+        hash = (hash * 33) + *str;
+        str++;
+    }
+
+    return hash;
+}
diff --git a/src/mesa/shader/hash_table.h b/src/mesa/shader/hash_table.h
new file mode 100644
index 0000000000..7b302f5dbe
--- /dev/null
+++ b/src/mesa/shader/hash_table.h
@@ -0,0 +1,117 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file hash_table.h
+ * \brief Implementation of a generic, opaque hash table data type.
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+#ifndef HASH_TABLE_H
+#define HASH_TABLE_H
+
+#include <string.h>
+
+struct hash_table;
+
+typedef unsigned (*hash_func_t)(const void *key);
+typedef int (*hash_compare_func_t)(const void *key1, const void *key2);
+
+/**
+ * Hash table constructor
+ *
+ * Creates a hash table with the specified number of buckets.  The supplied
+ * \c hash and \c compare routines are used when adding elements to the table
+ * and when searching for elements in the table.
+ *
+ * \param num_buckets  Number of buckets (bins) in the hash table.
+ * \param hash         Function used to compute hash value of input keys.
+ * \param compare      Function used to compare keys.
+ */
+extern struct hash_table *hash_table_ctor(unsigned num_buckets,
+    hash_func_t hash, hash_compare_func_t compare);
+
+
+/**
+ * Release all memory associated with a hash table
+ *
+ * \warning
+ * This function cannot release memory occupied either by keys or data.
+ */
+extern void hash_table_dtor(struct hash_table *ht);
+
+
+/**
+ * Flush all entries from a hash table
+ *
+ * \param ht  Table to be cleared of its entries.
+ */
+extern void hash_table_clear(struct hash_table *ht);
+
+
+/**
+ * Search a hash table for a specific element
+ *
+ * \param ht   Table to be searched
+ * \param key  Key of the desired element
+ *
+ * \return
+ * The \c data value supplied to \c hash_table_insert when the element with
+ * the matching key was added.  If no matching key exists in the table,
+ * \c NULL is returned.
+ */
+extern void *hash_table_find(struct hash_table *ht, const void *key);
+
+
+/**
+ * Add an element to a hash table
+ */
+extern void hash_table_insert(struct hash_table *ht, void *data,
+    const void *key);
+
+
+/**
+ * Compute hash value of a string
+ *
+ * Computes the hash value of a string using the DJB2 algorithm developed by
+ * Professor Daniel J. Bernstein.  It was published on comp.lang.c once upon
+ * a time.  I was unable to find the original posting in the archives.
+ *
+ * \param key  Pointer to a NUL terminated string to be hashed.
+ *
+ * \sa hash_table_string_compare
+ */
+extern unsigned hash_table_string_hash(const void *key);
+
+
+/**
+ * Compare two strings used as keys
+ *
+ * This is just a macro wrapper around \c strcmp.
+ *
+ * \sa hash_table_string_hash
+ */
+#define hash_table_string_compare ((hash_compare_func_t) strcmp)
+
+#endif /* HASH_TABLE_H */
diff --git a/src/mesa/shader/lex.yy.c b/src/mesa/shader/lex.yy.c
new file mode 100644
index 0000000000..709426f3a6
--- /dev/null
+++ b/src/mesa/shader/lex.yy.c
@@ -0,0 +1,3574 @@
+
+#line 3 "lex.yy.c"
+
+#define  YY_INT_ALIGNED short int
+
+/* A lexical scanner generated by flex */
+
+#define FLEX_SCANNER
+#define YY_FLEX_MAJOR_VERSION 2
+#define YY_FLEX_MINOR_VERSION 5
+#define YY_FLEX_SUBMINOR_VERSION 35
+#if YY_FLEX_SUBMINOR_VERSION > 0
+#define FLEX_BETA
+#endif
+
+/* First, we deal with  platform-specific or compiler-specific issues. */
+
+/* begin standard C headers. */
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include <stdlib.h>
+
+/* end standard C headers. */
+
+/* flex integer type definitions */
+
+#ifndef FLEXINT_H
+#define FLEXINT_H
+
+/* C99 systems have <inttypes.h>. Non-C99 systems may or may not. */
+
+#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
+
+/* C99 says to define __STDC_LIMIT_MACROS before including stdint.h,
+ * if you want the limit (max/min) macros for int types. 
+ */
+#ifndef __STDC_LIMIT_MACROS
+#define __STDC_LIMIT_MACROS 1
+#endif
+
+#include <inttypes.h>
+typedef int8_t flex_int8_t;
+typedef uint8_t flex_uint8_t;
+typedef int16_t flex_int16_t;
+typedef uint16_t flex_uint16_t;
+typedef int32_t flex_int32_t;
+typedef uint32_t flex_uint32_t;
+#else
+typedef signed char flex_int8_t;
+typedef short int flex_int16_t;
+typedef int flex_int32_t;
+typedef unsigned char flex_uint8_t; 
+typedef unsigned short int flex_uint16_t;
+typedef unsigned int flex_uint32_t;
+#endif /* ! C99 */
+
+/* Limits of integral types. */
+#ifndef INT8_MIN
+#define INT8_MIN               (-128)
+#endif
+#ifndef INT16_MIN
+#define INT16_MIN              (-32767-1)
+#endif
+#ifndef INT32_MIN
+#define INT32_MIN              (-2147483647-1)
+#endif
+#ifndef INT8_MAX
+#define INT8_MAX               (127)
+#endif
+#ifndef INT16_MAX
+#define INT16_MAX              (32767)
+#endif
+#ifndef INT32_MAX
+#define INT32_MAX              (2147483647)
+#endif
+#ifndef UINT8_MAX
+#define UINT8_MAX              (255U)
+#endif
+#ifndef UINT16_MAX
+#define UINT16_MAX             (65535U)
+#endif
+#ifndef UINT32_MAX
+#define UINT32_MAX             (4294967295U)
+#endif
+
+#endif /* ! FLEXINT_H */
+
+#ifdef __cplusplus
+
+/* The "const" storage-class-modifier is valid. */
+#define YY_USE_CONST
+
+#else	/* ! __cplusplus */
+
+/* C99 requires __STDC__ to be defined as 1. */
+#if defined (__STDC__)
+
+#define YY_USE_CONST
+
+#endif	/* defined (__STDC__) */
+#endif	/* ! __cplusplus */
+
+#ifdef YY_USE_CONST
+#define yyconst const
+#else
+#define yyconst
+#endif
+
+/* Returned upon end-of-file. */
+#define YY_NULL 0
+
+/* Promotes a possibly negative, possibly signed char to an unsigned
+ * integer for use as an array index.  If the signed char is negative,
+ * we want to instead treat it as an 8-bit unsigned char, hence the
+ * double cast.
+ */
+#define YY_SC_TO_UI(c) ((unsigned int) (unsigned char) c)
+
+/* An opaque pointer. */
+#ifndef YY_TYPEDEF_YY_SCANNER_T
+#define YY_TYPEDEF_YY_SCANNER_T
+typedef void* yyscan_t;
+#endif
+
+/* For convenience, these vars (plus the bison vars far below)
+   are macros in the reentrant scanner. */
+#define yyin yyg->yyin_r
+#define yyout yyg->yyout_r
+#define yyextra yyg->yyextra_r
+#define yyleng yyg->yyleng_r
+#define yytext yyg->yytext_r
+#define yylineno (YY_CURRENT_BUFFER_LVALUE->yy_bs_lineno)
+#define yycolumn (YY_CURRENT_BUFFER_LVALUE->yy_bs_column)
+#define yy_flex_debug yyg->yy_flex_debug_r
+
+/* Enter a start condition.  This macro really ought to take a parameter,
+ * but we do it the disgusting crufty way forced on us by the ()-less
+ * definition of BEGIN.
+ */
+#define BEGIN yyg->yy_start = 1 + 2 *
+
+/* Translate the current start state into a value that can be later handed
+ * to BEGIN to return to the state.  The YYSTATE alias is for lex
+ * compatibility.
+ */
+#define YY_START ((yyg->yy_start - 1) / 2)
+#define YYSTATE YY_START
+
+/* Action number for EOF rule of a given start state. */
+#define YY_STATE_EOF(state) (YY_END_OF_BUFFER + state + 1)
+
+/* Special action meaning "start processing a new file". */
+#define YY_NEW_FILE yyrestart(yyin ,yyscanner )
+
+#define YY_END_OF_BUFFER_CHAR 0
+
+/* Size of default input buffer. */
+#ifndef YY_BUF_SIZE
+#define YY_BUF_SIZE 16384
+#endif
+
+/* The state buf must be large enough to hold one state per character in the main buffer.
+ */
+#define YY_STATE_BUF_SIZE   ((YY_BUF_SIZE + 2) * sizeof(yy_state_type))
+
+#ifndef YY_TYPEDEF_YY_BUFFER_STATE
+#define YY_TYPEDEF_YY_BUFFER_STATE
+typedef struct yy_buffer_state *YY_BUFFER_STATE;
+#endif
+
+#define EOB_ACT_CONTINUE_SCAN 0
+#define EOB_ACT_END_OF_FILE 1
+#define EOB_ACT_LAST_MATCH 2
+
+    #define YY_LESS_LINENO(n)
+    
+/* Return all but the first "n" matched characters back to the input stream. */
+#define yyless(n) \
+	do \
+		{ \
+		/* Undo effects of setting up yytext. */ \
+        int yyless_macro_arg = (n); \
+        YY_LESS_LINENO(yyless_macro_arg);\
+		*yy_cp = yyg->yy_hold_char; \
+		YY_RESTORE_YY_MORE_OFFSET \
+		yyg->yy_c_buf_p = yy_cp = yy_bp + yyless_macro_arg - YY_MORE_ADJ; \
+		YY_DO_BEFORE_ACTION; /* set up yytext again */ \
+		} \
+	while ( 0 )
+
+#define unput(c) yyunput( c, yyg->yytext_ptr , yyscanner )
+
+#ifndef YY_TYPEDEF_YY_SIZE_T
+#define YY_TYPEDEF_YY_SIZE_T
+typedef size_t yy_size_t;
+#endif
+
+#ifndef YY_STRUCT_YY_BUFFER_STATE
+#define YY_STRUCT_YY_BUFFER_STATE
+struct yy_buffer_state
+	{
+	FILE *yy_input_file;
+
+	char *yy_ch_buf;		/* input buffer */
+	char *yy_buf_pos;		/* current position in input buffer */
+
+	/* Size of input buffer in bytes, not including room for EOB
+	 * characters.
+	 */
+	yy_size_t yy_buf_size;
+
+	/* Number of characters read into yy_ch_buf, not including EOB
+	 * characters.
+	 */
+	int yy_n_chars;
+
+	/* Whether we "own" the buffer - i.e., we know we created it,
+	 * and can realloc() it to grow it, and should free() it to
+	 * delete it.
+	 */
+	int yy_is_our_buffer;
+
+	/* Whether this is an "interactive" input source; if so, and
+	 * if we're using stdio for input, then we want to use getc()
+	 * instead of fread(), to make sure we stop fetching input after
+	 * each newline.
+	 */
+	int yy_is_interactive;
+
+	/* Whether we're considered to be at the beginning of a line.
+	 * If so, '^' rules will be active on the next match, otherwise
+	 * not.
+	 */
+	int yy_at_bol;
+
+    int yy_bs_lineno; /**< The line count. */
+    int yy_bs_column; /**< The column count. */
+    
+	/* Whether to try to fill the input buffer when we reach the
+	 * end of it.
+	 */
+	int yy_fill_buffer;
+
+	int yy_buffer_status;
+
+#define YY_BUFFER_NEW 0
+#define YY_BUFFER_NORMAL 1
+	/* When an EOF's been seen but there's still some text to process
+	 * then we mark the buffer as YY_EOF_PENDING, to indicate that we
+	 * shouldn't try reading from the input source any more.  We might
+	 * still have a bunch of tokens to match, though, because of
+	 * possible backing-up.
+	 *
+	 * When we actually see the EOF, we change the status to "new"
+	 * (via yyrestart()), so that the user can continue scanning by
+	 * just pointing yyin at a new input file.
+	 */
+#define YY_BUFFER_EOF_PENDING 2
+
+	};
+#endif /* !YY_STRUCT_YY_BUFFER_STATE */
+
+/* We provide macros for accessing buffer states in case in the
+ * future we want to put the buffer states in a more general
+ * "scanner state".
+ *
+ * Returns the top of the stack, or NULL.
+ */
+#define YY_CURRENT_BUFFER ( yyg->yy_buffer_stack \
+                          ? yyg->yy_buffer_stack[yyg->yy_buffer_stack_top] \
+                          : NULL)
+
+/* Same as previous macro, but useful when we know that the buffer stack is not
+ * NULL or when we need an lvalue. For internal use only.
+ */
+#define YY_CURRENT_BUFFER_LVALUE yyg->yy_buffer_stack[yyg->yy_buffer_stack_top]
+
+void yyrestart (FILE *input_file ,yyscan_t yyscanner );
+void yy_switch_to_buffer (YY_BUFFER_STATE new_buffer ,yyscan_t yyscanner );
+YY_BUFFER_STATE yy_create_buffer (FILE *file,int size ,yyscan_t yyscanner );
+void yy_delete_buffer (YY_BUFFER_STATE b ,yyscan_t yyscanner );
+void yy_flush_buffer (YY_BUFFER_STATE b ,yyscan_t yyscanner );
+void yypush_buffer_state (YY_BUFFER_STATE new_buffer ,yyscan_t yyscanner );
+void yypop_buffer_state (yyscan_t yyscanner );
+
+static void yyensure_buffer_stack (yyscan_t yyscanner );
+static void yy_load_buffer_state (yyscan_t yyscanner );
+static void yy_init_buffer (YY_BUFFER_STATE b,FILE *file ,yyscan_t yyscanner );
+
+#define YY_FLUSH_BUFFER yy_flush_buffer(YY_CURRENT_BUFFER ,yyscanner)
+
+YY_BUFFER_STATE yy_scan_buffer (char *base,yy_size_t size ,yyscan_t yyscanner );
+YY_BUFFER_STATE yy_scan_string (yyconst char *yy_str ,yyscan_t yyscanner );
+YY_BUFFER_STATE yy_scan_bytes (yyconst char *bytes,int len ,yyscan_t yyscanner );
+
+void *yyalloc (yy_size_t ,yyscan_t yyscanner );
+void *yyrealloc (void *,yy_size_t ,yyscan_t yyscanner );
+void yyfree (void * ,yyscan_t yyscanner );
+
+#define yy_new_buffer yy_create_buffer
+
+#define yy_set_interactive(is_interactive) \
+	{ \
+	if ( ! YY_CURRENT_BUFFER ){ \
+        yyensure_buffer_stack (yyscanner); \
+		YY_CURRENT_BUFFER_LVALUE =    \
+            yy_create_buffer(yyin,YY_BUF_SIZE ,yyscanner); \
+	} \
+	YY_CURRENT_BUFFER_LVALUE->yy_is_interactive = is_interactive; \
+	}
+
+#define yy_set_bol(at_bol) \
+	{ \
+	if ( ! YY_CURRENT_BUFFER ){\
+        yyensure_buffer_stack (yyscanner); \
+		YY_CURRENT_BUFFER_LVALUE =    \
+            yy_create_buffer(yyin,YY_BUF_SIZE ,yyscanner); \
+	} \
+	YY_CURRENT_BUFFER_LVALUE->yy_at_bol = at_bol; \
+	}
+
+#define YY_AT_BOL() (YY_CURRENT_BUFFER_LVALUE->yy_at_bol)
+
+/* Begin user sect3 */
+
+#define yywrap(n) 1
+#define YY_SKIP_YYWRAP
+
+typedef unsigned char YY_CHAR;
+
+typedef int yy_state_type;
+
+#define yytext_ptr yytext_r
+
+static yy_state_type yy_get_previous_state (yyscan_t yyscanner );
+static yy_state_type yy_try_NUL_trans (yy_state_type current_state  ,yyscan_t yyscanner);
+static int yy_get_next_buffer (yyscan_t yyscanner );
+static void yy_fatal_error (yyconst char msg[] ,yyscan_t yyscanner );
+
+/* Done after the current pattern has been matched and before the
+ * corresponding action - sets up yytext.
+ */
+#define YY_DO_BEFORE_ACTION \
+	yyg->yytext_ptr = yy_bp; \
+	yyleng = (size_t) (yy_cp - yy_bp); \
+	yyg->yy_hold_char = *yy_cp; \
+	*yy_cp = '\0'; \
+	yyg->yy_c_buf_p = yy_cp;
+
+#define YY_NUM_RULES 183
+#define YY_END_OF_BUFFER 184
+/* This struct is not used in this scanner,
+   but its presence is necessary. */
+struct yy_trans_info
+	{
+	flex_int32_t yy_verify;
+	flex_int32_t yy_nxt;
+	};
+static yyconst flex_int16_t yy_accept[675] =
+    {   0,
+        0,    0,  184,  182,  180,  179,  182,  182,  152,  178,
+      154,  154,  154,  154,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  180,    0,    0,  181,  152,    0,
+      153,  155,  175,  175,    0,    0,    0,    0,  175,    0,
+        0,    0,    0,    0,    0,    0,  132,  176,  133,  134,
+      166,  166,  166,  166,    0,  154,    0,  140,  141,  142,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,    0,    0,    0,    0,
+        0,    0,    0,    0,    0,  174,    0,    0,    0,    0,
+        0,    0,    0,    0,    0,  173,  173,    0,    0,    0,
+        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,
+      172,  172,  172,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,    0,  163,  163,  163,  164,  164,  165,  156,
+      155,  156,    0,  157,   11,   13,  152,   15,  152,  152,
+       16,   18,  152,   20,   22,   24,   26,    6,   28,   30,
+       31,   33,   35,   38,   36,   40,   41,   43,   45,   47,
+
+       49,   51,  152,  152,  152,   53,   55,  152,   57,   59,
+       61,  152,   63,   65,   67,   69,  152,   71,   73,   75,
+       77,  152,  152,  152,  152,  152,  152,    0,    0,    0,
+        0,  155,    0,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,   93,   94,   96,    0,  171,    0,    0,    0,
+        0,    0,    0,  110,    0,    0,    0,    0,    0,    0,
+        0,    0,    0,    0,  170,  169,  169,  122,    0,    0,
+        0,    0,    0,    0,    0,    0,    0,    0,  160,  160,
+      161,  162,    0,  158,  152,  152,  152,  152,  152,  152,
+      152,  152,  143,  152,  152,  152,  152,  152,  152,  152,
+
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  144,  152,  152,  152,  152,  152,  152,
+      152,  152,   10,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,    0,  177,    0,    0,    0,   86,   87,
+        0,    0,    0,    0,    0,    0,    0,   98,    0,    0,
+        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,    0,    0,    0,    0,  168,    0,    0,    0,
+      126,    0,  128,    0,    0,    0,    0,    0,    0,  167,
+      159,  152,  152,  152,    4,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+
+      152,  152,  152,  152,  152,  152,    9,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,   82,  152,  152,    0,    0,    0,
+        0,    0,   88,   89,    0,    0,    0,    0,   97,    0,
+        0,  101,  104,    0,    0,    0,    0,    0,    0,    0,
+      115,  116,    0,    0,    0,    0,  121,    0,    0,    0,
+        0,    0,    0,    0,    0,    0,    0,  152,  152,  152,
+      152,  152,  152,    5,  152,  152,  152,  152,  152,  152,
+      152,  152,  152,  152,  152,  152,  152,  152,  152,  152,
+      152,    7,    8,  152,  152,  152,  152,  152,  152,  152,
+
+      152,  152,  152,  152,  152,  152,  152,  152,  152,   83,
+      152,   79,    0,    0,    0,    0,  137,    0,    0,    0,
+        0,    0,    0,    0,    0,    0,  107,    0,  111,  112,
+        0,  114,    0,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,  130,  131,    0,    0,  138,   12,    3,   14,
+      148,  149,  152,   17,   19,   21,   23,   25,   27,   29,
+       32,   34,   39,   37,   42,   44,   46,   48,   50,   52,
+       54,   56,   58,   60,   62,  152,  152,  152,   64,   66,
+       68,   70,   72,   74,   76,   78,  152,   81,  139,    0,
+        0,   84,    0,   90,    0,    0,    0,   99,    0,    0,
+
+        0,    0,    0,    0,  113,    0,    0,  119,  106,    0,
+        0,    0,    0,    0,    0,  135,    0,  152,  145,  146,
+      152,   80,    0,    0,    0,    0,   92,   95,  100,    0,
+        0,  105,    0,    0,    0,  118,    0,    0,    0,    0,
+      127,  129,    0,  152,  152,    2,    1,    0,   91,    0,
+      103,    0,  109,  117,    0,    0,  124,  125,  136,  152,
+      147,    0,  102,    0,  120,  123,  152,   85,  108,  152,
+      152,  150,  151,    0
+    } ;
+
+static yyconst flex_int32_t yy_ec[256] =
+    {   0,
+        1,    1,    1,    1,    1,    1,    1,    1,    2,    3,
+        1,    1,    4,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    2,    5,    1,    6,    7,    1,    1,    1,    1,
+        1,    1,    8,    1,    8,    9,    1,   10,   11,   12,
+       13,   14,   15,   15,   15,   15,   15,    1,    1,    1,
+        1,    1,    1,    1,   16,   17,   18,   19,   20,   21,
+       22,   23,   24,    7,   25,   26,   27,   28,   29,   30,
+       31,   32,   33,   34,   35,   36,   37,   38,   39,   40,
+        1,    1,    1,    1,   41,    1,   42,   43,   44,   45,
+
+       46,   47,   48,   49,   50,   51,   52,   53,   54,   55,
+       56,   57,   58,   59,   60,   61,   62,   63,   64,   65,
+       66,   67,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1
+    } ;
+
+static yyconst flex_int32_t yy_meta[68] =
+    {   0,
+        1,    1,    1,    1,    1,    1,    2,    1,    3,    2,
+        2,    2,    2,    2,    2,    2,    2,    2,    2,    2,
+        2,    2,    2,    2,    2,    2,    2,    2,    2,    2,
+        2,    2,    2,    2,    2,    2,    2,    2,    2,    2,
+        2,    2,    2,    2,    2,    2,    2,    2,    2,    2,
+        2,    2,    2,    2,    2,    2,    2,    2,    2,    2,
+        2,    2,    2,    2,    2,    2,    2
+    } ;
+
+static yyconst flex_int16_t yy_base[678] =
+    {   0,
+        0,    0,  954,  955,   66,  955,  948,  949,    0,   69,
+       85,  128,  140,  152,  151,   58,   39,   48,   75,  927,
+      158,  160,   73,   59,   71,  170,   54,  920,  890,  889,
+      901,  885,  899,  898,  142,  927,  939,  955,    0,  206,
+      955,  189,  168,  171,   53,   27,   66,  119,  175,  899,
+      885,  123,  170,  883,  895,  183,  955,  198,  225,   99,
+      212,  219,  223,  227,  285,  297,  308,  955,  955,  955,
+      904,  917,  911,  165,  900,  903,  899,  914,  224,  896,
+      910,  194,  896,  909,  900,  913,  890,  901,  892,  294,
+      893,  884,  893,  884,  883,  884,  878,  884,  895,  881,
+
+      878,  890,  893,  880,  873,  889,  865,  193,  139,  885,
+      861,  846,  841,  858,  834,  839,  865,  167,  854,  259,
+      849,  325,  282,  851,  832,  302,  842,  838,  833,   43,
+      839,  825,  841,  838,  829,  305,  309,  831,  820,  834,
+      837,  819,  834,  821,  818,  825,  275,  833,  254,  299,
+      317,  327,  331,  810,  827,  828,  821,  803,  310,  804,
+      826,  817,  316,  327,  331,  335,  339,  343,  347,  955,
+      405,  416,  422,  428,  825,  240,  849,    0,  848,  831,
+      821,  820,  840,  818,  817,  816,  815,    0,  814,    0,
+      813,  812,    0,  811,  810,    0,  809,  808,  807,  806,
+
+      805,  804,  820,  813,  826,  800,  799,  805,  797,  796,
+      795,  816,  793,  792,  791,  790,  800,  788,  787,  786,
+      785,  777,  776,  761,  761,  760,  759,  802,  774,  762,
+      434,  442,  416,  766,  186,  763,  757,  757,  751,  764,
+      764,  749,  955,  955,  764,  752,  418,  759,  281,  756,
+      762,  308,  757,  955,  748,  755,  754,  757,  743,  742,
+      746,  741,  278,  746,  420,  428,  430,  955,  738,  736,
+      736,  744,  745,  727,  421,  732,  738,  419,  426,  430,
+      434,  438,  496,  502,  752,  764,  750,  749,  742,  756,
+      746,  745,    0,  744,  743,  742,  741,  740,  739,  738,
+
+      737,  736,  735,  734,  733,  732,  731,  730,  733,  726,
+      733,  726,  725,    0,  724,  723,  722,  725,  720,  719,
+      718,  717,    0,  716,  715,  714,  713,  691,  685,  690,
+      696,  679,  694,  315,  955,  693,  683,  687,  955,  955,
+      677,  686,  672,  689,  672,  675,  669,  955,  670,  669,
+      666,  673,  666,  674,  670,  680,  677,  659,  665,  672,
+      656,  655,  673,  655,  667,  666,  955,  665,  655,  659,
+      955,  646,  955,  651,  651,  659,  642,  643,  653,  955,
+      955,  685,  667,  683,    0,  507,  681,  681,  680,  679,
+      678,  677,  676,  675,  674,  673,  672,  671,  670,  669,
+
+      668,  667,  666,  665,  652,  645,    0,  662,  661,  660,
+      659,  658,  636,  656,  655,  654,  653,  652,  651,  650,
+      649,  618,  621,  601,    0,  602,  595,  602,  601,  602,
+      594,  612,  955,  955,  594,  592,  602,  595,  955,  590,
+      607,  330,  955,  598,  582,  583,  592,  583,  582,  582,
+      955,  581,  590,  580,  596,  593,  955,  592,  590,  579,
+      580,  576,  568,  575,  570,  571,  566,  592,  592,  590,
+      604,  603,  598,    0,  586,  585,  584,  583,  582,  581,
+      580,  579,  578,  577,  576,  575,  574,  573,  572,  571,
+      570,    0,    0,  569,  568,  567,  566,  565,  509,  564,
+
+      563,  562,  561,  560,  559,  558,  557,  535,  535,    0,
+      542,    0,  576,  575,  524,  542,  955,  537,  532,  525,
+      521,  533,  523,  521,  517,  533,  524,  523,  955,  955,
+      526,  955,  521,  514,  503,  514,  506,  510,  523,  518,
+      521,  503,  955,  955,  515,  504,  955,    0,    0,    0,
+        0,    0,  543,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,    0,    0,    0,    0,    0,    0,    0,    0,
+        0,    0,    0,    0,    0,  539,  538,  536,    0,    0,
+        0,    0,    0,    0,    0,    0,  494,    0,    0,  545,
+      544,  955,  491,  955,  495,  495,  504,  955,  488,  502,
+
+      483,  485,  482,  490,  955,  468,  479,  955,  955,  483,
+      479,  472,  470,  470,  483,  955,  467,  507,    0,    0,
+      507,    0,  514,  513,  472,  433,  955,  955,  955,  435,
+      435,  955,  429,  386,  377,  955,  366,  365,  323,  328,
+      955,  955,  339,  348,  337,  955,  955,  307,  955,  305,
+      955,  257,  955,  955,  247,  221,  955,  955,  955,  236,
+        0,  213,  955,  150,  955,  955,  232,  955,  955,  162,
+      138,    0,    0,  955,  541,  108,  544
+    } ;
+
+static yyconst flex_int16_t yy_def[678] =
+    {   0,
+      674,    1,  674,  674,  674,  674,  674,  675,  676,  674,
+      674,  674,  674,  674,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  674,  674,  675,  674,  676,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  677,  674,  674,  674,  674,  674,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  676,
+      676,  676,  676,  676,  676,  676,  676,  676,  676,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  676,  676,  676,
+      676,  676,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  676,  676,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  676,
+      676,  674,  674,  674,  674,  674,  676,  674,  674,  676,
+      676,  676,  676,    0,  674,  674,  674
+    } ;
+
+static yyconst flex_int16_t yy_nxt[1023] =
+    {   0,
+        4,    5,    6,    5,    7,    8,    9,    4,   10,   11,
+       12,   13,   14,   11,   11,   15,    9,   16,   17,   18,
+       19,    9,    9,    9,   20,   21,   22,    9,   23,   24,
+        9,   25,   26,   27,    9,    9,    9,   28,    9,    9,
+        9,    9,    9,    9,    9,    9,   29,    9,    9,    9,
+        9,    9,    9,    9,    9,    9,   30,    9,   31,   32,
+       33,    9,   34,    9,    9,    9,    9,   35,   79,   35,
+       40,   80,  129,  108,   96,   81,  130,   41,   42,   42,
+       42,   42,   42,   42,   76,   82,   77,   97,   98,  240,
+       99,  109,   78,   65,   66,   66,   66,   66,   66,   66,
+
+       83,  241,   94,  100,   67,  127,   84,   95,  128,   39,
+       43,   44,   45,   46,   47,   48,   49,   50,   51,  131,
+      132,   52,   53,   54,   55,   56,   57,   58,   59,   60,
+       67,  133,   61,   62,   63,   64,   65,   66,   66,   66,
+       66,   66,   66,   35,  160,   35,   68,   67,   65,   66,
+       66,   66,   66,   66,   66,  219,  673,  161,   69,   67,
+       65,   66,   66,   66,   66,   66,   66,   71,  220,   72,
+       70,   67,  140,   67,  134,   90,   73,  135,  141,   86,
+      672,   87,   74,   91,   75,   67,   88,  101,   92,   89,
+      178,  102,  103,  104,   93,  105,  179,   67,   42,   42,
+
+       42,   42,   42,   42,  106,  189,  107,   40,  122,  123,
+      123,  142,  126,  123,  669,  123,  136,  137,  123,  217,
+      124,  124,  123,  190,  147,  143,  123,  125,  125,  123,
+      218,  337,  144,  123,  122,  148,  184,  185,  149,  151,
+      152,  150,  670,  671,  338,  153,  186,  118,  119,   45,
+       46,   47,   48,  154,   50,   51,  123,  162,   52,   53,
+       54,   55,   56,   57,  120,   59,   60,  668,  155,  121,
+      156,  286,  667,  157,  158,  163,  163,  163,  163,  666,
+      287,  159,  164,  163,  165,  166,  167,  163,  163,  168,
+      169,  163,  163,  163,  171,  171,  171,  171,  171,  171,
+
+      230,  665,  664,  260,  172,   65,   66,   66,   66,   66,
+       66,   66,  198,  261,  154,  173,   67,  174,  174,  174,
+      174,  174,  174,  233,  233,  364,  349,  257,  365,  233,
+      172,  199,  231,  258,  232,  232,  232,  232,  232,  232,
+      233,  350,   67,  233,  233,  236,  233,  233,  262,  233,
+      247,  233,  233,  353,  263,  273,  233,  663,  233,  233,
+      233,  428,  662,  233,  233,  274,  354,  233,  265,  233,
+      661,  264,  266,  267,  233,  233,  660,  429,  233,  278,
+      278,  278,  278,  524,  659,  233,  525,  658,  657,  233,
+      278,  278,  278,  278,  279,  278,  278,  280,  281,  278,
+
+      278,  278,  278,  278,  278,  278,  282,  278,  278,  278,
+      278,  278,  278,  278,   42,   42,   42,   42,   42,   42,
+      656,  655,  654,  283,  122,  284,  284,  284,  284,  284,
+      284,  174,  174,  174,  174,  174,  174,  174,  174,  174,
+      174,  174,  174,  232,  232,  232,  232,  232,  232,  653,
+      122,  232,  232,  232,  232,  232,  232,  335,  335,  335,
+      335,  335,  335,  335,  374,  335,  375,  335,  376,  335,
+      335,  367,  335,  652,  335,  335,  335,  335,  335,  651,
+      650,  377,  380,  380,  380,  380,  335,  649,  335,  380,
+      380,  380,  380,  381,  380,  380,  380,  380,  380,  380,
+
+      380,  380,  380,  380,  380,  284,  284,  284,  284,  284,
+      284,  284,  284,  284,  284,  284,  284,  471,  472,  576,
+      577,  648,  647,  646,  645,  644,  643,  642,  641,  640,
+      639,  638,  637,  636,  635,  634,  633,  632,  631,  473,
+      578,   37,   37,   37,  170,  170,  630,  629,  628,  627,
+      626,  625,  624,  623,  622,  621,  620,  619,  618,  617,
+      616,  615,  614,  613,  612,  611,  610,  609,  608,  607,
+      606,  605,  604,  603,  602,  601,  600,  599,  598,  597,
+      596,  595,  594,  593,  592,  591,  590,  589,  588,  587,
+      586,  585,  584,  583,  582,  581,  580,  579,  575,  574,
+
+      573,  572,  571,  570,  569,  568,  567,  566,  565,  564,
+      563,  562,  561,  560,  559,  558,  557,  556,  555,  554,
+      553,  552,  551,  550,  549,  548,  547,  546,  545,  544,
+      543,  542,  541,  540,  539,  538,  537,  536,  535,  534,
+      533,  532,  531,  530,  529,  528,  527,  526,  523,  522,
+      521,  520,  519,  518,  517,  516,  515,  514,  513,  512,
+      511,  510,  509,  508,  507,  506,  505,  504,  503,  502,
+      501,  500,  499,  498,  497,  496,  495,  494,  493,  492,
+      491,  490,  489,  488,  487,  486,  485,  484,  483,  482,
+      481,  480,  479,  478,  477,  476,  475,  474,  470,  469,
+
+      468,  467,  466,  465,  464,  463,  462,  461,  460,  459,
+      458,  457,  456,  455,  454,  453,  452,  451,  450,  449,
+      448,  447,  446,  445,  444,  443,  442,  441,  440,  439,
+      438,  437,  436,  435,  434,  433,  432,  431,  430,  427,
+      426,  425,  424,  423,  422,  421,  420,  419,  418,  417,
+      416,  415,  414,  413,  412,  411,  410,  409,  408,  407,
+      406,  405,  404,  403,  402,  401,  400,  399,  398,  397,
+      396,  395,  394,  393,  392,  391,  390,  389,  388,  387,
+      386,  385,  384,  383,  382,  379,  378,  373,  372,  371,
+      370,  369,  368,  366,  363,  362,  361,  360,  359,  358,
+
+      357,  356,  355,  352,  351,  348,  347,  346,  345,  344,
+      343,  342,  341,  340,  339,  336,  264,  236,  334,  333,
+      332,  331,  330,  329,  328,  327,  326,  325,  324,  323,
+      322,  321,  320,  319,  318,  317,  316,  315,  314,  313,
+      312,  311,  310,  309,  308,  307,  306,  305,  304,  303,
+      302,  301,  300,  299,  298,  297,  296,  295,  294,  293,
+      292,  291,  290,  289,  288,  285,  277,  276,  275,  272,
+      271,  270,  269,  268,  259,  256,  255,  254,  253,  252,
+      251,  250,  249,  248,  246,  245,  244,  243,  242,  239,
+      238,  237,  235,  234,  162,  229,  228,  227,  226,  225,
+
+      224,  223,  222,  221,  216,  215,  214,  213,  212,  211,
+      210,  209,  208,  207,  206,  205,  204,  203,  202,  201,
+      200,  197,  196,  195,  194,  193,  192,  191,  188,  187,
+      183,  182,  181,  180,  177,  176,  175,  146,  145,  139,
+      138,   38,  117,  116,  115,  114,  113,  112,  111,  110,
+       85,   38,   36,  674,    3,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674
+    } ;
+
+static yyconst flex_int16_t yy_chk[1023] =
+    {   0,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    1,    1,    1,
+        1,    1,    1,    1,    1,    1,    1,    5,   17,    5,
+       10,   17,   46,   27,   24,   18,   46,   10,   10,   10,
+       10,   10,   10,   10,   16,   18,   16,   24,   25,  130,
+       25,   27,   16,   11,   11,   11,   11,   11,   11,   11,
+
+       19,  130,   23,   25,   11,   45,   19,   23,   45,  676,
+       10,   10,   10,   10,   10,   10,   10,   10,   10,   47,
+       47,   10,   10,   10,   10,   10,   10,   10,   10,   10,
+       11,   47,   10,   10,   10,   10,   12,   12,   12,   12,
+       12,   12,   12,   35,   60,   35,   12,   12,   13,   13,
+       13,   13,   13,   13,   13,  109,  671,   60,   13,   13,
+       14,   14,   14,   14,   14,   14,   14,   15,  109,   15,
+       14,   14,   52,   12,   48,   22,   15,   48,   52,   21,
+      670,   21,   15,   22,   15,   13,   21,   26,   22,   21,
+       74,   26,   26,   26,   22,   26,   74,   14,   42,   42,
+
+       42,   42,   42,   42,   26,   82,   26,   40,   42,   43,
+       43,   53,   44,   44,  664,   43,   49,   49,   44,  108,
+      118,   43,   49,   82,   56,   53,   43,  118,   43,   44,
+      108,  235,   53,   49,   42,   56,   79,   79,   56,   58,
+       58,   56,  667,  667,  235,   58,   79,   40,   40,   40,
+       40,   40,   40,   58,   40,   40,   58,   61,   40,   40,
+       40,   40,   40,   40,   40,   40,   40,  662,   59,   40,
+       59,  176,  660,   59,   59,   61,   61,   61,   61,  656,
+      176,   59,   62,   62,   62,   62,   63,   63,   63,   63,
+       64,   64,   64,   64,   65,   65,   65,   65,   65,   65,
+
+      120,  655,  652,  149,   65,   66,   66,   66,   66,   66,
+       66,   66,   90,  149,  120,   67,   66,   67,   67,   67,
+       67,   67,   67,  123,  123,  263,  249,  147,  263,  123,
+       65,   90,  122,  147,  122,  122,  122,  122,  122,  122,
+      123,  249,   66,  126,  126,  126,  136,  136,  150,  126,
+      137,  137,  136,  252,  150,  159,  137,  650,  151,  151,
+      126,  334,  648,  136,  151,  159,  252,  137,  152,  152,
+      645,  151,  153,  153,  152,  151,  644,  334,  153,  163,
+      163,  163,  163,  442,  643,  152,  442,  640,  639,  153,
+      164,  164,  164,  164,  165,  165,  165,  165,  166,  166,
+
+      166,  166,  167,  167,  167,  167,  168,  168,  168,  168,
+      169,  169,  169,  169,  171,  171,  171,  171,  171,  171,
+      638,  637,  635,  172,  171,  172,  172,  172,  172,  172,
+      172,  173,  173,  173,  173,  173,  173,  174,  174,  174,
+      174,  174,  174,  231,  231,  231,  231,  231,  231,  634,
+      171,  232,  232,  232,  232,  232,  232,  233,  233,  247,
+      247,  265,  265,  233,  275,  247,  275,  265,  275,  266,
+      266,  267,  267,  633,  233,  266,  247,  267,  265,  631,
+      630,  275,  278,  278,  278,  278,  266,  626,  267,  279,
+      279,  279,  279,  280,  280,  280,  280,  281,  281,  281,
+
+      281,  282,  282,  282,  282,  283,  283,  283,  283,  283,
+      283,  284,  284,  284,  284,  284,  284,  386,  386,  499,
+      499,  625,  624,  623,  621,  618,  617,  615,  614,  613,
+      612,  611,  610,  607,  606,  604,  603,  602,  601,  386,
+      499,  675,  675,  675,  677,  677,  600,  599,  597,  596,
+      595,  593,  591,  590,  587,  578,  577,  576,  553,  546,
+      545,  542,  541,  540,  539,  538,  537,  536,  535,  534,
+      533,  531,  528,  527,  526,  525,  524,  523,  522,  521,
+      520,  519,  518,  516,  515,  514,  513,  511,  509,  508,
+      507,  506,  505,  504,  503,  502,  501,  500,  498,  497,
+
+      496,  495,  494,  491,  490,  489,  488,  487,  486,  485,
+      484,  483,  482,  481,  480,  479,  478,  477,  476,  475,
+      473,  472,  471,  470,  469,  468,  467,  466,  465,  464,
+      463,  462,  461,  460,  459,  458,  456,  455,  454,  453,
+      452,  450,  449,  448,  447,  446,  445,  444,  441,  440,
+      438,  437,  436,  435,  432,  431,  430,  429,  428,  427,
+      426,  424,  423,  422,  421,  420,  419,  418,  417,  416,
+      415,  414,  413,  412,  411,  410,  409,  408,  406,  405,
+      404,  403,  402,  401,  400,  399,  398,  397,  396,  395,
+      394,  393,  392,  391,  390,  389,  388,  387,  384,  383,
+
+      382,  379,  378,  377,  376,  375,  374,  372,  370,  369,
+      368,  366,  365,  364,  363,  362,  361,  360,  359,  358,
+      357,  356,  355,  354,  353,  352,  351,  350,  349,  347,
+      346,  345,  344,  343,  342,  341,  338,  337,  336,  333,
+      332,  331,  330,  329,  328,  327,  326,  325,  324,  322,
+      321,  320,  319,  318,  317,  316,  315,  313,  312,  311,
+      310,  309,  308,  307,  306,  305,  304,  303,  302,  301,
+      300,  299,  298,  297,  296,  295,  294,  292,  291,  290,
+      289,  288,  287,  286,  285,  277,  276,  274,  273,  272,
+      271,  270,  269,  264,  262,  261,  260,  259,  258,  257,
+
+      256,  255,  253,  251,  250,  248,  246,  245,  242,  241,
+      240,  239,  238,  237,  236,  234,  230,  229,  228,  227,
+      226,  225,  224,  223,  222,  221,  220,  219,  218,  217,
+      216,  215,  214,  213,  212,  211,  210,  209,  208,  207,
+      206,  205,  204,  203,  202,  201,  200,  199,  198,  197,
+      195,  194,  192,  191,  189,  187,  186,  185,  184,  183,
+      182,  181,  180,  179,  177,  175,  162,  161,  160,  158,
+      157,  156,  155,  154,  148,  146,  145,  144,  143,  142,
+      141,  140,  139,  138,  135,  134,  133,  132,  131,  129,
+      128,  127,  125,  124,  121,  119,  117,  116,  115,  114,
+
+      113,  112,  111,  110,  107,  106,  105,  104,  103,  102,
+      101,  100,   99,   98,   97,   96,   95,   94,   93,   92,
+       91,   89,   88,   87,   86,   85,   84,   83,   81,   80,
+       78,   77,   76,   75,   73,   72,   71,   55,   54,   51,
+       50,   37,   36,   34,   33,   32,   31,   30,   29,   28,
+       20,    8,    7,    3,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674,  674,  674,  674,  674,  674,  674,  674,  674,
+      674,  674
+    } ;
+
+/* The intent behind this definition is that it'll catch
+ * any uses of REJECT which flex missed.
+ */
+#define REJECT reject_used_but_not_detected
+#define yymore() yymore_used_but_not_detected
+#define YY_MORE_ADJ 0
+#define YY_RESTORE_YY_MORE_OFFSET
+#line 1 "program_lexer.l"
+#line 2 "program_lexer.l"
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include "main/glheader.h"
+#include "prog_instruction.h"
+#include "prog_statevars.h"
+
+#include "program_parser.h"
+#include "program_parse.tab.h"
+
+#define require_ARB_vp (yyextra->mode == ARB_vertex)
+#define require_ARB_fp (yyextra->mode == ARB_fragment)
+#define require_shadow (yyextra->option.Shadow)
+#define require_rect   (yyextra->option.TexRect)
+#define require_texarray        (yyextra->option.TexArray)
+
+#define return_token_or_IDENTIFIER(condition, token)	\
+   do {							\
+      if (condition) {					\
+	 return token;					\
+      } else {						\
+	 yylval->string = strdup(yytext);		\
+	 return IDENTIFIER;				\
+      }							\
+   } while (0)
+
+#define return_token_or_DOT(condition, token)		\
+   do {							\
+      if (condition) {					\
+	 return token;					\
+      } else {						\
+	 yyless(1);					\
+	 return DOT;					\
+      }							\
+   } while (0)
+
+
+#define return_opcode(condition, token, opcode, sat)	\
+   do {							\
+      if (condition) {					\
+	 yylval->temp_inst.Opcode = OPCODE_ ## opcode;	\
+	 yylval->temp_inst.SaturateMode = SATURATE_ ## sat; \
+	 return token;					\
+      } else {						\
+	 yylval->string = strdup(yytext);		\
+	 return IDENTIFIER;				\
+      }							\
+   } while (0)
+
+#define SWIZZLE_INVAL  MAKE_SWIZZLE4(SWIZZLE_NIL, SWIZZLE_NIL, \
+				     SWIZZLE_NIL, SWIZZLE_NIL)
+
+static unsigned
+mask_from_char(char c)
+{
+   switch (c) {
+   case 'x':
+   case 'r':
+      return WRITEMASK_X;
+   case 'y':
+   case 'g':
+      return WRITEMASK_Y;
+   case 'z':
+   case 'b':
+      return WRITEMASK_Z;
+   case 'w':
+   case 'a':
+      return WRITEMASK_W;
+   }
+
+   return 0;
+}
+
+static unsigned
+swiz_from_char(char c)
+{
+   switch (c) {
+   case 'x':
+   case 'r':
+      return SWIZZLE_X;
+   case 'y':
+   case 'g':
+      return SWIZZLE_Y;
+   case 'z':
+   case 'b':
+      return SWIZZLE_Z;
+   case 'w':
+   case 'a':
+      return SWIZZLE_W;
+   }
+
+   return 0;
+}
+
+#define YY_USER_ACTION							\
+   do {									\
+      yylloc->first_column = yylloc->last_column;			\
+      yylloc->last_column += yyleng;					\
+      if ((yylloc->first_line == 1)					\
+	  && (yylloc->first_column == 1)) {				\
+	 yylloc->position = 1;						\
+      } else {								\
+	 yylloc->position += yylloc->last_column - yylloc->first_column; \
+      }									\
+   } while(0);
+
+#define YY_EXTRA_TYPE struct asm_parser_state *
+#line 1007 "lex.yy.c"
+
+#define INITIAL 0
+
+#ifndef YY_NO_UNISTD_H
+/* Special case for "unistd.h", since it is non-ANSI. We include it way
+ * down here because we want the user's section 1 to have been scanned first.
+ * The user has a chance to override it with an option.
+ */
+#include <unistd.h>
+#endif
+
+#ifndef YY_EXTRA_TYPE
+#define YY_EXTRA_TYPE void *
+#endif
+
+/* Holds the entire state of the reentrant scanner. */
+struct yyguts_t
+    {
+
+    /* User-defined. Not touched by flex. */
+    YY_EXTRA_TYPE yyextra_r;
+
+    /* The rest are the same as the globals declared in the non-reentrant scanner. */
+    FILE *yyin_r, *yyout_r;
+    size_t yy_buffer_stack_top; /**< index of top of stack. */
+    size_t yy_buffer_stack_max; /**< capacity of stack. */
+    YY_BUFFER_STATE * yy_buffer_stack; /**< Stack as an array. */
+    char yy_hold_char;
+    int yy_n_chars;
+    int yyleng_r;
+    char *yy_c_buf_p;
+    int yy_init;
+    int yy_start;
+    int yy_did_buffer_switch_on_eof;
+    int yy_start_stack_ptr;
+    int yy_start_stack_depth;
+    int *yy_start_stack;
+    yy_state_type yy_last_accepting_state;
+    char* yy_last_accepting_cpos;
+
+    int yylineno_r;
+    int yy_flex_debug_r;
+
+    char *yytext_r;
+    int yy_more_flag;
+    int yy_more_len;
+
+    YYSTYPE * yylval_r;
+
+    YYLTYPE * yylloc_r;
+
+    }; /* end struct yyguts_t */
+
+static int yy_init_globals (yyscan_t yyscanner );
+
+    /* This must go here because YYSTYPE and YYLTYPE are included
+     * from bison output in section 1.*/
+    #    define yylval yyg->yylval_r
+    
+    #    define yylloc yyg->yylloc_r
+    
+int yylex_init (yyscan_t* scanner);
+
+int yylex_init_extra (YY_EXTRA_TYPE user_defined,yyscan_t* scanner);
+
+/* Accessor methods to globals.
+   These are made visible to non-reentrant scanners for convenience. */
+
+int yylex_destroy (yyscan_t yyscanner );
+
+int yyget_debug (yyscan_t yyscanner );
+
+void yyset_debug (int debug_flag ,yyscan_t yyscanner );
+
+YY_EXTRA_TYPE yyget_extra (yyscan_t yyscanner );
+
+void yyset_extra (YY_EXTRA_TYPE user_defined ,yyscan_t yyscanner );
+
+FILE *yyget_in (yyscan_t yyscanner );
+
+void yyset_in  (FILE * in_str ,yyscan_t yyscanner );
+
+FILE *yyget_out (yyscan_t yyscanner );
+
+void yyset_out  (FILE * out_str ,yyscan_t yyscanner );
+
+int yyget_leng (yyscan_t yyscanner );
+
+char *yyget_text (yyscan_t yyscanner );
+
+int yyget_lineno (yyscan_t yyscanner );
+
+void yyset_lineno (int line_number ,yyscan_t yyscanner );
+
+YYSTYPE * yyget_lval (yyscan_t yyscanner );
+
+void yyset_lval (YYSTYPE * yylval_param ,yyscan_t yyscanner );
+
+       YYLTYPE *yyget_lloc (yyscan_t yyscanner );
+    
+        void yyset_lloc (YYLTYPE * yylloc_param ,yyscan_t yyscanner );
+    
+/* Macros after this point can all be overridden by user definitions in
+ * section 1.
+ */
+
+#ifndef YY_SKIP_YYWRAP
+#ifdef __cplusplus
+extern "C" int yywrap (yyscan_t yyscanner );
+#else
+extern int yywrap (yyscan_t yyscanner );
+#endif
+#endif
+
+    static void yyunput (int c,char *buf_ptr  ,yyscan_t yyscanner);
+    
+#ifndef yytext_ptr
+static void yy_flex_strncpy (char *,yyconst char *,int ,yyscan_t yyscanner);
+#endif
+
+#ifdef YY_NEED_STRLEN
+static int yy_flex_strlen (yyconst char * ,yyscan_t yyscanner);
+#endif
+
+#ifndef YY_NO_INPUT
+
+#ifdef __cplusplus
+static int yyinput (yyscan_t yyscanner );
+#else
+static int input (yyscan_t yyscanner );
+#endif
+
+#endif
+
+/* Amount of stuff to slurp up with each read. */
+#ifndef YY_READ_BUF_SIZE
+#define YY_READ_BUF_SIZE 8192
+#endif
+
+/* Copy whatever the last rule matched to the standard output. */
+#ifndef ECHO
+/* This used to be an fputs(), but since the string might contain NUL's,
+ * we now use fwrite().
+ */
+#define ECHO fwrite( yytext, yyleng, 1, yyout )
+#endif
+
+/* Gets input and stuffs it into "buf".  number of characters read, or YY_NULL,
+ * is returned in "result".
+ */
+#ifndef YY_INPUT
+#define YY_INPUT(buf,result,max_size) \
+	if ( YY_CURRENT_BUFFER_LVALUE->yy_is_interactive ) \
+		{ \
+		int c = '*'; \
+		unsigned n; \
+		for ( n = 0; n < max_size && \
+			     (c = getc( yyin )) != EOF && c != '\n'; ++n ) \
+			buf[n] = (char) c; \
+		if ( c == '\n' ) \
+			buf[n++] = (char) c; \
+		if ( c == EOF && ferror( yyin ) ) \
+			YY_FATAL_ERROR( "input in flex scanner failed" ); \
+		result = n; \
+		} \
+	else \
+		{ \
+		errno=0; \
+		while ( (result = fread(buf, 1, max_size, yyin))==0 && ferror(yyin)) \
+			{ \
+			if( errno != EINTR) \
+				{ \
+				YY_FATAL_ERROR( "input in flex scanner failed" ); \
+				break; \
+				} \
+			errno=0; \
+			clearerr(yyin); \
+			} \
+		}\
+\
+
+#endif
+
+/* No semi-colon after return; correct usage is to write "yyterminate();" -
+ * we don't want an extra ';' after the "return" because that will cause
+ * some compilers to complain about unreachable statements.
+ */
+#ifndef yyterminate
+#define yyterminate() return YY_NULL
+#endif
+
+/* Number of entries by which start-condition stack grows. */
+#ifndef YY_START_STACK_INCR
+#define YY_START_STACK_INCR 25
+#endif
+
+/* Report a fatal error. */
+#ifndef YY_FATAL_ERROR
+#define YY_FATAL_ERROR(msg) yy_fatal_error( msg , yyscanner)
+#endif
+
+/* end tables serialization structures and prototypes */
+
+/* Default declaration of generated scanner - a define so the user can
+ * easily add parameters.
+ */
+#ifndef YY_DECL
+#define YY_DECL_IS_OURS 1
+
+extern int yylex \
+               (YYSTYPE * yylval_param,YYLTYPE * yylloc_param ,yyscan_t yyscanner);
+
+#define YY_DECL int yylex \
+               (YYSTYPE * yylval_param, YYLTYPE * yylloc_param , yyscan_t yyscanner)
+#endif /* !YY_DECL */
+
+/* Code executed at the beginning of each rule, after yytext and yyleng
+ * have been set up.
+ */
+#ifndef YY_USER_ACTION
+#define YY_USER_ACTION
+#endif
+
+/* Code executed at the end of each rule. */
+#ifndef YY_BREAK
+#define YY_BREAK break;
+#endif
+
+#define YY_RULE_SETUP \
+	YY_USER_ACTION
+
+/** The main scanner function which does all the work.
+ */
+YY_DECL
+{
+	register yy_state_type yy_current_state;
+	register char *yy_cp, *yy_bp;
+	register int yy_act;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+#line 136 "program_lexer.l"
+
+
+#line 1251 "lex.yy.c"
+
+    yylval = yylval_param;
+
+    yylloc = yylloc_param;
+
+	if ( !yyg->yy_init )
+		{
+		yyg->yy_init = 1;
+
+#ifdef YY_USER_INIT
+		YY_USER_INIT;
+#endif
+
+		if ( ! yyg->yy_start )
+			yyg->yy_start = 1;	/* first start state */
+
+		if ( ! yyin )
+			yyin = stdin;
+
+		if ( ! yyout )
+			yyout = stdout;
+
+		if ( ! YY_CURRENT_BUFFER ) {
+			yyensure_buffer_stack (yyscanner);
+			YY_CURRENT_BUFFER_LVALUE =
+				yy_create_buffer(yyin,YY_BUF_SIZE ,yyscanner);
+		}
+
+		yy_load_buffer_state(yyscanner );
+		}
+
+	while ( 1 )		/* loops until end-of-file is reached */
+		{
+		yy_cp = yyg->yy_c_buf_p;
+
+		/* Support of yytext. */
+		*yy_cp = yyg->yy_hold_char;
+
+		/* yy_bp points to the position in yy_ch_buf of the start of
+		 * the current run.
+		 */
+		yy_bp = yy_cp;
+
+		yy_current_state = yyg->yy_start;
+yy_match:
+		do
+			{
+			register YY_CHAR yy_c = yy_ec[YY_SC_TO_UI(*yy_cp)];
+			if ( yy_accept[yy_current_state] )
+				{
+				yyg->yy_last_accepting_state = yy_current_state;
+				yyg->yy_last_accepting_cpos = yy_cp;
+				}
+			while ( yy_chk[yy_base[yy_current_state] + yy_c] != yy_current_state )
+				{
+				yy_current_state = (int) yy_def[yy_current_state];
+				if ( yy_current_state >= 675 )
+					yy_c = yy_meta[(unsigned int) yy_c];
+				}
+			yy_current_state = yy_nxt[yy_base[yy_current_state] + (unsigned int) yy_c];
+			++yy_cp;
+			}
+		while ( yy_base[yy_current_state] != 955 );
+
+yy_find_action:
+		yy_act = yy_accept[yy_current_state];
+		if ( yy_act == 0 )
+			{ /* have to back up */
+			yy_cp = yyg->yy_last_accepting_cpos;
+			yy_current_state = yyg->yy_last_accepting_state;
+			yy_act = yy_accept[yy_current_state];
+			}
+
+		YY_DO_BEFORE_ACTION;
+
+do_action:	/* This label is used only to access EOF actions. */
+
+		switch ( yy_act )
+	{ /* beginning of action switch */
+			case 0: /* must back up */
+			/* undo the effects of YY_DO_BEFORE_ACTION */
+			*yy_cp = yyg->yy_hold_char;
+			yy_cp = yyg->yy_last_accepting_cpos;
+			yy_current_state = yyg->yy_last_accepting_state;
+			goto yy_find_action;
+
+case 1:
+YY_RULE_SETUP
+#line 138 "program_lexer.l"
+{ return ARBvp_10; }
+	YY_BREAK
+case 2:
+YY_RULE_SETUP
+#line 139 "program_lexer.l"
+{ return ARBfp_10; }
+	YY_BREAK
+case 3:
+YY_RULE_SETUP
+#line 140 "program_lexer.l"
+{
+   yylval->integer = at_address;
+   return_token_or_IDENTIFIER(require_ARB_vp, ADDRESS);
+}
+	YY_BREAK
+case 4:
+YY_RULE_SETUP
+#line 144 "program_lexer.l"
+{ return ALIAS; }
+	YY_BREAK
+case 5:
+YY_RULE_SETUP
+#line 145 "program_lexer.l"
+{ return ATTRIB; }
+	YY_BREAK
+case 6:
+YY_RULE_SETUP
+#line 146 "program_lexer.l"
+{ return END; }
+	YY_BREAK
+case 7:
+YY_RULE_SETUP
+#line 147 "program_lexer.l"
+{ return OPTION; }
+	YY_BREAK
+case 8:
+YY_RULE_SETUP
+#line 148 "program_lexer.l"
+{ return OUTPUT; }
+	YY_BREAK
+case 9:
+YY_RULE_SETUP
+#line 149 "program_lexer.l"
+{ return PARAM; }
+	YY_BREAK
+case 10:
+YY_RULE_SETUP
+#line 150 "program_lexer.l"
+{ yylval->integer = at_temp; return TEMP; }
+	YY_BREAK
+case 11:
+YY_RULE_SETUP
+#line 152 "program_lexer.l"
+{ return_opcode(             1, VECTOR_OP, ABS, OFF); }
+	YY_BREAK
+case 12:
+YY_RULE_SETUP
+#line 153 "program_lexer.l"
+{ return_opcode(require_ARB_fp, VECTOR_OP, ABS, ZERO_ONE); }
+	YY_BREAK
+case 13:
+YY_RULE_SETUP
+#line 154 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, ADD, OFF); }
+	YY_BREAK
+case 14:
+YY_RULE_SETUP
+#line 155 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, ADD, ZERO_ONE); }
+	YY_BREAK
+case 15:
+YY_RULE_SETUP
+#line 156 "program_lexer.l"
+{ return_opcode(require_ARB_vp, ARL, ARL, OFF); }
+	YY_BREAK
+case 16:
+YY_RULE_SETUP
+#line 158 "program_lexer.l"
+{ return_opcode(require_ARB_fp, TRI_OP, CMP, OFF); }
+	YY_BREAK
+case 17:
+YY_RULE_SETUP
+#line 159 "program_lexer.l"
+{ return_opcode(require_ARB_fp, TRI_OP, CMP, ZERO_ONE); }
+	YY_BREAK
+case 18:
+YY_RULE_SETUP
+#line 160 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, COS, OFF); }
+	YY_BREAK
+case 19:
+YY_RULE_SETUP
+#line 161 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, COS, ZERO_ONE); }
+	YY_BREAK
+case 20:
+YY_RULE_SETUP
+#line 163 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, DP3, OFF); }
+	YY_BREAK
+case 21:
+YY_RULE_SETUP
+#line 164 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, DP3, ZERO_ONE); }
+	YY_BREAK
+case 22:
+YY_RULE_SETUP
+#line 165 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, DP4, OFF); }
+	YY_BREAK
+case 23:
+YY_RULE_SETUP
+#line 166 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, DP4, ZERO_ONE); }
+	YY_BREAK
+case 24:
+YY_RULE_SETUP
+#line 167 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, DPH, OFF); }
+	YY_BREAK
+case 25:
+YY_RULE_SETUP
+#line 168 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, DPH, ZERO_ONE); }
+	YY_BREAK
+case 26:
+YY_RULE_SETUP
+#line 169 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, DST, OFF); }
+	YY_BREAK
+case 27:
+YY_RULE_SETUP
+#line 170 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, DST, ZERO_ONE); }
+	YY_BREAK
+case 28:
+YY_RULE_SETUP
+#line 172 "program_lexer.l"
+{ return_opcode(             1, SCALAR_OP, EX2, OFF); }
+	YY_BREAK
+case 29:
+YY_RULE_SETUP
+#line 173 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, EX2, ZERO_ONE); }
+	YY_BREAK
+case 30:
+YY_RULE_SETUP
+#line 174 "program_lexer.l"
+{ return_opcode(require_ARB_vp, SCALAR_OP, EXP, OFF); }
+	YY_BREAK
+case 31:
+YY_RULE_SETUP
+#line 176 "program_lexer.l"
+{ return_opcode(             1, VECTOR_OP, FLR, OFF); }
+	YY_BREAK
+case 32:
+YY_RULE_SETUP
+#line 177 "program_lexer.l"
+{ return_opcode(require_ARB_fp, VECTOR_OP, FLR, ZERO_ONE); }
+	YY_BREAK
+case 33:
+YY_RULE_SETUP
+#line 178 "program_lexer.l"
+{ return_opcode(             1, VECTOR_OP, FRC, OFF); }
+	YY_BREAK
+case 34:
+YY_RULE_SETUP
+#line 179 "program_lexer.l"
+{ return_opcode(require_ARB_fp, VECTOR_OP, FRC, ZERO_ONE); }
+	YY_BREAK
+case 35:
+YY_RULE_SETUP
+#line 181 "program_lexer.l"
+{ return_opcode(require_ARB_fp, KIL, KIL, OFF); }
+	YY_BREAK
+case 36:
+YY_RULE_SETUP
+#line 183 "program_lexer.l"
+{ return_opcode(             1, VECTOR_OP, LIT, OFF); }
+	YY_BREAK
+case 37:
+YY_RULE_SETUP
+#line 184 "program_lexer.l"
+{ return_opcode(require_ARB_fp, VECTOR_OP, LIT, ZERO_ONE); }
+	YY_BREAK
+case 38:
+YY_RULE_SETUP
+#line 185 "program_lexer.l"
+{ return_opcode(             1, SCALAR_OP, LG2, OFF); }
+	YY_BREAK
+case 39:
+YY_RULE_SETUP
+#line 186 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, LG2, ZERO_ONE); }
+	YY_BREAK
+case 40:
+YY_RULE_SETUP
+#line 187 "program_lexer.l"
+{ return_opcode(require_ARB_vp, SCALAR_OP, LOG, OFF); }
+	YY_BREAK
+case 41:
+YY_RULE_SETUP
+#line 188 "program_lexer.l"
+{ return_opcode(require_ARB_fp, TRI_OP, LRP, OFF); }
+	YY_BREAK
+case 42:
+YY_RULE_SETUP
+#line 189 "program_lexer.l"
+{ return_opcode(require_ARB_fp, TRI_OP, LRP, ZERO_ONE); }
+	YY_BREAK
+case 43:
+YY_RULE_SETUP
+#line 191 "program_lexer.l"
+{ return_opcode(             1, TRI_OP, MAD, OFF); }
+	YY_BREAK
+case 44:
+YY_RULE_SETUP
+#line 192 "program_lexer.l"
+{ return_opcode(require_ARB_fp, TRI_OP, MAD, ZERO_ONE); }
+	YY_BREAK
+case 45:
+YY_RULE_SETUP
+#line 193 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, MAX, OFF); }
+	YY_BREAK
+case 46:
+YY_RULE_SETUP
+#line 194 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, MAX, ZERO_ONE); }
+	YY_BREAK
+case 47:
+YY_RULE_SETUP
+#line 195 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, MIN, OFF); }
+	YY_BREAK
+case 48:
+YY_RULE_SETUP
+#line 196 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, MIN, ZERO_ONE); }
+	YY_BREAK
+case 49:
+YY_RULE_SETUP
+#line 197 "program_lexer.l"
+{ return_opcode(             1, VECTOR_OP, MOV, OFF); }
+	YY_BREAK
+case 50:
+YY_RULE_SETUP
+#line 198 "program_lexer.l"
+{ return_opcode(require_ARB_fp, VECTOR_OP, MOV, ZERO_ONE); }
+	YY_BREAK
+case 51:
+YY_RULE_SETUP
+#line 199 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, MUL, OFF); }
+	YY_BREAK
+case 52:
+YY_RULE_SETUP
+#line 200 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, MUL, ZERO_ONE); }
+	YY_BREAK
+case 53:
+YY_RULE_SETUP
+#line 202 "program_lexer.l"
+{ return_opcode(             1, BINSC_OP, POW, OFF); }
+	YY_BREAK
+case 54:
+YY_RULE_SETUP
+#line 203 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BINSC_OP, POW, ZERO_ONE); }
+	YY_BREAK
+case 55:
+YY_RULE_SETUP
+#line 205 "program_lexer.l"
+{ return_opcode(             1, SCALAR_OP, RCP, OFF); }
+	YY_BREAK
+case 56:
+YY_RULE_SETUP
+#line 206 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, RCP, ZERO_ONE); }
+	YY_BREAK
+case 57:
+YY_RULE_SETUP
+#line 207 "program_lexer.l"
+{ return_opcode(             1, SCALAR_OP, RSQ, OFF); }
+	YY_BREAK
+case 58:
+YY_RULE_SETUP
+#line 208 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, RSQ, ZERO_ONE); }
+	YY_BREAK
+case 59:
+YY_RULE_SETUP
+#line 210 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, SCS, OFF); }
+	YY_BREAK
+case 60:
+YY_RULE_SETUP
+#line 211 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, SCS, ZERO_ONE); }
+	YY_BREAK
+case 61:
+YY_RULE_SETUP
+#line 212 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, SGE, OFF); }
+	YY_BREAK
+case 62:
+YY_RULE_SETUP
+#line 213 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, SGE, ZERO_ONE); }
+	YY_BREAK
+case 63:
+YY_RULE_SETUP
+#line 214 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, SIN, OFF); }
+	YY_BREAK
+case 64:
+YY_RULE_SETUP
+#line 215 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SCALAR_OP, SIN, ZERO_ONE); }
+	YY_BREAK
+case 65:
+YY_RULE_SETUP
+#line 216 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, SLT, OFF); }
+	YY_BREAK
+case 66:
+YY_RULE_SETUP
+#line 217 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, SLT, ZERO_ONE); }
+	YY_BREAK
+case 67:
+YY_RULE_SETUP
+#line 218 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, SUB, OFF); }
+	YY_BREAK
+case 68:
+YY_RULE_SETUP
+#line 219 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, SUB, ZERO_ONE); }
+	YY_BREAK
+case 69:
+YY_RULE_SETUP
+#line 220 "program_lexer.l"
+{ return_opcode(             1, SWZ, SWZ, OFF); }
+	YY_BREAK
+case 70:
+YY_RULE_SETUP
+#line 221 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SWZ, SWZ, ZERO_ONE); }
+	YY_BREAK
+case 71:
+YY_RULE_SETUP
+#line 223 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TEX, OFF); }
+	YY_BREAK
+case 72:
+YY_RULE_SETUP
+#line 224 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TEX, ZERO_ONE); }
+	YY_BREAK
+case 73:
+YY_RULE_SETUP
+#line 225 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TXB, OFF); }
+	YY_BREAK
+case 74:
+YY_RULE_SETUP
+#line 226 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TXB, ZERO_ONE); }
+	YY_BREAK
+case 75:
+YY_RULE_SETUP
+#line 227 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TXP, OFF); }
+	YY_BREAK
+case 76:
+YY_RULE_SETUP
+#line 228 "program_lexer.l"
+{ return_opcode(require_ARB_fp, SAMPLE_OP, TXP, ZERO_ONE); }
+	YY_BREAK
+case 77:
+YY_RULE_SETUP
+#line 230 "program_lexer.l"
+{ return_opcode(             1, BIN_OP, XPD, OFF); }
+	YY_BREAK
+case 78:
+YY_RULE_SETUP
+#line 231 "program_lexer.l"
+{ return_opcode(require_ARB_fp, BIN_OP, XPD, ZERO_ONE); }
+	YY_BREAK
+case 79:
+YY_RULE_SETUP
+#line 233 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_vp, VERTEX); }
+	YY_BREAK
+case 80:
+YY_RULE_SETUP
+#line 234 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, FRAGMENT); }
+	YY_BREAK
+case 81:
+YY_RULE_SETUP
+#line 235 "program_lexer.l"
+{ return PROGRAM; }
+	YY_BREAK
+case 82:
+YY_RULE_SETUP
+#line 236 "program_lexer.l"
+{ return STATE; }
+	YY_BREAK
+case 83:
+YY_RULE_SETUP
+#line 237 "program_lexer.l"
+{ return RESULT; }
+	YY_BREAK
+case 84:
+YY_RULE_SETUP
+#line 239 "program_lexer.l"
+{ return AMBIENT; }
+	YY_BREAK
+case 85:
+YY_RULE_SETUP
+#line 240 "program_lexer.l"
+{ return ATTENUATION; }
+	YY_BREAK
+case 86:
+YY_RULE_SETUP
+#line 241 "program_lexer.l"
+{ return BACK; }
+	YY_BREAK
+case 87:
+YY_RULE_SETUP
+#line 242 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, CLIP); }
+	YY_BREAK
+case 88:
+YY_RULE_SETUP
+#line 243 "program_lexer.l"
+{ return COLOR; }
+	YY_BREAK
+case 89:
+YY_RULE_SETUP
+#line 244 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_fp, DEPTH); }
+	YY_BREAK
+case 90:
+YY_RULE_SETUP
+#line 245 "program_lexer.l"
+{ return DIFFUSE; }
+	YY_BREAK
+case 91:
+YY_RULE_SETUP
+#line 246 "program_lexer.l"
+{ return DIRECTION; }
+	YY_BREAK
+case 92:
+YY_RULE_SETUP
+#line 247 "program_lexer.l"
+{ return EMISSION; }
+	YY_BREAK
+case 93:
+YY_RULE_SETUP
+#line 248 "program_lexer.l"
+{ return ENV; }
+	YY_BREAK
+case 94:
+YY_RULE_SETUP
+#line 249 "program_lexer.l"
+{ return EYE; }
+	YY_BREAK
+case 95:
+YY_RULE_SETUP
+#line 250 "program_lexer.l"
+{ return FOGCOORD; }
+	YY_BREAK
+case 96:
+YY_RULE_SETUP
+#line 251 "program_lexer.l"
+{ return FOG; }
+	YY_BREAK
+case 97:
+YY_RULE_SETUP
+#line 252 "program_lexer.l"
+{ return FRONT; }
+	YY_BREAK
+case 98:
+YY_RULE_SETUP
+#line 253 "program_lexer.l"
+{ return HALF; }
+	YY_BREAK
+case 99:
+YY_RULE_SETUP
+#line 254 "program_lexer.l"
+{ return INVERSE; }
+	YY_BREAK
+case 100:
+YY_RULE_SETUP
+#line 255 "program_lexer.l"
+{ return INVTRANS; }
+	YY_BREAK
+case 101:
+YY_RULE_SETUP
+#line 256 "program_lexer.l"
+{ return LIGHT; }
+	YY_BREAK
+case 102:
+YY_RULE_SETUP
+#line 257 "program_lexer.l"
+{ return LIGHTMODEL; }
+	YY_BREAK
+case 103:
+YY_RULE_SETUP
+#line 258 "program_lexer.l"
+{ return LIGHTPROD; }
+	YY_BREAK
+case 104:
+YY_RULE_SETUP
+#line 259 "program_lexer.l"
+{ return LOCAL; }
+	YY_BREAK
+case 105:
+YY_RULE_SETUP
+#line 260 "program_lexer.l"
+{ return MATERIAL; }
+	YY_BREAK
+case 106:
+YY_RULE_SETUP
+#line 261 "program_lexer.l"
+{ return MAT_PROGRAM; }
+	YY_BREAK
+case 107:
+YY_RULE_SETUP
+#line 262 "program_lexer.l"
+{ return MATRIX; }
+	YY_BREAK
+case 108:
+YY_RULE_SETUP
+#line 263 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, MATRIXINDEX); }
+	YY_BREAK
+case 109:
+YY_RULE_SETUP
+#line 264 "program_lexer.l"
+{ return MODELVIEW; }
+	YY_BREAK
+case 110:
+YY_RULE_SETUP
+#line 265 "program_lexer.l"
+{ return MVP; }
+	YY_BREAK
+case 111:
+YY_RULE_SETUP
+#line 266 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, NORMAL); }
+	YY_BREAK
+case 112:
+YY_RULE_SETUP
+#line 267 "program_lexer.l"
+{ return OBJECT; }
+	YY_BREAK
+case 113:
+YY_RULE_SETUP
+#line 268 "program_lexer.l"
+{ return PALETTE; }
+	YY_BREAK
+case 114:
+YY_RULE_SETUP
+#line 269 "program_lexer.l"
+{ return PARAMS; }
+	YY_BREAK
+case 115:
+YY_RULE_SETUP
+#line 270 "program_lexer.l"
+{ return PLANE; }
+	YY_BREAK
+case 116:
+YY_RULE_SETUP
+#line 271 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, POINT_TOK); }
+	YY_BREAK
+case 117:
+YY_RULE_SETUP
+#line 272 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, POINTSIZE); }
+	YY_BREAK
+case 118:
+YY_RULE_SETUP
+#line 273 "program_lexer.l"
+{ return POSITION; }
+	YY_BREAK
+case 119:
+YY_RULE_SETUP
+#line 274 "program_lexer.l"
+{ return PRIMARY; }
+	YY_BREAK
+case 120:
+YY_RULE_SETUP
+#line 275 "program_lexer.l"
+{ return PROJECTION; }
+	YY_BREAK
+case 121:
+YY_RULE_SETUP
+#line 276 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_fp, RANGE); }
+	YY_BREAK
+case 122:
+YY_RULE_SETUP
+#line 277 "program_lexer.l"
+{ return ROW; }
+	YY_BREAK
+case 123:
+YY_RULE_SETUP
+#line 278 "program_lexer.l"
+{ return SCENECOLOR; }
+	YY_BREAK
+case 124:
+YY_RULE_SETUP
+#line 279 "program_lexer.l"
+{ return SECONDARY; }
+	YY_BREAK
+case 125:
+YY_RULE_SETUP
+#line 280 "program_lexer.l"
+{ return SHININESS; }
+	YY_BREAK
+case 126:
+YY_RULE_SETUP
+#line 281 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, SIZE_TOK); }
+	YY_BREAK
+case 127:
+YY_RULE_SETUP
+#line 282 "program_lexer.l"
+{ return SPECULAR; }
+	YY_BREAK
+case 128:
+YY_RULE_SETUP
+#line 283 "program_lexer.l"
+{ return SPOT; }
+	YY_BREAK
+case 129:
+YY_RULE_SETUP
+#line 284 "program_lexer.l"
+{ return TEXCOORD; }
+	YY_BREAK
+case 130:
+YY_RULE_SETUP
+#line 285 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_fp, TEXENV); }
+	YY_BREAK
+case 131:
+YY_RULE_SETUP
+#line 286 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, TEXGEN); }
+	YY_BREAK
+case 132:
+YY_RULE_SETUP
+#line 287 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, TEXGEN_Q); }
+	YY_BREAK
+case 133:
+YY_RULE_SETUP
+#line 288 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, TEXGEN_S); }
+	YY_BREAK
+case 134:
+YY_RULE_SETUP
+#line 289 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, TEXGEN_T); }
+	YY_BREAK
+case 135:
+YY_RULE_SETUP
+#line 290 "program_lexer.l"
+{ return TEXTURE; }
+	YY_BREAK
+case 136:
+YY_RULE_SETUP
+#line 291 "program_lexer.l"
+{ return TRANSPOSE; }
+	YY_BREAK
+case 137:
+YY_RULE_SETUP
+#line 292 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, VTXATTRIB); }
+	YY_BREAK
+case 138:
+YY_RULE_SETUP
+#line 293 "program_lexer.l"
+{ return_token_or_DOT(require_ARB_vp, WEIGHT); }
+	YY_BREAK
+case 139:
+YY_RULE_SETUP
+#line 295 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, TEXTURE_UNIT); }
+	YY_BREAK
+case 140:
+YY_RULE_SETUP
+#line 296 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, TEX_1D); }
+	YY_BREAK
+case 141:
+YY_RULE_SETUP
+#line 297 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, TEX_2D); }
+	YY_BREAK
+case 142:
+YY_RULE_SETUP
+#line 298 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, TEX_3D); }
+	YY_BREAK
+case 143:
+YY_RULE_SETUP
+#line 299 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp, TEX_CUBE); }
+	YY_BREAK
+case 144:
+YY_RULE_SETUP
+#line 300 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_rect, TEX_RECT); }
+	YY_BREAK
+case 145:
+YY_RULE_SETUP
+#line 301 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_shadow, TEX_SHADOW1D); }
+	YY_BREAK
+case 146:
+YY_RULE_SETUP
+#line 302 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_shadow, TEX_SHADOW2D); }
+	YY_BREAK
+case 147:
+YY_RULE_SETUP
+#line 303 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_rect, TEX_SHADOWRECT); }
+	YY_BREAK
+case 148:
+YY_RULE_SETUP
+#line 304 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_texarray, TEX_ARRAY1D); }
+	YY_BREAK
+case 149:
+YY_RULE_SETUP
+#line 305 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_texarray, TEX_ARRAY2D); }
+	YY_BREAK
+case 150:
+YY_RULE_SETUP
+#line 306 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_texarray, TEX_ARRAYSHADOW1D); }
+	YY_BREAK
+case 151:
+YY_RULE_SETUP
+#line 307 "program_lexer.l"
+{ return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_texarray, TEX_ARRAYSHADOW2D); }
+	YY_BREAK
+case 152:
+YY_RULE_SETUP
+#line 309 "program_lexer.l"
+{
+   yylval->string = strdup(yytext);
+   return IDENTIFIER;
+}
+	YY_BREAK
+case 153:
+YY_RULE_SETUP
+#line 314 "program_lexer.l"
+{ return DOT_DOT; }
+	YY_BREAK
+case 154:
+YY_RULE_SETUP
+#line 316 "program_lexer.l"
+{
+   yylval->integer = strtol(yytext, NULL, 10);
+   return INTEGER;
+}
+	YY_BREAK
+case 155:
+YY_RULE_SETUP
+#line 320 "program_lexer.l"
+{
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+	YY_BREAK
+case 156:
+/* rule 156 can match eol */
+*yy_cp = yyg->yy_hold_char; /* undo effects of setting up yytext */
+yyg->yy_c_buf_p = yy_cp -= 1;
+YY_DO_BEFORE_ACTION; /* set up yytext again */
+YY_RULE_SETUP
+#line 324 "program_lexer.l"
+{
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+	YY_BREAK
+case 157:
+YY_RULE_SETUP
+#line 328 "program_lexer.l"
+{
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+	YY_BREAK
+case 158:
+YY_RULE_SETUP
+#line 332 "program_lexer.l"
+{
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+	YY_BREAK
+case 159:
+YY_RULE_SETUP
+#line 337 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_NOOP;
+   yylval->swiz_mask.mask = WRITEMASK_XYZW;
+   return MASK4;
+}
+	YY_BREAK
+case 160:
+YY_RULE_SETUP
+#line 343 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XY
+      | mask_from_char(yytext[3]);
+   return MASK3;
+}
+	YY_BREAK
+case 161:
+YY_RULE_SETUP
+#line 349 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XZW;
+   return MASK3;
+}
+	YY_BREAK
+case 162:
+YY_RULE_SETUP
+#line 354 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_YZW;
+   return MASK3;
+}
+	YY_BREAK
+case 163:
+YY_RULE_SETUP
+#line 360 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_X
+      | mask_from_char(yytext[2]);
+   return MASK2;
+}
+	YY_BREAK
+case 164:
+YY_RULE_SETUP
+#line 366 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_Y
+      | mask_from_char(yytext[2]);
+   return MASK2;
+}
+	YY_BREAK
+case 165:
+YY_RULE_SETUP
+#line 372 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_ZW;
+   return MASK2;
+}
+	YY_BREAK
+case 166:
+YY_RULE_SETUP
+#line 378 "program_lexer.l"
+{
+   const unsigned s = swiz_from_char(yytext[1]);
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(s, s, s, s);
+   yylval->swiz_mask.mask = mask_from_char(yytext[1]);
+   return MASK1; 
+}
+	YY_BREAK
+case 167:
+YY_RULE_SETUP
+#line 385 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(swiz_from_char(yytext[1]),
+					    swiz_from_char(yytext[2]),
+					    swiz_from_char(yytext[3]),
+					    swiz_from_char(yytext[4]));
+   yylval->swiz_mask.mask = 0;
+   return SWIZZLE;
+}
+	YY_BREAK
+case 168:
+YY_RULE_SETUP
+#line 394 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_NOOP;
+   yylval->swiz_mask.mask = WRITEMASK_XYZW;
+   return_token_or_DOT(require_ARB_fp, MASK4);
+}
+	YY_BREAK
+case 169:
+YY_RULE_SETUP
+#line 400 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XY
+      | mask_from_char(yytext[3]);
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+	YY_BREAK
+case 170:
+YY_RULE_SETUP
+#line 406 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XZW;
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+	YY_BREAK
+case 171:
+YY_RULE_SETUP
+#line 411 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_YZW;
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+	YY_BREAK
+case 172:
+YY_RULE_SETUP
+#line 417 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_X
+      | mask_from_char(yytext[2]);
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+	YY_BREAK
+case 173:
+YY_RULE_SETUP
+#line 423 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_Y
+      | mask_from_char(yytext[2]);
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+	YY_BREAK
+case 174:
+YY_RULE_SETUP
+#line 429 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_ZW;
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+	YY_BREAK
+case 175:
+YY_RULE_SETUP
+#line 435 "program_lexer.l"
+{
+   const unsigned s = swiz_from_char(yytext[1]);
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(s, s, s, s);
+   yylval->swiz_mask.mask = mask_from_char(yytext[1]);
+   return_token_or_DOT(require_ARB_fp, MASK1);
+}
+	YY_BREAK
+case 176:
+YY_RULE_SETUP
+#line 443 "program_lexer.l"
+{
+   if (require_ARB_vp) {
+      return TEXGEN_R;
+   } else {
+      yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X,
+						SWIZZLE_X, SWIZZLE_X);
+      yylval->swiz_mask.mask = WRITEMASK_X;
+      return MASK1;
+   }
+}
+	YY_BREAK
+case 177:
+YY_RULE_SETUP
+#line 454 "program_lexer.l"
+{
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(swiz_from_char(yytext[1]),
+					    swiz_from_char(yytext[2]),
+					    swiz_from_char(yytext[3]),
+					    swiz_from_char(yytext[4]));
+   yylval->swiz_mask.mask = 0;
+   return_token_or_DOT(require_ARB_fp, SWIZZLE);
+}
+	YY_BREAK
+case 178:
+YY_RULE_SETUP
+#line 463 "program_lexer.l"
+{ return DOT; }
+	YY_BREAK
+case 179:
+/* rule 179 can match eol */
+YY_RULE_SETUP
+#line 465 "program_lexer.l"
+{
+   yylloc->first_line++;
+   yylloc->first_column = 1;
+   yylloc->last_line++;
+   yylloc->last_column = 1;
+   yylloc->position++;
+}
+	YY_BREAK
+case 180:
+YY_RULE_SETUP
+#line 472 "program_lexer.l"
+/* eat whitespace */ ;
+	YY_BREAK
+case 181:
+*yy_cp = yyg->yy_hold_char; /* undo effects of setting up yytext */
+yyg->yy_c_buf_p = yy_cp -= 1;
+YY_DO_BEFORE_ACTION; /* set up yytext again */
+YY_RULE_SETUP
+#line 473 "program_lexer.l"
+/* eat comments */ ;
+	YY_BREAK
+case 182:
+YY_RULE_SETUP
+#line 474 "program_lexer.l"
+{ return yytext[0]; }
+	YY_BREAK
+case 183:
+YY_RULE_SETUP
+#line 475 "program_lexer.l"
+ECHO;
+	YY_BREAK
+#line 2383 "lex.yy.c"
+case YY_STATE_EOF(INITIAL):
+	yyterminate();
+
+	case YY_END_OF_BUFFER:
+		{
+		/* Amount of text matched not including the EOB char. */
+		int yy_amount_of_matched_text = (int) (yy_cp - yyg->yytext_ptr) - 1;
+
+		/* Undo the effects of YY_DO_BEFORE_ACTION. */
+		*yy_cp = yyg->yy_hold_char;
+		YY_RESTORE_YY_MORE_OFFSET
+
+		if ( YY_CURRENT_BUFFER_LVALUE->yy_buffer_status == YY_BUFFER_NEW )
+			{
+			/* We're scanning a new file or input source.  It's
+			 * possible that this happened because the user
+			 * just pointed yyin at a new source and called
+			 * yylex().  If so, then we have to assure
+			 * consistency between YY_CURRENT_BUFFER and our
+			 * globals.  Here is the right place to do so, because
+			 * this is the first action (other than possibly a
+			 * back-up) that will match for the new input source.
+			 */
+			yyg->yy_n_chars = YY_CURRENT_BUFFER_LVALUE->yy_n_chars;
+			YY_CURRENT_BUFFER_LVALUE->yy_input_file = yyin;
+			YY_CURRENT_BUFFER_LVALUE->yy_buffer_status = YY_BUFFER_NORMAL;
+			}
+
+		/* Note that here we test for yy_c_buf_p "<=" to the position
+		 * of the first EOB in the buffer, since yy_c_buf_p will
+		 * already have been incremented past the NUL character
+		 * (since all states make transitions on EOB to the
+		 * end-of-buffer state).  Contrast this with the test
+		 * in input().
+		 */
+		if ( yyg->yy_c_buf_p <= &YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars] )
+			{ /* This was really a NUL. */
+			yy_state_type yy_next_state;
+
+			yyg->yy_c_buf_p = yyg->yytext_ptr + yy_amount_of_matched_text;
+
+			yy_current_state = yy_get_previous_state( yyscanner );
+
+			/* Okay, we're now positioned to make the NUL
+			 * transition.  We couldn't have
+			 * yy_get_previous_state() go ahead and do it
+			 * for us because it doesn't know how to deal
+			 * with the possibility of jamming (and we don't
+			 * want to build jamming into it because then it
+			 * will run more slowly).
+			 */
+
+			yy_next_state = yy_try_NUL_trans( yy_current_state , yyscanner);
+
+			yy_bp = yyg->yytext_ptr + YY_MORE_ADJ;
+
+			if ( yy_next_state )
+				{
+				/* Consume the NUL. */
+				yy_cp = ++yyg->yy_c_buf_p;
+				yy_current_state = yy_next_state;
+				goto yy_match;
+				}
+
+			else
+				{
+				yy_cp = yyg->yy_c_buf_p;
+				goto yy_find_action;
+				}
+			}
+
+		else switch ( yy_get_next_buffer( yyscanner ) )
+			{
+			case EOB_ACT_END_OF_FILE:
+				{
+				yyg->yy_did_buffer_switch_on_eof = 0;
+
+				if ( yywrap(yyscanner ) )
+					{
+					/* Note: because we've taken care in
+					 * yy_get_next_buffer() to have set up
+					 * yytext, we can now set up
+					 * yy_c_buf_p so that if some total
+					 * hoser (like flex itself) wants to
+					 * call the scanner after we return the
+					 * YY_NULL, it'll still work - another
+					 * YY_NULL will get returned.
+					 */
+					yyg->yy_c_buf_p = yyg->yytext_ptr + YY_MORE_ADJ;
+
+					yy_act = YY_STATE_EOF(YY_START);
+					goto do_action;
+					}
+
+				else
+					{
+					if ( ! yyg->yy_did_buffer_switch_on_eof )
+						YY_NEW_FILE;
+					}
+				break;
+				}
+
+			case EOB_ACT_CONTINUE_SCAN:
+				yyg->yy_c_buf_p =
+					yyg->yytext_ptr + yy_amount_of_matched_text;
+
+				yy_current_state = yy_get_previous_state( yyscanner );
+
+				yy_cp = yyg->yy_c_buf_p;
+				yy_bp = yyg->yytext_ptr + YY_MORE_ADJ;
+				goto yy_match;
+
+			case EOB_ACT_LAST_MATCH:
+				yyg->yy_c_buf_p =
+				&YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars];
+
+				yy_current_state = yy_get_previous_state( yyscanner );
+
+				yy_cp = yyg->yy_c_buf_p;
+				yy_bp = yyg->yytext_ptr + YY_MORE_ADJ;
+				goto yy_find_action;
+			}
+		break;
+		}
+
+	default:
+		YY_FATAL_ERROR(
+			"fatal flex scanner internal error--no action found" );
+	} /* end of action switch */
+		} /* end of scanning one token */
+} /* end of yylex */
+
+/* yy_get_next_buffer - try to read in a new buffer
+ *
+ * Returns a code representing an action:
+ *	EOB_ACT_LAST_MATCH -
+ *	EOB_ACT_CONTINUE_SCAN - continue scanning from current position
+ *	EOB_ACT_END_OF_FILE - end of file
+ */
+static int yy_get_next_buffer (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+	register char *dest = YY_CURRENT_BUFFER_LVALUE->yy_ch_buf;
+	register char *source = yyg->yytext_ptr;
+	register int number_to_move, i;
+	int ret_val;
+
+	if ( yyg->yy_c_buf_p > &YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars + 1] )
+		YY_FATAL_ERROR(
+		"fatal flex scanner internal error--end of buffer missed" );
+
+	if ( YY_CURRENT_BUFFER_LVALUE->yy_fill_buffer == 0 )
+		{ /* Don't try to fill the buffer, so this is an EOF. */
+		if ( yyg->yy_c_buf_p - yyg->yytext_ptr - YY_MORE_ADJ == 1 )
+			{
+			/* We matched a single character, the EOB, so
+			 * treat this as a final EOF.
+			 */
+			return EOB_ACT_END_OF_FILE;
+			}
+
+		else
+			{
+			/* We matched some text prior to the EOB, first
+			 * process it.
+			 */
+			return EOB_ACT_LAST_MATCH;
+			}
+		}
+
+	/* Try to read more data. */
+
+	/* First move last chars to start of buffer. */
+	number_to_move = (int) (yyg->yy_c_buf_p - yyg->yytext_ptr) - 1;
+
+	for ( i = 0; i < number_to_move; ++i )
+		*(dest++) = *(source++);
+
+	if ( YY_CURRENT_BUFFER_LVALUE->yy_buffer_status == YY_BUFFER_EOF_PENDING )
+		/* don't do the read, it's not guaranteed to return an EOF,
+		 * just force an EOF
+		 */
+		YY_CURRENT_BUFFER_LVALUE->yy_n_chars = yyg->yy_n_chars = 0;
+
+	else
+		{
+			int num_to_read =
+			YY_CURRENT_BUFFER_LVALUE->yy_buf_size - number_to_move - 1;
+
+		while ( num_to_read <= 0 )
+			{ /* Not enough room in the buffer - grow it. */
+
+			/* just a shorter name for the current buffer */
+			YY_BUFFER_STATE b = YY_CURRENT_BUFFER;
+
+			int yy_c_buf_p_offset =
+				(int) (yyg->yy_c_buf_p - b->yy_ch_buf);
+
+			if ( b->yy_is_our_buffer )
+				{
+				int new_size = b->yy_buf_size * 2;
+
+				if ( new_size <= 0 )
+					b->yy_buf_size += b->yy_buf_size / 8;
+				else
+					b->yy_buf_size *= 2;
+
+				b->yy_ch_buf = (char *)
+					/* Include room in for 2 EOB chars. */
+					yyrealloc((void *) b->yy_ch_buf,b->yy_buf_size + 2 ,yyscanner );
+				}
+			else
+				/* Can't grow it, we don't own it. */
+				b->yy_ch_buf = 0;
+
+			if ( ! b->yy_ch_buf )
+				YY_FATAL_ERROR(
+				"fatal error - scanner input buffer overflow" );
+
+			yyg->yy_c_buf_p = &b->yy_ch_buf[yy_c_buf_p_offset];
+
+			num_to_read = YY_CURRENT_BUFFER_LVALUE->yy_buf_size -
+						number_to_move - 1;
+
+			}
+
+		if ( num_to_read > YY_READ_BUF_SIZE )
+			num_to_read = YY_READ_BUF_SIZE;
+
+		/* Read in more data. */
+		YY_INPUT( (&YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[number_to_move]),
+			yyg->yy_n_chars, (size_t) num_to_read );
+
+		YY_CURRENT_BUFFER_LVALUE->yy_n_chars = yyg->yy_n_chars;
+		}
+
+	if ( yyg->yy_n_chars == 0 )
+		{
+		if ( number_to_move == YY_MORE_ADJ )
+			{
+			ret_val = EOB_ACT_END_OF_FILE;
+			yyrestart(yyin  ,yyscanner);
+			}
+
+		else
+			{
+			ret_val = EOB_ACT_LAST_MATCH;
+			YY_CURRENT_BUFFER_LVALUE->yy_buffer_status =
+				YY_BUFFER_EOF_PENDING;
+			}
+		}
+
+	else
+		ret_val = EOB_ACT_CONTINUE_SCAN;
+
+	if ((yy_size_t) (yyg->yy_n_chars + number_to_move) > YY_CURRENT_BUFFER_LVALUE->yy_buf_size) {
+		/* Extend the array by 50%, plus the number we really need. */
+		yy_size_t new_size = yyg->yy_n_chars + number_to_move + (yyg->yy_n_chars >> 1);
+		YY_CURRENT_BUFFER_LVALUE->yy_ch_buf = (char *) yyrealloc((void *) YY_CURRENT_BUFFER_LVALUE->yy_ch_buf,new_size ,yyscanner );
+		if ( ! YY_CURRENT_BUFFER_LVALUE->yy_ch_buf )
+			YY_FATAL_ERROR( "out of dynamic memory in yy_get_next_buffer()" );
+	}
+
+	yyg->yy_n_chars += number_to_move;
+	YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars] = YY_END_OF_BUFFER_CHAR;
+	YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars + 1] = YY_END_OF_BUFFER_CHAR;
+
+	yyg->yytext_ptr = &YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[0];
+
+	return ret_val;
+}
+
+/* yy_get_previous_state - get the state just before the EOB char was reached */
+
+    static yy_state_type yy_get_previous_state (yyscan_t yyscanner)
+{
+	register yy_state_type yy_current_state;
+	register char *yy_cp;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	yy_current_state = yyg->yy_start;
+
+	for ( yy_cp = yyg->yytext_ptr + YY_MORE_ADJ; yy_cp < yyg->yy_c_buf_p; ++yy_cp )
+		{
+		register YY_CHAR yy_c = (*yy_cp ? yy_ec[YY_SC_TO_UI(*yy_cp)] : 1);
+		if ( yy_accept[yy_current_state] )
+			{
+			yyg->yy_last_accepting_state = yy_current_state;
+			yyg->yy_last_accepting_cpos = yy_cp;
+			}
+		while ( yy_chk[yy_base[yy_current_state] + yy_c] != yy_current_state )
+			{
+			yy_current_state = (int) yy_def[yy_current_state];
+			if ( yy_current_state >= 675 )
+				yy_c = yy_meta[(unsigned int) yy_c];
+			}
+		yy_current_state = yy_nxt[yy_base[yy_current_state] + (unsigned int) yy_c];
+		}
+
+	return yy_current_state;
+}
+
+/* yy_try_NUL_trans - try to make a transition on the NUL character
+ *
+ * synopsis
+ *	next_state = yy_try_NUL_trans( current_state );
+ */
+    static yy_state_type yy_try_NUL_trans  (yy_state_type yy_current_state , yyscan_t yyscanner)
+{
+	register int yy_is_jam;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner; /* This var may be unused depending upon options. */
+	register char *yy_cp = yyg->yy_c_buf_p;
+
+	register YY_CHAR yy_c = 1;
+	if ( yy_accept[yy_current_state] )
+		{
+		yyg->yy_last_accepting_state = yy_current_state;
+		yyg->yy_last_accepting_cpos = yy_cp;
+		}
+	while ( yy_chk[yy_base[yy_current_state] + yy_c] != yy_current_state )
+		{
+		yy_current_state = (int) yy_def[yy_current_state];
+		if ( yy_current_state >= 675 )
+			yy_c = yy_meta[(unsigned int) yy_c];
+		}
+	yy_current_state = yy_nxt[yy_base[yy_current_state] + (unsigned int) yy_c];
+	yy_is_jam = (yy_current_state == 674);
+
+	return yy_is_jam ? 0 : yy_current_state;
+}
+
+    static void yyunput (int c, register char * yy_bp , yyscan_t yyscanner)
+{
+	register char *yy_cp;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+    yy_cp = yyg->yy_c_buf_p;
+
+	/* undo effects of setting up yytext */
+	*yy_cp = yyg->yy_hold_char;
+
+	if ( yy_cp < YY_CURRENT_BUFFER_LVALUE->yy_ch_buf + 2 )
+		{ /* need to shift things up to make room */
+		/* +2 for EOB chars. */
+		register int number_to_move = yyg->yy_n_chars + 2;
+		register char *dest = &YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[
+					YY_CURRENT_BUFFER_LVALUE->yy_buf_size + 2];
+		register char *source =
+				&YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[number_to_move];
+
+		while ( source > YY_CURRENT_BUFFER_LVALUE->yy_ch_buf )
+			*--dest = *--source;
+
+		yy_cp += (int) (dest - source);
+		yy_bp += (int) (dest - source);
+		YY_CURRENT_BUFFER_LVALUE->yy_n_chars =
+			yyg->yy_n_chars = YY_CURRENT_BUFFER_LVALUE->yy_buf_size;
+
+		if ( yy_cp < YY_CURRENT_BUFFER_LVALUE->yy_ch_buf + 2 )
+			YY_FATAL_ERROR( "flex scanner push-back overflow" );
+		}
+
+	*--yy_cp = (char) c;
+
+	yyg->yytext_ptr = yy_bp;
+	yyg->yy_hold_char = *yy_cp;
+	yyg->yy_c_buf_p = yy_cp;
+}
+
+#ifndef YY_NO_INPUT
+#ifdef __cplusplus
+    static int yyinput (yyscan_t yyscanner)
+#else
+    static int input  (yyscan_t yyscanner)
+#endif
+
+{
+	int c;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	*yyg->yy_c_buf_p = yyg->yy_hold_char;
+
+	if ( *yyg->yy_c_buf_p == YY_END_OF_BUFFER_CHAR )
+		{
+		/* yy_c_buf_p now points to the character we want to return.
+		 * If this occurs *before* the EOB characters, then it's a
+		 * valid NUL; if not, then we've hit the end of the buffer.
+		 */
+		if ( yyg->yy_c_buf_p < &YY_CURRENT_BUFFER_LVALUE->yy_ch_buf[yyg->yy_n_chars] )
+			/* This was really a NUL. */
+			*yyg->yy_c_buf_p = '\0';
+
+		else
+			{ /* need more input */
+			int offset = yyg->yy_c_buf_p - yyg->yytext_ptr;
+			++yyg->yy_c_buf_p;
+
+			switch ( yy_get_next_buffer( yyscanner ) )
+				{
+				case EOB_ACT_LAST_MATCH:
+					/* This happens because yy_g_n_b()
+					 * sees that we've accumulated a
+					 * token and flags that we need to
+					 * try matching the token before
+					 * proceeding.  But for input(),
+					 * there's no matching to consider.
+					 * So convert the EOB_ACT_LAST_MATCH
+					 * to EOB_ACT_END_OF_FILE.
+					 */
+
+					/* Reset buffer status. */
+					yyrestart(yyin ,yyscanner);
+
+					/*FALLTHROUGH*/
+
+				case EOB_ACT_END_OF_FILE:
+					{
+					if ( yywrap(yyscanner ) )
+						return EOF;
+
+					if ( ! yyg->yy_did_buffer_switch_on_eof )
+						YY_NEW_FILE;
+#ifdef __cplusplus
+					return yyinput(yyscanner);
+#else
+					return input(yyscanner);
+#endif
+					}
+
+				case EOB_ACT_CONTINUE_SCAN:
+					yyg->yy_c_buf_p = yyg->yytext_ptr + offset;
+					break;
+				}
+			}
+		}
+
+	c = *(unsigned char *) yyg->yy_c_buf_p;	/* cast for 8-bit char's */
+	*yyg->yy_c_buf_p = '\0';	/* preserve yytext */
+	yyg->yy_hold_char = *++yyg->yy_c_buf_p;
+
+	return c;
+}
+#endif	/* ifndef YY_NO_INPUT */
+
+/** Immediately switch to a different input stream.
+ * @param input_file A readable stream.
+ * @param yyscanner The scanner object.
+ * @note This function does not reset the start condition to @c INITIAL .
+ */
+    void yyrestart  (FILE * input_file , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	if ( ! YY_CURRENT_BUFFER ){
+        yyensure_buffer_stack (yyscanner);
+		YY_CURRENT_BUFFER_LVALUE =
+            yy_create_buffer(yyin,YY_BUF_SIZE ,yyscanner);
+	}
+
+	yy_init_buffer(YY_CURRENT_BUFFER,input_file ,yyscanner);
+	yy_load_buffer_state(yyscanner );
+}
+
+/** Switch to a different input buffer.
+ * @param new_buffer The new input buffer.
+ * @param yyscanner The scanner object.
+ */
+    void yy_switch_to_buffer  (YY_BUFFER_STATE  new_buffer , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	/* TODO. We should be able to replace this entire function body
+	 * with
+	 *		yypop_buffer_state();
+	 *		yypush_buffer_state(new_buffer);
+     */
+	yyensure_buffer_stack (yyscanner);
+	if ( YY_CURRENT_BUFFER == new_buffer )
+		return;
+
+	if ( YY_CURRENT_BUFFER )
+		{
+		/* Flush out information for old buffer. */
+		*yyg->yy_c_buf_p = yyg->yy_hold_char;
+		YY_CURRENT_BUFFER_LVALUE->yy_buf_pos = yyg->yy_c_buf_p;
+		YY_CURRENT_BUFFER_LVALUE->yy_n_chars = yyg->yy_n_chars;
+		}
+
+	YY_CURRENT_BUFFER_LVALUE = new_buffer;
+	yy_load_buffer_state(yyscanner );
+
+	/* We don't actually know whether we did this switch during
+	 * EOF (yywrap()) processing, but the only time this flag
+	 * is looked at is after yywrap() is called, so it's safe
+	 * to go ahead and always set it.
+	 */
+	yyg->yy_did_buffer_switch_on_eof = 1;
+}
+
+static void yy_load_buffer_state  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+	yyg->yy_n_chars = YY_CURRENT_BUFFER_LVALUE->yy_n_chars;
+	yyg->yytext_ptr = yyg->yy_c_buf_p = YY_CURRENT_BUFFER_LVALUE->yy_buf_pos;
+	yyin = YY_CURRENT_BUFFER_LVALUE->yy_input_file;
+	yyg->yy_hold_char = *yyg->yy_c_buf_p;
+}
+
+/** Allocate and initialize an input buffer state.
+ * @param file A readable stream.
+ * @param size The character buffer size in bytes. When in doubt, use @c YY_BUF_SIZE.
+ * @param yyscanner The scanner object.
+ * @return the allocated buffer state.
+ */
+    YY_BUFFER_STATE yy_create_buffer  (FILE * file, int  size , yyscan_t yyscanner)
+{
+	YY_BUFFER_STATE b;
+    
+	b = (YY_BUFFER_STATE) yyalloc(sizeof( struct yy_buffer_state ) ,yyscanner );
+	if ( ! b )
+		YY_FATAL_ERROR( "out of dynamic memory in yy_create_buffer()" );
+
+	b->yy_buf_size = size;
+
+	/* yy_ch_buf has to be 2 characters longer than the size given because
+	 * we need to put in 2 end-of-buffer characters.
+	 */
+	b->yy_ch_buf = (char *) yyalloc(b->yy_buf_size + 2 ,yyscanner );
+	if ( ! b->yy_ch_buf )
+		YY_FATAL_ERROR( "out of dynamic memory in yy_create_buffer()" );
+
+	b->yy_is_our_buffer = 1;
+
+	yy_init_buffer(b,file ,yyscanner);
+
+	return b;
+}
+
+/** Destroy the buffer.
+ * @param b a buffer created with yy_create_buffer()
+ * @param yyscanner The scanner object.
+ */
+    void yy_delete_buffer (YY_BUFFER_STATE  b , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	if ( ! b )
+		return;
+
+	if ( b == YY_CURRENT_BUFFER ) /* Not sure if we should pop here. */
+		YY_CURRENT_BUFFER_LVALUE = (YY_BUFFER_STATE) 0;
+
+	if ( b->yy_is_our_buffer )
+		yyfree((void *) b->yy_ch_buf ,yyscanner );
+
+	yyfree((void *) b ,yyscanner );
+}
+
+/* Initializes or reinitializes a buffer.
+ * This function is sometimes called more than once on the same buffer,
+ * such as during a yyrestart() or at EOF.
+ */
+    static void yy_init_buffer  (YY_BUFFER_STATE  b, FILE * file , yyscan_t yyscanner)
+
+{
+	int oerrno = errno;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	yy_flush_buffer(b ,yyscanner);
+
+	b->yy_input_file = file;
+	b->yy_fill_buffer = 1;
+
+    /* If b is the current buffer, then yy_init_buffer was _probably_
+     * called from yyrestart() or through yy_get_next_buffer.
+     * In that case, we don't want to reset the lineno or column.
+     */
+    if (b != YY_CURRENT_BUFFER){
+        b->yy_bs_lineno = 1;
+        b->yy_bs_column = 0;
+    }
+
+        b->yy_is_interactive = 0;
+    
+	errno = oerrno;
+}
+
+/** Discard all buffered characters. On the next scan, YY_INPUT will be called.
+ * @param b the buffer state to be flushed, usually @c YY_CURRENT_BUFFER.
+ * @param yyscanner The scanner object.
+ */
+    void yy_flush_buffer (YY_BUFFER_STATE  b , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+	if ( ! b )
+		return;
+
+	b->yy_n_chars = 0;
+
+	/* We always need two end-of-buffer characters.  The first causes
+	 * a transition to the end-of-buffer state.  The second causes
+	 * a jam in that state.
+	 */
+	b->yy_ch_buf[0] = YY_END_OF_BUFFER_CHAR;
+	b->yy_ch_buf[1] = YY_END_OF_BUFFER_CHAR;
+
+	b->yy_buf_pos = &b->yy_ch_buf[0];
+
+	b->yy_at_bol = 1;
+	b->yy_buffer_status = YY_BUFFER_NEW;
+
+	if ( b == YY_CURRENT_BUFFER )
+		yy_load_buffer_state(yyscanner );
+}
+
+/** Pushes the new state onto the stack. The new state becomes
+ *  the current state. This function will allocate the stack
+ *  if necessary.
+ *  @param new_buffer The new state.
+ *  @param yyscanner The scanner object.
+ */
+void yypush_buffer_state (YY_BUFFER_STATE new_buffer , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+	if (new_buffer == NULL)
+		return;
+
+	yyensure_buffer_stack(yyscanner);
+
+	/* This block is copied from yy_switch_to_buffer. */
+	if ( YY_CURRENT_BUFFER )
+		{
+		/* Flush out information for old buffer. */
+		*yyg->yy_c_buf_p = yyg->yy_hold_char;
+		YY_CURRENT_BUFFER_LVALUE->yy_buf_pos = yyg->yy_c_buf_p;
+		YY_CURRENT_BUFFER_LVALUE->yy_n_chars = yyg->yy_n_chars;
+		}
+
+	/* Only push if top exists. Otherwise, replace top. */
+	if (YY_CURRENT_BUFFER)
+		yyg->yy_buffer_stack_top++;
+	YY_CURRENT_BUFFER_LVALUE = new_buffer;
+
+	/* copied from yy_switch_to_buffer. */
+	yy_load_buffer_state(yyscanner );
+	yyg->yy_did_buffer_switch_on_eof = 1;
+}
+
+/** Removes and deletes the top of the stack, if present.
+ *  The next element becomes the new top.
+ *  @param yyscanner The scanner object.
+ */
+void yypop_buffer_state (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+	if (!YY_CURRENT_BUFFER)
+		return;
+
+	yy_delete_buffer(YY_CURRENT_BUFFER ,yyscanner);
+	YY_CURRENT_BUFFER_LVALUE = NULL;
+	if (yyg->yy_buffer_stack_top > 0)
+		--yyg->yy_buffer_stack_top;
+
+	if (YY_CURRENT_BUFFER) {
+		yy_load_buffer_state(yyscanner );
+		yyg->yy_did_buffer_switch_on_eof = 1;
+	}
+}
+
+/* Allocates the stack if it does not exist.
+ *  Guarantees space for at least one push.
+ */
+static void yyensure_buffer_stack (yyscan_t yyscanner)
+{
+	int num_to_alloc;
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+	if (!yyg->yy_buffer_stack) {
+
+		/* First allocation is just for 2 elements, since we don't know if this
+		 * scanner will even need a stack. We use 2 instead of 1 to avoid an
+		 * immediate realloc on the next call.
+         */
+		num_to_alloc = 1;
+		yyg->yy_buffer_stack = (struct yy_buffer_state**)yyalloc
+								(num_to_alloc * sizeof(struct yy_buffer_state*)
+								, yyscanner);
+		if ( ! yyg->yy_buffer_stack )
+			YY_FATAL_ERROR( "out of dynamic memory in yyensure_buffer_stack()" );
+								  
+		memset(yyg->yy_buffer_stack, 0, num_to_alloc * sizeof(struct yy_buffer_state*));
+				
+		yyg->yy_buffer_stack_max = num_to_alloc;
+		yyg->yy_buffer_stack_top = 0;
+		return;
+	}
+
+	if (yyg->yy_buffer_stack_top >= (yyg->yy_buffer_stack_max) - 1){
+
+		/* Increase the buffer to prepare for a possible push. */
+		int grow_size = 8 /* arbitrary grow size */;
+
+		num_to_alloc = yyg->yy_buffer_stack_max + grow_size;
+		yyg->yy_buffer_stack = (struct yy_buffer_state**)yyrealloc
+								(yyg->yy_buffer_stack,
+								num_to_alloc * sizeof(struct yy_buffer_state*)
+								, yyscanner);
+		if ( ! yyg->yy_buffer_stack )
+			YY_FATAL_ERROR( "out of dynamic memory in yyensure_buffer_stack()" );
+
+		/* zero only the new slots.*/
+		memset(yyg->yy_buffer_stack + yyg->yy_buffer_stack_max, 0, grow_size * sizeof(struct yy_buffer_state*));
+		yyg->yy_buffer_stack_max = num_to_alloc;
+	}
+}
+
+/** Setup the input buffer state to scan directly from a user-specified character buffer.
+ * @param base the character buffer
+ * @param size the size in bytes of the character buffer
+ * @param yyscanner The scanner object.
+ * @return the newly allocated buffer state object. 
+ */
+YY_BUFFER_STATE yy_scan_buffer  (char * base, yy_size_t  size , yyscan_t yyscanner)
+{
+	YY_BUFFER_STATE b;
+    
+	if ( size < 2 ||
+	     base[size-2] != YY_END_OF_BUFFER_CHAR ||
+	     base[size-1] != YY_END_OF_BUFFER_CHAR )
+		/* They forgot to leave room for the EOB's. */
+		return 0;
+
+	b = (YY_BUFFER_STATE) yyalloc(sizeof( struct yy_buffer_state ) ,yyscanner );
+	if ( ! b )
+		YY_FATAL_ERROR( "out of dynamic memory in yy_scan_buffer()" );
+
+	b->yy_buf_size = size - 2;	/* "- 2" to take care of EOB's */
+	b->yy_buf_pos = b->yy_ch_buf = base;
+	b->yy_is_our_buffer = 0;
+	b->yy_input_file = 0;
+	b->yy_n_chars = b->yy_buf_size;
+	b->yy_is_interactive = 0;
+	b->yy_at_bol = 1;
+	b->yy_fill_buffer = 0;
+	b->yy_buffer_status = YY_BUFFER_NEW;
+
+	yy_switch_to_buffer(b ,yyscanner );
+
+	return b;
+}
+
+/** Setup the input buffer state to scan a string. The next call to yylex() will
+ * scan from a @e copy of @a str.
+ * @param yystr a NUL-terminated string to scan
+ * @param yyscanner The scanner object.
+ * @return the newly allocated buffer state object.
+ * @note If you want to scan bytes that may contain NUL values, then use
+ *       yy_scan_bytes() instead.
+ */
+YY_BUFFER_STATE yy_scan_string (yyconst char * yystr , yyscan_t yyscanner)
+{
+    
+	return yy_scan_bytes(yystr,strlen(yystr) ,yyscanner);
+}
+
+/** Setup the input buffer state to scan the given bytes. The next call to yylex() will
+ * scan from a @e copy of @a bytes.
+ * @param bytes the byte buffer to scan
+ * @param len the number of bytes in the buffer pointed to by @a bytes.
+ * @param yyscanner The scanner object.
+ * @return the newly allocated buffer state object.
+ */
+YY_BUFFER_STATE yy_scan_bytes  (yyconst char * yybytes, int  _yybytes_len , yyscan_t yyscanner)
+{
+	YY_BUFFER_STATE b;
+	char *buf;
+	yy_size_t n;
+	int i;
+    
+	/* Get memory for full buffer, including space for trailing EOB's. */
+	n = _yybytes_len + 2;
+	buf = (char *) yyalloc(n ,yyscanner );
+	if ( ! buf )
+		YY_FATAL_ERROR( "out of dynamic memory in yy_scan_bytes()" );
+
+	for ( i = 0; i < _yybytes_len; ++i )
+		buf[i] = yybytes[i];
+
+	buf[_yybytes_len] = buf[_yybytes_len+1] = YY_END_OF_BUFFER_CHAR;
+
+	b = yy_scan_buffer(buf,n ,yyscanner);
+	if ( ! b )
+		YY_FATAL_ERROR( "bad buffer in yy_scan_bytes()" );
+
+	/* It's okay to grow etc. this buffer, and we should throw it
+	 * away when we're done.
+	 */
+	b->yy_is_our_buffer = 1;
+
+	return b;
+}
+
+#ifndef YY_EXIT_FAILURE
+#define YY_EXIT_FAILURE 2
+#endif
+
+static void yy_fatal_error (yyconst char* msg , yyscan_t yyscanner)
+{
+    	(void) fprintf( stderr, "%s\n", msg );
+	exit( YY_EXIT_FAILURE );
+}
+
+/* Redefine yyless() so it works in section 3 code. */
+
+#undef yyless
+#define yyless(n) \
+	do \
+		{ \
+		/* Undo effects of setting up yytext. */ \
+        int yyless_macro_arg = (n); \
+        YY_LESS_LINENO(yyless_macro_arg);\
+		yytext[yyleng] = yyg->yy_hold_char; \
+		yyg->yy_c_buf_p = yytext + yyless_macro_arg; \
+		yyg->yy_hold_char = *yyg->yy_c_buf_p; \
+		*yyg->yy_c_buf_p = '\0'; \
+		yyleng = yyless_macro_arg; \
+		} \
+	while ( 0 )
+
+/* Accessor  methods (get/set functions) to struct members. */
+
+/** Get the user-defined data for this scanner.
+ * @param yyscanner The scanner object.
+ */
+YY_EXTRA_TYPE yyget_extra  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yyextra;
+}
+
+/** Get the current line number.
+ * @param yyscanner The scanner object.
+ */
+int yyget_lineno  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    
+        if (! YY_CURRENT_BUFFER)
+            return 0;
+    
+    return yylineno;
+}
+
+/** Get the current column number.
+ * @param yyscanner The scanner object.
+ */
+int yyget_column  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    
+        if (! YY_CURRENT_BUFFER)
+            return 0;
+    
+    return yycolumn;
+}
+
+/** Get the input stream.
+ * @param yyscanner The scanner object.
+ */
+FILE *yyget_in  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yyin;
+}
+
+/** Get the output stream.
+ * @param yyscanner The scanner object.
+ */
+FILE *yyget_out  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yyout;
+}
+
+/** Get the length of the current token.
+ * @param yyscanner The scanner object.
+ */
+int yyget_leng  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yyleng;
+}
+
+/** Get the current token.
+ * @param yyscanner The scanner object.
+ */
+
+char *yyget_text  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yytext;
+}
+
+/** Set the user-defined data. This data is never touched by the scanner.
+ * @param user_defined The data to be associated with this scanner.
+ * @param yyscanner The scanner object.
+ */
+void yyset_extra (YY_EXTRA_TYPE  user_defined , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yyextra = user_defined ;
+}
+
+/** Set the current line number.
+ * @param line_number
+ * @param yyscanner The scanner object.
+ */
+void yyset_lineno (int  line_number , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+        /* lineno is only valid if an input buffer exists. */
+        if (! YY_CURRENT_BUFFER )
+           yy_fatal_error( "yyset_lineno called with no buffer" , yyscanner); 
+    
+    yylineno = line_number;
+}
+
+/** Set the current column.
+ * @param line_number
+ * @param yyscanner The scanner object.
+ */
+void yyset_column (int  column_no , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+        /* column is only valid if an input buffer exists. */
+        if (! YY_CURRENT_BUFFER )
+           yy_fatal_error( "yyset_column called with no buffer" , yyscanner); 
+    
+    yycolumn = column_no;
+}
+
+/** Set the input stream. This does not discard the current
+ * input buffer.
+ * @param in_str A readable stream.
+ * @param yyscanner The scanner object.
+ * @see yy_switch_to_buffer
+ */
+void yyset_in (FILE *  in_str , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yyin = in_str ;
+}
+
+void yyset_out (FILE *  out_str , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yyout = out_str ;
+}
+
+int yyget_debug  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yy_flex_debug;
+}
+
+void yyset_debug (int  bdebug , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yy_flex_debug = bdebug ;
+}
+
+/* Accessor methods for yylval and yylloc */
+
+YYSTYPE * yyget_lval  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yylval;
+}
+
+void yyset_lval (YYSTYPE *  yylval_param , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yylval = yylval_param;
+}
+
+YYLTYPE *yyget_lloc  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    return yylloc;
+}
+    
+void yyset_lloc (YYLTYPE *  yylloc_param , yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    yylloc = yylloc_param;
+}
+    
+/* User-visible API */
+
+/* yylex_init is special because it creates the scanner itself, so it is
+ * the ONLY reentrant function that doesn't take the scanner as the last argument.
+ * That's why we explicitly handle the declaration, instead of using our macros.
+ */
+
+int yylex_init(yyscan_t* ptr_yy_globals)
+
+{
+    if (ptr_yy_globals == NULL){
+        errno = EINVAL;
+        return 1;
+    }
+
+    *ptr_yy_globals = (yyscan_t) yyalloc ( sizeof( struct yyguts_t ), NULL );
+
+    if (*ptr_yy_globals == NULL){
+        errno = ENOMEM;
+        return 1;
+    }
+
+    /* By setting to 0xAA, we expose bugs in yy_init_globals. Leave at 0x00 for releases. */
+    memset(*ptr_yy_globals,0x00,sizeof(struct yyguts_t));
+
+    return yy_init_globals ( *ptr_yy_globals );
+}
+
+/* yylex_init_extra has the same functionality as yylex_init, but follows the
+ * convention of taking the scanner as the last argument. Note however, that
+ * this is a *pointer* to a scanner, as it will be allocated by this call (and
+ * is the reason, too, why this function also must handle its own declaration).
+ * The user defined value in the first argument will be available to yyalloc in
+ * the yyextra field.
+ */
+
+int yylex_init_extra(YY_EXTRA_TYPE yy_user_defined,yyscan_t* ptr_yy_globals )
+
+{
+    struct yyguts_t dummy_yyguts;
+
+    yyset_extra (yy_user_defined, &dummy_yyguts);
+
+    if (ptr_yy_globals == NULL){
+        errno = EINVAL;
+        return 1;
+    }
+	
+    *ptr_yy_globals = (yyscan_t) yyalloc ( sizeof( struct yyguts_t ), &dummy_yyguts );
+	
+    if (*ptr_yy_globals == NULL){
+        errno = ENOMEM;
+        return 1;
+    }
+    
+    /* By setting to 0xAA, we expose bugs in
+    yy_init_globals. Leave at 0x00 for releases. */
+    memset(*ptr_yy_globals,0x00,sizeof(struct yyguts_t));
+    
+    yyset_extra (yy_user_defined, *ptr_yy_globals);
+    
+    return yy_init_globals ( *ptr_yy_globals );
+}
+
+static int yy_init_globals (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+    /* Initialization is the same as for the non-reentrant scanner.
+     * This function is called from yylex_destroy(), so don't allocate here.
+     */
+
+    yyg->yy_buffer_stack = 0;
+    yyg->yy_buffer_stack_top = 0;
+    yyg->yy_buffer_stack_max = 0;
+    yyg->yy_c_buf_p = (char *) 0;
+    yyg->yy_init = 0;
+    yyg->yy_start = 0;
+
+    yyg->yy_start_stack_ptr = 0;
+    yyg->yy_start_stack_depth = 0;
+    yyg->yy_start_stack =  NULL;
+
+/* Defined in main.c */
+#ifdef YY_STDINIT
+    yyin = stdin;
+    yyout = stdout;
+#else
+    yyin = (FILE *) 0;
+    yyout = (FILE *) 0;
+#endif
+
+    /* For future reference: Set errno on error, since we are called by
+     * yylex_init()
+     */
+    return 0;
+}
+
+/* yylex_destroy is for both reentrant and non-reentrant scanners. */
+int yylex_destroy  (yyscan_t yyscanner)
+{
+    struct yyguts_t * yyg = (struct yyguts_t*)yyscanner;
+
+    /* Pop the buffer stack, destroying each element. */
+	while(YY_CURRENT_BUFFER){
+		yy_delete_buffer(YY_CURRENT_BUFFER ,yyscanner );
+		YY_CURRENT_BUFFER_LVALUE = NULL;
+		yypop_buffer_state(yyscanner);
+	}
+
+	/* Destroy the stack itself. */
+	yyfree(yyg->yy_buffer_stack ,yyscanner);
+	yyg->yy_buffer_stack = NULL;
+
+    /* Destroy the start condition stack. */
+        yyfree(yyg->yy_start_stack ,yyscanner );
+        yyg->yy_start_stack = NULL;
+
+    /* Reset the globals. This is important in a non-reentrant scanner so the next time
+     * yylex() is called, initialization will occur. */
+    yy_init_globals( yyscanner);
+
+    /* Destroy the main struct (reentrant only). */
+    yyfree ( yyscanner , yyscanner );
+    yyscanner = NULL;
+    return 0;
+}
+
+/*
+ * Internal utility routines.
+ */
+
+#ifndef yytext_ptr
+static void yy_flex_strncpy (char* s1, yyconst char * s2, int n , yyscan_t yyscanner)
+{
+	register int i;
+	for ( i = 0; i < n; ++i )
+		s1[i] = s2[i];
+}
+#endif
+
+#ifdef YY_NEED_STRLEN
+static int yy_flex_strlen (yyconst char * s , yyscan_t yyscanner)
+{
+	register int n;
+	for ( n = 0; s[n]; ++n )
+		;
+
+	return n;
+}
+#endif
+
+void *yyalloc (yy_size_t  size , yyscan_t yyscanner)
+{
+	return (void *) malloc( size );
+}
+
+void *yyrealloc  (void * ptr, yy_size_t  size , yyscan_t yyscanner)
+{
+	/* The cast to (char *) in the following accommodates both
+	 * implementations that use char* generic pointers, and those
+	 * that use void* generic pointers.  It works with the latter
+	 * because both ANSI C and C++ allow castless assignment from
+	 * any pointer type to void*, and deal with argument conversions
+	 * as though doing an assignment.
+	 */
+	return (void *) realloc( (char *) ptr, size );
+}
+
+void yyfree (void * ptr , yyscan_t yyscanner)
+{
+	free( (char *) ptr );	/* see yyrealloc() for (char *) cast */
+}
+
+#define YYTABLES_NAME "yytables"
+
+#line 475 "program_lexer.l"
+
+
+
+void
+_mesa_program_lexer_ctor(void **scanner, struct asm_parser_state *state,
+			 const char *string, size_t len)
+{
+   yylex_init_extra(state,scanner);
+   yy_scan_bytes(string,len,*scanner);
+}
+
+void
+_mesa_program_lexer_dtor(void *scanner)
+{
+   yylex_destroy(scanner);
+}
+
diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c
index 5142c2a4a5..d6469b17be 100644
--- a/src/mesa/shader/nvprogram.c
+++ b/src/mesa/shader/nvprogram.c
@@ -354,6 +354,7 @@ _mesa_GetTrackMatrixivNV(GLenum target, GLuint address,
 void GLAPIENTRY
 _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -362,15 +363,17 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -395,6 +398,7 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
 void GLAPIENTRY
 _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -403,15 +407,17 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = (GLfloat) array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = (GLfloat) array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = (GLfloat) array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -436,6 +442,7 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
 void GLAPIENTRY
 _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -444,15 +451,17 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -467,7 +476,7 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
          params[3] = (GLint) ctx->Current.Attrib[index][3];
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
+         params[0] = array->BufferObj->Name;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribdvNV");
@@ -706,7 +715,7 @@ _mesa_ProgramNamedParameter4fNV(GLuint id, GLsizei len, const GLubyte *name,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    prog = _mesa_lookup_program(ctx, id);
    if (!prog || prog->Target != GL_FRAGMENT_PROGRAM_NV) {
diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c
index 68a59350a1..c8a762f8ff 100644
--- a/src/mesa/shader/prog_execute.c
+++ b/src/mesa/shader/prog_execute.c
@@ -225,6 +225,13 @@ fetch_vector4(const struct prog_src_register *source,
       result[2] = -result[2];
       result[3] = -result[3];
    }
+
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+#endif
 }
 
 
@@ -479,6 +486,13 @@ store_vector4(const struct prog_instruction *inst,
       }
    }
 
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+#endif
+
    if (writeMask & WRITEMASK_X)
       dst[0] = value[0];
    if (writeMask & WRITEMASK_Y)
@@ -832,10 +846,14 @@ _mesa_execute_program(GLcontext * ctx,
          break;
       case OPCODE_EX2:         /* Exponential base 2 */
          {
-            GLfloat a[4], result[4];
+            GLfloat a[4], result[4], val;
             fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = result[1] = result[2] = result[3] =
-               (GLfloat) _mesa_pow(2.0, a[0]);
+            val = (GLfloat) _mesa_pow(2.0, a[0]);
+            /*
+            if (IS_INF_OR_NAN(val))
+               val = 1.0e10;
+            */
+            result[0] = result[1] = result[2] = result[3] = val;
             store_vector4(inst, machine, result);
          }
          break;
@@ -904,6 +922,11 @@ _mesa_execute_program(GLcontext * ctx,
          {
             GLfloat a[4];
             fetch_vector4(&inst->SrcReg[0], machine, a);
+            if (DEBUG_PROG) {
+               printf("KIL if (%g %g %g %g) <= 0.0\n",
+                      a[0], a[1], a[2], a[3]);
+            }
+
             if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
                return GL_FALSE;
             }
@@ -911,12 +934,17 @@ _mesa_execute_program(GLcontext * ctx,
          break;
       case OPCODE_LG2:         /* log base 2 */
          {
-            GLfloat a[4], result[4];
+            GLfloat a[4], result[4], val;
             fetch_vector1(&inst->SrcReg[0], machine, a);
 	    /* The fast LOG2 macro doesn't meet the precision requirements.
 	     */
-            result[0] = result[1] = result[2] = result[3] =
-		(log(a[0]) * 1.442695F);
+            if (a[0] == 0.0F) {
+               val = 0.0F;
+            }
+            else {
+               val = log(a[0]) * 1.442695F;
+            }
+            result[0] = result[1] = result[2] = result[3] = val;
             store_vector4(inst, machine, result);
          }
          break;
diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c
index ae3a003fee..44c961927a 100644
--- a/src/mesa/shader/prog_instruction.c
+++ b/src/mesa/shader/prog_instruction.c
@@ -343,7 +343,10 @@ _mesa_opcode_string(gl_inst_opcode opcode)
 {
    if (opcode < MAX_OPCODE)
       return InstInfo[opcode].Name;
-   else
-      return "OP?";
+   else {
+      static char s[20];
+      _mesa_snprintf(s, sizeof(s), "OP%u", opcode);
+      return s;
+   }
 }
 
diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h
index 40ad998f79..39a221eeab 100644
--- a/src/mesa/shader/prog_instruction.h
+++ b/src/mesa/shader/prog_instruction.h
@@ -133,6 +133,7 @@
 #define NEGATE_Y    0x2
 #define NEGATE_Z    0x4
 #define NEGATE_W    0x8
+#define NEGATE_XYZ  0x7
 #define NEGATE_XYZW 0xf
 #define NEGATE_NONE 0x0
 /*@}*/
@@ -303,11 +304,11 @@ struct prog_dst_register
     * Condition code swizzle value.
     */
    GLuint CondSwizzle:12;
-   
+
    /**
     * Selects the condition code register to use for conditional destination
     * update masking.  In NV_fragmnet_program or NV_vertex_program2 mode, only
-    * condition code register 0 is available.  In NV_vertex_program3 mode, 
+    * condition code register 0 is available.  In NV_vertex_program3 mode,
     * condition code registers 0 and 1 are available.
     */
    GLuint CondSrc:1;
@@ -359,7 +360,7 @@ struct prog_instruction
     * NV_fragment_program, NV_fragment_program_option, NV_vertex_program3.
     */
    GLuint SaturateMode:2;
-   
+
    /**
     * Per-instruction selectable precision: FLOAT32, FLOAT16, FIXED12.
     *
@@ -374,7 +375,7 @@ struct prog_instruction
    /*@{*/
    /** Source texture unit. */
    GLuint TexSrcUnit:5;
-   
+
    /** Source texture target, one of TEXTURE_{1D,2D,3D,CUBE,RECT}_INDEX */
    GLuint TexSrcTarget:3;
 
diff --git a/src/mesa/shader/prog_optimize.c b/src/mesa/shader/prog_optimize.c
index 6ba2e76ff9..be903106a0 100644
--- a/src/mesa/shader/prog_optimize.c
+++ b/src/mesa/shader/prog_optimize.c
@@ -547,15 +547,13 @@ update_interval(GLint intBegin[], GLint intEnd[], GLuint index, GLuint ic)
 
 
 /**
- * Find the live intervals for each temporary register in the program.
- * For register R, the interval [A,B] indicates that R is referenced
- * from instruction A through instruction B.
- * Special consideration is needed for loops and subroutines.
- * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ * Find first/last instruction that references each temporary register.
  */
-static GLboolean
-find_live_intervals(struct gl_program *prog,
-                    struct interval_list *liveIntervals)
+GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS])
 {
    struct loop_info
    {
@@ -563,26 +561,15 @@ find_live_intervals(struct gl_program *prog,
    };
    struct loop_info loopStack[MAX_LOOP_NESTING];
    GLuint loopStackDepth = 0;
-   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
    GLuint i;
 
-   /*
-    * Note: we'll return GL_FALSE below if we find relative indexing
-    * into the TEMP register file.  We can't handle that yet.
-    * We also give up on subroutines for now.
-    */
-
-   if (dbg) {
-      _mesa_printf("Optimize: Begin find intervals\n");
-   }
-
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++){
       intBegin[i] = intEnd[i] = -1;
    }
 
    /* Scan instructions looking for temporary registers */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < numInstructions; i++) {
+      const struct prog_instruction *inst = instructions + i;
       if (inst->Opcode == OPCODE_BGNLOOP) {
          loopStack[loopStackDepth].Start = i;
          loopStack[loopStackDepth].End = inst->BranchTarget;
@@ -595,7 +582,7 @@ find_live_intervals(struct gl_program *prog,
          return GL_FALSE;
       }
       else {
-         const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
+         const GLuint numSrc = 3;/*_mesa_num_inst_src_regs(inst->Opcode);*/
          GLuint j;
          for (j = 0; j < numSrc; j++) {
             if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) {
@@ -624,6 +611,39 @@ find_live_intervals(struct gl_program *prog,
       }
    }
 
+   return GL_TRUE;
+}
+
+
+/**
+ * Find the live intervals for each temporary register in the program.
+ * For register R, the interval [A,B] indicates that R is referenced
+ * from instruction A through instruction B.
+ * Special consideration is needed for loops and subroutines.
+ * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ */
+static GLboolean
+find_live_intervals(struct gl_program *prog,
+                    struct interval_list *liveIntervals)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
+   GLuint i;
+
+   /*
+    * Note: we'll return GL_FALSE below if we find relative indexing
+    * into the TEMP register file.  We can't handle that yet.
+    * We also give up on subroutines for now.
+    */
+
+   if (dbg) {
+      _mesa_printf("Optimize: Begin find intervals\n");
+   }
+
+   /* build intermediate arrays */
+   if (!_mesa_find_temp_intervals(prog->Instructions, prog->NumInstructions,
+                                  intBegin, intEnd))
+      return GL_FALSE;
+
    /* Build live intervals list from intermediate arrays */
    liveIntervals->Num = 0;
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
@@ -792,8 +812,6 @@ _mesa_reallocate_registers(struct gl_program *prog)
 }
 
 
-
-
 /**
  * Apply optimizations to the given program to eliminate unnecessary
  * instructions, temp regs, etc.
diff --git a/src/mesa/shader/prog_optimize.h b/src/mesa/shader/prog_optimize.h
index d102cfd9fc..43894a2723 100644
--- a/src/mesa/shader/prog_optimize.h
+++ b/src/mesa/shader/prog_optimize.h
@@ -25,7 +25,19 @@
 #ifndef PROG_OPT_H
 #define PROG_OPT_H
 
+
+#include "main/config.h"
+
+
 struct gl_program;
+struct prog_instruction;
+
+
+extern GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS]);
 
 extern void
 _mesa_optimize_program(GLcontext *ctx, struct gl_program *program);
diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c
index bcd8c5d9dc..6b9e73b2cb 100644
--- a/src/mesa/shader/prog_parameter.c
+++ b/src/mesa/shader/prog_parameter.c
@@ -44,6 +44,34 @@ _mesa_new_parameter_list(void)
 }
 
 
+struct gl_program_parameter_list *
+_mesa_new_parameter_list_sized(unsigned size)
+{
+   struct gl_program_parameter_list *p = _mesa_new_parameter_list();
+
+   if ((p != NULL) && (size != 0)) {
+      p->Size = size;
+
+      /* alloc arrays */
+      p->Parameters = (struct gl_program_parameter *)
+	 _mesa_calloc(size * sizeof(struct gl_program_parameter));
+
+      p->ParameterValues = (GLfloat (*)[4])
+         _mesa_align_malloc(size * 4 *sizeof(GLfloat), 16);
+
+
+      if ((p->Parameters == NULL) || (p->ParameterValues == NULL)) {
+	 _mesa_free(p->Parameters);
+	 _mesa_align_free(p->ParameterValues);
+	 _mesa_free(p);
+	 p = NULL;
+      }
+   }
+
+   return p;
+}
+
+
 /**
  * Free a parameter list and all its parameters
  */
diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h
index 01f5a0e179..d1fcf47e61 100644
--- a/src/mesa/shader/prog_parameter.h
+++ b/src/mesa/shader/prog_parameter.h
@@ -84,6 +84,9 @@ struct gl_program_parameter_list
 extern struct gl_program_parameter_list *
 _mesa_new_parameter_list(void);
 
+extern struct gl_program_parameter_list *
+_mesa_new_parameter_list_sized(unsigned size);
+
 extern void
 _mesa_free_parameter_list(struct gl_program_parameter_list *paramList);
 
diff --git a/src/mesa/shader/prog_parameter_layout.c b/src/mesa/shader/prog_parameter_layout.c
new file mode 100644
index 0000000000..1c37b3a7a5
--- /dev/null
+++ b/src/mesa/shader/prog_parameter_layout.c
@@ -0,0 +1,217 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file prog_parameter_layout.c
+ * \brief Helper functions to layout storage for program parameters
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+#include "main/mtypes.h"
+#include "prog_parameter.h"
+#include "prog_parameter_layout.h"
+#include "prog_instruction.h"
+#include "program_parser.h"
+
+unsigned
+_mesa_combine_swizzles(unsigned base, unsigned applied)
+{
+   unsigned swiz = 0;
+   unsigned i;
+
+   for (i = 0; i < 4; i++) {
+      const unsigned s = GET_SWZ(applied, i);
+
+      swiz |= ((s <= SWIZZLE_W) ? GET_SWZ(base, s) : s) << (i * 3);
+   }
+
+   return swiz;
+}
+
+
+/**
+ * Copy indirect access array from one parameter list to another
+ *
+ * \param src   Parameter array copied from
+ * \param dst   Parameter array copied to
+ * \param first Index of first element in \c src to copy
+ * \param count Number of elements to copy
+ *
+ * \return
+ * The location in \c dst of the first element copied from \c src on
+ * success.  -1 on failure.
+ *
+ * \warning
+ * This function assumes that there is already enough space available in
+ * \c dst to hold all of the elements that will be copied over.
+ */
+static int
+copy_indirect_accessed_array(struct gl_program_parameter_list *src,
+			     struct gl_program_parameter_list *dst,
+			     unsigned first, unsigned count)
+{
+   const int base = dst->NumParameters;
+   unsigned i;
+   unsigned j;
+
+
+   for (i = first; i < (first + count); i++) {
+      struct gl_program_parameter *curr = & src->Parameters[i];
+
+
+      if (curr->Type == PROGRAM_CONSTANT) {
+	 j = dst->NumParameters;
+      } else {
+	 for (j = 0; j < dst->NumParameters; j++) {
+	    if (memcmp(dst->Parameters[j].StateIndexes, curr->StateIndexes, 
+		       sizeof(curr->StateIndexes)) == 0) {
+	       return -1;
+	    }
+	 }
+      }
+
+      assert(j == dst->NumParameters);
+
+      memcpy(& dst->Parameters[j], curr,
+	     sizeof(dst->Parameters[j]));
+      memcpy(dst->ParameterValues[j], src->ParameterValues[i],
+	     sizeof(GLfloat) * 4);
+      curr->Name = NULL;
+
+      dst->NumParameters++;
+   }
+
+   return base;
+}
+
+
+/**
+ * XXX description???
+ * \return GL_TRUE for success, GL_FALSE for failure
+ */
+GLboolean
+_mesa_layout_parameters(struct asm_parser_state *state)
+{
+   struct gl_program_parameter_list *layout;
+   struct asm_instruction *inst;
+   unsigned i;
+
+
+   layout =
+      _mesa_new_parameter_list_sized(state->prog->Parameters->NumParameters);
+
+
+   /* PASS 1:  Move any parameters that are accessed indirectly from the
+    * original parameter list to the new parameter list.
+    */
+   for (inst = state->inst_head; inst != NULL; inst = inst->next) {
+      for (i = 0; i < 3; i++) {
+	 if (inst->SrcReg[i].Base.RelAddr) {
+	    /* Only attempt to add the to the new parameter list once.
+	     */
+	    if (!inst->SrcReg[i].Symbol->pass1_done) {
+	       const int new_begin =
+		  copy_indirect_accessed_array(state->prog->Parameters, layout,
+		      inst->SrcReg[i].Symbol->param_binding_begin,
+		      inst->SrcReg[i].Symbol->param_binding_length);
+
+	       if (new_begin < 0) {
+		  return GL_FALSE;
+	       }
+
+	       inst->SrcReg[i].Symbol->param_binding_begin = new_begin;
+	       inst->SrcReg[i].Symbol->pass1_done = 1;
+	    }
+
+	    /* Previously the Index was just the offset from the parameter
+	     * array.  Now that the base of the parameter array is known, the
+	     * index can be updated to its actual value.
+	     */
+	    inst->Base.SrcReg[i] = inst->SrcReg[i].Base;
+	    inst->Base.SrcReg[i].Index +=
+	       inst->SrcReg[i].Symbol->param_binding_begin;
+	 }
+      }
+   }
+
+
+   /* PASS 2:  Move any parameters that are not accessed indirectly from the
+    * original parameter list to the new parameter list.
+    */
+   for (inst = state->inst_head; inst != NULL; inst = inst->next) {
+      for (i = 0; i < 3; i++) {
+	 const struct gl_program_parameter *p;
+	 const int idx = inst->SrcReg[i].Base.Index;
+	 unsigned swizzle = SWIZZLE_NOOP;
+
+
+	 /* All relative addressed operands were processed on the first
+	  * pass.  Just skip them here.
+	  */
+	 if (inst->SrcReg[i].Base.RelAddr) {
+	    continue;
+	 }
+
+
+	 if ((inst->SrcReg[i].Base.File <= PROGRAM_VARYING )
+	     || (inst->SrcReg[i].Base.File >= PROGRAM_WRITE_ONLY)) {
+	    continue;
+	 }
+
+	 inst->Base.SrcReg[i] = inst->SrcReg[i].Base;
+	 p = & state->prog->Parameters->Parameters[idx];
+
+	 switch (p->Type) {
+	 case PROGRAM_CONSTANT: {
+	    const float *const v =
+	       state->prog->Parameters->ParameterValues[idx];
+
+	    inst->Base.SrcReg[i].Index =
+	       _mesa_add_unnamed_constant(layout, v, p->Size, & swizzle);
+
+	    inst->Base.SrcReg[i].Swizzle = 
+	       _mesa_combine_swizzles(swizzle, inst->Base.SrcReg[i].Swizzle);
+	    break;
+	 }
+
+	 case PROGRAM_STATE_VAR:
+	    inst->Base.SrcReg[i].Index =
+	       _mesa_add_state_reference(layout, p->StateIndexes);
+	    break;
+
+	 default:
+	    break;
+	 }
+
+	 inst->SrcReg[i].Base.File = p->Type;
+	 inst->Base.SrcReg[i].File = p->Type;
+      }
+   }
+
+
+   _mesa_free_parameter_list(state->prog->Parameters);
+   state->prog->Parameters = layout;
+
+   return GL_TRUE;
+}
diff --git a/src/mesa/shader/prog_parameter_layout.h b/src/mesa/shader/prog_parameter_layout.h
new file mode 100644
index 0000000000..99a7b6c726
--- /dev/null
+++ b/src/mesa/shader/prog_parameter_layout.h
@@ -0,0 +1,42 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file prog_parameter_layout.h
+ * \brief Helper functions to layout storage for program parameters
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+#pragma once
+
+#ifndef PROG_PARAMETER_LAYOUT_H
+#define PROG_PARAMETER_LAYOUT_H
+
+extern unsigned _mesa_combine_swizzles(unsigned base, unsigned applied);
+
+struct asm_parser_state;
+
+extern GLboolean _mesa_layout_parameters(struct asm_parser_state *state);
+
+#endif /* PROG_PARAMETER_LAYOUT_H */
diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c
index 5346db7bf6..b71735aa80 100644
--- a/src/mesa/shader/prog_print.c
+++ b/src/mesa/shader/prog_print.c
@@ -75,7 +75,11 @@ file_string(gl_register_file f, gl_prog_print_mode mode)
    case PROGRAM_UNDEFINED:
       return "UNDEFINED";
    default:
-      return "Unknown program file!";
+      {
+         static char s[20];
+         _mesa_snprintf(s, sizeof(s), "FILE%u", f);
+         return s;
+      }
    }
 }
 
@@ -537,7 +541,7 @@ _mesa_print_alu_instruction(const struct prog_instruction *inst,
 /**
  * Print a single vertex/fragment program instruction.
  */
-static GLint
+GLint
 _mesa_fprint_instruction_opt(FILE *f,
                             const struct prog_instruction *inst,
                             GLint indent,
@@ -736,7 +740,10 @@ _mesa_fprint_instruction_opt(FILE *f,
                                 mode, prog);
       }
       else {
-         _mesa_fprintf(f, "Other opcode %d\n", inst->Opcode);
+         fprint_alu_instruction(f, inst,
+                                _mesa_opcode_string(inst->Opcode),
+                                3/*_mesa_num_inst_src_regs(inst->Opcode)*/,
+                                mode, prog);
       }
       break;
    }
@@ -814,6 +821,29 @@ _mesa_print_program(const struct gl_program *prog)
 
 
 /**
+ * Return binary representation of value (as a string).
+ * Insert a comma to separate each group of 8 bits.
+ * XXX move to imports.[ch] if useful elsewhere.
+ */
+static const char *
+binary(GLbitfield val)
+{
+   static char buf[50];
+   GLint i, len = 0;
+   for (i = 31; i >= 0; --i) {
+      if (val & (1 << i))
+         buf[len++] = '1';
+      else if (len > 0 || i == 0)
+         buf[len++] = '0';
+      if (len > 0 && ((i-1) % 8) == 7)
+         buf[len++] = ',';
+   }
+   buf[len] = '\0';
+   return buf;
+}
+
+
+/**
  * Print all of a program's parameters/fields to given file.
  */
 static void
@@ -823,13 +853,17 @@ _mesa_fprint_program_parameters(FILE *f,
 {
    GLuint i;
 
-   _mesa_fprintf(f, "InputsRead: 0x%x\n", prog->InputsRead);
-   _mesa_fprintf(f, "OutputsWritten: 0x%x\n", prog->OutputsWritten);
+   _mesa_fprintf(f, "InputsRead: 0x%x (0b%s)\n",
+                 prog->InputsRead, binary(prog->InputsRead));
+   _mesa_fprintf(f, "OutputsWritten: 0x%x (0b%s)\n",
+                 prog->OutputsWritten, binary(prog->OutputsWritten));
    _mesa_fprintf(f, "NumInstructions=%d\n", prog->NumInstructions);
    _mesa_fprintf(f, "NumTemporaries=%d\n", prog->NumTemporaries);
    _mesa_fprintf(f, "NumParameters=%d\n", prog->NumParameters);
    _mesa_fprintf(f, "NumAttributes=%d\n", prog->NumAttributes);
    _mesa_fprintf(f, "NumAddressRegs=%d\n", prog->NumAddressRegs);
+   _mesa_fprintf(f, "SamplersUsed: 0x%x (0b%s)\n",
+                 prog->SamplersUsed, binary(prog->SamplersUsed));
    _mesa_fprintf(f, "Samplers=[ ");
    for (i = 0; i < MAX_SAMPLERS; i++) {
       _mesa_fprintf(f, "%d ", prog->SamplerUnits[i]);
@@ -926,7 +960,7 @@ _mesa_write_shader_to_file(const struct gl_shader *shader)
       return;
    }
 
-   fprintf(f, "/* Shader %u source */\n", shader->Name);
+   fprintf(f, "/* Shader %u source, checksum %u */\n", shader->Name, shader->SourceChecksum);
    fputs(shader->Source, f);
    fprintf(f, "\n");
 
@@ -941,9 +975,45 @@ _mesa_write_shader_to_file(const struct gl_shader *shader)
       fprintf(f, "/*\n");
       _mesa_fprint_program_opt(f, shader->Program, PROG_PRINT_DEBUG, GL_TRUE);
       fprintf(f, "*/\n");
+      fprintf(f, "/* Parameters / constants */\n");
+      fprintf(f, "/*\n");
+      _mesa_fprint_parameter_list(f, shader->Program->Parameters);
+      fprintf(f, "*/\n");
    }
 
    fclose(f);
 }
 
 
+/**
+ * Append the shader's uniform info/values to the shader log file.
+ * The log file will typically have been created by the
+ * _mesa_write_shader_to_file function.
+ */
+void
+_mesa_append_uniforms_to_file(const struct gl_shader *shader,
+                              const struct gl_program *prog)
+{
+   const char *type;
+   char filename[100];
+   FILE *f;
+
+   if (shader->Type == GL_FRAGMENT_SHADER)
+      type = "frag";
+   else
+      type = "vert";
+
+   _mesa_snprintf(filename, sizeof(filename), "shader_%u.%s", shader->Name, type);
+   f = fopen(filename, "a"); /* append */
+   if (!f) {
+      fprintf(stderr, "Unable to open %s for appending\n", filename);
+      return;
+   }
+
+   fprintf(f, "/* First-draw parameters / constants */\n");
+   fprintf(f, "/*\n");
+   _mesa_fprint_parameter_list(f, prog->Parameters);
+   fprintf(f, "*/\n");
+
+   fclose(f);
+}
diff --git a/src/mesa/shader/prog_print.h b/src/mesa/shader/prog_print.h
index d55661cebb..fc286ded54 100644
--- a/src/mesa/shader/prog_print.h
+++ b/src/mesa/shader/prog_print.h
@@ -57,6 +57,13 @@ extern void
 _mesa_print_instruction(const struct prog_instruction *inst);
 
 extern GLint
+_mesa_fprint_instruction_opt(FILE *f,
+                            const struct prog_instruction *inst,
+                            GLint indent,
+                            gl_prog_print_mode mode,
+                            const struct gl_program *prog);
+
+extern GLint
 _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent,
                             gl_prog_print_mode mode,
                             const struct gl_program *prog);
@@ -79,5 +86,9 @@ _mesa_print_parameter_list(const struct gl_program_parameter_list *list);
 extern void
 _mesa_write_shader_to_file(const struct gl_shader *shader);
 
+extern void
+_mesa_append_uniforms_to_file(const struct gl_shader *shader,
+                              const struct gl_program *prog);
+
 
 #endif /* PROG_PRINT_H */
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index d270bf9e1c..963478fccd 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -62,6 +62,12 @@ _mesa_init_program(GLcontext *ctx)
    ASSERT(ctx->Const.FragmentProgram.MaxUniformComponents / 4
           <= (1 << INST_INDEX_BITS));
 
+   /* If this fails, increase prog_instruction::TexSrcUnit size */
+   ASSERT(MAX_TEXTURE_UNITS < (1 << 5));
+
+   /* If this fails, increase prog_instruction::TexSrcTarget size */
+   ASSERT(NUM_TEXTURE_TARGETS < (1 << 3));
+
    ctx->Program.ErrorPos = -1;
    ctx->Program.ErrorString = _mesa_strdup("");
 
@@ -820,3 +826,63 @@ _mesa_find_free_register(const struct gl_program *prog, GLuint regFile)
 
    return -1;
 }
+
+
+
+/**
+ * "Post-process" a GPU program.  This is intended to be used for debugging.
+ * Example actions include no-op'ing instructions or changing instruction
+ * behaviour.
+ */
+void
+_mesa_postprocess_program(GLcontext *ctx, struct gl_program *prog)
+{
+   static const GLfloat white[4] = { 0.5, 0.5, 0.5, 0.5 };
+   GLuint i;
+   GLuint whiteSwizzle;
+   GLint whiteIndex = _mesa_add_unnamed_constant(prog->Parameters,
+                                                 white, 4, &whiteSwizzle);
+
+   (void) whiteIndex;
+
+   for (i = 0; i < prog->NumInstructions; i++) {
+      struct prog_instruction *inst = prog->Instructions + i;
+      const GLuint n = _mesa_num_inst_src_regs(inst->Opcode);
+
+      (void) n;
+
+      if (_mesa_is_tex_instruction(inst->Opcode)) {
+#if 0
+         /* replace TEX/TXP/TXB with MOV */
+         inst->Opcode = OPCODE_MOV;
+         inst->DstReg.WriteMask = WRITEMASK_XYZW;
+         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+         inst->SrcReg[0].Negate = NEGATE_NONE;
+#endif
+
+#if 0
+         /* disable shadow texture mode */
+         inst->TexShadow = 0;
+#endif
+      }
+
+      if (inst->Opcode == OPCODE_TXP) {
+#if 0
+         inst->Opcode = OPCODE_MOV;
+         inst->DstReg.WriteMask = WRITEMASK_XYZW;
+         inst->SrcReg[0].File = PROGRAM_CONSTANT;
+         inst->SrcReg[0].Index = whiteIndex;
+         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+         inst->SrcReg[0].Negate = NEGATE_NONE;
+#endif
+#if 0
+         inst->TexShadow = 0;
+#endif
+#if 0
+         inst->Opcode = OPCODE_TEX;
+         inst->TexShadow = 0;
+#endif
+      }
+
+   }
+}
diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h
index 48176162c3..56a4191f57 100644
--- a/src/mesa/shader/program.h
+++ b/src/mesa/shader/program.h
@@ -122,6 +122,8 @@ _mesa_combine_programs(GLcontext *ctx,
 extern GLint
 _mesa_find_free_register(const struct gl_program *prog, GLuint regFile);
 
+extern void
+_mesa_postprocess_program(GLcontext *ctx, struct gl_program *prog);
 
 
 #endif /* PROGRAM_H */
diff --git a/src/mesa/shader/program_lexer.l b/src/mesa/shader/program_lexer.l
new file mode 100644
index 0000000000..d240217481
--- /dev/null
+++ b/src/mesa/shader/program_lexer.l
@@ -0,0 +1,489 @@
+%{
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include "main/glheader.h"
+#include "prog_instruction.h"
+#include "prog_statevars.h"
+
+#include "program_parser.h"
+#include "program_parse.tab.h"
+
+#define require_ARB_vp (yyextra->mode == ARB_vertex)
+#define require_ARB_fp (yyextra->mode == ARB_fragment)
+#define require_shadow (yyextra->option.Shadow)
+#define require_rect   (yyextra->option.TexRect)
+#define require_texarray        (yyextra->option.TexArray)
+
+#define return_token_or_IDENTIFIER(condition, token)	\
+   do {							\
+      if (condition) {					\
+	 return token;					\
+      } else {						\
+	 yylval->string = strdup(yytext);		\
+	 return IDENTIFIER;				\
+      }							\
+   } while (0)
+
+#define return_token_or_DOT(condition, token)		\
+   do {							\
+      if (condition) {					\
+	 return token;					\
+      } else {						\
+	 yyless(1);					\
+	 return DOT;					\
+      }							\
+   } while (0)
+
+
+#define return_opcode(condition, token, opcode, sat)	\
+   do {							\
+      if (condition) {					\
+	 yylval->temp_inst.Opcode = OPCODE_ ## opcode;	\
+	 yylval->temp_inst.SaturateMode = SATURATE_ ## sat; \
+	 return token;					\
+      } else {						\
+	 yylval->string = strdup(yytext);		\
+	 return IDENTIFIER;				\
+      }							\
+   } while (0)
+
+#define SWIZZLE_INVAL  MAKE_SWIZZLE4(SWIZZLE_NIL, SWIZZLE_NIL, \
+				     SWIZZLE_NIL, SWIZZLE_NIL)
+
+static unsigned
+mask_from_char(char c)
+{
+   switch (c) {
+   case 'x':
+   case 'r':
+      return WRITEMASK_X;
+   case 'y':
+   case 'g':
+      return WRITEMASK_Y;
+   case 'z':
+   case 'b':
+      return WRITEMASK_Z;
+   case 'w':
+   case 'a':
+      return WRITEMASK_W;
+   }
+
+   return 0;
+}
+
+static unsigned
+swiz_from_char(char c)
+{
+   switch (c) {
+   case 'x':
+   case 'r':
+      return SWIZZLE_X;
+   case 'y':
+   case 'g':
+      return SWIZZLE_Y;
+   case 'z':
+   case 'b':
+      return SWIZZLE_Z;
+   case 'w':
+   case 'a':
+      return SWIZZLE_W;
+   }
+
+   return 0;
+}
+
+#define YY_USER_ACTION							\
+   do {									\
+      yylloc->first_column = yylloc->last_column;			\
+      yylloc->last_column += yyleng;					\
+      if ((yylloc->first_line == 1)					\
+	  && (yylloc->first_column == 1)) {				\
+	 yylloc->position = 1;						\
+      } else {								\
+	 yylloc->position += yylloc->last_column - yylloc->first_column; \
+      }									\
+   } while(0);
+
+#define YY_EXTRA_TYPE struct asm_parser_state *
+%}
+
+num    [0-9]+
+exp    [Ee][-+]?[0-9]+
+frac   "."[0-9]+
+dot    "."[ \t]*
+
+%option bison-bridge bison-locations reentrant noyywrap
+%%
+
+"!!ARBvp1.0"              { return ARBvp_10; }
+"!!ARBfp1.0"              { return ARBfp_10; }
+ADDRESS                   {
+   yylval->integer = at_address;
+   return_token_or_IDENTIFIER(require_ARB_vp, ADDRESS);
+}
+ALIAS                     { return ALIAS; }
+ATTRIB                    { return ATTRIB; }
+END                       { return END; }
+OPTION                    { return OPTION; }
+OUTPUT                    { return OUTPUT; }
+PARAM                     { return PARAM; }
+TEMP                      { yylval->integer = at_temp; return TEMP; }
+
+ABS       { return_opcode(             1, VECTOR_OP, ABS, OFF); }
+ABS_SAT   { return_opcode(require_ARB_fp, VECTOR_OP, ABS, ZERO_ONE); }
+ADD       { return_opcode(             1, BIN_OP, ADD, OFF); }
+ADD_SAT   { return_opcode(require_ARB_fp, BIN_OP, ADD, ZERO_ONE); }
+ARL       { return_opcode(require_ARB_vp, ARL, ARL, OFF); }
+
+CMP       { return_opcode(require_ARB_fp, TRI_OP, CMP, OFF); }
+CMP_SAT   { return_opcode(require_ARB_fp, TRI_OP, CMP, ZERO_ONE); }
+COS       { return_opcode(require_ARB_fp, SCALAR_OP, COS, OFF); }
+COS_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, COS, ZERO_ONE); }
+
+DP3       { return_opcode(             1, BIN_OP, DP3, OFF); }
+DP3_SAT   { return_opcode(require_ARB_fp, BIN_OP, DP3, ZERO_ONE); }
+DP4       { return_opcode(             1, BIN_OP, DP4, OFF); }
+DP4_SAT   { return_opcode(require_ARB_fp, BIN_OP, DP4, ZERO_ONE); }
+DPH       { return_opcode(             1, BIN_OP, DPH, OFF); }
+DPH_SAT   { return_opcode(require_ARB_fp, BIN_OP, DPH, ZERO_ONE); }
+DST       { return_opcode(             1, BIN_OP, DST, OFF); }
+DST_SAT   { return_opcode(require_ARB_fp, BIN_OP, DST, ZERO_ONE); }
+
+EX2       { return_opcode(             1, SCALAR_OP, EX2, OFF); }
+EX2_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, EX2, ZERO_ONE); }
+EXP       { return_opcode(require_ARB_vp, SCALAR_OP, EXP, OFF); }
+
+FLR       { return_opcode(             1, VECTOR_OP, FLR, OFF); }
+FLR_SAT   { return_opcode(require_ARB_fp, VECTOR_OP, FLR, ZERO_ONE); }
+FRC       { return_opcode(             1, VECTOR_OP, FRC, OFF); }
+FRC_SAT   { return_opcode(require_ARB_fp, VECTOR_OP, FRC, ZERO_ONE); }
+
+KIL       { return_opcode(require_ARB_fp, KIL, KIL, OFF); }
+
+LIT       { return_opcode(             1, VECTOR_OP, LIT, OFF); }
+LIT_SAT   { return_opcode(require_ARB_fp, VECTOR_OP, LIT, ZERO_ONE); }
+LG2       { return_opcode(             1, SCALAR_OP, LG2, OFF); }
+LG2_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, LG2, ZERO_ONE); }
+LOG       { return_opcode(require_ARB_vp, SCALAR_OP, LOG, OFF); }
+LRP       { return_opcode(require_ARB_fp, TRI_OP, LRP, OFF); }
+LRP_SAT   { return_opcode(require_ARB_fp, TRI_OP, LRP, ZERO_ONE); }
+
+MAD       { return_opcode(             1, TRI_OP, MAD, OFF); }
+MAD_SAT   { return_opcode(require_ARB_fp, TRI_OP, MAD, ZERO_ONE); }
+MAX       { return_opcode(             1, BIN_OP, MAX, OFF); }
+MAX_SAT   { return_opcode(require_ARB_fp, BIN_OP, MAX, ZERO_ONE); }
+MIN       { return_opcode(             1, BIN_OP, MIN, OFF); }
+MIN_SAT   { return_opcode(require_ARB_fp, BIN_OP, MIN, ZERO_ONE); }
+MOV       { return_opcode(             1, VECTOR_OP, MOV, OFF); }
+MOV_SAT   { return_opcode(require_ARB_fp, VECTOR_OP, MOV, ZERO_ONE); }
+MUL       { return_opcode(             1, BIN_OP, MUL, OFF); }
+MUL_SAT   { return_opcode(require_ARB_fp, BIN_OP, MUL, ZERO_ONE); }
+
+POW       { return_opcode(             1, BINSC_OP, POW, OFF); }
+POW_SAT   { return_opcode(require_ARB_fp, BINSC_OP, POW, ZERO_ONE); }
+
+RCP       { return_opcode(             1, SCALAR_OP, RCP, OFF); }
+RCP_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, RCP, ZERO_ONE); }
+RSQ       { return_opcode(             1, SCALAR_OP, RSQ, OFF); }
+RSQ_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, RSQ, ZERO_ONE); }
+
+SCS       { return_opcode(require_ARB_fp, SCALAR_OP, SCS, OFF); }
+SCS_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, SCS, ZERO_ONE); }
+SGE       { return_opcode(             1, BIN_OP, SGE, OFF); }
+SGE_SAT   { return_opcode(require_ARB_fp, BIN_OP, SGE, ZERO_ONE); }
+SIN       { return_opcode(require_ARB_fp, SCALAR_OP, SIN, OFF); }
+SIN_SAT   { return_opcode(require_ARB_fp, SCALAR_OP, SIN, ZERO_ONE); }
+SLT       { return_opcode(             1, BIN_OP, SLT, OFF); }
+SLT_SAT   { return_opcode(require_ARB_fp, BIN_OP, SLT, ZERO_ONE); }
+SUB       { return_opcode(             1, BIN_OP, SUB, OFF); }
+SUB_SAT   { return_opcode(require_ARB_fp, BIN_OP, SUB, ZERO_ONE); }
+SWZ       { return_opcode(             1, SWZ, SWZ, OFF); }
+SWZ_SAT   { return_opcode(require_ARB_fp, SWZ, SWZ, ZERO_ONE); }
+
+TEX       { return_opcode(require_ARB_fp, SAMPLE_OP, TEX, OFF); }
+TEX_SAT   { return_opcode(require_ARB_fp, SAMPLE_OP, TEX, ZERO_ONE); }
+TXB       { return_opcode(require_ARB_fp, SAMPLE_OP, TXB, OFF); }
+TXB_SAT   { return_opcode(require_ARB_fp, SAMPLE_OP, TXB, ZERO_ONE); }
+TXP       { return_opcode(require_ARB_fp, SAMPLE_OP, TXP, OFF); }
+TXP_SAT   { return_opcode(require_ARB_fp, SAMPLE_OP, TXP, ZERO_ONE); }
+
+XPD       { return_opcode(             1, BIN_OP, XPD, OFF); }
+XPD_SAT   { return_opcode(require_ARB_fp, BIN_OP, XPD, ZERO_ONE); }
+
+vertex                    { return_token_or_IDENTIFIER(require_ARB_vp, VERTEX); }
+fragment                  { return_token_or_IDENTIFIER(require_ARB_fp, FRAGMENT); }
+program                   { return PROGRAM; }
+state                     { return STATE; }
+result                    { return RESULT; }
+
+{dot}ambient              { return AMBIENT; }
+{dot}attenuation          { return ATTENUATION; }
+{dot}back                 { return BACK; }
+{dot}clip                 { return_token_or_DOT(require_ARB_vp, CLIP); }
+{dot}color                { return COLOR; }
+{dot}depth                { return_token_or_DOT(require_ARB_fp, DEPTH); }
+{dot}diffuse              { return DIFFUSE; }
+{dot}direction            { return DIRECTION; }
+{dot}emission             { return EMISSION; }
+{dot}env                  { return ENV; }
+{dot}eye                  { return EYE; }
+{dot}fogcoord             { return FOGCOORD; }
+{dot}fog                  { return FOG; }
+{dot}front                { return FRONT; }
+{dot}half                 { return HALF; }
+{dot}inverse              { return INVERSE; }
+{dot}invtrans             { return INVTRANS; }
+{dot}light                { return LIGHT; }
+{dot}lightmodel           { return LIGHTMODEL; }
+{dot}lightprod            { return LIGHTPROD; }
+{dot}local                { return LOCAL; }
+{dot}material             { return MATERIAL; }
+{dot}program              { return MAT_PROGRAM; }
+{dot}matrix               { return MATRIX; }
+{dot}matrixindex          { return_token_or_DOT(require_ARB_vp, MATRIXINDEX); }
+{dot}modelview            { return MODELVIEW; }
+{dot}mvp                  { return MVP; }
+{dot}normal               { return_token_or_DOT(require_ARB_vp, NORMAL); }
+{dot}object               { return OBJECT; }
+{dot}palette              { return PALETTE; }
+{dot}params               { return PARAMS; }
+{dot}plane                { return PLANE; }
+{dot}point                { return_token_or_DOT(require_ARB_vp, POINT_TOK); }
+{dot}pointsize            { return_token_or_DOT(require_ARB_vp, POINTSIZE); }
+{dot}position             { return POSITION; }
+{dot}primary              { return PRIMARY; }
+{dot}projection           { return PROJECTION; }
+{dot}range                { return_token_or_DOT(require_ARB_fp, RANGE); }
+{dot}row                  { return ROW; }
+{dot}scenecolor           { return SCENECOLOR; }
+{dot}secondary            { return SECONDARY; }
+{dot}shininess            { return SHININESS; }
+{dot}size                 { return_token_or_DOT(require_ARB_vp, SIZE_TOK); }
+{dot}specular             { return SPECULAR; }
+{dot}spot                 { return SPOT; }
+{dot}texcoord             { return TEXCOORD; }
+{dot}texenv               { return_token_or_DOT(require_ARB_fp, TEXENV); }
+{dot}texgen               { return_token_or_DOT(require_ARB_vp, TEXGEN); }
+{dot}q                    { return_token_or_DOT(require_ARB_vp, TEXGEN_Q); }
+{dot}s                    { return_token_or_DOT(require_ARB_vp, TEXGEN_S); }
+{dot}t                    { return_token_or_DOT(require_ARB_vp, TEXGEN_T); }
+{dot}texture              { return TEXTURE; }
+{dot}transpose            { return TRANSPOSE; }
+{dot}attrib               { return_token_or_DOT(require_ARB_vp, VTXATTRIB); }
+{dot}weight               { return_token_or_DOT(require_ARB_vp, WEIGHT); }
+
+texture                   { return_token_or_IDENTIFIER(require_ARB_fp, TEXTURE_UNIT); }
+1D                        { return_token_or_IDENTIFIER(require_ARB_fp, TEX_1D); }
+2D                        { return_token_or_IDENTIFIER(require_ARB_fp, TEX_2D); }
+3D                        { return_token_or_IDENTIFIER(require_ARB_fp, TEX_3D); }
+CUBE                      { return_token_or_IDENTIFIER(require_ARB_fp, TEX_CUBE); }
+RECT                      { return_token_or_IDENTIFIER(require_ARB_fp && require_rect, TEX_RECT); }
+SHADOW1D                  { return_token_or_IDENTIFIER(require_ARB_fp && require_shadow, TEX_SHADOW1D); }
+SHADOW2D                  { return_token_or_IDENTIFIER(require_ARB_fp && require_shadow, TEX_SHADOW2D); }
+SHADOWRECT                { return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_rect, TEX_SHADOWRECT); }
+ARRAY1D                   { return_token_or_IDENTIFIER(require_ARB_fp && require_texarray, TEX_ARRAY1D); }
+ARRAY2D                   { return_token_or_IDENTIFIER(require_ARB_fp && require_texarray, TEX_ARRAY2D); }
+ARRAYSHADOW1D             { return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_texarray, TEX_ARRAYSHADOW1D); }
+ARRAYSHADOW2D             { return_token_or_IDENTIFIER(require_ARB_fp && require_shadow && require_texarray, TEX_ARRAYSHADOW2D); }
+
+[_a-zA-Z$][_a-zA-Z0-9$]*  {
+   yylval->string = strdup(yytext);
+   return IDENTIFIER;
+}
+
+".."                      { return DOT_DOT; }
+
+{num}                     {
+   yylval->integer = strtol(yytext, NULL, 10);
+   return INTEGER;
+}
+{num}?{frac}{exp}?        {
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+{num}"."/[^.]             {
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+{num}{exp}                {
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+{num}"."{exp}             {
+   yylval->real = strtod(yytext, NULL);
+   return REAL;
+}
+
+".xyzw"                   {
+   yylval->swiz_mask.swizzle = SWIZZLE_NOOP;
+   yylval->swiz_mask.mask = WRITEMASK_XYZW;
+   return MASK4;
+}
+
+".xy"[zw]                 {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XY
+      | mask_from_char(yytext[3]);
+   return MASK3;
+}
+".xzw"                    {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XZW;
+   return MASK3;
+}
+".yzw"                    {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_YZW;
+   return MASK3;
+}
+
+".x"[yzw]                 {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_X
+      | mask_from_char(yytext[2]);
+   return MASK2;
+}
+".y"[zw]                  {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_Y
+      | mask_from_char(yytext[2]);
+   return MASK2;
+}
+".zw"                     {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_ZW;
+   return MASK2;
+}
+
+"."[xyzw]                 {
+   const unsigned s = swiz_from_char(yytext[1]);
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(s, s, s, s);
+   yylval->swiz_mask.mask = mask_from_char(yytext[1]);
+   return MASK1; 
+}
+
+"."[xyzw]{4}              {
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(swiz_from_char(yytext[1]),
+					    swiz_from_char(yytext[2]),
+					    swiz_from_char(yytext[3]),
+					    swiz_from_char(yytext[4]));
+   yylval->swiz_mask.mask = 0;
+   return SWIZZLE;
+}
+
+".rgba"                   {
+   yylval->swiz_mask.swizzle = SWIZZLE_NOOP;
+   yylval->swiz_mask.mask = WRITEMASK_XYZW;
+   return_token_or_DOT(require_ARB_fp, MASK4);
+}
+
+".rg"[ba]                 {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XY
+      | mask_from_char(yytext[3]);
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+".rba"                    {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_XZW;
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+".gba"                    {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_YZW;
+   return_token_or_DOT(require_ARB_fp, MASK3);
+}
+
+".r"[gba]                 {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_X
+      | mask_from_char(yytext[2]);
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+".g"[ba]                  {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_Y
+      | mask_from_char(yytext[2]);
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+".ba"                     {
+   yylval->swiz_mask.swizzle = SWIZZLE_INVAL;
+   yylval->swiz_mask.mask = WRITEMASK_ZW;
+   return_token_or_DOT(require_ARB_fp, MASK2);
+}
+
+"."[gba]                  {
+   const unsigned s = swiz_from_char(yytext[1]);
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(s, s, s, s);
+   yylval->swiz_mask.mask = mask_from_char(yytext[1]);
+   return_token_or_DOT(require_ARB_fp, MASK1);
+}
+
+
+".r"                      {
+   if (require_ARB_vp) {
+      return TEXGEN_R;
+   } else {
+      yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X,
+						SWIZZLE_X, SWIZZLE_X);
+      yylval->swiz_mask.mask = WRITEMASK_X;
+      return MASK1;
+   }
+}
+
+"."[rgba]{4}              {
+   yylval->swiz_mask.swizzle = MAKE_SWIZZLE4(swiz_from_char(yytext[1]),
+					    swiz_from_char(yytext[2]),
+					    swiz_from_char(yytext[3]),
+					    swiz_from_char(yytext[4]));
+   yylval->swiz_mask.mask = 0;
+   return_token_or_DOT(require_ARB_fp, SWIZZLE);
+}
+
+"."                       { return DOT; }
+
+\n                        {
+   yylloc->first_line++;
+   yylloc->first_column = 1;
+   yylloc->last_line++;
+   yylloc->last_column = 1;
+   yylloc->position++;
+}
+[ \t\r]+                  /* eat whitespace */ ;
+#.*$                      /* eat comments */ ;
+.                         { return yytext[0]; }
+%%
+
+void
+_mesa_program_lexer_ctor(void **scanner, struct asm_parser_state *state,
+			 const char *string, size_t len)
+{
+   yylex_init_extra(state, scanner);
+   yy_scan_bytes(string, len, *scanner);
+}
+
+void
+_mesa_program_lexer_dtor(void *scanner)
+{
+   yylex_destroy(scanner);
+}
diff --git a/src/mesa/shader/program_parse.tab.c b/src/mesa/shader/program_parse.tab.c
new file mode 100644
index 0000000000..9f2d4de90f
--- /dev/null
+++ b/src/mesa/shader/program_parse.tab.c
@@ -0,0 +1,5249 @@
+
+/* A Bison parser, made by GNU Bison 2.4.1.  */
+
+/* Skeleton implementation for Bison's Yacc-like parsers in C
+   
+      Copyright (C) 1984, 1989, 1990, 2000, 2001, 2002, 2003, 2004, 2005, 2006
+   Free Software Foundation, Inc.
+   
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+/* As a special exception, you may create a larger work that contains
+   part or all of the Bison parser skeleton and distribute that work
+   under terms of your choice, so long as that work isn't itself a
+   parser generator using the skeleton or a modified version thereof
+   as a parser skeleton.  Alternatively, if you modify or redistribute
+   the parser skeleton itself, you may (at your option) remove this
+   special exception, which will cause the skeleton and the resulting
+   Bison output files to be licensed under the GNU General Public
+   License without this special exception.
+   
+   This special exception was added by the Free Software Foundation in
+   version 2.2 of Bison.  */
+
+/* C LALR(1) parser skeleton written by Richard Stallman, by
+   simplifying the original so-called "semantic" parser.  */
+
+/* All symbols defined below should begin with yy or YY, to avoid
+   infringing on user name space.  This should be done even for local
+   variables, as they might otherwise be expanded by user macros.
+   There are some unavoidable exceptions within include files to
+   define necessary library symbols; they are noted "INFRINGES ON
+   USER NAME SPACE" below.  */
+
+/* Identify Bison output.  */
+#define YYBISON 1
+
+/* Bison version.  */
+#define YYBISON_VERSION "2.4.1"
+
+/* Skeleton name.  */
+#define YYSKELETON_NAME "yacc.c"
+
+/* Pure parsers.  */
+#define YYPURE 1
+
+/* Push parsers.  */
+#define YYPUSH 0
+
+/* Pull parsers.  */
+#define YYPULL 1
+
+/* Using locations.  */
+#define YYLSP_NEEDED 1
+
+
+
+/* Copy the first part of user declarations.  */
+
+/* Line 189 of yacc.c  */
+#line 1 "program_parse.y"
+
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "main/mtypes.h"
+#include "main/imports.h"
+#include "program.h"
+#include "prog_parameter.h"
+#include "prog_parameter_layout.h"
+#include "prog_statevars.h"
+#include "prog_instruction.h"
+
+#include "symbol_table.h"
+#include "program_parser.h"
+
+extern void *yy_scan_string(char *);
+extern void yy_delete_buffer(void *);
+
+static struct asm_symbol *declare_variable(struct asm_parser_state *state,
+    char *name, enum asm_type t, struct YYLTYPE *locp);
+
+static int add_state_reference(struct gl_program_parameter_list *param_list,
+    const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_state(struct gl_program *prog,
+    struct asm_symbol *param_var, const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_param(struct gl_program *prog,
+    struct asm_symbol *param_var, const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_const(struct gl_program *prog,
+    struct asm_symbol *param_var, const struct asm_vector *vec);
+
+static int yyparse(struct asm_parser_state *state);
+
+static char *make_error_string(const char *fmt, ...);
+
+static void yyerror(struct YYLTYPE *locp, struct asm_parser_state *state,
+    const char *s);
+
+static int validate_inputs(struct YYLTYPE *locp,
+    struct asm_parser_state *state);
+
+static void init_dst_reg(struct prog_dst_register *r);
+
+static void init_src_reg(struct asm_src_register *r);
+
+static struct asm_instruction *asm_instruction_ctor(gl_inst_opcode op,
+    const struct prog_dst_register *dst, const struct asm_src_register *src0,
+    const struct asm_src_register *src1, const struct asm_src_register *src2);
+
+#ifndef FALSE
+#define FALSE 0
+#define TRUE (!FALSE)
+#endif
+
+#define YYLLOC_DEFAULT(Current, Rhs, N)					\
+   do {									\
+      if (YYID(N)) {							\
+	 (Current).first_line = YYRHSLOC(Rhs, 1).first_line;		\
+	 (Current).first_column = YYRHSLOC(Rhs, 1).first_column;	\
+	 (Current).position = YYRHSLOC(Rhs, 1).position;		\
+	 (Current).last_line = YYRHSLOC(Rhs, N).last_line;		\
+	 (Current).last_column = YYRHSLOC(Rhs, N).last_column;		\
+      } else {								\
+	 (Current).first_line = YYRHSLOC(Rhs, 0).last_line;		\
+	 (Current).last_line = (Current).first_line;			\
+	 (Current).first_column = YYRHSLOC(Rhs, 0).last_column;		\
+	 (Current).last_column = (Current).first_column;		\
+	 (Current).position = YYRHSLOC(Rhs, 0).position			\
+	    + (Current).first_column;					\
+      }									\
+   } while(YYID(0))
+
+#define YYLEX_PARAM state->scanner
+
+
+/* Line 189 of yacc.c  */
+#line 174 "program_parse.tab.c"
+
+/* Enabling traces.  */
+#ifndef YYDEBUG
+# define YYDEBUG 0
+#endif
+
+/* Enabling verbose error messages.  */
+#ifdef YYERROR_VERBOSE
+# undef YYERROR_VERBOSE
+# define YYERROR_VERBOSE 1
+#else
+# define YYERROR_VERBOSE 1
+#endif
+
+/* Enabling the token table.  */
+#ifndef YYTOKEN_TABLE
+# define YYTOKEN_TABLE 0
+#endif
+
+
+/* Tokens.  */
+#ifndef YYTOKENTYPE
+# define YYTOKENTYPE
+   /* Put the tokens into the symbol table, so that GDB and other debuggers
+      know about them.  */
+   enum yytokentype {
+     ARBvp_10 = 258,
+     ARBfp_10 = 259,
+     ADDRESS = 260,
+     ALIAS = 261,
+     ATTRIB = 262,
+     OPTION = 263,
+     OUTPUT = 264,
+     PARAM = 265,
+     TEMP = 266,
+     END = 267,
+     BIN_OP = 268,
+     BINSC_OP = 269,
+     SAMPLE_OP = 270,
+     SCALAR_OP = 271,
+     TRI_OP = 272,
+     VECTOR_OP = 273,
+     ARL = 274,
+     KIL = 275,
+     SWZ = 276,
+     INTEGER = 277,
+     REAL = 278,
+     AMBIENT = 279,
+     ATTENUATION = 280,
+     BACK = 281,
+     CLIP = 282,
+     COLOR = 283,
+     DEPTH = 284,
+     DIFFUSE = 285,
+     DIRECTION = 286,
+     EMISSION = 287,
+     ENV = 288,
+     EYE = 289,
+     FOG = 290,
+     FOGCOORD = 291,
+     FRAGMENT = 292,
+     FRONT = 293,
+     HALF = 294,
+     INVERSE = 295,
+     INVTRANS = 296,
+     LIGHT = 297,
+     LIGHTMODEL = 298,
+     LIGHTPROD = 299,
+     LOCAL = 300,
+     MATERIAL = 301,
+     MAT_PROGRAM = 302,
+     MATRIX = 303,
+     MATRIXINDEX = 304,
+     MODELVIEW = 305,
+     MVP = 306,
+     NORMAL = 307,
+     OBJECT = 308,
+     PALETTE = 309,
+     PARAMS = 310,
+     PLANE = 311,
+     POINT_TOK = 312,
+     POINTSIZE = 313,
+     POSITION = 314,
+     PRIMARY = 315,
+     PROGRAM = 316,
+     PROJECTION = 317,
+     RANGE = 318,
+     RESULT = 319,
+     ROW = 320,
+     SCENECOLOR = 321,
+     SECONDARY = 322,
+     SHININESS = 323,
+     SIZE_TOK = 324,
+     SPECULAR = 325,
+     SPOT = 326,
+     STATE = 327,
+     TEXCOORD = 328,
+     TEXENV = 329,
+     TEXGEN = 330,
+     TEXGEN_Q = 331,
+     TEXGEN_R = 332,
+     TEXGEN_S = 333,
+     TEXGEN_T = 334,
+     TEXTURE = 335,
+     TRANSPOSE = 336,
+     TEXTURE_UNIT = 337,
+     TEX_1D = 338,
+     TEX_2D = 339,
+     TEX_3D = 340,
+     TEX_CUBE = 341,
+     TEX_RECT = 342,
+     TEX_SHADOW1D = 343,
+     TEX_SHADOW2D = 344,
+     TEX_SHADOWRECT = 345,
+     TEX_ARRAY1D = 346,
+     TEX_ARRAY2D = 347,
+     TEX_ARRAYSHADOW1D = 348,
+     TEX_ARRAYSHADOW2D = 349,
+     VERTEX = 350,
+     VTXATTRIB = 351,
+     WEIGHT = 352,
+     IDENTIFIER = 353,
+     MASK4 = 354,
+     MASK3 = 355,
+     MASK2 = 356,
+     MASK1 = 357,
+     SWIZZLE = 358,
+     DOT_DOT = 359,
+     DOT = 360
+   };
+#endif
+
+
+
+#if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED
+typedef union YYSTYPE
+{
+
+/* Line 214 of yacc.c  */
+#line 107 "program_parse.y"
+
+   struct asm_instruction *inst;
+   struct asm_symbol *sym;
+   struct asm_symbol temp_sym;
+   struct asm_swizzle_mask swiz_mask;
+   struct asm_src_register src_reg;
+   struct prog_dst_register dst_reg;
+   struct prog_instruction temp_inst;
+   char *string;
+   unsigned result;
+   unsigned attrib;
+   int integer;
+   float real;
+   gl_state_index state[STATE_LENGTH];
+   int negate;
+   struct asm_vector vector;
+   gl_inst_opcode opcode;
+
+   struct {
+      unsigned swz;
+      unsigned rgba_valid:1;
+      unsigned xyzw_valid:1;
+      unsigned negate:1;
+   } ext_swizzle;
+
+
+
+/* Line 214 of yacc.c  */
+#line 343 "program_parse.tab.c"
+} YYSTYPE;
+# define YYSTYPE_IS_TRIVIAL 1
+# define yystype YYSTYPE /* obsolescent; will be withdrawn */
+# define YYSTYPE_IS_DECLARED 1
+#endif
+
+#if ! defined YYLTYPE && ! defined YYLTYPE_IS_DECLARED
+typedef struct YYLTYPE
+{
+  int first_line;
+  int first_column;
+  int last_line;
+  int last_column;
+} YYLTYPE;
+# define yyltype YYLTYPE /* obsolescent; will be withdrawn */
+# define YYLTYPE_IS_DECLARED 1
+# define YYLTYPE_IS_TRIVIAL 1
+#endif
+
+
+/* Copy the second part of user declarations.  */
+
+/* Line 264 of yacc.c  */
+#line 249 "program_parse.y"
+
+extern int yylex(YYSTYPE *yylval_param, YYLTYPE *yylloc_param,
+    void *yyscanner);
+
+
+/* Line 264 of yacc.c  */
+#line 374 "program_parse.tab.c"
+
+#ifdef short
+# undef short
+#endif
+
+#ifdef YYTYPE_UINT8
+typedef YYTYPE_UINT8 yytype_uint8;
+#else
+typedef unsigned char yytype_uint8;
+#endif
+
+#ifdef YYTYPE_INT8
+typedef YYTYPE_INT8 yytype_int8;
+#elif (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+typedef signed char yytype_int8;
+#else
+typedef short int yytype_int8;
+#endif
+
+#ifdef YYTYPE_UINT16
+typedef YYTYPE_UINT16 yytype_uint16;
+#else
+typedef unsigned short int yytype_uint16;
+#endif
+
+#ifdef YYTYPE_INT16
+typedef YYTYPE_INT16 yytype_int16;
+#else
+typedef short int yytype_int16;
+#endif
+
+#ifndef YYSIZE_T
+# ifdef __SIZE_TYPE__
+#  define YYSIZE_T __SIZE_TYPE__
+# elif defined size_t
+#  define YYSIZE_T size_t
+# elif ! defined YYSIZE_T && (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+#  include <stddef.h> /* INFRINGES ON USER NAME SPACE */
+#  define YYSIZE_T size_t
+# else
+#  define YYSIZE_T unsigned int
+# endif
+#endif
+
+#define YYSIZE_MAXIMUM ((YYSIZE_T) -1)
+
+#ifndef YY_
+# if YYENABLE_NLS
+#  if ENABLE_NLS
+#   include <libintl.h> /* INFRINGES ON USER NAME SPACE */
+#   define YY_(msgid) dgettext ("bison-runtime", msgid)
+#  endif
+# endif
+# ifndef YY_
+#  define YY_(msgid) msgid
+# endif
+#endif
+
+/* Suppress unused-variable warnings by "using" E.  */
+#if ! defined lint || defined __GNUC__
+# define YYUSE(e) ((void) (e))
+#else
+# define YYUSE(e) /* empty */
+#endif
+
+/* Identity function, used to suppress warnings about constant conditions.  */
+#ifndef lint
+# define YYID(n) (n)
+#else
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static int
+YYID (int yyi)
+#else
+static int
+YYID (yyi)
+    int yyi;
+#endif
+{
+  return yyi;
+}
+#endif
+
+#if ! defined yyoverflow || YYERROR_VERBOSE
+
+/* The parser invokes alloca or malloc; define the necessary symbols.  */
+
+# ifdef YYSTACK_USE_ALLOCA
+#  if YYSTACK_USE_ALLOCA
+#   ifdef __GNUC__
+#    define YYSTACK_ALLOC __builtin_alloca
+#   elif defined __BUILTIN_VA_ARG_INCR
+#    include <alloca.h> /* INFRINGES ON USER NAME SPACE */
+#   elif defined _AIX
+#    define YYSTACK_ALLOC __alloca
+#   elif defined _MSC_VER
+#    include <malloc.h> /* INFRINGES ON USER NAME SPACE */
+#    define alloca _alloca
+#   else
+#    define YYSTACK_ALLOC alloca
+#    if ! defined _ALLOCA_H && ! defined _STDLIB_H && (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+#     include <stdlib.h> /* INFRINGES ON USER NAME SPACE */
+#     ifndef _STDLIB_H
+#      define _STDLIB_H 1
+#     endif
+#    endif
+#   endif
+#  endif
+# endif
+
+# ifdef YYSTACK_ALLOC
+   /* Pacify GCC's `empty if-body' warning.  */
+#  define YYSTACK_FREE(Ptr) do { /* empty */; } while (YYID (0))
+#  ifndef YYSTACK_ALLOC_MAXIMUM
+    /* The OS might guarantee only one guard page at the bottom of the stack,
+       and a page size can be as small as 4096 bytes.  So we cannot safely
+       invoke alloca (N) if N exceeds 4096.  Use a slightly smaller number
+       to allow for a few compiler-allocated temporary stack slots.  */
+#   define YYSTACK_ALLOC_MAXIMUM 4032 /* reasonable circa 2006 */
+#  endif
+# else
+#  define YYSTACK_ALLOC YYMALLOC
+#  define YYSTACK_FREE YYFREE
+#  ifndef YYSTACK_ALLOC_MAXIMUM
+#   define YYSTACK_ALLOC_MAXIMUM YYSIZE_MAXIMUM
+#  endif
+#  if (defined __cplusplus && ! defined _STDLIB_H \
+       && ! ((defined YYMALLOC || defined malloc) \
+	     && (defined YYFREE || defined free)))
+#   include <stdlib.h> /* INFRINGES ON USER NAME SPACE */
+#   ifndef _STDLIB_H
+#    define _STDLIB_H 1
+#   endif
+#  endif
+#  ifndef YYMALLOC
+#   define YYMALLOC malloc
+#   if ! defined malloc && ! defined _STDLIB_H && (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+void *malloc (YYSIZE_T); /* INFRINGES ON USER NAME SPACE */
+#   endif
+#  endif
+#  ifndef YYFREE
+#   define YYFREE free
+#   if ! defined free && ! defined _STDLIB_H && (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+void free (void *); /* INFRINGES ON USER NAME SPACE */
+#   endif
+#  endif
+# endif
+#endif /* ! defined yyoverflow || YYERROR_VERBOSE */
+
+
+#if (! defined yyoverflow \
+     && (! defined __cplusplus \
+	 || (defined YYLTYPE_IS_TRIVIAL && YYLTYPE_IS_TRIVIAL \
+	     && defined YYSTYPE_IS_TRIVIAL && YYSTYPE_IS_TRIVIAL)))
+
+/* A type that is properly aligned for any stack member.  */
+union yyalloc
+{
+  yytype_int16 yyss_alloc;
+  YYSTYPE yyvs_alloc;
+  YYLTYPE yyls_alloc;
+};
+
+/* The size of the maximum gap between one aligned stack and the next.  */
+# define YYSTACK_GAP_MAXIMUM (sizeof (union yyalloc) - 1)
+
+/* The size of an array large to enough to hold all stacks, each with
+   N elements.  */
+# define YYSTACK_BYTES(N) \
+     ((N) * (sizeof (yytype_int16) + sizeof (YYSTYPE) + sizeof (YYLTYPE)) \
+      + 2 * YYSTACK_GAP_MAXIMUM)
+
+/* Copy COUNT objects from FROM to TO.  The source and destination do
+   not overlap.  */
+# ifndef YYCOPY
+#  if defined __GNUC__ && 1 < __GNUC__
+#   define YYCOPY(To, From, Count) \
+      __builtin_memcpy (To, From, (Count) * sizeof (*(From)))
+#  else
+#   define YYCOPY(To, From, Count)		\
+      do					\
+	{					\
+	  YYSIZE_T yyi;				\
+	  for (yyi = 0; yyi < (Count); yyi++)	\
+	    (To)[yyi] = (From)[yyi];		\
+	}					\
+      while (YYID (0))
+#  endif
+# endif
+
+/* Relocate STACK from its old location to the new one.  The
+   local variables YYSIZE and YYSTACKSIZE give the old and new number of
+   elements in the stack, and YYPTR gives the new location of the
+   stack.  Advance YYPTR to a properly aligned location for the next
+   stack.  */
+# define YYSTACK_RELOCATE(Stack_alloc, Stack)				\
+    do									\
+      {									\
+	YYSIZE_T yynewbytes;						\
+	YYCOPY (&yyptr->Stack_alloc, Stack, yysize);			\
+	Stack = &yyptr->Stack_alloc;					\
+	yynewbytes = yystacksize * sizeof (*Stack) + YYSTACK_GAP_MAXIMUM; \
+	yyptr += yynewbytes / sizeof (*yyptr);				\
+      }									\
+    while (YYID (0))
+
+#endif
+
+/* YYFINAL -- State number of the termination state.  */
+#define YYFINAL  5
+/* YYLAST -- Last index in YYTABLE.  */
+#define YYLAST   342
+
+/* YYNTOKENS -- Number of terminals.  */
+#define YYNTOKENS  115
+/* YYNNTS -- Number of nonterminals.  */
+#define YYNNTS  134
+/* YYNRULES -- Number of rules.  */
+#define YYNRULES  264
+/* YYNRULES -- Number of states.  */
+#define YYNSTATES  436
+
+/* YYTRANSLATE(YYLEX) -- Bison symbol number corresponding to YYLEX.  */
+#define YYUNDEFTOK  2
+#define YYMAXUTOK   360
+
+#define YYTRANSLATE(YYX)						\
+  ((unsigned int) (YYX) <= YYMAXUTOK ? yytranslate[YYX] : YYUNDEFTOK)
+
+/* YYTRANSLATE[YYLEX] -- Bison symbol number corresponding to YYLEX.  */
+static const yytype_uint8 yytranslate[] =
+{
+       0,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,   110,   107,   111,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,   106,
+       2,   112,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,   108,     2,   109,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,   113,     2,   114,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     1,     2,     3,     4,
+       5,     6,     7,     8,     9,    10,    11,    12,    13,    14,
+      15,    16,    17,    18,    19,    20,    21,    22,    23,    24,
+      25,    26,    27,    28,    29,    30,    31,    32,    33,    34,
+      35,    36,    37,    38,    39,    40,    41,    42,    43,    44,
+      45,    46,    47,    48,    49,    50,    51,    52,    53,    54,
+      55,    56,    57,    58,    59,    60,    61,    62,    63,    64,
+      65,    66,    67,    68,    69,    70,    71,    72,    73,    74,
+      75,    76,    77,    78,    79,    80,    81,    82,    83,    84,
+      85,    86,    87,    88,    89,    90,    91,    92,    93,    94,
+      95,    96,    97,    98,    99,   100,   101,   102,   103,   104,
+     105
+};
+
+#if YYDEBUG
+/* YYPRHS[YYN] -- Index of the first RHS symbol of rule number YYN in
+   YYRHS.  */
+static const yytype_uint16 yyprhs[] =
+{
+       0,     0,     3,     8,    10,    12,    15,    16,    20,    23,
+      24,    27,    30,    32,    34,    36,    38,    40,    42,    44,
+      46,    48,    50,    52,    57,    62,    67,    74,    81,    90,
+      99,   102,   105,   107,   109,   111,   113,   115,   117,   119,
+     121,   123,   125,   127,   129,   136,   140,   144,   147,   150,
+     158,   161,   163,   165,   167,   169,   174,   176,   178,   180,
+     182,   184,   186,   188,   192,   193,   196,   199,   201,   203,
+     205,   207,   209,   211,   213,   215,   217,   218,   220,   222,
+     224,   226,   227,   229,   231,   233,   235,   237,   239,   244,
+     247,   250,   252,   255,   257,   260,   262,   265,   270,   275,
+     277,   278,   282,   284,   286,   289,   291,   294,   296,   298,
+     302,   309,   310,   312,   315,   320,   322,   326,   328,   330,
+     332,   334,   336,   338,   340,   342,   344,   346,   349,   352,
+     355,   358,   361,   364,   367,   370,   373,   376,   379,   382,
+     386,   388,   390,   392,   398,   400,   402,   404,   407,   409,
+     411,   414,   416,   419,   426,   428,   432,   434,   436,   438,
+     440,   442,   447,   449,   451,   453,   455,   457,   459,   462,
+     464,   466,   472,   474,   477,   479,   481,   487,   490,   491,
+     498,   502,   503,   505,   507,   509,   511,   513,   516,   518,
+     520,   523,   528,   533,   534,   538,   540,   542,   544,   547,
+     549,   551,   553,   555,   561,   563,   567,   573,   579,   581,
+     585,   591,   593,   595,   597,   599,   601,   603,   605,   607,
+     609,   613,   619,   627,   637,   640,   643,   645,   647,   648,
+     649,   653,   654,   658,   662,   664,   669,   672,   675,   678,
+     681,   685,   688,   692,   693,   695,   697,   698,   700,   702,
+     703,   705,   707,   708,   710,   712,   713,   717,   718,   722,
+     723,   727,   729,   731,   733
+};
+
+/* YYRHS -- A `-1'-separated list of the rules' RHS.  */
+static const yytype_int16 yyrhs[] =
+{
+     116,     0,    -1,   117,   118,   120,    12,    -1,     3,    -1,
+       4,    -1,   118,   119,    -1,    -1,     8,    98,   106,    -1,
+     120,   121,    -1,    -1,   122,   106,    -1,   158,   106,    -1,
+     123,    -1,   124,    -1,   125,    -1,   126,    -1,   127,    -1,
+     128,    -1,   129,    -1,   130,    -1,   135,    -1,   131,    -1,
+     132,    -1,    19,   139,   107,   136,    -1,    18,   138,   107,
+     137,    -1,    16,   138,   107,   136,    -1,    14,   138,   107,
+     136,   107,   136,    -1,    13,   138,   107,   137,   107,   137,
+      -1,    17,   138,   107,   137,   107,   137,   107,   137,    -1,
+      15,   138,   107,   137,   107,   133,   107,   134,    -1,    20,
+     137,    -1,    82,   243,    -1,    83,    -1,    84,    -1,    85,
+      -1,    86,    -1,    87,    -1,    88,    -1,    89,    -1,    90,
+      -1,    91,    -1,    92,    -1,    93,    -1,    94,    -1,    21,
+     138,   107,   143,   107,   140,    -1,   229,   143,   155,    -1,
+     229,   143,   156,    -1,   144,   157,    -1,   152,   154,    -1,
+     141,   107,   141,   107,   141,   107,   141,    -1,   229,   142,
+      -1,    22,    -1,    98,    -1,    98,    -1,   160,    -1,   145,
+     108,   146,   109,    -1,   174,    -1,   236,    -1,    98,    -1,
+      98,    -1,   147,    -1,   148,    -1,    22,    -1,   152,   153,
+     149,    -1,    -1,   110,   150,    -1,   111,   151,    -1,    22,
+      -1,    22,    -1,    98,    -1,   102,    -1,   102,    -1,   102,
+      -1,   102,    -1,    99,    -1,   103,    -1,    -1,    99,    -1,
+     100,    -1,   101,    -1,   102,    -1,    -1,   159,    -1,   166,
+      -1,   230,    -1,   232,    -1,   235,    -1,   248,    -1,     7,
+      98,   112,   160,    -1,    95,   161,    -1,    37,   165,    -1,
+      59,    -1,    97,   163,    -1,    52,    -1,    28,   241,    -1,
+      36,    -1,    73,   242,    -1,    49,   108,   164,   109,    -1,
+      96,   108,   162,   109,    -1,    22,    -1,    -1,   108,   164,
+     109,    -1,    22,    -1,    59,    -1,    28,   241,    -1,    36,
+      -1,    73,   242,    -1,   167,    -1,   168,    -1,    10,    98,
+     170,    -1,    10,    98,   108,   169,   109,   171,    -1,    -1,
+      22,    -1,   112,   173,    -1,   112,   113,   172,   114,    -1,
+     175,    -1,   172,   107,   175,    -1,   177,    -1,   213,    -1,
+     223,    -1,   177,    -1,   213,    -1,   224,    -1,   176,    -1,
+     214,    -1,   223,    -1,   177,    -1,    72,   201,    -1,    72,
+     178,    -1,    72,   180,    -1,    72,   183,    -1,    72,   185,
+      -1,    72,   191,    -1,    72,   187,    -1,    72,   194,    -1,
+      72,   196,    -1,    72,   198,    -1,    72,   200,    -1,    72,
+     212,    -1,    46,   240,   179,    -1,   189,    -1,    32,    -1,
+      68,    -1,    42,   108,   190,   109,   181,    -1,   189,    -1,
+      59,    -1,    25,    -1,    71,   182,    -1,    39,    -1,    31,
+      -1,    43,   184,    -1,    24,    -1,   240,    66,    -1,    44,
+     108,   190,   109,   240,   186,    -1,   189,    -1,    74,   244,
+     188,    -1,    28,    -1,    24,    -1,    30,    -1,    70,    -1,
+      22,    -1,    75,   242,   192,   193,    -1,    34,    -1,    53,
+      -1,    78,    -1,    79,    -1,    77,    -1,    76,    -1,    35,
+     195,    -1,    28,    -1,    55,    -1,    27,   108,   197,   109,
+      56,    -1,    22,    -1,    57,   199,    -1,    69,    -1,    25,
+      -1,   203,    65,   108,   206,   109,    -1,   203,   202,    -1,
+      -1,    65,   108,   206,   104,   206,   109,    -1,    48,   207,
+     204,    -1,    -1,   205,    -1,    40,    -1,    81,    -1,    41,
+      -1,    22,    -1,    50,   208,    -1,    62,    -1,    51,    -1,
+      80,   242,    -1,    54,   108,   210,   109,    -1,    47,   108,
+     211,   109,    -1,    -1,   108,   209,   109,    -1,    22,    -1,
+      22,    -1,    22,    -1,    29,    63,    -1,   217,    -1,   220,
+      -1,   215,    -1,   218,    -1,    61,    33,   108,   216,   109,
+      -1,   221,    -1,   221,   104,   221,    -1,    61,    33,   108,
+     221,   109,    -1,    61,    45,   108,   219,   109,    -1,   222,
+      -1,   222,   104,   222,    -1,    61,    45,   108,   222,   109,
+      -1,    22,    -1,    22,    -1,   225,    -1,   227,    -1,   226,
+      -1,   227,    -1,   228,    -1,    23,    -1,    22,    -1,   113,
+     228,   114,    -1,   113,   228,   107,   228,   114,    -1,   113,
+     228,   107,   228,   107,   228,   114,    -1,   113,   228,   107,
+     228,   107,   228,   107,   228,   114,    -1,   229,    23,    -1,
+     229,    22,    -1,   110,    -1,   111,    -1,    -1,    -1,    11,
+     231,   234,    -1,    -1,     5,   233,   234,    -1,   234,   107,
+      98,    -1,    98,    -1,     9,    98,   112,   236,    -1,    64,
+      59,    -1,    64,    36,    -1,    64,   237,    -1,    64,    58,
+      -1,    64,    73,   242,    -1,    64,    29,    -1,    28,   238,
+     239,    -1,    -1,    38,    -1,    26,    -1,    -1,    60,    -1,
+      67,    -1,    -1,    38,    -1,    26,    -1,    -1,    60,    -1,
+      67,    -1,    -1,   108,   245,   109,    -1,    -1,   108,   246,
+     109,    -1,    -1,   108,   247,   109,    -1,    22,    -1,    22,
+      -1,    22,    -1,     6,    98,   112,    98,    -1
+};
+
+/* YYRLINE[YYN] -- source line where rule number YYN was defined.  */
+static const yytype_uint16 yyrline[] =
+{
+       0,   256,   256,   259,   267,   279,   280,   283,   305,   306,
+     309,   324,   327,   332,   339,   340,   341,   342,   343,   344,
+     345,   348,   349,   352,   358,   365,   372,   380,   387,   395,
+     440,   447,   453,   454,   455,   456,   457,   458,   459,   460,
+     461,   462,   463,   464,   467,   480,   493,   506,   528,   537,
+     570,   577,   592,   642,   684,   695,   716,   726,   732,   763,
+     780,   780,   782,   789,   801,   802,   803,   806,   818,   830,
+     848,   859,   871,   873,   874,   875,   876,   879,   879,   879,
+     879,   880,   883,   884,   885,   886,   887,   888,   891,   909,
+     913,   919,   923,   927,   931,   940,   949,   953,   958,   964,
+     975,   975,   976,   978,   982,   986,   990,   996,   996,   998,
+    1014,  1037,  1040,  1051,  1057,  1063,  1064,  1071,  1077,  1083,
+    1091,  1097,  1103,  1111,  1117,  1123,  1131,  1132,  1135,  1136,
+    1137,  1138,  1139,  1140,  1141,  1142,  1143,  1144,  1145,  1148,
+    1157,  1161,  1165,  1171,  1180,  1184,  1188,  1197,  1201,  1207,
+    1213,  1220,  1225,  1233,  1243,  1245,  1253,  1259,  1263,  1267,
+    1273,  1284,  1293,  1297,  1302,  1306,  1310,  1314,  1320,  1327,
+    1331,  1337,  1345,  1356,  1363,  1367,  1373,  1383,  1394,  1398,
+    1416,  1425,  1428,  1434,  1438,  1442,  1448,  1459,  1464,  1469,
+    1474,  1479,  1484,  1492,  1495,  1500,  1513,  1521,  1532,  1540,
+    1540,  1542,  1542,  1544,  1554,  1559,  1566,  1576,  1585,  1590,
+    1597,  1607,  1617,  1629,  1629,  1630,  1630,  1632,  1642,  1650,
+    1660,  1668,  1676,  1685,  1696,  1700,  1706,  1707,  1708,  1711,
+    1711,  1714,  1714,  1717,  1723,  1731,  1744,  1753,  1762,  1766,
+    1775,  1784,  1795,  1802,  1807,  1816,  1828,  1831,  1840,  1851,
+    1852,  1853,  1856,  1857,  1858,  1861,  1862,  1865,  1866,  1869,
+    1870,  1873,  1884,  1895,  1906
+};
+#endif
+
+#if YYDEBUG || YYERROR_VERBOSE || YYTOKEN_TABLE
+/* YYTNAME[SYMBOL-NUM] -- String name of the symbol SYMBOL-NUM.
+   First, the terminals, then, starting at YYNTOKENS, nonterminals.  */
+static const char *const yytname[] =
+{
+  "$end", "error", "$undefined", "ARBvp_10", "ARBfp_10", "ADDRESS",
+  "ALIAS", "ATTRIB", "OPTION", "OUTPUT", "PARAM", "TEMP", "END", "BIN_OP",
+  "BINSC_OP", "SAMPLE_OP", "SCALAR_OP", "TRI_OP", "VECTOR_OP", "ARL",
+  "KIL", "SWZ", "INTEGER", "REAL", "AMBIENT", "ATTENUATION", "BACK",
+  "CLIP", "COLOR", "DEPTH", "DIFFUSE", "DIRECTION", "EMISSION", "ENV",
+  "EYE", "FOG", "FOGCOORD", "FRAGMENT", "FRONT", "HALF", "INVERSE",
+  "INVTRANS", "LIGHT", "LIGHTMODEL", "LIGHTPROD", "LOCAL", "MATERIAL",
+  "MAT_PROGRAM", "MATRIX", "MATRIXINDEX", "MODELVIEW", "MVP", "NORMAL",
+  "OBJECT", "PALETTE", "PARAMS", "PLANE", "POINT_TOK", "POINTSIZE",
+  "POSITION", "PRIMARY", "PROGRAM", "PROJECTION", "RANGE", "RESULT", "ROW",
+  "SCENECOLOR", "SECONDARY", "SHININESS", "SIZE_TOK", "SPECULAR", "SPOT",
+  "STATE", "TEXCOORD", "TEXENV", "TEXGEN", "TEXGEN_Q", "TEXGEN_R",
+  "TEXGEN_S", "TEXGEN_T", "TEXTURE", "TRANSPOSE", "TEXTURE_UNIT", "TEX_1D",
+  "TEX_2D", "TEX_3D", "TEX_CUBE", "TEX_RECT", "TEX_SHADOW1D",
+  "TEX_SHADOW2D", "TEX_SHADOWRECT", "TEX_ARRAY1D", "TEX_ARRAY2D",
+  "TEX_ARRAYSHADOW1D", "TEX_ARRAYSHADOW2D", "VERTEX", "VTXATTRIB",
+  "WEIGHT", "IDENTIFIER", "MASK4", "MASK3", "MASK2", "MASK1", "SWIZZLE",
+  "DOT_DOT", "DOT", "';'", "','", "'['", "']'", "'+'", "'-'", "'='", "'{'",
+  "'}'", "$accept", "program", "language", "optionSequence", "option",
+  "statementSequence", "statement", "instruction", "ALU_instruction",
+  "TexInstruction", "ARL_instruction", "VECTORop_instruction",
+  "SCALARop_instruction", "BINSCop_instruction", "BINop_instruction",
+  "TRIop_instruction", "SAMPLE_instruction", "KIL_instruction",
+  "texImageUnit", "texTarget", "SWZ_instruction", "scalarSrcReg",
+  "swizzleSrcReg", "maskedDstReg", "maskedAddrReg", "extendedSwizzle",
+  "extSwizComp", "extSwizSel", "srcReg", "dstReg", "progParamArray",
+  "progParamArrayMem", "progParamArrayAbs", "progParamArrayRel",
+  "addrRegRelOffset", "addrRegPosOffset", "addrRegNegOffset", "addrReg",
+  "addrComponent", "addrWriteMask", "scalarSuffix", "swizzleSuffix",
+  "optionalMask", "namingStatement", "ATTRIB_statement", "attribBinding",
+  "vtxAttribItem", "vtxAttribNum", "vtxOptWeightNum", "vtxWeightNum",
+  "fragAttribItem", "PARAM_statement", "PARAM_singleStmt",
+  "PARAM_multipleStmt", "optArraySize", "paramSingleInit",
+  "paramMultipleInit", "paramMultInitList", "paramSingleItemDecl",
+  "paramSingleItemUse", "paramMultipleItem", "stateMultipleItem",
+  "stateSingleItem", "stateMaterialItem", "stateMatProperty",
+  "stateLightItem", "stateLightProperty", "stateSpotProperty",
+  "stateLightModelItem", "stateLModProperty", "stateLightProdItem",
+  "stateLProdProperty", "stateTexEnvItem", "stateTexEnvProperty",
+  "ambDiffSpecProperty", "stateLightNumber", "stateTexGenItem",
+  "stateTexGenType", "stateTexGenCoord", "stateFogItem",
+  "stateFogProperty", "stateClipPlaneItem", "stateClipPlaneNum",
+  "statePointItem", "statePointProperty", "stateMatrixRow",
+  "stateMatrixRows", "optMatrixRows", "stateMatrixItem",
+  "stateOptMatModifier", "stateMatModifier", "stateMatrixRowNum",
+  "stateMatrixName", "stateOptModMatNum", "stateModMatNum",
+  "statePaletteMatNum", "stateProgramMatNum", "stateDepthItem",
+  "programSingleItem", "programMultipleItem", "progEnvParams",
+  "progEnvParamNums", "progEnvParam", "progLocalParams",
+  "progLocalParamNums", "progLocalParam", "progEnvParamNum",
+  "progLocalParamNum", "paramConstDecl", "paramConstUse",
+  "paramConstScalarDecl", "paramConstScalarUse", "paramConstVector",
+  "signedFloatConstant", "optionalSign", "TEMP_statement", "@1",
+  "ADDRESS_statement", "@2", "varNameList", "OUTPUT_statement",
+  "resultBinding", "resultColBinding", "optResultFaceType",
+  "optResultColorType", "optFaceType", "optColorType",
+  "optTexCoordUnitNum", "optTexImageUnitNum", "optLegacyTexUnitNum",
+  "texCoordUnitNum", "texImageUnitNum", "legacyTexUnitNum",
+  "ALIAS_statement", 0
+};
+#endif
+
+# ifdef YYPRINT
+/* YYTOKNUM[YYLEX-NUM] -- Internal token number corresponding to
+   token YYLEX-NUM.  */
+static const yytype_uint16 yytoknum[] =
+{
+       0,   256,   257,   258,   259,   260,   261,   262,   263,   264,
+     265,   266,   267,   268,   269,   270,   271,   272,   273,   274,
+     275,   276,   277,   278,   279,   280,   281,   282,   283,   284,
+     285,   286,   287,   288,   289,   290,   291,   292,   293,   294,
+     295,   296,   297,   298,   299,   300,   301,   302,   303,   304,
+     305,   306,   307,   308,   309,   310,   311,   312,   313,   314,
+     315,   316,   317,   318,   319,   320,   321,   322,   323,   324,
+     325,   326,   327,   328,   329,   330,   331,   332,   333,   334,
+     335,   336,   337,   338,   339,   340,   341,   342,   343,   344,
+     345,   346,   347,   348,   349,   350,   351,   352,   353,   354,
+     355,   356,   357,   358,   359,   360,    59,    44,    91,    93,
+      43,    45,    61,   123,   125
+};
+# endif
+
+/* YYR1[YYN] -- Symbol number of symbol that rule YYN derives.  */
+static const yytype_uint8 yyr1[] =
+{
+       0,   115,   116,   117,   117,   118,   118,   119,   120,   120,
+     121,   121,   122,   122,   123,   123,   123,   123,   123,   123,
+     123,   124,   124,   125,   126,   127,   128,   129,   130,   131,
+     132,   133,   134,   134,   134,   134,   134,   134,   134,   134,
+     134,   134,   134,   134,   135,   136,   137,   138,   139,   140,
+     141,   142,   142,   143,   143,   143,   143,   144,   144,   145,
+     146,   146,   147,   148,   149,   149,   149,   150,   151,   152,
+     153,   154,   155,   156,   156,   156,   156,   157,   157,   157,
+     157,   157,   158,   158,   158,   158,   158,   158,   159,   160,
+     160,   161,   161,   161,   161,   161,   161,   161,   161,   162,
+     163,   163,   164,   165,   165,   165,   165,   166,   166,   167,
+     168,   169,   169,   170,   171,   172,   172,   173,   173,   173,
+     174,   174,   174,   175,   175,   175,   176,   176,   177,   177,
+     177,   177,   177,   177,   177,   177,   177,   177,   177,   178,
+     179,   179,   179,   180,   181,   181,   181,   181,   181,   182,
+     183,   184,   184,   185,   186,   187,   188,   189,   189,   189,
+     190,   191,   192,   192,   193,   193,   193,   193,   194,   195,
+     195,   196,   197,   198,   199,   199,   200,   201,   202,   202,
+     203,   204,   204,   205,   205,   205,   206,   207,   207,   207,
+     207,   207,   207,   208,   208,   209,   210,   211,   212,   213,
+     213,   214,   214,   215,   216,   216,   217,   218,   219,   219,
+     220,   221,   222,   223,   223,   224,   224,   225,   226,   226,
+     227,   227,   227,   227,   228,   228,   229,   229,   229,   231,
+     230,   233,   232,   234,   234,   235,   236,   236,   236,   236,
+     236,   236,   237,   238,   238,   238,   239,   239,   239,   240,
+     240,   240,   241,   241,   241,   242,   242,   243,   243,   244,
+     244,   245,   246,   247,   248
+};
+
+/* YYR2[YYN] -- Number of symbols composing right hand side of rule YYN.  */
+static const yytype_uint8 yyr2[] =
+{
+       0,     2,     4,     1,     1,     2,     0,     3,     2,     0,
+       2,     2,     1,     1,     1,     1,     1,     1,     1,     1,
+       1,     1,     1,     4,     4,     4,     6,     6,     8,     8,
+       2,     2,     1,     1,     1,     1,     1,     1,     1,     1,
+       1,     1,     1,     1,     6,     3,     3,     2,     2,     7,
+       2,     1,     1,     1,     1,     4,     1,     1,     1,     1,
+       1,     1,     1,     3,     0,     2,     2,     1,     1,     1,
+       1,     1,     1,     1,     1,     1,     0,     1,     1,     1,
+       1,     0,     1,     1,     1,     1,     1,     1,     4,     2,
+       2,     1,     2,     1,     2,     1,     2,     4,     4,     1,
+       0,     3,     1,     1,     2,     1,     2,     1,     1,     3,
+       6,     0,     1,     2,     4,     1,     3,     1,     1,     1,
+       1,     1,     1,     1,     1,     1,     1,     2,     2,     2,
+       2,     2,     2,     2,     2,     2,     2,     2,     2,     3,
+       1,     1,     1,     5,     1,     1,     1,     2,     1,     1,
+       2,     1,     2,     6,     1,     3,     1,     1,     1,     1,
+       1,     4,     1,     1,     1,     1,     1,     1,     2,     1,
+       1,     5,     1,     2,     1,     1,     5,     2,     0,     6,
+       3,     0,     1,     1,     1,     1,     1,     2,     1,     1,
+       2,     4,     4,     0,     3,     1,     1,     1,     2,     1,
+       1,     1,     1,     5,     1,     3,     5,     5,     1,     3,
+       5,     1,     1,     1,     1,     1,     1,     1,     1,     1,
+       3,     5,     7,     9,     2,     2,     1,     1,     0,     0,
+       3,     0,     3,     3,     1,     4,     2,     2,     2,     2,
+       3,     2,     3,     0,     1,     1,     0,     1,     1,     0,
+       1,     1,     0,     1,     1,     0,     3,     0,     3,     0,
+       3,     1,     1,     1,     4
+};
+
+/* YYDEFACT[STATE-NAME] -- Default rule to reduce with in state
+   STATE-NUM when YYTABLE doesn't specify something else to do.  Zero
+   means the default is an error.  */
+static const yytype_uint16 yydefact[] =
+{
+       0,     3,     4,     0,     6,     1,     9,     0,     5,     0,
+       0,   231,     0,     0,     0,     0,   229,     2,     0,     0,
+       0,     0,     0,     0,     0,   228,     0,     8,     0,    12,
+      13,    14,    15,    16,    17,    18,    19,    21,    22,    20,
+       0,    82,    83,   107,   108,    84,    85,    86,    87,     7,
+       0,     0,     0,     0,     0,     0,     0,    58,     0,    81,
+      57,     0,     0,     0,     0,     0,    69,     0,     0,   226,
+     227,    30,     0,     0,    10,    11,   234,   232,     0,     0,
+       0,   111,   228,   109,   230,   243,   241,   237,   239,   236,
+     255,   238,   228,    77,    78,    79,    80,    47,   228,   228,
+     228,   228,   228,   228,    71,    48,   219,   218,     0,     0,
+       0,     0,    53,   228,    76,     0,    54,    56,   120,   121,
+     199,   200,   122,   215,   216,     0,     0,   264,    88,   235,
+     112,     0,   113,   117,   118,   119,   213,   214,   217,     0,
+     245,   244,   246,     0,   240,     0,     0,     0,     0,    25,
+       0,    24,    23,   252,   105,   103,   255,    90,     0,     0,
+       0,     0,     0,     0,   249,     0,   249,     0,     0,   259,
+     255,   128,   129,   130,   131,   133,   132,   134,   135,   136,
+     137,     0,   138,   252,    95,     0,    93,    91,   255,     0,
+     100,    89,     0,    74,    73,    75,    46,     0,     0,   233,
+       0,   225,   224,   247,   248,   242,   261,     0,   228,   228,
+       0,     0,   228,   253,   254,   104,   106,     0,     0,     0,
+     198,   169,   170,   168,     0,   151,   251,   250,   150,     0,
+       0,     0,     0,   193,   189,     0,   188,   255,   181,   175,
+     174,   173,     0,     0,     0,     0,    94,     0,    96,     0,
+       0,    92,   228,   220,    62,     0,    60,    61,     0,   228,
+       0,   110,   256,    27,    26,    72,    45,   257,     0,     0,
+     211,     0,   212,     0,   172,     0,   160,     0,   152,     0,
+     157,   158,   141,   142,   159,   139,   140,     0,     0,   187,
+       0,   190,   183,   185,   184,   180,   182,   263,     0,   156,
+     155,   162,   163,     0,     0,   102,     0,    99,     0,     0,
+       0,    55,    70,    64,    44,     0,     0,   228,     0,    31,
+       0,   228,   206,   210,     0,     0,   249,   197,     0,   195,
+       0,   196,     0,   260,   167,   166,   164,   165,   161,   186,
+       0,    97,    98,   101,   228,   221,     0,     0,    63,   228,
+      51,    52,    50,     0,     0,     0,   115,   123,   126,   124,
+     201,   202,   125,   262,     0,    32,    33,    34,    35,    36,
+      37,    38,    39,    40,    41,    42,    43,    29,    28,   171,
+     146,   148,   145,     0,   143,   144,     0,   192,   194,   191,
+     176,     0,    67,    65,    68,    66,     0,     0,     0,   127,
+     178,   228,   114,   258,   149,   147,   153,   154,   228,   222,
+     228,     0,     0,     0,   177,   116,     0,     0,     0,   204,
+       0,   208,     0,   223,   228,   203,     0,   207,     0,     0,
+      49,   205,   209,     0,     0,   179
+};
+
+/* YYDEFGOTO[NTERM-NUM].  */
+static const yytype_int16 yydefgoto[] =
+{
+      -1,     3,     4,     6,     8,     9,    27,    28,    29,    30,
+      31,    32,    33,    34,    35,    36,    37,    38,   268,   377,
+      39,   146,    71,    58,    67,   314,   315,   352,   114,    59,
+     115,   255,   256,   257,   348,   393,   395,    68,   313,   105,
+     266,   196,    97,    40,    41,   116,   191,   308,   251,   306,
+     157,    42,    43,    44,   131,    83,   261,   355,   132,   117,
+     356,   357,   118,   171,   285,   172,   384,   405,   173,   228,
+     174,   406,   175,   300,   286,   277,   176,   303,   338,   177,
+     223,   178,   275,   179,   241,   180,   399,   414,   181,   295,
+     296,   340,   238,   289,   330,   332,   328,   182,   119,   359,
+     360,   418,   120,   361,   420,   121,   271,   273,   362,   122,
+     136,   123,   124,   138,    72,    45,    55,    46,    50,    77,
+      47,    60,    91,   142,   205,   229,   215,   144,   319,   243,
+     207,   364,   298,    48
+};
+
+/* YYPACT[STATE-NUM] -- Index in YYTABLE of the portion describing
+   STATE-NUM.  */
+#define YYPACT_NINF -369
+static const yytype_int16 yypact[] =
+{
+     143,  -369,  -369,    36,  -369,  -369,    45,   -39,  -369,   169,
+     -33,  -369,   -19,    -6,    -4,    12,  -369,  -369,   -34,   -34,
+     -34,   -34,   -34,   -34,    15,    62,   -34,  -369,    26,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+      60,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+      20,    56,   107,   110,    37,    20,    -3,  -369,   111,   109,
+    -369,   113,   114,   116,   117,   118,  -369,   119,   125,  -369,
+    -369,  -369,   -15,   121,  -369,  -369,  -369,   122,   132,   -18,
+     167,   210,   -11,  -369,   122,    63,  -369,  -369,  -369,  -369,
+     130,  -369,    62,  -369,  -369,  -369,  -369,  -369,    62,    62,
+      62,    62,    62,    62,  -369,  -369,  -369,  -369,     9,    72,
+      87,    -1,   131,    62,   104,   134,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,  -369,   -15,   142,  -369,  -369,  -369,
+    -369,   135,  -369,  -369,  -369,  -369,  -369,  -369,  -369,   182,
+    -369,  -369,    52,   219,  -369,   138,   139,   -15,   140,  -369,
+     141,  -369,  -369,    61,  -369,  -369,   130,  -369,   144,   145,
+     146,   180,    11,   147,    85,   148,    99,    89,    -2,   149,
+     130,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+    -369,   184,  -369,    61,  -369,   150,  -369,  -369,   130,   151,
+     152,  -369,    27,  -369,  -369,  -369,  -369,   -10,   154,  -369,
+     153,  -369,  -369,  -369,  -369,  -369,  -369,   155,    62,    62,
+     161,   168,    62,  -369,  -369,  -369,  -369,   229,   244,   246,
+    -369,  -369,  -369,  -369,   247,  -369,  -369,  -369,  -369,   204,
+     247,    17,   163,   164,  -369,   165,  -369,   130,    67,  -369,
+    -369,  -369,   252,   248,    18,   170,  -369,   253,  -369,   255,
+     253,  -369,    62,  -369,  -369,   171,  -369,  -369,   177,    62,
+     172,  -369,  -369,  -369,  -369,  -369,  -369,   173,   175,   176,
+    -369,   178,  -369,   179,  -369,   181,  -369,   183,  -369,   185,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,   262,   264,  -369,
+     267,  -369,  -369,  -369,  -369,  -369,  -369,  -369,   186,  -369,
+    -369,  -369,  -369,   136,   269,  -369,   187,  -369,   188,   190,
+      43,  -369,  -369,   106,  -369,   193,    -5,    -7,   271,  -369,
+     108,    62,  -369,  -369,   245,     4,    99,  -369,   194,  -369,
+     195,  -369,   196,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+     197,  -369,  -369,  -369,    62,  -369,   280,   285,  -369,    62,
+    -369,  -369,  -369,    93,    87,    53,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,   199,  -369,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,   278,  -369,  -369,     8,  -369,  -369,  -369,
+    -369,    57,  -369,  -369,  -369,  -369,   203,   205,   206,  -369,
+     250,    -7,  -369,  -369,  -369,  -369,  -369,  -369,    62,  -369,
+      62,   229,   244,   208,  -369,  -369,   198,   211,   202,   213,
+     214,   218,   269,  -369,    62,  -369,   229,  -369,   244,    54,
+    -369,  -369,  -369,   269,   215,  -369
+};
+
+/* YYPGOTO[NTERM-NUM].  */
+static const yytype_int16 yypgoto[] =
+{
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+    -369,   -94,   -88,   133,  -369,  -369,  -334,  -369,   -85,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,   128,  -369,  -369,
+    -369,  -369,  -369,  -369,  -369,   251,  -369,  -369,  -369,    77,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+     -72,  -369,   -81,  -369,  -369,  -369,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,  -305,   101,  -369,  -369,  -369,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -369,  -369,   -22,  -369,
+    -369,  -336,  -369,  -369,  -369,  -369,  -369,  -369,   254,  -369,
+    -369,  -369,  -369,  -369,  -369,  -369,  -342,  -368,   256,  -369,
+    -369,  -369,   -80,  -110,   -82,  -369,  -369,  -369,  -369,   279,
+    -369,   257,  -369,  -369,  -369,  -161,   156,  -146,  -369,  -369,
+    -369,  -369,  -369,  -369
+};
+
+/* YYTABLE[YYPACT[STATE-NUM]].  What to do in state STATE-NUM.  If
+   positive, shift that token.  If negative, reduce the rule which
+   number is the opposite.  If zero, do what YYDEFACT says.
+   If YYTABLE_NINF, syntax error.  */
+#define YYTABLE_NINF -60
+static const yytype_int16 yytable[] =
+{
+     139,   133,   137,   192,   145,   231,   149,   106,   107,   152,
+     216,   148,   254,   150,   151,   396,   147,   350,   147,   108,
+     385,   147,   108,   239,   244,    85,    86,   183,   280,   380,
+      56,   139,   280,    87,   281,   184,     5,   153,   281,   221,
+     198,   280,   248,   381,   421,   154,   109,   281,   185,   282,
+     109,   186,   301,     7,   353,    88,    89,   110,   187,    10,
+     432,   110,   210,   382,    57,   354,   222,   240,   155,   419,
+      90,   302,   188,    49,   284,   383,   417,   111,   284,    51,
+     111,   407,   156,   112,   431,   283,   429,   284,    66,   140,
+     430,   291,    52,   351,    53,   189,   190,   434,   113,    69,
+      70,   141,   113,    69,    70,   158,   113,   292,   293,   225,
+      54,   226,   203,    66,   160,   264,   161,   159,    76,   204,
+     263,   213,   162,   227,   269,   226,   397,   147,   214,   163,
+     164,   165,    74,   166,   252,   167,   232,   227,   398,   233,
+     234,   253,   310,   235,   168,    81,     1,     2,   294,    82,
+     344,   236,    61,    62,    63,    64,    65,   345,   433,    73,
+     401,   169,   170,   390,   408,   386,    75,   402,    78,   237,
+     139,   409,    69,    70,    11,    12,    13,   316,    14,    15,
+      16,    17,    18,    19,    20,    21,    22,    23,    24,    25,
+      26,   365,   366,   367,   368,   369,   370,   371,   372,   373,
+     374,   375,   376,   193,   201,   202,   194,   195,    93,    94,
+      95,    96,   334,   335,   336,   337,   346,   347,    92,    79,
+      98,    99,    80,   100,   101,   102,   103,   104,   125,   126,
+     127,    56,   130,   378,   391,   139,   358,   137,   143,   -59,
+     199,   206,   197,   220,   200,   208,   209,   211,   212,   245,
+     267,   270,   217,   218,   219,   224,   230,   242,   247,   249,
+     250,   259,   139,   265,   262,   260,   272,   316,   274,   276,
+     278,   287,   288,   290,   297,   305,   299,   307,   304,   312,
+     311,   318,   320,   321,   327,   317,   329,   322,   323,   331,
+     324,   339,   325,   363,   326,   333,   341,   342,   416,   343,
+     349,   379,   392,   387,   388,   389,   390,   394,   403,   404,
+     410,   425,   423,   411,   412,   413,   422,   426,   424,   139,
+     358,   137,   428,   427,   435,   258,   139,   309,   316,   415,
+     128,   279,   400,     0,    84,     0,   134,   129,   135,   246,
+       0,     0,   316
+};
+
+static const yytype_int16 yycheck[] =
+{
+      82,    82,    82,   113,    92,   166,   100,    22,    23,   103,
+     156,    99,    22,   101,   102,   349,    98,    22,   100,    37,
+     325,   103,    37,    25,   170,    28,    29,    28,    24,    25,
+      64,   113,    24,    36,    30,    36,     0,    28,    30,    28,
+     125,    24,   188,    39,   412,    36,    61,    30,    49,    32,
+      61,    52,    34,     8,    61,    58,    59,    72,    59,    98,
+     428,    72,   147,    59,    98,    72,    55,    69,    59,   411,
+      73,    53,    73,   106,    70,    71,   410,    95,    70,    98,
+      95,   386,    73,    98,   426,    68,   422,    70,    98,    26,
+     424,   237,    98,    98,    98,    96,    97,   433,   113,   110,
+     111,    38,   113,   110,   111,    33,   113,    40,    41,    24,
+      98,    26,    60,    98,    27,   209,    29,    45,    98,    67,
+     208,    60,    35,    38,   212,    26,    33,   209,    67,    42,
+      43,    44,   106,    46,   107,    48,    47,    38,    45,    50,
+      51,   114,   252,    54,    57,   108,     3,     4,    81,   112,
+     107,    62,    19,    20,    21,    22,    23,   114,   104,    26,
+     107,    74,    75,   109,   107,   326,   106,   114,   112,    80,
+     252,   114,   110,   111,     5,     6,     7,   259,     9,    10,
+      11,    12,    13,    14,    15,    16,    17,    18,    19,    20,
+      21,    83,    84,    85,    86,    87,    88,    89,    90,    91,
+      92,    93,    94,    99,    22,    23,   102,   103,    99,   100,
+     101,   102,    76,    77,    78,    79,   110,   111,   107,   112,
+     107,   107,   112,   107,   107,   107,   107,   102,   107,   107,
+      98,    64,    22,   321,   344,   317,   317,   317,   108,   108,
+      98,    22,   108,    63,   109,   107,   107,   107,   107,    65,
+      82,    22,   108,   108,   108,   108,   108,   108,   108,   108,
+     108,   107,   344,   102,   109,   112,    22,   349,    22,    22,
+      66,   108,   108,   108,    22,    22,    28,    22,   108,   102,
+     109,   108,   107,   107,    22,   113,    22,   109,   109,    22,
+     109,    22,   109,    22,   109,   109,   109,   109,   408,   109,
+     107,    56,    22,   109,   109,   109,   109,    22,   109,    31,
+     107,   109,   114,   108,   108,    65,   108,   104,   107,   401,
+     401,   401,   104,   109,   109,   197,   408,   250,   410,   401,
+      79,   230,   354,    -1,    55,    -1,    82,    80,    82,   183,
+      -1,    -1,   424
+};
+
+/* YYSTOS[STATE-NUM] -- The (internal number of the) accessing
+   symbol of state STATE-NUM.  */
+static const yytype_uint8 yystos[] =
+{
+       0,     3,     4,   116,   117,     0,   118,     8,   119,   120,
+      98,     5,     6,     7,     9,    10,    11,    12,    13,    14,
+      15,    16,    17,    18,    19,    20,    21,   121,   122,   123,
+     124,   125,   126,   127,   128,   129,   130,   131,   132,   135,
+     158,   159,   166,   167,   168,   230,   232,   235,   248,   106,
+     233,    98,    98,    98,    98,   231,    64,    98,   138,   144,
+     236,   138,   138,   138,   138,   138,    98,   139,   152,   110,
+     111,   137,   229,   138,   106,   106,    98,   234,   112,   112,
+     112,   108,   112,   170,   234,    28,    29,    36,    58,    59,
+      73,   237,   107,    99,   100,   101,   102,   157,   107,   107,
+     107,   107,   107,   107,   102,   154,    22,    23,    37,    61,
+      72,    95,    98,   113,   143,   145,   160,   174,   177,   213,
+     217,   220,   224,   226,   227,   107,   107,    98,   160,   236,
+      22,   169,   173,   177,   213,   223,   225,   227,   228,   229,
+      26,    38,   238,   108,   242,   137,   136,   229,   137,   136,
+     137,   137,   136,    28,    36,    59,    73,   165,    33,    45,
+      27,    29,    35,    42,    43,    44,    46,    48,    57,    74,
+      75,   178,   180,   183,   185,   187,   191,   194,   196,   198,
+     200,   203,   212,    28,    36,    49,    52,    59,    73,    96,
+      97,   161,   228,    99,   102,   103,   156,   108,   143,    98,
+     109,    22,    23,    60,    67,   239,    22,   245,   107,   107,
+     143,   107,   107,    60,    67,   241,   242,   108,   108,   108,
+      63,    28,    55,   195,   108,    24,    26,    38,   184,   240,
+     108,   240,    47,    50,    51,    54,    62,    80,   207,    25,
+      69,   199,   108,   244,   242,    65,   241,   108,   242,   108,
+     108,   163,   107,   114,    22,   146,   147,   148,   152,   107,
+     112,   171,   109,   137,   136,   102,   155,    82,   133,   137,
+      22,   221,    22,   222,    22,   197,    22,   190,    66,   190,
+      24,    30,    32,    68,    70,   179,   189,   108,   108,   208,
+     108,   242,    40,    41,    81,   204,   205,    22,   247,    28,
+     188,    34,    53,   192,   108,    22,   164,    22,   162,   164,
+     228,   109,   102,   153,   140,   141,   229,   113,   108,   243,
+     107,   107,   109,   109,   109,   109,   109,    22,   211,    22,
+     209,    22,   210,   109,    76,    77,    78,    79,   193,    22,
+     206,   109,   109,   109,   107,   114,   110,   111,   149,   107,
+      22,    98,   142,    61,    72,   172,   175,   176,   177,   214,
+     215,   218,   223,    22,   246,    83,    84,    85,    86,    87,
+      88,    89,    90,    91,    92,    93,    94,   134,   137,    56,
+      25,    39,    59,    71,   181,   189,   240,   109,   109,   109,
+     109,   228,    22,   150,    22,   151,   141,    33,    45,   201,
+     203,   107,   114,   109,    31,   182,   186,   189,   107,   114,
+     107,   108,   108,    65,   202,   175,   228,   141,   216,   221,
+     219,   222,   108,   114,   107,   109,   104,   109,   104,   206,
+     141,   221,   222,   104,   206,   109
+};
+
+#define yyerrok		(yyerrstatus = 0)
+#define yyclearin	(yychar = YYEMPTY)
+#define YYEMPTY		(-2)
+#define YYEOF		0
+
+#define YYACCEPT	goto yyacceptlab
+#define YYABORT		goto yyabortlab
+#define YYERROR		goto yyerrorlab
+
+
+/* Like YYERROR except do call yyerror.  This remains here temporarily
+   to ease the transition to the new meaning of YYERROR, for GCC.
+   Once GCC version 2 has supplanted version 1, this can go.  */
+
+#define YYFAIL		goto yyerrlab
+
+#define YYRECOVERING()  (!!yyerrstatus)
+
+#define YYBACKUP(Token, Value)					\
+do								\
+  if (yychar == YYEMPTY && yylen == 1)				\
+    {								\
+      yychar = (Token);						\
+      yylval = (Value);						\
+      yytoken = YYTRANSLATE (yychar);				\
+      YYPOPSTACK (1);						\
+      goto yybackup;						\
+    }								\
+  else								\
+    {								\
+      yyerror (&yylloc, state, YY_("syntax error: cannot back up")); \
+      YYERROR;							\
+    }								\
+while (YYID (0))
+
+
+#define YYTERROR	1
+#define YYERRCODE	256
+
+
+/* YYLLOC_DEFAULT -- Set CURRENT to span from RHS[1] to RHS[N].
+   If N is 0, then set CURRENT to the empty location which ends
+   the previous symbol: RHS[0] (always defined).  */
+
+#define YYRHSLOC(Rhs, K) ((Rhs)[K])
+#ifndef YYLLOC_DEFAULT
+# define YYLLOC_DEFAULT(Current, Rhs, N)				\
+    do									\
+      if (YYID (N))                                                    \
+	{								\
+	  (Current).first_line   = YYRHSLOC (Rhs, 1).first_line;	\
+	  (Current).first_column = YYRHSLOC (Rhs, 1).first_column;	\
+	  (Current).last_line    = YYRHSLOC (Rhs, N).last_line;		\
+	  (Current).last_column  = YYRHSLOC (Rhs, N).last_column;	\
+	}								\
+      else								\
+	{								\
+	  (Current).first_line   = (Current).last_line   =		\
+	    YYRHSLOC (Rhs, 0).last_line;				\
+	  (Current).first_column = (Current).last_column =		\
+	    YYRHSLOC (Rhs, 0).last_column;				\
+	}								\
+    while (YYID (0))
+#endif
+
+
+/* YY_LOCATION_PRINT -- Print the location on the stream.
+   This macro was not mandated originally: define only if we know
+   we won't break user code: when these are the locations we know.  */
+
+#ifndef YY_LOCATION_PRINT
+# if YYLTYPE_IS_TRIVIAL
+#  define YY_LOCATION_PRINT(File, Loc)			\
+     fprintf (File, "%d.%d-%d.%d",			\
+	      (Loc).first_line, (Loc).first_column,	\
+	      (Loc).last_line,  (Loc).last_column)
+# else
+#  define YY_LOCATION_PRINT(File, Loc) ((void) 0)
+# endif
+#endif
+
+
+/* YYLEX -- calling `yylex' with the right arguments.  */
+
+#ifdef YYLEX_PARAM
+# define YYLEX yylex (&yylval, &yylloc, YYLEX_PARAM)
+#else
+# define YYLEX yylex (&yylval, &yylloc, scanner)
+#endif
+
+/* Enable debugging if requested.  */
+#if YYDEBUG
+
+# ifndef YYFPRINTF
+#  include <stdio.h> /* INFRINGES ON USER NAME SPACE */
+#  define YYFPRINTF fprintf
+# endif
+
+# define YYDPRINTF(Args)			\
+do {						\
+  if (yydebug)					\
+    YYFPRINTF Args;				\
+} while (YYID (0))
+
+# define YY_SYMBOL_PRINT(Title, Type, Value, Location)			  \
+do {									  \
+  if (yydebug)								  \
+    {									  \
+      YYFPRINTF (stderr, "%s ", Title);					  \
+      yy_symbol_print (stderr,						  \
+		  Type, Value, Location, state); \
+      YYFPRINTF (stderr, "\n");						  \
+    }									  \
+} while (YYID (0))
+
+
+/*--------------------------------.
+| Print this symbol on YYOUTPUT.  |
+`--------------------------------*/
+
+/*ARGSUSED*/
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static void
+yy_symbol_value_print (FILE *yyoutput, int yytype, YYSTYPE const * const yyvaluep, YYLTYPE const * const yylocationp, struct asm_parser_state *state)
+#else
+static void
+yy_symbol_value_print (yyoutput, yytype, yyvaluep, yylocationp, state)
+    FILE *yyoutput;
+    int yytype;
+    YYSTYPE const * const yyvaluep;
+    YYLTYPE const * const yylocationp;
+    struct asm_parser_state *state;
+#endif
+{
+  if (!yyvaluep)
+    return;
+  YYUSE (yylocationp);
+  YYUSE (state);
+# ifdef YYPRINT
+  if (yytype < YYNTOKENS)
+    YYPRINT (yyoutput, yytoknum[yytype], *yyvaluep);
+# else
+  YYUSE (yyoutput);
+# endif
+  switch (yytype)
+    {
+      default:
+	break;
+    }
+}
+
+
+/*--------------------------------.
+| Print this symbol on YYOUTPUT.  |
+`--------------------------------*/
+
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static void
+yy_symbol_print (FILE *yyoutput, int yytype, YYSTYPE const * const yyvaluep, YYLTYPE const * const yylocationp, struct asm_parser_state *state)
+#else
+static void
+yy_symbol_print (yyoutput, yytype, yyvaluep, yylocationp, state)
+    FILE *yyoutput;
+    int yytype;
+    YYSTYPE const * const yyvaluep;
+    YYLTYPE const * const yylocationp;
+    struct asm_parser_state *state;
+#endif
+{
+  if (yytype < YYNTOKENS)
+    YYFPRINTF (yyoutput, "token %s (", yytname[yytype]);
+  else
+    YYFPRINTF (yyoutput, "nterm %s (", yytname[yytype]);
+
+  YY_LOCATION_PRINT (yyoutput, *yylocationp);
+  YYFPRINTF (yyoutput, ": ");
+  yy_symbol_value_print (yyoutput, yytype, yyvaluep, yylocationp, state);
+  YYFPRINTF (yyoutput, ")");
+}
+
+/*------------------------------------------------------------------.
+| yy_stack_print -- Print the state stack from its BOTTOM up to its |
+| TOP (included).                                                   |
+`------------------------------------------------------------------*/
+
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static void
+yy_stack_print (yytype_int16 *yybottom, yytype_int16 *yytop)
+#else
+static void
+yy_stack_print (yybottom, yytop)
+    yytype_int16 *yybottom;
+    yytype_int16 *yytop;
+#endif
+{
+  YYFPRINTF (stderr, "Stack now");
+  for (; yybottom <= yytop; yybottom++)
+    {
+      int yybot = *yybottom;
+      YYFPRINTF (stderr, " %d", yybot);
+    }
+  YYFPRINTF (stderr, "\n");
+}
+
+# define YY_STACK_PRINT(Bottom, Top)				\
+do {								\
+  if (yydebug)							\
+    yy_stack_print ((Bottom), (Top));				\
+} while (YYID (0))
+
+
+/*------------------------------------------------.
+| Report that the YYRULE is going to be reduced.  |
+`------------------------------------------------*/
+
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static void
+yy_reduce_print (YYSTYPE *yyvsp, YYLTYPE *yylsp, int yyrule, struct asm_parser_state *state)
+#else
+static void
+yy_reduce_print (yyvsp, yylsp, yyrule, state)
+    YYSTYPE *yyvsp;
+    YYLTYPE *yylsp;
+    int yyrule;
+    struct asm_parser_state *state;
+#endif
+{
+  int yynrhs = yyr2[yyrule];
+  int yyi;
+  unsigned long int yylno = yyrline[yyrule];
+  YYFPRINTF (stderr, "Reducing stack by rule %d (line %lu):\n",
+	     yyrule - 1, yylno);
+  /* The symbols being reduced.  */
+  for (yyi = 0; yyi < yynrhs; yyi++)
+    {
+      YYFPRINTF (stderr, "   $%d = ", yyi + 1);
+      yy_symbol_print (stderr, yyrhs[yyprhs[yyrule] + yyi],
+		       &(yyvsp[(yyi + 1) - (yynrhs)])
+		       , &(yylsp[(yyi + 1) - (yynrhs)])		       , state);
+      YYFPRINTF (stderr, "\n");
+    }
+}
+
+# define YY_REDUCE_PRINT(Rule)		\
+do {					\
+  if (yydebug)				\
+    yy_reduce_print (yyvsp, yylsp, Rule, state); \
+} while (YYID (0))
+
+/* Nonzero means print parse trace.  It is left uninitialized so that
+   multiple parsers can coexist.  */
+int yydebug;
+#else /* !YYDEBUG */
+# define YYDPRINTF(Args)
+# define YY_SYMBOL_PRINT(Title, Type, Value, Location)
+# define YY_STACK_PRINT(Bottom, Top)
+# define YY_REDUCE_PRINT(Rule)
+#endif /* !YYDEBUG */
+
+
+/* YYINITDEPTH -- initial size of the parser's stacks.  */
+#ifndef	YYINITDEPTH
+# define YYINITDEPTH 200
+#endif
+
+/* YYMAXDEPTH -- maximum size the stacks can grow to (effective only
+   if the built-in stack extension method is used).
+
+   Do not make this value too large; the results are undefined if
+   YYSTACK_ALLOC_MAXIMUM < YYSTACK_BYTES (YYMAXDEPTH)
+   evaluated with infinite-precision integer arithmetic.  */
+
+#ifndef YYMAXDEPTH
+# define YYMAXDEPTH 10000
+#endif
+
+
+
+#if YYERROR_VERBOSE
+
+# ifndef yystrlen
+#  if defined __GLIBC__ && defined _STRING_H
+#   define yystrlen strlen
+#  else
+/* Return the length of YYSTR.  */
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static YYSIZE_T
+yystrlen (const char *yystr)
+#else
+static YYSIZE_T
+yystrlen (yystr)
+    const char *yystr;
+#endif
+{
+  YYSIZE_T yylen;
+  for (yylen = 0; yystr[yylen]; yylen++)
+    continue;
+  return yylen;
+}
+#  endif
+# endif
+
+# ifndef yystpcpy
+#  if defined __GLIBC__ && defined _STRING_H && defined _GNU_SOURCE
+#   define yystpcpy stpcpy
+#  else
+/* Copy YYSRC to YYDEST, returning the address of the terminating '\0' in
+   YYDEST.  */
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static char *
+yystpcpy (char *yydest, const char *yysrc)
+#else
+static char *
+yystpcpy (yydest, yysrc)
+    char *yydest;
+    const char *yysrc;
+#endif
+{
+  char *yyd = yydest;
+  const char *yys = yysrc;
+
+  while ((*yyd++ = *yys++) != '\0')
+    continue;
+
+  return yyd - 1;
+}
+#  endif
+# endif
+
+# ifndef yytnamerr
+/* Copy to YYRES the contents of YYSTR after stripping away unnecessary
+   quotes and backslashes, so that it's suitable for yyerror.  The
+   heuristic is that double-quoting is unnecessary unless the string
+   contains an apostrophe, a comma, or backslash (other than
+   backslash-backslash).  YYSTR is taken from yytname.  If YYRES is
+   null, do not copy; instead, return the length of what the result
+   would have been.  */
+static YYSIZE_T
+yytnamerr (char *yyres, const char *yystr)
+{
+  if (*yystr == '"')
+    {
+      YYSIZE_T yyn = 0;
+      char const *yyp = yystr;
+
+      for (;;)
+	switch (*++yyp)
+	  {
+	  case '\'':
+	  case ',':
+	    goto do_not_strip_quotes;
+
+	  case '\\':
+	    if (*++yyp != '\\')
+	      goto do_not_strip_quotes;
+	    /* Fall through.  */
+	  default:
+	    if (yyres)
+	      yyres[yyn] = *yyp;
+	    yyn++;
+	    break;
+
+	  case '"':
+	    if (yyres)
+	      yyres[yyn] = '\0';
+	    return yyn;
+	  }
+    do_not_strip_quotes: ;
+    }
+
+  if (! yyres)
+    return yystrlen (yystr);
+
+  return yystpcpy (yyres, yystr) - yyres;
+}
+# endif
+
+/* Copy into YYRESULT an error message about the unexpected token
+   YYCHAR while in state YYSTATE.  Return the number of bytes copied,
+   including the terminating null byte.  If YYRESULT is null, do not
+   copy anything; just return the number of bytes that would be
+   copied.  As a special case, return 0 if an ordinary "syntax error"
+   message will do.  Return YYSIZE_MAXIMUM if overflow occurs during
+   size calculation.  */
+static YYSIZE_T
+yysyntax_error (char *yyresult, int yystate, int yychar)
+{
+  int yyn = yypact[yystate];
+
+  if (! (YYPACT_NINF < yyn && yyn <= YYLAST))
+    return 0;
+  else
+    {
+      int yytype = YYTRANSLATE (yychar);
+      YYSIZE_T yysize0 = yytnamerr (0, yytname[yytype]);
+      YYSIZE_T yysize = yysize0;
+      YYSIZE_T yysize1;
+      int yysize_overflow = 0;
+      enum { YYERROR_VERBOSE_ARGS_MAXIMUM = 5 };
+      char const *yyarg[YYERROR_VERBOSE_ARGS_MAXIMUM];
+      int yyx;
+
+# if 0
+      /* This is so xgettext sees the translatable formats that are
+	 constructed on the fly.  */
+      YY_("syntax error, unexpected %s");
+      YY_("syntax error, unexpected %s, expecting %s");
+      YY_("syntax error, unexpected %s, expecting %s or %s");
+      YY_("syntax error, unexpected %s, expecting %s or %s or %s");
+      YY_("syntax error, unexpected %s, expecting %s or %s or %s or %s");
+# endif
+      char *yyfmt;
+      char const *yyf;
+      static char const yyunexpected[] = "syntax error, unexpected %s";
+      static char const yyexpecting[] = ", expecting %s";
+      static char const yyor[] = " or %s";
+      char yyformat[sizeof yyunexpected
+		    + sizeof yyexpecting - 1
+		    + ((YYERROR_VERBOSE_ARGS_MAXIMUM - 2)
+		       * (sizeof yyor - 1))];
+      char const *yyprefix = yyexpecting;
+
+      /* Start YYX at -YYN if negative to avoid negative indexes in
+	 YYCHECK.  */
+      int yyxbegin = yyn < 0 ? -yyn : 0;
+
+      /* Stay within bounds of both yycheck and yytname.  */
+      int yychecklim = YYLAST - yyn + 1;
+      int yyxend = yychecklim < YYNTOKENS ? yychecklim : YYNTOKENS;
+      int yycount = 1;
+
+      yyarg[0] = yytname[yytype];
+      yyfmt = yystpcpy (yyformat, yyunexpected);
+
+      for (yyx = yyxbegin; yyx < yyxend; ++yyx)
+	if (yycheck[yyx + yyn] == yyx && yyx != YYTERROR)
+	  {
+	    if (yycount == YYERROR_VERBOSE_ARGS_MAXIMUM)
+	      {
+		yycount = 1;
+		yysize = yysize0;
+		yyformat[sizeof yyunexpected - 1] = '\0';
+		break;
+	      }
+	    yyarg[yycount++] = yytname[yyx];
+	    yysize1 = yysize + yytnamerr (0, yytname[yyx]);
+	    yysize_overflow |= (yysize1 < yysize);
+	    yysize = yysize1;
+	    yyfmt = yystpcpy (yyfmt, yyprefix);
+	    yyprefix = yyor;
+	  }
+
+      yyf = YY_(yyformat);
+      yysize1 = yysize + yystrlen (yyf);
+      yysize_overflow |= (yysize1 < yysize);
+      yysize = yysize1;
+
+      if (yysize_overflow)
+	return YYSIZE_MAXIMUM;
+
+      if (yyresult)
+	{
+	  /* Avoid sprintf, as that infringes on the user's name space.
+	     Don't have undefined behavior even if the translation
+	     produced a string with the wrong number of "%s"s.  */
+	  char *yyp = yyresult;
+	  int yyi = 0;
+	  while ((*yyp = *yyf) != '\0')
+	    {
+	      if (*yyp == '%' && yyf[1] == 's' && yyi < yycount)
+		{
+		  yyp += yytnamerr (yyp, yyarg[yyi++]);
+		  yyf += 2;
+		}
+	      else
+		{
+		  yyp++;
+		  yyf++;
+		}
+	    }
+	}
+      return yysize;
+    }
+}
+#endif /* YYERROR_VERBOSE */
+
+
+/*-----------------------------------------------.
+| Release the memory associated to this symbol.  |
+`-----------------------------------------------*/
+
+/*ARGSUSED*/
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+static void
+yydestruct (const char *yymsg, int yytype, YYSTYPE *yyvaluep, YYLTYPE *yylocationp, struct asm_parser_state *state)
+#else
+static void
+yydestruct (yymsg, yytype, yyvaluep, yylocationp, state)
+    const char *yymsg;
+    int yytype;
+    YYSTYPE *yyvaluep;
+    YYLTYPE *yylocationp;
+    struct asm_parser_state *state;
+#endif
+{
+  YYUSE (yyvaluep);
+  YYUSE (yylocationp);
+  YYUSE (state);
+
+  if (!yymsg)
+    yymsg = "Deleting";
+  YY_SYMBOL_PRINT (yymsg, yytype, yyvaluep, yylocationp);
+
+  switch (yytype)
+    {
+
+      default:
+	break;
+    }
+}
+
+/* Prevent warnings from -Wmissing-prototypes.  */
+#ifdef YYPARSE_PARAM
+#if defined __STDC__ || defined __cplusplus
+int yyparse (void *YYPARSE_PARAM);
+#else
+int yyparse ();
+#endif
+#else /* ! YYPARSE_PARAM */
+#if defined __STDC__ || defined __cplusplus
+int yyparse (struct asm_parser_state *state);
+#else
+int yyparse ();
+#endif
+#endif /* ! YYPARSE_PARAM */
+
+
+
+
+
+/*-------------------------.
+| yyparse or yypush_parse.  |
+`-------------------------*/
+
+#ifdef YYPARSE_PARAM
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+int
+yyparse (void *YYPARSE_PARAM)
+#else
+int
+yyparse (YYPARSE_PARAM)
+    void *YYPARSE_PARAM;
+#endif
+#else /* ! YYPARSE_PARAM */
+#if (defined __STDC__ || defined __C99__FUNC__ \
+     || defined __cplusplus || defined _MSC_VER)
+int
+yyparse (struct asm_parser_state *state)
+#else
+int
+yyparse (state)
+    struct asm_parser_state *state;
+#endif
+#endif
+{
+/* The lookahead symbol.  */
+int yychar;
+
+/* The semantic value of the lookahead symbol.  */
+YYSTYPE yylval;
+
+/* Location data for the lookahead symbol.  */
+YYLTYPE yylloc;
+
+    /* Number of syntax errors so far.  */
+    int yynerrs;
+
+    int yystate;
+    /* Number of tokens to shift before error messages enabled.  */
+    int yyerrstatus;
+
+    /* The stacks and their tools:
+       `yyss': related to states.
+       `yyvs': related to semantic values.
+       `yyls': related to locations.
+
+       Refer to the stacks thru separate pointers, to allow yyoverflow
+       to reallocate them elsewhere.  */
+
+    /* The state stack.  */
+    yytype_int16 yyssa[YYINITDEPTH];
+    yytype_int16 *yyss;
+    yytype_int16 *yyssp;
+
+    /* The semantic value stack.  */
+    YYSTYPE yyvsa[YYINITDEPTH];
+    YYSTYPE *yyvs;
+    YYSTYPE *yyvsp;
+
+    /* The location stack.  */
+    YYLTYPE yylsa[YYINITDEPTH];
+    YYLTYPE *yyls;
+    YYLTYPE *yylsp;
+
+    /* The locations where the error started and ended.  */
+    YYLTYPE yyerror_range[2];
+
+    YYSIZE_T yystacksize;
+
+  int yyn;
+  int yyresult;
+  /* Lookahead token as an internal (translated) token number.  */
+  int yytoken;
+  /* The variables used to return semantic value and location from the
+     action routines.  */
+  YYSTYPE yyval;
+  YYLTYPE yyloc;
+
+#if YYERROR_VERBOSE
+  /* Buffer for error messages, and its allocated size.  */
+  char yymsgbuf[128];
+  char *yymsg = yymsgbuf;
+  YYSIZE_T yymsg_alloc = sizeof yymsgbuf;
+#endif
+
+#define YYPOPSTACK(N)   (yyvsp -= (N), yyssp -= (N), yylsp -= (N))
+
+  /* The number of symbols on the RHS of the reduced rule.
+     Keep to zero when no symbol should be popped.  */
+  int yylen = 0;
+
+  yytoken = 0;
+  yyss = yyssa;
+  yyvs = yyvsa;
+  yyls = yylsa;
+  yystacksize = YYINITDEPTH;
+
+  YYDPRINTF ((stderr, "Starting parse\n"));
+
+  yystate = 0;
+  yyerrstatus = 0;
+  yynerrs = 0;
+  yychar = YYEMPTY; /* Cause a token to be read.  */
+
+  /* Initialize stack pointers.
+     Waste one element of value and location stack
+     so that they stay on the same level as the state stack.
+     The wasted elements are never initialized.  */
+  yyssp = yyss;
+  yyvsp = yyvs;
+  yylsp = yyls;
+
+#if YYLTYPE_IS_TRIVIAL
+  /* Initialize the default location before parsing starts.  */
+  yylloc.first_line   = yylloc.last_line   = 1;
+  yylloc.first_column = yylloc.last_column = 1;
+#endif
+
+  goto yysetstate;
+
+/*------------------------------------------------------------.
+| yynewstate -- Push a new state, which is found in yystate.  |
+`------------------------------------------------------------*/
+ yynewstate:
+  /* In all cases, when you get here, the value and location stacks
+     have just been pushed.  So pushing a state here evens the stacks.  */
+  yyssp++;
+
+ yysetstate:
+  *yyssp = yystate;
+
+  if (yyss + yystacksize - 1 <= yyssp)
+    {
+      /* Get the current used size of the three stacks, in elements.  */
+      YYSIZE_T yysize = yyssp - yyss + 1;
+
+#ifdef yyoverflow
+      {
+	/* Give user a chance to reallocate the stack.  Use copies of
+	   these so that the &'s don't force the real ones into
+	   memory.  */
+	YYSTYPE *yyvs1 = yyvs;
+	yytype_int16 *yyss1 = yyss;
+	YYLTYPE *yyls1 = yyls;
+
+	/* Each stack pointer address is followed by the size of the
+	   data in use in that stack, in bytes.  This used to be a
+	   conditional around just the two extra args, but that might
+	   be undefined if yyoverflow is a macro.  */
+	yyoverflow (YY_("memory exhausted"),
+		    &yyss1, yysize * sizeof (*yyssp),
+		    &yyvs1, yysize * sizeof (*yyvsp),
+		    &yyls1, yysize * sizeof (*yylsp),
+		    &yystacksize);
+
+	yyls = yyls1;
+	yyss = yyss1;
+	yyvs = yyvs1;
+      }
+#else /* no yyoverflow */
+# ifndef YYSTACK_RELOCATE
+      goto yyexhaustedlab;
+# else
+      /* Extend the stack our own way.  */
+      if (YYMAXDEPTH <= yystacksize)
+	goto yyexhaustedlab;
+      yystacksize *= 2;
+      if (YYMAXDEPTH < yystacksize)
+	yystacksize = YYMAXDEPTH;
+
+      {
+	yytype_int16 *yyss1 = yyss;
+	union yyalloc *yyptr =
+	  (union yyalloc *) YYSTACK_ALLOC (YYSTACK_BYTES (yystacksize));
+	if (! yyptr)
+	  goto yyexhaustedlab;
+	YYSTACK_RELOCATE (yyss_alloc, yyss);
+	YYSTACK_RELOCATE (yyvs_alloc, yyvs);
+	YYSTACK_RELOCATE (yyls_alloc, yyls);
+#  undef YYSTACK_RELOCATE
+	if (yyss1 != yyssa)
+	  YYSTACK_FREE (yyss1);
+      }
+# endif
+#endif /* no yyoverflow */
+
+      yyssp = yyss + yysize - 1;
+      yyvsp = yyvs + yysize - 1;
+      yylsp = yyls + yysize - 1;
+
+      YYDPRINTF ((stderr, "Stack size increased to %lu\n",
+		  (unsigned long int) yystacksize));
+
+      if (yyss + yystacksize - 1 <= yyssp)
+	YYABORT;
+    }
+
+  YYDPRINTF ((stderr, "Entering state %d\n", yystate));
+
+  if (yystate == YYFINAL)
+    YYACCEPT;
+
+  goto yybackup;
+
+/*-----------.
+| yybackup.  |
+`-----------*/
+yybackup:
+
+  /* Do appropriate processing given the current state.  Read a
+     lookahead token if we need one and don't already have one.  */
+
+  /* First try to decide what to do without reference to lookahead token.  */
+  yyn = yypact[yystate];
+  if (yyn == YYPACT_NINF)
+    goto yydefault;
+
+  /* Not known => get a lookahead token if don't already have one.  */
+
+  /* YYCHAR is either YYEMPTY or YYEOF or a valid lookahead symbol.  */
+  if (yychar == YYEMPTY)
+    {
+      YYDPRINTF ((stderr, "Reading a token: "));
+      yychar = YYLEX;
+    }
+
+  if (yychar <= YYEOF)
+    {
+      yychar = yytoken = YYEOF;
+      YYDPRINTF ((stderr, "Now at end of input.\n"));
+    }
+  else
+    {
+      yytoken = YYTRANSLATE (yychar);
+      YY_SYMBOL_PRINT ("Next token is", yytoken, &yylval, &yylloc);
+    }
+
+  /* If the proper action on seeing token YYTOKEN is to reduce or to
+     detect an error, take that action.  */
+  yyn += yytoken;
+  if (yyn < 0 || YYLAST < yyn || yycheck[yyn] != yytoken)
+    goto yydefault;
+  yyn = yytable[yyn];
+  if (yyn <= 0)
+    {
+      if (yyn == 0 || yyn == YYTABLE_NINF)
+	goto yyerrlab;
+      yyn = -yyn;
+      goto yyreduce;
+    }
+
+  /* Count tokens shifted since error; after three, turn off error
+     status.  */
+  if (yyerrstatus)
+    yyerrstatus--;
+
+  /* Shift the lookahead token.  */
+  YY_SYMBOL_PRINT ("Shifting", yytoken, &yylval, &yylloc);
+
+  /* Discard the shifted token.  */
+  yychar = YYEMPTY;
+
+  yystate = yyn;
+  *++yyvsp = yylval;
+  *++yylsp = yylloc;
+  goto yynewstate;
+
+
+/*-----------------------------------------------------------.
+| yydefault -- do the default action for the current state.  |
+`-----------------------------------------------------------*/
+yydefault:
+  yyn = yydefact[yystate];
+  if (yyn == 0)
+    goto yyerrlab;
+  goto yyreduce;
+
+
+/*-----------------------------.
+| yyreduce -- Do a reduction.  |
+`-----------------------------*/
+yyreduce:
+  /* yyn is the number of a rule to reduce with.  */
+  yylen = yyr2[yyn];
+
+  /* If YYLEN is nonzero, implement the default value of the action:
+     `$$ = $1'.
+
+     Otherwise, the following line sets YYVAL to garbage.
+     This behavior is undocumented and Bison
+     users should not rely upon it.  Assigning to YYVAL
+     unconditionally makes the parser a bit smaller, and it avoids a
+     GCC warning that YYVAL may be used uninitialized.  */
+  yyval = yyvsp[1-yylen];
+
+  /* Default location.  */
+  YYLLOC_DEFAULT (yyloc, (yylsp - yylen), yylen);
+  YY_REDUCE_PRINT (yyn);
+  switch (yyn)
+    {
+        case 3:
+
+/* Line 1455 of yacc.c  */
+#line 260 "program_parse.y"
+    {
+	   if (state->prog->Target != GL_VERTEX_PROGRAM_ARB) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid fragment program header");
+
+	   }
+	   state->mode = ARB_vertex;
+	;}
+    break;
+
+  case 4:
+
+/* Line 1455 of yacc.c  */
+#line 268 "program_parse.y"
+    {
+	   if (state->prog->Target != GL_FRAGMENT_PROGRAM_ARB) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid vertex program header");
+	   }
+	   state->mode = ARB_fragment;
+
+	   state->option.TexRect =
+	      (state->ctx->Extensions.NV_texture_rectangle != GL_FALSE);
+	;}
+    break;
+
+  case 7:
+
+/* Line 1455 of yacc.c  */
+#line 284 "program_parse.y"
+    {
+	   int valid = 0;
+
+	   if (state->mode == ARB_vertex) {
+	      valid = _mesa_ARBvp_parse_option(state, (yyvsp[(2) - (3)].string));
+	   } else if (state->mode == ARB_fragment) {
+	      valid = _mesa_ARBfp_parse_option(state, (yyvsp[(2) - (3)].string));
+	   }
+
+
+	   if (!valid) {
+	      const char *const err_str = (state->mode == ARB_vertex)
+		 ? "invalid ARB vertex program option"
+		 : "invalid ARB fragment program option";
+
+	      yyerror(& (yylsp[(2) - (3)]), state, err_str);
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 10:
+
+/* Line 1455 of yacc.c  */
+#line 310 "program_parse.y"
+    {
+	   if ((yyvsp[(1) - (2)].inst) != NULL) {
+	      if (state->inst_tail == NULL) {
+		 state->inst_head = (yyvsp[(1) - (2)].inst);
+	      } else {
+		 state->inst_tail->next = (yyvsp[(1) - (2)].inst);
+	      }
+
+	      state->inst_tail = (yyvsp[(1) - (2)].inst);
+	      (yyvsp[(1) - (2)].inst)->next = NULL;
+
+	      state->prog->NumInstructions++;
+	   }
+	;}
+    break;
+
+  case 12:
+
+/* Line 1455 of yacc.c  */
+#line 328 "program_parse.y"
+    {
+	   (yyval.inst) = (yyvsp[(1) - (1)].inst);
+	   state->prog->NumAluInstructions++;
+	;}
+    break;
+
+  case 13:
+
+/* Line 1455 of yacc.c  */
+#line 333 "program_parse.y"
+    {
+	   (yyval.inst) = (yyvsp[(1) - (1)].inst);
+	   state->prog->NumTexInstructions++;
+	;}
+    break;
+
+  case 23:
+
+/* Line 1455 of yacc.c  */
+#line 353 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor(OPCODE_ARL, & (yyvsp[(2) - (4)].dst_reg), & (yyvsp[(4) - (4)].src_reg), NULL, NULL);
+	;}
+    break;
+
+  case 24:
+
+/* Line 1455 of yacc.c  */
+#line 359 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (4)].temp_inst).Opcode, & (yyvsp[(2) - (4)].dst_reg), & (yyvsp[(4) - (4)].src_reg), NULL, NULL);
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (4)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 25:
+
+/* Line 1455 of yacc.c  */
+#line 366 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (4)].temp_inst).Opcode, & (yyvsp[(2) - (4)].dst_reg), & (yyvsp[(4) - (4)].src_reg), NULL, NULL);
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (4)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 26:
+
+/* Line 1455 of yacc.c  */
+#line 373 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (6)].temp_inst).Opcode, & (yyvsp[(2) - (6)].dst_reg), & (yyvsp[(4) - (6)].src_reg), & (yyvsp[(6) - (6)].src_reg), NULL);
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (6)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 27:
+
+/* Line 1455 of yacc.c  */
+#line 381 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (6)].temp_inst).Opcode, & (yyvsp[(2) - (6)].dst_reg), & (yyvsp[(4) - (6)].src_reg), & (yyvsp[(6) - (6)].src_reg), NULL);
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (6)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 28:
+
+/* Line 1455 of yacc.c  */
+#line 389 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (8)].temp_inst).Opcode, & (yyvsp[(2) - (8)].dst_reg), & (yyvsp[(4) - (8)].src_reg), & (yyvsp[(6) - (8)].src_reg), & (yyvsp[(8) - (8)].src_reg));
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (8)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 29:
+
+/* Line 1455 of yacc.c  */
+#line 396 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor((yyvsp[(1) - (8)].temp_inst).Opcode, & (yyvsp[(2) - (8)].dst_reg), & (yyvsp[(4) - (8)].src_reg), NULL, NULL);
+	   if ((yyval.inst) != NULL) {
+	      const GLbitfield tex_mask = (1U << (yyvsp[(6) - (8)].integer));
+	      GLbitfield shadow_tex = 0;
+	      GLbitfield target_mask = 0;
+
+
+	      (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (8)].temp_inst).SaturateMode;
+	      (yyval.inst)->Base.TexSrcUnit = (yyvsp[(6) - (8)].integer);
+
+	      if ((yyvsp[(8) - (8)].integer) < 0) {
+		 shadow_tex = tex_mask;
+
+		 (yyval.inst)->Base.TexSrcTarget = -(yyvsp[(8) - (8)].integer);
+		 (yyval.inst)->Base.TexShadow = 1;
+	      } else {
+		 (yyval.inst)->Base.TexSrcTarget = (yyvsp[(8) - (8)].integer);
+	      }
+
+	      target_mask = (1U << (yyval.inst)->Base.TexSrcTarget);
+
+	      /* If this texture unit was previously accessed and that access
+	       * had a different texture target, generate an error.
+	       *
+	       * If this texture unit was previously accessed and that access
+	       * had a different shadow mode, generate an error.
+	       */
+	      if ((state->prog->TexturesUsed[(yyvsp[(6) - (8)].integer)] != 0)
+		  && ((state->prog->TexturesUsed[(yyvsp[(6) - (8)].integer)] != target_mask)
+		      || ((state->prog->ShadowSamplers & tex_mask)
+			  != shadow_tex))) {
+		 yyerror(& (yylsp[(8) - (8)]), state,
+			 "multiple targets used on one texture image unit");
+		 YYERROR;
+	      }
+
+
+	      state->prog->TexturesUsed[(yyvsp[(6) - (8)].integer)] |= target_mask;
+	      state->prog->ShadowSamplers |= shadow_tex;
+	   }
+	;}
+    break;
+
+  case 30:
+
+/* Line 1455 of yacc.c  */
+#line 441 "program_parse.y"
+    {
+	   (yyval.inst) = asm_instruction_ctor(OPCODE_KIL, NULL, & (yyvsp[(2) - (2)].src_reg), NULL, NULL);
+	   state->fragment.UsesKill = 1;
+	;}
+    break;
+
+  case 31:
+
+/* Line 1455 of yacc.c  */
+#line 448 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 32:
+
+/* Line 1455 of yacc.c  */
+#line 453 "program_parse.y"
+    { (yyval.integer) = TEXTURE_1D_INDEX; ;}
+    break;
+
+  case 33:
+
+/* Line 1455 of yacc.c  */
+#line 454 "program_parse.y"
+    { (yyval.integer) = TEXTURE_2D_INDEX; ;}
+    break;
+
+  case 34:
+
+/* Line 1455 of yacc.c  */
+#line 455 "program_parse.y"
+    { (yyval.integer) = TEXTURE_3D_INDEX; ;}
+    break;
+
+  case 35:
+
+/* Line 1455 of yacc.c  */
+#line 456 "program_parse.y"
+    { (yyval.integer) = TEXTURE_CUBE_INDEX; ;}
+    break;
+
+  case 36:
+
+/* Line 1455 of yacc.c  */
+#line 457 "program_parse.y"
+    { (yyval.integer) = TEXTURE_RECT_INDEX; ;}
+    break;
+
+  case 37:
+
+/* Line 1455 of yacc.c  */
+#line 458 "program_parse.y"
+    { (yyval.integer) = -TEXTURE_1D_INDEX; ;}
+    break;
+
+  case 38:
+
+/* Line 1455 of yacc.c  */
+#line 459 "program_parse.y"
+    { (yyval.integer) = -TEXTURE_2D_INDEX; ;}
+    break;
+
+  case 39:
+
+/* Line 1455 of yacc.c  */
+#line 460 "program_parse.y"
+    { (yyval.integer) = -TEXTURE_RECT_INDEX; ;}
+    break;
+
+  case 40:
+
+/* Line 1455 of yacc.c  */
+#line 461 "program_parse.y"
+    { (yyval.integer) = TEXTURE_1D_ARRAY_INDEX; ;}
+    break;
+
+  case 41:
+
+/* Line 1455 of yacc.c  */
+#line 462 "program_parse.y"
+    { (yyval.integer) = TEXTURE_2D_ARRAY_INDEX; ;}
+    break;
+
+  case 42:
+
+/* Line 1455 of yacc.c  */
+#line 463 "program_parse.y"
+    { (yyval.integer) = -TEXTURE_1D_ARRAY_INDEX; ;}
+    break;
+
+  case 43:
+
+/* Line 1455 of yacc.c  */
+#line 464 "program_parse.y"
+    { (yyval.integer) = -TEXTURE_2D_ARRAY_INDEX; ;}
+    break;
+
+  case 44:
+
+/* Line 1455 of yacc.c  */
+#line 468 "program_parse.y"
+    {
+	   /* FIXME: Is this correct?  Should the extenedSwizzle be applied
+	    * FIXME: to the existing swizzle?
+	    */
+	   (yyvsp[(4) - (6)].src_reg).Base.Swizzle = (yyvsp[(6) - (6)].swiz_mask).swizzle;
+	   (yyvsp[(4) - (6)].src_reg).Base.Negate = (yyvsp[(6) - (6)].swiz_mask).mask;
+
+	   (yyval.inst) = asm_instruction_ctor(OPCODE_SWZ, & (yyvsp[(2) - (6)].dst_reg), & (yyvsp[(4) - (6)].src_reg), NULL, NULL);
+	   (yyval.inst)->Base.SaturateMode = (yyvsp[(1) - (6)].temp_inst).SaturateMode;
+	;}
+    break;
+
+  case 45:
+
+/* Line 1455 of yacc.c  */
+#line 481 "program_parse.y"
+    {
+	   (yyval.src_reg) = (yyvsp[(2) - (3)].src_reg);
+
+	   if ((yyvsp[(1) - (3)].negate)) {
+	      (yyval.src_reg).Base.Negate = ~(yyval.src_reg).Base.Negate;
+	   }
+
+	   (yyval.src_reg).Base.Swizzle = _mesa_combine_swizzles((yyval.src_reg).Base.Swizzle,
+						    (yyvsp[(3) - (3)].swiz_mask).swizzle);
+	;}
+    break;
+
+  case 46:
+
+/* Line 1455 of yacc.c  */
+#line 494 "program_parse.y"
+    {
+	   (yyval.src_reg) = (yyvsp[(2) - (3)].src_reg);
+
+	   if ((yyvsp[(1) - (3)].negate)) {
+	      (yyval.src_reg).Base.Negate = ~(yyval.src_reg).Base.Negate;
+	   }
+
+	   (yyval.src_reg).Base.Swizzle = _mesa_combine_swizzles((yyval.src_reg).Base.Swizzle,
+						    (yyvsp[(3) - (3)].swiz_mask).swizzle);
+	;}
+    break;
+
+  case 47:
+
+/* Line 1455 of yacc.c  */
+#line 507 "program_parse.y"
+    {
+	   (yyval.dst_reg) = (yyvsp[(1) - (2)].dst_reg);
+	   (yyval.dst_reg).WriteMask = (yyvsp[(2) - (2)].swiz_mask).mask;
+
+	   if ((yyval.dst_reg).File == PROGRAM_OUTPUT) {
+	      /* Technically speaking, this should check that it is in
+	       * vertex program mode.  However, PositionInvariant can never be
+	       * set in fragment program mode, so it is somewhat irrelevant.
+	       */
+	      if (state->option.PositionInvariant
+	       && ((yyval.dst_reg).Index == VERT_RESULT_HPOS)) {
+		 yyerror(& (yylsp[(1) - (2)]), state, "position-invariant programs cannot "
+			 "write position");
+		 YYERROR;
+	      }
+
+	      state->prog->OutputsWritten |= (1U << (yyval.dst_reg).Index);
+	   }
+	;}
+    break;
+
+  case 48:
+
+/* Line 1455 of yacc.c  */
+#line 529 "program_parse.y"
+    {
+	   init_dst_reg(& (yyval.dst_reg));
+	   (yyval.dst_reg).File = PROGRAM_ADDRESS;
+	   (yyval.dst_reg).Index = 0;
+	   (yyval.dst_reg).WriteMask = (yyvsp[(2) - (2)].swiz_mask).mask;
+	;}
+    break;
+
+  case 49:
+
+/* Line 1455 of yacc.c  */
+#line 538 "program_parse.y"
+    {
+	   const unsigned xyzw_valid =
+	      ((yyvsp[(1) - (7)].ext_swizzle).xyzw_valid << 0)
+	      | ((yyvsp[(3) - (7)].ext_swizzle).xyzw_valid << 1)
+	      | ((yyvsp[(5) - (7)].ext_swizzle).xyzw_valid << 2)
+	      | ((yyvsp[(7) - (7)].ext_swizzle).xyzw_valid << 3);
+	   const unsigned rgba_valid =
+	      ((yyvsp[(1) - (7)].ext_swizzle).rgba_valid << 0)
+	      | ((yyvsp[(3) - (7)].ext_swizzle).rgba_valid << 1)
+	      | ((yyvsp[(5) - (7)].ext_swizzle).rgba_valid << 2)
+	      | ((yyvsp[(7) - (7)].ext_swizzle).rgba_valid << 3);
+
+	   /* All of the swizzle components have to be valid in either RGBA
+	    * or XYZW.  Note that 0 and 1 are valid in both, so both masks
+	    * can have some bits set.
+	    *
+	    * We somewhat deviate from the spec here.  It would be really hard
+	    * to figure out which component is the error, and there probably
+	    * isn't a lot of benefit.
+	    */
+	   if ((rgba_valid != 0x0f) && (xyzw_valid != 0x0f)) {
+	      yyerror(& (yylsp[(1) - (7)]), state, "cannot combine RGBA and XYZW swizzle "
+		      "components");
+	      YYERROR;
+	   }
+
+	   (yyval.swiz_mask).swizzle = MAKE_SWIZZLE4((yyvsp[(1) - (7)].ext_swizzle).swz, (yyvsp[(3) - (7)].ext_swizzle).swz, (yyvsp[(5) - (7)].ext_swizzle).swz, (yyvsp[(7) - (7)].ext_swizzle).swz);
+	   (yyval.swiz_mask).mask = ((yyvsp[(1) - (7)].ext_swizzle).negate) | ((yyvsp[(3) - (7)].ext_swizzle).negate << 1) | ((yyvsp[(5) - (7)].ext_swizzle).negate << 2)
+	      | ((yyvsp[(7) - (7)].ext_swizzle).negate << 3);
+	;}
+    break;
+
+  case 50:
+
+/* Line 1455 of yacc.c  */
+#line 571 "program_parse.y"
+    {
+	   (yyval.ext_swizzle) = (yyvsp[(2) - (2)].ext_swizzle);
+	   (yyval.ext_swizzle).negate = ((yyvsp[(1) - (2)].negate)) ? 1 : 0;
+	;}
+    break;
+
+  case 51:
+
+/* Line 1455 of yacc.c  */
+#line 578 "program_parse.y"
+    {
+	   if (((yyvsp[(1) - (1)].integer) != 0) && ((yyvsp[(1) - (1)].integer) != 1)) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid extended swizzle selector");
+	      YYERROR;
+	   }
+
+	   (yyval.ext_swizzle).swz = ((yyvsp[(1) - (1)].integer) == 0) ? SWIZZLE_ZERO : SWIZZLE_ONE;
+
+	   /* 0 and 1 are valid for both RGBA swizzle names and XYZW
+	    * swizzle names.
+	    */
+	   (yyval.ext_swizzle).xyzw_valid = 1;
+	   (yyval.ext_swizzle).rgba_valid = 1;
+	;}
+    break;
+
+  case 52:
+
+/* Line 1455 of yacc.c  */
+#line 593 "program_parse.y"
+    {
+	   if (strlen((yyvsp[(1) - (1)].string)) > 1) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid extended swizzle selector");
+	      YYERROR;
+	   }
+
+	   switch ((yyvsp[(1) - (1)].string)[0]) {
+	   case 'x':
+	      (yyval.ext_swizzle).swz = SWIZZLE_X;
+	      (yyval.ext_swizzle).xyzw_valid = 1;
+	      break;
+	   case 'y':
+	      (yyval.ext_swizzle).swz = SWIZZLE_Y;
+	      (yyval.ext_swizzle).xyzw_valid = 1;
+	      break;
+	   case 'z':
+	      (yyval.ext_swizzle).swz = SWIZZLE_Z;
+	      (yyval.ext_swizzle).xyzw_valid = 1;
+	      break;
+	   case 'w':
+	      (yyval.ext_swizzle).swz = SWIZZLE_W;
+	      (yyval.ext_swizzle).xyzw_valid = 1;
+	      break;
+
+	   case 'r':
+	      (yyval.ext_swizzle).swz = SWIZZLE_X;
+	      (yyval.ext_swizzle).rgba_valid = 1;
+	      break;
+	   case 'g':
+	      (yyval.ext_swizzle).swz = SWIZZLE_Y;
+	      (yyval.ext_swizzle).rgba_valid = 1;
+	      break;
+	   case 'b':
+	      (yyval.ext_swizzle).swz = SWIZZLE_Z;
+	      (yyval.ext_swizzle).rgba_valid = 1;
+	      break;
+	   case 'a':
+	      (yyval.ext_swizzle).swz = SWIZZLE_W;
+	      (yyval.ext_swizzle).rgba_valid = 1;
+	      break;
+
+	   default:
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid extended swizzle selector");
+	      YYERROR;
+	      break;
+	   }
+	;}
+    break;
+
+  case 53:
+
+/* Line 1455 of yacc.c  */
+#line 643 "program_parse.y"
+    {
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
+
+	   if (s == NULL) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_param) && (s->type != at_temp)
+		      && (s->type != at_attrib)) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type == at_param) && s->param_is_array) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "non-array access to array PARAM");
+	      YYERROR;
+	   }
+
+	   init_src_reg(& (yyval.src_reg));
+	   switch (s->type) {
+	   case at_temp:
+	      (yyval.src_reg).Base.File = PROGRAM_TEMPORARY;
+	      (yyval.src_reg).Base.Index = s->temp_binding;
+	      break;
+	   case at_param:
+	      (yyval.src_reg).Base.File = s->param_binding_type;
+	      (yyval.src_reg).Base.Index = s->param_binding_begin;
+	      break;
+	   case at_attrib:
+	      (yyval.src_reg).Base.File = PROGRAM_INPUT;
+	      (yyval.src_reg).Base.Index = s->attrib_binding;
+	      state->prog->InputsRead |= (1U << (yyval.src_reg).Base.Index);
+
+	      if (!validate_inputs(& (yylsp[(1) - (1)]), state)) {
+		 YYERROR;
+	      }
+	      break;
+
+	   default:
+	      YYERROR;
+	      break;
+	   }
+	;}
+    break;
+
+  case 54:
+
+/* Line 1455 of yacc.c  */
+#line 685 "program_parse.y"
+    {
+	   init_src_reg(& (yyval.src_reg));
+	   (yyval.src_reg).Base.File = PROGRAM_INPUT;
+	   (yyval.src_reg).Base.Index = (yyvsp[(1) - (1)].attrib);
+	   state->prog->InputsRead |= (1U << (yyval.src_reg).Base.Index);
+
+	   if (!validate_inputs(& (yylsp[(1) - (1)]), state)) {
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 55:
+
+/* Line 1455 of yacc.c  */
+#line 696 "program_parse.y"
+    {
+	   if (! (yyvsp[(3) - (4)].src_reg).Base.RelAddr
+	       && ((unsigned) (yyvsp[(3) - (4)].src_reg).Base.Index >= (yyvsp[(1) - (4)].sym)->param_binding_length)) {
+	      yyerror(& (yylsp[(3) - (4)]), state, "out of bounds array access");
+	      YYERROR;
+	   }
+
+	   init_src_reg(& (yyval.src_reg));
+	   (yyval.src_reg).Base.File = (yyvsp[(1) - (4)].sym)->param_binding_type;
+
+	   if ((yyvsp[(3) - (4)].src_reg).Base.RelAddr) {
+	      (yyvsp[(1) - (4)].sym)->param_accessed_indirectly = 1;
+
+	      (yyval.src_reg).Base.RelAddr = 1;
+	      (yyval.src_reg).Base.Index = (yyvsp[(3) - (4)].src_reg).Base.Index;
+	      (yyval.src_reg).Symbol = (yyvsp[(1) - (4)].sym);
+	   } else {
+	      (yyval.src_reg).Base.Index = (yyvsp[(1) - (4)].sym)->param_binding_begin + (yyvsp[(3) - (4)].src_reg).Base.Index;
+	   }
+	;}
+    break;
+
+  case 56:
+
+/* Line 1455 of yacc.c  */
+#line 717 "program_parse.y"
+    {
+	   init_src_reg(& (yyval.src_reg));
+	   (yyval.src_reg).Base.File = ((yyvsp[(1) - (1)].temp_sym).name != NULL) 
+	      ? (yyvsp[(1) - (1)].temp_sym).param_binding_type
+	      : PROGRAM_CONSTANT;
+	   (yyval.src_reg).Base.Index = (yyvsp[(1) - (1)].temp_sym).param_binding_begin;
+	;}
+    break;
+
+  case 57:
+
+/* Line 1455 of yacc.c  */
+#line 727 "program_parse.y"
+    {
+	   init_dst_reg(& (yyval.dst_reg));
+	   (yyval.dst_reg).File = PROGRAM_OUTPUT;
+	   (yyval.dst_reg).Index = (yyvsp[(1) - (1)].result);
+	;}
+    break;
+
+  case 58:
+
+/* Line 1455 of yacc.c  */
+#line 733 "program_parse.y"
+    {
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
+
+	   if (s == NULL) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_output) && (s->type != at_temp)) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid operand variable");
+	      YYERROR;
+	   }
+
+	   init_dst_reg(& (yyval.dst_reg));
+	   switch (s->type) {
+	   case at_temp:
+	      (yyval.dst_reg).File = PROGRAM_TEMPORARY;
+	      (yyval.dst_reg).Index = s->temp_binding;
+	      break;
+	   case at_output:
+	      (yyval.dst_reg).File = PROGRAM_OUTPUT;
+	      (yyval.dst_reg).Index = s->output_binding;
+	      break;
+	   default:
+	      (yyval.dst_reg).File = s->param_binding_type;
+	      (yyval.dst_reg).Index = s->param_binding_begin;
+	      break;
+	   }
+	;}
+    break;
+
+  case 59:
+
+/* Line 1455 of yacc.c  */
+#line 764 "program_parse.y"
+    {
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
+
+	   if (s == NULL) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_param) || !s->param_is_array) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "array access to non-PARAM variable");
+	      YYERROR;
+	   } else {
+	      (yyval.sym) = s;
+	   }
+	;}
+    break;
+
+  case 62:
+
+/* Line 1455 of yacc.c  */
+#line 783 "program_parse.y"
+    {
+	   init_src_reg(& (yyval.src_reg));
+	   (yyval.src_reg).Base.Index = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 63:
+
+/* Line 1455 of yacc.c  */
+#line 790 "program_parse.y"
+    {
+	   /* FINISHME: Add support for multiple address registers.
+	    */
+	   /* FINISHME: Add support for 4-component address registers.
+	    */
+	   init_src_reg(& (yyval.src_reg));
+	   (yyval.src_reg).Base.RelAddr = 1;
+	   (yyval.src_reg).Base.Index = (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 64:
+
+/* Line 1455 of yacc.c  */
+#line 801 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 65:
+
+/* Line 1455 of yacc.c  */
+#line 802 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(2) - (2)].integer); ;}
+    break;
+
+  case 66:
+
+/* Line 1455 of yacc.c  */
+#line 803 "program_parse.y"
+    { (yyval.integer) = -(yyvsp[(2) - (2)].integer); ;}
+    break;
+
+  case 67:
+
+/* Line 1455 of yacc.c  */
+#line 807 "program_parse.y"
+    {
+	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 63)) {
+	      yyerror(& (yylsp[(1) - (1)]), state,
+		      "relative address offset too large (positive)");
+	      YYERROR;
+	   } else {
+	      (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	   }
+	;}
+    break;
+
+  case 68:
+
+/* Line 1455 of yacc.c  */
+#line 819 "program_parse.y"
+    {
+	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 64)) {
+	      yyerror(& (yylsp[(1) - (1)]), state,
+		      "relative address offset too large (negative)");
+	      YYERROR;
+	   } else {
+	      (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	   }
+	;}
+    break;
+
+  case 69:
+
+/* Line 1455 of yacc.c  */
+#line 831 "program_parse.y"
+    {
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
+
+	   if (s == NULL) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid array member");
+	      YYERROR;
+	   } else if (s->type != at_address) {
+	      yyerror(& (yylsp[(1) - (1)]), state,
+		      "invalid variable for indexed array access");
+	      YYERROR;
+	   } else {
+	      (yyval.sym) = s;
+	   }
+	;}
+    break;
+
+  case 70:
+
+/* Line 1455 of yacc.c  */
+#line 849 "program_parse.y"
+    {
+	   if ((yyvsp[(1) - (1)].swiz_mask).mask != WRITEMASK_X) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid address component selector");
+	      YYERROR;
+	   } else {
+	      (yyval.swiz_mask) = (yyvsp[(1) - (1)].swiz_mask);
+	   }
+	;}
+    break;
+
+  case 71:
+
+/* Line 1455 of yacc.c  */
+#line 860 "program_parse.y"
+    {
+	   if ((yyvsp[(1) - (1)].swiz_mask).mask != WRITEMASK_X) {
+	      yyerror(& (yylsp[(1) - (1)]), state,
+		      "address register write mask must be \".x\"");
+	      YYERROR;
+	   } else {
+	      (yyval.swiz_mask) = (yyvsp[(1) - (1)].swiz_mask);
+	   }
+	;}
+    break;
+
+  case 76:
+
+/* Line 1455 of yacc.c  */
+#line 876 "program_parse.y"
+    { (yyval.swiz_mask).swizzle = SWIZZLE_NOOP; (yyval.swiz_mask).mask = WRITEMASK_XYZW; ;}
+    break;
+
+  case 81:
+
+/* Line 1455 of yacc.c  */
+#line 880 "program_parse.y"
+    { (yyval.swiz_mask).swizzle = SWIZZLE_NOOP; (yyval.swiz_mask).mask = WRITEMASK_XYZW; ;}
+    break;
+
+  case 88:
+
+/* Line 1455 of yacc.c  */
+#line 892 "program_parse.y"
+    {
+	   struct asm_symbol *const s =
+	      declare_variable(state, (yyvsp[(2) - (4)].string), at_attrib, & (yylsp[(2) - (4)]));
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->attrib_binding = (yyvsp[(4) - (4)].attrib);
+	      state->InputsBound |= (1U << s->attrib_binding);
+
+	      if (!validate_inputs(& (yylsp[(4) - (4)]), state)) {
+		 YYERROR;
+	      }
+	   }
+	;}
+    break;
+
+  case 89:
+
+/* Line 1455 of yacc.c  */
+#line 910 "program_parse.y"
+    {
+	   (yyval.attrib) = (yyvsp[(2) - (2)].attrib);
+	;}
+    break;
+
+  case 90:
+
+/* Line 1455 of yacc.c  */
+#line 914 "program_parse.y"
+    {
+	   (yyval.attrib) = (yyvsp[(2) - (2)].attrib);
+	;}
+    break;
+
+  case 91:
+
+/* Line 1455 of yacc.c  */
+#line 920 "program_parse.y"
+    {
+	   (yyval.attrib) = VERT_ATTRIB_POS;
+	;}
+    break;
+
+  case 92:
+
+/* Line 1455 of yacc.c  */
+#line 924 "program_parse.y"
+    {
+	   (yyval.attrib) = VERT_ATTRIB_WEIGHT;
+	;}
+    break;
+
+  case 93:
+
+/* Line 1455 of yacc.c  */
+#line 928 "program_parse.y"
+    {
+	   (yyval.attrib) = VERT_ATTRIB_NORMAL;
+	;}
+    break;
+
+  case 94:
+
+/* Line 1455 of yacc.c  */
+#line 932 "program_parse.y"
+    {
+	   if (!state->ctx->Extensions.EXT_secondary_color) {
+	      yyerror(& (yylsp[(2) - (2)]), state, "GL_EXT_secondary_color not supported");
+	      YYERROR;
+	   }
+
+	   (yyval.attrib) = VERT_ATTRIB_COLOR0 + (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 95:
+
+/* Line 1455 of yacc.c  */
+#line 941 "program_parse.y"
+    {
+	   if (!state->ctx->Extensions.EXT_fog_coord) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "GL_EXT_fog_coord not supported");
+	      YYERROR;
+	   }
+
+	   (yyval.attrib) = VERT_ATTRIB_FOG;
+	;}
+    break;
+
+  case 96:
+
+/* Line 1455 of yacc.c  */
+#line 950 "program_parse.y"
+    {
+	   (yyval.attrib) = VERT_ATTRIB_TEX0 + (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 97:
+
+/* Line 1455 of yacc.c  */
+#line 954 "program_parse.y"
+    {
+	   yyerror(& (yylsp[(1) - (4)]), state, "GL_ARB_matrix_palette not supported");
+	   YYERROR;
+	;}
+    break;
+
+  case 98:
+
+/* Line 1455 of yacc.c  */
+#line 959 "program_parse.y"
+    {
+	   (yyval.attrib) = VERT_ATTRIB_GENERIC0 + (yyvsp[(3) - (4)].integer);
+	;}
+    break;
+
+  case 99:
+
+/* Line 1455 of yacc.c  */
+#line 965 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxAttribs) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid vertex attribute reference");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 103:
+
+/* Line 1455 of yacc.c  */
+#line 979 "program_parse.y"
+    {
+	   (yyval.attrib) = FRAG_ATTRIB_WPOS;
+	;}
+    break;
+
+  case 104:
+
+/* Line 1455 of yacc.c  */
+#line 983 "program_parse.y"
+    {
+	   (yyval.attrib) = FRAG_ATTRIB_COL0 + (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 105:
+
+/* Line 1455 of yacc.c  */
+#line 987 "program_parse.y"
+    {
+	   (yyval.attrib) = FRAG_ATTRIB_FOGC;
+	;}
+    break;
+
+  case 106:
+
+/* Line 1455 of yacc.c  */
+#line 991 "program_parse.y"
+    {
+	   (yyval.attrib) = FRAG_ATTRIB_TEX0 + (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 109:
+
+/* Line 1455 of yacc.c  */
+#line 999 "program_parse.y"
+    {
+	   struct asm_symbol *const s =
+	      declare_variable(state, (yyvsp[(2) - (3)].string), at_param, & (yylsp[(2) - (3)]));
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->param_binding_type = (yyvsp[(3) - (3)].temp_sym).param_binding_type;
+	      s->param_binding_begin = (yyvsp[(3) - (3)].temp_sym).param_binding_begin;
+	      s->param_binding_length = (yyvsp[(3) - (3)].temp_sym).param_binding_length;
+	      s->param_is_array = 0;
+	   }
+	;}
+    break;
+
+  case 110:
+
+/* Line 1455 of yacc.c  */
+#line 1015 "program_parse.y"
+    {
+	   if (((yyvsp[(4) - (6)].integer) != 0) && ((unsigned) (yyvsp[(4) - (6)].integer) != (yyvsp[(6) - (6)].temp_sym).param_binding_length)) {
+	      yyerror(& (yylsp[(4) - (6)]), state, 
+		      "parameter array size and number of bindings must match");
+	      YYERROR;
+	   } else {
+	      struct asm_symbol *const s =
+		 declare_variable(state, (yyvsp[(2) - (6)].string), (yyvsp[(6) - (6)].temp_sym).type, & (yylsp[(2) - (6)]));
+
+	      if (s == NULL) {
+		 YYERROR;
+	      } else {
+		 s->param_binding_type = (yyvsp[(6) - (6)].temp_sym).param_binding_type;
+		 s->param_binding_begin = (yyvsp[(6) - (6)].temp_sym).param_binding_begin;
+		 s->param_binding_length = (yyvsp[(6) - (6)].temp_sym).param_binding_length;
+		 s->param_is_array = 1;
+	      }
+	   }
+	;}
+    break;
+
+  case 111:
+
+/* Line 1455 of yacc.c  */
+#line 1037 "program_parse.y"
+    {
+	   (yyval.integer) = 0;
+	;}
+    break;
+
+  case 112:
+
+/* Line 1455 of yacc.c  */
+#line 1041 "program_parse.y"
+    {
+	   if (((yyvsp[(1) - (1)].integer) < 1) || ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxParameters)) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid parameter array size");
+	      YYERROR;
+	   } else {
+	      (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	   }
+	;}
+    break;
+
+  case 113:
+
+/* Line 1455 of yacc.c  */
+#line 1052 "program_parse.y"
+    {
+	   (yyval.temp_sym) = (yyvsp[(2) - (2)].temp_sym);
+	;}
+    break;
+
+  case 114:
+
+/* Line 1455 of yacc.c  */
+#line 1058 "program_parse.y"
+    {
+	   (yyval.temp_sym) = (yyvsp[(3) - (4)].temp_sym);
+	;}
+    break;
+
+  case 116:
+
+/* Line 1455 of yacc.c  */
+#line 1065 "program_parse.y"
+    {
+	   (yyvsp[(1) - (3)].temp_sym).param_binding_length += (yyvsp[(3) - (3)].temp_sym).param_binding_length;
+	   (yyval.temp_sym) = (yyvsp[(1) - (3)].temp_sym);
+	;}
+    break;
+
+  case 117:
+
+/* Line 1455 of yacc.c  */
+#line 1072 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 118:
+
+/* Line 1455 of yacc.c  */
+#line 1078 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 119:
+
+/* Line 1455 of yacc.c  */
+#line 1084 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & (yyval.temp_sym), & (yyvsp[(1) - (1)].vector));
+	;}
+    break;
+
+  case 120:
+
+/* Line 1455 of yacc.c  */
+#line 1092 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 121:
+
+/* Line 1455 of yacc.c  */
+#line 1098 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 122:
+
+/* Line 1455 of yacc.c  */
+#line 1104 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & (yyval.temp_sym), & (yyvsp[(1) - (1)].vector));
+	;}
+    break;
+
+  case 123:
+
+/* Line 1455 of yacc.c  */
+#line 1112 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 124:
+
+/* Line 1455 of yacc.c  */
+#line 1118 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & (yyval.temp_sym), (yyvsp[(1) - (1)].state));
+	;}
+    break;
+
+  case 125:
+
+/* Line 1455 of yacc.c  */
+#line 1124 "program_parse.y"
+    {
+	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
+	   (yyval.temp_sym).param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & (yyval.temp_sym), & (yyvsp[(1) - (1)].vector));
+	;}
+    break;
+
+  case 126:
+
+/* Line 1455 of yacc.c  */
+#line 1131 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(1) - (1)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 127:
+
+/* Line 1455 of yacc.c  */
+#line 1132 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 128:
+
+/* Line 1455 of yacc.c  */
+#line 1135 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 129:
+
+/* Line 1455 of yacc.c  */
+#line 1136 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 130:
+
+/* Line 1455 of yacc.c  */
+#line 1137 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 131:
+
+/* Line 1455 of yacc.c  */
+#line 1138 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 132:
+
+/* Line 1455 of yacc.c  */
+#line 1139 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 133:
+
+/* Line 1455 of yacc.c  */
+#line 1140 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 134:
+
+/* Line 1455 of yacc.c  */
+#line 1141 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 135:
+
+/* Line 1455 of yacc.c  */
+#line 1142 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 136:
+
+/* Line 1455 of yacc.c  */
+#line 1143 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 137:
+
+/* Line 1455 of yacc.c  */
+#line 1144 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 138:
+
+/* Line 1455 of yacc.c  */
+#line 1145 "program_parse.y"
+    { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
+    break;
+
+  case 139:
+
+/* Line 1455 of yacc.c  */
+#line 1149 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_MATERIAL;
+	   (yyval.state)[1] = (yyvsp[(2) - (3)].integer);
+	   (yyval.state)[2] = (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 140:
+
+/* Line 1455 of yacc.c  */
+#line 1158 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 141:
+
+/* Line 1455 of yacc.c  */
+#line 1162 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_EMISSION;
+	;}
+    break;
+
+  case 142:
+
+/* Line 1455 of yacc.c  */
+#line 1166 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_SHININESS;
+	;}
+    break;
+
+  case 143:
+
+/* Line 1455 of yacc.c  */
+#line 1172 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_LIGHT;
+	   (yyval.state)[1] = (yyvsp[(3) - (5)].integer);
+	   (yyval.state)[2] = (yyvsp[(5) - (5)].integer);
+	;}
+    break;
+
+  case 144:
+
+/* Line 1455 of yacc.c  */
+#line 1181 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 145:
+
+/* Line 1455 of yacc.c  */
+#line 1185 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_POSITION;
+	;}
+    break;
+
+  case 146:
+
+/* Line 1455 of yacc.c  */
+#line 1189 "program_parse.y"
+    {
+	   if (!state->ctx->Extensions.EXT_point_parameters) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "GL_ARB_point_parameters not supported");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = STATE_ATTENUATION;
+	;}
+    break;
+
+  case 147:
+
+/* Line 1455 of yacc.c  */
+#line 1198 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 148:
+
+/* Line 1455 of yacc.c  */
+#line 1202 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_HALF_VECTOR;
+	;}
+    break;
+
+  case 149:
+
+/* Line 1455 of yacc.c  */
+#line 1208 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_SPOT_DIRECTION;
+	;}
+    break;
+
+  case 150:
+
+/* Line 1455 of yacc.c  */
+#line 1214 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(2) - (2)].state)[0];
+	   (yyval.state)[1] = (yyvsp[(2) - (2)].state)[1];
+	;}
+    break;
+
+  case 151:
+
+/* Line 1455 of yacc.c  */
+#line 1221 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_LIGHTMODEL_AMBIENT;
+	;}
+    break;
+
+  case 152:
+
+/* Line 1455 of yacc.c  */
+#line 1226 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_LIGHTMODEL_SCENECOLOR;
+	   (yyval.state)[1] = (yyvsp[(1) - (2)].integer);
+	;}
+    break;
+
+  case 153:
+
+/* Line 1455 of yacc.c  */
+#line 1234 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_LIGHTPROD;
+	   (yyval.state)[1] = (yyvsp[(3) - (6)].integer);
+	   (yyval.state)[2] = (yyvsp[(5) - (6)].integer);
+	   (yyval.state)[3] = (yyvsp[(6) - (6)].integer);
+	;}
+    break;
+
+  case 155:
+
+/* Line 1455 of yacc.c  */
+#line 1246 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = (yyvsp[(3) - (3)].integer);
+	   (yyval.state)[1] = (yyvsp[(2) - (3)].integer);
+	;}
+    break;
+
+  case 156:
+
+/* Line 1455 of yacc.c  */
+#line 1254 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXENV_COLOR;
+	;}
+    break;
+
+  case 157:
+
+/* Line 1455 of yacc.c  */
+#line 1260 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_AMBIENT;
+	;}
+    break;
+
+  case 158:
+
+/* Line 1455 of yacc.c  */
+#line 1264 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_DIFFUSE;
+	;}
+    break;
+
+  case 159:
+
+/* Line 1455 of yacc.c  */
+#line 1268 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_SPECULAR;
+	;}
+    break;
+
+  case 160:
+
+/* Line 1455 of yacc.c  */
+#line 1274 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxLights) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid light selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 161:
+
+/* Line 1455 of yacc.c  */
+#line 1285 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_TEXGEN;
+	   (yyval.state)[1] = (yyvsp[(2) - (4)].integer);
+	   (yyval.state)[2] = (yyvsp[(3) - (4)].integer) + (yyvsp[(4) - (4)].integer);
+	;}
+    break;
+
+  case 162:
+
+/* Line 1455 of yacc.c  */
+#line 1294 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_EYE_S;
+	;}
+    break;
+
+  case 163:
+
+/* Line 1455 of yacc.c  */
+#line 1298 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_OBJECT_S;
+	;}
+    break;
+
+  case 164:
+
+/* Line 1455 of yacc.c  */
+#line 1303 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_EYE_S - STATE_TEXGEN_EYE_S;
+	;}
+    break;
+
+  case 165:
+
+/* Line 1455 of yacc.c  */
+#line 1307 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_EYE_T - STATE_TEXGEN_EYE_S;
+	;}
+    break;
+
+  case 166:
+
+/* Line 1455 of yacc.c  */
+#line 1311 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_EYE_R - STATE_TEXGEN_EYE_S;
+	;}
+    break;
+
+  case 167:
+
+/* Line 1455 of yacc.c  */
+#line 1315 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_TEXGEN_EYE_Q - STATE_TEXGEN_EYE_S;
+	;}
+    break;
+
+  case 168:
+
+/* Line 1455 of yacc.c  */
+#line 1321 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 169:
+
+/* Line 1455 of yacc.c  */
+#line 1328 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_FOG_COLOR;
+	;}
+    break;
+
+  case 170:
+
+/* Line 1455 of yacc.c  */
+#line 1332 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_FOG_PARAMS;
+	;}
+    break;
+
+  case 171:
+
+/* Line 1455 of yacc.c  */
+#line 1338 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_CLIPPLANE;
+	   (yyval.state)[1] = (yyvsp[(3) - (5)].integer);
+	;}
+    break;
+
+  case 172:
+
+/* Line 1455 of yacc.c  */
+#line 1346 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxClipPlanes) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid clip plane selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 173:
+
+/* Line 1455 of yacc.c  */
+#line 1357 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 174:
+
+/* Line 1455 of yacc.c  */
+#line 1364 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_POINT_SIZE;
+	;}
+    break;
+
+  case 175:
+
+/* Line 1455 of yacc.c  */
+#line 1368 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_POINT_ATTENUATION;
+	;}
+    break;
+
+  case 176:
+
+/* Line 1455 of yacc.c  */
+#line 1374 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (5)].state)[0];
+	   (yyval.state)[1] = (yyvsp[(1) - (5)].state)[1];
+	   (yyval.state)[2] = (yyvsp[(4) - (5)].integer);
+	   (yyval.state)[3] = (yyvsp[(4) - (5)].integer);
+	   (yyval.state)[4] = (yyvsp[(1) - (5)].state)[2];
+	;}
+    break;
+
+  case 177:
+
+/* Line 1455 of yacc.c  */
+#line 1384 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (2)].state)[0];
+	   (yyval.state)[1] = (yyvsp[(1) - (2)].state)[1];
+	   (yyval.state)[2] = (yyvsp[(2) - (2)].state)[2];
+	   (yyval.state)[3] = (yyvsp[(2) - (2)].state)[3];
+	   (yyval.state)[4] = (yyvsp[(1) - (2)].state)[2];
+	;}
+    break;
+
+  case 178:
+
+/* Line 1455 of yacc.c  */
+#line 1394 "program_parse.y"
+    {
+	   (yyval.state)[2] = 0;
+	   (yyval.state)[3] = 3;
+	;}
+    break;
+
+  case 179:
+
+/* Line 1455 of yacc.c  */
+#line 1399 "program_parse.y"
+    {
+	   /* It seems logical that the matrix row range specifier would have
+	    * to specify a range or more than one row (i.e., $5 > $3).
+	    * However, the ARB_vertex_program spec says "a program will fail
+	    * to load if <a> is greater than <b>."  This means that $3 == $5
+	    * is valid.
+	    */
+	   if ((yyvsp[(3) - (6)].integer) > (yyvsp[(5) - (6)].integer)) {
+	      yyerror(& (yylsp[(3) - (6)]), state, "invalid matrix row range");
+	      YYERROR;
+	   }
+
+	   (yyval.state)[2] = (yyvsp[(3) - (6)].integer);
+	   (yyval.state)[3] = (yyvsp[(5) - (6)].integer);
+	;}
+    break;
+
+  case 180:
+
+/* Line 1455 of yacc.c  */
+#line 1417 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(2) - (3)].state)[0];
+	   (yyval.state)[1] = (yyvsp[(2) - (3)].state)[1];
+	   (yyval.state)[2] = (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 181:
+
+/* Line 1455 of yacc.c  */
+#line 1425 "program_parse.y"
+    {
+	   (yyval.integer) = 0;
+	;}
+    break;
+
+  case 182:
+
+/* Line 1455 of yacc.c  */
+#line 1429 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 183:
+
+/* Line 1455 of yacc.c  */
+#line 1435 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_MATRIX_INVERSE;
+	;}
+    break;
+
+  case 184:
+
+/* Line 1455 of yacc.c  */
+#line 1439 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_MATRIX_TRANSPOSE;
+	;}
+    break;
+
+  case 185:
+
+/* Line 1455 of yacc.c  */
+#line 1443 "program_parse.y"
+    {
+	   (yyval.integer) = STATE_MATRIX_INVTRANS;
+	;}
+    break;
+
+  case 186:
+
+/* Line 1455 of yacc.c  */
+#line 1449 "program_parse.y"
+    {
+	   if ((yyvsp[(1) - (1)].integer) > 3) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid matrix row reference");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 187:
+
+/* Line 1455 of yacc.c  */
+#line 1460 "program_parse.y"
+    {
+	   (yyval.state)[0] = STATE_MODELVIEW_MATRIX;
+	   (yyval.state)[1] = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 188:
+
+/* Line 1455 of yacc.c  */
+#line 1465 "program_parse.y"
+    {
+	   (yyval.state)[0] = STATE_PROJECTION_MATRIX;
+	   (yyval.state)[1] = 0;
+	;}
+    break;
+
+  case 189:
+
+/* Line 1455 of yacc.c  */
+#line 1470 "program_parse.y"
+    {
+	   (yyval.state)[0] = STATE_MVP_MATRIX;
+	   (yyval.state)[1] = 0;
+	;}
+    break;
+
+  case 190:
+
+/* Line 1455 of yacc.c  */
+#line 1475 "program_parse.y"
+    {
+	   (yyval.state)[0] = STATE_TEXTURE_MATRIX;
+	   (yyval.state)[1] = (yyvsp[(2) - (2)].integer);
+	;}
+    break;
+
+  case 191:
+
+/* Line 1455 of yacc.c  */
+#line 1480 "program_parse.y"
+    {
+	   yyerror(& (yylsp[(1) - (4)]), state, "GL_ARB_matrix_palette not supported");
+	   YYERROR;
+	;}
+    break;
+
+  case 192:
+
+/* Line 1455 of yacc.c  */
+#line 1485 "program_parse.y"
+    {
+	   (yyval.state)[0] = STATE_PROGRAM_MATRIX;
+	   (yyval.state)[1] = (yyvsp[(3) - (4)].integer);
+	;}
+    break;
+
+  case 193:
+
+/* Line 1455 of yacc.c  */
+#line 1492 "program_parse.y"
+    {
+	   (yyval.integer) = 0;
+	;}
+    break;
+
+  case 194:
+
+/* Line 1455 of yacc.c  */
+#line 1496 "program_parse.y"
+    {
+	   (yyval.integer) = (yyvsp[(2) - (3)].integer);
+	;}
+    break;
+
+  case 195:
+
+/* Line 1455 of yacc.c  */
+#line 1501 "program_parse.y"
+    {
+	   /* Since GL_ARB_vertex_blend isn't supported, only modelview matrix
+	    * zero is valid.
+	    */
+	   if ((yyvsp[(1) - (1)].integer) != 0) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid modelview matrix index");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 196:
+
+/* Line 1455 of yacc.c  */
+#line 1514 "program_parse.y"
+    {
+	   /* Since GL_ARB_matrix_palette isn't supported, just let any value
+	    * through here.  The error will be generated later.
+	    */
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 197:
+
+/* Line 1455 of yacc.c  */
+#line 1522 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxProgramMatrices) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program matrix selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 198:
+
+/* Line 1455 of yacc.c  */
+#line 1533 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = STATE_DEPTH_RANGE;
+	;}
+    break;
+
+  case 203:
+
+/* Line 1455 of yacc.c  */
+#line 1545 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = state->state_param_enum;
+	   (yyval.state)[1] = STATE_ENV;
+	   (yyval.state)[2] = (yyvsp[(4) - (5)].state)[0];
+	   (yyval.state)[3] = (yyvsp[(4) - (5)].state)[1];
+	;}
+    break;
+
+  case 204:
+
+/* Line 1455 of yacc.c  */
+#line 1555 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (1)].integer);
+	   (yyval.state)[1] = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 205:
+
+/* Line 1455 of yacc.c  */
+#line 1560 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (3)].integer);
+	   (yyval.state)[1] = (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 206:
+
+/* Line 1455 of yacc.c  */
+#line 1567 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = state->state_param_enum;
+	   (yyval.state)[1] = STATE_ENV;
+	   (yyval.state)[2] = (yyvsp[(4) - (5)].integer);
+	   (yyval.state)[3] = (yyvsp[(4) - (5)].integer);
+	;}
+    break;
+
+  case 207:
+
+/* Line 1455 of yacc.c  */
+#line 1577 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = state->state_param_enum;
+	   (yyval.state)[1] = STATE_LOCAL;
+	   (yyval.state)[2] = (yyvsp[(4) - (5)].state)[0];
+	   (yyval.state)[3] = (yyvsp[(4) - (5)].state)[1];
+	;}
+    break;
+
+  case 208:
+
+/* Line 1455 of yacc.c  */
+#line 1586 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (1)].integer);
+	   (yyval.state)[1] = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 209:
+
+/* Line 1455 of yacc.c  */
+#line 1591 "program_parse.y"
+    {
+	   (yyval.state)[0] = (yyvsp[(1) - (3)].integer);
+	   (yyval.state)[1] = (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 210:
+
+/* Line 1455 of yacc.c  */
+#line 1598 "program_parse.y"
+    {
+	   memset((yyval.state), 0, sizeof((yyval.state)));
+	   (yyval.state)[0] = state->state_param_enum;
+	   (yyval.state)[1] = STATE_LOCAL;
+	   (yyval.state)[2] = (yyvsp[(4) - (5)].integer);
+	   (yyval.state)[3] = (yyvsp[(4) - (5)].integer);
+	;}
+    break;
+
+  case 211:
+
+/* Line 1455 of yacc.c  */
+#line 1608 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxEnvParams) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid environment parameter reference");
+	      YYERROR;
+	   }
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 212:
+
+/* Line 1455 of yacc.c  */
+#line 1618 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxLocalParams) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid local parameter reference");
+	      YYERROR;
+	   }
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 217:
+
+/* Line 1455 of yacc.c  */
+#line 1633 "program_parse.y"
+    {
+	   (yyval.vector).count = 4;
+	   (yyval.vector).data[0] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[1] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[2] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[3] = (yyvsp[(1) - (1)].real);
+	;}
+    break;
+
+  case 218:
+
+/* Line 1455 of yacc.c  */
+#line 1643 "program_parse.y"
+    {
+	   (yyval.vector).count = 1;
+	   (yyval.vector).data[0] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[1] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[2] = (yyvsp[(1) - (1)].real);
+	   (yyval.vector).data[3] = (yyvsp[(1) - (1)].real);
+	;}
+    break;
+
+  case 219:
+
+/* Line 1455 of yacc.c  */
+#line 1651 "program_parse.y"
+    {
+	   (yyval.vector).count = 1;
+	   (yyval.vector).data[0] = (float) (yyvsp[(1) - (1)].integer);
+	   (yyval.vector).data[1] = (float) (yyvsp[(1) - (1)].integer);
+	   (yyval.vector).data[2] = (float) (yyvsp[(1) - (1)].integer);
+	   (yyval.vector).data[3] = (float) (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 220:
+
+/* Line 1455 of yacc.c  */
+#line 1661 "program_parse.y"
+    {
+	   (yyval.vector).count = 4;
+	   (yyval.vector).data[0] = (yyvsp[(2) - (3)].real);
+	   (yyval.vector).data[1] = 0.0f;
+	   (yyval.vector).data[2] = 0.0f;
+	   (yyval.vector).data[3] = 1.0f;
+	;}
+    break;
+
+  case 221:
+
+/* Line 1455 of yacc.c  */
+#line 1669 "program_parse.y"
+    {
+	   (yyval.vector).count = 4;
+	   (yyval.vector).data[0] = (yyvsp[(2) - (5)].real);
+	   (yyval.vector).data[1] = (yyvsp[(4) - (5)].real);
+	   (yyval.vector).data[2] = 0.0f;
+	   (yyval.vector).data[3] = 1.0f;
+	;}
+    break;
+
+  case 222:
+
+/* Line 1455 of yacc.c  */
+#line 1678 "program_parse.y"
+    {
+	   (yyval.vector).count = 4;
+	   (yyval.vector).data[0] = (yyvsp[(2) - (7)].real);
+	   (yyval.vector).data[1] = (yyvsp[(4) - (7)].real);
+	   (yyval.vector).data[2] = (yyvsp[(6) - (7)].real);
+	   (yyval.vector).data[3] = 1.0f;
+	;}
+    break;
+
+  case 223:
+
+/* Line 1455 of yacc.c  */
+#line 1687 "program_parse.y"
+    {
+	   (yyval.vector).count = 4;
+	   (yyval.vector).data[0] = (yyvsp[(2) - (9)].real);
+	   (yyval.vector).data[1] = (yyvsp[(4) - (9)].real);
+	   (yyval.vector).data[2] = (yyvsp[(6) - (9)].real);
+	   (yyval.vector).data[3] = (yyvsp[(8) - (9)].real);
+	;}
+    break;
+
+  case 224:
+
+/* Line 1455 of yacc.c  */
+#line 1697 "program_parse.y"
+    {
+	   (yyval.real) = ((yyvsp[(1) - (2)].negate)) ? -(yyvsp[(2) - (2)].real) : (yyvsp[(2) - (2)].real);
+	;}
+    break;
+
+  case 225:
+
+/* Line 1455 of yacc.c  */
+#line 1701 "program_parse.y"
+    {
+	   (yyval.real) = (float)(((yyvsp[(1) - (2)].negate)) ? -(yyvsp[(2) - (2)].integer) : (yyvsp[(2) - (2)].integer));
+	;}
+    break;
+
+  case 226:
+
+/* Line 1455 of yacc.c  */
+#line 1706 "program_parse.y"
+    { (yyval.negate) = FALSE; ;}
+    break;
+
+  case 227:
+
+/* Line 1455 of yacc.c  */
+#line 1707 "program_parse.y"
+    { (yyval.negate) = TRUE;  ;}
+    break;
+
+  case 228:
+
+/* Line 1455 of yacc.c  */
+#line 1708 "program_parse.y"
+    { (yyval.negate) = FALSE; ;}
+    break;
+
+  case 229:
+
+/* Line 1455 of yacc.c  */
+#line 1711 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(1) - (1)].integer); ;}
+    break;
+
+  case 231:
+
+/* Line 1455 of yacc.c  */
+#line 1714 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(1) - (1)].integer); ;}
+    break;
+
+  case 233:
+
+/* Line 1455 of yacc.c  */
+#line 1718 "program_parse.y"
+    {
+	   if (!declare_variable(state, (yyvsp[(3) - (3)].string), (yyvsp[(0) - (3)].integer), & (yylsp[(3) - (3)]))) {
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 234:
+
+/* Line 1455 of yacc.c  */
+#line 1724 "program_parse.y"
+    {
+	   if (!declare_variable(state, (yyvsp[(1) - (1)].string), (yyvsp[(0) - (1)].integer), & (yylsp[(1) - (1)]))) {
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 235:
+
+/* Line 1455 of yacc.c  */
+#line 1732 "program_parse.y"
+    {
+	   struct asm_symbol *const s =
+	      declare_variable(state, (yyvsp[(2) - (4)].string), at_output, & (yylsp[(2) - (4)]));
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->output_binding = (yyvsp[(4) - (4)].result);
+	   }
+	;}
+    break;
+
+  case 236:
+
+/* Line 1455 of yacc.c  */
+#line 1745 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.result) = VERT_RESULT_HPOS;
+	   } else {
+	      yyerror(& (yylsp[(2) - (2)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 237:
+
+/* Line 1455 of yacc.c  */
+#line 1754 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.result) = VERT_RESULT_FOGC;
+	   } else {
+	      yyerror(& (yylsp[(2) - (2)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 238:
+
+/* Line 1455 of yacc.c  */
+#line 1763 "program_parse.y"
+    {
+	   (yyval.result) = (yyvsp[(2) - (2)].result);
+	;}
+    break;
+
+  case 239:
+
+/* Line 1455 of yacc.c  */
+#line 1767 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.result) = VERT_RESULT_PSIZ;
+	   } else {
+	      yyerror(& (yylsp[(2) - (2)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 240:
+
+/* Line 1455 of yacc.c  */
+#line 1776 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.result) = VERT_RESULT_TEX0 + (yyvsp[(3) - (3)].integer);
+	   } else {
+	      yyerror(& (yylsp[(2) - (3)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 241:
+
+/* Line 1455 of yacc.c  */
+#line 1785 "program_parse.y"
+    {
+	   if (state->mode == ARB_fragment) {
+	      (yyval.result) = FRAG_RESULT_DEPTH;
+	   } else {
+	      yyerror(& (yylsp[(2) - (2)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 242:
+
+/* Line 1455 of yacc.c  */
+#line 1796 "program_parse.y"
+    {
+	   (yyval.result) = (yyvsp[(2) - (3)].integer) + (yyvsp[(3) - (3)].integer);
+	;}
+    break;
+
+  case 243:
+
+/* Line 1455 of yacc.c  */
+#line 1802 "program_parse.y"
+    {
+	   (yyval.integer) = (state->mode == ARB_vertex)
+	      ? VERT_RESULT_COL0
+	      : FRAG_RESULT_COLOR;
+	;}
+    break;
+
+  case 244:
+
+/* Line 1455 of yacc.c  */
+#line 1808 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.integer) = VERT_RESULT_COL0;
+	   } else {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 245:
+
+/* Line 1455 of yacc.c  */
+#line 1817 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.integer) = VERT_RESULT_BFC0;
+	   } else {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 246:
+
+/* Line 1455 of yacc.c  */
+#line 1828 "program_parse.y"
+    {
+	   (yyval.integer) = 0; 
+	;}
+    break;
+
+  case 247:
+
+/* Line 1455 of yacc.c  */
+#line 1832 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.integer) = 0;
+	   } else {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 248:
+
+/* Line 1455 of yacc.c  */
+#line 1841 "program_parse.y"
+    {
+	   if (state->mode == ARB_vertex) {
+	      (yyval.integer) = 1;
+	   } else {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program result name");
+	      YYERROR;
+	   }
+	;}
+    break;
+
+  case 249:
+
+/* Line 1455 of yacc.c  */
+#line 1851 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 250:
+
+/* Line 1455 of yacc.c  */
+#line 1852 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 251:
+
+/* Line 1455 of yacc.c  */
+#line 1853 "program_parse.y"
+    { (yyval.integer) = 1; ;}
+    break;
+
+  case 252:
+
+/* Line 1455 of yacc.c  */
+#line 1856 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 253:
+
+/* Line 1455 of yacc.c  */
+#line 1857 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 254:
+
+/* Line 1455 of yacc.c  */
+#line 1858 "program_parse.y"
+    { (yyval.integer) = 1; ;}
+    break;
+
+  case 255:
+
+/* Line 1455 of yacc.c  */
+#line 1861 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 256:
+
+/* Line 1455 of yacc.c  */
+#line 1862 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
+    break;
+
+  case 257:
+
+/* Line 1455 of yacc.c  */
+#line 1865 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 258:
+
+/* Line 1455 of yacc.c  */
+#line 1866 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
+    break;
+
+  case 259:
+
+/* Line 1455 of yacc.c  */
+#line 1869 "program_parse.y"
+    { (yyval.integer) = 0; ;}
+    break;
+
+  case 260:
+
+/* Line 1455 of yacc.c  */
+#line 1870 "program_parse.y"
+    { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
+    break;
+
+  case 261:
+
+/* Line 1455 of yacc.c  */
+#line 1874 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureCoordUnits) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture coordinate unit selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 262:
+
+/* Line 1455 of yacc.c  */
+#line 1885 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureImageUnits) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture image unit selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 263:
+
+/* Line 1455 of yacc.c  */
+#line 1896 "program_parse.y"
+    {
+	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureUnits) {
+	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture unit selector");
+	      YYERROR;
+	   }
+
+	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
+	;}
+    break;
+
+  case 264:
+
+/* Line 1455 of yacc.c  */
+#line 1907 "program_parse.y"
+    {
+	   struct asm_symbol *exist = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(2) - (4)].string));
+	   struct asm_symbol *target = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(4) - (4)].string));
+
+
+	   if (exist != NULL) {
+	      yyerror(& (yylsp[(2) - (4)]), state, "redeclared identifier");
+	      YYERROR;
+	   } else if (target == NULL) {
+	      yyerror(& (yylsp[(4) - (4)]), state,
+		      "undefined variable binding in ALIAS statement");
+	      YYERROR;
+	   } else {
+	      _mesa_symbol_table_add_symbol(state->st, 0, (yyvsp[(2) - (4)].string), target);
+	   }
+	;}
+    break;
+
+
+
+/* Line 1455 of yacc.c  */
+#line 4577 "program_parse.tab.c"
+      default: break;
+    }
+  YY_SYMBOL_PRINT ("-> $$ =", yyr1[yyn], &yyval, &yyloc);
+
+  YYPOPSTACK (yylen);
+  yylen = 0;
+  YY_STACK_PRINT (yyss, yyssp);
+
+  *++yyvsp = yyval;
+  *++yylsp = yyloc;
+
+  /* Now `shift' the result of the reduction.  Determine what state
+     that goes to, based on the state we popped back to and the rule
+     number reduced by.  */
+
+  yyn = yyr1[yyn];
+
+  yystate = yypgoto[yyn - YYNTOKENS] + *yyssp;
+  if (0 <= yystate && yystate <= YYLAST && yycheck[yystate] == *yyssp)
+    yystate = yytable[yystate];
+  else
+    yystate = yydefgoto[yyn - YYNTOKENS];
+
+  goto yynewstate;
+
+
+/*------------------------------------.
+| yyerrlab -- here on detecting error |
+`------------------------------------*/
+yyerrlab:
+  /* If not already recovering from an error, report this error.  */
+  if (!yyerrstatus)
+    {
+      ++yynerrs;
+#if ! YYERROR_VERBOSE
+      yyerror (&yylloc, state, YY_("syntax error"));
+#else
+      {
+	YYSIZE_T yysize = yysyntax_error (0, yystate, yychar);
+	if (yymsg_alloc < yysize && yymsg_alloc < YYSTACK_ALLOC_MAXIMUM)
+	  {
+	    YYSIZE_T yyalloc = 2 * yysize;
+	    if (! (yysize <= yyalloc && yyalloc <= YYSTACK_ALLOC_MAXIMUM))
+	      yyalloc = YYSTACK_ALLOC_MAXIMUM;
+	    if (yymsg != yymsgbuf)
+	      YYSTACK_FREE (yymsg);
+	    yymsg = (char *) YYSTACK_ALLOC (yyalloc);
+	    if (yymsg)
+	      yymsg_alloc = yyalloc;
+	    else
+	      {
+		yymsg = yymsgbuf;
+		yymsg_alloc = sizeof yymsgbuf;
+	      }
+	  }
+
+	if (0 < yysize && yysize <= yymsg_alloc)
+	  {
+	    (void) yysyntax_error (yymsg, yystate, yychar);
+	    yyerror (&yylloc, state, yymsg);
+	  }
+	else
+	  {
+	    yyerror (&yylloc, state, YY_("syntax error"));
+	    if (yysize != 0)
+	      goto yyexhaustedlab;
+	  }
+      }
+#endif
+    }
+
+  yyerror_range[0] = yylloc;
+
+  if (yyerrstatus == 3)
+    {
+      /* If just tried and failed to reuse lookahead token after an
+	 error, discard it.  */
+
+      if (yychar <= YYEOF)
+	{
+	  /* Return failure if at end of input.  */
+	  if (yychar == YYEOF)
+	    YYABORT;
+	}
+      else
+	{
+	  yydestruct ("Error: discarding",
+		      yytoken, &yylval, &yylloc, state);
+	  yychar = YYEMPTY;
+	}
+    }
+
+  /* Else will try to reuse lookahead token after shifting the error
+     token.  */
+  goto yyerrlab1;
+
+
+/*---------------------------------------------------.
+| yyerrorlab -- error raised explicitly by YYERROR.  |
+`---------------------------------------------------*/
+yyerrorlab:
+
+  /* Pacify compilers like GCC when the user code never invokes
+     YYERROR and the label yyerrorlab therefore never appears in user
+     code.  */
+  if (/*CONSTCOND*/ 0)
+     goto yyerrorlab;
+
+  yyerror_range[0] = yylsp[1-yylen];
+  /* Do not reclaim the symbols of the rule which action triggered
+     this YYERROR.  */
+  YYPOPSTACK (yylen);
+  yylen = 0;
+  YY_STACK_PRINT (yyss, yyssp);
+  yystate = *yyssp;
+  goto yyerrlab1;
+
+
+/*-------------------------------------------------------------.
+| yyerrlab1 -- common code for both syntax error and YYERROR.  |
+`-------------------------------------------------------------*/
+yyerrlab1:
+  yyerrstatus = 3;	/* Each real token shifted decrements this.  */
+
+  for (;;)
+    {
+      yyn = yypact[yystate];
+      if (yyn != YYPACT_NINF)
+	{
+	  yyn += YYTERROR;
+	  if (0 <= yyn && yyn <= YYLAST && yycheck[yyn] == YYTERROR)
+	    {
+	      yyn = yytable[yyn];
+	      if (0 < yyn)
+		break;
+	    }
+	}
+
+      /* Pop the current state because it cannot handle the error token.  */
+      if (yyssp == yyss)
+	YYABORT;
+
+      yyerror_range[0] = *yylsp;
+      yydestruct ("Error: popping",
+		  yystos[yystate], yyvsp, yylsp, state);
+      YYPOPSTACK (1);
+      yystate = *yyssp;
+      YY_STACK_PRINT (yyss, yyssp);
+    }
+
+  *++yyvsp = yylval;
+
+  yyerror_range[1] = yylloc;
+  /* Using YYLLOC is tempting, but would change the location of
+     the lookahead.  YYLOC is available though.  */
+  YYLLOC_DEFAULT (yyloc, (yyerror_range - 1), 2);
+  *++yylsp = yyloc;
+
+  /* Shift the error token.  */
+  YY_SYMBOL_PRINT ("Shifting", yystos[yyn], yyvsp, yylsp);
+
+  yystate = yyn;
+  goto yynewstate;
+
+
+/*-------------------------------------.
+| yyacceptlab -- YYACCEPT comes here.  |
+`-------------------------------------*/
+yyacceptlab:
+  yyresult = 0;
+  goto yyreturn;
+
+/*-----------------------------------.
+| yyabortlab -- YYABORT comes here.  |
+`-----------------------------------*/
+yyabortlab:
+  yyresult = 1;
+  goto yyreturn;
+
+#if !defined(yyoverflow) || YYERROR_VERBOSE
+/*-------------------------------------------------.
+| yyexhaustedlab -- memory exhaustion comes here.  |
+`-------------------------------------------------*/
+yyexhaustedlab:
+  yyerror (&yylloc, state, YY_("memory exhausted"));
+  yyresult = 2;
+  /* Fall through.  */
+#endif
+
+yyreturn:
+  if (yychar != YYEMPTY)
+     yydestruct ("Cleanup: discarding lookahead",
+		 yytoken, &yylval, &yylloc, state);
+  /* Do not reclaim the symbols of the rule which action triggered
+     this YYABORT or YYACCEPT.  */
+  YYPOPSTACK (yylen);
+  YY_STACK_PRINT (yyss, yyssp);
+  while (yyssp != yyss)
+    {
+      yydestruct ("Cleanup: popping",
+		  yystos[*yyssp], yyvsp, yylsp, state);
+      YYPOPSTACK (1);
+    }
+#ifndef yyoverflow
+  if (yyss != yyssa)
+    YYSTACK_FREE (yyss);
+#endif
+#if YYERROR_VERBOSE
+  if (yymsg != yymsgbuf)
+    YYSTACK_FREE (yymsg);
+#endif
+  /* Make sure YYID is used.  */
+  return YYID (yyresult);
+}
+
+
+
+/* Line 1675 of yacc.c  */
+#line 1927 "program_parse.y"
+
+
+struct asm_instruction *
+asm_instruction_ctor(gl_inst_opcode op,
+		     const struct prog_dst_register *dst,
+		     const struct asm_src_register *src0,
+		     const struct asm_src_register *src1,
+		     const struct asm_src_register *src2)
+{
+   struct asm_instruction *inst = calloc(1, sizeof(struct asm_instruction));
+
+   if (inst) {
+      _mesa_init_instructions(& inst->Base, 1);
+      inst->Base.Opcode = op;
+
+      /* In the core ARB extensions only the KIL instruction doesn't have a
+       * destination register.
+       */
+      if (dst == NULL) {
+	 init_dst_reg(& inst->Base.DstReg);
+      } else {
+	 inst->Base.DstReg = *dst;
+      }
+
+      inst->Base.SrcReg[0] = src0->Base;
+      inst->SrcReg[0] = *src0;
+
+      if (src1 != NULL) {
+	 inst->Base.SrcReg[1] = src1->Base;
+	 inst->SrcReg[1] = *src1;
+      } else {
+	 init_src_reg(& inst->SrcReg[1]);
+      }
+
+      if (src2 != NULL) {
+	 inst->Base.SrcReg[2] = src2->Base;
+	 inst->SrcReg[2] = *src2;
+      } else {
+	 init_src_reg(& inst->SrcReg[2]);
+      }
+   }
+
+   return inst;
+}
+
+
+void
+init_dst_reg(struct prog_dst_register *r)
+{
+   memset(r, 0, sizeof(*r));
+   r->File = PROGRAM_UNDEFINED;
+   r->WriteMask = WRITEMASK_XYZW;
+   r->CondMask = COND_TR;
+   r->CondSwizzle = SWIZZLE_NOOP;
+}
+
+
+void
+init_src_reg(struct asm_src_register *r)
+{
+   memset(r, 0, sizeof(*r));
+   r->Base.File = PROGRAM_UNDEFINED;
+   r->Base.Swizzle = SWIZZLE_NOOP;
+   r->Symbol = NULL;
+}
+
+
+/**
+ * Validate the set of inputs used by a program
+ *
+ * Validates that legal sets of inputs are used by the program.  In this case
+ * "used" included both reading the input or binding the input to a name using
+ * the \c ATTRIB command.
+ *
+ * \return
+ * \c TRUE if the combination of inputs used is valid, \c FALSE otherwise.
+ */
+int
+validate_inputs(struct YYLTYPE *locp, struct asm_parser_state *state)
+{
+   const int inputs = state->prog->InputsRead | state->InputsBound;
+
+   if (((inputs & 0x0ffff) & (inputs >> 16)) != 0) {
+      yyerror(locp, state, "illegal use of generic attribute and name attribute");
+      return 0;
+   }
+
+   return 1;
+}
+
+
+struct asm_symbol *
+declare_variable(struct asm_parser_state *state, char *name, enum asm_type t,
+		 struct YYLTYPE *locp)
+{
+   struct asm_symbol *s = NULL;
+   struct asm_symbol *exist = (struct asm_symbol *)
+      _mesa_symbol_table_find_symbol(state->st, 0, name);
+
+
+   if (exist != NULL) {
+      yyerror(locp, state, "redeclared identifier");
+   } else {
+      s = calloc(1, sizeof(struct asm_symbol));
+      s->name = name;
+      s->type = t;
+
+      switch (t) {
+      case at_temp:
+	 if (state->prog->NumTemporaries >= state->limits->MaxTemps) {
+	    yyerror(locp, state, "too many temporaries declared");
+	    free(s);
+	    return NULL;
+	 }
+
+	 s->temp_binding = state->prog->NumTemporaries;
+	 state->prog->NumTemporaries++;
+	 break;
+
+      case at_address:
+	 if (state->prog->NumAddressRegs >= state->limits->MaxAddressRegs) {
+	    yyerror(locp, state, "too many address registers declared");
+	    free(s);
+	    return NULL;
+	 }
+
+	 /* FINISHME: Add support for multiple address registers.
+	  */
+	 state->prog->NumAddressRegs++;
+	 break;
+
+      default:
+	 break;
+      }
+
+      _mesa_symbol_table_add_symbol(state->st, 0, s->name, s);
+      s->next = state->sym;
+      state->sym = s;
+   }
+
+   return s;
+}
+
+
+int add_state_reference(struct gl_program_parameter_list *param_list,
+			const gl_state_index tokens[STATE_LENGTH])
+{
+   const GLuint size = 4; /* XXX fix */
+   char *name;
+   GLint index;
+
+   name = _mesa_program_state_string(tokens);
+   index = _mesa_add_parameter(param_list, PROGRAM_STATE_VAR, name,
+                               size, GL_NONE, NULL, tokens, 0x0);
+   param_list->StateFlags |= _mesa_program_state_flags(tokens);
+
+   /* free name string here since we duplicated it in add_parameter() */
+   _mesa_free(name);
+
+   return index;
+}
+
+
+int
+initialize_symbol_from_state(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const gl_state_index tokens[STATE_LENGTH])
+{
+   int idx = -1;
+   gl_state_index state_tokens[STATE_LENGTH];
+
+
+   memcpy(state_tokens, tokens, sizeof(state_tokens));
+
+   param_var->type = at_param;
+   param_var->param_binding_type = PROGRAM_STATE_VAR;
+
+   /* If we are adding a STATE_MATRIX that has multiple rows, we need to
+    * unroll it and call add_state_reference() for each row
+    */
+   if ((state_tokens[0] == STATE_MODELVIEW_MATRIX ||
+	state_tokens[0] == STATE_PROJECTION_MATRIX ||
+	state_tokens[0] == STATE_MVP_MATRIX ||
+	state_tokens[0] == STATE_TEXTURE_MATRIX ||
+	state_tokens[0] == STATE_PROGRAM_MATRIX)
+       && (state_tokens[2] != state_tokens[3])) {
+      int row;
+      const int first_row = state_tokens[2];
+      const int last_row = state_tokens[3];
+
+      for (row = first_row; row <= last_row; row++) {
+	 state_tokens[2] = state_tokens[3] = row;
+
+	 idx = add_state_reference(prog->Parameters, state_tokens);
+	 if (param_var->param_binding_begin == ~0U)
+	    param_var->param_binding_begin = idx;
+	 param_var->param_binding_length++;
+      }
+   }
+   else {
+      idx = add_state_reference(prog->Parameters, state_tokens);
+      if (param_var->param_binding_begin == ~0U)
+	 param_var->param_binding_begin = idx;
+      param_var->param_binding_length++;
+   }
+
+   return idx;
+}
+
+
+int
+initialize_symbol_from_param(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const gl_state_index tokens[STATE_LENGTH])
+{
+   int idx = -1;
+   gl_state_index state_tokens[STATE_LENGTH];
+
+
+   memcpy(state_tokens, tokens, sizeof(state_tokens));
+
+   assert((state_tokens[0] == STATE_VERTEX_PROGRAM)
+	  || (state_tokens[0] == STATE_FRAGMENT_PROGRAM));
+   assert((state_tokens[1] == STATE_ENV)
+	  || (state_tokens[1] == STATE_LOCAL));
+
+   param_var->type = at_param;
+   param_var->param_binding_type = (state_tokens[1] == STATE_ENV)
+     ? PROGRAM_ENV_PARAM : PROGRAM_LOCAL_PARAM;
+
+   /* If we are adding a STATE_ENV or STATE_LOCAL that has multiple elements,
+    * we need to unroll it and call add_state_reference() for each row
+    */
+   if (state_tokens[2] != state_tokens[3]) {
+      int row;
+      const int first_row = state_tokens[2];
+      const int last_row = state_tokens[3];
+
+      for (row = first_row; row <= last_row; row++) {
+	 state_tokens[2] = state_tokens[3] = row;
+
+	 idx = add_state_reference(prog->Parameters, state_tokens);
+	 if (param_var->param_binding_begin == ~0U)
+	    param_var->param_binding_begin = idx;
+	 param_var->param_binding_length++;
+      }
+   }
+   else {
+      idx = add_state_reference(prog->Parameters, state_tokens);
+      if (param_var->param_binding_begin == ~0U)
+	 param_var->param_binding_begin = idx;
+      param_var->param_binding_length++;
+   }
+
+   return idx;
+}
+
+
+int
+initialize_symbol_from_const(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const struct asm_vector *vec)
+{
+   const int idx = _mesa_add_parameter(prog->Parameters, PROGRAM_CONSTANT,
+				       NULL, vec->count, GL_NONE, vec->data,
+				       NULL, 0x0);
+
+   param_var->type = at_param;
+   param_var->param_binding_type = PROGRAM_CONSTANT;
+
+   if (param_var->param_binding_begin == ~0U)
+      param_var->param_binding_begin = idx;
+   param_var->param_binding_length++;
+
+   return idx;
+}
+
+
+char *
+make_error_string(const char *fmt, ...)
+{
+   int length;
+   char *str;
+   va_list args;
+
+   va_start(args, fmt);
+
+   /* Call vsnprintf once to determine how large the final string is.  Call it
+    * again to do the actual formatting.  from the vsnprintf manual page:
+    *
+    *    Upon successful return, these functions return the number of
+    *    characters printed  (not including the trailing '\0' used to end
+    *    output to strings).
+    */
+   length = 1 + vsnprintf(NULL, 0, fmt, args);
+
+   str = _mesa_malloc(length);
+   if (str) {
+      vsnprintf(str, length, fmt, args);
+   }
+
+   va_end(args);
+
+   return str;
+}
+
+
+void
+yyerror(YYLTYPE *locp, struct asm_parser_state *state, const char *s)
+{
+   char *err_str;
+
+
+   err_str = make_error_string("glProgramStringARB(%s)\n", s);
+   if (err_str) {
+      _mesa_error(state->ctx, GL_INVALID_OPERATION, err_str);
+      _mesa_free(err_str);
+   }
+
+   err_str = make_error_string("line %u, char %u: error: %s\n",
+			       locp->first_line, locp->first_column, s);
+   _mesa_set_program_error(state->ctx, locp->position, err_str);
+
+   if (err_str) {
+      _mesa_free(err_str);
+   }
+}
+
+
+GLboolean
+_mesa_parse_arb_program(GLcontext *ctx, GLenum target, const GLubyte *str,
+			GLsizei len, struct asm_parser_state *state)
+{
+   struct asm_instruction *inst;
+   unsigned i;
+   GLubyte *strz;
+   GLboolean result = GL_FALSE;
+   void *temp;
+   struct asm_symbol *sym;
+
+   state->ctx = ctx;
+   state->prog->Target = target;
+   state->prog->Parameters = _mesa_new_parameter_list();
+
+   /* Make a copy of the program string and force it to be NUL-terminated.
+    */
+   strz = (GLubyte *) _mesa_malloc(len + 1);
+   if (strz == NULL) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glProgramStringARB");
+      return GL_FALSE;
+   }
+   _mesa_memcpy (strz, str, len);
+   strz[len] = '\0';
+
+   state->prog->String = strz;
+
+   state->st = _mesa_symbol_table_ctor();
+
+   state->limits = (target == GL_VERTEX_PROGRAM_ARB)
+      ? & ctx->Const.VertexProgram
+      : & ctx->Const.FragmentProgram;
+
+   state->MaxTextureImageUnits = ctx->Const.MaxTextureImageUnits;
+   state->MaxTextureCoordUnits = ctx->Const.MaxTextureCoordUnits;
+   state->MaxTextureUnits = ctx->Const.MaxTextureUnits;
+   state->MaxClipPlanes = ctx->Const.MaxClipPlanes;
+   state->MaxLights = ctx->Const.MaxLights;
+   state->MaxProgramMatrices = ctx->Const.MaxProgramMatrices;
+
+   state->state_param_enum = (target == GL_VERTEX_PROGRAM_ARB)
+      ? STATE_VERTEX_PROGRAM : STATE_FRAGMENT_PROGRAM;
+
+   _mesa_set_program_error(ctx, -1, NULL);
+
+   _mesa_program_lexer_ctor(& state->scanner, state, (const char *) str, len);
+   yyparse(state);
+   _mesa_program_lexer_dtor(state->scanner);
+
+
+   if (ctx->Program.ErrorPos != -1) {
+      goto error;
+   }
+
+   if (! _mesa_layout_parameters(state)) {
+      struct YYLTYPE loc;
+
+      loc.first_line = 0;
+      loc.first_column = 0;
+      loc.position = len;
+
+      yyerror(& loc, state, "invalid PARAM usage");
+      goto error;
+   }
+
+
+   
+   /* Add one instruction to store the "END" instruction.
+    */
+   state->prog->Instructions =
+      _mesa_alloc_instructions(state->prog->NumInstructions + 1);
+   inst = state->inst_head;
+   for (i = 0; i < state->prog->NumInstructions; i++) {
+      struct asm_instruction *const temp = inst->next;
+
+      state->prog->Instructions[i] = inst->Base;
+      inst = temp;
+   }
+
+   /* Finally, tag on an OPCODE_END instruction */
+   {
+      const GLuint numInst = state->prog->NumInstructions;
+      _mesa_init_instructions(state->prog->Instructions + numInst, 1);
+      state->prog->Instructions[numInst].Opcode = OPCODE_END;
+   }
+   state->prog->NumInstructions++;
+
+   state->prog->NumParameters = state->prog->Parameters->NumParameters;
+   state->prog->NumAttributes = _mesa_bitcount(state->prog->InputsRead);
+
+   /*
+    * Initialize native counts to logical counts.  The device driver may
+    * change them if program is translated into a hardware program.
+    */
+   state->prog->NumNativeInstructions = state->prog->NumInstructions;
+   state->prog->NumNativeTemporaries = state->prog->NumTemporaries;
+   state->prog->NumNativeParameters = state->prog->NumParameters;
+   state->prog->NumNativeAttributes = state->prog->NumAttributes;
+   state->prog->NumNativeAddressRegs = state->prog->NumAddressRegs;
+
+   result = GL_TRUE;
+
+error:
+   for (inst = state->inst_head; inst != NULL; inst = temp) {
+      temp = inst->next;
+      _mesa_free(inst);
+   }
+
+   state->inst_head = NULL;
+   state->inst_tail = NULL;
+
+   for (sym = state->sym; sym != NULL; sym = temp) {
+      temp = sym->next;
+
+      _mesa_free((void *) sym->name);
+      _mesa_free(sym);
+   }
+   state->sym = NULL;
+
+   _mesa_symbol_table_dtor(state->st);
+   state->st = NULL;
+
+   return result;
+}
+
diff --git a/src/mesa/shader/program_parse.tab.h b/src/mesa/shader/program_parse.tab.h
new file mode 100644
index 0000000000..dabb3bf1a3
--- /dev/null
+++ b/src/mesa/shader/program_parse.tab.h
@@ -0,0 +1,207 @@
+
+/* A Bison parser, made by GNU Bison 2.4.1.  */
+
+/* Skeleton interface for Bison's Yacc-like parsers in C
+   
+      Copyright (C) 1984, 1989, 1990, 2000, 2001, 2002, 2003, 2004, 2005, 2006
+   Free Software Foundation, Inc.
+   
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+   
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+   
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+/* As a special exception, you may create a larger work that contains
+   part or all of the Bison parser skeleton and distribute that work
+   under terms of your choice, so long as that work isn't itself a
+   parser generator using the skeleton or a modified version thereof
+   as a parser skeleton.  Alternatively, if you modify or redistribute
+   the parser skeleton itself, you may (at your option) remove this
+   special exception, which will cause the skeleton and the resulting
+   Bison output files to be licensed under the GNU General Public
+   License without this special exception.
+   
+   This special exception was added by the Free Software Foundation in
+   version 2.2 of Bison.  */
+
+
+/* Tokens.  */
+#ifndef YYTOKENTYPE
+# define YYTOKENTYPE
+   /* Put the tokens into the symbol table, so that GDB and other debuggers
+      know about them.  */
+   enum yytokentype {
+     ARBvp_10 = 258,
+     ARBfp_10 = 259,
+     ADDRESS = 260,
+     ALIAS = 261,
+     ATTRIB = 262,
+     OPTION = 263,
+     OUTPUT = 264,
+     PARAM = 265,
+     TEMP = 266,
+     END = 267,
+     BIN_OP = 268,
+     BINSC_OP = 269,
+     SAMPLE_OP = 270,
+     SCALAR_OP = 271,
+     TRI_OP = 272,
+     VECTOR_OP = 273,
+     ARL = 274,
+     KIL = 275,
+     SWZ = 276,
+     INTEGER = 277,
+     REAL = 278,
+     AMBIENT = 279,
+     ATTENUATION = 280,
+     BACK = 281,
+     CLIP = 282,
+     COLOR = 283,
+     DEPTH = 284,
+     DIFFUSE = 285,
+     DIRECTION = 286,
+     EMISSION = 287,
+     ENV = 288,
+     EYE = 289,
+     FOG = 290,
+     FOGCOORD = 291,
+     FRAGMENT = 292,
+     FRONT = 293,
+     HALF = 294,
+     INVERSE = 295,
+     INVTRANS = 296,
+     LIGHT = 297,
+     LIGHTMODEL = 298,
+     LIGHTPROD = 299,
+     LOCAL = 300,
+     MATERIAL = 301,
+     MAT_PROGRAM = 302,
+     MATRIX = 303,
+     MATRIXINDEX = 304,
+     MODELVIEW = 305,
+     MVP = 306,
+     NORMAL = 307,
+     OBJECT = 308,
+     PALETTE = 309,
+     PARAMS = 310,
+     PLANE = 311,
+     POINT_TOK = 312,
+     POINTSIZE = 313,
+     POSITION = 314,
+     PRIMARY = 315,
+     PROGRAM = 316,
+     PROJECTION = 317,
+     RANGE = 318,
+     RESULT = 319,
+     ROW = 320,
+     SCENECOLOR = 321,
+     SECONDARY = 322,
+     SHININESS = 323,
+     SIZE_TOK = 324,
+     SPECULAR = 325,
+     SPOT = 326,
+     STATE = 327,
+     TEXCOORD = 328,
+     TEXENV = 329,
+     TEXGEN = 330,
+     TEXGEN_Q = 331,
+     TEXGEN_R = 332,
+     TEXGEN_S = 333,
+     TEXGEN_T = 334,
+     TEXTURE = 335,
+     TRANSPOSE = 336,
+     TEXTURE_UNIT = 337,
+     TEX_1D = 338,
+     TEX_2D = 339,
+     TEX_3D = 340,
+     TEX_CUBE = 341,
+     TEX_RECT = 342,
+     TEX_SHADOW1D = 343,
+     TEX_SHADOW2D = 344,
+     TEX_SHADOWRECT = 345,
+     TEX_ARRAY1D = 346,
+     TEX_ARRAY2D = 347,
+     TEX_ARRAYSHADOW1D = 348,
+     TEX_ARRAYSHADOW2D = 349,
+     VERTEX = 350,
+     VTXATTRIB = 351,
+     WEIGHT = 352,
+     IDENTIFIER = 353,
+     MASK4 = 354,
+     MASK3 = 355,
+     MASK2 = 356,
+     MASK1 = 357,
+     SWIZZLE = 358,
+     DOT_DOT = 359,
+     DOT = 360
+   };
+#endif
+
+
+
+#if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED
+typedef union YYSTYPE
+{
+
+/* Line 1676 of yacc.c  */
+#line 107 "program_parse.y"
+
+   struct asm_instruction *inst;
+   struct asm_symbol *sym;
+   struct asm_symbol temp_sym;
+   struct asm_swizzle_mask swiz_mask;
+   struct asm_src_register src_reg;
+   struct prog_dst_register dst_reg;
+   struct prog_instruction temp_inst;
+   char *string;
+   unsigned result;
+   unsigned attrib;
+   int integer;
+   float real;
+   gl_state_index state[STATE_LENGTH];
+   int negate;
+   struct asm_vector vector;
+   gl_inst_opcode opcode;
+
+   struct {
+      unsigned swz;
+      unsigned rgba_valid:1;
+      unsigned xyzw_valid:1;
+      unsigned negate:1;
+   } ext_swizzle;
+
+
+
+/* Line 1676 of yacc.c  */
+#line 185 "program_parse.tab.h"
+} YYSTYPE;
+# define YYSTYPE_IS_TRIVIAL 1
+# define yystype YYSTYPE /* obsolescent; will be withdrawn */
+# define YYSTYPE_IS_DECLARED 1
+#endif
+
+
+
+#if ! defined YYLTYPE && ! defined YYLTYPE_IS_DECLARED
+typedef struct YYLTYPE
+{
+  int first_line;
+  int first_column;
+  int last_line;
+  int last_column;
+} YYLTYPE;
+# define yyltype YYLTYPE /* obsolescent; will be withdrawn */
+# define YYLTYPE_IS_DECLARED 1
+# define YYLTYPE_IS_TRIVIAL 1
+#endif
+
+
+
diff --git a/src/mesa/shader/program_parse.y b/src/mesa/shader/program_parse.y
new file mode 100644
index 0000000000..06c1915fbe
--- /dev/null
+++ b/src/mesa/shader/program_parse.y
@@ -0,0 +1,2379 @@
+%{
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "main/mtypes.h"
+#include "main/imports.h"
+#include "program.h"
+#include "prog_parameter.h"
+#include "prog_parameter_layout.h"
+#include "prog_statevars.h"
+#include "prog_instruction.h"
+
+#include "symbol_table.h"
+#include "program_parser.h"
+
+extern void *yy_scan_string(char *);
+extern void yy_delete_buffer(void *);
+
+static struct asm_symbol *declare_variable(struct asm_parser_state *state,
+    char *name, enum asm_type t, struct YYLTYPE *locp);
+
+static int add_state_reference(struct gl_program_parameter_list *param_list,
+    const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_state(struct gl_program *prog,
+    struct asm_symbol *param_var, const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_param(struct gl_program *prog,
+    struct asm_symbol *param_var, const gl_state_index tokens[STATE_LENGTH]);
+
+static int initialize_symbol_from_const(struct gl_program *prog,
+    struct asm_symbol *param_var, const struct asm_vector *vec);
+
+static int yyparse(struct asm_parser_state *state);
+
+static char *make_error_string(const char *fmt, ...);
+
+static void yyerror(struct YYLTYPE *locp, struct asm_parser_state *state,
+    const char *s);
+
+static int validate_inputs(struct YYLTYPE *locp,
+    struct asm_parser_state *state);
+
+static void init_dst_reg(struct prog_dst_register *r);
+
+static void init_src_reg(struct asm_src_register *r);
+
+static struct asm_instruction *asm_instruction_ctor(gl_inst_opcode op,
+    const struct prog_dst_register *dst, const struct asm_src_register *src0,
+    const struct asm_src_register *src1, const struct asm_src_register *src2);
+
+#ifndef FALSE
+#define FALSE 0
+#define TRUE (!FALSE)
+#endif
+
+#define YYLLOC_DEFAULT(Current, Rhs, N)					\
+   do {									\
+      if (YYID(N)) {							\
+	 (Current).first_line = YYRHSLOC(Rhs, 1).first_line;		\
+	 (Current).first_column = YYRHSLOC(Rhs, 1).first_column;	\
+	 (Current).position = YYRHSLOC(Rhs, 1).position;		\
+	 (Current).last_line = YYRHSLOC(Rhs, N).last_line;		\
+	 (Current).last_column = YYRHSLOC(Rhs, N).last_column;		\
+      } else {								\
+	 (Current).first_line = YYRHSLOC(Rhs, 0).last_line;		\
+	 (Current).last_line = (Current).first_line;			\
+	 (Current).first_column = YYRHSLOC(Rhs, 0).last_column;		\
+	 (Current).last_column = (Current).first_column;		\
+	 (Current).position = YYRHSLOC(Rhs, 0).position			\
+	    + (Current).first_column;					\
+      }									\
+   } while(YYID(0))
+
+#define YYLEX_PARAM state->scanner
+%}
+
+%pure-parser
+%locations
+%parse-param { struct asm_parser_state *state }
+%error-verbose
+%lex-param { void *scanner }
+
+%union {
+   struct asm_instruction *inst;
+   struct asm_symbol *sym;
+   struct asm_symbol temp_sym;
+   struct asm_swizzle_mask swiz_mask;
+   struct asm_src_register src_reg;
+   struct prog_dst_register dst_reg;
+   struct prog_instruction temp_inst;
+   char *string;
+   unsigned result;
+   unsigned attrib;
+   int integer;
+   float real;
+   gl_state_index state[STATE_LENGTH];
+   int negate;
+   struct asm_vector vector;
+   gl_inst_opcode opcode;
+
+   struct {
+      unsigned swz;
+      unsigned rgba_valid:1;
+      unsigned xyzw_valid:1;
+      unsigned negate:1;
+   } ext_swizzle;
+}
+
+%token ARBvp_10 ARBfp_10
+
+/* Tokens for assembler pseudo-ops */
+%token <integer> ADDRESS
+%token ALIAS ATTRIB
+%token OPTION OUTPUT
+%token PARAM
+%token <integer> TEMP
+%token END
+
+ /* Tokens for instructions */
+%token <temp_inst> BIN_OP BINSC_OP SAMPLE_OP SCALAR_OP TRI_OP VECTOR_OP
+%token <temp_inst> ARL KIL SWZ
+
+%token <integer> INTEGER
+%token <real> REAL
+
+%token AMBIENT ATTENUATION
+%token BACK
+%token CLIP COLOR
+%token DEPTH DIFFUSE DIRECTION
+%token EMISSION ENV EYE
+%token FOG FOGCOORD FRAGMENT FRONT
+%token HALF
+%token INVERSE INVTRANS
+%token LIGHT LIGHTMODEL LIGHTPROD LOCAL
+%token MATERIAL MAT_PROGRAM MATRIX MATRIXINDEX MODELVIEW MVP
+%token NORMAL
+%token OBJECT
+%token PALETTE PARAMS PLANE POINT_TOK POINTSIZE POSITION PRIMARY PROGRAM PROJECTION
+%token RANGE RESULT ROW
+%token SCENECOLOR SECONDARY SHININESS SIZE_TOK SPECULAR SPOT STATE
+%token TEXCOORD TEXENV TEXGEN TEXGEN_Q TEXGEN_R TEXGEN_S TEXGEN_T TEXTURE TRANSPOSE
+%token TEXTURE_UNIT TEX_1D TEX_2D TEX_3D TEX_CUBE TEX_RECT
+%token TEX_SHADOW1D TEX_SHADOW2D TEX_SHADOWRECT
+%token TEX_ARRAY1D TEX_ARRAY2D TEX_ARRAYSHADOW1D TEX_ARRAYSHADOW2D 
+%token VERTEX VTXATTRIB
+%token WEIGHT
+
+%token <string> IDENTIFIER
+%token <swiz_mask> MASK4 MASK3 MASK2 MASK1 SWIZZLE
+%token DOT_DOT
+%token DOT
+
+%type <inst> instruction ALU_instruction TexInstruction
+%type <inst> ARL_instruction VECTORop_instruction
+%type <inst> SCALARop_instruction BINSCop_instruction BINop_instruction
+%type <inst> TRIop_instruction SWZ_instruction SAMPLE_instruction
+%type <inst> KIL_instruction
+
+%type <dst_reg> dstReg maskedDstReg maskedAddrReg
+%type <src_reg> srcReg scalarSrcReg swizzleSrcReg
+%type <swiz_mask> scalarSuffix swizzleSuffix extendedSwizzle
+%type <ext_swizzle> extSwizComp extSwizSel
+%type <swiz_mask> optionalMask
+
+%type <sym> progParamArray
+%type <integer> addrRegRelOffset addrRegPosOffset addrRegNegOffset
+%type <src_reg> progParamArrayMem progParamArrayAbs progParamArrayRel
+%type <sym> addrReg
+%type <swiz_mask> addrComponent addrWriteMask
+
+%type <result> resultBinding resultColBinding
+%type <integer> optFaceType optColorType
+%type <integer> optResultFaceType optResultColorType
+
+%type <integer> optTexImageUnitNum texImageUnitNum
+%type <integer> optTexCoordUnitNum texCoordUnitNum
+%type <integer> optLegacyTexUnitNum legacyTexUnitNum
+%type <integer> texImageUnit texTarget
+%type <integer> vtxAttribNum
+
+%type <attrib> attribBinding vtxAttribItem fragAttribItem
+
+%type <temp_sym> paramSingleInit paramSingleItemDecl
+%type <integer> optArraySize
+
+%type <state> stateSingleItem stateMultipleItem
+%type <state> stateMaterialItem
+%type <state> stateLightItem stateLightModelItem stateLightProdItem
+%type <state> stateTexGenItem stateFogItem stateClipPlaneItem statePointItem
+%type <state> stateMatrixItem stateMatrixRow stateMatrixRows
+%type <state> stateTexEnvItem stateDepthItem
+
+%type <state> stateLModProperty
+%type <state> stateMatrixName optMatrixRows
+
+%type <integer> stateMatProperty
+%type <integer> stateLightProperty stateSpotProperty
+%type <integer> stateLightNumber stateLProdProperty
+%type <integer> stateTexGenType stateTexGenCoord
+%type <integer> stateTexEnvProperty
+%type <integer> stateFogProperty
+%type <integer> stateClipPlaneNum
+%type <integer> statePointProperty
+
+%type <integer> stateOptMatModifier stateMatModifier stateMatrixRowNum
+%type <integer> stateOptModMatNum stateModMatNum statePaletteMatNum 
+%type <integer> stateProgramMatNum
+
+%type <integer> ambDiffSpecProperty
+
+%type <state> programSingleItem progEnvParam progLocalParam
+%type <state> programMultipleItem progEnvParams progLocalParams
+
+%type <temp_sym> paramMultipleInit paramMultInitList paramMultipleItem
+%type <temp_sym> paramSingleItemUse
+
+%type <integer> progEnvParamNum progLocalParamNum
+%type <state> progEnvParamNums progLocalParamNums
+
+%type <vector> paramConstDecl paramConstUse
+%type <vector> paramConstScalarDecl paramConstScalarUse paramConstVector
+%type <real> signedFloatConstant
+%type <negate> optionalSign
+
+%{
+extern int yylex(YYSTYPE *yylval_param, YYLTYPE *yylloc_param,
+    void *yyscanner);
+%}
+
+%%
+
+program: language optionSequence statementSequence END
+	;
+
+language: ARBvp_10
+	{
+	   if (state->prog->Target != GL_VERTEX_PROGRAM_ARB) {
+	      yyerror(& @1, state, "invalid fragment program header");
+
+	   }
+	   state->mode = ARB_vertex;
+	}
+	| ARBfp_10
+	{
+	   if (state->prog->Target != GL_FRAGMENT_PROGRAM_ARB) {
+	      yyerror(& @1, state, "invalid vertex program header");
+	   }
+	   state->mode = ARB_fragment;
+
+	   state->option.TexRect =
+	      (state->ctx->Extensions.NV_texture_rectangle != GL_FALSE);
+	}
+	;
+
+optionSequence: optionSequence option
+	|
+	;
+
+option: OPTION IDENTIFIER ';'
+	{
+	   int valid = 0;
+
+	   if (state->mode == ARB_vertex) {
+	      valid = _mesa_ARBvp_parse_option(state, $2);
+	   } else if (state->mode == ARB_fragment) {
+	      valid = _mesa_ARBfp_parse_option(state, $2);
+	   }
+
+
+	   if (!valid) {
+	      const char *const err_str = (state->mode == ARB_vertex)
+		 ? "invalid ARB vertex program option"
+		 : "invalid ARB fragment program option";
+
+	      yyerror(& @2, state, err_str);
+	      YYERROR;
+	   }
+	}
+	;
+
+statementSequence: statementSequence statement
+	|
+	;
+
+statement: instruction ';'
+	{
+	   if ($1 != NULL) {
+	      if (state->inst_tail == NULL) {
+		 state->inst_head = $1;
+	      } else {
+		 state->inst_tail->next = $1;
+	      }
+
+	      state->inst_tail = $1;
+	      $1->next = NULL;
+
+	      state->prog->NumInstructions++;
+	   }
+	}
+	| namingStatement ';'
+	;
+
+instruction: ALU_instruction
+	{
+	   $$ = $1;
+	   state->prog->NumAluInstructions++;
+	}
+	| TexInstruction
+	{
+	   $$ = $1;
+	   state->prog->NumTexInstructions++;
+	}
+	;
+
+ALU_instruction: ARL_instruction
+	| VECTORop_instruction
+	| SCALARop_instruction
+	| BINSCop_instruction
+	| BINop_instruction
+	| TRIop_instruction
+	| SWZ_instruction
+	;
+
+TexInstruction: SAMPLE_instruction
+	| KIL_instruction
+	;
+
+ARL_instruction: ARL maskedAddrReg ',' scalarSrcReg
+	{
+	   $$ = asm_instruction_ctor(OPCODE_ARL, & $2, & $4, NULL, NULL);
+	}
+	;
+
+VECTORop_instruction: VECTOR_OP maskedDstReg ',' swizzleSrcReg
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, NULL, NULL);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+SCALARop_instruction: SCALAR_OP maskedDstReg ',' scalarSrcReg
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, NULL, NULL);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+BINSCop_instruction: BINSC_OP maskedDstReg ',' scalarSrcReg ',' scalarSrcReg
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, & $6, NULL);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+
+BINop_instruction: BIN_OP maskedDstReg ',' swizzleSrcReg ',' swizzleSrcReg
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, & $6, NULL);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+TRIop_instruction: TRI_OP maskedDstReg ','
+                   swizzleSrcReg ',' swizzleSrcReg ',' swizzleSrcReg
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, & $6, & $8);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+SAMPLE_instruction: SAMPLE_OP maskedDstReg ',' swizzleSrcReg ',' texImageUnit ',' texTarget
+	{
+	   $$ = asm_instruction_ctor($1.Opcode, & $2, & $4, NULL, NULL);
+	   if ($$ != NULL) {
+	      const GLbitfield tex_mask = (1U << $6);
+	      GLbitfield shadow_tex = 0;
+	      GLbitfield target_mask = 0;
+
+
+	      $$->Base.SaturateMode = $1.SaturateMode;
+	      $$->Base.TexSrcUnit = $6;
+
+	      if ($8 < 0) {
+		 shadow_tex = tex_mask;
+
+		 $$->Base.TexSrcTarget = -$8;
+		 $$->Base.TexShadow = 1;
+	      } else {
+		 $$->Base.TexSrcTarget = $8;
+	      }
+
+	      target_mask = (1U << $$->Base.TexSrcTarget);
+
+	      /* If this texture unit was previously accessed and that access
+	       * had a different texture target, generate an error.
+	       *
+	       * If this texture unit was previously accessed and that access
+	       * had a different shadow mode, generate an error.
+	       */
+	      if ((state->prog->TexturesUsed[$6] != 0)
+		  && ((state->prog->TexturesUsed[$6] != target_mask)
+		      || ((state->prog->ShadowSamplers & tex_mask)
+			  != shadow_tex))) {
+		 yyerror(& @8, state,
+			 "multiple targets used on one texture image unit");
+		 YYERROR;
+	      }
+
+
+	      state->prog->TexturesUsed[$6] |= target_mask;
+	      state->prog->ShadowSamplers |= shadow_tex;
+	   }
+	}
+	;
+
+KIL_instruction: KIL swizzleSrcReg
+	{
+	   $$ = asm_instruction_ctor(OPCODE_KIL, NULL, & $2, NULL, NULL);
+	   state->fragment.UsesKill = 1;
+	}
+	;
+
+texImageUnit: TEXTURE_UNIT optTexImageUnitNum
+	{
+	   $$ = $2;
+	}
+	;
+
+texTarget: TEX_1D  { $$ = TEXTURE_1D_INDEX; }
+	| TEX_2D   { $$ = TEXTURE_2D_INDEX; }
+	| TEX_3D   { $$ = TEXTURE_3D_INDEX; }
+	| TEX_CUBE { $$ = TEXTURE_CUBE_INDEX; }
+	| TEX_RECT { $$ = TEXTURE_RECT_INDEX; }
+	| TEX_SHADOW1D   { $$ = -TEXTURE_1D_INDEX; }
+	| TEX_SHADOW2D   { $$ = -TEXTURE_2D_INDEX; }
+	| TEX_SHADOWRECT { $$ = -TEXTURE_RECT_INDEX; }
+	| TEX_ARRAY1D         { $$ = TEXTURE_1D_ARRAY_INDEX; }
+	| TEX_ARRAY2D         { $$ = TEXTURE_2D_ARRAY_INDEX; }
+	| TEX_ARRAYSHADOW1D   { $$ = -TEXTURE_1D_ARRAY_INDEX; }
+	| TEX_ARRAYSHADOW2D   { $$ = -TEXTURE_2D_ARRAY_INDEX; }
+	;
+
+SWZ_instruction: SWZ maskedDstReg ',' srcReg ',' extendedSwizzle
+	{
+	   /* FIXME: Is this correct?  Should the extenedSwizzle be applied
+	    * FIXME: to the existing swizzle?
+	    */
+	   $4.Base.Swizzle = $6.swizzle;
+	   $4.Base.Negate = $6.mask;
+
+	   $$ = asm_instruction_ctor(OPCODE_SWZ, & $2, & $4, NULL, NULL);
+	   $$->Base.SaturateMode = $1.SaturateMode;
+	}
+	;
+
+scalarSrcReg: optionalSign srcReg scalarSuffix
+	{
+	   $$ = $2;
+
+	   if ($1) {
+	      $$.Base.Negate = ~$$.Base.Negate;
+	   }
+
+	   $$.Base.Swizzle = _mesa_combine_swizzles($$.Base.Swizzle,
+						    $3.swizzle);
+	}
+	;
+
+swizzleSrcReg: optionalSign srcReg swizzleSuffix
+	{
+	   $$ = $2;
+
+	   if ($1) {
+	      $$.Base.Negate = ~$$.Base.Negate;
+	   }
+
+	   $$.Base.Swizzle = _mesa_combine_swizzles($$.Base.Swizzle,
+						    $3.swizzle);
+	}
+	;
+
+maskedDstReg: dstReg optionalMask
+	{
+	   $$ = $1;
+	   $$.WriteMask = $2.mask;
+
+	   if ($$.File == PROGRAM_OUTPUT) {
+	      /* Technically speaking, this should check that it is in
+	       * vertex program mode.  However, PositionInvariant can never be
+	       * set in fragment program mode, so it is somewhat irrelevant.
+	       */
+	      if (state->option.PositionInvariant
+	       && ($$.Index == VERT_RESULT_HPOS)) {
+		 yyerror(& @1, state, "position-invariant programs cannot "
+			 "write position");
+		 YYERROR;
+	      }
+
+	      state->prog->OutputsWritten |= (1U << $$.Index);
+	   }
+	}
+	;
+
+maskedAddrReg: addrReg addrWriteMask
+	{
+	   init_dst_reg(& $$);
+	   $$.File = PROGRAM_ADDRESS;
+	   $$.Index = 0;
+	   $$.WriteMask = $2.mask;
+	}
+	;
+
+extendedSwizzle: extSwizComp ',' extSwizComp ',' extSwizComp ',' extSwizComp
+	{
+	   const unsigned xyzw_valid =
+	      ($1.xyzw_valid << 0)
+	      | ($3.xyzw_valid << 1)
+	      | ($5.xyzw_valid << 2)
+	      | ($7.xyzw_valid << 3);
+	   const unsigned rgba_valid =
+	      ($1.rgba_valid << 0)
+	      | ($3.rgba_valid << 1)
+	      | ($5.rgba_valid << 2)
+	      | ($7.rgba_valid << 3);
+
+	   /* All of the swizzle components have to be valid in either RGBA
+	    * or XYZW.  Note that 0 and 1 are valid in both, so both masks
+	    * can have some bits set.
+	    *
+	    * We somewhat deviate from the spec here.  It would be really hard
+	    * to figure out which component is the error, and there probably
+	    * isn't a lot of benefit.
+	    */
+	   if ((rgba_valid != 0x0f) && (xyzw_valid != 0x0f)) {
+	      yyerror(& @1, state, "cannot combine RGBA and XYZW swizzle "
+		      "components");
+	      YYERROR;
+	   }
+
+	   $$.swizzle = MAKE_SWIZZLE4($1.swz, $3.swz, $5.swz, $7.swz);
+	   $$.mask = ($1.negate) | ($3.negate << 1) | ($5.negate << 2)
+	      | ($7.negate << 3);
+	}
+	;
+
+extSwizComp: optionalSign extSwizSel
+	{
+	   $$ = $2;
+	   $$.negate = ($1) ? 1 : 0;
+	}
+	;
+
+extSwizSel: INTEGER
+	{
+	   if (($1 != 0) && ($1 != 1)) {
+	      yyerror(& @1, state, "invalid extended swizzle selector");
+	      YYERROR;
+	   }
+
+	   $$.swz = ($1 == 0) ? SWIZZLE_ZERO : SWIZZLE_ONE;
+
+	   /* 0 and 1 are valid for both RGBA swizzle names and XYZW
+	    * swizzle names.
+	    */
+	   $$.xyzw_valid = 1;
+	   $$.rgba_valid = 1;
+	}
+	| IDENTIFIER
+	{
+	   if (strlen($1) > 1) {
+	      yyerror(& @1, state, "invalid extended swizzle selector");
+	      YYERROR;
+	   }
+
+	   switch ($1[0]) {
+	   case 'x':
+	      $$.swz = SWIZZLE_X;
+	      $$.xyzw_valid = 1;
+	      break;
+	   case 'y':
+	      $$.swz = SWIZZLE_Y;
+	      $$.xyzw_valid = 1;
+	      break;
+	   case 'z':
+	      $$.swz = SWIZZLE_Z;
+	      $$.xyzw_valid = 1;
+	      break;
+	   case 'w':
+	      $$.swz = SWIZZLE_W;
+	      $$.xyzw_valid = 1;
+	      break;
+
+	   case 'r':
+	      $$.swz = SWIZZLE_X;
+	      $$.rgba_valid = 1;
+	      break;
+	   case 'g':
+	      $$.swz = SWIZZLE_Y;
+	      $$.rgba_valid = 1;
+	      break;
+	   case 'b':
+	      $$.swz = SWIZZLE_Z;
+	      $$.rgba_valid = 1;
+	      break;
+	   case 'a':
+	      $$.swz = SWIZZLE_W;
+	      $$.rgba_valid = 1;
+	      break;
+
+	   default:
+	      yyerror(& @1, state, "invalid extended swizzle selector");
+	      YYERROR;
+	      break;
+	   }
+	}
+	;
+
+srcReg: IDENTIFIER /* temporaryReg | progParamSingle */
+	{
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $1);
+
+	   if (s == NULL) {
+	      yyerror(& @1, state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_param) && (s->type != at_temp)
+		      && (s->type != at_attrib)) {
+	      yyerror(& @1, state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type == at_param) && s->param_is_array) {
+	      yyerror(& @1, state, "non-array access to array PARAM");
+	      YYERROR;
+	   }
+
+	   init_src_reg(& $$);
+	   switch (s->type) {
+	   case at_temp:
+	      $$.Base.File = PROGRAM_TEMPORARY;
+	      $$.Base.Index = s->temp_binding;
+	      break;
+	   case at_param:
+	      $$.Base.File = s->param_binding_type;
+	      $$.Base.Index = s->param_binding_begin;
+	      break;
+	   case at_attrib:
+	      $$.Base.File = PROGRAM_INPUT;
+	      $$.Base.Index = s->attrib_binding;
+	      state->prog->InputsRead |= (1U << $$.Base.Index);
+
+	      if (!validate_inputs(& @1, state)) {
+		 YYERROR;
+	      }
+	      break;
+
+	   default:
+	      YYERROR;
+	      break;
+	   }
+	}
+	| attribBinding
+	{
+	   init_src_reg(& $$);
+	   $$.Base.File = PROGRAM_INPUT;
+	   $$.Base.Index = $1;
+	   state->prog->InputsRead |= (1U << $$.Base.Index);
+
+	   if (!validate_inputs(& @1, state)) {
+	      YYERROR;
+	   }
+	}
+	| progParamArray '[' progParamArrayMem ']'
+	{
+	   if (! $3.Base.RelAddr
+	       && ((unsigned) $3.Base.Index >= $1->param_binding_length)) {
+	      yyerror(& @3, state, "out of bounds array access");
+	      YYERROR;
+	   }
+
+	   init_src_reg(& $$);
+	   $$.Base.File = $1->param_binding_type;
+
+	   if ($3.Base.RelAddr) {
+	      $1->param_accessed_indirectly = 1;
+
+	      $$.Base.RelAddr = 1;
+	      $$.Base.Index = $3.Base.Index;
+	      $$.Symbol = $1;
+	   } else {
+	      $$.Base.Index = $1->param_binding_begin + $3.Base.Index;
+	   }
+	}
+	| paramSingleItemUse
+	{
+	   init_src_reg(& $$);
+	   $$.Base.File = ($1.name != NULL) 
+	      ? $1.param_binding_type
+	      : PROGRAM_CONSTANT;
+	   $$.Base.Index = $1.param_binding_begin;
+	}
+	;
+
+dstReg: resultBinding
+	{
+	   init_dst_reg(& $$);
+	   $$.File = PROGRAM_OUTPUT;
+	   $$.Index = $1;
+	}
+	| IDENTIFIER /* temporaryReg | vertexResultReg */
+	{
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $1);
+
+	   if (s == NULL) {
+	      yyerror(& @1, state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_output) && (s->type != at_temp)) {
+	      yyerror(& @1, state, "invalid operand variable");
+	      YYERROR;
+	   }
+
+	   init_dst_reg(& $$);
+	   switch (s->type) {
+	   case at_temp:
+	      $$.File = PROGRAM_TEMPORARY;
+	      $$.Index = s->temp_binding;
+	      break;
+	   case at_output:
+	      $$.File = PROGRAM_OUTPUT;
+	      $$.Index = s->output_binding;
+	      break;
+	   default:
+	      $$.File = s->param_binding_type;
+	      $$.Index = s->param_binding_begin;
+	      break;
+	   }
+	}
+	;
+
+progParamArray: IDENTIFIER
+	{
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $1);
+
+	   if (s == NULL) {
+	      yyerror(& @1, state, "invalid operand variable");
+	      YYERROR;
+	   } else if ((s->type != at_param) || !s->param_is_array) {
+	      yyerror(& @1, state, "array access to non-PARAM variable");
+	      YYERROR;
+	   } else {
+	      $$ = s;
+	   }
+	}
+	;
+
+progParamArrayMem: progParamArrayAbs | progParamArrayRel;
+
+progParamArrayAbs: INTEGER
+	{
+	   init_src_reg(& $$);
+	   $$.Base.Index = $1;
+	}
+	;
+
+progParamArrayRel: addrReg addrComponent addrRegRelOffset
+	{
+	   /* FINISHME: Add support for multiple address registers.
+	    */
+	   /* FINISHME: Add support for 4-component address registers.
+	    */
+	   init_src_reg(& $$);
+	   $$.Base.RelAddr = 1;
+	   $$.Base.Index = $3;
+	}
+	;
+
+addrRegRelOffset:              { $$ = 0; }
+	| '+' addrRegPosOffset { $$ = $2; }
+	| '-' addrRegNegOffset { $$ = -$2; }
+	;
+
+addrRegPosOffset: INTEGER
+	{
+	   if (($1 < 0) || ($1 > 63)) {
+	      yyerror(& @1, state,
+		      "relative address offset too large (positive)");
+	      YYERROR;
+	   } else {
+	      $$ = $1;
+	   }
+	}
+	;
+
+addrRegNegOffset: INTEGER
+	{
+	   if (($1 < 0) || ($1 > 64)) {
+	      yyerror(& @1, state,
+		      "relative address offset too large (negative)");
+	      YYERROR;
+	   } else {
+	      $$ = $1;
+	   }
+	}
+	;
+
+addrReg: IDENTIFIER
+	{
+	   struct asm_symbol *const s = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $1);
+
+	   if (s == NULL) {
+	      yyerror(& @1, state, "invalid array member");
+	      YYERROR;
+	   } else if (s->type != at_address) {
+	      yyerror(& @1, state,
+		      "invalid variable for indexed array access");
+	      YYERROR;
+	   } else {
+	      $$ = s;
+	   }
+	}
+	;
+
+addrComponent: MASK1
+	{
+	   if ($1.mask != WRITEMASK_X) {
+	      yyerror(& @1, state, "invalid address component selector");
+	      YYERROR;
+	   } else {
+	      $$ = $1;
+	   }
+	}
+	;
+
+addrWriteMask: MASK1
+	{
+	   if ($1.mask != WRITEMASK_X) {
+	      yyerror(& @1, state,
+		      "address register write mask must be \".x\"");
+	      YYERROR;
+	   } else {
+	      $$ = $1;
+	   }
+	}
+	;
+
+scalarSuffix: MASK1;
+
+swizzleSuffix: MASK1
+	| MASK4
+	| SWIZZLE
+	|              { $$.swizzle = SWIZZLE_NOOP; $$.mask = WRITEMASK_XYZW; }
+	;
+
+optionalMask: MASK4 | MASK3 | MASK2 | MASK1 
+	|              { $$.swizzle = SWIZZLE_NOOP; $$.mask = WRITEMASK_XYZW; }
+	;
+
+namingStatement: ATTRIB_statement
+	| PARAM_statement
+	| TEMP_statement
+	| ADDRESS_statement
+	| OUTPUT_statement
+	| ALIAS_statement
+	;
+
+ATTRIB_statement: ATTRIB IDENTIFIER '=' attribBinding
+	{
+	   struct asm_symbol *const s =
+	      declare_variable(state, $2, at_attrib, & @2);
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->attrib_binding = $4;
+	      state->InputsBound |= (1U << s->attrib_binding);
+
+	      if (!validate_inputs(& @4, state)) {
+		 YYERROR;
+	      }
+	   }
+	}
+	;
+
+attribBinding: VERTEX vtxAttribItem
+	{
+	   $$ = $2;
+	}
+	| FRAGMENT fragAttribItem
+	{
+	   $$ = $2;
+	}
+	;
+
+vtxAttribItem: POSITION
+	{
+	   $$ = VERT_ATTRIB_POS;
+	}
+	| WEIGHT vtxOptWeightNum
+	{
+	   $$ = VERT_ATTRIB_WEIGHT;
+	}
+	| NORMAL
+	{
+	   $$ = VERT_ATTRIB_NORMAL;
+	}
+	| COLOR optColorType
+	{
+	   if (!state->ctx->Extensions.EXT_secondary_color) {
+	      yyerror(& @2, state, "GL_EXT_secondary_color not supported");
+	      YYERROR;
+	   }
+
+	   $$ = VERT_ATTRIB_COLOR0 + $2;
+	}
+	| FOGCOORD
+	{
+	   if (!state->ctx->Extensions.EXT_fog_coord) {
+	      yyerror(& @1, state, "GL_EXT_fog_coord not supported");
+	      YYERROR;
+	   }
+
+	   $$ = VERT_ATTRIB_FOG;
+	}
+	| TEXCOORD optTexCoordUnitNum
+	{
+	   $$ = VERT_ATTRIB_TEX0 + $2;
+	}
+	| MATRIXINDEX '[' vtxWeightNum ']'
+	{
+	   yyerror(& @1, state, "GL_ARB_matrix_palette not supported");
+	   YYERROR;
+	}
+	| VTXATTRIB '[' vtxAttribNum ']'
+	{
+	   $$ = VERT_ATTRIB_GENERIC0 + $3;
+	}
+	;
+
+vtxAttribNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->limits->MaxAttribs) {
+	      yyerror(& @1, state, "invalid vertex attribute reference");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+vtxOptWeightNum:  | '[' vtxWeightNum ']';
+vtxWeightNum: INTEGER;
+
+fragAttribItem: POSITION
+	{
+	   $$ = FRAG_ATTRIB_WPOS;
+	}
+	| COLOR optColorType
+	{
+	   $$ = FRAG_ATTRIB_COL0 + $2;
+	}
+	| FOGCOORD
+	{
+	   $$ = FRAG_ATTRIB_FOGC;
+	}
+	| TEXCOORD optTexCoordUnitNum
+	{
+	   $$ = FRAG_ATTRIB_TEX0 + $2;
+	}
+	;
+
+PARAM_statement: PARAM_singleStmt | PARAM_multipleStmt;
+
+PARAM_singleStmt: PARAM IDENTIFIER paramSingleInit
+	{
+	   struct asm_symbol *const s =
+	      declare_variable(state, $2, at_param, & @2);
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->param_binding_type = $3.param_binding_type;
+	      s->param_binding_begin = $3.param_binding_begin;
+	      s->param_binding_length = $3.param_binding_length;
+	      s->param_is_array = 0;
+	   }
+	}
+	;
+
+PARAM_multipleStmt: PARAM IDENTIFIER '[' optArraySize ']' paramMultipleInit
+	{
+	   if (($4 != 0) && ((unsigned) $4 != $6.param_binding_length)) {
+	      yyerror(& @4, state, 
+		      "parameter array size and number of bindings must match");
+	      YYERROR;
+	   } else {
+	      struct asm_symbol *const s =
+		 declare_variable(state, $2, $6.type, & @2);
+
+	      if (s == NULL) {
+		 YYERROR;
+	      } else {
+		 s->param_binding_type = $6.param_binding_type;
+		 s->param_binding_begin = $6.param_binding_begin;
+		 s->param_binding_length = $6.param_binding_length;
+		 s->param_is_array = 1;
+	      }
+	   }
+	}
+	;
+
+optArraySize:
+	{
+	   $$ = 0;
+	}
+	| INTEGER
+        {
+	   if (($1 < 1) || ((unsigned) $1 >= state->limits->MaxParameters)) {
+	      yyerror(& @1, state, "invalid parameter array size");
+	      YYERROR;
+	   } else {
+	      $$ = $1;
+	   }
+	}
+	;
+
+paramSingleInit: '=' paramSingleItemDecl
+	{
+	   $$ = $2;
+	}
+	;
+
+paramMultipleInit: '=' '{' paramMultInitList '}'
+	{
+	   $$ = $3;
+	}
+	;
+
+paramMultInitList: paramMultipleItem
+	| paramMultInitList ',' paramMultipleItem
+	{
+	   $1.param_binding_length += $3.param_binding_length;
+	   $$ = $1;
+	}
+	;
+
+paramSingleItemDecl: stateSingleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & $$, $1);
+	}
+	| programSingleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & $$, $1);
+	}
+	| paramConstDecl
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & $$, & $1);
+	}
+	;
+
+paramSingleItemUse: stateSingleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & $$, $1);
+	}
+	| programSingleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & $$, $1);
+	}
+	| paramConstUse
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & $$, & $1);
+	}
+	;
+
+paramMultipleItem: stateMultipleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_state(state->prog, & $$, $1);
+	}
+	| programMultipleItem
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_param(state->prog, & $$, $1);
+	}
+	| paramConstDecl
+	{
+	   memset(& $$, 0, sizeof($$));
+	   $$.param_binding_begin = ~0;
+	   initialize_symbol_from_const(state->prog, & $$, & $1);
+	}
+	;
+
+stateMultipleItem: stateSingleItem        { memcpy($$, $1, sizeof($$)); }
+	| STATE stateMatrixRows           { memcpy($$, $2, sizeof($$)); }
+	;
+
+stateSingleItem: STATE stateMaterialItem  { memcpy($$, $2, sizeof($$)); }
+	| STATE stateLightItem            { memcpy($$, $2, sizeof($$)); }
+	| STATE stateLightModelItem       { memcpy($$, $2, sizeof($$)); }
+	| STATE stateLightProdItem        { memcpy($$, $2, sizeof($$)); }
+	| STATE stateTexGenItem           { memcpy($$, $2, sizeof($$)); }
+	| STATE stateTexEnvItem           { memcpy($$, $2, sizeof($$)); }
+	| STATE stateFogItem              { memcpy($$, $2, sizeof($$)); }
+	| STATE stateClipPlaneItem        { memcpy($$, $2, sizeof($$)); }
+	| STATE statePointItem            { memcpy($$, $2, sizeof($$)); }
+	| STATE stateMatrixRow            { memcpy($$, $2, sizeof($$)); }
+	| STATE stateDepthItem            { memcpy($$, $2, sizeof($$)); }
+	;
+
+stateMaterialItem: MATERIAL optFaceType stateMatProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_MATERIAL;
+	   $$[1] = $2;
+	   $$[2] = $3;
+	}
+	;
+
+stateMatProperty: ambDiffSpecProperty
+	{
+	   $$ = $1;
+	}
+	| EMISSION
+	{
+	   $$ = STATE_EMISSION;
+	}
+	| SHININESS
+	{
+	   $$ = STATE_SHININESS;
+	}
+	;
+
+stateLightItem: LIGHT '[' stateLightNumber ']' stateLightProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_LIGHT;
+	   $$[1] = $3;
+	   $$[2] = $5;
+	}
+	;
+
+stateLightProperty: ambDiffSpecProperty
+	{
+	   $$ = $1;
+	}
+	| POSITION
+	{
+	   $$ = STATE_POSITION;
+	}
+	| ATTENUATION
+	{
+	   if (!state->ctx->Extensions.EXT_point_parameters) {
+	      yyerror(& @1, state, "GL_ARB_point_parameters not supported");
+	      YYERROR;
+	   }
+
+	   $$ = STATE_ATTENUATION;
+	}
+	| SPOT stateSpotProperty
+	{
+	   $$ = $2;
+	}
+	| HALF
+	{
+	   $$ = STATE_HALF_VECTOR;
+	}
+	;
+
+stateSpotProperty: DIRECTION
+	{
+	   $$ = STATE_SPOT_DIRECTION;
+	}
+	;
+
+stateLightModelItem: LIGHTMODEL stateLModProperty
+	{
+	   $$[0] = $2[0];
+	   $$[1] = $2[1];
+	}
+	;
+
+stateLModProperty: AMBIENT
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_LIGHTMODEL_AMBIENT;
+	}
+	| optFaceType SCENECOLOR
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_LIGHTMODEL_SCENECOLOR;
+	   $$[1] = $1;
+	}
+	;
+
+stateLightProdItem: LIGHTPROD '[' stateLightNumber ']' optFaceType stateLProdProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_LIGHTPROD;
+	   $$[1] = $3;
+	   $$[2] = $5;
+	   $$[3] = $6;
+	}
+	;
+
+stateLProdProperty: ambDiffSpecProperty;
+
+stateTexEnvItem: TEXENV optLegacyTexUnitNum stateTexEnvProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = $3;
+	   $$[1] = $2;
+	}
+	;
+
+stateTexEnvProperty: COLOR
+	{
+	   $$ = STATE_TEXENV_COLOR;
+	}
+	;
+
+ambDiffSpecProperty: AMBIENT
+	{
+	   $$ = STATE_AMBIENT;
+	}
+	| DIFFUSE
+	{
+	   $$ = STATE_DIFFUSE;
+	}
+	| SPECULAR
+	{
+	   $$ = STATE_SPECULAR;
+	}
+	;
+
+stateLightNumber: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxLights) {
+	      yyerror(& @1, state, "invalid light selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+stateTexGenItem: TEXGEN optTexCoordUnitNum stateTexGenType stateTexGenCoord
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_TEXGEN;
+	   $$[1] = $2;
+	   $$[2] = $3 + $4;
+	}
+	;
+
+stateTexGenType: EYE
+	{
+	   $$ = STATE_TEXGEN_EYE_S;
+	}
+	| OBJECT
+	{
+	   $$ = STATE_TEXGEN_OBJECT_S;
+	}
+	;
+stateTexGenCoord: TEXGEN_S
+	{
+	   $$ = STATE_TEXGEN_EYE_S - STATE_TEXGEN_EYE_S;
+	}
+	| TEXGEN_T
+	{
+	   $$ = STATE_TEXGEN_EYE_T - STATE_TEXGEN_EYE_S;
+	}
+	| TEXGEN_R
+	{
+	   $$ = STATE_TEXGEN_EYE_R - STATE_TEXGEN_EYE_S;
+	}
+	| TEXGEN_Q
+	{
+	   $$ = STATE_TEXGEN_EYE_Q - STATE_TEXGEN_EYE_S;
+	}
+	;
+
+stateFogItem: FOG stateFogProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = $2;
+	}
+	;
+
+stateFogProperty: COLOR
+	{
+	   $$ = STATE_FOG_COLOR;
+	}
+	| PARAMS
+	{
+	   $$ = STATE_FOG_PARAMS;
+	}
+	;
+
+stateClipPlaneItem: CLIP '[' stateClipPlaneNum ']' PLANE
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_CLIPPLANE;
+	   $$[1] = $3;
+	}
+	;
+
+stateClipPlaneNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxClipPlanes) {
+	      yyerror(& @1, state, "invalid clip plane selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+statePointItem: POINT_TOK statePointProperty
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = $2;
+	}
+	;
+
+statePointProperty: SIZE_TOK
+	{
+	   $$ = STATE_POINT_SIZE;
+	}
+	| ATTENUATION
+	{
+	   $$ = STATE_POINT_ATTENUATION;
+	}
+	;
+
+stateMatrixRow: stateMatrixItem ROW '[' stateMatrixRowNum ']'
+	{
+	   $$[0] = $1[0];
+	   $$[1] = $1[1];
+	   $$[2] = $4;
+	   $$[3] = $4;
+	   $$[4] = $1[2];
+	}
+	;
+
+stateMatrixRows: stateMatrixItem optMatrixRows
+	{
+	   $$[0] = $1[0];
+	   $$[1] = $1[1];
+	   $$[2] = $2[2];
+	   $$[3] = $2[3];
+	   $$[4] = $1[2];
+	}
+	;
+
+optMatrixRows:
+	{
+	   $$[2] = 0;
+	   $$[3] = 3;
+	}
+	| ROW '[' stateMatrixRowNum DOT_DOT stateMatrixRowNum ']'
+	{
+	   /* It seems logical that the matrix row range specifier would have
+	    * to specify a range or more than one row (i.e., $5 > $3).
+	    * However, the ARB_vertex_program spec says "a program will fail
+	    * to load if <a> is greater than <b>."  This means that $3 == $5
+	    * is valid.
+	    */
+	   if ($3 > $5) {
+	      yyerror(& @3, state, "invalid matrix row range");
+	      YYERROR;
+	   }
+
+	   $$[2] = $3;
+	   $$[3] = $5;
+	}
+	;
+
+stateMatrixItem: MATRIX stateMatrixName stateOptMatModifier
+	{
+	   $$[0] = $2[0];
+	   $$[1] = $2[1];
+	   $$[2] = $3;
+	}
+	;
+
+stateOptMatModifier: 
+	{
+	   $$ = 0;
+	}
+	| stateMatModifier
+	{
+	   $$ = $1;
+	}
+	;
+
+stateMatModifier: INVERSE 
+	{
+	   $$ = STATE_MATRIX_INVERSE;
+	}
+	| TRANSPOSE 
+	{
+	   $$ = STATE_MATRIX_TRANSPOSE;
+	}
+	| INVTRANS
+	{
+	   $$ = STATE_MATRIX_INVTRANS;
+	}
+	;
+
+stateMatrixRowNum: INTEGER
+	{
+	   if ($1 > 3) {
+	      yyerror(& @1, state, "invalid matrix row reference");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+stateMatrixName: MODELVIEW stateOptModMatNum
+	{
+	   $$[0] = STATE_MODELVIEW_MATRIX;
+	   $$[1] = $2;
+	}
+	| PROJECTION
+	{
+	   $$[0] = STATE_PROJECTION_MATRIX;
+	   $$[1] = 0;
+	}
+	| MVP
+	{
+	   $$[0] = STATE_MVP_MATRIX;
+	   $$[1] = 0;
+	}
+	| TEXTURE optTexCoordUnitNum
+	{
+	   $$[0] = STATE_TEXTURE_MATRIX;
+	   $$[1] = $2;
+	}
+	| PALETTE '[' statePaletteMatNum ']'
+	{
+	   yyerror(& @1, state, "GL_ARB_matrix_palette not supported");
+	   YYERROR;
+	}
+	| MAT_PROGRAM '[' stateProgramMatNum ']'
+	{
+	   $$[0] = STATE_PROGRAM_MATRIX;
+	   $$[1] = $3;
+	}
+	;
+
+stateOptModMatNum:
+	{
+	   $$ = 0;
+	}
+	| '[' stateModMatNum ']'
+	{
+	   $$ = $2;
+	}
+	;
+stateModMatNum: INTEGER
+	{
+	   /* Since GL_ARB_vertex_blend isn't supported, only modelview matrix
+	    * zero is valid.
+	    */
+	   if ($1 != 0) {
+	      yyerror(& @1, state, "invalid modelview matrix index");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+statePaletteMatNum: INTEGER
+	{
+	   /* Since GL_ARB_matrix_palette isn't supported, just let any value
+	    * through here.  The error will be generated later.
+	    */
+	   $$ = $1;
+	}
+	;
+stateProgramMatNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxProgramMatrices) {
+	      yyerror(& @1, state, "invalid program matrix selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+stateDepthItem: DEPTH RANGE
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = STATE_DEPTH_RANGE;
+	}
+	;
+
+
+programSingleItem: progEnvParam | progLocalParam;
+
+programMultipleItem: progEnvParams | progLocalParams;
+
+progEnvParams: PROGRAM ENV '[' progEnvParamNums ']'
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = state->state_param_enum;
+	   $$[1] = STATE_ENV;
+	   $$[2] = $4[0];
+	   $$[3] = $4[1];
+	}
+	;
+
+progEnvParamNums: progEnvParamNum
+	{
+	   $$[0] = $1;
+	   $$[1] = $1;
+	}
+	| progEnvParamNum DOT_DOT progEnvParamNum
+	{
+	   $$[0] = $1;
+	   $$[1] = $3;
+	}
+	;
+
+progEnvParam: PROGRAM ENV '[' progEnvParamNum ']'
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = state->state_param_enum;
+	   $$[1] = STATE_ENV;
+	   $$[2] = $4;
+	   $$[3] = $4;
+	}
+	;
+
+progLocalParams: PROGRAM LOCAL '[' progLocalParamNums ']'
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = state->state_param_enum;
+	   $$[1] = STATE_LOCAL;
+	   $$[2] = $4[0];
+	   $$[3] = $4[1];
+	}
+
+progLocalParamNums: progLocalParamNum
+	{
+	   $$[0] = $1;
+	   $$[1] = $1;
+	}
+	| progLocalParamNum DOT_DOT progLocalParamNum
+	{
+	   $$[0] = $1;
+	   $$[1] = $3;
+	}
+	;
+
+progLocalParam: PROGRAM LOCAL '[' progLocalParamNum ']'
+	{
+	   memset($$, 0, sizeof($$));
+	   $$[0] = state->state_param_enum;
+	   $$[1] = STATE_LOCAL;
+	   $$[2] = $4;
+	   $$[3] = $4;
+	}
+	;
+
+progEnvParamNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->limits->MaxEnvParams) {
+	      yyerror(& @1, state, "invalid environment parameter reference");
+	      YYERROR;
+	   }
+	   $$ = $1;
+	}
+	;
+
+progLocalParamNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->limits->MaxLocalParams) {
+	      yyerror(& @1, state, "invalid local parameter reference");
+	      YYERROR;
+	   }
+	   $$ = $1;
+	}
+	;
+
+
+
+paramConstDecl: paramConstScalarDecl | paramConstVector;
+paramConstUse: paramConstScalarUse | paramConstVector;
+
+paramConstScalarDecl: signedFloatConstant
+	{
+	   $$.count = 4;
+	   $$.data[0] = $1;
+	   $$.data[1] = $1;
+	   $$.data[2] = $1;
+	   $$.data[3] = $1;
+	}
+	;
+
+paramConstScalarUse: REAL
+	{
+	   $$.count = 1;
+	   $$.data[0] = $1;
+	   $$.data[1] = $1;
+	   $$.data[2] = $1;
+	   $$.data[3] = $1;
+	}
+	| INTEGER
+	{
+	   $$.count = 1;
+	   $$.data[0] = (float) $1;
+	   $$.data[1] = (float) $1;
+	   $$.data[2] = (float) $1;
+	   $$.data[3] = (float) $1;
+	}
+	;
+
+paramConstVector: '{' signedFloatConstant '}'
+	{
+	   $$.count = 4;
+	   $$.data[0] = $2;
+	   $$.data[1] = 0.0f;
+	   $$.data[2] = 0.0f;
+	   $$.data[3] = 1.0f;
+	}
+	| '{' signedFloatConstant ',' signedFloatConstant '}'
+	{
+	   $$.count = 4;
+	   $$.data[0] = $2;
+	   $$.data[1] = $4;
+	   $$.data[2] = 0.0f;
+	   $$.data[3] = 1.0f;
+	}
+	| '{' signedFloatConstant ',' signedFloatConstant ','
+              signedFloatConstant '}'
+	{
+	   $$.count = 4;
+	   $$.data[0] = $2;
+	   $$.data[1] = $4;
+	   $$.data[2] = $6;
+	   $$.data[3] = 1.0f;
+	}
+	| '{' signedFloatConstant ',' signedFloatConstant ','
+              signedFloatConstant ',' signedFloatConstant '}'
+	{
+	   $$.count = 4;
+	   $$.data[0] = $2;
+	   $$.data[1] = $4;
+	   $$.data[2] = $6;
+	   $$.data[3] = $8;
+	}
+	;
+
+signedFloatConstant: optionalSign REAL
+	{
+	   $$ = ($1) ? -$2 : $2;
+	}
+	| optionalSign INTEGER
+	{
+	   $$ = (float)(($1) ? -$2 : $2);
+	}
+	;
+
+optionalSign: '+'        { $$ = FALSE; }
+	| '-'            { $$ = TRUE;  }
+	|                { $$ = FALSE; }
+	;
+
+TEMP_statement: TEMP { $<integer>$ = $1; } varNameList
+	;
+
+ADDRESS_statement: ADDRESS { $<integer>$ = $1; } varNameList
+	;
+
+varNameList: varNameList ',' IDENTIFIER
+	{
+	   if (!declare_variable(state, $3, $<integer>0, & @3)) {
+	      YYERROR;
+	   }
+	}
+	| IDENTIFIER
+	{
+	   if (!declare_variable(state, $1, $<integer>0, & @1)) {
+	      YYERROR;
+	   }
+	}
+	;
+
+OUTPUT_statement: OUTPUT IDENTIFIER '=' resultBinding
+	{
+	   struct asm_symbol *const s =
+	      declare_variable(state, $2, at_output, & @2);
+
+	   if (s == NULL) {
+	      YYERROR;
+	   } else {
+	      s->output_binding = $4;
+	   }
+	}
+	;
+
+resultBinding: RESULT POSITION
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_HPOS;
+	   } else {
+	      yyerror(& @2, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| RESULT FOGCOORD
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_FOGC;
+	   } else {
+	      yyerror(& @2, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| RESULT resultColBinding
+	{
+	   $$ = $2;
+	}
+	| RESULT POINTSIZE
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_PSIZ;
+	   } else {
+	      yyerror(& @2, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| RESULT TEXCOORD optTexCoordUnitNum
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_TEX0 + $3;
+	   } else {
+	      yyerror(& @2, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| RESULT DEPTH
+	{
+	   if (state->mode == ARB_fragment) {
+	      $$ = FRAG_RESULT_DEPTH;
+	   } else {
+	      yyerror(& @2, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	;
+
+resultColBinding: COLOR optResultFaceType optResultColorType
+	{
+	   $$ = $2 + $3;
+	}
+	;
+
+optResultFaceType:
+	{
+	   $$ = (state->mode == ARB_vertex)
+	      ? VERT_RESULT_COL0
+	      : FRAG_RESULT_COLOR;
+	}
+	| FRONT
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_COL0;
+	   } else {
+	      yyerror(& @1, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| BACK
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = VERT_RESULT_BFC0;
+	   } else {
+	      yyerror(& @1, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	;
+
+optResultColorType:
+	{
+	   $$ = 0; 
+	}
+	| PRIMARY
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = 0;
+	   } else {
+	      yyerror(& @1, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	| SECONDARY
+	{
+	   if (state->mode == ARB_vertex) {
+	      $$ = 1;
+	   } else {
+	      yyerror(& @1, state, "invalid program result name");
+	      YYERROR;
+	   }
+	}
+	;
+
+optFaceType:    { $$ = 0; }
+	| FRONT	{ $$ = 0; }
+	| BACK  { $$ = 1; }
+	;
+
+optColorType:       { $$ = 0; }
+	| PRIMARY   { $$ = 0; }
+	| SECONDARY { $$ = 1; }
+	;
+
+optTexCoordUnitNum:                { $$ = 0; }
+	| '[' texCoordUnitNum ']'  { $$ = $2; }
+	;
+
+optTexImageUnitNum:                { $$ = 0; }
+	| '[' texImageUnitNum ']'  { $$ = $2; }
+	;
+
+optLegacyTexUnitNum:               { $$ = 0; }
+	| '[' legacyTexUnitNum ']' { $$ = $2; }
+	;
+
+texCoordUnitNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxTextureCoordUnits) {
+	      yyerror(& @1, state, "invalid texture coordinate unit selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+texImageUnitNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxTextureImageUnits) {
+	      yyerror(& @1, state, "invalid texture image unit selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+legacyTexUnitNum: INTEGER
+	{
+	   if ((unsigned) $1 >= state->MaxTextureUnits) {
+	      yyerror(& @1, state, "invalid texture unit selector");
+	      YYERROR;
+	   }
+
+	   $$ = $1;
+	}
+	;
+
+ALIAS_statement: ALIAS IDENTIFIER '=' IDENTIFIER
+	{
+	   struct asm_symbol *exist = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $2);
+	   struct asm_symbol *target = (struct asm_symbol *)
+	      _mesa_symbol_table_find_symbol(state->st, 0, $4);
+
+
+	   if (exist != NULL) {
+	      yyerror(& @2, state, "redeclared identifier");
+	      YYERROR;
+	   } else if (target == NULL) {
+	      yyerror(& @4, state,
+		      "undefined variable binding in ALIAS statement");
+	      YYERROR;
+	   } else {
+	      _mesa_symbol_table_add_symbol(state->st, 0, $2, target);
+	   }
+	}
+	;
+
+%%
+
+struct asm_instruction *
+asm_instruction_ctor(gl_inst_opcode op,
+		     const struct prog_dst_register *dst,
+		     const struct asm_src_register *src0,
+		     const struct asm_src_register *src1,
+		     const struct asm_src_register *src2)
+{
+   struct asm_instruction *inst = calloc(1, sizeof(struct asm_instruction));
+
+   if (inst) {
+      _mesa_init_instructions(& inst->Base, 1);
+      inst->Base.Opcode = op;
+
+      /* In the core ARB extensions only the KIL instruction doesn't have a
+       * destination register.
+       */
+      if (dst == NULL) {
+	 init_dst_reg(& inst->Base.DstReg);
+      } else {
+	 inst->Base.DstReg = *dst;
+      }
+
+      inst->Base.SrcReg[0] = src0->Base;
+      inst->SrcReg[0] = *src0;
+
+      if (src1 != NULL) {
+	 inst->Base.SrcReg[1] = src1->Base;
+	 inst->SrcReg[1] = *src1;
+      } else {
+	 init_src_reg(& inst->SrcReg[1]);
+      }
+
+      if (src2 != NULL) {
+	 inst->Base.SrcReg[2] = src2->Base;
+	 inst->SrcReg[2] = *src2;
+      } else {
+	 init_src_reg(& inst->SrcReg[2]);
+      }
+   }
+
+   return inst;
+}
+
+
+void
+init_dst_reg(struct prog_dst_register *r)
+{
+   memset(r, 0, sizeof(*r));
+   r->File = PROGRAM_UNDEFINED;
+   r->WriteMask = WRITEMASK_XYZW;
+   r->CondMask = COND_TR;
+   r->CondSwizzle = SWIZZLE_NOOP;
+}
+
+
+void
+init_src_reg(struct asm_src_register *r)
+{
+   memset(r, 0, sizeof(*r));
+   r->Base.File = PROGRAM_UNDEFINED;
+   r->Base.Swizzle = SWIZZLE_NOOP;
+   r->Symbol = NULL;
+}
+
+
+/**
+ * Validate the set of inputs used by a program
+ *
+ * Validates that legal sets of inputs are used by the program.  In this case
+ * "used" included both reading the input or binding the input to a name using
+ * the \c ATTRIB command.
+ *
+ * \return
+ * \c TRUE if the combination of inputs used is valid, \c FALSE otherwise.
+ */
+int
+validate_inputs(struct YYLTYPE *locp, struct asm_parser_state *state)
+{
+   const int inputs = state->prog->InputsRead | state->InputsBound;
+
+   if (((inputs & 0x0ffff) & (inputs >> 16)) != 0) {
+      yyerror(locp, state, "illegal use of generic attribute and name attribute");
+      return 0;
+   }
+
+   return 1;
+}
+
+
+struct asm_symbol *
+declare_variable(struct asm_parser_state *state, char *name, enum asm_type t,
+		 struct YYLTYPE *locp)
+{
+   struct asm_symbol *s = NULL;
+   struct asm_symbol *exist = (struct asm_symbol *)
+      _mesa_symbol_table_find_symbol(state->st, 0, name);
+
+
+   if (exist != NULL) {
+      yyerror(locp, state, "redeclared identifier");
+   } else {
+      s = calloc(1, sizeof(struct asm_symbol));
+      s->name = name;
+      s->type = t;
+
+      switch (t) {
+      case at_temp:
+	 if (state->prog->NumTemporaries >= state->limits->MaxTemps) {
+	    yyerror(locp, state, "too many temporaries declared");
+	    free(s);
+	    return NULL;
+	 }
+
+	 s->temp_binding = state->prog->NumTemporaries;
+	 state->prog->NumTemporaries++;
+	 break;
+
+      case at_address:
+	 if (state->prog->NumAddressRegs >= state->limits->MaxAddressRegs) {
+	    yyerror(locp, state, "too many address registers declared");
+	    free(s);
+	    return NULL;
+	 }
+
+	 /* FINISHME: Add support for multiple address registers.
+	  */
+	 state->prog->NumAddressRegs++;
+	 break;
+
+      default:
+	 break;
+      }
+
+      _mesa_symbol_table_add_symbol(state->st, 0, s->name, s);
+      s->next = state->sym;
+      state->sym = s;
+   }
+
+   return s;
+}
+
+
+int add_state_reference(struct gl_program_parameter_list *param_list,
+			const gl_state_index tokens[STATE_LENGTH])
+{
+   const GLuint size = 4; /* XXX fix */
+   char *name;
+   GLint index;
+
+   name = _mesa_program_state_string(tokens);
+   index = _mesa_add_parameter(param_list, PROGRAM_STATE_VAR, name,
+                               size, GL_NONE, NULL, tokens, 0x0);
+   param_list->StateFlags |= _mesa_program_state_flags(tokens);
+
+   /* free name string here since we duplicated it in add_parameter() */
+   _mesa_free(name);
+
+   return index;
+}
+
+
+int
+initialize_symbol_from_state(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const gl_state_index tokens[STATE_LENGTH])
+{
+   int idx = -1;
+   gl_state_index state_tokens[STATE_LENGTH];
+
+
+   memcpy(state_tokens, tokens, sizeof(state_tokens));
+
+   param_var->type = at_param;
+   param_var->param_binding_type = PROGRAM_STATE_VAR;
+
+   /* If we are adding a STATE_MATRIX that has multiple rows, we need to
+    * unroll it and call add_state_reference() for each row
+    */
+   if ((state_tokens[0] == STATE_MODELVIEW_MATRIX ||
+	state_tokens[0] == STATE_PROJECTION_MATRIX ||
+	state_tokens[0] == STATE_MVP_MATRIX ||
+	state_tokens[0] == STATE_TEXTURE_MATRIX ||
+	state_tokens[0] == STATE_PROGRAM_MATRIX)
+       && (state_tokens[2] != state_tokens[3])) {
+      int row;
+      const int first_row = state_tokens[2];
+      const int last_row = state_tokens[3];
+
+      for (row = first_row; row <= last_row; row++) {
+	 state_tokens[2] = state_tokens[3] = row;
+
+	 idx = add_state_reference(prog->Parameters, state_tokens);
+	 if (param_var->param_binding_begin == ~0U)
+	    param_var->param_binding_begin = idx;
+	 param_var->param_binding_length++;
+      }
+   }
+   else {
+      idx = add_state_reference(prog->Parameters, state_tokens);
+      if (param_var->param_binding_begin == ~0U)
+	 param_var->param_binding_begin = idx;
+      param_var->param_binding_length++;
+   }
+
+   return idx;
+}
+
+
+int
+initialize_symbol_from_param(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const gl_state_index tokens[STATE_LENGTH])
+{
+   int idx = -1;
+   gl_state_index state_tokens[STATE_LENGTH];
+
+
+   memcpy(state_tokens, tokens, sizeof(state_tokens));
+
+   assert((state_tokens[0] == STATE_VERTEX_PROGRAM)
+	  || (state_tokens[0] == STATE_FRAGMENT_PROGRAM));
+   assert((state_tokens[1] == STATE_ENV)
+	  || (state_tokens[1] == STATE_LOCAL));
+
+   param_var->type = at_param;
+   param_var->param_binding_type = (state_tokens[1] == STATE_ENV)
+     ? PROGRAM_ENV_PARAM : PROGRAM_LOCAL_PARAM;
+
+   /* If we are adding a STATE_ENV or STATE_LOCAL that has multiple elements,
+    * we need to unroll it and call add_state_reference() for each row
+    */
+   if (state_tokens[2] != state_tokens[3]) {
+      int row;
+      const int first_row = state_tokens[2];
+      const int last_row = state_tokens[3];
+
+      for (row = first_row; row <= last_row; row++) {
+	 state_tokens[2] = state_tokens[3] = row;
+
+	 idx = add_state_reference(prog->Parameters, state_tokens);
+	 if (param_var->param_binding_begin == ~0U)
+	    param_var->param_binding_begin = idx;
+	 param_var->param_binding_length++;
+      }
+   }
+   else {
+      idx = add_state_reference(prog->Parameters, state_tokens);
+      if (param_var->param_binding_begin == ~0U)
+	 param_var->param_binding_begin = idx;
+      param_var->param_binding_length++;
+   }
+
+   return idx;
+}
+
+
+int
+initialize_symbol_from_const(struct gl_program *prog,
+			     struct asm_symbol *param_var, 
+			     const struct asm_vector *vec)
+{
+   const int idx = _mesa_add_parameter(prog->Parameters, PROGRAM_CONSTANT,
+				       NULL, vec->count, GL_NONE, vec->data,
+				       NULL, 0x0);
+
+   param_var->type = at_param;
+   param_var->param_binding_type = PROGRAM_CONSTANT;
+
+   if (param_var->param_binding_begin == ~0U)
+      param_var->param_binding_begin = idx;
+   param_var->param_binding_length++;
+
+   return idx;
+}
+
+
+char *
+make_error_string(const char *fmt, ...)
+{
+   int length;
+   char *str;
+   va_list args;
+
+   va_start(args, fmt);
+
+   /* Call vsnprintf once to determine how large the final string is.  Call it
+    * again to do the actual formatting.  from the vsnprintf manual page:
+    *
+    *    Upon successful return, these functions return the number of
+    *    characters printed  (not including the trailing '\0' used to end
+    *    output to strings).
+    */
+   length = 1 + vsnprintf(NULL, 0, fmt, args);
+
+   str = _mesa_malloc(length);
+   if (str) {
+      vsnprintf(str, length, fmt, args);
+   }
+
+   va_end(args);
+
+   return str;
+}
+
+
+void
+yyerror(YYLTYPE *locp, struct asm_parser_state *state, const char *s)
+{
+   char *err_str;
+
+
+   err_str = make_error_string("glProgramStringARB(%s)\n", s);
+   if (err_str) {
+      _mesa_error(state->ctx, GL_INVALID_OPERATION, err_str);
+      _mesa_free(err_str);
+   }
+
+   err_str = make_error_string("line %u, char %u: error: %s\n",
+			       locp->first_line, locp->first_column, s);
+   _mesa_set_program_error(state->ctx, locp->position, err_str);
+
+   if (err_str) {
+      _mesa_free(err_str);
+   }
+}
+
+
+GLboolean
+_mesa_parse_arb_program(GLcontext *ctx, GLenum target, const GLubyte *str,
+			GLsizei len, struct asm_parser_state *state)
+{
+   struct asm_instruction *inst;
+   unsigned i;
+   GLubyte *strz;
+   GLboolean result = GL_FALSE;
+   void *temp;
+   struct asm_symbol *sym;
+
+   state->ctx = ctx;
+   state->prog->Target = target;
+   state->prog->Parameters = _mesa_new_parameter_list();
+
+   /* Make a copy of the program string and force it to be NUL-terminated.
+    */
+   strz = (GLubyte *) _mesa_malloc(len + 1);
+   if (strz == NULL) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glProgramStringARB");
+      return GL_FALSE;
+   }
+   _mesa_memcpy (strz, str, len);
+   strz[len] = '\0';
+
+   state->prog->String = strz;
+
+   state->st = _mesa_symbol_table_ctor();
+
+   state->limits = (target == GL_VERTEX_PROGRAM_ARB)
+      ? & ctx->Const.VertexProgram
+      : & ctx->Const.FragmentProgram;
+
+   state->MaxTextureImageUnits = ctx->Const.MaxTextureImageUnits;
+   state->MaxTextureCoordUnits = ctx->Const.MaxTextureCoordUnits;
+   state->MaxTextureUnits = ctx->Const.MaxTextureUnits;
+   state->MaxClipPlanes = ctx->Const.MaxClipPlanes;
+   state->MaxLights = ctx->Const.MaxLights;
+   state->MaxProgramMatrices = ctx->Const.MaxProgramMatrices;
+
+   state->state_param_enum = (target == GL_VERTEX_PROGRAM_ARB)
+      ? STATE_VERTEX_PROGRAM : STATE_FRAGMENT_PROGRAM;
+
+   _mesa_set_program_error(ctx, -1, NULL);
+
+   _mesa_program_lexer_ctor(& state->scanner, state, (const char *) str, len);
+   yyparse(state);
+   _mesa_program_lexer_dtor(state->scanner);
+
+
+   if (ctx->Program.ErrorPos != -1) {
+      goto error;
+   }
+
+   if (! _mesa_layout_parameters(state)) {
+      struct YYLTYPE loc;
+
+      loc.first_line = 0;
+      loc.first_column = 0;
+      loc.position = len;
+
+      yyerror(& loc, state, "invalid PARAM usage");
+      goto error;
+   }
+
+
+   
+   /* Add one instruction to store the "END" instruction.
+    */
+   state->prog->Instructions =
+      _mesa_alloc_instructions(state->prog->NumInstructions + 1);
+   inst = state->inst_head;
+   for (i = 0; i < state->prog->NumInstructions; i++) {
+      struct asm_instruction *const temp = inst->next;
+
+      state->prog->Instructions[i] = inst->Base;
+      inst = temp;
+   }
+
+   /* Finally, tag on an OPCODE_END instruction */
+   {
+      const GLuint numInst = state->prog->NumInstructions;
+      _mesa_init_instructions(state->prog->Instructions + numInst, 1);
+      state->prog->Instructions[numInst].Opcode = OPCODE_END;
+   }
+   state->prog->NumInstructions++;
+
+   state->prog->NumParameters = state->prog->Parameters->NumParameters;
+   state->prog->NumAttributes = _mesa_bitcount(state->prog->InputsRead);
+
+   /*
+    * Initialize native counts to logical counts.  The device driver may
+    * change them if program is translated into a hardware program.
+    */
+   state->prog->NumNativeInstructions = state->prog->NumInstructions;
+   state->prog->NumNativeTemporaries = state->prog->NumTemporaries;
+   state->prog->NumNativeParameters = state->prog->NumParameters;
+   state->prog->NumNativeAttributes = state->prog->NumAttributes;
+   state->prog->NumNativeAddressRegs = state->prog->NumAddressRegs;
+
+   result = GL_TRUE;
+
+error:
+   for (inst = state->inst_head; inst != NULL; inst = temp) {
+      temp = inst->next;
+      _mesa_free(inst);
+   }
+
+   state->inst_head = NULL;
+   state->inst_tail = NULL;
+
+   for (sym = state->sym; sym != NULL; sym = temp) {
+      temp = sym->next;
+
+      _mesa_free((void *) sym->name);
+      _mesa_free(sym);
+   }
+   state->sym = NULL;
+
+   _mesa_symbol_table_dtor(state->st);
+   state->st = NULL;
+
+   return result;
+}
diff --git a/src/mesa/shader/program_parse_extra.c b/src/mesa/shader/program_parse_extra.c
new file mode 100644
index 0000000000..8e4be606f1
--- /dev/null
+++ b/src/mesa/shader/program_parse_extra.c
@@ -0,0 +1,117 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <string.h>
+#include "main/mtypes.h"
+#include "prog_instruction.h"
+#include "program_parser.h"
+
+
+/**
+ * Extra assembly-level parser routines
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+int
+_mesa_ARBvp_parse_option(struct asm_parser_state *state, const char *option)
+{
+   if (strcmp(option, "ARB_position_invariant") == 0) {
+      state->option.PositionInvariant = 1;
+      return 1;
+   }
+
+   return 0;
+}
+
+
+int
+_mesa_ARBfp_parse_option(struct asm_parser_state *state, const char *option)
+{
+   /* All of the options currently supported start with "ARB_".  The code is
+    * currently structured with nested if-statements because eventually options
+    * that start with "NV_" will be supported.  This structure will result in
+    * less churn when those options are added.
+    */
+   if (strncmp(option, "ARB_", 4) == 0) {
+      /* Advance the pointer past the "ARB_" prefix.
+       */
+      option += 4;
+
+
+      if (strncmp(option, "fog_", 4) == 0) {
+	 option += 4;
+
+	 if (state->option.Fog == OPTION_NONE) {
+	    if (strcmp(option, "exp") == 0) {
+	       state->option.Fog = OPTION_FOG_EXP;
+	       return 1;
+	    } else if (strcmp(option, "exp2") == 0) {
+	       state->option.Fog = OPTION_FOG_EXP2;
+	       return 1;
+	    } else if (strcmp(option, "linear") == 0) {
+	       state->option.Fog = OPTION_FOG_LINEAR;
+	       return 1;
+	    }
+	 }
+
+	 return 0;
+      } else if (strncmp(option, "precision_hint_", 15) == 0) {
+	 option += 15;
+
+	 if (state->option.PrecisionHint == OPTION_NONE) {
+	    if (strcmp(option, "nicest") == 0) {
+	       state->option.PrecisionHint = OPTION_NICEST;
+	       return 1;
+	    } else if (strcmp(option, "fastest") == 0) {
+	       state->option.PrecisionHint = OPTION_FASTEST;
+	       return 1;
+	    }
+	 }
+
+	 return 0;
+      } else if (strcmp(option, "draw_buffers") == 0) {
+	 /* Don't need to check extension availability because all Mesa-based
+	  * drivers support GL_ARB_draw_buffers.
+	  */
+	 state->option.DrawBuffers = 1;
+	 return 1;
+      } else if (strcmp(option, "fragment_program_shadow") == 0) {
+	 if (state->ctx->Extensions.ARB_fragment_program_shadow) {
+	    state->option.Shadow = 1;
+	    return 1;
+	 }
+      }
+   } else if (strncmp(option, "MESA_", 5) == 0) {
+      option += 5;
+
+      if (strcmp(option, "texture_array") == 0) {
+	 if (state->ctx->Extensions.MESA_texture_array) {
+	    state->option.TexArray = 1;
+	    return 1;
+	 }
+      }
+   }
+
+   return 0;
+}
diff --git a/src/mesa/shader/program_parser.h b/src/mesa/shader/program_parser.h
new file mode 100644
index 0000000000..fa47d84565
--- /dev/null
+++ b/src/mesa/shader/program_parser.h
@@ -0,0 +1,266 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#pragma once
+
+#include "main/config.h"
+
+#ifndef MTYPES_H
+struct __GLcontextRec;
+typedef struct __GLcontextRec GLcontext;
+#endif
+
+enum asm_type {
+   at_none,
+   at_address,
+   at_attrib,
+   at_param,
+   at_temp,
+   at_output,
+};
+
+struct asm_symbol {
+   struct asm_symbol *next;    /**< List linkage for freeing. */
+   const char *name;
+   enum asm_type type;
+   unsigned attrib_binding;
+   unsigned output_binding;   /**< Output / result register number. */
+
+   /**
+    * One of PROGRAM_STATE_VAR, PROGRAM_LOCAL_PARAM, or PROGRAM_ENV_PARAM.
+    */
+   unsigned param_binding_type;
+
+   /** 
+    * Offset into the program_parameter_list where the tokens representing our
+    * bound state (or constants) start.
+    */
+   unsigned param_binding_begin;
+
+   /* This is how many entries in the the program_parameter_list we take up
+    * with our state tokens or constants. Note that this is _not_ the same as
+    * the number of param registers we eventually use.
+    */
+   unsigned param_binding_length;
+
+   /**
+    * Index of the temp register assigned to this variable.
+    */
+   unsigned temp_binding;
+
+   /**
+    * Flag whether or not a PARAM is an array
+    */
+   unsigned param_is_array:1;
+
+
+   /**
+    * Flag whether or not a PARAM array is accessed indirectly
+    */
+   unsigned param_accessed_indirectly:1;
+
+
+   /**
+    * \brief Is first pass of parameter layout done with this variable?
+    *
+    * The parameter layout routine operates in two passes.  This flag tracks
+    * whether or not the first pass has handled this variable.
+    *
+    * \sa _mesa_layout_parameters
+    */
+   unsigned pass1_done:1;
+};
+
+
+struct asm_vector {
+   unsigned count;
+   float    data[4];
+};
+
+
+struct asm_swizzle_mask {
+   unsigned swizzle:12;
+   unsigned mask:4;
+};
+
+
+struct asm_src_register {
+   struct prog_src_register Base;
+
+   /**
+    * Symbol associated with indirect access to parameter arrays.
+    *
+    * If \c Base::RelAddr is 1, this will point to the symbol for the parameter
+    * that is being dereferenced.  Further, \c Base::Index will be the offset
+    * from the address register being used.
+    */
+   struct asm_symbol *Symbol;
+};
+
+
+struct asm_instruction {
+   struct prog_instruction Base;
+   struct asm_instruction *next;
+   struct asm_src_register SrcReg[3];
+};
+
+
+struct asm_parser_state {
+   GLcontext *ctx;
+   struct gl_program *prog;
+
+   /**
+    * Per-program target limits
+    */
+   struct gl_program_constants *limits;
+
+   struct _mesa_symbol_table *st;
+
+   /**
+    * Linked list of symbols
+    *
+    * This list is \b only used when cleaning up compiler state and freeing
+    * memory.
+    */
+   struct asm_symbol *sym;
+
+   /**
+    * State for the lexer.
+    */
+   void *scanner;
+
+   /**
+    * Linked list of instructions generated during parsing.
+    */
+   /*@{*/
+   struct asm_instruction *inst_head;
+   struct asm_instruction *inst_tail;
+   /*@}*/
+
+
+   /**
+    * Selected limits copied from gl_constants
+    *
+    * These are limits from the GL context, but various bits in the program
+    * must be validated against these values.
+    */
+   /*@{*/
+   unsigned MaxTextureCoordUnits;
+   unsigned MaxTextureImageUnits;
+   unsigned MaxTextureUnits;
+   unsigned MaxClipPlanes;
+   unsigned MaxLights;
+   unsigned MaxProgramMatrices;
+   /*@}*/
+
+   /**
+    * Value to use in state vector accessors for environment and local
+    * parameters
+    */
+   unsigned state_param_enum;
+
+
+   /**
+    * Input attributes bound to specific names
+    *
+    * This is only needed so that errors can be properly produced when
+    * multiple ATTRIB statements bind illegal combinations of vertex
+    * attributes.
+    */
+   unsigned InputsBound;
+
+   enum {
+      invalid_mode = 0,
+      ARB_vertex,
+      ARB_fragment
+   } mode;
+
+   struct {
+      unsigned PositionInvariant:1;
+      unsigned Fog:2;
+      unsigned PrecisionHint:2;
+      unsigned DrawBuffers:1;
+      unsigned Shadow:1;
+      unsigned TexRect:1;
+      unsigned TexArray:1;
+   } option;
+
+   struct {
+      unsigned UsesKill:1;
+   } fragment;
+};
+
+#define OPTION_NONE        0
+#define OPTION_FOG_EXP     1
+#define OPTION_FOG_EXP2    2
+#define OPTION_FOG_LINEAR  3
+#define OPTION_NICEST      1
+#define OPTION_FASTEST     2
+
+typedef struct YYLTYPE {
+   int first_line;
+   int first_column;
+   int last_line;
+   int last_column;
+   int position;
+} YYLTYPE;
+
+#define YYLTYPE_IS_DECLARED 1
+#define YYLTYPE_IS_TRIVIAL 1
+
+
+extern GLboolean _mesa_parse_arb_program(GLcontext *ctx, GLenum target,
+    const GLubyte *str, GLsizei len, struct asm_parser_state *state);
+
+
+
+/* From program_lexer.l. */
+extern void _mesa_program_lexer_dtor(void *scanner);
+
+extern void _mesa_program_lexer_ctor(void **scanner,
+    struct asm_parser_state *state, const char *string, size_t len);
+
+
+/**
+ *\name From program_parse_extra.c
+ */
+/*@{*/
+
+/**
+ * Parses and processes an option string to an ARB vertex program
+ *
+ * \return
+ * Non-zero on success, zero on failure.
+ */
+extern int _mesa_ARBvp_parse_option(struct asm_parser_state *state,
+    const char *option);
+
+/**
+ * Parses and processes an option string to an ARB fragment program
+ *
+ * \return
+ * Non-zero on success, zero on failure.
+ */
+extern int _mesa_ARBfp_parse_option(struct asm_parser_state *state,
+    const char *option);
+
+/*@}*/
diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c
index ac5fe0f691..f70c75cec8 100644
--- a/src/mesa/shader/programopt.c
+++ b/src/mesa/shader/programopt.c
@@ -396,7 +396,6 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
    fprog->Base.Instructions = newInst;
    fprog->Base.NumInstructions = inst - newInst;
    fprog->Base.InputsRead |= FRAG_BIT_FOGC;
-   fprog->UsesFogFragCoord = GL_TRUE;
    /* XXX do this?  fprog->FogOption = GL_NONE; */
 }
 
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 4d5d1bef3d..178b7d0dba 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 2004-2008  Brian Paul   All Rights Reserved.
  * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
@@ -1431,6 +1431,9 @@ _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source)
    }
    sh->Source = source;
    sh->CompileStatus = GL_FALSE;
+#ifdef DEBUG
+   sh->SourceChecksum = _mesa_str_checksum(sh->Source);
+#endif
 }
 
 
@@ -1503,7 +1506,7 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
 
    if (program) {
       shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram");
@@ -1521,10 +1524,15 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
          GLuint i;
          _mesa_printf("Use Shader %u\n", shProg->Name);
          for (i = 0; i < shProg->NumShaders; i++) {
-            _mesa_printf(" shader %u, type 0x%x\n",
+            _mesa_printf(" shader %u, type 0x%x, checksum %u\n",
                          shProg->Shaders[i]->Name,
-                         shProg->Shaders[i]->Type);
+                         shProg->Shaders[i]->Type,
+                         shProg->Shaders[i]->SourceChecksum);
          }
+         if (shProg->VertexProgram)
+            printf(" vert prog %u\n", shProg->VertexProgram->Base.Id);
+         if (shProg->FragmentProgram)
+            printf(" frag prog %u\n", shProg->FragmentProgram->Base.Id);
       }
    }
    else {
@@ -1560,10 +1568,11 @@ _mesa_update_shader_textures_used(struct gl_program *prog)
 
    for (s = 0; s < MAX_SAMPLERS; s++) {
       if (prog->SamplersUsed & (1 << s)) {
-         GLuint u = prog->SamplerUnits[s];
-         GLuint t = prog->SamplerTargets[s];
-         assert(u < MAX_TEXTURE_IMAGE_UNITS);
-         prog->TexturesUsed[u] |= (1 << t);
+         GLuint unit = prog->SamplerUnits[s];
+         GLuint tgt = prog->SamplerTargets[s];
+         assert(unit < MAX_TEXTURE_IMAGE_UNITS);
+         assert(tgt < NUM_TEXTURE_TARGETS);
+         prog->TexturesUsed[unit] |= (1 << tgt);
       }
    }
 }
@@ -1618,10 +1627,8 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
                     GLenum type, GLsizei count, GLint elems,
                     const void *values)
 {
-   struct gl_program_parameter *param =
+   const struct gl_program_parameter *param =
       &program->Parameters->Parameters[index];
-   const GLboolean isUniformBool = is_boolean_type(param->DataType);
-   const GLboolean areIntValues = is_integer_type(type);
 
    assert(offset >= 0);
    assert(elems >= 1);
@@ -1642,17 +1649,10 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
       GLboolean changed = GL_FALSE;
       GLint i;
 
-      /* data type for setting samplers must be int */
-      if (type != GL_INT) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glUniform(only glUniform1i can be used "
-                     "to set sampler uniforms)");
-         return;
-      }
+      /* this should have been caught by the compatible_types() check */
+      ASSERT(type == GL_INT);
 
-      /* XXX arrays of samplers haven't been tested much, but it's not a
-       * common thing...
-       */
+      /* loop over number of samplers to change */
       for (i = 0; i < count; i++) {
          GLuint sampler =
             (GLuint) program->Parameters->ParameterValues[index + offset + i][0];
@@ -1691,9 +1691,11 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
    }
    else {
       /* ordinary uniform variable */
-      GLsizei k, i;
+      const GLboolean isUniformBool = is_boolean_type(param->DataType);
+      const GLboolean areIntValues = is_integer_type(type);
       const GLint slots = (param->Size + 3) / 4;
       const GLint typeSize = sizeof_glsl_type(param->DataType);
+      GLsizei k, i;
 
       if (param->Size > typeSize) {
          /* an array */
@@ -1820,7 +1822,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
@@ -1960,7 +1962,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
@@ -1988,19 +1990,74 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows,
 }
 
 
-static void
-_mesa_validate_program(GLcontext *ctx, GLuint program)
+/**
+ * Validate a program's samplers.
+ * Specifically, check that there aren't two samplers of different types
+ * pointing to the same texture unit.
+ * \return GL_TRUE if valid, GL_FALSE if invalid
+ */
+static GLboolean
+validate_samplers(GLcontext *ctx, const struct gl_program *prog, char *errMsg)
 {
-   struct gl_shader_program *shProg;
+   static const char *targetName[] = {
+      "TEXTURE_2D_ARRAY",
+      "TEXTURE_1D_ARRAY",
+      "TEXTURE_CUBE",
+      "TEXTURE_3D",
+      "TEXTURE_RECT",
+      "TEXTURE_2D",
+      "TEXTURE_1D",
+   };
+   GLint targetUsed[MAX_TEXTURE_IMAGE_UNITS];
+   GLbitfield samplersUsed = prog->SamplersUsed;
+   GLuint i;
 
-   shProg = _mesa_lookup_shader_program_err(ctx, program, "glValidateProgram");
-   if (!shProg) {
-      return;
+   assert(Elements(targetName) == NUM_TEXTURE_TARGETS);
+
+   if (samplersUsed == 0x0)
+      return GL_TRUE;
+
+   for (i = 0; i < Elements(targetUsed); i++)
+      targetUsed[i] = -1;
+
+   /* walk over bits which are set in 'samplers' */
+   while (samplersUsed) {
+      GLuint unit;
+      gl_texture_index target;
+      GLint sampler = _mesa_ffs(samplersUsed) - 1;
+      assert(sampler >= 0);
+      assert(sampler < MAX_TEXTURE_IMAGE_UNITS);
+      unit = prog->SamplerUnits[sampler];
+      target = prog->SamplerTargets[sampler];
+      if (targetUsed[unit] != -1 && targetUsed[unit] != target) {
+         _mesa_snprintf(errMsg, 100,
+                       "Texture unit %d is accessed both as %s and %s",
+                       unit, targetName[targetUsed[unit]], targetName[target]);
+         return GL_FALSE;
+      }
+      targetUsed[unit] = target;
+      samplersUsed ^= (1 << sampler);
    }
 
+   return GL_TRUE;
+}
+
+
+/**
+ * Do validation of the given shader program.
+ * \param errMsg  returns error message if validation fails.
+ * \return GL_TRUE if valid, GL_FALSE if invalid (and set errMsg)
+ */
+GLboolean
+_mesa_validate_shader_program(GLcontext *ctx,
+                              const struct gl_shader_program *shProg,
+                              char *errMsg)
+{
+   const struct gl_vertex_program *vp = shProg->VertexProgram;
+   const struct gl_fragment_program *fp = shProg->FragmentProgram;
+
    if (!shProg->LinkStatus) {
-      shProg->Validated = GL_FALSE;
-      return;
+      return GL_FALSE;
    }
 
    /* From the GL spec, a program is invalid if any of these are true:
@@ -2018,7 +2075,44 @@ _mesa_validate_program(GLcontext *ctx, GLuint program)
      image units allowed.
    */
 
-   shProg->Validated = GL_TRUE;
+
+   /*
+    * Check: any two active samplers in the current program object are of
+    * different types, but refer to the same texture image unit,
+    */
+   if (vp && !validate_samplers(ctx, &vp->Base, errMsg)) {
+      return GL_FALSE;
+   }
+   if (fp && !validate_samplers(ctx, &fp->Base, errMsg)) {
+      return GL_FALSE;
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Called via glValidateProgram()
+ */
+static void
+_mesa_validate_program(GLcontext *ctx, GLuint program)
+{
+   struct gl_shader_program *shProg;
+   char errMsg[100];
+
+   shProg = _mesa_lookup_shader_program_err(ctx, program, "glValidateProgram");
+   if (!shProg) {
+      return;
+   }
+
+   shProg->Validated = _mesa_validate_shader_program(ctx, shProg, errMsg);
+   if (!shProg->Validated) {
+      /* update info log */
+      if (shProg->InfoLog) {
+         _mesa_free(shProg->InfoLog);
+      }
+      shProg->InfoLog = _mesa_strdup(errMsg);
+   }
 }
 
 
diff --git a/src/mesa/shader/shader_api.h b/src/mesa/shader/shader_api.h
index ec1996ee98..d08d47373e 100644
--- a/src/mesa/shader/shader_api.h
+++ b/src/mesa/shader/shader_api.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.6
  *
  * Copyright (C) 2004-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -87,6 +88,11 @@ extern void
 _mesa_use_program(GLcontext *ctx, GLuint program);
 
 
+extern GLboolean
+_mesa_validate_shader_program(GLcontext *ctx,
+                              const struct gl_shader_program *shProg,
+                              char *errMsg);
+
 extern void
 _mesa_init_glsl_driver_functions(struct dd_function_table *driver);
 
diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c
index 4e67241f3b..bef0f85653 100644
--- a/src/mesa/shader/slang/slang_builtin.c
+++ b/src/mesa/shader/slang/slang_builtin.c
@@ -710,3 +710,186 @@ _slang_alloc_statevar(slang_ir_node *n,
    *direct = GL_FALSE;
    return alloc_state_var_array(n->Var, paramList);
 }
+
+
+
+
+#define SWIZZLE_ZWWW MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W)
+
+
+/** Predefined shader inputs */
+struct input_info
+{
+   const char *Name;
+   GLuint Attrib;
+   GLenum Type;
+   GLuint Swizzle;
+};
+
+/** Predefined vertex shader inputs/attributes */
+static const struct input_info vertInputs[] = {
+   { "gl_Vertex", VERT_ATTRIB_POS, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_Normal", VERT_ATTRIB_NORMAL, GL_FLOAT_VEC3, SWIZZLE_NOOP },
+   { "gl_Color", VERT_ATTRIB_COLOR0, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_SecondaryColor", VERT_ATTRIB_COLOR1, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_FogCoord", VERT_ATTRIB_FOG, GL_FLOAT, SWIZZLE_XXXX },
+   { "gl_MultiTexCoord0", VERT_ATTRIB_TEX0, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord1", VERT_ATTRIB_TEX1, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord2", VERT_ATTRIB_TEX2, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord3", VERT_ATTRIB_TEX3, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord4", VERT_ATTRIB_TEX4, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord5", VERT_ATTRIB_TEX5, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord6", VERT_ATTRIB_TEX6, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_MultiTexCoord7", VERT_ATTRIB_TEX7, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { NULL, 0, SWIZZLE_NOOP }
+};
+
+/** Predefined fragment shader inputs */
+static const struct input_info fragInputs[] = {
+   { "gl_FragCoord", FRAG_ATTRIB_WPOS, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_Color", FRAG_ATTRIB_COL0, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_SecondaryColor", FRAG_ATTRIB_COL1, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   { "gl_TexCoord", FRAG_ATTRIB_TEX0, GL_FLOAT_VEC4, SWIZZLE_NOOP },
+   /* note: we're packing several quantities into the fogcoord vector */
+   { "gl_FogFragCoord", FRAG_ATTRIB_FOGC, GL_FLOAT, SWIZZLE_XXXX },
+   { "gl_FrontFacing", FRAG_ATTRIB_FACE, GL_FLOAT, SWIZZLE_XXXX },
+   { "gl_PointCoord", FRAG_ATTRIB_PNTC, GL_FLOAT_VEC2, SWIZZLE_XYZW },
+   { NULL, 0, SWIZZLE_NOOP }
+};
+
+
+/**
+ * Return the VERT_ATTRIB_* or FRAG_ATTRIB_* value that corresponds to
+ * a vertex or fragment program input variable.  Return -1 if the input
+ * name is invalid.
+ * XXX return size too
+ */
+GLint
+_slang_input_index(const char *name, GLenum target, GLuint *swizzleOut)
+{
+   const struct input_info *inputs;
+   GLuint i;
+
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB:
+      inputs = vertInputs;
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      inputs = fragInputs;
+      break;
+   /* XXX geom program */
+   default:
+      _mesa_problem(NULL, "bad target in _slang_input_index");
+      return -1;
+   }
+
+   ASSERT(MAX_TEXTURE_COORD_UNITS == 8); /* if this fails, fix vertInputs above */
+
+   for (i = 0; inputs[i].Name; i++) {
+      if (strcmp(inputs[i].Name, name) == 0) {
+         /* found */
+         *swizzleOut = inputs[i].Swizzle;
+         return inputs[i].Attrib;
+      }
+   }
+   return -1;
+}
+
+
+/**
+ * Return name of the given vertex attribute (VERT_ATTRIB_x).
+ */
+const char *
+_slang_vert_attrib_name(GLuint attrib)
+{
+   GLuint i;
+   assert(attrib < VERT_ATTRIB_GENERIC0);
+   for (i = 0; vertInputs[i].Name; i++) {
+      if (vertInputs[i].Attrib == attrib)
+         return vertInputs[i].Name;
+   }
+   return NULL;
+}
+
+
+/**
+ * Return type (GL_FLOAT, GL_FLOAT_VEC2, etc) of the given vertex
+ * attribute (VERT_ATTRIB_x).
+ */
+GLenum
+_slang_vert_attrib_type(GLuint attrib)
+{
+   GLuint i;
+   assert(attrib < VERT_ATTRIB_GENERIC0);
+   for (i = 0; vertInputs[i].Name; i++) {
+      if (vertInputs[i].Attrib == attrib)
+         return vertInputs[i].Type;
+   }
+   return GL_NONE;
+}
+
+
+
+
+
+/** Predefined shader output info */
+struct output_info
+{
+   const char *Name;
+   GLuint Attrib;
+};
+
+/** Predefined vertex shader outputs */
+static const struct output_info vertOutputs[] = {
+   { "gl_Position", VERT_RESULT_HPOS },
+   { "gl_FrontColor", VERT_RESULT_COL0 },
+   { "gl_BackColor", VERT_RESULT_BFC0 },
+   { "gl_FrontSecondaryColor", VERT_RESULT_COL1 },
+   { "gl_BackSecondaryColor", VERT_RESULT_BFC1 },
+   { "gl_TexCoord", VERT_RESULT_TEX0 },
+   { "gl_FogFragCoord", VERT_RESULT_FOGC },
+   { "gl_PointSize", VERT_RESULT_PSIZ },
+   { NULL, 0 }
+};
+
+/** Predefined fragment shader outputs */
+static const struct output_info fragOutputs[] = {
+   { "gl_FragColor", FRAG_RESULT_COLOR },
+   { "gl_FragDepth", FRAG_RESULT_DEPTH },
+   { "gl_FragData", FRAG_RESULT_DATA0 },
+   { NULL, 0 }
+};
+
+
+/**
+ * Return the VERT_RESULT_* or FRAG_RESULT_* value that corresponds to
+ * a vertex or fragment program output variable.  Return -1 for an invalid
+ * output name.
+ */
+GLint
+_slang_output_index(const char *name, GLenum target)
+{
+   const struct output_info *outputs;
+   GLuint i;
+
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB:
+      outputs = vertOutputs;
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      outputs = fragOutputs;
+      break;
+   /* XXX geom program */
+   default:
+      _mesa_problem(NULL, "bad target in _slang_output_index");
+      return -1;
+   }
+
+   for (i = 0; outputs[i].Name; i++) {
+      if (strcmp(outputs[i].Name, name) == 0) {
+         /* found */
+         return outputs[i].Attrib;
+      }
+   }
+   return -1;
+}
diff --git a/src/mesa/shader/slang/slang_builtin.h b/src/mesa/shader/slang/slang_builtin.h
index 7f6fe80fcc..f814d11ac7 100644
--- a/src/mesa/shader/slang/slang_builtin.h
+++ b/src/mesa/shader/slang/slang_builtin.h
@@ -37,4 +37,18 @@ _slang_alloc_statevar(slang_ir_node *n,
                       GLboolean *direct);
 
 
+extern GLint
+_slang_input_index(const char *name, GLenum target, GLuint *swizzleOut);
+
+extern GLint
+_slang_output_index(const char *name, GLenum target);
+
+
+extern const char *
+_slang_vert_attrib_name(GLuint attrib);
+
+extern GLenum
+_slang_vert_attrib_type(GLuint attrib);
+
+
 #endif /* SLANG_BUILTIN_H */
diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c
index 2b7e781f98..349f432dec 100644
--- a/src/mesa/shader/slang/slang_codegen.c
+++ b/src/mesa/shader/slang/slang_codegen.c
@@ -46,6 +46,7 @@
 #include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "slang_typeinfo.h"
+#include "slang_builtin.h"
 #include "slang_codegen.h"
 #include "slang_compile.h"
 #include "slang_label.h"
@@ -342,109 +343,6 @@ slang_operation_identifier(slang_operation *oper,
 }
 
 
-#define SWIZZLE_ZWWW MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W)
-
-/**
- * Return the VERT_ATTRIB_* or FRAG_ATTRIB_* value that corresponds to
- * a vertex or fragment program input variable.  Return -1 if the input
- * name is invalid.
- * XXX return size too
- */
-static GLint
-_slang_input_index(const char *name, GLenum target, GLuint *swizzleOut)
-{
-   struct input_info {
-      const char *Name;
-      GLuint Attrib;
-      GLuint Swizzle;
-   };
-   static const struct input_info vertInputs[] = {
-      { "gl_Vertex", VERT_ATTRIB_POS, SWIZZLE_NOOP },
-      { "gl_Normal", VERT_ATTRIB_NORMAL, SWIZZLE_NOOP },
-      { "gl_Color", VERT_ATTRIB_COLOR0, SWIZZLE_NOOP },
-      { "gl_SecondaryColor", VERT_ATTRIB_COLOR1, SWIZZLE_NOOP },
-      { "gl_FogCoord", VERT_ATTRIB_FOG, SWIZZLE_XXXX },
-      { "gl_MultiTexCoord0", VERT_ATTRIB_TEX0, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord1", VERT_ATTRIB_TEX1, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord2", VERT_ATTRIB_TEX2, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord3", VERT_ATTRIB_TEX3, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord4", VERT_ATTRIB_TEX4, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord5", VERT_ATTRIB_TEX5, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord6", VERT_ATTRIB_TEX6, SWIZZLE_NOOP },
-      { "gl_MultiTexCoord7", VERT_ATTRIB_TEX7, SWIZZLE_NOOP },
-      { NULL, 0, SWIZZLE_NOOP }
-   };
-   static const struct input_info fragInputs[] = {
-      { "gl_FragCoord", FRAG_ATTRIB_WPOS, SWIZZLE_NOOP },
-      { "gl_Color", FRAG_ATTRIB_COL0, SWIZZLE_NOOP },
-      { "gl_SecondaryColor", FRAG_ATTRIB_COL1, SWIZZLE_NOOP },
-      { "gl_TexCoord", FRAG_ATTRIB_TEX0, SWIZZLE_NOOP },
-      /* note: we're packing several quantities into the fogcoord vector */
-      { "gl_FogFragCoord", FRAG_ATTRIB_FOGC, SWIZZLE_XXXX },
-      { "gl_FrontFacing", FRAG_ATTRIB_FOGC, SWIZZLE_YYYY }, /*XXX*/
-      { "gl_PointCoord", FRAG_ATTRIB_FOGC, SWIZZLE_ZWWW },
-      { NULL, 0, SWIZZLE_NOOP }
-   };
-   GLuint i;
-   const struct input_info *inputs
-      = (target == GL_VERTEX_PROGRAM_ARB) ? vertInputs : fragInputs;
-
-   ASSERT(MAX_TEXTURE_COORD_UNITS == 8); /* if this fails, fix vertInputs above */
-
-   for (i = 0; inputs[i].Name; i++) {
-      if (strcmp(inputs[i].Name, name) == 0) {
-         /* found */
-         *swizzleOut = inputs[i].Swizzle;
-         return inputs[i].Attrib;
-      }
-   }
-   return -1;
-}
-
-
-/**
- * Return the VERT_RESULT_* or FRAG_RESULT_* value that corresponds to
- * a vertex or fragment program output variable.  Return -1 for an invalid
- * output name.
- */
-static GLint
-_slang_output_index(const char *name, GLenum target)
-{
-   struct output_info {
-      const char *Name;
-      GLuint Attrib;
-   };
-   static const struct output_info vertOutputs[] = {
-      { "gl_Position", VERT_RESULT_HPOS },
-      { "gl_FrontColor", VERT_RESULT_COL0 },
-      { "gl_BackColor", VERT_RESULT_BFC0 },
-      { "gl_FrontSecondaryColor", VERT_RESULT_COL1 },
-      { "gl_BackSecondaryColor", VERT_RESULT_BFC1 },
-      { "gl_TexCoord", VERT_RESULT_TEX0 },
-      { "gl_FogFragCoord", VERT_RESULT_FOGC },
-      { "gl_PointSize", VERT_RESULT_PSIZ },
-      { NULL, 0 }
-   };
-   static const struct output_info fragOutputs[] = {
-      { "gl_FragColor", FRAG_RESULT_COLOR },
-      { "gl_FragDepth", FRAG_RESULT_DEPTH },
-      { "gl_FragData", FRAG_RESULT_DATA0 },
-      { NULL, 0 }
-   };
-   GLuint i;
-   const struct output_info *outputs
-      = (target == GL_VERTEX_PROGRAM_ARB) ? vertOutputs : fragOutputs;
-
-   for (i = 0; outputs[i].Name; i++) {
-      if (strcmp(outputs[i].Name, name) == 0) {
-         /* found */
-         return outputs[i].Attrib;
-      }
-   }
-   return -1;
-}
-
-
 /**
  * Called when we begin code/IR generation for a new while/do/for loop.
  */
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index 3617ce1fb6..8f2b40d5df 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -40,6 +40,7 @@
 #include "shader/prog_statevars.h"
 #include "shader/prog_uniform.h"
 #include "shader/shader_api.h"
+#include "slang_builtin.h"
 #include "slang_link.h"
 
 
@@ -287,6 +288,7 @@ link_uniform_vars(GLcontext *ctx,
    for (i = 0; i < prog->NumInstructions; i++) {
       struct prog_instruction *inst = prog->Instructions + i;
       if (_mesa_is_tex_instruction(inst->Opcode)) {
+         /* here, inst->TexSrcUnit is really the sampler unit */
          const GLint oldSampNum = inst->TexSrcUnit;
 
 #if 0
@@ -294,7 +296,6 @@ link_uniform_vars(GLcontext *ctx,
                 inst->TexSrcUnit, samplerMap[ inst->TexSrcUnit ]);
 #endif
 
-         /* here, texUnit is really samplerUnit */
          if (oldSampNum < Elements(samplerMap)) {
             const GLuint newSampNum = samplerMap[oldSampNum];
             inst->TexSrcUnit = newSampNum;
@@ -327,6 +328,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
    GLint attribMap[MAX_VERTEX_GENERIC_ATTRIBS];
    GLuint i, j;
    GLbitfield usedAttributes; /* generics only, not legacy attributes */
+   GLbitfield inputsRead = 0x0;
 
    assert(origProg != linkedProg);
    assert(origProg->Target == GL_VERTEX_PROGRAM_ARB);
@@ -370,6 +372,10 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
    for (i = 0; i < linkedProg->NumInstructions; i++) {
       struct prog_instruction *inst = linkedProg->Instructions + i;
       for (j = 0; j < 3; j++) {
+         if (inst->SrcReg[j].File == PROGRAM_INPUT) {
+            inputsRead |= (1 << inst->SrcReg[j].Index);
+         }
+
          if (inst->SrcReg[j].File == PROGRAM_INPUT &&
              inst->SrcReg[j].Index >= VERT_ATTRIB_GENERIC0) {
             /*
@@ -431,6 +437,20 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
       }
    }
 
+   /* Handle pre-defined attributes here (gl_Vertex, gl_Normal, etc).
+    * When the user queries the active attributes we need to include both
+    * the user-defined attributes and the built-in ones.
+    */
+   for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_GENERIC0; i++) {
+      if (inputsRead & (1 << i)) {
+         _mesa_add_attribute(linkedProg->Attributes,
+                             _slang_vert_attrib_name(i),
+                             4, /* size in floats */
+                             _slang_vert_attrib_type(i),
+                             -1 /* attrib/input */);
+      }
+   }
+
    return GL_TRUE;
 }
 
@@ -484,20 +504,6 @@ _slang_update_inputs_outputs(struct gl_program *prog)
       for (j = 0; j < numSrc; j++) {
          if (inst->SrcReg[j].File == PROGRAM_INPUT) {
             prog->InputsRead |= 1 << inst->SrcReg[j].Index;
-            if (prog->Target == GL_FRAGMENT_PROGRAM_ARB &&
-                inst->SrcReg[j].Index == FRAG_ATTRIB_FOGC) {
-               /* The fragment shader FOGC input is used for fog,
-                * front-facing and sprite/point coord.
-                */
-               struct gl_fragment_program *fp = fragment_program(prog);
-               const GLint swz = GET_SWZ(inst->SrcReg[j].Swizzle, 0);
-               if (swz == SWIZZLE_X)
-                  fp->UsesFogFragCoord = GL_TRUE;
-               else if (swz == SWIZZLE_Y)
-                  fp->UsesFrontFacing = GL_TRUE;
-               else if (swz == SWIZZLE_Z || swz == SWIZZLE_W)
-                  fp->UsesPointCoord = GL_TRUE;
-            }
          }
          else if (inst->SrcReg[j].File == PROGRAM_ADDRESS) {
             maxAddrReg = MAX2(maxAddrReg, (GLuint) (inst->SrcReg[j].Index + 1));
@@ -745,6 +751,8 @@ _slang_link(GLcontext *ctx,
       struct gl_vertex_program *linked_vprog =
          vertex_program(_mesa_clone_program(ctx, &vertProg->Base));
       shProg->VertexProgram = linked_vprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->VertexProgram->Base.Id = shProg->Name;
       ASSERT(shProg->VertexProgram->Base.RefCount == 1);
    }
 
@@ -753,6 +761,8 @@ _slang_link(GLcontext *ctx,
       struct gl_fragment_program *linked_fprog = 
          fragment_program(_mesa_clone_program(ctx, &fragProg->Base));
       shProg->FragmentProgram = linked_fprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->FragmentProgram->Base.Id = shProg->Name;
       ASSERT(shProg->FragmentProgram->Base.RefCount == 1);
    }
 
@@ -867,6 +877,14 @@ _slang_link(GLcontext *ctx,
       }
    }
 
+   /* Debug: */
+   if (0) {
+      if (shProg->VertexProgram)
+         _mesa_postprocess_program(ctx, &shProg->VertexProgram->Base);
+      if (shProg->FragmentProgram)
+         _mesa_postprocess_program(ctx, &shProg->FragmentProgram->Base);
+   }
+
    if (ctx->Shader.Flags & GLSL_DUMP) {
       _mesa_printf("Varying vars:\n");
       _mesa_print_parameter_list(shProg->Varying);
diff --git a/src/mesa/shader/symbol_table.c b/src/mesa/shader/symbol_table.c
new file mode 100644
index 0000000000..7a9aa7b8f6
--- /dev/null
+++ b/src/mesa/shader/symbol_table.c
@@ -0,0 +1,350 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+#include <assert.h>
+
+#include "symbol_table.h"
+#include "hash_table.h"
+
+struct symbol {
+    /**
+     * Link to the next symbol in the table with the same name
+     *
+     * The linked list of symbols with the same name is ordered by scope
+     * from inner-most to outer-most.
+     */
+    struct symbol *next_with_same_name;
+
+
+    /**
+     * Link to the next symbol in the table with the same scope
+     *
+     * The linked list of symbols with the same scope is unordered.  Symbols
+     * in this list my have unique names.
+     */
+    struct symbol *next_with_same_scope;
+
+
+    /**
+     * Header information for the list of symbols with the same name.
+     */
+    struct symbol_header *hdr;
+
+
+    /**
+     * Name space of the symbol
+     *
+     * Name space are arbitrary user assigned integers.  No two symbols can
+     * exist in the same name space at the same scope level.
+     */
+    int name_space;
+
+    
+    /**
+     * Arbitrary user supplied data.
+     */
+    void *data;
+};
+
+
+/**
+ */
+struct symbol_header {
+    /** Symbol name. */
+    const char *name;
+
+    /** Linked list of symbols with the same name. */
+    struct symbol *symbols;
+};
+
+
+/**
+ * Element of the scope stack.
+ */
+struct scope_level {
+    /** Link to next (inner) scope level. */
+    struct scope_level *next;
+    
+    /** Linked list of symbols with the same scope. */
+    struct symbol *symbols;
+};
+
+
+/**
+ *
+ */
+struct _mesa_symbol_table {
+    /** Hash table containing all symbols in the symbol table. */
+    struct hash_table *ht;
+
+    /** Top of scope stack. */
+    struct scope_level *current_scope;
+};
+
+
+struct _mesa_symbol_table_iterator {
+    /**
+     * Name space of symbols returned by this iterator.
+     */
+    int name_space;
+
+
+    /**
+     * Currently iterated symbol
+     *
+     * The next call to \c _mesa_symbol_table_iterator_get will return this
+     * value.  It will also update this value to the value that should be
+     * returned by the next call.
+     */
+    struct symbol *curr;
+};
+
+
+static void
+check_symbol_table(struct _mesa_symbol_table *table)
+{
+#if 1
+    struct scope_level *scope;
+
+    for (scope = table->current_scope; scope != NULL; scope = scope->next) {
+        struct symbol *sym;
+
+        for (sym = scope->symbols
+             ; sym != NULL
+             ; sym = sym->next_with_same_name) {
+            const struct symbol_header *const hdr = sym->hdr;
+            struct symbol *sym2;
+
+            for (sym2 = hdr->symbols
+                 ; sym2 != NULL
+                 ; sym2 = sym2->next_with_same_name) {
+                assert(sym2->hdr == hdr);
+            }
+        }
+    }
+#endif
+}
+
+void
+_mesa_symbol_table_pop_scope(struct _mesa_symbol_table *table)
+{
+    struct scope_level *const scope = table->current_scope;
+    struct symbol *sym = scope->symbols;
+
+    table->current_scope = scope->next;
+
+    free(scope);
+
+    while (sym != NULL) {
+        struct symbol *const next = sym->next_with_same_scope;
+        struct symbol_header *const hdr = sym->hdr;
+
+        assert(hdr->symbols == sym);
+
+        hdr->symbols = sym->next_with_same_name;
+
+        free(sym);
+
+        sym = next;
+    }
+
+    check_symbol_table(table);
+}
+
+
+void
+_mesa_symbol_table_push_scope(struct _mesa_symbol_table *table)
+{
+    struct scope_level *const scope = calloc(1, sizeof(*scope));
+    
+    scope->next = table->current_scope;
+    table->current_scope = scope;
+}
+
+
+static struct symbol_header *
+find_symbol(struct _mesa_symbol_table *table, const char *name)
+{
+    return (struct symbol_header *) hash_table_find(table->ht, name);
+}
+
+
+struct _mesa_symbol_table_iterator *
+_mesa_symbol_table_iterator_ctor(struct _mesa_symbol_table *table,
+                                 int name_space, const char *name)
+{
+    struct _mesa_symbol_table_iterator *iter = calloc(1, sizeof(*iter));
+    struct symbol_header *const hdr = find_symbol(table, name);
+    
+    iter->name_space = name_space;
+
+    if (hdr != NULL) {
+        struct symbol *sym;
+
+        for (sym = hdr->symbols; sym != NULL; sym = sym->next_with_same_name) {
+            assert(sym->hdr == hdr);
+
+            if ((name_space == -1) || (sym->name_space == name_space)) {
+                iter->curr = sym;
+                break;
+            }
+        }
+    }
+
+    return iter;
+}
+
+
+void
+_mesa_symbol_table_iterator_dtor(struct _mesa_symbol_table_iterator *iter)
+{
+    free(iter);
+}
+
+
+void *
+_mesa_symbol_table_iterator_get(struct _mesa_symbol_table_iterator *iter)
+{
+    return (iter->curr == NULL) ? NULL : iter->curr->data;
+}
+
+
+int
+_mesa_symbol_table_iterator_next(struct _mesa_symbol_table_iterator *iter)
+{
+    struct symbol_header *hdr;
+
+    if (iter->curr == NULL) {
+        return 0;
+    }
+
+    hdr = iter->curr->hdr;
+    iter->curr = iter->curr->next_with_same_name;
+
+    while (iter->curr != NULL) {
+        assert(iter->curr->hdr == hdr);
+
+        if ((iter->name_space == -1)
+            || (iter->curr->name_space == iter->name_space)) {
+            return 1;
+        }
+
+        iter->curr = iter->curr->next_with_same_name;
+    }
+
+    return 0;
+}
+
+
+void *
+_mesa_symbol_table_find_symbol(struct _mesa_symbol_table *table,
+                               int name_space, const char *name)
+{
+    struct symbol_header *const hdr = find_symbol(table, name);
+
+    if (hdr != NULL) {
+        struct symbol *sym;
+
+
+        for (sym = hdr->symbols; sym != NULL; sym = sym->next_with_same_name) {
+            assert(sym->hdr == hdr);
+
+            if ((name_space == -1) || (sym->name_space == name_space)) {
+                return sym->data;
+            }
+        }
+    }
+
+    return NULL;
+}
+
+
+int
+_mesa_symbol_table_add_symbol(struct _mesa_symbol_table *table,
+                              int name_space, const char *name,
+                              void *declaration)
+{
+    struct symbol_header *hdr;
+    struct symbol *sym;
+
+    check_symbol_table(table);
+
+    hdr = find_symbol(table, name);
+
+    check_symbol_table(table);
+
+    if (hdr == NULL) {
+        hdr = calloc(1, sizeof(*hdr));
+        hdr->name = name;
+
+        hash_table_insert(table->ht, hdr, name);
+    }
+
+    check_symbol_table(table);
+
+    sym = calloc(1, sizeof(*sym));
+    sym->next_with_same_name = hdr->symbols;
+    sym->next_with_same_scope = table->current_scope->symbols;
+    sym->hdr = hdr;
+    sym->name_space = name_space;
+    sym->data = declaration;
+
+    assert(sym->hdr == hdr);
+
+    hdr->symbols = sym;
+    table->current_scope->symbols = sym;
+
+    check_symbol_table(table);
+    return 0;
+}
+
+
+struct _mesa_symbol_table *
+_mesa_symbol_table_ctor(void)
+{
+    struct _mesa_symbol_table *table = calloc(1, sizeof(*table));
+
+    if (table != NULL) {
+       table->ht = hash_table_ctor(32, hash_table_string_hash,
+				   hash_table_string_compare);
+
+       _mesa_symbol_table_push_scope(table);
+    }
+
+    return table;
+}
+
+
+void
+_mesa_symbol_table_dtor(struct _mesa_symbol_table *table)
+{
+   while (table->current_scope != NULL) {
+      _mesa_symbol_table_pop_scope(table);
+   }
+
+   hash_table_dtor(table->ht);
+   free(table);
+}
diff --git a/src/mesa/shader/symbol_table.h b/src/mesa/shader/symbol_table.h
new file mode 100644
index 0000000000..0c054ef139
--- /dev/null
+++ b/src/mesa/shader/symbol_table.h
@@ -0,0 +1,55 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#ifndef MESA_SYMBOL_TABLE_H
+#define MESA_SYMBOL_TABLE_H
+
+struct _mesa_symbol_table;
+struct _mesa_symbol_table_iterator;
+
+extern void _mesa_symbol_table_push_scope(struct _mesa_symbol_table *table);
+
+extern void _mesa_symbol_table_pop_scope(struct _mesa_symbol_table *table);
+
+extern int _mesa_symbol_table_add_symbol(struct _mesa_symbol_table *symtab,
+    int name_space, const char *name, void *declaration);
+
+extern void *_mesa_symbol_table_find_symbol(
+    struct _mesa_symbol_table *symtab, int name_space, const char *name);
+
+extern struct _mesa_symbol_table *_mesa_symbol_table_ctor(void);
+
+extern void _mesa_symbol_table_dtor(struct _mesa_symbol_table *);
+
+extern struct _mesa_symbol_table_iterator *_mesa_symbol_table_iterator_ctor(
+    struct _mesa_symbol_table *table, int name_space, const char *name);
+
+extern void _mesa_symbol_table_iterator_dtor(
+    struct _mesa_symbol_table_iterator *);
+
+extern void *_mesa_symbol_table_iterator_get(
+    struct _mesa_symbol_table_iterator *iter);
+
+extern int _mesa_symbol_table_iterator_next(
+    struct _mesa_symbol_table_iterator *iter);
+
+#endif /* MESA_SYMBOL_TABLE_H */
diff --git a/src/mesa/sources.mak b/src/mesa/sources.mak
index 78ca60f3ca..fa2a6307a4 100644
--- a/src/mesa/sources.mak
+++ b/src/mesa/sources.mak
@@ -61,6 +61,7 @@ MAIN_SOURCES = \
 	main/shared.c \
 	main/state.c \
 	main/stencil.c \
+	main/syncobj.c \
 	main/texcompress.c \
 	main/texcompress_s3tc.c \
 	main/texcompress_fxt1.c \
@@ -218,21 +219,27 @@ SHADER_SOURCES = \
 	shader/arbprogram.c \
 	shader/atifragshader.c \
 	shader/grammar/grammar_mesa.c \
+	shader/hash_table.c \
+	shader/lex.yy.c \
 	shader/nvfragparse.c \
 	shader/nvprogram.c \
 	shader/nvvertparse.c \
 	shader/program.c \
+	shader/program_parse.tab.c \
+	shader/program_parse_extra.c \
 	shader/prog_cache.c \
 	shader/prog_execute.c \
 	shader/prog_instruction.c \
 	shader/prog_noise.c \
 	shader/prog_optimize.c \
 	shader/prog_parameter.c \
+	shader/prog_parameter_layout.c \
 	shader/prog_print.c \
 	shader/prog_statevars.c \
 	shader/prog_uniform.c \
 	shader/programopt.c \
-	shader/shader_api.c \
+	shader/symbol_table.c \
+	shader/shader_api.c
 
 SLANG_SOURCES =	\
 	shader/slang/slang_builtin.c	\
@@ -303,7 +310,8 @@ SPARC_API =			\
 	sparc/glapi_sparc.S
 
 COMMON_DRIVER_SOURCES =			\
-	drivers/common/driverfuncs.c
+	drivers/common/driverfuncs.c	\
+	drivers/common/meta.c
 
 
 # Sources for building non-Gallium drivers
@@ -327,6 +335,7 @@ MESA_GALLIUM_SOURCES = \
 	$(VBO_SOURCES)		\
 	$(STATETRACKER_SOURCES)	\
 	$(SHADER_SOURCES)	\
+	ppc/common_ppc.c	\
 	x86/common_x86.c	\
 	$(SLANG_SOURCES)
 
@@ -363,5 +372,4 @@ INCLUDE_DIRS = \
 	-I$(TOP)/include \
 	-I$(TOP)/src/mesa \
 	-I$(TOP)/src/gallium/include \
-	-I$(TOP)/src/gallium/drivers \
 	-I$(TOP)/src/gallium/auxiliary
diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S
index 493e094cde..5b2c9e4a9a 100644
--- a/src/mesa/sparc/glapi_sparc.S
+++ b/src/mesa/sparc/glapi_sparc.S
@@ -759,23 +759,35 @@ gl_dispatch_functions_start:
 	GL_STUB(glGetAttribLocationARB, _gloffset_GetAttribLocationARB)
 	GL_STUB(glDrawBuffersARB, _gloffset_DrawBuffersARB)
 	GL_STUB(glRenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample)
+	GL_STUB(glFlushMappedBufferRange, _gloffset_FlushMappedBufferRange)
+	GL_STUB(glMapBufferRange, _gloffset_MapBufferRange)
+	GL_STUB(glBindVertexArray, _gloffset_BindVertexArray)
+	GL_STUB(glGenVertexArrays, _gloffset_GenVertexArrays)
+	GL_STUB(glCopyBufferSubData, _gloffset_CopyBufferSubData)
+	GL_STUB(glClientWaitSync, _gloffset_ClientWaitSync)
+	GL_STUB(glDeleteSync, _gloffset_DeleteSync)
+	GL_STUB(glFenceSync, _gloffset_FenceSync)
+	GL_STUB(glGetInteger64v, _gloffset_GetInteger64v)
+	GL_STUB(glGetSynciv, _gloffset_GetSynciv)
+	GL_STUB(glIsSync, _gloffset_IsSync)
+	GL_STUB(glWaitSync, _gloffset_WaitSync)
 	GL_STUB(glPolygonOffsetEXT, _gloffset_PolygonOffsetEXT)
-	GL_STUB(gl_dispatch_stub_563, _gloffset_GetPixelTexGenParameterfvSGIS)
-	HIDDEN(gl_dispatch_stub_563)
-	GL_STUB(gl_dispatch_stub_564, _gloffset_GetPixelTexGenParameterivSGIS)
-	HIDDEN(gl_dispatch_stub_564)
-	GL_STUB(gl_dispatch_stub_565, _gloffset_PixelTexGenParameterfSGIS)
-	HIDDEN(gl_dispatch_stub_565)
-	GL_STUB(gl_dispatch_stub_566, _gloffset_PixelTexGenParameterfvSGIS)
-	HIDDEN(gl_dispatch_stub_566)
-	GL_STUB(gl_dispatch_stub_567, _gloffset_PixelTexGenParameteriSGIS)
-	HIDDEN(gl_dispatch_stub_567)
-	GL_STUB(gl_dispatch_stub_568, _gloffset_PixelTexGenParameterivSGIS)
-	HIDDEN(gl_dispatch_stub_568)
-	GL_STUB(gl_dispatch_stub_569, _gloffset_SampleMaskSGIS)
-	HIDDEN(gl_dispatch_stub_569)
-	GL_STUB(gl_dispatch_stub_570, _gloffset_SamplePatternSGIS)
-	HIDDEN(gl_dispatch_stub_570)
+	GL_STUB(gl_dispatch_stub_575, _gloffset_GetPixelTexGenParameterfvSGIS)
+	HIDDEN(gl_dispatch_stub_575)
+	GL_STUB(gl_dispatch_stub_576, _gloffset_GetPixelTexGenParameterivSGIS)
+	HIDDEN(gl_dispatch_stub_576)
+	GL_STUB(gl_dispatch_stub_577, _gloffset_PixelTexGenParameterfSGIS)
+	HIDDEN(gl_dispatch_stub_577)
+	GL_STUB(gl_dispatch_stub_578, _gloffset_PixelTexGenParameterfvSGIS)
+	HIDDEN(gl_dispatch_stub_578)
+	GL_STUB(gl_dispatch_stub_579, _gloffset_PixelTexGenParameteriSGIS)
+	HIDDEN(gl_dispatch_stub_579)
+	GL_STUB(gl_dispatch_stub_580, _gloffset_PixelTexGenParameterivSGIS)
+	HIDDEN(gl_dispatch_stub_580)
+	GL_STUB(gl_dispatch_stub_581, _gloffset_SampleMaskSGIS)
+	HIDDEN(gl_dispatch_stub_581)
+	GL_STUB(gl_dispatch_stub_582, _gloffset_SamplePatternSGIS)
+	HIDDEN(gl_dispatch_stub_582)
 	GL_STUB(glColorPointerEXT, _gloffset_ColorPointerEXT)
 	GL_STUB(glEdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT)
 	GL_STUB(glIndexPointerEXT, _gloffset_IndexPointerEXT)
@@ -786,10 +798,10 @@ gl_dispatch_functions_start:
 	GL_STUB(glPointParameterfvEXT, _gloffset_PointParameterfvEXT)
 	GL_STUB(glLockArraysEXT, _gloffset_LockArraysEXT)
 	GL_STUB(glUnlockArraysEXT, _gloffset_UnlockArraysEXT)
-	GL_STUB(gl_dispatch_stub_581, _gloffset_CullParameterdvEXT)
-	HIDDEN(gl_dispatch_stub_581)
-	GL_STUB(gl_dispatch_stub_582, _gloffset_CullParameterfvEXT)
-	HIDDEN(gl_dispatch_stub_582)
+	GL_STUB(gl_dispatch_stub_593, _gloffset_CullParameterdvEXT)
+	HIDDEN(gl_dispatch_stub_593)
+	GL_STUB(gl_dispatch_stub_594, _gloffset_CullParameterfvEXT)
+	HIDDEN(gl_dispatch_stub_594)
 	GL_STUB(glSecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT)
 	GL_STUB(glSecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT)
 	GL_STUB(glSecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT)
@@ -814,8 +826,8 @@ gl_dispatch_functions_start:
 	GL_STUB(glFogCoorddvEXT, _gloffset_FogCoorddvEXT)
 	GL_STUB(glFogCoordfEXT, _gloffset_FogCoordfEXT)
 	GL_STUB(glFogCoordfvEXT, _gloffset_FogCoordfvEXT)
-	GL_STUB(gl_dispatch_stub_607, _gloffset_PixelTexGenSGIX)
-	HIDDEN(gl_dispatch_stub_607)
+	GL_STUB(gl_dispatch_stub_619, _gloffset_PixelTexGenSGIX)
+	HIDDEN(gl_dispatch_stub_619)
 	GL_STUB(glBlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT)
 	GL_STUB(glFlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV)
 	GL_STUB(glVertexArrayRangeNV, _gloffset_VertexArrayRangeNV)
@@ -857,24 +869,24 @@ gl_dispatch_functions_start:
 	GL_STUB(glWindowPos4ivMESA, _gloffset_WindowPos4ivMESA)
 	GL_STUB(glWindowPos4sMESA, _gloffset_WindowPos4sMESA)
 	GL_STUB(glWindowPos4svMESA, _gloffset_WindowPos4svMESA)
-	GL_STUB(gl_dispatch_stub_649, _gloffset_MultiModeDrawArraysIBM)
-	HIDDEN(gl_dispatch_stub_649)
-	GL_STUB(gl_dispatch_stub_650, _gloffset_MultiModeDrawElementsIBM)
-	HIDDEN(gl_dispatch_stub_650)
-	GL_STUB(gl_dispatch_stub_651, _gloffset_DeleteFencesNV)
-	HIDDEN(gl_dispatch_stub_651)
-	GL_STUB(gl_dispatch_stub_652, _gloffset_FinishFenceNV)
-	HIDDEN(gl_dispatch_stub_652)
-	GL_STUB(gl_dispatch_stub_653, _gloffset_GenFencesNV)
-	HIDDEN(gl_dispatch_stub_653)
-	GL_STUB(gl_dispatch_stub_654, _gloffset_GetFenceivNV)
-	HIDDEN(gl_dispatch_stub_654)
-	GL_STUB(gl_dispatch_stub_655, _gloffset_IsFenceNV)
-	HIDDEN(gl_dispatch_stub_655)
-	GL_STUB(gl_dispatch_stub_656, _gloffset_SetFenceNV)
-	HIDDEN(gl_dispatch_stub_656)
-	GL_STUB(gl_dispatch_stub_657, _gloffset_TestFenceNV)
-	HIDDEN(gl_dispatch_stub_657)
+	GL_STUB(gl_dispatch_stub_661, _gloffset_MultiModeDrawArraysIBM)
+	HIDDEN(gl_dispatch_stub_661)
+	GL_STUB(gl_dispatch_stub_662, _gloffset_MultiModeDrawElementsIBM)
+	HIDDEN(gl_dispatch_stub_662)
+	GL_STUB(gl_dispatch_stub_663, _gloffset_DeleteFencesNV)
+	HIDDEN(gl_dispatch_stub_663)
+	GL_STUB(gl_dispatch_stub_664, _gloffset_FinishFenceNV)
+	HIDDEN(gl_dispatch_stub_664)
+	GL_STUB(gl_dispatch_stub_665, _gloffset_GenFencesNV)
+	HIDDEN(gl_dispatch_stub_665)
+	GL_STUB(gl_dispatch_stub_666, _gloffset_GetFenceivNV)
+	HIDDEN(gl_dispatch_stub_666)
+	GL_STUB(gl_dispatch_stub_667, _gloffset_IsFenceNV)
+	HIDDEN(gl_dispatch_stub_667)
+	GL_STUB(gl_dispatch_stub_668, _gloffset_SetFenceNV)
+	HIDDEN(gl_dispatch_stub_668)
+	GL_STUB(gl_dispatch_stub_669, _gloffset_TestFenceNV)
+	HIDDEN(gl_dispatch_stub_669)
 	GL_STUB(glAreProgramsResidentNV, _gloffset_AreProgramsResidentNV)
 	GL_STUB(glBindProgramNV, _gloffset_BindProgramNV)
 	GL_STUB(glDeleteProgramsNV, _gloffset_DeleteProgramsNV)
@@ -955,26 +967,26 @@ gl_dispatch_functions_start:
 	GL_STUB(glSetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI)
 	GL_STUB(glPointParameteriNV, _gloffset_PointParameteriNV)
 	GL_STUB(glPointParameterivNV, _gloffset_PointParameterivNV)
-	GL_STUB(gl_dispatch_stub_738, _gloffset_ActiveStencilFaceEXT)
-	HIDDEN(gl_dispatch_stub_738)
-	GL_STUB(gl_dispatch_stub_739, _gloffset_BindVertexArrayAPPLE)
-	HIDDEN(gl_dispatch_stub_739)
-	GL_STUB(gl_dispatch_stub_740, _gloffset_DeleteVertexArraysAPPLE)
-	HIDDEN(gl_dispatch_stub_740)
-	GL_STUB(gl_dispatch_stub_741, _gloffset_GenVertexArraysAPPLE)
-	HIDDEN(gl_dispatch_stub_741)
-	GL_STUB(gl_dispatch_stub_742, _gloffset_IsVertexArrayAPPLE)
-	HIDDEN(gl_dispatch_stub_742)
+	GL_STUB(gl_dispatch_stub_750, _gloffset_ActiveStencilFaceEXT)
+	HIDDEN(gl_dispatch_stub_750)
+	GL_STUB(gl_dispatch_stub_751, _gloffset_BindVertexArrayAPPLE)
+	HIDDEN(gl_dispatch_stub_751)
+	GL_STUB(gl_dispatch_stub_752, _gloffset_DeleteVertexArraysAPPLE)
+	HIDDEN(gl_dispatch_stub_752)
+	GL_STUB(gl_dispatch_stub_753, _gloffset_GenVertexArraysAPPLE)
+	HIDDEN(gl_dispatch_stub_753)
+	GL_STUB(gl_dispatch_stub_754, _gloffset_IsVertexArrayAPPLE)
+	HIDDEN(gl_dispatch_stub_754)
 	GL_STUB(glGetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV)
 	GL_STUB(glGetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV)
 	GL_STUB(glProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV)
 	GL_STUB(glProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV)
 	GL_STUB(glProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV)
 	GL_STUB(glProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV)
-	GL_STUB(gl_dispatch_stub_749, _gloffset_DepthBoundsEXT)
-	HIDDEN(gl_dispatch_stub_749)
-	GL_STUB(gl_dispatch_stub_750, _gloffset_BlendEquationSeparateEXT)
-	HIDDEN(gl_dispatch_stub_750)
+	GL_STUB(gl_dispatch_stub_761, _gloffset_DepthBoundsEXT)
+	HIDDEN(gl_dispatch_stub_761)
+	GL_STUB(gl_dispatch_stub_762, _gloffset_BlendEquationSeparateEXT)
+	HIDDEN(gl_dispatch_stub_762)
 	GL_STUB(glBindFramebufferEXT, _gloffset_BindFramebufferEXT)
 	GL_STUB(glBindRenderbufferEXT, _gloffset_BindRenderbufferEXT)
 	GL_STUB(glCheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT)
@@ -992,19 +1004,28 @@ gl_dispatch_functions_start:
 	GL_STUB(glIsFramebufferEXT, _gloffset_IsFramebufferEXT)
 	GL_STUB(glIsRenderbufferEXT, _gloffset_IsRenderbufferEXT)
 	GL_STUB(glRenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT)
-	GL_STUB(gl_dispatch_stub_768, _gloffset_BlitFramebufferEXT)
-	HIDDEN(gl_dispatch_stub_768)
+	GL_STUB(gl_dispatch_stub_780, _gloffset_BlitFramebufferEXT)
+	HIDDEN(gl_dispatch_stub_780)
+	GL_STUB(gl_dispatch_stub_781, _gloffset_BufferParameteriAPPLE)
+	HIDDEN(gl_dispatch_stub_781)
+	GL_STUB(gl_dispatch_stub_782, _gloffset_FlushMappedBufferRangeAPPLE)
+	HIDDEN(gl_dispatch_stub_782)
 	GL_STUB(glFramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT)
-	GL_STUB(gl_dispatch_stub_770, _gloffset_StencilFuncSeparateATI)
-	HIDDEN(gl_dispatch_stub_770)
-	GL_STUB(gl_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT)
-	HIDDEN(gl_dispatch_stub_771)
-	GL_STUB(gl_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT)
-	HIDDEN(gl_dispatch_stub_772)
-	GL_STUB(gl_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT)
-	HIDDEN(gl_dispatch_stub_773)
-	GL_STUB(gl_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT)
-	HIDDEN(gl_dispatch_stub_774)
+	GL_STUB(glProvokingVertexEXT, _gloffset_ProvokingVertexEXT)
+	GL_STUB(gl_dispatch_stub_785, _gloffset_GetTexParameterPointervAPPLE)
+	HIDDEN(gl_dispatch_stub_785)
+	GL_STUB(gl_dispatch_stub_786, _gloffset_TextureRangeAPPLE)
+	HIDDEN(gl_dispatch_stub_786)
+	GL_STUB(gl_dispatch_stub_787, _gloffset_StencilFuncSeparateATI)
+	HIDDEN(gl_dispatch_stub_787)
+	GL_STUB(gl_dispatch_stub_788, _gloffset_ProgramEnvParameters4fvEXT)
+	HIDDEN(gl_dispatch_stub_788)
+	GL_STUB(gl_dispatch_stub_789, _gloffset_ProgramLocalParameters4fvEXT)
+	HIDDEN(gl_dispatch_stub_789)
+	GL_STUB(gl_dispatch_stub_790, _gloffset_GetQueryObjecti64vEXT)
+	HIDDEN(gl_dispatch_stub_790)
+	GL_STUB(gl_dispatch_stub_791, _gloffset_GetQueryObjectui64vEXT)
+	HIDDEN(gl_dispatch_stub_791)
 	GL_STUB_ALIAS(glArrayElementEXT, glArrayElement)
 	GL_STUB_ALIAS(glBindTextureEXT, glBindTexture)
 	GL_STUB_ALIAS(glDrawArraysEXT, glDrawArrays)
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 3ba7b26928..5d4d8eee02 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -24,11 +24,12 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * 
  **************************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <keith@tungstengraphics.com>
-  *   Brian Paul
-  */
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *   Brian Paul
+ */
 
 #include "main/imports.h"
 #include "shader/prog_parameter.h"
@@ -44,19 +45,21 @@
 #include "st_program.h"
 #include "st_inlines.h"
 
+
 /**
  * Pass the given program parameters to the graphics pipe as a
  * constant buffer.
- * \param id  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
+ * \param shader_type  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
  */
 void st_upload_constants( struct st_context *st,
                           struct gl_program_parameter_list *params,
-                          unsigned id)
+                          unsigned shader_type)
 {
    struct pipe_context *pipe = st->pipe;
-   struct pipe_constant_buffer *cbuf = &st->state.constants[id];
+   struct pipe_constant_buffer *cbuf = &st->state.constants[shader_type];
 
-   assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT);
+   assert(shader_type == PIPE_SHADER_VERTEX ||
+          shader_type == PIPE_SHADER_FRAGMENT);
 
    /* update constants */
    if (params && params->NumParameters) {
@@ -68,13 +71,14 @@ void st_upload_constants( struct st_context *st,
        * avoid gratuitous rendering synchronization.
        */
       pipe_buffer_reference(&cbuf->buffer, NULL );
-      cbuf->buffer = pipe_buffer_create(pipe->screen, 16, PIPE_BUFFER_USAGE_CONSTANT,
+      cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                        PIPE_BUFFER_USAGE_CONSTANT,
 					paramBytes );
 
-      if (0)
-      {
-	 printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
-                __FUNCTION__, id, params->NumParameters, params->StateFlags);
+      if (0) {
+	 debug_printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
+                      __FUNCTION__, shader_type, params->NumParameters,
+                      params->StateFlags);
          _mesa_print_parameter_list(params);
       }
 
@@ -84,15 +88,16 @@ void st_upload_constants( struct st_context *st,
 				       0, paramBytes,
 				       params->ParameterValues);
 
-      st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf);
+      st->pipe->set_constant_buffer(st->pipe, shader_type, 0, cbuf);
    }
    else {
-      st->constants.tracked_state[id].dirty.mesa = 0;
-      //  st->pipe->set_constant_buffer(st->pipe, id, 0, NULL);
+      st->constants.tracked_state[shader_type].dirty.mesa = 0x0;
    }
 }
 
-/* Vertex shader:
+
+/**
+ * Vertex shader:
  */
 static void update_vs_constants(struct st_context *st )
 {
@@ -102,6 +107,7 @@ static void update_vs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_VERTEX );
 }
 
+
 const struct st_tracked_state st_update_vs_constants = {
    "st_update_vs_constants",				/* name */
    {							/* dirty */
@@ -111,7 +117,10 @@ const struct st_tracked_state st_update_vs_constants = {
    update_vs_constants					/* update */
 };
 
-/* Fragment shader:
+
+
+/**
+ * Fragment shader:
  */
 static void update_fs_constants(struct st_context *st )
 {
@@ -121,6 +130,7 @@ static void update_fs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_FRAGMENT );
 }
 
+
 const struct st_tracked_state st_update_fs_constants = {
    "st_update_fs_constants",				/* name */
    {							/* dirty */
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index 3ef919a45e..5209a6a0c9 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -122,6 +122,7 @@ update_framebuffer_state( struct st_context *st )
                                    strb->surface);
             framebuffer->nr_cbufs++;
          }
+         strb->defined = GL_TRUE; /* we'll be drawing something */
       }
    }
    for (i = framebuffer->nr_cbufs; i < PIPE_MAX_COLOR_BUFS; i++) {
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
index 5c7206409c..36b28cb4df 100644
--- a/src/mesa/state_tracker/st_atom_rasterizer.c
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -102,6 +102,9 @@ static void update_raster_state( struct st_context *st )
    if (ctx->Light.ShadeModel == GL_FLAT)
       raster->flatshade = 1;
 
+   if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION_EXT)
+      raster->flatshade_first = 1;
+
    /* _NEW_LIGHT | _NEW_PROGRAM
     *
     * Back-face colors can come from traditional lighting (when
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 8b3bb5cc03..ee649be885 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -139,23 +139,6 @@ find_translated_vp(struct st_context *st,
          if (fragInputsRead & (1 << inAttr)) {
             stfp->input_to_slot[inAttr] = numIn;
             numIn++;
-            if (((1 << inAttr) & FRAG_BIT_FOGC)) {
-               /* leave placeholders for the
-                * extra registers we extract from fog */
-               if (stfp->Base.UsesFrontFacing) {
-                  if (!stfp->Base.UsesFogFragCoord)
-                     --stfp->input_to_slot[inAttr];
-                  else
-                     ++numIn;
-               }
-               if (stfp->Base.UsesPointCoord) {
-                  if (!stfp->Base.UsesFrontFacing &&
-                      !stfp->Base.UsesFogFragCoord)
-                     stfp->input_to_slot[inAttr] -= 2;
-                  else
-                     ++numIn;
-               }
-            }
          }
          else {
             stfp->input_to_slot[inAttr] = UNUSED;
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index ccf972f805..902fb38d1a 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -241,7 +241,7 @@ combined_bitmap_fragment_program(GLcontext *ctx)
 /**
  * Copy user-provide bitmap bits into texture buffer, expanding
  * bits into texels.
- * "On" bits will set texels to 0xff.
+ * "On" bits will set texels to 0x0.
  * "Off" bits will not modify texels.
  * Note that the image is actually going to be upside down in
  * the texture.  We deal with that with texcoords.
@@ -253,63 +253,10 @@ unpack_bitmap(struct st_context *st,
               const GLubyte *bitmap,
               ubyte *destBuffer, uint destStride)
 {
-   GLint row, col;
-
-#define SET_PIXEL(COL, ROW) \
-   destBuffer[(py + (ROW)) * destStride + px + (COL)] = 0x0;
-
-   for (row = 0; row < height; row++) {
-      const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack,
-                 bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
-
-      if (unpack->LsbFirst) {
-         /* Lsb first */
-         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
-         for (col = 0; col < width; col++) {
-
-            if (*src & mask) {
-               SET_PIXEL(col, row);
-            }
-
-            if (mask == 128U) {
-               src++;
-               mask = 1U;
-            }
-            else {
-               mask = mask << 1;
-            }
-         }
-
-         /* get ready for next row */
-         if (mask != 1)
-            src++;
-      }
-      else {
-         /* Msb first */
-         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
-         for (col = 0; col < width; col++) {
-
-            if (*src & mask) {
-               SET_PIXEL(col, row);
-            }
-
-            if (mask == 1U) {
-               src++;
-               mask = 128U;
-            }
-            else {
-               mask = mask >> 1;
-            }
-         }
-
-         /* get ready for next row */
-         if (mask != 128)
-            src++;
-      }
-
-   } /* row */
+   destBuffer += py * destStride + px;
 
-#undef SET_PIXEL
+   _mesa_expand_bitmap(width, height, unpack, bitmap,
+                       destBuffer, destStride, 0x0);
 }
 
 
@@ -328,7 +275,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
    struct pipe_texture *pt;
 
    /* PBO source... */
-   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   bitmap = _mesa_map_pbo_source(ctx, unpack, bitmap);
    if (!bitmap) {
       return NULL;
    }
@@ -340,7 +287,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
                           0, width, height, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt) {
-      _mesa_unmap_bitmap_pbo(ctx, unpack);
+      _mesa_unmap_pbo_source(ctx, unpack);
       return NULL;
    }
 
@@ -355,7 +302,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
    unpack_bitmap(ctx->st, 0, 0, width, height, unpack, bitmap,
                  dest, transfer->stride);
 
-   _mesa_unmap_bitmap_pbo(ctx, unpack);
+   _mesa_unmap_pbo_source(ctx, unpack);
 
    /* Release transfer */
    screen->transfer_unmap(screen, transfer);
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
index a627c0374a..8e09d0b932 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.c
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -26,8 +26,14 @@
  **************************************************************************/
 
 
+/**
+ * Functions for pixel buffer objects and vertex/element buffer objects.
+ */
+
+
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "main/arrayobj.h"
 #include "main/bufferobj.h"
 
 #include "st_inlines.h"
@@ -39,14 +45,6 @@
 #include "pipe/p_inlines.h"
 
 
-
-/* Pixel buffers and Vertex/index buffers are handled through these
- * mesa callbacks.  Framebuffer/Renderbuffer objects are
- * created/managed elsewhere.
- */
-
-
-
 /**
  * There is some duplication between mesa's bufferobjects and our
  * bufmgr buffers.  Both have an integer handle and a hashtable to
@@ -100,8 +98,10 @@ st_bufferobj_subdata(GLcontext *ctx,
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
-   if (offset >= st_obj->size || size > (st_obj->size - offset))
-      return;
+   /* we may be called from VBO code, so double-check params here */
+   ASSERT(offset >= 0);
+   ASSERT(size >= 0);
+   ASSERT(offset + size <= obj->Size);
 
    st_cond_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer,
 				   offset, size, data);
@@ -120,8 +120,10 @@ st_bufferobj_get_subdata(GLcontext *ctx,
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
-   if (offset >= st_obj->size || size > (st_obj->size - offset))
-      return;
+   /* we may be called from VBO code, so double-check params here */
+   ASSERT(offset >= 0);
+   ASSERT(size >= 0);
+   ASSERT(offset + size <= obj->Size);
 
    st_cond_flush_pipe_buffer_read(st_context(ctx), st_obj->buffer,
 				  offset, size, data);
@@ -132,9 +134,10 @@ st_bufferobj_get_subdata(GLcontext *ctx,
  * Allocate space for and store data in a buffer object.  Any data that was
  * previously stored in the buffer object is lost.  If data is NULL,
  * memory will be allocated, but no copy will occur.
- * Called via glBufferDataARB().
+ * Called via ctx->Driver.BufferData().
+ * \return GL_TRUE for success, GL_FALSE if out of memory
  */
-static void
+static GLboolean
 st_bufferobj_data(GLcontext *ctx,
 		  GLenum target,
 		  GLsizeiptrARB size,
@@ -170,15 +173,13 @@ st_bufferobj_data(GLcontext *ctx,
    st_obj->buffer = pipe_buffer_create( pipe->screen, 32, buffer_usage, size );
 
    if (!st_obj->buffer) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBufferDataARB");
-      return;
+      return GL_FALSE;
    }
 
-   st_obj->size = size;
-
    if (data)
       st_no_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer, 0,
 				    size, data);
+   return GL_TRUE;
 }
 
 
@@ -190,7 +191,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
                  struct gl_buffer_object *obj)
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags;
+   uint flags;
 
    switch (access) {
    case GL_WRITE_ONLY:
@@ -209,7 +210,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
    obj->Pointer = st_cond_flush_pipe_buffer_map(st_context(ctx),
 						st_obj->buffer,
 						flags);
-   if(obj->Pointer) {
+   if (obj->Pointer) {
       obj->Offset = 0;
       obj->Length = obj->Size;
    }
@@ -217,7 +218,6 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
 }
 
 
-
 /**
  * Called via glMapBufferRange().
  */
@@ -228,8 +228,7 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
 {
    struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags = 0;
-   char *map;
+   uint flags = 0x0;
 
    if (access & GL_MAP_WRITE_BIT)
       flags |= PIPE_BUFFER_USAGE_CPU_WRITE;
@@ -251,14 +250,15 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
    assert(offset < obj->Size);
    assert(offset + length <= obj->Size);
 
-   map = obj->Pointer = pipe_buffer_map_range(pipe->screen, st_obj->buffer, offset, length, flags);
-   if(obj->Pointer) {
+   obj->Pointer = pipe_buffer_map_range(pipe->screen, st_obj->buffer, offset, length, flags);
+   if (obj->Pointer) {
+      obj->Pointer = (ubyte *) obj->Pointer + offset;
       obj->Offset = offset;
       obj->Length = length;
-      map += offset;
+      obj->AccessFlags = access;
    }
    
-   return map;
+   return obj->Pointer;
 }
 
 
@@ -297,6 +297,43 @@ st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
 }
 
 
+/**
+ * Called via glCopyBufferSubData().
+ */
+static void
+st_copy_buffer_subdata(GLcontext *ctx,
+                       struct gl_buffer_object *src,
+                       struct gl_buffer_object *dst,
+                       GLintptr readOffset, GLintptr writeOffset,
+                       GLsizeiptr size)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *srcObj = st_buffer_object(src);
+   struct st_buffer_object *dstObj = st_buffer_object(dst);
+   ubyte *srcPtr, *dstPtr;
+
+   /* buffer should not already be mapped */
+   assert(!src->Pointer);
+   assert(!dst->Pointer);
+
+   srcPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            srcObj->buffer,
+                                            readOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_READ);
+
+   dstPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            dstObj->buffer,
+                                            writeOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   if (srcPtr && dstPtr)
+      _mesa_memcpy(dstPtr + writeOffset, srcPtr + readOffset, size);
+
+   pipe_buffer_unmap(pipe->screen, srcObj->buffer);
+   pipe_buffer_unmap(pipe->screen, dstObj->buffer);
+}
+
+
 void
 st_init_bufferobject_functions(struct dd_function_table *functions)
 {
@@ -309,4 +346,9 @@ st_init_bufferobject_functions(struct dd_function_table *functions)
    functions->MapBufferRange = st_bufferobj_map_range;
    functions->FlushMappedBufferRange = st_bufferobj_flush_mapped_range;
    functions->UnmapBuffer = st_bufferobj_unmap;
+   functions->CopyBufferSubData = st_copy_buffer_subdata;
+
+   /* For GL_APPLE_vertex_array_object */
+   functions->NewArrayObject = _mesa_new_array_object;
+   functions->DeleteArrayObject = _mesa_delete_array_object;
 }
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h
index dcbb5a5233..fda6d05dd3 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.h
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.h
@@ -40,22 +40,14 @@ struct st_buffer_object
 {
    struct gl_buffer_object Base;
    struct pipe_buffer *buffer;  
-   GLsizeiptrARB size;
 };
 
 
-/* Are the obj->Name tests necessary?  Unfortunately yes, mesa
- * allocates a couple of gl_buffer_object structs statically, and the
- * Name == 0 test is the only way to identify them and avoid casting
- * them erroneously to our structs.
- */
+/** cast wrapper */
 static INLINE struct st_buffer_object *
 st_buffer_object(struct gl_buffer_object *obj)
 {
-   if (obj->Name)
-      return (struct st_buffer_object *) obj;
-   else
-      return NULL;
+   return (struct st_buffer_object *) obj;
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index adb349c1f1..a9cafbf8cd 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -60,6 +60,7 @@
 #include "pipe/p_inlines.h"
 #include "util/u_tile.h"
 #include "util/u_draw_quad.h"
+#include "util/u_math.h"
 #include "shader/prog_instruction.h"
 #include "cso_cache/cso_context.h"
 
@@ -341,6 +342,7 @@ make_texture(struct st_context *st,
    enum pipe_format pipeFormat;
    GLuint cpp;
    GLenum baseFormat;
+   int ptw, pth;
 
    baseFormat = _mesa_base_format(format);
 
@@ -351,14 +353,35 @@ make_texture(struct st_context *st,
    assert(pipeFormat);
    cpp = st_sizeof_format(pipeFormat);
 
-   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
    if (!pixels)
       return NULL;
 
-   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1,
+   /* Need to use POT texture? */
+   ptw = width;
+   pth = height;
+   if (!screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) {
+      int l2pt, maxSize;
+
+      l2pt = util_logbase2(width);
+      if (1<<l2pt != width) {
+         ptw = 1<<(l2pt+1);
+      }
+      l2pt = util_logbase2(height);
+      if (1<<l2pt != height) {
+         pth = 1<<(l2pt+1);
+      }
+
+      /* Check against maximum texture size */
+      maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+      assert(ptw <= maxSize);
+      assert(pth <= maxSize);
+   }
+
+   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, ptw, pth, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt) {
-      _mesa_unmap_drawpix_pbo(ctx, unpack);
+      _mesa_unmap_pbo_source(ctx, unpack);
       return NULL;
    }
 
@@ -405,7 +428,7 @@ make_texture(struct st_context *st,
       ctx->_ImageTransferState = imageTransferStateSave;
    }
 
-   _mesa_unmap_drawpix_pbo(ctx, unpack);
+   _mesa_unmap_pbo_source(ctx, unpack);
 
    return pt;
 }
@@ -420,7 +443,7 @@ make_texture(struct st_context *st,
 static void
 draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
           GLfloat x1, GLfloat y1, const GLfloat *color,
-          GLboolean invertTex)
+          GLboolean invertTex, GLfloat maxXcoord, GLfloat maxYcoord)
 {
    struct st_context *st = ctx->st;
    struct pipe_context *pipe = ctx->st->pipe;
@@ -435,8 +458,9 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
       const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
       const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
       const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
-      const GLfloat sLeft = 0.0f, sRight = 1.0f;
-      const GLfloat tTop = invertTex, tBot = 1.0f - tTop;
+      const GLfloat sLeft = 0.0f, sRight = maxXcoord;
+      const GLfloat tTop = invertTex ? maxYcoord : 0.0f;
+      const GLfloat tBot = invertTex ? 0.0f : maxYcoord;
       GLuint tex, i;
 
       /* upper-left */
@@ -608,7 +632,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    y0 = (GLfloat) y;
    y1 = y + height * ctx->Pixel.ZoomY;
 
-   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex);
+   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
+	     (GLfloat) width / pt->width[0],
+	     (GLfloat) height / pt->height[0]);
 
    /* restore state */
    cso_restore_rasterizer(cso);
@@ -648,14 +674,14 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       usage = PIPE_TRANSFER_READ_WRITE;
    else
       usage = PIPE_TRANSFER_WRITE;
-   
+
    pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, 0, 0, 0,
 				       usage, x, y,
 				       width, height);
 
    stmap = screen->transfer_map(screen, pt);
 
-   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
    assert(pixels);
 
    /* if width > MAX_WIDTH, have to process image in chunks */
@@ -749,7 +775,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       skipPixels += spanWidth;
    }
 
-   _mesa_unmap_drawpix_pbo(ctx, unpack);
+   _mesa_unmap_pbo_source(ctx, unpack);
 
    /* unmap the stencil buffer */
    screen->transfer_unmap(screen, pt);
@@ -778,8 +804,8 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
       return;
    }
 
-   _mesa_set_vp_override( ctx, TRUE );
-   _mesa_update_state( ctx );
+   /* Mesa state should be up to date by now */
+   assert(ctx->NewState == 0x0);
 
    st_validate_state(st);
 
@@ -807,8 +833,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
          pipe_texture_reference(&pt, NULL);
       }
    }
-
-   _mesa_set_vp_override( ctx, FALSE );
 }
 
 
@@ -853,7 +877,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
 
    assert(ptDraw->block.width == 1);
    assert(ptDraw->block.height == 1);
-   
+
    /* map the stencil buffer */
    drawMap = screen->transfer_map(screen, ptDraw);
 
@@ -927,6 +951,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    struct pipe_texture *pt;
    GLfloat *color;
    enum pipe_format srcFormat, texFormat;
+   int ptw, pth;
 
    pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
 
@@ -1008,13 +1033,34 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
          height -= -srcy;
          srcy = 0;
       }
-      
+
       if (height < 0)
          return;
    }
 
+   /* Need to use POT texture? */
+   ptw = width;
+   pth = height;
+   if (!screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) {
+      int l2pt, maxSize;
+
+      l2pt = util_logbase2(width);
+      if (1<<l2pt != width) {
+         ptw = 1<<(l2pt+1);
+      }
+      l2pt = util_logbase2(height);
+      if (1<<l2pt != height) {
+         pth = 1<<(l2pt+1);
+      }
+
+      /* Check against maximum texture size */
+      maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+      assert(ptw <= maxSize);
+      assert(pth <= maxSize);
+   }
+
    pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0,
-                          width, height, 1,
+                          ptw, pth, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt)
       return;
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 21ddf2fc7a..a96602878e 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -100,6 +100,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    strb->Base.Height = height;
    init_renderbuffer_bits(strb, format);
 
+   strb->defined = GL_FALSE;  /* undefined contents now */
+
    if(strb->software) {
       struct pipe_format_block block;
       size_t size;
@@ -444,6 +446,35 @@ st_finish_render_texture(GLcontext *ctx,
 
 
 /**
+ * Validate a renderbuffer attachment for a particular usage.
+ */
+
+static GLboolean
+st_validate_attachment(struct pipe_screen *screen,
+		       const struct gl_renderbuffer_attachment *att,
+		       GLuint usage)
+{
+   const struct st_texture_object *stObj =
+      st_texture_object(att->Texture);
+
+   /**
+    * Only validate texture attachments for now, since
+    * st_renderbuffer_alloc_storage makes sure that
+    * the format is supported.
+    */
+
+   if (att->Type != GL_TEXTURE)
+      return GL_TRUE;
+
+   if (!stObj)
+      return GL_FALSE;
+
+   return screen->is_format_supported(screen, stObj->pt->format,
+				      PIPE_TEXTURE_2D,
+				      usage, 0);
+}
+
+/**
  * Check that the framebuffer configuration is valid in terms of what
  * the driver can support.
  *
@@ -452,17 +483,169 @@ st_finish_render_texture(GLcontext *ctx,
 static void
 st_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
 {
+   struct pipe_screen *screen = ctx->st->pipe->screen;
    const struct gl_renderbuffer *depthRb =
       fb->Attachment[BUFFER_DEPTH].Renderbuffer;
    const struct gl_renderbuffer *stencilRb =
       fb->Attachment[BUFFER_STENCIL].Renderbuffer;
+   GLuint i;
 
    if (stencilRb && depthRb && stencilRb != depthRb) {
       fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+      return;
+   }
+
+   if (!st_validate_attachment(screen,
+			       &fb->Attachment[BUFFER_DEPTH],
+			       PIPE_TEXTURE_USAGE_DEPTH_STENCIL)) {
+      fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+      return;
+   }
+   if (!st_validate_attachment(screen,
+			       &fb->Attachment[BUFFER_STENCIL],
+			       PIPE_TEXTURE_USAGE_DEPTH_STENCIL)) {
+      fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+      return;
+   }
+   for (i = 0; i < ctx->Const.MaxColorAttachments; i++) {
+      if (!st_validate_attachment(screen,
+				  &fb->Attachment[BUFFER_COLOR0 + i],
+				  PIPE_TEXTURE_USAGE_RENDER_TARGET)) {
+	 fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+	 return;
+      }
+   }
+}
+
+
+/**
+ * Copy back color buffer to front color buffer.
+ */
+static void
+copy_back_to_front(struct st_context *st,
+                   struct gl_framebuffer *fb,
+                   gl_buffer_index frontIndex,
+                   gl_buffer_index backIndex)
+
+{
+   struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+   struct pipe_surface *surf_front, *surf_back;
+
+   (void) st_get_framebuffer_surface(stfb, frontIndex, &surf_front);
+   (void) st_get_framebuffer_surface(stfb, backIndex, &surf_back);
+
+   if (surf_front && surf_back) {
+      st->pipe->surface_copy(st->pipe,
+                             surf_front, 0, 0,  /* dest */
+                             surf_back, 0, 0,   /* src */
+                             fb->Width, fb->Height);
+   }
+}
+
+
+/**
+ * Check if we're drawing into, or read from, a front color buffer.  If the
+ * front buffer is missing, create it now.
+ *
+ * The back color buffer must exist since we'll use its format/samples info
+ * for creating the front buffer.
+ *
+ * \param frontIndex  either BUFFER_FRONT_LEFT or BUFFER_FRONT_RIGHT
+ * \param backIndex  either BUFFER_BACK_LEFT or BUFFER_BACK_RIGHT
+ */
+static void
+check_create_front_buffer(GLcontext *ctx, struct gl_framebuffer *fb,
+                          gl_buffer_index frontIndex,
+                          gl_buffer_index backIndex)
+{
+   if (fb->Attachment[frontIndex].Renderbuffer == NULL) {
+      GLboolean create = GL_FALSE;
+
+      /* check if drawing to or reading from front buffer */
+      if (fb->_ColorReadBufferIndex == frontIndex) {
+         create = GL_TRUE;
+      }
+      else {
+         GLuint b;
+         for (b = 0; b < fb->_NumColorDrawBuffers; b++) {
+            if (fb->_ColorDrawBufferIndexes[b] == frontIndex) {
+               create = GL_TRUE;
+               break;
+            }
+         }
+      }
+
+      if (create) {
+         struct st_renderbuffer *back;
+         struct gl_renderbuffer *front;
+         enum pipe_format colorFormat;
+         uint samples;
+
+         if (0)
+            _mesa_debug(ctx, "Allocate new front buffer\n");
+
+         /* get back renderbuffer info */
+         back = st_renderbuffer(fb->Attachment[backIndex].Renderbuffer);
+         colorFormat = back->format;
+         samples = back->Base.NumSamples;
+
+         /* create front renderbuffer */
+         front = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
+         _mesa_add_renderbuffer(fb, frontIndex, front);
+
+         /* alloc texture/surface for new front buffer */
+         front->AllocStorage(ctx, front, front->InternalFormat,
+                             fb->Width, fb->Height);
+
+         /* initialize the front color buffer contents by copying
+          * the back buffer.
+          */
+         copy_back_to_front(ctx->st, fb, frontIndex, backIndex);
+      }
    }
 }
 
 
+/**
+ * If front left/right color buffers are missing, create them now.
+ */
+static void
+check_create_front_buffers(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   /* check if we need to create the front left buffer now */
+   check_create_front_buffer(ctx, fb, BUFFER_FRONT_LEFT, BUFFER_BACK_LEFT);
+
+   if (fb->Visual.stereoMode) {
+      check_create_front_buffer(ctx, fb, BUFFER_FRONT_RIGHT, BUFFER_BACK_RIGHT);
+   }
+
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via glDrawBuffer.
+ */
+static void
+st_DrawBuffers(GLcontext *ctx, GLsizei count, const GLenum *buffers)
+{
+   (void) count;
+   (void) buffers;
+   check_create_front_buffers(ctx, ctx->DrawBuffer);
+}
+
+
+/**
+ * Called via glReadBuffer.
+ */
+static void
+st_ReadBuffer(GLcontext *ctx, GLenum buffer)
+{
+   (void) buffer;
+   check_create_front_buffers(ctx, ctx->ReadBuffer);
+}
+
+
 void st_init_fbo_functions(struct dd_function_table *functions)
 {
    functions->NewFramebuffer = st_new_framebuffer;
@@ -475,4 +658,7 @@ void st_init_fbo_functions(struct dd_function_table *functions)
    /* no longer needed by core Mesa, drivers handle resizes...
    functions->ResizeBuffers = st_resize_buffers;
    */
+
+   functions->DrawBuffers = st_DrawBuffers;
+   functions->ReadBuffer = st_ReadBuffer;
 }
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index 9a199550d9..bea6eb89c3 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -40,6 +40,7 @@ struct st_renderbuffer
    struct pipe_texture *texture;
    struct pipe_surface *surface; /* temporary view into texture */
    enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
+   GLboolean defined;        /**< defined contents? */
 
    /**
     * Used only when hardware accumulation buffers are not supported.
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
index f7de4e7a4d..1329f807bc 100644
--- a/src/mesa/state_tracker/st_cb_flush.c
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -47,10 +47,19 @@
 #include "util/u_blit.h"
 
 
+/** Check if we have a front color buffer and if it's been drawn to. */
 static INLINE GLboolean
 is_front_buffer_dirty(struct st_context *st)
 {
-   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+   if (st->frontbuffer_status == FRONT_STATUS_DIRTY) {
+      return GL_TRUE;
+   }
+   else {
+      GLframebuffer *fb = st->ctx->DrawBuffer;
+      struct st_renderbuffer *strb
+         = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+      return strb && strb->defined;
+   }
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
index 7dd2352739..d82b2a2035 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.c
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -194,9 +194,10 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw)
    rs->stage.destroy = rastpos_destroy;
    rs->ctx = ctx;
 
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(rs->array); i++) {
       rs->array[i].Size = 4;
       rs->array[i].Type = GL_FLOAT;
+      rs->array[i].Format = GL_RGBA;
       rs->array[i].Stride = 0;
       rs->array[i].StrideB = 0;
       rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i];
@@ -250,7 +251,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4])
    rs->array[0].Ptr = (GLubyte *) v;
 
    /* draw the point */
-   st_feedback_draw_vbo(ctx, rs->arrays, &rs->prim, 1, NULL, 0, 1);
+   st_feedback_draw_vbo(ctx, rs->arrays, &rs->prim, 1, NULL, GL_TRUE, 0, 1);
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index ccf1a0b563..75424aa2e7 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -353,7 +353,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
       return;
    }
 
-   dest = _mesa_map_readpix_pbo(ctx, &clippedPacking, dest);
+   dest = _mesa_map_pbo_dest(ctx, &clippedPacking, dest);
    if (!dest)
       return;
 
@@ -380,7 +380,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
    if (st_fast_readpixels(ctx, strb, x, y, width, height,
                           format, type, pack, dest)) {
       /* success! */
-      _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+      _mesa_unmap_pbo_dest(ctx, &clippedPacking);
       return;
    }
 
@@ -534,7 +534,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
 
    screen->tex_transfer_destroy(trans);
 
-   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+   _mesa_unmap_pbo_dest(ctx, &clippedPacking);
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 2db28ef0a4..abaf9d2c35 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -60,6 +60,7 @@
 #include "util/u_tile.h"
 #include "util/u_blit.h"
 #include "util/u_surface.h"
+#include "util/u_math.h"
 
 
 #define DBG if (0) printf
@@ -238,16 +239,6 @@ do_memcpy(void *dest, const void *src, size_t n)
 }
 
 
-static INLINE unsigned
-logbase2(unsigned n)
-{
-   unsigned log2 = 0;
-   while (n >>= 1)
-      ++log2;
-   return log2;
-}
-
-
 /**
  * Return default texture usage bitmask for the given texture format.
  */
@@ -341,9 +332,9 @@ guess_and_alloc_texture(struct st_context *st,
       lastLevel = firstLevel;
    }
    else {
-      GLuint l2width = logbase2(width);
-      GLuint l2height = logbase2(height);
-      GLuint l2depth = logbase2(depth);
+      GLuint l2width = util_logbase2(width);
+      GLuint l2height = util_logbase2(height);
+      GLuint l2depth = util_logbase2(depth);
       lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth);
    }
 
@@ -533,6 +524,12 @@ st_TexImage(GLcontext * ctx,
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
 
+   /* switch to "normal" */
+   if (stObj->surface_based) {
+      _mesa_clear_texture_object(ctx, texObj);
+      stObj->surface_based = GL_FALSE;
+   }
+
    /* gallium does not support texture borders, strip it off */
    if (border) {
       strip_texture_border(border, &width, &height, &depth, unpack, &unpackNB);
@@ -874,7 +871,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level,
 					     PIPE_TRANSFER_READ,
 					     0, 0, width, height);
 
-   pixels = _mesa_map_readpix_pbo(ctx, &ctx->Pack, pixels);
+   pixels = _mesa_map_pbo_dest(ctx, &ctx->Pack, pixels);
 
    /* copy/pack data into user buffer */
    if (st_equal_formats(stImage->pt->format, format, type)) {
@@ -907,7 +904,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level,
       }
    }
 
-   _mesa_unmap_readpix_pbo(ctx, &ctx->Pack);
+   _mesa_unmap_pbo_dest(ctx, &ctx->Pack);
 
    /* destroy the temp / dest surface */
    util_destroy_rgba_surface(dst_texture, dst_surface);
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index 6ffed56d9a..18adb35e87 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -45,6 +45,7 @@ struct blit_state;
 struct bitmap_cache;
 
 
+/** XXX we'd like to get rid of these */
 #define FRONT_STATUS_UNDEFINED    0
 #define FRONT_STATUS_DIRTY        1
 #define FRONT_STATUS_COPY_OF_BACK 2
@@ -111,7 +112,7 @@ struct st_context
       struct gl_fragment_program *fragment_program;
    } cb;
 
-   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ (XXX to be removed) */
 
    char vendor[100];
    char renderer[100];
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 225541a30b..503a5f34a3 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -159,12 +159,21 @@ static GLuint fixed_types[4] = {
  * Return a PIPE_FORMAT_x for the given GL datatype and size.
  */
 GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized)
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized)
 {
    assert((type >= GL_BYTE && type <= GL_DOUBLE) ||
           type == GL_FIXED);
    assert(size >= 1);
    assert(size <= 4);
+   assert(format == GL_RGBA || format == GL_BGRA);
+
+   if (format == GL_BGRA) {
+      /* this is an odd-ball case */
+      assert(type == GL_UNSIGNED_BYTE);
+      assert(normalized);
+      return PIPE_FORMAT_B8G8R8A8_UNORM;
+   }
 
    if (normalized) {
       switch (type) {
@@ -220,8 +229,10 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
       struct st_buffer_object *stobj = st_buffer_object(array->BufferObj);
       ubyte *map;
 
-      if (!stobj)
+      if (!stobj || stobj->Base.Name == 0) {
+         /* edge flags are not in a VBO */
          return NULL;
+      }
 
       vec = (unsigned *) _mesa_calloc(sizeof(unsigned) * ((count + 31) / 32));
       if (!vec)
@@ -392,6 +403,7 @@ setup_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format =
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
@@ -479,6 +491,7 @@ setup_non_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format
          = st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                  arrays[mesaAttr]->Size,
+                                 arrays[mesaAttr]->Format,
                                  arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
@@ -520,6 +533,7 @@ st_draw_vbo(GLcontext *ctx,
             const struct _mesa_prim *prims,
             GLuint nr_prims,
             const struct _mesa_index_buffer *ib,
+	    GLboolean index_bounds_valid,
             GLuint min_index,
             GLuint max_index)
 {
@@ -532,6 +546,10 @@ st_draw_vbo(GLcontext *ctx,
    unsigned num_vbuffers, num_velements;
    GLboolean userSpace;
 
+   /* Gallium probably doesn't want this in some cases. */
+   if (!index_bounds_valid)
+      vbo_get_minmax_index(ctx, prims, ib, &min_index, &max_index);
+
    /* sanity check for pointer arithmetic below */
    assert(sizeof(arrays[0]->Ptr[0]) == 1);
 
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
index da04fce8e2..3e0face656 100644
--- a/src/mesa/state_tracker/st_draw.h
+++ b/src/mesa/state_tracker/st_draw.h
@@ -47,6 +47,7 @@ st_draw_vbo(GLcontext *ctx,
             const struct _mesa_prim *prims,
             GLuint nr_prims,
             const struct _mesa_index_buffer *ib,
+	    GLboolean index_bounds_valid,
             GLuint min_index,
             GLuint max_index);
 
@@ -56,13 +57,15 @@ st_feedback_draw_vbo(GLcontext *ctx,
                      const struct _mesa_prim *prims,
                      GLuint nr_prims,
                      const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
                      GLuint min_index,
                      GLuint max_index);
 
 /* Internal function:
  */
 extern GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized);
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized);
 
 
 /**
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index 32502a9cda..b2d682ef64 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -96,6 +96,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
                      const struct _mesa_prim *prims,
                      GLuint nr_prims,
                      const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
                      GLuint min_index,
                      GLuint max_index)
 {
@@ -114,6 +115,9 @@ st_feedback_draw_vbo(GLcontext *ctx,
 
    st_validate_state(ctx->st);
 
+   if (!index_bounds_valid)
+      vbo_get_minmax_index(ctx, prims, ib, &min_index, &max_index);
+
    /* must get these after state validation! */
    vp = ctx->st->vp;
    vs = &st->vp->state;
@@ -178,6 +182,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
       velements[attr].src_format = 
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
 
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index 9f6564dbc5..3f835d38dd 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -143,19 +143,21 @@ void st_init_extensions(struct st_context *st)
    /*
     * Extensions that are supported by all Gallium drivers:
     */
-   ctx->Extensions.ARB_multisample = GL_TRUE;
+   ctx->Extensions.ARB_copy_buffer = GL_TRUE;
    ctx->Extensions.ARB_fragment_program = GL_TRUE;
+   ctx->Extensions.ARB_map_buffer_range = GL_TRUE;
+   ctx->Extensions.ARB_multisample = GL_TRUE;
    ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */
    ctx->Extensions.ARB_texture_compression = GL_TRUE;
    ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
    ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
    ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE;
    ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE;
-   ctx->Extensions.ARB_vertex_program = GL_TRUE;
+   ctx->Extensions.ARB_vertex_array_object = GL_TRUE;
    ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+   ctx->Extensions.ARB_vertex_program = GL_TRUE;
 
    ctx->Extensions.EXT_blend_color = GL_TRUE;
-   ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_func_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_logic_op = GL_TRUE;
    ctx->Extensions.EXT_blend_minmax = GL_TRUE;
@@ -166,15 +168,20 @@ void st_init_extensions(struct st_context *st)
    ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
    ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
    ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_provoking_vertex = GL_TRUE;
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
    ctx->Extensions.EXT_texture_env_add = GL_TRUE;
    ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
    ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
    ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
+   ctx->Extensions.EXT_vertex_array_bgra = GL_TRUE;
+
+   ctx->Extensions.APPLE_vertex_array_object = GL_TRUE;
 
    ctx->Extensions.NV_blend_square = GL_TRUE;
    ctx->Extensions.NV_texgen_reflection = GL_TRUE;
+   ctx->Extensions.NV_texture_env_combine4 = GL_TRUE;
 
    ctx->Extensions.SGI_color_matrix = GL_TRUE;
    ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
@@ -198,6 +205,10 @@ void st_init_extensions(struct st_context *st)
       ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
    }
 
+   if (screen->get_param(screen, PIPE_CAP_BLEND_EQUATION_SEPARATE)) {
+      ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
+   }
+
    if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_CLAMP) > 0) {
       ctx->Extensions.EXT_texture_mirror_clamp = GL_TRUE;
    }
@@ -284,4 +295,10 @@ void st_init_extensions(struct st_context *st)
                                    PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
       ctx->Extensions.MESA_ycbcr_texture = GL_TRUE;
    }
+
+   /* GL_ARB_framebuffer_object */
+   if (ctx->Extensions.EXT_packed_depth_stencil) {
+      /* we support always support GL_EXT_framebuffer_blit */
+      ctx->Extensions.ARB_framebuffer_object = GL_TRUE;
+   }
 }
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
index b243c249e3..dcb90a3107 100644
--- a/src/mesa/state_tracker/st_format.c
+++ b/src/mesa/state_tracker/st_format.c
@@ -435,10 +435,10 @@ st_choose_format(struct pipe_context *pipe, GLenum internalFormat,
    case GL_RGB5:
    case GL_RGB4:
    case GL_R3_G3_B2:
-      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
-         return PIPE_FORMAT_A1R5G5B5_UNORM;
       if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, target, tex_usage, geom_flags ))
          return PIPE_FORMAT_R5G6B5_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A1R5G5B5_UNORM;
       return default_rgba_format( screen, target, tex_usage, geom_flags );
 
    case GL_ALPHA:
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index fe99fc0892..ca32b2e573 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -58,19 +58,19 @@ st_create_framebuffer( const __GLcontextModes *visual,
 
       _mesa_initialize_framebuffer(&stfb->Base, visual);
 
-      {
-         /* fake frontbuffer */
-         /* XXX allocation should only happen in the unusual case
-            it's actually needed */
+      if (visual->doubleBufferMode) {
          struct gl_renderbuffer *rb
             = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
       }
-
-      if (visual->doubleBufferMode) {
+      else {
+         /* Only allocate front buffer right now if we're single buffered.
+          * If double-buffered, allocate front buffer on demand later.
+          * See check_create_front_buffers().
+          */
          struct gl_renderbuffer *rb
             = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
       }
 
       if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
@@ -286,6 +286,115 @@ st_notify_swapbuffers(struct st_framebuffer *stfb)
 }
 
 
+/**
+ * Swap the front/back color buffers.  Exchange the front/back pointers
+ * and update some derived state.
+ * No need to call st_notify_swapbuffers() first.
+ *
+ * For a single-buffered framebuffer, no swap occurs, but we still return
+ * the pointer(s) to the front color buffer(s).
+ *
+ * \param front_left  returns pointer to front-left renderbuffer after swap
+ * \param front_right  returns pointer to front-right renderbuffer after swap
+ */
+void
+st_swapbuffers(struct st_framebuffer *stfb,
+               struct pipe_surface **front_left,
+               struct pipe_surface **front_right)
+{
+   struct gl_framebuffer *fb = &stfb->Base;
+
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+   }
+
+   if (!fb->Visual.doubleBufferMode) {
+      /* single buffer mode - return pointers to front surfaces */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+      return;
+   }
+
+   /* swap left buffers */
+   if (fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer = rbTemp;
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front buffer, display the back buffer */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+   }
+
+   /* swap right buffers (for stereo) */
+   if (fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer = rbTemp;
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front right buffer, display back right buffer (if exists) */
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+   }
+
+   /* Update the _ColorDrawBuffers[] array and _ColorReadBuffer pointer */
+   _mesa_update_framebuffer(ctx);
+
+   /* Make sure we draw into the new back surface */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
 void *st_framebuffer_private( struct st_framebuffer *stfb )
 {
    return stfb->Private;
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index e150dff9bb..2ab12d3cf3 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -112,27 +112,6 @@ map_register_file_index(
 {
    switch( file ) {
    case TGSI_FILE_INPUT:
-      if (procType == TGSI_PROCESSOR_FRAGMENT &&
-          index == FRAG_ATTRIB_FOGC) {
-         if (GET_SWZ(*swizzle, 0) == SWIZZLE_X) {
-            /* do nothing we're, ok */
-         } else if (GET_SWZ(*swizzle, 0) == SWIZZLE_Y) {
-            /* replace the swizzle with xxxx */
-            *swizzle = MAKE_SWIZZLE4(SWIZZLE_X,
-                                     SWIZZLE_X,
-                                     SWIZZLE_X,
-                                     SWIZZLE_X);
-            /* register after fog */
-            return inputMapping[index] + 1;
-         } else {
-            *swizzle = MAKE_SWIZZLE4(SWIZZLE_Z,
-                                     SWIZZLE_W,
-                                     SWIZZLE_Z,
-                                     SWIZZLE_W);
-            /* register after frontface */
-            return inputMapping[index] + 2;
-         }
-      }
       /* inputs are mapped according to the user-defined map */
       return inputMapping[index];
 
@@ -158,12 +137,6 @@ map_texture_target(
     GLuint textarget,
     GLboolean shadow )
 {
-#if 1
-   /* XXX remove this line after we've checked that the rest of gallium
-    * can handle the TGSI_TEXTURE_SHADOWx tokens.
-    */
-   shadow = GL_FALSE;
-#endif
    switch( textarget ) {
    case TEXTURE_1D_INDEX:
       if (shadow)
@@ -225,11 +198,15 @@ static struct tgsi_full_immediate
 make_immediate(const float *value, uint size)
 {
    struct tgsi_full_immediate imm;
+   unsigned i;
 
    imm = tgsi_default_full_immediate();
    imm.Immediate.NrTokens += size;
    imm.Immediate.DataType = TGSI_IMM_FLOAT32;
-   imm.u.Pointer = value;
+
+   for (i = 0; i < size; i++)
+      imm.u[i].Float = value[i];
+
    return imm;
 }
 
@@ -370,7 +347,7 @@ compile_instruction(
       fullinst->Instruction.Opcode = TGSI_OPCODE_ADD;
       break;
    case OPCODE_BGNLOOP:
-      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2;
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP;
       fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
       break;
    case OPCODE_BGNSUB:
@@ -428,7 +405,7 @@ compile_instruction(
       fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF;
       break;
    case OPCODE_ENDLOOP:
-      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2;
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP;
       fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
       break;
    case OPCODE_ENDSUB:
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 18d1046831..d2da20ae42 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -151,6 +151,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_TEX5:
          case VERT_ATTRIB_TEX6:
          case VERT_ATTRIB_TEX7:
+            assert(slot < Elements(vs_input_semantic_name));
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
             break;
@@ -171,6 +172,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_GENERIC14:
          case VERT_ATTRIB_GENERIC15:
             assert(attr < VERT_ATTRIB_MAX);
+            assert(slot < Elements(vs_input_semantic_name));
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
             break;
@@ -198,6 +200,7 @@ st_translate_vertex_program(struct st_context *st,
 
    /* initialize output semantics to defaults */
    for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) {
+      assert(i < Elements(vs_output_semantic_name));
       vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC;
       vs_output_semantic_index[i] = 0;
       output_flags[i] = 0x0;
@@ -273,6 +276,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_RESULT_VAR0:
             /* fall-through */
          default:
+            assert(slot < Elements(vs_output_semantic_name));
             if (outputSemanticName) {
                /* use provided semantic into */
                assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT);
@@ -286,6 +290,7 @@ st_translate_vertex_program(struct st_context *st,
             }
          }
 
+         assert(slot < Elements(output_flags));
          output_flags[slot] = stvp->Base.Base.OutputFlags[attr];
       }
    }
@@ -303,6 +308,26 @@ st_translate_vertex_program(struct st_context *st,
       outputMapping = defaultOutputMapping;
    }
 
+#if 0 /* debug */
+   {
+      GLuint i;
+      printf("outputMapping? %d\n", outputMapping ? 1 : 0);
+      if (outputMapping) {
+         printf("attr -> slot\n");
+         for (i = 0; i < 16;  i++) {
+            printf(" %2d       %3d\n", i, outputMapping[i]);
+         }
+      }
+      printf("slot    sem_name  sem_index\n");
+      for (i = 0; i < vs_num_outputs; i++) {
+         printf(" %2d         %d         %d\n",
+                i,
+                vs_output_semantic_name[i],
+                vs_output_semantic_index[i]);
+      }
+   }
+#endif
+
    /* free old shader state, if any */
    if (stvp->state.tokens) {
       _mesa_free((void *) stvp->state.tokens);
@@ -433,34 +458,20 @@ st_translate_fragment_program(struct st_context *st,
             stfp->input_semantic_index[slot] = 1;
             interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
             break;
-         case FRAG_ATTRIB_FOGC: {
-            int extra_decls = 0;
-            if (stfp->Base.UsesFogFragCoord) {
-               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
-               stfp->input_semantic_index[slot] = 0;
-               interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
-               input_flags[slot] = stfp->Base.Base.InputFlags[attr];
-               ++extra_decls;
-	    }
-            if (stfp->Base.UsesFrontFacing) {
-               GLint idx = slot + extra_decls;
-               stfp->input_semantic_name[idx] = TGSI_SEMANTIC_FACE;
-               stfp->input_semantic_index[idx] = 0;
-               interpMode[idx] = TGSI_INTERPOLATE_CONSTANT;
-               input_flags[idx] = stfp->Base.Base.InputFlags[attr];
-               ++extra_decls;
-            }
-            if (stfp->Base.UsesPointCoord) {
-               GLint idx = slot + extra_decls;
-               stfp->input_semantic_name[idx] = TGSI_SEMANTIC_GENERIC;
-               stfp->input_semantic_index[idx] = num_generic++;
-               interpMode[idx] = TGSI_INTERPOLATE_PERSPECTIVE;
-               input_flags[idx] = stfp->Base.Base.InputFlags[attr];
-               ++extra_decls;
-            }
-            fs_num_inputs += extra_decls - 1;
-            continue;
-         }
+         case FRAG_ATTRIB_FOGC:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+            break;
+         case FRAG_ATTRIB_FACE:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FACE;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_CONSTANT;
+            break;
+         case FRAG_ATTRIB_PNTC:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
             break;
          case FRAG_ATTRIB_TEX0:
          case FRAG_ATTRIB_TEX1:
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
index c411687bb6..a5fdac32d1 100644
--- a/src/mesa/state_tracker/st_public.h
+++ b/src/mesa/state_tracker/st_public.h
@@ -103,7 +103,13 @@ void st_finish( struct st_context *st );
 
 void st_notify_swapbuffers(struct st_framebuffer *stfb);
 
-int st_set_teximage(struct pipe_texture *pt, int target);
+void st_swapbuffers(struct st_framebuffer *stfb,
+                    struct pipe_surface **front_left,
+                    struct pipe_surface **front_right);
+
+int st_bind_texture_surface(struct pipe_surface *ps, int target, int level,
+                            enum pipe_format format);
+int st_unbind_texture_surface(struct pipe_surface *ps, int target, int level);
 
 /** Redirect rendering into stfb's surface to a texture image */
 int st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
index 10faa633ea..bbc2830e69 100644
--- a/src/mesa/state_tracker/st_texture.c
+++ b/src/mesa/state_tracker/st_texture.c
@@ -32,6 +32,7 @@
 #include "st_cb_fbo.h"
 #include "st_inlines.h"
 #include "main/enums.h"
+#include "main/texobj.h"
 #include "main/teximage.h"
 #include "main/texstore.h"
 
@@ -239,7 +240,7 @@ st_surface_data(struct pipe_context *pipe,
    struct pipe_screen *screen = pipe->screen;
    void *map = screen->transfer_map(screen, dst);
 
-   pipe_copy_rect(map,
+   util_copy_rect(map,
                   &dst->block,
                   dst->stride,
                   dstx, dsty, 
@@ -353,25 +354,95 @@ st_texture_image_copy(struct pipe_context *pipe,
    }
 }
 
-/** Bind a pipe surface for use as a texture image */
+
+/**
+ * Bind a pipe surface to a texture object.  After the call,
+ * the texture object is marked dirty and will be (re-)validated.
+ *
+ * If this is the first surface bound, the texture object is said to
+ * switch from normal to surface based.  It will be cleared first in
+ * this case.
+ *
+ * \param ps      pipe surface to be unbound
+ * \param target  texture target
+ * \param level   image level
+ * \param format  internal format of the texture
+ */
 int
-st_set_teximage(struct pipe_texture *pt, int target)
+st_bind_texture_surface(struct pipe_surface *ps, int target, int level,
+                        enum pipe_format format)
 {
    GET_CURRENT_CONTEXT(ctx);
    const GLuint unit = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
    struct gl_texture_object *texObj;
    struct gl_texture_image *texImage;
+   struct st_texture_object *stObj;
    struct st_texture_image *stImage;
-   int internalFormat;
+   GLenum internalFormat;
 
-   switch (pt->format) {
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-      internalFormat = GL_RGBA8;
+   switch (target) {
+   case ST_TEXTURE_2D:
+      target = GL_TEXTURE_2D;
+      break;
+   case ST_TEXTURE_RECT:
+      target = GL_TEXTURE_RECTANGLE_ARB;
       break;
    default:
       return 0;
-   };
+   }
+
+   /* map pipe format to base format for now */
+   if (pf_get_component_bits(format, PIPE_FORMAT_COMP_A) > 0)
+      internalFormat = GL_RGBA;
+   else
+      internalFormat = GL_RGB;
+
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   _mesa_lock_texture(ctx, texObj);
+
+   stObj = st_texture_object(texObj);
+   /* switch to surface based */
+   if (!stObj->surface_based) {
+      _mesa_clear_texture_object(ctx, texObj);
+      stObj->surface_based = GL_TRUE;
+   }
+
+   texImage = _mesa_get_tex_image(ctx, texObj, target, level);
+   stImage = st_texture_image(texImage);
+
+   _mesa_init_teximage_fields(ctx, target, texImage,
+                              ps->width, ps->height, 1, 0, internalFormat);
+   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat,
+                                                GL_RGBA, GL_UNSIGNED_BYTE);
+   _mesa_set_fetch_functions(texImage, 2);
+   pipe_texture_reference(&stImage->pt, ps->texture);
+
+   _mesa_dirty_texobj(ctx, texObj, GL_TRUE);
+   _mesa_unlock_texture(ctx, texObj);
+   
+   return 1;
+}
+
+
+/**
+ * Unbind a pipe surface from a texture object.  After the call,
+ * the texture object is marked dirty and will be (re-)validated.
+ *
+ * \param ps      pipe surface to be unbound
+ * \param target  texture target
+ * \param level   image level
+ */
+int
+st_unbind_texture_surface(struct pipe_surface *ps, int target, int level)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   struct st_texture_object *stObj;
+   struct st_texture_image *stImage;
 
    switch (target) {
    case ST_TEXTURE_2D:
@@ -385,21 +456,28 @@ st_set_teximage(struct pipe_texture *pt, int target)
    }
 
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
-   texImage = _mesa_get_tex_image(ctx, texObj, target, 0);
+
+   _mesa_lock_texture(ctx, texObj);
+
+   texImage = _mesa_get_tex_image(ctx, texObj, target, level);
+   stObj = st_texture_object(texObj);
    stImage = st_texture_image(texImage);
-   
-   _mesa_init_teximage_fields(ctx, GL_TEXTURE_2D, texImage, pt->width[0],
-                              pt->height[0], 1, 0, internalFormat);
 
-   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, GL_RGBA,
-                                                GL_UNSIGNED_BYTE);
-   _mesa_set_fetch_functions(texImage, 2);
+   /* Make sure the pipe surface is still bound.  The texture object is still
+    * considered surface based even if this is the last bound surface. */
+   if (stImage->pt == ps->texture) {
+      pipe_texture_reference(&stImage->pt, NULL);
+      _mesa_clear_texture_image(ctx, texImage);
 
-   pipe_texture_reference(&stImage->pt, pt);
+      _mesa_dirty_texobj(ctx, texObj, GL_TRUE);
+   }
 
+   _mesa_unlock_texture(ctx, texObj);
+   
    return 1;
 }
 
+
 /** Redirect rendering into stfb's surface to a texture image */
 int
 st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
index b9d447cb56..60868ce067 100644
--- a/src/mesa/state_tracker/st_texture.h
+++ b/src/mesa/state_tracker/st_texture.h
@@ -69,6 +69,11 @@ struct st_texture_object
    struct pipe_texture *pt;
 
    GLboolean teximage_realloc;
+
+   /* True if there is/was a surface bound to this texture object.  It helps
+    * track whether the texture object is surface based or not.
+    */
+   GLboolean surface_based;
 };
 
 
diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c
index 5e7822cf32..3dbdf2a61a 100644
--- a/src/mesa/swrast/s_bitmap.c
+++ b/src/mesa/swrast/s_bitmap.c
@@ -56,7 +56,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
 
    ASSERT(ctx->RenderMode == GL_RENDER);
 
-   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   bitmap = (const GLubyte *) _mesa_map_pbo_source(ctx, unpack, bitmap);
    if (!bitmap)
       return;
 
@@ -133,7 +133,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
 
    swrast_render_finish(ctx);
 
-   _mesa_unmap_bitmap_pbo(ctx, unpack);
+   _mesa_unmap_pbo_source(ctx, unpack);
 }
 
 
diff --git a/src/mesa/swrast/s_blit.c b/src/mesa/swrast/s_blit.c
index 4a95c222d5..8303e4debc 100644
--- a/src/mesa/swrast/s_blit.c
+++ b/src/mesa/swrast/s_blit.c
@@ -105,7 +105,7 @@ static void
 blit_nearest(GLcontext *ctx,
              GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
              GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
-             GLenum buffer)
+             GLbitfield buffer)
 {
    struct gl_renderbuffer *readRb, *drawRb;
 
@@ -457,7 +457,7 @@ static void
 simple_blit(GLcontext *ctx,
             GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
             GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
-            GLenum buffer)
+            GLbitfield buffer)
 {
    struct gl_renderbuffer *readRb, *drawRb;
    const GLint width = srcX1 - srcX0;
@@ -560,7 +560,7 @@ _swrast_BlitFramebuffer(GLcontext *ctx,
                         GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
                         GLbitfield mask, GLenum filter)
 {
-   static const GLint buffers[3] = {
+   static const GLbitfield buffers[3] = {
       GL_COLOR_BUFFER_BIT,
       GL_DEPTH_BUFFER_BIT,
       GL_STENCIL_BUFFER_BIT
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index a9ef8e685f..6970b2e9cb 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -831,11 +831,13 @@ _swrast_DrawPixels( GLcontext *ctx,
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLboolean save_vp_override = ctx->VertexProgram._Overriden;
 
-   /* We are creating fragments directly, without going through vertex programs.
+   /* We are creating fragments directly, without going through vertex
+    * programs.
     *
-    * This override flag tells the fragment processing code that its input comes
-    * from a non-standard source, and it may therefore not rely on optimizations
-    * that assume e.g. constant color if there is no color vertex array.
+    * This override flag tells the fragment processing code that its input
+    * comes from a non-standard source, and it may therefore not rely on
+    * optimizations that assume e.g. constant color if there is no color
+    * vertex array.
     */
    _mesa_set_vp_override(ctx, GL_TRUE);
 
@@ -847,7 +849,7 @@ _swrast_DrawPixels( GLcontext *ctx,
    if (swrast->NewState)
       _swrast_validate_derived( ctx );
 
-    pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+    pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
     if (!pixels) {
        swrast_render_finish(ctx);
        _mesa_set_vp_override(ctx, save_vp_override);
@@ -892,58 +894,5 @@ _swrast_DrawPixels( GLcontext *ctx,
    swrast_render_finish(ctx);
    _mesa_set_vp_override(ctx, save_vp_override);
 
-   _mesa_unmap_drawpix_pbo(ctx, unpack);
+   _mesa_unmap_pbo_source(ctx, unpack);
 }
-
-
-
-#if 0  /* experimental */
-/*
- * Execute glDrawDepthPixelsMESA().
- */
-void
-_swrast_DrawDepthPixelsMESA( GLcontext *ctx,
-                             GLint x, GLint y,
-                             GLsizei width, GLsizei height,
-                             GLenum colorFormat, GLenum colorType,
-                             const GLvoid *colors,
-                             GLenum depthType, const GLvoid *depths,
-                             const struct gl_pixelstore_attrib *unpack )
-{
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
-
-   if (swrast->NewState)
-      _swrast_validate_derived( ctx );
-
-   swrast_render_start(ctx);
-
-   switch (colorFormat) {
-   case GL_COLOR_INDEX:
-      if (ctx->Visual.rgbMode)
-	 draw_rgba_pixels(ctx, x,y, width, height, colorFormat, colorType,
-                          unpack, colors);
-      else
-	 draw_index_pixels(ctx, x, y, width, height, colorType,
-                           unpack, colors);
-      break;
-   case GL_RED:
-   case GL_GREEN:
-   case GL_BLUE:
-   case GL_ALPHA:
-   case GL_LUMINANCE:
-   case GL_LUMINANCE_ALPHA:
-   case GL_RGB:
-   case GL_BGR:
-   case GL_RGBA:
-   case GL_BGRA:
-   case GL_ABGR_EXT:
-      draw_rgba_pixels(ctx, x, y, width, height, colorFormat, colorType,
-                       unpack, colors);
-      break;
-   default:
-      _mesa_problem(ctx, "unexpected format in glDrawDepthPixelsMESA");
-   }
-
-   swrast_render_finish(ctx);
-}
-#endif
diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c
index b71fb9eae9..613a91b0ec 100644
--- a/src/mesa/swrast/s_fragprog.c
+++ b/src/mesa/swrast/s_fragprog.c
@@ -157,9 +157,8 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine,
 
    /* if running a GLSL program (not ARB_fragment_program) */
    if (ctx->Shader.CurrentProgram) {
-      /* Store front/back facing value in register FOGC.Y */
-      machine->Attribs[FRAG_ATTRIB_FOGC][col][1] = 1.0 - span->facing;
-      /* Note FOGC.ZW is gl_PointCoord if drawing a sprite */
+      /* Store front/back facing value */
+      machine->Attribs[FRAG_ATTRIB_FACE][col][0] = 1.0 - span->facing;
    }
 
    machine->CurElement = col;
diff --git a/src/mesa/swrast/s_imaging.c b/src/mesa/swrast/s_imaging.c
index d6be3aa022..3578b713f6 100644
--- a/src/mesa/swrast/s_imaging.c
+++ b/src/mesa/swrast/s_imaging.c
@@ -60,7 +60,7 @@ _swrast_CopyColorTable( GLcontext *ctx,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorTable(target, internalformat, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -94,7 +94,7 @@ _swrast_CopyColorSubTable( GLcontext *ctx,GLenum target, GLsizei start,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorSubTable(target, start, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -126,7 +126,7 @@ _swrast_CopyConvolutionFilter1D(GLcontext *ctx, GLenum target,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    /* store as convolution filter */
    _mesa_ConvolutionFilter1D(target, internalFormat, width,
@@ -178,12 +178,12 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
    ctx->Unpack.SkipImages = 0;
    ctx->Unpack.SwapBytes = GL_FALSE;
    ctx->Unpack.LsbFirst = GL_FALSE;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
    ctx->NewState |= _NEW_PACKUNPACK;
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ConvolutionFilter2D(target, internalFormat, width, height,
                              GL_RGBA, CHAN_TYPE, rgba);
diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c
index 0a3ad97a71..50ec2063a5 100644
--- a/src/mesa/swrast/s_points.c
+++ b/src/mesa/swrast/s_points.c
@@ -139,9 +139,10 @@ sprite_point(GLcontext *ctx, const SWvertex *vert)
       }
 
       ATTRIB_LOOP_BEGIN
-         if (attr >= FRAG_ATTRIB_TEX0 && attr < FRAG_ATTRIB_VAR0) {
+         if (attr >= FRAG_ATTRIB_TEX0 && attr <= FRAG_ATTRIB_TEX7) {
+            /* a texcoord attribute */
             const GLuint u = attr - FRAG_ATTRIB_TEX0;
-            /* a texcoord */
+            ASSERT(u < Elements(ctx->Point.CoordReplace));
             if (ctx->Point.CoordReplace[u]) {
                tCoords[numTcoords++] = attr;
 
@@ -170,15 +171,15 @@ sprite_point(GLcontext *ctx, const SWvertex *vert)
                continue;
             }
          }
-         else if (attr == FRAG_ATTRIB_FOGC) {
-            /* GLSL gl_PointCoord is stored in fog.zw */
-            span.attrStart[FRAG_ATTRIB_FOGC][2] = 0.0;
-            span.attrStart[FRAG_ATTRIB_FOGC][3] = 0.0; /* t0 set below */
-            span.attrStepX[FRAG_ATTRIB_FOGC][2] = dsdx;
-            span.attrStepX[FRAG_ATTRIB_FOGC][3] = 0.0;
-            span.attrStepY[FRAG_ATTRIB_FOGC][2] = 0.0;
-            span.attrStepY[FRAG_ATTRIB_FOGC][3] = dtdy;
-            tCoords[numTcoords++] = FRAG_ATTRIB_FOGC;
+         else if (attr == FRAG_ATTRIB_PNTC) {
+            /* GLSL gl_PointCoord.xy (.zw undefined) */
+            span.attrStart[FRAG_ATTRIB_PNTC][0] = 0.0;
+            span.attrStart[FRAG_ATTRIB_PNTC][1] = 0.0; /* t0 set below */
+            span.attrStepX[FRAG_ATTRIB_PNTC][0] = dsdx;
+            span.attrStepX[FRAG_ATTRIB_PNTC][1] = 0.0;
+            span.attrStepY[FRAG_ATTRIB_PNTC][0] = 0.0;
+            span.attrStepY[FRAG_ATTRIB_PNTC][1] = dtdy;
+            tCoords[numTcoords++] = FRAG_ATTRIB_PNTC;
             continue;
          }
          /* use vertex's texcoord/attrib */
@@ -221,10 +222,7 @@ sprite_point(GLcontext *ctx, const SWvertex *vert)
          GLuint i;
          /* setup texcoord T for this row */
          for (i = 0; i < numTcoords; i++) {
-            if (tCoords[i] == FRAG_ATTRIB_FOGC)
-               span.attrStart[FRAG_ATTRIB_FOGC][3] = tcoord;
-            else
-               span.attrStart[tCoords[i]][1] = tcoord;
+            span.attrStart[tCoords[i]][1] = tcoord;
          }
 
          /* these might get changed by span clipping */
diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c
index e901fc6b5d..48b9408d24 100644
--- a/src/mesa/swrast/s_readpix.c
+++ b/src/mesa/swrast/s_readpix.c
@@ -574,7 +574,7 @@ _swrast_ReadPixels( GLcontext *ctx,
       return;
    }
 
-   pixels = _mesa_map_readpix_pbo(ctx, &clippedPacking, pixels);
+   pixels = _mesa_map_pbo_dest(ctx, &clippedPacking, pixels);
    if (!pixels)
       return;
   
@@ -616,5 +616,5 @@ _swrast_ReadPixels( GLcontext *ctx,
 
    swrast_render_finish(ctx);
 
-   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+   _mesa_unmap_pbo_dest(ctx, &clippedPacking);
 }
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index fa8ca1d0e2..0e2793b474 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -1297,7 +1297,6 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
           span->primitive == GL_LINE ||
 	  span->primitive == GL_POLYGON ||
           span->primitive == GL_BITMAP);
-   ASSERT(span->end <= MAX_WIDTH);
 
    /* Fragment write masks */
    if (span->arrayMask & SPAN_MASK) {
@@ -1310,12 +1309,12 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    }
 
    /* Clip to window/scissor box */
-   if ((swrast->_RasterMask & CLIP_BIT) || (span->primitive != GL_POLYGON)) {
-      if (!clip_span(ctx, span)) {
-	 return;
-      }
+   if (!clip_span(ctx, span)) {
+      return;
    }
 
+   ASSERT(span->end <= MAX_WIDTH);
+
 #ifdef DEBUG
    /* Make sure all fragments are within window bounds */
    if (span->arrayMask & SPAN_XY) {
@@ -1356,15 +1355,6 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    if (ctx->Stencil._Enabled || ctx->Depth.Test) {
       if (!(span->arrayMask & SPAN_Z))
          _swrast_span_interpolate_z(ctx, span);
-
-      if ((span->arrayMask & SPAN_XY) == 0) {
-         if (span->x < fb->_Xmin || span->x + span->end > fb->_Xmax ||
-             span->y < fb->_Ymin || span->y >= fb->_Ymax) {
-            printf("Bad span clipping at %d, %d\n", span->x, span->y);
-            return;
-         }
-      }
-
       if (ctx->Stencil._Enabled) {
          /* Combined Z/stencil tests */
          if (!_swrast_stencil_and_ztest_span(ctx, span)) {
diff --git a/src/mesa/swrast/s_stencil.c b/src/mesa/swrast/s_stencil.c
index 2e84ddec71..e9e9d3a4f1 100644
--- a/src/mesa/swrast/s_stencil.c
+++ b/src/mesa/swrast/s_stencil.c
@@ -231,8 +231,9 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
    GLubyte fail[MAX_WIDTH];
    GLboolean allfail = GL_FALSE;
    GLuint i;
-   GLstencil r, s;
    const GLuint valueMask = ctx->Stencil.ValueMask[face];
+   const GLstencil r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+   GLstencil s;
 
    ASSERT(n <= MAX_WIDTH);
 
@@ -260,7 +261,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 allfail = GL_TRUE;
 	 break;
       case GL_LESS:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
@@ -279,7 +279,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 }
 	 break;
       case GL_LEQUAL:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
@@ -298,7 +297,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 }
 	 break;
       case GL_GREATER:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
@@ -317,7 +315,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 }
 	 break;
       case GL_GEQUAL:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
@@ -336,7 +333,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 }
 	 break;
       case GL_EQUAL:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
@@ -355,7 +351,6 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 	 }
 	 break;
       case GL_NOTEQUAL:
-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
 	 for (i=0;i<n;i++) {
 	    if (mask[i]) {
 	       s = (GLstencil) (stencil[i] & valueMask);
diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c
index dd59314cd9..6b1f934647 100644
--- a/src/mesa/swrast/s_texfilter.c
+++ b/src/mesa/swrast/s_texfilter.c
@@ -1329,7 +1329,7 @@ static void
 opt_sample_rgb_2d(GLcontext *ctx,
                   const struct gl_texture_object *tObj,
                   GLuint n, const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLchan rgba[][4])
+                  const GLfloat lambda[], GLfloat rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1351,9 +1351,9 @@ opt_sample_rgb_2d(GLcontext *ctx,
       GLint j = IFLOOR(texcoords[k][1] * height) & rowMask;
       GLint pos = (j << shift) | i;
       GLchan *texel = ((GLchan *) img->Data) + 3*pos;
-      rgba[k][RCOMP] = texel[0];
-      rgba[k][GCOMP] = texel[1];
-      rgba[k][BCOMP] = texel[2];
+      rgba[k][RCOMP] = CHAN_TO_FLOAT(texel[0]);
+      rgba[k][GCOMP] = CHAN_TO_FLOAT(texel[1]);
+      rgba[k][BCOMP] = CHAN_TO_FLOAT(texel[2]);
    }
 }
 
@@ -1370,7 +1370,7 @@ static void
 opt_sample_rgba_2d(GLcontext *ctx,
                    const struct gl_texture_object *tObj,
                    GLuint n, const GLfloat texcoords[][4],
-                   const GLfloat lambda[], GLchan rgba[][4])
+                   const GLfloat lambda[], GLfloat rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1392,7 +1392,10 @@ opt_sample_rgba_2d(GLcontext *ctx,
       const GLint row = IFLOOR(texcoords[i][1] * height) & rowMask;
       const GLint pos = (row << shift) | col;
       const GLchan *texel = ((GLchan *) img->Data) + (pos << 2);    /* pos*4 */
-      COPY_4V(rgba[i], texel);
+      rgba[i][RCOMP] = CHAN_TO_FLOAT(texel[0]);
+      rgba[i][GCOMP] = CHAN_TO_FLOAT(texel[1]);
+      rgba[i][BCOMP] = CHAN_TO_FLOAT(texel[2]);
+      rgba[i][ACOMP] = CHAN_TO_FLOAT(texel[3]);
    }
 }
 
@@ -1425,7 +1428,6 @@ sample_lambda_2d(GLcontext *ctx,
       case GL_NEAREST:
          if (repeatNoBorderPOT) {
             switch (tImg->TexFormat->MesaFormat) {
-#if 0
             case MESA_FORMAT_RGB:
                opt_sample_rgb_2d(ctx, tObj, m, texcoords + minStart,
                                  NULL, rgba + minStart);
@@ -1434,7 +1436,6 @@ sample_lambda_2d(GLcontext *ctx,
 	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + minStart,
                                   NULL, rgba + minStart);
                break;
-#endif
             default:
                sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
                                  NULL, rgba + minStart );
@@ -1484,7 +1485,6 @@ sample_lambda_2d(GLcontext *ctx,
       case GL_NEAREST:
          if (repeatNoBorderPOT) {
             switch (tImg->TexFormat->MesaFormat) {
-#if 0
             case MESA_FORMAT_RGB:
                opt_sample_rgb_2d(ctx, tObj, m, texcoords + magStart,
                                  NULL, rgba + magStart);
@@ -1493,7 +1493,6 @@ sample_lambda_2d(GLcontext *ctx,
 	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + magStart,
                                   NULL, rgba + magStart);
                break;
-#endif
             default:
                sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
                                  NULL, rgba + magStart );
@@ -3180,7 +3179,6 @@ _swrast_choose_texture_sample_func( GLcontext *ctx,
          }
          else {
             /* check for a few optimized cases */
-#if 0
             const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
             ASSERT(t->MinFilter == GL_NEAREST);
             if (t->WrapS == GL_REPEAT &&
@@ -3197,10 +3195,6 @@ _swrast_choose_texture_sample_func( GLcontext *ctx,
                      img->TexFormat->MesaFormat == MESA_FORMAT_RGBA) {
                return &opt_sample_rgba_2d;
             }
-#else
-            if (0)
-               ;
-#endif
             else {
                return &sample_nearest_2d;
             }
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index f69b122046..f2771cde09 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -81,7 +81,7 @@ _tnl_CreateContext( GLcontext *ctx )
    tnl->nr_blocks = 0;
 
    /* plug in the VBO drawing function */
-   vbo_set_draw_func(ctx, _tnl_draw_prims);
+   vbo_set_draw_func(ctx, _tnl_vbo_draw_prims);
 
    _math_init_transformation();
    _math_init_translate();
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
index 2ec65d5323..c64c2c2077 100644
--- a/src/mesa/tnl/t_draw.c
+++ b/src/mesa/tnl/t_draw.c
@@ -360,6 +360,20 @@ static void unmap_vbos( GLcontext *ctx,
 }
 
 
+void _tnl_vbo_draw_prims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLboolean index_bounds_valid,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+   if (!index_bounds_valid)
+      vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+
+   _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
 
 /* This is the main entrypoint into the slimmed-down software tnl
  * module.  In a regular swtnl driver, this can be plugged straight
@@ -393,7 +407,7 @@ void _tnl_draw_prims( GLcontext *ctx,
        */
       vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, 
 			min_index, max_index,
-			_tnl_draw_prims );
+			_tnl_vbo_draw_prims );
       return;
    }
    else if (max_index > max) {
@@ -411,7 +425,7 @@ void _tnl_draw_prims( GLcontext *ctx,
        */
       vbo_split_prims( ctx, arrays, prim, nr_prims, ib, 
 		       0, max_index,
-		       _tnl_draw_prims,
+		       _tnl_vbo_draw_prims,
 		       &limits );
    }
    else {
diff --git a/src/mesa/tnl/t_vb_cliptmp.h b/src/mesa/tnl/t_vb_cliptmp.h
index 788fe329ed..618b8b3130 100644
--- a/src/mesa/tnl/t_vb_cliptmp.h
+++ b/src/mesa/tnl/t_vb_cliptmp.h
@@ -127,7 +127,7 @@ TAG(clip_line)( GLcontext *ctx, GLuint v0, GLuint v1, GLubyte mask )
    GLuint p;
    const GLuint v0_orig = v0;
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       LINE_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       LINE_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       LINE_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -199,7 +199,24 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
 
    ASSIGN_3V(inlist, v2, v0, v1 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (0) {
+      /* print pre-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("pre clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+         assert(!IS_INF_OR_NAN(coord[j][0]));
+         assert(!IS_INF_OR_NAN(coord[j][1]));
+         assert(!IS_INF_OR_NAN(coord[j][2]));
+         assert(!IS_INF_OR_NAN(coord[j][3]));
+      }
+   }
+
+
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -227,6 +244,18 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
       }
    }
 
+   if (0) {
+      /* print post-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("post clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+      }
+   }
+
    tnl->Driver.Render.ClippedPolygon( ctx, inlist, n );
 }
 
@@ -250,7 +279,7 @@ TAG(clip_quad)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint v3,
 
    ASSIGN_4V(inlist, v3, v0, v1, v2 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index 1795f62c32..dc954bcba1 100644
--- a/src/mesa/tnl/t_vb_program.c
+++ b/src/mesa/tnl/t_vb_program.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.6
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -46,6 +47,16 @@
 #include "tnl/t_pipeline.h"
 
 
+#ifdef NAN_CHECK
+/** Check for NaNs and very large values */
+static INLINE void
+check_float(float x)
+{
+   assert(!IS_INF_OR_NAN(x));
+   assert(1.0e-15 <= x && x <= 1.0e15);
+}
+#endif
+
 
 /*!
  * Private storage for the vertex program pipeline stage.
@@ -207,7 +218,7 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine)
 {
    /* Input registers get initialized from the current vertex attribs */
    MEMCPY(machine->VertAttribs, ctx->Current.Attrib,
-          MAX_VERTEX_PROGRAM_ATTRIBS * 4 * sizeof(GLfloat));
+          MAX_VERTEX_GENERIC_ATTRIBS * 4 * sizeof(GLfloat));
 
    if (ctx->VertexProgram._Current->IsNVProgram) {
       GLuint i;
@@ -351,6 +362,12 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
 	    const GLuint size = VB->AttribPtr[attr]->size;
 	    const GLuint stride = VB->AttribPtr[attr]->stride;
 	    const GLfloat *data = (GLfloat *) (ptr + stride * i);
+#ifdef NAN_CHECK
+            check_float(data[0]);
+            check_float(data[1]);
+            check_float(data[2]);
+            check_float(data[3]);
+#endif
 	    COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
 	 }
       }
@@ -361,8 +378,17 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
       /* copy the output registers into the VB->attribs arrays */
       for (j = 0; j < numOutputs; j++) {
          const GLuint attr = outputs[j];
+#ifdef NAN_CHECK
+         check_float(machine.Outputs[attr][0]);
+         check_float(machine.Outputs[attr][1]);
+         check_float(machine.Outputs[attr][2]);
+         check_float(machine.Outputs[attr][3]);
+#endif
          COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
       }
+#ifdef NAN_CHECK
+      ASSERT(machine.Outputs[0][3] != 0.0F);
+#endif
 #if 0
       printf("HPOS: %f %f %f %f\n",
              machine.Outputs[0][0], 
diff --git a/src/mesa/tnl/t_vb_rendertmp.h b/src/mesa/tnl/t_vb_rendertmp.h
index 2b5f4e93b2..75f6f55bdc 100644
--- a/src/mesa/tnl/t_vb_rendertmp.h
+++ b/src/mesa/tnl/t_vb_rendertmp.h
@@ -82,7 +82,10 @@ static void TAG(render_lines)( GLcontext *ctx,
    INIT(GL_LINES);
    for (j=start+1; j<count; j+=2 ) {
       RESET_STIPPLE;
-      RENDER_LINE( ELT(j-1), ELT(j) );
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
    }
    POSTFIX;
 }
@@ -103,9 +106,12 @@ static void TAG(render_line_strip)( GLcontext *ctx,
       RESET_STIPPLE;
    }
 
-   for (j=start+1; j<count; j++ )
-      RENDER_LINE( ELT(j-1), ELT(j) );
-
+   for (j=start+1; j<count; j++ ) {
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
+   }
    POSTFIX;
 }
 
@@ -125,15 +131,24 @@ static void TAG(render_line_loop)( GLcontext *ctx,
    if (start+1 < count) {
       if (TEST_PRIM_BEGIN(flags)) {
 	 RESET_STIPPLE;
-	 RENDER_LINE( ELT(start), ELT(start+1) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(start), ELT(start+1) );
+         else
+            RENDER_LINE( ELT(start+1), ELT(start) );
       }
 
       for ( i = start+2 ; i < count ; i++) {
-	 RENDER_LINE( ELT(i-1), ELT(i) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(i-1), ELT(i) );
+         else
+            RENDER_LINE( ELT(i), ELT(i-1) );
       }
 
       if ( TEST_PRIM_END(flags)) {
-	 RENDER_LINE( ELT(count-1), ELT(start) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(count-1), ELT(start) );
+         else
+            RENDER_LINE( ELT(start), ELT(count-1) );
       }
    }
 
@@ -156,11 +171,17 @@ static void TAG(render_triangles)( GLcontext *ctx,
 	 /* Leave the edgeflags as supplied by the user.
 	  */
 	 RESET_STIPPLE;
-	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
       }
    } else {
       for (j=start+2; j<count; j+=3) {
-	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
       }
    }
    POSTFIX;
@@ -180,26 +201,38 @@ static void TAG(render_tri_strip)( GLcontext *ctx,
    INIT(GL_TRIANGLE_STRIP);
    if (NEED_EDGEFLAG_SETUP) {
       for (j=start+2;j<count;j++,parity^=1) {
-	 GLuint ej2 = ELT(j-2+parity);
-	 GLuint ej1 = ELT(j-1-parity);
-	 GLuint ej = ELT(j);
-	 GLboolean ef2 = EDGEFLAG_GET( ej2 );
-	 GLboolean ef1 = EDGEFLAG_GET( ej1 );
-	 GLboolean ef = EDGEFLAG_GET( ej );
+         GLuint ej2, ej1, ej;
+         GLboolean ef2, ef1, ef;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT) {
+            ej2 = ELT(j-2+parity);
+            ej1 = ELT(j-1-parity);
+            ej = ELT(j);
+         }
+         else {
+            ej2 = ELT(j-1+parity);
+            ej1 = ELT(j-parity);
+            ej = ELT(j-2);
+         }
+	 ef2 = EDGEFLAG_GET( ej2 );
+	 ef1 = EDGEFLAG_GET( ej1 );
+	 ef = EDGEFLAG_GET( ej );
 	 if (TEST_PRIM_BEGIN(flags)) {
 	    RESET_STIPPLE;
 	 }
 	 EDGEFLAG_SET( ej2, GL_TRUE );
 	 EDGEFLAG_SET( ej1, GL_TRUE );
 	 EDGEFLAG_SET( ej, GL_TRUE );
-	 RENDER_TRI( ej2, ej1, ej );
+         RENDER_TRI( ej2, ej1, ej );
 	 EDGEFLAG_SET( ej2, ef2 );
 	 EDGEFLAG_SET( ej1, ef1 );
 	 EDGEFLAG_SET( ej, ef );
       }
    } else {
       for (j=start+2; j<count ; j++, parity^=1) {
-	 RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1+parity), ELT(j-parity), ELT(j-2) );
       }
    }
    POSTFIX;
@@ -232,14 +265,20 @@ static void TAG(render_tri_fan)( GLcontext *ctx,
 	 EDGEFLAG_SET( ejs, GL_TRUE );
 	 EDGEFLAG_SET( ej1, GL_TRUE );
 	 EDGEFLAG_SET( ej, GL_TRUE );
-	 RENDER_TRI( ejs, ej1, ej);
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ejs, ej1, ej);
+         else
+            RENDER_TRI( ej, ejs, ej1);
 	 EDGEFLAG_SET( ejs, efs );
 	 EDGEFLAG_SET( ej1, ef1 );
 	 EDGEFLAG_SET( ej, ef );
       }
    } else {
       for (j=start+2;j<count;j++) {
-	 RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j), ELT(start), ELT(j-1) );
       }
    }
 
@@ -331,11 +370,19 @@ static void TAG(render_quads)( GLcontext *ctx,
 	 /* Use user-specified edgeflags for quads.
 	  */
 	 RESET_STIPPLE;
-	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
       }
    } else {
       for (j=start+3; j<count; j+=4) {
-	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
       }
    }
    POSTFIX;
@@ -367,7 +414,11 @@ static void TAG(render_quad_strip)( GLcontext *ctx,
 	 EDGEFLAG_SET( ELT(j-2), GL_TRUE );
 	 EDGEFLAG_SET( ELT(j-1), GL_TRUE );
 	 EDGEFLAG_SET( ELT(j), GL_TRUE );
-	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
 	 EDGEFLAG_SET( ELT(j-3), ef3 );
 	 EDGEFLAG_SET( ELT(j-2), ef2 );
 	 EDGEFLAG_SET( ELT(j-1), ef1 );
@@ -375,7 +426,11 @@ static void TAG(render_quad_strip)( GLcontext *ctx,
       }
    } else {
       for (j=start+3;j<count;j+=2) {
-	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
       }
    }
    POSTFIX;
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index 4d628aa9a6..9c66d3b019 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -81,6 +81,16 @@ _tnl_draw_prims( GLcontext *ctx,
 		 GLuint min_index,
 		 GLuint max_index);
 
+void
+_tnl_vbo_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
+		     GLuint min_index,
+		     GLuint max_index);
+
 extern void
 _mesa_load_tracked_matrices(GLcontext *ctx);
 
diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h
index 5362226c2f..5986e93576 100644
--- a/src/mesa/vbo/vbo.h
+++ b/src/mesa/vbo/vbo.h
@@ -69,6 +69,7 @@ typedef void (*vbo_draw_func)( GLcontext *ctx,
 			       const struct _mesa_prim *prims,
 			       GLuint nr_prims,
 			       const struct _mesa_index_buffer *ib,
+			       GLboolean index_bounds_valid,
 			       GLuint min_index,
 			       GLuint max_index );
 
@@ -112,7 +113,10 @@ void vbo_rebase_prims( GLcontext *ctx,
 		       GLuint min_index,
 		       GLuint max_index,
 		       vbo_draw_func draw );
-
+void
+vbo_get_minmax_index(GLcontext *ctx, const struct _mesa_prim *prim,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint *min_index, GLuint *max_index);
 
 void vbo_use_buffer_objects(GLcontext *ctx);
 
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
index ca8190fd05..75c32e0b9b 100644
--- a/src/mesa/vbo/vbo_context.c
+++ b/src/mesa/vbo/vbo_context.c
@@ -28,6 +28,7 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 #include "main/api_arrayelt.h"
+#include "main/bufferobj.h"
 #include "math/m_eval.h"
 #include "vbo.h"
 #include "vbo_context.h"
@@ -81,7 +82,8 @@ static void init_legacy_currval(GLcontext *ctx)
       cl->Type = GL_FLOAT;
       cl->Format = GL_RGBA;
       cl->Ptr = (const void *)ctx->Current.Attrib[i];
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -106,7 +108,8 @@ static void init_generic_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -150,7 +153,8 @@ static void init_mat_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -211,7 +215,7 @@ GLboolean _vbo_CreateContext( GLcontext *ctx )
       for (i = 0; i < 4; i++)
 	 vbo->map_vp_none[28+i] = i;	
       
-      for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      for (i = 0; i < Elements(vbo->map_vp_arb); i++)
 	 vbo->map_vp_arb[i] = i;
    }
 
diff --git a/src/mesa/vbo/vbo_context.h b/src/mesa/vbo/vbo_context.h
index 5c85161caa..8b726dc8ac 100644
--- a/src/mesa/vbo/vbo_context.h
+++ b/src/mesa/vbo/vbo_context.h
@@ -68,8 +68,8 @@ struct vbo_context {
    struct gl_client_array *generic_currval;
    struct gl_client_array *mat_currval;
 
-   GLuint map_vp_none[32];
-   GLuint map_vp_arb[32];
+   GLuint map_vp_none[VERT_ATTRIB_MAX];
+   GLuint map_vp_arb[VERT_ATTRIB_MAX];
 
    GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
    GLfloat CurrentFloatEdgeFlag;
@@ -92,13 +92,13 @@ static INLINE struct vbo_context *vbo_context(GLcontext *ctx)
    return (struct vbo_context *)(ctx->swtnl_im);
 }
 
-enum {
-   VP_NONE = 1,
-   VP_NV,
-   VP_ARB
-};
 
-static INLINE GLuint get_program_mode( GLcontext *ctx )
+/**
+ * Return VP_x token to indicate whether we're running fixed-function
+ * vertex transformation, an NV vertex program or ARB vertex program/shader.
+ */
+static INLINE enum vp_mode
+get_program_mode( GLcontext *ctx )
 {
    if (!ctx->VertexProgram._Current)
       return VP_NONE;
diff --git a/src/mesa/vbo/vbo_exec.h b/src/mesa/vbo/vbo_exec.h
index 100bb8a5de..e0f44892cf 100644
--- a/src/mesa/vbo/vbo_exec.h
+++ b/src/mesa/vbo/vbo_exec.h
@@ -48,6 +48,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define ERROR_ATTRIB 16
 
 
+/** Current vertex program mode */
+enum vp_mode {
+   VP_NONE,   /**< fixed function */
+   VP_NV,     /**< NV vertex program */
+   VP_ARB     /**< ARB vertex program or GLSL vertex shader */
+};
 
 
 struct vbo_exec_eval1_map {
@@ -103,7 +109,7 @@ struct vbo_exec_context
        * values are squashed down to the 32 attributes passed to the
        * vertex program below:
        */
-      GLuint program_mode;
+      enum vp_mode program_mode;
       GLuint enabled_flags;
       const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
    } vtx;
@@ -116,7 +122,7 @@ struct vbo_exec_context
    } eval;
 
    struct {
-      GLuint program_mode;
+      enum vp_mode program_mode;
       GLuint enabled_flags;
       GLuint array_obj;
 
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 34e849aaab..387d4ee3d4 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -486,23 +486,6 @@ static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j )
 }
 
 
-/**
- * Check if programs/shaders are enabled and valid at glBegin time.
- */
-GLboolean 
-vbo_validate_shaders(GLcontext *ctx)
-{
-   if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-       (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-      return GL_FALSE;
-   }
-   if (ctx->Shader.CurrentProgram && !ctx->Shader.CurrentProgram->LinkStatus) {
-      return GL_FALSE;
-   }
-   return GL_TRUE;
-}
-
-
 /* Build a list of primitives on the fly.  Keep
  * ctx->Driver.CurrentExecPrimitive uptodate as well.
  */
@@ -521,9 +504,7 @@ static void GLAPIENTRY vbo_exec_Begin( GLenum mode )
 	 return;
       }
 
-      if (!vbo_validate_shaders(ctx)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glBegin (invalid vertex/fragment program)");
+      if (!_mesa_valid_to_render(ctx, "glBegin")) {
          return;
       }
 
@@ -675,7 +656,7 @@ void vbo_use_buffer_objects(GLcontext *ctx)
    GLsizei size = VBO_VERT_BUFFER_SIZE;
 
    /* Make sure this func is only used once */
-   assert(exec->vtx.bufferobj == ctx->Array.NullBufferObj);
+   assert(exec->vtx.bufferobj == ctx->Shared->NullBufferObj);
    if (exec->vtx.buffer_map) {
       _mesa_align_free(exec->vtx.buffer_map);
       exec->vtx.buffer_map = NULL;
@@ -701,7 +682,7 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
     */
    _mesa_reference_buffer_object(ctx,
                                  &exec->vtx.bufferobj,
-                                 ctx->Array.NullBufferObj);
+                                 ctx->Shared->NullBufferObj);
 
    ASSERT(!exec->vtx.buffer_map);
    exec->vtx.buffer_map = (GLfloat *)ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE, 64);
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index de66cdd92f..12911f5750 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -1,6 +1,7 @@
 /**************************************************************************
  * 
  * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -32,20 +33,36 @@
 #include "main/api_noop.h"
 #include "main/varray.h"
 #include "main/bufferobj.h"
+#include "main/macros.h"
 #include "glapi/dispatch.h"
 
 #include "vbo_context.h"
 
-/* Compute min and max elements for drawelements calls.
+
+/**
+ * Compute min and max elements for glDraw[Range]Elements() calls.
  */
-static void get_minmax_index( GLuint count, GLuint type, 
-			      const GLvoid *indices,
-			      GLuint *min_index,
-			      GLuint *max_index)
+void
+vbo_get_minmax_index(GLcontext *ctx,
+		     const struct _mesa_prim *prim,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint *min_index, GLuint *max_index)
 {
    GLuint i;
+   GLsizei count = prim->count;
+   const void *indices;
+
+   if (_mesa_is_bufferobj(ib->obj)) {
+      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
+                                                GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                                GL_READ_ONLY,
+                                                ib->obj);
+      indices = ADD_POINTERS(map, ib->ptr);
+   } else {
+      indices = ib->ptr;
+   }
 
-   switch(type) {
+   switch (ib->type) {
    case GL_UNSIGNED_INT: {
       const GLuint *ui_indices = (const GLuint *)indices;
       GLuint max_ui = ui_indices[count-1];
@@ -86,12 +103,208 @@ static void get_minmax_index( GLuint count, GLuint type,
       assert(0);
       break;
    }
+
+   if (_mesa_is_bufferobj(ib->obj)) {
+      ctx->Driver.UnmapBuffer(ctx,
+			      GL_ELEMENT_ARRAY_BUFFER_ARB,
+			      ib->obj);
+   }
+}
+
+
+/**
+ * Check that element 'j' of the array has reasonable data.
+ * Map VBO if needed.
+ */
+static void
+check_array_data(GLcontext *ctx, struct gl_client_array *array,
+                 GLuint attrib, GLuint j)
+{
+   if (array->Enabled) {
+      const void *data = array->Ptr;
+      if (_mesa_is_bufferobj(array->BufferObj)) {
+         if (!array->BufferObj->Pointer) {
+            /* need to map now */
+            array->BufferObj->Pointer = ctx->Driver.MapBuffer(ctx,
+                                                              GL_ARRAY_BUFFER_ARB,
+                                                              GL_READ_ONLY,
+                                                              array->BufferObj);
+         }
+         data = ADD_POINTERS(data, array->BufferObj->Pointer);
+      }
+      switch (array->Type) {
+      case GL_FLOAT:
+         {
+            GLfloat *f = (GLfloat *) ((GLubyte *) data + array->StrideB * j);
+            GLuint k;
+            for (k = 0; k < array->Size; k++) {
+               if (IS_INF_OR_NAN(f[k]) ||
+                   f[k] >= 1.0e20 || f[k] <= -1.0e10) {
+                  _mesa_printf("Bad array data:\n");
+                  _mesa_printf("  Element[%u].%u = %f\n", j, k, f[k]);
+                  _mesa_printf("  Array %u at %p\n", attrib, (void* ) array);
+                  _mesa_printf("  Type 0x%x, Size %d, Stride %d\n",
+                               array->Type, array->Size, array->Stride);
+                  _mesa_printf("  Address/offset %p in Buffer Object %u\n",
+                               array->Ptr, array->BufferObj->Name);
+                  f[k] = 1.0; /* XXX replace the bad value! */
+               }
+               //assert(!IS_INF_OR_NAN(f[k]));
+            }
+         }
+         break;
+      default:
+         ;
+      }
+   }
+}
+
+
+/**
+ * Unmap the buffer object referenced by given array, if mapped.
+ */
+static void
+unmap_array_buffer(GLcontext *ctx, struct gl_client_array *array)
+{
+   if (array->Enabled &&
+       _mesa_is_bufferobj(array->BufferObj) &&
+       _mesa_bufferobj_mapped(array->BufferObj)) {
+      ctx->Driver.UnmapBuffer(ctx,
+                              GL_ARRAY_BUFFER_ARB,
+                              array->BufferObj);
+   }
 }
 
 
-/* Just translate the arrayobj into a sane layout.
+/**
+ * Examine the array's data for NaNs, etc.
  */
-static void bind_array_obj( GLcontext *ctx )
+static void
+check_draw_elements_data(GLcontext *ctx, GLsizei count, GLenum elemType,
+                         const void *elements)
+{
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   const void *elemMap;
+   GLint i, k;
+
+   if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
+      elemMap = ctx->Driver.MapBuffer(ctx,
+                                      GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                      GL_READ_ONLY,
+                                      ctx->Array.ElementArrayBufferObj);
+      elements = ADD_POINTERS(elements, elemMap);
+   }
+
+   for (i = 0; i < count; i++) {
+      GLuint j;
+
+      /* j = element[i] */
+      switch (elemType) {
+      case GL_UNSIGNED_BYTE:
+         j = ((const GLubyte *) elements)[i];
+         break;
+      case GL_UNSIGNED_SHORT:
+         j = ((const GLushort *) elements)[i];
+         break;
+      case GL_UNSIGNED_INT:
+         j = ((const GLuint *) elements)[i];
+         break;
+      default:
+         assert(0);
+      }
+
+      /* check element j of each enabled array */
+      check_array_data(ctx, &arrayObj->Vertex, VERT_ATTRIB_POS, j);
+      check_array_data(ctx, &arrayObj->Normal, VERT_ATTRIB_NORMAL, j);
+      check_array_data(ctx, &arrayObj->Color, VERT_ATTRIB_COLOR0, j);
+      check_array_data(ctx, &arrayObj->SecondaryColor, VERT_ATTRIB_COLOR1, j);
+      for (k = 0; k < Elements(arrayObj->TexCoord); k++) {
+         check_array_data(ctx, &arrayObj->TexCoord[k], VERT_ATTRIB_TEX0 + k, j);
+      }
+      for (k = 0; k < Elements(arrayObj->VertexAttrib); k++) {
+         check_array_data(ctx, &arrayObj->VertexAttrib[k], VERT_ATTRIB_GENERIC0 + k, j);
+      }
+   }
+
+   if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
+      ctx->Driver.UnmapBuffer(ctx,
+			      GL_ELEMENT_ARRAY_BUFFER_ARB,
+			      ctx->Array.ElementArrayBufferObj);
+   }
+
+   unmap_array_buffer(ctx, &arrayObj->Vertex);
+   unmap_array_buffer(ctx, &arrayObj->Normal);
+   unmap_array_buffer(ctx, &arrayObj->Color);
+   for (k = 0; k < Elements(arrayObj->TexCoord); k++) {
+      unmap_array_buffer(ctx, &arrayObj->TexCoord[k]);
+   }
+   for (k = 0; k < Elements(arrayObj->VertexAttrib); k++) {
+      unmap_array_buffer(ctx, &arrayObj->VertexAttrib[k]);
+   }
+}
+
+
+/**
+ * Check array data, looking for NaNs, etc.
+ */
+static void
+check_draw_arrays_data(GLcontext *ctx, GLint start, GLsizei count)
+{
+   /* TO DO */
+}
+
+
+/**
+ * Print info/data for glDrawArrays().
+ */
+static void
+print_draw_arrays(GLcontext *ctx, struct vbo_exec_context *exec,
+                  GLenum mode, GLint start, GLsizei count)
+{
+   int i;
+
+   _mesa_printf("vbo_exec_DrawArrays(mode 0x%x, start %d, count %d):\n",
+                mode, start, count);
+
+   for (i = 0; i < 32; i++) {
+      GLuint bufName = exec->array.inputs[i]->BufferObj->Name;
+      GLint stride = exec->array.inputs[i]->Stride;
+      _mesa_printf("attr %2d: size %d stride %d  enabled %d  "
+                   "ptr %p  Bufobj %u\n",
+                   i,
+                   exec->array.inputs[i]->Size,
+                   stride,
+                   /*exec->array.inputs[i]->Enabled,*/
+                   exec->array.legacy_array[i]->Enabled,
+                   exec->array.inputs[i]->Ptr,
+                   bufName);
+
+      if (bufName) {
+         struct gl_buffer_object *buf = _mesa_lookup_bufferobj(ctx, bufName);
+         GLubyte *p = ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
+                                            GL_READ_ONLY_ARB, buf);
+         int offset = (int) (GLintptr) exec->array.inputs[i]->Ptr;
+         float *f = (float *) (p + offset);
+         int *k = (int *) f;
+         int i;
+         int n = (count * stride) / 4;
+         if (n > 32)
+            n = 32;
+         _mesa_printf("  Data at offset %d:\n", offset);
+         for (i = 0; i < n; i++) {
+            _mesa_printf("    float[%d] = 0x%08x %f\n", i, k[i], f[i]);
+         }
+         ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, buf);
+      }
+   }
+}
+
+
+/**
+ * Just translate the arrayobj into a sane layout.
+ */
+static void
+bind_array_obj(GLcontext *ctx)
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
@@ -103,7 +316,7 @@ static void bind_array_obj( GLcontext *ctx )
     * go away.
     */
    exec->array.legacy_array[VERT_ATTRIB_POS] = &arrayObj->Vertex;
-   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &vbo->legacy_currval[VERT_ATTRIB_WEIGHT];
+   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &arrayObj->Weight;
    exec->array.legacy_array[VERT_ATTRIB_NORMAL] = &arrayObj->Normal;
    exec->array.legacy_array[VERT_ATTRIB_COLOR0] = &arrayObj->Color;
    exec->array.legacy_array[VERT_ATTRIB_COLOR1] = &arrayObj->SecondaryColor;
@@ -115,11 +328,10 @@ static void bind_array_obj( GLcontext *ctx )
    }
    exec->array.legacy_array[VERT_ATTRIB_EDGEFLAG] = &arrayObj->EdgeFlag;
 
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
       exec->array.legacy_array[VERT_ATTRIB_TEX0 + i] = &arrayObj->TexCoord[i];
 
-   for (i = 0; i < MAX_VERTEX_GENERIC_ATTRIBS; i++) {
-      assert(i < Elements(arrayObj->VertexAttrib));
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++) {
       assert(i < Elements(exec->array.generic_array));
       exec->array.generic_array[i] = &arrayObj->VertexAttrib[i];
    }
@@ -127,7 +339,9 @@ static void bind_array_obj( GLcontext *ctx )
    exec->array.array_obj = arrayObj->Name;
 }
 
-static void recalculate_input_bindings( GLcontext *ctx )
+
+static void
+recalculate_input_bindings(GLcontext *ctx)
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
@@ -140,9 +354,10 @@ static void recalculate_input_bindings( GLcontext *ctx )
 
    switch (exec->array.program_mode) {
    case VP_NONE:
-      /* When no vertex program is active, we put the material values
-       * into the generic slots.  This is the only situation where
-       * material values are available as per-vertex attributes.
+      /* When no vertex program is active (or the vertex program is generated
+       * from fixed-function state).  We put the material values into the
+       * generic slots.  This is the only situation where material values
+       * are available as per-vertex attributes.
        */
       for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
@@ -165,8 +380,8 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
          const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
       }
-
       break;
+
    case VP_NV:
       /* NV_vertex_program - attribute arrays alias and override
        * conventional, legacy arrays.  No materials, and the generic
@@ -190,11 +405,11 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
          const_inputs |= 1 << i;
       }
-
       break;
+
    case VP_ARB:
-      /* ARB_vertex_program - Only the attribute zero (position) array
-       * aliases and overrides the legacy position array.  
+      /* GL_ARB_vertex_program or GLSL vertex shader - Only the generic[0]
+       * attribute array aliases and overrides the legacy position array.  
        *
        * Otherwise, legacy attributes available in the legacy slots,
        * generic attributes in the generic slots and materials are not
@@ -209,7 +424,6 @@ static void recalculate_input_bindings( GLcontext *ctx )
          const_inputs |= 1 << 0;
       }
 
-
       for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
@@ -234,7 +448,9 @@ static void recalculate_input_bindings( GLcontext *ctx )
    _mesa_set_varying_vp_inputs( ctx, ~const_inputs );
 }
 
-static void bind_arrays( GLcontext *ctx )
+
+static void
+bind_arrays(GLcontext *ctx)
 {
 #if 0
    if (ctx->Array.ArrayObj.Name != exec->array.array_obj) {
@@ -274,14 +490,22 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
    if (ctx->NewState)
       _mesa_update_state( ctx );
       
-   if (!vbo_validate_shaders(ctx)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawArrays(bad shader)");
+   if (!_mesa_valid_to_render(ctx, "glDrawArrays")) {
       return;
    }
 
+#if 0
+   check_draw_arrays_data(ctx, start, count);
+#else
+   (void) check_draw_arrays_data;
+#endif
+
    bind_arrays( ctx );
 
-   /* Again...
+   /* Again... because we may have changed the bitmask of per-vertex varying
+    * attributes.  If we regenerate the fixed-function vertex program now
+    * we may be able to prune down the number of vertex attributes which we
+    * need in the shader.
     */
    if (ctx->NewState)
       _mesa_update_state( ctx );
@@ -295,83 +519,102 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
    prim[0].count = count;
    prim[0].indexed = 0;
 
-   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count - 1 );
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL,
+                    GL_TRUE, start, start + count - 1 );
 
 #if 0
-   {
-      int i;
-
-      _mesa_printf("vbo_exec_DrawArrays(mode 0x%x, start %d, count %d):\n",
-                   mode, start, count);
-
-      for (i = 0; i < 32; i++) {
-         GLuint bufName = exec->array.inputs[i]->BufferObj->Name;
-         GLint stride = exec->array.inputs[i]->Stride;
-         _mesa_printf("attr %2d: size %d stride %d  enabled %d  "
-                      "ptr %p  Bufobj %u\n",
-                      i,
-                      exec->array.inputs[i]->Size,
-                      stride,
-                      /*exec->array.inputs[i]->Enabled,*/
-                      exec->array.legacy_array[i]->Enabled,
-                      exec->array.inputs[i]->Ptr,
-                      bufName);
-         
-         if (bufName) {
-            struct gl_buffer_object *buf = _mesa_lookup_bufferobj(ctx, bufName);
-            GLubyte *p = ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
-                                            GL_READ_ONLY_ARB, buf);
-            int offset = (int) exec->array.inputs[i]->Ptr;
-            float *f = (float *) (p + offset);
-            int *k = (int *) f;
-            int i;
-            int n = (count * stride) / 4;
-            if (n > 32)
-               n = 32;
-            _mesa_printf("  Data at offset %d:\n", offset);
-            for (i = 0; i < n; i++) {
-               _mesa_printf("    float[%d] = 0x%08x %f\n", i, k[i], f[i]);
-            }
-            ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, buf);
-         }
-      }
-   }
+   print_draw_arrays(ctx, exec, mode, start, count);
+#else
+   (void) print_draw_arrays;
 #endif
 }
 
 
+/**
+ * Map GL_ELEMENT_ARRAY_BUFFER and print contents.
+ */
+static void
+dump_element_buffer(GLcontext *ctx, GLenum type)
+{
+   const GLvoid *map = ctx->Driver.MapBuffer(ctx,
+                                             GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                             GL_READ_ONLY,
+                                             ctx->Array.ElementArrayBufferObj);
+   switch (type) {
+   case GL_UNSIGNED_BYTE:
+      {
+         const GLubyte *us = (const GLubyte *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size; i++) {
+            _mesa_printf("%02x ", us[i]);
+            if (i % 32 == 31)
+               _mesa_printf("\n");
+         }
+         _mesa_printf("\n");
+      }
+      break;
+   case GL_UNSIGNED_SHORT:
+      {
+         const GLushort *us = (const GLushort *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size / 2; i++) {
+            _mesa_printf("%04x ", us[i]);
+            if (i % 16 == 15)
+               _mesa_printf("\n");
+         }
+         _mesa_printf("\n");
+      }
+      break;
+   case GL_UNSIGNED_INT:
+      {
+         const GLuint *us = (const GLuint *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size / 4; i++) {
+            _mesa_printf("%08x ", us[i]);
+            if (i % 8 == 7)
+               _mesa_printf("\n");
+         }
+         _mesa_printf("\n");
+      }
+      break;
+   default:
+      ;
+   }
+
+   ctx->Driver.UnmapBuffer(ctx,
+                           GL_ELEMENT_ARRAY_BUFFER_ARB,
+                           ctx->Array.ElementArrayBufferObj);
+}
 
-static void GLAPIENTRY
-vbo_exec_DrawRangeElements(GLenum mode,
-			   GLuint start, GLuint end,
-			   GLsizei count, GLenum type, const GLvoid *indices)
+/* Inner support for both _mesa_DrawElements and _mesa_DrawRangeElements */
+static void
+vbo_validated_drawrangeelements(GLcontext *ctx, GLenum mode,
+				GLboolean index_bounds_valid,
+				GLuint start, GLuint end,
+				GLsizei count, GLenum type,
+				const GLvoid *indices)
 {
-   GET_CURRENT_CONTEXT(ctx);
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
    struct _mesa_index_buffer ib;
    struct _mesa_prim prim[1];
 
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
-      return;
-
    FLUSH_CURRENT( ctx, 0 );
 
    if (ctx->NewState)
       _mesa_update_state( ctx );
 
-   if (!vbo_validate_shaders(ctx)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawRangeElements(bad shader)");
+   if (!_mesa_valid_to_render(ctx, "glDraw[Range]Elements")) {
       return;
    }
 
-   bind_arrays( ctx );
-
    if (ctx->NewState)
       _mesa_update_state( ctx );
 
+   bind_arrays( ctx );
+
    ib.count = count;
-   ib.type = type; 
+   ib.type = type;
    ib.obj = ctx->Array.ElementArrayBufferObj;
    ib.ptr = indices;
 
@@ -415,66 +658,240 @@ vbo_exec_DrawRangeElements(GLenum mode,
     * for the latter case elsewhere.
     */
 
-   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, &ib, start, end );
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, &ib,
+		    index_bounds_valid, start, end );
+}
+
+static void GLAPIENTRY
+vbo_exec_DrawRangeElements(GLenum mode,
+			   GLuint start, GLuint end,
+			   GLsizei count, GLenum type, const GLvoid *indices)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count,
+                                          type, indices ))
+      return;
+
+   if (end >= ctx->Array.ArrayObj->_MaxElement) {
+      /* the max element is out of bounds of one or more enabled arrays */
+      _mesa_warning(ctx, "glDraw[Range]Elements(start %u, end %u, count %d, "
+                    "type 0x%x, indices=%p)\n"
+                    "\tindex=%u is out of bounds (max=%u)  "
+                    "Element Buffer %u (size %d)",
+                    start, end, count, type, indices, end,
+                    ctx->Array.ArrayObj->_MaxElement - 1,
+                    ctx->Array.ElementArrayBufferObj->Name,
+                    ctx->Array.ElementArrayBufferObj->Size);
+
+      if (0)
+         dump_element_buffer(ctx, type);
+
+      if (0)
+         _mesa_print_arrays(ctx);
+      return;
+   }
+   else if (0) {
+      _mesa_printf("glDraw[Range]Elements"
+                   "(start %u, end %u, type 0x%x, count %d) ElemBuf %u\n",
+                   start, end, type, count,
+                   ctx->Array.ElementArrayBufferObj->Name);
+   }
+
+#if 0
+   check_draw_elements_data(ctx, count, type, indices);
+#else
+   (void) check_draw_elements_data;
+#endif
+
+   vbo_validated_drawrangeelements(ctx, mode, GL_TRUE, start, end,
+				   count, type, indices);
 }
 
+
 static void GLAPIENTRY
-vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
+vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                      const GLvoid *indices)
 {
    GET_CURRENT_CONTEXT(ctx);
-   GLuint min_index = 0;
-   GLuint max_index = 0;
 
    if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
       return;
 
-   if (!vbo_validate_shaders(ctx)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawElements(bad shader)");
+   vbo_validated_drawrangeelements(ctx, mode, GL_FALSE, ~0, ~0,
+				   count, type, indices);
+}
+
+/* Inner support for both _mesa_DrawElements and _mesa_DrawRangeElements */
+static void
+vbo_validated_multidrawelements(GLcontext *ctx, GLenum mode,
+				const GLsizei *count, GLenum type,
+				const GLvoid **indices, GLsizei primcount)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct _mesa_index_buffer ib;
+   struct _mesa_prim *prim;
+   unsigned int index_type_size = 0;
+   uintptr_t min_index_ptr, max_index_ptr;
+   GLboolean fallback = GL_FALSE;
+   int i;
+
+   if (primcount == 0)
+      return;
+
+   FLUSH_CURRENT( ctx, 0 );
+
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
+   if (!_mesa_valid_to_render(ctx, "glMultiDrawElements")) {
       return;
    }
 
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
-						 GL_ELEMENT_ARRAY_BUFFER_ARB,
-						 GL_READ_ONLY,
-						 ctx->Array.ElementArrayBufferObj);
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
+   prim = _mesa_calloc(primcount * sizeof(*prim));
+   if (prim == NULL) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMultiDrawElements");
+      return;
+   }
 
-      get_minmax_index(count, type, ADD_POINTERS(map, indices), &min_index, &max_index);
+   /* Decide if we can do this all as one set of primitives sharing the
+    * same index buffer, or if we have to reset the index pointer per primitive.
+    */
+   bind_arrays( ctx );
 
-      ctx->Driver.UnmapBuffer(ctx,
-			      GL_ELEMENT_ARRAY_BUFFER_ARB,
-			      ctx->Array.ElementArrayBufferObj);
+   switch (type) {
+   case GL_UNSIGNED_INT:
+      index_type_size = 4;
+      break;
+   case GL_UNSIGNED_SHORT:
+      index_type_size = 2;
+      break;
+   case GL_UNSIGNED_BYTE:
+      index_type_size = 1;
+      break;
+   default:
+      assert(0);
    }
-   else {
-      get_minmax_index(count, type, indices, &min_index, &max_index);
+
+   min_index_ptr = (uintptr_t)indices[0];
+   max_index_ptr = 0;
+   for (i = 0; i < primcount; i++) {
+      min_index_ptr = MIN2(min_index_ptr, (uintptr_t)indices[i]);
+      max_index_ptr = MAX2(max_index_ptr, (uintptr_t)indices[i] +
+			   index_type_size * count[i]);
    }
 
-   vbo_exec_DrawRangeElements(mode, min_index, max_index, count, type, indices);
+   /* Check if we can handle this thing as a bunch of index offsets from the
+    * same index pointer.  If we can't, then we have to fall back to doing
+    * a draw_prims per primitive.
+    */
+   if (index_type_size != 1) {
+      for (i = 0; i < primcount; i++) {
+	 if ((((uintptr_t)indices[i] - min_index_ptr) % index_type_size) != 0) {
+	    fallback = GL_TRUE;
+	    break;
+	 }
+      }
+   }
+
+   /* If the index buffer isn't in a VBO, then treating the application's
+    * subranges of the index buffer as one large index buffer may lead to
+    * us reading unmapped memory.
+    */
+   if (!_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj))
+      fallback = GL_TRUE;
+
+   if (!fallback) {
+      ib.count = (max_index_ptr - min_index_ptr) / index_type_size;
+      ib.type = type;
+      ib.obj = ctx->Array.ElementArrayBufferObj;
+      ib.ptr = (void *)min_index_ptr;
+
+      for (i = 0; i < primcount; i++) {
+	 prim[i].begin = (i == 0);
+	 prim[i].end = (i == primcount - 1);
+	 prim[i].weak = 0;
+	 prim[i].pad = 0;
+	 prim[i].mode = mode;
+	 prim[i].start = ((uintptr_t)indices[i] - min_index_ptr) / index_type_size;
+	 prim[i].count = count[i];
+	 prim[i].indexed = 1;
+      }
+
+      vbo->draw_prims(ctx, exec->array.inputs, prim, primcount, &ib,
+		      GL_FALSE, ~0, ~0);
+   } else {
+      for (i = 0; i < primcount; i++) {
+	 ib.count = count[i];
+	 ib.type = type;
+	 ib.obj = ctx->Array.ElementArrayBufferObj;
+	 ib.ptr = indices[i];
+
+
+	 prim[0].begin = 1;
+	 prim[0].end = 1;
+	 prim[0].weak = 0;
+	 prim[0].pad = 0;
+	 prim[0].mode = mode;
+	 prim[0].start = 0;
+	 prim[0].count = count[i];
+	 prim[0].indexed = 1;
+      }
+
+      vbo->draw_prims(ctx, exec->array.inputs, prim, 1, &ib,
+		      GL_FALSE, ~0, ~0);
+   }
+   _mesa_free(prim);
 }
 
+static void GLAPIENTRY
+vbo_exec_MultiDrawElements(GLenum mode,
+			   const GLsizei *count, GLenum type,
+			   const GLvoid **indices,
+			   GLsizei primcount)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLint i;
 
-/***********************************************************************
- * Initialization
- */
+   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
+   for (i = 0; i < primcount; i++) {
+      if (!_mesa_validate_DrawElements( ctx, mode, count[i], type, indices[i] ))
+	 return;
+   }
+
+   vbo_validated_multidrawelements(ctx, mode, count, type, indices, primcount);
+}
 
 
 
-void vbo_exec_array_init( struct vbo_exec_context *exec )
+/***********************************************************************
+ * Initialization
+ */
+
+void
+vbo_exec_array_init( struct vbo_exec_context *exec )
 {
 #if 1
    exec->vtxfmt.DrawArrays = vbo_exec_DrawArrays;
    exec->vtxfmt.DrawElements = vbo_exec_DrawElements;
    exec->vtxfmt.DrawRangeElements = vbo_exec_DrawRangeElements;
+   exec->vtxfmt.MultiDrawElementsEXT = vbo_exec_MultiDrawElements;
 #else
    exec->vtxfmt.DrawArrays = _mesa_noop_DrawArrays;
    exec->vtxfmt.DrawElements = _mesa_noop_DrawElements;
    exec->vtxfmt.DrawRangeElements = _mesa_noop_DrawRangeElements;
+   exec->vtxfmt.MultiDrawElementsEXT = _mesa_noop_MultiDrawElements;
 #endif
 }
 
 
-void vbo_exec_array_destroy( struct vbo_exec_context *exec )
+void
+vbo_exec_array_destroy( struct vbo_exec_context *exec )
 {
    /* nothing to do */
 }
@@ -504,3 +921,11 @@ _mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
 {
    vbo_exec_DrawRangeElements(mode, start, end, count, type, indices);
 }
+
+/* GL_EXT_multi_draw_arrays */
+void GLAPIENTRY
+_mesa_MultiDrawElementsEXT(GLenum mode, const GLsizei *count, GLenum type,
+			   const GLvoid **indices, GLsizei primcount)
+{
+   vbo_exec_MultiDrawElements(mode, count, type, indices, primcount);
+}
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 98177a4175..0c258c535e 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -35,7 +35,8 @@
 #include "vbo_context.h"
 
 
-static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
+static void
+vbo_exec_debug_verts( struct vbo_exec_context *exec )
 {
    GLuint count = exec->vtx.vert_count;
    GLuint i;
@@ -64,19 +65,19 @@ static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
  * NOTE: Need to have calculated primitives by this point -- do it on the fly.
  * NOTE: Old 'parity' issue is gone.
  */
-static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
+static GLuint
+vbo_copy_vertices( struct vbo_exec_context *exec )
 {
    GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
    GLuint ovf, i;
    GLuint sz = exec->vtx.vertex_size;
    GLfloat *dst = exec->vtx.copied.buffer;
-   GLfloat *src = (exec->vtx.buffer_map + 
-		   exec->vtx.prim[exec->vtx.prim_count-1].start * 
-		   exec->vtx.vertex_size);
+   const GLfloat *src = (exec->vtx.buffer_map + 
+                         exec->vtx.prim[exec->vtx.prim_count-1].start * 
+                         exec->vtx.vertex_size);
 
 
-   switch( exec->ctx->Driver.CurrentExecPrimitive )
-   {
+   switch (exec->ctx->Driver.CurrentExecPrimitive) {
    case GL_POINTS:
       return 0;
    case GL_LINES:
@@ -95,8 +96,9 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
       return i;
    case GL_LINE_STRIP:
-      if (nr == 0) 
+      if (nr == 0) {
 	 return 0;
+      }
       else {
 	 _mesa_memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
 	 return 1;
@@ -104,12 +106,14 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
    case GL_LINE_LOOP:
    case GL_TRIANGLE_FAN:
    case GL_POLYGON:
-      if (nr == 0) 
+      if (nr == 0) {
 	 return 0;
+      }
       else if (nr == 1) {
 	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
 	 return 1;
-      } else {
+      }
+      else {
 	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
 	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
 	 return 2;
@@ -122,9 +126,15 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
       /* fallthrough */
    case GL_QUAD_STRIP:
       switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
+      case 0:
+         ovf = 0;
+         break;
+      case 1:
+         ovf = 1;
+         break;
+      default:
+         ovf = 2 + (nr & 1);
+         break;
       }
       for (i = 0 ; i < ovf ; i++)
 	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
@@ -141,13 +151,14 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 
 /* TODO: populate these as the vertex is defined:
  */
-static void vbo_exec_bind_arrays( GLcontext *ctx )
+static void
+vbo_exec_bind_arrays( GLcontext *ctx )
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
    struct gl_client_array *arrays = exec->vtx.arrays;
-   GLuint count = exec->vtx.vert_count;
-   GLubyte *data = (GLubyte *)exec->vtx.buffer_map;
+   const GLuint count = exec->vtx.vert_count;
+   const GLubyte *data = (GLubyte *) exec->vtx.buffer_map;
    const GLuint *map;
    GLuint attr;
    GLbitfield varying_inputs = 0x0;
@@ -203,11 +214,13 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
          /* override the default array set above */
          exec->vtx.inputs[attr] = &arrays[attr];
 
-         if (exec->vtx.bufferobj->Name) {
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
             /* a real buffer obj: Ptr is an offset, not a pointer*/
             GLsizeiptr offset;
             assert(exec->vtx.bufferobj->Pointer);  /* buf should be mapped */
-            offset = (GLbyte *) data - (GLbyte *) exec->vtx.bufferobj->Pointer;
+            offset = (GLbyte *) data -
+	       (GLbyte *) exec->vtx.bufferobj->Pointer +
+	       exec->vtx.bufferobj->Offset;
             assert(offset >= 0);
             arrays[attr].Ptr = (void *) offset;
          }
@@ -227,7 +240,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 	 arrays[attr]._MaxElement = count; /* ??? */
 
 	 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
-         varying_inputs |= 1<<attr;
+         varying_inputs |= 1 << attr;
       }
    }
 
@@ -235,18 +248,19 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 }
 
 
-static void vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
+static void
+vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
 {
    GLenum target = GL_ARRAY_BUFFER_ARB;
 
-   if (exec->vtx.bufferobj->Name) {
+   if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
       GLcontext *ctx = exec->ctx;
       
-      if(ctx->Driver.FlushMappedBufferRange) {
+      if (ctx->Driver.FlushMappedBufferRange) {
          GLintptr offset = exec->vtx.buffer_used - exec->vtx.bufferobj->Offset;
          GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) * sizeof(float);
 
-         if(length)
+         if (length)
             ctx->Driver.FlushMappedBufferRange(ctx, target,
                                                offset, length,
                                                exec->vtx.bufferobj);
@@ -265,7 +279,9 @@ static void vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
    }
 }
 
-void vbo_exec_vtx_map( struct vbo_exec_context *exec )
+
+void
+vbo_exec_vtx_map( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
    const GLenum target = GL_ARRAY_BUFFER_ARB;
@@ -277,7 +293,7 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
                               MESA_MAP_NOWAIT_BIT;
    const GLenum usage = GL_STREAM_DRAW_ARB;
    
-   if (exec->vtx.bufferobj->Name == 0)
+   if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
       return;
 
    if (exec->vtx.buffer_map != NULL) {
@@ -287,8 +303,7 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
    }
 
    if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024 &&
-       ctx->Driver.MapBufferRange)
-   {
+       ctx->Driver.MapBufferRange) {
       exec->vtx.buffer_map = 
          (GLfloat *)ctx->Driver.MapBufferRange(ctx, 
                                                target, 
@@ -314,13 +329,15 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
                                                   0, VBO_VERT_BUFFER_SIZE,
                                                   accessRange,
                                                   exec->vtx.bufferobj);
-      else
+      if (!exec->vtx.buffer_map)
          exec->vtx.buffer_map =
             (GLfloat *)ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj);
+      assert(exec->vtx.buffer_map);
       exec->vtx.buffer_ptr = exec->vtx.buffer_map;
    }
 
-   if (0) _mesa_printf("map %d..\n", exec->vtx.buffer_used);
+   if (0)
+      _mesa_printf("map %d..\n", exec->vtx.buffer_used);
 }
 
 
@@ -328,13 +345,12 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
 /**
  * Execute the buffer and save copied verts.
  */
-void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
-                         GLboolean unmap )
+void
+vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap )
 {
    if (0)
       vbo_exec_debug_verts( exec );
 
-
    if (exec->vtx.prim_count && 
        exec->vtx.vert_count) {
 
@@ -351,33 +367,35 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
          if (ctx->NewState)
             _mesa_update_state( ctx );
 
-         if (exec->vtx.bufferobj->Name) {
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
             vbo_exec_vtx_unmap( exec );
          }
 
-         if (0) _mesa_printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
-                      exec->vtx.vert_count);
+         if (0)
+            _mesa_printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
+                         exec->vtx.vert_count);
 
 	 vbo_context(ctx)->draw_prims( ctx, 
 				       exec->vtx.inputs, 
 				       exec->vtx.prim, 
 				       exec->vtx.prim_count,
 				       NULL,
+				       GL_TRUE,
 				       0,
 				       exec->vtx.vert_count - 1);
 
 	 /* If using a real VBO, get new storage -- unless asked not to.
           */
-         if (exec->vtx.bufferobj->Name && !unmap) {
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !unmap) {
             vbo_exec_vtx_map( exec );
-          }
+         }
       }
    }
 
    /* May have to unmap explicitly if we didn't draw:
     */
    if (unmap && 
-       exec->vtx.bufferobj->Name &&
+       _mesa_is_bufferobj(exec->vtx.bufferobj) &&
        exec->vtx.buffer_map) {
       vbo_exec_vtx_unmap( exec );
    }
@@ -389,7 +407,6 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
       exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) / 
                             (exec->vtx.vertex_size * sizeof(GLfloat)));
 
-
    exec->vtx.buffer_ptr = exec->vtx.buffer_map;
    exec->vtx.prim_count = 0;
    exec->vtx.vert_count = 0;
diff --git a/src/mesa/vbo/vbo_rebase.c b/src/mesa/vbo/vbo_rebase.c
index dae778e741..3bf7ef580f 100644
--- a/src/mesa/vbo/vbo_rebase.c
+++ b/src/mesa/vbo/vbo_rebase.c
@@ -161,7 +161,7 @@ void vbo_rebase_prims( GLcontext *ctx,
 				 GL_ELEMENT_ARRAY_BUFFER,
 				 ib->obj);
 
-      tmp_ib.obj = ctx->Array.NullBufferObj;
+      tmp_ib.obj = ctx->Shared->NullBufferObj;
       tmp_ib.ptr = tmp_indices;
       tmp_ib.count = ib->count;
       tmp_ib.type = ib->type;
@@ -208,6 +208,7 @@ void vbo_rebase_prims( GLcontext *ctx,
 	 prim, 
 	 nr_prims, 
 	 ib, 
+	 GL_TRUE,
 	 0, 
 	 max_index - min_index );
    
diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c
index a7f2706206..1771510d84 100644
--- a/src/mesa/vbo/vbo_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -73,6 +73,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/dlist.h"
 #include "main/enums.h"
 #include "main/macros.h"
+#include "main/api_noop.h"
 #include "main/api_validate.h"
 #include "main/api_arrayelt.h"
 #include "main/vtxfmt.h"
@@ -911,7 +912,7 @@ static void GLAPIENTRY _save_OBE_DrawElements(GLenum mode, GLsizei count, GLenum
 
    _ae_map_vbos( ctx );
 
-   if (ctx->Array.ElementArrayBufferObj->Name)
+   if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj))
       indices = ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Pointer, indices);
 
    vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK );
@@ -1038,6 +1039,8 @@ static void _save_vtxfmt_init( GLcontext *ctx )
    vfmt->DrawArrays = _save_DrawArrays;
    vfmt->DrawElements = _save_DrawElements;
    vfmt->DrawRangeElements = _save_DrawRangeElements;
+   /* Loops back into vfmt->DrawElements */
+   vfmt->MultiDrawElementsEXT = _mesa_noop_MultiDrawElements;
 
 }
 
@@ -1158,10 +1161,10 @@ static void vbo_print_vertex_list( GLcontext *ctx, void *data )
    GLuint i;
    (void) ctx;
 
-   _mesa_debug(NULL, "VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
-               node->count,
-	       node->prim_count,
-	       node->vertex_size);
+   _mesa_printf("VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
+		node->count,
+		node->prim_count,
+		node->vertex_size);
 
    for (i = 0 ; i < node->prim_count ; i++) {
       struct _mesa_prim *prim = &node->prim[i];
@@ -1228,6 +1231,8 @@ void vbo_save_api_init( struct vbo_save_context *save )
    ctx->ListState.ListVtxfmt.DrawArrays = _save_OBE_DrawArrays;
    ctx->ListState.ListVtxfmt.DrawElements = _save_OBE_DrawElements;
    ctx->ListState.ListVtxfmt.DrawRangeElements = _save_OBE_DrawRangeElements;
+   /* loops back into _save_OBE_DrawElements */
+   ctx->ListState.ListVtxfmt.MultiDrawElementsEXT = _mesa_noop_MultiDrawElements;
    _mesa_install_save_vtxfmt( ctx, &ctx->ListState.ListVtxfmt );
 }
 
diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index 5110648c28..d834fa1f2e 100644
--- a/src/mesa/vbo/vbo_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -279,6 +279,7 @@ void vbo_save_playback_vertex_list( GLcontext *ctx, void *data )
 				    node->prim, 
 				    node->prim_count,
 				    NULL,
+				    GL_TRUE,
 				    0,	/* Node is a VBO, so this is ok */
 				    node->count - 1);
    }
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
index b9a5c56987..8ec180d550 100644
--- a/src/mesa/vbo/vbo_split_copy.c
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -30,7 +30,9 @@
  */
 
 #include "main/glheader.h"
+#include "main/bufferobj.h"
 #include "main/imports.h"
+#include "main/image.h"
 #include "main/macros.h"
 #include "main/enums.h"
 #include "main/mtypes.h"
@@ -41,7 +43,8 @@
 
 #define ELT_TABLE_SIZE 16
 
-/* Used for vertex-level splitting of indexed buffers.  Note that
+/**
+ * Used for vertex-level splitting of indexed buffers.  Note that
  * non-indexed primitives may be converted to indexed in some cases
  * (eg loops, fans) in order to use this splitting path.
  */
@@ -73,23 +76,21 @@ struct copy_context {
    GLuint *translated_elt_buf;
    const GLuint *srcelt;
 
-   /* A baby hash table to avoid re-emitting (some) duplicate
+   /** A baby hash table to avoid re-emitting (some) duplicate
     * vertices when splitting indexed primitives.
     */
    struct { 
       GLuint in;
       GLuint out;
    } vert_cache[ELT_TABLE_SIZE];
-      
 
    GLuint vertex_size;
    GLubyte *dstbuf;
-   GLubyte *dstptr;		/* dstptr == dstbuf + dstelt_max * vertsize */
-   GLuint dstbuf_size;	/* in vertices */
-   GLuint dstbuf_nr;		/* count of emitted vertices, also the
-				 * largest value in dstelt.  Our
-				 * MaxIndex.
-				 */
+   GLubyte *dstptr;     /**< dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;  /**< in vertices */
+   GLuint dstbuf_nr;    /**< count of emitted vertices, also the largest value
+                         * in dstelt.  Our MaxIndex.
+                         */
 
    GLuint *dstelt;
    GLuint dstelt_nr;
@@ -102,32 +103,19 @@ struct copy_context {
 };
 
 
-static GLuint type_size( GLenum type )
-{
-   switch(type) {
-   case GL_BYTE: return sizeof(GLbyte);
-   case GL_UNSIGNED_BYTE: return sizeof(GLubyte);
-   case GL_SHORT: return sizeof(GLshort);
-   case GL_UNSIGNED_SHORT: return sizeof(GLushort);
-   case GL_INT: return sizeof(GLint);
-   case GL_UNSIGNED_INT: return sizeof(GLuint);
-   case GL_FLOAT: return sizeof(GLfloat);
-   case GL_DOUBLE: return sizeof(GLdouble);
-   default: return 0;
-   }
-}
-
 static GLuint attr_size( const struct gl_client_array *array )
 {
-   return array->Size * type_size(array->Type);
+   return array->Size * _mesa_sizeof_type(array->Type);
 }
 
 
-/* Starts returning true slightly before the buffer fills, to ensure
+/**
+ * Starts returning true slightly before the buffer fills, to ensure
  * that there is sufficient room for any remaining vertices to finish
  * off the prim:
  */
-static GLboolean check_flush( struct copy_context *copy )
+static GLboolean
+check_flush( struct copy_context *copy )
 {
    GLenum mode = copy->dstprim[copy->dstprim_nr].mode; 
 
@@ -145,7 +133,44 @@ static GLboolean check_flush( struct copy_context *copy )
    return GL_FALSE;
 }
 
-static void flush( struct copy_context *copy )
+
+/**
+ * Dump the parameters/info for a vbo->draw() call.
+ */
+static void
+dump_draw_info(GLcontext *ctx,
+               const struct gl_client_array **arrays,
+               const struct _mesa_prim *prims,
+               GLuint nr_prims,
+               const struct _mesa_index_buffer *ib,
+               GLuint min_index,
+               GLuint max_index)
+{
+   GLuint i, j;
+
+   _mesa_printf("VBO Draw:\n");
+   for (i = 0; i < nr_prims; i++) {
+      _mesa_printf("Prim %u of %u\n", i, nr_prims);
+      _mesa_printf("  Prim mode 0x%x\n", prims[i].mode);
+      _mesa_printf("  IB: %p\n", (void*) ib);
+      for (j = 0; j < VERT_ATTRIB_MAX; j++) {
+         _mesa_printf("    array %d at %p:\n", j, (void*) arrays[j]);
+         _mesa_printf("      enabled %d, ptr %p, size %d, type 0x%x, stride %d\n",
+                      arrays[j]->Enabled, arrays[j]->Ptr,
+                      arrays[j]->Size, arrays[j]->Type, arrays[j]->StrideB);
+         if (0) {
+            GLint k = prims[i].start + prims[i].count - 1;
+            GLfloat *last = (GLfloat *) (arrays[j]->Ptr + arrays[j]->Stride * k);
+            _mesa_printf("        last: %f %f %f\n",
+                         last[0], last[1], last[2]);
+         }
+      }
+   }
+}
+
+
+static void
+flush( struct copy_context *copy )
 {
    GLuint i;
 
@@ -153,11 +178,24 @@ static void flush( struct copy_context *copy )
     */
    copy->dstib.count = copy->dstelt_nr;
 
+#if 0
+   dump_draw_info(copy->ctx,
+                  copy->dstarray_ptr,
+                  copy->dstprim,
+                  copy->dstprim_nr,
+                  &copy->dstib,
+                  0,
+                  copy->dstbuf_nr);
+#else
+   (void) dump_draw_info;
+#endif
+
    copy->draw( copy->ctx,
 	       copy->dstarray_ptr,
 	       copy->dstprim,
 	       copy->dstprim_nr,
 	       &copy->dstib,
+	       GL_TRUE,
 	       0,
 	       copy->dstbuf_nr );
 
@@ -175,8 +213,11 @@ static void flush( struct copy_context *copy )
 }
 
 
-
-static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+/**
+ * Called at begin of each primitive during replay.
+ */
+static void
+begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -187,10 +228,12 @@ static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag
 }
 
 
-/* Use a hashtable to attempt to identify recently-emitted vertices
+/**
+ * Use a hashtable to attempt to identify recently-emitted vertices
  * and avoid re-emitting them.
  */
-static GLuint elt(struct copy_context *copy, GLuint elt_idx)
+static GLuint
+elt(struct copy_context *copy, GLuint elt_idx)
 {
    GLuint elt = copy->srcelt[elt_idx];
    GLuint slot = elt & (ELT_TABLE_SIZE-1);
@@ -213,6 +256,17 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
 	 memcpy(csr, srcptr, copy->varying[i].size);
 	 csr += copy->varying[i].size;
 
+#ifdef NAN_CHECK
+         if (srcarray->Type == GL_FLOAT) {
+            GLuint k;
+            GLfloat *f = (GLfloat *) srcptr;
+            for (k = 0; k < srcarray->Size; k++) {
+               assert(!IS_INF_OR_NAN(f[k]));
+               assert(f[k] <= 1.0e20 && f[k] >= -1.0e20);
+            }
+         }
+#endif
+
 	 if (0) 
 	 {
 	    const GLuint *f = (const GLuint *)srcptr;
@@ -222,7 +276,6 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
 	       _mesa_printf("%x ", f[j]);
 	    _mesa_printf("\n");
 	 }
-	       
       }
 
       copy->vert_cache[slot].in = elt;
@@ -230,9 +283,8 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
       copy->dstptr += copy->vertex_size;
 
       assert(csr == copy->dstptr);
-      assert(copy->dstptr == (copy->dstbuf + 
-				    copy->dstbuf_nr * 
-				    copy->vertex_size));
+      assert(copy->dstptr == (copy->dstbuf +
+                              copy->dstbuf_nr * copy->vertex_size));
    }
 /*    else */
 /*       _mesa_printf("  --> reuse vertex\n"); */
@@ -242,7 +294,12 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
    return check_flush(copy);
 }
 
-static void end( struct copy_context *copy, GLboolean end_flag )
+
+/**
+ * Called at end of each primitive during replay.
+ */
+static void
+end( struct copy_context *copy, GLboolean end_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -257,8 +314,8 @@ static void end( struct copy_context *copy, GLboolean end_flag )
 }
 
 
-
-static void replay_elts( struct copy_context *copy )
+static void
+replay_elts( struct copy_context *copy )
 {
    GLuint i, j, k;
    GLboolean split;
@@ -362,7 +419,8 @@ static void replay_elts( struct copy_context *copy )
 }
 
 
-static void replay_init( struct copy_context *copy )
+static void
+replay_init( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -387,11 +445,8 @@ static void replay_init( struct copy_context *copy )
 	 copy->varying[j].size = attr_size(copy->array[i]);
 	 copy->vertex_size += attr_size(copy->array[i]);
       
-	 if (vbo->Name && !vbo->Pointer) 
-	    ctx->Driver.MapBuffer(ctx,
-				  GL_ARRAY_BUFFER_ARB, 
-				  GL_WRITE_ONLY, /* XXX */
-				  vbo);
+	 if (_mesa_is_bufferobj(vbo) && !_mesa_bufferobj_mapped(vbo)) 
+	    ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY, vbo);
 
 	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
 						 copy->array[i]->Ptr);
@@ -404,13 +459,13 @@ static void replay_init( struct copy_context *copy )
     * caller convert non-indexed prims to indexed.  Could alternately
     * do it internally.
     */
-   if (copy->ib->obj->Name && !copy->ib->obj->Pointer) 
-      ctx->Driver.MapBuffer(ctx, 
-			    GL_ARRAY_BUFFER_ARB, /* XXX */
-			    GL_WRITE_ONLY, /* XXX */
+   if (_mesa_is_bufferobj(copy->ib->obj) &&
+       !_mesa_bufferobj_mapped(copy->ib->obj)) 
+      ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY,
 			    copy->ib->obj);
 
-   srcptr = (const GLubyte *)ADD_POINTERS(copy->ib->obj->Pointer, copy->ib->ptr);
+   srcptr = (const GLubyte *) ADD_POINTERS(copy->ib->obj->Pointer,
+                                           copy->ib->ptr);
 
    switch (copy->ib->type) {
    case GL_UNSIGNED_BYTE:
@@ -434,7 +489,6 @@ static void replay_init( struct copy_context *copy )
       copy->srcelt = (const GLuint *)srcptr;
       break;
    }
-   
 
    /* Figure out the maximum allowed vertex buffer size:
     */
@@ -449,8 +503,7 @@ static void replay_init( struct copy_context *copy )
     *
     * XXX:  This should be a VBO!
     */
-   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * 
-			       copy->vertex_size);   
+   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * copy->vertex_size);   
    copy->dstptr = copy->dstbuf;
 
    /* Setup new vertex arrays to point into the output buffer: 
@@ -467,7 +520,7 @@ static void replay_init( struct copy_context *copy )
       dst->Ptr = copy->dstbuf + offset;
       dst->Enabled = GL_TRUE;
       dst->Normalized = src->Normalized; 
-      dst->BufferObj = ctx->Array.NullBufferObj;
+      dst->BufferObj = ctx->Shared->NullBufferObj;
       dst->_MaxElement = copy->dstbuf_size; /* may be less! */
 
       offset += copy->varying[i].size;
@@ -487,12 +540,16 @@ static void replay_init( struct copy_context *copy )
     */
    copy->dstib.count = 0;	/* duplicates dstelt_nr */
    copy->dstib.type = GL_UNSIGNED_INT;
-   copy->dstib.obj = ctx->Array.NullBufferObj;
+   copy->dstib.obj = ctx->Shared->NullBufferObj;
    copy->dstib.ptr = copy->dstelt;
 }
 
 
-static void replay_finish( struct copy_context *copy )
+/**
+ * Free up everything allocated during split/replay.
+ */
+static void
+replay_finish( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -502,25 +559,27 @@ static void replay_finish( struct copy_context *copy )
    _mesa_free(copy->translated_elt_buf);
    _mesa_free(copy->dstbuf);
    _mesa_free(copy->dstelt);
-   
+
    /* Unmap VBO's 
     */
    for (i = 0; i < copy->nr_varying; i++) {
       struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
-
-      if (vbo->Name && vbo->Pointer) 
-	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vbo);
+      if (_mesa_is_bufferobj(vbo) && _mesa_bufferobj_mapped(vbo)) 
+	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, vbo);
    }
 
    /* Unmap index buffer:
     */
-   if (copy->ib->obj->Name && copy->ib->obj->Pointer) {
-      ctx->Driver.UnmapBuffer(ctx, 
-			      GL_ARRAY_BUFFER_ARB, /* XXX */
-			      copy->ib->obj);
+   if (_mesa_is_bufferobj(copy->ib->obj) &&
+       _mesa_bufferobj_mapped(copy->ib->obj)) {
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, copy->ib->obj);
    }
 }
 
+
+/**
+ * Split VBO into smaller pieces, draw the pieces.
+ */
 void vbo_split_copy( GLcontext *ctx,
 		     const struct gl_client_array *arrays[],
 		     const struct _mesa_prim *prim,
@@ -546,13 +605,11 @@ void vbo_split_copy( GLcontext *ctx,
    copy.draw = draw;
    copy.limits = limits;
 
-
    /* Clear the vertex cache:
     */
    for (i = 0; i < ELT_TABLE_SIZE; i++)
       copy.vert_cache[i].in = ~0;
 
-
    replay_init(&copy);
    replay_elts(&copy);
    replay_finish(&copy);
diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c
index fbc856e93b..da84eaa6ea 100644
--- a/src/mesa/vbo/vbo_split_inplace.c
+++ b/src/mesa/vbo/vbo_split_inplace.c
@@ -59,39 +59,23 @@ struct split_context {
 static void flush_vertex( struct split_context *split )
 {
    GLuint min_index, max_index;
+   GLuint i;
 
    if (!split->dstprim_nr) 
       return;
 
-   if (split->ib) {
-      /* This should basically be multipass rendering over the same
-       * unchanging set of VBO's.  Would like the driver not to
-       * re-upload the data, or swtnl not to re-transform the
-       * vertices.
-       */
-      assert(split->max_index - split->min_index < split->limits->max_verts);
-      min_index = split->min_index;
-      max_index = split->max_index;
-   }
-   else {
-      /* Non-indexed rendering.  Cannot assume that the primitives are
-       * ordered by increasing vertex, because of entrypoints like
-       * MultiDrawArrays.
-       */
-      GLuint i;
-      min_index = split->dstprim[0].start;
-      max_index = min_index + split->dstprim[0].count - 1;
+   min_index = split->dstprim[0].start;
+   max_index = min_index + split->dstprim[0].count - 1;
 
-      for (i = 1; i < split->dstprim_nr; i++) {
-	 GLuint tmp_min = split->dstprim[i].start;
-	 GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
+   for (i = 1; i < split->dstprim_nr; i++) {
+      GLuint tmp_min = split->dstprim[i].start;
+      GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
 
-	 if (tmp_min < min_index) 
-	    min_index = tmp_min;
+      if (tmp_min < min_index)
+	 min_index = tmp_min;
 
-	 if (tmp_max > max_index) 
-	    max_index = tmp_max;
-      }
+      if (tmp_max > max_index)
+	 max_index = tmp_max;
    }
 
    assert(max_index >= min_index);
@@ -101,6 +85,7 @@ static void flush_vertex( struct split_context *split )
 		split->dstprim,
 		split->dstprim_nr,
 		NULL,
+		GL_TRUE,
 		min_index,
 		max_index);
 
@@ -221,7 +206,7 @@ static void split_prims( struct split_context *split)
 
 	 ib.count = count;
 	 ib.type = GL_UNSIGNED_INT;
-	 ib.obj = split->ctx->Array.NullBufferObj;
+	 ib.obj = split->ctx->Shared->NullBufferObj;
 	 ib.ptr = elts;
 	    
 	 tmpprim = *prim;
diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S
index fe05549c1d..db917f3982 100644
--- a/src/mesa/x86-64/glapi_x86-64.S
+++ b/src/mesa/x86-64/glapi_x86-64.S
@@ -45,7 +45,7 @@
 #  define GL_PREFIX(n) GLNAME(CONCAT(gl,n))
 # endif
 
-#if defined(PTHREADS) || defined(USE_XTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)
+#if defined(PTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)
 #  define THREADS
 #endif
 
@@ -21157,12 +21157,456 @@ GL_PREFIX(RenderbufferStorageMultisample):
 	.size	GL_PREFIX(RenderbufferStorageMultisample), .-GL_PREFIX(RenderbufferStorageMultisample)
 
 	.p2align	4,,15
+	.globl	GL_PREFIX(FlushMappedBufferRange)
+	.type	GL_PREFIX(FlushMappedBufferRange), @function
+GL_PREFIX(FlushMappedBufferRange):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4496(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(FlushMappedBufferRange), .-GL_PREFIX(FlushMappedBufferRange)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(MapBufferRange)
+	.type	GL_PREFIX(MapBufferRange), @function
+GL_PREFIX(MapBufferRange):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4504(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(MapBufferRange), .-GL_PREFIX(MapBufferRange)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(BindVertexArray)
+	.type	GL_PREFIX(BindVertexArray), @function
+GL_PREFIX(BindVertexArray):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4512(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(BindVertexArray), .-GL_PREFIX(BindVertexArray)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(GenVertexArrays)
+	.type	GL_PREFIX(GenVertexArrays), @function
+GL_PREFIX(GenVertexArrays):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4520(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(GenVertexArrays), .-GL_PREFIX(GenVertexArrays)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(CopyBufferSubData)
+	.type	GL_PREFIX(CopyBufferSubData), @function
+GL_PREFIX(CopyBufferSubData):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_x86_64_get_dispatch@PLT
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4528(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_glapi_get_dispatch
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(CopyBufferSubData), .-GL_PREFIX(CopyBufferSubData)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(ClientWaitSync)
+	.type	GL_PREFIX(ClientWaitSync), @function
+GL_PREFIX(ClientWaitSync):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4536(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4536(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4536(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4536(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(ClientWaitSync), .-GL_PREFIX(ClientWaitSync)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(DeleteSync)
+	.type	GL_PREFIX(DeleteSync), @function
+GL_PREFIX(DeleteSync):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4544(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	4544(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4544(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	4544(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(DeleteSync), .-GL_PREFIX(DeleteSync)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(FenceSync)
+	.type	GL_PREFIX(FenceSync), @function
+GL_PREFIX(FenceSync):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4552(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4552(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4552(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4552(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(FenceSync), .-GL_PREFIX(FenceSync)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(GetInteger64v)
+	.type	GL_PREFIX(GetInteger64v), @function
+GL_PREFIX(GetInteger64v):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4560(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(GetInteger64v), .-GL_PREFIX(GetInteger64v)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(GetSynciv)
+	.type	GL_PREFIX(GetSynciv), @function
+GL_PREFIX(GetSynciv):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4568(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_x86_64_get_dispatch@PLT
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4568(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4568(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_glapi_get_dispatch
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4568(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(GetSynciv), .-GL_PREFIX(GetSynciv)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(IsSync)
+	.type	GL_PREFIX(IsSync), @function
+GL_PREFIX(IsSync):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4576(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	4576(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4576(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	4576(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(IsSync), .-GL_PREFIX(IsSync)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(WaitSync)
+	.type	GL_PREFIX(WaitSync), @function
+GL_PREFIX(WaitSync):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4584(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4584(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4584(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4584(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(WaitSync), .-GL_PREFIX(WaitSync)
+
+	.p2align	4,,15
 	.globl	GL_PREFIX(PolygonOffsetEXT)
 	.type	GL_PREFIX(PolygonOffsetEXT), @function
 GL_PREFIX(PolygonOffsetEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4496(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21172,13 +21616,13 @@ GL_PREFIX(PolygonOffsetEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4496(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4496(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21188,19 +21632,19 @@ GL_PREFIX(PolygonOffsetEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4496(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PolygonOffsetEXT), .-GL_PREFIX(PolygonOffsetEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_563)
-	.type	GL_PREFIX(_dispatch_stub_563), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_563))
-GL_PREFIX(_dispatch_stub_563):
+	.globl	GL_PREFIX(_dispatch_stub_575)
+	.type	GL_PREFIX(_dispatch_stub_575), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_575))
+GL_PREFIX(_dispatch_stub_575):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4504(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21210,13 +21654,13 @@ GL_PREFIX(_dispatch_stub_563):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4504(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4504(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21226,19 +21670,19 @@ GL_PREFIX(_dispatch_stub_563):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4504(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_563), .-GL_PREFIX(_dispatch_stub_563)
+	.size	GL_PREFIX(_dispatch_stub_575), .-GL_PREFIX(_dispatch_stub_575)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_564)
-	.type	GL_PREFIX(_dispatch_stub_564), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_564))
-GL_PREFIX(_dispatch_stub_564):
+	.globl	GL_PREFIX(_dispatch_stub_576)
+	.type	GL_PREFIX(_dispatch_stub_576), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_576))
+GL_PREFIX(_dispatch_stub_576):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4512(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21248,13 +21692,13 @@ GL_PREFIX(_dispatch_stub_564):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4512(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4512(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21264,19 +21708,19 @@ GL_PREFIX(_dispatch_stub_564):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4512(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_564), .-GL_PREFIX(_dispatch_stub_564)
+	.size	GL_PREFIX(_dispatch_stub_576), .-GL_PREFIX(_dispatch_stub_576)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_565)
-	.type	GL_PREFIX(_dispatch_stub_565), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_565))
-GL_PREFIX(_dispatch_stub_565):
+	.globl	GL_PREFIX(_dispatch_stub_577)
+	.type	GL_PREFIX(_dispatch_stub_577), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_577))
+GL_PREFIX(_dispatch_stub_577):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4520(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21286,13 +21730,13 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4520(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4520(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21302,19 +21746,19 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4520(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_565), .-GL_PREFIX(_dispatch_stub_565)
+	.size	GL_PREFIX(_dispatch_stub_577), .-GL_PREFIX(_dispatch_stub_577)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_566)
-	.type	GL_PREFIX(_dispatch_stub_566), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_566))
-GL_PREFIX(_dispatch_stub_566):
+	.globl	GL_PREFIX(_dispatch_stub_578)
+	.type	GL_PREFIX(_dispatch_stub_578), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_578))
+GL_PREFIX(_dispatch_stub_578):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4528(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21324,13 +21768,13 @@ GL_PREFIX(_dispatch_stub_566):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4528(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4528(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21340,19 +21784,19 @@ GL_PREFIX(_dispatch_stub_566):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4528(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_566), .-GL_PREFIX(_dispatch_stub_566)
+	.size	GL_PREFIX(_dispatch_stub_578), .-GL_PREFIX(_dispatch_stub_578)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_567)
-	.type	GL_PREFIX(_dispatch_stub_567), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_567))
-GL_PREFIX(_dispatch_stub_567):
+	.globl	GL_PREFIX(_dispatch_stub_579)
+	.type	GL_PREFIX(_dispatch_stub_579), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_579))
+GL_PREFIX(_dispatch_stub_579):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4536(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21362,13 +21806,13 @@ GL_PREFIX(_dispatch_stub_567):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4536(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4536(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21378,19 +21822,19 @@ GL_PREFIX(_dispatch_stub_567):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4536(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_567), .-GL_PREFIX(_dispatch_stub_567)
+	.size	GL_PREFIX(_dispatch_stub_579), .-GL_PREFIX(_dispatch_stub_579)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_568)
-	.type	GL_PREFIX(_dispatch_stub_568), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_568))
-GL_PREFIX(_dispatch_stub_568):
+	.globl	GL_PREFIX(_dispatch_stub_580)
+	.type	GL_PREFIX(_dispatch_stub_580), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_580))
+GL_PREFIX(_dispatch_stub_580):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4544(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21400,13 +21844,13 @@ GL_PREFIX(_dispatch_stub_568):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4544(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4544(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21416,19 +21860,19 @@ GL_PREFIX(_dispatch_stub_568):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4544(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_568), .-GL_PREFIX(_dispatch_stub_568)
+	.size	GL_PREFIX(_dispatch_stub_580), .-GL_PREFIX(_dispatch_stub_580)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_569)
-	.type	GL_PREFIX(_dispatch_stub_569), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_569))
-GL_PREFIX(_dispatch_stub_569):
+	.globl	GL_PREFIX(_dispatch_stub_581)
+	.type	GL_PREFIX(_dispatch_stub_581), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_581))
+GL_PREFIX(_dispatch_stub_581):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4552(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21438,13 +21882,13 @@ GL_PREFIX(_dispatch_stub_569):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4552(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4552(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21454,40 +21898,40 @@ GL_PREFIX(_dispatch_stub_569):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4552(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_569), .-GL_PREFIX(_dispatch_stub_569)
+	.size	GL_PREFIX(_dispatch_stub_581), .-GL_PREFIX(_dispatch_stub_581)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_570)
-	.type	GL_PREFIX(_dispatch_stub_570), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_570))
-GL_PREFIX(_dispatch_stub_570):
+	.globl	GL_PREFIX(_dispatch_stub_582)
+	.type	GL_PREFIX(_dispatch_stub_582), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_582))
+GL_PREFIX(_dispatch_stub_582):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4560(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4560(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_570), .-GL_PREFIX(_dispatch_stub_570)
+	.size	GL_PREFIX(_dispatch_stub_582), .-GL_PREFIX(_dispatch_stub_582)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(ColorPointerEXT)
@@ -21495,7 +21939,7 @@ GL_PREFIX(_dispatch_stub_570):
 GL_PREFIX(ColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4568(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21509,13 +21953,13 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4568(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21529,7 +21973,7 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorPointerEXT), .-GL_PREFIX(ColorPointerEXT)
@@ -21540,7 +21984,7 @@ GL_PREFIX(ColorPointerEXT):
 GL_PREFIX(EdgeFlagPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4576(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21550,13 +21994,13 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4576(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21566,7 +22010,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EdgeFlagPointerEXT), .-GL_PREFIX(EdgeFlagPointerEXT)
@@ -21577,7 +22021,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 GL_PREFIX(IndexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4584(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21591,13 +22035,13 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4584(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21611,7 +22055,7 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IndexPointerEXT), .-GL_PREFIX(IndexPointerEXT)
@@ -21622,7 +22066,7 @@ GL_PREFIX(IndexPointerEXT):
 GL_PREFIX(NormalPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4592(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21636,13 +22080,13 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4592(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21656,7 +22100,7 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(NormalPointerEXT), .-GL_PREFIX(NormalPointerEXT)
@@ -21667,7 +22111,7 @@ GL_PREFIX(NormalPointerEXT):
 GL_PREFIX(TexCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4600(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21681,13 +22125,13 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4600(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21701,7 +22145,7 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexCoordPointerEXT), .-GL_PREFIX(TexCoordPointerEXT)
@@ -21712,7 +22156,7 @@ GL_PREFIX(TexCoordPointerEXT):
 GL_PREFIX(VertexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4608(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21726,13 +22170,13 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4608(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4608(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21746,7 +22190,7 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4608(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexPointerEXT), .-GL_PREFIX(VertexPointerEXT)
@@ -21757,7 +22201,7 @@ GL_PREFIX(VertexPointerEXT):
 GL_PREFIX(PointParameterfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4616(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21767,13 +22211,13 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4616(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4616(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21783,7 +22227,7 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4616(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfEXT), .-GL_PREFIX(PointParameterfEXT)
@@ -21794,7 +22238,7 @@ GL_PREFIX(PointParameterfEXT):
 GL_PREFIX(PointParameterfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4624(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21804,13 +22248,13 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4624(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21820,7 +22264,7 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfvEXT), .-GL_PREFIX(PointParameterfvEXT)
@@ -21831,7 +22275,7 @@ GL_PREFIX(PointParameterfvEXT):
 GL_PREFIX(LockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4632(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21841,13 +22285,13 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4632(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4632(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21857,7 +22301,7 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4632(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LockArraysEXT), .-GL_PREFIX(LockArraysEXT)
@@ -21868,37 +22312,37 @@ GL_PREFIX(LockArraysEXT):
 GL_PREFIX(UnlockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4640(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4640(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4640(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4640(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(UnlockArraysEXT), .-GL_PREFIX(UnlockArraysEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_581)
-	.type	GL_PREFIX(_dispatch_stub_581), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_581))
-GL_PREFIX(_dispatch_stub_581):
+	.globl	GL_PREFIX(_dispatch_stub_593)
+	.type	GL_PREFIX(_dispatch_stub_593), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_593))
+GL_PREFIX(_dispatch_stub_593):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4648(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21908,13 +22352,13 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4648(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21924,19 +22368,19 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_581), .-GL_PREFIX(_dispatch_stub_581)
+	.size	GL_PREFIX(_dispatch_stub_593), .-GL_PREFIX(_dispatch_stub_593)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_582)
-	.type	GL_PREFIX(_dispatch_stub_582), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_582))
-GL_PREFIX(_dispatch_stub_582):
+	.globl	GL_PREFIX(_dispatch_stub_594)
+	.type	GL_PREFIX(_dispatch_stub_594), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_594))
+GL_PREFIX(_dispatch_stub_594):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4656(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21946,13 +22390,13 @@ GL_PREFIX(_dispatch_stub_582):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4656(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21962,10 +22406,10 @@ GL_PREFIX(_dispatch_stub_582):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_582), .-GL_PREFIX(_dispatch_stub_582)
+	.size	GL_PREFIX(_dispatch_stub_594), .-GL_PREFIX(_dispatch_stub_594)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(SecondaryColor3bEXT)
@@ -21973,7 +22417,7 @@ GL_PREFIX(_dispatch_stub_582):
 GL_PREFIX(SecondaryColor3bEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4664(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21983,13 +22427,13 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4664(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21999,7 +22443,7 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bEXT), .-GL_PREFIX(SecondaryColor3bEXT)
@@ -22010,25 +22454,25 @@ GL_PREFIX(SecondaryColor3bEXT):
 GL_PREFIX(SecondaryColor3bvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4672(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4672(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4672(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4672(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bvEXT), .-GL_PREFIX(SecondaryColor3bvEXT)
@@ -22039,7 +22483,7 @@ GL_PREFIX(SecondaryColor3bvEXT):
 GL_PREFIX(SecondaryColor3dEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4680(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22051,13 +22495,13 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4680(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4680(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22069,7 +22513,7 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4680(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dEXT), .-GL_PREFIX(SecondaryColor3dEXT)
@@ -22080,25 +22524,25 @@ GL_PREFIX(SecondaryColor3dEXT):
 GL_PREFIX(SecondaryColor3dvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4688(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4688(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4688(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4688(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dvEXT), .-GL_PREFIX(SecondaryColor3dvEXT)
@@ -22109,7 +22553,7 @@ GL_PREFIX(SecondaryColor3dvEXT):
 GL_PREFIX(SecondaryColor3fEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4696(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22121,13 +22565,13 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4696(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4696(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22139,7 +22583,7 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4696(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fEXT), .-GL_PREFIX(SecondaryColor3fEXT)
@@ -22150,25 +22594,25 @@ GL_PREFIX(SecondaryColor3fEXT):
 GL_PREFIX(SecondaryColor3fvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4704(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4704(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fvEXT), .-GL_PREFIX(SecondaryColor3fvEXT)
@@ -22179,7 +22623,7 @@ GL_PREFIX(SecondaryColor3fvEXT):
 GL_PREFIX(SecondaryColor3iEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4712(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22189,13 +22633,13 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4712(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22205,7 +22649,7 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3iEXT), .-GL_PREFIX(SecondaryColor3iEXT)
@@ -22216,25 +22660,25 @@ GL_PREFIX(SecondaryColor3iEXT):
 GL_PREFIX(SecondaryColor3ivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4720(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4720(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ivEXT), .-GL_PREFIX(SecondaryColor3ivEXT)
@@ -22245,7 +22689,7 @@ GL_PREFIX(SecondaryColor3ivEXT):
 GL_PREFIX(SecondaryColor3sEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4728(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22255,13 +22699,13 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4728(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22271,7 +22715,7 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3sEXT), .-GL_PREFIX(SecondaryColor3sEXT)
@@ -22282,25 +22726,25 @@ GL_PREFIX(SecondaryColor3sEXT):
 GL_PREFIX(SecondaryColor3svEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4736(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4736(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3svEXT), .-GL_PREFIX(SecondaryColor3svEXT)
@@ -22311,7 +22755,7 @@ GL_PREFIX(SecondaryColor3svEXT):
 GL_PREFIX(SecondaryColor3ubEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4744(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22321,13 +22765,13 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4744(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22337,7 +22781,7 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubEXT), .-GL_PREFIX(SecondaryColor3ubEXT)
@@ -22348,25 +22792,25 @@ GL_PREFIX(SecondaryColor3ubEXT):
 GL_PREFIX(SecondaryColor3ubvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4752(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4752(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubvEXT), .-GL_PREFIX(SecondaryColor3ubvEXT)
@@ -22377,7 +22821,7 @@ GL_PREFIX(SecondaryColor3ubvEXT):
 GL_PREFIX(SecondaryColor3uiEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4760(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22387,13 +22831,13 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4760(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22403,7 +22847,7 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uiEXT), .-GL_PREFIX(SecondaryColor3uiEXT)
@@ -22414,25 +22858,25 @@ GL_PREFIX(SecondaryColor3uiEXT):
 GL_PREFIX(SecondaryColor3uivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4768(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4768(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uivEXT), .-GL_PREFIX(SecondaryColor3uivEXT)
@@ -22443,7 +22887,7 @@ GL_PREFIX(SecondaryColor3uivEXT):
 GL_PREFIX(SecondaryColor3usEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4776(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22453,13 +22897,13 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4776(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22469,7 +22913,7 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usEXT), .-GL_PREFIX(SecondaryColor3usEXT)
@@ -22480,25 +22924,25 @@ GL_PREFIX(SecondaryColor3usEXT):
 GL_PREFIX(SecondaryColor3usvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4784(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4784(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usvEXT), .-GL_PREFIX(SecondaryColor3usvEXT)
@@ -22509,7 +22953,7 @@ GL_PREFIX(SecondaryColor3usvEXT):
 GL_PREFIX(SecondaryColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4792(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22523,13 +22967,13 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4792(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22543,7 +22987,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColorPointerEXT), .-GL_PREFIX(SecondaryColorPointerEXT)
@@ -22554,7 +22998,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 GL_PREFIX(MultiDrawArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4800(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22568,13 +23012,13 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4800(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22588,7 +23032,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawArraysEXT), .-GL_PREFIX(MultiDrawArraysEXT)
@@ -22599,7 +23043,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 GL_PREFIX(MultiDrawElementsEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4808(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22613,13 +23057,13 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4808(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22633,7 +23077,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawElementsEXT), .-GL_PREFIX(MultiDrawElementsEXT)
@@ -22644,7 +23088,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 GL_PREFIX(FogCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4816(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22654,13 +23098,13 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4816(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4816(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22670,7 +23114,7 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4816(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordPointerEXT), .-GL_PREFIX(FogCoordPointerEXT)
@@ -22681,7 +23125,7 @@ GL_PREFIX(FogCoordPointerEXT):
 GL_PREFIX(FogCoorddEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4824(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22689,13 +23133,13 @@ GL_PREFIX(FogCoorddEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4824(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4824(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22703,7 +23147,7 @@ GL_PREFIX(FogCoorddEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4824(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddEXT), .-GL_PREFIX(FogCoorddEXT)
@@ -22714,25 +23158,25 @@ GL_PREFIX(FogCoorddEXT):
 GL_PREFIX(FogCoorddvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4832(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4832(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4832(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4832(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddvEXT), .-GL_PREFIX(FogCoorddvEXT)
@@ -22743,7 +23187,7 @@ GL_PREFIX(FogCoorddvEXT):
 GL_PREFIX(FogCoordfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4840(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22751,13 +23195,13 @@ GL_PREFIX(FogCoordfEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4840(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4840(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22765,7 +23209,7 @@ GL_PREFIX(FogCoordfEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4840(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfEXT), .-GL_PREFIX(FogCoordfEXT)
@@ -22776,58 +23220,58 @@ GL_PREFIX(FogCoordfEXT):
 GL_PREFIX(FogCoordfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4848(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4848(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfvEXT), .-GL_PREFIX(FogCoordfvEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_607)
-	.type	GL_PREFIX(_dispatch_stub_607), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_607))
-GL_PREFIX(_dispatch_stub_607):
+	.globl	GL_PREFIX(_dispatch_stub_619)
+	.type	GL_PREFIX(_dispatch_stub_619), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_619))
+GL_PREFIX(_dispatch_stub_619):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4856(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4856(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_607), .-GL_PREFIX(_dispatch_stub_607)
+	.size	GL_PREFIX(_dispatch_stub_619), .-GL_PREFIX(_dispatch_stub_619)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BlendFuncSeparateEXT)
@@ -22835,7 +23279,7 @@ GL_PREFIX(_dispatch_stub_607):
 GL_PREFIX(BlendFuncSeparateEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4864(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22849,13 +23293,13 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4864(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4864(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22869,7 +23313,7 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4864(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BlendFuncSeparateEXT), .-GL_PREFIX(BlendFuncSeparateEXT)
@@ -22880,25 +23324,25 @@ GL_PREFIX(BlendFuncSeparateEXT):
 GL_PREFIX(FlushVertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4872(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4872(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4872(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4872(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FlushVertexArrayRangeNV), .-GL_PREFIX(FlushVertexArrayRangeNV)
@@ -22909,7 +23353,7 @@ GL_PREFIX(FlushVertexArrayRangeNV):
 GL_PREFIX(VertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4880(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22919,13 +23363,13 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4880(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22935,7 +23379,7 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexArrayRangeNV), .-GL_PREFIX(VertexArrayRangeNV)
@@ -22946,7 +23390,7 @@ GL_PREFIX(VertexArrayRangeNV):
 GL_PREFIX(CombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4888(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22964,13 +23408,13 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4888(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22988,7 +23432,7 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerInputNV), .-GL_PREFIX(CombinerInputNV)
@@ -22999,7 +23443,7 @@ GL_PREFIX(CombinerInputNV):
 GL_PREFIX(CombinerOutputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4896(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23017,13 +23461,13 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4896(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4896(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23041,7 +23485,7 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4896(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerOutputNV), .-GL_PREFIX(CombinerOutputNV)
@@ -23052,7 +23496,7 @@ GL_PREFIX(CombinerOutputNV):
 GL_PREFIX(CombinerParameterfNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4904(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23062,13 +23506,13 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4904(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4904(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23078,7 +23522,7 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4904(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfNV), .-GL_PREFIX(CombinerParameterfNV)
@@ -23089,7 +23533,7 @@ GL_PREFIX(CombinerParameterfNV):
 GL_PREFIX(CombinerParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4912(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23099,13 +23543,13 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4912(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23115,7 +23559,7 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfvNV), .-GL_PREFIX(CombinerParameterfvNV)
@@ -23126,7 +23570,7 @@ GL_PREFIX(CombinerParameterfvNV):
 GL_PREFIX(CombinerParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4920(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23136,13 +23580,13 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4920(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23152,7 +23596,7 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameteriNV), .-GL_PREFIX(CombinerParameteriNV)
@@ -23163,7 +23607,7 @@ GL_PREFIX(CombinerParameteriNV):
 GL_PREFIX(CombinerParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4928(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23173,13 +23617,13 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4928(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23189,7 +23633,7 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterivNV), .-GL_PREFIX(CombinerParameterivNV)
@@ -23200,7 +23644,7 @@ GL_PREFIX(CombinerParameterivNV):
 GL_PREFIX(FinalCombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4936(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23214,13 +23658,13 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4936(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23234,7 +23678,7 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FinalCombinerInputNV), .-GL_PREFIX(FinalCombinerInputNV)
@@ -23245,7 +23689,7 @@ GL_PREFIX(FinalCombinerInputNV):
 GL_PREFIX(GetCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4944(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23259,13 +23703,13 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4944(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23279,7 +23723,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterfvNV), .-GL_PREFIX(GetCombinerInputParameterfvNV)
@@ -23290,7 +23734,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 GL_PREFIX(GetCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4952(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23304,13 +23748,13 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4952(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23324,7 +23768,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterivNV), .-GL_PREFIX(GetCombinerInputParameterivNV)
@@ -23335,7 +23779,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 GL_PREFIX(GetCombinerOutputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4960(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23349,13 +23793,13 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4960(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23369,7 +23813,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterfvNV), .-GL_PREFIX(GetCombinerOutputParameterfvNV)
@@ -23380,7 +23824,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 GL_PREFIX(GetCombinerOutputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4968(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23394,13 +23838,13 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4968(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23414,7 +23858,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterivNV), .-GL_PREFIX(GetCombinerOutputParameterivNV)
@@ -23425,7 +23869,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4976(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23435,13 +23879,13 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4976(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23451,7 +23895,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterfvNV), .-GL_PREFIX(GetFinalCombinerInputParameterfvNV)
@@ -23462,7 +23906,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 GL_PREFIX(GetFinalCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4984(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23472,13 +23916,13 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4984(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4984(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23488,7 +23932,7 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4984(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterivNV), .-GL_PREFIX(GetFinalCombinerInputParameterivNV)
@@ -23499,25 +23943,25 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 GL_PREFIX(ResizeBuffersMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4992(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4992(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4992(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4992(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ResizeBuffersMESA), .-GL_PREFIX(ResizeBuffersMESA)
@@ -23528,7 +23972,7 @@ GL_PREFIX(ResizeBuffersMESA):
 GL_PREFIX(WindowPos2dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5000(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23538,13 +23982,13 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5000(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5000(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23554,7 +23998,7 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5000(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dMESA), .-GL_PREFIX(WindowPos2dMESA)
@@ -23565,25 +24009,25 @@ GL_PREFIX(WindowPos2dMESA):
 GL_PREFIX(WindowPos2dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5008(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5008(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5008(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5008(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dvMESA), .-GL_PREFIX(WindowPos2dvMESA)
@@ -23594,7 +24038,7 @@ GL_PREFIX(WindowPos2dvMESA):
 GL_PREFIX(WindowPos2fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5016(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23604,13 +24048,13 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5016(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5016(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23620,7 +24064,7 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5016(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fMESA), .-GL_PREFIX(WindowPos2fMESA)
@@ -23631,25 +24075,25 @@ GL_PREFIX(WindowPos2fMESA):
 GL_PREFIX(WindowPos2fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5024(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5024(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fvMESA), .-GL_PREFIX(WindowPos2fvMESA)
@@ -23660,7 +24104,7 @@ GL_PREFIX(WindowPos2fvMESA):
 GL_PREFIX(WindowPos2iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5032(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23670,13 +24114,13 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5032(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23686,7 +24130,7 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2iMESA), .-GL_PREFIX(WindowPos2iMESA)
@@ -23697,25 +24141,25 @@ GL_PREFIX(WindowPos2iMESA):
 GL_PREFIX(WindowPos2ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5040(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5040(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2ivMESA), .-GL_PREFIX(WindowPos2ivMESA)
@@ -23726,7 +24170,7 @@ GL_PREFIX(WindowPos2ivMESA):
 GL_PREFIX(WindowPos2sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5048(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23736,13 +24180,13 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5048(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23752,7 +24196,7 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2sMESA), .-GL_PREFIX(WindowPos2sMESA)
@@ -23763,25 +24207,25 @@ GL_PREFIX(WindowPos2sMESA):
 GL_PREFIX(WindowPos2svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5056(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5056(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5056(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5056(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2svMESA), .-GL_PREFIX(WindowPos2svMESA)
@@ -23792,7 +24236,7 @@ GL_PREFIX(WindowPos2svMESA):
 GL_PREFIX(WindowPos3dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5064(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23804,13 +24248,13 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5064(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5064(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23822,7 +24266,7 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5064(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dMESA), .-GL_PREFIX(WindowPos3dMESA)
@@ -23833,25 +24277,25 @@ GL_PREFIX(WindowPos3dMESA):
 GL_PREFIX(WindowPos3dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5072(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5072(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5072(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5072(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dvMESA), .-GL_PREFIX(WindowPos3dvMESA)
@@ -23862,7 +24306,7 @@ GL_PREFIX(WindowPos3dvMESA):
 GL_PREFIX(WindowPos3fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5080(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23874,13 +24318,13 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5080(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5080(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23892,7 +24336,7 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5080(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fMESA), .-GL_PREFIX(WindowPos3fMESA)
@@ -23903,25 +24347,25 @@ GL_PREFIX(WindowPos3fMESA):
 GL_PREFIX(WindowPos3fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5088(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5088(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fvMESA), .-GL_PREFIX(WindowPos3fvMESA)
@@ -23932,7 +24376,7 @@ GL_PREFIX(WindowPos3fvMESA):
 GL_PREFIX(WindowPos3iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5096(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23942,13 +24386,13 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5096(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23958,7 +24402,7 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3iMESA), .-GL_PREFIX(WindowPos3iMESA)
@@ -23969,25 +24413,25 @@ GL_PREFIX(WindowPos3iMESA):
 GL_PREFIX(WindowPos3ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5104(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5104(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3ivMESA), .-GL_PREFIX(WindowPos3ivMESA)
@@ -23998,7 +24442,7 @@ GL_PREFIX(WindowPos3ivMESA):
 GL_PREFIX(WindowPos3sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5112(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24008,13 +24452,13 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5112(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24024,7 +24468,7 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3sMESA), .-GL_PREFIX(WindowPos3sMESA)
@@ -24035,25 +24479,25 @@ GL_PREFIX(WindowPos3sMESA):
 GL_PREFIX(WindowPos3svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5120(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5120(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5120(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5120(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3svMESA), .-GL_PREFIX(WindowPos3svMESA)
@@ -24064,7 +24508,7 @@ GL_PREFIX(WindowPos3svMESA):
 GL_PREFIX(WindowPos4dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5128(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24078,13 +24522,13 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5128(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5128(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24098,7 +24542,7 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5128(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dMESA), .-GL_PREFIX(WindowPos4dMESA)
@@ -24109,25 +24553,25 @@ GL_PREFIX(WindowPos4dMESA):
 GL_PREFIX(WindowPos4dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5136(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5136(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5136(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5136(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dvMESA), .-GL_PREFIX(WindowPos4dvMESA)
@@ -24138,7 +24582,7 @@ GL_PREFIX(WindowPos4dvMESA):
 GL_PREFIX(WindowPos4fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5144(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24152,13 +24596,13 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5144(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5144(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24172,7 +24616,7 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5144(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fMESA), .-GL_PREFIX(WindowPos4fMESA)
@@ -24183,25 +24627,25 @@ GL_PREFIX(WindowPos4fMESA):
 GL_PREFIX(WindowPos4fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5152(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5152(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fvMESA), .-GL_PREFIX(WindowPos4fvMESA)
@@ -24212,7 +24656,7 @@ GL_PREFIX(WindowPos4fvMESA):
 GL_PREFIX(WindowPos4iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5160(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24226,13 +24670,13 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5160(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24246,7 +24690,7 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4iMESA), .-GL_PREFIX(WindowPos4iMESA)
@@ -24257,25 +24701,25 @@ GL_PREFIX(WindowPos4iMESA):
 GL_PREFIX(WindowPos4ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5168(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5168(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4ivMESA), .-GL_PREFIX(WindowPos4ivMESA)
@@ -24286,7 +24730,7 @@ GL_PREFIX(WindowPos4ivMESA):
 GL_PREFIX(WindowPos4sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5176(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24300,13 +24744,13 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5176(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5176(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24320,7 +24764,7 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5176(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4sMESA), .-GL_PREFIX(WindowPos4sMESA)
@@ -24331,37 +24775,37 @@ GL_PREFIX(WindowPos4sMESA):
 GL_PREFIX(WindowPos4svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5184(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5184(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5184(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5184(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4svMESA), .-GL_PREFIX(WindowPos4svMESA)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_649)
-	.type	GL_PREFIX(_dispatch_stub_649), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_649))
-GL_PREFIX(_dispatch_stub_649):
+	.globl	GL_PREFIX(_dispatch_stub_661)
+	.type	GL_PREFIX(_dispatch_stub_661), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_661))
+GL_PREFIX(_dispatch_stub_661):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5192(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24375,13 +24819,13 @@ GL_PREFIX(_dispatch_stub_649):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5192(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5192(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24395,19 +24839,19 @@ GL_PREFIX(_dispatch_stub_649):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5192(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_649), .-GL_PREFIX(_dispatch_stub_649)
+	.size	GL_PREFIX(_dispatch_stub_661), .-GL_PREFIX(_dispatch_stub_661)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_650)
-	.type	GL_PREFIX(_dispatch_stub_650), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_650))
-GL_PREFIX(_dispatch_stub_650):
+	.globl	GL_PREFIX(_dispatch_stub_662)
+	.type	GL_PREFIX(_dispatch_stub_662), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_662))
+GL_PREFIX(_dispatch_stub_662):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5200(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24425,13 +24869,13 @@ GL_PREFIX(_dispatch_stub_650):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5200(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5200(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24449,19 +24893,19 @@ GL_PREFIX(_dispatch_stub_650):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5200(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_650), .-GL_PREFIX(_dispatch_stub_650)
+	.size	GL_PREFIX(_dispatch_stub_662), .-GL_PREFIX(_dispatch_stub_662)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_651)
-	.type	GL_PREFIX(_dispatch_stub_651), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_651))
-GL_PREFIX(_dispatch_stub_651):
+	.globl	GL_PREFIX(_dispatch_stub_663)
+	.type	GL_PREFIX(_dispatch_stub_663), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_663))
+GL_PREFIX(_dispatch_stub_663):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5208(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24471,13 +24915,13 @@ GL_PREFIX(_dispatch_stub_651):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5208(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5208(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24487,49 +24931,49 @@ GL_PREFIX(_dispatch_stub_651):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5208(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_651), .-GL_PREFIX(_dispatch_stub_651)
+	.size	GL_PREFIX(_dispatch_stub_663), .-GL_PREFIX(_dispatch_stub_663)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_652)
-	.type	GL_PREFIX(_dispatch_stub_652), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_652))
-GL_PREFIX(_dispatch_stub_652):
+	.globl	GL_PREFIX(_dispatch_stub_664)
+	.type	GL_PREFIX(_dispatch_stub_664), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_664))
+GL_PREFIX(_dispatch_stub_664):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5216(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5216(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5216(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5216(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_652), .-GL_PREFIX(_dispatch_stub_652)
+	.size	GL_PREFIX(_dispatch_stub_664), .-GL_PREFIX(_dispatch_stub_664)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_653)
-	.type	GL_PREFIX(_dispatch_stub_653), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_653))
-GL_PREFIX(_dispatch_stub_653):
+	.globl	GL_PREFIX(_dispatch_stub_665)
+	.type	GL_PREFIX(_dispatch_stub_665), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_665))
+GL_PREFIX(_dispatch_stub_665):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5224(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24539,13 +24983,13 @@ GL_PREFIX(_dispatch_stub_653):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5224(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5224(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24555,19 +24999,19 @@ GL_PREFIX(_dispatch_stub_653):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5224(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_653), .-GL_PREFIX(_dispatch_stub_653)
+	.size	GL_PREFIX(_dispatch_stub_665), .-GL_PREFIX(_dispatch_stub_665)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_654)
-	.type	GL_PREFIX(_dispatch_stub_654), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_654))
-GL_PREFIX(_dispatch_stub_654):
+	.globl	GL_PREFIX(_dispatch_stub_666)
+	.type	GL_PREFIX(_dispatch_stub_666), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_666))
+GL_PREFIX(_dispatch_stub_666):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5232(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24577,13 +25021,13 @@ GL_PREFIX(_dispatch_stub_654):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5232(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5232(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24593,49 +25037,49 @@ GL_PREFIX(_dispatch_stub_654):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5232(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_654), .-GL_PREFIX(_dispatch_stub_654)
+	.size	GL_PREFIX(_dispatch_stub_666), .-GL_PREFIX(_dispatch_stub_666)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_655)
-	.type	GL_PREFIX(_dispatch_stub_655), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_655))
-GL_PREFIX(_dispatch_stub_655):
+	.globl	GL_PREFIX(_dispatch_stub_667)
+	.type	GL_PREFIX(_dispatch_stub_667), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_667))
+GL_PREFIX(_dispatch_stub_667):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5240(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5240(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5240(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5240(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_655), .-GL_PREFIX(_dispatch_stub_655)
+	.size	GL_PREFIX(_dispatch_stub_667), .-GL_PREFIX(_dispatch_stub_667)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_656)
-	.type	GL_PREFIX(_dispatch_stub_656), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_656))
-GL_PREFIX(_dispatch_stub_656):
+	.globl	GL_PREFIX(_dispatch_stub_668)
+	.type	GL_PREFIX(_dispatch_stub_668), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_668))
+GL_PREFIX(_dispatch_stub_668):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5248(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24645,13 +25089,13 @@ GL_PREFIX(_dispatch_stub_656):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5248(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5248(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24661,40 +25105,40 @@ GL_PREFIX(_dispatch_stub_656):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5248(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_656), .-GL_PREFIX(_dispatch_stub_656)
+	.size	GL_PREFIX(_dispatch_stub_668), .-GL_PREFIX(_dispatch_stub_668)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_657)
-	.type	GL_PREFIX(_dispatch_stub_657), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_657))
-GL_PREFIX(_dispatch_stub_657):
+	.globl	GL_PREFIX(_dispatch_stub_669)
+	.type	GL_PREFIX(_dispatch_stub_669), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_669))
+GL_PREFIX(_dispatch_stub_669):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5256(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5256(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_657), .-GL_PREFIX(_dispatch_stub_657)
+	.size	GL_PREFIX(_dispatch_stub_669), .-GL_PREFIX(_dispatch_stub_669)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(AreProgramsResidentNV)
@@ -24702,7 +25146,7 @@ GL_PREFIX(_dispatch_stub_657):
 GL_PREFIX(AreProgramsResidentNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5264(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24712,13 +25156,13 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5264(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24728,7 +25172,7 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AreProgramsResidentNV), .-GL_PREFIX(AreProgramsResidentNV)
@@ -24739,7 +25183,7 @@ GL_PREFIX(AreProgramsResidentNV):
 GL_PREFIX(BindProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5272(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24749,13 +25193,13 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5272(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24765,7 +25209,7 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindProgramNV), .-GL_PREFIX(BindProgramNV)
@@ -24776,7 +25220,7 @@ GL_PREFIX(BindProgramNV):
 GL_PREFIX(DeleteProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5280(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24786,13 +25230,13 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5280(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24802,7 +25246,7 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteProgramsNV), .-GL_PREFIX(DeleteProgramsNV)
@@ -24813,7 +25257,7 @@ GL_PREFIX(DeleteProgramsNV):
 GL_PREFIX(ExecuteProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5288(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24823,13 +25267,13 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5288(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24839,7 +25283,7 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ExecuteProgramNV), .-GL_PREFIX(ExecuteProgramNV)
@@ -24850,7 +25294,7 @@ GL_PREFIX(ExecuteProgramNV):
 GL_PREFIX(GenProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5296(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24860,13 +25304,13 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5296(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24876,7 +25320,7 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenProgramsNV), .-GL_PREFIX(GenProgramsNV)
@@ -24887,7 +25331,7 @@ GL_PREFIX(GenProgramsNV):
 GL_PREFIX(GetProgramParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5304(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24901,13 +25345,13 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5304(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24921,7 +25365,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterdvNV), .-GL_PREFIX(GetProgramParameterdvNV)
@@ -24932,7 +25376,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 GL_PREFIX(GetProgramParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5312(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24946,13 +25390,13 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5312(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24966,7 +25410,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterfvNV), .-GL_PREFIX(GetProgramParameterfvNV)
@@ -24977,7 +25421,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 GL_PREFIX(GetProgramStringNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5320(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24987,13 +25431,13 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5320(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25003,7 +25447,7 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramStringNV), .-GL_PREFIX(GetProgramStringNV)
@@ -25014,7 +25458,7 @@ GL_PREFIX(GetProgramStringNV):
 GL_PREFIX(GetProgramivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5328(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25024,13 +25468,13 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5328(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25040,7 +25484,7 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramivNV), .-GL_PREFIX(GetProgramivNV)
@@ -25051,7 +25495,7 @@ GL_PREFIX(GetProgramivNV):
 GL_PREFIX(GetTrackMatrixivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5336(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25065,13 +25509,13 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5336(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25085,7 +25529,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTrackMatrixivNV), .-GL_PREFIX(GetTrackMatrixivNV)
@@ -25096,7 +25540,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 GL_PREFIX(GetVertexAttribPointervNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5344(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25106,13 +25550,13 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5344(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25122,7 +25566,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribPointervNV), .-GL_PREFIX(GetVertexAttribPointervNV)
@@ -25133,7 +25577,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 GL_PREFIX(GetVertexAttribdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5352(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25143,13 +25587,13 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5352(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25159,7 +25603,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribdvNV), .-GL_PREFIX(GetVertexAttribdvNV)
@@ -25170,7 +25614,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 GL_PREFIX(GetVertexAttribfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5360(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25180,13 +25624,13 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5360(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25196,7 +25640,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribfvNV), .-GL_PREFIX(GetVertexAttribfvNV)
@@ -25207,7 +25651,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 GL_PREFIX(GetVertexAttribivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5368(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25217,13 +25661,13 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5368(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25233,7 +25677,7 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribivNV), .-GL_PREFIX(GetVertexAttribivNV)
@@ -25244,25 +25688,25 @@ GL_PREFIX(GetVertexAttribivNV):
 GL_PREFIX(IsProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5376(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5376(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsProgramNV), .-GL_PREFIX(IsProgramNV)
@@ -25273,7 +25717,7 @@ GL_PREFIX(IsProgramNV):
 GL_PREFIX(LoadProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5384(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25287,13 +25731,13 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5384(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25307,7 +25751,7 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LoadProgramNV), .-GL_PREFIX(LoadProgramNV)
@@ -25318,7 +25762,7 @@ GL_PREFIX(LoadProgramNV):
 GL_PREFIX(ProgramParameters4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5392(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25332,13 +25776,13 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5392(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25352,7 +25796,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4dvNV), .-GL_PREFIX(ProgramParameters4dvNV)
@@ -25363,7 +25807,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 GL_PREFIX(ProgramParameters4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5400(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25377,13 +25821,13 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5400(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25397,7 +25841,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4fvNV), .-GL_PREFIX(ProgramParameters4fvNV)
@@ -25408,7 +25852,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 GL_PREFIX(RequestResidentProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5408(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25418,13 +25862,13 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5408(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25434,7 +25878,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RequestResidentProgramsNV), .-GL_PREFIX(RequestResidentProgramsNV)
@@ -25445,7 +25889,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 GL_PREFIX(TrackMatrixNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5416(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25459,13 +25903,13 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5416(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5416(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25479,7 +25923,7 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5416(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TrackMatrixNV), .-GL_PREFIX(TrackMatrixNV)
@@ -25490,7 +25934,7 @@ GL_PREFIX(TrackMatrixNV):
 GL_PREFIX(VertexAttrib1dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5424(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25500,13 +25944,13 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5424(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5424(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25516,7 +25960,7 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5424(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dNV), .-GL_PREFIX(VertexAttrib1dNV)
@@ -25527,7 +25971,7 @@ GL_PREFIX(VertexAttrib1dNV):
 GL_PREFIX(VertexAttrib1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5432(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25537,13 +25981,13 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5432(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5432(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25553,7 +25997,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5432(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dvNV), .-GL_PREFIX(VertexAttrib1dvNV)
@@ -25564,7 +26008,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 GL_PREFIX(VertexAttrib1fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5440(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25574,13 +26018,13 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5440(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5440(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25590,7 +26034,7 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5440(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fNV), .-GL_PREFIX(VertexAttrib1fNV)
@@ -25601,7 +26045,7 @@ GL_PREFIX(VertexAttrib1fNV):
 GL_PREFIX(VertexAttrib1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5448(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25611,13 +26055,13 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5448(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25627,7 +26071,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fvNV), .-GL_PREFIX(VertexAttrib1fvNV)
@@ -25638,7 +26082,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 GL_PREFIX(VertexAttrib1sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5456(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25648,13 +26092,13 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5456(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25664,7 +26108,7 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1sNV), .-GL_PREFIX(VertexAttrib1sNV)
@@ -25675,7 +26119,7 @@ GL_PREFIX(VertexAttrib1sNV):
 GL_PREFIX(VertexAttrib1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5464(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25685,13 +26129,13 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5464(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5464(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25701,7 +26145,7 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5464(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1svNV), .-GL_PREFIX(VertexAttrib1svNV)
@@ -25712,7 +26156,7 @@ GL_PREFIX(VertexAttrib1svNV):
 GL_PREFIX(VertexAttrib2dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5472(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25724,13 +26168,13 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5472(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5472(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25742,7 +26186,7 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5472(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dNV), .-GL_PREFIX(VertexAttrib2dNV)
@@ -25753,7 +26197,7 @@ GL_PREFIX(VertexAttrib2dNV):
 GL_PREFIX(VertexAttrib2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5480(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25763,13 +26207,13 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5480(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5480(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25779,7 +26223,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5480(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dvNV), .-GL_PREFIX(VertexAttrib2dvNV)
@@ -25790,7 +26234,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 GL_PREFIX(VertexAttrib2fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5488(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25802,13 +26246,13 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5488(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5488(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25820,7 +26264,7 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5488(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fNV), .-GL_PREFIX(VertexAttrib2fNV)
@@ -25831,7 +26275,7 @@ GL_PREFIX(VertexAttrib2fNV):
 GL_PREFIX(VertexAttrib2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5496(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25841,13 +26285,13 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5496(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25857,7 +26301,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fvNV), .-GL_PREFIX(VertexAttrib2fvNV)
@@ -25868,7 +26312,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 GL_PREFIX(VertexAttrib2sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5504(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25878,13 +26322,13 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5504(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25894,7 +26338,7 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2sNV), .-GL_PREFIX(VertexAttrib2sNV)
@@ -25905,7 +26349,7 @@ GL_PREFIX(VertexAttrib2sNV):
 GL_PREFIX(VertexAttrib2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5512(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25915,13 +26359,13 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5512(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5512(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25931,7 +26375,7 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5512(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2svNV), .-GL_PREFIX(VertexAttrib2svNV)
@@ -25942,7 +26386,7 @@ GL_PREFIX(VertexAttrib2svNV):
 GL_PREFIX(VertexAttrib3dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5520(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -25956,13 +26400,13 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5520(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5520(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -25976,7 +26420,7 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5520(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dNV), .-GL_PREFIX(VertexAttrib3dNV)
@@ -25987,7 +26431,7 @@ GL_PREFIX(VertexAttrib3dNV):
 GL_PREFIX(VertexAttrib3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5528(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25997,13 +26441,13 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5528(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5528(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26013,7 +26457,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5528(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dvNV), .-GL_PREFIX(VertexAttrib3dvNV)
@@ -26024,7 +26468,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 GL_PREFIX(VertexAttrib3fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5536(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26038,13 +26482,13 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5536(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5536(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26058,7 +26502,7 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5536(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fNV), .-GL_PREFIX(VertexAttrib3fNV)
@@ -26069,7 +26513,7 @@ GL_PREFIX(VertexAttrib3fNV):
 GL_PREFIX(VertexAttrib3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5544(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26079,13 +26523,13 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5544(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26095,7 +26539,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fvNV), .-GL_PREFIX(VertexAttrib3fvNV)
@@ -26106,7 +26550,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 GL_PREFIX(VertexAttrib3sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5552(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26120,13 +26564,13 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5552(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26140,7 +26584,7 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3sNV), .-GL_PREFIX(VertexAttrib3sNV)
@@ -26151,7 +26595,7 @@ GL_PREFIX(VertexAttrib3sNV):
 GL_PREFIX(VertexAttrib3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5560(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26161,13 +26605,13 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5560(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5560(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26177,7 +26621,7 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5560(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3svNV), .-GL_PREFIX(VertexAttrib3svNV)
@@ -26188,7 +26632,7 @@ GL_PREFIX(VertexAttrib3svNV):
 GL_PREFIX(VertexAttrib4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5568(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26204,13 +26648,13 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5568(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5568(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26226,7 +26670,7 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5568(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dNV), .-GL_PREFIX(VertexAttrib4dNV)
@@ -26237,7 +26681,7 @@ GL_PREFIX(VertexAttrib4dNV):
 GL_PREFIX(VertexAttrib4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5576(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26247,13 +26691,13 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5576(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5576(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26263,7 +26707,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5576(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dvNV), .-GL_PREFIX(VertexAttrib4dvNV)
@@ -26274,7 +26718,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 GL_PREFIX(VertexAttrib4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5584(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26290,13 +26734,13 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5584(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5584(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26312,7 +26756,7 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5584(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fNV), .-GL_PREFIX(VertexAttrib4fNV)
@@ -26323,7 +26767,7 @@ GL_PREFIX(VertexAttrib4fNV):
 GL_PREFIX(VertexAttrib4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5592(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26333,13 +26777,13 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5592(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26349,7 +26793,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fvNV), .-GL_PREFIX(VertexAttrib4fvNV)
@@ -26360,7 +26804,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 GL_PREFIX(VertexAttrib4sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5600(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26374,13 +26818,13 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5600(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26394,7 +26838,7 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4sNV), .-GL_PREFIX(VertexAttrib4sNV)
@@ -26405,7 +26849,7 @@ GL_PREFIX(VertexAttrib4sNV):
 GL_PREFIX(VertexAttrib4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5608(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26415,13 +26859,13 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5608(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26431,7 +26875,7 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4svNV), .-GL_PREFIX(VertexAttrib4svNV)
@@ -26442,7 +26886,7 @@ GL_PREFIX(VertexAttrib4svNV):
 GL_PREFIX(VertexAttrib4ubNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5616(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26456,13 +26900,13 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5616(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26476,7 +26920,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubNV), .-GL_PREFIX(VertexAttrib4ubNV)
@@ -26487,7 +26931,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 GL_PREFIX(VertexAttrib4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5624(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26497,13 +26941,13 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5624(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26513,7 +26957,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubvNV), .-GL_PREFIX(VertexAttrib4ubvNV)
@@ -26524,7 +26968,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 GL_PREFIX(VertexAttribPointerNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5632(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26538,13 +26982,13 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5632(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26558,7 +27002,7 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribPointerNV), .-GL_PREFIX(VertexAttribPointerNV)
@@ -26569,7 +27013,7 @@ GL_PREFIX(VertexAttribPointerNV):
 GL_PREFIX(VertexAttribs1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5640(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26579,13 +27023,13 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5640(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26595,7 +27039,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1dvNV), .-GL_PREFIX(VertexAttribs1dvNV)
@@ -26606,7 +27050,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 GL_PREFIX(VertexAttribs1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5648(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26616,13 +27060,13 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5648(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26632,7 +27076,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1fvNV), .-GL_PREFIX(VertexAttribs1fvNV)
@@ -26643,7 +27087,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 GL_PREFIX(VertexAttribs1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5656(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26653,13 +27097,13 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5656(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26669,7 +27113,7 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1svNV), .-GL_PREFIX(VertexAttribs1svNV)
@@ -26680,7 +27124,7 @@ GL_PREFIX(VertexAttribs1svNV):
 GL_PREFIX(VertexAttribs2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5664(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26690,13 +27134,13 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5664(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26706,7 +27150,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2dvNV), .-GL_PREFIX(VertexAttribs2dvNV)
@@ -26717,7 +27161,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 GL_PREFIX(VertexAttribs2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5672(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26727,13 +27171,13 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5672(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26743,7 +27187,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2fvNV), .-GL_PREFIX(VertexAttribs2fvNV)
@@ -26754,7 +27198,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 GL_PREFIX(VertexAttribs2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5680(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26764,13 +27208,13 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5680(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26780,7 +27224,7 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2svNV), .-GL_PREFIX(VertexAttribs2svNV)
@@ -26791,7 +27235,7 @@ GL_PREFIX(VertexAttribs2svNV):
 GL_PREFIX(VertexAttribs3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5688(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26801,13 +27245,13 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5688(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26817,7 +27261,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3dvNV), .-GL_PREFIX(VertexAttribs3dvNV)
@@ -26828,7 +27272,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 GL_PREFIX(VertexAttribs3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5696(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26838,13 +27282,13 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5696(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26854,7 +27298,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3fvNV), .-GL_PREFIX(VertexAttribs3fvNV)
@@ -26865,7 +27309,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 GL_PREFIX(VertexAttribs3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5704(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26875,13 +27319,13 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5704(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26891,7 +27335,7 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3svNV), .-GL_PREFIX(VertexAttribs3svNV)
@@ -26902,7 +27346,7 @@ GL_PREFIX(VertexAttribs3svNV):
 GL_PREFIX(VertexAttribs4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5712(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26912,13 +27356,13 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5712(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26928,7 +27372,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4dvNV), .-GL_PREFIX(VertexAttribs4dvNV)
@@ -26939,7 +27383,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 GL_PREFIX(VertexAttribs4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5720(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26949,13 +27393,13 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5720(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26965,7 +27409,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4fvNV), .-GL_PREFIX(VertexAttribs4fvNV)
@@ -26976,7 +27420,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 GL_PREFIX(VertexAttribs4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5728(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26986,13 +27430,13 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5728(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27002,7 +27446,7 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4svNV), .-GL_PREFIX(VertexAttribs4svNV)
@@ -27013,7 +27457,7 @@ GL_PREFIX(VertexAttribs4svNV):
 GL_PREFIX(VertexAttribs4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5736(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27023,13 +27467,13 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5736(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27039,7 +27483,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4ubvNV), .-GL_PREFIX(VertexAttribs4ubvNV)
@@ -27050,7 +27494,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 GL_PREFIX(GetTexBumpParameterfvATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5744(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27060,13 +27504,13 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5744(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27076,7 +27520,7 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTexBumpParameterfvATI), .-GL_PREFIX(GetTexBumpParameterfvATI)
@@ -27087,7 +27531,7 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 GL_PREFIX(GetTexBumpParameterivATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5752(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27097,13 +27541,13 @@ GL_PREFIX(GetTexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5752(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27113,7 +27557,7 @@ GL_PREFIX(GetTexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTexBumpParameterivATI), .-GL_PREFIX(GetTexBumpParameterivATI)
@@ -27124,7 +27568,7 @@ GL_PREFIX(GetTexBumpParameterivATI):
 GL_PREFIX(TexBumpParameterfvATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5760(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27134,13 +27578,13 @@ GL_PREFIX(TexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5760(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5760(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27150,7 +27594,7 @@ GL_PREFIX(TexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5760(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexBumpParameterfvATI), .-GL_PREFIX(TexBumpParameterfvATI)
@@ -27161,7 +27605,7 @@ GL_PREFIX(TexBumpParameterfvATI):
 GL_PREFIX(TexBumpParameterivATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5768(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27171,13 +27615,13 @@ GL_PREFIX(TexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5768(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27187,7 +27631,7 @@ GL_PREFIX(TexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexBumpParameterivATI), .-GL_PREFIX(TexBumpParameterivATI)
@@ -27198,7 +27642,7 @@ GL_PREFIX(TexBumpParameterivATI):
 GL_PREFIX(AlphaFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5776(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27216,13 +27660,13 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5776(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27240,7 +27684,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp1ATI), .-GL_PREFIX(AlphaFragmentOp1ATI)
@@ -27251,7 +27695,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 GL_PREFIX(AlphaFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5784(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27269,13 +27713,13 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5784(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27293,7 +27737,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp2ATI), .-GL_PREFIX(AlphaFragmentOp2ATI)
@@ -27304,7 +27748,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 GL_PREFIX(AlphaFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5792(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27322,13 +27766,13 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5792(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27346,7 +27790,7 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp3ATI), .-GL_PREFIX(AlphaFragmentOp3ATI)
@@ -27357,25 +27801,25 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 GL_PREFIX(BeginFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5800(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5800(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5800(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5800(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BeginFragmentShaderATI), .-GL_PREFIX(BeginFragmentShaderATI)
@@ -27386,25 +27830,25 @@ GL_PREFIX(BeginFragmentShaderATI):
 GL_PREFIX(BindFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5808(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5808(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5808(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5808(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFragmentShaderATI), .-GL_PREFIX(BindFragmentShaderATI)
@@ -27415,7 +27859,7 @@ GL_PREFIX(BindFragmentShaderATI):
 GL_PREFIX(ColorFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5816(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27433,13 +27877,13 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5816(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27457,7 +27901,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp1ATI), .-GL_PREFIX(ColorFragmentOp1ATI)
@@ -27468,7 +27912,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 GL_PREFIX(ColorFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5824(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27486,13 +27930,13 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5824(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27510,7 +27954,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp2ATI), .-GL_PREFIX(ColorFragmentOp2ATI)
@@ -27521,7 +27965,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 GL_PREFIX(ColorFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5832(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27539,13 +27983,13 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5832(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27563,7 +28007,7 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp3ATI), .-GL_PREFIX(ColorFragmentOp3ATI)
@@ -27574,25 +28018,25 @@ GL_PREFIX(ColorFragmentOp3ATI):
 GL_PREFIX(DeleteFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5840(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5840(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFragmentShaderATI), .-GL_PREFIX(DeleteFragmentShaderATI)
@@ -27603,25 +28047,25 @@ GL_PREFIX(DeleteFragmentShaderATI):
 GL_PREFIX(EndFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5848(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5848(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5848(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5848(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EndFragmentShaderATI), .-GL_PREFIX(EndFragmentShaderATI)
@@ -27632,25 +28076,25 @@ GL_PREFIX(EndFragmentShaderATI):
 GL_PREFIX(GenFragmentShadersATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5856(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5856(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5856(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5856(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFragmentShadersATI), .-GL_PREFIX(GenFragmentShadersATI)
@@ -27661,7 +28105,7 @@ GL_PREFIX(GenFragmentShadersATI):
 GL_PREFIX(PassTexCoordATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5864(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27671,13 +28115,13 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5864(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5864(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27687,7 +28131,7 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5864(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PassTexCoordATI), .-GL_PREFIX(PassTexCoordATI)
@@ -27698,7 +28142,7 @@ GL_PREFIX(PassTexCoordATI):
 GL_PREFIX(SampleMapATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5872(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27708,13 +28152,13 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5872(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5872(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27724,7 +28168,7 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5872(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SampleMapATI), .-GL_PREFIX(SampleMapATI)
@@ -27735,7 +28179,7 @@ GL_PREFIX(SampleMapATI):
 GL_PREFIX(SetFragmentShaderConstantATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5880(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27745,13 +28189,13 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5880(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5880(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27761,7 +28205,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5880(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SetFragmentShaderConstantATI), .-GL_PREFIX(SetFragmentShaderConstantATI)
@@ -27772,7 +28216,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 GL_PREFIX(PointParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5888(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27782,13 +28226,13 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5888(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5888(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27798,7 +28242,7 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5888(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameteriNV), .-GL_PREFIX(PointParameteriNV)
@@ -27809,7 +28253,7 @@ GL_PREFIX(PointParameteriNV):
 GL_PREFIX(PointParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5896(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27819,13 +28263,13 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5896(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5896(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27835,79 +28279,79 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5896(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterivNV), .-GL_PREFIX(PointParameterivNV)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_738)
-	.type	GL_PREFIX(_dispatch_stub_738), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_738))
-GL_PREFIX(_dispatch_stub_738):
+	.globl	GL_PREFIX(_dispatch_stub_750)
+	.type	GL_PREFIX(_dispatch_stub_750), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_750))
+GL_PREFIX(_dispatch_stub_750):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5904(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5904(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_738), .-GL_PREFIX(_dispatch_stub_738)
+	.size	GL_PREFIX(_dispatch_stub_750), .-GL_PREFIX(_dispatch_stub_750)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_739)
-	.type	GL_PREFIX(_dispatch_stub_739), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_739))
-GL_PREFIX(_dispatch_stub_739):
+	.globl	GL_PREFIX(_dispatch_stub_751)
+	.type	GL_PREFIX(_dispatch_stub_751), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_751))
+GL_PREFIX(_dispatch_stub_751):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5912(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5912(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_739), .-GL_PREFIX(_dispatch_stub_739)
+	.size	GL_PREFIX(_dispatch_stub_751), .-GL_PREFIX(_dispatch_stub_751)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_740)
-	.type	GL_PREFIX(_dispatch_stub_740), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_740))
-GL_PREFIX(_dispatch_stub_740):
+	.globl	GL_PREFIX(_dispatch_stub_752)
+	.type	GL_PREFIX(_dispatch_stub_752), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_752))
+GL_PREFIX(_dispatch_stub_752):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5920(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27917,13 +28361,13 @@ GL_PREFIX(_dispatch_stub_740):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5920(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5920(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27933,19 +28377,19 @@ GL_PREFIX(_dispatch_stub_740):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5920(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_740), .-GL_PREFIX(_dispatch_stub_740)
+	.size	GL_PREFIX(_dispatch_stub_752), .-GL_PREFIX(_dispatch_stub_752)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_741)
-	.type	GL_PREFIX(_dispatch_stub_741), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_741))
-GL_PREFIX(_dispatch_stub_741):
+	.globl	GL_PREFIX(_dispatch_stub_753)
+	.type	GL_PREFIX(_dispatch_stub_753), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_753))
+GL_PREFIX(_dispatch_stub_753):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5928(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27955,13 +28399,13 @@ GL_PREFIX(_dispatch_stub_741):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5928(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27971,40 +28415,40 @@ GL_PREFIX(_dispatch_stub_741):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_741), .-GL_PREFIX(_dispatch_stub_741)
+	.size	GL_PREFIX(_dispatch_stub_753), .-GL_PREFIX(_dispatch_stub_753)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_742)
-	.type	GL_PREFIX(_dispatch_stub_742), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_742))
-GL_PREFIX(_dispatch_stub_742):
+	.globl	GL_PREFIX(_dispatch_stub_754)
+	.type	GL_PREFIX(_dispatch_stub_754), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_754))
+GL_PREFIX(_dispatch_stub_754):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5936(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5936(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5936(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5936(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_742), .-GL_PREFIX(_dispatch_stub_742)
+	.size	GL_PREFIX(_dispatch_stub_754), .-GL_PREFIX(_dispatch_stub_754)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(GetProgramNamedParameterdvNV)
@@ -28012,7 +28456,7 @@ GL_PREFIX(_dispatch_stub_742):
 GL_PREFIX(GetProgramNamedParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5944(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28026,13 +28470,13 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5944(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28046,7 +28490,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterdvNV), .-GL_PREFIX(GetProgramNamedParameterdvNV)
@@ -28057,7 +28501,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 GL_PREFIX(GetProgramNamedParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5952(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28071,13 +28515,13 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5952(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28091,7 +28535,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterfvNV), .-GL_PREFIX(GetProgramNamedParameterfvNV)
@@ -28102,7 +28546,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 GL_PREFIX(ProgramNamedParameter4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5960(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -28122,13 +28566,13 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5960(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5960(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -28148,7 +28592,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5960(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dNV), .-GL_PREFIX(ProgramNamedParameter4dNV)
@@ -28159,7 +28603,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 GL_PREFIX(ProgramNamedParameter4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5968(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28173,13 +28617,13 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5968(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28193,7 +28637,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dvNV), .-GL_PREFIX(ProgramNamedParameter4dvNV)
@@ -28204,7 +28648,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 GL_PREFIX(ProgramNamedParameter4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5976(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -28224,13 +28668,13 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5976(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5976(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -28250,7 +28694,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5976(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fNV), .-GL_PREFIX(ProgramNamedParameter4fNV)
@@ -28261,7 +28705,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 GL_PREFIX(ProgramNamedParameter4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5984(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28275,13 +28719,13 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5984(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28295,19 +28739,19 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fvNV), .-GL_PREFIX(ProgramNamedParameter4fvNV)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_749)
-	.type	GL_PREFIX(_dispatch_stub_749), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_749))
-GL_PREFIX(_dispatch_stub_749):
+	.globl	GL_PREFIX(_dispatch_stub_761)
+	.type	GL_PREFIX(_dispatch_stub_761), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_761))
+GL_PREFIX(_dispatch_stub_761):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5992(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28317,13 +28761,13 @@ GL_PREFIX(_dispatch_stub_749):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5992(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28333,19 +28777,19 @@ GL_PREFIX(_dispatch_stub_749):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_749), .-GL_PREFIX(_dispatch_stub_749)
+	.size	GL_PREFIX(_dispatch_stub_761), .-GL_PREFIX(_dispatch_stub_761)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_750)
-	.type	GL_PREFIX(_dispatch_stub_750), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_750))
-GL_PREFIX(_dispatch_stub_750):
+	.globl	GL_PREFIX(_dispatch_stub_762)
+	.type	GL_PREFIX(_dispatch_stub_762), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_762))
+GL_PREFIX(_dispatch_stub_762):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6000(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28355,13 +28799,13 @@ GL_PREFIX(_dispatch_stub_750):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6000(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28371,10 +28815,10 @@ GL_PREFIX(_dispatch_stub_750):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_750), .-GL_PREFIX(_dispatch_stub_750)
+	.size	GL_PREFIX(_dispatch_stub_762), .-GL_PREFIX(_dispatch_stub_762)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BindFramebufferEXT)
@@ -28382,7 +28826,7 @@ GL_PREFIX(_dispatch_stub_750):
 GL_PREFIX(BindFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6008(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28392,13 +28836,13 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6008(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28408,7 +28852,7 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFramebufferEXT), .-GL_PREFIX(BindFramebufferEXT)
@@ -28419,7 +28863,7 @@ GL_PREFIX(BindFramebufferEXT):
 GL_PREFIX(BindRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6016(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28429,13 +28873,13 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6016(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28445,7 +28889,7 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindRenderbufferEXT), .-GL_PREFIX(BindRenderbufferEXT)
@@ -28456,25 +28900,25 @@ GL_PREFIX(BindRenderbufferEXT):
 GL_PREFIX(CheckFramebufferStatusEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6024(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6024(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CheckFramebufferStatusEXT), .-GL_PREFIX(CheckFramebufferStatusEXT)
@@ -28485,7 +28929,7 @@ GL_PREFIX(CheckFramebufferStatusEXT):
 GL_PREFIX(DeleteFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6032(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28495,13 +28939,13 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6032(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28511,7 +28955,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFramebuffersEXT), .-GL_PREFIX(DeleteFramebuffersEXT)
@@ -28522,7 +28966,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 GL_PREFIX(DeleteRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6040(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28532,13 +28976,13 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6040(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28548,7 +28992,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteRenderbuffersEXT), .-GL_PREFIX(DeleteRenderbuffersEXT)
@@ -28559,7 +29003,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 GL_PREFIX(FramebufferRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6048(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28573,13 +29017,13 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6048(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28593,7 +29037,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferRenderbufferEXT), .-GL_PREFIX(FramebufferRenderbufferEXT)
@@ -28604,7 +29048,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 GL_PREFIX(FramebufferTexture1DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6056(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28618,13 +29062,13 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6056(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28638,7 +29082,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture1DEXT), .-GL_PREFIX(FramebufferTexture1DEXT)
@@ -28649,7 +29093,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 GL_PREFIX(FramebufferTexture2DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6064(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28663,13 +29107,13 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6064(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28683,7 +29127,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture2DEXT), .-GL_PREFIX(FramebufferTexture2DEXT)
@@ -28694,7 +29138,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 GL_PREFIX(FramebufferTexture3DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6072(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28712,13 +29156,13 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6072(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28736,7 +29180,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture3DEXT), .-GL_PREFIX(FramebufferTexture3DEXT)
@@ -28747,7 +29191,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 GL_PREFIX(GenFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6080(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28757,13 +29201,13 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6080(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28773,7 +29217,7 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFramebuffersEXT), .-GL_PREFIX(GenFramebuffersEXT)
@@ -28784,7 +29228,7 @@ GL_PREFIX(GenFramebuffersEXT):
 GL_PREFIX(GenRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6088(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28794,13 +29238,13 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6088(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28810,7 +29254,7 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenRenderbuffersEXT), .-GL_PREFIX(GenRenderbuffersEXT)
@@ -28821,25 +29265,25 @@ GL_PREFIX(GenRenderbuffersEXT):
 GL_PREFIX(GenerateMipmapEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6096(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6096(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenerateMipmapEXT), .-GL_PREFIX(GenerateMipmapEXT)
@@ -28850,7 +29294,7 @@ GL_PREFIX(GenerateMipmapEXT):
 GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6104(%rax), %r11
+	movq	6200(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28864,13 +29308,13 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6200(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6104(%rax), %r11
+	movq	6200(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28884,7 +29328,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6200(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFramebufferAttachmentParameterivEXT), .-GL_PREFIX(GetFramebufferAttachmentParameterivEXT)
@@ -28895,7 +29339,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 GL_PREFIX(GetRenderbufferParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6112(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28905,13 +29349,13 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6112(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28921,7 +29365,7 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetRenderbufferParameterivEXT), .-GL_PREFIX(GetRenderbufferParameterivEXT)
@@ -28932,25 +29376,25 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 GL_PREFIX(IsFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6120(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6120(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsFramebufferEXT), .-GL_PREFIX(IsFramebufferEXT)
@@ -28961,25 +29405,25 @@ GL_PREFIX(IsFramebufferEXT):
 GL_PREFIX(IsRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6128(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6128(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsRenderbufferEXT), .-GL_PREFIX(IsRenderbufferEXT)
@@ -28990,7 +29434,7 @@ GL_PREFIX(IsRenderbufferEXT):
 GL_PREFIX(RenderbufferStorageEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6136(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29004,13 +29448,13 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6136(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29024,19 +29468,19 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RenderbufferStorageEXT), .-GL_PREFIX(RenderbufferStorageEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_768)
-	.type	GL_PREFIX(_dispatch_stub_768), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_768))
-GL_PREFIX(_dispatch_stub_768):
+	.globl	GL_PREFIX(_dispatch_stub_780)
+	.type	GL_PREFIX(_dispatch_stub_780), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_780))
+GL_PREFIX(_dispatch_stub_780):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6144(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29054,13 +29498,13 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6144(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29078,10 +29522,86 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_768), .-GL_PREFIX(_dispatch_stub_768)
+	.size	GL_PREFIX(_dispatch_stub_780), .-GL_PREFIX(_dispatch_stub_780)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_781)
+	.type	GL_PREFIX(_dispatch_stub_781), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_781))
+GL_PREFIX(_dispatch_stub_781):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6248(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6248(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6248(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6248(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_781), .-GL_PREFIX(_dispatch_stub_781)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_782)
+	.type	GL_PREFIX(_dispatch_stub_782), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_782))
+GL_PREFIX(_dispatch_stub_782):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6256(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6256(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6256(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6256(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_782), .-GL_PREFIX(_dispatch_stub_782)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(FramebufferTextureLayerEXT)
@@ -29089,7 +29609,7 @@ GL_PREFIX(_dispatch_stub_768):
 GL_PREFIX(FramebufferTextureLayerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6152(%rax), %r11
+	movq	6264(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29103,13 +29623,13 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6264(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6152(%rax), %r11
+	movq	6264(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29123,19 +29643,124 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6264(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTextureLayerEXT), .-GL_PREFIX(FramebufferTextureLayerEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_770)
-	.type	GL_PREFIX(_dispatch_stub_770), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_770))
-GL_PREFIX(_dispatch_stub_770):
+	.globl	GL_PREFIX(ProvokingVertexEXT)
+	.type	GL_PREFIX(ProvokingVertexEXT), @function
+GL_PREFIX(ProvokingVertexEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6160(%rax), %r11
+	movq	6272(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	6272(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6272(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	6272(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(ProvokingVertexEXT), .-GL_PREFIX(ProvokingVertexEXT)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_785)
+	.type	GL_PREFIX(_dispatch_stub_785), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_785))
+GL_PREFIX(_dispatch_stub_785):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6280(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6280(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6280(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6280(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_785), .-GL_PREFIX(_dispatch_stub_785)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_786)
+	.type	GL_PREFIX(_dispatch_stub_786), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_786))
+GL_PREFIX(_dispatch_stub_786):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6288(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6288(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6288(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	6288(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_786), .-GL_PREFIX(_dispatch_stub_786)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_787)
+	.type	GL_PREFIX(_dispatch_stub_787), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_787))
+GL_PREFIX(_dispatch_stub_787):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6296(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29149,13 +29774,13 @@ GL_PREFIX(_dispatch_stub_770):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6160(%rax), %r11
+	movq	6296(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6160(%rax), %r11
+	movq	6296(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29169,19 +29794,19 @@ GL_PREFIX(_dispatch_stub_770):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6160(%rax), %r11
+	movq	6296(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_770), .-GL_PREFIX(_dispatch_stub_770)
+	.size	GL_PREFIX(_dispatch_stub_787), .-GL_PREFIX(_dispatch_stub_787)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_771)
-	.type	GL_PREFIX(_dispatch_stub_771), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_771))
-GL_PREFIX(_dispatch_stub_771):
+	.globl	GL_PREFIX(_dispatch_stub_788)
+	.type	GL_PREFIX(_dispatch_stub_788), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_788))
+GL_PREFIX(_dispatch_stub_788):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6168(%rax), %r11
+	movq	6304(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29195,13 +29820,13 @@ GL_PREFIX(_dispatch_stub_771):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6168(%rax), %r11
+	movq	6304(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6168(%rax), %r11
+	movq	6304(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29215,19 +29840,19 @@ GL_PREFIX(_dispatch_stub_771):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6168(%rax), %r11
+	movq	6304(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_771), .-GL_PREFIX(_dispatch_stub_771)
+	.size	GL_PREFIX(_dispatch_stub_788), .-GL_PREFIX(_dispatch_stub_788)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_772)
-	.type	GL_PREFIX(_dispatch_stub_772), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_772))
-GL_PREFIX(_dispatch_stub_772):
+	.globl	GL_PREFIX(_dispatch_stub_789)
+	.type	GL_PREFIX(_dispatch_stub_789), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_789))
+GL_PREFIX(_dispatch_stub_789):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6176(%rax), %r11
+	movq	6312(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29241,13 +29866,13 @@ GL_PREFIX(_dispatch_stub_772):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6176(%rax), %r11
+	movq	6312(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6176(%rax), %r11
+	movq	6312(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29261,19 +29886,19 @@ GL_PREFIX(_dispatch_stub_772):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6176(%rax), %r11
+	movq	6312(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_772), .-GL_PREFIX(_dispatch_stub_772)
+	.size	GL_PREFIX(_dispatch_stub_789), .-GL_PREFIX(_dispatch_stub_789)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_773)
-	.type	GL_PREFIX(_dispatch_stub_773), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_773))
-GL_PREFIX(_dispatch_stub_773):
+	.globl	GL_PREFIX(_dispatch_stub_790)
+	.type	GL_PREFIX(_dispatch_stub_790), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_790))
+GL_PREFIX(_dispatch_stub_790):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6184(%rax), %r11
+	movq	6320(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29283,13 +29908,13 @@ GL_PREFIX(_dispatch_stub_773):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6184(%rax), %r11
+	movq	6320(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6184(%rax), %r11
+	movq	6320(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29299,19 +29924,19 @@ GL_PREFIX(_dispatch_stub_773):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6184(%rax), %r11
+	movq	6320(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773)
+	.size	GL_PREFIX(_dispatch_stub_790), .-GL_PREFIX(_dispatch_stub_790)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_774)
-	.type	GL_PREFIX(_dispatch_stub_774), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_774))
-GL_PREFIX(_dispatch_stub_774):
+	.globl	GL_PREFIX(_dispatch_stub_791)
+	.type	GL_PREFIX(_dispatch_stub_791), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_791))
+GL_PREFIX(_dispatch_stub_791):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6192(%rax), %r11
+	movq	6328(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29321,13 +29946,13 @@ GL_PREFIX(_dispatch_stub_774):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6192(%rax), %r11
+	movq	6328(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6192(%rax), %r11
+	movq	6328(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29337,10 +29962,10 @@ GL_PREFIX(_dispatch_stub_774):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6192(%rax), %r11
+	movq	6328(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_774), .-GL_PREFIX(_dispatch_stub_774)
+	.size	GL_PREFIX(_dispatch_stub_791), .-GL_PREFIX(_dispatch_stub_791)
 
 	.globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement)
 	.globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture)
@@ -29595,7 +30220,9 @@ GL_PREFIX(_dispatch_stub_774):
 	.globl GL_PREFIX(IsProgramARB) ; .set GL_PREFIX(IsProgramARB), GL_PREFIX(IsProgramNV)
 	.globl GL_PREFIX(PointParameteri) ; .set GL_PREFIX(PointParameteri), GL_PREFIX(PointParameteriNV)
 	.globl GL_PREFIX(PointParameteriv) ; .set GL_PREFIX(PointParameteriv), GL_PREFIX(PointParameterivNV)
-	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_750)
+	.globl GL_PREFIX(DeleteVertexArrays) ; .set GL_PREFIX(DeleteVertexArrays), GL_PREFIX(_dispatch_stub_752)
+	.globl GL_PREFIX(IsVertexArray) ; .set GL_PREFIX(IsVertexArray), GL_PREFIX(_dispatch_stub_754)
+	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_762)
 	.globl GL_PREFIX(BindFramebuffer) ; .set GL_PREFIX(BindFramebuffer), GL_PREFIX(BindFramebufferEXT)
 	.globl GL_PREFIX(BindRenderbuffer) ; .set GL_PREFIX(BindRenderbuffer), GL_PREFIX(BindRenderbufferEXT)
 	.globl GL_PREFIX(CheckFramebufferStatus) ; .set GL_PREFIX(CheckFramebufferStatus), GL_PREFIX(CheckFramebufferStatusEXT)
@@ -29613,7 +30240,7 @@ GL_PREFIX(_dispatch_stub_774):
 	.globl GL_PREFIX(IsFramebuffer) ; .set GL_PREFIX(IsFramebuffer), GL_PREFIX(IsFramebufferEXT)
 	.globl GL_PREFIX(IsRenderbuffer) ; .set GL_PREFIX(IsRenderbuffer), GL_PREFIX(IsRenderbufferEXT)
 	.globl GL_PREFIX(RenderbufferStorage) ; .set GL_PREFIX(RenderbufferStorage), GL_PREFIX(RenderbufferStorageEXT)
-	.globl GL_PREFIX(BlitFramebuffer) ; .set GL_PREFIX(BlitFramebuffer), GL_PREFIX(_dispatch_stub_768)
+	.globl GL_PREFIX(BlitFramebuffer) ; .set GL_PREFIX(BlitFramebuffer), GL_PREFIX(_dispatch_stub_780)
 	.globl GL_PREFIX(FramebufferTextureLayer) ; .set GL_PREFIX(FramebufferTextureLayer), GL_PREFIX(FramebufferTextureLayerEXT)
 
 #if defined(GLX_USE_TLS) && defined(__linux__)
diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S
index 7aa344f214..46794adaf7 100644
--- a/src/mesa/x86/glapi_x86.S
+++ b/src/mesa/x86/glapi_x86.S
@@ -52,7 +52,7 @@
 #define GLOBL_FN(x) GLOBL x
 #endif
 
-#if defined(PTHREADS) || defined(USE_XTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)
+#if defined(PTHREADS) || defined(SOLARIS_THREADS) || defined(WIN32_THREADS) || defined(BEOS_THREADS)
 #  define THREADS
 #endif
 
@@ -713,23 +713,35 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(GetAttribLocationARB, _gloffset_GetAttribLocationARB, GetAttribLocationARB@8)
 	GL_STUB(DrawBuffersARB, _gloffset_DrawBuffersARB, DrawBuffersARB@8)
 	GL_STUB(RenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample, RenderbufferStorageMultisample@20)
+	GL_STUB(FlushMappedBufferRange, _gloffset_FlushMappedBufferRange, FlushMappedBufferRange@12)
+	GL_STUB(MapBufferRange, _gloffset_MapBufferRange, MapBufferRange@16)
+	GL_STUB(BindVertexArray, _gloffset_BindVertexArray, BindVertexArray@4)
+	GL_STUB(GenVertexArrays, _gloffset_GenVertexArrays, GenVertexArrays@8)
+	GL_STUB(CopyBufferSubData, _gloffset_CopyBufferSubData, CopyBufferSubData@20)
+	GL_STUB(ClientWaitSync, _gloffset_ClientWaitSync, ClientWaitSync@12)
+	GL_STUB(DeleteSync, _gloffset_DeleteSync, DeleteSync@4)
+	GL_STUB(FenceSync, _gloffset_FenceSync, FenceSync@8)
+	GL_STUB(GetInteger64v, _gloffset_GetInteger64v, GetInteger64v@8)
+	GL_STUB(GetSynciv, _gloffset_GetSynciv, GetSynciv@20)
+	GL_STUB(IsSync, _gloffset_IsSync, IsSync@4)
+	GL_STUB(WaitSync, _gloffset_WaitSync, WaitSync@12)
 	GL_STUB(PolygonOffsetEXT, _gloffset_PolygonOffsetEXT, PolygonOffsetEXT@8)
-	GL_STUB(_dispatch_stub_563, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_563@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_563, _dispatch_stub_563@8))
-	GL_STUB(_dispatch_stub_564, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_564@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_564, _dispatch_stub_564@8))
-	GL_STUB(_dispatch_stub_565, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_565@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_565, _dispatch_stub_565@8))
-	GL_STUB(_dispatch_stub_566, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_566@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_566, _dispatch_stub_566@8))
-	GL_STUB(_dispatch_stub_567, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_567@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_567, _dispatch_stub_567@8))
-	GL_STUB(_dispatch_stub_568, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_568@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_568, _dispatch_stub_568@8))
-	GL_STUB(_dispatch_stub_569, _gloffset_SampleMaskSGIS, _dispatch_stub_569@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_569, _dispatch_stub_569@8))
-	GL_STUB(_dispatch_stub_570, _gloffset_SamplePatternSGIS, _dispatch_stub_570@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_570, _dispatch_stub_570@4))
+	GL_STUB(_dispatch_stub_575, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_575@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_575, _dispatch_stub_575@8))
+	GL_STUB(_dispatch_stub_576, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_576@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_576, _dispatch_stub_576@8))
+	GL_STUB(_dispatch_stub_577, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_577@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_577, _dispatch_stub_577@8))
+	GL_STUB(_dispatch_stub_578, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_578@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_578, _dispatch_stub_578@8))
+	GL_STUB(_dispatch_stub_579, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_579@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_579, _dispatch_stub_579@8))
+	GL_STUB(_dispatch_stub_580, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_580@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_580, _dispatch_stub_580@8))
+	GL_STUB(_dispatch_stub_581, _gloffset_SampleMaskSGIS, _dispatch_stub_581@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_581, _dispatch_stub_581@8))
+	GL_STUB(_dispatch_stub_582, _gloffset_SamplePatternSGIS, _dispatch_stub_582@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_582, _dispatch_stub_582@4))
 	GL_STUB(ColorPointerEXT, _gloffset_ColorPointerEXT, ColorPointerEXT@20)
 	GL_STUB(EdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT, EdgeFlagPointerEXT@12)
 	GL_STUB(IndexPointerEXT, _gloffset_IndexPointerEXT, IndexPointerEXT@16)
@@ -740,10 +752,10 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(PointParameterfvEXT, _gloffset_PointParameterfvEXT, PointParameterfvEXT@8)
 	GL_STUB(LockArraysEXT, _gloffset_LockArraysEXT, LockArraysEXT@8)
 	GL_STUB(UnlockArraysEXT, _gloffset_UnlockArraysEXT, UnlockArraysEXT@0)
-	GL_STUB(_dispatch_stub_581, _gloffset_CullParameterdvEXT, _dispatch_stub_581@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_581, _dispatch_stub_581@8))
-	GL_STUB(_dispatch_stub_582, _gloffset_CullParameterfvEXT, _dispatch_stub_582@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_582, _dispatch_stub_582@8))
+	GL_STUB(_dispatch_stub_593, _gloffset_CullParameterdvEXT, _dispatch_stub_593@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_593, _dispatch_stub_593@8))
+	GL_STUB(_dispatch_stub_594, _gloffset_CullParameterfvEXT, _dispatch_stub_594@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_594, _dispatch_stub_594@8))
 	GL_STUB(SecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT, SecondaryColor3bEXT@12)
 	GL_STUB(SecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT, SecondaryColor3bvEXT@4)
 	GL_STUB(SecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT, SecondaryColor3dEXT@24)
@@ -768,8 +780,8 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(FogCoorddvEXT, _gloffset_FogCoorddvEXT, FogCoorddvEXT@4)
 	GL_STUB(FogCoordfEXT, _gloffset_FogCoordfEXT, FogCoordfEXT@4)
 	GL_STUB(FogCoordfvEXT, _gloffset_FogCoordfvEXT, FogCoordfvEXT@4)
-	GL_STUB(_dispatch_stub_607, _gloffset_PixelTexGenSGIX, _dispatch_stub_607@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_607, _dispatch_stub_607@4))
+	GL_STUB(_dispatch_stub_619, _gloffset_PixelTexGenSGIX, _dispatch_stub_619@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_619, _dispatch_stub_619@4))
 	GL_STUB(BlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparateEXT@16)
 	GL_STUB(FlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV, FlushVertexArrayRangeNV@0)
 	GL_STUB(VertexArrayRangeNV, _gloffset_VertexArrayRangeNV, VertexArrayRangeNV@8)
@@ -811,24 +823,24 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(WindowPos4ivMESA, _gloffset_WindowPos4ivMESA, WindowPos4ivMESA@4)
 	GL_STUB(WindowPos4sMESA, _gloffset_WindowPos4sMESA, WindowPos4sMESA@16)
 	GL_STUB(WindowPos4svMESA, _gloffset_WindowPos4svMESA, WindowPos4svMESA@4)
-	GL_STUB(_dispatch_stub_649, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_649@20)
-	HIDDEN(GL_PREFIX(_dispatch_stub_649, _dispatch_stub_649@20))
-	GL_STUB(_dispatch_stub_650, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_650@24)
-	HIDDEN(GL_PREFIX(_dispatch_stub_650, _dispatch_stub_650@24))
-	GL_STUB(_dispatch_stub_651, _gloffset_DeleteFencesNV, _dispatch_stub_651@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_651, _dispatch_stub_651@8))
-	GL_STUB(_dispatch_stub_652, _gloffset_FinishFenceNV, _dispatch_stub_652@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_652, _dispatch_stub_652@4))
-	GL_STUB(_dispatch_stub_653, _gloffset_GenFencesNV, _dispatch_stub_653@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_653, _dispatch_stub_653@8))
-	GL_STUB(_dispatch_stub_654, _gloffset_GetFenceivNV, _dispatch_stub_654@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_654, _dispatch_stub_654@12))
-	GL_STUB(_dispatch_stub_655, _gloffset_IsFenceNV, _dispatch_stub_655@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_655, _dispatch_stub_655@4))
-	GL_STUB(_dispatch_stub_656, _gloffset_SetFenceNV, _dispatch_stub_656@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_656, _dispatch_stub_656@8))
-	GL_STUB(_dispatch_stub_657, _gloffset_TestFenceNV, _dispatch_stub_657@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_657, _dispatch_stub_657@4))
+	GL_STUB(_dispatch_stub_661, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_661@20)
+	HIDDEN(GL_PREFIX(_dispatch_stub_661, _dispatch_stub_661@20))
+	GL_STUB(_dispatch_stub_662, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_662@24)
+	HIDDEN(GL_PREFIX(_dispatch_stub_662, _dispatch_stub_662@24))
+	GL_STUB(_dispatch_stub_663, _gloffset_DeleteFencesNV, _dispatch_stub_663@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_663, _dispatch_stub_663@8))
+	GL_STUB(_dispatch_stub_664, _gloffset_FinishFenceNV, _dispatch_stub_664@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_664, _dispatch_stub_664@4))
+	GL_STUB(_dispatch_stub_665, _gloffset_GenFencesNV, _dispatch_stub_665@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_665, _dispatch_stub_665@8))
+	GL_STUB(_dispatch_stub_666, _gloffset_GetFenceivNV, _dispatch_stub_666@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_666, _dispatch_stub_666@12))
+	GL_STUB(_dispatch_stub_667, _gloffset_IsFenceNV, _dispatch_stub_667@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_667, _dispatch_stub_667@4))
+	GL_STUB(_dispatch_stub_668, _gloffset_SetFenceNV, _dispatch_stub_668@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_668, _dispatch_stub_668@8))
+	GL_STUB(_dispatch_stub_669, _gloffset_TestFenceNV, _dispatch_stub_669@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_669, _dispatch_stub_669@4))
 	GL_STUB(AreProgramsResidentNV, _gloffset_AreProgramsResidentNV, AreProgramsResidentNV@12)
 	GL_STUB(BindProgramNV, _gloffset_BindProgramNV, BindProgramNV@8)
 	GL_STUB(DeleteProgramsNV, _gloffset_DeleteProgramsNV, DeleteProgramsNV@8)
@@ -909,26 +921,26 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(SetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI, SetFragmentShaderConstantATI@8)
 	GL_STUB(PointParameteriNV, _gloffset_PointParameteriNV, PointParameteriNV@8)
 	GL_STUB(PointParameterivNV, _gloffset_PointParameterivNV, PointParameterivNV@8)
-	GL_STUB(_dispatch_stub_738, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_738@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_738, _dispatch_stub_738@4))
-	GL_STUB(_dispatch_stub_739, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_739@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_739, _dispatch_stub_739@4))
-	GL_STUB(_dispatch_stub_740, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_740@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_740, _dispatch_stub_740@8))
-	GL_STUB(_dispatch_stub_741, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_741@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_741, _dispatch_stub_741@8))
-	GL_STUB(_dispatch_stub_742, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_742@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_742, _dispatch_stub_742@4))
+	GL_STUB(_dispatch_stub_750, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_750@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_750, _dispatch_stub_750@4))
+	GL_STUB(_dispatch_stub_751, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_751@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_751, _dispatch_stub_751@4))
+	GL_STUB(_dispatch_stub_752, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_752@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_752, _dispatch_stub_752@8))
+	GL_STUB(_dispatch_stub_753, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_753@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_753, _dispatch_stub_753@8))
+	GL_STUB(_dispatch_stub_754, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_754@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_754, _dispatch_stub_754@4))
 	GL_STUB(GetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV, GetProgramNamedParameterdvNV@16)
 	GL_STUB(GetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV, GetProgramNamedParameterfvNV@16)
 	GL_STUB(ProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV, ProgramNamedParameter4dNV@44)
 	GL_STUB(ProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV, ProgramNamedParameter4dvNV@16)
 	GL_STUB(ProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV, ProgramNamedParameter4fNV@28)
 	GL_STUB(ProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV, ProgramNamedParameter4fvNV@16)
-	GL_STUB(_dispatch_stub_749, _gloffset_DepthBoundsEXT, _dispatch_stub_749@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_749, _dispatch_stub_749@16))
-	GL_STUB(_dispatch_stub_750, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_750@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_750, _dispatch_stub_750@8))
+	GL_STUB(_dispatch_stub_761, _gloffset_DepthBoundsEXT, _dispatch_stub_761@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_761, _dispatch_stub_761@16))
+	GL_STUB(_dispatch_stub_762, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_762@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_762, _dispatch_stub_762@8))
 	GL_STUB(BindFramebufferEXT, _gloffset_BindFramebufferEXT, BindFramebufferEXT@8)
 	GL_STUB(BindRenderbufferEXT, _gloffset_BindRenderbufferEXT, BindRenderbufferEXT@8)
 	GL_STUB(CheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT, CheckFramebufferStatusEXT@4)
@@ -946,19 +958,28 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(IsFramebufferEXT, _gloffset_IsFramebufferEXT, IsFramebufferEXT@4)
 	GL_STUB(IsRenderbufferEXT, _gloffset_IsRenderbufferEXT, IsRenderbufferEXT@4)
 	GL_STUB(RenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT, RenderbufferStorageEXT@16)
-	GL_STUB(_dispatch_stub_768, _gloffset_BlitFramebufferEXT, _dispatch_stub_768@40)
-	HIDDEN(GL_PREFIX(_dispatch_stub_768, _dispatch_stub_768@40))
+	GL_STUB(_dispatch_stub_780, _gloffset_BlitFramebufferEXT, _dispatch_stub_780@40)
+	HIDDEN(GL_PREFIX(_dispatch_stub_780, _dispatch_stub_780@40))
+	GL_STUB(_dispatch_stub_781, _gloffset_BufferParameteriAPPLE, _dispatch_stub_781@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_781, _dispatch_stub_781@12))
+	GL_STUB(_dispatch_stub_782, _gloffset_FlushMappedBufferRangeAPPLE, _dispatch_stub_782@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_782, _dispatch_stub_782@12))
 	GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20)
-	GL_STUB(_dispatch_stub_770, _gloffset_StencilFuncSeparateATI, _dispatch_stub_770@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@16))
-	GL_STUB(_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_771@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@16))
-	GL_STUB(_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_772@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@16))
-	GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_773@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@12))
-	GL_STUB(_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_774@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_774, _dispatch_stub_774@12))
+	GL_STUB(ProvokingVertexEXT, _gloffset_ProvokingVertexEXT, ProvokingVertexEXT@4)
+	GL_STUB(_dispatch_stub_785, _gloffset_GetTexParameterPointervAPPLE, _dispatch_stub_785@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_785, _dispatch_stub_785@12))
+	GL_STUB(_dispatch_stub_786, _gloffset_TextureRangeAPPLE, _dispatch_stub_786@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_786, _dispatch_stub_786@12))
+	GL_STUB(_dispatch_stub_787, _gloffset_StencilFuncSeparateATI, _dispatch_stub_787@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_787, _dispatch_stub_787@16))
+	GL_STUB(_dispatch_stub_788, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_788@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_788, _dispatch_stub_788@16))
+	GL_STUB(_dispatch_stub_789, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_789@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_789, _dispatch_stub_789@16))
+	GL_STUB(_dispatch_stub_790, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_790@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_790, _dispatch_stub_790@12))
+	GL_STUB(_dispatch_stub_791, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_791@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_791, _dispatch_stub_791@12))
 	GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4)
 	GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8)
 	GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12)